apply clang-format
Change-Id: Ib8c9eb6263d6eba6b9d7b2e402b7e83a78c86be9
diff --git a/vp10/common/alloccommon.c b/vp10/common/alloccommon.c
index 5469a7a..fb6d81d 100644
--- a/vp10/common/alloccommon.c
+++ b/vp10/common/alloccommon.c
@@ -35,8 +35,7 @@
for (i = 0; i < NUM_PING_PONG_BUFFERS; ++i) {
cm->seg_map_array[i] = (uint8_t *)vpx_calloc(seg_map_size, 1);
- if (cm->seg_map_array[i] == NULL)
- return 1;
+ if (cm->seg_map_array[i] == NULL) return 1;
}
cm->seg_map_alloc_size = seg_map_size;
@@ -97,15 +96,13 @@
new_mi_size = cm->mi_stride * calc_mi_size(cm->mi_rows);
if (cm->mi_alloc_size < new_mi_size) {
cm->free_mi(cm);
- if (cm->alloc_mi(cm, new_mi_size))
- goto fail;
+ if (cm->alloc_mi(cm, new_mi_size)) goto fail;
}
if (cm->seg_map_alloc_size < cm->mi_rows * cm->mi_cols) {
// Create the segmentation map structure and set to 0.
free_seg_map(cm);
- if (alloc_seg_map(cm, cm->mi_rows * cm->mi_cols))
- goto fail;
+ if (alloc_seg_map(cm, cm->mi_rows * cm->mi_cols)) goto fail;
}
if (cm->above_context_alloc_cols < cm->mi_cols) {
@@ -124,7 +121,7 @@
return 0;
- fail:
+fail:
vp10_free_context_buffers(cm);
return 1;
}
diff --git a/vp10/common/alloccommon.h b/vp10/common/alloccommon.h
index 9d35e5e..bbe2ff6 100644
--- a/vp10/common/alloccommon.h
+++ b/vp10/common/alloccommon.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_COMMON_ALLOCCOMMON_H_
#define VP10_COMMON_ALLOCCOMMON_H_
diff --git a/vp10/common/arm/neon/iht4x4_add_neon.c b/vp10/common/arm/neon/iht4x4_add_neon.c
index bd3e8b3..d074bc9 100644
--- a/vp10/common/arm/neon/iht4x4_add_neon.c
+++ b/vp10/common/arm/neon/iht4x4_add_neon.c
@@ -23,226 +23,211 @@
static int16_t cospi_16_64 = 0x2d41;
static int16_t cospi_24_64 = 0x187e;
-static INLINE void TRANSPOSE4X4(
- int16x8_t *q8s16,
- int16x8_t *q9s16) {
- int32x4_t q8s32, q9s32;
- int16x4x2_t d0x2s16, d1x2s16;
- int32x4x2_t q0x2s32;
+static INLINE void TRANSPOSE4X4(int16x8_t *q8s16, int16x8_t *q9s16) {
+ int32x4_t q8s32, q9s32;
+ int16x4x2_t d0x2s16, d1x2s16;
+ int32x4x2_t q0x2s32;
- d0x2s16 = vtrn_s16(vget_low_s16(*q8s16), vget_high_s16(*q8s16));
- d1x2s16 = vtrn_s16(vget_low_s16(*q9s16), vget_high_s16(*q9s16));
+ d0x2s16 = vtrn_s16(vget_low_s16(*q8s16), vget_high_s16(*q8s16));
+ d1x2s16 = vtrn_s16(vget_low_s16(*q9s16), vget_high_s16(*q9s16));
- q8s32 = vreinterpretq_s32_s16(vcombine_s16(d0x2s16.val[0], d0x2s16.val[1]));
- q9s32 = vreinterpretq_s32_s16(vcombine_s16(d1x2s16.val[0], d1x2s16.val[1]));
- q0x2s32 = vtrnq_s32(q8s32, q9s32);
+ q8s32 = vreinterpretq_s32_s16(vcombine_s16(d0x2s16.val[0], d0x2s16.val[1]));
+ q9s32 = vreinterpretq_s32_s16(vcombine_s16(d1x2s16.val[0], d1x2s16.val[1]));
+ q0x2s32 = vtrnq_s32(q8s32, q9s32);
- *q8s16 = vreinterpretq_s16_s32(q0x2s32.val[0]);
- *q9s16 = vreinterpretq_s16_s32(q0x2s32.val[1]);
- return;
+ *q8s16 = vreinterpretq_s16_s32(q0x2s32.val[0]);
+ *q9s16 = vreinterpretq_s16_s32(q0x2s32.val[1]);
+ return;
}
-static INLINE void GENERATE_COSINE_CONSTANTS(
- int16x4_t *d0s16,
- int16x4_t *d1s16,
- int16x4_t *d2s16) {
- *d0s16 = vdup_n_s16(cospi_8_64);
- *d1s16 = vdup_n_s16(cospi_16_64);
- *d2s16 = vdup_n_s16(cospi_24_64);
- return;
+static INLINE void GENERATE_COSINE_CONSTANTS(int16x4_t *d0s16, int16x4_t *d1s16,
+ int16x4_t *d2s16) {
+ *d0s16 = vdup_n_s16(cospi_8_64);
+ *d1s16 = vdup_n_s16(cospi_16_64);
+ *d2s16 = vdup_n_s16(cospi_24_64);
+ return;
}
-static INLINE void GENERATE_SINE_CONSTANTS(
- int16x4_t *d3s16,
- int16x4_t *d4s16,
- int16x4_t *d5s16,
- int16x8_t *q3s16) {
- *d3s16 = vdup_n_s16(sinpi_1_9);
- *d4s16 = vdup_n_s16(sinpi_2_9);
- *q3s16 = vdupq_n_s16(sinpi_3_9);
- *d5s16 = vdup_n_s16(sinpi_4_9);
- return;
+static INLINE void GENERATE_SINE_CONSTANTS(int16x4_t *d3s16, int16x4_t *d4s16,
+ int16x4_t *d5s16, int16x8_t *q3s16) {
+ *d3s16 = vdup_n_s16(sinpi_1_9);
+ *d4s16 = vdup_n_s16(sinpi_2_9);
+ *q3s16 = vdupq_n_s16(sinpi_3_9);
+ *d5s16 = vdup_n_s16(sinpi_4_9);
+ return;
}
-static INLINE void IDCT4x4_1D(
- int16x4_t *d0s16,
- int16x4_t *d1s16,
- int16x4_t *d2s16,
- int16x8_t *q8s16,
- int16x8_t *q9s16) {
- int16x4_t d16s16, d17s16, d18s16, d19s16, d23s16, d24s16;
- int16x4_t d26s16, d27s16, d28s16, d29s16;
- int32x4_t q10s32, q13s32, q14s32, q15s32;
- int16x8_t q13s16, q14s16;
+static INLINE void IDCT4x4_1D(int16x4_t *d0s16, int16x4_t *d1s16,
+ int16x4_t *d2s16, int16x8_t *q8s16,
+ int16x8_t *q9s16) {
+ int16x4_t d16s16, d17s16, d18s16, d19s16, d23s16, d24s16;
+ int16x4_t d26s16, d27s16, d28s16, d29s16;
+ int32x4_t q10s32, q13s32, q14s32, q15s32;
+ int16x8_t q13s16, q14s16;
- d16s16 = vget_low_s16(*q8s16);
- d17s16 = vget_high_s16(*q8s16);
- d18s16 = vget_low_s16(*q9s16);
- d19s16 = vget_high_s16(*q9s16);
+ d16s16 = vget_low_s16(*q8s16);
+ d17s16 = vget_high_s16(*q8s16);
+ d18s16 = vget_low_s16(*q9s16);
+ d19s16 = vget_high_s16(*q9s16);
- d23s16 = vadd_s16(d16s16, d18s16);
- d24s16 = vsub_s16(d16s16, d18s16);
+ d23s16 = vadd_s16(d16s16, d18s16);
+ d24s16 = vsub_s16(d16s16, d18s16);
- q15s32 = vmull_s16(d17s16, *d2s16);
- q10s32 = vmull_s16(d17s16, *d0s16);
- q13s32 = vmull_s16(d23s16, *d1s16);
- q14s32 = vmull_s16(d24s16, *d1s16);
- q15s32 = vmlsl_s16(q15s32, d19s16, *d0s16);
- q10s32 = vmlal_s16(q10s32, d19s16, *d2s16);
+ q15s32 = vmull_s16(d17s16, *d2s16);
+ q10s32 = vmull_s16(d17s16, *d0s16);
+ q13s32 = vmull_s16(d23s16, *d1s16);
+ q14s32 = vmull_s16(d24s16, *d1s16);
+ q15s32 = vmlsl_s16(q15s32, d19s16, *d0s16);
+ q10s32 = vmlal_s16(q10s32, d19s16, *d2s16);
- d26s16 = vqrshrn_n_s32(q13s32, 14);
- d27s16 = vqrshrn_n_s32(q14s32, 14);
- d29s16 = vqrshrn_n_s32(q15s32, 14);
- d28s16 = vqrshrn_n_s32(q10s32, 14);
+ d26s16 = vqrshrn_n_s32(q13s32, 14);
+ d27s16 = vqrshrn_n_s32(q14s32, 14);
+ d29s16 = vqrshrn_n_s32(q15s32, 14);
+ d28s16 = vqrshrn_n_s32(q10s32, 14);
- q13s16 = vcombine_s16(d26s16, d27s16);
- q14s16 = vcombine_s16(d28s16, d29s16);
- *q8s16 = vaddq_s16(q13s16, q14s16);
- *q9s16 = vsubq_s16(q13s16, q14s16);
- *q9s16 = vcombine_s16(vget_high_s16(*q9s16),
- vget_low_s16(*q9s16)); // vswp
- return;
+ q13s16 = vcombine_s16(d26s16, d27s16);
+ q14s16 = vcombine_s16(d28s16, d29s16);
+ *q8s16 = vaddq_s16(q13s16, q14s16);
+ *q9s16 = vsubq_s16(q13s16, q14s16);
+ *q9s16 = vcombine_s16(vget_high_s16(*q9s16), vget_low_s16(*q9s16)); // vswp
+ return;
}
-static INLINE void IADST4x4_1D(
- int16x4_t *d3s16,
- int16x4_t *d4s16,
- int16x4_t *d5s16,
- int16x8_t *q3s16,
- int16x8_t *q8s16,
- int16x8_t *q9s16) {
- int16x4_t d6s16, d16s16, d17s16, d18s16, d19s16;
- int32x4_t q8s32, q9s32, q10s32, q11s32, q12s32, q13s32, q14s32, q15s32;
+static INLINE void IADST4x4_1D(int16x4_t *d3s16, int16x4_t *d4s16,
+ int16x4_t *d5s16, int16x8_t *q3s16,
+ int16x8_t *q8s16, int16x8_t *q9s16) {
+ int16x4_t d6s16, d16s16, d17s16, d18s16, d19s16;
+ int32x4_t q8s32, q9s32, q10s32, q11s32, q12s32, q13s32, q14s32, q15s32;
- d6s16 = vget_low_s16(*q3s16);
+ d6s16 = vget_low_s16(*q3s16);
- d16s16 = vget_low_s16(*q8s16);
- d17s16 = vget_high_s16(*q8s16);
- d18s16 = vget_low_s16(*q9s16);
- d19s16 = vget_high_s16(*q9s16);
+ d16s16 = vget_low_s16(*q8s16);
+ d17s16 = vget_high_s16(*q8s16);
+ d18s16 = vget_low_s16(*q9s16);
+ d19s16 = vget_high_s16(*q9s16);
- q10s32 = vmull_s16(*d3s16, d16s16);
- q11s32 = vmull_s16(*d4s16, d16s16);
- q12s32 = vmull_s16(d6s16, d17s16);
- q13s32 = vmull_s16(*d5s16, d18s16);
- q14s32 = vmull_s16(*d3s16, d18s16);
- q15s32 = vmovl_s16(d16s16);
- q15s32 = vaddw_s16(q15s32, d19s16);
- q8s32 = vmull_s16(*d4s16, d19s16);
- q15s32 = vsubw_s16(q15s32, d18s16);
- q9s32 = vmull_s16(*d5s16, d19s16);
+ q10s32 = vmull_s16(*d3s16, d16s16);
+ q11s32 = vmull_s16(*d4s16, d16s16);
+ q12s32 = vmull_s16(d6s16, d17s16);
+ q13s32 = vmull_s16(*d5s16, d18s16);
+ q14s32 = vmull_s16(*d3s16, d18s16);
+ q15s32 = vmovl_s16(d16s16);
+ q15s32 = vaddw_s16(q15s32, d19s16);
+ q8s32 = vmull_s16(*d4s16, d19s16);
+ q15s32 = vsubw_s16(q15s32, d18s16);
+ q9s32 = vmull_s16(*d5s16, d19s16);
- q10s32 = vaddq_s32(q10s32, q13s32);
- q10s32 = vaddq_s32(q10s32, q8s32);
- q11s32 = vsubq_s32(q11s32, q14s32);
- q8s32 = vdupq_n_s32(sinpi_3_9);
- q11s32 = vsubq_s32(q11s32, q9s32);
- q15s32 = vmulq_s32(q15s32, q8s32);
+ q10s32 = vaddq_s32(q10s32, q13s32);
+ q10s32 = vaddq_s32(q10s32, q8s32);
+ q11s32 = vsubq_s32(q11s32, q14s32);
+ q8s32 = vdupq_n_s32(sinpi_3_9);
+ q11s32 = vsubq_s32(q11s32, q9s32);
+ q15s32 = vmulq_s32(q15s32, q8s32);
- q13s32 = vaddq_s32(q10s32, q12s32);
- q10s32 = vaddq_s32(q10s32, q11s32);
- q14s32 = vaddq_s32(q11s32, q12s32);
- q10s32 = vsubq_s32(q10s32, q12s32);
+ q13s32 = vaddq_s32(q10s32, q12s32);
+ q10s32 = vaddq_s32(q10s32, q11s32);
+ q14s32 = vaddq_s32(q11s32, q12s32);
+ q10s32 = vsubq_s32(q10s32, q12s32);
- d16s16 = vqrshrn_n_s32(q13s32, 14);
- d17s16 = vqrshrn_n_s32(q14s32, 14);
- d18s16 = vqrshrn_n_s32(q15s32, 14);
- d19s16 = vqrshrn_n_s32(q10s32, 14);
+ d16s16 = vqrshrn_n_s32(q13s32, 14);
+ d17s16 = vqrshrn_n_s32(q14s32, 14);
+ d18s16 = vqrshrn_n_s32(q15s32, 14);
+ d19s16 = vqrshrn_n_s32(q10s32, 14);
- *q8s16 = vcombine_s16(d16s16, d17s16);
- *q9s16 = vcombine_s16(d18s16, d19s16);
- return;
+ *q8s16 = vcombine_s16(d16s16, d17s16);
+ *q9s16 = vcombine_s16(d18s16, d19s16);
+ return;
}
void vp10_iht4x4_16_add_neon(const tran_low_t *input, uint8_t *dest,
- int dest_stride, int tx_type) {
- uint8x8_t d26u8, d27u8;
- int16x4_t d0s16, d1s16, d2s16, d3s16, d4s16, d5s16;
- uint32x2_t d26u32, d27u32;
- int16x8_t q3s16, q8s16, q9s16;
- uint16x8_t q8u16, q9u16;
+ int dest_stride, int tx_type) {
+ uint8x8_t d26u8, d27u8;
+ int16x4_t d0s16, d1s16, d2s16, d3s16, d4s16, d5s16;
+ uint32x2_t d26u32, d27u32;
+ int16x8_t q3s16, q8s16, q9s16;
+ uint16x8_t q8u16, q9u16;
- d26u32 = d27u32 = vdup_n_u32(0);
+ d26u32 = d27u32 = vdup_n_u32(0);
- q8s16 = vld1q_s16(input);
- q9s16 = vld1q_s16(input + 8);
+ q8s16 = vld1q_s16(input);
+ q9s16 = vld1q_s16(input + 8);
- TRANSPOSE4X4(&q8s16, &q9s16);
+ TRANSPOSE4X4(&q8s16, &q9s16);
- switch (tx_type) {
- case 0: // idct_idct is not supported. Fall back to C
- vp10_iht4x4_16_add_c(input, dest, dest_stride, tx_type);
- return;
- break;
- case 1: // iadst_idct
- // generate constants
- GENERATE_COSINE_CONSTANTS(&d0s16, &d1s16, &d2s16);
- GENERATE_SINE_CONSTANTS(&d3s16, &d4s16, &d5s16, &q3s16);
+ switch (tx_type) {
+ case 0: // idct_idct is not supported. Fall back to C
+ vp10_iht4x4_16_add_c(input, dest, dest_stride, tx_type);
+ return;
+ break;
+ case 1: // iadst_idct
+ // generate constants
+ GENERATE_COSINE_CONSTANTS(&d0s16, &d1s16, &d2s16);
+ GENERATE_SINE_CONSTANTS(&d3s16, &d4s16, &d5s16, &q3s16);
- // first transform rows
- IDCT4x4_1D(&d0s16, &d1s16, &d2s16, &q8s16, &q9s16);
+ // first transform rows
+ IDCT4x4_1D(&d0s16, &d1s16, &d2s16, &q8s16, &q9s16);
- // transpose the matrix
- TRANSPOSE4X4(&q8s16, &q9s16);
+ // transpose the matrix
+ TRANSPOSE4X4(&q8s16, &q9s16);
- // then transform columns
- IADST4x4_1D(&d3s16, &d4s16, &d5s16, &q3s16, &q8s16, &q9s16);
- break;
- case 2: // idct_iadst
- // generate constantsyy
- GENERATE_COSINE_CONSTANTS(&d0s16, &d1s16, &d2s16);
- GENERATE_SINE_CONSTANTS(&d3s16, &d4s16, &d5s16, &q3s16);
+ // then transform columns
+ IADST4x4_1D(&d3s16, &d4s16, &d5s16, &q3s16, &q8s16, &q9s16);
+ break;
+ case 2: // idct_iadst
+ // generate constantsyy
+ GENERATE_COSINE_CONSTANTS(&d0s16, &d1s16, &d2s16);
+ GENERATE_SINE_CONSTANTS(&d3s16, &d4s16, &d5s16, &q3s16);
- // first transform rows
- IADST4x4_1D(&d3s16, &d4s16, &d5s16, &q3s16, &q8s16, &q9s16);
+ // first transform rows
+ IADST4x4_1D(&d3s16, &d4s16, &d5s16, &q3s16, &q8s16, &q9s16);
- // transpose the matrix
- TRANSPOSE4X4(&q8s16, &q9s16);
+ // transpose the matrix
+ TRANSPOSE4X4(&q8s16, &q9s16);
- // then transform columns
- IDCT4x4_1D(&d0s16, &d1s16, &d2s16, &q8s16, &q9s16);
- break;
- case 3: // iadst_iadst
- // generate constants
- GENERATE_SINE_CONSTANTS(&d3s16, &d4s16, &d5s16, &q3s16);
+ // then transform columns
+ IDCT4x4_1D(&d0s16, &d1s16, &d2s16, &q8s16, &q9s16);
+ break;
+ case 3: // iadst_iadst
+ // generate constants
+ GENERATE_SINE_CONSTANTS(&d3s16, &d4s16, &d5s16, &q3s16);
- // first transform rows
- IADST4x4_1D(&d3s16, &d4s16, &d5s16, &q3s16, &q8s16, &q9s16);
+ // first transform rows
+ IADST4x4_1D(&d3s16, &d4s16, &d5s16, &q3s16, &q8s16, &q9s16);
- // transpose the matrix
- TRANSPOSE4X4(&q8s16, &q9s16);
+ // transpose the matrix
+ TRANSPOSE4X4(&q8s16, &q9s16);
- // then transform columns
- IADST4x4_1D(&d3s16, &d4s16, &d5s16, &q3s16, &q8s16, &q9s16);
- break;
- default: // iadst_idct
- assert(0);
- break;
- }
+ // then transform columns
+ IADST4x4_1D(&d3s16, &d4s16, &d5s16, &q3s16, &q8s16, &q9s16);
+ break;
+ default: // iadst_idct
+ assert(0);
+ break;
+ }
- q8s16 = vrshrq_n_s16(q8s16, 4);
- q9s16 = vrshrq_n_s16(q9s16, 4);
+ q8s16 = vrshrq_n_s16(q8s16, 4);
+ q9s16 = vrshrq_n_s16(q9s16, 4);
- d26u32 = vld1_lane_u32((const uint32_t *)dest, d26u32, 0);
- dest += dest_stride;
- d26u32 = vld1_lane_u32((const uint32_t *)dest, d26u32, 1);
- dest += dest_stride;
- d27u32 = vld1_lane_u32((const uint32_t *)dest, d27u32, 0);
- dest += dest_stride;
- d27u32 = vld1_lane_u32((const uint32_t *)dest, d27u32, 1);
+ d26u32 = vld1_lane_u32((const uint32_t *)dest, d26u32, 0);
+ dest += dest_stride;
+ d26u32 = vld1_lane_u32((const uint32_t *)dest, d26u32, 1);
+ dest += dest_stride;
+ d27u32 = vld1_lane_u32((const uint32_t *)dest, d27u32, 0);
+ dest += dest_stride;
+ d27u32 = vld1_lane_u32((const uint32_t *)dest, d27u32, 1);
- q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_u32(d26u32));
- q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_u32(d27u32));
+ q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_u32(d26u32));
+ q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_u32(d27u32));
- d26u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
- d27u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
+ d26u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
+ d27u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
- vst1_lane_u32((uint32_t *)dest, vreinterpret_u32_u8(d27u8), 1);
- dest -= dest_stride;
- vst1_lane_u32((uint32_t *)dest, vreinterpret_u32_u8(d27u8), 0);
- dest -= dest_stride;
- vst1_lane_u32((uint32_t *)dest, vreinterpret_u32_u8(d26u8), 1);
- dest -= dest_stride;
- vst1_lane_u32((uint32_t *)dest, vreinterpret_u32_u8(d26u8), 0);
- return;
+ vst1_lane_u32((uint32_t *)dest, vreinterpret_u32_u8(d27u8), 1);
+ dest -= dest_stride;
+ vst1_lane_u32((uint32_t *)dest, vreinterpret_u32_u8(d27u8), 0);
+ dest -= dest_stride;
+ vst1_lane_u32((uint32_t *)dest, vreinterpret_u32_u8(d26u8), 1);
+ dest -= dest_stride;
+ vst1_lane_u32((uint32_t *)dest, vreinterpret_u32_u8(d26u8), 0);
+ return;
}
diff --git a/vp10/common/arm/neon/iht8x8_add_neon.c b/vp10/common/arm/neon/iht8x8_add_neon.c
index 82d7ccc..7e1c83f 100644
--- a/vp10/common/arm/neon/iht8x8_add_neon.c
+++ b/vp10/common/arm/neon/iht8x8_add_neon.c
@@ -31,594 +31,577 @@
static int16_t cospi_28_64 = 3196;
static int16_t cospi_30_64 = 1606;
-static INLINE void TRANSPOSE8X8(
- int16x8_t *q8s16,
- int16x8_t *q9s16,
- int16x8_t *q10s16,
- int16x8_t *q11s16,
- int16x8_t *q12s16,
- int16x8_t *q13s16,
- int16x8_t *q14s16,
- int16x8_t *q15s16) {
- int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
- int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
- int32x4x2_t q0x2s32, q1x2s32, q2x2s32, q3x2s32;
- int16x8x2_t q0x2s16, q1x2s16, q2x2s16, q3x2s16;
+static INLINE void TRANSPOSE8X8(int16x8_t *q8s16, int16x8_t *q9s16,
+ int16x8_t *q10s16, int16x8_t *q11s16,
+ int16x8_t *q12s16, int16x8_t *q13s16,
+ int16x8_t *q14s16, int16x8_t *q15s16) {
+ int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
+ int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
+ int32x4x2_t q0x2s32, q1x2s32, q2x2s32, q3x2s32;
+ int16x8x2_t q0x2s16, q1x2s16, q2x2s16, q3x2s16;
- d16s16 = vget_low_s16(*q8s16);
- d17s16 = vget_high_s16(*q8s16);
- d18s16 = vget_low_s16(*q9s16);
- d19s16 = vget_high_s16(*q9s16);
- d20s16 = vget_low_s16(*q10s16);
- d21s16 = vget_high_s16(*q10s16);
- d22s16 = vget_low_s16(*q11s16);
- d23s16 = vget_high_s16(*q11s16);
- d24s16 = vget_low_s16(*q12s16);
- d25s16 = vget_high_s16(*q12s16);
- d26s16 = vget_low_s16(*q13s16);
- d27s16 = vget_high_s16(*q13s16);
- d28s16 = vget_low_s16(*q14s16);
- d29s16 = vget_high_s16(*q14s16);
- d30s16 = vget_low_s16(*q15s16);
- d31s16 = vget_high_s16(*q15s16);
+ d16s16 = vget_low_s16(*q8s16);
+ d17s16 = vget_high_s16(*q8s16);
+ d18s16 = vget_low_s16(*q9s16);
+ d19s16 = vget_high_s16(*q9s16);
+ d20s16 = vget_low_s16(*q10s16);
+ d21s16 = vget_high_s16(*q10s16);
+ d22s16 = vget_low_s16(*q11s16);
+ d23s16 = vget_high_s16(*q11s16);
+ d24s16 = vget_low_s16(*q12s16);
+ d25s16 = vget_high_s16(*q12s16);
+ d26s16 = vget_low_s16(*q13s16);
+ d27s16 = vget_high_s16(*q13s16);
+ d28s16 = vget_low_s16(*q14s16);
+ d29s16 = vget_high_s16(*q14s16);
+ d30s16 = vget_low_s16(*q15s16);
+ d31s16 = vget_high_s16(*q15s16);
- *q8s16 = vcombine_s16(d16s16, d24s16); // vswp d17, d24
- *q9s16 = vcombine_s16(d18s16, d26s16); // vswp d19, d26
- *q10s16 = vcombine_s16(d20s16, d28s16); // vswp d21, d28
- *q11s16 = vcombine_s16(d22s16, d30s16); // vswp d23, d30
- *q12s16 = vcombine_s16(d17s16, d25s16);
- *q13s16 = vcombine_s16(d19s16, d27s16);
- *q14s16 = vcombine_s16(d21s16, d29s16);
- *q15s16 = vcombine_s16(d23s16, d31s16);
+ *q8s16 = vcombine_s16(d16s16, d24s16); // vswp d17, d24
+ *q9s16 = vcombine_s16(d18s16, d26s16); // vswp d19, d26
+ *q10s16 = vcombine_s16(d20s16, d28s16); // vswp d21, d28
+ *q11s16 = vcombine_s16(d22s16, d30s16); // vswp d23, d30
+ *q12s16 = vcombine_s16(d17s16, d25s16);
+ *q13s16 = vcombine_s16(d19s16, d27s16);
+ *q14s16 = vcombine_s16(d21s16, d29s16);
+ *q15s16 = vcombine_s16(d23s16, d31s16);
- q0x2s32 = vtrnq_s32(vreinterpretq_s32_s16(*q8s16),
- vreinterpretq_s32_s16(*q10s16));
- q1x2s32 = vtrnq_s32(vreinterpretq_s32_s16(*q9s16),
- vreinterpretq_s32_s16(*q11s16));
- q2x2s32 = vtrnq_s32(vreinterpretq_s32_s16(*q12s16),
- vreinterpretq_s32_s16(*q14s16));
- q3x2s32 = vtrnq_s32(vreinterpretq_s32_s16(*q13s16),
- vreinterpretq_s32_s16(*q15s16));
+ q0x2s32 =
+ vtrnq_s32(vreinterpretq_s32_s16(*q8s16), vreinterpretq_s32_s16(*q10s16));
+ q1x2s32 =
+ vtrnq_s32(vreinterpretq_s32_s16(*q9s16), vreinterpretq_s32_s16(*q11s16));
+ q2x2s32 =
+ vtrnq_s32(vreinterpretq_s32_s16(*q12s16), vreinterpretq_s32_s16(*q14s16));
+ q3x2s32 =
+ vtrnq_s32(vreinterpretq_s32_s16(*q13s16), vreinterpretq_s32_s16(*q15s16));
- q0x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q0x2s32.val[0]), // q8
- vreinterpretq_s16_s32(q1x2s32.val[0])); // q9
- q1x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q0x2s32.val[1]), // q10
- vreinterpretq_s16_s32(q1x2s32.val[1])); // q11
- q2x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q2x2s32.val[0]), // q12
- vreinterpretq_s16_s32(q3x2s32.val[0])); // q13
- q3x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q2x2s32.val[1]), // q14
- vreinterpretq_s16_s32(q3x2s32.val[1])); // q15
+ q0x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q0x2s32.val[0]), // q8
+ vreinterpretq_s16_s32(q1x2s32.val[0])); // q9
+ q1x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q0x2s32.val[1]), // q10
+ vreinterpretq_s16_s32(q1x2s32.val[1])); // q11
+ q2x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q2x2s32.val[0]), // q12
+ vreinterpretq_s16_s32(q3x2s32.val[0])); // q13
+ q3x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q2x2s32.val[1]), // q14
+ vreinterpretq_s16_s32(q3x2s32.val[1])); // q15
- *q8s16 = q0x2s16.val[0];
- *q9s16 = q0x2s16.val[1];
- *q10s16 = q1x2s16.val[0];
- *q11s16 = q1x2s16.val[1];
- *q12s16 = q2x2s16.val[0];
- *q13s16 = q2x2s16.val[1];
- *q14s16 = q3x2s16.val[0];
- *q15s16 = q3x2s16.val[1];
- return;
+ *q8s16 = q0x2s16.val[0];
+ *q9s16 = q0x2s16.val[1];
+ *q10s16 = q1x2s16.val[0];
+ *q11s16 = q1x2s16.val[1];
+ *q12s16 = q2x2s16.val[0];
+ *q13s16 = q2x2s16.val[1];
+ *q14s16 = q3x2s16.val[0];
+ *q15s16 = q3x2s16.val[1];
+ return;
}
-static INLINE void IDCT8x8_1D(
- int16x8_t *q8s16,
- int16x8_t *q9s16,
- int16x8_t *q10s16,
- int16x8_t *q11s16,
- int16x8_t *q12s16,
- int16x8_t *q13s16,
- int16x8_t *q14s16,
- int16x8_t *q15s16) {
- int16x4_t d0s16, d1s16, d2s16, d3s16;
- int16x4_t d8s16, d9s16, d10s16, d11s16, d12s16, d13s16, d14s16, d15s16;
- int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
- int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
- int16x8_t q0s16, q1s16, q2s16, q3s16, q4s16, q5s16, q6s16, q7s16;
- int32x4_t q2s32, q3s32, q5s32, q6s32, q8s32, q9s32;
- int32x4_t q10s32, q11s32, q12s32, q13s32, q15s32;
+static INLINE void IDCT8x8_1D(int16x8_t *q8s16, int16x8_t *q9s16,
+ int16x8_t *q10s16, int16x8_t *q11s16,
+ int16x8_t *q12s16, int16x8_t *q13s16,
+ int16x8_t *q14s16, int16x8_t *q15s16) {
+ int16x4_t d0s16, d1s16, d2s16, d3s16;
+ int16x4_t d8s16, d9s16, d10s16, d11s16, d12s16, d13s16, d14s16, d15s16;
+ int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
+ int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
+ int16x8_t q0s16, q1s16, q2s16, q3s16, q4s16, q5s16, q6s16, q7s16;
+ int32x4_t q2s32, q3s32, q5s32, q6s32, q8s32, q9s32;
+ int32x4_t q10s32, q11s32, q12s32, q13s32, q15s32;
- d0s16 = vdup_n_s16(cospi_28_64);
- d1s16 = vdup_n_s16(cospi_4_64);
- d2s16 = vdup_n_s16(cospi_12_64);
- d3s16 = vdup_n_s16(cospi_20_64);
+ d0s16 = vdup_n_s16(cospi_28_64);
+ d1s16 = vdup_n_s16(cospi_4_64);
+ d2s16 = vdup_n_s16(cospi_12_64);
+ d3s16 = vdup_n_s16(cospi_20_64);
- d16s16 = vget_low_s16(*q8s16);
- d17s16 = vget_high_s16(*q8s16);
- d18s16 = vget_low_s16(*q9s16);
- d19s16 = vget_high_s16(*q9s16);
- d20s16 = vget_low_s16(*q10s16);
- d21s16 = vget_high_s16(*q10s16);
- d22s16 = vget_low_s16(*q11s16);
- d23s16 = vget_high_s16(*q11s16);
- d24s16 = vget_low_s16(*q12s16);
- d25s16 = vget_high_s16(*q12s16);
- d26s16 = vget_low_s16(*q13s16);
- d27s16 = vget_high_s16(*q13s16);
- d28s16 = vget_low_s16(*q14s16);
- d29s16 = vget_high_s16(*q14s16);
- d30s16 = vget_low_s16(*q15s16);
- d31s16 = vget_high_s16(*q15s16);
+ d16s16 = vget_low_s16(*q8s16);
+ d17s16 = vget_high_s16(*q8s16);
+ d18s16 = vget_low_s16(*q9s16);
+ d19s16 = vget_high_s16(*q9s16);
+ d20s16 = vget_low_s16(*q10s16);
+ d21s16 = vget_high_s16(*q10s16);
+ d22s16 = vget_low_s16(*q11s16);
+ d23s16 = vget_high_s16(*q11s16);
+ d24s16 = vget_low_s16(*q12s16);
+ d25s16 = vget_high_s16(*q12s16);
+ d26s16 = vget_low_s16(*q13s16);
+ d27s16 = vget_high_s16(*q13s16);
+ d28s16 = vget_low_s16(*q14s16);
+ d29s16 = vget_high_s16(*q14s16);
+ d30s16 = vget_low_s16(*q15s16);
+ d31s16 = vget_high_s16(*q15s16);
- q2s32 = vmull_s16(d18s16, d0s16);
- q3s32 = vmull_s16(d19s16, d0s16);
- q5s32 = vmull_s16(d26s16, d2s16);
- q6s32 = vmull_s16(d27s16, d2s16);
+ q2s32 = vmull_s16(d18s16, d0s16);
+ q3s32 = vmull_s16(d19s16, d0s16);
+ q5s32 = vmull_s16(d26s16, d2s16);
+ q6s32 = vmull_s16(d27s16, d2s16);
- q2s32 = vmlsl_s16(q2s32, d30s16, d1s16);
- q3s32 = vmlsl_s16(q3s32, d31s16, d1s16);
- q5s32 = vmlsl_s16(q5s32, d22s16, d3s16);
- q6s32 = vmlsl_s16(q6s32, d23s16, d3s16);
+ q2s32 = vmlsl_s16(q2s32, d30s16, d1s16);
+ q3s32 = vmlsl_s16(q3s32, d31s16, d1s16);
+ q5s32 = vmlsl_s16(q5s32, d22s16, d3s16);
+ q6s32 = vmlsl_s16(q6s32, d23s16, d3s16);
- d8s16 = vqrshrn_n_s32(q2s32, 14);
- d9s16 = vqrshrn_n_s32(q3s32, 14);
- d10s16 = vqrshrn_n_s32(q5s32, 14);
- d11s16 = vqrshrn_n_s32(q6s32, 14);
- q4s16 = vcombine_s16(d8s16, d9s16);
- q5s16 = vcombine_s16(d10s16, d11s16);
+ d8s16 = vqrshrn_n_s32(q2s32, 14);
+ d9s16 = vqrshrn_n_s32(q3s32, 14);
+ d10s16 = vqrshrn_n_s32(q5s32, 14);
+ d11s16 = vqrshrn_n_s32(q6s32, 14);
+ q4s16 = vcombine_s16(d8s16, d9s16);
+ q5s16 = vcombine_s16(d10s16, d11s16);
- q2s32 = vmull_s16(d18s16, d1s16);
- q3s32 = vmull_s16(d19s16, d1s16);
- q9s32 = vmull_s16(d26s16, d3s16);
- q13s32 = vmull_s16(d27s16, d3s16);
+ q2s32 = vmull_s16(d18s16, d1s16);
+ q3s32 = vmull_s16(d19s16, d1s16);
+ q9s32 = vmull_s16(d26s16, d3s16);
+ q13s32 = vmull_s16(d27s16, d3s16);
- q2s32 = vmlal_s16(q2s32, d30s16, d0s16);
- q3s32 = vmlal_s16(q3s32, d31s16, d0s16);
- q9s32 = vmlal_s16(q9s32, d22s16, d2s16);
- q13s32 = vmlal_s16(q13s32, d23s16, d2s16);
+ q2s32 = vmlal_s16(q2s32, d30s16, d0s16);
+ q3s32 = vmlal_s16(q3s32, d31s16, d0s16);
+ q9s32 = vmlal_s16(q9s32, d22s16, d2s16);
+ q13s32 = vmlal_s16(q13s32, d23s16, d2s16);
- d14s16 = vqrshrn_n_s32(q2s32, 14);
- d15s16 = vqrshrn_n_s32(q3s32, 14);
- d12s16 = vqrshrn_n_s32(q9s32, 14);
- d13s16 = vqrshrn_n_s32(q13s32, 14);
- q6s16 = vcombine_s16(d12s16, d13s16);
- q7s16 = vcombine_s16(d14s16, d15s16);
+ d14s16 = vqrshrn_n_s32(q2s32, 14);
+ d15s16 = vqrshrn_n_s32(q3s32, 14);
+ d12s16 = vqrshrn_n_s32(q9s32, 14);
+ d13s16 = vqrshrn_n_s32(q13s32, 14);
+ q6s16 = vcombine_s16(d12s16, d13s16);
+ q7s16 = vcombine_s16(d14s16, d15s16);
- d0s16 = vdup_n_s16(cospi_16_64);
+ d0s16 = vdup_n_s16(cospi_16_64);
- q2s32 = vmull_s16(d16s16, d0s16);
- q3s32 = vmull_s16(d17s16, d0s16);
- q13s32 = vmull_s16(d16s16, d0s16);
- q15s32 = vmull_s16(d17s16, d0s16);
+ q2s32 = vmull_s16(d16s16, d0s16);
+ q3s32 = vmull_s16(d17s16, d0s16);
+ q13s32 = vmull_s16(d16s16, d0s16);
+ q15s32 = vmull_s16(d17s16, d0s16);
- q2s32 = vmlal_s16(q2s32, d24s16, d0s16);
- q3s32 = vmlal_s16(q3s32, d25s16, d0s16);
- q13s32 = vmlsl_s16(q13s32, d24s16, d0s16);
- q15s32 = vmlsl_s16(q15s32, d25s16, d0s16);
+ q2s32 = vmlal_s16(q2s32, d24s16, d0s16);
+ q3s32 = vmlal_s16(q3s32, d25s16, d0s16);
+ q13s32 = vmlsl_s16(q13s32, d24s16, d0s16);
+ q15s32 = vmlsl_s16(q15s32, d25s16, d0s16);
- d0s16 = vdup_n_s16(cospi_24_64);
- d1s16 = vdup_n_s16(cospi_8_64);
+ d0s16 = vdup_n_s16(cospi_24_64);
+ d1s16 = vdup_n_s16(cospi_8_64);
- d18s16 = vqrshrn_n_s32(q2s32, 14);
- d19s16 = vqrshrn_n_s32(q3s32, 14);
- d22s16 = vqrshrn_n_s32(q13s32, 14);
- d23s16 = vqrshrn_n_s32(q15s32, 14);
- *q9s16 = vcombine_s16(d18s16, d19s16);
- *q11s16 = vcombine_s16(d22s16, d23s16);
+ d18s16 = vqrshrn_n_s32(q2s32, 14);
+ d19s16 = vqrshrn_n_s32(q3s32, 14);
+ d22s16 = vqrshrn_n_s32(q13s32, 14);
+ d23s16 = vqrshrn_n_s32(q15s32, 14);
+ *q9s16 = vcombine_s16(d18s16, d19s16);
+ *q11s16 = vcombine_s16(d22s16, d23s16);
- q2s32 = vmull_s16(d20s16, d0s16);
- q3s32 = vmull_s16(d21s16, d0s16);
- q8s32 = vmull_s16(d20s16, d1s16);
- q12s32 = vmull_s16(d21s16, d1s16);
+ q2s32 = vmull_s16(d20s16, d0s16);
+ q3s32 = vmull_s16(d21s16, d0s16);
+ q8s32 = vmull_s16(d20s16, d1s16);
+ q12s32 = vmull_s16(d21s16, d1s16);
- q2s32 = vmlsl_s16(q2s32, d28s16, d1s16);
- q3s32 = vmlsl_s16(q3s32, d29s16, d1s16);
- q8s32 = vmlal_s16(q8s32, d28s16, d0s16);
- q12s32 = vmlal_s16(q12s32, d29s16, d0s16);
+ q2s32 = vmlsl_s16(q2s32, d28s16, d1s16);
+ q3s32 = vmlsl_s16(q3s32, d29s16, d1s16);
+ q8s32 = vmlal_s16(q8s32, d28s16, d0s16);
+ q12s32 = vmlal_s16(q12s32, d29s16, d0s16);
- d26s16 = vqrshrn_n_s32(q2s32, 14);
- d27s16 = vqrshrn_n_s32(q3s32, 14);
- d30s16 = vqrshrn_n_s32(q8s32, 14);
- d31s16 = vqrshrn_n_s32(q12s32, 14);
- *q13s16 = vcombine_s16(d26s16, d27s16);
- *q15s16 = vcombine_s16(d30s16, d31s16);
+ d26s16 = vqrshrn_n_s32(q2s32, 14);
+ d27s16 = vqrshrn_n_s32(q3s32, 14);
+ d30s16 = vqrshrn_n_s32(q8s32, 14);
+ d31s16 = vqrshrn_n_s32(q12s32, 14);
+ *q13s16 = vcombine_s16(d26s16, d27s16);
+ *q15s16 = vcombine_s16(d30s16, d31s16);
- q0s16 = vaddq_s16(*q9s16, *q15s16);
- q1s16 = vaddq_s16(*q11s16, *q13s16);
- q2s16 = vsubq_s16(*q11s16, *q13s16);
- q3s16 = vsubq_s16(*q9s16, *q15s16);
+ q0s16 = vaddq_s16(*q9s16, *q15s16);
+ q1s16 = vaddq_s16(*q11s16, *q13s16);
+ q2s16 = vsubq_s16(*q11s16, *q13s16);
+ q3s16 = vsubq_s16(*q9s16, *q15s16);
- *q13s16 = vsubq_s16(q4s16, q5s16);
- q4s16 = vaddq_s16(q4s16, q5s16);
- *q14s16 = vsubq_s16(q7s16, q6s16);
- q7s16 = vaddq_s16(q7s16, q6s16);
- d26s16 = vget_low_s16(*q13s16);
- d27s16 = vget_high_s16(*q13s16);
- d28s16 = vget_low_s16(*q14s16);
- d29s16 = vget_high_s16(*q14s16);
+ *q13s16 = vsubq_s16(q4s16, q5s16);
+ q4s16 = vaddq_s16(q4s16, q5s16);
+ *q14s16 = vsubq_s16(q7s16, q6s16);
+ q7s16 = vaddq_s16(q7s16, q6s16);
+ d26s16 = vget_low_s16(*q13s16);
+ d27s16 = vget_high_s16(*q13s16);
+ d28s16 = vget_low_s16(*q14s16);
+ d29s16 = vget_high_s16(*q14s16);
- d16s16 = vdup_n_s16(cospi_16_64);
+ d16s16 = vdup_n_s16(cospi_16_64);
- q9s32 = vmull_s16(d28s16, d16s16);
- q10s32 = vmull_s16(d29s16, d16s16);
- q11s32 = vmull_s16(d28s16, d16s16);
- q12s32 = vmull_s16(d29s16, d16s16);
+ q9s32 = vmull_s16(d28s16, d16s16);
+ q10s32 = vmull_s16(d29s16, d16s16);
+ q11s32 = vmull_s16(d28s16, d16s16);
+ q12s32 = vmull_s16(d29s16, d16s16);
- q9s32 = vmlsl_s16(q9s32, d26s16, d16s16);
- q10s32 = vmlsl_s16(q10s32, d27s16, d16s16);
- q11s32 = vmlal_s16(q11s32, d26s16, d16s16);
- q12s32 = vmlal_s16(q12s32, d27s16, d16s16);
+ q9s32 = vmlsl_s16(q9s32, d26s16, d16s16);
+ q10s32 = vmlsl_s16(q10s32, d27s16, d16s16);
+ q11s32 = vmlal_s16(q11s32, d26s16, d16s16);
+ q12s32 = vmlal_s16(q12s32, d27s16, d16s16);
- d10s16 = vqrshrn_n_s32(q9s32, 14);
- d11s16 = vqrshrn_n_s32(q10s32, 14);
- d12s16 = vqrshrn_n_s32(q11s32, 14);
- d13s16 = vqrshrn_n_s32(q12s32, 14);
- q5s16 = vcombine_s16(d10s16, d11s16);
- q6s16 = vcombine_s16(d12s16, d13s16);
+ d10s16 = vqrshrn_n_s32(q9s32, 14);
+ d11s16 = vqrshrn_n_s32(q10s32, 14);
+ d12s16 = vqrshrn_n_s32(q11s32, 14);
+ d13s16 = vqrshrn_n_s32(q12s32, 14);
+ q5s16 = vcombine_s16(d10s16, d11s16);
+ q6s16 = vcombine_s16(d12s16, d13s16);
- *q8s16 = vaddq_s16(q0s16, q7s16);
- *q9s16 = vaddq_s16(q1s16, q6s16);
- *q10s16 = vaddq_s16(q2s16, q5s16);
- *q11s16 = vaddq_s16(q3s16, q4s16);
- *q12s16 = vsubq_s16(q3s16, q4s16);
- *q13s16 = vsubq_s16(q2s16, q5s16);
- *q14s16 = vsubq_s16(q1s16, q6s16);
- *q15s16 = vsubq_s16(q0s16, q7s16);
- return;
+ *q8s16 = vaddq_s16(q0s16, q7s16);
+ *q9s16 = vaddq_s16(q1s16, q6s16);
+ *q10s16 = vaddq_s16(q2s16, q5s16);
+ *q11s16 = vaddq_s16(q3s16, q4s16);
+ *q12s16 = vsubq_s16(q3s16, q4s16);
+ *q13s16 = vsubq_s16(q2s16, q5s16);
+ *q14s16 = vsubq_s16(q1s16, q6s16);
+ *q15s16 = vsubq_s16(q0s16, q7s16);
+ return;
}
-static INLINE void IADST8X8_1D(
- int16x8_t *q8s16,
- int16x8_t *q9s16,
- int16x8_t *q10s16,
- int16x8_t *q11s16,
- int16x8_t *q12s16,
- int16x8_t *q13s16,
- int16x8_t *q14s16,
- int16x8_t *q15s16) {
- int16x4_t d0s16, d1s16, d2s16, d3s16, d4s16, d5s16, d6s16, d7s16;
- int16x4_t d8s16, d9s16, d10s16, d11s16, d12s16, d13s16, d14s16, d15s16;
- int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
- int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
- int16x8_t q2s16, q4s16, q5s16, q6s16;
- int32x4_t q0s32, q1s32, q2s32, q3s32, q4s32, q5s32, q6s32, q7s32, q8s32;
- int32x4_t q9s32, q10s32, q11s32, q12s32, q13s32, q14s32, q15s32;
+static INLINE void IADST8X8_1D(int16x8_t *q8s16, int16x8_t *q9s16,
+ int16x8_t *q10s16, int16x8_t *q11s16,
+ int16x8_t *q12s16, int16x8_t *q13s16,
+ int16x8_t *q14s16, int16x8_t *q15s16) {
+ int16x4_t d0s16, d1s16, d2s16, d3s16, d4s16, d5s16, d6s16, d7s16;
+ int16x4_t d8s16, d9s16, d10s16, d11s16, d12s16, d13s16, d14s16, d15s16;
+ int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
+ int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
+ int16x8_t q2s16, q4s16, q5s16, q6s16;
+ int32x4_t q0s32, q1s32, q2s32, q3s32, q4s32, q5s32, q6s32, q7s32, q8s32;
+ int32x4_t q9s32, q10s32, q11s32, q12s32, q13s32, q14s32, q15s32;
- d16s16 = vget_low_s16(*q8s16);
- d17s16 = vget_high_s16(*q8s16);
- d18s16 = vget_low_s16(*q9s16);
- d19s16 = vget_high_s16(*q9s16);
- d20s16 = vget_low_s16(*q10s16);
- d21s16 = vget_high_s16(*q10s16);
- d22s16 = vget_low_s16(*q11s16);
- d23s16 = vget_high_s16(*q11s16);
- d24s16 = vget_low_s16(*q12s16);
- d25s16 = vget_high_s16(*q12s16);
- d26s16 = vget_low_s16(*q13s16);
- d27s16 = vget_high_s16(*q13s16);
- d28s16 = vget_low_s16(*q14s16);
- d29s16 = vget_high_s16(*q14s16);
- d30s16 = vget_low_s16(*q15s16);
- d31s16 = vget_high_s16(*q15s16);
+ d16s16 = vget_low_s16(*q8s16);
+ d17s16 = vget_high_s16(*q8s16);
+ d18s16 = vget_low_s16(*q9s16);
+ d19s16 = vget_high_s16(*q9s16);
+ d20s16 = vget_low_s16(*q10s16);
+ d21s16 = vget_high_s16(*q10s16);
+ d22s16 = vget_low_s16(*q11s16);
+ d23s16 = vget_high_s16(*q11s16);
+ d24s16 = vget_low_s16(*q12s16);
+ d25s16 = vget_high_s16(*q12s16);
+ d26s16 = vget_low_s16(*q13s16);
+ d27s16 = vget_high_s16(*q13s16);
+ d28s16 = vget_low_s16(*q14s16);
+ d29s16 = vget_high_s16(*q14s16);
+ d30s16 = vget_low_s16(*q15s16);
+ d31s16 = vget_high_s16(*q15s16);
- d14s16 = vdup_n_s16(cospi_2_64);
- d15s16 = vdup_n_s16(cospi_30_64);
+ d14s16 = vdup_n_s16(cospi_2_64);
+ d15s16 = vdup_n_s16(cospi_30_64);
- q1s32 = vmull_s16(d30s16, d14s16);
- q2s32 = vmull_s16(d31s16, d14s16);
- q3s32 = vmull_s16(d30s16, d15s16);
- q4s32 = vmull_s16(d31s16, d15s16);
+ q1s32 = vmull_s16(d30s16, d14s16);
+ q2s32 = vmull_s16(d31s16, d14s16);
+ q3s32 = vmull_s16(d30s16, d15s16);
+ q4s32 = vmull_s16(d31s16, d15s16);
- d30s16 = vdup_n_s16(cospi_18_64);
- d31s16 = vdup_n_s16(cospi_14_64);
+ d30s16 = vdup_n_s16(cospi_18_64);
+ d31s16 = vdup_n_s16(cospi_14_64);
- q1s32 = vmlal_s16(q1s32, d16s16, d15s16);
- q2s32 = vmlal_s16(q2s32, d17s16, d15s16);
- q3s32 = vmlsl_s16(q3s32, d16s16, d14s16);
- q4s32 = vmlsl_s16(q4s32, d17s16, d14s16);
+ q1s32 = vmlal_s16(q1s32, d16s16, d15s16);
+ q2s32 = vmlal_s16(q2s32, d17s16, d15s16);
+ q3s32 = vmlsl_s16(q3s32, d16s16, d14s16);
+ q4s32 = vmlsl_s16(q4s32, d17s16, d14s16);
- q5s32 = vmull_s16(d22s16, d30s16);
- q6s32 = vmull_s16(d23s16, d30s16);
- q7s32 = vmull_s16(d22s16, d31s16);
- q8s32 = vmull_s16(d23s16, d31s16);
+ q5s32 = vmull_s16(d22s16, d30s16);
+ q6s32 = vmull_s16(d23s16, d30s16);
+ q7s32 = vmull_s16(d22s16, d31s16);
+ q8s32 = vmull_s16(d23s16, d31s16);
- q5s32 = vmlal_s16(q5s32, d24s16, d31s16);
- q6s32 = vmlal_s16(q6s32, d25s16, d31s16);
- q7s32 = vmlsl_s16(q7s32, d24s16, d30s16);
- q8s32 = vmlsl_s16(q8s32, d25s16, d30s16);
+ q5s32 = vmlal_s16(q5s32, d24s16, d31s16);
+ q6s32 = vmlal_s16(q6s32, d25s16, d31s16);
+ q7s32 = vmlsl_s16(q7s32, d24s16, d30s16);
+ q8s32 = vmlsl_s16(q8s32, d25s16, d30s16);
- q11s32 = vaddq_s32(q1s32, q5s32);
- q12s32 = vaddq_s32(q2s32, q6s32);
- q1s32 = vsubq_s32(q1s32, q5s32);
- q2s32 = vsubq_s32(q2s32, q6s32);
+ q11s32 = vaddq_s32(q1s32, q5s32);
+ q12s32 = vaddq_s32(q2s32, q6s32);
+ q1s32 = vsubq_s32(q1s32, q5s32);
+ q2s32 = vsubq_s32(q2s32, q6s32);
- d22s16 = vqrshrn_n_s32(q11s32, 14);
- d23s16 = vqrshrn_n_s32(q12s32, 14);
- *q11s16 = vcombine_s16(d22s16, d23s16);
+ d22s16 = vqrshrn_n_s32(q11s32, 14);
+ d23s16 = vqrshrn_n_s32(q12s32, 14);
+ *q11s16 = vcombine_s16(d22s16, d23s16);
- q12s32 = vaddq_s32(q3s32, q7s32);
- q15s32 = vaddq_s32(q4s32, q8s32);
- q3s32 = vsubq_s32(q3s32, q7s32);
- q4s32 = vsubq_s32(q4s32, q8s32);
+ q12s32 = vaddq_s32(q3s32, q7s32);
+ q15s32 = vaddq_s32(q4s32, q8s32);
+ q3s32 = vsubq_s32(q3s32, q7s32);
+ q4s32 = vsubq_s32(q4s32, q8s32);
- d2s16 = vqrshrn_n_s32(q1s32, 14);
- d3s16 = vqrshrn_n_s32(q2s32, 14);
- d24s16 = vqrshrn_n_s32(q12s32, 14);
- d25s16 = vqrshrn_n_s32(q15s32, 14);
- d6s16 = vqrshrn_n_s32(q3s32, 14);
- d7s16 = vqrshrn_n_s32(q4s32, 14);
- *q12s16 = vcombine_s16(d24s16, d25s16);
+ d2s16 = vqrshrn_n_s32(q1s32, 14);
+ d3s16 = vqrshrn_n_s32(q2s32, 14);
+ d24s16 = vqrshrn_n_s32(q12s32, 14);
+ d25s16 = vqrshrn_n_s32(q15s32, 14);
+ d6s16 = vqrshrn_n_s32(q3s32, 14);
+ d7s16 = vqrshrn_n_s32(q4s32, 14);
+ *q12s16 = vcombine_s16(d24s16, d25s16);
- d0s16 = vdup_n_s16(cospi_10_64);
- d1s16 = vdup_n_s16(cospi_22_64);
- q4s32 = vmull_s16(d26s16, d0s16);
- q5s32 = vmull_s16(d27s16, d0s16);
- q2s32 = vmull_s16(d26s16, d1s16);
- q6s32 = vmull_s16(d27s16, d1s16);
+ d0s16 = vdup_n_s16(cospi_10_64);
+ d1s16 = vdup_n_s16(cospi_22_64);
+ q4s32 = vmull_s16(d26s16, d0s16);
+ q5s32 = vmull_s16(d27s16, d0s16);
+ q2s32 = vmull_s16(d26s16, d1s16);
+ q6s32 = vmull_s16(d27s16, d1s16);
- d30s16 = vdup_n_s16(cospi_26_64);
- d31s16 = vdup_n_s16(cospi_6_64);
+ d30s16 = vdup_n_s16(cospi_26_64);
+ d31s16 = vdup_n_s16(cospi_6_64);
- q4s32 = vmlal_s16(q4s32, d20s16, d1s16);
- q5s32 = vmlal_s16(q5s32, d21s16, d1s16);
- q2s32 = vmlsl_s16(q2s32, d20s16, d0s16);
- q6s32 = vmlsl_s16(q6s32, d21s16, d0s16);
+ q4s32 = vmlal_s16(q4s32, d20s16, d1s16);
+ q5s32 = vmlal_s16(q5s32, d21s16, d1s16);
+ q2s32 = vmlsl_s16(q2s32, d20s16, d0s16);
+ q6s32 = vmlsl_s16(q6s32, d21s16, d0s16);
- q0s32 = vmull_s16(d18s16, d30s16);
- q13s32 = vmull_s16(d19s16, d30s16);
+ q0s32 = vmull_s16(d18s16, d30s16);
+ q13s32 = vmull_s16(d19s16, d30s16);
- q0s32 = vmlal_s16(q0s32, d28s16, d31s16);
- q13s32 = vmlal_s16(q13s32, d29s16, d31s16);
+ q0s32 = vmlal_s16(q0s32, d28s16, d31s16);
+ q13s32 = vmlal_s16(q13s32, d29s16, d31s16);
- q10s32 = vmull_s16(d18s16, d31s16);
- q9s32 = vmull_s16(d19s16, d31s16);
+ q10s32 = vmull_s16(d18s16, d31s16);
+ q9s32 = vmull_s16(d19s16, d31s16);
- q10s32 = vmlsl_s16(q10s32, d28s16, d30s16);
- q9s32 = vmlsl_s16(q9s32, d29s16, d30s16);
+ q10s32 = vmlsl_s16(q10s32, d28s16, d30s16);
+ q9s32 = vmlsl_s16(q9s32, d29s16, d30s16);
- q14s32 = vaddq_s32(q2s32, q10s32);
- q15s32 = vaddq_s32(q6s32, q9s32);
- q2s32 = vsubq_s32(q2s32, q10s32);
- q6s32 = vsubq_s32(q6s32, q9s32);
+ q14s32 = vaddq_s32(q2s32, q10s32);
+ q15s32 = vaddq_s32(q6s32, q9s32);
+ q2s32 = vsubq_s32(q2s32, q10s32);
+ q6s32 = vsubq_s32(q6s32, q9s32);
- d28s16 = vqrshrn_n_s32(q14s32, 14);
- d29s16 = vqrshrn_n_s32(q15s32, 14);
- d4s16 = vqrshrn_n_s32(q2s32, 14);
- d5s16 = vqrshrn_n_s32(q6s32, 14);
- *q14s16 = vcombine_s16(d28s16, d29s16);
+ d28s16 = vqrshrn_n_s32(q14s32, 14);
+ d29s16 = vqrshrn_n_s32(q15s32, 14);
+ d4s16 = vqrshrn_n_s32(q2s32, 14);
+ d5s16 = vqrshrn_n_s32(q6s32, 14);
+ *q14s16 = vcombine_s16(d28s16, d29s16);
- q9s32 = vaddq_s32(q4s32, q0s32);
- q10s32 = vaddq_s32(q5s32, q13s32);
- q4s32 = vsubq_s32(q4s32, q0s32);
- q5s32 = vsubq_s32(q5s32, q13s32);
+ q9s32 = vaddq_s32(q4s32, q0s32);
+ q10s32 = vaddq_s32(q5s32, q13s32);
+ q4s32 = vsubq_s32(q4s32, q0s32);
+ q5s32 = vsubq_s32(q5s32, q13s32);
- d30s16 = vdup_n_s16(cospi_8_64);
- d31s16 = vdup_n_s16(cospi_24_64);
+ d30s16 = vdup_n_s16(cospi_8_64);
+ d31s16 = vdup_n_s16(cospi_24_64);
- d18s16 = vqrshrn_n_s32(q9s32, 14);
- d19s16 = vqrshrn_n_s32(q10s32, 14);
- d8s16 = vqrshrn_n_s32(q4s32, 14);
- d9s16 = vqrshrn_n_s32(q5s32, 14);
- *q9s16 = vcombine_s16(d18s16, d19s16);
+ d18s16 = vqrshrn_n_s32(q9s32, 14);
+ d19s16 = vqrshrn_n_s32(q10s32, 14);
+ d8s16 = vqrshrn_n_s32(q4s32, 14);
+ d9s16 = vqrshrn_n_s32(q5s32, 14);
+ *q9s16 = vcombine_s16(d18s16, d19s16);
- q5s32 = vmull_s16(d2s16, d30s16);
- q6s32 = vmull_s16(d3s16, d30s16);
- q7s32 = vmull_s16(d2s16, d31s16);
- q0s32 = vmull_s16(d3s16, d31s16);
+ q5s32 = vmull_s16(d2s16, d30s16);
+ q6s32 = vmull_s16(d3s16, d30s16);
+ q7s32 = vmull_s16(d2s16, d31s16);
+ q0s32 = vmull_s16(d3s16, d31s16);
- q5s32 = vmlal_s16(q5s32, d6s16, d31s16);
- q6s32 = vmlal_s16(q6s32, d7s16, d31s16);
- q7s32 = vmlsl_s16(q7s32, d6s16, d30s16);
- q0s32 = vmlsl_s16(q0s32, d7s16, d30s16);
+ q5s32 = vmlal_s16(q5s32, d6s16, d31s16);
+ q6s32 = vmlal_s16(q6s32, d7s16, d31s16);
+ q7s32 = vmlsl_s16(q7s32, d6s16, d30s16);
+ q0s32 = vmlsl_s16(q0s32, d7s16, d30s16);
- q1s32 = vmull_s16(d4s16, d30s16);
- q3s32 = vmull_s16(d5s16, d30s16);
- q10s32 = vmull_s16(d4s16, d31s16);
- q2s32 = vmull_s16(d5s16, d31s16);
+ q1s32 = vmull_s16(d4s16, d30s16);
+ q3s32 = vmull_s16(d5s16, d30s16);
+ q10s32 = vmull_s16(d4s16, d31s16);
+ q2s32 = vmull_s16(d5s16, d31s16);
- q1s32 = vmlsl_s16(q1s32, d8s16, d31s16);
- q3s32 = vmlsl_s16(q3s32, d9s16, d31s16);
- q10s32 = vmlal_s16(q10s32, d8s16, d30s16);
- q2s32 = vmlal_s16(q2s32, d9s16, d30s16);
+ q1s32 = vmlsl_s16(q1s32, d8s16, d31s16);
+ q3s32 = vmlsl_s16(q3s32, d9s16, d31s16);
+ q10s32 = vmlal_s16(q10s32, d8s16, d30s16);
+ q2s32 = vmlal_s16(q2s32, d9s16, d30s16);
- *q8s16 = vaddq_s16(*q11s16, *q9s16);
- *q11s16 = vsubq_s16(*q11s16, *q9s16);
- q4s16 = vaddq_s16(*q12s16, *q14s16);
- *q12s16 = vsubq_s16(*q12s16, *q14s16);
+ *q8s16 = vaddq_s16(*q11s16, *q9s16);
+ *q11s16 = vsubq_s16(*q11s16, *q9s16);
+ q4s16 = vaddq_s16(*q12s16, *q14s16);
+ *q12s16 = vsubq_s16(*q12s16, *q14s16);
- q14s32 = vaddq_s32(q5s32, q1s32);
- q15s32 = vaddq_s32(q6s32, q3s32);
- q5s32 = vsubq_s32(q5s32, q1s32);
- q6s32 = vsubq_s32(q6s32, q3s32);
+ q14s32 = vaddq_s32(q5s32, q1s32);
+ q15s32 = vaddq_s32(q6s32, q3s32);
+ q5s32 = vsubq_s32(q5s32, q1s32);
+ q6s32 = vsubq_s32(q6s32, q3s32);
- d18s16 = vqrshrn_n_s32(q14s32, 14);
- d19s16 = vqrshrn_n_s32(q15s32, 14);
- d10s16 = vqrshrn_n_s32(q5s32, 14);
- d11s16 = vqrshrn_n_s32(q6s32, 14);
- *q9s16 = vcombine_s16(d18s16, d19s16);
+ d18s16 = vqrshrn_n_s32(q14s32, 14);
+ d19s16 = vqrshrn_n_s32(q15s32, 14);
+ d10s16 = vqrshrn_n_s32(q5s32, 14);
+ d11s16 = vqrshrn_n_s32(q6s32, 14);
+ *q9s16 = vcombine_s16(d18s16, d19s16);
- q1s32 = vaddq_s32(q7s32, q10s32);
- q3s32 = vaddq_s32(q0s32, q2s32);
- q7s32 = vsubq_s32(q7s32, q10s32);
- q0s32 = vsubq_s32(q0s32, q2s32);
+ q1s32 = vaddq_s32(q7s32, q10s32);
+ q3s32 = vaddq_s32(q0s32, q2s32);
+ q7s32 = vsubq_s32(q7s32, q10s32);
+ q0s32 = vsubq_s32(q0s32, q2s32);
- d28s16 = vqrshrn_n_s32(q1s32, 14);
- d29s16 = vqrshrn_n_s32(q3s32, 14);
- d14s16 = vqrshrn_n_s32(q7s32, 14);
- d15s16 = vqrshrn_n_s32(q0s32, 14);
- *q14s16 = vcombine_s16(d28s16, d29s16);
+ d28s16 = vqrshrn_n_s32(q1s32, 14);
+ d29s16 = vqrshrn_n_s32(q3s32, 14);
+ d14s16 = vqrshrn_n_s32(q7s32, 14);
+ d15s16 = vqrshrn_n_s32(q0s32, 14);
+ *q14s16 = vcombine_s16(d28s16, d29s16);
- d30s16 = vdup_n_s16(cospi_16_64);
+ d30s16 = vdup_n_s16(cospi_16_64);
- d22s16 = vget_low_s16(*q11s16);
- d23s16 = vget_high_s16(*q11s16);
- q2s32 = vmull_s16(d22s16, d30s16);
- q3s32 = vmull_s16(d23s16, d30s16);
- q13s32 = vmull_s16(d22s16, d30s16);
- q1s32 = vmull_s16(d23s16, d30s16);
+ d22s16 = vget_low_s16(*q11s16);
+ d23s16 = vget_high_s16(*q11s16);
+ q2s32 = vmull_s16(d22s16, d30s16);
+ q3s32 = vmull_s16(d23s16, d30s16);
+ q13s32 = vmull_s16(d22s16, d30s16);
+ q1s32 = vmull_s16(d23s16, d30s16);
- d24s16 = vget_low_s16(*q12s16);
- d25s16 = vget_high_s16(*q12s16);
- q2s32 = vmlal_s16(q2s32, d24s16, d30s16);
- q3s32 = vmlal_s16(q3s32, d25s16, d30s16);
- q13s32 = vmlsl_s16(q13s32, d24s16, d30s16);
- q1s32 = vmlsl_s16(q1s32, d25s16, d30s16);
+ d24s16 = vget_low_s16(*q12s16);
+ d25s16 = vget_high_s16(*q12s16);
+ q2s32 = vmlal_s16(q2s32, d24s16, d30s16);
+ q3s32 = vmlal_s16(q3s32, d25s16, d30s16);
+ q13s32 = vmlsl_s16(q13s32, d24s16, d30s16);
+ q1s32 = vmlsl_s16(q1s32, d25s16, d30s16);
- d4s16 = vqrshrn_n_s32(q2s32, 14);
- d5s16 = vqrshrn_n_s32(q3s32, 14);
- d24s16 = vqrshrn_n_s32(q13s32, 14);
- d25s16 = vqrshrn_n_s32(q1s32, 14);
- q2s16 = vcombine_s16(d4s16, d5s16);
- *q12s16 = vcombine_s16(d24s16, d25s16);
+ d4s16 = vqrshrn_n_s32(q2s32, 14);
+ d5s16 = vqrshrn_n_s32(q3s32, 14);
+ d24s16 = vqrshrn_n_s32(q13s32, 14);
+ d25s16 = vqrshrn_n_s32(q1s32, 14);
+ q2s16 = vcombine_s16(d4s16, d5s16);
+ *q12s16 = vcombine_s16(d24s16, d25s16);
- q13s32 = vmull_s16(d10s16, d30s16);
- q1s32 = vmull_s16(d11s16, d30s16);
- q11s32 = vmull_s16(d10s16, d30s16);
- q0s32 = vmull_s16(d11s16, d30s16);
+ q13s32 = vmull_s16(d10s16, d30s16);
+ q1s32 = vmull_s16(d11s16, d30s16);
+ q11s32 = vmull_s16(d10s16, d30s16);
+ q0s32 = vmull_s16(d11s16, d30s16);
- q13s32 = vmlal_s16(q13s32, d14s16, d30s16);
- q1s32 = vmlal_s16(q1s32, d15s16, d30s16);
- q11s32 = vmlsl_s16(q11s32, d14s16, d30s16);
- q0s32 = vmlsl_s16(q0s32, d15s16, d30s16);
+ q13s32 = vmlal_s16(q13s32, d14s16, d30s16);
+ q1s32 = vmlal_s16(q1s32, d15s16, d30s16);
+ q11s32 = vmlsl_s16(q11s32, d14s16, d30s16);
+ q0s32 = vmlsl_s16(q0s32, d15s16, d30s16);
- d20s16 = vqrshrn_n_s32(q13s32, 14);
- d21s16 = vqrshrn_n_s32(q1s32, 14);
- d12s16 = vqrshrn_n_s32(q11s32, 14);
- d13s16 = vqrshrn_n_s32(q0s32, 14);
- *q10s16 = vcombine_s16(d20s16, d21s16);
- q6s16 = vcombine_s16(d12s16, d13s16);
+ d20s16 = vqrshrn_n_s32(q13s32, 14);
+ d21s16 = vqrshrn_n_s32(q1s32, 14);
+ d12s16 = vqrshrn_n_s32(q11s32, 14);
+ d13s16 = vqrshrn_n_s32(q0s32, 14);
+ *q10s16 = vcombine_s16(d20s16, d21s16);
+ q6s16 = vcombine_s16(d12s16, d13s16);
- q5s16 = vdupq_n_s16(0);
+ q5s16 = vdupq_n_s16(0);
- *q9s16 = vsubq_s16(q5s16, *q9s16);
- *q11s16 = vsubq_s16(q5s16, q2s16);
- *q13s16 = vsubq_s16(q5s16, q6s16);
- *q15s16 = vsubq_s16(q5s16, q4s16);
- return;
+ *q9s16 = vsubq_s16(q5s16, *q9s16);
+ *q11s16 = vsubq_s16(q5s16, q2s16);
+ *q13s16 = vsubq_s16(q5s16, q6s16);
+ *q15s16 = vsubq_s16(q5s16, q4s16);
+ return;
}
void vp10_iht8x8_64_add_neon(const tran_low_t *input, uint8_t *dest,
- int dest_stride, int tx_type) {
- int i;
- uint8_t *d1, *d2;
- uint8x8_t d0u8, d1u8, d2u8, d3u8;
- uint64x1_t d0u64, d1u64, d2u64, d3u64;
- int16x8_t q8s16, q9s16, q10s16, q11s16, q12s16, q13s16, q14s16, q15s16;
- uint16x8_t q8u16, q9u16, q10u16, q11u16;
+ int dest_stride, int tx_type) {
+ int i;
+ uint8_t *d1, *d2;
+ uint8x8_t d0u8, d1u8, d2u8, d3u8;
+ uint64x1_t d0u64, d1u64, d2u64, d3u64;
+ int16x8_t q8s16, q9s16, q10s16, q11s16, q12s16, q13s16, q14s16, q15s16;
+ uint16x8_t q8u16, q9u16, q10u16, q11u16;
- q8s16 = vld1q_s16(input);
- q9s16 = vld1q_s16(input + 8);
- q10s16 = vld1q_s16(input + 8 * 2);
- q11s16 = vld1q_s16(input + 8 * 3);
- q12s16 = vld1q_s16(input + 8 * 4);
- q13s16 = vld1q_s16(input + 8 * 5);
- q14s16 = vld1q_s16(input + 8 * 6);
- q15s16 = vld1q_s16(input + 8 * 7);
+ q8s16 = vld1q_s16(input);
+ q9s16 = vld1q_s16(input + 8);
+ q10s16 = vld1q_s16(input + 8 * 2);
+ q11s16 = vld1q_s16(input + 8 * 3);
+ q12s16 = vld1q_s16(input + 8 * 4);
+ q13s16 = vld1q_s16(input + 8 * 5);
+ q14s16 = vld1q_s16(input + 8 * 6);
+ q15s16 = vld1q_s16(input + 8 * 7);
- TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
+ TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
- switch (tx_type) {
- case 0: // idct_idct is not supported. Fall back to C
- vp10_iht8x8_64_add_c(input, dest, dest_stride, tx_type);
- return;
- break;
- case 1: // iadst_idct
- // generate IDCT constants
- // GENERATE_IDCT_CONSTANTS
+ switch (tx_type) {
+ case 0: // idct_idct is not supported. Fall back to C
+ vp10_iht8x8_64_add_c(input, dest, dest_stride, tx_type);
+ return;
+ break;
+ case 1: // iadst_idct
+ // generate IDCT constants
+ // GENERATE_IDCT_CONSTANTS
- // first transform rows
- IDCT8x8_1D(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
+ // first transform rows
+ IDCT8x8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
- // transpose the matrix
- TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
+ // transpose the matrix
+ TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
- // generate IADST constants
- // GENERATE_IADST_CONSTANTS
+ // generate IADST constants
+ // GENERATE_IADST_CONSTANTS
- // then transform columns
- IADST8X8_1D(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
- break;
- case 2: // idct_iadst
- // generate IADST constants
- // GENERATE_IADST_CONSTANTS
+ // then transform columns
+ IADST8X8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
+ break;
+ case 2: // idct_iadst
+ // generate IADST constants
+ // GENERATE_IADST_CONSTANTS
- // first transform rows
- IADST8X8_1D(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
+ // first transform rows
+ IADST8X8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
- // transpose the matrix
- TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
+ // transpose the matrix
+ TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
- // generate IDCT constants
- // GENERATE_IDCT_CONSTANTS
+ // generate IDCT constants
+ // GENERATE_IDCT_CONSTANTS
- // then transform columns
- IDCT8x8_1D(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
- break;
- case 3: // iadst_iadst
- // generate IADST constants
- // GENERATE_IADST_CONSTANTS
+ // then transform columns
+ IDCT8x8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
+ break;
+ case 3: // iadst_iadst
+ // generate IADST constants
+ // GENERATE_IADST_CONSTANTS
- // first transform rows
- IADST8X8_1D(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
+ // first transform rows
+ IADST8X8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
- // transpose the matrix
- TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
+ // transpose the matrix
+ TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
- // then transform columns
- IADST8X8_1D(&q8s16, &q9s16, &q10s16, &q11s16,
- &q12s16, &q13s16, &q14s16, &q15s16);
- break;
- default: // iadst_idct
- assert(0);
- break;
+ // then transform columns
+ IADST8X8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
+ &q15s16);
+ break;
+ default: // iadst_idct
+ assert(0);
+ break;
+ }
+
+ q8s16 = vrshrq_n_s16(q8s16, 5);
+ q9s16 = vrshrq_n_s16(q9s16, 5);
+ q10s16 = vrshrq_n_s16(q10s16, 5);
+ q11s16 = vrshrq_n_s16(q11s16, 5);
+ q12s16 = vrshrq_n_s16(q12s16, 5);
+ q13s16 = vrshrq_n_s16(q13s16, 5);
+ q14s16 = vrshrq_n_s16(q14s16, 5);
+ q15s16 = vrshrq_n_s16(q15s16, 5);
+
+ for (d1 = d2 = dest, i = 0; i < 2; i++) {
+ if (i != 0) {
+ q8s16 = q12s16;
+ q9s16 = q13s16;
+ q10s16 = q14s16;
+ q11s16 = q15s16;
}
- q8s16 = vrshrq_n_s16(q8s16, 5);
- q9s16 = vrshrq_n_s16(q9s16, 5);
- q10s16 = vrshrq_n_s16(q10s16, 5);
- q11s16 = vrshrq_n_s16(q11s16, 5);
- q12s16 = vrshrq_n_s16(q12s16, 5);
- q13s16 = vrshrq_n_s16(q13s16, 5);
- q14s16 = vrshrq_n_s16(q14s16, 5);
- q15s16 = vrshrq_n_s16(q15s16, 5);
+ d0u64 = vld1_u64((uint64_t *)d1);
+ d1 += dest_stride;
+ d1u64 = vld1_u64((uint64_t *)d1);
+ d1 += dest_stride;
+ d2u64 = vld1_u64((uint64_t *)d1);
+ d1 += dest_stride;
+ d3u64 = vld1_u64((uint64_t *)d1);
+ d1 += dest_stride;
- for (d1 = d2 = dest, i = 0; i < 2; i++) {
- if (i != 0) {
- q8s16 = q12s16;
- q9s16 = q13s16;
- q10s16 = q14s16;
- q11s16 = q15s16;
- }
+ q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_u64(d0u64));
+ q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_u64(d1u64));
+ q10u16 =
+ vaddw_u8(vreinterpretq_u16_s16(q10s16), vreinterpret_u8_u64(d2u64));
+ q11u16 =
+ vaddw_u8(vreinterpretq_u16_s16(q11s16), vreinterpret_u8_u64(d3u64));
- d0u64 = vld1_u64((uint64_t *)d1);
- d1 += dest_stride;
- d1u64 = vld1_u64((uint64_t *)d1);
- d1 += dest_stride;
- d2u64 = vld1_u64((uint64_t *)d1);
- d1 += dest_stride;
- d3u64 = vld1_u64((uint64_t *)d1);
- d1 += dest_stride;
+ d0u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
+ d1u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
+ d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
+ d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
- q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16),
- vreinterpret_u8_u64(d0u64));
- q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16),
- vreinterpret_u8_u64(d1u64));
- q10u16 = vaddw_u8(vreinterpretq_u16_s16(q10s16),
- vreinterpret_u8_u64(d2u64));
- q11u16 = vaddw_u8(vreinterpretq_u16_s16(q11s16),
- vreinterpret_u8_u64(d3u64));
-
- d0u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
- d1u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
- d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
- d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
-
- vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d0u8));
- d2 += dest_stride;
- vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d1u8));
- d2 += dest_stride;
- vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
- d2 += dest_stride;
- vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
- d2 += dest_stride;
- }
- return;
+ vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d0u8));
+ d2 += dest_stride;
+ vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d1u8));
+ d2 += dest_stride;
+ vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
+ d2 += dest_stride;
+ vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
+ d2 += dest_stride;
+ }
+ return;
}
diff --git a/vp10/common/blockd.c b/vp10/common/blockd.c
index b6f910f..031ca1e 100644
--- a/vp10/common/blockd.c
+++ b/vp10/common/blockd.c
@@ -11,10 +11,9 @@
#include "vp10/common/blockd.h"
PREDICTION_MODE vp10_left_block_mode(const MODE_INFO *cur_mi,
- const MODE_INFO *left_mi, int b) {
+ const MODE_INFO *left_mi, int b) {
if (b == 0 || b == 2) {
- if (!left_mi || is_inter_block(&left_mi->mbmi))
- return DC_PRED;
+ if (!left_mi || is_inter_block(&left_mi->mbmi)) return DC_PRED;
return get_y_mode(left_mi, b + 1);
} else {
@@ -24,10 +23,9 @@
}
PREDICTION_MODE vp10_above_block_mode(const MODE_INFO *cur_mi,
- const MODE_INFO *above_mi, int b) {
+ const MODE_INFO *above_mi, int b) {
if (b == 0 || b == 1) {
- if (!above_mi || is_inter_block(&above_mi->mbmi))
- return DC_PRED;
+ if (!above_mi || is_inter_block(&above_mi->mbmi)) return DC_PRED;
return get_y_mode(above_mi, b + 2);
} else {
@@ -40,12 +38,11 @@
const MACROBLOCKD *const xd, BLOCK_SIZE bsize, int plane,
foreach_transformed_block_visitor visit, void *arg) {
const struct macroblockd_plane *const pd = &xd->plane[plane];
- const MB_MODE_INFO* mbmi = &xd->mi[0]->mbmi;
+ const MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
// block and transform sizes, in number of 4x4 blocks log 2 ("*_b")
// 4x4=0, 8x8=2, 16x16=4, 32x32=6, 64x64=8
// transform size varies per plane, look it up in a common way.
- const TX_SIZE tx_size = plane ? get_uv_tx_size(mbmi, pd)
- : mbmi->tx_size;
+ const TX_SIZE tx_size = plane ? get_uv_tx_size(mbmi, pd) : mbmi->tx_size;
const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
const int num_4x4_w = num_4x4_blocks_wide_lookup[plane_bsize];
const int num_4x4_h = num_4x4_blocks_high_lookup[plane_bsize];
@@ -55,10 +52,13 @@
// If mb_to_right_edge is < 0 we are in a situation in which
// the current block size extends into the UMV and we won't
// visit the sub blocks that are wholly within the UMV.
- const int max_blocks_wide = num_4x4_w + (xd->mb_to_right_edge >= 0 ? 0 :
- xd->mb_to_right_edge >> (5 + pd->subsampling_x));
- const int max_blocks_high = num_4x4_h + (xd->mb_to_bottom_edge >= 0 ? 0 :
- xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
+ const int max_blocks_wide =
+ num_4x4_w + (xd->mb_to_right_edge >= 0 ? 0 : xd->mb_to_right_edge >>
+ (5 + pd->subsampling_x));
+ const int max_blocks_high =
+ num_4x4_h + (xd->mb_to_bottom_edge >= 0
+ ? 0
+ : xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
const int extra_step = ((num_4x4_w - max_blocks_wide) >> tx_size) * step;
// Keep track of the row and column of the blocks we use so that we know
@@ -73,10 +73,10 @@
}
}
-void vp10_foreach_transformed_block(const MACROBLOCKD* const xd,
- BLOCK_SIZE bsize,
- foreach_transformed_block_visitor visit,
- void *arg) {
+void vp10_foreach_transformed_block(const MACROBLOCKD *const xd,
+ BLOCK_SIZE bsize,
+ foreach_transformed_block_visitor visit,
+ void *arg) {
int plane;
for (plane = 0; plane < MAX_MB_PLANE; ++plane)
@@ -84,8 +84,8 @@
}
void vp10_set_contexts(const MACROBLOCKD *xd, struct macroblockd_plane *pd,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size, int has_eob,
- int aoff, int loff) {
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size, int has_eob,
+ int aoff, int loff) {
ENTROPY_CONTEXT *const a = pd->above_context + aoff;
ENTROPY_CONTEXT *const l = pd->left_context + loff;
const int tx_size_in_blocks = 1 << tx_size;
@@ -99,10 +99,8 @@
if (above_contexts + aoff > blocks_wide)
above_contexts = blocks_wide - aoff;
- for (i = 0; i < above_contexts; ++i)
- a[i] = has_eob;
- for (i = above_contexts; i < tx_size_in_blocks; ++i)
- a[i] = 0;
+ for (i = 0; i < above_contexts; ++i) a[i] = has_eob;
+ for (i = above_contexts; i < tx_size_in_blocks; ++i) a[i] = 0;
} else {
memset(a, has_eob, sizeof(ENTROPY_CONTEXT) * tx_size_in_blocks);
}
@@ -113,13 +111,10 @@
const int blocks_high = num_4x4_blocks_high_lookup[plane_bsize] +
(xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
int left_contexts = tx_size_in_blocks;
- if (left_contexts + loff > blocks_high)
- left_contexts = blocks_high - loff;
+ if (left_contexts + loff > blocks_high) left_contexts = blocks_high - loff;
- for (i = 0; i < left_contexts; ++i)
- l[i] = has_eob;
- for (i = left_contexts; i < tx_size_in_blocks; ++i)
- l[i] = 0;
+ for (i = 0; i < left_contexts; ++i) l[i] = has_eob;
+ for (i = left_contexts; i < tx_size_in_blocks; ++i) l[i] = 0;
} else {
memset(l, has_eob, sizeof(ENTROPY_CONTEXT) * tx_size_in_blocks);
}
diff --git a/vp10/common/blockd.h b/vp10/common/blockd.h
index a4a8f8d..2b52860 100644
--- a/vp10/common/blockd.h
+++ b/vp10/common/blockd.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_COMMON_BLOCKD_H_
#define VP10_COMMON_BLOCKD_H_
@@ -55,12 +54,12 @@
// decoder implementation modules critically rely on the defined entry values
// specified herein. They should be refactored concurrently.
-#define NONE -1
-#define INTRA_FRAME 0
-#define LAST_FRAME 1
-#define GOLDEN_FRAME 2
-#define ALTREF_FRAME 3
-#define MAX_REF_FRAMES 4
+#define NONE -1
+#define INTRA_FRAME 0
+#define LAST_FRAME 1
+#define GOLDEN_FRAME 2
+#define ALTREF_FRAME 3
+#define MAX_REF_FRAMES 4
typedef int8_t MV_REFERENCE_FRAME;
// This structure now relates to 8x8 block regions.
@@ -94,8 +93,7 @@
} MODE_INFO;
static INLINE PREDICTION_MODE get_y_mode(const MODE_INFO *mi, int block) {
- return mi->mbmi.sb_type < BLOCK_8X8 ? mi->bmi[block].as_mode
- : mi->mbmi.mode;
+ return mi->mbmi.sb_type < BLOCK_8X8 ? mi->bmi[block].as_mode : mi->mbmi.mode;
}
static INLINE int is_inter_block(const MB_MODE_INFO *mbmi) {
@@ -107,15 +105,12 @@
}
PREDICTION_MODE vp10_left_block_mode(const MODE_INFO *cur_mi,
- const MODE_INFO *left_mi, int b);
+ const MODE_INFO *left_mi, int b);
PREDICTION_MODE vp10_above_block_mode(const MODE_INFO *cur_mi,
- const MODE_INFO *above_mi, int b);
+ const MODE_INFO *above_mi, int b);
-enum mv_precision {
- MV_PRECISION_Q3,
- MV_PRECISION_Q4
-};
+enum mv_precision { MV_PRECISION_Q3, MV_PRECISION_Q4 };
struct buf_2d {
uint8_t *buf;
@@ -143,7 +138,7 @@
const int16_t *dequant;
};
-#define BLOCK_OFFSET(x, i) ((x) + (i) * 16)
+#define BLOCK_OFFSET(x, i) ((x) + (i)*16)
typedef struct RefBuffer {
// TODO(dkovalev): idx is not really required and should be removed, now it
@@ -226,7 +221,7 @@
const MODE_INFO *const mi = xd->mi[0];
const MB_MODE_INFO *const mbmi = &mi->mbmi;
- (void) block_idx;
+ (void)block_idx;
if (plane_type != PLANE_TYPE_Y || xd->lossless[mbmi->segment_id] ||
mbmi->tx_size >= TX_32X32)
return DCT_DCT;
@@ -252,8 +247,8 @@
pd->subsampling_y);
}
-static INLINE BLOCK_SIZE get_plane_block_size(BLOCK_SIZE bsize,
- const struct macroblockd_plane *pd) {
+static INLINE BLOCK_SIZE
+get_plane_block_size(BLOCK_SIZE bsize, const struct macroblockd_plane *pd) {
return ss_size_lookup[bsize][pd->subsampling_x][pd->subsampling_y];
}
@@ -272,21 +267,20 @@
typedef void (*foreach_transformed_block_visitor)(int plane, int block,
int blk_row, int blk_col,
BLOCK_SIZE plane_bsize,
- TX_SIZE tx_size,
- void *arg);
+ TX_SIZE tx_size, void *arg);
void vp10_foreach_transformed_block_in_plane(
const MACROBLOCKD *const xd, BLOCK_SIZE bsize, int plane,
foreach_transformed_block_visitor visit, void *arg);
-
-void vp10_foreach_transformed_block(
- const MACROBLOCKD* const xd, BLOCK_SIZE bsize,
- foreach_transformed_block_visitor visit, void *arg);
+void vp10_foreach_transformed_block(const MACROBLOCKD *const xd,
+ BLOCK_SIZE bsize,
+ foreach_transformed_block_visitor visit,
+ void *arg);
void vp10_set_contexts(const MACROBLOCKD *xd, struct macroblockd_plane *pd,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size, int has_eob,
- int aoff, int loff);
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size, int has_eob,
+ int aoff, int loff);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/common/common.h b/vp10/common/common.h
index a984113..ed2f423 100644
--- a/vp10/common/common.h
+++ b/vp10/common/common.h
@@ -26,15 +26,17 @@
#endif
// Only need this for fixed-size arrays, for structs just assign.
-#define vp10_copy(dest, src) { \
+#define vp10_copy(dest, src) \
+ { \
assert(sizeof(dest) == sizeof(src)); \
- memcpy(dest, src, sizeof(src)); \
+ memcpy(dest, src, sizeof(src)); \
}
// Use this for variably-sized arrays.
-#define vp10_copy_array(dest, src, n) { \
- assert(sizeof(*dest) == sizeof(*src)); \
- memcpy(dest, src, n * sizeof(*src)); \
+#define vp10_copy_array(dest, src, n) \
+ { \
+ assert(sizeof(*dest) == sizeof(*src)); \
+ memcpy(dest, src, n * sizeof(*src)); \
}
#define vp10_zero(dest) memset(&(dest), 0, sizeof(dest))
@@ -45,19 +47,21 @@
}
#if CONFIG_DEBUG
-#define CHECK_MEM_ERROR(cm, lval, expr) do { \
- lval = (expr); \
- if (!lval) \
- vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR, \
- "Failed to allocate "#lval" at %s:%d", \
- __FILE__, __LINE__); \
+#define CHECK_MEM_ERROR(cm, lval, expr) \
+ do { \
+ lval = (expr); \
+ if (!lval) \
+ vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR, \
+ "Failed to allocate " #lval " at %s:%d", __FILE__, \
+ __LINE__); \
} while (0)
#else
-#define CHECK_MEM_ERROR(cm, lval, expr) do { \
- lval = (expr); \
- if (!lval) \
- vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR, \
- "Failed to allocate "#lval); \
+#define CHECK_MEM_ERROR(cm, lval, expr) \
+ do { \
+ lval = (expr); \
+ if (!lval) \
+ vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR, \
+ "Failed to allocate " #lval); \
} while (0)
#endif
// TODO(yaowu: validate the usage of these codes or develop new ones.)
@@ -67,7 +71,6 @@
#define VPX_FRAME_MARKER 0x2
-
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/vp10/common/common_data.h b/vp10/common/common_data.h
index 334489c..8863feb 100644
--- a/vp10/common/common_data.h
+++ b/vp10/common/common_data.h
@@ -20,132 +20,132 @@
#endif
// Log 2 conversion lookup tables for block width and height
-static const uint8_t b_width_log2_lookup[BLOCK_SIZES] =
- {0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4};
-static const uint8_t b_height_log2_lookup[BLOCK_SIZES] =
- {0, 1, 0, 1, 2, 1, 2, 3, 2, 3, 4, 3, 4};
-static const uint8_t num_4x4_blocks_wide_lookup[BLOCK_SIZES] =
- {1, 1, 2, 2, 2, 4, 4, 4, 8, 8, 8, 16, 16};
-static const uint8_t num_4x4_blocks_high_lookup[BLOCK_SIZES] =
- {1, 2, 1, 2, 4, 2, 4, 8, 4, 8, 16, 8, 16};
+static const uint8_t b_width_log2_lookup[BLOCK_SIZES] = { 0, 0, 1, 1, 1, 2, 2,
+ 2, 3, 3, 3, 4, 4 };
+static const uint8_t b_height_log2_lookup[BLOCK_SIZES] = { 0, 1, 0, 1, 2, 1, 2,
+ 3, 2, 3, 4, 3, 4 };
+static const uint8_t num_4x4_blocks_wide_lookup[BLOCK_SIZES] = {
+ 1, 1, 2, 2, 2, 4, 4, 4, 8, 8, 8, 16, 16
+};
+static const uint8_t num_4x4_blocks_high_lookup[BLOCK_SIZES] = {
+ 1, 2, 1, 2, 4, 2, 4, 8, 4, 8, 16, 8, 16
+};
// Log 2 conversion lookup tables for modeinfo width and height
-static const uint8_t mi_width_log2_lookup[BLOCK_SIZES] =
- {0, 0, 0, 0, 0, 1, 1, 1, 2, 2, 2, 3, 3};
-static const uint8_t mi_height_log2_lookup[BLOCK_SIZES] =
- {0, 0, 0, 0, 1, 0, 1, 2, 1, 2, 3, 2, 3};
-static const uint8_t num_8x8_blocks_wide_lookup[BLOCK_SIZES] =
- {1, 1, 1, 1, 1, 2, 2, 2, 4, 4, 4, 8, 8};
-static const uint8_t num_8x8_blocks_high_lookup[BLOCK_SIZES] =
- {1, 1, 1, 1, 2, 1, 2, 4, 2, 4, 8, 4, 8};
+static const uint8_t mi_width_log2_lookup[BLOCK_SIZES] = { 0, 0, 0, 0, 0, 1, 1,
+ 1, 2, 2, 2, 3, 3 };
+static const uint8_t mi_height_log2_lookup[BLOCK_SIZES] = { 0, 0, 0, 0, 1, 0, 1,
+ 2, 1, 2, 3, 2, 3 };
+static const uint8_t num_8x8_blocks_wide_lookup[BLOCK_SIZES] = {
+ 1, 1, 1, 1, 1, 2, 2, 2, 4, 4, 4, 8, 8
+};
+static const uint8_t num_8x8_blocks_high_lookup[BLOCK_SIZES] = {
+ 1, 1, 1, 1, 2, 1, 2, 4, 2, 4, 8, 4, 8
+};
// VPXMIN(3, VPXMIN(b_width_log2(bsize), b_height_log2(bsize)))
-static const uint8_t size_group_lookup[BLOCK_SIZES] =
- {0, 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 3};
+static const uint8_t size_group_lookup[BLOCK_SIZES] = { 0, 0, 0, 1, 1, 1, 2,
+ 2, 2, 3, 3, 3, 3 };
-static const uint8_t num_pels_log2_lookup[BLOCK_SIZES] =
- {4, 5, 5, 6, 7, 7, 8, 9, 9, 10, 11, 11, 12};
+static const uint8_t num_pels_log2_lookup[BLOCK_SIZES] = {
+ 4, 5, 5, 6, 7, 7, 8, 9, 9, 10, 11, 11, 12
+};
static const PARTITION_TYPE partition_lookup[][BLOCK_SIZES] = {
- { // 4X4
+ { // 4X4
// 4X4, 4X8,8X4,8X8,8X16,16X8,16X16,16X32,32X16,32X32,32X64,64X32,64X64
- PARTITION_NONE, PARTITION_INVALID, PARTITION_INVALID,
- PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
- PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
- PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
- PARTITION_INVALID
- }, { // 8X8
+ PARTITION_NONE, PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
+ PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
+ PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
+ PARTITION_INVALID },
+ { // 8X8
// 4X4, 4X8,8X4,8X8,8X16,16X8,16X16,16X32,32X16,32X32,32X64,64X32,64X64
PARTITION_SPLIT, PARTITION_VERT, PARTITION_HORZ, PARTITION_NONE,
- PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
- PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
- PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID
- }, { // 16X16
+ PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
+ PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
+ PARTITION_INVALID },
+ { // 16X16
// 4X4, 4X8,8X4,8X8,8X16,16X8,16X16,16X32,32X16,32X32,32X64,64X32,64X64
PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT,
PARTITION_VERT, PARTITION_HORZ, PARTITION_NONE, PARTITION_INVALID,
- PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
- PARTITION_INVALID, PARTITION_INVALID
- }, { // 32X32
+ PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID, PARTITION_INVALID,
+ PARTITION_INVALID },
+ { // 32X32
// 4X4, 4X8,8X4,8X8,8X16,16X8,16X16,16X32,32X16,32X32,32X64,64X32,64X64
PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT,
PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_VERT,
- PARTITION_HORZ, PARTITION_NONE, PARTITION_INVALID,
- PARTITION_INVALID, PARTITION_INVALID
- }, { // 64X64
+ PARTITION_HORZ, PARTITION_NONE, PARTITION_INVALID, PARTITION_INVALID,
+ PARTITION_INVALID },
+ { // 64X64
// 4X4, 4X8,8X4,8X8,8X16,16X8,16X16,16X32,32X16,32X32,32X64,64X32,64X64
PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT,
PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_SPLIT,
PARTITION_SPLIT, PARTITION_SPLIT, PARTITION_VERT, PARTITION_HORZ,
- PARTITION_NONE
- }
+ PARTITION_NONE }
};
static const BLOCK_SIZE subsize_lookup[PARTITION_TYPES][BLOCK_SIZES] = {
- { // PARTITION_NONE
- BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
- BLOCK_8X8, BLOCK_8X16, BLOCK_16X8,
- BLOCK_16X16, BLOCK_16X32, BLOCK_32X16,
- BLOCK_32X32, BLOCK_32X64, BLOCK_64X32,
- BLOCK_64X64,
- }, { // PARTITION_HORZ
- BLOCK_INVALID, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_8X4, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_16X8, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_32X16, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_64X32,
- }, { // PARTITION_VERT
- BLOCK_INVALID, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_4X8, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_8X16, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_16X32, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_32X64,
- }, { // PARTITION_SPLIT
- BLOCK_INVALID, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_4X4, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_8X8, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_16X16, BLOCK_INVALID, BLOCK_INVALID,
- BLOCK_32X32,
+ {
+ // PARTITION_NONE
+ BLOCK_4X4, BLOCK_4X8, BLOCK_8X4, BLOCK_8X8, BLOCK_8X16, BLOCK_16X8,
+ BLOCK_16X16, BLOCK_16X32, BLOCK_32X16, BLOCK_32X32, BLOCK_32X64,
+ BLOCK_64X32, BLOCK_64X64,
+ },
+ {
+ // PARTITION_HORZ
+ BLOCK_INVALID, BLOCK_INVALID, BLOCK_INVALID, BLOCK_8X4, BLOCK_INVALID,
+ BLOCK_INVALID, BLOCK_16X8, BLOCK_INVALID, BLOCK_INVALID, BLOCK_32X16,
+ BLOCK_INVALID, BLOCK_INVALID, BLOCK_64X32,
+ },
+ {
+ // PARTITION_VERT
+ BLOCK_INVALID, BLOCK_INVALID, BLOCK_INVALID, BLOCK_4X8, BLOCK_INVALID,
+ BLOCK_INVALID, BLOCK_8X16, BLOCK_INVALID, BLOCK_INVALID, BLOCK_16X32,
+ BLOCK_INVALID, BLOCK_INVALID, BLOCK_32X64,
+ },
+ {
+ // PARTITION_SPLIT
+ BLOCK_INVALID, BLOCK_INVALID, BLOCK_INVALID, BLOCK_4X4, BLOCK_INVALID,
+ BLOCK_INVALID, BLOCK_8X8, BLOCK_INVALID, BLOCK_INVALID, BLOCK_16X16,
+ BLOCK_INVALID, BLOCK_INVALID, BLOCK_32X32,
}
};
static const TX_SIZE max_txsize_lookup[BLOCK_SIZES] = {
- TX_4X4, TX_4X4, TX_4X4,
- TX_8X8, TX_8X8, TX_8X8,
- TX_16X16, TX_16X16, TX_16X16,
- TX_32X32, TX_32X32, TX_32X32, TX_32X32
+ TX_4X4, TX_4X4, TX_4X4, TX_8X8, TX_8X8, TX_8X8, TX_16X16,
+ TX_16X16, TX_16X16, TX_32X32, TX_32X32, TX_32X32, TX_32X32
};
static const BLOCK_SIZE txsize_to_bsize[TX_SIZES] = {
- BLOCK_4X4, // TX_4X4
- BLOCK_8X8, // TX_8X8
- BLOCK_16X16, // TX_16X16
- BLOCK_32X32, // TX_32X32
+ BLOCK_4X4, // TX_4X4
+ BLOCK_8X8, // TX_8X8
+ BLOCK_16X16, // TX_16X16
+ BLOCK_32X32, // TX_32X32
};
static const TX_SIZE tx_mode_to_biggest_tx_size[TX_MODES] = {
- TX_4X4, // ONLY_4X4
- TX_8X8, // ALLOW_8X8
+ TX_4X4, // ONLY_4X4
+ TX_8X8, // ALLOW_8X8
TX_16X16, // ALLOW_16X16
TX_32X32, // ALLOW_32X32
TX_32X32, // TX_MODE_SELECT
};
static const BLOCK_SIZE ss_size_lookup[BLOCK_SIZES][2][2] = {
-// ss_x == 0 ss_x == 0 ss_x == 1 ss_x == 1
-// ss_y == 0 ss_y == 1 ss_y == 0 ss_y == 1
- {{BLOCK_4X4, BLOCK_INVALID}, {BLOCK_INVALID, BLOCK_INVALID}},
- {{BLOCK_4X8, BLOCK_4X4}, {BLOCK_INVALID, BLOCK_INVALID}},
- {{BLOCK_8X4, BLOCK_INVALID}, {BLOCK_4X4, BLOCK_INVALID}},
- {{BLOCK_8X8, BLOCK_8X4}, {BLOCK_4X8, BLOCK_4X4}},
- {{BLOCK_8X16, BLOCK_8X8}, {BLOCK_INVALID, BLOCK_4X8}},
- {{BLOCK_16X8, BLOCK_INVALID}, {BLOCK_8X8, BLOCK_8X4}},
- {{BLOCK_16X16, BLOCK_16X8}, {BLOCK_8X16, BLOCK_8X8}},
- {{BLOCK_16X32, BLOCK_16X16}, {BLOCK_INVALID, BLOCK_8X16}},
- {{BLOCK_32X16, BLOCK_INVALID}, {BLOCK_16X16, BLOCK_16X8}},
- {{BLOCK_32X32, BLOCK_32X16}, {BLOCK_16X32, BLOCK_16X16}},
- {{BLOCK_32X64, BLOCK_32X32}, {BLOCK_INVALID, BLOCK_16X32}},
- {{BLOCK_64X32, BLOCK_INVALID}, {BLOCK_32X32, BLOCK_32X16}},
- {{BLOCK_64X64, BLOCK_64X32}, {BLOCK_32X64, BLOCK_32X32}},
+ // ss_x == 0 ss_x == 0 ss_x == 1 ss_x == 1
+ // ss_y == 0 ss_y == 1 ss_y == 0 ss_y == 1
+ { { BLOCK_4X4, BLOCK_INVALID }, { BLOCK_INVALID, BLOCK_INVALID } },
+ { { BLOCK_4X8, BLOCK_4X4 }, { BLOCK_INVALID, BLOCK_INVALID } },
+ { { BLOCK_8X4, BLOCK_INVALID }, { BLOCK_4X4, BLOCK_INVALID } },
+ { { BLOCK_8X8, BLOCK_8X4 }, { BLOCK_4X8, BLOCK_4X4 } },
+ { { BLOCK_8X16, BLOCK_8X8 }, { BLOCK_INVALID, BLOCK_4X8 } },
+ { { BLOCK_16X8, BLOCK_INVALID }, { BLOCK_8X8, BLOCK_8X4 } },
+ { { BLOCK_16X16, BLOCK_16X8 }, { BLOCK_8X16, BLOCK_8X8 } },
+ { { BLOCK_16X32, BLOCK_16X16 }, { BLOCK_INVALID, BLOCK_8X16 } },
+ { { BLOCK_32X16, BLOCK_INVALID }, { BLOCK_16X16, BLOCK_16X8 } },
+ { { BLOCK_32X32, BLOCK_32X16 }, { BLOCK_16X32, BLOCK_16X16 } },
+ { { BLOCK_32X64, BLOCK_32X32 }, { BLOCK_INVALID, BLOCK_16X32 } },
+ { { BLOCK_64X32, BLOCK_INVALID }, { BLOCK_32X32, BLOCK_32X16 } },
+ { { BLOCK_64X64, BLOCK_64X32 }, { BLOCK_32X64, BLOCK_32X32 } },
};
// Generates 4 bit field in which each bit set to 1 represents
@@ -154,20 +154,20 @@
static const struct {
PARTITION_CONTEXT above;
PARTITION_CONTEXT left;
-} partition_context_lookup[BLOCK_SIZES]= {
- {15, 15}, // 4X4 - {0b1111, 0b1111}
- {15, 14}, // 4X8 - {0b1111, 0b1110}
- {14, 15}, // 8X4 - {0b1110, 0b1111}
- {14, 14}, // 8X8 - {0b1110, 0b1110}
- {14, 12}, // 8X16 - {0b1110, 0b1100}
- {12, 14}, // 16X8 - {0b1100, 0b1110}
- {12, 12}, // 16X16 - {0b1100, 0b1100}
- {12, 8 }, // 16X32 - {0b1100, 0b1000}
- {8, 12}, // 32X16 - {0b1000, 0b1100}
- {8, 8 }, // 32X32 - {0b1000, 0b1000}
- {8, 0 }, // 32X64 - {0b1000, 0b0000}
- {0, 8 }, // 64X32 - {0b0000, 0b1000}
- {0, 0 }, // 64X64 - {0b0000, 0b0000}
+} partition_context_lookup[BLOCK_SIZES] = {
+ { 15, 15 }, // 4X4 - {0b1111, 0b1111}
+ { 15, 14 }, // 4X8 - {0b1111, 0b1110}
+ { 14, 15 }, // 8X4 - {0b1110, 0b1111}
+ { 14, 14 }, // 8X8 - {0b1110, 0b1110}
+ { 14, 12 }, // 8X16 - {0b1110, 0b1100}
+ { 12, 14 }, // 16X8 - {0b1100, 0b1110}
+ { 12, 12 }, // 16X16 - {0b1100, 0b1100}
+ { 12, 8 }, // 16X32 - {0b1100, 0b1000}
+ { 8, 12 }, // 32X16 - {0b1000, 0b1100}
+ { 8, 8 }, // 32X32 - {0b1000, 0b1000}
+ { 8, 0 }, // 32X64 - {0b1000, 0b0000}
+ { 0, 8 }, // 64X32 - {0b0000, 0b1000}
+ { 0, 0 }, // 64X64 - {0b0000, 0b0000}
};
#ifdef __cplusplus
diff --git a/vp10/common/debugmodes.c b/vp10/common/debugmodes.c
index 10fc4d6..18bbbec 100644
--- a/vp10/common/debugmodes.c
+++ b/vp10/common/debugmodes.c
@@ -34,9 +34,7 @@
for (mi_row = 0; mi_row < rows; mi_row++) {
fprintf(file, "%c ", prefix);
for (mi_col = 0; mi_col < cols; mi_col++) {
- fprintf(file, "%2d ",
- *((int*) ((char *) (&mi[0]->mbmi) +
- member_offset)));
+ fprintf(file, "%2d ", *((int *)((char *)(&mi[0]->mbmi) + member_offset)));
mi++;
}
fprintf(file, "\n");
@@ -79,7 +77,7 @@
fprintf(mvs, "V ");
for (mi_col = 0; mi_col < cols; mi_col++) {
fprintf(mvs, "%4d:%4d ", mi[0]->mbmi.mv[0].as_mv.row,
- mi[0]->mbmi.mv[0].as_mv.col);
+ mi[0]->mbmi.mv[0].as_mv.col);
mi++;
}
fprintf(mvs, "\n");
diff --git a/vp10/common/entropy.c b/vp10/common/entropy.c
index c5930e6..e2b73cb 100644
--- a/vp10/common/entropy.c
+++ b/vp10/common/entropy.c
@@ -17,14 +17,22 @@
// Unconstrained Node Tree
const vpx_tree_index vp10_coef_con_tree[TREE_SIZE(ENTROPY_TOKENS)] = {
- 2, 6, // 0 = LOW_VAL
- -TWO_TOKEN, 4, // 1 = TWO
- -THREE_TOKEN, -FOUR_TOKEN, // 2 = THREE
- 8, 10, // 3 = HIGH_LOW
- -CATEGORY1_TOKEN, -CATEGORY2_TOKEN, // 4 = CAT_ONE
- 12, 14, // 5 = CAT_THREEFOUR
- -CATEGORY3_TOKEN, -CATEGORY4_TOKEN, // 6 = CAT_THREE
- -CATEGORY5_TOKEN, -CATEGORY6_TOKEN // 7 = CAT_FIVE
+ 2,
+ 6, // 0 = LOW_VAL
+ -TWO_TOKEN,
+ 4, // 1 = TWO
+ -THREE_TOKEN,
+ -FOUR_TOKEN, // 2 = THREE
+ 8,
+ 10, // 3 = HIGH_LOW
+ -CATEGORY1_TOKEN,
+ -CATEGORY2_TOKEN, // 4 = CAT_ONE
+ 12,
+ 14, // 5 = CAT_THREEFOUR
+ -CATEGORY3_TOKEN,
+ -CATEGORY4_TOKEN, // 6 = CAT_THREE
+ -CATEGORY5_TOKEN,
+ -CATEGORY6_TOKEN // 7 = CAT_FIVE
};
const vpx_prob vp10_cat1_prob[] = { 159 };
@@ -32,9 +40,8 @@
const vpx_prob vp10_cat3_prob[] = { 173, 148, 140 };
const vpx_prob vp10_cat4_prob[] = { 176, 155, 140, 135 };
const vpx_prob vp10_cat5_prob[] = { 180, 157, 141, 134, 130 };
-const vpx_prob vp10_cat6_prob[] = {
- 254, 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129
-};
+const vpx_prob vp10_cat6_prob[] = { 254, 254, 254, 252, 249, 243, 230,
+ 196, 177, 153, 140, 133, 130, 129 };
#if CONFIG_VPX_HIGHBITDEPTH
const vpx_prob vp10_cat1_prob_high10[] = { 159 };
const vpx_prob vp10_cat2_prob_high10[] = { 165, 145 };
@@ -42,96 +49,68 @@
const vpx_prob vp10_cat4_prob_high10[] = { 176, 155, 140, 135 };
const vpx_prob vp10_cat5_prob_high10[] = { 180, 157, 141, 134, 130 };
const vpx_prob vp10_cat6_prob_high10[] = {
- 255, 255, 254, 254, 254, 252, 249, 243,
- 230, 196, 177, 153, 140, 133, 130, 129
+ 255, 255, 254, 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129
};
const vpx_prob vp10_cat1_prob_high12[] = { 159 };
const vpx_prob vp10_cat2_prob_high12[] = { 165, 145 };
const vpx_prob vp10_cat3_prob_high12[] = { 173, 148, 140 };
const vpx_prob vp10_cat4_prob_high12[] = { 176, 155, 140, 135 };
const vpx_prob vp10_cat5_prob_high12[] = { 180, 157, 141, 134, 130 };
-const vpx_prob vp10_cat6_prob_high12[] = {
- 255, 255, 255, 255, 254, 254, 254, 252, 249,
- 243, 230, 196, 177, 153, 140, 133, 130, 129
-};
+const vpx_prob vp10_cat6_prob_high12[] = { 255, 255, 255, 255, 254, 254,
+ 254, 252, 249, 243, 230, 196,
+ 177, 153, 140, 133, 130, 129 };
#endif
const uint8_t vp10_coefband_trans_8x8plus[1024] = {
- 0, 1, 1, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4,
- 4, 4, 4, 4, 4, 5,
+ 0, 1, 1, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5,
// beyond MAXBAND_INDEX+1 all values are filled as 5
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
};
const uint8_t vp10_coefband_trans_4x4[16] = {
0, 1, 1, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 5, 5, 5,
};
-const uint8_t vp10_pt_energy_class[ENTROPY_TOKENS] = {
- 0, 1, 2, 3, 3, 4, 4, 5, 5, 5, 5, 5
-};
+const uint8_t vp10_pt_energy_class[ENTROPY_TOKENS] = { 0, 1, 2, 3, 3, 4,
+ 4, 5, 5, 5, 5, 5 };
// Model obtained from a 2-sided zero-centerd distribuition derived
// from a Pareto distribution. The cdf of the distribution is:
@@ -148,597 +127,921 @@
// vp10_pareto8_full[l][node] = (vp10_pareto8_full[l-1][node] +
// vp10_pareto8_full[l+1][node] ) >> 1;
const vpx_prob vp10_pareto8_full[COEFF_PROB_MODELS][MODEL_NODES] = {
- { 3, 86, 128, 6, 86, 23, 88, 29},
- { 6, 86, 128, 11, 87, 42, 91, 52},
- { 9, 86, 129, 17, 88, 61, 94, 76},
- { 12, 86, 129, 22, 88, 77, 97, 93},
- { 15, 87, 129, 28, 89, 93, 100, 110},
- { 17, 87, 129, 33, 90, 105, 103, 123},
- { 20, 88, 130, 38, 91, 118, 106, 136},
- { 23, 88, 130, 43, 91, 128, 108, 146},
- { 26, 89, 131, 48, 92, 139, 111, 156},
- { 28, 89, 131, 53, 93, 147, 114, 163},
- { 31, 90, 131, 58, 94, 156, 117, 171},
- { 34, 90, 131, 62, 94, 163, 119, 177},
- { 37, 90, 132, 66, 95, 171, 122, 184},
- { 39, 90, 132, 70, 96, 177, 124, 189},
- { 42, 91, 132, 75, 97, 183, 127, 194},
- { 44, 91, 132, 79, 97, 188, 129, 198},
- { 47, 92, 133, 83, 98, 193, 132, 202},
- { 49, 92, 133, 86, 99, 197, 134, 205},
- { 52, 93, 133, 90, 100, 201, 137, 208},
- { 54, 93, 133, 94, 100, 204, 139, 211},
- { 57, 94, 134, 98, 101, 208, 142, 214},
- { 59, 94, 134, 101, 102, 211, 144, 216},
- { 62, 94, 135, 105, 103, 214, 146, 218},
- { 64, 94, 135, 108, 103, 216, 148, 220},
- { 66, 95, 135, 111, 104, 219, 151, 222},
- { 68, 95, 135, 114, 105, 221, 153, 223},
- { 71, 96, 136, 117, 106, 224, 155, 225},
- { 73, 96, 136, 120, 106, 225, 157, 226},
- { 76, 97, 136, 123, 107, 227, 159, 228},
- { 78, 97, 136, 126, 108, 229, 160, 229},
- { 80, 98, 137, 129, 109, 231, 162, 231},
- { 82, 98, 137, 131, 109, 232, 164, 232},
- { 84, 98, 138, 134, 110, 234, 166, 233},
- { 86, 98, 138, 137, 111, 235, 168, 234},
- { 89, 99, 138, 140, 112, 236, 170, 235},
- { 91, 99, 138, 142, 112, 237, 171, 235},
- { 93, 100, 139, 145, 113, 238, 173, 236},
- { 95, 100, 139, 147, 114, 239, 174, 237},
- { 97, 101, 140, 149, 115, 240, 176, 238},
- { 99, 101, 140, 151, 115, 241, 177, 238},
- {101, 102, 140, 154, 116, 242, 179, 239},
- {103, 102, 140, 156, 117, 242, 180, 239},
- {105, 103, 141, 158, 118, 243, 182, 240},
- {107, 103, 141, 160, 118, 243, 183, 240},
- {109, 104, 141, 162, 119, 244, 185, 241},
- {111, 104, 141, 164, 119, 244, 186, 241},
- {113, 104, 142, 166, 120, 245, 187, 242},
- {114, 104, 142, 168, 121, 245, 188, 242},
- {116, 105, 143, 170, 122, 246, 190, 243},
- {118, 105, 143, 171, 122, 246, 191, 243},
- {120, 106, 143, 173, 123, 247, 192, 244},
- {121, 106, 143, 175, 124, 247, 193, 244},
- {123, 107, 144, 177, 125, 248, 195, 244},
- {125, 107, 144, 178, 125, 248, 196, 244},
- {127, 108, 145, 180, 126, 249, 197, 245},
- {128, 108, 145, 181, 127, 249, 198, 245},
- {130, 109, 145, 183, 128, 249, 199, 245},
- {132, 109, 145, 184, 128, 249, 200, 245},
- {134, 110, 146, 186, 129, 250, 201, 246},
- {135, 110, 146, 187, 130, 250, 202, 246},
- {137, 111, 147, 189, 131, 251, 203, 246},
- {138, 111, 147, 190, 131, 251, 204, 246},
- {140, 112, 147, 192, 132, 251, 205, 247},
- {141, 112, 147, 193, 132, 251, 206, 247},
- {143, 113, 148, 194, 133, 251, 207, 247},
- {144, 113, 148, 195, 134, 251, 207, 247},
- {146, 114, 149, 197, 135, 252, 208, 248},
- {147, 114, 149, 198, 135, 252, 209, 248},
- {149, 115, 149, 199, 136, 252, 210, 248},
- {150, 115, 149, 200, 137, 252, 210, 248},
- {152, 115, 150, 201, 138, 252, 211, 248},
- {153, 115, 150, 202, 138, 252, 212, 248},
- {155, 116, 151, 204, 139, 253, 213, 249},
- {156, 116, 151, 205, 139, 253, 213, 249},
- {158, 117, 151, 206, 140, 253, 214, 249},
- {159, 117, 151, 207, 141, 253, 215, 249},
- {161, 118, 152, 208, 142, 253, 216, 249},
- {162, 118, 152, 209, 142, 253, 216, 249},
- {163, 119, 153, 210, 143, 253, 217, 249},
- {164, 119, 153, 211, 143, 253, 217, 249},
- {166, 120, 153, 212, 144, 254, 218, 250},
- {167, 120, 153, 212, 145, 254, 219, 250},
- {168, 121, 154, 213, 146, 254, 220, 250},
- {169, 121, 154, 214, 146, 254, 220, 250},
- {171, 122, 155, 215, 147, 254, 221, 250},
- {172, 122, 155, 216, 147, 254, 221, 250},
- {173, 123, 155, 217, 148, 254, 222, 250},
- {174, 123, 155, 217, 149, 254, 222, 250},
- {176, 124, 156, 218, 150, 254, 223, 250},
- {177, 124, 156, 219, 150, 254, 223, 250},
- {178, 125, 157, 220, 151, 254, 224, 251},
- {179, 125, 157, 220, 151, 254, 224, 251},
- {180, 126, 157, 221, 152, 254, 225, 251},
- {181, 126, 157, 221, 152, 254, 225, 251},
- {183, 127, 158, 222, 153, 254, 226, 251},
- {184, 127, 158, 223, 154, 254, 226, 251},
- {185, 128, 159, 224, 155, 255, 227, 251},
- {186, 128, 159, 224, 155, 255, 227, 251},
- {187, 129, 160, 225, 156, 255, 228, 251},
- {188, 130, 160, 225, 156, 255, 228, 251},
- {189, 131, 160, 226, 157, 255, 228, 251},
- {190, 131, 160, 226, 158, 255, 228, 251},
- {191, 132, 161, 227, 159, 255, 229, 251},
- {192, 132, 161, 227, 159, 255, 229, 251},
- {193, 133, 162, 228, 160, 255, 230, 252},
- {194, 133, 162, 229, 160, 255, 230, 252},
- {195, 134, 163, 230, 161, 255, 231, 252},
- {196, 134, 163, 230, 161, 255, 231, 252},
- {197, 135, 163, 231, 162, 255, 231, 252},
- {198, 135, 163, 231, 162, 255, 231, 252},
- {199, 136, 164, 232, 163, 255, 232, 252},
- {200, 136, 164, 232, 164, 255, 232, 252},
- {201, 137, 165, 233, 165, 255, 233, 252},
- {201, 137, 165, 233, 165, 255, 233, 252},
- {202, 138, 166, 233, 166, 255, 233, 252},
- {203, 138, 166, 233, 166, 255, 233, 252},
- {204, 139, 166, 234, 167, 255, 234, 252},
- {205, 139, 166, 234, 167, 255, 234, 252},
- {206, 140, 167, 235, 168, 255, 235, 252},
- {206, 140, 167, 235, 168, 255, 235, 252},
- {207, 141, 168, 236, 169, 255, 235, 252},
- {208, 141, 168, 236, 170, 255, 235, 252},
- {209, 142, 169, 237, 171, 255, 236, 252},
- {209, 143, 169, 237, 171, 255, 236, 252},
- {210, 144, 169, 237, 172, 255, 236, 252},
- {211, 144, 169, 237, 172, 255, 236, 252},
- {212, 145, 170, 238, 173, 255, 237, 252},
- {213, 145, 170, 238, 173, 255, 237, 252},
- {214, 146, 171, 239, 174, 255, 237, 253},
- {214, 146, 171, 239, 174, 255, 237, 253},
- {215, 147, 172, 240, 175, 255, 238, 253},
- {215, 147, 172, 240, 175, 255, 238, 253},
- {216, 148, 173, 240, 176, 255, 238, 253},
- {217, 148, 173, 240, 176, 255, 238, 253},
- {218, 149, 173, 241, 177, 255, 239, 253},
- {218, 149, 173, 241, 178, 255, 239, 253},
- {219, 150, 174, 241, 179, 255, 239, 253},
- {219, 151, 174, 241, 179, 255, 239, 253},
- {220, 152, 175, 242, 180, 255, 240, 253},
- {221, 152, 175, 242, 180, 255, 240, 253},
- {222, 153, 176, 242, 181, 255, 240, 253},
- {222, 153, 176, 242, 181, 255, 240, 253},
- {223, 154, 177, 243, 182, 255, 240, 253},
- {223, 154, 177, 243, 182, 255, 240, 253},
- {224, 155, 178, 244, 183, 255, 241, 253},
- {224, 155, 178, 244, 183, 255, 241, 253},
- {225, 156, 178, 244, 184, 255, 241, 253},
- {225, 157, 178, 244, 184, 255, 241, 253},
- {226, 158, 179, 244, 185, 255, 242, 253},
- {227, 158, 179, 244, 185, 255, 242, 253},
- {228, 159, 180, 245, 186, 255, 242, 253},
- {228, 159, 180, 245, 186, 255, 242, 253},
- {229, 160, 181, 245, 187, 255, 242, 253},
- {229, 160, 181, 245, 187, 255, 242, 253},
- {230, 161, 182, 246, 188, 255, 243, 253},
- {230, 162, 182, 246, 188, 255, 243, 253},
- {231, 163, 183, 246, 189, 255, 243, 253},
- {231, 163, 183, 246, 189, 255, 243, 253},
- {232, 164, 184, 247, 190, 255, 243, 253},
- {232, 164, 184, 247, 190, 255, 243, 253},
- {233, 165, 185, 247, 191, 255, 244, 253},
- {233, 165, 185, 247, 191, 255, 244, 253},
- {234, 166, 185, 247, 192, 255, 244, 253},
- {234, 167, 185, 247, 192, 255, 244, 253},
- {235, 168, 186, 248, 193, 255, 244, 253},
- {235, 168, 186, 248, 193, 255, 244, 253},
- {236, 169, 187, 248, 194, 255, 244, 253},
- {236, 169, 187, 248, 194, 255, 244, 253},
- {236, 170, 188, 248, 195, 255, 245, 253},
- {236, 170, 188, 248, 195, 255, 245, 253},
- {237, 171, 189, 249, 196, 255, 245, 254},
- {237, 172, 189, 249, 196, 255, 245, 254},
- {238, 173, 190, 249, 197, 255, 245, 254},
- {238, 173, 190, 249, 197, 255, 245, 254},
- {239, 174, 191, 249, 198, 255, 245, 254},
- {239, 174, 191, 249, 198, 255, 245, 254},
- {240, 175, 192, 249, 199, 255, 246, 254},
- {240, 176, 192, 249, 199, 255, 246, 254},
- {240, 177, 193, 250, 200, 255, 246, 254},
- {240, 177, 193, 250, 200, 255, 246, 254},
- {241, 178, 194, 250, 201, 255, 246, 254},
- {241, 178, 194, 250, 201, 255, 246, 254},
- {242, 179, 195, 250, 202, 255, 246, 254},
- {242, 180, 195, 250, 202, 255, 246, 254},
- {242, 181, 196, 250, 203, 255, 247, 254},
- {242, 181, 196, 250, 203, 255, 247, 254},
- {243, 182, 197, 251, 204, 255, 247, 254},
- {243, 183, 197, 251, 204, 255, 247, 254},
- {244, 184, 198, 251, 205, 255, 247, 254},
- {244, 184, 198, 251, 205, 255, 247, 254},
- {244, 185, 199, 251, 206, 255, 247, 254},
- {244, 185, 199, 251, 206, 255, 247, 254},
- {245, 186, 200, 251, 207, 255, 247, 254},
- {245, 187, 200, 251, 207, 255, 247, 254},
- {246, 188, 201, 252, 207, 255, 248, 254},
- {246, 188, 201, 252, 207, 255, 248, 254},
- {246, 189, 202, 252, 208, 255, 248, 254},
- {246, 190, 202, 252, 208, 255, 248, 254},
- {247, 191, 203, 252, 209, 255, 248, 254},
- {247, 191, 203, 252, 209, 255, 248, 254},
- {247, 192, 204, 252, 210, 255, 248, 254},
- {247, 193, 204, 252, 210, 255, 248, 254},
- {248, 194, 205, 252, 211, 255, 248, 254},
- {248, 194, 205, 252, 211, 255, 248, 254},
- {248, 195, 206, 252, 212, 255, 249, 254},
- {248, 196, 206, 252, 212, 255, 249, 254},
- {249, 197, 207, 253, 213, 255, 249, 254},
- {249, 197, 207, 253, 213, 255, 249, 254},
- {249, 198, 208, 253, 214, 255, 249, 254},
- {249, 199, 209, 253, 214, 255, 249, 254},
- {250, 200, 210, 253, 215, 255, 249, 254},
- {250, 200, 210, 253, 215, 255, 249, 254},
- {250, 201, 211, 253, 215, 255, 249, 254},
- {250, 202, 211, 253, 215, 255, 249, 254},
- {250, 203, 212, 253, 216, 255, 249, 254},
- {250, 203, 212, 253, 216, 255, 249, 254},
- {251, 204, 213, 253, 217, 255, 250, 254},
- {251, 205, 213, 253, 217, 255, 250, 254},
- {251, 206, 214, 254, 218, 255, 250, 254},
- {251, 206, 215, 254, 218, 255, 250, 254},
- {252, 207, 216, 254, 219, 255, 250, 254},
- {252, 208, 216, 254, 219, 255, 250, 254},
- {252, 209, 217, 254, 220, 255, 250, 254},
- {252, 210, 217, 254, 220, 255, 250, 254},
- {252, 211, 218, 254, 221, 255, 250, 254},
- {252, 212, 218, 254, 221, 255, 250, 254},
- {253, 213, 219, 254, 222, 255, 250, 254},
- {253, 213, 220, 254, 222, 255, 250, 254},
- {253, 214, 221, 254, 223, 255, 250, 254},
- {253, 215, 221, 254, 223, 255, 250, 254},
- {253, 216, 222, 254, 224, 255, 251, 254},
- {253, 217, 223, 254, 224, 255, 251, 254},
- {253, 218, 224, 254, 225, 255, 251, 254},
- {253, 219, 224, 254, 225, 255, 251, 254},
- {254, 220, 225, 254, 225, 255, 251, 254},
- {254, 221, 226, 254, 225, 255, 251, 254},
- {254, 222, 227, 255, 226, 255, 251, 254},
- {254, 223, 227, 255, 226, 255, 251, 254},
- {254, 224, 228, 255, 227, 255, 251, 254},
- {254, 225, 229, 255, 227, 255, 251, 254},
- {254, 226, 230, 255, 228, 255, 251, 254},
- {254, 227, 230, 255, 229, 255, 251, 254},
- {255, 228, 231, 255, 230, 255, 251, 254},
- {255, 229, 232, 255, 230, 255, 251, 254},
- {255, 230, 233, 255, 231, 255, 252, 254},
- {255, 231, 234, 255, 231, 255, 252, 254},
- {255, 232, 235, 255, 232, 255, 252, 254},
- {255, 233, 236, 255, 232, 255, 252, 254},
- {255, 235, 237, 255, 233, 255, 252, 254},
- {255, 236, 238, 255, 234, 255, 252, 254},
- {255, 238, 240, 255, 235, 255, 252, 255},
- {255, 239, 241, 255, 235, 255, 252, 254},
- {255, 241, 243, 255, 236, 255, 252, 254},
- {255, 243, 245, 255, 237, 255, 252, 254},
- {255, 246, 247, 255, 239, 255, 253, 255},
+ { 3, 86, 128, 6, 86, 23, 88, 29 },
+ { 6, 86, 128, 11, 87, 42, 91, 52 },
+ { 9, 86, 129, 17, 88, 61, 94, 76 },
+ { 12, 86, 129, 22, 88, 77, 97, 93 },
+ { 15, 87, 129, 28, 89, 93, 100, 110 },
+ { 17, 87, 129, 33, 90, 105, 103, 123 },
+ { 20, 88, 130, 38, 91, 118, 106, 136 },
+ { 23, 88, 130, 43, 91, 128, 108, 146 },
+ { 26, 89, 131, 48, 92, 139, 111, 156 },
+ { 28, 89, 131, 53, 93, 147, 114, 163 },
+ { 31, 90, 131, 58, 94, 156, 117, 171 },
+ { 34, 90, 131, 62, 94, 163, 119, 177 },
+ { 37, 90, 132, 66, 95, 171, 122, 184 },
+ { 39, 90, 132, 70, 96, 177, 124, 189 },
+ { 42, 91, 132, 75, 97, 183, 127, 194 },
+ { 44, 91, 132, 79, 97, 188, 129, 198 },
+ { 47, 92, 133, 83, 98, 193, 132, 202 },
+ { 49, 92, 133, 86, 99, 197, 134, 205 },
+ { 52, 93, 133, 90, 100, 201, 137, 208 },
+ { 54, 93, 133, 94, 100, 204, 139, 211 },
+ { 57, 94, 134, 98, 101, 208, 142, 214 },
+ { 59, 94, 134, 101, 102, 211, 144, 216 },
+ { 62, 94, 135, 105, 103, 214, 146, 218 },
+ { 64, 94, 135, 108, 103, 216, 148, 220 },
+ { 66, 95, 135, 111, 104, 219, 151, 222 },
+ { 68, 95, 135, 114, 105, 221, 153, 223 },
+ { 71, 96, 136, 117, 106, 224, 155, 225 },
+ { 73, 96, 136, 120, 106, 225, 157, 226 },
+ { 76, 97, 136, 123, 107, 227, 159, 228 },
+ { 78, 97, 136, 126, 108, 229, 160, 229 },
+ { 80, 98, 137, 129, 109, 231, 162, 231 },
+ { 82, 98, 137, 131, 109, 232, 164, 232 },
+ { 84, 98, 138, 134, 110, 234, 166, 233 },
+ { 86, 98, 138, 137, 111, 235, 168, 234 },
+ { 89, 99, 138, 140, 112, 236, 170, 235 },
+ { 91, 99, 138, 142, 112, 237, 171, 235 },
+ { 93, 100, 139, 145, 113, 238, 173, 236 },
+ { 95, 100, 139, 147, 114, 239, 174, 237 },
+ { 97, 101, 140, 149, 115, 240, 176, 238 },
+ { 99, 101, 140, 151, 115, 241, 177, 238 },
+ { 101, 102, 140, 154, 116, 242, 179, 239 },
+ { 103, 102, 140, 156, 117, 242, 180, 239 },
+ { 105, 103, 141, 158, 118, 243, 182, 240 },
+ { 107, 103, 141, 160, 118, 243, 183, 240 },
+ { 109, 104, 141, 162, 119, 244, 185, 241 },
+ { 111, 104, 141, 164, 119, 244, 186, 241 },
+ { 113, 104, 142, 166, 120, 245, 187, 242 },
+ { 114, 104, 142, 168, 121, 245, 188, 242 },
+ { 116, 105, 143, 170, 122, 246, 190, 243 },
+ { 118, 105, 143, 171, 122, 246, 191, 243 },
+ { 120, 106, 143, 173, 123, 247, 192, 244 },
+ { 121, 106, 143, 175, 124, 247, 193, 244 },
+ { 123, 107, 144, 177, 125, 248, 195, 244 },
+ { 125, 107, 144, 178, 125, 248, 196, 244 },
+ { 127, 108, 145, 180, 126, 249, 197, 245 },
+ { 128, 108, 145, 181, 127, 249, 198, 245 },
+ { 130, 109, 145, 183, 128, 249, 199, 245 },
+ { 132, 109, 145, 184, 128, 249, 200, 245 },
+ { 134, 110, 146, 186, 129, 250, 201, 246 },
+ { 135, 110, 146, 187, 130, 250, 202, 246 },
+ { 137, 111, 147, 189, 131, 251, 203, 246 },
+ { 138, 111, 147, 190, 131, 251, 204, 246 },
+ { 140, 112, 147, 192, 132, 251, 205, 247 },
+ { 141, 112, 147, 193, 132, 251, 206, 247 },
+ { 143, 113, 148, 194, 133, 251, 207, 247 },
+ { 144, 113, 148, 195, 134, 251, 207, 247 },
+ { 146, 114, 149, 197, 135, 252, 208, 248 },
+ { 147, 114, 149, 198, 135, 252, 209, 248 },
+ { 149, 115, 149, 199, 136, 252, 210, 248 },
+ { 150, 115, 149, 200, 137, 252, 210, 248 },
+ { 152, 115, 150, 201, 138, 252, 211, 248 },
+ { 153, 115, 150, 202, 138, 252, 212, 248 },
+ { 155, 116, 151, 204, 139, 253, 213, 249 },
+ { 156, 116, 151, 205, 139, 253, 213, 249 },
+ { 158, 117, 151, 206, 140, 253, 214, 249 },
+ { 159, 117, 151, 207, 141, 253, 215, 249 },
+ { 161, 118, 152, 208, 142, 253, 216, 249 },
+ { 162, 118, 152, 209, 142, 253, 216, 249 },
+ { 163, 119, 153, 210, 143, 253, 217, 249 },
+ { 164, 119, 153, 211, 143, 253, 217, 249 },
+ { 166, 120, 153, 212, 144, 254, 218, 250 },
+ { 167, 120, 153, 212, 145, 254, 219, 250 },
+ { 168, 121, 154, 213, 146, 254, 220, 250 },
+ { 169, 121, 154, 214, 146, 254, 220, 250 },
+ { 171, 122, 155, 215, 147, 254, 221, 250 },
+ { 172, 122, 155, 216, 147, 254, 221, 250 },
+ { 173, 123, 155, 217, 148, 254, 222, 250 },
+ { 174, 123, 155, 217, 149, 254, 222, 250 },
+ { 176, 124, 156, 218, 150, 254, 223, 250 },
+ { 177, 124, 156, 219, 150, 254, 223, 250 },
+ { 178, 125, 157, 220, 151, 254, 224, 251 },
+ { 179, 125, 157, 220, 151, 254, 224, 251 },
+ { 180, 126, 157, 221, 152, 254, 225, 251 },
+ { 181, 126, 157, 221, 152, 254, 225, 251 },
+ { 183, 127, 158, 222, 153, 254, 226, 251 },
+ { 184, 127, 158, 223, 154, 254, 226, 251 },
+ { 185, 128, 159, 224, 155, 255, 227, 251 },
+ { 186, 128, 159, 224, 155, 255, 227, 251 },
+ { 187, 129, 160, 225, 156, 255, 228, 251 },
+ { 188, 130, 160, 225, 156, 255, 228, 251 },
+ { 189, 131, 160, 226, 157, 255, 228, 251 },
+ { 190, 131, 160, 226, 158, 255, 228, 251 },
+ { 191, 132, 161, 227, 159, 255, 229, 251 },
+ { 192, 132, 161, 227, 159, 255, 229, 251 },
+ { 193, 133, 162, 228, 160, 255, 230, 252 },
+ { 194, 133, 162, 229, 160, 255, 230, 252 },
+ { 195, 134, 163, 230, 161, 255, 231, 252 },
+ { 196, 134, 163, 230, 161, 255, 231, 252 },
+ { 197, 135, 163, 231, 162, 255, 231, 252 },
+ { 198, 135, 163, 231, 162, 255, 231, 252 },
+ { 199, 136, 164, 232, 163, 255, 232, 252 },
+ { 200, 136, 164, 232, 164, 255, 232, 252 },
+ { 201, 137, 165, 233, 165, 255, 233, 252 },
+ { 201, 137, 165, 233, 165, 255, 233, 252 },
+ { 202, 138, 166, 233, 166, 255, 233, 252 },
+ { 203, 138, 166, 233, 166, 255, 233, 252 },
+ { 204, 139, 166, 234, 167, 255, 234, 252 },
+ { 205, 139, 166, 234, 167, 255, 234, 252 },
+ { 206, 140, 167, 235, 168, 255, 235, 252 },
+ { 206, 140, 167, 235, 168, 255, 235, 252 },
+ { 207, 141, 168, 236, 169, 255, 235, 252 },
+ { 208, 141, 168, 236, 170, 255, 235, 252 },
+ { 209, 142, 169, 237, 171, 255, 236, 252 },
+ { 209, 143, 169, 237, 171, 255, 236, 252 },
+ { 210, 144, 169, 237, 172, 255, 236, 252 },
+ { 211, 144, 169, 237, 172, 255, 236, 252 },
+ { 212, 145, 170, 238, 173, 255, 237, 252 },
+ { 213, 145, 170, 238, 173, 255, 237, 252 },
+ { 214, 146, 171, 239, 174, 255, 237, 253 },
+ { 214, 146, 171, 239, 174, 255, 237, 253 },
+ { 215, 147, 172, 240, 175, 255, 238, 253 },
+ { 215, 147, 172, 240, 175, 255, 238, 253 },
+ { 216, 148, 173, 240, 176, 255, 238, 253 },
+ { 217, 148, 173, 240, 176, 255, 238, 253 },
+ { 218, 149, 173, 241, 177, 255, 239, 253 },
+ { 218, 149, 173, 241, 178, 255, 239, 253 },
+ { 219, 150, 174, 241, 179, 255, 239, 253 },
+ { 219, 151, 174, 241, 179, 255, 239, 253 },
+ { 220, 152, 175, 242, 180, 255, 240, 253 },
+ { 221, 152, 175, 242, 180, 255, 240, 253 },
+ { 222, 153, 176, 242, 181, 255, 240, 253 },
+ { 222, 153, 176, 242, 181, 255, 240, 253 },
+ { 223, 154, 177, 243, 182, 255, 240, 253 },
+ { 223, 154, 177, 243, 182, 255, 240, 253 },
+ { 224, 155, 178, 244, 183, 255, 241, 253 },
+ { 224, 155, 178, 244, 183, 255, 241, 253 },
+ { 225, 156, 178, 244, 184, 255, 241, 253 },
+ { 225, 157, 178, 244, 184, 255, 241, 253 },
+ { 226, 158, 179, 244, 185, 255, 242, 253 },
+ { 227, 158, 179, 244, 185, 255, 242, 253 },
+ { 228, 159, 180, 245, 186, 255, 242, 253 },
+ { 228, 159, 180, 245, 186, 255, 242, 253 },
+ { 229, 160, 181, 245, 187, 255, 242, 253 },
+ { 229, 160, 181, 245, 187, 255, 242, 253 },
+ { 230, 161, 182, 246, 188, 255, 243, 253 },
+ { 230, 162, 182, 246, 188, 255, 243, 253 },
+ { 231, 163, 183, 246, 189, 255, 243, 253 },
+ { 231, 163, 183, 246, 189, 255, 243, 253 },
+ { 232, 164, 184, 247, 190, 255, 243, 253 },
+ { 232, 164, 184, 247, 190, 255, 243, 253 },
+ { 233, 165, 185, 247, 191, 255, 244, 253 },
+ { 233, 165, 185, 247, 191, 255, 244, 253 },
+ { 234, 166, 185, 247, 192, 255, 244, 253 },
+ { 234, 167, 185, 247, 192, 255, 244, 253 },
+ { 235, 168, 186, 248, 193, 255, 244, 253 },
+ { 235, 168, 186, 248, 193, 255, 244, 253 },
+ { 236, 169, 187, 248, 194, 255, 244, 253 },
+ { 236, 169, 187, 248, 194, 255, 244, 253 },
+ { 236, 170, 188, 248, 195, 255, 245, 253 },
+ { 236, 170, 188, 248, 195, 255, 245, 253 },
+ { 237, 171, 189, 249, 196, 255, 245, 254 },
+ { 237, 172, 189, 249, 196, 255, 245, 254 },
+ { 238, 173, 190, 249, 197, 255, 245, 254 },
+ { 238, 173, 190, 249, 197, 255, 245, 254 },
+ { 239, 174, 191, 249, 198, 255, 245, 254 },
+ { 239, 174, 191, 249, 198, 255, 245, 254 },
+ { 240, 175, 192, 249, 199, 255, 246, 254 },
+ { 240, 176, 192, 249, 199, 255, 246, 254 },
+ { 240, 177, 193, 250, 200, 255, 246, 254 },
+ { 240, 177, 193, 250, 200, 255, 246, 254 },
+ { 241, 178, 194, 250, 201, 255, 246, 254 },
+ { 241, 178, 194, 250, 201, 255, 246, 254 },
+ { 242, 179, 195, 250, 202, 255, 246, 254 },
+ { 242, 180, 195, 250, 202, 255, 246, 254 },
+ { 242, 181, 196, 250, 203, 255, 247, 254 },
+ { 242, 181, 196, 250, 203, 255, 247, 254 },
+ { 243, 182, 197, 251, 204, 255, 247, 254 },
+ { 243, 183, 197, 251, 204, 255, 247, 254 },
+ { 244, 184, 198, 251, 205, 255, 247, 254 },
+ { 244, 184, 198, 251, 205, 255, 247, 254 },
+ { 244, 185, 199, 251, 206, 255, 247, 254 },
+ { 244, 185, 199, 251, 206, 255, 247, 254 },
+ { 245, 186, 200, 251, 207, 255, 247, 254 },
+ { 245, 187, 200, 251, 207, 255, 247, 254 },
+ { 246, 188, 201, 252, 207, 255, 248, 254 },
+ { 246, 188, 201, 252, 207, 255, 248, 254 },
+ { 246, 189, 202, 252, 208, 255, 248, 254 },
+ { 246, 190, 202, 252, 208, 255, 248, 254 },
+ { 247, 191, 203, 252, 209, 255, 248, 254 },
+ { 247, 191, 203, 252, 209, 255, 248, 254 },
+ { 247, 192, 204, 252, 210, 255, 248, 254 },
+ { 247, 193, 204, 252, 210, 255, 248, 254 },
+ { 248, 194, 205, 252, 211, 255, 248, 254 },
+ { 248, 194, 205, 252, 211, 255, 248, 254 },
+ { 248, 195, 206, 252, 212, 255, 249, 254 },
+ { 248, 196, 206, 252, 212, 255, 249, 254 },
+ { 249, 197, 207, 253, 213, 255, 249, 254 },
+ { 249, 197, 207, 253, 213, 255, 249, 254 },
+ { 249, 198, 208, 253, 214, 255, 249, 254 },
+ { 249, 199, 209, 253, 214, 255, 249, 254 },
+ { 250, 200, 210, 253, 215, 255, 249, 254 },
+ { 250, 200, 210, 253, 215, 255, 249, 254 },
+ { 250, 201, 211, 253, 215, 255, 249, 254 },
+ { 250, 202, 211, 253, 215, 255, 249, 254 },
+ { 250, 203, 212, 253, 216, 255, 249, 254 },
+ { 250, 203, 212, 253, 216, 255, 249, 254 },
+ { 251, 204, 213, 253, 217, 255, 250, 254 },
+ { 251, 205, 213, 253, 217, 255, 250, 254 },
+ { 251, 206, 214, 254, 218, 255, 250, 254 },
+ { 251, 206, 215, 254, 218, 255, 250, 254 },
+ { 252, 207, 216, 254, 219, 255, 250, 254 },
+ { 252, 208, 216, 254, 219, 255, 250, 254 },
+ { 252, 209, 217, 254, 220, 255, 250, 254 },
+ { 252, 210, 217, 254, 220, 255, 250, 254 },
+ { 252, 211, 218, 254, 221, 255, 250, 254 },
+ { 252, 212, 218, 254, 221, 255, 250, 254 },
+ { 253, 213, 219, 254, 222, 255, 250, 254 },
+ { 253, 213, 220, 254, 222, 255, 250, 254 },
+ { 253, 214, 221, 254, 223, 255, 250, 254 },
+ { 253, 215, 221, 254, 223, 255, 250, 254 },
+ { 253, 216, 222, 254, 224, 255, 251, 254 },
+ { 253, 217, 223, 254, 224, 255, 251, 254 },
+ { 253, 218, 224, 254, 225, 255, 251, 254 },
+ { 253, 219, 224, 254, 225, 255, 251, 254 },
+ { 254, 220, 225, 254, 225, 255, 251, 254 },
+ { 254, 221, 226, 254, 225, 255, 251, 254 },
+ { 254, 222, 227, 255, 226, 255, 251, 254 },
+ { 254, 223, 227, 255, 226, 255, 251, 254 },
+ { 254, 224, 228, 255, 227, 255, 251, 254 },
+ { 254, 225, 229, 255, 227, 255, 251, 254 },
+ { 254, 226, 230, 255, 228, 255, 251, 254 },
+ { 254, 227, 230, 255, 229, 255, 251, 254 },
+ { 255, 228, 231, 255, 230, 255, 251, 254 },
+ { 255, 229, 232, 255, 230, 255, 251, 254 },
+ { 255, 230, 233, 255, 231, 255, 252, 254 },
+ { 255, 231, 234, 255, 231, 255, 252, 254 },
+ { 255, 232, 235, 255, 232, 255, 252, 254 },
+ { 255, 233, 236, 255, 232, 255, 252, 254 },
+ { 255, 235, 237, 255, 233, 255, 252, 254 },
+ { 255, 236, 238, 255, 234, 255, 252, 254 },
+ { 255, 238, 240, 255, 235, 255, 252, 255 },
+ { 255, 239, 241, 255, 235, 255, 252, 254 },
+ { 255, 241, 243, 255, 236, 255, 252, 254 },
+ { 255, 243, 245, 255, 237, 255, 252, 254 },
+ { 255, 246, 247, 255, 239, 255, 253, 255 },
};
static const vp10_coeff_probs_model default_coef_probs_4x4[PLANE_TYPES] = {
- { // Y plane
- { // Intra
- { // Band 0
- { 195, 29, 183 }, { 84, 49, 136 }, { 8, 42, 71 }
- }, { // Band 1
- { 31, 107, 169 }, { 35, 99, 159 }, { 17, 82, 140 },
- { 8, 66, 114 }, { 2, 44, 76 }, { 1, 19, 32 }
- }, { // Band 2
- { 40, 132, 201 }, { 29, 114, 187 }, { 13, 91, 157 },
- { 7, 75, 127 }, { 3, 58, 95 }, { 1, 28, 47 }
- }, { // Band 3
- { 69, 142, 221 }, { 42, 122, 201 }, { 15, 91, 159 },
- { 6, 67, 121 }, { 1, 42, 77 }, { 1, 17, 31 }
- }, { // Band 4
- { 102, 148, 228 }, { 67, 117, 204 }, { 17, 82, 154 },
- { 6, 59, 114 }, { 2, 39, 75 }, { 1, 15, 29 }
- }, { // Band 5
- { 156, 57, 233 }, { 119, 57, 212 }, { 58, 48, 163 },
- { 29, 40, 124 }, { 12, 30, 81 }, { 3, 12, 31 }
- }
- }, { // Inter
- { // Band 0
- { 191, 107, 226 }, { 124, 117, 204 }, { 25, 99, 155 }
- }, { // Band 1
- { 29, 148, 210 }, { 37, 126, 194 }, { 8, 93, 157 },
- { 2, 68, 118 }, { 1, 39, 69 }, { 1, 17, 33 }
- }, { // Band 2
- { 41, 151, 213 }, { 27, 123, 193 }, { 3, 82, 144 },
- { 1, 58, 105 }, { 1, 32, 60 }, { 1, 13, 26 }
- }, { // Band 3
- { 59, 159, 220 }, { 23, 126, 198 }, { 4, 88, 151 },
- { 1, 66, 114 }, { 1, 38, 71 }, { 1, 18, 34 }
- }, { // Band 4
- { 114, 136, 232 }, { 51, 114, 207 }, { 11, 83, 155 },
- { 3, 56, 105 }, { 1, 33, 65 }, { 1, 17, 34 }
- }, { // Band 5
- { 149, 65, 234 }, { 121, 57, 215 }, { 61, 49, 166 },
- { 28, 36, 114 }, { 12, 25, 76 }, { 3, 16, 42 }
- }
- }
- }, { // UV plane
- { // Intra
- { // Band 0
- { 214, 49, 220 }, { 132, 63, 188 }, { 42, 65, 137 }
- }, { // Band 1
- { 85, 137, 221 }, { 104, 131, 216 }, { 49, 111, 192 },
- { 21, 87, 155 }, { 2, 49, 87 }, { 1, 16, 28 }
- }, { // Band 2
- { 89, 163, 230 }, { 90, 137, 220 }, { 29, 100, 183 },
- { 10, 70, 135 }, { 2, 42, 81 }, { 1, 17, 33 }
- }, { // Band 3
- { 108, 167, 237 }, { 55, 133, 222 }, { 15, 97, 179 },
- { 4, 72, 135 }, { 1, 45, 85 }, { 1, 19, 38 }
- }, { // Band 4
- { 124, 146, 240 }, { 66, 124, 224 }, { 17, 88, 175 },
- { 4, 58, 122 }, { 1, 36, 75 }, { 1, 18, 37 }
- }, { // Band 5
- { 141, 79, 241 }, { 126, 70, 227 }, { 66, 58, 182 },
- { 30, 44, 136 }, { 12, 34, 96 }, { 2, 20, 47 }
- }
- }, { // Inter
- { // Band 0
- { 229, 99, 249 }, { 143, 111, 235 }, { 46, 109, 192 }
- }, { // Band 1
- { 82, 158, 236 }, { 94, 146, 224 }, { 25, 117, 191 },
- { 9, 87, 149 }, { 3, 56, 99 }, { 1, 33, 57 }
- }, { // Band 2
- { 83, 167, 237 }, { 68, 145, 222 }, { 10, 103, 177 },
- { 2, 72, 131 }, { 1, 41, 79 }, { 1, 20, 39 }
- }, { // Band 3
- { 99, 167, 239 }, { 47, 141, 224 }, { 10, 104, 178 },
- { 2, 73, 133 }, { 1, 44, 85 }, { 1, 22, 47 }
- }, { // Band 4
- { 127, 145, 243 }, { 71, 129, 228 }, { 17, 93, 177 },
- { 3, 61, 124 }, { 1, 41, 84 }, { 1, 21, 52 }
- }, { // Band 5
- { 157, 78, 244 }, { 140, 72, 231 }, { 69, 58, 184 },
- { 31, 44, 137 }, { 14, 38, 105 }, { 8, 23, 61 }
- }
- }
- }
+ { // Y plane
+ { // Intra
+ { // Band 0
+ { 195, 29, 183 },
+ { 84, 49, 136 },
+ { 8, 42, 71 } },
+ { // Band 1
+ { 31, 107, 169 },
+ { 35, 99, 159 },
+ { 17, 82, 140 },
+ { 8, 66, 114 },
+ { 2, 44, 76 },
+ { 1, 19, 32 } },
+ { // Band 2
+ { 40, 132, 201 },
+ { 29, 114, 187 },
+ { 13, 91, 157 },
+ { 7, 75, 127 },
+ { 3, 58, 95 },
+ { 1, 28, 47 } },
+ { // Band 3
+ { 69, 142, 221 },
+ { 42, 122, 201 },
+ { 15, 91, 159 },
+ { 6, 67, 121 },
+ { 1, 42, 77 },
+ { 1, 17, 31 } },
+ { // Band 4
+ { 102, 148, 228 },
+ { 67, 117, 204 },
+ { 17, 82, 154 },
+ { 6, 59, 114 },
+ { 2, 39, 75 },
+ { 1, 15, 29 } },
+ { // Band 5
+ { 156, 57, 233 },
+ { 119, 57, 212 },
+ { 58, 48, 163 },
+ { 29, 40, 124 },
+ { 12, 30, 81 },
+ { 3, 12, 31 } } },
+ { // Inter
+ { // Band 0
+ { 191, 107, 226 },
+ { 124, 117, 204 },
+ { 25, 99, 155 } },
+ { // Band 1
+ { 29, 148, 210 },
+ { 37, 126, 194 },
+ { 8, 93, 157 },
+ { 2, 68, 118 },
+ { 1, 39, 69 },
+ { 1, 17, 33 } },
+ { // Band 2
+ { 41, 151, 213 },
+ { 27, 123, 193 },
+ { 3, 82, 144 },
+ { 1, 58, 105 },
+ { 1, 32, 60 },
+ { 1, 13, 26 } },
+ { // Band 3
+ { 59, 159, 220 },
+ { 23, 126, 198 },
+ { 4, 88, 151 },
+ { 1, 66, 114 },
+ { 1, 38, 71 },
+ { 1, 18, 34 } },
+ { // Band 4
+ { 114, 136, 232 },
+ { 51, 114, 207 },
+ { 11, 83, 155 },
+ { 3, 56, 105 },
+ { 1, 33, 65 },
+ { 1, 17, 34 } },
+ { // Band 5
+ { 149, 65, 234 },
+ { 121, 57, 215 },
+ { 61, 49, 166 },
+ { 28, 36, 114 },
+ { 12, 25, 76 },
+ { 3, 16, 42 } } } },
+ { // UV plane
+ { // Intra
+ { // Band 0
+ { 214, 49, 220 },
+ { 132, 63, 188 },
+ { 42, 65, 137 } },
+ { // Band 1
+ { 85, 137, 221 },
+ { 104, 131, 216 },
+ { 49, 111, 192 },
+ { 21, 87, 155 },
+ { 2, 49, 87 },
+ { 1, 16, 28 } },
+ { // Band 2
+ { 89, 163, 230 },
+ { 90, 137, 220 },
+ { 29, 100, 183 },
+ { 10, 70, 135 },
+ { 2, 42, 81 },
+ { 1, 17, 33 } },
+ { // Band 3
+ { 108, 167, 237 },
+ { 55, 133, 222 },
+ { 15, 97, 179 },
+ { 4, 72, 135 },
+ { 1, 45, 85 },
+ { 1, 19, 38 } },
+ { // Band 4
+ { 124, 146, 240 },
+ { 66, 124, 224 },
+ { 17, 88, 175 },
+ { 4, 58, 122 },
+ { 1, 36, 75 },
+ { 1, 18, 37 } },
+ { // Band 5
+ { 141, 79, 241 },
+ { 126, 70, 227 },
+ { 66, 58, 182 },
+ { 30, 44, 136 },
+ { 12, 34, 96 },
+ { 2, 20, 47 } } },
+ { // Inter
+ { // Band 0
+ { 229, 99, 249 },
+ { 143, 111, 235 },
+ { 46, 109, 192 } },
+ { // Band 1
+ { 82, 158, 236 },
+ { 94, 146, 224 },
+ { 25, 117, 191 },
+ { 9, 87, 149 },
+ { 3, 56, 99 },
+ { 1, 33, 57 } },
+ { // Band 2
+ { 83, 167, 237 },
+ { 68, 145, 222 },
+ { 10, 103, 177 },
+ { 2, 72, 131 },
+ { 1, 41, 79 },
+ { 1, 20, 39 } },
+ { // Band 3
+ { 99, 167, 239 },
+ { 47, 141, 224 },
+ { 10, 104, 178 },
+ { 2, 73, 133 },
+ { 1, 44, 85 },
+ { 1, 22, 47 } },
+ { // Band 4
+ { 127, 145, 243 },
+ { 71, 129, 228 },
+ { 17, 93, 177 },
+ { 3, 61, 124 },
+ { 1, 41, 84 },
+ { 1, 21, 52 } },
+ { // Band 5
+ { 157, 78, 244 },
+ { 140, 72, 231 },
+ { 69, 58, 184 },
+ { 31, 44, 137 },
+ { 14, 38, 105 },
+ { 8, 23, 61 } } } }
};
static const vp10_coeff_probs_model default_coef_probs_8x8[PLANE_TYPES] = {
- { // Y plane
- { // Intra
- { // Band 0
- { 125, 34, 187 }, { 52, 41, 133 }, { 6, 31, 56 }
- }, { // Band 1
- { 37, 109, 153 }, { 51, 102, 147 }, { 23, 87, 128 },
- { 8, 67, 101 }, { 1, 41, 63 }, { 1, 19, 29 }
- }, { // Band 2
- { 31, 154, 185 }, { 17, 127, 175 }, { 6, 96, 145 },
- { 2, 73, 114 }, { 1, 51, 82 }, { 1, 28, 45 }
- }, { // Band 3
- { 23, 163, 200 }, { 10, 131, 185 }, { 2, 93, 148 },
- { 1, 67, 111 }, { 1, 41, 69 }, { 1, 14, 24 }
- }, { // Band 4
- { 29, 176, 217 }, { 12, 145, 201 }, { 3, 101, 156 },
- { 1, 69, 111 }, { 1, 39, 63 }, { 1, 14, 23 }
- }, { // Band 5
- { 57, 192, 233 }, { 25, 154, 215 }, { 6, 109, 167 },
- { 3, 78, 118 }, { 1, 48, 69 }, { 1, 21, 29 }
- }
- }, { // Inter
- { // Band 0
- { 202, 105, 245 }, { 108, 106, 216 }, { 18, 90, 144 }
- }, { // Band 1
- { 33, 172, 219 }, { 64, 149, 206 }, { 14, 117, 177 },
- { 5, 90, 141 }, { 2, 61, 95 }, { 1, 37, 57 }
- }, { // Band 2
- { 33, 179, 220 }, { 11, 140, 198 }, { 1, 89, 148 },
- { 1, 60, 104 }, { 1, 33, 57 }, { 1, 12, 21 }
- }, { // Band 3
- { 30, 181, 221 }, { 8, 141, 198 }, { 1, 87, 145 },
- { 1, 58, 100 }, { 1, 31, 55 }, { 1, 12, 20 }
- }, { // Band 4
- { 32, 186, 224 }, { 7, 142, 198 }, { 1, 86, 143 },
- { 1, 58, 100 }, { 1, 31, 55 }, { 1, 12, 22 }
- }, { // Band 5
- { 57, 192, 227 }, { 20, 143, 204 }, { 3, 96, 154 },
- { 1, 68, 112 }, { 1, 42, 69 }, { 1, 19, 32 }
- }
- }
- }, { // UV plane
- { // Intra
- { // Band 0
- { 212, 35, 215 }, { 113, 47, 169 }, { 29, 48, 105 }
- }, { // Band 1
- { 74, 129, 203 }, { 106, 120, 203 }, { 49, 107, 178 },
- { 19, 84, 144 }, { 4, 50, 84 }, { 1, 15, 25 }
- }, { // Band 2
- { 71, 172, 217 }, { 44, 141, 209 }, { 15, 102, 173 },
- { 6, 76, 133 }, { 2, 51, 89 }, { 1, 24, 42 }
- }, { // Band 3
- { 64, 185, 231 }, { 31, 148, 216 }, { 8, 103, 175 },
- { 3, 74, 131 }, { 1, 46, 81 }, { 1, 18, 30 }
- }, { // Band 4
- { 65, 196, 235 }, { 25, 157, 221 }, { 5, 105, 174 },
- { 1, 67, 120 }, { 1, 38, 69 }, { 1, 15, 30 }
- }, { // Band 5
- { 65, 204, 238 }, { 30, 156, 224 }, { 7, 107, 177 },
- { 2, 70, 124 }, { 1, 42, 73 }, { 1, 18, 34 }
- }
- }, { // Inter
- { // Band 0
- { 225, 86, 251 }, { 144, 104, 235 }, { 42, 99, 181 }
- }, { // Band 1
- { 85, 175, 239 }, { 112, 165, 229 }, { 29, 136, 200 },
- { 12, 103, 162 }, { 6, 77, 123 }, { 2, 53, 84 }
- }, { // Band 2
- { 75, 183, 239 }, { 30, 155, 221 }, { 3, 106, 171 },
- { 1, 74, 128 }, { 1, 44, 76 }, { 1, 17, 28 }
- }, { // Band 3
- { 73, 185, 240 }, { 27, 159, 222 }, { 2, 107, 172 },
- { 1, 75, 127 }, { 1, 42, 73 }, { 1, 17, 29 }
- }, { // Band 4
- { 62, 190, 238 }, { 21, 159, 222 }, { 2, 107, 172 },
- { 1, 72, 122 }, { 1, 40, 71 }, { 1, 18, 32 }
- }, { // Band 5
- { 61, 199, 240 }, { 27, 161, 226 }, { 4, 113, 180 },
- { 1, 76, 129 }, { 1, 46, 80 }, { 1, 23, 41 }
- }
- }
- }
+ { // Y plane
+ { // Intra
+ { // Band 0
+ { 125, 34, 187 },
+ { 52, 41, 133 },
+ { 6, 31, 56 } },
+ { // Band 1
+ { 37, 109, 153 },
+ { 51, 102, 147 },
+ { 23, 87, 128 },
+ { 8, 67, 101 },
+ { 1, 41, 63 },
+ { 1, 19, 29 } },
+ { // Band 2
+ { 31, 154, 185 },
+ { 17, 127, 175 },
+ { 6, 96, 145 },
+ { 2, 73, 114 },
+ { 1, 51, 82 },
+ { 1, 28, 45 } },
+ { // Band 3
+ { 23, 163, 200 },
+ { 10, 131, 185 },
+ { 2, 93, 148 },
+ { 1, 67, 111 },
+ { 1, 41, 69 },
+ { 1, 14, 24 } },
+ { // Band 4
+ { 29, 176, 217 },
+ { 12, 145, 201 },
+ { 3, 101, 156 },
+ { 1, 69, 111 },
+ { 1, 39, 63 },
+ { 1, 14, 23 } },
+ { // Band 5
+ { 57, 192, 233 },
+ { 25, 154, 215 },
+ { 6, 109, 167 },
+ { 3, 78, 118 },
+ { 1, 48, 69 },
+ { 1, 21, 29 } } },
+ { // Inter
+ { // Band 0
+ { 202, 105, 245 },
+ { 108, 106, 216 },
+ { 18, 90, 144 } },
+ { // Band 1
+ { 33, 172, 219 },
+ { 64, 149, 206 },
+ { 14, 117, 177 },
+ { 5, 90, 141 },
+ { 2, 61, 95 },
+ { 1, 37, 57 } },
+ { // Band 2
+ { 33, 179, 220 },
+ { 11, 140, 198 },
+ { 1, 89, 148 },
+ { 1, 60, 104 },
+ { 1, 33, 57 },
+ { 1, 12, 21 } },
+ { // Band 3
+ { 30, 181, 221 },
+ { 8, 141, 198 },
+ { 1, 87, 145 },
+ { 1, 58, 100 },
+ { 1, 31, 55 },
+ { 1, 12, 20 } },
+ { // Band 4
+ { 32, 186, 224 },
+ { 7, 142, 198 },
+ { 1, 86, 143 },
+ { 1, 58, 100 },
+ { 1, 31, 55 },
+ { 1, 12, 22 } },
+ { // Band 5
+ { 57, 192, 227 },
+ { 20, 143, 204 },
+ { 3, 96, 154 },
+ { 1, 68, 112 },
+ { 1, 42, 69 },
+ { 1, 19, 32 } } } },
+ { // UV plane
+ { // Intra
+ { // Band 0
+ { 212, 35, 215 },
+ { 113, 47, 169 },
+ { 29, 48, 105 } },
+ { // Band 1
+ { 74, 129, 203 },
+ { 106, 120, 203 },
+ { 49, 107, 178 },
+ { 19, 84, 144 },
+ { 4, 50, 84 },
+ { 1, 15, 25 } },
+ { // Band 2
+ { 71, 172, 217 },
+ { 44, 141, 209 },
+ { 15, 102, 173 },
+ { 6, 76, 133 },
+ { 2, 51, 89 },
+ { 1, 24, 42 } },
+ { // Band 3
+ { 64, 185, 231 },
+ { 31, 148, 216 },
+ { 8, 103, 175 },
+ { 3, 74, 131 },
+ { 1, 46, 81 },
+ { 1, 18, 30 } },
+ { // Band 4
+ { 65, 196, 235 },
+ { 25, 157, 221 },
+ { 5, 105, 174 },
+ { 1, 67, 120 },
+ { 1, 38, 69 },
+ { 1, 15, 30 } },
+ { // Band 5
+ { 65, 204, 238 },
+ { 30, 156, 224 },
+ { 7, 107, 177 },
+ { 2, 70, 124 },
+ { 1, 42, 73 },
+ { 1, 18, 34 } } },
+ { // Inter
+ { // Band 0
+ { 225, 86, 251 },
+ { 144, 104, 235 },
+ { 42, 99, 181 } },
+ { // Band 1
+ { 85, 175, 239 },
+ { 112, 165, 229 },
+ { 29, 136, 200 },
+ { 12, 103, 162 },
+ { 6, 77, 123 },
+ { 2, 53, 84 } },
+ { // Band 2
+ { 75, 183, 239 },
+ { 30, 155, 221 },
+ { 3, 106, 171 },
+ { 1, 74, 128 },
+ { 1, 44, 76 },
+ { 1, 17, 28 } },
+ { // Band 3
+ { 73, 185, 240 },
+ { 27, 159, 222 },
+ { 2, 107, 172 },
+ { 1, 75, 127 },
+ { 1, 42, 73 },
+ { 1, 17, 29 } },
+ { // Band 4
+ { 62, 190, 238 },
+ { 21, 159, 222 },
+ { 2, 107, 172 },
+ { 1, 72, 122 },
+ { 1, 40, 71 },
+ { 1, 18, 32 } },
+ { // Band 5
+ { 61, 199, 240 },
+ { 27, 161, 226 },
+ { 4, 113, 180 },
+ { 1, 76, 129 },
+ { 1, 46, 80 },
+ { 1, 23, 41 } } } }
};
static const vp10_coeff_probs_model default_coef_probs_16x16[PLANE_TYPES] = {
- { // Y plane
- { // Intra
- { // Band 0
- { 7, 27, 153 }, { 5, 30, 95 }, { 1, 16, 30 }
- }, { // Band 1
- { 50, 75, 127 }, { 57, 75, 124 }, { 27, 67, 108 },
- { 10, 54, 86 }, { 1, 33, 52 }, { 1, 12, 18 }
- }, { // Band 2
- { 43, 125, 151 }, { 26, 108, 148 }, { 7, 83, 122 },
- { 2, 59, 89 }, { 1, 38, 60 }, { 1, 17, 27 }
- }, { // Band 3
- { 23, 144, 163 }, { 13, 112, 154 }, { 2, 75, 117 },
- { 1, 50, 81 }, { 1, 31, 51 }, { 1, 14, 23 }
- }, { // Band 4
- { 18, 162, 185 }, { 6, 123, 171 }, { 1, 78, 125 },
- { 1, 51, 86 }, { 1, 31, 54 }, { 1, 14, 23 }
- }, { // Band 5
- { 15, 199, 227 }, { 3, 150, 204 }, { 1, 91, 146 },
- { 1, 55, 95 }, { 1, 30, 53 }, { 1, 11, 20 }
- }
- }, { // Inter
- { // Band 0
- { 19, 55, 240 }, { 19, 59, 196 }, { 3, 52, 105 }
- }, { // Band 1
- { 41, 166, 207 }, { 104, 153, 199 }, { 31, 123, 181 },
- { 14, 101, 152 }, { 5, 72, 106 }, { 1, 36, 52 }
- }, { // Band 2
- { 35, 176, 211 }, { 12, 131, 190 }, { 2, 88, 144 },
- { 1, 60, 101 }, { 1, 36, 60 }, { 1, 16, 28 }
- }, { // Band 3
- { 28, 183, 213 }, { 8, 134, 191 }, { 1, 86, 142 },
- { 1, 56, 96 }, { 1, 30, 53 }, { 1, 12, 20 }
- }, { // Band 4
- { 20, 190, 215 }, { 4, 135, 192 }, { 1, 84, 139 },
- { 1, 53, 91 }, { 1, 28, 49 }, { 1, 11, 20 }
- }, { // Band 5
- { 13, 196, 216 }, { 2, 137, 192 }, { 1, 86, 143 },
- { 1, 57, 99 }, { 1, 32, 56 }, { 1, 13, 24 }
- }
- }
- }, { // UV plane
- { // Intra
- { // Band 0
- { 211, 29, 217 }, { 96, 47, 156 }, { 22, 43, 87 }
- }, { // Band 1
- { 78, 120, 193 }, { 111, 116, 186 }, { 46, 102, 164 },
- { 15, 80, 128 }, { 2, 49, 76 }, { 1, 18, 28 }
- }, { // Band 2
- { 71, 161, 203 }, { 42, 132, 192 }, { 10, 98, 150 },
- { 3, 69, 109 }, { 1, 44, 70 }, { 1, 18, 29 }
- }, { // Band 3
- { 57, 186, 211 }, { 30, 140, 196 }, { 4, 93, 146 },
- { 1, 62, 102 }, { 1, 38, 65 }, { 1, 16, 27 }
- }, { // Band 4
- { 47, 199, 217 }, { 14, 145, 196 }, { 1, 88, 142 },
- { 1, 57, 98 }, { 1, 36, 62 }, { 1, 15, 26 }
- }, { // Band 5
- { 26, 219, 229 }, { 5, 155, 207 }, { 1, 94, 151 },
- { 1, 60, 104 }, { 1, 36, 62 }, { 1, 16, 28 }
- }
- }, { // Inter
- { // Band 0
- { 233, 29, 248 }, { 146, 47, 220 }, { 43, 52, 140 }
- }, { // Band 1
- { 100, 163, 232 }, { 179, 161, 222 }, { 63, 142, 204 },
- { 37, 113, 174 }, { 26, 89, 137 }, { 18, 68, 97 }
- }, { // Band 2
- { 85, 181, 230 }, { 32, 146, 209 }, { 7, 100, 164 },
- { 3, 71, 121 }, { 1, 45, 77 }, { 1, 18, 30 }
- }, { // Band 3
- { 65, 187, 230 }, { 20, 148, 207 }, { 2, 97, 159 },
- { 1, 68, 116 }, { 1, 40, 70 }, { 1, 14, 29 }
- }, { // Band 4
- { 40, 194, 227 }, { 8, 147, 204 }, { 1, 94, 155 },
- { 1, 65, 112 }, { 1, 39, 66 }, { 1, 14, 26 }
- }, { // Band 5
- { 16, 208, 228 }, { 3, 151, 207 }, { 1, 98, 160 },
- { 1, 67, 117 }, { 1, 41, 74 }, { 1, 17, 31 }
- }
- }
- }
+ { // Y plane
+ { // Intra
+ { // Band 0
+ { 7, 27, 153 },
+ { 5, 30, 95 },
+ { 1, 16, 30 } },
+ { // Band 1
+ { 50, 75, 127 },
+ { 57, 75, 124 },
+ { 27, 67, 108 },
+ { 10, 54, 86 },
+ { 1, 33, 52 },
+ { 1, 12, 18 } },
+ { // Band 2
+ { 43, 125, 151 },
+ { 26, 108, 148 },
+ { 7, 83, 122 },
+ { 2, 59, 89 },
+ { 1, 38, 60 },
+ { 1, 17, 27 } },
+ { // Band 3
+ { 23, 144, 163 },
+ { 13, 112, 154 },
+ { 2, 75, 117 },
+ { 1, 50, 81 },
+ { 1, 31, 51 },
+ { 1, 14, 23 } },
+ { // Band 4
+ { 18, 162, 185 },
+ { 6, 123, 171 },
+ { 1, 78, 125 },
+ { 1, 51, 86 },
+ { 1, 31, 54 },
+ { 1, 14, 23 } },
+ { // Band 5
+ { 15, 199, 227 },
+ { 3, 150, 204 },
+ { 1, 91, 146 },
+ { 1, 55, 95 },
+ { 1, 30, 53 },
+ { 1, 11, 20 } } },
+ { // Inter
+ { // Band 0
+ { 19, 55, 240 },
+ { 19, 59, 196 },
+ { 3, 52, 105 } },
+ { // Band 1
+ { 41, 166, 207 },
+ { 104, 153, 199 },
+ { 31, 123, 181 },
+ { 14, 101, 152 },
+ { 5, 72, 106 },
+ { 1, 36, 52 } },
+ { // Band 2
+ { 35, 176, 211 },
+ { 12, 131, 190 },
+ { 2, 88, 144 },
+ { 1, 60, 101 },
+ { 1, 36, 60 },
+ { 1, 16, 28 } },
+ { // Band 3
+ { 28, 183, 213 },
+ { 8, 134, 191 },
+ { 1, 86, 142 },
+ { 1, 56, 96 },
+ { 1, 30, 53 },
+ { 1, 12, 20 } },
+ { // Band 4
+ { 20, 190, 215 },
+ { 4, 135, 192 },
+ { 1, 84, 139 },
+ { 1, 53, 91 },
+ { 1, 28, 49 },
+ { 1, 11, 20 } },
+ { // Band 5
+ { 13, 196, 216 },
+ { 2, 137, 192 },
+ { 1, 86, 143 },
+ { 1, 57, 99 },
+ { 1, 32, 56 },
+ { 1, 13, 24 } } } },
+ { // UV plane
+ { // Intra
+ { // Band 0
+ { 211, 29, 217 },
+ { 96, 47, 156 },
+ { 22, 43, 87 } },
+ { // Band 1
+ { 78, 120, 193 },
+ { 111, 116, 186 },
+ { 46, 102, 164 },
+ { 15, 80, 128 },
+ { 2, 49, 76 },
+ { 1, 18, 28 } },
+ { // Band 2
+ { 71, 161, 203 },
+ { 42, 132, 192 },
+ { 10, 98, 150 },
+ { 3, 69, 109 },
+ { 1, 44, 70 },
+ { 1, 18, 29 } },
+ { // Band 3
+ { 57, 186, 211 },
+ { 30, 140, 196 },
+ { 4, 93, 146 },
+ { 1, 62, 102 },
+ { 1, 38, 65 },
+ { 1, 16, 27 } },
+ { // Band 4
+ { 47, 199, 217 },
+ { 14, 145, 196 },
+ { 1, 88, 142 },
+ { 1, 57, 98 },
+ { 1, 36, 62 },
+ { 1, 15, 26 } },
+ { // Band 5
+ { 26, 219, 229 },
+ { 5, 155, 207 },
+ { 1, 94, 151 },
+ { 1, 60, 104 },
+ { 1, 36, 62 },
+ { 1, 16, 28 } } },
+ { // Inter
+ { // Band 0
+ { 233, 29, 248 },
+ { 146, 47, 220 },
+ { 43, 52, 140 } },
+ { // Band 1
+ { 100, 163, 232 },
+ { 179, 161, 222 },
+ { 63, 142, 204 },
+ { 37, 113, 174 },
+ { 26, 89, 137 },
+ { 18, 68, 97 } },
+ { // Band 2
+ { 85, 181, 230 },
+ { 32, 146, 209 },
+ { 7, 100, 164 },
+ { 3, 71, 121 },
+ { 1, 45, 77 },
+ { 1, 18, 30 } },
+ { // Band 3
+ { 65, 187, 230 },
+ { 20, 148, 207 },
+ { 2, 97, 159 },
+ { 1, 68, 116 },
+ { 1, 40, 70 },
+ { 1, 14, 29 } },
+ { // Band 4
+ { 40, 194, 227 },
+ { 8, 147, 204 },
+ { 1, 94, 155 },
+ { 1, 65, 112 },
+ { 1, 39, 66 },
+ { 1, 14, 26 } },
+ { // Band 5
+ { 16, 208, 228 },
+ { 3, 151, 207 },
+ { 1, 98, 160 },
+ { 1, 67, 117 },
+ { 1, 41, 74 },
+ { 1, 17, 31 } } } }
};
static const vp10_coeff_probs_model default_coef_probs_32x32[PLANE_TYPES] = {
- { // Y plane
- { // Intra
- { // Band 0
- { 17, 38, 140 }, { 7, 34, 80 }, { 1, 17, 29 }
- }, { // Band 1
- { 37, 75, 128 }, { 41, 76, 128 }, { 26, 66, 116 },
- { 12, 52, 94 }, { 2, 32, 55 }, { 1, 10, 16 }
- }, { // Band 2
- { 50, 127, 154 }, { 37, 109, 152 }, { 16, 82, 121 },
- { 5, 59, 85 }, { 1, 35, 54 }, { 1, 13, 20 }
- }, { // Band 3
- { 40, 142, 167 }, { 17, 110, 157 }, { 2, 71, 112 },
- { 1, 44, 72 }, { 1, 27, 45 }, { 1, 11, 17 }
- }, { // Band 4
- { 30, 175, 188 }, { 9, 124, 169 }, { 1, 74, 116 },
- { 1, 48, 78 }, { 1, 30, 49 }, { 1, 11, 18 }
- }, { // Band 5
- { 10, 222, 223 }, { 2, 150, 194 }, { 1, 83, 128 },
- { 1, 48, 79 }, { 1, 27, 45 }, { 1, 11, 17 }
- }
- }, { // Inter
- { // Band 0
- { 36, 41, 235 }, { 29, 36, 193 }, { 10, 27, 111 }
- }, { // Band 1
- { 85, 165, 222 }, { 177, 162, 215 }, { 110, 135, 195 },
- { 57, 113, 168 }, { 23, 83, 120 }, { 10, 49, 61 }
- }, { // Band 2
- { 85, 190, 223 }, { 36, 139, 200 }, { 5, 90, 146 },
- { 1, 60, 103 }, { 1, 38, 65 }, { 1, 18, 30 }
- }, { // Band 3
- { 72, 202, 223 }, { 23, 141, 199 }, { 2, 86, 140 },
- { 1, 56, 97 }, { 1, 36, 61 }, { 1, 16, 27 }
- }, { // Band 4
- { 55, 218, 225 }, { 13, 145, 200 }, { 1, 86, 141 },
- { 1, 57, 99 }, { 1, 35, 61 }, { 1, 13, 22 }
- }, { // Band 5
- { 15, 235, 212 }, { 1, 132, 184 }, { 1, 84, 139 },
- { 1, 57, 97 }, { 1, 34, 56 }, { 1, 14, 23 }
- }
- }
- }, { // UV plane
- { // Intra
- { // Band 0
- { 181, 21, 201 }, { 61, 37, 123 }, { 10, 38, 71 }
- }, { // Band 1
- { 47, 106, 172 }, { 95, 104, 173 }, { 42, 93, 159 },
- { 18, 77, 131 }, { 4, 50, 81 }, { 1, 17, 23 }
- }, { // Band 2
- { 62, 147, 199 }, { 44, 130, 189 }, { 28, 102, 154 },
- { 18, 75, 115 }, { 2, 44, 65 }, { 1, 12, 19 }
- }, { // Band 3
- { 55, 153, 210 }, { 24, 130, 194 }, { 3, 93, 146 },
- { 1, 61, 97 }, { 1, 31, 50 }, { 1, 10, 16 }
- }, { // Band 4
- { 49, 186, 223 }, { 17, 148, 204 }, { 1, 96, 142 },
- { 1, 53, 83 }, { 1, 26, 44 }, { 1, 11, 17 }
- }, { // Band 5
- { 13, 217, 212 }, { 2, 136, 180 }, { 1, 78, 124 },
- { 1, 50, 83 }, { 1, 29, 49 }, { 1, 14, 23 }
- }
- }, { // Inter
- { // Band 0
- { 197, 13, 247 }, { 82, 17, 222 }, { 25, 17, 162 }
- }, { // Band 1
- { 126, 186, 247 }, { 234, 191, 243 }, { 176, 177, 234 },
- { 104, 158, 220 }, { 66, 128, 186 }, { 55, 90, 137 }
- }, { // Band 2
- { 111, 197, 242 }, { 46, 158, 219 }, { 9, 104, 171 },
- { 2, 65, 125 }, { 1, 44, 80 }, { 1, 17, 91 }
- }, { // Band 3
- { 104, 208, 245 }, { 39, 168, 224 }, { 3, 109, 162 },
- { 1, 79, 124 }, { 1, 50, 102 }, { 1, 43, 102 }
- }, { // Band 4
- { 84, 220, 246 }, { 31, 177, 231 }, { 2, 115, 180 },
- { 1, 79, 134 }, { 1, 55, 77 }, { 1, 60, 79 }
- }, { // Band 5
- { 43, 243, 240 }, { 8, 180, 217 }, { 1, 115, 166 },
- { 1, 84, 121 }, { 1, 51, 67 }, { 1, 16, 6 }
- }
- }
- }
+ { // Y plane
+ { // Intra
+ { // Band 0
+ { 17, 38, 140 },
+ { 7, 34, 80 },
+ { 1, 17, 29 } },
+ { // Band 1
+ { 37, 75, 128 },
+ { 41, 76, 128 },
+ { 26, 66, 116 },
+ { 12, 52, 94 },
+ { 2, 32, 55 },
+ { 1, 10, 16 } },
+ { // Band 2
+ { 50, 127, 154 },
+ { 37, 109, 152 },
+ { 16, 82, 121 },
+ { 5, 59, 85 },
+ { 1, 35, 54 },
+ { 1, 13, 20 } },
+ { // Band 3
+ { 40, 142, 167 },
+ { 17, 110, 157 },
+ { 2, 71, 112 },
+ { 1, 44, 72 },
+ { 1, 27, 45 },
+ { 1, 11, 17 } },
+ { // Band 4
+ { 30, 175, 188 },
+ { 9, 124, 169 },
+ { 1, 74, 116 },
+ { 1, 48, 78 },
+ { 1, 30, 49 },
+ { 1, 11, 18 } },
+ { // Band 5
+ { 10, 222, 223 },
+ { 2, 150, 194 },
+ { 1, 83, 128 },
+ { 1, 48, 79 },
+ { 1, 27, 45 },
+ { 1, 11, 17 } } },
+ { // Inter
+ { // Band 0
+ { 36, 41, 235 },
+ { 29, 36, 193 },
+ { 10, 27, 111 } },
+ { // Band 1
+ { 85, 165, 222 },
+ { 177, 162, 215 },
+ { 110, 135, 195 },
+ { 57, 113, 168 },
+ { 23, 83, 120 },
+ { 10, 49, 61 } },
+ { // Band 2
+ { 85, 190, 223 },
+ { 36, 139, 200 },
+ { 5, 90, 146 },
+ { 1, 60, 103 },
+ { 1, 38, 65 },
+ { 1, 18, 30 } },
+ { // Band 3
+ { 72, 202, 223 },
+ { 23, 141, 199 },
+ { 2, 86, 140 },
+ { 1, 56, 97 },
+ { 1, 36, 61 },
+ { 1, 16, 27 } },
+ { // Band 4
+ { 55, 218, 225 },
+ { 13, 145, 200 },
+ { 1, 86, 141 },
+ { 1, 57, 99 },
+ { 1, 35, 61 },
+ { 1, 13, 22 } },
+ { // Band 5
+ { 15, 235, 212 },
+ { 1, 132, 184 },
+ { 1, 84, 139 },
+ { 1, 57, 97 },
+ { 1, 34, 56 },
+ { 1, 14, 23 } } } },
+ { // UV plane
+ { // Intra
+ { // Band 0
+ { 181, 21, 201 },
+ { 61, 37, 123 },
+ { 10, 38, 71 } },
+ { // Band 1
+ { 47, 106, 172 },
+ { 95, 104, 173 },
+ { 42, 93, 159 },
+ { 18, 77, 131 },
+ { 4, 50, 81 },
+ { 1, 17, 23 } },
+ { // Band 2
+ { 62, 147, 199 },
+ { 44, 130, 189 },
+ { 28, 102, 154 },
+ { 18, 75, 115 },
+ { 2, 44, 65 },
+ { 1, 12, 19 } },
+ { // Band 3
+ { 55, 153, 210 },
+ { 24, 130, 194 },
+ { 3, 93, 146 },
+ { 1, 61, 97 },
+ { 1, 31, 50 },
+ { 1, 10, 16 } },
+ { // Band 4
+ { 49, 186, 223 },
+ { 17, 148, 204 },
+ { 1, 96, 142 },
+ { 1, 53, 83 },
+ { 1, 26, 44 },
+ { 1, 11, 17 } },
+ { // Band 5
+ { 13, 217, 212 },
+ { 2, 136, 180 },
+ { 1, 78, 124 },
+ { 1, 50, 83 },
+ { 1, 29, 49 },
+ { 1, 14, 23 } } },
+ { // Inter
+ { // Band 0
+ { 197, 13, 247 },
+ { 82, 17, 222 },
+ { 25, 17, 162 } },
+ { // Band 1
+ { 126, 186, 247 },
+ { 234, 191, 243 },
+ { 176, 177, 234 },
+ { 104, 158, 220 },
+ { 66, 128, 186 },
+ { 55, 90, 137 } },
+ { // Band 2
+ { 111, 197, 242 },
+ { 46, 158, 219 },
+ { 9, 104, 171 },
+ { 2, 65, 125 },
+ { 1, 44, 80 },
+ { 1, 17, 91 } },
+ { // Band 3
+ { 104, 208, 245 },
+ { 39, 168, 224 },
+ { 3, 109, 162 },
+ { 1, 79, 124 },
+ { 1, 50, 102 },
+ { 1, 43, 102 } },
+ { // Band 4
+ { 84, 220, 246 },
+ { 31, 177, 231 },
+ { 2, 115, 180 },
+ { 1, 79, 134 },
+ { 1, 55, 77 },
+ { 1, 60, 79 } },
+ { // Band 5
+ { 43, 243, 240 },
+ { 8, 180, 217 },
+ { 1, 115, 166 },
+ { 1, 84, 121 },
+ { 1, 51, 67 },
+ { 1, 16, 6 } } } }
};
static void extend_to_full_distribution(vpx_prob *probs, vpx_prob p) {
@@ -775,7 +1078,7 @@
vp10_coeff_probs_model *const probs = cm->fc->coef_probs[tx_size];
const vp10_coeff_probs_model *const pre_probs = pre_fc->coef_probs[tx_size];
vp10_coeff_count_model *counts = cm->counts.coef[tx_size];
- unsigned int (*eob_counts)[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS] =
+ unsigned int(*eob_counts)[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS] =
cm->counts.eob_branch[tx_size];
int i, j, k, l, m;
@@ -788,14 +1091,12 @@
const int n2 = counts[i][j][k][l][TWO_TOKEN];
const int neob = counts[i][j][k][l][EOB_MODEL_TOKEN];
const unsigned int branch_ct[UNCONSTRAINED_NODES][2] = {
- { neob, eob_counts[i][j][k][l] - neob },
- { n0, n1 + n2 },
- { n1, n2 }
+ { neob, eob_counts[i][j][k][l] - neob }, { n0, n1 + n2 }, { n1, n2 }
};
for (m = 0; m < UNCONSTRAINED_NODES; ++m)
- probs[i][j][k][l][m] = merge_probs(pre_probs[i][j][k][l][m],
- branch_ct[m],
- count_sat, update_factor);
+ probs[i][j][k][l][m] =
+ merge_probs(pre_probs[i][j][k][l][m], branch_ct[m], count_sat,
+ update_factor);
}
}
@@ -807,7 +1108,7 @@
update_factor = COEF_MAX_UPDATE_FACTOR_KEY;
count_sat = COEF_COUNT_SAT_KEY;
} else if (cm->last_frame_type == KEY_FRAME) {
- update_factor = COEF_MAX_UPDATE_FACTOR_AFTER_KEY; /* adapt quickly */
+ update_factor = COEF_MAX_UPDATE_FACTOR_AFTER_KEY; /* adapt quickly */
count_sat = COEF_COUNT_SAT_AFTER_KEY;
} else {
update_factor = COEF_MAX_UPDATE_FACTOR;
diff --git a/vp10/common/entropy.h b/vp10/common/entropy.h
index dd62ed2..3dd137c 100644
--- a/vp10/common/entropy.h
+++ b/vp10/common/entropy.h
@@ -21,22 +21,22 @@
extern "C" {
#endif
-#define DIFF_UPDATE_PROB 252
-#define GROUP_DIFF_UPDATE_PROB 252
+#define DIFF_UPDATE_PROB 252
+#define GROUP_DIFF_UPDATE_PROB 252
// Coefficient token alphabet
-#define ZERO_TOKEN 0 // 0 Extra Bits 0+0
-#define ONE_TOKEN 1 // 1 Extra Bits 0+1
-#define TWO_TOKEN 2 // 2 Extra Bits 0+1
-#define THREE_TOKEN 3 // 3 Extra Bits 0+1
-#define FOUR_TOKEN 4 // 4 Extra Bits 0+1
+#define ZERO_TOKEN 0 // 0 Extra Bits 0+0
+#define ONE_TOKEN 1 // 1 Extra Bits 0+1
+#define TWO_TOKEN 2 // 2 Extra Bits 0+1
+#define THREE_TOKEN 3 // 3 Extra Bits 0+1
+#define FOUR_TOKEN 4 // 4 Extra Bits 0+1
#define CATEGORY1_TOKEN 5 // 5-6 Extra Bits 1+1
#define CATEGORY2_TOKEN 6 // 7-10 Extra Bits 2+1
#define CATEGORY3_TOKEN 7 // 11-18 Extra Bits 3+1
#define CATEGORY4_TOKEN 8 // 19-34 Extra Bits 4+1
#define CATEGORY5_TOKEN 9 // 35-66 Extra Bits 5+1
#define CATEGORY6_TOKEN 10 // 67+ Extra Bits 14+1
-#define EOB_TOKEN 11 // EOB Extra Bits 0+0
+#define EOB_TOKEN 11 // EOB Extra Bits 0+0
#define ENTROPY_TOKENS 12
@@ -44,12 +44,12 @@
DECLARE_ALIGNED(16, extern const uint8_t, vp10_pt_energy_class[ENTROPY_TOKENS]);
-#define CAT1_MIN_VAL 5
-#define CAT2_MIN_VAL 7
-#define CAT3_MIN_VAL 11
-#define CAT4_MIN_VAL 19
-#define CAT5_MIN_VAL 35
-#define CAT6_MIN_VAL 67
+#define CAT1_MIN_VAL 5
+#define CAT2_MIN_VAL 7
+#define CAT3_MIN_VAL 11
+#define CAT4_MIN_VAL 19
+#define CAT5_MIN_VAL 35
+#define CAT6_MIN_VAL 67
// Extra bit probabilities.
DECLARE_ALIGNED(16, extern const uint8_t, vp10_cat1_prob[1]);
@@ -91,10 +91,10 @@
extern const vp10_extra_bit vp10_extra_bits_high12[ENTROPY_TOKENS];
#endif // CONFIG_VPX_HIGHBITDEPTH
-#define DCT_MAX_VALUE 16384
+#define DCT_MAX_VALUE 16384
#if CONFIG_VPX_HIGHBITDEPTH
-#define DCT_MAX_VALUE_HIGH10 65536
-#define DCT_MAX_VALUE_HIGH12 262144
+#define DCT_MAX_VALUE_HIGH10 65536
+#define DCT_MAX_VALUE_HIGH12 262144
#endif // CONFIG_VPX_HIGHBITDEPTH
/* Coefficients are predicted via a 3-dimensional probability table. */
@@ -125,13 +125,13 @@
// #define ENTROPY_STATS
-typedef unsigned int vp10_coeff_count[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS]
- [ENTROPY_TOKENS];
-typedef unsigned int vp10_coeff_stats[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS]
- [ENTROPY_NODES][2];
+typedef unsigned int
+ vp10_coeff_count[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS][ENTROPY_TOKENS];
+typedef unsigned int
+ vp10_coeff_stats[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS][ENTROPY_NODES][2];
-#define SUBEXP_PARAM 4 /* Subexponential code parameter */
-#define MODULUS_PARAM 13 /* Modulus parameter */
+#define SUBEXP_PARAM 4 /* Subexponential code parameter */
+#define MODULUS_PARAM 13 /* Modulus parameter */
struct VP10Common;
void vp10_default_coef_probs(struct VP10Common *cm);
@@ -156,20 +156,19 @@
#define COEFF_PROB_MODELS 255
-#define UNCONSTRAINED_NODES 3
+#define UNCONSTRAINED_NODES 3
-#define PIVOT_NODE 2 // which node is pivot
+#define PIVOT_NODE 2 // which node is pivot
#define MODEL_NODES (ENTROPY_NODES - UNCONSTRAINED_NODES)
extern const vpx_tree_index vp10_coef_con_tree[TREE_SIZE(ENTROPY_TOKENS)];
extern const vpx_prob vp10_pareto8_full[COEFF_PROB_MODELS][MODEL_NODES];
-typedef vpx_prob vp10_coeff_probs_model[REF_TYPES][COEF_BANDS]
- [COEFF_CONTEXTS][UNCONSTRAINED_NODES];
+typedef vpx_prob vp10_coeff_probs_model[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS]
+ [UNCONSTRAINED_NODES];
-typedef unsigned int vp10_coeff_count_model[REF_TYPES][COEF_BANDS]
- [COEFF_CONTEXTS]
- [UNCONSTRAINED_NODES + 1];
+typedef unsigned int vp10_coeff_count_model
+ [REF_TYPES][COEF_BANDS][COEFF_CONTEXTS][UNCONSTRAINED_NODES + 1];
void vp10_model_to_full_probs(const vpx_prob *model, vpx_prob *full);
@@ -191,19 +190,17 @@
break;
case TX_8X8:
above_ec = !!*(const uint16_t *)a;
- left_ec = !!*(const uint16_t *)l;
+ left_ec = !!*(const uint16_t *)l;
break;
case TX_16X16:
above_ec = !!*(const uint32_t *)a;
- left_ec = !!*(const uint32_t *)l;
+ left_ec = !!*(const uint32_t *)l;
break;
case TX_32X32:
above_ec = !!*(const uint64_t *)a;
- left_ec = !!*(const uint64_t *)l;
+ left_ec = !!*(const uint64_t *)l;
break;
- default:
- assert(0 && "Invalid transform size.");
- break;
+ default: assert(0 && "Invalid transform size."); break;
}
return combine_entropy_contexts(above_ec, left_ec);
diff --git a/vp10/common/entropymode.c b/vp10/common/entropymode.c
index 78f3650..b9ad5c9 100644
--- a/vp10/common/entropymode.c
+++ b/vp10/common/entropymode.c
@@ -13,239 +13,254 @@
#include "vp10/common/onyxc_int.h"
#include "vp10/common/seg_common.h"
-const vpx_prob vp10_kf_y_mode_prob[INTRA_MODES][INTRA_MODES][INTRA_MODES - 1] = {
- { // above = dc
- { 137, 30, 42, 148, 151, 207, 70, 52, 91 }, // left = dc
- { 92, 45, 102, 136, 116, 180, 74, 90, 100 }, // left = v
- { 73, 32, 19, 187, 222, 215, 46, 34, 100 }, // left = h
- { 91, 30, 32, 116, 121, 186, 93, 86, 94 }, // left = d45
- { 72, 35, 36, 149, 68, 206, 68, 63, 105 }, // left = d135
- { 73, 31, 28, 138, 57, 124, 55, 122, 151 }, // left = d117
- { 67, 23, 21, 140, 126, 197, 40, 37, 171 }, // left = d153
- { 86, 27, 28, 128, 154, 212, 45, 43, 53 }, // left = d207
- { 74, 32, 27, 107, 86, 160, 63, 134, 102 }, // left = d63
- { 59, 67, 44, 140, 161, 202, 78, 67, 119 } // left = tm
- }, { // above = v
- { 63, 36, 126, 146, 123, 158, 60, 90, 96 }, // left = dc
- { 43, 46, 168, 134, 107, 128, 69, 142, 92 }, // left = v
- { 44, 29, 68, 159, 201, 177, 50, 57, 77 }, // left = h
- { 58, 38, 76, 114, 97, 172, 78, 133, 92 }, // left = d45
- { 46, 41, 76, 140, 63, 184, 69, 112, 57 }, // left = d135
- { 38, 32, 85, 140, 46, 112, 54, 151, 133 }, // left = d117
- { 39, 27, 61, 131, 110, 175, 44, 75, 136 }, // left = d153
- { 52, 30, 74, 113, 130, 175, 51, 64, 58 }, // left = d207
- { 47, 35, 80, 100, 74, 143, 64, 163, 74 }, // left = d63
- { 36, 61, 116, 114, 128, 162, 80, 125, 82 } // left = tm
- }, { // above = h
- { 82, 26, 26, 171, 208, 204, 44, 32, 105 }, // left = dc
- { 55, 44, 68, 166, 179, 192, 57, 57, 108 }, // left = v
- { 42, 26, 11, 199, 241, 228, 23, 15, 85 }, // left = h
- { 68, 42, 19, 131, 160, 199, 55, 52, 83 }, // left = d45
- { 58, 50, 25, 139, 115, 232, 39, 52, 118 }, // left = d135
- { 50, 35, 33, 153, 104, 162, 64, 59, 131 }, // left = d117
- { 44, 24, 16, 150, 177, 202, 33, 19, 156 }, // left = d153
- { 55, 27, 12, 153, 203, 218, 26, 27, 49 }, // left = d207
- { 53, 49, 21, 110, 116, 168, 59, 80, 76 }, // left = d63
- { 38, 72, 19, 168, 203, 212, 50, 50, 107 } // left = tm
- }, { // above = d45
- { 103, 26, 36, 129, 132, 201, 83, 80, 93 }, // left = dc
- { 59, 38, 83, 112, 103, 162, 98, 136, 90 }, // left = v
- { 62, 30, 23, 158, 200, 207, 59, 57, 50 }, // left = h
- { 67, 30, 29, 84, 86, 191, 102, 91, 59 }, // left = d45
- { 60, 32, 33, 112, 71, 220, 64, 89, 104 }, // left = d135
- { 53, 26, 34, 130, 56, 149, 84, 120, 103 }, // left = d117
- { 53, 21, 23, 133, 109, 210, 56, 77, 172 }, // left = d153
- { 77, 19, 29, 112, 142, 228, 55, 66, 36 }, // left = d207
- { 61, 29, 29, 93, 97, 165, 83, 175, 162 }, // left = d63
- { 47, 47, 43, 114, 137, 181, 100, 99, 95 } // left = tm
- }, { // above = d135
- { 69, 23, 29, 128, 83, 199, 46, 44, 101 }, // left = dc
- { 53, 40, 55, 139, 69, 183, 61, 80, 110 }, // left = v
- { 40, 29, 19, 161, 180, 207, 43, 24, 91 }, // left = h
- { 60, 34, 19, 105, 61, 198, 53, 64, 89 }, // left = d45
- { 52, 31, 22, 158, 40, 209, 58, 62, 89 }, // left = d135
- { 44, 31, 29, 147, 46, 158, 56, 102, 198 }, // left = d117
- { 35, 19, 12, 135, 87, 209, 41, 45, 167 }, // left = d153
- { 55, 25, 21, 118, 95, 215, 38, 39, 66 }, // left = d207
- { 51, 38, 25, 113, 58, 164, 70, 93, 97 }, // left = d63
- { 47, 54, 34, 146, 108, 203, 72, 103, 151 } // left = tm
- }, { // above = d117
- { 64, 19, 37, 156, 66, 138, 49, 95, 133 }, // left = dc
- { 46, 27, 80, 150, 55, 124, 55, 121, 135 }, // left = v
- { 36, 23, 27, 165, 149, 166, 54, 64, 118 }, // left = h
- { 53, 21, 36, 131, 63, 163, 60, 109, 81 }, // left = d45
- { 40, 26, 35, 154, 40, 185, 51, 97, 123 }, // left = d135
- { 35, 19, 34, 179, 19, 97, 48, 129, 124 }, // left = d117
- { 36, 20, 26, 136, 62, 164, 33, 77, 154 }, // left = d153
- { 45, 18, 32, 130, 90, 157, 40, 79, 91 }, // left = d207
- { 45, 26, 28, 129, 45, 129, 49, 147, 123 }, // left = d63
- { 38, 44, 51, 136, 74, 162, 57, 97, 121 } // left = tm
- }, { // above = d153
- { 75, 17, 22, 136, 138, 185, 32, 34, 166 }, // left = dc
- { 56, 39, 58, 133, 117, 173, 48, 53, 187 }, // left = v
- { 35, 21, 12, 161, 212, 207, 20, 23, 145 }, // left = h
- { 56, 29, 19, 117, 109, 181, 55, 68, 112 }, // left = d45
- { 47, 29, 17, 153, 64, 220, 59, 51, 114 }, // left = d135
- { 46, 16, 24, 136, 76, 147, 41, 64, 172 }, // left = d117
- { 34, 17, 11, 108, 152, 187, 13, 15, 209 }, // left = d153
- { 51, 24, 14, 115, 133, 209, 32, 26, 104 }, // left = d207
- { 55, 30, 18, 122, 79, 179, 44, 88, 116 }, // left = d63
- { 37, 49, 25, 129, 168, 164, 41, 54, 148 } // left = tm
- }, { // above = d207
- { 82, 22, 32, 127, 143, 213, 39, 41, 70 }, // left = dc
- { 62, 44, 61, 123, 105, 189, 48, 57, 64 }, // left = v
- { 47, 25, 17, 175, 222, 220, 24, 30, 86 }, // left = h
- { 68, 36, 17, 106, 102, 206, 59, 74, 74 }, // left = d45
- { 57, 39, 23, 151, 68, 216, 55, 63, 58 }, // left = d135
- { 49, 30, 35, 141, 70, 168, 82, 40, 115 }, // left = d117
- { 51, 25, 15, 136, 129, 202, 38, 35, 139 }, // left = d153
- { 68, 26, 16, 111, 141, 215, 29, 28, 28 }, // left = d207
- { 59, 39, 19, 114, 75, 180, 77, 104, 42 }, // left = d63
- { 40, 61, 26, 126, 152, 206, 61, 59, 93 } // left = tm
- }, { // above = d63
- { 78, 23, 39, 111, 117, 170, 74, 124, 94 }, // left = dc
- { 48, 34, 86, 101, 92, 146, 78, 179, 134 }, // left = v
- { 47, 22, 24, 138, 187, 178, 68, 69, 59 }, // left = h
- { 56, 25, 33, 105, 112, 187, 95, 177, 129 }, // left = d45
- { 48, 31, 27, 114, 63, 183, 82, 116, 56 }, // left = d135
- { 43, 28, 37, 121, 63, 123, 61, 192, 169 }, // left = d117
- { 42, 17, 24, 109, 97, 177, 56, 76, 122 }, // left = d153
- { 58, 18, 28, 105, 139, 182, 70, 92, 63 }, // left = d207
- { 46, 23, 32, 74, 86, 150, 67, 183, 88 }, // left = d63
- { 36, 38, 48, 92, 122, 165, 88, 137, 91 } // left = tm
- }, { // above = tm
- { 65, 70, 60, 155, 159, 199, 61, 60, 81 }, // left = dc
- { 44, 78, 115, 132, 119, 173, 71, 112, 93 }, // left = v
- { 39, 38, 21, 184, 227, 206, 42, 32, 64 }, // left = h
- { 58, 47, 36, 124, 137, 193, 80, 82, 78 }, // left = d45
- { 49, 50, 35, 144, 95, 205, 63, 78, 59 }, // left = d135
- { 41, 53, 52, 148, 71, 142, 65, 128, 51 }, // left = d117
- { 40, 36, 28, 143, 143, 202, 40, 55, 137 }, // left = d153
- { 52, 34, 29, 129, 183, 227, 42, 35, 43 }, // left = d207
- { 42, 44, 44, 104, 105, 164, 64, 130, 80 }, // left = d63
- { 43, 81, 53, 140, 169, 204, 68, 84, 72 } // left = tm
- }
-};
+const vpx_prob vp10_kf_y_mode_prob[INTRA_MODES][INTRA_MODES][INTRA_MODES - 1] =
+ { {
+ // above = dc
+ { 137, 30, 42, 148, 151, 207, 70, 52, 91 }, // left = dc
+ { 92, 45, 102, 136, 116, 180, 74, 90, 100 }, // left = v
+ { 73, 32, 19, 187, 222, 215, 46, 34, 100 }, // left = h
+ { 91, 30, 32, 116, 121, 186, 93, 86, 94 }, // left = d45
+ { 72, 35, 36, 149, 68, 206, 68, 63, 105 }, // left = d135
+ { 73, 31, 28, 138, 57, 124, 55, 122, 151 }, // left = d117
+ { 67, 23, 21, 140, 126, 197, 40, 37, 171 }, // left = d153
+ { 86, 27, 28, 128, 154, 212, 45, 43, 53 }, // left = d207
+ { 74, 32, 27, 107, 86, 160, 63, 134, 102 }, // left = d63
+ { 59, 67, 44, 140, 161, 202, 78, 67, 119 } // left = tm
+ },
+ {
+ // above = v
+ { 63, 36, 126, 146, 123, 158, 60, 90, 96 }, // left = dc
+ { 43, 46, 168, 134, 107, 128, 69, 142, 92 }, // left = v
+ { 44, 29, 68, 159, 201, 177, 50, 57, 77 }, // left = h
+ { 58, 38, 76, 114, 97, 172, 78, 133, 92 }, // left = d45
+ { 46, 41, 76, 140, 63, 184, 69, 112, 57 }, // left = d135
+ { 38, 32, 85, 140, 46, 112, 54, 151, 133 }, // left = d117
+ { 39, 27, 61, 131, 110, 175, 44, 75, 136 }, // left = d153
+ { 52, 30, 74, 113, 130, 175, 51, 64, 58 }, // left = d207
+ { 47, 35, 80, 100, 74, 143, 64, 163, 74 }, // left = d63
+ { 36, 61, 116, 114, 128, 162, 80, 125, 82 } // left = tm
+ },
+ {
+ // above = h
+ { 82, 26, 26, 171, 208, 204, 44, 32, 105 }, // left = dc
+ { 55, 44, 68, 166, 179, 192, 57, 57, 108 }, // left = v
+ { 42, 26, 11, 199, 241, 228, 23, 15, 85 }, // left = h
+ { 68, 42, 19, 131, 160, 199, 55, 52, 83 }, // left = d45
+ { 58, 50, 25, 139, 115, 232, 39, 52, 118 }, // left = d135
+ { 50, 35, 33, 153, 104, 162, 64, 59, 131 }, // left = d117
+ { 44, 24, 16, 150, 177, 202, 33, 19, 156 }, // left = d153
+ { 55, 27, 12, 153, 203, 218, 26, 27, 49 }, // left = d207
+ { 53, 49, 21, 110, 116, 168, 59, 80, 76 }, // left = d63
+ { 38, 72, 19, 168, 203, 212, 50, 50, 107 } // left = tm
+ },
+ {
+ // above = d45
+ { 103, 26, 36, 129, 132, 201, 83, 80, 93 }, // left = dc
+ { 59, 38, 83, 112, 103, 162, 98, 136, 90 }, // left = v
+ { 62, 30, 23, 158, 200, 207, 59, 57, 50 }, // left = h
+ { 67, 30, 29, 84, 86, 191, 102, 91, 59 }, // left = d45
+ { 60, 32, 33, 112, 71, 220, 64, 89, 104 }, // left = d135
+ { 53, 26, 34, 130, 56, 149, 84, 120, 103 }, // left = d117
+ { 53, 21, 23, 133, 109, 210, 56, 77, 172 }, // left = d153
+ { 77, 19, 29, 112, 142, 228, 55, 66, 36 }, // left = d207
+ { 61, 29, 29, 93, 97, 165, 83, 175, 162 }, // left = d63
+ { 47, 47, 43, 114, 137, 181, 100, 99, 95 } // left = tm
+ },
+ {
+ // above = d135
+ { 69, 23, 29, 128, 83, 199, 46, 44, 101 }, // left = dc
+ { 53, 40, 55, 139, 69, 183, 61, 80, 110 }, // left = v
+ { 40, 29, 19, 161, 180, 207, 43, 24, 91 }, // left = h
+ { 60, 34, 19, 105, 61, 198, 53, 64, 89 }, // left = d45
+ { 52, 31, 22, 158, 40, 209, 58, 62, 89 }, // left = d135
+ { 44, 31, 29, 147, 46, 158, 56, 102, 198 }, // left = d117
+ { 35, 19, 12, 135, 87, 209, 41, 45, 167 }, // left = d153
+ { 55, 25, 21, 118, 95, 215, 38, 39, 66 }, // left = d207
+ { 51, 38, 25, 113, 58, 164, 70, 93, 97 }, // left = d63
+ { 47, 54, 34, 146, 108, 203, 72, 103, 151 } // left = tm
+ },
+ {
+ // above = d117
+ { 64, 19, 37, 156, 66, 138, 49, 95, 133 }, // left = dc
+ { 46, 27, 80, 150, 55, 124, 55, 121, 135 }, // left = v
+ { 36, 23, 27, 165, 149, 166, 54, 64, 118 }, // left = h
+ { 53, 21, 36, 131, 63, 163, 60, 109, 81 }, // left = d45
+ { 40, 26, 35, 154, 40, 185, 51, 97, 123 }, // left = d135
+ { 35, 19, 34, 179, 19, 97, 48, 129, 124 }, // left = d117
+ { 36, 20, 26, 136, 62, 164, 33, 77, 154 }, // left = d153
+ { 45, 18, 32, 130, 90, 157, 40, 79, 91 }, // left = d207
+ { 45, 26, 28, 129, 45, 129, 49, 147, 123 }, // left = d63
+ { 38, 44, 51, 136, 74, 162, 57, 97, 121 } // left = tm
+ },
+ {
+ // above = d153
+ { 75, 17, 22, 136, 138, 185, 32, 34, 166 }, // left = dc
+ { 56, 39, 58, 133, 117, 173, 48, 53, 187 }, // left = v
+ { 35, 21, 12, 161, 212, 207, 20, 23, 145 }, // left = h
+ { 56, 29, 19, 117, 109, 181, 55, 68, 112 }, // left = d45
+ { 47, 29, 17, 153, 64, 220, 59, 51, 114 }, // left = d135
+ { 46, 16, 24, 136, 76, 147, 41, 64, 172 }, // left = d117
+ { 34, 17, 11, 108, 152, 187, 13, 15, 209 }, // left = d153
+ { 51, 24, 14, 115, 133, 209, 32, 26, 104 }, // left = d207
+ { 55, 30, 18, 122, 79, 179, 44, 88, 116 }, // left = d63
+ { 37, 49, 25, 129, 168, 164, 41, 54, 148 } // left = tm
+ },
+ {
+ // above = d207
+ { 82, 22, 32, 127, 143, 213, 39, 41, 70 }, // left = dc
+ { 62, 44, 61, 123, 105, 189, 48, 57, 64 }, // left = v
+ { 47, 25, 17, 175, 222, 220, 24, 30, 86 }, // left = h
+ { 68, 36, 17, 106, 102, 206, 59, 74, 74 }, // left = d45
+ { 57, 39, 23, 151, 68, 216, 55, 63, 58 }, // left = d135
+ { 49, 30, 35, 141, 70, 168, 82, 40, 115 }, // left = d117
+ { 51, 25, 15, 136, 129, 202, 38, 35, 139 }, // left = d153
+ { 68, 26, 16, 111, 141, 215, 29, 28, 28 }, // left = d207
+ { 59, 39, 19, 114, 75, 180, 77, 104, 42 }, // left = d63
+ { 40, 61, 26, 126, 152, 206, 61, 59, 93 } // left = tm
+ },
+ {
+ // above = d63
+ { 78, 23, 39, 111, 117, 170, 74, 124, 94 }, // left = dc
+ { 48, 34, 86, 101, 92, 146, 78, 179, 134 }, // left = v
+ { 47, 22, 24, 138, 187, 178, 68, 69, 59 }, // left = h
+ { 56, 25, 33, 105, 112, 187, 95, 177, 129 }, // left = d45
+ { 48, 31, 27, 114, 63, 183, 82, 116, 56 }, // left = d135
+ { 43, 28, 37, 121, 63, 123, 61, 192, 169 }, // left = d117
+ { 42, 17, 24, 109, 97, 177, 56, 76, 122 }, // left = d153
+ { 58, 18, 28, 105, 139, 182, 70, 92, 63 }, // left = d207
+ { 46, 23, 32, 74, 86, 150, 67, 183, 88 }, // left = d63
+ { 36, 38, 48, 92, 122, 165, 88, 137, 91 } // left = tm
+ },
+ {
+ // above = tm
+ { 65, 70, 60, 155, 159, 199, 61, 60, 81 }, // left = dc
+ { 44, 78, 115, 132, 119, 173, 71, 112, 93 }, // left = v
+ { 39, 38, 21, 184, 227, 206, 42, 32, 64 }, // left = h
+ { 58, 47, 36, 124, 137, 193, 80, 82, 78 }, // left = d45
+ { 49, 50, 35, 144, 95, 205, 63, 78, 59 }, // left = d135
+ { 41, 53, 52, 148, 71, 142, 65, 128, 51 }, // left = d117
+ { 40, 36, 28, 143, 143, 202, 40, 55, 137 }, // left = d153
+ { 52, 34, 29, 129, 183, 227, 42, 35, 43 }, // left = d207
+ { 42, 44, 44, 104, 105, 164, 64, 130, 80 }, // left = d63
+ { 43, 81, 53, 140, 169, 204, 68, 84, 72 } // left = tm
+ } };
#if !CONFIG_MISC_FIXES
const vpx_prob vp10_kf_uv_mode_prob[INTRA_MODES][INTRA_MODES - 1] = {
- { 144, 11, 54, 157, 195, 130, 46, 58, 108 }, // y = dc
- { 118, 15, 123, 148, 131, 101, 44, 93, 131 }, // y = v
- { 113, 12, 23, 188, 226, 142, 26, 32, 125 }, // y = h
- { 120, 11, 50, 123, 163, 135, 64, 77, 103 }, // y = d45
- { 113, 9, 36, 155, 111, 157, 32, 44, 161 }, // y = d135
- { 116, 9, 55, 176, 76, 96, 37, 61, 149 }, // y = d117
- { 115, 9, 28, 141, 161, 167, 21, 25, 193 }, // y = d153
- { 120, 12, 32, 145, 195, 142, 32, 38, 86 }, // y = d207
- { 116, 12, 64, 120, 140, 125, 49, 115, 121 }, // y = d63
- { 102, 19, 66, 162, 182, 122, 35, 59, 128 } // y = tm
+ { 144, 11, 54, 157, 195, 130, 46, 58, 108 }, // y = dc
+ { 118, 15, 123, 148, 131, 101, 44, 93, 131 }, // y = v
+ { 113, 12, 23, 188, 226, 142, 26, 32, 125 }, // y = h
+ { 120, 11, 50, 123, 163, 135, 64, 77, 103 }, // y = d45
+ { 113, 9, 36, 155, 111, 157, 32, 44, 161 }, // y = d135
+ { 116, 9, 55, 176, 76, 96, 37, 61, 149 }, // y = d117
+ { 115, 9, 28, 141, 161, 167, 21, 25, 193 }, // y = d153
+ { 120, 12, 32, 145, 195, 142, 32, 38, 86 }, // y = d207
+ { 116, 12, 64, 120, 140, 125, 49, 115, 121 }, // y = d63
+ { 102, 19, 66, 162, 182, 122, 35, 59, 128 } // y = tm
};
#endif
static const vpx_prob default_if_y_probs[BLOCK_SIZE_GROUPS][INTRA_MODES - 1] = {
- { 65, 32, 18, 144, 162, 194, 41, 51, 98 }, // block_size < 8x8
- { 132, 68, 18, 165, 217, 196, 45, 40, 78 }, // block_size < 16x16
- { 173, 80, 19, 176, 240, 193, 64, 35, 46 }, // block_size < 32x32
- { 221, 135, 38, 194, 248, 121, 96, 85, 29 } // block_size >= 32x32
+ { 65, 32, 18, 144, 162, 194, 41, 51, 98 }, // block_size < 8x8
+ { 132, 68, 18, 165, 217, 196, 45, 40, 78 }, // block_size < 16x16
+ { 173, 80, 19, 176, 240, 193, 64, 35, 46 }, // block_size < 32x32
+ { 221, 135, 38, 194, 248, 121, 96, 85, 29 } // block_size >= 32x32
};
static const vpx_prob default_uv_probs[INTRA_MODES][INTRA_MODES - 1] = {
- { 120, 7, 76, 176, 208, 126, 28, 54, 103 }, // y = dc
- { 48, 12, 154, 155, 139, 90, 34, 117, 119 }, // y = v
- { 67, 6, 25, 204, 243, 158, 13, 21, 96 }, // y = h
- { 97, 5, 44, 131, 176, 139, 48, 68, 97 }, // y = d45
- { 83, 5, 42, 156, 111, 152, 26, 49, 152 }, // y = d135
- { 80, 5, 58, 178, 74, 83, 33, 62, 145 }, // y = d117
- { 86, 5, 32, 154, 192, 168, 14, 22, 163 }, // y = d153
- { 85, 5, 32, 156, 216, 148, 19, 29, 73 }, // y = d207
- { 77, 7, 64, 116, 132, 122, 37, 126, 120 }, // y = d63
- { 101, 21, 107, 181, 192, 103, 19, 67, 125 } // y = tm
+ { 120, 7, 76, 176, 208, 126, 28, 54, 103 }, // y = dc
+ { 48, 12, 154, 155, 139, 90, 34, 117, 119 }, // y = v
+ { 67, 6, 25, 204, 243, 158, 13, 21, 96 }, // y = h
+ { 97, 5, 44, 131, 176, 139, 48, 68, 97 }, // y = d45
+ { 83, 5, 42, 156, 111, 152, 26, 49, 152 }, // y = d135
+ { 80, 5, 58, 178, 74, 83, 33, 62, 145 }, // y = d117
+ { 86, 5, 32, 154, 192, 168, 14, 22, 163 }, // y = d153
+ { 85, 5, 32, 156, 216, 148, 19, 29, 73 }, // y = d207
+ { 77, 7, 64, 116, 132, 122, 37, 126, 120 }, // y = d63
+ { 101, 21, 107, 181, 192, 103, 19, 67, 125 } // y = tm
};
#if !CONFIG_MISC_FIXES
-const vpx_prob vp10_kf_partition_probs[PARTITION_CONTEXTS]
- [PARTITION_TYPES - 1] = {
- // 8x8 -> 4x4
- { 158, 97, 94 }, // a/l both not split
- { 93, 24, 99 }, // a split, l not split
- { 85, 119, 44 }, // l split, a not split
- { 62, 59, 67 }, // a/l both split
- // 16x16 -> 8x8
- { 149, 53, 53 }, // a/l both not split
- { 94, 20, 48 }, // a split, l not split
- { 83, 53, 24 }, // l split, a not split
- { 52, 18, 18 }, // a/l both split
- // 32x32 -> 16x16
- { 150, 40, 39 }, // a/l both not split
- { 78, 12, 26 }, // a split, l not split
- { 67, 33, 11 }, // l split, a not split
- { 24, 7, 5 }, // a/l both split
- // 64x64 -> 32x32
- { 174, 35, 49 }, // a/l both not split
- { 68, 11, 27 }, // a split, l not split
- { 57, 15, 9 }, // l split, a not split
- { 12, 3, 3 }, // a/l both split
-};
+const vpx_prob
+ vp10_kf_partition_probs[PARTITION_CONTEXTS][PARTITION_TYPES - 1] = {
+ // 8x8 -> 4x4
+ { 158, 97, 94 }, // a/l both not split
+ { 93, 24, 99 }, // a split, l not split
+ { 85, 119, 44 }, // l split, a not split
+ { 62, 59, 67 }, // a/l both split
+ // 16x16 -> 8x8
+ { 149, 53, 53 }, // a/l both not split
+ { 94, 20, 48 }, // a split, l not split
+ { 83, 53, 24 }, // l split, a not split
+ { 52, 18, 18 }, // a/l both split
+ // 32x32 -> 16x16
+ { 150, 40, 39 }, // a/l both not split
+ { 78, 12, 26 }, // a split, l not split
+ { 67, 33, 11 }, // l split, a not split
+ { 24, 7, 5 }, // a/l both split
+ // 64x64 -> 32x32
+ { 174, 35, 49 }, // a/l both not split
+ { 68, 11, 27 }, // a split, l not split
+ { 57, 15, 9 }, // l split, a not split
+ { 12, 3, 3 }, // a/l both split
+ };
#endif
-static const vpx_prob default_partition_probs[PARTITION_CONTEXTS]
- [PARTITION_TYPES - 1] = {
- // 8x8 -> 4x4
- { 199, 122, 141 }, // a/l both not split
- { 147, 63, 159 }, // a split, l not split
- { 148, 133, 118 }, // l split, a not split
- { 121, 104, 114 }, // a/l both split
- // 16x16 -> 8x8
- { 174, 73, 87 }, // a/l both not split
- { 92, 41, 83 }, // a split, l not split
- { 82, 99, 50 }, // l split, a not split
- { 53, 39, 39 }, // a/l both split
- // 32x32 -> 16x16
- { 177, 58, 59 }, // a/l both not split
- { 68, 26, 63 }, // a split, l not split
- { 52, 79, 25 }, // l split, a not split
- { 17, 14, 12 }, // a/l both split
- // 64x64 -> 32x32
- { 222, 34, 30 }, // a/l both not split
- { 72, 16, 44 }, // a split, l not split
- { 58, 32, 12 }, // l split, a not split
- { 10, 7, 6 }, // a/l both split
-};
+static const vpx_prob
+ default_partition_probs[PARTITION_CONTEXTS][PARTITION_TYPES - 1] = {
+ // 8x8 -> 4x4
+ { 199, 122, 141 }, // a/l both not split
+ { 147, 63, 159 }, // a split, l not split
+ { 148, 133, 118 }, // l split, a not split
+ { 121, 104, 114 }, // a/l both split
+ // 16x16 -> 8x8
+ { 174, 73, 87 }, // a/l both not split
+ { 92, 41, 83 }, // a split, l not split
+ { 82, 99, 50 }, // l split, a not split
+ { 53, 39, 39 }, // a/l both split
+ // 32x32 -> 16x16
+ { 177, 58, 59 }, // a/l both not split
+ { 68, 26, 63 }, // a split, l not split
+ { 52, 79, 25 }, // l split, a not split
+ { 17, 14, 12 }, // a/l both split
+ // 64x64 -> 32x32
+ { 222, 34, 30 }, // a/l both not split
+ { 72, 16, 44 }, // a split, l not split
+ { 58, 32, 12 }, // l split, a not split
+ { 10, 7, 6 }, // a/l both split
+ };
-static const vpx_prob default_inter_mode_probs[INTER_MODE_CONTEXTS]
- [INTER_MODES - 1] = {
- {2, 173, 34}, // 0 = both zero mv
- {7, 145, 85}, // 1 = one zero mv + one a predicted mv
- {7, 166, 63}, // 2 = two predicted mvs
- {7, 94, 66}, // 3 = one predicted/zero and one new mv
- {8, 64, 46}, // 4 = two new mvs
- {17, 81, 31}, // 5 = one intra neighbour + x
- {25, 29, 30}, // 6 = two intra neighbours
-};
+static const vpx_prob
+ default_inter_mode_probs[INTER_MODE_CONTEXTS][INTER_MODES - 1] = {
+ { 2, 173, 34 }, // 0 = both zero mv
+ { 7, 145, 85 }, // 1 = one zero mv + one a predicted mv
+ { 7, 166, 63 }, // 2 = two predicted mvs
+ { 7, 94, 66 }, // 3 = one predicted/zero and one new mv
+ { 8, 64, 46 }, // 4 = two new mvs
+ { 17, 81, 31 }, // 5 = one intra neighbour + x
+ { 25, 29, 30 }, // 6 = two intra neighbours
+ };
/* Array indices are identical to previously-existing INTRAMODECONTEXTNODES. */
const vpx_tree_index vp10_intra_mode_tree[TREE_SIZE(INTRA_MODES)] = {
- -DC_PRED, 2, /* 0 = DC_NODE */
- -TM_PRED, 4, /* 1 = TM_NODE */
- -V_PRED, 6, /* 2 = V_NODE */
- 8, 12, /* 3 = COM_NODE */
- -H_PRED, 10, /* 4 = H_NODE */
- -D135_PRED, -D117_PRED, /* 5 = D135_NODE */
- -D45_PRED, 14, /* 6 = D45_NODE */
- -D63_PRED, 16, /* 7 = D63_NODE */
- -D153_PRED, -D207_PRED /* 8 = D153_NODE */
+ -DC_PRED, 2, /* 0 = DC_NODE */
+ -TM_PRED, 4, /* 1 = TM_NODE */
+ -V_PRED, 6, /* 2 = V_NODE */
+ 8, 12, /* 3 = COM_NODE */
+ -H_PRED, 10, /* 4 = H_NODE */
+ -D135_PRED, -D117_PRED, /* 5 = D135_NODE */
+ -D45_PRED, 14, /* 6 = D45_NODE */
+ -D63_PRED, 16, /* 7 = D63_NODE */
+ -D153_PRED, -D207_PRED /* 8 = D153_NODE */
};
const vpx_tree_index vp10_inter_mode_tree[TREE_SIZE(INTER_MODES)] = {
- -INTER_OFFSET(ZEROMV), 2,
- -INTER_OFFSET(NEARESTMV), 4,
- -INTER_OFFSET(NEARMV), -INTER_OFFSET(NEWMV)
+ -INTER_OFFSET(ZEROMV), 2, -INTER_OFFSET(NEARESTMV), 4, -INTER_OFFSET(NEARMV),
+ -INTER_OFFSET(NEWMV)
};
const vpx_tree_index vp10_partition_tree[TREE_SIZE(PARTITION_TYPES)] = {
- -PARTITION_NONE, 2,
- -PARTITION_HORZ, 4,
- -PARTITION_VERT, -PARTITION_SPLIT
+ -PARTITION_NONE, 2, -PARTITION_HORZ, 4, -PARTITION_VERT, -PARTITION_SPLIT
};
static const vpx_prob default_intra_inter_p[INTRA_INTER_CONTEXTS] = {
@@ -253,47 +268,36 @@
};
static const vpx_prob default_comp_inter_p[COMP_INTER_CONTEXTS] = {
- 239, 183, 119, 96, 41
+ 239, 183, 119, 96, 41
};
-static const vpx_prob default_comp_ref_p[REF_CONTEXTS] = {
- 50, 126, 123, 221, 226
-};
+static const vpx_prob default_comp_ref_p[REF_CONTEXTS] = { 50, 126, 123, 221,
+ 226 };
static const vpx_prob default_single_ref_p[REF_CONTEXTS][2] = {
- { 33, 16 },
- { 77, 74 },
- { 142, 142 },
- { 172, 170 },
- { 238, 247 }
+ { 33, 16 }, { 77, 74 }, { 142, 142 }, { 172, 170 }, { 238, 247 }
};
-static const struct tx_probs default_tx_probs = {
- { { 3, 136, 37 },
- { 5, 52, 13 } },
+static const struct tx_probs default_tx_probs = { { { 3, 136, 37 },
+ { 5, 52, 13 } },
- { { 20, 152 },
- { 15, 101 } },
+ { { 20, 152 }, { 15, 101 } },
- { { 100 },
- { 66 } }
-};
+ { { 100 }, { 66 } } };
void vp10_tx_counts_to_branch_counts_32x32(const unsigned int *tx_count_32x32p,
- unsigned int (*ct_32x32p)[2]) {
+ unsigned int (*ct_32x32p)[2]) {
ct_32x32p[0][0] = tx_count_32x32p[TX_4X4];
- ct_32x32p[0][1] = tx_count_32x32p[TX_8X8] +
- tx_count_32x32p[TX_16X16] +
+ ct_32x32p[0][1] = tx_count_32x32p[TX_8X8] + tx_count_32x32p[TX_16X16] +
tx_count_32x32p[TX_32X32];
ct_32x32p[1][0] = tx_count_32x32p[TX_8X8];
- ct_32x32p[1][1] = tx_count_32x32p[TX_16X16] +
- tx_count_32x32p[TX_32X32];
+ ct_32x32p[1][1] = tx_count_32x32p[TX_16X16] + tx_count_32x32p[TX_32X32];
ct_32x32p[2][0] = tx_count_32x32p[TX_16X16];
ct_32x32p[2][1] = tx_count_32x32p[TX_32X32];
}
void vp10_tx_counts_to_branch_counts_16x16(const unsigned int *tx_count_16x16p,
- unsigned int (*ct_16x16p)[2]) {
+ unsigned int (*ct_16x16p)[2]) {
ct_16x16p[0][0] = tx_count_16x16p[TX_4X4];
ct_16x16p[0][1] = tx_count_16x16p[TX_8X8] + tx_count_16x16p[TX_16X16];
ct_16x16p[1][0] = tx_count_16x16p[TX_8X8];
@@ -301,49 +305,38 @@
}
void vp10_tx_counts_to_branch_counts_8x8(const unsigned int *tx_count_8x8p,
- unsigned int (*ct_8x8p)[2]) {
+ unsigned int (*ct_8x8p)[2]) {
ct_8x8p[0][0] = tx_count_8x8p[TX_4X4];
ct_8x8p[0][1] = tx_count_8x8p[TX_8X8];
}
-static const vpx_prob default_skip_probs[SKIP_CONTEXTS] = {
- 192, 128, 64
-};
+static const vpx_prob default_skip_probs[SKIP_CONTEXTS] = { 192, 128, 64 };
-static const vpx_prob default_switchable_interp_prob[SWITCHABLE_FILTER_CONTEXTS]
- [SWITCHABLE_FILTERS - 1] = {
- { 235, 162, },
- { 36, 255, },
- { 34, 3, },
- { 149, 144, },
-};
+static const vpx_prob default_switchable_interp_prob
+ [SWITCHABLE_FILTER_CONTEXTS][SWITCHABLE_FILTERS - 1] = {
+ { 235, 162 }, { 36, 255 }, { 34, 3 }, { 149, 144 },
+ };
#if CONFIG_MISC_FIXES
// FIXME(someone) need real defaults here
static const struct segmentation_probs default_seg_probs = {
- { 128, 128, 128, 128, 128, 128, 128 },
- { 128, 128, 128 },
+ { 128, 128, 128, 128, 128, 128, 128 }, { 128, 128, 128 },
};
#endif
const vpx_tree_index vp10_ext_tx_tree[TREE_SIZE(TX_TYPES)] = {
- -DCT_DCT, 2,
- -ADST_ADST, 4,
- -ADST_DCT, -DCT_ADST
+ -DCT_DCT, 2, -ADST_ADST, 4, -ADST_DCT, -DCT_ADST
};
-static const vpx_prob default_intra_ext_tx_prob[EXT_TX_SIZES]
- [TX_TYPES][TX_TYPES - 1] = {
- {{240, 85, 128}, {4, 1, 248}, {4, 1, 8}, {4, 248, 128}},
- {{244, 85, 128}, {8, 2, 248}, {8, 2, 8}, {8, 248, 128}},
- {{248, 85, 128}, {16, 4, 248}, {16, 4, 8}, {16, 248, 128}},
-};
+static const vpx_prob
+ default_intra_ext_tx_prob[EXT_TX_SIZES][TX_TYPES][TX_TYPES - 1] = {
+ { { 240, 85, 128 }, { 4, 1, 248 }, { 4, 1, 8 }, { 4, 248, 128 } },
+ { { 244, 85, 128 }, { 8, 2, 248 }, { 8, 2, 8 }, { 8, 248, 128 } },
+ { { 248, 85, 128 }, { 16, 4, 248 }, { 16, 4, 8 }, { 16, 248, 128 } },
+ };
-static const vpx_prob default_inter_ext_tx_prob[EXT_TX_SIZES]
- [TX_TYPES - 1] = {
- {160, 85, 128},
- {176, 85, 128},
- {192, 85, 128},
+static const vpx_prob default_inter_ext_tx_prob[EXT_TX_SIZES][TX_TYPES - 1] = {
+ { 160, 85, 128 }, { 176, 85, 128 }, { 192, 85, 128 },
};
static void init_mode_probs(FRAME_CONTEXT *fc) {
@@ -366,11 +359,8 @@
vp10_copy(fc->inter_ext_tx_prob, default_inter_ext_tx_prob);
}
-const vpx_tree_index vp10_switchable_interp_tree
- [TREE_SIZE(SWITCHABLE_FILTERS)] = {
- -EIGHTTAP, 2,
- -EIGHTTAP_SMOOTH, -EIGHTTAP_SHARP
-};
+const vpx_tree_index vp10_switchable_interp_tree[TREE_SIZE(
+ SWITCHABLE_FILTERS)] = { -EIGHTTAP, 2, -EIGHTTAP_SMOOTH, -EIGHTTAP_SHARP };
void vp10_adapt_inter_frame_probs(VP10_COMMON *cm) {
int i, j;
@@ -382,11 +372,11 @@
fc->intra_inter_prob[i] = mode_mv_merge_probs(pre_fc->intra_inter_prob[i],
counts->intra_inter[i]);
for (i = 0; i < COMP_INTER_CONTEXTS; i++)
- fc->comp_inter_prob[i] = mode_mv_merge_probs(pre_fc->comp_inter_prob[i],
- counts->comp_inter[i]);
+ fc->comp_inter_prob[i] =
+ mode_mv_merge_probs(pre_fc->comp_inter_prob[i], counts->comp_inter[i]);
for (i = 0; i < REF_CONTEXTS; i++)
- fc->comp_ref_prob[i] = mode_mv_merge_probs(pre_fc->comp_ref_prob[i],
- counts->comp_ref[i]);
+ fc->comp_ref_prob[i] =
+ mode_mv_merge_probs(pre_fc->comp_ref_prob[i], counts->comp_ref[i]);
for (i = 0; i < REF_CONTEXTS; i++)
for (j = 0; j < 2; j++)
fc->single_ref_prob[i][j] = mode_mv_merge_probs(
@@ -394,11 +384,11 @@
for (i = 0; i < INTER_MODE_CONTEXTS; i++)
vpx_tree_merge_probs(vp10_inter_mode_tree, pre_fc->inter_mode_probs[i],
- counts->inter_mode[i], fc->inter_mode_probs[i]);
+ counts->inter_mode[i], fc->inter_mode_probs[i]);
for (i = 0; i < BLOCK_SIZE_GROUPS; i++)
vpx_tree_merge_probs(vp10_intra_mode_tree, pre_fc->y_mode_prob[i],
- counts->y_mode[i], fc->y_mode_prob[i]);
+ counts->y_mode[i], fc->y_mode_prob[i]);
#if !CONFIG_MISC_FIXES
for (i = 0; i < INTRA_MODES; ++i)
@@ -412,10 +402,9 @@
if (cm->interp_filter == SWITCHABLE) {
for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++)
- vpx_tree_merge_probs(vp10_switchable_interp_tree,
- pre_fc->switchable_interp_prob[i],
- counts->switchable_interp[i],
- fc->switchable_interp_prob[i]);
+ vpx_tree_merge_probs(
+ vp10_switchable_interp_tree, pre_fc->switchable_interp_prob[i],
+ counts->switchable_interp[i], fc->switchable_interp_prob[i]);
}
}
@@ -434,15 +423,17 @@
for (i = 0; i < TX_SIZE_CONTEXTS; ++i) {
vp10_tx_counts_to_branch_counts_8x8(counts->tx.p8x8[i], branch_ct_8x8p);
for (j = 0; j < TX_SIZES - 3; ++j)
- fc->tx_probs.p8x8[i][j] = mode_mv_merge_probs(
- pre_fc->tx_probs.p8x8[i][j], branch_ct_8x8p[j]);
+ fc->tx_probs.p8x8[i][j] =
+ mode_mv_merge_probs(pre_fc->tx_probs.p8x8[i][j], branch_ct_8x8p[j]);
- vp10_tx_counts_to_branch_counts_16x16(counts->tx.p16x16[i], branch_ct_16x16p);
+ vp10_tx_counts_to_branch_counts_16x16(counts->tx.p16x16[i],
+ branch_ct_16x16p);
for (j = 0; j < TX_SIZES - 2; ++j)
fc->tx_probs.p16x16[i][j] = mode_mv_merge_probs(
pre_fc->tx_probs.p16x16[i][j], branch_ct_16x16p[j]);
- vp10_tx_counts_to_branch_counts_32x32(counts->tx.p32x32[i], branch_ct_32x32p);
+ vp10_tx_counts_to_branch_counts_32x32(counts->tx.p32x32[i],
+ branch_ct_32x32p);
for (j = 0; j < TX_SIZES - 1; ++j)
fc->tx_probs.p32x32[i][j] = mode_mv_merge_probs(
pre_fc->tx_probs.p32x32[i][j], branch_ct_32x32p[j]);
@@ -450,29 +441,26 @@
}
for (i = 0; i < SKIP_CONTEXTS; ++i)
- fc->skip_probs[i] = mode_mv_merge_probs(
- pre_fc->skip_probs[i], counts->skip[i]);
+ fc->skip_probs[i] =
+ mode_mv_merge_probs(pre_fc->skip_probs[i], counts->skip[i]);
for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
int j;
for (j = 0; j < TX_TYPES; ++j)
- vpx_tree_merge_probs(vp10_ext_tx_tree,
- pre_fc->intra_ext_tx_prob[i][j],
+ vpx_tree_merge_probs(vp10_ext_tx_tree, pre_fc->intra_ext_tx_prob[i][j],
counts->intra_ext_tx[i][j],
fc->intra_ext_tx_prob[i][j]);
}
for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
- vpx_tree_merge_probs(vp10_ext_tx_tree,
- pre_fc->inter_ext_tx_prob[i],
- counts->inter_ext_tx[i],
- fc->inter_ext_tx_prob[i]);
+ vpx_tree_merge_probs(vp10_ext_tx_tree, pre_fc->inter_ext_tx_prob[i],
+ counts->inter_ext_tx[i], fc->inter_ext_tx_prob[i]);
}
#if CONFIG_MISC_FIXES
if (cm->seg.temporal_update) {
for (i = 0; i < PREDICTION_PROBS; i++)
- fc->seg.pred_probs[i] = mode_mv_merge_probs(pre_fc->seg.pred_probs[i],
- counts->seg.pred[i]);
+ fc->seg.pred_probs[i] =
+ mode_mv_merge_probs(pre_fc->seg.pred_probs[i], counts->seg.pred[i]);
vpx_tree_merge_probs(vp10_segment_tree, pre_fc->seg.tree_probs,
counts->seg.tree_mispred, fc->seg.tree_probs);
@@ -535,8 +523,7 @@
if (cm->frame_type == KEY_FRAME || cm->error_resilient_mode ||
cm->reset_frame_context == RESET_FRAME_CONTEXT_ALL) {
// Reset all frame contexts.
- for (i = 0; i < FRAME_CONTEXTS; ++i)
- cm->frame_contexts[i] = *cm->fc;
+ for (i = 0; i < FRAME_CONTEXTS; ++i) cm->frame_contexts[i] = *cm->fc;
} else if (cm->reset_frame_context == RESET_FRAME_CONTEXT_CURRENT) {
// Reset only the frame context specified in the frame header.
cm->frame_contexts[cm->frame_context_idx] = *cm->fc;
diff --git a/vp10/common/entropymode.h b/vp10/common/entropymode.h
index 611d3ad..377762a 100644
--- a/vp10/common/entropymode.h
+++ b/vp10/common/entropymode.h
@@ -25,7 +25,7 @@
#define TX_SIZE_CONTEXTS 2
-#define INTER_OFFSET(mode) ((mode) - NEARESTMV)
+#define INTER_OFFSET(mode) ((mode)-NEARESTMV)
struct VP10Common;
@@ -53,8 +53,9 @@
vpx_prob uv_mode_prob[INTRA_MODES][INTRA_MODES - 1];
vpx_prob partition_prob[PARTITION_CONTEXTS][PARTITION_TYPES - 1];
vp10_coeff_probs_model coef_probs[TX_SIZES][PLANE_TYPES];
- vpx_prob switchable_interp_prob[SWITCHABLE_FILTER_CONTEXTS]
- [SWITCHABLE_FILTERS - 1];
+ vpx_prob
+ switchable_interp_prob[SWITCHABLE_FILTER_CONTEXTS][SWITCHABLE_FILTERS -
+ 1];
vpx_prob inter_mode_probs[INTER_MODE_CONTEXTS][INTER_MODES - 1];
vpx_prob intra_inter_prob[INTRA_INTER_CONTEXTS];
vpx_prob comp_inter_prob[COMP_INTER_CONTEXTS];
@@ -77,10 +78,10 @@
unsigned int uv_mode[INTRA_MODES][INTRA_MODES];
unsigned int partition[PARTITION_CONTEXTS][PARTITION_TYPES];
vp10_coeff_count_model coef[TX_SIZES][PLANE_TYPES];
- unsigned int eob_branch[TX_SIZES][PLANE_TYPES][REF_TYPES]
- [COEF_BANDS][COEFF_CONTEXTS];
- unsigned int switchable_interp[SWITCHABLE_FILTER_CONTEXTS]
- [SWITCHABLE_FILTERS];
+ unsigned int
+ eob_branch[TX_SIZES][PLANE_TYPES][REF_TYPES][COEF_BANDS][COEFF_CONTEXTS];
+ unsigned int
+ switchable_interp[SWITCHABLE_FILTER_CONTEXTS][SWITCHABLE_FILTERS];
unsigned int inter_mode[INTER_MODE_CONTEXTS][INTER_MODES];
unsigned int intra_inter[INTRA_INTER_CONTEXTS][2];
unsigned int comp_inter[COMP_INTER_CONTEXTS][2];
@@ -96,20 +97,19 @@
unsigned int inter_ext_tx[EXT_TX_SIZES][TX_TYPES];
} FRAME_COUNTS;
-extern const vpx_prob vp10_kf_y_mode_prob[INTRA_MODES][INTRA_MODES]
- [INTRA_MODES - 1];
+extern const vpx_prob
+ vp10_kf_y_mode_prob[INTRA_MODES][INTRA_MODES][INTRA_MODES - 1];
#if !CONFIG_MISC_FIXES
extern const vpx_prob vp10_kf_uv_mode_prob[INTRA_MODES][INTRA_MODES - 1];
-extern const vpx_prob vp10_kf_partition_probs[PARTITION_CONTEXTS]
- [PARTITION_TYPES - 1];
+extern const vpx_prob
+ vp10_kf_partition_probs[PARTITION_CONTEXTS][PARTITION_TYPES - 1];
#endif
extern const vpx_tree_index vp10_intra_mode_tree[TREE_SIZE(INTRA_MODES)];
extern const vpx_tree_index vp10_inter_mode_tree[TREE_SIZE(INTER_MODES)];
extern const vpx_tree_index vp10_partition_tree[TREE_SIZE(PARTITION_TYPES)];
-extern const vpx_tree_index vp10_switchable_interp_tree
- [TREE_SIZE(SWITCHABLE_FILTERS)];
-
+extern const vpx_tree_index
+ vp10_switchable_interp_tree[TREE_SIZE(SWITCHABLE_FILTERS)];
void vp10_setup_past_independence(struct VP10Common *cm);
@@ -117,14 +117,13 @@
void vp10_adapt_inter_frame_probs(struct VP10Common *cm);
void vp10_tx_counts_to_branch_counts_32x32(const unsigned int *tx_count_32x32p,
- unsigned int (*ct_32x32p)[2]);
+ unsigned int (*ct_32x32p)[2]);
void vp10_tx_counts_to_branch_counts_16x16(const unsigned int *tx_count_16x16p,
- unsigned int (*ct_16x16p)[2]);
+ unsigned int (*ct_16x16p)[2]);
void vp10_tx_counts_to_branch_counts_8x8(const unsigned int *tx_count_8x8p,
- unsigned int (*ct_8x8p)[2]);
+ unsigned int (*ct_8x8p)[2]);
-extern const vpx_tree_index
- vp10_ext_tx_tree[TREE_SIZE(TX_TYPES)];
+extern const vpx_tree_index vp10_ext_tx_tree[TREE_SIZE(TX_TYPES)];
static INLINE int vp10_ceil_log2(int n) {
int i = 1, p = 2;
diff --git a/vp10/common/entropymv.c b/vp10/common/entropymv.c
index a9946ee..a984241 100644
--- a/vp10/common/entropymv.c
+++ b/vp10/common/entropymv.c
@@ -15,11 +15,10 @@
#define COMPANDED_MVREF_THRESH 8
const vpx_tree_index vp10_mv_joint_tree[TREE_SIZE(MV_JOINTS)] = {
- -MV_JOINT_ZERO, 2,
- -MV_JOINT_HNZVZ, 4,
- -MV_JOINT_HZVNZ, -MV_JOINT_HNZVNZ
+ -MV_JOINT_ZERO, 2, -MV_JOINT_HNZVZ, 4, -MV_JOINT_HZVNZ, -MV_JOINT_HNZVNZ
};
+/* clang-format off */
const vpx_tree_index vp10_mv_class_tree[TREE_SIZE(MV_CLASSES)] = {
-MV_CLASS_0, 2,
-MV_CLASS_1, 4,
@@ -32,87 +31,82 @@
-MV_CLASS_7, -MV_CLASS_8,
-MV_CLASS_9, -MV_CLASS_10,
};
+/* clang-format on */
const vpx_tree_index vp10_mv_class0_tree[TREE_SIZE(CLASS0_SIZE)] = {
-0, -1,
};
-const vpx_tree_index vp10_mv_fp_tree[TREE_SIZE(MV_FP_SIZE)] = {
- -0, 2,
- -1, 4,
- -2, -3
-};
+const vpx_tree_index vp10_mv_fp_tree[TREE_SIZE(MV_FP_SIZE)] = { -0, 2, -1,
+ 4, -2, -3 };
static const nmv_context default_nmv_context = {
- {32, 64, 96},
- {
- { // Vertical component
- 128, // sign
- {224, 144, 192, 168, 192, 176, 192, 198, 198, 245}, // class
- {216}, // class0
- {136, 140, 148, 160, 176, 192, 224, 234, 234, 240}, // bits
- {{128, 128, 64}, {96, 112, 64}}, // class0_fp
- {64, 96, 64}, // fp
- 160, // class0_hp bit
- 128, // hp
+ { 32, 64, 96 },
+ { {
+ // Vertical component
+ 128, // sign
+ { 224, 144, 192, 168, 192, 176, 192, 198, 198, 245 }, // class
+ { 216 }, // class0
+ { 136, 140, 148, 160, 176, 192, 224, 234, 234, 240 }, // bits
+ { { 128, 128, 64 }, { 96, 112, 64 } }, // class0_fp
+ { 64, 96, 64 }, // fp
+ 160, // class0_hp bit
+ 128, // hp
},
- { // Horizontal component
- 128, // sign
- {216, 128, 176, 160, 176, 176, 192, 198, 198, 208}, // class
- {208}, // class0
- {136, 140, 148, 160, 176, 192, 224, 234, 234, 240}, // bits
- {{128, 128, 64}, {96, 112, 64}}, // class0_fp
- {64, 96, 64}, // fp
- 160, // class0_hp bit
- 128, // hp
- }
- },
+ {
+ // Horizontal component
+ 128, // sign
+ { 216, 128, 176, 160, 176, 176, 192, 198, 198, 208 }, // class
+ { 208 }, // class0
+ { 136, 140, 148, 160, 176, 192, 224, 234, 234, 240 }, // bits
+ { { 128, 128, 64 }, { 96, 112, 64 } }, // class0_fp
+ { 64, 96, 64 }, // fp
+ 160, // class0_hp bit
+ 128, // hp
+ } },
};
static const uint8_t log_in_base_2[] = {
- 0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
- 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6,
- 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
- 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
- 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
- 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
- 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
- 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
- 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
- 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
- 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 10
+ 0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+ 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 10
};
static INLINE int mv_class_base(MV_CLASS_TYPE c) {
@@ -120,16 +114,16 @@
}
MV_CLASS_TYPE vp10_get_mv_class(int z, int *offset) {
- const MV_CLASS_TYPE c = (z >= CLASS0_SIZE * 4096) ?
- MV_CLASS_10 : (MV_CLASS_TYPE)log_in_base_2[z >> 3];
- if (offset)
- *offset = z - mv_class_base(c);
+ const MV_CLASS_TYPE c = (z >= CLASS0_SIZE * 4096)
+ ? MV_CLASS_10
+ : (MV_CLASS_TYPE)log_in_base_2[z >> 3];
+ if (offset) *offset = z - mv_class_base(c);
return c;
}
int vp10_use_mv_hp(const MV *ref) {
#if CONFIG_MISC_FIXES
- (void) ref;
+ (void)ref;
return 1;
#else
return (abs(ref->row) >> 3) < COMPANDED_MVREF_THRESH &&
@@ -137,20 +131,20 @@
#endif
}
-static void inc_mv_component(int v, nmv_component_counts *comp_counts,
- int incr, int usehp) {
+static void inc_mv_component(int v, nmv_component_counts *comp_counts, int incr,
+ int usehp) {
int s, z, c, o, d, e, f;
- assert(v != 0); /* should not be zero */
+ assert(v != 0); /* should not be zero */
s = v < 0;
comp_counts->sign[s] += incr;
- z = (s ? -v : v) - 1; /* magnitude - 1 */
+ z = (s ? -v : v) - 1; /* magnitude - 1 */
c = vp10_get_mv_class(z, &o);
comp_counts->classes[c] += incr;
- d = (o >> 3); /* int mv data */
- f = (o >> 1) & 3; /* fractional pel mv data */
- e = (o & 1); /* high precision mv data */
+ d = (o >> 3); /* int mv data */
+ f = (o >> 1) & 3; /* fractional pel mv data */
+ e = (o & 1); /* high precision mv data */
if (c == MV_CLASS_0) {
comp_counts->class0[d] += incr;
@@ -159,8 +153,7 @@
} else {
int i;
int b = c + CLASS0_BITS - 1; // number of bits
- for (i = 0; i < b; ++i)
- comp_counts->bits[i][((d >> i) & 1)] += incr;
+ for (i = 0; i < b; ++i) comp_counts->bits[i][((d >> i) & 1)] += incr;
comp_counts->fp[f] += incr;
comp_counts->hp[e] += usehp * incr;
}
@@ -220,6 +213,4 @@
}
}
-void vp10_init_mv_probs(VP10_COMMON *cm) {
- cm->fc->nmvc = default_nmv_context;
-}
+void vp10_init_mv_probs(VP10_COMMON *cm) { cm->fc->nmvc = default_nmv_context; }
diff --git a/vp10/common/entropymv.h b/vp10/common/entropymv.h
index d1eb95c..737a88c 100644
--- a/vp10/common/entropymv.h
+++ b/vp10/common/entropymv.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_COMMON_ENTROPYMV_H_
#define VP10_COMMON_ENTROPYMV_H_
@@ -32,12 +31,12 @@
#define MV_UPDATE_PROB 252
/* Symbols for coding which components are zero jointly */
-#define MV_JOINTS 4
+#define MV_JOINTS 4
typedef enum {
- MV_JOINT_ZERO = 0, /* Zero vector */
- MV_JOINT_HNZVZ = 1, /* Vert zero, hor nonzero */
- MV_JOINT_HZVNZ = 2, /* Hor zero, vert nonzero */
- MV_JOINT_HNZVNZ = 3, /* Both components nonzero */
+ MV_JOINT_ZERO = 0, /* Zero vector */
+ MV_JOINT_HNZVZ = 1, /* Vert zero, hor nonzero */
+ MV_JOINT_HZVNZ = 2, /* Hor zero, vert nonzero */
+ MV_JOINT_HNZVNZ = 3, /* Both components nonzero */
} MV_JOINT_TYPE;
static INLINE int mv_joint_vertical(MV_JOINT_TYPE type) {
@@ -49,33 +48,33 @@
}
/* Symbols for coding magnitude class of nonzero components */
-#define MV_CLASSES 11
+#define MV_CLASSES 11
typedef enum {
- MV_CLASS_0 = 0, /* (0, 2] integer pel */
- MV_CLASS_1 = 1, /* (2, 4] integer pel */
- MV_CLASS_2 = 2, /* (4, 8] integer pel */
- MV_CLASS_3 = 3, /* (8, 16] integer pel */
- MV_CLASS_4 = 4, /* (16, 32] integer pel */
- MV_CLASS_5 = 5, /* (32, 64] integer pel */
- MV_CLASS_6 = 6, /* (64, 128] integer pel */
- MV_CLASS_7 = 7, /* (128, 256] integer pel */
- MV_CLASS_8 = 8, /* (256, 512] integer pel */
- MV_CLASS_9 = 9, /* (512, 1024] integer pel */
- MV_CLASS_10 = 10, /* (1024,2048] integer pel */
+ MV_CLASS_0 = 0, /* (0, 2] integer pel */
+ MV_CLASS_1 = 1, /* (2, 4] integer pel */
+ MV_CLASS_2 = 2, /* (4, 8] integer pel */
+ MV_CLASS_3 = 3, /* (8, 16] integer pel */
+ MV_CLASS_4 = 4, /* (16, 32] integer pel */
+ MV_CLASS_5 = 5, /* (32, 64] integer pel */
+ MV_CLASS_6 = 6, /* (64, 128] integer pel */
+ MV_CLASS_7 = 7, /* (128, 256] integer pel */
+ MV_CLASS_8 = 8, /* (256, 512] integer pel */
+ MV_CLASS_9 = 9, /* (512, 1024] integer pel */
+ MV_CLASS_10 = 10, /* (1024,2048] integer pel */
} MV_CLASS_TYPE;
-#define CLASS0_BITS 1 /* bits at integer precision for class 0 */
-#define CLASS0_SIZE (1 << CLASS0_BITS)
+#define CLASS0_BITS 1 /* bits at integer precision for class 0 */
+#define CLASS0_SIZE (1 << CLASS0_BITS)
#define MV_OFFSET_BITS (MV_CLASSES + CLASS0_BITS - 2)
#define MV_FP_SIZE 4
-#define MV_MAX_BITS (MV_CLASSES + CLASS0_BITS + 2)
-#define MV_MAX ((1 << MV_MAX_BITS) - 1)
-#define MV_VALS ((MV_MAX << 1) + 1)
+#define MV_MAX_BITS (MV_CLASSES + CLASS0_BITS + 2)
+#define MV_MAX ((1 << MV_MAX_BITS) - 1)
+#define MV_VALS ((MV_MAX << 1) + 1)
#define MV_IN_USE_BITS 14
-#define MV_UPP ((1 << MV_IN_USE_BITS) - 1)
-#define MV_LOW (-(1 << MV_IN_USE_BITS))
+#define MV_UPP ((1 << MV_IN_USE_BITS) - 1)
+#define MV_LOW (-(1 << MV_IN_USE_BITS))
extern const vpx_tree_index vp10_mv_joint_tree[];
extern const vpx_tree_index vp10_mv_class_tree[];
diff --git a/vp10/common/enums.h b/vp10/common/enums.h
index 9b10379..c08b6b9 100644
--- a/vp10/common/enums.h
+++ b/vp10/common/enums.h
@@ -21,7 +21,7 @@
#define MI_SIZE_LOG2 3
#define MI_BLOCK_SIZE_LOG2 (6 - MI_SIZE_LOG2) // 64 = 2^6
-#define MI_SIZE (1 << MI_SIZE_LOG2) // pixels per mi-unit
+#define MI_SIZE (1 << MI_SIZE_LOG2) // pixels per mi-unit
#define MI_BLOCK_SIZE (1 << MI_BLOCK_SIZE_LOG2) // mi-units per max block
#define MI_MASK (MI_BLOCK_SIZE - 1)
@@ -41,20 +41,20 @@
MAX_PROFILES
} BITSTREAM_PROFILE;
-#define BLOCK_4X4 0
-#define BLOCK_4X8 1
-#define BLOCK_8X4 2
-#define BLOCK_8X8 3
-#define BLOCK_8X16 4
-#define BLOCK_16X8 5
-#define BLOCK_16X16 6
-#define BLOCK_16X32 7
-#define BLOCK_32X16 8
-#define BLOCK_32X32 9
-#define BLOCK_32X64 10
-#define BLOCK_64X32 11
-#define BLOCK_64X64 12
-#define BLOCK_SIZES 13
+#define BLOCK_4X4 0
+#define BLOCK_4X8 1
+#define BLOCK_8X4 2
+#define BLOCK_8X8 3
+#define BLOCK_8X16 4
+#define BLOCK_16X8 5
+#define BLOCK_16X16 6
+#define BLOCK_16X32 7
+#define BLOCK_32X16 8
+#define BLOCK_32X32 9
+#define BLOCK_32X64 10
+#define BLOCK_64X32 11
+#define BLOCK_64X64 12
+#define BLOCK_SIZES 13
#define BLOCK_INVALID BLOCK_SIZES
typedef uint8_t BLOCK_SIZE;
@@ -68,36 +68,36 @@
} PARTITION_TYPE;
typedef char PARTITION_CONTEXT;
-#define PARTITION_PLOFFSET 4 // number of probability models per block size
+#define PARTITION_PLOFFSET 4 // number of probability models per block size
#define PARTITION_CONTEXTS (4 * PARTITION_PLOFFSET)
// block transform size
typedef uint8_t TX_SIZE;
-#define TX_4X4 ((TX_SIZE)0) // 4x4 transform
-#define TX_8X8 ((TX_SIZE)1) // 8x8 transform
-#define TX_16X16 ((TX_SIZE)2) // 16x16 transform
-#define TX_32X32 ((TX_SIZE)3) // 32x32 transform
+#define TX_4X4 ((TX_SIZE)0) // 4x4 transform
+#define TX_8X8 ((TX_SIZE)1) // 8x8 transform
+#define TX_16X16 ((TX_SIZE)2) // 16x16 transform
+#define TX_32X32 ((TX_SIZE)3) // 32x32 transform
#define TX_SIZES ((TX_SIZE)4)
// frame transform mode
typedef enum {
- ONLY_4X4 = 0, // only 4x4 transform used
- ALLOW_8X8 = 1, // allow block transform size up to 8x8
- ALLOW_16X16 = 2, // allow block transform size up to 16x16
- ALLOW_32X32 = 3, // allow block transform size up to 32x32
- TX_MODE_SELECT = 4, // transform specified for each block
- TX_MODES = 5,
+ ONLY_4X4 = 0, // only 4x4 transform used
+ ALLOW_8X8 = 1, // allow block transform size up to 8x8
+ ALLOW_16X16 = 2, // allow block transform size up to 16x16
+ ALLOW_32X32 = 3, // allow block transform size up to 32x32
+ TX_MODE_SELECT = 4, // transform specified for each block
+ TX_MODES = 5,
} TX_MODE;
typedef enum {
- DCT_DCT = 0, // DCT in both horizontal and vertical
- ADST_DCT = 1, // ADST in vertical, DCT in horizontal
- DCT_ADST = 2, // DCT in vertical, ADST in horizontal
- ADST_ADST = 3, // ADST in both directions
+ DCT_DCT = 0, // DCT in both horizontal and vertical
+ ADST_DCT = 1, // ADST in vertical, DCT in horizontal
+ DCT_ADST = 2, // DCT in vertical, ADST in horizontal
+ ADST_ADST = 3, // ADST in both directions
TX_TYPES = 4
} TX_TYPE;
-#define EXT_TX_SIZES 3 // number of sizes that use extended transforms
+#define EXT_TX_SIZES 3 // number of sizes that use extended transforms
typedef enum {
VPX_LAST_FLAG = 1 << 0,
@@ -105,26 +105,22 @@
VPX_ALT_FLAG = 1 << 2,
} VPX_REFFRAME;
-typedef enum {
- PLANE_TYPE_Y = 0,
- PLANE_TYPE_UV = 1,
- PLANE_TYPES
-} PLANE_TYPE;
+typedef enum { PLANE_TYPE_Y = 0, PLANE_TYPE_UV = 1, PLANE_TYPES } PLANE_TYPE;
-#define DC_PRED 0 // Average of above and left pixels
-#define V_PRED 1 // Vertical
-#define H_PRED 2 // Horizontal
-#define D45_PRED 3 // Directional 45 deg = round(arctan(1/1) * 180/pi)
-#define D135_PRED 4 // Directional 135 deg = 180 - 45
-#define D117_PRED 5 // Directional 117 deg = 180 - 63
-#define D153_PRED 6 // Directional 153 deg = 180 - 27
-#define D207_PRED 7 // Directional 207 deg = 180 + 27
-#define D63_PRED 8 // Directional 63 deg = round(arctan(2/1) * 180/pi)
-#define TM_PRED 9 // True-motion
+#define DC_PRED 0 // Average of above and left pixels
+#define V_PRED 1 // Vertical
+#define H_PRED 2 // Horizontal
+#define D45_PRED 3 // Directional 45 deg = round(arctan(1/1) * 180/pi)
+#define D135_PRED 4 // Directional 135 deg = 180 - 45
+#define D117_PRED 5 // Directional 117 deg = 180 - 63
+#define D153_PRED 6 // Directional 153 deg = 180 - 27
+#define D207_PRED 7 // Directional 207 deg = 180 + 27
+#define D63_PRED 8 // Directional 63 deg = round(arctan(2/1) * 180/pi)
+#define TM_PRED 9 // True-motion
#define NEARESTMV 10
-#define NEARMV 11
-#define ZEROMV 12
-#define NEWMV 13
+#define NEARMV 11
+#define ZEROMV 12
+#define NEWMV 13
#define MB_MODE_COUNT 14
typedef uint8_t PREDICTION_MODE;
diff --git a/vp10/common/filter.c b/vp10/common/filter.c
index dda279f..8597efc 100644
--- a/vp10/common/filter.c
+++ b/vp10/common/filter.c
@@ -14,91 +14,55 @@
DECLARE_ALIGNED(256, static const InterpKernel,
bilinear_filters[SUBPEL_SHIFTS]) = {
- { 0, 0, 0, 128, 0, 0, 0, 0 },
- { 0, 0, 0, 120, 8, 0, 0, 0 },
- { 0, 0, 0, 112, 16, 0, 0, 0 },
- { 0, 0, 0, 104, 24, 0, 0, 0 },
- { 0, 0, 0, 96, 32, 0, 0, 0 },
- { 0, 0, 0, 88, 40, 0, 0, 0 },
- { 0, 0, 0, 80, 48, 0, 0, 0 },
- { 0, 0, 0, 72, 56, 0, 0, 0 },
- { 0, 0, 0, 64, 64, 0, 0, 0 },
- { 0, 0, 0, 56, 72, 0, 0, 0 },
- { 0, 0, 0, 48, 80, 0, 0, 0 },
- { 0, 0, 0, 40, 88, 0, 0, 0 },
- { 0, 0, 0, 32, 96, 0, 0, 0 },
- { 0, 0, 0, 24, 104, 0, 0, 0 },
- { 0, 0, 0, 16, 112, 0, 0, 0 },
- { 0, 0, 0, 8, 120, 0, 0, 0 }
+ { 0, 0, 0, 128, 0, 0, 0, 0 }, { 0, 0, 0, 120, 8, 0, 0, 0 },
+ { 0, 0, 0, 112, 16, 0, 0, 0 }, { 0, 0, 0, 104, 24, 0, 0, 0 },
+ { 0, 0, 0, 96, 32, 0, 0, 0 }, { 0, 0, 0, 88, 40, 0, 0, 0 },
+ { 0, 0, 0, 80, 48, 0, 0, 0 }, { 0, 0, 0, 72, 56, 0, 0, 0 },
+ { 0, 0, 0, 64, 64, 0, 0, 0 }, { 0, 0, 0, 56, 72, 0, 0, 0 },
+ { 0, 0, 0, 48, 80, 0, 0, 0 }, { 0, 0, 0, 40, 88, 0, 0, 0 },
+ { 0, 0, 0, 32, 96, 0, 0, 0 }, { 0, 0, 0, 24, 104, 0, 0, 0 },
+ { 0, 0, 0, 16, 112, 0, 0, 0 }, { 0, 0, 0, 8, 120, 0, 0, 0 }
};
// Lagrangian interpolation filter
DECLARE_ALIGNED(256, static const InterpKernel,
sub_pel_filters_8[SUBPEL_SHIFTS]) = {
- { 0, 0, 0, 128, 0, 0, 0, 0},
- { 0, 1, -5, 126, 8, -3, 1, 0},
- { -1, 3, -10, 122, 18, -6, 2, 0},
- { -1, 4, -13, 118, 27, -9, 3, -1},
- { -1, 4, -16, 112, 37, -11, 4, -1},
- { -1, 5, -18, 105, 48, -14, 4, -1},
- { -1, 5, -19, 97, 58, -16, 5, -1},
- { -1, 6, -19, 88, 68, -18, 5, -1},
- { -1, 6, -19, 78, 78, -19, 6, -1},
- { -1, 5, -18, 68, 88, -19, 6, -1},
- { -1, 5, -16, 58, 97, -19, 5, -1},
- { -1, 4, -14, 48, 105, -18, 5, -1},
- { -1, 4, -11, 37, 112, -16, 4, -1},
- { -1, 3, -9, 27, 118, -13, 4, -1},
- { 0, 2, -6, 18, 122, -10, 3, -1},
- { 0, 1, -3, 8, 126, -5, 1, 0}
+ { 0, 0, 0, 128, 0, 0, 0, 0 }, { 0, 1, -5, 126, 8, -3, 1, 0 },
+ { -1, 3, -10, 122, 18, -6, 2, 0 }, { -1, 4, -13, 118, 27, -9, 3, -1 },
+ { -1, 4, -16, 112, 37, -11, 4, -1 }, { -1, 5, -18, 105, 48, -14, 4, -1 },
+ { -1, 5, -19, 97, 58, -16, 5, -1 }, { -1, 6, -19, 88, 68, -18, 5, -1 },
+ { -1, 6, -19, 78, 78, -19, 6, -1 }, { -1, 5, -18, 68, 88, -19, 6, -1 },
+ { -1, 5, -16, 58, 97, -19, 5, -1 }, { -1, 4, -14, 48, 105, -18, 5, -1 },
+ { -1, 4, -11, 37, 112, -16, 4, -1 }, { -1, 3, -9, 27, 118, -13, 4, -1 },
+ { 0, 2, -6, 18, 122, -10, 3, -1 }, { 0, 1, -3, 8, 126, -5, 1, 0 }
};
// DCT based filter
DECLARE_ALIGNED(256, static const InterpKernel,
sub_pel_filters_8s[SUBPEL_SHIFTS]) = {
- {0, 0, 0, 128, 0, 0, 0, 0},
- {-1, 3, -7, 127, 8, -3, 1, 0},
- {-2, 5, -13, 125, 17, -6, 3, -1},
- {-3, 7, -17, 121, 27, -10, 5, -2},
- {-4, 9, -20, 115, 37, -13, 6, -2},
- {-4, 10, -23, 108, 48, -16, 8, -3},
- {-4, 10, -24, 100, 59, -19, 9, -3},
- {-4, 11, -24, 90, 70, -21, 10, -4},
- {-4, 11, -23, 80, 80, -23, 11, -4},
- {-4, 10, -21, 70, 90, -24, 11, -4},
- {-3, 9, -19, 59, 100, -24, 10, -4},
- {-3, 8, -16, 48, 108, -23, 10, -4},
- {-2, 6, -13, 37, 115, -20, 9, -4},
- {-2, 5, -10, 27, 121, -17, 7, -3},
- {-1, 3, -6, 17, 125, -13, 5, -2},
- {0, 1, -3, 8, 127, -7, 3, -1}
+ { 0, 0, 0, 128, 0, 0, 0, 0 }, { -1, 3, -7, 127, 8, -3, 1, 0 },
+ { -2, 5, -13, 125, 17, -6, 3, -1 }, { -3, 7, -17, 121, 27, -10, 5, -2 },
+ { -4, 9, -20, 115, 37, -13, 6, -2 }, { -4, 10, -23, 108, 48, -16, 8, -3 },
+ { -4, 10, -24, 100, 59, -19, 9, -3 }, { -4, 11, -24, 90, 70, -21, 10, -4 },
+ { -4, 11, -23, 80, 80, -23, 11, -4 }, { -4, 10, -21, 70, 90, -24, 11, -4 },
+ { -3, 9, -19, 59, 100, -24, 10, -4 }, { -3, 8, -16, 48, 108, -23, 10, -4 },
+ { -2, 6, -13, 37, 115, -20, 9, -4 }, { -2, 5, -10, 27, 121, -17, 7, -3 },
+ { -1, 3, -6, 17, 125, -13, 5, -2 }, { 0, 1, -3, 8, 127, -7, 3, -1 }
};
// freqmultiplier = 0.5
DECLARE_ALIGNED(256, static const InterpKernel,
sub_pel_filters_8lp[SUBPEL_SHIFTS]) = {
- { 0, 0, 0, 128, 0, 0, 0, 0},
- {-3, -1, 32, 64, 38, 1, -3, 0},
- {-2, -2, 29, 63, 41, 2, -3, 0},
- {-2, -2, 26, 63, 43, 4, -4, 0},
- {-2, -3, 24, 62, 46, 5, -4, 0},
- {-2, -3, 21, 60, 49, 7, -4, 0},
- {-1, -4, 18, 59, 51, 9, -4, 0},
- {-1, -4, 16, 57, 53, 12, -4, -1},
- {-1, -4, 14, 55, 55, 14, -4, -1},
- {-1, -4, 12, 53, 57, 16, -4, -1},
- { 0, -4, 9, 51, 59, 18, -4, -1},
- { 0, -4, 7, 49, 60, 21, -3, -2},
- { 0, -4, 5, 46, 62, 24, -3, -2},
- { 0, -4, 4, 43, 63, 26, -2, -2},
- { 0, -3, 2, 41, 63, 29, -2, -2},
- { 0, -3, 1, 38, 64, 32, -1, -3}
+ { 0, 0, 0, 128, 0, 0, 0, 0 }, { -3, -1, 32, 64, 38, 1, -3, 0 },
+ { -2, -2, 29, 63, 41, 2, -3, 0 }, { -2, -2, 26, 63, 43, 4, -4, 0 },
+ { -2, -3, 24, 62, 46, 5, -4, 0 }, { -2, -3, 21, 60, 49, 7, -4, 0 },
+ { -1, -4, 18, 59, 51, 9, -4, 0 }, { -1, -4, 16, 57, 53, 12, -4, -1 },
+ { -1, -4, 14, 55, 55, 14, -4, -1 }, { -1, -4, 12, 53, 57, 16, -4, -1 },
+ { 0, -4, 9, 51, 59, 18, -4, -1 }, { 0, -4, 7, 49, 60, 21, -3, -2 },
+ { 0, -4, 5, 46, 62, 24, -3, -2 }, { 0, -4, 4, 43, 63, 26, -2, -2 },
+ { 0, -3, 2, 41, 63, 29, -2, -2 }, { 0, -3, 1, 38, 64, 32, -1, -3 }
};
-
const InterpKernel *vp10_filter_kernels[4] = {
- sub_pel_filters_8,
- sub_pel_filters_8lp,
- sub_pel_filters_8s,
- bilinear_filters
+ sub_pel_filters_8, sub_pel_filters_8lp, sub_pel_filters_8s, bilinear_filters
};
diff --git a/vp10/common/filter.h b/vp10/common/filter.h
index 826cd03..c0a25d3 100644
--- a/vp10/common/filter.h
+++ b/vp10/common/filter.h
@@ -16,16 +16,15 @@
#include "vpx_dsp/vpx_filter.h"
#include "vpx_ports/mem.h"
-
#ifdef __cplusplus
extern "C" {
#endif
-#define EIGHTTAP 0
-#define EIGHTTAP_SMOOTH 1
-#define EIGHTTAP_SHARP 2
-#define SWITCHABLE_FILTERS 3 /* Number of switchable filters */
-#define BILINEAR 3
+#define EIGHTTAP 0
+#define EIGHTTAP_SMOOTH 1
+#define EIGHTTAP_SHARP 2
+#define SWITCHABLE_FILTERS 3 /* Number of switchable filters */
+#define BILINEAR 3
// The codec can operate in four possible inter prediction filter mode:
// 8-tap, 8-tap-smooth, 8-tap-sharp, and switching between the three.
#define SWITCHABLE_FILTER_CONTEXTS (SWITCHABLE_FILTERS + 1)
diff --git a/vp10/common/frame_buffers.c b/vp10/common/frame_buffers.c
index 31ff960..564fa9c 100644
--- a/vp10/common/frame_buffers.c
+++ b/vp10/common/frame_buffers.c
@@ -19,9 +19,8 @@
list->num_internal_frame_buffers =
VPX_MAXIMUM_REF_BUFFERS + VPX_MAXIMUM_WORK_BUFFERS;
- list->int_fb =
- (InternalFrameBuffer *)vpx_calloc(list->num_internal_frame_buffers,
- sizeof(*list->int_fb));
+ list->int_fb = (InternalFrameBuffer *)vpx_calloc(
+ list->num_internal_frame_buffers, sizeof(*list->int_fb));
return (list->int_fb == NULL);
}
@@ -39,27 +38,23 @@
}
int vp10_get_frame_buffer(void *cb_priv, size_t min_size,
- vpx_codec_frame_buffer_t *fb) {
+ vpx_codec_frame_buffer_t *fb) {
int i;
InternalFrameBufferList *const int_fb_list =
(InternalFrameBufferList *)cb_priv;
- if (int_fb_list == NULL)
- return -1;
+ if (int_fb_list == NULL) return -1;
// Find a free frame buffer.
for (i = 0; i < int_fb_list->num_internal_frame_buffers; ++i) {
- if (!int_fb_list->int_fb[i].in_use)
- break;
+ if (!int_fb_list->int_fb[i].in_use) break;
}
- if (i == int_fb_list->num_internal_frame_buffers)
- return -1;
+ if (i == int_fb_list->num_internal_frame_buffers) return -1;
if (int_fb_list->int_fb[i].size < min_size) {
int_fb_list->int_fb[i].data =
(uint8_t *)vpx_realloc(int_fb_list->int_fb[i].data, min_size);
- if (!int_fb_list->int_fb[i].data)
- return -1;
+ if (!int_fb_list->int_fb[i].data) return -1;
// This memset is needed for fixing valgrind error from C loop filter
// due to access uninitialized memory in frame border. It could be
@@ -80,7 +75,6 @@
int vp10_release_frame_buffer(void *cb_priv, vpx_codec_frame_buffer_t *fb) {
InternalFrameBuffer *const int_fb = (InternalFrameBuffer *)fb->priv;
(void)cb_priv;
- if (int_fb)
- int_fb->in_use = 0;
+ if (int_fb) int_fb->in_use = 0;
return 0;
}
diff --git a/vp10/common/frame_buffers.h b/vp10/common/frame_buffers.h
index 729ebaf..fa6a2e2 100644
--- a/vp10/common/frame_buffers.h
+++ b/vp10/common/frame_buffers.h
@@ -40,7 +40,7 @@
// |min_size| is the minimum size in bytes needed to decode the next frame.
// |fb| pointer to the frame buffer.
int vp10_get_frame_buffer(void *cb_priv, size_t min_size,
- vpx_codec_frame_buffer_t *fb);
+ vpx_codec_frame_buffer_t *fb);
// Callback used by libvpx when there are no references to the frame buffer.
// |cb_priv| is not used. |fb| pointer to the frame buffer.
diff --git a/vp10/common/idct.c b/vp10/common/idct.c
index 20a120e..9870925 100644
--- a/vp10/common/idct.c
+++ b/vp10/common/idct.c
@@ -18,12 +18,12 @@
#include "vpx_ports/mem.h"
void vp10_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride,
- int tx_type) {
+ int tx_type) {
const transform_2d IHT_4[] = {
- { idct4_c, idct4_c }, // DCT_DCT = 0
- { iadst4_c, idct4_c }, // ADST_DCT = 1
- { idct4_c, iadst4_c }, // DCT_ADST = 2
- { iadst4_c, iadst4_c } // ADST_ADST = 3
+ { idct4_c, idct4_c }, // DCT_DCT = 0
+ { iadst4_c, idct4_c }, // ADST_DCT = 1
+ { idct4_c, iadst4_c }, // DCT_ADST = 2
+ { iadst4_c, iadst4_c } // ADST_ADST = 3
};
int i, j;
@@ -34,14 +34,13 @@
// inverse transform row vectors
for (i = 0; i < 4; ++i) {
IHT_4[tx_type].rows(input, outptr);
- input += 4;
+ input += 4;
outptr += 4;
}
// inverse transform column vectors
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j * 4 + i];
+ for (j = 0; j < 4; ++j) temp_in[j] = out[j * 4 + i];
IHT_4[tx_type].cols(temp_in, temp_out);
for (j = 0; j < 4; ++j) {
dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
@@ -51,14 +50,14 @@
}
static const transform_2d IHT_8[] = {
- { idct8_c, idct8_c }, // DCT_DCT = 0
- { iadst8_c, idct8_c }, // ADST_DCT = 1
- { idct8_c, iadst8_c }, // DCT_ADST = 2
- { iadst8_c, iadst8_c } // ADST_ADST = 3
+ { idct8_c, idct8_c }, // DCT_DCT = 0
+ { iadst8_c, idct8_c }, // ADST_DCT = 1
+ { idct8_c, iadst8_c }, // DCT_ADST = 2
+ { iadst8_c, iadst8_c } // ADST_ADST = 3
};
void vp10_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride,
- int tx_type) {
+ int tx_type) {
int i, j;
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
@@ -74,8 +73,7 @@
// inverse transform column vectors
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
ht.cols(temp_in, temp_out);
for (j = 0; j < 8; ++j) {
dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
@@ -85,14 +83,14 @@
}
static const transform_2d IHT_16[] = {
- { idct16_c, idct16_c }, // DCT_DCT = 0
- { iadst16_c, idct16_c }, // ADST_DCT = 1
- { idct16_c, iadst16_c }, // DCT_ADST = 2
- { iadst16_c, iadst16_c } // ADST_ADST = 3
+ { idct16_c, idct16_c }, // DCT_DCT = 0
+ { iadst16_c, idct16_c }, // ADST_DCT = 1
+ { idct16_c, iadst16_c }, // DCT_ADST = 2
+ { iadst16_c, iadst16_c } // ADST_ADST = 3
};
void vp10_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest, int stride,
- int tx_type) {
+ int tx_type) {
int i, j;
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
@@ -108,8 +106,7 @@
// Columns
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
ht.cols(temp_in, temp_out);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
@@ -120,16 +117,15 @@
// idct
void vp10_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob) {
+ int eob) {
if (eob > 1)
vpx_idct4x4_16_add(input, dest, stride);
else
vpx_idct4x4_1_add(input, dest, stride);
}
-
void vp10_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob) {
+ int eob) {
if (eob > 1)
vpx_iwht4x4_16_add(input, dest, stride);
else
@@ -137,7 +133,7 @@
}
void vp10_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob) {
+ int eob) {
// If dc is 1, then input[0] is the reconstructed value, do not need
// dequantization. Also, when dc is 1, dc is counted in eobs, namely eobs >=1.
@@ -155,11 +151,10 @@
}
void vp10_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob) {
+ int eob) {
/* The calculation can be simplified if there are not many non-zero dct
* coefficients. Use eobs to separate different cases. */
- if (eob == 1)
- /* DC only DCT coefficient. */
+ if (eob == 1) /* DC only DCT coefficient. */
vpx_idct16x16_1_add(input, dest, stride);
else if (eob <= 10)
vpx_idct16x16_10_add(input, dest, stride);
@@ -168,7 +163,7 @@
}
void vp10_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob) {
+ int eob) {
if (eob == 1)
vpx_idct32x32_1_add(input, dest, stride);
else if (eob <= 34)
@@ -178,87 +173,63 @@
vpx_idct32x32_1024_add(input, dest, stride);
}
-void vp10_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, TX_TYPE tx_type, int lossless) {
+void vp10_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, TX_TYPE tx_type, int lossless) {
if (lossless) {
assert(tx_type == DCT_DCT);
vp10_iwht4x4_add(input, dest, stride, eob);
} else {
switch (tx_type) {
- case DCT_DCT:
- vp10_idct4x4_add(input, dest, stride, eob);
- break;
+ case DCT_DCT: vp10_idct4x4_add(input, dest, stride, eob); break;
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- vp10_iht4x4_16_add(input, dest, stride, tx_type);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: vp10_iht4x4_16_add(input, dest, stride, tx_type); break;
+ default: assert(0); break;
}
}
}
-void vp10_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, TX_TYPE tx_type) {
+void vp10_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, TX_TYPE tx_type) {
switch (tx_type) {
- case DCT_DCT:
- vp10_idct8x8_add(input, dest, stride, eob);
- break;
+ case DCT_DCT: vp10_idct8x8_add(input, dest, stride, eob); break;
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- vp10_iht8x8_64_add(input, dest, stride, tx_type);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: vp10_iht8x8_64_add(input, dest, stride, tx_type); break;
+ default: assert(0); break;
}
}
-void vp10_inv_txfm_add_16x16(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, TX_TYPE tx_type) {
+void vp10_inv_txfm_add_16x16(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, TX_TYPE tx_type) {
switch (tx_type) {
- case DCT_DCT:
- vp10_idct16x16_add(input, dest, stride, eob);
- break;
+ case DCT_DCT: vp10_idct16x16_add(input, dest, stride, eob); break;
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- vp10_iht16x16_256_add(input, dest, stride, tx_type);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: vp10_iht16x16_256_add(input, dest, stride, tx_type); break;
+ default: assert(0); break;
}
}
-void vp10_inv_txfm_add_32x32(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, TX_TYPE tx_type) {
+void vp10_inv_txfm_add_32x32(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, TX_TYPE tx_type) {
switch (tx_type) {
- case DCT_DCT:
- vp10_idct32x32_add(input, dest, stride, eob);
- break;
+ case DCT_DCT: vp10_idct32x32_add(input, dest, stride, eob); break;
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- assert(0);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: assert(0); break;
+ default: assert(0); break;
}
}
#if CONFIG_VPX_HIGHBITDEPTH
void vp10_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int tx_type, int bd) {
+ int stride, int tx_type, int bd) {
const highbd_transform_2d IHT_4[] = {
- { vpx_highbd_idct4_c, vpx_highbd_idct4_c }, // DCT_DCT = 0
- { vpx_highbd_iadst4_c, vpx_highbd_idct4_c }, // ADST_DCT = 1
- { vpx_highbd_idct4_c, vpx_highbd_iadst4_c }, // DCT_ADST = 2
- { vpx_highbd_iadst4_c, vpx_highbd_iadst4_c } // ADST_ADST = 3
+ { vpx_highbd_idct4_c, vpx_highbd_idct4_c }, // DCT_DCT = 0
+ { vpx_highbd_iadst4_c, vpx_highbd_idct4_c }, // ADST_DCT = 1
+ { vpx_highbd_idct4_c, vpx_highbd_iadst4_c }, // DCT_ADST = 2
+ { vpx_highbd_iadst4_c, vpx_highbd_iadst4_c } // ADST_ADST = 3
};
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
@@ -270,14 +241,13 @@
// Inverse transform row vectors.
for (i = 0; i < 4; ++i) {
IHT_4[tx_type].rows(input, outptr, bd);
- input += 4;
+ input += 4;
outptr += 4;
}
// Inverse transform column vectors.
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j * 4 + i];
+ for (j = 0; j < 4; ++j) temp_in[j] = out[j * 4 + i];
IHT_4[tx_type].cols(temp_in, temp_out, bd);
for (j = 0; j < 4; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -287,14 +257,14 @@
}
static const highbd_transform_2d HIGH_IHT_8[] = {
- { vpx_highbd_idct8_c, vpx_highbd_idct8_c }, // DCT_DCT = 0
- { vpx_highbd_iadst8_c, vpx_highbd_idct8_c }, // ADST_DCT = 1
- { vpx_highbd_idct8_c, vpx_highbd_iadst8_c }, // DCT_ADST = 2
- { vpx_highbd_iadst8_c, vpx_highbd_iadst8_c } // ADST_ADST = 3
+ { vpx_highbd_idct8_c, vpx_highbd_idct8_c }, // DCT_DCT = 0
+ { vpx_highbd_iadst8_c, vpx_highbd_idct8_c }, // ADST_DCT = 1
+ { vpx_highbd_idct8_c, vpx_highbd_iadst8_c }, // DCT_ADST = 2
+ { vpx_highbd_iadst8_c, vpx_highbd_iadst8_c } // ADST_ADST = 3
};
void vp10_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int tx_type, int bd) {
+ int stride, int tx_type, int bd) {
int i, j;
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
@@ -311,8 +281,7 @@
// Inverse transform column vectors.
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
ht.cols(temp_in, temp_out, bd);
for (j = 0; j < 8; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -322,14 +291,14 @@
}
static const highbd_transform_2d HIGH_IHT_16[] = {
- { vpx_highbd_idct16_c, vpx_highbd_idct16_c }, // DCT_DCT = 0
- { vpx_highbd_iadst16_c, vpx_highbd_idct16_c }, // ADST_DCT = 1
- { vpx_highbd_idct16_c, vpx_highbd_iadst16_c }, // DCT_ADST = 2
- { vpx_highbd_iadst16_c, vpx_highbd_iadst16_c } // ADST_ADST = 3
+ { vpx_highbd_idct16_c, vpx_highbd_idct16_c }, // DCT_DCT = 0
+ { vpx_highbd_iadst16_c, vpx_highbd_idct16_c }, // ADST_DCT = 1
+ { vpx_highbd_idct16_c, vpx_highbd_iadst16_c }, // DCT_ADST = 2
+ { vpx_highbd_iadst16_c, vpx_highbd_iadst16_c } // ADST_ADST = 3
};
void vp10_highbd_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int tx_type, int bd) {
+ int stride, int tx_type, int bd) {
int i, j;
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
@@ -346,8 +315,7 @@
// Columns
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
ht.cols(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -358,16 +326,15 @@
// idct
void vp10_highbd_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd) {
+ int eob, int bd) {
if (eob > 1)
vpx_highbd_idct4x4_16_add(input, dest, stride, bd);
else
vpx_highbd_idct4x4_1_add(input, dest, stride, bd);
}
-
void vp10_highbd_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd) {
+ int eob, int bd) {
if (eob > 1)
vpx_highbd_iwht4x4_16_add(input, dest, stride, bd);
else
@@ -375,7 +342,7 @@
}
void vp10_highbd_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd) {
+ int eob, int bd) {
// If dc is 1, then input[0] is the reconstructed value, do not need
// dequantization. Also, when dc is 1, dc is counted in eobs, namely eobs >=1.
@@ -394,7 +361,7 @@
}
void vp10_highbd_idct16x16_add(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, int bd) {
+ int stride, int eob, int bd) {
// The calculation can be simplified if there are not many non-zero dct
// coefficients. Use eobs to separate different cases.
// DC only DCT coefficient.
@@ -408,7 +375,7 @@
}
void vp10_highbd_idct32x32_add(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, int bd) {
+ int stride, int eob, int bd) {
// Non-zero coeff only in upper-left 8x8
if (eob == 1) {
vpx_highbd_idct32x32_1_add(input, dest, stride, bd);
@@ -433,11 +400,9 @@
case ADST_DCT:
case DCT_ADST:
case ADST_ADST:
- vp10_highbd_iht4x4_16_add(input, dest, stride, tx_type, bd);
- break;
- default:
- assert(0);
+ vp10_highbd_iht4x4_16_add(input, dest, stride, tx_type, bd);
break;
+ default: assert(0); break;
}
}
}
@@ -446,17 +411,13 @@
int stride, int eob, int bd,
TX_TYPE tx_type) {
switch (tx_type) {
- case DCT_DCT:
- vp10_highbd_idct8x8_add(input, dest, stride, eob, bd);
- break;
+ case DCT_DCT: vp10_highbd_idct8x8_add(input, dest, stride, eob, bd); break;
case ADST_DCT:
case DCT_ADST:
case ADST_ADST:
vp10_highbd_iht8x8_64_add(input, dest, stride, tx_type, bd);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
@@ -472,9 +433,7 @@
case ADST_ADST:
vp10_highbd_iht16x16_256_add(input, dest, stride, tx_type, bd);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
@@ -487,12 +446,8 @@
break;
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- assert(0);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: assert(0); break;
+ default: assert(0); break;
}
}
#endif // CONFIG_VPX_HIGHBITDEPTH
diff --git a/vp10/common/idct.h b/vp10/common/idct.h
index 410fa14..a73e61f 100644
--- a/vp10/common/idct.h
+++ b/vp10/common/idct.h
@@ -24,14 +24,14 @@
extern "C" {
#endif
-typedef void (*transform_1d)(const tran_low_t*, tran_low_t*);
+typedef void (*transform_1d)(const tran_low_t *, tran_low_t *);
typedef struct {
transform_1d cols, rows; // vertical and horizontal
} transform_2d;
#if CONFIG_VPX_HIGHBITDEPTH
-typedef void (*highbd_transform_1d)(const tran_low_t*, tran_low_t*, int bd);
+typedef void (*highbd_transform_1d)(const tran_low_t *, tran_low_t *, int bd);
typedef struct {
highbd_transform_1d cols, rows; // vertical and horizontal
@@ -39,30 +39,30 @@
#endif // CONFIG_VPX_HIGHBITDEPTH
void vp10_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob);
+ int eob);
void vp10_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob);
+ int eob);
-void vp10_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, TX_TYPE tx_type, int lossless);
-void vp10_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, TX_TYPE tx_type);
-void vp10_inv_txfm_add_16x16(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, TX_TYPE tx_type);
-void vp10_inv_txfm_add_32x32(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, TX_TYPE tx_type);
+void vp10_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, TX_TYPE tx_type, int lossless);
+void vp10_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, TX_TYPE tx_type);
+void vp10_inv_txfm_add_16x16(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, TX_TYPE tx_type);
+void vp10_inv_txfm_add_32x32(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, TX_TYPE tx_type);
#if CONFIG_VPX_HIGHBITDEPTH
void vp10_highbd_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd);
+ int eob, int bd);
void vp10_highbd_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd);
+ int eob, int bd);
void vp10_highbd_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd);
+ int eob, int bd);
void vp10_highbd_idct16x16_add(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, int bd);
+ int stride, int eob, int bd);
void vp10_highbd_idct32x32_add(const tran_low_t *input, uint8_t *dest,
- int stride, int eob, int bd);
+ int stride, int eob, int bd);
void vp10_highbd_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest,
int stride, int eob, int bd, TX_TYPE tx_type,
int lossless);
diff --git a/vp10/common/loopfilter.c b/vp10/common/loopfilter.c
index 4c1716e..a17939d 100644
--- a/vp10/common/loopfilter.c
+++ b/vp10/common/loopfilter.c
@@ -36,7 +36,7 @@
// 10101010
//
// A loopfilter should be applied to every other 8x8 horizontally.
-static const uint64_t left_64x64_txform_mask[TX_SIZES]= {
+static const uint64_t left_64x64_txform_mask[TX_SIZES] = {
0xffffffffffffffffULL, // TX_4X4
0xffffffffffffffffULL, // TX_8x8
0x5555555555555555ULL, // TX_16x16
@@ -60,7 +60,7 @@
// 00000000
//
// A loopfilter should be applied to every other 4 the row vertically.
-static const uint64_t above_64x64_txform_mask[TX_SIZES]= {
+static const uint64_t above_64x64_txform_mask[TX_SIZES] = {
0xffffffffffffffffULL, // TX_4X4
0xffffffffffffffffULL, // TX_8x8
0x00ff00ff00ff00ffULL, // TX_16x16
@@ -134,18 +134,18 @@
};
// These are used for masking the left and above borders.
-static const uint64_t left_border = 0x1111111111111111ULL;
+static const uint64_t left_border = 0x1111111111111111ULL;
static const uint64_t above_border = 0x000000ff000000ffULL;
// 16 bit masks for uv transform sizes.
-static const uint16_t left_64x64_txform_mask_uv[TX_SIZES]= {
+static const uint16_t left_64x64_txform_mask_uv[TX_SIZES] = {
0xffff, // TX_4X4
0xffff, // TX_8x8
0x5555, // TX_16x16
0x1111, // TX_32x32
};
-static const uint16_t above_64x64_txform_mask_uv[TX_SIZES]= {
+static const uint16_t above_64x64_txform_mask_uv[TX_SIZES] = {
0xffff, // TX_4X4
0xffff, // TX_8x8
0x0f0f, // TX_16x16
@@ -201,7 +201,7 @@
0x00ff, // BLOCK_64X32,
0xffff, // BLOCK_64X64
};
-static const uint16_t left_border_uv = 0x1111;
+static const uint16_t left_border_uv = 0x1111;
static const uint16_t above_border_uv = 0x000f;
static const int mode_lf_lut[MB_MODE_COUNT] = {
@@ -222,8 +222,7 @@
block_inside_limit = (9 - sharpness_lvl);
}
- if (block_inside_limit < 1)
- block_inside_limit = 1;
+ if (block_inside_limit < 1) block_inside_limit = 1;
memset(lfi->lfthr[lvl].lim, block_inside_limit, SIMD_WIDTH);
memset(lfi->lfthr[lvl].mblim, (2 * (lvl + 2) + block_inside_limit),
@@ -233,8 +232,8 @@
static uint8_t get_filter_level(const loop_filter_info_n *lfi_n,
const MB_MODE_INFO *mbmi) {
- return lfi_n->lvl[mbmi->segment_id][mbmi->ref_frame[0]]
- [mode_lf_lut[mbmi->mode]];
+ return lfi_n
+ ->lvl[mbmi->segment_id][mbmi->ref_frame[0]][mode_lf_lut[mbmi->mode]];
}
void vp10_loop_filter_init(VP10_COMMON *cm) {
@@ -271,9 +270,9 @@
int lvl_seg = default_filt_lvl;
if (segfeature_active(seg, seg_id, SEG_LVL_ALT_LF)) {
const int data = get_segdata(seg, seg_id, SEG_LVL_ALT_LF);
- lvl_seg = clamp(seg->abs_delta == SEGMENT_ABSDATA ?
- data : default_filt_lvl + data,
- 0, MAX_LOOP_FILTER);
+ lvl_seg = clamp(
+ seg->abs_delta == SEGMENT_ABSDATA ? data : default_filt_lvl + data, 0,
+ MAX_LOOP_FILTER);
}
if (!lf->mode_ref_delta_enabled) {
@@ -287,8 +286,8 @@
for (ref = LAST_FRAME; ref < MAX_REF_FRAMES; ++ref) {
for (mode = 0; mode < MAX_MODE_LF_DELTAS; ++mode) {
- const int inter_lvl = lvl_seg + lf->ref_deltas[ref] * scale
- + lf->mode_deltas[mode] * scale;
+ const int inter_lvl = lvl_seg + lf->ref_deltas[ref] * scale +
+ lf->mode_deltas[mode] * scale;
lfi->lvl[seg_id][ref][mode] = clamp(inter_lvl, 0, MAX_LOOP_FILTER);
}
}
@@ -296,9 +295,8 @@
}
}
-static void filter_selectively_vert_row2(int subsampling_factor,
- uint8_t *s, int pitch,
- unsigned int mask_16x16_l,
+static void filter_selectively_vert_row2(int subsampling_factor, uint8_t *s,
+ int pitch, unsigned int mask_16x16_l,
unsigned int mask_8x8_l,
unsigned int mask_4x4_l,
unsigned int mask_4x4_int_l,
@@ -331,11 +329,10 @@
vpx_lpf_vertical_16_dual(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr);
} else if (mask_16x16_0 & 1) {
- vpx_lpf_vertical_16(s, pitch, lfi0->mblim, lfi0->lim,
- lfi0->hev_thr);
+ vpx_lpf_vertical_16(s, pitch, lfi0->mblim, lfi0->lim, lfi0->hev_thr);
} else {
- vpx_lpf_vertical_16(s + 8 *pitch, pitch, lfi1->mblim,
- lfi1->lim, lfi1->hev_thr);
+ vpx_lpf_vertical_16(s + 8 * pitch, pitch, lfi1->mblim, lfi1->lim,
+ lfi1->hev_thr);
}
}
@@ -396,14 +393,11 @@
}
#if CONFIG_VPX_HIGHBITDEPTH
-static void highbd_filter_selectively_vert_row2(int subsampling_factor,
- uint16_t *s, int pitch,
- unsigned int mask_16x16_l,
- unsigned int mask_8x8_l,
- unsigned int mask_4x4_l,
- unsigned int mask_4x4_int_l,
- const loop_filter_info_n *lfi_n,
- const uint8_t *lfl, int bd) {
+static void highbd_filter_selectively_vert_row2(
+ int subsampling_factor, uint16_t *s, int pitch, unsigned int mask_16x16_l,
+ unsigned int mask_8x8_l, unsigned int mask_4x4_l,
+ unsigned int mask_4x4_int_l, const loop_filter_info_n *lfi_n,
+ const uint8_t *lfl, int bd) {
const int mask_shift = subsampling_factor ? 4 : 8;
const int mask_cutoff = subsampling_factor ? 0xf : 0xff;
const int lfl_forward = subsampling_factor ? 4 : 8;
@@ -419,7 +413,7 @@
unsigned int mask;
for (mask = mask_16x16_0 | mask_8x8_0 | mask_4x4_0 | mask_4x4_int_0 |
- mask_16x16_1 | mask_8x8_1 | mask_4x4_1 | mask_4x4_int_1;
+ mask_16x16_1 | mask_8x8_1 | mask_4x4_1 | mask_4x4_int_1;
mask; mask >>= 1) {
const loop_filter_thresh *lfi0 = lfi_n->lfthr + *lfl;
const loop_filter_thresh *lfi1 = lfi_n->lfthr + *(lfl + lfl_forward);
@@ -434,7 +428,7 @@
vpx_highbd_lpf_vertical_16(s, pitch, lfi0->mblim, lfi0->lim,
lfi0->hev_thr, bd);
} else {
- vpx_highbd_lpf_vertical_16(s + 8 *pitch, pitch, lfi1->mblim,
+ vpx_highbd_lpf_vertical_16(s + 8 * pitch, pitch, lfi1->mblim,
lfi1->lim, lfi1->hev_thr, bd);
}
}
@@ -496,30 +490,27 @@
}
#endif // CONFIG_VPX_HIGHBITDEPTH
-static void filter_selectively_horiz(uint8_t *s, int pitch,
- unsigned int mask_16x16,
- unsigned int mask_8x8,
- unsigned int mask_4x4,
- unsigned int mask_4x4_int,
- const loop_filter_info_n *lfi_n,
- const uint8_t *lfl) {
+static void filter_selectively_horiz(
+ uint8_t *s, int pitch, unsigned int mask_16x16, unsigned int mask_8x8,
+ unsigned int mask_4x4, unsigned int mask_4x4_int,
+ const loop_filter_info_n *lfi_n, const uint8_t *lfl) {
unsigned int mask;
int count;
- for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int;
- mask; mask >>= count) {
+ for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int; mask;
+ mask >>= count) {
const loop_filter_thresh *lfi = lfi_n->lfthr + *lfl;
count = 1;
if (mask & 1) {
if (mask_16x16 & 1) {
if ((mask_16x16 & 3) == 3) {
- vpx_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, 2);
+ vpx_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr,
+ 2);
count = 2;
} else {
- vpx_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, 1);
+ vpx_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr,
+ 1);
}
} else if (mask_8x8 & 1) {
if ((mask_8x8 & 3) == 3) {
@@ -593,18 +584,15 @@
}
#if CONFIG_VPX_HIGHBITDEPTH
-static void highbd_filter_selectively_horiz(uint16_t *s, int pitch,
- unsigned int mask_16x16,
- unsigned int mask_8x8,
- unsigned int mask_4x4,
- unsigned int mask_4x4_int,
- const loop_filter_info_n *lfi_n,
- const uint8_t *lfl, int bd) {
+static void highbd_filter_selectively_horiz(
+ uint16_t *s, int pitch, unsigned int mask_16x16, unsigned int mask_8x8,
+ unsigned int mask_4x4, unsigned int mask_4x4_int,
+ const loop_filter_info_n *lfi_n, const uint8_t *lfl, int bd) {
unsigned int mask;
int count;
- for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int;
- mask; mask >>= count) {
+ for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int; mask;
+ mask >>= count) {
const loop_filter_thresh *lfi = lfi_n->lfthr + *lfl;
count = 1;
@@ -628,10 +616,9 @@
lfin->hev_thr, bd);
if ((mask_4x4_int & 3) == 3) {
- vpx_highbd_lpf_horizontal_4_dual(s + 4 * pitch, pitch, lfi->mblim,
- lfi->lim, lfi->hev_thr,
- lfin->mblim, lfin->lim,
- lfin->hev_thr, bd);
+ vpx_highbd_lpf_horizontal_4_dual(
+ s + 4 * pitch, pitch, lfi->mblim, lfi->lim, lfi->hev_thr,
+ lfin->mblim, lfin->lim, lfin->hev_thr, bd);
} else {
if (mask_4x4_int & 1) {
vpx_highbd_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim,
@@ -660,10 +647,9 @@
lfi->hev_thr, lfin->mblim, lfin->lim,
lfin->hev_thr, bd);
if ((mask_4x4_int & 3) == 3) {
- vpx_highbd_lpf_horizontal_4_dual(s + 4 * pitch, pitch, lfi->mblim,
- lfi->lim, lfi->hev_thr,
- lfin->mblim, lfin->lim,
- lfin->hev_thr, bd);
+ vpx_highbd_lpf_horizontal_4_dual(
+ s + 4 * pitch, pitch, lfi->mblim, lfi->lim, lfi->hev_thr,
+ lfin->mblim, lfin->lim, lfin->hev_thr, bd);
} else {
if (mask_4x4_int & 1) {
vpx_highbd_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim,
@@ -707,8 +693,7 @@
// TODO(JBB) Need another function for different resolution color..
static void build_masks(const loop_filter_info_n *const lfi_n,
const MODE_INFO *mi, const int shift_y,
- const int shift_uv,
- LOOP_FILTER_MASK *lfm) {
+ const int shift_uv, LOOP_FILTER_MASK *lfm) {
const MB_MODE_INFO *mbmi = &mi->mbmi;
const BLOCK_SIZE block_size = mbmi->sb_type;
const TX_SIZE tx_size_y = mbmi->tx_size;
@@ -756,29 +741,28 @@
*left_y |= left_prediction_mask[block_size] << shift_y;
*left_uv |= left_prediction_mask_uv[block_size] << shift_uv;
- // If the block has no coefficients and is not intra we skip applying
- // the loop filter on block edges.
+// If the block has no coefficients and is not intra we skip applying
+// the loop filter on block edges.
#if CONFIG_MISC_FIXES
- if ((mbmi->skip || mbmi->has_no_coeffs) && is_inter_block(mbmi))
- return;
+ if ((mbmi->skip || mbmi->has_no_coeffs) && is_inter_block(mbmi)) return;
#else
- if (mbmi->skip && is_inter_block(mbmi))
- return;
+ if (mbmi->skip && is_inter_block(mbmi)) return;
#endif
// Here we are adding a mask for the transform size. The transform
// size mask is set to be correct for a 64x64 prediction block size. We
// mask to match the size of the block we are working on and then shift it
// into place..
- *above_y |= (size_mask[block_size] &
- above_64x64_txform_mask[tx_size_y]) << shift_y;
- *above_uv |= (size_mask_uv[block_size] &
- above_64x64_txform_mask_uv[tx_size_uv]) << shift_uv;
+ *above_y |= (size_mask[block_size] & above_64x64_txform_mask[tx_size_y])
+ << shift_y;
+ *above_uv |=
+ (size_mask_uv[block_size] & above_64x64_txform_mask_uv[tx_size_uv])
+ << shift_uv;
- *left_y |= (size_mask[block_size] &
- left_64x64_txform_mask[tx_size_y]) << shift_y;
- *left_uv |= (size_mask_uv[block_size] &
- left_64x64_txform_mask_uv[tx_size_uv]) << shift_uv;
+ *left_y |= (size_mask[block_size] & left_64x64_txform_mask[tx_size_y])
+ << shift_y;
+ *left_uv |= (size_mask_uv[block_size] & left_64x64_txform_mask_uv[tx_size_uv])
+ << shift_uv;
// Here we are trying to determine what to do with the internal 4x4 block
// boundaries. These differ from the 4x4 boundaries on the outside edge of
@@ -822,18 +806,16 @@
*left_y |= left_prediction_mask[block_size] << shift_y;
#if CONFIG_MISC_FIXES
- if ((mbmi->skip || mbmi->has_no_coeffs) && is_inter_block(mbmi))
- return;
+ if ((mbmi->skip || mbmi->has_no_coeffs) && is_inter_block(mbmi)) return;
#else
- if (mbmi->skip && is_inter_block(mbmi))
- return;
+ if (mbmi->skip && is_inter_block(mbmi)) return;
#endif
- *above_y |= (size_mask[block_size] &
- above_64x64_txform_mask[tx_size_y]) << shift_y;
+ *above_y |= (size_mask[block_size] & above_64x64_txform_mask[tx_size_y])
+ << shift_y;
- *left_y |= (size_mask[block_size] &
- left_64x64_txform_mask[tx_size_y]) << shift_y;
+ *left_y |= (size_mask[block_size] & left_64x64_txform_mask[tx_size_y])
+ << shift_y;
if (tx_size_y == TX_4X4)
*int_4x4_y |= (size_mask[block_size] & 0xffffffffffffffffULL) << shift_y;
@@ -843,8 +825,8 @@
// by mi_row, mi_col.
// TODO(JBB): This function only works for yv12.
void vp10_setup_mask(VP10_COMMON *const cm, const int mi_row, const int mi_col,
- MODE_INFO **mi, const int mode_info_stride,
- LOOP_FILTER_MASK *lfm) {
+ MODE_INFO **mi, const int mode_info_stride,
+ LOOP_FILTER_MASK *lfm) {
int idx_32, idx_16, idx_8;
const loop_filter_info_n *const lfi_n = &cm->lf_info;
MODE_INFO **mip = mi;
@@ -854,26 +836,28 @@
// added to the mi ptr as we go through each loop. It helps us to avoid
// setting up special row and column counters for each index. The last step
// brings us out back to the starting position.
- const int offset_32[] = {4, (mode_info_stride << 2) - 4, 4,
- -(mode_info_stride << 2) - 4};
- const int offset_16[] = {2, (mode_info_stride << 1) - 2, 2,
- -(mode_info_stride << 1) - 2};
- const int offset[] = {1, mode_info_stride - 1, 1, -mode_info_stride - 1};
+ const int offset_32[] = { 4, (mode_info_stride << 2) - 4, 4,
+ -(mode_info_stride << 2) - 4 };
+ const int offset_16[] = { 2, (mode_info_stride << 1) - 2, 2,
+ -(mode_info_stride << 1) - 2 };
+ const int offset[] = { 1, mode_info_stride - 1, 1, -mode_info_stride - 1 };
// Following variables represent shifts to position the current block
// mask over the appropriate block. A shift of 36 to the left will move
// the bits for the final 32 by 32 block in the 64x64 up 4 rows and left
// 4 rows to the appropriate spot.
- const int shift_32_y[] = {0, 4, 32, 36};
- const int shift_16_y[] = {0, 2, 16, 18};
- const int shift_8_y[] = {0, 1, 8, 9};
- const int shift_32_uv[] = {0, 2, 8, 10};
- const int shift_16_uv[] = {0, 1, 4, 5};
+ const int shift_32_y[] = { 0, 4, 32, 36 };
+ const int shift_16_y[] = { 0, 2, 16, 18 };
+ const int shift_8_y[] = { 0, 1, 8, 9 };
+ const int shift_32_uv[] = { 0, 2, 8, 10 };
+ const int shift_16_uv[] = { 0, 1, 4, 5 };
int i;
- const int max_rows = (mi_row + MI_BLOCK_SIZE > cm->mi_rows ?
- cm->mi_rows - mi_row : MI_BLOCK_SIZE);
- const int max_cols = (mi_col + MI_BLOCK_SIZE > cm->mi_cols ?
- cm->mi_cols - mi_col : MI_BLOCK_SIZE);
+ const int max_rows =
+ (mi_row + MI_BLOCK_SIZE > cm->mi_rows ? cm->mi_rows - mi_row
+ : MI_BLOCK_SIZE);
+ const int max_cols =
+ (mi_col + MI_BLOCK_SIZE > cm->mi_cols ? cm->mi_cols - mi_col
+ : MI_BLOCK_SIZE);
vp10_zero(*lfm);
assert(mip[0] != NULL);
@@ -882,21 +866,17 @@
// loop and storing lfm in the mbmi structure so that we don't have to go
// through the recursive loop structure multiple times.
switch (mip[0]->mbmi.sb_type) {
- case BLOCK_64X64:
- build_masks(lfi_n, mip[0] , 0, 0, lfm);
- break;
+ case BLOCK_64X64: build_masks(lfi_n, mip[0], 0, 0, lfm); break;
case BLOCK_64X32:
build_masks(lfi_n, mip[0], 0, 0, lfm);
mip2 = mip + mode_info_stride * 4;
- if (4 >= max_rows)
- break;
+ if (4 >= max_rows) break;
build_masks(lfi_n, mip2[0], 32, 8, lfm);
break;
case BLOCK_32X64:
build_masks(lfi_n, mip[0], 0, 0, lfm);
mip2 = mip + 4;
- if (4 >= max_cols)
- break;
+ if (4 >= max_cols) break;
build_masks(lfi_n, mip2[0], 4, 2, lfm);
break;
default:
@@ -913,15 +893,13 @@
break;
case BLOCK_32X16:
build_masks(lfi_n, mip[0], shift_y, shift_uv, lfm);
- if (mi_32_row_offset + 2 >= max_rows)
- continue;
+ if (mi_32_row_offset + 2 >= max_rows) continue;
mip2 = mip + mode_info_stride * 2;
build_masks(lfi_n, mip2[0], shift_y + 16, shift_uv + 4, lfm);
break;
case BLOCK_16X32:
build_masks(lfi_n, mip[0], shift_y, shift_uv, lfm);
- if (mi_32_col_offset + 2 >= max_cols)
- continue;
+ if (mi_32_col_offset + 2 >= max_cols) continue;
mip2 = mip + 2;
build_masks(lfi_n, mip2[0], shift_y + 2, shift_uv + 1, lfm);
break;
@@ -929,10 +907,10 @@
for (idx_16 = 0; idx_16 < 4; mip += offset_16[idx_16], ++idx_16) {
const int shift_y = shift_32_y[idx_32] + shift_16_y[idx_16];
const int shift_uv = shift_32_uv[idx_32] + shift_16_uv[idx_16];
- const int mi_16_col_offset = mi_32_col_offset +
- ((idx_16 & 1) << 1);
- const int mi_16_row_offset = mi_32_row_offset +
- ((idx_16 >> 1) << 1);
+ const int mi_16_col_offset =
+ mi_32_col_offset + ((idx_16 & 1) << 1);
+ const int mi_16_row_offset =
+ mi_32_row_offset + ((idx_16 >> 1) << 1);
if (mi_16_col_offset >= max_cols || mi_16_row_offset >= max_rows)
continue;
@@ -943,32 +921,28 @@
break;
case BLOCK_16X8:
build_masks(lfi_n, mip[0], shift_y, shift_uv, lfm);
- if (mi_16_row_offset + 1 >= max_rows)
- continue;
+ if (mi_16_row_offset + 1 >= max_rows) continue;
mip2 = mip + mode_info_stride;
- build_y_mask(lfi_n, mip2[0], shift_y+8, lfm);
+ build_y_mask(lfi_n, mip2[0], shift_y + 8, lfm);
break;
case BLOCK_8X16:
build_masks(lfi_n, mip[0], shift_y, shift_uv, lfm);
- if (mi_16_col_offset +1 >= max_cols)
- continue;
+ if (mi_16_col_offset + 1 >= max_cols) continue;
mip2 = mip + 1;
- build_y_mask(lfi_n, mip2[0], shift_y+1, lfm);
+ build_y_mask(lfi_n, mip2[0], shift_y + 1, lfm);
break;
default: {
- const int shift_y = shift_32_y[idx_32] +
- shift_16_y[idx_16] +
- shift_8_y[0];
+ const int shift_y =
+ shift_32_y[idx_32] + shift_16_y[idx_16] + shift_8_y[0];
build_masks(lfi_n, mip[0], shift_y, shift_uv, lfm);
mip += offset[0];
for (idx_8 = 1; idx_8 < 4; mip += offset[idx_8], ++idx_8) {
const int shift_y = shift_32_y[idx_32] +
- shift_16_y[idx_16] +
- shift_8_y[idx_8];
- const int mi_8_col_offset = mi_16_col_offset +
- ((idx_8 & 1));
- const int mi_8_row_offset = mi_16_row_offset +
- ((idx_8 >> 1));
+ shift_16_y[idx_16] + shift_8_y[idx_8];
+ const int mi_8_col_offset =
+ mi_16_col_offset + ((idx_8 & 1));
+ const int mi_8_row_offset =
+ mi_16_row_offset + ((idx_8 >> 1));
if (mi_8_col_offset >= max_cols ||
mi_8_row_offset >= max_rows)
@@ -1008,8 +982,8 @@
const uint64_t rows = cm->mi_rows - mi_row;
// Each pixel inside the border gets a 1,
- const uint64_t mask_y = (((uint64_t) 1 << (rows << 3)) - 1);
- const uint16_t mask_uv = (((uint16_t) 1 << (((rows + 1) >> 1) << 2)) - 1);
+ const uint64_t mask_y = (((uint64_t)1 << (rows << 3)) - 1);
+ const uint16_t mask_uv = (((uint16_t)1 << (((rows + 1) >> 1) << 2)) - 1);
// Remove values completely outside our border.
for (i = 0; i < TX_32X32; i++) {
@@ -1042,7 +1016,7 @@
// Each pixel inside the border gets a 1, the multiply copies the border
// to where we need it.
- const uint64_t mask_y = (((1 << columns) - 1)) * 0x0101010101010101ULL;
+ const uint64_t mask_y = (((1 << columns) - 1)) * 0x0101010101010101ULL;
const uint16_t mask_uv = ((1 << ((columns + 1) >> 1)) - 1) * 0x1111;
// Internal edges are not applied on the last column of the image so
@@ -1088,7 +1062,7 @@
assert(!(lfm->left_y[TX_16X16] & lfm->left_y[TX_4X4]));
assert(!(lfm->left_y[TX_8X8] & lfm->left_y[TX_4X4]));
assert(!(lfm->int_4x4_y & lfm->left_y[TX_16X16]));
- assert(!(lfm->left_uv[TX_16X16]&lfm->left_uv[TX_8X8]));
+ assert(!(lfm->left_uv[TX_16X16] & lfm->left_uv[TX_8X8]));
assert(!(lfm->left_uv[TX_16X16] & lfm->left_uv[TX_4X4]));
assert(!(lfm->left_uv[TX_8X8] & lfm->left_uv[TX_4X4]));
#if CONFIG_MISC_FIXES
@@ -1110,17 +1084,14 @@
#endif
}
-static void filter_selectively_vert(uint8_t *s, int pitch,
- unsigned int mask_16x16,
- unsigned int mask_8x8,
- unsigned int mask_4x4,
- unsigned int mask_4x4_int,
- const loop_filter_info_n *lfi_n,
- const uint8_t *lfl) {
+static void filter_selectively_vert(
+ uint8_t *s, int pitch, unsigned int mask_16x16, unsigned int mask_8x8,
+ unsigned int mask_4x4, unsigned int mask_4x4_int,
+ const loop_filter_info_n *lfi_n, const uint8_t *lfl) {
unsigned int mask;
- for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int;
- mask; mask >>= 1) {
+ for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int; mask;
+ mask >>= 1) {
const loop_filter_thresh *lfi = lfi_n->lfthr + *lfl;
if (mask & 1) {
@@ -1144,29 +1115,26 @@
}
#if CONFIG_VPX_HIGHBITDEPTH
-static void highbd_filter_selectively_vert(uint16_t *s, int pitch,
- unsigned int mask_16x16,
- unsigned int mask_8x8,
- unsigned int mask_4x4,
- unsigned int mask_4x4_int,
- const loop_filter_info_n *lfi_n,
- const uint8_t *lfl, int bd) {
+static void highbd_filter_selectively_vert(
+ uint16_t *s, int pitch, unsigned int mask_16x16, unsigned int mask_8x8,
+ unsigned int mask_4x4, unsigned int mask_4x4_int,
+ const loop_filter_info_n *lfi_n, const uint8_t *lfl, int bd) {
unsigned int mask;
- for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int;
- mask; mask >>= 1) {
+ for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int; mask;
+ mask >>= 1) {
const loop_filter_thresh *lfi = lfi_n->lfthr + *lfl;
if (mask & 1) {
if (mask_16x16 & 1) {
- vpx_highbd_lpf_vertical_16(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, bd);
+ vpx_highbd_lpf_vertical_16(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr,
+ bd);
} else if (mask_8x8 & 1) {
- vpx_highbd_lpf_vertical_8(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, 1, bd);
+ vpx_highbd_lpf_vertical_8(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr,
+ 1, bd);
} else if (mask_4x4 & 1) {
- vpx_highbd_lpf_vertical_4(s, pitch, lfi->mblim, lfi->lim,
- lfi->hev_thr, 1, bd);
+ vpx_highbd_lpf_vertical_4(s, pitch, lfi->mblim, lfi->lim, lfi->hev_thr,
+ 1, bd);
}
}
if (mask_4x4_int & 1)
@@ -1183,20 +1151,20 @@
#endif // CONFIG_VPX_HIGHBITDEPTH
void vp10_filter_block_plane_non420(VP10_COMMON *cm,
- struct macroblockd_plane *plane,
- MODE_INFO **mi_8x8,
- int mi_row, int mi_col) {
+ struct macroblockd_plane *plane,
+ MODE_INFO **mi_8x8, int mi_row,
+ int mi_col) {
const int ss_x = plane->subsampling_x;
const int ss_y = plane->subsampling_y;
const int row_step = 1 << ss_y;
const int col_step = 1 << ss_x;
const int row_step_stride = cm->mi_stride * row_step;
struct buf_2d *const dst = &plane->dst;
- uint8_t* const dst0 = dst->buf;
- unsigned int mask_16x16[MI_BLOCK_SIZE] = {0};
- unsigned int mask_8x8[MI_BLOCK_SIZE] = {0};
- unsigned int mask_4x4[MI_BLOCK_SIZE] = {0};
- unsigned int mask_4x4_int[MI_BLOCK_SIZE] = {0};
+ uint8_t *const dst0 = dst->buf;
+ unsigned int mask_16x16[MI_BLOCK_SIZE] = { 0 };
+ unsigned int mask_8x8[MI_BLOCK_SIZE] = { 0 };
+ unsigned int mask_4x4[MI_BLOCK_SIZE] = { 0 };
+ unsigned int mask_4x4_int[MI_BLOCK_SIZE] = { 0 };
uint8_t lfl[MI_BLOCK_SIZE * MI_BLOCK_SIZE];
int r, c;
@@ -1212,22 +1180,26 @@
const BLOCK_SIZE sb_type = mi[0].mbmi.sb_type;
const int skip_this = mi[0].mbmi.skip && is_inter_block(&mi[0].mbmi);
// left edge of current unit is block/partition edge -> no skip
- const int block_edge_left = (num_4x4_blocks_wide_lookup[sb_type] > 1) ?
- !(c & (num_8x8_blocks_wide_lookup[sb_type] - 1)) : 1;
+ const int block_edge_left =
+ (num_4x4_blocks_wide_lookup[sb_type] > 1)
+ ? !(c & (num_8x8_blocks_wide_lookup[sb_type] - 1))
+ : 1;
const int skip_this_c = skip_this && !block_edge_left;
// top edge of current unit is block/partition edge -> no skip
- const int block_edge_above = (num_4x4_blocks_high_lookup[sb_type] > 1) ?
- !(r & (num_8x8_blocks_high_lookup[sb_type] - 1)) : 1;
+ const int block_edge_above =
+ (num_4x4_blocks_high_lookup[sb_type] > 1)
+ ? !(r & (num_8x8_blocks_high_lookup[sb_type] - 1))
+ : 1;
const int skip_this_r = skip_this && !block_edge_above;
const TX_SIZE tx_size = (plane->plane_type == PLANE_TYPE_UV)
- ? get_uv_tx_size(&mi[0].mbmi, plane)
- : mi[0].mbmi.tx_size;
+ ? get_uv_tx_size(&mi[0].mbmi, plane)
+ : mi[0].mbmi.tx_size;
const int skip_border_4x4_c = ss_x && mi_col + c == cm->mi_cols - 1;
const int skip_border_4x4_r = ss_y && mi_row + r == cm->mi_rows - 1;
// Filter level can vary per MI
if (!(lfl[(r << 3) + (c >> ss_x)] =
- get_filter_level(&cm->lf_info, &mi[0].mbmi)))
+ get_filter_level(&cm->lf_info, &mi[0].mbmi)))
continue;
// Build masks based on the transform size of each block
@@ -1282,29 +1254,21 @@
border_mask = ~(mi_col == 0);
#if CONFIG_VPX_HIGHBITDEPTH
if (cm->use_highbitdepth) {
- highbd_filter_selectively_vert(CONVERT_TO_SHORTPTR(dst->buf),
- dst->stride,
- mask_16x16_c & border_mask,
- mask_8x8_c & border_mask,
- mask_4x4_c & border_mask,
- mask_4x4_int[r],
- &cm->lf_info, &lfl[r << 3],
- (int)cm->bit_depth);
+ highbd_filter_selectively_vert(
+ CONVERT_TO_SHORTPTR(dst->buf), dst->stride,
+ mask_16x16_c & border_mask, mask_8x8_c & border_mask,
+ mask_4x4_c & border_mask, mask_4x4_int[r], &cm->lf_info, &lfl[r << 3],
+ (int)cm->bit_depth);
} else {
- filter_selectively_vert(dst->buf, dst->stride,
- mask_16x16_c & border_mask,
+ filter_selectively_vert(dst->buf, dst->stride, mask_16x16_c & border_mask,
mask_8x8_c & border_mask,
- mask_4x4_c & border_mask,
- mask_4x4_int[r],
+ mask_4x4_c & border_mask, mask_4x4_int[r],
&cm->lf_info, &lfl[r << 3]);
}
#else
- filter_selectively_vert(dst->buf, dst->stride,
- mask_16x16_c & border_mask,
- mask_8x8_c & border_mask,
- mask_4x4_c & border_mask,
- mask_4x4_int[r],
- &cm->lf_info, &lfl[r << 3]);
+ filter_selectively_vert(dst->buf, dst->stride, mask_16x16_c & border_mask,
+ mask_8x8_c & border_mask, mask_4x4_c & border_mask,
+ mask_4x4_int[r], &cm->lf_info, &lfl[r << 3]);
#endif // CONFIG_VPX_HIGHBITDEPTH
dst->buf += 8 * dst->stride;
mi_8x8 += row_step_stride;
@@ -1332,37 +1296,26 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (cm->use_highbitdepth) {
highbd_filter_selectively_horiz(CONVERT_TO_SHORTPTR(dst->buf),
- dst->stride,
- mask_16x16_r,
- mask_8x8_r,
- mask_4x4_r,
- mask_4x4_int_r,
- &cm->lf_info, &lfl[r << 3],
- (int)cm->bit_depth);
+ dst->stride, mask_16x16_r, mask_8x8_r,
+ mask_4x4_r, mask_4x4_int_r, &cm->lf_info,
+ &lfl[r << 3], (int)cm->bit_depth);
} else {
- filter_selectively_horiz(dst->buf, dst->stride,
- mask_16x16_r,
- mask_8x8_r,
- mask_4x4_r,
- mask_4x4_int_r,
- &cm->lf_info, &lfl[r << 3]);
+ filter_selectively_horiz(dst->buf, dst->stride, mask_16x16_r, mask_8x8_r,
+ mask_4x4_r, mask_4x4_int_r, &cm->lf_info,
+ &lfl[r << 3]);
}
#else
- filter_selectively_horiz(dst->buf, dst->stride,
- mask_16x16_r,
- mask_8x8_r,
- mask_4x4_r,
- mask_4x4_int_r,
- &cm->lf_info, &lfl[r << 3]);
+ filter_selectively_horiz(dst->buf, dst->stride, mask_16x16_r, mask_8x8_r,
+ mask_4x4_r, mask_4x4_int_r, &cm->lf_info,
+ &lfl[r << 3]);
#endif // CONFIG_VPX_HIGHBITDEPTH
dst->buf += 8 * dst->stride;
}
}
void vp10_filter_block_plane_ss00(VP10_COMMON *const cm,
- struct macroblockd_plane *const plane,
- int mi_row,
- LOOP_FILTER_MASK *lfm) {
+ struct macroblockd_plane *const plane,
+ int mi_row, LOOP_FILTER_MASK *lfm) {
struct buf_2d *const dst = &plane->dst;
uint8_t *const dst0 = dst->buf;
int r;
@@ -1452,9 +1405,8 @@
}
void vp10_filter_block_plane_ss11(VP10_COMMON *const cm,
- struct macroblockd_plane *const plane,
- int mi_row,
- LOOP_FILTER_MASK *lfm) {
+ struct macroblockd_plane *const plane,
+ int mi_row, LOOP_FILTER_MASK *lfm) {
struct buf_2d *const dst = &plane->dst;
uint8_t *const dst0 = dst->buf;
int r, c;
@@ -1493,16 +1445,15 @@
mask_16x16_l, mask_8x8_l, mask_4x4_l, mask_4x4_int_l, &cm->lf_info,
&lfm->lfl_uv[r << 1], (int)cm->bit_depth);
} else {
- filter_selectively_vert_row2(
- plane->subsampling_x, dst->buf, dst->stride,
- mask_16x16_l, mask_8x8_l, mask_4x4_l, mask_4x4_int_l, &cm->lf_info,
- &lfm->lfl_uv[r << 1]);
+ filter_selectively_vert_row2(plane->subsampling_x, dst->buf,
+ dst->stride, mask_16x16_l, mask_8x8_l,
+ mask_4x4_l, mask_4x4_int_l, &cm->lf_info,
+ &lfm->lfl_uv[r << 1]);
}
#else
filter_selectively_vert_row2(
- plane->subsampling_x, dst->buf, dst->stride,
- mask_16x16_l, mask_8x8_l, mask_4x4_l, mask_4x4_int_l, &cm->lf_info,
- &lfm->lfl_uv[r << 1]);
+ plane->subsampling_x, dst->buf, dst->stride, mask_16x16_l, mask_8x8_l,
+ mask_4x4_l, mask_4x4_int_l, &cm->lf_info, &lfm->lfl_uv[r << 1]);
#endif // CONFIG_VPX_HIGHBITDEPTH
dst->buf += 16 * dst->stride;
@@ -1567,10 +1518,9 @@
}
}
-void vp10_loop_filter_rows(YV12_BUFFER_CONFIG *frame_buffer,
- VP10_COMMON *cm,
- struct macroblockd_plane planes[MAX_MB_PLANE],
- int start, int stop, int y_only) {
+void vp10_loop_filter_rows(YV12_BUFFER_CONFIG *frame_buffer, VP10_COMMON *cm,
+ struct macroblockd_plane planes[MAX_MB_PLANE],
+ int start, int stop, int y_only) {
const int num_planes = y_only ? 1 : MAX_MB_PLANE;
enum lf_path path;
LOOP_FILTER_MASK lfm;
@@ -1594,8 +1544,7 @@
vp10_setup_dst_planes(planes, frame_buffer, mi_row, mi_col);
// TODO(JBB): Make setup_mask work for non 420.
- vp10_setup_mask(cm, mi_row, mi_col, mi + mi_col, cm->mi_stride,
- &lfm);
+ vp10_setup_mask(cm, mi_row, mi_col, mi + mi_col, cm->mi_stride, &lfm);
vp10_filter_block_plane_ss00(cm, &planes[0], mi_row, &lfm);
for (plane = 1; plane < num_planes; ++plane) {
@@ -1608,7 +1557,7 @@
break;
case LF_PATH_SLOW:
vp10_filter_block_plane_non420(cm, &planes[plane], mi + mi_col,
- mi_row, mi_col);
+ mi_row, mi_col);
break;
}
}
@@ -1616,10 +1565,9 @@
}
}
-void vp10_loop_filter_frame(YV12_BUFFER_CONFIG *frame,
- VP10_COMMON *cm, MACROBLOCKD *xd,
- int frame_filter_level,
- int y_only, int partial_frame) {
+void vp10_loop_filter_frame(YV12_BUFFER_CONFIG *frame, VP10_COMMON *cm,
+ MACROBLOCKD *xd, int frame_filter_level, int y_only,
+ int partial_frame) {
int start_mi_row, end_mi_row, mi_rows_to_filter;
if (!frame_filter_level) return;
start_mi_row = 0;
@@ -1631,9 +1579,7 @@
}
end_mi_row = start_mi_row + mi_rows_to_filter;
vp10_loop_filter_frame_init(cm, frame_filter_level);
- vp10_loop_filter_rows(frame, cm, xd->plane,
- start_mi_row, end_mi_row,
- y_only);
+ vp10_loop_filter_rows(frame, cm, xd->plane, start_mi_row, end_mi_row, y_only);
}
void vp10_loop_filter_data_reset(
@@ -1651,6 +1597,6 @@
int vp10_loop_filter_worker(LFWorkerData *const lf_data, void *unused) {
(void)unused;
vp10_loop_filter_rows(lf_data->frame_buffer, lf_data->cm, lf_data->planes,
- lf_data->start, lf_data->stop, lf_data->y_only);
+ lf_data->start, lf_data->stop, lf_data->y_only);
return 1;
}
diff --git a/vp10/common/loopfilter.h b/vp10/common/loopfilter.h
index 51aae07..18aa7a1 100644
--- a/vp10/common/loopfilter.h
+++ b/vp10/common/loopfilter.h
@@ -26,7 +26,7 @@
#define SIMD_WIDTH 16
-#define MAX_MODE_LF_DELTAS 2
+#define MAX_MODE_LF_DELTAS 2
enum lf_path {
LF_PATH_420,
@@ -97,44 +97,39 @@
// This function sets up the bit masks for the entire 64x64 region represented
// by mi_row, mi_col.
-void vp10_setup_mask(struct VP10Common *const cm,
- const int mi_row, const int mi_col,
- MODE_INFO **mi_8x8, const int mode_info_stride,
- LOOP_FILTER_MASK *lfm);
+void vp10_setup_mask(struct VP10Common *const cm, const int mi_row,
+ const int mi_col, MODE_INFO **mi_8x8,
+ const int mode_info_stride, LOOP_FILTER_MASK *lfm);
void vp10_filter_block_plane_ss00(struct VP10Common *const cm,
- struct macroblockd_plane *const plane,
- int mi_row,
- LOOP_FILTER_MASK *lfm);
+ struct macroblockd_plane *const plane,
+ int mi_row, LOOP_FILTER_MASK *lfm);
void vp10_filter_block_plane_ss11(struct VP10Common *const cm,
- struct macroblockd_plane *const plane,
- int mi_row,
- LOOP_FILTER_MASK *lfm);
+ struct macroblockd_plane *const plane,
+ int mi_row, LOOP_FILTER_MASK *lfm);
void vp10_filter_block_plane_non420(struct VP10Common *cm,
- struct macroblockd_plane *plane,
- MODE_INFO **mi_8x8,
- int mi_row, int mi_col);
+ struct macroblockd_plane *plane,
+ MODE_INFO **mi_8x8, int mi_row, int mi_col);
void vp10_loop_filter_init(struct VP10Common *cm);
// Update the loop filter for the current frame.
-// This should be called before vp10_loop_filter_rows(), vp10_loop_filter_frame()
+// This should be called before vp10_loop_filter_rows(),
+// vp10_loop_filter_frame()
// calls this function directly.
void vp10_loop_filter_frame_init(struct VP10Common *cm, int default_filt_lvl);
-void vp10_loop_filter_frame(YV12_BUFFER_CONFIG *frame,
- struct VP10Common *cm,
- struct macroblockd *mbd,
- int filter_level,
- int y_only, int partial_frame);
+void vp10_loop_filter_frame(YV12_BUFFER_CONFIG *frame, struct VP10Common *cm,
+ struct macroblockd *mbd, int filter_level,
+ int y_only, int partial_frame);
// Apply the loop filter to [start, stop) macro block rows in frame_buffer.
void vp10_loop_filter_rows(YV12_BUFFER_CONFIG *frame_buffer,
- struct VP10Common *cm,
- struct macroblockd_plane planes[MAX_MB_PLANE],
- int start, int stop, int y_only);
+ struct VP10Common *cm,
+ struct macroblockd_plane planes[MAX_MB_PLANE],
+ int start, int stop, int y_only);
typedef struct LoopFilterWorkerData {
YV12_BUFFER_CONFIG *frame_buffer;
diff --git a/vp10/common/mips/dspr2/itrans16_dspr2.c b/vp10/common/mips/dspr2/itrans16_dspr2.c
index 3d1bd3d..0fcae87 100644
--- a/vp10/common/mips/dspr2/itrans16_dspr2.c
+++ b/vp10/common/mips/dspr2/itrans16_dspr2.c
@@ -21,27 +21,23 @@
#include "vpx_ports/mem.h"
#if HAVE_DSPR2
-void vp10_iht16x16_256_add_dspr2(const int16_t *input, uint8_t *dest,
- int pitch, int tx_type) {
+void vp10_iht16x16_256_add_dspr2(const int16_t *input, uint8_t *dest, int pitch,
+ int tx_type) {
int i, j;
- DECLARE_ALIGNED(32, int16_t, out[16 * 16]);
+ DECLARE_ALIGNED(32, int16_t, out[16 * 16]);
int16_t *outptr = out;
int16_t temp_out[16];
uint32_t pos = 45;
/* bit positon for extract from acc */
- __asm__ __volatile__ (
- "wrdsp %[pos], 1 \n\t"
- :
- : [pos] "r" (pos)
- );
+ __asm__ __volatile__("wrdsp %[pos], 1 \n\t" : : [pos] "r"(pos));
switch (tx_type) {
- case DCT_DCT: // DCT in both horizontal and vertical
+ case DCT_DCT: // DCT in both horizontal and vertical
idct16_rows_dspr2(input, outptr, 16);
idct16_cols_add_blk_dspr2(out, dest, pitch);
break;
- case ADST_DCT: // ADST in vertical, DCT in horizontal
+ case ADST_DCT: // ADST in vertical, DCT in horizontal
idct16_rows_dspr2(input, outptr, 16);
outptr = out;
@@ -50,13 +46,12 @@
iadst16_dspr2(outptr, temp_out);
for (j = 0; j < 16; ++j)
- dest[j * pitch + i] =
- clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
- + dest[j * pitch + i]);
+ dest[j * pitch + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6) +
+ dest[j * pitch + i]);
outptr += 16;
}
break;
- case DCT_ADST: // DCT in vertical, ADST in horizontal
+ case DCT_ADST: // DCT in vertical, ADST in horizontal
{
int16_t temp_in[16 * 16];
@@ -70,13 +65,11 @@
}
for (i = 0; i < 16; ++i)
- for (j = 0; j < 16; ++j)
- temp_in[j * 16 + i] = out[i * 16 + j];
+ for (j = 0; j < 16; ++j) temp_in[j * 16 + i] = out[i * 16 + j];
idct16_cols_add_blk_dspr2(temp_in, dest, pitch);
- }
- break;
- case ADST_ADST: // ADST in both directions
+ } break;
+ case ADST_ADST: // ADST in both directions
{
int16_t temp_in[16];
@@ -90,19 +83,14 @@
}
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
iadst16_dspr2(temp_in, temp_out);
for (j = 0; j < 16; ++j)
- dest[j * pitch + i] =
- clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
- + dest[j * pitch + i]);
+ dest[j * pitch + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6) +
+ dest[j * pitch + i]);
}
- }
- break;
- default:
- printf("vp10_short_iht16x16_add_dspr2 : Invalid tx_type\n");
- break;
+ } break;
+ default: printf("vp10_short_iht16x16_add_dspr2 : Invalid tx_type\n"); break;
}
}
#endif // #if HAVE_DSPR2
diff --git a/vp10/common/mips/dspr2/itrans4_dspr2.c b/vp10/common/mips/dspr2/itrans4_dspr2.c
index 5249287..9d10d5e 100644
--- a/vp10/common/mips/dspr2/itrans4_dspr2.c
+++ b/vp10/common/mips/dspr2/itrans4_dspr2.c
@@ -22,7 +22,7 @@
#if HAVE_DSPR2
void vp10_iht4x4_16_add_dspr2(const int16_t *input, uint8_t *dest,
- int dest_stride, int tx_type) {
+ int dest_stride, int tx_type) {
int i, j;
DECLARE_ALIGNED(32, int16_t, out[4 * 4]);
int16_t *outptr = out;
@@ -30,14 +30,12 @@
uint32_t pos = 45;
/* bit positon for extract from acc */
- __asm__ __volatile__ (
- "wrdsp %[pos], 1 \n\t"
- :
- : [pos] "r" (pos)
- );
+ __asm__ __volatile__("wrdsp %[pos], 1 \n\t"
+ :
+ : [pos] "r"(pos));
switch (tx_type) {
- case DCT_DCT: // DCT in both horizontal and vertical
+ case DCT_DCT: // DCT in both horizontal and vertical
vpx_idct4_rows_dspr2(input, outptr);
vpx_idct4_columns_add_blk_dspr2(&out[0], dest, dest_stride);
break;
@@ -50,9 +48,8 @@
iadst4_dspr2(outptr, temp_out);
for (j = 0; j < 4; ++j)
- dest[j * dest_stride + i] =
- clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 4)
- + dest[j * dest_stride + i]);
+ dest[j * dest_stride + i] = clip_pixel(
+ ROUND_POWER_OF_TWO(temp_out[j], 4) + dest[j * dest_stride + i]);
outptr += 4;
}
@@ -60,7 +57,7 @@
case DCT_ADST: // DCT in vertical, ADST in horizontal
for (i = 0; i < 4; ++i) {
iadst4_dspr2(input, outptr);
- input += 4;
+ input += 4;
outptr += 4;
}
@@ -74,24 +71,20 @@
case ADST_ADST: // ADST in both directions
for (i = 0; i < 4; ++i) {
iadst4_dspr2(input, outptr);
- input += 4;
+ input += 4;
outptr += 4;
}
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j * 4 + i];
+ for (j = 0; j < 4; ++j) temp_in[j] = out[j * 4 + i];
iadst4_dspr2(temp_in, temp_out);
for (j = 0; j < 4; ++j)
- dest[j * dest_stride + i] =
- clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 4)
- + dest[j * dest_stride + i]);
+ dest[j * dest_stride + i] = clip_pixel(
+ ROUND_POWER_OF_TWO(temp_out[j], 4) + dest[j * dest_stride + i]);
}
break;
- default:
- printf("vp10_short_iht4x4_add_dspr2 : Invalid tx_type\n");
- break;
+ default: printf("vp10_short_iht4x4_add_dspr2 : Invalid tx_type\n"); break;
}
}
#endif // #if HAVE_DSPR2
diff --git a/vp10/common/mips/dspr2/itrans8_dspr2.c b/vp10/common/mips/dspr2/itrans8_dspr2.c
index b25b93a..3ebf8ce 100644
--- a/vp10/common/mips/dspr2/itrans8_dspr2.c
+++ b/vp10/common/mips/dspr2/itrans8_dspr2.c
@@ -21,7 +21,7 @@
#if HAVE_DSPR2
void vp10_iht8x8_64_add_dspr2(const int16_t *input, uint8_t *dest,
- int dest_stride, int tx_type) {
+ int dest_stride, int tx_type) {
int i, j;
DECLARE_ALIGNED(32, int16_t, out[8 * 8]);
int16_t *outptr = out;
@@ -29,30 +29,25 @@
uint32_t pos = 45;
/* bit positon for extract from acc */
- __asm__ __volatile__ (
- "wrdsp %[pos], 1 \n\t"
- :
- : [pos] "r" (pos)
- );
+ __asm__ __volatile__("wrdsp %[pos], 1 \n\t" : : [pos] "r"(pos));
switch (tx_type) {
- case DCT_DCT: // DCT in both horizontal and vertical
+ case DCT_DCT: // DCT in both horizontal and vertical
idct8_rows_dspr2(input, outptr, 8);
idct8_columns_add_blk_dspr2(&out[0], dest, dest_stride);
break;
- case ADST_DCT: // ADST in vertical, DCT in horizontal
+ case ADST_DCT: // ADST in vertical, DCT in horizontal
idct8_rows_dspr2(input, outptr, 8);
for (i = 0; i < 8; ++i) {
iadst8_dspr2(&out[i * 8], temp_out);
for (j = 0; j < 8; ++j)
- dest[j * dest_stride + i] =
- clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 5)
- + dest[j * dest_stride + i]);
+ dest[j * dest_stride + i] = clip_pixel(
+ ROUND_POWER_OF_TWO(temp_out[j], 5) + dest[j * dest_stride + i]);
}
break;
- case DCT_ADST: // DCT in vertical, ADST in horizontal
+ case DCT_ADST: // DCT in vertical, ADST in horizontal
for (i = 0; i < 8; ++i) {
iadst8_dspr2(input, outptr);
input += 8;
@@ -66,7 +61,7 @@
}
idct8_columns_add_blk_dspr2(&temp_in[0], dest, dest_stride);
break;
- case ADST_ADST: // ADST in both directions
+ case ADST_ADST: // ADST in both directions
for (i = 0; i < 8; ++i) {
iadst8_dspr2(input, outptr);
input += 8;
@@ -74,20 +69,16 @@
}
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
iadst8_dspr2(temp_in, temp_out);
for (j = 0; j < 8; ++j)
- dest[j * dest_stride + i] =
- clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 5)
- + dest[j * dest_stride + i]);
+ dest[j * dest_stride + i] = clip_pixel(
+ ROUND_POWER_OF_TWO(temp_out[j], 5) + dest[j * dest_stride + i]);
}
break;
- default:
- printf("vp10_short_iht8x8_add_dspr2 : Invalid tx_type\n");
- break;
+ default: printf("vp10_short_iht8x8_add_dspr2 : Invalid tx_type\n"); break;
}
}
#endif // #if HAVE_DSPR2
diff --git a/vp10/common/mips/msa/idct16x16_msa.c b/vp10/common/mips/msa/idct16x16_msa.c
index a89e41b..c73ef37 100644
--- a/vp10/common/mips/msa/idct16x16_msa.c
+++ b/vp10/common/mips/msa/idct16x16_msa.c
@@ -14,7 +14,7 @@
#include "vpx_dsp/mips/inv_txfm_msa.h"
void vp10_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst,
- int32_t dst_stride, int32_t tx_type) {
+ int32_t dst_stride, int32_t tx_type) {
int32_t i;
DECLARE_ALIGNED(32, int16_t, out[16 * 16]);
int16_t *out_ptr = &out[0];
@@ -74,8 +74,6 @@
(dst + (i << 3)), dst_stride);
}
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
diff --git a/vp10/common/mips/msa/idct4x4_msa.c b/vp10/common/mips/msa/idct4x4_msa.c
index e38889f..ea4091b 100644
--- a/vp10/common/mips/msa/idct4x4_msa.c
+++ b/vp10/common/mips/msa/idct4x4_msa.c
@@ -14,7 +14,7 @@
#include "vpx_dsp/mips/inv_txfm_msa.h"
void vp10_iht4x4_16_add_msa(const int16_t *input, uint8_t *dst,
- int32_t dst_stride, int32_t tx_type) {
+ int32_t dst_stride, int32_t tx_type) {
v8i16 in0, in1, in2, in3;
/* load vector elements of 4x4 block */
@@ -50,9 +50,7 @@
TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
VPX_IADST4x4(in0, in1, in2, in3, in0, in1, in2, in3);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
/* final rounding (add 2^3, divide by 2^4) and shift */
diff --git a/vp10/common/mips/msa/idct8x8_msa.c b/vp10/common/mips/msa/idct8x8_msa.c
index ede6751..c62e82d 100644
--- a/vp10/common/mips/msa/idct8x8_msa.c
+++ b/vp10/common/mips/msa/idct8x8_msa.c
@@ -14,59 +14,57 @@
#include "vpx_dsp/mips/inv_txfm_msa.h"
void vp10_iht8x8_64_add_msa(const int16_t *input, uint8_t *dst,
- int32_t dst_stride, int32_t tx_type) {
+ int32_t dst_stride, int32_t tx_type) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
/* load vector elements of 8x8 block */
LD_SH8(input, 8, in0, in1, in2, in3, in4, in5, in6, in7);
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
+ in4, in5, in6, in7);
switch (tx_type) {
case DCT_DCT:
/* DCT in horizontal */
- VPX_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ VPX_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
+ in4, in5, in6, in7);
/* DCT in vertical */
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- VPX_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2,
+ in3, in4, in5, in6, in7);
+ VPX_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
+ in4, in5, in6, in7);
break;
case ADST_DCT:
/* DCT in horizontal */
- VPX_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ VPX_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
+ in4, in5, in6, in7);
/* ADST in vertical */
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2,
+ in3, in4, in5, in6, in7);
+ VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
break;
case DCT_ADST:
/* ADST in horizontal */
- VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
/* DCT in vertical */
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- VPX_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2,
+ in3, in4, in5, in6, in7);
+ VPX_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
+ in4, in5, in6, in7);
break;
case ADST_ADST:
/* ADST in horizontal */
- VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
/* ADST in vertical */
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2,
+ in3, in4, in5, in6, in7);
+ VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
/* final rounding (add 2^4, divide by 2^5) and shift */
diff --git a/vp10/common/mv.h b/vp10/common/mv.h
index b4971a5..96656aa 100644
--- a/vp10/common/mv.h
+++ b/vp10/common/mv.h
@@ -39,11 +39,11 @@
}
static INLINE int is_equal_mv(const MV *a, const MV *b) {
- return *((const uint32_t *)a) == *((const uint32_t *)b);
+ return *((const uint32_t *)a) == *((const uint32_t *)b);
}
-static INLINE void clamp_mv(MV *mv, int min_col, int max_col,
- int min_row, int max_row) {
+static INLINE void clamp_mv(MV *mv, int min_col, int max_col, int min_row,
+ int max_row) {
mv->col = clamp(mv->col, min_col, max_col);
mv->row = clamp(mv->row, min_row, max_row);
}
diff --git a/vp10/common/mvref_common.c b/vp10/common/mvref_common.c
index 1ef80c2..198bbf5 100644
--- a/vp10/common/mvref_common.c
+++ b/vp10/common/mvref_common.c
@@ -15,17 +15,18 @@
// to try and find candidate reference vectors.
static void find_mv_refs_idx(const VP10_COMMON *cm, const MACROBLOCKD *xd,
MODE_INFO *mi, MV_REFERENCE_FRAME ref_frame,
- int_mv *mv_ref_list,
- int block, int mi_row, int mi_col,
- find_mv_refs_sync sync, void *const data,
- uint8_t *mode_context) {
+ int_mv *mv_ref_list, int block, int mi_row,
+ int mi_col, find_mv_refs_sync sync,
+ void *const data, uint8_t *mode_context) {
const int *ref_sign_bias = cm->ref_frame_sign_bias;
int i, refmv_count = 0;
const POSITION *const mv_ref_search = mv_ref_blocks[mi->mbmi.sb_type];
int different_ref_found = 0;
int context_counter = 0;
- const MV_REF *const prev_frame_mvs = cm->use_prev_frame_mvs ?
- cm->prev_frame->mvs + mi_row * cm->mi_cols + mi_col : NULL;
+ const MV_REF *const prev_frame_mvs =
+ cm->use_prev_frame_mvs
+ ? cm->prev_frame->mvs + mi_row * cm->mi_cols + mi_col
+ : NULL;
const TileInfo *const tile = &xd->tile;
const int bw = num_8x8_blocks_wide_lookup[mi->mbmi.sb_type] << 3;
const int bh = num_8x8_blocks_high_lookup[mi->mbmi.sb_type] << 3;
@@ -41,8 +42,8 @@
for (i = 0; i < 2; ++i) {
const POSITION *const mv_ref = &mv_ref_search[i];
if (is_inside(tile, mi_col, mi_row, cm->mi_rows, mv_ref)) {
- const MODE_INFO *const candidate_mi = xd->mi[mv_ref->col + mv_ref->row *
- xd->mi_stride];
+ const MODE_INFO *const candidate_mi =
+ xd->mi[mv_ref->col + mv_ref->row * xd->mi_stride];
const MB_MODE_INFO *const candidate = &candidate_mi->mbmi;
// Keep counts for entropy encoding.
context_counter += mode_2_counter[candidate->mode];
@@ -63,27 +64,27 @@
for (; i < MVREF_NEIGHBOURS; ++i) {
const POSITION *const mv_ref = &mv_ref_search[i];
if (is_inside(tile, mi_col, mi_row, cm->mi_rows, mv_ref)) {
- const MB_MODE_INFO *const candidate = &xd->mi[mv_ref->col + mv_ref->row *
- xd->mi_stride]->mbmi;
+ const MB_MODE_INFO *const candidate =
+ &xd->mi[mv_ref->col + mv_ref->row * xd->mi_stride]->mbmi;
different_ref_found = 1;
if (candidate->ref_frame[0] == ref_frame)
- ADD_MV_REF_LIST(candidate->mv[0], refmv_count, mv_ref_list,
- bw, bh, xd, Done);
+ ADD_MV_REF_LIST(candidate->mv[0], refmv_count, mv_ref_list, bw, bh, xd,
+ Done);
else if (candidate->ref_frame[1] == ref_frame)
- ADD_MV_REF_LIST(candidate->mv[1], refmv_count, mv_ref_list,
- bw, bh, xd, Done);
+ ADD_MV_REF_LIST(candidate->mv[1], refmv_count, mv_ref_list, bw, bh, xd,
+ Done);
}
}
- // TODO(hkuang): Remove this sync after fixing pthread_cond_broadcast
- // on windows platform. The sync here is unncessary if use_perv_frame_mvs
- // is 0. But after removing it, there will be hang in the unit test on windows
- // due to several threads waiting for a thread's signal.
+// TODO(hkuang): Remove this sync after fixing pthread_cond_broadcast
+// on windows platform. The sync here is unncessary if use_perv_frame_mvs
+// is 0. But after removing it, there will be hang in the unit test on windows
+// due to several threads waiting for a thread's signal.
#if defined(_WIN32) && !HAVE_PTHREAD_H
- if (cm->frame_parallel_decode && sync != NULL) {
- sync(data, mi_row);
- }
+ if (cm->frame_parallel_decode && sync != NULL) {
+ sync(data, mi_row);
+ }
#endif
// Check the last frame's mode and mv info.
@@ -94,11 +95,11 @@
}
if (prev_frame_mvs->ref_frame[0] == ref_frame) {
- ADD_MV_REF_LIST(prev_frame_mvs->mv[0], refmv_count, mv_ref_list,
- bw, bh, xd, Done);
+ ADD_MV_REF_LIST(prev_frame_mvs->mv[0], refmv_count, mv_ref_list, bw, bh,
+ xd, Done);
} else if (prev_frame_mvs->ref_frame[1] == ref_frame) {
- ADD_MV_REF_LIST(prev_frame_mvs->mv[1], refmv_count, mv_ref_list,
- bw, bh, xd, Done);
+ ADD_MV_REF_LIST(prev_frame_mvs->mv[1], refmv_count, mv_ref_list, bw, bh,
+ xd, Done);
}
}
@@ -109,8 +110,8 @@
for (i = 0; i < MVREF_NEIGHBOURS; ++i) {
const POSITION *mv_ref = &mv_ref_search[i];
if (is_inside(tile, mi_col, mi_row, cm->mi_rows, mv_ref)) {
- const MB_MODE_INFO *const candidate = &xd->mi[mv_ref->col + mv_ref->row
- * xd->mi_stride]->mbmi;
+ const MB_MODE_INFO *const candidate =
+ &xd->mi[mv_ref->col + mv_ref->row * xd->mi_stride]->mbmi;
// If the candidate is INTRA we don't want to consider its mv.
IF_DIFF_REF_FRAME_ADD_MV(candidate, ref_frame, ref_sign_bias,
@@ -147,13 +148,13 @@
}
}
- Done:
+Done:
mode_context[ref_frame] = counter_to_context[context_counter];
#if CONFIG_MISC_FIXES
for (i = refmv_count; i < MAX_MV_REF_CANDIDATES; ++i)
- mv_ref_list[i].as_int = 0;
+ mv_ref_list[i].as_int = 0;
#else
// Clamp vectors
for (i = 0; i < MAX_MV_REF_CANDIDATES; ++i)
@@ -162,28 +163,24 @@
}
void vp10_find_mv_refs(const VP10_COMMON *cm, const MACROBLOCKD *xd,
- MODE_INFO *mi, MV_REFERENCE_FRAME ref_frame,
- int_mv *mv_ref_list,
- int mi_row, int mi_col,
- find_mv_refs_sync sync, void *const data,
- uint8_t *mode_context) {
- find_mv_refs_idx(cm, xd, mi, ref_frame, mv_ref_list, -1,
- mi_row, mi_col, sync, data, mode_context);
+ MODE_INFO *mi, MV_REFERENCE_FRAME ref_frame,
+ int_mv *mv_ref_list, int mi_row, int mi_col,
+ find_mv_refs_sync sync, void *const data,
+ uint8_t *mode_context) {
+ find_mv_refs_idx(cm, xd, mi, ref_frame, mv_ref_list, -1, mi_row, mi_col, sync,
+ data, mode_context);
}
static void lower_mv_precision(MV *mv, int allow_hp) {
const int use_hp = allow_hp && vp10_use_mv_hp(mv);
if (!use_hp) {
- if (mv->row & 1)
- mv->row += (mv->row > 0 ? -1 : 1);
- if (mv->col & 1)
- mv->col += (mv->col > 0 ? -1 : 1);
+ if (mv->row & 1) mv->row += (mv->row > 0 ? -1 : 1);
+ if (mv->col & 1) mv->col += (mv->col > 0 ? -1 : 1);
}
}
-void vp10_find_best_ref_mvs(int allow_hp,
- int_mv *mvlist, int_mv *nearest_mv,
- int_mv *near_mv) {
+void vp10_find_best_ref_mvs(int allow_hp, int_mv *mvlist, int_mv *nearest_mv,
+ int_mv *near_mv) {
int i;
// Make sure all the candidates are properly clamped etc
for (i = 0; i < MAX_MV_REF_CANDIDATES; ++i) {
@@ -193,10 +190,10 @@
*near_mv = mvlist[1];
}
-void vp10_append_sub8x8_mvs_for_idx(VP10_COMMON *cm, MACROBLOCKD *xd,
- int block, int ref, int mi_row, int mi_col,
- int_mv *nearest_mv, int_mv *near_mv,
- uint8_t *mode_context) {
+void vp10_append_sub8x8_mvs_for_idx(VP10_COMMON *cm, MACROBLOCKD *xd, int block,
+ int ref, int mi_row, int mi_col,
+ int_mv *nearest_mv, int_mv *near_mv,
+ uint8_t *mode_context) {
int_mv mv_list[MAX_MV_REF_CANDIDATES];
MODE_INFO *const mi = xd->mi[0];
b_mode_info *bmi = mi->bmi;
@@ -204,8 +201,8 @@
assert(MAX_MV_REF_CANDIDATES == 2);
- find_mv_refs_idx(cm, xd, mi, mi->mbmi.ref_frame[ref], mv_list, block,
- mi_row, mi_col, NULL, NULL, mode_context);
+ find_mv_refs_idx(cm, xd, mi, mi->mbmi.ref_frame[ref], mv_list, block, mi_row,
+ mi_col, NULL, NULL, mode_context);
near_mv->as_int = 0;
switch (block) {
@@ -237,7 +234,6 @@
}
break;
}
- default:
- assert(0 && "Invalid block index.");
+ default: assert(0 && "Invalid block index.");
}
}
diff --git a/vp10/common/mvref_common.h b/vp10/common/mvref_common.h
index 0a98866..44e4bb5 100644
--- a/vp10/common/mvref_common.h
+++ b/vp10/common/mvref_common.h
@@ -61,61 +61,149 @@
// 2. However the actual count can never be greater than 2 so the highest
// counter we need is 18. 9 is an invalid counter that's never used.
static const int counter_to_context[19] = {
- BOTH_PREDICTED, // 0
- NEW_PLUS_NON_INTRA, // 1
- BOTH_NEW, // 2
- ZERO_PLUS_PREDICTED, // 3
- NEW_PLUS_NON_INTRA, // 4
- INVALID_CASE, // 5
- BOTH_ZERO, // 6
- INVALID_CASE, // 7
- INVALID_CASE, // 8
+ BOTH_PREDICTED, // 0
+ NEW_PLUS_NON_INTRA, // 1
+ BOTH_NEW, // 2
+ ZERO_PLUS_PREDICTED, // 3
+ NEW_PLUS_NON_INTRA, // 4
+ INVALID_CASE, // 5
+ BOTH_ZERO, // 6
+ INVALID_CASE, // 7
+ INVALID_CASE, // 8
INTRA_PLUS_NON_INTRA, // 9
INTRA_PLUS_NON_INTRA, // 10
- INVALID_CASE, // 11
+ INVALID_CASE, // 11
INTRA_PLUS_NON_INTRA, // 12
- INVALID_CASE, // 13
- INVALID_CASE, // 14
- INVALID_CASE, // 15
- INVALID_CASE, // 16
- INVALID_CASE, // 17
- BOTH_INTRA // 18
+ INVALID_CASE, // 13
+ INVALID_CASE, // 14
+ INVALID_CASE, // 15
+ INVALID_CASE, // 16
+ INVALID_CASE, // 17
+ BOTH_INTRA // 18
};
static const POSITION mv_ref_blocks[BLOCK_SIZES][MVREF_NEIGHBOURS] = {
// 4X4
- {{-1, 0}, {0, -1}, {-1, -1}, {-2, 0}, {0, -2}, {-2, -1}, {-1, -2}, {-2, -2}},
+ { { -1, 0 },
+ { 0, -1 },
+ { -1, -1 },
+ { -2, 0 },
+ { 0, -2 },
+ { -2, -1 },
+ { -1, -2 },
+ { -2, -2 } },
// 4X8
- {{-1, 0}, {0, -1}, {-1, -1}, {-2, 0}, {0, -2}, {-2, -1}, {-1, -2}, {-2, -2}},
+ { { -1, 0 },
+ { 0, -1 },
+ { -1, -1 },
+ { -2, 0 },
+ { 0, -2 },
+ { -2, -1 },
+ { -1, -2 },
+ { -2, -2 } },
// 8X4
- {{-1, 0}, {0, -1}, {-1, -1}, {-2, 0}, {0, -2}, {-2, -1}, {-1, -2}, {-2, -2}},
+ { { -1, 0 },
+ { 0, -1 },
+ { -1, -1 },
+ { -2, 0 },
+ { 0, -2 },
+ { -2, -1 },
+ { -1, -2 },
+ { -2, -2 } },
// 8X8
- {{-1, 0}, {0, -1}, {-1, -1}, {-2, 0}, {0, -2}, {-2, -1}, {-1, -2}, {-2, -2}},
+ { { -1, 0 },
+ { 0, -1 },
+ { -1, -1 },
+ { -2, 0 },
+ { 0, -2 },
+ { -2, -1 },
+ { -1, -2 },
+ { -2, -2 } },
// 8X16
- {{0, -1}, {-1, 0}, {1, -1}, {-1, -1}, {0, -2}, {-2, 0}, {-2, -1}, {-1, -2}},
+ { { 0, -1 },
+ { -1, 0 },
+ { 1, -1 },
+ { -1, -1 },
+ { 0, -2 },
+ { -2, 0 },
+ { -2, -1 },
+ { -1, -2 } },
// 16X8
- {{-1, 0}, {0, -1}, {-1, 1}, {-1, -1}, {-2, 0}, {0, -2}, {-1, -2}, {-2, -1}},
+ { { -1, 0 },
+ { 0, -1 },
+ { -1, 1 },
+ { -1, -1 },
+ { -2, 0 },
+ { 0, -2 },
+ { -1, -2 },
+ { -2, -1 } },
// 16X16
- {{-1, 0}, {0, -1}, {-1, 1}, {1, -1}, {-1, -1}, {-3, 0}, {0, -3}, {-3, -3}},
+ { { -1, 0 },
+ { 0, -1 },
+ { -1, 1 },
+ { 1, -1 },
+ { -1, -1 },
+ { -3, 0 },
+ { 0, -3 },
+ { -3, -3 } },
// 16X32
- {{0, -1}, {-1, 0}, {2, -1}, {-1, -1}, {-1, 1}, {0, -3}, {-3, 0}, {-3, -3}},
+ { { 0, -1 },
+ { -1, 0 },
+ { 2, -1 },
+ { -1, -1 },
+ { -1, 1 },
+ { 0, -3 },
+ { -3, 0 },
+ { -3, -3 } },
// 32X16
- {{-1, 0}, {0, -1}, {-1, 2}, {-1, -1}, {1, -1}, {-3, 0}, {0, -3}, {-3, -3}},
+ { { -1, 0 },
+ { 0, -1 },
+ { -1, 2 },
+ { -1, -1 },
+ { 1, -1 },
+ { -3, 0 },
+ { 0, -3 },
+ { -3, -3 } },
// 32X32
- {{-1, 1}, {1, -1}, {-1, 2}, {2, -1}, {-1, -1}, {-3, 0}, {0, -3}, {-3, -3}},
+ { { -1, 1 },
+ { 1, -1 },
+ { -1, 2 },
+ { 2, -1 },
+ { -1, -1 },
+ { -3, 0 },
+ { 0, -3 },
+ { -3, -3 } },
// 32X64
- {{0, -1}, {-1, 0}, {4, -1}, {-1, 2}, {-1, -1}, {0, -3}, {-3, 0}, {2, -1}},
+ { { 0, -1 },
+ { -1, 0 },
+ { 4, -1 },
+ { -1, 2 },
+ { -1, -1 },
+ { 0, -3 },
+ { -3, 0 },
+ { 2, -1 } },
// 64X32
- {{-1, 0}, {0, -1}, {-1, 4}, {2, -1}, {-1, -1}, {-3, 0}, {0, -3}, {-1, 2}},
+ { { -1, 0 },
+ { 0, -1 },
+ { -1, 4 },
+ { 2, -1 },
+ { -1, -1 },
+ { -3, 0 },
+ { 0, -3 },
+ { -1, 2 } },
// 64X64
- {{-1, 3}, {3, -1}, {-1, 4}, {4, -1}, {-1, -1}, {-1, 0}, {0, -1}, {-1, 6}}
+ { { -1, 3 },
+ { 3, -1 },
+ { -1, 4 },
+ { 4, -1 },
+ { -1, -1 },
+ { -1, 0 },
+ { 0, -1 },
+ { -1, 6 } }
};
static const int idx_n_column_to_subblock[4][2] = {
- {1, 2},
- {1, 3},
- {3, 2},
- {3, 3}
+ { 1, 2 }, { 1, 3 }, { 3, 2 }, { 3, 3 }
};
// clamp_mv_ref
@@ -128,16 +216,15 @@
static INLINE void clamp_mv_ref(MV *mv, int bw, int bh, const MACROBLOCKD *xd) {
#if CONFIG_MISC_FIXES
clamp_mv(mv, xd->mb_to_left_edge - bw * 8 - MV_BORDER,
- xd->mb_to_right_edge + bw * 8 + MV_BORDER,
- xd->mb_to_top_edge - bh * 8 - MV_BORDER,
- xd->mb_to_bottom_edge + bh * 8 + MV_BORDER);
+ xd->mb_to_right_edge + bw * 8 + MV_BORDER,
+ xd->mb_to_top_edge - bh * 8 - MV_BORDER,
+ xd->mb_to_bottom_edge + bh * 8 + MV_BORDER);
#else
- (void) bw;
- (void) bh;
+ (void)bw;
+ (void)bh;
clamp_mv(mv, xd->mb_to_left_edge - MV_BORDER,
- xd->mb_to_right_edge + MV_BORDER,
- xd->mb_to_top_edge - MV_BORDER,
- xd->mb_to_bottom_edge + MV_BORDER);
+ xd->mb_to_right_edge + MV_BORDER, xd->mb_to_top_edge - MV_BORDER,
+ xd->mb_to_bottom_edge + MV_BORDER);
#endif
}
@@ -146,12 +233,12 @@
static INLINE int_mv get_sub_block_mv(const MODE_INFO *candidate, int which_mv,
int search_col, int block_idx) {
return block_idx >= 0 && candidate->mbmi.sb_type < BLOCK_8X8
- ? candidate->bmi[idx_n_column_to_subblock[block_idx][search_col == 0]]
- .as_mv[which_mv]
- : candidate->mbmi.mv[which_mv];
+ ? candidate
+ ->bmi[idx_n_column_to_subblock[block_idx][search_col == 0]]
+ .as_mv[which_mv]
+ : candidate->mbmi.mv[which_mv];
}
-
// Performs mv sign inversion if indicated by the reference frame combination.
static INLINE int_mv scale_mv(const MB_MODE_INFO *mbmi, int ref,
const MV_REFERENCE_FRAME this_ref_frame,
@@ -167,47 +254,47 @@
#if CONFIG_MISC_FIXES
#define CLIP_IN_ADD(mv, bw, bh, xd) clamp_mv_ref(mv, bw, bh, xd)
#else
-#define CLIP_IN_ADD(mv, bw, bh, xd) do {} while (0)
+#define CLIP_IN_ADD(mv, bw, bh, xd) \
+ do { \
+ } while (0)
#endif
// This macro is used to add a motion vector mv_ref list if it isn't
// already in the list. If it's the second motion vector it will also
// skip all additional processing and jump to done!
-#define ADD_MV_REF_LIST(mv, refmv_count, mv_ref_list, bw, bh, xd, Done) \
- do { \
- (mv_ref_list)[(refmv_count)] = (mv); \
- CLIP_IN_ADD(&(mv_ref_list)[(refmv_count)].as_mv, (bw), (bh), (xd)); \
+#define ADD_MV_REF_LIST(mv, refmv_count, mv_ref_list, bw, bh, xd, Done) \
+ do { \
+ (mv_ref_list)[(refmv_count)] = (mv); \
+ CLIP_IN_ADD(&(mv_ref_list)[(refmv_count)].as_mv, (bw), (bh), (xd)); \
if (refmv_count && (mv_ref_list)[1].as_int != (mv_ref_list)[0].as_int) { \
- (refmv_count) = 2; \
- goto Done; \
- } \
- (refmv_count) = 1; \
+ (refmv_count) = 2; \
+ goto Done; \
+ } \
+ (refmv_count) = 1; \
} while (0)
// If either reference frame is different, not INTRA, and they
// are different from each other scale and add the mv to our list.
#define IF_DIFF_REF_FRAME_ADD_MV(mbmi, ref_frame, ref_sign_bias, refmv_count, \
- mv_ref_list, bw, bh, xd, Done) \
- do { \
- if (is_inter_block(mbmi)) { \
- if ((mbmi)->ref_frame[0] != ref_frame) \
- ADD_MV_REF_LIST(scale_mv((mbmi), 0, ref_frame, ref_sign_bias), \
- refmv_count, mv_ref_list, bw, bh, xd, Done); \
- if (has_second_ref(mbmi) && \
- (CONFIG_MISC_FIXES || \
- (mbmi)->mv[1].as_int != (mbmi)->mv[0].as_int) && \
- (mbmi)->ref_frame[1] != ref_frame) \
- ADD_MV_REF_LIST(scale_mv((mbmi), 1, ref_frame, ref_sign_bias), \
- refmv_count, mv_ref_list, bw, bh, xd, Done); \
- } \
+ mv_ref_list, bw, bh, xd, Done) \
+ do { \
+ if (is_inter_block(mbmi)) { \
+ if ((mbmi)->ref_frame[0] != ref_frame) \
+ ADD_MV_REF_LIST(scale_mv((mbmi), 0, ref_frame, ref_sign_bias), \
+ refmv_count, mv_ref_list, bw, bh, xd, Done); \
+ if (has_second_ref(mbmi) && \
+ (CONFIG_MISC_FIXES || \
+ (mbmi)->mv[1].as_int != (mbmi)->mv[0].as_int) && \
+ (mbmi)->ref_frame[1] != ref_frame) \
+ ADD_MV_REF_LIST(scale_mv((mbmi), 1, ref_frame, ref_sign_bias), \
+ refmv_count, mv_ref_list, bw, bh, xd, Done); \
+ } \
} while (0)
-
// Checks that the given mi_row, mi_col and search point
// are inside the borders of the tile.
-static INLINE int is_inside(const TileInfo *const tile,
- int mi_col, int mi_row, int mi_rows,
- const POSITION *mi_pos) {
+static INLINE int is_inside(const TileInfo *const tile, int mi_col, int mi_row,
+ int mi_rows, const POSITION *mi_pos) {
return !(mi_row + mi_pos->row < 0 ||
mi_col + mi_pos->col < tile->mi_col_start ||
mi_row + mi_pos->row >= mi_rows ||
@@ -216,21 +303,21 @@
typedef void (*find_mv_refs_sync)(void *const data, int mi_row);
void vp10_find_mv_refs(const VP10_COMMON *cm, const MACROBLOCKD *xd,
- MODE_INFO *mi, MV_REFERENCE_FRAME ref_frame,
- int_mv *mv_ref_list, int mi_row, int mi_col,
- find_mv_refs_sync sync, void *const data,
- uint8_t *mode_context);
+ MODE_INFO *mi, MV_REFERENCE_FRAME ref_frame,
+ int_mv *mv_ref_list, int mi_row, int mi_col,
+ find_mv_refs_sync sync, void *const data,
+ uint8_t *mode_context);
// check a list of motion vectors by sad score using a number rows of pixels
// above and a number cols of pixels in the left to select the one with best
// score to use as ref motion vector
-void vp10_find_best_ref_mvs(int allow_hp,
- int_mv *mvlist, int_mv *nearest_mv, int_mv *near_mv);
+void vp10_find_best_ref_mvs(int allow_hp, int_mv *mvlist, int_mv *nearest_mv,
+ int_mv *near_mv);
-void vp10_append_sub8x8_mvs_for_idx(VP10_COMMON *cm, MACROBLOCKD *xd,
- int block, int ref, int mi_row, int mi_col,
- int_mv *nearest_mv, int_mv *near_mv,
- uint8_t *mode_context);
+void vp10_append_sub8x8_mvs_for_idx(VP10_COMMON *cm, MACROBLOCKD *xd, int block,
+ int ref, int mi_row, int mi_col,
+ int_mv *nearest_mv, int_mv *near_mv,
+ uint8_t *mode_context);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/common/onyxc_int.h b/vp10/common/onyxc_int.h
index 067b261..df31696 100644
--- a/vp10/common/onyxc_int.h
+++ b/vp10/common/onyxc_int.h
@@ -47,10 +47,10 @@
#define NUM_PING_PONG_BUFFERS 2
typedef enum {
- SINGLE_REFERENCE = 0,
- COMPOUND_REFERENCE = 1,
+ SINGLE_REFERENCE = 0,
+ COMPOUND_REFERENCE = 1,
REFERENCE_MODE_SELECT = 2,
- REFERENCE_MODES = 3,
+ REFERENCE_MODES = 3,
} REFERENCE_MODE;
typedef enum {
@@ -103,9 +103,9 @@
} RefCntBuffer;
typedef struct BufferPool {
- // Protect BufferPool from being accessed by several FrameWorkers at
- // the same time during frame parallel decode.
- // TODO(hkuang): Try to use atomic variable instead of locking the whole pool.
+// Protect BufferPool from being accessed by several FrameWorkers at
+// the same time during frame parallel decode.
+// TODO(hkuang): Try to use atomic variable instead of locking the whole pool.
#if CONFIG_MULTITHREAD
pthread_mutex_t pool_mutex;
#endif
@@ -123,7 +123,7 @@
} BufferPool;
typedef struct VP10Common {
- struct vpx_internal_error_info error;
+ struct vpx_internal_error_info error;
vpx_color_space_t color_space;
int color_range;
int width;
@@ -163,7 +163,7 @@
int new_fb_idx;
- FRAME_TYPE last_frame_type; /* last frame's frame type for motion search.*/
+ FRAME_TYPE last_frame_type; /* last frame's frame type for motion search.*/
FRAME_TYPE frame_type;
int show_frame;
@@ -239,7 +239,7 @@
// a frame decode
REFRESH_FRAME_CONTEXT_MODE refresh_frame_context;
- int ref_frame_sign_bias[MAX_REF_FRAMES]; /* Two state 0, 1 */
+ int ref_frame_sign_bias[MAX_REF_FRAMES]; /* Two state 0, 1 */
struct loopfilter lf;
struct segmentation seg;
@@ -254,9 +254,9 @@
MV_REFERENCE_FRAME comp_var_ref[2];
REFERENCE_MODE reference_mode;
- FRAME_CONTEXT *fc; /* this frame entropy */
- FRAME_CONTEXT *frame_contexts; // FRAME_CONTEXTS
- unsigned int frame_context_idx; /* Context to use/update */
+ FRAME_CONTEXT *fc; /* this frame entropy */
+ FRAME_CONTEXT *frame_contexts; // FRAME_CONTEXTS
+ unsigned int frame_context_idx; /* Context to use/update */
FRAME_COUNTS counts;
unsigned int current_video_frame;
@@ -313,10 +313,8 @@
}
static INLINE YV12_BUFFER_CONFIG *get_ref_frame(VP10_COMMON *cm, int index) {
- if (index < 0 || index >= REF_FRAMES)
- return NULL;
- if (cm->ref_frame_map[index] < 0)
- return NULL;
+ if (index < 0 || index >= REF_FRAMES) return NULL;
+ if (cm->ref_frame_map[index] < 0) return NULL;
assert(cm->ref_frame_map[index] < FRAME_BUFFERS);
return &cm->buffer_pool->frame_bufs[cm->ref_frame_map[index]].buf;
}
@@ -331,8 +329,7 @@
lock_buffer_pool(cm->buffer_pool);
for (i = 0; i < FRAME_BUFFERS; ++i)
- if (frame_bufs[i].ref_count == 0)
- break;
+ if (frame_bufs[i].ref_count == 0) break;
if (i != FRAME_BUFFERS) {
frame_bufs[i].ref_count = 1;
@@ -365,12 +362,13 @@
}
static INLINE void vp10_init_macroblockd(VP10_COMMON *cm, MACROBLOCKD *xd,
- tran_low_t *dqcoeff) {
+ tran_low_t *dqcoeff) {
int i;
for (i = 0; i < MAX_MB_PLANE; ++i) {
xd->plane[i].dqcoeff = dqcoeff;
- xd->above_context[i] = cm->above_context +
+ xd->above_context[i] =
+ cm->above_context +
i * sizeof(*cm->above_context) * 2 * mi_cols_aligned_to_sb(cm->mi_cols);
if (xd->plane[i].plane_type == PLANE_TYPE_Y) {
@@ -403,17 +401,16 @@
}
static INLINE void set_mi_row_col(MACROBLOCKD *xd, const TileInfo *const tile,
- int mi_row, int bh,
- int mi_col, int bw,
+ int mi_row, int bh, int mi_col, int bw,
int mi_rows, int mi_cols) {
- xd->mb_to_top_edge = -((mi_row * MI_SIZE) * 8);
+ xd->mb_to_top_edge = -((mi_row * MI_SIZE) * 8);
xd->mb_to_bottom_edge = ((mi_rows - bh - mi_row) * MI_SIZE) * 8;
- xd->mb_to_left_edge = -((mi_col * MI_SIZE) * 8);
- xd->mb_to_right_edge = ((mi_cols - bw - mi_col) * MI_SIZE) * 8;
+ xd->mb_to_left_edge = -((mi_col * MI_SIZE) * 8);
+ xd->mb_to_right_edge = ((mi_cols - bw - mi_col) * MI_SIZE) * 8;
// Are edges available for intra prediction?
- xd->up_available = (mi_row != 0);
- xd->left_available = (mi_col > tile->mi_col_start);
+ xd->up_available = (mi_row != 0);
+ xd->left_available = (mi_col > tile->mi_col_start);
if (xd->up_available) {
xd->above_mi = xd->mi[-xd->mi_stride];
// above_mi may be NULL in encoder's first pass.
@@ -443,9 +440,8 @@
return cm->kf_y_prob[above][left];
}
-static INLINE void update_partition_context(MACROBLOCKD *xd,
- int mi_row, int mi_col,
- BLOCK_SIZE subsize,
+static INLINE void update_partition_context(MACROBLOCKD *xd, int mi_row,
+ int mi_col, BLOCK_SIZE subsize,
BLOCK_SIZE bsize) {
PARTITION_CONTEXT *const above_ctx = xd->above_seg_context + mi_col;
PARTITION_CONTEXT *const left_ctx = xd->left_seg_context + (mi_row & MI_MASK);
@@ -460,13 +456,12 @@
memset(left_ctx, partition_context_lookup[subsize].left, bs);
}
-static INLINE int partition_plane_context(const MACROBLOCKD *xd,
- int mi_row, int mi_col,
- BLOCK_SIZE bsize) {
+static INLINE int partition_plane_context(const MACROBLOCKD *xd, int mi_row,
+ int mi_col, BLOCK_SIZE bsize) {
const PARTITION_CONTEXT *above_ctx = xd->above_seg_context + mi_col;
const PARTITION_CONTEXT *left_ctx = xd->left_seg_context + (mi_row & MI_MASK);
const int bsl = mi_width_log2_lookup[bsize];
- int above = (*above_ctx >> bsl) & 1 , left = (*left_ctx >> bsl) & 1;
+ int above = (*above_ctx >> bsl) & 1, left = (*left_ctx >> bsl) & 1;
assert(b_width_log2_lookup[bsize] == b_height_log2_lookup[bsize]);
assert(bsl >= 0);
diff --git a/vp10/common/pred_common.c b/vp10/common/pred_common.c
index 2e79e0d..b3a4531 100644
--- a/vp10/common/pred_common.c
+++ b/vp10/common/pred_common.c
@@ -20,11 +20,13 @@
// left of the entries correpsonding to real macroblocks.
// The prediction flags in these dummy entries are initialised to 0.
const MB_MODE_INFO *const left_mbmi = xd->left_mbmi;
- const int left_type = xd->left_available && is_inter_block(left_mbmi) ?
- left_mbmi->interp_filter : SWITCHABLE_FILTERS;
+ const int left_type = xd->left_available && is_inter_block(left_mbmi)
+ ? left_mbmi->interp_filter
+ : SWITCHABLE_FILTERS;
const MB_MODE_INFO *const above_mbmi = xd->above_mbmi;
- const int above_type = xd->up_available && is_inter_block(above_mbmi) ?
- above_mbmi->interp_filter : SWITCHABLE_FILTERS;
+ const int above_type = xd->up_available && is_inter_block(above_mbmi)
+ ? above_mbmi->interp_filter
+ : SWITCHABLE_FILTERS;
if (left_type == above_type)
return left_type;
@@ -52,8 +54,7 @@
if (has_above && has_left) { // both edges available
const int above_intra = !is_inter_block(above_mbmi);
const int left_intra = !is_inter_block(left_mbmi);
- return left_intra && above_intra ? 3
- : left_intra || above_intra;
+ return left_intra && above_intra ? 3 : left_intra || above_intra;
} else if (has_above || has_left) { // one edge available
return 2 * !is_inter_block(has_above ? above_mbmi : left_mbmi);
} else {
@@ -62,7 +63,7 @@
}
int vp10_get_reference_mode_context(const VP10_COMMON *cm,
- const MACROBLOCKD *xd) {
+ const MACROBLOCKD *xd) {
int ctx;
const MB_MODE_INFO *const above_mbmi = xd->above_mbmi;
const MB_MODE_INFO *const left_mbmi = xd->left_mbmi;
@@ -105,7 +106,7 @@
// Returns a context number for the given MB prediction signal
int vp10_get_pred_context_comp_ref_p(const VP10_COMMON *cm,
- const MACROBLOCKD *xd) {
+ const MACROBLOCKD *xd) {
int pred_context;
const MB_MODE_INFO *const above_mbmi = xd->above_mbmi;
const MB_MODE_INFO *const left_mbmi = xd->left_mbmi;
@@ -131,15 +132,15 @@
if (!has_second_ref(edge_mbmi)) // single pred (1/3)
pred_context = 1 + 2 * (edge_mbmi->ref_frame[0] != cm->comp_var_ref[1]);
else // comp pred (1/3)
- pred_context = 1 + 2 * (edge_mbmi->ref_frame[var_ref_idx]
- != cm->comp_var_ref[1]);
+ pred_context =
+ 1 + 2 * (edge_mbmi->ref_frame[var_ref_idx] != cm->comp_var_ref[1]);
} else { // inter/inter
const int l_sg = !has_second_ref(left_mbmi);
const int a_sg = !has_second_ref(above_mbmi);
- const MV_REFERENCE_FRAME vrfa = a_sg ? above_mbmi->ref_frame[0]
- : above_mbmi->ref_frame[var_ref_idx];
- const MV_REFERENCE_FRAME vrfl = l_sg ? left_mbmi->ref_frame[0]
- : left_mbmi->ref_frame[var_ref_idx];
+ const MV_REFERENCE_FRAME vrfa =
+ a_sg ? above_mbmi->ref_frame[0] : above_mbmi->ref_frame[var_ref_idx];
+ const MV_REFERENCE_FRAME vrfl =
+ l_sg ? left_mbmi->ref_frame[0] : left_mbmi->ref_frame[var_ref_idx];
if (vrfa == vrfl && cm->comp_var_ref[1] == vrfa) {
pred_context = 0;
@@ -173,8 +174,8 @@
pred_context = 2;
} else {
if (has_second_ref(edge_mbmi))
- pred_context = 4 * (edge_mbmi->ref_frame[var_ref_idx]
- != cm->comp_var_ref[1]);
+ pred_context =
+ 4 * (edge_mbmi->ref_frame[var_ref_idx] != cm->comp_var_ref[1]);
else
pred_context = 3 * (edge_mbmi->ref_frame[0] != cm->comp_var_ref[1]);
}
@@ -277,8 +278,9 @@
else
pred_context = 4 * (edge_mbmi->ref_frame[0] == GOLDEN_FRAME);
} else {
- pred_context = 1 + 2 * (edge_mbmi->ref_frame[0] == GOLDEN_FRAME ||
- edge_mbmi->ref_frame[1] == GOLDEN_FRAME);
+ pred_context = 1 +
+ 2 * (edge_mbmi->ref_frame[0] == GOLDEN_FRAME ||
+ edge_mbmi->ref_frame[1] == GOLDEN_FRAME);
}
} else { // inter/inter
const int above_has_second = has_second_ref(above_mbmi);
@@ -290,10 +292,9 @@
if (above_has_second && left_has_second) {
if (above0 == left0 && above1 == left1)
- pred_context = 3 * (above0 == GOLDEN_FRAME ||
- above1 == GOLDEN_FRAME ||
- left0 == GOLDEN_FRAME ||
- left1 == GOLDEN_FRAME);
+ pred_context =
+ 3 * (above0 == GOLDEN_FRAME || above1 == GOLDEN_FRAME ||
+ left0 == GOLDEN_FRAME || left1 == GOLDEN_FRAME);
else
pred_context = 2;
} else if (above_has_second || left_has_second) {
@@ -311,12 +312,12 @@
if (above0 == LAST_FRAME && left0 == LAST_FRAME) {
pred_context = 3;
} else if (above0 == LAST_FRAME || left0 == LAST_FRAME) {
- const MV_REFERENCE_FRAME edge0 = (above0 == LAST_FRAME) ? left0
- : above0;
+ const MV_REFERENCE_FRAME edge0 =
+ (above0 == LAST_FRAME) ? left0 : above0;
pred_context = 4 * (edge0 == GOLDEN_FRAME);
} else {
- pred_context = 2 * (above0 == GOLDEN_FRAME) +
- 2 * (left0 == GOLDEN_FRAME);
+ pred_context =
+ 2 * (above0 == GOLDEN_FRAME) + 2 * (left0 == GOLDEN_FRAME);
}
}
}
diff --git a/vp10/common/pred_common.h b/vp10/common/pred_common.h
index d6d7146..dfb86f9 100644
--- a/vp10/common/pred_common.h
+++ b/vp10/common/pred_common.h
@@ -20,8 +20,8 @@
#endif
static INLINE int get_segment_id(const VP10_COMMON *cm,
- const uint8_t *segment_ids,
- BLOCK_SIZE bsize, int mi_row, int mi_col) {
+ const uint8_t *segment_ids, BLOCK_SIZE bsize,
+ int mi_row, int mi_col) {
const int mi_offset = mi_row * cm->mi_cols + mi_col;
const int bw = num_8x8_blocks_wide_lookup[bsize];
const int bh = num_8x8_blocks_high_lookup[bsize];
@@ -41,8 +41,8 @@
static INLINE int vp10_get_pred_context_seg_id(const MACROBLOCKD *xd) {
const MODE_INFO *const above_mi = xd->above_mi;
const MODE_INFO *const left_mi = xd->left_mi;
- const int above_sip = (above_mi != NULL) ?
- above_mi->mbmi.seg_id_predicted : 0;
+ const int above_sip =
+ (above_mi != NULL) ? above_mi->mbmi.seg_id_predicted : 0;
const int left_sip = (left_mi != NULL) ? left_mi->mbmi.seg_id_predicted : 0;
return above_sip + left_sip;
@@ -62,7 +62,7 @@
}
static INLINE vpx_prob vp10_get_skip_prob(const VP10_COMMON *cm,
- const MACROBLOCKD *xd) {
+ const MACROBLOCKD *xd) {
return cm->fc->skip_probs[vp10_get_skip_context(xd)];
}
@@ -71,7 +71,7 @@
int vp10_get_intra_inter_context(const MACROBLOCKD *xd);
static INLINE vpx_prob vp10_get_intra_inter_prob(const VP10_COMMON *cm,
- const MACROBLOCKD *xd) {
+ const MACROBLOCKD *xd) {
return cm->fc->intra_inter_prob[vp10_get_intra_inter_context(xd)];
}
@@ -79,15 +79,15 @@
const MACROBLOCKD *xd);
static INLINE vpx_prob vp10_get_reference_mode_prob(const VP10_COMMON *cm,
- const MACROBLOCKD *xd) {
+ const MACROBLOCKD *xd) {
return cm->fc->comp_inter_prob[vp10_get_reference_mode_context(cm, xd)];
}
int vp10_get_pred_context_comp_ref_p(const VP10_COMMON *cm,
- const MACROBLOCKD *xd);
+ const MACROBLOCKD *xd);
static INLINE vpx_prob vp10_get_pred_prob_comp_ref_p(const VP10_COMMON *cm,
- const MACROBLOCKD *xd) {
+ const MACROBLOCKD *xd) {
const int pred_context = vp10_get_pred_context_comp_ref_p(cm, xd);
return cm->fc->comp_ref_prob[pred_context];
}
@@ -95,14 +95,14 @@
int vp10_get_pred_context_single_ref_p1(const MACROBLOCKD *xd);
static INLINE vpx_prob vp10_get_pred_prob_single_ref_p1(const VP10_COMMON *cm,
- const MACROBLOCKD *xd) {
+ const MACROBLOCKD *xd) {
return cm->fc->single_ref_prob[vp10_get_pred_context_single_ref_p1(xd)][0];
}
int vp10_get_pred_context_single_ref_p2(const MACROBLOCKD *xd);
static INLINE vpx_prob vp10_get_pred_prob_single_ref_p2(const VP10_COMMON *cm,
- const MACROBLOCKD *xd) {
+ const MACROBLOCKD *xd) {
return cm->fc->single_ref_prob[vp10_get_pred_context_single_ref_p2(xd)][1];
}
@@ -116,15 +116,13 @@
const MB_MODE_INFO *const left_mbmi = xd->left_mbmi;
const int has_above = xd->up_available;
const int has_left = xd->left_available;
- int above_ctx = (has_above && !above_mbmi->skip) ? (int)above_mbmi->tx_size
- : max_tx_size;
- int left_ctx = (has_left && !left_mbmi->skip) ? (int)left_mbmi->tx_size
- : max_tx_size;
- if (!has_left)
- left_ctx = above_ctx;
+ int above_ctx =
+ (has_above && !above_mbmi->skip) ? (int)above_mbmi->tx_size : max_tx_size;
+ int left_ctx =
+ (has_left && !left_mbmi->skip) ? (int)left_mbmi->tx_size : max_tx_size;
+ if (!has_left) left_ctx = above_ctx;
- if (!has_above)
- above_ctx = left_ctx;
+ if (!has_above) above_ctx = left_ctx;
return (above_ctx + left_ctx) > max_tx_size;
}
@@ -132,15 +130,10 @@
static INLINE const vpx_prob *get_tx_probs(TX_SIZE max_tx_size, int ctx,
const struct tx_probs *tx_probs) {
switch (max_tx_size) {
- case TX_8X8:
- return tx_probs->p8x8[ctx];
- case TX_16X16:
- return tx_probs->p16x16[ctx];
- case TX_32X32:
- return tx_probs->p32x32[ctx];
- default:
- assert(0 && "Invalid max_tx_size.");
- return NULL;
+ case TX_8X8: return tx_probs->p8x8[ctx];
+ case TX_16X16: return tx_probs->p16x16[ctx];
+ case TX_32X32: return tx_probs->p32x32[ctx];
+ default: assert(0 && "Invalid max_tx_size."); return NULL;
}
}
@@ -153,15 +146,10 @@
static INLINE unsigned int *get_tx_counts(TX_SIZE max_tx_size, int ctx,
struct tx_counts *tx_counts) {
switch (max_tx_size) {
- case TX_8X8:
- return tx_counts->p8x8[ctx];
- case TX_16X16:
- return tx_counts->p16x16[ctx];
- case TX_32X32:
- return tx_counts->p32x32[ctx];
- default:
- assert(0 && "Invalid max_tx_size.");
- return NULL;
+ case TX_8X8: return tx_counts->p8x8[ctx];
+ case TX_16X16: return tx_counts->p16x16[ctx];
+ case TX_32X32: return tx_counts->p32x32[ctx];
+ default: assert(0 && "Invalid max_tx_size."); return NULL;
}
}
diff --git a/vp10/common/quant_common.c b/vp10/common/quant_common.c
index 968a816..20bc8c5 100644
--- a/vp10/common/quant_common.c
+++ b/vp10/common/quant_common.c
@@ -13,234 +13,166 @@
#include "vp10/common/seg_common.h"
static const int16_t dc_qlookup[QINDEX_RANGE] = {
- 4, 8, 8, 9, 10, 11, 12, 12,
- 13, 14, 15, 16, 17, 18, 19, 19,
- 20, 21, 22, 23, 24, 25, 26, 26,
- 27, 28, 29, 30, 31, 32, 32, 33,
- 34, 35, 36, 37, 38, 38, 39, 40,
- 41, 42, 43, 43, 44, 45, 46, 47,
- 48, 48, 49, 50, 51, 52, 53, 53,
- 54, 55, 56, 57, 57, 58, 59, 60,
- 61, 62, 62, 63, 64, 65, 66, 66,
- 67, 68, 69, 70, 70, 71, 72, 73,
- 74, 74, 75, 76, 77, 78, 78, 79,
- 80, 81, 81, 82, 83, 84, 85, 85,
- 87, 88, 90, 92, 93, 95, 96, 98,
- 99, 101, 102, 104, 105, 107, 108, 110,
- 111, 113, 114, 116, 117, 118, 120, 121,
- 123, 125, 127, 129, 131, 134, 136, 138,
- 140, 142, 144, 146, 148, 150, 152, 154,
- 156, 158, 161, 164, 166, 169, 172, 174,
- 177, 180, 182, 185, 187, 190, 192, 195,
- 199, 202, 205, 208, 211, 214, 217, 220,
- 223, 226, 230, 233, 237, 240, 243, 247,
- 250, 253, 257, 261, 265, 269, 272, 276,
- 280, 284, 288, 292, 296, 300, 304, 309,
- 313, 317, 322, 326, 330, 335, 340, 344,
- 349, 354, 359, 364, 369, 374, 379, 384,
- 389, 395, 400, 406, 411, 417, 423, 429,
- 435, 441, 447, 454, 461, 467, 475, 482,
- 489, 497, 505, 513, 522, 530, 539, 549,
- 559, 569, 579, 590, 602, 614, 626, 640,
- 654, 668, 684, 700, 717, 736, 755, 775,
- 796, 819, 843, 869, 896, 925, 955, 988,
- 1022, 1058, 1098, 1139, 1184, 1232, 1282, 1336,
+ 4, 8, 8, 9, 10, 11, 12, 12, 13, 14, 15, 16, 17, 18,
+ 19, 19, 20, 21, 22, 23, 24, 25, 26, 26, 27, 28, 29, 30,
+ 31, 32, 32, 33, 34, 35, 36, 37, 38, 38, 39, 40, 41, 42,
+ 43, 43, 44, 45, 46, 47, 48, 48, 49, 50, 51, 52, 53, 53,
+ 54, 55, 56, 57, 57, 58, 59, 60, 61, 62, 62, 63, 64, 65,
+ 66, 66, 67, 68, 69, 70, 70, 71, 72, 73, 74, 74, 75, 76,
+ 77, 78, 78, 79, 80, 81, 81, 82, 83, 84, 85, 85, 87, 88,
+ 90, 92, 93, 95, 96, 98, 99, 101, 102, 104, 105, 107, 108, 110,
+ 111, 113, 114, 116, 117, 118, 120, 121, 123, 125, 127, 129, 131, 134,
+ 136, 138, 140, 142, 144, 146, 148, 150, 152, 154, 156, 158, 161, 164,
+ 166, 169, 172, 174, 177, 180, 182, 185, 187, 190, 192, 195, 199, 202,
+ 205, 208, 211, 214, 217, 220, 223, 226, 230, 233, 237, 240, 243, 247,
+ 250, 253, 257, 261, 265, 269, 272, 276, 280, 284, 288, 292, 296, 300,
+ 304, 309, 313, 317, 322, 326, 330, 335, 340, 344, 349, 354, 359, 364,
+ 369, 374, 379, 384, 389, 395, 400, 406, 411, 417, 423, 429, 435, 441,
+ 447, 454, 461, 467, 475, 482, 489, 497, 505, 513, 522, 530, 539, 549,
+ 559, 569, 579, 590, 602, 614, 626, 640, 654, 668, 684, 700, 717, 736,
+ 755, 775, 796, 819, 843, 869, 896, 925, 955, 988, 1022, 1058, 1098, 1139,
+ 1184, 1232, 1282, 1336,
};
#if CONFIG_VPX_HIGHBITDEPTH
static const int16_t dc_qlookup_10[QINDEX_RANGE] = {
- 4, 9, 10, 13, 15, 17, 20, 22,
- 25, 28, 31, 34, 37, 40, 43, 47,
- 50, 53, 57, 60, 64, 68, 71, 75,
- 78, 82, 86, 90, 93, 97, 101, 105,
- 109, 113, 116, 120, 124, 128, 132, 136,
- 140, 143, 147, 151, 155, 159, 163, 166,
- 170, 174, 178, 182, 185, 189, 193, 197,
- 200, 204, 208, 212, 215, 219, 223, 226,
- 230, 233, 237, 241, 244, 248, 251, 255,
- 259, 262, 266, 269, 273, 276, 280, 283,
- 287, 290, 293, 297, 300, 304, 307, 310,
- 314, 317, 321, 324, 327, 331, 334, 337,
- 343, 350, 356, 362, 369, 375, 381, 387,
- 394, 400, 406, 412, 418, 424, 430, 436,
- 442, 448, 454, 460, 466, 472, 478, 484,
- 490, 499, 507, 516, 525, 533, 542, 550,
- 559, 567, 576, 584, 592, 601, 609, 617,
- 625, 634, 644, 655, 666, 676, 687, 698,
- 708, 718, 729, 739, 749, 759, 770, 782,
- 795, 807, 819, 831, 844, 856, 868, 880,
- 891, 906, 920, 933, 947, 961, 975, 988,
- 1001, 1015, 1030, 1045, 1061, 1076, 1090, 1105,
- 1120, 1137, 1153, 1170, 1186, 1202, 1218, 1236,
- 1253, 1271, 1288, 1306, 1323, 1342, 1361, 1379,
- 1398, 1416, 1436, 1456, 1476, 1496, 1516, 1537,
- 1559, 1580, 1601, 1624, 1647, 1670, 1692, 1717,
- 1741, 1766, 1791, 1817, 1844, 1871, 1900, 1929,
- 1958, 1990, 2021, 2054, 2088, 2123, 2159, 2197,
- 2236, 2276, 2319, 2363, 2410, 2458, 2508, 2561,
- 2616, 2675, 2737, 2802, 2871, 2944, 3020, 3102,
- 3188, 3280, 3375, 3478, 3586, 3702, 3823, 3953,
- 4089, 4236, 4394, 4559, 4737, 4929, 5130, 5347,
+ 4, 9, 10, 13, 15, 17, 20, 22, 25, 28, 31, 34, 37,
+ 40, 43, 47, 50, 53, 57, 60, 64, 68, 71, 75, 78, 82,
+ 86, 90, 93, 97, 101, 105, 109, 113, 116, 120, 124, 128, 132,
+ 136, 140, 143, 147, 151, 155, 159, 163, 166, 170, 174, 178, 182,
+ 185, 189, 193, 197, 200, 204, 208, 212, 215, 219, 223, 226, 230,
+ 233, 237, 241, 244, 248, 251, 255, 259, 262, 266, 269, 273, 276,
+ 280, 283, 287, 290, 293, 297, 300, 304, 307, 310, 314, 317, 321,
+ 324, 327, 331, 334, 337, 343, 350, 356, 362, 369, 375, 381, 387,
+ 394, 400, 406, 412, 418, 424, 430, 436, 442, 448, 454, 460, 466,
+ 472, 478, 484, 490, 499, 507, 516, 525, 533, 542, 550, 559, 567,
+ 576, 584, 592, 601, 609, 617, 625, 634, 644, 655, 666, 676, 687,
+ 698, 708, 718, 729, 739, 749, 759, 770, 782, 795, 807, 819, 831,
+ 844, 856, 868, 880, 891, 906, 920, 933, 947, 961, 975, 988, 1001,
+ 1015, 1030, 1045, 1061, 1076, 1090, 1105, 1120, 1137, 1153, 1170, 1186, 1202,
+ 1218, 1236, 1253, 1271, 1288, 1306, 1323, 1342, 1361, 1379, 1398, 1416, 1436,
+ 1456, 1476, 1496, 1516, 1537, 1559, 1580, 1601, 1624, 1647, 1670, 1692, 1717,
+ 1741, 1766, 1791, 1817, 1844, 1871, 1900, 1929, 1958, 1990, 2021, 2054, 2088,
+ 2123, 2159, 2197, 2236, 2276, 2319, 2363, 2410, 2458, 2508, 2561, 2616, 2675,
+ 2737, 2802, 2871, 2944, 3020, 3102, 3188, 3280, 3375, 3478, 3586, 3702, 3823,
+ 3953, 4089, 4236, 4394, 4559, 4737, 4929, 5130, 5347,
};
static const int16_t dc_qlookup_12[QINDEX_RANGE] = {
- 4, 12, 18, 25, 33, 41, 50, 60,
- 70, 80, 91, 103, 115, 127, 140, 153,
- 166, 180, 194, 208, 222, 237, 251, 266,
- 281, 296, 312, 327, 343, 358, 374, 390,
- 405, 421, 437, 453, 469, 484, 500, 516,
- 532, 548, 564, 580, 596, 611, 627, 643,
- 659, 674, 690, 706, 721, 737, 752, 768,
- 783, 798, 814, 829, 844, 859, 874, 889,
- 904, 919, 934, 949, 964, 978, 993, 1008,
- 1022, 1037, 1051, 1065, 1080, 1094, 1108, 1122,
- 1136, 1151, 1165, 1179, 1192, 1206, 1220, 1234,
- 1248, 1261, 1275, 1288, 1302, 1315, 1329, 1342,
- 1368, 1393, 1419, 1444, 1469, 1494, 1519, 1544,
- 1569, 1594, 1618, 1643, 1668, 1692, 1717, 1741,
- 1765, 1789, 1814, 1838, 1862, 1885, 1909, 1933,
- 1957, 1992, 2027, 2061, 2096, 2130, 2165, 2199,
- 2233, 2267, 2300, 2334, 2367, 2400, 2434, 2467,
- 2499, 2532, 2575, 2618, 2661, 2704, 2746, 2788,
- 2830, 2872, 2913, 2954, 2995, 3036, 3076, 3127,
- 3177, 3226, 3275, 3324, 3373, 3421, 3469, 3517,
- 3565, 3621, 3677, 3733, 3788, 3843, 3897, 3951,
- 4005, 4058, 4119, 4181, 4241, 4301, 4361, 4420,
- 4479, 4546, 4612, 4677, 4742, 4807, 4871, 4942,
- 5013, 5083, 5153, 5222, 5291, 5367, 5442, 5517,
- 5591, 5665, 5745, 5825, 5905, 5984, 6063, 6149,
- 6234, 6319, 6404, 6495, 6587, 6678, 6769, 6867,
- 6966, 7064, 7163, 7269, 7376, 7483, 7599, 7715,
- 7832, 7958, 8085, 8214, 8352, 8492, 8635, 8788,
- 8945, 9104, 9275, 9450, 9639, 9832, 10031, 10245,
- 10465, 10702, 10946, 11210, 11482, 11776, 12081, 12409,
- 12750, 13118, 13501, 13913, 14343, 14807, 15290, 15812,
- 16356, 16943, 17575, 18237, 18949, 19718, 20521, 21387,
+ 4, 12, 18, 25, 33, 41, 50, 60, 70, 80, 91,
+ 103, 115, 127, 140, 153, 166, 180, 194, 208, 222, 237,
+ 251, 266, 281, 296, 312, 327, 343, 358, 374, 390, 405,
+ 421, 437, 453, 469, 484, 500, 516, 532, 548, 564, 580,
+ 596, 611, 627, 643, 659, 674, 690, 706, 721, 737, 752,
+ 768, 783, 798, 814, 829, 844, 859, 874, 889, 904, 919,
+ 934, 949, 964, 978, 993, 1008, 1022, 1037, 1051, 1065, 1080,
+ 1094, 1108, 1122, 1136, 1151, 1165, 1179, 1192, 1206, 1220, 1234,
+ 1248, 1261, 1275, 1288, 1302, 1315, 1329, 1342, 1368, 1393, 1419,
+ 1444, 1469, 1494, 1519, 1544, 1569, 1594, 1618, 1643, 1668, 1692,
+ 1717, 1741, 1765, 1789, 1814, 1838, 1862, 1885, 1909, 1933, 1957,
+ 1992, 2027, 2061, 2096, 2130, 2165, 2199, 2233, 2267, 2300, 2334,
+ 2367, 2400, 2434, 2467, 2499, 2532, 2575, 2618, 2661, 2704, 2746,
+ 2788, 2830, 2872, 2913, 2954, 2995, 3036, 3076, 3127, 3177, 3226,
+ 3275, 3324, 3373, 3421, 3469, 3517, 3565, 3621, 3677, 3733, 3788,
+ 3843, 3897, 3951, 4005, 4058, 4119, 4181, 4241, 4301, 4361, 4420,
+ 4479, 4546, 4612, 4677, 4742, 4807, 4871, 4942, 5013, 5083, 5153,
+ 5222, 5291, 5367, 5442, 5517, 5591, 5665, 5745, 5825, 5905, 5984,
+ 6063, 6149, 6234, 6319, 6404, 6495, 6587, 6678, 6769, 6867, 6966,
+ 7064, 7163, 7269, 7376, 7483, 7599, 7715, 7832, 7958, 8085, 8214,
+ 8352, 8492, 8635, 8788, 8945, 9104, 9275, 9450, 9639, 9832, 10031,
+ 10245, 10465, 10702, 10946, 11210, 11482, 11776, 12081, 12409, 12750, 13118,
+ 13501, 13913, 14343, 14807, 15290, 15812, 16356, 16943, 17575, 18237, 18949,
+ 19718, 20521, 21387,
};
#endif
static const int16_t ac_qlookup[QINDEX_RANGE] = {
- 4, 8, 9, 10, 11, 12, 13, 14,
- 15, 16, 17, 18, 19, 20, 21, 22,
- 23, 24, 25, 26, 27, 28, 29, 30,
- 31, 32, 33, 34, 35, 36, 37, 38,
- 39, 40, 41, 42, 43, 44, 45, 46,
- 47, 48, 49, 50, 51, 52, 53, 54,
- 55, 56, 57, 58, 59, 60, 61, 62,
- 63, 64, 65, 66, 67, 68, 69, 70,
- 71, 72, 73, 74, 75, 76, 77, 78,
- 79, 80, 81, 82, 83, 84, 85, 86,
- 87, 88, 89, 90, 91, 92, 93, 94,
- 95, 96, 97, 98, 99, 100, 101, 102,
- 104, 106, 108, 110, 112, 114, 116, 118,
- 120, 122, 124, 126, 128, 130, 132, 134,
- 136, 138, 140, 142, 144, 146, 148, 150,
- 152, 155, 158, 161, 164, 167, 170, 173,
- 176, 179, 182, 185, 188, 191, 194, 197,
- 200, 203, 207, 211, 215, 219, 223, 227,
- 231, 235, 239, 243, 247, 251, 255, 260,
- 265, 270, 275, 280, 285, 290, 295, 300,
- 305, 311, 317, 323, 329, 335, 341, 347,
- 353, 359, 366, 373, 380, 387, 394, 401,
- 408, 416, 424, 432, 440, 448, 456, 465,
- 474, 483, 492, 501, 510, 520, 530, 540,
- 550, 560, 571, 582, 593, 604, 615, 627,
- 639, 651, 663, 676, 689, 702, 715, 729,
- 743, 757, 771, 786, 801, 816, 832, 848,
- 864, 881, 898, 915, 933, 951, 969, 988,
- 1007, 1026, 1046, 1066, 1087, 1108, 1129, 1151,
- 1173, 1196, 1219, 1243, 1267, 1292, 1317, 1343,
- 1369, 1396, 1423, 1451, 1479, 1508, 1537, 1567,
- 1597, 1628, 1660, 1692, 1725, 1759, 1793, 1828,
+ 4, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
+ 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32,
+ 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45,
+ 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58,
+ 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71,
+ 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84,
+ 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97,
+ 98, 99, 100, 101, 102, 104, 106, 108, 110, 112, 114, 116, 118,
+ 120, 122, 124, 126, 128, 130, 132, 134, 136, 138, 140, 142, 144,
+ 146, 148, 150, 152, 155, 158, 161, 164, 167, 170, 173, 176, 179,
+ 182, 185, 188, 191, 194, 197, 200, 203, 207, 211, 215, 219, 223,
+ 227, 231, 235, 239, 243, 247, 251, 255, 260, 265, 270, 275, 280,
+ 285, 290, 295, 300, 305, 311, 317, 323, 329, 335, 341, 347, 353,
+ 359, 366, 373, 380, 387, 394, 401, 408, 416, 424, 432, 440, 448,
+ 456, 465, 474, 483, 492, 501, 510, 520, 530, 540, 550, 560, 571,
+ 582, 593, 604, 615, 627, 639, 651, 663, 676, 689, 702, 715, 729,
+ 743, 757, 771, 786, 801, 816, 832, 848, 864, 881, 898, 915, 933,
+ 951, 969, 988, 1007, 1026, 1046, 1066, 1087, 1108, 1129, 1151, 1173, 1196,
+ 1219, 1243, 1267, 1292, 1317, 1343, 1369, 1396, 1423, 1451, 1479, 1508, 1537,
+ 1567, 1597, 1628, 1660, 1692, 1725, 1759, 1793, 1828,
};
#if CONFIG_VPX_HIGHBITDEPTH
static const int16_t ac_qlookup_10[QINDEX_RANGE] = {
- 4, 9, 11, 13, 16, 18, 21, 24,
- 27, 30, 33, 37, 40, 44, 48, 51,
- 55, 59, 63, 67, 71, 75, 79, 83,
- 88, 92, 96, 100, 105, 109, 114, 118,
- 122, 127, 131, 136, 140, 145, 149, 154,
- 158, 163, 168, 172, 177, 181, 186, 190,
- 195, 199, 204, 208, 213, 217, 222, 226,
- 231, 235, 240, 244, 249, 253, 258, 262,
- 267, 271, 275, 280, 284, 289, 293, 297,
- 302, 306, 311, 315, 319, 324, 328, 332,
- 337, 341, 345, 349, 354, 358, 362, 367,
- 371, 375, 379, 384, 388, 392, 396, 401,
- 409, 417, 425, 433, 441, 449, 458, 466,
- 474, 482, 490, 498, 506, 514, 523, 531,
- 539, 547, 555, 563, 571, 579, 588, 596,
- 604, 616, 628, 640, 652, 664, 676, 688,
- 700, 713, 725, 737, 749, 761, 773, 785,
- 797, 809, 825, 841, 857, 873, 889, 905,
- 922, 938, 954, 970, 986, 1002, 1018, 1038,
- 1058, 1078, 1098, 1118, 1138, 1158, 1178, 1198,
- 1218, 1242, 1266, 1290, 1314, 1338, 1362, 1386,
- 1411, 1435, 1463, 1491, 1519, 1547, 1575, 1603,
- 1631, 1663, 1695, 1727, 1759, 1791, 1823, 1859,
- 1895, 1931, 1967, 2003, 2039, 2079, 2119, 2159,
- 2199, 2239, 2283, 2327, 2371, 2415, 2459, 2507,
- 2555, 2603, 2651, 2703, 2755, 2807, 2859, 2915,
- 2971, 3027, 3083, 3143, 3203, 3263, 3327, 3391,
- 3455, 3523, 3591, 3659, 3731, 3803, 3876, 3952,
- 4028, 4104, 4184, 4264, 4348, 4432, 4516, 4604,
- 4692, 4784, 4876, 4972, 5068, 5168, 5268, 5372,
- 5476, 5584, 5692, 5804, 5916, 6032, 6148, 6268,
- 6388, 6512, 6640, 6768, 6900, 7036, 7172, 7312,
+ 4, 9, 11, 13, 16, 18, 21, 24, 27, 30, 33, 37, 40,
+ 44, 48, 51, 55, 59, 63, 67, 71, 75, 79, 83, 88, 92,
+ 96, 100, 105, 109, 114, 118, 122, 127, 131, 136, 140, 145, 149,
+ 154, 158, 163, 168, 172, 177, 181, 186, 190, 195, 199, 204, 208,
+ 213, 217, 222, 226, 231, 235, 240, 244, 249, 253, 258, 262, 267,
+ 271, 275, 280, 284, 289, 293, 297, 302, 306, 311, 315, 319, 324,
+ 328, 332, 337, 341, 345, 349, 354, 358, 362, 367, 371, 375, 379,
+ 384, 388, 392, 396, 401, 409, 417, 425, 433, 441, 449, 458, 466,
+ 474, 482, 490, 498, 506, 514, 523, 531, 539, 547, 555, 563, 571,
+ 579, 588, 596, 604, 616, 628, 640, 652, 664, 676, 688, 700, 713,
+ 725, 737, 749, 761, 773, 785, 797, 809, 825, 841, 857, 873, 889,
+ 905, 922, 938, 954, 970, 986, 1002, 1018, 1038, 1058, 1078, 1098, 1118,
+ 1138, 1158, 1178, 1198, 1218, 1242, 1266, 1290, 1314, 1338, 1362, 1386, 1411,
+ 1435, 1463, 1491, 1519, 1547, 1575, 1603, 1631, 1663, 1695, 1727, 1759, 1791,
+ 1823, 1859, 1895, 1931, 1967, 2003, 2039, 2079, 2119, 2159, 2199, 2239, 2283,
+ 2327, 2371, 2415, 2459, 2507, 2555, 2603, 2651, 2703, 2755, 2807, 2859, 2915,
+ 2971, 3027, 3083, 3143, 3203, 3263, 3327, 3391, 3455, 3523, 3591, 3659, 3731,
+ 3803, 3876, 3952, 4028, 4104, 4184, 4264, 4348, 4432, 4516, 4604, 4692, 4784,
+ 4876, 4972, 5068, 5168, 5268, 5372, 5476, 5584, 5692, 5804, 5916, 6032, 6148,
+ 6268, 6388, 6512, 6640, 6768, 6900, 7036, 7172, 7312,
};
static const int16_t ac_qlookup_12[QINDEX_RANGE] = {
- 4, 13, 19, 27, 35, 44, 54, 64,
- 75, 87, 99, 112, 126, 139, 154, 168,
- 183, 199, 214, 230, 247, 263, 280, 297,
- 314, 331, 349, 366, 384, 402, 420, 438,
- 456, 475, 493, 511, 530, 548, 567, 586,
- 604, 623, 642, 660, 679, 698, 716, 735,
- 753, 772, 791, 809, 828, 846, 865, 884,
- 902, 920, 939, 957, 976, 994, 1012, 1030,
- 1049, 1067, 1085, 1103, 1121, 1139, 1157, 1175,
- 1193, 1211, 1229, 1246, 1264, 1282, 1299, 1317,
- 1335, 1352, 1370, 1387, 1405, 1422, 1440, 1457,
- 1474, 1491, 1509, 1526, 1543, 1560, 1577, 1595,
- 1627, 1660, 1693, 1725, 1758, 1791, 1824, 1856,
- 1889, 1922, 1954, 1987, 2020, 2052, 2085, 2118,
- 2150, 2183, 2216, 2248, 2281, 2313, 2346, 2378,
- 2411, 2459, 2508, 2556, 2605, 2653, 2701, 2750,
- 2798, 2847, 2895, 2943, 2992, 3040, 3088, 3137,
- 3185, 3234, 3298, 3362, 3426, 3491, 3555, 3619,
- 3684, 3748, 3812, 3876, 3941, 4005, 4069, 4149,
- 4230, 4310, 4390, 4470, 4550, 4631, 4711, 4791,
- 4871, 4967, 5064, 5160, 5256, 5352, 5448, 5544,
- 5641, 5737, 5849, 5961, 6073, 6185, 6297, 6410,
- 6522, 6650, 6778, 6906, 7034, 7162, 7290, 7435,
- 7579, 7723, 7867, 8011, 8155, 8315, 8475, 8635,
- 8795, 8956, 9132, 9308, 9484, 9660, 9836, 10028,
- 10220, 10412, 10604, 10812, 11020, 11228, 11437, 11661,
- 11885, 12109, 12333, 12573, 12813, 13053, 13309, 13565,
- 13821, 14093, 14365, 14637, 14925, 15213, 15502, 15806,
- 16110, 16414, 16734, 17054, 17390, 17726, 18062, 18414,
- 18766, 19134, 19502, 19886, 20270, 20670, 21070, 21486,
- 21902, 22334, 22766, 23214, 23662, 24126, 24590, 25070,
- 25551, 26047, 26559, 27071, 27599, 28143, 28687, 29247,
+ 4, 13, 19, 27, 35, 44, 54, 64, 75, 87, 99,
+ 112, 126, 139, 154, 168, 183, 199, 214, 230, 247, 263,
+ 280, 297, 314, 331, 349, 366, 384, 402, 420, 438, 456,
+ 475, 493, 511, 530, 548, 567, 586, 604, 623, 642, 660,
+ 679, 698, 716, 735, 753, 772, 791, 809, 828, 846, 865,
+ 884, 902, 920, 939, 957, 976, 994, 1012, 1030, 1049, 1067,
+ 1085, 1103, 1121, 1139, 1157, 1175, 1193, 1211, 1229, 1246, 1264,
+ 1282, 1299, 1317, 1335, 1352, 1370, 1387, 1405, 1422, 1440, 1457,
+ 1474, 1491, 1509, 1526, 1543, 1560, 1577, 1595, 1627, 1660, 1693,
+ 1725, 1758, 1791, 1824, 1856, 1889, 1922, 1954, 1987, 2020, 2052,
+ 2085, 2118, 2150, 2183, 2216, 2248, 2281, 2313, 2346, 2378, 2411,
+ 2459, 2508, 2556, 2605, 2653, 2701, 2750, 2798, 2847, 2895, 2943,
+ 2992, 3040, 3088, 3137, 3185, 3234, 3298, 3362, 3426, 3491, 3555,
+ 3619, 3684, 3748, 3812, 3876, 3941, 4005, 4069, 4149, 4230, 4310,
+ 4390, 4470, 4550, 4631, 4711, 4791, 4871, 4967, 5064, 5160, 5256,
+ 5352, 5448, 5544, 5641, 5737, 5849, 5961, 6073, 6185, 6297, 6410,
+ 6522, 6650, 6778, 6906, 7034, 7162, 7290, 7435, 7579, 7723, 7867,
+ 8011, 8155, 8315, 8475, 8635, 8795, 8956, 9132, 9308, 9484, 9660,
+ 9836, 10028, 10220, 10412, 10604, 10812, 11020, 11228, 11437, 11661, 11885,
+ 12109, 12333, 12573, 12813, 13053, 13309, 13565, 13821, 14093, 14365, 14637,
+ 14925, 15213, 15502, 15806, 16110, 16414, 16734, 17054, 17390, 17726, 18062,
+ 18414, 18766, 19134, 19502, 19886, 20270, 20670, 21070, 21486, 21902, 22334,
+ 22766, 23214, 23662, 24126, 24590, 25070, 25551, 26047, 26559, 27071, 27599,
+ 28143, 28687, 29247,
};
#endif
int16_t vp10_dc_quant(int qindex, int delta, vpx_bit_depth_t bit_depth) {
#if CONFIG_VPX_HIGHBITDEPTH
switch (bit_depth) {
- case VPX_BITS_8:
- return dc_qlookup[clamp(qindex + delta, 0, MAXQ)];
- case VPX_BITS_10:
- return dc_qlookup_10[clamp(qindex + delta, 0, MAXQ)];
- case VPX_BITS_12:
- return dc_qlookup_12[clamp(qindex + delta, 0, MAXQ)];
+ case VPX_BITS_8: return dc_qlookup[clamp(qindex + delta, 0, MAXQ)];
+ case VPX_BITS_10: return dc_qlookup_10[clamp(qindex + delta, 0, MAXQ)];
+ case VPX_BITS_12: return dc_qlookup_12[clamp(qindex + delta, 0, MAXQ)];
default:
assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
return -1;
}
#else
- (void) bit_depth;
+ (void)bit_depth;
return dc_qlookup[clamp(qindex + delta, 0, MAXQ)];
#endif
}
@@ -248,31 +180,27 @@
int16_t vp10_ac_quant(int qindex, int delta, vpx_bit_depth_t bit_depth) {
#if CONFIG_VPX_HIGHBITDEPTH
switch (bit_depth) {
- case VPX_BITS_8:
- return ac_qlookup[clamp(qindex + delta, 0, MAXQ)];
- case VPX_BITS_10:
- return ac_qlookup_10[clamp(qindex + delta, 0, MAXQ)];
- case VPX_BITS_12:
- return ac_qlookup_12[clamp(qindex + delta, 0, MAXQ)];
+ case VPX_BITS_8: return ac_qlookup[clamp(qindex + delta, 0, MAXQ)];
+ case VPX_BITS_10: return ac_qlookup_10[clamp(qindex + delta, 0, MAXQ)];
+ case VPX_BITS_12: return ac_qlookup_12[clamp(qindex + delta, 0, MAXQ)];
default:
assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
return -1;
}
#else
- (void) bit_depth;
+ (void)bit_depth;
return ac_qlookup[clamp(qindex + delta, 0, MAXQ)];
#endif
}
int vp10_get_qindex(const struct segmentation *seg, int segment_id,
- int base_qindex) {
+ int base_qindex) {
if (segfeature_active(seg, segment_id, SEG_LVL_ALT_Q)) {
const int data = get_segdata(seg, segment_id, SEG_LVL_ALT_Q);
- const int seg_qindex = seg->abs_delta == SEGMENT_ABSDATA ?
- data : base_qindex + data;
+ const int seg_qindex =
+ seg->abs_delta == SEGMENT_ABSDATA ? data : base_qindex + data;
return clamp(seg_qindex, 0, MAXQ);
} else {
return base_qindex;
}
}
-
diff --git a/vp10/common/quant_common.h b/vp10/common/quant_common.h
index 6813e17..c6bb24f 100644
--- a/vp10/common/quant_common.h
+++ b/vp10/common/quant_common.h
@@ -27,7 +27,7 @@
int16_t vp10_ac_quant(int qindex, int delta, vpx_bit_depth_t bit_depth);
int vp10_get_qindex(const struct segmentation *seg, int segment_id,
- int base_qindex);
+ int base_qindex);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/common/reconinter.c b/vp10/common/reconinter.c
index ddfb8fa..f857e94 100644
--- a/vp10/common/reconinter.c
+++ b/vp10/common/reconinter.c
@@ -20,14 +20,11 @@
#include "vp10/common/reconintra.h"
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_highbd_build_inter_predictor(const uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- const MV *src_mv,
- const struct scale_factors *sf,
- int w, int h, int ref,
- const InterpKernel *kernel,
- enum mv_precision precision,
- int x, int y, int bd) {
+void vp10_highbd_build_inter_predictor(
+ const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride,
+ const MV *src_mv, const struct scale_factors *sf, int w, int h, int ref,
+ const InterpKernel *kernel, enum mv_precision precision, int x, int y,
+ int bd) {
const int is_q4 = precision == MV_PRECISION_Q4;
const MV mv_q4 = { is_q4 ? src_mv->row : src_mv->row * 2,
is_q4 ? src_mv->col : src_mv->col * 2 };
@@ -37,19 +34,16 @@
src += (mv.row >> SUBPEL_BITS) * src_stride + (mv.col >> SUBPEL_BITS);
- high_inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y,
- sf, w, h, ref, kernel, sf->x_step_q4, sf->y_step_q4, bd);
+ high_inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y, sf,
+ w, h, ref, kernel, sf->x_step_q4, sf->y_step_q4, bd);
}
#endif // CONFIG_VPX_HIGHBITDEPTH
void vp10_build_inter_predictor(const uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- const MV *src_mv,
- const struct scale_factors *sf,
- int w, int h, int ref,
- const InterpKernel *kernel,
- enum mv_precision precision,
- int x, int y) {
+ uint8_t *dst, int dst_stride, const MV *src_mv,
+ const struct scale_factors *sf, int w, int h,
+ int ref, const InterpKernel *kernel,
+ enum mv_precision precision, int x, int y) {
const int is_q4 = precision == MV_PRECISION_Q4;
const MV mv_q4 = { is_q4 ? src_mv->row : src_mv->row * 2,
is_q4 ? src_mv->col : src_mv->col * 2 };
@@ -59,14 +53,13 @@
src += (mv.row >> SUBPEL_BITS) * src_stride + (mv.col >> SUBPEL_BITS);
- inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y,
- sf, w, h, ref, kernel, sf->x_step_q4, sf->y_step_q4);
+ inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y, sf, w,
+ h, ref, kernel, sf->x_step_q4, sf->y_step_q4);
}
-void build_inter_predictors(MACROBLOCKD *xd, int plane, int block,
- int bw, int bh,
- int x, int y, int w, int h,
- int mi_x, int mi_y) {
+void build_inter_predictors(MACROBLOCKD *xd, int plane, int block, int bw,
+ int bh, int x, int y, int w, int h, int mi_x,
+ int mi_y) {
struct macroblockd_plane *const pd = &xd->plane[plane];
const MODE_INFO *mi = xd->mi[0];
const int is_compound = has_second_ref(&mi->mbmi);
@@ -79,17 +72,16 @@
struct buf_2d *const dst_buf = &pd->dst;
uint8_t *const dst = dst_buf->buf + dst_buf->stride * y + x;
const MV mv = mi->mbmi.sb_type < BLOCK_8X8
- ? average_split_mvs(pd, mi, ref, block)
- : mi->mbmi.mv[ref].as_mv;
+ ? average_split_mvs(pd, mi, ref, block)
+ : mi->mbmi.mv[ref].as_mv;
// TODO(jkoleszar): This clamping is done in the incorrect place for the
// scaling case. It needs to be done on the scaled MV, not the pre-scaling
// MV. Note however that it performs the subsampling aware scaling so
// that the result is always q4.
// mv_precision precision is MV_PRECISION_Q4.
- const MV mv_q4 = clamp_mv_to_umv_border_sb(xd, &mv, bw, bh,
- pd->subsampling_x,
- pd->subsampling_y);
+ const MV mv_q4 = clamp_mv_to_umv_border_sb(
+ xd, &mv, bw, bh, pd->subsampling_x, pd->subsampling_y);
uint8_t *pre;
MV32 scaled_mv;
@@ -109,28 +101,26 @@
}
subpel_x = scaled_mv.col & SUBPEL_MASK;
subpel_y = scaled_mv.row & SUBPEL_MASK;
- pre += (scaled_mv.row >> SUBPEL_BITS) * pre_buf->stride
- + (scaled_mv.col >> SUBPEL_BITS);
+ pre += (scaled_mv.row >> SUBPEL_BITS) * pre_buf->stride +
+ (scaled_mv.col >> SUBPEL_BITS);
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- high_inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride,
- subpel_x, subpel_y, sf, w, h, ref, kernel, xs, ys,
- xd->bd);
+ high_inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride, subpel_x,
+ subpel_y, sf, w, h, ref, kernel, xs, ys, xd->bd);
} else {
- inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride,
- subpel_x, subpel_y, sf, w, h, ref, kernel, xs, ys);
+ inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride, subpel_x,
+ subpel_y, sf, w, h, ref, kernel, xs, ys);
}
#else
- inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride,
- subpel_x, subpel_y, sf, w, h, ref, kernel, xs, ys);
+ inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride, subpel_x,
+ subpel_y, sf, w, h, ref, kernel, xs, ys);
#endif // CONFIG_VPX_HIGHBITDEPTH
}
}
-void vp10_build_inter_predictor_sub8x8(MACROBLOCKD *xd, int plane,
- int i, int ir, int ic,
- int mi_row, int mi_col) {
+void vp10_build_inter_predictor_sub8x8(MACROBLOCKD *xd, int plane, int i,
+ int ir, int ic, int mi_row, int mi_col) {
struct macroblockd_plane *const pd = &xd->plane[plane];
MODE_INFO *const mi = xd->mi[0];
const BLOCK_SIZE plane_bsize = get_plane_block_size(mi->mbmi.sb_type, pd);
@@ -146,31 +136,25 @@
const uint8_t *pre =
&pd->pre[ref].buf[(ir * pd->pre[ref].stride + ic) << 2];
#if CONFIG_VPX_HIGHBITDEPTH
- if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- vp10_highbd_build_inter_predictor(pre, pd->pre[ref].stride,
- dst, pd->dst.stride,
- &mi->bmi[i].as_mv[ref].as_mv,
- &xd->block_refs[ref]->sf, width, height,
- ref, kernel, MV_PRECISION_Q3,
- mi_col * MI_SIZE + 4 * ic,
- mi_row * MI_SIZE + 4 * ir, xd->bd);
- } else {
- vp10_build_inter_predictor(pre, pd->pre[ref].stride,
- dst, pd->dst.stride,
- &mi->bmi[i].as_mv[ref].as_mv,
- &xd->block_refs[ref]->sf, width, height, ref,
- kernel, MV_PRECISION_Q3,
- mi_col * MI_SIZE + 4 * ic,
- mi_row * MI_SIZE + 4 * ir);
- }
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp10_highbd_build_inter_predictor(
+ pre, pd->pre[ref].stride, dst, pd->dst.stride,
+ &mi->bmi[i].as_mv[ref].as_mv, &xd->block_refs[ref]->sf, width, height,
+ ref, kernel, MV_PRECISION_Q3, mi_col * MI_SIZE + 4 * ic,
+ mi_row * MI_SIZE + 4 * ir, xd->bd);
+ } else {
+ vp10_build_inter_predictor(
+ pre, pd->pre[ref].stride, dst, pd->dst.stride,
+ &mi->bmi[i].as_mv[ref].as_mv, &xd->block_refs[ref]->sf, width, height,
+ ref, kernel, MV_PRECISION_Q3, mi_col * MI_SIZE + 4 * ic,
+ mi_row * MI_SIZE + 4 * ir);
+ }
#else
- vp10_build_inter_predictor(pre, pd->pre[ref].stride,
- dst, pd->dst.stride,
- &mi->bmi[i].as_mv[ref].as_mv,
- &xd->block_refs[ref]->sf, width, height, ref,
- kernel, MV_PRECISION_Q3,
- mi_col * MI_SIZE + 4 * ic,
- mi_row * MI_SIZE + 4 * ir);
+ vp10_build_inter_predictor(
+ pre, pd->pre[ref].stride, dst, pd->dst.stride,
+ &mi->bmi[i].as_mv[ref].as_mv, &xd->block_refs[ref]->sf, width, height,
+ ref, kernel, MV_PRECISION_Q3, mi_col * MI_SIZE + 4 * ic,
+ mi_row * MI_SIZE + 4 * ir);
#endif // CONFIG_VPX_HIGHBITDEPTH
}
}
@@ -200,44 +184,43 @@
assert(pw * num_4x4_w == bw && ph * num_4x4_h == bh);
for (y = 0; y < num_4x4_h; ++y)
for (x = 0; x < num_4x4_w; ++x)
- build_inter_predictors(xd, plane, y * 2 + x, bw, bh,
- 4 * x, 4 * y, pw, ph, mi_x, mi_y);
+ build_inter_predictors(xd, plane, y * 2 + x, bw, bh, 4 * x, 4 * y, pw,
+ ph, mi_x, mi_y);
} else {
- build_inter_predictors(xd, plane, 0, bw, bh,
- 0, 0, bw, bh, mi_x, mi_y);
+ build_inter_predictors(xd, plane, 0, bw, bh, 0, 0, bw, bh, mi_x, mi_y);
}
}
}
void vp10_build_inter_predictors_sby(MACROBLOCKD *xd, int mi_row, int mi_col,
- BLOCK_SIZE bsize) {
+ BLOCK_SIZE bsize) {
build_inter_predictors_for_planes(xd, bsize, mi_row, mi_col, 0, 0);
}
void vp10_build_inter_predictors_sbp(MACROBLOCKD *xd, int mi_row, int mi_col,
- BLOCK_SIZE bsize, int plane) {
+ BLOCK_SIZE bsize, int plane) {
build_inter_predictors_for_planes(xd, bsize, mi_row, mi_col, plane, plane);
}
void vp10_build_inter_predictors_sbuv(MACROBLOCKD *xd, int mi_row, int mi_col,
- BLOCK_SIZE bsize) {
+ BLOCK_SIZE bsize) {
build_inter_predictors_for_planes(xd, bsize, mi_row, mi_col, 1,
MAX_MB_PLANE - 1);
}
void vp10_build_inter_predictors_sb(MACROBLOCKD *xd, int mi_row, int mi_col,
- BLOCK_SIZE bsize) {
+ BLOCK_SIZE bsize) {
build_inter_predictors_for_planes(xd, bsize, mi_row, mi_col, 0,
MAX_MB_PLANE - 1);
}
void vp10_setup_dst_planes(struct macroblockd_plane planes[MAX_MB_PLANE],
- const YV12_BUFFER_CONFIG *src,
- int mi_row, int mi_col) {
+ const YV12_BUFFER_CONFIG *src, int mi_row,
+ int mi_col) {
uint8_t *const buffers[MAX_MB_PLANE] = { src->y_buffer, src->u_buffer,
- src->v_buffer};
+ src->v_buffer };
const int strides[MAX_MB_PLANE] = { src->y_stride, src->uv_stride,
- src->uv_stride};
+ src->uv_stride };
int i;
for (i = 0; i < MAX_MB_PLANE; ++i) {
@@ -248,15 +231,14 @@
}
void vp10_setup_pre_planes(MACROBLOCKD *xd, int idx,
- const YV12_BUFFER_CONFIG *src,
- int mi_row, int mi_col,
- const struct scale_factors *sf) {
+ const YV12_BUFFER_CONFIG *src, int mi_row,
+ int mi_col, const struct scale_factors *sf) {
if (src != NULL) {
int i;
uint8_t *const buffers[MAX_MB_PLANE] = { src->y_buffer, src->u_buffer,
- src->v_buffer};
+ src->v_buffer };
const int strides[MAX_MB_PLANE] = { src->y_stride, src->uv_stride,
- src->uv_stride};
+ src->uv_stride };
for (i = 0; i < MAX_MB_PLANE; ++i) {
struct macroblockd_plane *const pd = &xd->plane[i];
setup_pred_plane(&pd->pre[idx], buffers[i], strides[i], mi_row, mi_col,
diff --git a/vp10/common/reconinter.h b/vp10/common/reconinter.h
index 451ea08..6af16ae 100644
--- a/vp10/common/reconinter.h
+++ b/vp10/common/reconinter.h
@@ -22,29 +22,23 @@
static INLINE void inter_predictor(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
- const int subpel_x,
- const int subpel_y,
- const struct scale_factors *sf,
- int w, int h, int ref,
- const InterpKernel *kernel,
- int xs, int ys) {
+ const int subpel_x, const int subpel_y,
+ const struct scale_factors *sf, int w, int h,
+ int ref, const InterpKernel *kernel, int xs,
+ int ys) {
sf->predict[subpel_x != 0][subpel_y != 0][ref](
- src, src_stride, dst, dst_stride,
- kernel[subpel_x], xs, kernel[subpel_y], ys, w, h);
+ src, src_stride, dst, dst_stride, kernel[subpel_x], xs, kernel[subpel_y],
+ ys, w, h);
}
#if CONFIG_VPX_HIGHBITDEPTH
-static INLINE void high_inter_predictor(const uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- const int subpel_x,
- const int subpel_y,
- const struct scale_factors *sf,
- int w, int h, int ref,
- const InterpKernel *kernel,
- int xs, int ys, int bd) {
+static INLINE void high_inter_predictor(
+ const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride,
+ const int subpel_x, const int subpel_y, const struct scale_factors *sf,
+ int w, int h, int ref, const InterpKernel *kernel, int xs, int ys, int bd) {
sf->highbd_predict[subpel_x != 0][subpel_y != 0][ref](
- src, src_stride, dst, dst_stride,
- kernel[subpel_x], xs, kernel[subpel_y], ys, w, h, bd);
+ src, src_stride, dst, dst_stride, kernel[subpel_x], xs, kernel[subpel_y],
+ ys, w, h, bd);
}
#endif // CONFIG_VPX_HIGHBITDEPTH
@@ -53,14 +47,14 @@
}
static MV mi_mv_pred_q4(const MODE_INFO *mi, int idx) {
- MV res = { round_mv_comp_q4(mi->bmi[0].as_mv[idx].as_mv.row +
- mi->bmi[1].as_mv[idx].as_mv.row +
- mi->bmi[2].as_mv[idx].as_mv.row +
- mi->bmi[3].as_mv[idx].as_mv.row),
- round_mv_comp_q4(mi->bmi[0].as_mv[idx].as_mv.col +
- mi->bmi[1].as_mv[idx].as_mv.col +
- mi->bmi[2].as_mv[idx].as_mv.col +
- mi->bmi[3].as_mv[idx].as_mv.col) };
+ MV res = {
+ round_mv_comp_q4(
+ mi->bmi[0].as_mv[idx].as_mv.row + mi->bmi[1].as_mv[idx].as_mv.row +
+ mi->bmi[2].as_mv[idx].as_mv.row + mi->bmi[3].as_mv[idx].as_mv.row),
+ round_mv_comp_q4(
+ mi->bmi[0].as_mv[idx].as_mv.col + mi->bmi[1].as_mv[idx].as_mv.col +
+ mi->bmi[2].as_mv[idx].as_mv.col + mi->bmi[3].as_mv[idx].as_mv.col)
+ };
return res;
}
@@ -78,8 +72,8 @@
// TODO(jkoleszar): yet another mv clamping function :-(
static INLINE MV clamp_mv_to_umv_border_sb(const MACROBLOCKD *xd,
- const MV *src_mv,
- int bw, int bh, int ss_x, int ss_y) {
+ const MV *src_mv, int bw, int bh,
+ int ss_x, int ss_y) {
// If the MV points so far into the UMV border that no visible pixels
// are used for reconstruction, the subpel part of the MV can be
// discarded and the MV limited to 16 pixels with equivalent results.
@@ -87,15 +81,12 @@
const int spel_right = spel_left - SUBPEL_SHIFTS;
const int spel_top = (VPX_INTERP_EXTEND + bh) << SUBPEL_BITS;
const int spel_bottom = spel_top - SUBPEL_SHIFTS;
- MV clamped_mv = {
- src_mv->row * (1 << (1 - ss_y)),
- src_mv->col * (1 << (1 - ss_x))
- };
+ MV clamped_mv = { src_mv->row * (1 << (1 - ss_y)),
+ src_mv->col * (1 << (1 - ss_x)) };
assert(ss_x <= 1);
assert(ss_y <= 1);
- clamp_mv(&clamped_mv,
- xd->mb_to_left_edge * (1 << (1 - ss_x)) - spel_left,
+ clamp_mv(&clamped_mv, xd->mb_to_left_edge * (1 << (1 - ss_x)) - spel_left,
xd->mb_to_right_edge * (1 << (1 - ss_x)) + spel_right,
xd->mb_to_top_edge * (1 << (1 - ss_y)) - spel_top,
xd->mb_to_bottom_edge * (1 << (1 - ss_y)) + spel_bottom);
@@ -106,65 +97,48 @@
static INLINE MV average_split_mvs(const struct macroblockd_plane *pd,
const MODE_INFO *mi, int ref, int block) {
const int ss_idx = ((pd->subsampling_x > 0) << 1) | (pd->subsampling_y > 0);
- MV res = {0, 0};
+ MV res = { 0, 0 };
switch (ss_idx) {
- case 0:
- res = mi->bmi[block].as_mv[ref].as_mv;
- break;
- case 1:
- res = mi_mv_pred_q2(mi, ref, block, block + 2);
- break;
- case 2:
- res = mi_mv_pred_q2(mi, ref, block, block + 1);
- break;
- case 3:
- res = mi_mv_pred_q4(mi, ref);
- break;
- default:
- assert(ss_idx <= 3 && ss_idx >= 0);
+ case 0: res = mi->bmi[block].as_mv[ref].as_mv; break;
+ case 1: res = mi_mv_pred_q2(mi, ref, block, block + 2); break;
+ case 2: res = mi_mv_pred_q2(mi, ref, block, block + 1); break;
+ case 3: res = mi_mv_pred_q4(mi, ref); break;
+ default: assert(ss_idx <= 3 && ss_idx >= 0);
}
return res;
}
-void build_inter_predictors(MACROBLOCKD *xd, int plane, int block,
- int bw, int bh,
- int x, int y, int w, int h,
- int mi_x, int mi_y);
+void build_inter_predictors(MACROBLOCKD *xd, int plane, int block, int bw,
+ int bh, int x, int y, int w, int h, int mi_x,
+ int mi_y);
-void vp10_build_inter_predictor_sub8x8(MACROBLOCKD *xd, int plane,
- int i, int ir, int ic,
- int mi_row, int mi_col);
+void vp10_build_inter_predictor_sub8x8(MACROBLOCKD *xd, int plane, int i,
+ int ir, int ic, int mi_row, int mi_col);
void vp10_build_inter_predictors_sby(MACROBLOCKD *xd, int mi_row, int mi_col,
- BLOCK_SIZE bsize);
-
-void vp10_build_inter_predictors_sbp(MACROBLOCKD *xd, int mi_row, int mi_col,
- BLOCK_SIZE bsize, int plane);
-
-void vp10_build_inter_predictors_sbuv(MACROBLOCKD *xd, int mi_row, int mi_col,
BLOCK_SIZE bsize);
+void vp10_build_inter_predictors_sbp(MACROBLOCKD *xd, int mi_row, int mi_col,
+ BLOCK_SIZE bsize, int plane);
+
+void vp10_build_inter_predictors_sbuv(MACROBLOCKD *xd, int mi_row, int mi_col,
+ BLOCK_SIZE bsize);
+
void vp10_build_inter_predictors_sb(MACROBLOCKD *xd, int mi_row, int mi_col,
- BLOCK_SIZE bsize);
+ BLOCK_SIZE bsize);
void vp10_build_inter_predictor(const uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- const MV *mv_q3,
- const struct scale_factors *sf,
- int w, int h, int do_avg,
- const InterpKernel *kernel,
- enum mv_precision precision,
- int x, int y);
+ uint8_t *dst, int dst_stride, const MV *mv_q3,
+ const struct scale_factors *sf, int w, int h,
+ int do_avg, const InterpKernel *kernel,
+ enum mv_precision precision, int x, int y);
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_highbd_build_inter_predictor(const uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- const MV *mv_q3,
- const struct scale_factors *sf,
- int w, int h, int do_avg,
- const InterpKernel *kernel,
- enum mv_precision precision,
- int x, int y, int bd);
+void vp10_highbd_build_inter_predictor(
+ const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride,
+ const MV *mv_q3, const struct scale_factors *sf, int w, int h, int do_avg,
+ const InterpKernel *kernel, enum mv_precision precision, int x, int y,
+ int bd);
#endif
static INLINE int scaled_buffer_offset(int x_offset, int y_offset, int stride,
@@ -174,9 +148,8 @@
return y * stride + x;
}
-static INLINE void setup_pred_plane(struct buf_2d *dst,
- uint8_t *src, int stride,
- int mi_row, int mi_col,
+static INLINE void setup_pred_plane(struct buf_2d *dst, uint8_t *src,
+ int stride, int mi_row, int mi_col,
const struct scale_factors *scale,
int subsampling_x, int subsampling_y) {
const int x = (MI_SIZE * mi_col) >> subsampling_x;
@@ -186,12 +159,12 @@
}
void vp10_setup_dst_planes(struct macroblockd_plane planes[MAX_MB_PLANE],
- const YV12_BUFFER_CONFIG *src,
- int mi_row, int mi_col);
+ const YV12_BUFFER_CONFIG *src, int mi_row,
+ int mi_col);
void vp10_setup_pre_planes(MACROBLOCKD *xd, int idx,
- const YV12_BUFFER_CONFIG *src, int mi_row, int mi_col,
- const struct scale_factors *sf);
+ const YV12_BUFFER_CONFIG *src, int mi_row,
+ int mi_col, const struct scale_factors *sf);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/common/reconintra.c b/vp10/common/reconintra.c
index 0d53794..8ad3c63 100644
--- a/vp10/common/reconintra.c
+++ b/vp10/common/reconintra.c
@@ -50,16 +50,16 @@
};
static const uint8_t extend_modes[INTRA_MODES] = {
- NEED_ABOVE | NEED_LEFT, // DC
- NEED_ABOVE, // V
- NEED_LEFT, // H
- NEED_ABOVERIGHT, // D45
- NEED_LEFT | NEED_ABOVE, // D135
- NEED_LEFT | NEED_ABOVE, // D117
- NEED_LEFT | NEED_ABOVE, // D153
- NEED_LEFT, // D207
- NEED_ABOVERIGHT, // D63
- NEED_LEFT | NEED_ABOVE, // TM
+ NEED_ABOVE | NEED_LEFT, // DC
+ NEED_ABOVE, // V
+ NEED_LEFT, // H
+ NEED_ABOVERIGHT, // D45
+ NEED_LEFT | NEED_ABOVE, // D135
+ NEED_LEFT | NEED_ABOVE, // D117
+ NEED_LEFT | NEED_ABOVE, // D153
+ NEED_LEFT, // D207
+ NEED_ABOVERIGHT, // D63
+ NEED_LEFT | NEED_ABOVE, // TM
};
#endif
@@ -68,61 +68,40 @@
static const uint8_t orders_64x32[2] = { 0, 1 };
static const uint8_t orders_32x64[2] = { 0, 1 };
static const uint8_t orders_32x32[4] = {
- 0, 1,
- 2, 3,
+ 0, 1, 2, 3,
};
static const uint8_t orders_32x16[8] = {
- 0, 2,
- 1, 3,
- 4, 6,
- 5, 7,
+ 0, 2, 1, 3, 4, 6, 5, 7,
};
static const uint8_t orders_16x32[8] = {
- 0, 1, 2, 3,
- 4, 5, 6, 7,
+ 0, 1, 2, 3, 4, 5, 6, 7,
};
static const uint8_t orders_16x16[16] = {
- 0, 1, 4, 5,
- 2, 3, 6, 7,
- 8, 9, 12, 13,
- 10, 11, 14, 15,
+ 0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15,
};
static const uint8_t orders_16x8[32] = {
- 0, 2, 8, 10,
- 1, 3, 9, 11,
- 4, 6, 12, 14,
- 5, 7, 13, 15,
- 16, 18, 24, 26,
- 17, 19, 25, 27,
- 20, 22, 28, 30,
- 21, 23, 29, 31,
+ 0, 2, 8, 10, 1, 3, 9, 11, 4, 6, 12, 14, 5, 7, 13, 15,
+ 16, 18, 24, 26, 17, 19, 25, 27, 20, 22, 28, 30, 21, 23, 29, 31,
};
static const uint8_t orders_8x16[32] = {
- 0, 1, 2, 3, 8, 9, 10, 11,
- 4, 5, 6, 7, 12, 13, 14, 15,
- 16, 17, 18, 19, 24, 25, 26, 27,
- 20, 21, 22, 23, 28, 29, 30, 31,
+ 0, 1, 2, 3, 8, 9, 10, 11, 4, 5, 6, 7, 12, 13, 14, 15,
+ 16, 17, 18, 19, 24, 25, 26, 27, 20, 21, 22, 23, 28, 29, 30, 31,
};
static const uint8_t orders_8x8[64] = {
- 0, 1, 4, 5, 16, 17, 20, 21,
- 2, 3, 6, 7, 18, 19, 22, 23,
- 8, 9, 12, 13, 24, 25, 28, 29,
- 10, 11, 14, 15, 26, 27, 30, 31,
- 32, 33, 36, 37, 48, 49, 52, 53,
- 34, 35, 38, 39, 50, 51, 54, 55,
- 40, 41, 44, 45, 56, 57, 60, 61,
- 42, 43, 46, 47, 58, 59, 62, 63,
+ 0, 1, 4, 5, 16, 17, 20, 21, 2, 3, 6, 7, 18, 19, 22, 23,
+ 8, 9, 12, 13, 24, 25, 28, 29, 10, 11, 14, 15, 26, 27, 30, 31,
+ 32, 33, 36, 37, 48, 49, 52, 53, 34, 35, 38, 39, 50, 51, 54, 55,
+ 40, 41, 44, 45, 56, 57, 60, 61, 42, 43, 46, 47, 58, 59, 62, 63,
};
static const uint8_t *const orders[BLOCK_SIZES] = {
- orders_8x8, orders_8x8, orders_8x8, orders_8x8,
- orders_8x16, orders_16x8, orders_16x16,
- orders_16x32, orders_32x16, orders_32x32,
+ orders_8x8, orders_8x8, orders_8x8, orders_8x8, orders_8x16,
+ orders_16x8, orders_16x16, orders_16x32, orders_32x16, orders_32x32,
orders_32x64, orders_64x32, orders_64x64,
};
static int vp10_has_right(BLOCK_SIZE bsize, int mi_row, int mi_col,
- int right_available,
- TX_SIZE txsz, int y, int x, int ss_x) {
+ int right_available, TX_SIZE txsz, int y, int x,
+ int ss_x) {
if (y == 0) {
int wl = mi_width_log2_lookup[bsize];
int hl = mi_height_log2_lookup[bsize];
@@ -131,17 +110,14 @@
const uint8_t *order = orders[bsize];
int my_order, tr_order;
- if (x + step < w)
- return 1;
+ if (x + step < w) return 1;
mi_row = (mi_row & 7) >> hl;
mi_col = (mi_col & 7) >> wl;
- if (mi_row == 0)
- return right_available;
+ if (mi_row == 0) return right_available;
- if (((mi_col + 1) << wl) >= 8)
- return 0;
+ if (((mi_col + 1) << wl) >= 8) return 0;
my_order = order[((mi_row + 0) << (3 - wl)) + mi_col + 0];
tr_order = order[((mi_row - 1) << (3 - wl)) + mi_col + 1];
@@ -157,8 +133,8 @@
}
static int vp10_has_bottom(BLOCK_SIZE bsize, int mi_row, int mi_col,
- int bottom_available, TX_SIZE txsz,
- int y, int x, int ss_y) {
+ int bottom_available, TX_SIZE txsz, int y, int x,
+ int ss_y) {
if (x == 0) {
int wl = mi_width_log2_lookup[bsize];
int hl = mi_height_log2_lookup[bsize];
@@ -174,11 +150,9 @@
return bottom_available &&
(mi_row << (hl + !ss_y)) + y + step < (8 << !ss_y);
- if (((mi_row + 1) << hl) >= 8)
- return 0;
+ if (((mi_row + 1) << hl) >= 8) return 0;
- if (y + step < h)
- return 1;
+ if (y + step < h) return 1;
my_order = order[((mi_row + 0) << (3 - wl)) + mi_col + 0];
bl_order = order[((mi_row + 1) << (3 - wl)) + mi_col - 1];
@@ -205,12 +179,12 @@
#endif // CONFIG_VPX_HIGHBITDEPTH
static void vp10_init_intra_predictors_internal(void) {
-#define INIT_NO_4X4(p, type) \
- p[TX_8X8] = vpx_##type##_predictor_8x8; \
+#define INIT_NO_4X4(p, type) \
+ p[TX_8X8] = vpx_##type##_predictor_8x8; \
p[TX_16X16] = vpx_##type##_predictor_16x16; \
p[TX_32X32] = vpx_##type##_predictor_32x32
-#define INIT_ALL_SIZES(p, type) \
+#define INIT_ALL_SIZES(p, type) \
p[TX_4X4] = vpx_##type##_predictor_4x4; \
INIT_NO_4X4(p, type)
@@ -263,29 +237,23 @@
#if CONFIG_MISC_FIXES
static INLINE void memset16(uint16_t *dst, int val, int n) {
- while (n--)
- *dst++ = val;
+ while (n--) *dst++ = val;
}
#endif
#if CONFIG_VPX_HIGHBITDEPTH
static void build_intra_predictors_high(const MACROBLOCKD *xd,
- const uint8_t *ref8,
- int ref_stride,
- uint8_t *dst8,
- int dst_stride,
- PREDICTION_MODE mode,
- TX_SIZE tx_size,
+ const uint8_t *ref8, int ref_stride,
+ uint8_t *dst8, int dst_stride,
+ PREDICTION_MODE mode, TX_SIZE tx_size,
#if CONFIG_MISC_FIXES
int n_top_px, int n_topright_px,
int n_left_px, int n_bottomleft_px,
#else
- int up_available,
- int left_available,
+ int up_available, int left_available,
int right_available,
#endif
- int x, int y,
- int plane, int bd) {
+ int x, int y, int plane, int bd) {
int i;
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
@@ -309,27 +277,26 @@
const int need_above = extend_modes[mode] & NEED_ABOVE;
const int need_aboveright = extend_modes[mode] & NEED_ABOVERIGHT;
int base = 128 << (bd - 8);
- // 127 127 127 .. 127 127 127 127 127 127
- // 129 A B .. Y Z
- // 129 C D .. W X
- // 129 E F .. U V
- // 129 G H .. S T T T T T
+// 127 127 127 .. 127 127 127 127 127 127
+// 129 A B .. Y Z
+// 129 C D .. W X
+// 129 E F .. U V
+// 129 G H .. S T T T T T
#if CONFIG_MISC_FIXES
- (void) x;
- (void) y;
- (void) plane;
- (void) need_left;
- (void) need_above;
- (void) need_aboveright;
+ (void)x;
+ (void)y;
+ (void)plane;
+ (void)need_left;
+ (void)need_above;
+ (void)need_aboveright;
// NEED_LEFT
if (extend_modes[mode] & NEED_LEFT) {
const int need_bottom = !!(extend_modes[mode] & NEED_BOTTOMLEFT);
i = 0;
if (n_left_px > 0) {
- for (; i < n_left_px; i++)
- left_col[i] = ref[i * ref_stride - 1];
+ for (; i < n_left_px; i++) left_col[i] = ref[i * ref_stride - 1];
if (need_bottom && n_bottomleft_px > 0) {
assert(i == bs);
for (; i < bs + n_bottomleft_px; i++)
@@ -361,8 +328,8 @@
}
if (extend_modes[mode] & NEED_ABOVELEFT) {
- above_row[-1] = n_top_px > 0 ?
- (n_left_px > 0 ? above_ref[-1] : base + 1) : base - 1;
+ above_row[-1] =
+ n_top_px > 0 ? (n_left_px > 0 ? above_ref[-1] : base + 1) : base - 1;
}
#else
// Get current frame pointer, width and height.
@@ -384,8 +351,7 @@
if (xd->mb_to_bottom_edge < 0) {
/* slower path if the block needs border extension */
if (y0 + bs <= frame_height) {
- for (i = 0; i < bs; ++i)
- left_col[i] = ref[i * ref_stride - 1];
+ for (i = 0; i < bs; ++i) left_col[i] = ref[i * ref_stride - 1];
} else {
const int extend_bottom = frame_height - y0;
for (i = 0; i < extend_bottom; ++i)
@@ -395,8 +361,7 @@
}
} else {
/* faster path if the block does not need extension */
- for (i = 0; i < bs; ++i)
- left_col[i] = ref[i * ref_stride - 1];
+ for (i = 0; i < bs; ++i) left_col[i] = ref[i * ref_stride - 1];
}
} else {
// TODO(Peter): this value should probably change for high bitdepth
@@ -488,13 +453,11 @@
// predict
if (mode == DC_PRED) {
#if CONFIG_MISC_FIXES
- dc_pred_high[n_left_px > 0][n_top_px > 0][tx_size](dst, dst_stride,
- const_above_row,
- left_col, xd->bd);
+ dc_pred_high[n_left_px > 0][n_top_px > 0][tx_size](
+ dst, dst_stride, const_above_row, left_col, xd->bd);
#else
- dc_pred_high[left_available][up_available][tx_size](dst, dst_stride,
- const_above_row,
- left_col, xd->bd);
+ dc_pred_high[left_available][up_available][tx_size](
+ dst, dst_stride, const_above_row, left_col, xd->bd);
#endif
} else {
pred_high[mode][tx_size](dst, dst_stride, const_above_row, left_col,
@@ -529,18 +492,18 @@
const uint8_t *const_above_row = above_row;
const int bs = 4 << tx_size;
- // 127 127 127 .. 127 127 127 127 127 127
- // 129 A B .. Y Z
- // 129 C D .. W X
- // 129 E F .. U V
- // 129 G H .. S T T T T T
- // ..
+// 127 127 127 .. 127 127 127 127 127 127
+// 129 A B .. Y Z
+// 129 C D .. W X
+// 129 E F .. U V
+// 129 G H .. S T T T T T
+// ..
#if CONFIG_MISC_FIXES
- (void) xd;
- (void) x;
- (void) y;
- (void) plane;
+ (void)xd;
+ (void)x;
+ (void)y;
+ (void)plane;
assert(n_top_px >= 0);
assert(n_topright_px >= 0);
assert(n_left_px >= 0);
@@ -566,8 +529,7 @@
const int need_bottom = !!(extend_modes[mode] & NEED_BOTTOMLEFT);
i = 0;
if (n_left_px > 0) {
- for (; i < n_left_px; i++)
- left_col[i] = ref[i * ref_stride - 1];
+ for (; i < n_left_px; i++) left_col[i] = ref[i * ref_stride - 1];
if (need_bottom && n_bottomleft_px > 0) {
assert(i == bs);
for (; i < bs + n_bottomleft_px; i++)
@@ -583,8 +545,7 @@
if (xd->mb_to_bottom_edge < 0) {
/* slower path if the block needs border extension */
if (y0 + bs <= frame_height) {
- for (i = 0; i < bs; ++i)
- left_col[i] = ref[i * ref_stride - 1];
+ for (i = 0; i < bs; ++i) left_col[i] = ref[i * ref_stride - 1];
} else {
const int extend_bottom = frame_height - y0;
for (i = 0; i < extend_bottom; ++i)
@@ -594,8 +555,7 @@
}
} else {
/* faster path if the block does not need extension */
- for (i = 0; i < bs; ++i)
- left_col[i] = ref[i * ref_stride - 1];
+ for (i = 0; i < bs; ++i) left_col[i] = ref[i * ref_stride - 1];
}
} else {
memset(left_col, 129, bs);
@@ -715,10 +675,9 @@
}
void vp10_predict_intra_block(const MACROBLOCKD *xd, int bwl_in, int bhl_in,
- TX_SIZE tx_size, PREDICTION_MODE mode,
- const uint8_t *ref, int ref_stride,
- uint8_t *dst, int dst_stride,
- int aoff, int loff, int plane) {
+ TX_SIZE tx_size, PREDICTION_MODE mode,
+ const uint8_t *ref, int ref_stride, uint8_t *dst,
+ int dst_stride, int aoff, int loff, int plane) {
const int txw = (1 << tx_size);
const int have_top = loff || xd->up_available;
const int have_left = aoff || xd->left_available;
@@ -733,14 +692,11 @@
const struct macroblockd_plane *const pd = &xd->plane[plane];
const int right_available =
mi_col + (bw >> !pd->subsampling_x) < xd->tile.mi_col_end;
- const int have_right = vp10_has_right(bsize, mi_row, mi_col,
- right_available,
- tx_size, loff, aoff,
- pd->subsampling_x);
- const int have_bottom = vp10_has_bottom(bsize, mi_row, mi_col,
- xd->mb_to_bottom_edge > 0,
- tx_size, loff, aoff,
- pd->subsampling_y);
+ const int have_right = vp10_has_right(bsize, mi_row, mi_col, right_available,
+ tx_size, loff, aoff, pd->subsampling_x);
+ const int have_bottom =
+ vp10_has_bottom(bsize, mi_row, mi_col, xd->mb_to_bottom_edge > 0, tx_size,
+ loff, aoff, pd->subsampling_y);
const int wpx = 4 * bw;
const int hpx = 4 * bh;
const int txpx = 4 * txw;
@@ -757,8 +713,7 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
build_intra_predictors_high(xd, ref, ref_stride, dst, dst_stride, mode,
- tx_size,
- have_top ? VPXMIN(txpx, xr + txpx) : 0,
+ tx_size, have_top ? VPXMIN(txpx, xr + txpx) : 0,
have_top && have_right ? VPXMIN(txpx, xr) : 0,
have_left ? VPXMIN(txpx, yd + txpx) : 0,
have_bottom && have_left ? VPXMIN(txpx, yd) : 0,
@@ -766,20 +721,19 @@
return;
}
#endif
- build_intra_predictors(xd, ref, ref_stride, dst, dst_stride, mode,
- tx_size,
+ build_intra_predictors(xd, ref, ref_stride, dst, dst_stride, mode, tx_size,
have_top ? VPXMIN(txpx, xr + txpx) : 0,
have_top && have_right ? VPXMIN(txpx, xr) : 0,
have_left ? VPXMIN(txpx, yd + txpx) : 0,
- have_bottom && have_left ? VPXMIN(txpx, yd) : 0,
- x, y, plane);
+ have_bottom && have_left ? VPXMIN(txpx, yd) : 0, x, y,
+ plane);
#else // CONFIG_MISC_FIXES
- (void) bhl_in;
+ (void)bhl_in;
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
build_intra_predictors_high(xd, ref, ref_stride, dst, dst_stride, mode,
- tx_size, have_top, have_left, have_right,
- x, y, plane, xd->bd);
+ tx_size, have_top, have_left, have_right, x, y,
+ plane, xd->bd);
return;
}
#endif
diff --git a/vp10/common/reconintra.h b/vp10/common/reconintra.h
index f451fb8..a38f058 100644
--- a/vp10/common/reconintra.h
+++ b/vp10/common/reconintra.h
@@ -21,10 +21,9 @@
void vp10_init_intra_predictors(void);
void vp10_predict_intra_block(const MACROBLOCKD *xd, int bwl_in, int bhl_in,
- TX_SIZE tx_size, PREDICTION_MODE mode,
- const uint8_t *ref, int ref_stride,
- uint8_t *dst, int dst_stride,
- int aoff, int loff, int plane);
+ TX_SIZE tx_size, PREDICTION_MODE mode,
+ const uint8_t *ref, int ref_stride, uint8_t *dst,
+ int dst_stride, int aoff, int loff, int plane);
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/vp10/common/scale.c b/vp10/common/scale.c
index c4287bf..d654b4a 100644
--- a/vp10/common/scale.c
+++ b/vp10/common/scale.c
@@ -22,7 +22,7 @@
}
static int unscaled_value(int val, const struct scale_factors *sf) {
- (void) sf;
+ (void)sf;
return val;
}
@@ -37,22 +37,18 @@
MV32 vp10_scale_mv(const MV *mv, int x, int y, const struct scale_factors *sf) {
const int x_off_q4 = scaled_x(x << SUBPEL_BITS, sf) & SUBPEL_MASK;
const int y_off_q4 = scaled_y(y << SUBPEL_BITS, sf) & SUBPEL_MASK;
- const MV32 res = {
- scaled_y(mv->row, sf) + y_off_q4,
- scaled_x(mv->col, sf) + x_off_q4
- };
+ const MV32 res = { scaled_y(mv->row, sf) + y_off_q4,
+ scaled_x(mv->col, sf) + x_off_q4 };
return res;
}
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_setup_scale_factors_for_frame(struct scale_factors *sf,
- int other_w, int other_h,
- int this_w, int this_h,
- int use_highbd) {
+void vp10_setup_scale_factors_for_frame(struct scale_factors *sf, int other_w,
+ int other_h, int this_w, int this_h,
+ int use_highbd) {
#else
-void vp10_setup_scale_factors_for_frame(struct scale_factors *sf,
- int other_w, int other_h,
- int this_w, int this_h) {
+void vp10_setup_scale_factors_for_frame(struct scale_factors *sf, int other_w,
+ int other_h, int this_w, int this_h) {
#endif
if (!valid_ref_frame_size(other_w, other_h, this_w, this_h)) {
sf->x_scale_fp = REF_INVALID_SCALE;
diff --git a/vp10/common/scale.h b/vp10/common/scale.h
index 709b5ad..70e301b 100644
--- a/vp10/common/scale.h
+++ b/vp10/common/scale.h
@@ -23,8 +23,8 @@
#define REF_INVALID_SCALE -1
struct scale_factors {
- int x_scale_fp; // horizontal fixed point scale factor
- int y_scale_fp; // vertical fixed point scale factor
+ int x_scale_fp; // horizontal fixed point scale factor
+ int y_scale_fp; // vertical fixed point scale factor
int x_step_q4;
int y_step_q4;
@@ -40,14 +40,12 @@
MV32 vp10_scale_mv(const MV *mv, int x, int y, const struct scale_factors *sf);
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_setup_scale_factors_for_frame(struct scale_factors *sf,
- int other_w, int other_h,
- int this_w, int this_h,
- int use_high);
+void vp10_setup_scale_factors_for_frame(struct scale_factors *sf, int other_w,
+ int other_h, int this_w, int this_h,
+ int use_high);
#else
-void vp10_setup_scale_factors_for_frame(struct scale_factors *sf,
- int other_w, int other_h,
- int this_w, int this_h);
+void vp10_setup_scale_factors_for_frame(struct scale_factors *sf, int other_w,
+ int other_h, int this_w, int this_h);
#endif
static INLINE int vp10_is_valid_scale(const struct scale_factors *sf) {
@@ -61,11 +59,9 @@
}
static INLINE int valid_ref_frame_size(int ref_width, int ref_height,
- int this_width, int this_height) {
- return 2 * this_width >= ref_width &&
- 2 * this_height >= ref_height &&
- this_width <= 16 * ref_width &&
- this_height <= 16 * ref_height;
+ int this_width, int this_height) {
+ return 2 * this_width >= ref_width && 2 * this_height >= ref_height &&
+ this_width <= 16 * ref_width && this_height <= 16 * ref_height;
}
#ifdef __cplusplus
diff --git a/vp10/common/scan.c b/vp10/common/scan.c
index 7217f6d..8edeea8 100644
--- a/vp10/common/scan.c
+++ b/vp10/common/scan.c
@@ -13,220 +13,181 @@
#include "vp10/common/scan.h"
DECLARE_ALIGNED(16, static const int16_t, default_scan_4x4[16]) = {
- 0, 4, 1, 5,
- 8, 2, 12, 9,
- 3, 6, 13, 10,
- 7, 14, 11, 15,
+ 0, 4, 1, 5, 8, 2, 12, 9, 3, 6, 13, 10, 7, 14, 11, 15,
};
DECLARE_ALIGNED(16, static const int16_t, col_scan_4x4[16]) = {
- 0, 4, 8, 1,
- 12, 5, 9, 2,
- 13, 6, 10, 3,
- 7, 14, 11, 15,
+ 0, 4, 8, 1, 12, 5, 9, 2, 13, 6, 10, 3, 7, 14, 11, 15,
};
DECLARE_ALIGNED(16, static const int16_t, row_scan_4x4[16]) = {
- 0, 1, 4, 2,
- 5, 3, 6, 8,
- 9, 7, 12, 10,
- 13, 11, 14, 15,
+ 0, 1, 4, 2, 5, 3, 6, 8, 9, 7, 12, 10, 13, 11, 14, 15,
};
DECLARE_ALIGNED(16, static const int16_t, default_scan_8x8[64]) = {
- 0, 8, 1, 16, 9, 2, 17, 24,
- 10, 3, 18, 25, 32, 11, 4, 26,
- 33, 19, 40, 12, 34, 27, 5, 41,
- 20, 48, 13, 35, 42, 28, 21, 6,
- 49, 56, 36, 43, 29, 7, 14, 50,
- 57, 44, 22, 37, 15, 51, 58, 30,
- 45, 23, 52, 59, 38, 31, 60, 53,
- 46, 39, 61, 54, 47, 62, 55, 63,
+ 0, 8, 1, 16, 9, 2, 17, 24, 10, 3, 18, 25, 32, 11, 4, 26,
+ 33, 19, 40, 12, 34, 27, 5, 41, 20, 48, 13, 35, 42, 28, 21, 6,
+ 49, 56, 36, 43, 29, 7, 14, 50, 57, 44, 22, 37, 15, 51, 58, 30,
+ 45, 23, 52, 59, 38, 31, 60, 53, 46, 39, 61, 54, 47, 62, 55, 63,
};
DECLARE_ALIGNED(16, static const int16_t, col_scan_8x8[64]) = {
- 0, 8, 16, 1, 24, 9, 32, 17,
- 2, 40, 25, 10, 33, 18, 48, 3,
- 26, 41, 11, 56, 19, 34, 4, 49,
- 27, 42, 12, 35, 20, 57, 50, 28,
- 5, 43, 13, 36, 58, 51, 21, 44,
- 6, 29, 59, 37, 14, 52, 22, 7,
- 45, 60, 30, 15, 38, 53, 23, 46,
- 31, 61, 39, 54, 47, 62, 55, 63,
+ 0, 8, 16, 1, 24, 9, 32, 17, 2, 40, 25, 10, 33, 18, 48, 3,
+ 26, 41, 11, 56, 19, 34, 4, 49, 27, 42, 12, 35, 20, 57, 50, 28,
+ 5, 43, 13, 36, 58, 51, 21, 44, 6, 29, 59, 37, 14, 52, 22, 7,
+ 45, 60, 30, 15, 38, 53, 23, 46, 31, 61, 39, 54, 47, 62, 55, 63,
};
DECLARE_ALIGNED(16, static const int16_t, row_scan_8x8[64]) = {
- 0, 1, 2, 8, 9, 3, 16, 10,
- 4, 17, 11, 24, 5, 18, 25, 12,
- 19, 26, 32, 6, 13, 20, 33, 27,
- 7, 34, 40, 21, 28, 41, 14, 35,
- 48, 42, 29, 36, 49, 22, 43, 15,
- 56, 37, 50, 44, 30, 57, 23, 51,
- 58, 45, 38, 52, 31, 59, 53, 46,
- 60, 39, 61, 47, 54, 55, 62, 63,
+ 0, 1, 2, 8, 9, 3, 16, 10, 4, 17, 11, 24, 5, 18, 25, 12,
+ 19, 26, 32, 6, 13, 20, 33, 27, 7, 34, 40, 21, 28, 41, 14, 35,
+ 48, 42, 29, 36, 49, 22, 43, 15, 56, 37, 50, 44, 30, 57, 23, 51,
+ 58, 45, 38, 52, 31, 59, 53, 46, 60, 39, 61, 47, 54, 55, 62, 63,
};
DECLARE_ALIGNED(16, static const int16_t, default_scan_16x16[256]) = {
- 0, 16, 1, 32, 17, 2, 48, 33, 18, 3, 64, 34, 49, 19, 65, 80,
- 50, 4, 35, 66, 20, 81, 96, 51, 5, 36, 82, 97, 67, 112, 21, 52,
- 98, 37, 83, 113, 6, 68, 128, 53, 22, 99, 114, 84, 7, 129, 38, 69,
- 100, 115, 144, 130, 85, 54, 23, 8, 145, 39, 70, 116, 101, 131, 160, 146,
- 55, 86, 24, 71, 132, 117, 161, 40, 9, 102, 147, 176, 162, 87, 56, 25,
- 133, 118, 177, 148, 72, 103, 41, 163, 10, 192, 178, 88, 57, 134, 149, 119,
- 26, 164, 73, 104, 193, 42, 179, 208, 11, 135, 89, 165, 120, 150, 58, 194,
- 180, 27, 74, 209, 105, 151, 136, 43, 90, 224, 166, 195, 181, 121, 210, 59,
- 12, 152, 106, 167, 196, 75, 137, 225, 211, 240, 182, 122, 91, 28, 197, 13,
- 226, 168, 183, 153, 44, 212, 138, 107, 241, 60, 29, 123, 198, 184, 227, 169,
- 242, 76, 213, 154, 45, 92, 14, 199, 139, 61, 228, 214, 170, 185, 243, 108,
- 77, 155, 30, 15, 200, 229, 124, 215, 244, 93, 46, 186, 171, 201, 109, 140,
- 230, 62, 216, 245, 31, 125, 78, 156, 231, 47, 187, 202, 217, 94, 246, 141,
- 63, 232, 172, 110, 247, 157, 79, 218, 203, 126, 233, 188, 248, 95, 173, 142,
- 219, 111, 249, 234, 158, 127, 189, 204, 250, 235, 143, 174, 220, 205, 159,
- 251,
+ 0, 16, 1, 32, 17, 2, 48, 33, 18, 3, 64, 34, 49, 19, 65,
+ 80, 50, 4, 35, 66, 20, 81, 96, 51, 5, 36, 82, 97, 67, 112,
+ 21, 52, 98, 37, 83, 113, 6, 68, 128, 53, 22, 99, 114, 84, 7,
+ 129, 38, 69, 100, 115, 144, 130, 85, 54, 23, 8, 145, 39, 70, 116,
+ 101, 131, 160, 146, 55, 86, 24, 71, 132, 117, 161, 40, 9, 102, 147,
+ 176, 162, 87, 56, 25, 133, 118, 177, 148, 72, 103, 41, 163, 10, 192,
+ 178, 88, 57, 134, 149, 119, 26, 164, 73, 104, 193, 42, 179, 208, 11,
+ 135, 89, 165, 120, 150, 58, 194, 180, 27, 74, 209, 105, 151, 136, 43,
+ 90, 224, 166, 195, 181, 121, 210, 59, 12, 152, 106, 167, 196, 75, 137,
+ 225, 211, 240, 182, 122, 91, 28, 197, 13, 226, 168, 183, 153, 44, 212,
+ 138, 107, 241, 60, 29, 123, 198, 184, 227, 169, 242, 76, 213, 154, 45,
+ 92, 14, 199, 139, 61, 228, 214, 170, 185, 243, 108, 77, 155, 30, 15,
+ 200, 229, 124, 215, 244, 93, 46, 186, 171, 201, 109, 140, 230, 62, 216,
+ 245, 31, 125, 78, 156, 231, 47, 187, 202, 217, 94, 246, 141, 63, 232,
+ 172, 110, 247, 157, 79, 218, 203, 126, 233, 188, 248, 95, 173, 142, 219,
+ 111, 249, 234, 158, 127, 189, 204, 250, 235, 143, 174, 220, 205, 159, 251,
190, 221, 175, 236, 237, 191, 206, 252, 222, 253, 207, 238, 223, 254, 239,
255,
};
DECLARE_ALIGNED(16, static const int16_t, col_scan_16x16[256]) = {
- 0, 16, 32, 48, 1, 64, 17, 80, 33, 96, 49, 2, 65, 112, 18, 81,
- 34, 128, 50, 97, 3, 66, 144, 19, 113, 35, 82, 160, 98, 51, 129, 4,
- 67, 176, 20, 114, 145, 83, 36, 99, 130, 52, 192, 5, 161, 68, 115, 21,
- 146, 84, 208, 177, 37, 131, 100, 53, 162, 224, 69, 6, 116, 193, 147, 85,
- 22, 240, 132, 38, 178, 101, 163, 54, 209, 117, 70, 7, 148, 194, 86, 179,
- 225, 23, 133, 39, 164, 8, 102, 210, 241, 55, 195, 118, 149, 71, 180, 24,
- 87, 226, 134, 165, 211, 40, 103, 56, 72, 150, 196, 242, 119, 9, 181, 227,
- 88, 166, 25, 135, 41, 104, 212, 57, 151, 197, 120, 73, 243, 182, 136, 167,
- 213, 89, 10, 228, 105, 152, 198, 26, 42, 121, 183, 244, 168, 58, 137, 229,
- 74, 214, 90, 153, 199, 184, 11, 106, 245, 27, 122, 230, 169, 43, 215, 59,
- 200, 138, 185, 246, 75, 12, 91, 154, 216, 231, 107, 28, 44, 201, 123, 170,
- 60, 247, 232, 76, 139, 13, 92, 217, 186, 248, 155, 108, 29, 124, 45, 202,
- 233, 171, 61, 14, 77, 140, 15, 249, 93, 30, 187, 156, 218, 46, 109, 125,
- 62, 172, 78, 203, 31, 141, 234, 94, 47, 188, 63, 157, 110, 250, 219, 79,
- 126, 204, 173, 142, 95, 189, 111, 235, 158, 220, 251, 127, 174, 143, 205,
- 236,
+ 0, 16, 32, 48, 1, 64, 17, 80, 33, 96, 49, 2, 65, 112, 18,
+ 81, 34, 128, 50, 97, 3, 66, 144, 19, 113, 35, 82, 160, 98, 51,
+ 129, 4, 67, 176, 20, 114, 145, 83, 36, 99, 130, 52, 192, 5, 161,
+ 68, 115, 21, 146, 84, 208, 177, 37, 131, 100, 53, 162, 224, 69, 6,
+ 116, 193, 147, 85, 22, 240, 132, 38, 178, 101, 163, 54, 209, 117, 70,
+ 7, 148, 194, 86, 179, 225, 23, 133, 39, 164, 8, 102, 210, 241, 55,
+ 195, 118, 149, 71, 180, 24, 87, 226, 134, 165, 211, 40, 103, 56, 72,
+ 150, 196, 242, 119, 9, 181, 227, 88, 166, 25, 135, 41, 104, 212, 57,
+ 151, 197, 120, 73, 243, 182, 136, 167, 213, 89, 10, 228, 105, 152, 198,
+ 26, 42, 121, 183, 244, 168, 58, 137, 229, 74, 214, 90, 153, 199, 184,
+ 11, 106, 245, 27, 122, 230, 169, 43, 215, 59, 200, 138, 185, 246, 75,
+ 12, 91, 154, 216, 231, 107, 28, 44, 201, 123, 170, 60, 247, 232, 76,
+ 139, 13, 92, 217, 186, 248, 155, 108, 29, 124, 45, 202, 233, 171, 61,
+ 14, 77, 140, 15, 249, 93, 30, 187, 156, 218, 46, 109, 125, 62, 172,
+ 78, 203, 31, 141, 234, 94, 47, 188, 63, 157, 110, 250, 219, 79, 126,
+ 204, 173, 142, 95, 189, 111, 235, 158, 220, 251, 127, 174, 143, 205, 236,
159, 190, 221, 252, 175, 206, 237, 191, 253, 222, 238, 207, 254, 223, 239,
255,
};
DECLARE_ALIGNED(16, static const int16_t, row_scan_16x16[256]) = {
- 0, 1, 2, 16, 3, 17, 4, 18, 32, 5, 33, 19, 6, 34, 48, 20,
- 49, 7, 35, 21, 50, 64, 8, 36, 65, 22, 51, 37, 80, 9, 66, 52,
- 23, 38, 81, 67, 10, 53, 24, 82, 68, 96, 39, 11, 54, 83, 97, 69,
- 25, 98, 84, 40, 112, 55, 12, 70, 99, 113, 85, 26, 41, 56, 114, 100,
- 13, 71, 128, 86, 27, 115, 101, 129, 42, 57, 72, 116, 14, 87, 130, 102,
- 144, 73, 131, 117, 28, 58, 15, 88, 43, 145, 103, 132, 146, 118, 74, 160,
- 89, 133, 104, 29, 59, 147, 119, 44, 161, 148, 90, 105, 134, 162, 120, 176,
- 75, 135, 149, 30, 60, 163, 177, 45, 121, 91, 106, 164, 178, 150, 192, 136,
- 165, 179, 31, 151, 193, 76, 122, 61, 137, 194, 107, 152, 180, 208, 46, 166,
- 167, 195, 92, 181, 138, 209, 123, 153, 224, 196, 77, 168, 210, 182, 240, 108,
- 197, 62, 154, 225, 183, 169, 211, 47, 139, 93, 184, 226, 212, 241, 198, 170,
- 124, 155, 199, 78, 213, 185, 109, 227, 200, 63, 228, 242, 140, 214, 171, 186,
- 156, 229, 243, 125, 94, 201, 244, 215, 216, 230, 141, 187, 202, 79, 172, 110,
- 157, 245, 217, 231, 95, 246, 232, 126, 203, 247, 233, 173, 218, 142, 111,
- 158,
- 188, 248, 127, 234, 219, 249, 189, 204, 143, 174, 159, 250, 235, 205, 220,
- 175,
+ 0, 1, 2, 16, 3, 17, 4, 18, 32, 5, 33, 19, 6, 34, 48,
+ 20, 49, 7, 35, 21, 50, 64, 8, 36, 65, 22, 51, 37, 80, 9,
+ 66, 52, 23, 38, 81, 67, 10, 53, 24, 82, 68, 96, 39, 11, 54,
+ 83, 97, 69, 25, 98, 84, 40, 112, 55, 12, 70, 99, 113, 85, 26,
+ 41, 56, 114, 100, 13, 71, 128, 86, 27, 115, 101, 129, 42, 57, 72,
+ 116, 14, 87, 130, 102, 144, 73, 131, 117, 28, 58, 15, 88, 43, 145,
+ 103, 132, 146, 118, 74, 160, 89, 133, 104, 29, 59, 147, 119, 44, 161,
+ 148, 90, 105, 134, 162, 120, 176, 75, 135, 149, 30, 60, 163, 177, 45,
+ 121, 91, 106, 164, 178, 150, 192, 136, 165, 179, 31, 151, 193, 76, 122,
+ 61, 137, 194, 107, 152, 180, 208, 46, 166, 167, 195, 92, 181, 138, 209,
+ 123, 153, 224, 196, 77, 168, 210, 182, 240, 108, 197, 62, 154, 225, 183,
+ 169, 211, 47, 139, 93, 184, 226, 212, 241, 198, 170, 124, 155, 199, 78,
+ 213, 185, 109, 227, 200, 63, 228, 242, 140, 214, 171, 186, 156, 229, 243,
+ 125, 94, 201, 244, 215, 216, 230, 141, 187, 202, 79, 172, 110, 157, 245,
+ 217, 231, 95, 246, 232, 126, 203, 247, 233, 173, 218, 142, 111, 158, 188,
+ 248, 127, 234, 219, 249, 189, 204, 143, 174, 159, 250, 235, 205, 220, 175,
190, 251, 221, 191, 206, 236, 207, 237, 252, 222, 253, 223, 238, 239, 254,
255,
};
DECLARE_ALIGNED(16, static const int16_t, default_scan_32x32[1024]) = {
- 0, 32, 1, 64, 33, 2, 96, 65, 34, 128, 3, 97, 66, 160,
- 129, 35, 98, 4, 67, 130, 161, 192, 36, 99, 224, 5, 162, 193,
- 68, 131, 37, 100,
- 225, 194, 256, 163, 69, 132, 6, 226, 257, 288, 195, 101, 164, 38,
- 258, 7, 227, 289, 133, 320, 70, 196, 165, 290, 259, 228, 39, 321,
- 102, 352, 8, 197,
- 71, 134, 322, 291, 260, 353, 384, 229, 166, 103, 40, 354, 323, 292,
- 135, 385, 198, 261, 72, 9, 416, 167, 386, 355, 230, 324, 104, 293,
- 41, 417, 199, 136,
- 262, 387, 448, 325, 356, 10, 73, 418, 231, 168, 449, 294, 388, 105,
- 419, 263, 42, 200, 357, 450, 137, 480, 74, 326, 232, 11, 389, 169,
- 295, 420, 106, 451,
- 481, 358, 264, 327, 201, 43, 138, 512, 482, 390, 296, 233, 170, 421,
- 75, 452, 359, 12, 513, 265, 483, 328, 107, 202, 514, 544, 422, 391,
- 453, 139, 44, 234,
- 484, 297, 360, 171, 76, 515, 545, 266, 329, 454, 13, 423, 203, 108,
- 546, 485, 576, 298, 235, 140, 361, 330, 172, 547, 45, 455, 267, 577,
- 486, 77, 204, 362,
- 608, 14, 299, 578, 109, 236, 487, 609, 331, 141, 579, 46, 15, 173,
- 610, 363, 78, 205, 16, 110, 237, 611, 142, 47, 174, 79, 206, 17,
- 111, 238, 48, 143,
- 80, 175, 112, 207, 49, 18, 239, 81, 113, 19, 50, 82, 114, 51,
- 83, 115, 640, 516, 392, 268, 144, 20, 672, 641, 548, 517, 424,
- 393, 300, 269, 176, 145,
- 52, 21, 704, 673, 642, 580, 549, 518, 456, 425, 394, 332, 301,
- 270, 208, 177, 146, 84, 53, 22, 736, 705, 674, 643, 612, 581,
- 550, 519, 488, 457, 426, 395,
- 364, 333, 302, 271, 240, 209, 178, 147, 116, 85, 54, 23, 737,
- 706, 675, 613, 582, 551, 489, 458, 427, 365, 334, 303, 241,
- 210, 179, 117, 86, 55, 738, 707,
- 614, 583, 490, 459, 366, 335, 242, 211, 118, 87, 739, 615, 491,
- 367, 243, 119, 768, 644, 520, 396, 272, 148, 24, 800, 769, 676,
- 645, 552, 521, 428, 397, 304,
- 273, 180, 149, 56, 25, 832, 801, 770, 708, 677, 646, 584, 553,
- 522, 460, 429, 398, 336, 305, 274, 212, 181, 150, 88, 57, 26,
- 864, 833, 802, 771, 740, 709,
- 678, 647, 616, 585, 554, 523, 492, 461, 430, 399, 368, 337, 306,
- 275, 244, 213, 182, 151, 120, 89, 58, 27, 865, 834, 803, 741,
- 710, 679, 617, 586, 555, 493,
- 462, 431, 369, 338, 307, 245, 214, 183, 121, 90, 59, 866, 835,
- 742, 711, 618, 587, 494, 463, 370, 339, 246, 215, 122, 91, 867,
- 743, 619, 495, 371, 247, 123,
- 896, 772, 648, 524, 400, 276, 152, 28, 928, 897, 804, 773, 680,
- 649, 556, 525, 432, 401, 308, 277, 184, 153, 60, 29, 960, 929,
- 898, 836, 805, 774, 712, 681,
- 650, 588, 557, 526, 464, 433, 402, 340, 309, 278, 216, 185, 154,
- 92, 61, 30, 992, 961, 930, 899, 868, 837, 806, 775, 744, 713, 682,
- 651, 620, 589, 558, 527,
- 496, 465, 434, 403, 372, 341, 310, 279, 248, 217, 186, 155, 124,
- 93, 62, 31, 993, 962, 931, 869, 838, 807, 745, 714, 683, 621, 590,
- 559, 497, 466, 435, 373,
- 342, 311, 249, 218, 187, 125, 94, 63, 994, 963, 870, 839, 746, 715,
- 622, 591, 498, 467, 374, 343, 250, 219, 126, 95, 995, 871, 747, 623,
- 499, 375, 251, 127,
- 900, 776, 652, 528, 404, 280, 156, 932, 901, 808, 777, 684, 653, 560,
- 529, 436, 405, 312, 281, 188, 157, 964, 933, 902, 840, 809, 778, 716,
- 685, 654, 592, 561,
- 530, 468, 437, 406, 344, 313, 282, 220, 189, 158, 996, 965, 934, 903,
- 872, 841, 810, 779, 748, 717, 686, 655, 624, 593, 562, 531, 500, 469,
- 438, 407, 376, 345,
- 314, 283, 252, 221, 190, 159, 997, 966, 935, 873, 842, 811, 749, 718,
- 687, 625, 594, 563, 501, 470, 439, 377, 346, 315, 253, 222, 191, 998,
- 967, 874, 843, 750,
- 719, 626, 595, 502, 471, 378, 347, 254, 223, 999, 875, 751, 627, 503,
- 379, 255, 904, 780, 656, 532, 408, 284, 936, 905, 812, 781, 688, 657,
- 564, 533, 440, 409,
- 316, 285, 968, 937, 906, 844, 813, 782, 720, 689, 658, 596, 565, 534,
- 472, 441, 410, 348, 317, 286, 1000, 969, 938, 907, 876, 845, 814, 783,
- 752, 721, 690, 659,
- 628, 597, 566, 535, 504, 473, 442, 411, 380, 349, 318, 287, 1001, 970,
- 939, 877, 846, 815, 753, 722, 691, 629, 598, 567, 505, 474, 443, 381,
- 350, 319, 1002, 971,
- 878, 847, 754, 723, 630, 599, 506, 475, 382, 351, 1003, 879, 755, 631,
- 507, 383, 908, 784, 660, 536, 412, 940, 909, 816, 785, 692, 661, 568,
- 537, 444, 413, 972,
- 941, 910, 848, 817, 786, 724, 693, 662, 600, 569, 538, 476, 445, 414,
- 1004, 973, 942, 911, 880, 849, 818, 787, 756, 725, 694, 663, 632, 601,
- 570, 539, 508, 477,
- 446, 415, 1005, 974, 943, 881, 850, 819, 757, 726, 695, 633, 602, 571,
- 509, 478, 447, 1006, 975, 882, 851, 758, 727, 634, 603, 510, 479,
- 1007, 883, 759, 635, 511,
- 912, 788, 664, 540, 944, 913, 820, 789, 696, 665, 572, 541, 976, 945,
- 914, 852, 821, 790, 728, 697, 666, 604, 573, 542, 1008, 977, 946, 915,
- 884, 853, 822, 791,
- 760, 729, 698, 667, 636, 605, 574, 543, 1009, 978, 947, 885, 854, 823,
- 761, 730, 699, 637, 606, 575, 1010, 979, 886, 855, 762, 731, 638, 607,
- 1011, 887, 763, 639,
- 916, 792, 668, 948, 917, 824, 793, 700, 669, 980, 949, 918, 856, 825,
- 794, 732, 701, 670, 1012, 981, 950, 919, 888, 857, 826, 795, 764, 733,
- 702, 671, 1013, 982,
- 951, 889, 858, 827, 765, 734, 703, 1014, 983, 890, 859, 766, 735, 1015,
- 891, 767, 920, 796, 952, 921, 828, 797, 984, 953, 922, 860, 829, 798,
- 1016, 985, 954, 923,
- 892, 861, 830, 799, 1017, 986, 955, 893, 862, 831, 1018, 987, 894, 863,
- 1019, 895, 924, 956, 925, 988, 957, 926, 1020, 989, 958, 927, 1021,
- 990, 959, 1022, 991, 1023,
+ 0, 32, 1, 64, 33, 2, 96, 65, 34, 128, 3, 97, 66,
+ 160, 129, 35, 98, 4, 67, 130, 161, 192, 36, 99, 224, 5,
+ 162, 193, 68, 131, 37, 100, 225, 194, 256, 163, 69, 132, 6,
+ 226, 257, 288, 195, 101, 164, 38, 258, 7, 227, 289, 133, 320,
+ 70, 196, 165, 290, 259, 228, 39, 321, 102, 352, 8, 197, 71,
+ 134, 322, 291, 260, 353, 384, 229, 166, 103, 40, 354, 323, 292,
+ 135, 385, 198, 261, 72, 9, 416, 167, 386, 355, 230, 324, 104,
+ 293, 41, 417, 199, 136, 262, 387, 448, 325, 356, 10, 73, 418,
+ 231, 168, 449, 294, 388, 105, 419, 263, 42, 200, 357, 450, 137,
+ 480, 74, 326, 232, 11, 389, 169, 295, 420, 106, 451, 481, 358,
+ 264, 327, 201, 43, 138, 512, 482, 390, 296, 233, 170, 421, 75,
+ 452, 359, 12, 513, 265, 483, 328, 107, 202, 514, 544, 422, 391,
+ 453, 139, 44, 234, 484, 297, 360, 171, 76, 515, 545, 266, 329,
+ 454, 13, 423, 203, 108, 546, 485, 576, 298, 235, 140, 361, 330,
+ 172, 547, 45, 455, 267, 577, 486, 77, 204, 362, 608, 14, 299,
+ 578, 109, 236, 487, 609, 331, 141, 579, 46, 15, 173, 610, 363,
+ 78, 205, 16, 110, 237, 611, 142, 47, 174, 79, 206, 17, 111,
+ 238, 48, 143, 80, 175, 112, 207, 49, 18, 239, 81, 113, 19,
+ 50, 82, 114, 51, 83, 115, 640, 516, 392, 268, 144, 20, 672,
+ 641, 548, 517, 424, 393, 300, 269, 176, 145, 52, 21, 704, 673,
+ 642, 580, 549, 518, 456, 425, 394, 332, 301, 270, 208, 177, 146,
+ 84, 53, 22, 736, 705, 674, 643, 612, 581, 550, 519, 488, 457,
+ 426, 395, 364, 333, 302, 271, 240, 209, 178, 147, 116, 85, 54,
+ 23, 737, 706, 675, 613, 582, 551, 489, 458, 427, 365, 334, 303,
+ 241, 210, 179, 117, 86, 55, 738, 707, 614, 583, 490, 459, 366,
+ 335, 242, 211, 118, 87, 739, 615, 491, 367, 243, 119, 768, 644,
+ 520, 396, 272, 148, 24, 800, 769, 676, 645, 552, 521, 428, 397,
+ 304, 273, 180, 149, 56, 25, 832, 801, 770, 708, 677, 646, 584,
+ 553, 522, 460, 429, 398, 336, 305, 274, 212, 181, 150, 88, 57,
+ 26, 864, 833, 802, 771, 740, 709, 678, 647, 616, 585, 554, 523,
+ 492, 461, 430, 399, 368, 337, 306, 275, 244, 213, 182, 151, 120,
+ 89, 58, 27, 865, 834, 803, 741, 710, 679, 617, 586, 555, 493,
+ 462, 431, 369, 338, 307, 245, 214, 183, 121, 90, 59, 866, 835,
+ 742, 711, 618, 587, 494, 463, 370, 339, 246, 215, 122, 91, 867,
+ 743, 619, 495, 371, 247, 123, 896, 772, 648, 524, 400, 276, 152,
+ 28, 928, 897, 804, 773, 680, 649, 556, 525, 432, 401, 308, 277,
+ 184, 153, 60, 29, 960, 929, 898, 836, 805, 774, 712, 681, 650,
+ 588, 557, 526, 464, 433, 402, 340, 309, 278, 216, 185, 154, 92,
+ 61, 30, 992, 961, 930, 899, 868, 837, 806, 775, 744, 713, 682,
+ 651, 620, 589, 558, 527, 496, 465, 434, 403, 372, 341, 310, 279,
+ 248, 217, 186, 155, 124, 93, 62, 31, 993, 962, 931, 869, 838,
+ 807, 745, 714, 683, 621, 590, 559, 497, 466, 435, 373, 342, 311,
+ 249, 218, 187, 125, 94, 63, 994, 963, 870, 839, 746, 715, 622,
+ 591, 498, 467, 374, 343, 250, 219, 126, 95, 995, 871, 747, 623,
+ 499, 375, 251, 127, 900, 776, 652, 528, 404, 280, 156, 932, 901,
+ 808, 777, 684, 653, 560, 529, 436, 405, 312, 281, 188, 157, 964,
+ 933, 902, 840, 809, 778, 716, 685, 654, 592, 561, 530, 468, 437,
+ 406, 344, 313, 282, 220, 189, 158, 996, 965, 934, 903, 872, 841,
+ 810, 779, 748, 717, 686, 655, 624, 593, 562, 531, 500, 469, 438,
+ 407, 376, 345, 314, 283, 252, 221, 190, 159, 997, 966, 935, 873,
+ 842, 811, 749, 718, 687, 625, 594, 563, 501, 470, 439, 377, 346,
+ 315, 253, 222, 191, 998, 967, 874, 843, 750, 719, 626, 595, 502,
+ 471, 378, 347, 254, 223, 999, 875, 751, 627, 503, 379, 255, 904,
+ 780, 656, 532, 408, 284, 936, 905, 812, 781, 688, 657, 564, 533,
+ 440, 409, 316, 285, 968, 937, 906, 844, 813, 782, 720, 689, 658,
+ 596, 565, 534, 472, 441, 410, 348, 317, 286, 1000, 969, 938, 907,
+ 876, 845, 814, 783, 752, 721, 690, 659, 628, 597, 566, 535, 504,
+ 473, 442, 411, 380, 349, 318, 287, 1001, 970, 939, 877, 846, 815,
+ 753, 722, 691, 629, 598, 567, 505, 474, 443, 381, 350, 319, 1002,
+ 971, 878, 847, 754, 723, 630, 599, 506, 475, 382, 351, 1003, 879,
+ 755, 631, 507, 383, 908, 784, 660, 536, 412, 940, 909, 816, 785,
+ 692, 661, 568, 537, 444, 413, 972, 941, 910, 848, 817, 786, 724,
+ 693, 662, 600, 569, 538, 476, 445, 414, 1004, 973, 942, 911, 880,
+ 849, 818, 787, 756, 725, 694, 663, 632, 601, 570, 539, 508, 477,
+ 446, 415, 1005, 974, 943, 881, 850, 819, 757, 726, 695, 633, 602,
+ 571, 509, 478, 447, 1006, 975, 882, 851, 758, 727, 634, 603, 510,
+ 479, 1007, 883, 759, 635, 511, 912, 788, 664, 540, 944, 913, 820,
+ 789, 696, 665, 572, 541, 976, 945, 914, 852, 821, 790, 728, 697,
+ 666, 604, 573, 542, 1008, 977, 946, 915, 884, 853, 822, 791, 760,
+ 729, 698, 667, 636, 605, 574, 543, 1009, 978, 947, 885, 854, 823,
+ 761, 730, 699, 637, 606, 575, 1010, 979, 886, 855, 762, 731, 638,
+ 607, 1011, 887, 763, 639, 916, 792, 668, 948, 917, 824, 793, 700,
+ 669, 980, 949, 918, 856, 825, 794, 732, 701, 670, 1012, 981, 950,
+ 919, 888, 857, 826, 795, 764, 733, 702, 671, 1013, 982, 951, 889,
+ 858, 827, 765, 734, 703, 1014, 983, 890, 859, 766, 735, 1015, 891,
+ 767, 920, 796, 952, 921, 828, 797, 984, 953, 922, 860, 829, 798,
+ 1016, 985, 954, 923, 892, 861, 830, 799, 1017, 986, 955, 893, 862,
+ 831, 1018, 987, 894, 863, 1019, 895, 924, 956, 925, 988, 957, 926,
+ 1020, 989, 958, 927, 1021, 990, 959, 1022, 991, 1023,
};
// Neighborhood 5-tuples for various scans and blocksizes,
@@ -235,301 +196,321 @@
// -1 indicates the neighbor does not exist.
DECLARE_ALIGNED(16, static const int16_t,
default_scan_4x4_neighbors[17 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 0, 0, 1, 4, 4, 4, 1, 1, 8, 8, 5, 8, 2, 2, 2, 5, 9, 12, 6, 9,
- 3, 6, 10, 13, 7, 10, 11, 14, 0, 0,
+ 0, 0, 0, 0, 0, 0, 1, 4, 4, 4, 1, 1, 8, 8, 5, 8, 2,
+ 2, 2, 5, 9, 12, 6, 9, 3, 6, 10, 13, 7, 10, 11, 14, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
col_scan_4x4_neighbors[17 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 4, 4, 0, 0, 8, 8, 1, 1, 5, 5, 1, 1, 9, 9, 2, 2, 6, 6, 2, 2, 3,
- 3, 10, 10, 7, 7, 11, 11, 0, 0,
+ 0, 0, 0, 0, 4, 4, 0, 0, 8, 8, 1, 1, 5, 5, 1, 1, 9,
+ 9, 2, 2, 6, 6, 2, 2, 3, 3, 10, 10, 7, 7, 11, 11, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
row_scan_4x4_neighbors[17 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 0, 0, 1, 1, 4, 4, 2, 2, 5, 5, 4, 4, 8, 8, 6, 6, 8, 8, 9, 9, 12,
- 12, 10, 10, 13, 13, 14, 14, 0, 0,
+ 0, 0, 0, 0, 0, 0, 1, 1, 4, 4, 2, 2, 5, 5, 4, 4, 8,
+ 8, 6, 6, 8, 8, 9, 9, 12, 12, 10, 10, 13, 13, 14, 14, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
col_scan_8x8_neighbors[65 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 8, 8, 0, 0, 16, 16, 1, 1, 24, 24, 9, 9, 1, 1, 32, 32, 17, 17, 2,
- 2, 25, 25, 10, 10, 40, 40, 2, 2, 18, 18, 33, 33, 3, 3, 48, 48, 11, 11, 26,
- 26, 3, 3, 41, 41, 19, 19, 34, 34, 4, 4, 27, 27, 12, 12, 49, 49, 42, 42, 20,
- 20, 4, 4, 35, 35, 5, 5, 28, 28, 50, 50, 43, 43, 13, 13, 36, 36, 5, 5, 21, 21,
- 51, 51, 29, 29, 6, 6, 44, 44, 14, 14, 6, 6, 37, 37, 52, 52, 22, 22, 7, 7, 30,
- 30, 45, 45, 15, 15, 38, 38, 23, 23, 53, 53, 31, 31, 46, 46, 39, 39, 54, 54,
- 47, 47, 55, 55, 0, 0,
+ 0, 0, 0, 0, 8, 8, 0, 0, 16, 16, 1, 1, 24, 24, 9, 9, 1, 1, 32,
+ 32, 17, 17, 2, 2, 25, 25, 10, 10, 40, 40, 2, 2, 18, 18, 33, 33, 3, 3,
+ 48, 48, 11, 11, 26, 26, 3, 3, 41, 41, 19, 19, 34, 34, 4, 4, 27, 27, 12,
+ 12, 49, 49, 42, 42, 20, 20, 4, 4, 35, 35, 5, 5, 28, 28, 50, 50, 43, 43,
+ 13, 13, 36, 36, 5, 5, 21, 21, 51, 51, 29, 29, 6, 6, 44, 44, 14, 14, 6,
+ 6, 37, 37, 52, 52, 22, 22, 7, 7, 30, 30, 45, 45, 15, 15, 38, 38, 23, 23,
+ 53, 53, 31, 31, 46, 46, 39, 39, 54, 54, 47, 47, 55, 55, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
row_scan_8x8_neighbors[65 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 1, 1, 0, 0, 8, 8, 2, 2, 8, 8, 9, 9, 3, 3, 16, 16, 10, 10, 16, 16,
- 4, 4, 17, 17, 24, 24, 11, 11, 18, 18, 25, 25, 24, 24, 5, 5, 12, 12, 19, 19,
- 32, 32, 26, 26, 6, 6, 33, 33, 32, 32, 20, 20, 27, 27, 40, 40, 13, 13, 34, 34,
- 40, 40, 41, 41, 28, 28, 35, 35, 48, 48, 21, 21, 42, 42, 14, 14, 48, 48, 36,
- 36, 49, 49, 43, 43, 29, 29, 56, 56, 22, 22, 50, 50, 57, 57, 44, 44, 37, 37,
- 51, 51, 30, 30, 58, 58, 52, 52, 45, 45, 59, 59, 38, 38, 60, 60, 46, 46, 53,
- 53, 54, 54, 61, 61, 62, 62, 0, 0,
+ 0, 0, 0, 0, 1, 1, 0, 0, 8, 8, 2, 2, 8, 8, 9, 9, 3, 3, 16,
+ 16, 10, 10, 16, 16, 4, 4, 17, 17, 24, 24, 11, 11, 18, 18, 25, 25, 24, 24,
+ 5, 5, 12, 12, 19, 19, 32, 32, 26, 26, 6, 6, 33, 33, 32, 32, 20, 20, 27,
+ 27, 40, 40, 13, 13, 34, 34, 40, 40, 41, 41, 28, 28, 35, 35, 48, 48, 21, 21,
+ 42, 42, 14, 14, 48, 48, 36, 36, 49, 49, 43, 43, 29, 29, 56, 56, 22, 22, 50,
+ 50, 57, 57, 44, 44, 37, 37, 51, 51, 30, 30, 58, 58, 52, 52, 45, 45, 59, 59,
+ 38, 38, 60, 60, 46, 46, 53, 53, 54, 54, 61, 61, 62, 62, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
default_scan_8x8_neighbors[65 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 0, 0, 8, 8, 1, 8, 1, 1, 9, 16, 16, 16, 2, 9, 2, 2, 10, 17, 17,
- 24, 24, 24, 3, 10, 3, 3, 18, 25, 25, 32, 11, 18, 32, 32, 4, 11, 26, 33, 19,
- 26, 4, 4, 33, 40, 12, 19, 40, 40, 5, 12, 27, 34, 34, 41, 20, 27, 13, 20, 5,
- 5, 41, 48, 48, 48, 28, 35, 35, 42, 21, 28, 6, 6, 6, 13, 42, 49, 49, 56, 36,
- 43, 14, 21, 29, 36, 7, 14, 43, 50, 50, 57, 22, 29, 37, 44, 15, 22, 44, 51,
- 51, 58, 30, 37, 23, 30, 52, 59, 45, 52, 38, 45, 31, 38, 53, 60, 46, 53, 39,
- 46, 54, 61, 47, 54, 55, 62, 0, 0,
+ 0, 0, 0, 0, 0, 0, 8, 8, 1, 8, 1, 1, 9, 16, 16, 16, 2, 9, 2,
+ 2, 10, 17, 17, 24, 24, 24, 3, 10, 3, 3, 18, 25, 25, 32, 11, 18, 32, 32,
+ 4, 11, 26, 33, 19, 26, 4, 4, 33, 40, 12, 19, 40, 40, 5, 12, 27, 34, 34,
+ 41, 20, 27, 13, 20, 5, 5, 41, 48, 48, 48, 28, 35, 35, 42, 21, 28, 6, 6,
+ 6, 13, 42, 49, 49, 56, 36, 43, 14, 21, 29, 36, 7, 14, 43, 50, 50, 57, 22,
+ 29, 37, 44, 15, 22, 44, 51, 51, 58, 30, 37, 23, 30, 52, 59, 45, 52, 38, 45,
+ 31, 38, 53, 60, 46, 53, 39, 46, 54, 61, 47, 54, 55, 62, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
col_scan_16x16_neighbors[257 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 16, 16, 32, 32, 0, 0, 48, 48, 1, 1, 64, 64,
- 17, 17, 80, 80, 33, 33, 1, 1, 49, 49, 96, 96, 2, 2, 65, 65,
- 18, 18, 112, 112, 34, 34, 81, 81, 2, 2, 50, 50, 128, 128, 3, 3,
- 97, 97, 19, 19, 66, 66, 144, 144, 82, 82, 35, 35, 113, 113, 3, 3,
- 51, 51, 160, 160, 4, 4, 98, 98, 129, 129, 67, 67, 20, 20, 83, 83,
- 114, 114, 36, 36, 176, 176, 4, 4, 145, 145, 52, 52, 99, 99, 5, 5,
- 130, 130, 68, 68, 192, 192, 161, 161, 21, 21, 115, 115, 84, 84, 37, 37,
- 146, 146, 208, 208, 53, 53, 5, 5, 100, 100, 177, 177, 131, 131, 69, 69,
- 6, 6, 224, 224, 116, 116, 22, 22, 162, 162, 85, 85, 147, 147, 38, 38,
- 193, 193, 101, 101, 54, 54, 6, 6, 132, 132, 178, 178, 70, 70, 163, 163,
- 209, 209, 7, 7, 117, 117, 23, 23, 148, 148, 7, 7, 86, 86, 194, 194,
- 225, 225, 39, 39, 179, 179, 102, 102, 133, 133, 55, 55, 164, 164, 8, 8,
- 71, 71, 210, 210, 118, 118, 149, 149, 195, 195, 24, 24, 87, 87, 40, 40,
- 56, 56, 134, 134, 180, 180, 226, 226, 103, 103, 8, 8, 165, 165, 211, 211,
- 72, 72, 150, 150, 9, 9, 119, 119, 25, 25, 88, 88, 196, 196, 41, 41,
- 135, 135, 181, 181, 104, 104, 57, 57, 227, 227, 166, 166, 120, 120, 151, 151,
- 197, 197, 73, 73, 9, 9, 212, 212, 89, 89, 136, 136, 182, 182, 10, 10,
- 26, 26, 105, 105, 167, 167, 228, 228, 152, 152, 42, 42, 121, 121, 213, 213,
- 58, 58, 198, 198, 74, 74, 137, 137, 183, 183, 168, 168, 10, 10, 90, 90,
- 229, 229, 11, 11, 106, 106, 214, 214, 153, 153, 27, 27, 199, 199, 43, 43,
- 184, 184, 122, 122, 169, 169, 230, 230, 59, 59, 11, 11, 75, 75, 138, 138,
- 200, 200, 215, 215, 91, 91, 12, 12, 28, 28, 185, 185, 107, 107, 154, 154,
- 44, 44, 231, 231, 216, 216, 60, 60, 123, 123, 12, 12, 76, 76, 201, 201,
- 170, 170, 232, 232, 139, 139, 92, 92, 13, 13, 108, 108, 29, 29, 186, 186,
- 217, 217, 155, 155, 45, 45, 13, 13, 61, 61, 124, 124, 14, 14, 233, 233,
- 77, 77, 14, 14, 171, 171, 140, 140, 202, 202, 30, 30, 93, 93, 109, 109,
- 46, 46, 156, 156, 62, 62, 187, 187, 15, 15, 125, 125, 218, 218, 78, 78,
- 31, 31, 172, 172, 47, 47, 141, 141, 94, 94, 234, 234, 203, 203, 63, 63,
- 110, 110, 188, 188, 157, 157, 126, 126, 79, 79, 173, 173, 95, 95, 219, 219,
- 142, 142, 204, 204, 235, 235, 111, 111, 158, 158, 127, 127, 189, 189, 220,
- 220, 143, 143, 174, 174, 205, 205, 236, 236, 159, 159, 190, 190, 221, 221,
- 175, 175, 237, 237, 206, 206, 222, 222, 191, 191, 238, 238, 207, 207, 223,
- 223, 239, 239, 0, 0,
+ 0, 0, 0, 0, 16, 16, 32, 32, 0, 0, 48, 48, 1, 1, 64,
+ 64, 17, 17, 80, 80, 33, 33, 1, 1, 49, 49, 96, 96, 2, 2,
+ 65, 65, 18, 18, 112, 112, 34, 34, 81, 81, 2, 2, 50, 50, 128,
+ 128, 3, 3, 97, 97, 19, 19, 66, 66, 144, 144, 82, 82, 35, 35,
+ 113, 113, 3, 3, 51, 51, 160, 160, 4, 4, 98, 98, 129, 129, 67,
+ 67, 20, 20, 83, 83, 114, 114, 36, 36, 176, 176, 4, 4, 145, 145,
+ 52, 52, 99, 99, 5, 5, 130, 130, 68, 68, 192, 192, 161, 161, 21,
+ 21, 115, 115, 84, 84, 37, 37, 146, 146, 208, 208, 53, 53, 5, 5,
+ 100, 100, 177, 177, 131, 131, 69, 69, 6, 6, 224, 224, 116, 116, 22,
+ 22, 162, 162, 85, 85, 147, 147, 38, 38, 193, 193, 101, 101, 54, 54,
+ 6, 6, 132, 132, 178, 178, 70, 70, 163, 163, 209, 209, 7, 7, 117,
+ 117, 23, 23, 148, 148, 7, 7, 86, 86, 194, 194, 225, 225, 39, 39,
+ 179, 179, 102, 102, 133, 133, 55, 55, 164, 164, 8, 8, 71, 71, 210,
+ 210, 118, 118, 149, 149, 195, 195, 24, 24, 87, 87, 40, 40, 56, 56,
+ 134, 134, 180, 180, 226, 226, 103, 103, 8, 8, 165, 165, 211, 211, 72,
+ 72, 150, 150, 9, 9, 119, 119, 25, 25, 88, 88, 196, 196, 41, 41,
+ 135, 135, 181, 181, 104, 104, 57, 57, 227, 227, 166, 166, 120, 120, 151,
+ 151, 197, 197, 73, 73, 9, 9, 212, 212, 89, 89, 136, 136, 182, 182,
+ 10, 10, 26, 26, 105, 105, 167, 167, 228, 228, 152, 152, 42, 42, 121,
+ 121, 213, 213, 58, 58, 198, 198, 74, 74, 137, 137, 183, 183, 168, 168,
+ 10, 10, 90, 90, 229, 229, 11, 11, 106, 106, 214, 214, 153, 153, 27,
+ 27, 199, 199, 43, 43, 184, 184, 122, 122, 169, 169, 230, 230, 59, 59,
+ 11, 11, 75, 75, 138, 138, 200, 200, 215, 215, 91, 91, 12, 12, 28,
+ 28, 185, 185, 107, 107, 154, 154, 44, 44, 231, 231, 216, 216, 60, 60,
+ 123, 123, 12, 12, 76, 76, 201, 201, 170, 170, 232, 232, 139, 139, 92,
+ 92, 13, 13, 108, 108, 29, 29, 186, 186, 217, 217, 155, 155, 45, 45,
+ 13, 13, 61, 61, 124, 124, 14, 14, 233, 233, 77, 77, 14, 14, 171,
+ 171, 140, 140, 202, 202, 30, 30, 93, 93, 109, 109, 46, 46, 156, 156,
+ 62, 62, 187, 187, 15, 15, 125, 125, 218, 218, 78, 78, 31, 31, 172,
+ 172, 47, 47, 141, 141, 94, 94, 234, 234, 203, 203, 63, 63, 110, 110,
+ 188, 188, 157, 157, 126, 126, 79, 79, 173, 173, 95, 95, 219, 219, 142,
+ 142, 204, 204, 235, 235, 111, 111, 158, 158, 127, 127, 189, 189, 220, 220,
+ 143, 143, 174, 174, 205, 205, 236, 236, 159, 159, 190, 190, 221, 221, 175,
+ 175, 237, 237, 206, 206, 222, 222, 191, 191, 238, 238, 207, 207, 223, 223,
+ 239, 239, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
row_scan_16x16_neighbors[257 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 1, 1, 0, 0, 2, 2, 16, 16, 3, 3, 17, 17,
- 16, 16, 4, 4, 32, 32, 18, 18, 5, 5, 33, 33, 32, 32, 19, 19,
- 48, 48, 6, 6, 34, 34, 20, 20, 49, 49, 48, 48, 7, 7, 35, 35,
- 64, 64, 21, 21, 50, 50, 36, 36, 64, 64, 8, 8, 65, 65, 51, 51,
- 22, 22, 37, 37, 80, 80, 66, 66, 9, 9, 52, 52, 23, 23, 81, 81,
- 67, 67, 80, 80, 38, 38, 10, 10, 53, 53, 82, 82, 96, 96, 68, 68,
- 24, 24, 97, 97, 83, 83, 39, 39, 96, 96, 54, 54, 11, 11, 69, 69,
- 98, 98, 112, 112, 84, 84, 25, 25, 40, 40, 55, 55, 113, 113, 99, 99,
- 12, 12, 70, 70, 112, 112, 85, 85, 26, 26, 114, 114, 100, 100, 128, 128,
- 41, 41, 56, 56, 71, 71, 115, 115, 13, 13, 86, 86, 129, 129, 101, 101,
- 128, 128, 72, 72, 130, 130, 116, 116, 27, 27, 57, 57, 14, 14, 87, 87,
- 42, 42, 144, 144, 102, 102, 131, 131, 145, 145, 117, 117, 73, 73, 144, 144,
- 88, 88, 132, 132, 103, 103, 28, 28, 58, 58, 146, 146, 118, 118, 43, 43,
- 160, 160, 147, 147, 89, 89, 104, 104, 133, 133, 161, 161, 119, 119, 160, 160,
- 74, 74, 134, 134, 148, 148, 29, 29, 59, 59, 162, 162, 176, 176, 44, 44,
- 120, 120, 90, 90, 105, 105, 163, 163, 177, 177, 149, 149, 176, 176, 135, 135,
- 164, 164, 178, 178, 30, 30, 150, 150, 192, 192, 75, 75, 121, 121, 60, 60,
- 136, 136, 193, 193, 106, 106, 151, 151, 179, 179, 192, 192, 45, 45, 165, 165,
- 166, 166, 194, 194, 91, 91, 180, 180, 137, 137, 208, 208, 122, 122, 152, 152,
- 208, 208, 195, 195, 76, 76, 167, 167, 209, 209, 181, 181, 224, 224, 107, 107,
- 196, 196, 61, 61, 153, 153, 224, 224, 182, 182, 168, 168, 210, 210, 46, 46,
- 138, 138, 92, 92, 183, 183, 225, 225, 211, 211, 240, 240, 197, 197, 169, 169,
- 123, 123, 154, 154, 198, 198, 77, 77, 212, 212, 184, 184, 108, 108, 226, 226,
- 199, 199, 62, 62, 227, 227, 241, 241, 139, 139, 213, 213, 170, 170, 185, 185,
- 155, 155, 228, 228, 242, 242, 124, 124, 93, 93, 200, 200, 243, 243, 214, 214,
- 215, 215, 229, 229, 140, 140, 186, 186, 201, 201, 78, 78, 171, 171, 109, 109,
- 156, 156, 244, 244, 216, 216, 230, 230, 94, 94, 245, 245, 231, 231, 125, 125,
- 202, 202, 246, 246, 232, 232, 172, 172, 217, 217, 141, 141, 110, 110, 157,
- 157, 187, 187, 247, 247, 126, 126, 233, 233, 218, 218, 248, 248, 188, 188,
- 203, 203, 142, 142, 173, 173, 158, 158, 249, 249, 234, 234, 204, 204, 219,
- 219, 174, 174, 189, 189, 250, 250, 220, 220, 190, 190, 205, 205, 235, 235,
- 206, 206, 236, 236, 251, 251, 221, 221, 252, 252, 222, 222, 237, 237, 238,
- 238, 253, 253, 254, 254, 0, 0,
+ 0, 0, 0, 0, 1, 1, 0, 0, 2, 2, 16, 16, 3, 3, 17,
+ 17, 16, 16, 4, 4, 32, 32, 18, 18, 5, 5, 33, 33, 32, 32,
+ 19, 19, 48, 48, 6, 6, 34, 34, 20, 20, 49, 49, 48, 48, 7,
+ 7, 35, 35, 64, 64, 21, 21, 50, 50, 36, 36, 64, 64, 8, 8,
+ 65, 65, 51, 51, 22, 22, 37, 37, 80, 80, 66, 66, 9, 9, 52,
+ 52, 23, 23, 81, 81, 67, 67, 80, 80, 38, 38, 10, 10, 53, 53,
+ 82, 82, 96, 96, 68, 68, 24, 24, 97, 97, 83, 83, 39, 39, 96,
+ 96, 54, 54, 11, 11, 69, 69, 98, 98, 112, 112, 84, 84, 25, 25,
+ 40, 40, 55, 55, 113, 113, 99, 99, 12, 12, 70, 70, 112, 112, 85,
+ 85, 26, 26, 114, 114, 100, 100, 128, 128, 41, 41, 56, 56, 71, 71,
+ 115, 115, 13, 13, 86, 86, 129, 129, 101, 101, 128, 128, 72, 72, 130,
+ 130, 116, 116, 27, 27, 57, 57, 14, 14, 87, 87, 42, 42, 144, 144,
+ 102, 102, 131, 131, 145, 145, 117, 117, 73, 73, 144, 144, 88, 88, 132,
+ 132, 103, 103, 28, 28, 58, 58, 146, 146, 118, 118, 43, 43, 160, 160,
+ 147, 147, 89, 89, 104, 104, 133, 133, 161, 161, 119, 119, 160, 160, 74,
+ 74, 134, 134, 148, 148, 29, 29, 59, 59, 162, 162, 176, 176, 44, 44,
+ 120, 120, 90, 90, 105, 105, 163, 163, 177, 177, 149, 149, 176, 176, 135,
+ 135, 164, 164, 178, 178, 30, 30, 150, 150, 192, 192, 75, 75, 121, 121,
+ 60, 60, 136, 136, 193, 193, 106, 106, 151, 151, 179, 179, 192, 192, 45,
+ 45, 165, 165, 166, 166, 194, 194, 91, 91, 180, 180, 137, 137, 208, 208,
+ 122, 122, 152, 152, 208, 208, 195, 195, 76, 76, 167, 167, 209, 209, 181,
+ 181, 224, 224, 107, 107, 196, 196, 61, 61, 153, 153, 224, 224, 182, 182,
+ 168, 168, 210, 210, 46, 46, 138, 138, 92, 92, 183, 183, 225, 225, 211,
+ 211, 240, 240, 197, 197, 169, 169, 123, 123, 154, 154, 198, 198, 77, 77,
+ 212, 212, 184, 184, 108, 108, 226, 226, 199, 199, 62, 62, 227, 227, 241,
+ 241, 139, 139, 213, 213, 170, 170, 185, 185, 155, 155, 228, 228, 242, 242,
+ 124, 124, 93, 93, 200, 200, 243, 243, 214, 214, 215, 215, 229, 229, 140,
+ 140, 186, 186, 201, 201, 78, 78, 171, 171, 109, 109, 156, 156, 244, 244,
+ 216, 216, 230, 230, 94, 94, 245, 245, 231, 231, 125, 125, 202, 202, 246,
+ 246, 232, 232, 172, 172, 217, 217, 141, 141, 110, 110, 157, 157, 187, 187,
+ 247, 247, 126, 126, 233, 233, 218, 218, 248, 248, 188, 188, 203, 203, 142,
+ 142, 173, 173, 158, 158, 249, 249, 234, 234, 204, 204, 219, 219, 174, 174,
+ 189, 189, 250, 250, 220, 220, 190, 190, 205, 205, 235, 235, 206, 206, 236,
+ 236, 251, 251, 221, 221, 252, 252, 222, 222, 237, 237, 238, 238, 253, 253,
+ 254, 254, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
default_scan_16x16_neighbors[257 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 0, 0, 16, 16, 1, 16, 1, 1, 32, 32, 17, 32,
- 2, 17, 2, 2, 48, 48, 18, 33, 33, 48, 3, 18, 49, 64, 64, 64,
- 34, 49, 3, 3, 19, 34, 50, 65, 4, 19, 65, 80, 80, 80, 35, 50,
- 4, 4, 20, 35, 66, 81, 81, 96, 51, 66, 96, 96, 5, 20, 36, 51,
- 82, 97, 21, 36, 67, 82, 97, 112, 5, 5, 52, 67, 112, 112, 37, 52,
- 6, 21, 83, 98, 98, 113, 68, 83, 6, 6, 113, 128, 22, 37, 53, 68,
- 84, 99, 99, 114, 128, 128, 114, 129, 69, 84, 38, 53, 7, 22, 7, 7,
- 129, 144, 23, 38, 54, 69, 100, 115, 85, 100, 115, 130, 144, 144, 130, 145,
- 39, 54, 70, 85, 8, 23, 55, 70, 116, 131, 101, 116, 145, 160, 24, 39,
- 8, 8, 86, 101, 131, 146, 160, 160, 146, 161, 71, 86, 40, 55, 9, 24,
- 117, 132, 102, 117, 161, 176, 132, 147, 56, 71, 87, 102, 25, 40, 147, 162,
- 9, 9, 176, 176, 162, 177, 72, 87, 41, 56, 118, 133, 133, 148, 103, 118,
- 10, 25, 148, 163, 57, 72, 88, 103, 177, 192, 26, 41, 163, 178, 192, 192,
- 10, 10, 119, 134, 73, 88, 149, 164, 104, 119, 134, 149, 42, 57, 178, 193,
- 164, 179, 11, 26, 58, 73, 193, 208, 89, 104, 135, 150, 120, 135, 27, 42,
- 74, 89, 208, 208, 150, 165, 179, 194, 165, 180, 105, 120, 194, 209, 43, 58,
- 11, 11, 136, 151, 90, 105, 151, 166, 180, 195, 59, 74, 121, 136, 209, 224,
- 195, 210, 224, 224, 166, 181, 106, 121, 75, 90, 12, 27, 181, 196, 12, 12,
- 210, 225, 152, 167, 167, 182, 137, 152, 28, 43, 196, 211, 122, 137, 91, 106,
- 225, 240, 44, 59, 13, 28, 107, 122, 182, 197, 168, 183, 211, 226, 153, 168,
- 226, 241, 60, 75, 197, 212, 138, 153, 29, 44, 76, 91, 13, 13, 183, 198,
- 123, 138, 45, 60, 212, 227, 198, 213, 154, 169, 169, 184, 227, 242, 92, 107,
- 61, 76, 139, 154, 14, 29, 14, 14, 184, 199, 213, 228, 108, 123, 199, 214,
- 228, 243, 77, 92, 30, 45, 170, 185, 155, 170, 185, 200, 93, 108, 124, 139,
- 214, 229, 46, 61, 200, 215, 229, 244, 15, 30, 109, 124, 62, 77, 140, 155,
- 215, 230, 31, 46, 171, 186, 186, 201, 201, 216, 78, 93, 230, 245, 125, 140,
- 47, 62, 216, 231, 156, 171, 94, 109, 231, 246, 141, 156, 63, 78, 202, 217,
- 187, 202, 110, 125, 217, 232, 172, 187, 232, 247, 79, 94, 157, 172, 126, 141,
- 203, 218, 95, 110, 233, 248, 218, 233, 142, 157, 111, 126, 173, 188, 188, 203,
- 234, 249, 219, 234, 127, 142, 158, 173, 204, 219, 189, 204, 143, 158, 235,
- 250, 174, 189, 205, 220, 159, 174, 220, 235, 221, 236, 175, 190, 190, 205,
- 236, 251, 206, 221, 237, 252, 191, 206, 222, 237, 207, 222, 238, 253, 223,
- 238, 239, 254, 0, 0,
+ 0, 0, 0, 0, 0, 0, 16, 16, 1, 16, 1, 1, 32, 32, 17,
+ 32, 2, 17, 2, 2, 48, 48, 18, 33, 33, 48, 3, 18, 49, 64,
+ 64, 64, 34, 49, 3, 3, 19, 34, 50, 65, 4, 19, 65, 80, 80,
+ 80, 35, 50, 4, 4, 20, 35, 66, 81, 81, 96, 51, 66, 96, 96,
+ 5, 20, 36, 51, 82, 97, 21, 36, 67, 82, 97, 112, 5, 5, 52,
+ 67, 112, 112, 37, 52, 6, 21, 83, 98, 98, 113, 68, 83, 6, 6,
+ 113, 128, 22, 37, 53, 68, 84, 99, 99, 114, 128, 128, 114, 129, 69,
+ 84, 38, 53, 7, 22, 7, 7, 129, 144, 23, 38, 54, 69, 100, 115,
+ 85, 100, 115, 130, 144, 144, 130, 145, 39, 54, 70, 85, 8, 23, 55,
+ 70, 116, 131, 101, 116, 145, 160, 24, 39, 8, 8, 86, 101, 131, 146,
+ 160, 160, 146, 161, 71, 86, 40, 55, 9, 24, 117, 132, 102, 117, 161,
+ 176, 132, 147, 56, 71, 87, 102, 25, 40, 147, 162, 9, 9, 176, 176,
+ 162, 177, 72, 87, 41, 56, 118, 133, 133, 148, 103, 118, 10, 25, 148,
+ 163, 57, 72, 88, 103, 177, 192, 26, 41, 163, 178, 192, 192, 10, 10,
+ 119, 134, 73, 88, 149, 164, 104, 119, 134, 149, 42, 57, 178, 193, 164,
+ 179, 11, 26, 58, 73, 193, 208, 89, 104, 135, 150, 120, 135, 27, 42,
+ 74, 89, 208, 208, 150, 165, 179, 194, 165, 180, 105, 120, 194, 209, 43,
+ 58, 11, 11, 136, 151, 90, 105, 151, 166, 180, 195, 59, 74, 121, 136,
+ 209, 224, 195, 210, 224, 224, 166, 181, 106, 121, 75, 90, 12, 27, 181,
+ 196, 12, 12, 210, 225, 152, 167, 167, 182, 137, 152, 28, 43, 196, 211,
+ 122, 137, 91, 106, 225, 240, 44, 59, 13, 28, 107, 122, 182, 197, 168,
+ 183, 211, 226, 153, 168, 226, 241, 60, 75, 197, 212, 138, 153, 29, 44,
+ 76, 91, 13, 13, 183, 198, 123, 138, 45, 60, 212, 227, 198, 213, 154,
+ 169, 169, 184, 227, 242, 92, 107, 61, 76, 139, 154, 14, 29, 14, 14,
+ 184, 199, 213, 228, 108, 123, 199, 214, 228, 243, 77, 92, 30, 45, 170,
+ 185, 155, 170, 185, 200, 93, 108, 124, 139, 214, 229, 46, 61, 200, 215,
+ 229, 244, 15, 30, 109, 124, 62, 77, 140, 155, 215, 230, 31, 46, 171,
+ 186, 186, 201, 201, 216, 78, 93, 230, 245, 125, 140, 47, 62, 216, 231,
+ 156, 171, 94, 109, 231, 246, 141, 156, 63, 78, 202, 217, 187, 202, 110,
+ 125, 217, 232, 172, 187, 232, 247, 79, 94, 157, 172, 126, 141, 203, 218,
+ 95, 110, 233, 248, 218, 233, 142, 157, 111, 126, 173, 188, 188, 203, 234,
+ 249, 219, 234, 127, 142, 158, 173, 204, 219, 189, 204, 143, 158, 235, 250,
+ 174, 189, 205, 220, 159, 174, 220, 235, 221, 236, 175, 190, 190, 205, 236,
+ 251, 206, 221, 237, 252, 191, 206, 222, 237, 207, 222, 238, 253, 223, 238,
+ 239, 254, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t,
default_scan_32x32_neighbors[1025 * MAX_NEIGHBORS]) = {
- 0, 0, 0, 0, 0, 0, 32, 32, 1, 32, 1, 1, 64, 64, 33, 64,
- 2, 33, 96, 96, 2, 2, 65, 96, 34, 65, 128, 128, 97, 128, 3, 34,
- 66, 97, 3, 3, 35, 66, 98, 129, 129, 160, 160, 160, 4, 35, 67, 98,
- 192, 192, 4, 4, 130, 161, 161, 192, 36, 67, 99, 130, 5, 36, 68, 99,
- 193, 224, 162, 193, 224, 224, 131, 162, 37, 68, 100, 131, 5, 5, 194, 225,
- 225, 256, 256, 256, 163, 194, 69, 100, 132, 163, 6, 37, 226, 257, 6, 6,
- 195, 226, 257, 288, 101, 132, 288, 288, 38, 69, 164, 195, 133, 164, 258, 289,
- 227, 258, 196, 227, 7, 38, 289, 320, 70, 101, 320, 320, 7, 7, 165, 196,
- 39, 70, 102, 133, 290, 321, 259, 290, 228, 259, 321, 352, 352, 352, 197, 228,
- 134, 165, 71, 102, 8, 39, 322, 353, 291, 322, 260, 291, 103, 134, 353, 384,
- 166, 197, 229, 260, 40, 71, 8, 8, 384, 384, 135, 166, 354, 385, 323, 354,
- 198, 229, 292, 323, 72, 103, 261, 292, 9, 40, 385, 416, 167, 198, 104, 135,
- 230, 261, 355, 386, 416, 416, 293, 324, 324, 355, 9, 9, 41, 72, 386, 417,
- 199, 230, 136, 167, 417, 448, 262, 293, 356, 387, 73, 104, 387, 418, 231, 262,
- 10, 41, 168, 199, 325, 356, 418, 449, 105, 136, 448, 448, 42, 73, 294, 325,
- 200, 231, 10, 10, 357, 388, 137, 168, 263, 294, 388, 419, 74, 105, 419, 450,
- 449, 480, 326, 357, 232, 263, 295, 326, 169, 200, 11, 42, 106, 137, 480, 480,
- 450, 481, 358, 389, 264, 295, 201, 232, 138, 169, 389, 420, 43, 74, 420, 451,
- 327, 358, 11, 11, 481, 512, 233, 264, 451, 482, 296, 327, 75, 106, 170, 201,
- 482, 513, 512, 512, 390, 421, 359, 390, 421, 452, 107, 138, 12, 43, 202, 233,
- 452, 483, 265, 296, 328, 359, 139, 170, 44, 75, 483, 514, 513, 544, 234, 265,
- 297, 328, 422, 453, 12, 12, 391, 422, 171, 202, 76, 107, 514, 545, 453, 484,
- 544, 544, 266, 297, 203, 234, 108, 139, 329, 360, 298, 329, 140, 171, 515,
- 546, 13, 44, 423, 454, 235, 266, 545, 576, 454, 485, 45, 76, 172, 203, 330,
- 361, 576, 576, 13, 13, 267, 298, 546, 577, 77, 108, 204, 235, 455, 486, 577,
- 608, 299, 330, 109, 140, 547, 578, 14, 45, 14, 14, 141, 172, 578, 609, 331,
- 362, 46, 77, 173, 204, 15, 15, 78, 109, 205, 236, 579, 610, 110, 141, 15, 46,
- 142, 173, 47, 78, 174, 205, 16, 16, 79, 110, 206, 237, 16, 47, 111, 142,
- 48, 79, 143, 174, 80, 111, 175, 206, 17, 48, 17, 17, 207, 238, 49, 80,
- 81, 112, 18, 18, 18, 49, 50, 81, 82, 113, 19, 50, 51, 82, 83, 114, 608, 608,
- 484, 515, 360, 391, 236, 267, 112, 143, 19, 19, 640, 640, 609, 640, 516, 547,
- 485, 516, 392, 423, 361, 392, 268, 299, 237, 268, 144, 175, 113, 144, 20, 51,
- 20, 20, 672, 672, 641, 672, 610, 641, 548, 579, 517, 548, 486, 517, 424, 455,
- 393, 424, 362, 393, 300, 331, 269, 300, 238, 269, 176, 207, 145, 176, 114,
- 145, 52, 83, 21, 52, 21, 21, 704, 704, 673, 704, 642, 673, 611, 642, 580,
- 611, 549, 580, 518, 549, 487, 518, 456, 487, 425, 456, 394, 425, 363, 394,
- 332, 363, 301, 332, 270, 301, 239, 270, 208, 239, 177, 208, 146, 177, 115,
- 146, 84, 115, 53, 84, 22, 53, 22, 22, 705, 736, 674, 705, 643, 674, 581, 612,
- 550, 581, 519, 550, 457, 488, 426, 457, 395, 426, 333, 364, 302, 333, 271,
- 302, 209, 240, 178, 209, 147, 178, 85, 116, 54, 85, 23, 54, 706, 737, 675,
- 706, 582, 613, 551, 582, 458, 489, 427, 458, 334, 365, 303, 334, 210, 241,
- 179, 210, 86, 117, 55, 86, 707, 738, 583, 614, 459, 490, 335, 366, 211, 242,
- 87, 118, 736, 736, 612, 643, 488, 519, 364, 395, 240, 271, 116, 147, 23, 23,
- 768, 768, 737, 768, 644, 675, 613, 644, 520, 551, 489, 520, 396, 427, 365,
- 396, 272, 303, 241, 272, 148, 179, 117, 148, 24, 55, 24, 24, 800, 800, 769,
- 800, 738, 769, 676, 707, 645, 676, 614, 645, 552, 583, 521, 552, 490, 521,
- 428, 459, 397, 428, 366, 397, 304, 335, 273, 304, 242, 273, 180, 211, 149,
- 180, 118, 149, 56, 87, 25, 56, 25, 25, 832, 832, 801, 832, 770, 801, 739,
- 770, 708, 739, 677, 708, 646, 677, 615, 646, 584, 615, 553, 584, 522, 553,
- 491, 522, 460, 491, 429, 460, 398, 429, 367, 398, 336, 367, 305, 336, 274,
- 305, 243, 274, 212, 243, 181, 212, 150, 181, 119, 150, 88, 119, 57, 88, 26,
- 57, 26, 26, 833, 864, 802, 833, 771, 802, 709, 740, 678, 709, 647, 678, 585,
- 616, 554, 585, 523, 554, 461, 492, 430, 461, 399, 430, 337, 368, 306, 337,
- 275, 306, 213, 244, 182, 213, 151, 182, 89, 120, 58, 89, 27, 58, 834, 865,
- 803, 834, 710, 741, 679, 710, 586, 617, 555, 586, 462, 493, 431, 462, 338,
- 369, 307, 338, 214, 245, 183, 214, 90, 121, 59, 90, 835, 866, 711, 742, 587,
- 618, 463, 494, 339, 370, 215, 246, 91, 122, 864, 864, 740, 771, 616, 647,
- 492, 523, 368, 399, 244, 275, 120, 151, 27, 27, 896, 896, 865, 896, 772, 803,
- 741, 772, 648, 679, 617, 648, 524, 555, 493, 524, 400, 431, 369, 400, 276,
- 307, 245, 276, 152, 183, 121, 152, 28, 59, 28, 28, 928, 928, 897, 928, 866,
- 897, 804, 835, 773, 804, 742, 773, 680, 711, 649, 680, 618, 649, 556, 587,
- 525, 556, 494, 525, 432, 463, 401, 432, 370, 401, 308, 339, 277, 308, 246,
- 277, 184, 215, 153, 184, 122, 153, 60, 91, 29, 60, 29, 29, 960, 960, 929,
- 960, 898, 929, 867, 898, 836, 867, 805, 836, 774, 805, 743, 774, 712, 743,
- 681, 712, 650, 681, 619, 650, 588, 619, 557, 588, 526, 557, 495, 526, 464,
- 495, 433, 464, 402, 433, 371, 402, 340, 371, 309, 340, 278, 309, 247, 278,
- 216, 247, 185, 216, 154, 185, 123, 154, 92, 123, 61, 92, 30, 61, 30, 30,
- 961, 992, 930, 961, 899, 930, 837, 868, 806, 837, 775, 806, 713, 744, 682,
- 713, 651, 682, 589, 620, 558, 589, 527, 558, 465, 496, 434, 465, 403, 434,
- 341, 372, 310, 341, 279, 310, 217, 248, 186, 217, 155, 186, 93, 124, 62, 93,
- 31, 62, 962, 993, 931, 962, 838, 869, 807, 838, 714, 745, 683, 714, 590, 621,
- 559, 590, 466, 497, 435, 466, 342, 373, 311, 342, 218, 249, 187, 218, 94,
- 125, 63, 94, 963, 994, 839, 870, 715, 746, 591, 622, 467, 498, 343, 374, 219,
- 250, 95, 126, 868, 899, 744, 775, 620, 651, 496, 527, 372, 403, 248, 279,
- 124, 155, 900, 931, 869, 900, 776, 807, 745, 776, 652, 683, 621, 652, 528,
- 559, 497, 528, 404, 435, 373, 404, 280, 311, 249, 280, 156, 187, 125, 156,
- 932, 963, 901, 932, 870, 901, 808, 839, 777, 808, 746, 777, 684, 715, 653,
- 684, 622, 653, 560, 591, 529, 560, 498, 529, 436, 467, 405, 436, 374, 405,
- 312, 343, 281, 312, 250, 281, 188, 219, 157, 188, 126, 157, 964, 995, 933,
- 964, 902, 933, 871, 902, 840, 871, 809, 840, 778, 809, 747, 778, 716, 747,
- 685, 716, 654, 685, 623, 654, 592, 623, 561, 592, 530, 561, 499, 530, 468,
- 499, 437, 468, 406, 437, 375, 406, 344, 375, 313, 344, 282, 313, 251, 282,
- 220, 251, 189, 220, 158, 189, 127, 158, 965, 996, 934, 965, 903, 934, 841,
- 872, 810, 841, 779, 810, 717, 748, 686, 717, 655, 686, 593, 624, 562, 593,
- 531, 562, 469, 500, 438, 469, 407, 438, 345, 376, 314, 345, 283, 314, 221,
- 252, 190, 221, 159, 190, 966, 997, 935, 966, 842, 873, 811, 842, 718, 749,
- 687, 718, 594, 625, 563, 594, 470, 501, 439, 470, 346, 377, 315, 346, 222,
- 253, 191, 222, 967, 998, 843, 874, 719, 750, 595, 626, 471, 502, 347, 378,
- 223, 254, 872, 903, 748, 779, 624, 655, 500, 531, 376, 407, 252, 283, 904,
- 935, 873, 904, 780, 811, 749, 780, 656, 687, 625, 656, 532, 563, 501, 532,
- 408, 439, 377, 408, 284, 315, 253, 284, 936, 967, 905, 936, 874, 905, 812,
- 843, 781, 812, 750, 781, 688, 719, 657, 688, 626, 657, 564, 595, 533, 564,
- 502, 533, 440, 471, 409, 440, 378, 409, 316, 347, 285, 316, 254, 285, 968,
- 999, 937, 968, 906, 937, 875, 906, 844, 875, 813, 844, 782, 813, 751, 782,
- 720, 751, 689, 720, 658, 689, 627, 658, 596, 627, 565, 596, 534, 565, 503,
- 534, 472, 503, 441, 472, 410, 441, 379, 410, 348, 379, 317, 348, 286, 317,
- 255, 286, 969, 1000, 938, 969, 907, 938, 845, 876, 814, 845, 783, 814, 721,
- 752, 690, 721, 659, 690, 597, 628, 566, 597, 535, 566, 473, 504, 442, 473,
- 411, 442, 349, 380, 318, 349, 287, 318, 970, 1001, 939, 970, 846, 877, 815,
- 846, 722, 753, 691, 722, 598, 629, 567, 598, 474, 505, 443, 474, 350, 381,
- 319, 350, 971, 1002, 847, 878, 723, 754, 599, 630, 475, 506, 351, 382, 876,
- 907, 752, 783, 628, 659, 504, 535, 380, 411, 908, 939, 877, 908, 784, 815,
- 753, 784, 660, 691, 629, 660, 536, 567, 505, 536, 412, 443, 381, 412, 940,
- 971, 909, 940, 878, 909, 816, 847, 785, 816, 754, 785, 692, 723, 661, 692,
- 630, 661, 568, 599, 537, 568, 506, 537, 444, 475, 413, 444, 382, 413, 972,
- 1003, 941, 972, 910, 941, 879, 910, 848, 879, 817, 848, 786, 817, 755, 786,
- 724, 755, 693, 724, 662, 693, 631, 662, 600, 631, 569, 600, 538, 569, 507,
- 538, 476, 507, 445, 476, 414, 445, 383, 414, 973, 1004, 942, 973, 911, 942,
- 849, 880, 818, 849, 787, 818, 725, 756, 694, 725, 663, 694, 601, 632, 570,
- 601, 539, 570, 477, 508, 446, 477, 415, 446, 974, 1005, 943, 974, 850, 881,
- 819, 850, 726, 757, 695, 726, 602, 633, 571, 602, 478, 509, 447, 478, 975,
- 1006, 851, 882, 727, 758, 603, 634, 479, 510, 880, 911, 756, 787, 632, 663,
- 508, 539, 912, 943, 881, 912, 788, 819, 757, 788, 664, 695, 633, 664, 540,
- 571, 509, 540, 944, 975, 913, 944, 882, 913, 820, 851, 789, 820, 758, 789,
- 696, 727, 665, 696, 634, 665, 572, 603, 541, 572, 510, 541, 976, 1007, 945,
- 976, 914, 945, 883, 914, 852, 883, 821, 852, 790, 821, 759, 790, 728, 759,
- 697, 728, 666, 697, 635, 666, 604, 635, 573, 604, 542, 573, 511, 542, 977,
- 1008, 946, 977, 915, 946, 853, 884, 822, 853, 791, 822, 729, 760, 698, 729,
- 667, 698, 605, 636, 574, 605, 543, 574, 978, 1009, 947, 978, 854, 885, 823,
- 854, 730, 761, 699, 730, 606, 637, 575, 606, 979, 1010, 855, 886, 731, 762,
- 607, 638, 884, 915, 760, 791, 636, 667, 916, 947, 885, 916, 792, 823, 761,
- 792, 668, 699, 637, 668, 948, 979, 917, 948, 886, 917, 824, 855, 793, 824,
- 762, 793, 700, 731, 669, 700, 638, 669, 980, 1011, 949, 980, 918, 949, 887,
- 918, 856, 887, 825, 856, 794, 825, 763, 794, 732, 763, 701, 732, 670, 701,
- 639, 670, 981, 1012, 950, 981, 919, 950, 857, 888, 826, 857, 795, 826, 733,
- 764, 702, 733, 671, 702, 982, 1013, 951, 982, 858, 889, 827, 858, 734, 765,
- 703, 734, 983, 1014, 859, 890, 735, 766, 888, 919, 764, 795, 920, 951, 889,
- 920, 796, 827, 765, 796, 952, 983, 921, 952, 890, 921, 828, 859, 797, 828,
- 766, 797, 984, 1015, 953, 984, 922, 953, 891, 922, 860, 891, 829, 860, 798,
- 829, 767, 798, 985, 1016, 954, 985, 923, 954, 861, 892, 830, 861, 799, 830,
- 986, 1017, 955, 986, 862, 893, 831, 862, 987, 1018, 863, 894, 892, 923, 924,
- 955, 893, 924, 956, 987, 925, 956, 894, 925, 988, 1019, 957, 988, 926, 957,
- 895, 926, 989, 1020, 958, 989, 927, 958, 990, 1021, 959, 990, 991, 1022, 0, 0,
+ 0, 0, 0, 0, 0, 0, 32, 32, 1, 32, 1, 1, 64, 64,
+ 33, 64, 2, 33, 96, 96, 2, 2, 65, 96, 34, 65, 128, 128,
+ 97, 128, 3, 34, 66, 97, 3, 3, 35, 66, 98, 129, 129, 160,
+ 160, 160, 4, 35, 67, 98, 192, 192, 4, 4, 130, 161, 161, 192,
+ 36, 67, 99, 130, 5, 36, 68, 99, 193, 224, 162, 193, 224, 224,
+ 131, 162, 37, 68, 100, 131, 5, 5, 194, 225, 225, 256, 256, 256,
+ 163, 194, 69, 100, 132, 163, 6, 37, 226, 257, 6, 6, 195, 226,
+ 257, 288, 101, 132, 288, 288, 38, 69, 164, 195, 133, 164, 258, 289,
+ 227, 258, 196, 227, 7, 38, 289, 320, 70, 101, 320, 320, 7, 7,
+ 165, 196, 39, 70, 102, 133, 290, 321, 259, 290, 228, 259, 321, 352,
+ 352, 352, 197, 228, 134, 165, 71, 102, 8, 39, 322, 353, 291, 322,
+ 260, 291, 103, 134, 353, 384, 166, 197, 229, 260, 40, 71, 8, 8,
+ 384, 384, 135, 166, 354, 385, 323, 354, 198, 229, 292, 323, 72, 103,
+ 261, 292, 9, 40, 385, 416, 167, 198, 104, 135, 230, 261, 355, 386,
+ 416, 416, 293, 324, 324, 355, 9, 9, 41, 72, 386, 417, 199, 230,
+ 136, 167, 417, 448, 262, 293, 356, 387, 73, 104, 387, 418, 231, 262,
+ 10, 41, 168, 199, 325, 356, 418, 449, 105, 136, 448, 448, 42, 73,
+ 294, 325, 200, 231, 10, 10, 357, 388, 137, 168, 263, 294, 388, 419,
+ 74, 105, 419, 450, 449, 480, 326, 357, 232, 263, 295, 326, 169, 200,
+ 11, 42, 106, 137, 480, 480, 450, 481, 358, 389, 264, 295, 201, 232,
+ 138, 169, 389, 420, 43, 74, 420, 451, 327, 358, 11, 11, 481, 512,
+ 233, 264, 451, 482, 296, 327, 75, 106, 170, 201, 482, 513, 512, 512,
+ 390, 421, 359, 390, 421, 452, 107, 138, 12, 43, 202, 233, 452, 483,
+ 265, 296, 328, 359, 139, 170, 44, 75, 483, 514, 513, 544, 234, 265,
+ 297, 328, 422, 453, 12, 12, 391, 422, 171, 202, 76, 107, 514, 545,
+ 453, 484, 544, 544, 266, 297, 203, 234, 108, 139, 329, 360, 298, 329,
+ 140, 171, 515, 546, 13, 44, 423, 454, 235, 266, 545, 576, 454, 485,
+ 45, 76, 172, 203, 330, 361, 576, 576, 13, 13, 267, 298, 546, 577,
+ 77, 108, 204, 235, 455, 486, 577, 608, 299, 330, 109, 140, 547, 578,
+ 14, 45, 14, 14, 141, 172, 578, 609, 331, 362, 46, 77, 173, 204,
+ 15, 15, 78, 109, 205, 236, 579, 610, 110, 141, 15, 46, 142, 173,
+ 47, 78, 174, 205, 16, 16, 79, 110, 206, 237, 16, 47, 111, 142,
+ 48, 79, 143, 174, 80, 111, 175, 206, 17, 48, 17, 17, 207, 238,
+ 49, 80, 81, 112, 18, 18, 18, 49, 50, 81, 82, 113, 19, 50,
+ 51, 82, 83, 114, 608, 608, 484, 515, 360, 391, 236, 267, 112, 143,
+ 19, 19, 640, 640, 609, 640, 516, 547, 485, 516, 392, 423, 361, 392,
+ 268, 299, 237, 268, 144, 175, 113, 144, 20, 51, 20, 20, 672, 672,
+ 641, 672, 610, 641, 548, 579, 517, 548, 486, 517, 424, 455, 393, 424,
+ 362, 393, 300, 331, 269, 300, 238, 269, 176, 207, 145, 176, 114, 145,
+ 52, 83, 21, 52, 21, 21, 704, 704, 673, 704, 642, 673, 611, 642,
+ 580, 611, 549, 580, 518, 549, 487, 518, 456, 487, 425, 456, 394, 425,
+ 363, 394, 332, 363, 301, 332, 270, 301, 239, 270, 208, 239, 177, 208,
+ 146, 177, 115, 146, 84, 115, 53, 84, 22, 53, 22, 22, 705, 736,
+ 674, 705, 643, 674, 581, 612, 550, 581, 519, 550, 457, 488, 426, 457,
+ 395, 426, 333, 364, 302, 333, 271, 302, 209, 240, 178, 209, 147, 178,
+ 85, 116, 54, 85, 23, 54, 706, 737, 675, 706, 582, 613, 551, 582,
+ 458, 489, 427, 458, 334, 365, 303, 334, 210, 241, 179, 210, 86, 117,
+ 55, 86, 707, 738, 583, 614, 459, 490, 335, 366, 211, 242, 87, 118,
+ 736, 736, 612, 643, 488, 519, 364, 395, 240, 271, 116, 147, 23, 23,
+ 768, 768, 737, 768, 644, 675, 613, 644, 520, 551, 489, 520, 396, 427,
+ 365, 396, 272, 303, 241, 272, 148, 179, 117, 148, 24, 55, 24, 24,
+ 800, 800, 769, 800, 738, 769, 676, 707, 645, 676, 614, 645, 552, 583,
+ 521, 552, 490, 521, 428, 459, 397, 428, 366, 397, 304, 335, 273, 304,
+ 242, 273, 180, 211, 149, 180, 118, 149, 56, 87, 25, 56, 25, 25,
+ 832, 832, 801, 832, 770, 801, 739, 770, 708, 739, 677, 708, 646, 677,
+ 615, 646, 584, 615, 553, 584, 522, 553, 491, 522, 460, 491, 429, 460,
+ 398, 429, 367, 398, 336, 367, 305, 336, 274, 305, 243, 274, 212, 243,
+ 181, 212, 150, 181, 119, 150, 88, 119, 57, 88, 26, 57, 26, 26,
+ 833, 864, 802, 833, 771, 802, 709, 740, 678, 709, 647, 678, 585, 616,
+ 554, 585, 523, 554, 461, 492, 430, 461, 399, 430, 337, 368, 306, 337,
+ 275, 306, 213, 244, 182, 213, 151, 182, 89, 120, 58, 89, 27, 58,
+ 834, 865, 803, 834, 710, 741, 679, 710, 586, 617, 555, 586, 462, 493,
+ 431, 462, 338, 369, 307, 338, 214, 245, 183, 214, 90, 121, 59, 90,
+ 835, 866, 711, 742, 587, 618, 463, 494, 339, 370, 215, 246, 91, 122,
+ 864, 864, 740, 771, 616, 647, 492, 523, 368, 399, 244, 275, 120, 151,
+ 27, 27, 896, 896, 865, 896, 772, 803, 741, 772, 648, 679, 617, 648,
+ 524, 555, 493, 524, 400, 431, 369, 400, 276, 307, 245, 276, 152, 183,
+ 121, 152, 28, 59, 28, 28, 928, 928, 897, 928, 866, 897, 804, 835,
+ 773, 804, 742, 773, 680, 711, 649, 680, 618, 649, 556, 587, 525, 556,
+ 494, 525, 432, 463, 401, 432, 370, 401, 308, 339, 277, 308, 246, 277,
+ 184, 215, 153, 184, 122, 153, 60, 91, 29, 60, 29, 29, 960, 960,
+ 929, 960, 898, 929, 867, 898, 836, 867, 805, 836, 774, 805, 743, 774,
+ 712, 743, 681, 712, 650, 681, 619, 650, 588, 619, 557, 588, 526, 557,
+ 495, 526, 464, 495, 433, 464, 402, 433, 371, 402, 340, 371, 309, 340,
+ 278, 309, 247, 278, 216, 247, 185, 216, 154, 185, 123, 154, 92, 123,
+ 61, 92, 30, 61, 30, 30, 961, 992, 930, 961, 899, 930, 837, 868,
+ 806, 837, 775, 806, 713, 744, 682, 713, 651, 682, 589, 620, 558, 589,
+ 527, 558, 465, 496, 434, 465, 403, 434, 341, 372, 310, 341, 279, 310,
+ 217, 248, 186, 217, 155, 186, 93, 124, 62, 93, 31, 62, 962, 993,
+ 931, 962, 838, 869, 807, 838, 714, 745, 683, 714, 590, 621, 559, 590,
+ 466, 497, 435, 466, 342, 373, 311, 342, 218, 249, 187, 218, 94, 125,
+ 63, 94, 963, 994, 839, 870, 715, 746, 591, 622, 467, 498, 343, 374,
+ 219, 250, 95, 126, 868, 899, 744, 775, 620, 651, 496, 527, 372, 403,
+ 248, 279, 124, 155, 900, 931, 869, 900, 776, 807, 745, 776, 652, 683,
+ 621, 652, 528, 559, 497, 528, 404, 435, 373, 404, 280, 311, 249, 280,
+ 156, 187, 125, 156, 932, 963, 901, 932, 870, 901, 808, 839, 777, 808,
+ 746, 777, 684, 715, 653, 684, 622, 653, 560, 591, 529, 560, 498, 529,
+ 436, 467, 405, 436, 374, 405, 312, 343, 281, 312, 250, 281, 188, 219,
+ 157, 188, 126, 157, 964, 995, 933, 964, 902, 933, 871, 902, 840, 871,
+ 809, 840, 778, 809, 747, 778, 716, 747, 685, 716, 654, 685, 623, 654,
+ 592, 623, 561, 592, 530, 561, 499, 530, 468, 499, 437, 468, 406, 437,
+ 375, 406, 344, 375, 313, 344, 282, 313, 251, 282, 220, 251, 189, 220,
+ 158, 189, 127, 158, 965, 996, 934, 965, 903, 934, 841, 872, 810, 841,
+ 779, 810, 717, 748, 686, 717, 655, 686, 593, 624, 562, 593, 531, 562,
+ 469, 500, 438, 469, 407, 438, 345, 376, 314, 345, 283, 314, 221, 252,
+ 190, 221, 159, 190, 966, 997, 935, 966, 842, 873, 811, 842, 718, 749,
+ 687, 718, 594, 625, 563, 594, 470, 501, 439, 470, 346, 377, 315, 346,
+ 222, 253, 191, 222, 967, 998, 843, 874, 719, 750, 595, 626, 471, 502,
+ 347, 378, 223, 254, 872, 903, 748, 779, 624, 655, 500, 531, 376, 407,
+ 252, 283, 904, 935, 873, 904, 780, 811, 749, 780, 656, 687, 625, 656,
+ 532, 563, 501, 532, 408, 439, 377, 408, 284, 315, 253, 284, 936, 967,
+ 905, 936, 874, 905, 812, 843, 781, 812, 750, 781, 688, 719, 657, 688,
+ 626, 657, 564, 595, 533, 564, 502, 533, 440, 471, 409, 440, 378, 409,
+ 316, 347, 285, 316, 254, 285, 968, 999, 937, 968, 906, 937, 875, 906,
+ 844, 875, 813, 844, 782, 813, 751, 782, 720, 751, 689, 720, 658, 689,
+ 627, 658, 596, 627, 565, 596, 534, 565, 503, 534, 472, 503, 441, 472,
+ 410, 441, 379, 410, 348, 379, 317, 348, 286, 317, 255, 286, 969, 1000,
+ 938, 969, 907, 938, 845, 876, 814, 845, 783, 814, 721, 752, 690, 721,
+ 659, 690, 597, 628, 566, 597, 535, 566, 473, 504, 442, 473, 411, 442,
+ 349, 380, 318, 349, 287, 318, 970, 1001, 939, 970, 846, 877, 815, 846,
+ 722, 753, 691, 722, 598, 629, 567, 598, 474, 505, 443, 474, 350, 381,
+ 319, 350, 971, 1002, 847, 878, 723, 754, 599, 630, 475, 506, 351, 382,
+ 876, 907, 752, 783, 628, 659, 504, 535, 380, 411, 908, 939, 877, 908,
+ 784, 815, 753, 784, 660, 691, 629, 660, 536, 567, 505, 536, 412, 443,
+ 381, 412, 940, 971, 909, 940, 878, 909, 816, 847, 785, 816, 754, 785,
+ 692, 723, 661, 692, 630, 661, 568, 599, 537, 568, 506, 537, 444, 475,
+ 413, 444, 382, 413, 972, 1003, 941, 972, 910, 941, 879, 910, 848, 879,
+ 817, 848, 786, 817, 755, 786, 724, 755, 693, 724, 662, 693, 631, 662,
+ 600, 631, 569, 600, 538, 569, 507, 538, 476, 507, 445, 476, 414, 445,
+ 383, 414, 973, 1004, 942, 973, 911, 942, 849, 880, 818, 849, 787, 818,
+ 725, 756, 694, 725, 663, 694, 601, 632, 570, 601, 539, 570, 477, 508,
+ 446, 477, 415, 446, 974, 1005, 943, 974, 850, 881, 819, 850, 726, 757,
+ 695, 726, 602, 633, 571, 602, 478, 509, 447, 478, 975, 1006, 851, 882,
+ 727, 758, 603, 634, 479, 510, 880, 911, 756, 787, 632, 663, 508, 539,
+ 912, 943, 881, 912, 788, 819, 757, 788, 664, 695, 633, 664, 540, 571,
+ 509, 540, 944, 975, 913, 944, 882, 913, 820, 851, 789, 820, 758, 789,
+ 696, 727, 665, 696, 634, 665, 572, 603, 541, 572, 510, 541, 976, 1007,
+ 945, 976, 914, 945, 883, 914, 852, 883, 821, 852, 790, 821, 759, 790,
+ 728, 759, 697, 728, 666, 697, 635, 666, 604, 635, 573, 604, 542, 573,
+ 511, 542, 977, 1008, 946, 977, 915, 946, 853, 884, 822, 853, 791, 822,
+ 729, 760, 698, 729, 667, 698, 605, 636, 574, 605, 543, 574, 978, 1009,
+ 947, 978, 854, 885, 823, 854, 730, 761, 699, 730, 606, 637, 575, 606,
+ 979, 1010, 855, 886, 731, 762, 607, 638, 884, 915, 760, 791, 636, 667,
+ 916, 947, 885, 916, 792, 823, 761, 792, 668, 699, 637, 668, 948, 979,
+ 917, 948, 886, 917, 824, 855, 793, 824, 762, 793, 700, 731, 669, 700,
+ 638, 669, 980, 1011, 949, 980, 918, 949, 887, 918, 856, 887, 825, 856,
+ 794, 825, 763, 794, 732, 763, 701, 732, 670, 701, 639, 670, 981, 1012,
+ 950, 981, 919, 950, 857, 888, 826, 857, 795, 826, 733, 764, 702, 733,
+ 671, 702, 982, 1013, 951, 982, 858, 889, 827, 858, 734, 765, 703, 734,
+ 983, 1014, 859, 890, 735, 766, 888, 919, 764, 795, 920, 951, 889, 920,
+ 796, 827, 765, 796, 952, 983, 921, 952, 890, 921, 828, 859, 797, 828,
+ 766, 797, 984, 1015, 953, 984, 922, 953, 891, 922, 860, 891, 829, 860,
+ 798, 829, 767, 798, 985, 1016, 954, 985, 923, 954, 861, 892, 830, 861,
+ 799, 830, 986, 1017, 955, 986, 862, 893, 831, 862, 987, 1018, 863, 894,
+ 892, 923, 924, 955, 893, 924, 956, 987, 925, 956, 894, 925, 988, 1019,
+ 957, 988, 926, 957, 895, 926, 989, 1020, 958, 989, 927, 958, 990, 1021,
+ 959, 990, 991, 1022, 0, 0,
};
DECLARE_ALIGNED(16, static const int16_t, vp10_default_iscan_4x4[16]) = {
@@ -545,183 +526,205 @@
};
DECLARE_ALIGNED(16, static const int16_t, vp10_col_iscan_8x8[64]) = {
- 0, 3, 8, 15, 22, 32, 40, 47, 1, 5, 11, 18, 26, 34, 44, 51,
- 2, 7, 13, 20, 28, 38, 46, 54, 4, 10, 16, 24, 31, 41, 50, 56,
- 6, 12, 21, 27, 35, 43, 52, 58, 9, 17, 25, 33, 39, 48, 55, 60,
+ 0, 3, 8, 15, 22, 32, 40, 47, 1, 5, 11, 18, 26, 34, 44, 51,
+ 2, 7, 13, 20, 28, 38, 46, 54, 4, 10, 16, 24, 31, 41, 50, 56,
+ 6, 12, 21, 27, 35, 43, 52, 58, 9, 17, 25, 33, 39, 48, 55, 60,
14, 23, 30, 37, 45, 53, 59, 62, 19, 29, 36, 42, 49, 57, 61, 63,
};
DECLARE_ALIGNED(16, static const int16_t, vp10_row_iscan_8x8[64]) = {
- 0, 1, 2, 5, 8, 12, 19, 24, 3, 4, 7, 10, 15, 20, 30, 39,
- 6, 9, 13, 16, 21, 27, 37, 46, 11, 14, 17, 23, 28, 34, 44, 52,
+ 0, 1, 2, 5, 8, 12, 19, 24, 3, 4, 7, 10, 15, 20, 30, 39,
+ 6, 9, 13, 16, 21, 27, 37, 46, 11, 14, 17, 23, 28, 34, 44, 52,
18, 22, 25, 31, 35, 41, 50, 57, 26, 29, 33, 38, 43, 49, 55, 59,
32, 36, 42, 47, 51, 54, 60, 61, 40, 45, 48, 53, 56, 58, 62, 63,
};
DECLARE_ALIGNED(16, static const int16_t, vp10_default_iscan_8x8[64]) = {
- 0, 2, 5, 9, 14, 22, 31, 37, 1, 4, 8, 13, 19, 26, 38, 44,
- 3, 6, 10, 17, 24, 30, 42, 49, 7, 11, 15, 21, 29, 36, 47, 53,
+ 0, 2, 5, 9, 14, 22, 31, 37, 1, 4, 8, 13, 19, 26, 38, 44,
+ 3, 6, 10, 17, 24, 30, 42, 49, 7, 11, 15, 21, 29, 36, 47, 53,
12, 16, 20, 27, 34, 43, 52, 57, 18, 23, 28, 35, 41, 48, 56, 60,
25, 32, 39, 45, 50, 55, 59, 62, 33, 40, 46, 51, 54, 58, 61, 63,
};
DECLARE_ALIGNED(16, static const int16_t, vp10_col_iscan_16x16[256]) = {
- 0, 4, 11, 20, 31, 43, 59, 75, 85, 109, 130, 150, 165, 181, 195, 198,
- 1, 6, 14, 23, 34, 47, 64, 81, 95, 114, 135, 153, 171, 188, 201, 212,
- 2, 8, 16, 25, 38, 52, 67, 83, 101, 116, 136, 157, 172, 190, 205, 216,
- 3, 10, 18, 29, 41, 55, 71, 89, 103, 119, 141, 159, 176, 194, 208, 218,
- 5, 12, 21, 32, 45, 58, 74, 93, 104, 123, 144, 164, 179, 196, 210, 223,
- 7, 15, 26, 37, 49, 63, 78, 96, 112, 129, 146, 166, 182, 200, 215, 228,
- 9, 19, 28, 39, 54, 69, 86, 102, 117, 132, 151, 170, 187, 206, 220, 230,
- 13, 24, 35, 46, 60, 73, 91, 108, 122, 137, 154, 174, 189, 207, 224, 235,
- 17, 30, 40, 53, 66, 82, 98, 115, 126, 142, 161, 180, 197, 213, 227, 237,
- 22, 36, 48, 62, 76, 92, 105, 120, 133, 147, 167, 186, 203, 219, 232, 240,
- 27, 44, 56, 70, 84, 99, 113, 127, 140, 156, 175, 193, 209, 226, 236, 244,
- 33, 51, 68, 79, 94, 110, 125, 138, 149, 162, 184, 202, 217, 229, 241, 247,
- 42, 61, 77, 90, 106, 121, 134, 148, 160, 173, 191, 211, 225, 238, 245, 251,
- 50, 72, 87, 100, 118, 128, 145, 158, 168, 183, 204, 222, 233, 242, 249, 253,
- 57, 80, 97, 111, 131, 143, 155, 169, 178, 192, 214, 231, 239, 246, 250, 254,
+ 0, 4, 11, 20, 31, 43, 59, 75, 85, 109, 130, 150, 165, 181, 195, 198,
+ 1, 6, 14, 23, 34, 47, 64, 81, 95, 114, 135, 153, 171, 188, 201, 212,
+ 2, 8, 16, 25, 38, 52, 67, 83, 101, 116, 136, 157, 172, 190, 205, 216,
+ 3, 10, 18, 29, 41, 55, 71, 89, 103, 119, 141, 159, 176, 194, 208, 218,
+ 5, 12, 21, 32, 45, 58, 74, 93, 104, 123, 144, 164, 179, 196, 210, 223,
+ 7, 15, 26, 37, 49, 63, 78, 96, 112, 129, 146, 166, 182, 200, 215, 228,
+ 9, 19, 28, 39, 54, 69, 86, 102, 117, 132, 151, 170, 187, 206, 220, 230,
+ 13, 24, 35, 46, 60, 73, 91, 108, 122, 137, 154, 174, 189, 207, 224, 235,
+ 17, 30, 40, 53, 66, 82, 98, 115, 126, 142, 161, 180, 197, 213, 227, 237,
+ 22, 36, 48, 62, 76, 92, 105, 120, 133, 147, 167, 186, 203, 219, 232, 240,
+ 27, 44, 56, 70, 84, 99, 113, 127, 140, 156, 175, 193, 209, 226, 236, 244,
+ 33, 51, 68, 79, 94, 110, 125, 138, 149, 162, 184, 202, 217, 229, 241, 247,
+ 42, 61, 77, 90, 106, 121, 134, 148, 160, 173, 191, 211, 225, 238, 245, 251,
+ 50, 72, 87, 100, 118, 128, 145, 158, 168, 183, 204, 222, 233, 242, 249, 253,
+ 57, 80, 97, 111, 131, 143, 155, 169, 178, 192, 214, 231, 239, 246, 250, 254,
65, 88, 107, 124, 139, 152, 163, 177, 185, 199, 221, 234, 243, 248, 252, 255,
};
DECLARE_ALIGNED(16, static const int16_t, vp10_row_iscan_16x16[256]) = {
- 0, 1, 2, 4, 6, 9, 12, 17, 22, 29, 36, 43, 54, 64, 76, 86,
- 3, 5, 7, 11, 15, 19, 25, 32, 38, 48, 59, 68, 84, 99, 115, 130,
- 8, 10, 13, 18, 23, 27, 33, 42, 51, 60, 72, 88, 103, 119, 142, 167,
- 14, 16, 20, 26, 31, 37, 44, 53, 61, 73, 85, 100, 116, 135, 161, 185,
- 21, 24, 30, 35, 40, 47, 55, 65, 74, 81, 94, 112, 133, 154, 179, 205,
- 28, 34, 39, 45, 50, 58, 67, 77, 87, 96, 106, 121, 146, 169, 196, 212,
- 41, 46, 49, 56, 63, 70, 79, 90, 98, 107, 122, 138, 159, 182, 207, 222,
- 52, 57, 62, 69, 75, 83, 93, 102, 110, 120, 134, 150, 176, 195, 215, 226,
- 66, 71, 78, 82, 91, 97, 108, 113, 127, 136, 148, 168, 188, 202, 221, 232,
- 80, 89, 92, 101, 105, 114, 125, 131, 139, 151, 162, 177, 192, 208, 223, 234,
- 95, 104, 109, 117, 123, 128, 143, 144, 155, 165, 175, 190, 206, 219, 233, 239,
- 111, 118, 124, 129, 140, 147, 157, 164, 170, 181, 191, 203, 224, 230, 240,
- 243, 126, 132, 137, 145, 153, 160, 174, 178, 184, 197, 204, 216, 231, 237,
- 244, 246, 141, 149, 156, 166, 172, 180, 189, 199, 200, 210, 220, 228, 238,
- 242, 249, 251, 152, 163, 171, 183, 186, 193, 201, 211, 214, 218, 227, 236,
- 245, 247, 252, 253, 158, 173, 187, 194, 198, 209, 213, 217, 225, 229, 235,
- 241, 248, 250, 254, 255,
+ 0, 1, 2, 4, 6, 9, 12, 17, 22, 29, 36, 43, 54, 64, 76,
+ 86, 3, 5, 7, 11, 15, 19, 25, 32, 38, 48, 59, 68, 84, 99,
+ 115, 130, 8, 10, 13, 18, 23, 27, 33, 42, 51, 60, 72, 88, 103,
+ 119, 142, 167, 14, 16, 20, 26, 31, 37, 44, 53, 61, 73, 85, 100,
+ 116, 135, 161, 185, 21, 24, 30, 35, 40, 47, 55, 65, 74, 81, 94,
+ 112, 133, 154, 179, 205, 28, 34, 39, 45, 50, 58, 67, 77, 87, 96,
+ 106, 121, 146, 169, 196, 212, 41, 46, 49, 56, 63, 70, 79, 90, 98,
+ 107, 122, 138, 159, 182, 207, 222, 52, 57, 62, 69, 75, 83, 93, 102,
+ 110, 120, 134, 150, 176, 195, 215, 226, 66, 71, 78, 82, 91, 97, 108,
+ 113, 127, 136, 148, 168, 188, 202, 221, 232, 80, 89, 92, 101, 105, 114,
+ 125, 131, 139, 151, 162, 177, 192, 208, 223, 234, 95, 104, 109, 117, 123,
+ 128, 143, 144, 155, 165, 175, 190, 206, 219, 233, 239, 111, 118, 124, 129,
+ 140, 147, 157, 164, 170, 181, 191, 203, 224, 230, 240, 243, 126, 132, 137,
+ 145, 153, 160, 174, 178, 184, 197, 204, 216, 231, 237, 244, 246, 141, 149,
+ 156, 166, 172, 180, 189, 199, 200, 210, 220, 228, 238, 242, 249, 251, 152,
+ 163, 171, 183, 186, 193, 201, 211, 214, 218, 227, 236, 245, 247, 252, 253,
+ 158, 173, 187, 194, 198, 209, 213, 217, 225, 229, 235, 241, 248, 250, 254,
+ 255,
};
DECLARE_ALIGNED(16, static const int16_t, vp10_default_iscan_16x16[256]) = {
- 0, 2, 5, 9, 17, 24, 36, 44, 55, 72, 88, 104, 128, 143, 166, 179,
- 1, 4, 8, 13, 20, 30, 40, 54, 66, 79, 96, 113, 141, 154, 178, 196,
- 3, 7, 11, 18, 25, 33, 46, 57, 71, 86, 101, 119, 148, 164, 186, 201,
- 6, 12, 16, 23, 31, 39, 53, 64, 78, 92, 110, 127, 153, 169, 193, 208,
- 10, 14, 19, 28, 37, 47, 58, 67, 84, 98, 114, 133, 161, 176, 198, 214,
- 15, 21, 26, 34, 43, 52, 65, 77, 91, 106, 120, 140, 165, 185, 205, 221,
- 22, 27, 32, 41, 48, 60, 73, 85, 99, 116, 130, 151, 175, 190, 211, 225,
- 29, 35, 42, 49, 59, 69, 81, 95, 108, 125, 139, 155, 182, 197, 217, 229,
- 38, 45, 51, 61, 68, 80, 93, 105, 118, 134, 150, 168, 191, 207, 223, 234,
- 50, 56, 63, 74, 83, 94, 109, 117, 129, 147, 163, 177, 199, 213, 228, 238,
- 62, 70, 76, 87, 97, 107, 122, 131, 145, 159, 172, 188, 210, 222, 235, 242,
- 75, 82, 90, 102, 112, 124, 138, 146, 157, 173, 187, 202, 219, 230, 240, 245,
- 89, 100, 111, 123, 132, 142, 156, 167, 180, 189, 203, 216, 231, 237, 246, 250,
- 103, 115, 126, 136, 149, 162, 171, 183, 194, 204, 215, 224, 236, 241, 248,
- 252, 121, 135, 144, 158, 170, 181, 192, 200, 209, 218, 227, 233, 243, 244,
- 251, 254, 137, 152, 160, 174, 184, 195, 206, 212, 220, 226, 232, 239, 247,
- 249, 253, 255,
+ 0, 2, 5, 9, 17, 24, 36, 44, 55, 72, 88, 104, 128, 143, 166,
+ 179, 1, 4, 8, 13, 20, 30, 40, 54, 66, 79, 96, 113, 141, 154,
+ 178, 196, 3, 7, 11, 18, 25, 33, 46, 57, 71, 86, 101, 119, 148,
+ 164, 186, 201, 6, 12, 16, 23, 31, 39, 53, 64, 78, 92, 110, 127,
+ 153, 169, 193, 208, 10, 14, 19, 28, 37, 47, 58, 67, 84, 98, 114,
+ 133, 161, 176, 198, 214, 15, 21, 26, 34, 43, 52, 65, 77, 91, 106,
+ 120, 140, 165, 185, 205, 221, 22, 27, 32, 41, 48, 60, 73, 85, 99,
+ 116, 130, 151, 175, 190, 211, 225, 29, 35, 42, 49, 59, 69, 81, 95,
+ 108, 125, 139, 155, 182, 197, 217, 229, 38, 45, 51, 61, 68, 80, 93,
+ 105, 118, 134, 150, 168, 191, 207, 223, 234, 50, 56, 63, 74, 83, 94,
+ 109, 117, 129, 147, 163, 177, 199, 213, 228, 238, 62, 70, 76, 87, 97,
+ 107, 122, 131, 145, 159, 172, 188, 210, 222, 235, 242, 75, 82, 90, 102,
+ 112, 124, 138, 146, 157, 173, 187, 202, 219, 230, 240, 245, 89, 100, 111,
+ 123, 132, 142, 156, 167, 180, 189, 203, 216, 231, 237, 246, 250, 103, 115,
+ 126, 136, 149, 162, 171, 183, 194, 204, 215, 224, 236, 241, 248, 252, 121,
+ 135, 144, 158, 170, 181, 192, 200, 209, 218, 227, 233, 243, 244, 251, 254,
+ 137, 152, 160, 174, 184, 195, 206, 212, 220, 226, 232, 239, 247, 249, 253,
+ 255,
};
DECLARE_ALIGNED(16, static const int16_t, vp10_default_iscan_32x32[1024]) = {
- 0, 2, 5, 10, 17, 25, 38, 47, 62, 83, 101, 121, 145, 170, 193, 204,
- 210, 219, 229, 233, 245, 257, 275, 299, 342, 356, 377, 405, 455, 471, 495,
- 527, 1, 4, 8, 15, 22, 30, 45, 58, 74, 92, 112, 133, 158, 184, 203, 215, 222,
- 228, 234, 237, 256, 274, 298, 317, 355, 376, 404, 426, 470, 494, 526, 551,
- 3, 7, 12, 18, 28, 36, 52, 64, 82, 102, 118, 142, 164, 189, 208, 217, 224,
- 231, 235, 238, 273, 297, 316, 329, 375, 403, 425, 440, 493, 525, 550, 567,
- 6, 11, 16, 23, 31, 43, 60, 73, 90, 109, 126, 150, 173, 196, 211, 220, 226,
- 232, 236, 239, 296, 315, 328, 335, 402, 424, 439, 447, 524, 549, 566, 575,
- 9, 14, 19, 29, 37, 50, 65, 78, 95, 116, 134, 157, 179, 201, 214, 223, 244,
- 255, 272, 295, 341, 354, 374, 401, 454, 469, 492, 523, 582, 596, 617, 645,
- 13, 20, 26, 35, 44, 54, 72, 85, 105, 123, 140, 163, 182, 205, 216, 225,
- 254, 271, 294, 314, 353, 373, 400, 423, 468, 491, 522, 548, 595, 616, 644,
- 666, 21, 27, 33, 42, 53, 63, 80, 94, 113, 132, 151, 172, 190, 209, 218, 227,
- 270, 293, 313, 327, 372, 399, 422, 438, 490, 521, 547, 565, 615, 643, 665,
- 680, 24, 32, 39, 48, 57, 71, 88, 104, 120, 139, 159, 178, 197, 212, 221, 230,
- 292, 312, 326, 334, 398, 421, 437, 446, 520, 546, 564, 574, 642, 664, 679,
- 687, 34, 40, 46, 56, 68, 81, 96, 111, 130, 147, 167, 186, 243, 253, 269, 291,
- 340, 352, 371, 397, 453, 467, 489, 519, 581, 594, 614, 641, 693, 705, 723,
- 747, 41, 49, 55, 67, 77, 91, 107, 124, 138, 161, 177, 194, 252, 268, 290,
- 311, 351, 370, 396, 420, 466, 488, 518, 545, 593, 613, 640, 663, 704, 722,
- 746, 765, 51, 59, 66, 76, 89, 99, 119, 131, 149, 168, 181, 200, 267, 289,
- 310, 325, 369, 395, 419, 436, 487, 517, 544, 563, 612, 639, 662, 678, 721,
- 745, 764, 777, 61, 69, 75, 87, 100, 114, 129, 144, 162, 180, 191, 207, 288,
- 309, 324, 333, 394, 418, 435, 445, 516, 543, 562, 573, 638, 661, 677, 686,
- 744, 763, 776, 783, 70, 79, 86, 97, 108, 122, 137, 155, 242, 251, 266, 287,
- 339, 350, 368, 393, 452, 465, 486, 515, 580, 592, 611, 637, 692, 703, 720,
- 743, 788, 798, 813, 833, 84, 93, 103, 110, 125, 141, 154, 171, 250, 265, 286,
- 308, 349, 367, 392, 417, 464, 485, 514, 542, 591, 610, 636, 660, 702, 719,
- 742, 762, 797, 812, 832, 848, 98, 106, 115, 127, 143, 156, 169, 185, 264,
- 285, 307, 323, 366, 391, 416, 434, 484, 513, 541, 561, 609, 635, 659, 676,
- 718, 741, 761, 775, 811, 831, 847, 858, 117, 128, 136, 148, 160, 175, 188,
- 198, 284, 306, 322, 332, 390, 415, 433, 444, 512, 540, 560, 572, 634, 658,
- 675, 685, 740, 760, 774, 782, 830, 846, 857, 863, 135, 146, 152, 165, 241,
- 249, 263, 283, 338, 348, 365, 389, 451, 463, 483, 511, 579, 590, 608, 633,
- 691, 701, 717, 739, 787, 796, 810, 829, 867, 875, 887, 903, 153, 166, 174,
- 183, 248, 262, 282, 305, 347, 364, 388, 414, 462, 482, 510, 539, 589, 607,
- 632, 657, 700, 716, 738, 759, 795, 809, 828, 845, 874, 886, 902, 915, 176,
- 187, 195, 202, 261, 281, 304, 321, 363, 387, 413, 432, 481, 509, 538, 559,
- 606, 631, 656, 674, 715, 737, 758, 773, 808, 827, 844, 856, 885, 901, 914,
- 923, 192, 199, 206, 213, 280, 303, 320, 331, 386, 412, 431, 443, 508, 537,
- 558, 571, 630, 655, 673, 684, 736, 757, 772, 781, 826, 843, 855, 862, 900,
- 913, 922, 927, 240, 247, 260, 279, 337, 346, 362, 385, 450, 461, 480, 507,
- 578, 588, 605, 629, 690, 699, 714, 735, 786, 794, 807, 825, 866, 873, 884,
- 899, 930, 936, 945, 957, 246, 259, 278, 302, 345, 361, 384, 411, 460, 479,
- 506, 536, 587, 604, 628, 654, 698, 713, 734, 756, 793, 806, 824, 842, 872,
- 883, 898, 912, 935, 944, 956, 966, 258, 277, 301, 319, 360, 383, 410, 430,
- 478, 505, 535, 557, 603, 627, 653, 672, 712, 733, 755, 771, 805, 823, 841,
- 854, 882, 897, 911, 921, 943, 955, 965, 972, 276, 300, 318, 330, 382, 409,
- 429, 442, 504, 534, 556, 570, 626, 652, 671, 683, 732, 754, 770, 780, 822,
- 840, 853, 861, 896, 910, 920, 926, 954, 964, 971, 975, 336, 344, 359, 381,
- 449, 459, 477, 503, 577, 586, 602, 625, 689, 697, 711, 731, 785, 792, 804,
- 821, 865, 871, 881, 895, 929, 934, 942, 953, 977, 981, 987, 995, 343, 358,
- 380, 408, 458, 476, 502, 533, 585, 601, 624, 651, 696, 710, 730, 753, 791,
- 803, 820, 839, 870, 880, 894, 909, 933, 941, 952, 963, 980, 986, 994, 1001,
- 357, 379, 407, 428, 475, 501, 532, 555, 600, 623, 650, 670, 709, 729, 752,
- 769, 802, 819, 838, 852, 879, 893, 908, 919, 940, 951, 962, 970, 985, 993,
- 1000, 1005, 378, 406, 427, 441, 500, 531, 554, 569, 622, 649, 669, 682, 728,
- 751, 768, 779, 818, 837, 851, 860, 892, 907, 918, 925, 950, 961, 969, 974,
- 992, 999, 1004, 1007, 448, 457, 474, 499, 576, 584, 599, 621, 688, 695, 708,
- 727, 784, 790, 801, 817, 864, 869, 878, 891, 928, 932, 939, 949, 976, 979,
- 984, 991, 1008, 1010, 1013, 1017, 456, 473, 498, 530, 583, 598, 620, 648,
- 694, 707, 726, 750, 789, 800, 816, 836, 868, 877, 890, 906, 931, 938, 948,
- 960, 978, 983, 990, 998, 1009, 1012, 1016, 1020, 472, 497, 529, 553, 597,
- 619, 647, 668, 706, 725, 749, 767, 799, 815, 835, 850, 876, 889, 905, 917,
- 937, 947, 959, 968, 982, 989, 997, 1003, 1011, 1015, 1019, 1022, 496, 528,
- 552, 568, 618, 646, 667, 681, 724, 748, 766, 778, 814, 834, 849, 859, 888,
- 904, 916, 924, 946, 958, 967, 973, 988, 996, 1002, 1006, 1014, 1018, 1021,
- 1023,
+ 0, 2, 5, 10, 17, 25, 38, 47, 62, 83, 101, 121, 145,
+ 170, 193, 204, 210, 219, 229, 233, 245, 257, 275, 299, 342, 356,
+ 377, 405, 455, 471, 495, 527, 1, 4, 8, 15, 22, 30, 45,
+ 58, 74, 92, 112, 133, 158, 184, 203, 215, 222, 228, 234, 237,
+ 256, 274, 298, 317, 355, 376, 404, 426, 470, 494, 526, 551, 3,
+ 7, 12, 18, 28, 36, 52, 64, 82, 102, 118, 142, 164, 189,
+ 208, 217, 224, 231, 235, 238, 273, 297, 316, 329, 375, 403, 425,
+ 440, 493, 525, 550, 567, 6, 11, 16, 23, 31, 43, 60, 73,
+ 90, 109, 126, 150, 173, 196, 211, 220, 226, 232, 236, 239, 296,
+ 315, 328, 335, 402, 424, 439, 447, 524, 549, 566, 575, 9, 14,
+ 19, 29, 37, 50, 65, 78, 95, 116, 134, 157, 179, 201, 214,
+ 223, 244, 255, 272, 295, 341, 354, 374, 401, 454, 469, 492, 523,
+ 582, 596, 617, 645, 13, 20, 26, 35, 44, 54, 72, 85, 105,
+ 123, 140, 163, 182, 205, 216, 225, 254, 271, 294, 314, 353, 373,
+ 400, 423, 468, 491, 522, 548, 595, 616, 644, 666, 21, 27, 33,
+ 42, 53, 63, 80, 94, 113, 132, 151, 172, 190, 209, 218, 227,
+ 270, 293, 313, 327, 372, 399, 422, 438, 490, 521, 547, 565, 615,
+ 643, 665, 680, 24, 32, 39, 48, 57, 71, 88, 104, 120, 139,
+ 159, 178, 197, 212, 221, 230, 292, 312, 326, 334, 398, 421, 437,
+ 446, 520, 546, 564, 574, 642, 664, 679, 687, 34, 40, 46, 56,
+ 68, 81, 96, 111, 130, 147, 167, 186, 243, 253, 269, 291, 340,
+ 352, 371, 397, 453, 467, 489, 519, 581, 594, 614, 641, 693, 705,
+ 723, 747, 41, 49, 55, 67, 77, 91, 107, 124, 138, 161, 177,
+ 194, 252, 268, 290, 311, 351, 370, 396, 420, 466, 488, 518, 545,
+ 593, 613, 640, 663, 704, 722, 746, 765, 51, 59, 66, 76, 89,
+ 99, 119, 131, 149, 168, 181, 200, 267, 289, 310, 325, 369, 395,
+ 419, 436, 487, 517, 544, 563, 612, 639, 662, 678, 721, 745, 764,
+ 777, 61, 69, 75, 87, 100, 114, 129, 144, 162, 180, 191, 207,
+ 288, 309, 324, 333, 394, 418, 435, 445, 516, 543, 562, 573, 638,
+ 661, 677, 686, 744, 763, 776, 783, 70, 79, 86, 97, 108, 122,
+ 137, 155, 242, 251, 266, 287, 339, 350, 368, 393, 452, 465, 486,
+ 515, 580, 592, 611, 637, 692, 703, 720, 743, 788, 798, 813, 833,
+ 84, 93, 103, 110, 125, 141, 154, 171, 250, 265, 286, 308, 349,
+ 367, 392, 417, 464, 485, 514, 542, 591, 610, 636, 660, 702, 719,
+ 742, 762, 797, 812, 832, 848, 98, 106, 115, 127, 143, 156, 169,
+ 185, 264, 285, 307, 323, 366, 391, 416, 434, 484, 513, 541, 561,
+ 609, 635, 659, 676, 718, 741, 761, 775, 811, 831, 847, 858, 117,
+ 128, 136, 148, 160, 175, 188, 198, 284, 306, 322, 332, 390, 415,
+ 433, 444, 512, 540, 560, 572, 634, 658, 675, 685, 740, 760, 774,
+ 782, 830, 846, 857, 863, 135, 146, 152, 165, 241, 249, 263, 283,
+ 338, 348, 365, 389, 451, 463, 483, 511, 579, 590, 608, 633, 691,
+ 701, 717, 739, 787, 796, 810, 829, 867, 875, 887, 903, 153, 166,
+ 174, 183, 248, 262, 282, 305, 347, 364, 388, 414, 462, 482, 510,
+ 539, 589, 607, 632, 657, 700, 716, 738, 759, 795, 809, 828, 845,
+ 874, 886, 902, 915, 176, 187, 195, 202, 261, 281, 304, 321, 363,
+ 387, 413, 432, 481, 509, 538, 559, 606, 631, 656, 674, 715, 737,
+ 758, 773, 808, 827, 844, 856, 885, 901, 914, 923, 192, 199, 206,
+ 213, 280, 303, 320, 331, 386, 412, 431, 443, 508, 537, 558, 571,
+ 630, 655, 673, 684, 736, 757, 772, 781, 826, 843, 855, 862, 900,
+ 913, 922, 927, 240, 247, 260, 279, 337, 346, 362, 385, 450, 461,
+ 480, 507, 578, 588, 605, 629, 690, 699, 714, 735, 786, 794, 807,
+ 825, 866, 873, 884, 899, 930, 936, 945, 957, 246, 259, 278, 302,
+ 345, 361, 384, 411, 460, 479, 506, 536, 587, 604, 628, 654, 698,
+ 713, 734, 756, 793, 806, 824, 842, 872, 883, 898, 912, 935, 944,
+ 956, 966, 258, 277, 301, 319, 360, 383, 410, 430, 478, 505, 535,
+ 557, 603, 627, 653, 672, 712, 733, 755, 771, 805, 823, 841, 854,
+ 882, 897, 911, 921, 943, 955, 965, 972, 276, 300, 318, 330, 382,
+ 409, 429, 442, 504, 534, 556, 570, 626, 652, 671, 683, 732, 754,
+ 770, 780, 822, 840, 853, 861, 896, 910, 920, 926, 954, 964, 971,
+ 975, 336, 344, 359, 381, 449, 459, 477, 503, 577, 586, 602, 625,
+ 689, 697, 711, 731, 785, 792, 804, 821, 865, 871, 881, 895, 929,
+ 934, 942, 953, 977, 981, 987, 995, 343, 358, 380, 408, 458, 476,
+ 502, 533, 585, 601, 624, 651, 696, 710, 730, 753, 791, 803, 820,
+ 839, 870, 880, 894, 909, 933, 941, 952, 963, 980, 986, 994, 1001,
+ 357, 379, 407, 428, 475, 501, 532, 555, 600, 623, 650, 670, 709,
+ 729, 752, 769, 802, 819, 838, 852, 879, 893, 908, 919, 940, 951,
+ 962, 970, 985, 993, 1000, 1005, 378, 406, 427, 441, 500, 531, 554,
+ 569, 622, 649, 669, 682, 728, 751, 768, 779, 818, 837, 851, 860,
+ 892, 907, 918, 925, 950, 961, 969, 974, 992, 999, 1004, 1007, 448,
+ 457, 474, 499, 576, 584, 599, 621, 688, 695, 708, 727, 784, 790,
+ 801, 817, 864, 869, 878, 891, 928, 932, 939, 949, 976, 979, 984,
+ 991, 1008, 1010, 1013, 1017, 456, 473, 498, 530, 583, 598, 620, 648,
+ 694, 707, 726, 750, 789, 800, 816, 836, 868, 877, 890, 906, 931,
+ 938, 948, 960, 978, 983, 990, 998, 1009, 1012, 1016, 1020, 472, 497,
+ 529, 553, 597, 619, 647, 668, 706, 725, 749, 767, 799, 815, 835,
+ 850, 876, 889, 905, 917, 937, 947, 959, 968, 982, 989, 997, 1003,
+ 1011, 1015, 1019, 1022, 496, 528, 552, 568, 618, 646, 667, 681, 724,
+ 748, 766, 778, 814, 834, 849, 859, 888, 904, 916, 924, 946, 958,
+ 967, 973, 988, 996, 1002, 1006, 1014, 1018, 1021, 1023,
};
const scan_order vp10_default_scan_orders[TX_SIZES] = {
- {default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
- {default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
- {default_scan_16x16, vp10_default_iscan_16x16, default_scan_16x16_neighbors},
- {default_scan_32x32, vp10_default_iscan_32x32, default_scan_32x32_neighbors},
+ { default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors },
+ { default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors },
+ { default_scan_16x16, vp10_default_iscan_16x16,
+ default_scan_16x16_neighbors },
+ { default_scan_32x32, vp10_default_iscan_32x32,
+ default_scan_32x32_neighbors },
};
const scan_order vp10_scan_orders[TX_SIZES][TX_TYPES] = {
- { // TX_4X4
- {default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
- {row_scan_4x4, vp10_row_iscan_4x4, row_scan_4x4_neighbors},
- {col_scan_4x4, vp10_col_iscan_4x4, col_scan_4x4_neighbors},
- {default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors}
- }, { // TX_8X8
- {default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
- {row_scan_8x8, vp10_row_iscan_8x8, row_scan_8x8_neighbors},
- {col_scan_8x8, vp10_col_iscan_8x8, col_scan_8x8_neighbors},
- {default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors}
- }, { // TX_16X16
- {default_scan_16x16, vp10_default_iscan_16x16, default_scan_16x16_neighbors},
- {row_scan_16x16, vp10_row_iscan_16x16, row_scan_16x16_neighbors},
- {col_scan_16x16, vp10_col_iscan_16x16, col_scan_16x16_neighbors},
- {default_scan_16x16, vp10_default_iscan_16x16, default_scan_16x16_neighbors}
- }, { // TX_32X32
- {default_scan_32x32, vp10_default_iscan_32x32, default_scan_32x32_neighbors},
- {default_scan_32x32, vp10_default_iscan_32x32, default_scan_32x32_neighbors},
- {default_scan_32x32, vp10_default_iscan_32x32, default_scan_32x32_neighbors},
- {default_scan_32x32, vp10_default_iscan_32x32, default_scan_32x32_neighbors},
+ { // TX_4X4
+ { default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors },
+ { row_scan_4x4, vp10_row_iscan_4x4, row_scan_4x4_neighbors },
+ { col_scan_4x4, vp10_col_iscan_4x4, col_scan_4x4_neighbors },
+ { default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors } },
+ { // TX_8X8
+ { default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors },
+ { row_scan_8x8, vp10_row_iscan_8x8, row_scan_8x8_neighbors },
+ { col_scan_8x8, vp10_col_iscan_8x8, col_scan_8x8_neighbors },
+ { default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors } },
+ { // TX_16X16
+ { default_scan_16x16, vp10_default_iscan_16x16,
+ default_scan_16x16_neighbors },
+ { row_scan_16x16, vp10_row_iscan_16x16, row_scan_16x16_neighbors },
+ { col_scan_16x16, vp10_col_iscan_16x16, col_scan_16x16_neighbors },
+ { default_scan_16x16, vp10_default_iscan_16x16,
+ default_scan_16x16_neighbors } },
+ {
+ // TX_32X32
+ { default_scan_32x32, vp10_default_iscan_32x32,
+ default_scan_32x32_neighbors },
+ { default_scan_32x32, vp10_default_iscan_32x32,
+ default_scan_32x32_neighbors },
+ { default_scan_32x32, vp10_default_iscan_32x32,
+ default_scan_32x32_neighbors },
+ { default_scan_32x32, vp10_default_iscan_32x32,
+ default_scan_32x32_neighbors },
}
};
diff --git a/vp10/common/scan.h b/vp10/common/scan.h
index f5a020f..acff121 100644
--- a/vp10/common/scan.h
+++ b/vp10/common/scan.h
@@ -35,7 +35,8 @@
static INLINE int get_coef_context(const int16_t *neighbors,
const uint8_t *token_cache, int c) {
return (1 + token_cache[neighbors[MAX_NEIGHBORS * c + 0]] +
- token_cache[neighbors[MAX_NEIGHBORS * c + 1]]) >> 1;
+ token_cache[neighbors[MAX_NEIGHBORS * c + 1]]) >>
+ 1;
}
static INLINE const scan_order *get_scan(TX_SIZE tx_size, TX_TYPE tx_type) {
diff --git a/vp10/common/seg_common.c b/vp10/common/seg_common.c
index 1bf09b9..7e5dcc2 100644
--- a/vp10/common/seg_common.c
+++ b/vp10/common/seg_common.c
@@ -17,8 +17,8 @@
static const int seg_feature_data_signed[SEG_LVL_MAX] = { 1, 1, 0, 0 };
-static const int seg_feature_data_max[SEG_LVL_MAX] = {
- MAXQ, MAX_LOOP_FILTER, 3, 0 };
+static const int seg_feature_data_max[SEG_LVL_MAX] = { MAXQ, MAX_LOOP_FILTER, 3,
+ 0 };
// These functions provide access to new segment level features.
// Eventually these function may be "optimized out" but for the moment,
@@ -31,7 +31,7 @@
}
void vp10_enable_segfeature(struct segmentation *seg, int segment_id,
- SEG_LVL_FEATURES feature_id) {
+ SEG_LVL_FEATURES feature_id) {
seg->feature_mask[segment_id] |= 1 << feature_id;
}
@@ -44,7 +44,7 @@
}
void vp10_set_segdata(struct segmentation *seg, int segment_id,
- SEG_LVL_FEATURES feature_id, int seg_data) {
+ SEG_LVL_FEATURES feature_id, int seg_data) {
assert(seg_data <= seg_feature_data_max[feature_id]);
if (seg_data < 0) {
assert(seg_feature_data_signed[feature_id]);
@@ -55,9 +55,7 @@
}
const vpx_tree_index vp10_segment_tree[TREE_SIZE(MAX_SEGMENTS)] = {
- 2, 4, 6, 8, 10, 12,
- 0, -1, -2, -3, -4, -5, -6, -7
+ 2, 4, 6, 8, 10, 12, 0, -1, -2, -3, -4, -5, -6, -7
};
-
// TBD? Functions to read and write segment data with range / validity checking
diff --git a/vp10/common/seg_common.h b/vp10/common/seg_common.h
index cd38e8e..4f8b80e 100644
--- a/vp10/common/seg_common.h
+++ b/vp10/common/seg_common.h
@@ -17,24 +17,23 @@
extern "C" {
#endif
-#define SEGMENT_DELTADATA 0
-#define SEGMENT_ABSDATA 1
+#define SEGMENT_DELTADATA 0
+#define SEGMENT_ABSDATA 1
-#define MAX_SEGMENTS 8
-#define SEG_TREE_PROBS (MAX_SEGMENTS-1)
+#define MAX_SEGMENTS 8
+#define SEG_TREE_PROBS (MAX_SEGMENTS - 1)
#define PREDICTION_PROBS 3
// Segment level features.
typedef enum {
- SEG_LVL_ALT_Q = 0, // Use alternate Quantizer ....
- SEG_LVL_ALT_LF = 1, // Use alternate loop filter value...
- SEG_LVL_REF_FRAME = 2, // Optional Segment reference frame
- SEG_LVL_SKIP = 3, // Optional Segment (0,0) + skip mode
- SEG_LVL_MAX = 4 // Number of features supported
+ SEG_LVL_ALT_Q = 0, // Use alternate Quantizer ....
+ SEG_LVL_ALT_LF = 1, // Use alternate loop filter value...
+ SEG_LVL_REF_FRAME = 2, // Optional Segment reference frame
+ SEG_LVL_SKIP = 3, // Optional Segment (0,0) + skip mode
+ SEG_LVL_MAX = 4 // Number of features supported
} SEG_LVL_FEATURES;
-
struct segmentation {
uint8_t enabled;
uint8_t update_map;
@@ -54,24 +53,20 @@
static INLINE int segfeature_active(const struct segmentation *seg,
int segment_id,
SEG_LVL_FEATURES feature_id) {
- return seg->enabled &&
- (seg->feature_mask[segment_id] & (1 << feature_id));
+ return seg->enabled && (seg->feature_mask[segment_id] & (1 << feature_id));
}
void vp10_clearall_segfeatures(struct segmentation *seg);
-void vp10_enable_segfeature(struct segmentation *seg,
- int segment_id,
- SEG_LVL_FEATURES feature_id);
+void vp10_enable_segfeature(struct segmentation *seg, int segment_id,
+ SEG_LVL_FEATURES feature_id);
int vp10_seg_feature_data_max(SEG_LVL_FEATURES feature_id);
int vp10_is_segfeature_signed(SEG_LVL_FEATURES feature_id);
-void vp10_set_segdata(struct segmentation *seg,
- int segment_id,
- SEG_LVL_FEATURES feature_id,
- int seg_data);
+void vp10_set_segdata(struct segmentation *seg, int segment_id,
+ SEG_LVL_FEATURES feature_id, int seg_data);
static INLINE int get_segdata(const struct segmentation *seg, int segment_id,
SEG_LVL_FEATURES feature_id) {
@@ -85,4 +80,3 @@
#endif
#endif // VP10_COMMON_SEG_COMMON_H_
-
diff --git a/vp10/common/thread_common.c b/vp10/common/thread_common.c
index bb4c8bb..3678c3b 100644
--- a/vp10/common/thread_common.c
+++ b/vp10/common/thread_common.c
@@ -29,8 +29,7 @@
}
}
- if (!locked)
- pthread_mutex_lock(mutex);
+ if (!locked) pthread_mutex_lock(mutex);
}
#endif // CONFIG_MULTITHREAD
@@ -64,8 +63,7 @@
if (c < sb_cols - 1) {
cur = c;
- if (c % nsync)
- sig = 0;
+ if (c % nsync) sig = 0;
} else {
cur = sb_cols + nsync;
}
@@ -87,12 +85,10 @@
}
// Implement row loopfiltering for each thread.
-static INLINE
-void thread_loop_filter_rows(const YV12_BUFFER_CONFIG *const frame_buffer,
- VP10_COMMON *const cm,
- struct macroblockd_plane planes[MAX_MB_PLANE],
- int start, int stop, int y_only,
- VP10LfSync *const lf_sync) {
+static INLINE void thread_loop_filter_rows(
+ const YV12_BUFFER_CONFIG *const frame_buffer, VP10_COMMON *const cm,
+ struct macroblockd_plane planes[MAX_MB_PLANE], int start, int stop,
+ int y_only, VP10LfSync *const lf_sync) {
const int num_planes = y_only ? 1 : MAX_MB_PLANE;
const int sb_cols = mi_cols_aligned_to_sb(cm->mi_cols) >> MI_BLOCK_SIZE_LOG2;
int mi_row, mi_col;
@@ -121,8 +117,7 @@
vp10_setup_dst_planes(planes, frame_buffer, mi_row, mi_col);
// TODO(JBB): Make setup_mask work for non 420.
- vp10_setup_mask(cm, mi_row, mi_col, mi + mi_col, cm->mi_stride,
- &lfm);
+ vp10_setup_mask(cm, mi_row, mi_col, mi + mi_col, cm->mi_stride, &lfm);
vp10_filter_block_plane_ss00(cm, &planes[0], mi_row, &lfm);
for (plane = 1; plane < num_planes; ++plane) {
@@ -135,7 +130,7 @@
break;
case LF_PATH_SLOW:
vp10_filter_block_plane_non420(cm, &planes[plane], mi + mi_col,
- mi_row, mi_col);
+ mi_row, mi_col);
break;
}
}
@@ -154,8 +149,7 @@
return 1;
}
-static void loop_filter_rows_mt(YV12_BUFFER_CONFIG *frame,
- VP10_COMMON *cm,
+static void loop_filter_rows_mt(YV12_BUFFER_CONFIG *frame, VP10_COMMON *cm,
struct macroblockd_plane planes[MAX_MB_PLANE],
int start, int stop, int y_only,
VPxWorker *workers, int nworkers,
@@ -214,13 +208,11 @@
}
}
-void vp10_loop_filter_frame_mt(YV12_BUFFER_CONFIG *frame,
- VP10_COMMON *cm,
- struct macroblockd_plane planes[MAX_MB_PLANE],
- int frame_filter_level,
- int y_only, int partial_frame,
- VPxWorker *workers, int num_workers,
- VP10LfSync *lf_sync) {
+void vp10_loop_filter_frame_mt(YV12_BUFFER_CONFIG *frame, VP10_COMMON *cm,
+ struct macroblockd_plane planes[MAX_MB_PLANE],
+ int frame_filter_level, int y_only,
+ int partial_frame, VPxWorker *workers,
+ int num_workers, VP10LfSync *lf_sync) {
int start_mi_row, end_mi_row, mi_rows_to_filter;
if (!frame_filter_level) return;
@@ -235,8 +227,8 @@
end_mi_row = start_mi_row + mi_rows_to_filter;
vp10_loop_filter_frame_init(cm, frame_filter_level);
- loop_filter_rows_mt(frame, cm, planes, start_mi_row, end_mi_row,
- y_only, workers, num_workers, lf_sync);
+ loop_filter_rows_mt(frame, cm, planes, start_mi_row, end_mi_row, y_only,
+ workers, num_workers, lf_sync);
}
// Set up nsync by width.
@@ -255,7 +247,7 @@
// Allocate memory for lf row synchronization
void vp10_loop_filter_alloc(VP10LfSync *lf_sync, VP10_COMMON *cm, int rows,
- int width, int num_workers) {
+ int width, int num_workers) {
lf_sync->rows = rows;
#if CONFIG_MULTITHREAD
{
@@ -319,7 +311,7 @@
// Accumulate frame counts.
void vp10_accumulate_frame_counts(VP10_COMMON *cm, FRAME_COUNTS *counts,
- int is_dec) {
+ int is_dec) {
int i, j, k, l, m;
for (i = 0; i < BLOCK_SIZE_GROUPS; i++)
@@ -355,11 +347,11 @@
for (m = 0; m < COEFF_CONTEXTS; m++)
cm->counts.eob_branch[i][j][k][l][m] +=
counts->eob_branch[i][j][k][l][m];
- // In the encoder, cm->counts.coef is only updated at frame
- // level, so not need to accumulate it here.
- // for (n = 0; n < UNCONSTRAINED_NODES + 1; n++)
- // cm->counts.coef[i][j][k][l][m][n] +=
- // counts->coef[i][j][k][l][m][n];
+ // In the encoder, cm->counts.coef is only updated at frame
+ // level, so not need to accumulate it here.
+ // for (n = 0; n < UNCONSTRAINED_NODES + 1; n++)
+ // cm->counts.coef[i][j][k][l][m][n] +=
+ // counts->coef[i][j][k][l][m][n];
}
for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++)
@@ -381,11 +373,10 @@
for (i = 0; i < REF_CONTEXTS; i++)
for (j = 0; j < 2; j++)
for (k = 0; k < 2; k++)
- cm->counts.single_ref[i][j][k] += counts->single_ref[i][j][k];
+ cm->counts.single_ref[i][j][k] += counts->single_ref[i][j][k];
for (i = 0; i < REF_CONTEXTS; i++)
- for (j = 0; j < 2; j++)
- cm->counts.comp_ref[i][j] += counts->comp_ref[i][j];
+ for (j = 0; j < 2; j++) cm->counts.comp_ref[i][j] += counts->comp_ref[i][j];
for (i = 0; i < TX_SIZE_CONTEXTS; i++) {
for (j = 0; j < TX_SIZES; j++)
@@ -402,8 +393,7 @@
cm->counts.tx.tx_totals[i] += counts->tx.tx_totals[i];
for (i = 0; i < SKIP_CONTEXTS; i++)
- for (j = 0; j < 2; j++)
- cm->counts.skip[i][j] += counts->skip[i][j];
+ for (j = 0; j < 2; j++) cm->counts.skip[i][j] += counts->skip[i][j];
for (i = 0; i < MV_JOINTS; i++)
cm->counts.mv.joints[i] += counts->mv.joints[i];
@@ -418,8 +408,7 @@
comps->hp[i] += comps_t->hp[i];
}
- for (i = 0; i < MV_CLASSES; i++)
- comps->classes[i] += comps_t->classes[i];
+ for (i = 0; i < MV_CLASSES; i++) comps->classes[i] += comps_t->classes[i];
for (i = 0; i < CLASS0_SIZE; i++) {
comps->class0[i] += comps_t->class0[i];
@@ -428,11 +417,9 @@
}
for (i = 0; i < MV_OFFSET_BITS; i++)
- for (j = 0; j < 2; j++)
- comps->bits[i][j] += comps_t->bits[i][j];
+ for (j = 0; j < 2; j++) comps->bits[i][j] += comps_t->bits[i][j];
- for (i = 0; i < MV_FP_SIZE; i++)
- comps->fp[i] += comps_t->fp[i];
+ for (i = 0; i < MV_FP_SIZE; i++) comps->fp[i] += comps_t->fp[i];
}
for (i = 0; i < EXT_TX_SIZES; i++) {
@@ -448,8 +435,7 @@
#if CONFIG_MISC_FIXES
for (i = 0; i < PREDICTION_PROBS; i++)
- for (j = 0; j < 2; j++)
- cm->counts.seg.pred[i][j] += counts->seg.pred[i][j];
+ for (j = 0; j < 2; j++) cm->counts.seg.pred[i][j] += counts->seg.pred[i][j];
for (i = 0; i < MAX_SEGMENTS; i++) {
cm->counts.seg.tree_total[i] += counts->seg.tree_total[i];
diff --git a/vp10/common/thread_common.h b/vp10/common/thread_common.h
index 908065e..1ceb680 100644
--- a/vp10/common/thread_common.h
+++ b/vp10/common/thread_common.h
@@ -47,16 +47,14 @@
void vp10_loop_filter_dealloc(VP10LfSync *lf_sync);
// Multi-threaded loopfilter that uses the tile threads.
-void vp10_loop_filter_frame_mt(YV12_BUFFER_CONFIG *frame,
- struct VP10Common *cm,
- struct macroblockd_plane planes[MAX_MB_PLANE],
- int frame_filter_level,
- int y_only, int partial_frame,
- VPxWorker *workers, int num_workers,
- VP10LfSync *lf_sync);
+void vp10_loop_filter_frame_mt(YV12_BUFFER_CONFIG *frame, struct VP10Common *cm,
+ struct macroblockd_plane planes[MAX_MB_PLANE],
+ int frame_filter_level, int y_only,
+ int partial_frame, VPxWorker *workers,
+ int num_workers, VP10LfSync *lf_sync);
void vp10_accumulate_frame_counts(struct VP10Common *cm,
- struct FRAME_COUNTS *counts, int is_dec);
+ struct FRAME_COUNTS *counts, int is_dec);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/common/tile_common.c b/vp10/common/tile_common.c
index 4d92b4c..e38a704 100644
--- a/vp10/common/tile_common.c
+++ b/vp10/common/tile_common.c
@@ -38,20 +38,18 @@
static int get_min_log2_tile_cols(const int sb64_cols) {
int min_log2 = 0;
- while ((MAX_TILE_WIDTH_B64 << min_log2) < sb64_cols)
- ++min_log2;
+ while ((MAX_TILE_WIDTH_B64 << min_log2) < sb64_cols) ++min_log2;
return min_log2;
}
static int get_max_log2_tile_cols(const int sb64_cols) {
int max_log2 = 1;
- while ((sb64_cols >> max_log2) >= MIN_TILE_WIDTH_B64)
- ++max_log2;
+ while ((sb64_cols >> max_log2) >= MIN_TILE_WIDTH_B64) ++max_log2;
return max_log2 - 1;
}
-void vp10_get_tile_n_bits(int mi_cols,
- int *min_log2_tile_cols, int *max_log2_tile_cols) {
+void vp10_get_tile_n_bits(int mi_cols, int *min_log2_tile_cols,
+ int *max_log2_tile_cols) {
const int sb64_cols = mi_cols_aligned_to_sb(mi_cols) >> MI_BLOCK_SIZE_LOG2;
*min_log2_tile_cols = get_min_log2_tile_cols(sb64_cols);
*max_log2_tile_cols = get_max_log2_tile_cols(sb64_cols);
diff --git a/vp10/common/tile_common.h b/vp10/common/tile_common.h
index 09cf060..6591e86 100644
--- a/vp10/common/tile_common.h
+++ b/vp10/common/tile_common.h
@@ -24,14 +24,14 @@
// initializes 'tile->mi_(row|col)_(start|end)' for (row, col) based on
// 'cm->log2_tile_(rows|cols)' & 'cm->mi_(rows|cols)'
-void vp10_tile_init(TileInfo *tile, const struct VP10Common *cm,
- int row, int col);
+void vp10_tile_init(TileInfo *tile, const struct VP10Common *cm, int row,
+ int col);
void vp10_tile_set_row(TileInfo *tile, const struct VP10Common *cm, int row);
void vp10_tile_set_col(TileInfo *tile, const struct VP10Common *cm, int col);
-void vp10_get_tile_n_bits(int mi_cols,
- int *min_log2_tile_cols, int *max_log2_tile_cols);
+void vp10_get_tile_n_bits(int mi_cols, int *min_log2_tile_cols,
+ int *max_log2_tile_cols);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/common/vp10_fwd_txfm.c b/vp10/common/vp10_fwd_txfm.c
index 805f559..8ec2316 100644
--- a/vp10/common/vp10_fwd_txfm.c
+++ b/vp10/common/vp10_fwd_txfm.c
@@ -71,8 +71,7 @@
{
int i, j;
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- output[j + i * 4] = (output[j + i * 4] + 1) >> 2;
+ for (j = 0; j < 4; ++j) output[j + i * 4] = (output[j + i * 4] + 1) >> 2;
}
}
}
@@ -81,15 +80,14 @@
int r, c;
tran_low_t sum = 0;
for (r = 0; r < 4; ++r)
- for (c = 0; c < 4; ++c)
- sum += input[r * stride + c];
+ for (c = 0; c < 4; ++c) sum += input[r * stride + c];
output[0] = sum << 1;
output[1] = 0;
}
-void vp10_fdct8x8_c(const int16_t *input,
- tran_low_t *final_output, int stride) {
+void vp10_fdct8x8_c(const int16_t *input, tran_low_t *final_output,
+ int stride) {
int i, j;
tran_low_t intermediate[64];
int pass;
@@ -134,8 +132,8 @@
x3 = s0 - s3;
t0 = (x0 + x1) * cospi_16_64;
t1 = (x0 - x1) * cospi_16_64;
- t2 = x2 * cospi_24_64 + x3 * cospi_8_64;
- t3 = -x2 * cospi_8_64 + x3 * cospi_24_64;
+ t2 = x2 * cospi_24_64 + x3 * cospi_8_64;
+ t3 = -x2 * cospi_8_64 + x3 * cospi_24_64;
output[0] = (tran_low_t)fdct_round_shift(t0);
output[2] = (tran_low_t)fdct_round_shift(t2);
output[4] = (tran_low_t)fdct_round_shift(t1);
@@ -154,24 +152,23 @@
x3 = s7 + t3;
// Stage 4
- t0 = x0 * cospi_28_64 + x3 * cospi_4_64;
- t1 = x1 * cospi_12_64 + x2 * cospi_20_64;
+ t0 = x0 * cospi_28_64 + x3 * cospi_4_64;
+ t1 = x1 * cospi_12_64 + x2 * cospi_20_64;
t2 = x2 * cospi_12_64 + x1 * -cospi_20_64;
- t3 = x3 * cospi_28_64 + x0 * -cospi_4_64;
+ t3 = x3 * cospi_28_64 + x0 * -cospi_4_64;
output[1] = (tran_low_t)fdct_round_shift(t0);
output[3] = (tran_low_t)fdct_round_shift(t2);
output[5] = (tran_low_t)fdct_round_shift(t1);
output[7] = (tran_low_t)fdct_round_shift(t3);
output += 8;
}
- in = intermediate;
+ in = intermediate;
output = final_output;
}
// Rows
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- final_output[j + i * 8] /= 2;
+ for (j = 0; j < 8; ++j) final_output[j + i * 8] /= 2;
}
}
@@ -179,8 +176,7 @@
int r, c;
tran_low_t sum = 0;
for (r = 0; r < 8; ++r)
- for (c = 0; c < 8; ++c)
- sum += input[r * stride + c];
+ for (c = 0; c < 8; ++c) sum += input[r * stride + c];
output[0] = sum;
output[1] = 0;
@@ -216,11 +212,11 @@
input[3] = (in_pass0[3 * stride] + in_pass0[12 * stride]) * 4;
input[4] = (in_pass0[4 * stride] + in_pass0[11 * stride]) * 4;
input[5] = (in_pass0[5 * stride] + in_pass0[10 * stride]) * 4;
- input[6] = (in_pass0[6 * stride] + in_pass0[ 9 * stride]) * 4;
- input[7] = (in_pass0[7 * stride] + in_pass0[ 8 * stride]) * 4;
+ input[6] = (in_pass0[6 * stride] + in_pass0[9 * stride]) * 4;
+ input[7] = (in_pass0[7 * stride] + in_pass0[8 * stride]) * 4;
// Calculate input for the next 8 results.
- step1[0] = (in_pass0[7 * stride] - in_pass0[ 8 * stride]) * 4;
- step1[1] = (in_pass0[6 * stride] - in_pass0[ 9 * stride]) * 4;
+ step1[0] = (in_pass0[7 * stride] - in_pass0[8 * stride]) * 4;
+ step1[1] = (in_pass0[6 * stride] - in_pass0[9 * stride]) * 4;
step1[2] = (in_pass0[5 * stride] - in_pass0[10 * stride]) * 4;
step1[3] = (in_pass0[4 * stride] - in_pass0[11 * stride]) * 4;
step1[4] = (in_pass0[3 * stride] - in_pass0[12 * stride]) * 4;
@@ -235,11 +231,11 @@
input[3] = ((in[3 * 16] + 1) >> 2) + ((in[12 * 16] + 1) >> 2);
input[4] = ((in[4 * 16] + 1) >> 2) + ((in[11 * 16] + 1) >> 2);
input[5] = ((in[5 * 16] + 1) >> 2) + ((in[10 * 16] + 1) >> 2);
- input[6] = ((in[6 * 16] + 1) >> 2) + ((in[ 9 * 16] + 1) >> 2);
- input[7] = ((in[7 * 16] + 1) >> 2) + ((in[ 8 * 16] + 1) >> 2);
+ input[6] = ((in[6 * 16] + 1) >> 2) + ((in[9 * 16] + 1) >> 2);
+ input[7] = ((in[7 * 16] + 1) >> 2) + ((in[8 * 16] + 1) >> 2);
// Calculate input for the next 8 results.
- step1[0] = ((in[7 * 16] + 1) >> 2) - ((in[ 8 * 16] + 1) >> 2);
- step1[1] = ((in[6 * 16] + 1) >> 2) - ((in[ 9 * 16] + 1) >> 2);
+ step1[0] = ((in[7 * 16] + 1) >> 2) - ((in[8 * 16] + 1) >> 2);
+ step1[1] = ((in[6 * 16] + 1) >> 2) - ((in[9 * 16] + 1) >> 2);
step1[2] = ((in[5 * 16] + 1) >> 2) - ((in[10 * 16] + 1) >> 2);
step1[3] = ((in[4 * 16] + 1) >> 2) - ((in[11 * 16] + 1) >> 2);
step1[4] = ((in[3 * 16] + 1) >> 2) - ((in[12 * 16] + 1) >> 2);
@@ -270,7 +266,7 @@
x3 = s0 - s3;
t0 = (x0 + x1) * cospi_16_64;
t1 = (x0 - x1) * cospi_16_64;
- t2 = x3 * cospi_8_64 + x2 * cospi_24_64;
+ t2 = x3 * cospi_8_64 + x2 * cospi_24_64;
t3 = x3 * cospi_24_64 - x2 * cospi_8_64;
out[0] = (tran_low_t)fdct_round_shift(t0);
out[4] = (tran_low_t)fdct_round_shift(t2);
@@ -290,10 +286,10 @@
x3 = s7 + t3;
// Stage 4
- t0 = x0 * cospi_28_64 + x3 * cospi_4_64;
- t1 = x1 * cospi_12_64 + x2 * cospi_20_64;
+ t0 = x0 * cospi_28_64 + x3 * cospi_4_64;
+ t1 = x1 * cospi_12_64 + x2 * cospi_20_64;
t2 = x2 * cospi_12_64 + x1 * -cospi_20_64;
- t3 = x3 * cospi_28_64 + x0 * -cospi_4_64;
+ t3 = x3 * cospi_28_64 + x0 * -cospi_4_64;
out[2] = (tran_low_t)fdct_round_shift(t0);
out[6] = (tran_low_t)fdct_round_shift(t2);
out[10] = (tran_low_t)fdct_round_shift(t1);
@@ -320,12 +316,12 @@
step3[6] = step1[6] + step2[5];
step3[7] = step1[7] + step2[4];
// step 4
- temp1 = step3[1] * -cospi_8_64 + step3[6] * cospi_24_64;
- temp2 = step3[2] * cospi_24_64 + step3[5] * cospi_8_64;
+ temp1 = step3[1] * -cospi_8_64 + step3[6] * cospi_24_64;
+ temp2 = step3[2] * cospi_24_64 + step3[5] * cospi_8_64;
step2[1] = fdct_round_shift(temp1);
step2[2] = fdct_round_shift(temp2);
temp1 = step3[2] * cospi_8_64 - step3[5] * cospi_24_64;
- temp2 = step3[1] * cospi_24_64 + step3[6] * cospi_8_64;
+ temp2 = step3[1] * cospi_24_64 + step3[6] * cospi_8_64;
step2[5] = fdct_round_shift(temp1);
step2[6] = fdct_round_shift(temp2);
// step 5
@@ -338,20 +334,20 @@
step1[6] = step3[7] - step2[6];
step1[7] = step3[7] + step2[6];
// step 6
- temp1 = step1[0] * cospi_30_64 + step1[7] * cospi_2_64;
+ temp1 = step1[0] * cospi_30_64 + step1[7] * cospi_2_64;
temp2 = step1[1] * cospi_14_64 + step1[6] * cospi_18_64;
out[1] = (tran_low_t)fdct_round_shift(temp1);
out[9] = (tran_low_t)fdct_round_shift(temp2);
temp1 = step1[2] * cospi_22_64 + step1[5] * cospi_10_64;
- temp2 = step1[3] * cospi_6_64 + step1[4] * cospi_26_64;
+ temp2 = step1[3] * cospi_6_64 + step1[4] * cospi_26_64;
out[5] = (tran_low_t)fdct_round_shift(temp1);
out[13] = (tran_low_t)fdct_round_shift(temp2);
- temp1 = step1[3] * -cospi_26_64 + step1[4] * cospi_6_64;
+ temp1 = step1[3] * -cospi_26_64 + step1[4] * cospi_6_64;
temp2 = step1[2] * -cospi_10_64 + step1[5] * cospi_22_64;
out[3] = (tran_low_t)fdct_round_shift(temp1);
out[11] = (tran_low_t)fdct_round_shift(temp2);
temp1 = step1[1] * -cospi_18_64 + step1[6] * cospi_14_64;
- temp2 = step1[0] * -cospi_2_64 + step1[7] * cospi_30_64;
+ temp2 = step1[0] * -cospi_2_64 + step1[7] * cospi_30_64;
out[7] = (tran_low_t)fdct_round_shift(temp1);
out[15] = (tran_low_t)fdct_round_shift(temp2);
}
@@ -370,8 +366,7 @@
int r, c;
tran_low_t sum = 0;
for (r = 0; r < 16; ++r)
- for (c = 0; c < 16; ++c)
- sum += input[r * stride + c];
+ for (c = 0; c < 16; ++c) sum += input[r * stride + c];
output[0] = sum >> 1;
output[1] = 0;
@@ -678,36 +673,36 @@
step[31] = output[31] + output[30];
// Final stage --- outputs indices are bit-reversed.
- output[0] = step[0];
+ output[0] = step[0];
output[16] = step[1];
- output[8] = step[2];
+ output[8] = step[2];
output[24] = step[3];
- output[4] = step[4];
+ output[4] = step[4];
output[20] = step[5];
output[12] = step[6];
output[28] = step[7];
- output[2] = step[8];
+ output[2] = step[8];
output[18] = step[9];
output[10] = step[10];
output[26] = step[11];
- output[6] = step[12];
+ output[6] = step[12];
output[22] = step[13];
output[14] = step[14];
output[30] = step[15];
- output[1] = dct_32_round(step[16] * cospi_31_64 + step[31] * cospi_1_64);
+ output[1] = dct_32_round(step[16] * cospi_31_64 + step[31] * cospi_1_64);
output[17] = dct_32_round(step[17] * cospi_15_64 + step[30] * cospi_17_64);
- output[9] = dct_32_round(step[18] * cospi_23_64 + step[29] * cospi_9_64);
+ output[9] = dct_32_round(step[18] * cospi_23_64 + step[29] * cospi_9_64);
output[25] = dct_32_round(step[19] * cospi_7_64 + step[28] * cospi_25_64);
- output[5] = dct_32_round(step[20] * cospi_27_64 + step[27] * cospi_5_64);
+ output[5] = dct_32_round(step[20] * cospi_27_64 + step[27] * cospi_5_64);
output[21] = dct_32_round(step[21] * cospi_11_64 + step[26] * cospi_21_64);
output[13] = dct_32_round(step[22] * cospi_19_64 + step[25] * cospi_13_64);
output[29] = dct_32_round(step[23] * cospi_3_64 + step[24] * cospi_29_64);
- output[3] = dct_32_round(step[24] * cospi_3_64 + step[23] * -cospi_29_64);
+ output[3] = dct_32_round(step[24] * cospi_3_64 + step[23] * -cospi_29_64);
output[19] = dct_32_round(step[25] * cospi_19_64 + step[22] * -cospi_13_64);
output[11] = dct_32_round(step[26] * cospi_11_64 + step[21] * -cospi_21_64);
output[27] = dct_32_round(step[27] * cospi_27_64 + step[20] * -cospi_5_64);
- output[7] = dct_32_round(step[28] * cospi_7_64 + step[19] * -cospi_25_64);
+ output[7] = dct_32_round(step[28] * cospi_7_64 + step[19] * -cospi_25_64);
output[23] = dct_32_round(step[29] * cospi_23_64 + step[18] * -cospi_9_64);
output[15] = dct_32_round(step[30] * cospi_15_64 + step[17] * -cospi_17_64);
output[31] = dct_32_round(step[31] * cospi_31_64 + step[16] * -cospi_1_64);
@@ -720,8 +715,7 @@
// Columns
for (i = 0; i < 32; ++i) {
tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = input[j * stride + i] * 4;
+ for (j = 0; j < 32; ++j) temp_in[j] = input[j * stride + i] * 4;
vp10_fdct32(temp_in, temp_out, 0);
for (j = 0; j < 32; ++j)
output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
@@ -730,8 +724,7 @@
// Rows
for (i = 0; i < 32; ++i) {
tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = output[j + i * 32];
+ for (j = 0; j < 32; ++j) temp_in[j] = output[j + i * 32];
vp10_fdct32(temp_in, temp_out, 0);
for (j = 0; j < 32; ++j)
out[j + i * 32] =
@@ -749,8 +742,7 @@
// Columns
for (i = 0; i < 32; ++i) {
tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = input[j * stride + i] * 4;
+ for (j = 0; j < 32; ++j) temp_in[j] = input[j * stride + i] * 4;
vp10_fdct32(temp_in, temp_out, 0);
for (j = 0; j < 32; ++j)
// TODO(cd): see quality impact of only doing
@@ -762,11 +754,9 @@
// Rows
for (i = 0; i < 32; ++i) {
tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = output[j + i * 32];
+ for (j = 0; j < 32; ++j) temp_in[j] = output[j + i * 32];
vp10_fdct32(temp_in, temp_out, 1);
- for (j = 0; j < 32; ++j)
- out[j + i * 32] = (tran_low_t)temp_out[j];
+ for (j = 0; j < 32; ++j) out[j + i * 32] = (tran_low_t)temp_out[j];
}
}
@@ -774,8 +764,7 @@
int r, c;
tran_low_t sum = 0;
for (r = 0; r < 32; ++r)
- for (c = 0; c < 32; ++c)
- sum += input[r * stride + c];
+ for (c = 0; c < 32; ++c) sum += input[r * stride + c];
output[0] = sum >> 3;
output[1] = 0;
@@ -783,42 +772,42 @@
#if CONFIG_VPX_HIGHBITDEPTH
void vp10_highbd_fdct4x4_c(const int16_t *input, tran_low_t *output,
- int stride) {
+ int stride) {
vp10_fdct4x4_c(input, output, stride);
}
void vp10_highbd_fdct8x8_c(const int16_t *input, tran_low_t *final_output,
- int stride) {
+ int stride) {
vp10_fdct8x8_c(input, final_output, stride);
}
void vp10_highbd_fdct8x8_1_c(const int16_t *input, tran_low_t *final_output,
- int stride) {
+ int stride) {
vp10_fdct8x8_1_c(input, final_output, stride);
}
void vp10_highbd_fdct16x16_c(const int16_t *input, tran_low_t *output,
- int stride) {
+ int stride) {
vp10_fdct16x16_c(input, output, stride);
}
void vp10_highbd_fdct16x16_1_c(const int16_t *input, tran_low_t *output,
- int stride) {
+ int stride) {
vp10_fdct16x16_1_c(input, output, stride);
}
-void vp10_highbd_fdct32x32_c(const int16_t *input,
- tran_low_t *out, int stride) {
+void vp10_highbd_fdct32x32_c(const int16_t *input, tran_low_t *out,
+ int stride) {
vp10_fdct32x32_c(input, out, stride);
}
void vp10_highbd_fdct32x32_rd_c(const int16_t *input, tran_low_t *out,
- int stride) {
+ int stride) {
vp10_fdct32x32_rd_c(input, out, stride);
}
-void vp10_highbd_fdct32x32_1_c(const int16_t *input,
- tran_low_t *out, int stride) {
+void vp10_highbd_fdct32x32_1_c(const int16_t *input, tran_low_t *out,
+ int stride) {
vp10_fdct32x32_1_c(input, out, stride);
}
#endif // CONFIG_VPX_HIGHBITDEPTH
diff --git a/vp10/common/vp10_inv_txfm.c b/vp10/common/vp10_inv_txfm.c
index cdfa3c1..cd5eea3 100644
--- a/vp10/common/vp10_inv_txfm.c
+++ b/vp10/common/vp10_inv_txfm.c
@@ -15,8 +15,8 @@
#include "vp10/common/vp10_inv_txfm.h"
void vp10_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
-/* 4-point reversible, orthonormal inverse Walsh-Hadamard in 3.5 adds,
- 0.5 shifts per pixel. */
+ /* 4-point reversible, orthonormal inverse Walsh-Hadamard in 3.5 adds,
+ 0.5 shifts per pixel. */
int i;
tran_low_t output[16];
tran_high_t a1, b1, c1, d1, e1;
@@ -66,8 +66,7 @@
}
}
-void vp10_iwht4x4_1_add_c(const tran_low_t *in,
- uint8_t *dest,
+void vp10_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest,
int dest_stride) {
int i;
tran_high_t a1, e1;
@@ -129,8 +128,7 @@
// Columns
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j * 4 + i];
+ for (j = 0; j < 4; ++j) temp_in[j] = out[j * 4 + i];
vp10_idct4_c(temp_in, temp_out);
for (j = 0; j < 4; ++j) {
dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
@@ -140,7 +138,7 @@
}
void vp10_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest,
- int dest_stride) {
+ int dest_stride) {
int i;
tran_high_t a1;
tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), 8);
@@ -225,8 +223,7 @@
// Then transform columns
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
vp10_idct8_c(temp_in, temp_out);
for (j = 0; j < 8; ++j) {
dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
@@ -242,8 +239,7 @@
out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), 8);
a1 = ROUND_POWER_OF_TWO(out, 5);
for (j = 0; j < 8; ++j) {
- for (i = 0; i < 8; ++i)
- dest[i] = clip_pixel_add(dest[i], a1);
+ for (i = 0; i < 8; ++i) dest[i] = clip_pixel_add(dest[i], a1);
dest += stride;
}
}
@@ -298,20 +294,20 @@
tran_high_t x7 = input[6];
if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) {
- output[0] = output[1] = output[2] = output[3] = output[4]
- = output[5] = output[6] = output[7] = 0;
+ output[0] = output[1] = output[2] = output[3] = output[4] = output[5] =
+ output[6] = output[7] = 0;
return;
}
// stage 1
- s0 = (int)(cospi_2_64 * x0 + cospi_30_64 * x1);
- s1 = (int)(cospi_30_64 * x0 - cospi_2_64 * x1);
+ s0 = (int)(cospi_2_64 * x0 + cospi_30_64 * x1);
+ s1 = (int)(cospi_30_64 * x0 - cospi_2_64 * x1);
s2 = (int)(cospi_10_64 * x2 + cospi_22_64 * x3);
s3 = (int)(cospi_22_64 * x2 - cospi_10_64 * x3);
s4 = (int)(cospi_18_64 * x4 + cospi_14_64 * x5);
s5 = (int)(cospi_14_64 * x4 - cospi_18_64 * x5);
- s6 = (int)(cospi_26_64 * x6 + cospi_6_64 * x7);
- s7 = (int)(cospi_6_64 * x6 - cospi_26_64 * x7);
+ s6 = (int)(cospi_26_64 * x6 + cospi_6_64 * x7);
+ s7 = (int)(cospi_6_64 * x6 - cospi_26_64 * x7);
x0 = WRAPLOW(dct_const_round_shift(s0 + s4), 8);
x1 = WRAPLOW(dct_const_round_shift(s1 + s5), 8);
@@ -378,8 +374,7 @@
// Then transform columns
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
vp10_idct8_c(temp_in, temp_out);
for (j = 0; j < 8; ++j) {
dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
@@ -393,22 +388,22 @@
tran_high_t temp1, temp2;
// stage 1
- step1[0] = input[0/2];
- step1[1] = input[16/2];
- step1[2] = input[8/2];
- step1[3] = input[24/2];
- step1[4] = input[4/2];
- step1[5] = input[20/2];
- step1[6] = input[12/2];
- step1[7] = input[28/2];
- step1[8] = input[2/2];
- step1[9] = input[18/2];
- step1[10] = input[10/2];
- step1[11] = input[26/2];
- step1[12] = input[6/2];
- step1[13] = input[22/2];
- step1[14] = input[14/2];
- step1[15] = input[30/2];
+ step1[0] = input[0 / 2];
+ step1[1] = input[16 / 2];
+ step1[2] = input[8 / 2];
+ step1[3] = input[24 / 2];
+ step1[4] = input[4 / 2];
+ step1[5] = input[20 / 2];
+ step1[6] = input[12 / 2];
+ step1[7] = input[28 / 2];
+ step1[8] = input[2 / 2];
+ step1[9] = input[18 / 2];
+ step1[10] = input[10 / 2];
+ step1[11] = input[26 / 2];
+ step1[12] = input[6 / 2];
+ step1[13] = input[22 / 2];
+ step1[14] = input[14 / 2];
+ step1[15] = input[30 / 2];
// stage 2
step2[0] = step1[0];
@@ -554,7 +549,7 @@
}
void vp10_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest,
- int stride) {
+ int stride) {
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
int i, j;
@@ -569,8 +564,7 @@
// Then transform columns
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
vp10_idct16_c(temp_in, temp_out);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
@@ -600,21 +594,20 @@
tran_high_t x14 = input[1];
tran_high_t x15 = input[14];
- if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8
- | x9 | x10 | x11 | x12 | x13 | x14 | x15)) {
- output[0] = output[1] = output[2] = output[3] = output[4]
- = output[5] = output[6] = output[7] = output[8]
- = output[9] = output[10] = output[11] = output[12]
- = output[13] = output[14] = output[15] = 0;
+ if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8 | x9 | x10 | x11 | x12 |
+ x13 | x14 | x15)) {
+ output[0] = output[1] = output[2] = output[3] = output[4] = output[5] =
+ output[6] = output[7] = output[8] = output[9] = output[10] =
+ output[11] = output[12] = output[13] = output[14] = output[15] = 0;
return;
}
// stage 1
- s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
+ s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
- s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
+ s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
- s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
+ s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
@@ -623,9 +616,9 @@
s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
- s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
+ s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
- s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
+ s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
x0 = WRAPLOW(dct_const_round_shift(s0 + s8), 8);
x1 = WRAPLOW(dct_const_round_shift(s1 + s9), 8);
@@ -653,14 +646,14 @@
s5 = x5;
s6 = x6;
s7 = x7;
- s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
- s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
- s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
- s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
- s12 = - x12 * cospi_28_64 + x13 * cospi_4_64;
- s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
- s14 = - x14 * cospi_12_64 + x15 * cospi_20_64;
- s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
+ s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
+ s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
+ s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
+ s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
+ s12 = -x12 * cospi_28_64 + x13 * cospi_4_64;
+ s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
+ s14 = -x14 * cospi_12_64 + x15 * cospi_20_64;
+ s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
x0 = WRAPLOW(s0 + s4, 8);
x1 = WRAPLOW(s1 + s5, 8);
@@ -684,18 +677,18 @@
s1 = x1;
s2 = x2;
s3 = x3;
- s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
+ s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
s5 = x4 * cospi_24_64 - x5 * cospi_8_64;
- s6 = - x6 * cospi_24_64 + x7 * cospi_8_64;
- s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
+ s6 = -x6 * cospi_24_64 + x7 * cospi_8_64;
+ s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
s8 = x8;
s9 = x9;
s10 = x10;
s11 = x11;
- s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
+ s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
s13 = x12 * cospi_24_64 - x13 * cospi_8_64;
- s14 = - x14 * cospi_24_64 + x15 * cospi_8_64;
- s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
+ s14 = -x14 * cospi_24_64 + x15 * cospi_8_64;
+ s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
x0 = WRAPLOW(check_range(s0 + s2), 8);
x1 = WRAPLOW(check_range(s1 + s3), 8);
@@ -715,13 +708,13 @@
x15 = WRAPLOW(dct_const_round_shift(s13 - s15), 8);
// stage 4
- s2 = (- cospi_16_64) * (x2 + x3);
+ s2 = (-cospi_16_64) * (x2 + x3);
s3 = cospi_16_64 * (x2 - x3);
s6 = cospi_16_64 * (x6 + x7);
- s7 = cospi_16_64 * (- x6 + x7);
+ s7 = cospi_16_64 * (-x6 + x7);
s10 = cospi_16_64 * (x10 + x11);
- s11 = cospi_16_64 * (- x10 + x11);
- s14 = (- cospi_16_64) * (x14 + x15);
+ s11 = cospi_16_64 * (-x10 + x11);
+ s14 = (-cospi_16_64) * (x14 + x15);
s15 = cospi_16_64 * (x14 - x15);
x2 = WRAPLOW(dct_const_round_shift(s2), 8);
@@ -752,7 +745,7 @@
}
void vp10_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest,
- int stride) {
+ int stride) {
tran_low_t out[16 * 16] = { 0 };
tran_low_t *outptr = out;
int i, j;
@@ -768,19 +761,16 @@
// Then transform columns
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j*16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
vp10_idct16_c(temp_in, temp_out);
for (j = 0; j < 16; ++j) {
- dest[j * stride + i] = clip_pixel_add(
- dest[j * stride + i],
- ROUND_POWER_OF_TWO(temp_out[j], 6));
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 6));
}
}
}
-void vp10_idct16x16_1_add_c(const tran_low_t *input,
- uint8_t *dest,
+void vp10_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest,
int stride) {
int i, j;
tran_high_t a1;
@@ -788,8 +778,7 @@
out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), 8);
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 16; ++j) {
- for (i = 0; i < 16; ++i)
- dest[i] = clip_pixel_add(dest[i], a1);
+ for (i = 0; i < 16; ++i) dest[i] = clip_pixel_add(dest[i], a1);
dest += stride;
}
}
@@ -1162,7 +1151,7 @@
}
void vp10_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest,
- int stride) {
+ int stride) {
tran_low_t out[32 * 32];
tran_low_t *outptr = out;
int i, j;
@@ -1171,8 +1160,7 @@
// Rows
for (i = 0; i < 32; ++i) {
int16_t zero_coeff[16];
- for (j = 0; j < 16; ++j)
- zero_coeff[j] = input[2 * j] | input[2 * j + 1];
+ for (j = 0; j < 16; ++j) zero_coeff[j] = input[2 * j] | input[2 * j + 1];
for (j = 0; j < 8; ++j)
zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];
for (j = 0; j < 4; ++j)
@@ -1190,8 +1178,7 @@
// Columns
for (i = 0; i < 32; ++i) {
- for (j = 0; j < 32; ++j)
- temp_in[j] = out[j * 32 + i];
+ for (j = 0; j < 32; ++j) temp_in[j] = out[j * 32 + i];
vp10_idct32_c(temp_in, temp_out);
for (j = 0; j < 32; ++j) {
dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
@@ -1201,8 +1188,8 @@
}
void vp10_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest,
- int stride) {
- tran_low_t out[32 * 32] = {0};
+ int stride) {
+ tran_low_t out[32 * 32] = { 0 };
tran_low_t *outptr = out;
int i, j;
tran_low_t temp_in[32], temp_out[32];
@@ -1217,19 +1204,16 @@
// Columns
for (i = 0; i < 32; ++i) {
- for (j = 0; j < 32; ++j)
- temp_in[j] = out[j * 32 + i];
+ for (j = 0; j < 32; ++j) temp_in[j] = out[j * 32 + i];
vp10_idct32_c(temp_in, temp_out);
for (j = 0; j < 32; ++j) {
- dest[j * stride + i] = clip_pixel_add(
- dest[j * stride + i],
- ROUND_POWER_OF_TWO(temp_out[j], 6));
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 6));
}
}
}
-void vp10_idct32x32_1_add_c(const tran_low_t *input,
- uint8_t *dest,
+void vp10_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest,
int stride) {
int i, j;
tran_high_t a1;
@@ -1239,15 +1223,14 @@
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 32; ++j) {
- for (i = 0; i < 32; ++i)
- dest[i] = clip_pixel_add(dest[i], a1);
+ for (i = 0; i < 32; ++i) dest[i] = clip_pixel_add(dest[i], a1);
dest += stride;
}
}
#if CONFIG_VPX_HIGHBITDEPTH
void vp10_highbd_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
/* 4-point reversible, orthonormal inverse Walsh-Hadamard in 3.5 adds,
0.5 shifts per pixel. */
int i;
@@ -1301,14 +1284,14 @@
}
void vp10_highbd_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest8,
- int dest_stride, int bd) {
+ int dest_stride, int bd) {
int i;
tran_high_t a1, e1;
tran_low_t tmp[4];
const tran_low_t *ip = in;
tran_low_t *op = tmp;
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
- (void) bd;
+ (void)bd;
a1 = ip[0] >> UNIT_QUANT_SHIFT;
e1 = a1 >> 1;
@@ -1320,14 +1303,14 @@
for (i = 0; i < 4; i++) {
e1 = ip[0] >> 1;
a1 = ip[0] - e1;
- dest[dest_stride * 0] = highbd_clip_pixel_add(
- dest[dest_stride * 0], a1, bd);
- dest[dest_stride * 1] = highbd_clip_pixel_add(
- dest[dest_stride * 1], e1, bd);
- dest[dest_stride * 2] = highbd_clip_pixel_add(
- dest[dest_stride * 2], e1, bd);
- dest[dest_stride * 3] = highbd_clip_pixel_add(
- dest[dest_stride * 3], e1, bd);
+ dest[dest_stride * 0] =
+ highbd_clip_pixel_add(dest[dest_stride * 0], a1, bd);
+ dest[dest_stride * 1] =
+ highbd_clip_pixel_add(dest[dest_stride * 1], e1, bd);
+ dest[dest_stride * 2] =
+ highbd_clip_pixel_add(dest[dest_stride * 2], e1, bd);
+ dest[dest_stride * 3] =
+ highbd_clip_pixel_add(dest[dest_stride * 3], e1, bd);
ip++;
dest++;
}
@@ -1336,7 +1319,7 @@
void vp10_highbd_idct4_c(const tran_low_t *input, tran_low_t *output, int bd) {
tran_low_t step[4];
tran_high_t temp1, temp2;
- (void) bd;
+ (void)bd;
// stage 1
temp1 = (input[0] + input[2]) * cospi_16_64;
temp2 = (input[0] - input[2]) * cospi_16_64;
@@ -1355,7 +1338,7 @@
}
void vp10_highbd_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[4 * 4];
tran_low_t *outptr = out;
int i, j;
@@ -1371,8 +1354,7 @@
// Columns
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j * 4 + i];
+ for (j = 0; j < 4; ++j) temp_in[j] = out[j * 4 + i];
vp10_highbd_idct4_c(temp_in, temp_out, bd);
for (j = 0; j < 4; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -1382,11 +1364,11 @@
}
void vp10_highbd_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest8,
- int dest_stride, int bd) {
+ int dest_stride, int bd) {
int i;
tran_high_t a1;
- tran_low_t out = WRAPLOW(
- highbd_dct_const_round_shift(input[0] * cospi_16_64, bd), bd);
+ tran_low_t out =
+ WRAPLOW(highbd_dct_const_round_shift(input[0] * cospi_16_64, bd), bd);
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
out = WRAPLOW(highbd_dct_const_round_shift(out * cospi_16_64, bd), bd);
@@ -1447,7 +1429,7 @@
}
void vp10_highbd_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
int i, j;
@@ -1463,8 +1445,7 @@
// Then transform columns.
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
vp10_highbd_idct8_c(temp_in, temp_out, bd);
for (j = 0; j < 8; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -1474,17 +1455,16 @@
}
void vp10_highbd_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
int i, j;
tran_high_t a1;
- tran_low_t out = WRAPLOW(
- highbd_dct_const_round_shift(input[0] * cospi_16_64, bd), bd);
+ tran_low_t out =
+ WRAPLOW(highbd_dct_const_round_shift(input[0] * cospi_16_64, bd), bd);
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
out = WRAPLOW(highbd_dct_const_round_shift(out * cospi_16_64, bd), bd);
a1 = ROUND_POWER_OF_TWO(out, 5);
for (j = 0; j < 8; ++j) {
- for (i = 0; i < 8; ++i)
- dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
+ for (i = 0; i < 8; ++i) dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
dest += stride;
}
}
@@ -1496,7 +1476,7 @@
tran_low_t x1 = input[1];
tran_low_t x2 = input[2];
tran_low_t x3 = input[3];
- (void) bd;
+ (void)bd;
if (!(x0 | x1 | x2 | x3)) {
memset(output, 0, 4 * sizeof(*output));
@@ -1538,7 +1518,7 @@
tran_low_t x5 = input[4];
tran_low_t x6 = input[1];
tran_low_t x7 = input[6];
- (void) bd;
+ (void)bd;
if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) {
memset(output, 0, 8 * sizeof(*output));
@@ -1546,14 +1526,14 @@
}
// stage 1
- s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
- s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
+ s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
+ s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
s2 = cospi_10_64 * x2 + cospi_22_64 * x3;
s3 = cospi_22_64 * x2 - cospi_10_64 * x3;
s4 = cospi_18_64 * x4 + cospi_14_64 * x5;
s5 = cospi_14_64 * x4 - cospi_18_64 * x5;
- s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
- s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
+ s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
+ s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
x0 = WRAPLOW(highbd_dct_const_round_shift(s0 + s4, bd), bd);
x1 = WRAPLOW(highbd_dct_const_round_shift(s1 + s5, bd), bd);
@@ -1569,10 +1549,10 @@
s1 = x1;
s2 = x2;
s3 = x3;
- s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
- s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
- s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
- s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
+ s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
+ s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
+ s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
+ s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
x0 = WRAPLOW(s0 + s2, bd);
x1 = WRAPLOW(s1 + s3, bd);
@@ -1605,7 +1585,7 @@
}
void vp10_highbd_idct8x8_10_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[8 * 8] = { 0 };
tran_low_t *outptr = out;
int i, j;
@@ -1621,8 +1601,7 @@
}
// Then transform columns.
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
vp10_highbd_idct8_c(temp_in, temp_out, bd);
for (j = 0; j < 8; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -1634,25 +1613,25 @@
void vp10_highbd_idct16_c(const tran_low_t *input, tran_low_t *output, int bd) {
tran_low_t step1[16], step2[16];
tran_high_t temp1, temp2;
- (void) bd;
+ (void)bd;
// stage 1
- step1[0] = input[0/2];
- step1[1] = input[16/2];
- step1[2] = input[8/2];
- step1[3] = input[24/2];
- step1[4] = input[4/2];
- step1[5] = input[20/2];
- step1[6] = input[12/2];
- step1[7] = input[28/2];
- step1[8] = input[2/2];
- step1[9] = input[18/2];
- step1[10] = input[10/2];
- step1[11] = input[26/2];
- step1[12] = input[6/2];
- step1[13] = input[22/2];
- step1[14] = input[14/2];
- step1[15] = input[30/2];
+ step1[0] = input[0 / 2];
+ step1[1] = input[16 / 2];
+ step1[2] = input[8 / 2];
+ step1[3] = input[24 / 2];
+ step1[4] = input[4 / 2];
+ step1[5] = input[20 / 2];
+ step1[6] = input[12 / 2];
+ step1[7] = input[28 / 2];
+ step1[8] = input[2 / 2];
+ step1[9] = input[18 / 2];
+ step1[10] = input[10 / 2];
+ step1[11] = input[26 / 2];
+ step1[12] = input[6 / 2];
+ step1[13] = input[22 / 2];
+ step1[14] = input[14 / 2];
+ step1[15] = input[30 / 2];
// stage 2
step2[0] = step1[0];
@@ -1798,7 +1777,7 @@
}
void vp10_highbd_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
int i, j;
@@ -1814,20 +1793,16 @@
// Then transform columns.
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
vp10_highbd_idct16_c(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
- dest[j * stride + i],
- ROUND_POWER_OF_TWO(temp_out[j], 6),
- bd);
+ dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
}
}
}
-void vp10_highbd_iadst16_c(const tran_low_t *input,
- tran_low_t *output,
+void vp10_highbd_iadst16_c(const tran_low_t *input, tran_low_t *output,
int bd) {
tran_high_t s0, s1, s2, s3, s4, s5, s6, s7, s8;
tran_high_t s9, s10, s11, s12, s13, s14, s15;
@@ -1848,20 +1823,20 @@
tran_low_t x13 = input[12];
tran_low_t x14 = input[1];
tran_low_t x15 = input[14];
- (void) bd;
+ (void)bd;
- if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8
- | x9 | x10 | x11 | x12 | x13 | x14 | x15)) {
+ if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8 | x9 | x10 | x11 | x12 |
+ x13 | x14 | x15)) {
memset(output, 0, 16 * sizeof(*output));
return;
}
// stage 1
- s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
+ s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
- s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
+ s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
- s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
+ s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
@@ -1870,9 +1845,9 @@
s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
- s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
+ s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
- s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
+ s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
x0 = WRAPLOW(highbd_dct_const_round_shift(s0 + s8, bd), bd);
x1 = WRAPLOW(highbd_dct_const_round_shift(s1 + s9, bd), bd);
@@ -1882,8 +1857,8 @@
x5 = WRAPLOW(highbd_dct_const_round_shift(s5 + s13, bd), bd);
x6 = WRAPLOW(highbd_dct_const_round_shift(s6 + s14, bd), bd);
x7 = WRAPLOW(highbd_dct_const_round_shift(s7 + s15, bd), bd);
- x8 = WRAPLOW(highbd_dct_const_round_shift(s0 - s8, bd), bd);
- x9 = WRAPLOW(highbd_dct_const_round_shift(s1 - s9, bd), bd);
+ x8 = WRAPLOW(highbd_dct_const_round_shift(s0 - s8, bd), bd);
+ x9 = WRAPLOW(highbd_dct_const_round_shift(s1 - s9, bd), bd);
x10 = WRAPLOW(highbd_dct_const_round_shift(s2 - s10, bd), bd);
x11 = WRAPLOW(highbd_dct_const_round_shift(s3 - s11, bd), bd);
x12 = WRAPLOW(highbd_dct_const_round_shift(s4 - s12, bd), bd);
@@ -1962,13 +1937,13 @@
x15 = WRAPLOW(highbd_dct_const_round_shift(s13 - s15, bd), bd);
// stage 4
- s2 = (- cospi_16_64) * (x2 + x3);
+ s2 = (-cospi_16_64) * (x2 + x3);
s3 = cospi_16_64 * (x2 - x3);
s6 = cospi_16_64 * (x6 + x7);
s7 = cospi_16_64 * (-x6 + x7);
s10 = cospi_16_64 * (x10 + x11);
s11 = cospi_16_64 * (-x10 + x11);
- s14 = (- cospi_16_64) * (x14 + x15);
+ s14 = (-cospi_16_64) * (x14 + x15);
s15 = cospi_16_64 * (x14 - x15);
x2 = WRAPLOW(highbd_dct_const_round_shift(s2, bd), bd);
@@ -1999,7 +1974,7 @@
}
void vp10_highbd_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[16 * 16] = { 0 };
tran_low_t *outptr = out;
int i, j;
@@ -2016,8 +1991,7 @@
// Then transform columns.
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j*16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
vp10_highbd_idct16_c(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -2027,27 +2001,26 @@
}
void vp10_highbd_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
int i, j;
tran_high_t a1;
- tran_low_t out = WRAPLOW(
- highbd_dct_const_round_shift(input[0] * cospi_16_64, bd), bd);
+ tran_low_t out =
+ WRAPLOW(highbd_dct_const_round_shift(input[0] * cospi_16_64, bd), bd);
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
out = WRAPLOW(highbd_dct_const_round_shift(out * cospi_16_64, bd), bd);
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 16; ++j) {
- for (i = 0; i < 16; ++i)
- dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
+ for (i = 0; i < 16; ++i) dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
dest += stride;
}
}
-static void highbd_idct32_c(const tran_low_t *input,
- tran_low_t *output, int bd) {
+static void highbd_idct32_c(const tran_low_t *input, tran_low_t *output,
+ int bd) {
tran_low_t step1[32], step2[32];
tran_high_t temp1, temp2;
- (void) bd;
+ (void)bd;
// stage 1
step1[0] = input[0];
@@ -2413,7 +2386,7 @@
}
void vp10_highbd_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[32 * 32];
tran_low_t *outptr = out;
int i, j;
@@ -2423,8 +2396,7 @@
// Rows
for (i = 0; i < 32; ++i) {
tran_low_t zero_coeff[16];
- for (j = 0; j < 16; ++j)
- zero_coeff[j] = input[2 * j] | input[2 * j + 1];
+ for (j = 0; j < 16; ++j) zero_coeff[j] = input[2 * j] | input[2 * j + 1];
for (j = 0; j < 8; ++j)
zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];
for (j = 0; j < 4; ++j)
@@ -2442,8 +2414,7 @@
// Columns
for (i = 0; i < 32; ++i) {
- for (j = 0; j < 32; ++j)
- temp_in[j] = out[j * 32 + i];
+ for (j = 0; j < 32; ++j) temp_in[j] = out[j * 32 + i];
highbd_idct32_c(temp_in, temp_out, bd);
for (j = 0; j < 32; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -2453,8 +2424,8 @@
}
void vp10_highbd_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
- tran_low_t out[32 * 32] = {0};
+ int stride, int bd) {
+ tran_low_t out[32 * 32] = { 0 };
tran_low_t *outptr = out;
int i, j;
tran_low_t temp_in[32], temp_out[32];
@@ -2469,8 +2440,7 @@
}
// Columns
for (i = 0; i < 32; ++i) {
- for (j = 0; j < 32; ++j)
- temp_in[j] = out[j * 32 + i];
+ for (j = 0; j < 32; ++j) temp_in[j] = out[j * 32 + i];
highbd_idct32_c(temp_in, temp_out, bd);
for (j = 0; j < 32; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -2480,19 +2450,18 @@
}
void vp10_highbd_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
int i, j;
int a1;
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
- tran_low_t out = WRAPLOW(
- highbd_dct_const_round_shift(input[0] * cospi_16_64, bd), bd);
+ tran_low_t out =
+ WRAPLOW(highbd_dct_const_round_shift(input[0] * cospi_16_64, bd), bd);
out = WRAPLOW(highbd_dct_const_round_shift(out * cospi_16_64, bd), bd);
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 32; ++j) {
- for (i = 0; i < 32; ++i)
- dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
+ for (i = 0; i < 32; ++i) dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
dest += stride;
}
}
diff --git a/vp10/common/vp10_inv_txfm.h b/vp10/common/vp10_inv_txfm.h
index 7f71dd5..ec981e8 100644
--- a/vp10/common/vp10_inv_txfm.h
+++ b/vp10/common/vp10_inv_txfm.h
@@ -41,8 +41,7 @@
}
#if CONFIG_VPX_HIGHBITDEPTH
-static INLINE tran_low_t highbd_check_range(tran_high_t input,
- int bd) {
+static INLINE tran_low_t highbd_check_range(tran_high_t input, int bd) {
#if CONFIG_COEFFICIENT_RANGE_CHECKING
// For valid highbitdepth streams, intermediate stage coefficients will
// stay within the ranges:
@@ -53,9 +52,9 @@
const int32_t int_min = -int_max - 1;
assert(int_min <= input);
assert(input <= int_max);
- (void) int_min;
+ (void)int_min;
#endif // CONFIG_COEFFICIENT_RANGE_CHECKING
- (void) bd;
+ (void)bd;
return (tran_low_t)input;
}
diff --git a/vp10/common/vp10_rtcd.c b/vp10/common/vp10_rtcd.c
index 36b294a..0fd8ab0 100644
--- a/vp10/common/vp10_rtcd.c
+++ b/vp10/common/vp10_rtcd.c
@@ -13,7 +13,7 @@
#include "vpx_ports/vpx_once.h"
void vp10_rtcd() {
- // TODO(JBB): Remove this once, by insuring that both the encoder and
- // decoder setup functions are protected by once();
- once(setup_rtcd_internal);
+ // TODO(JBB): Remove this once, by insuring that both the encoder and
+ // decoder setup functions are protected by once();
+ once(setup_rtcd_internal);
}
diff --git a/vp10/common/x86/idct_intrin_sse2.c b/vp10/common/x86/idct_intrin_sse2.c
index a2c674b..792e2f9 100644
--- a/vp10/common/x86/idct_intrin_sse2.c
+++ b/vp10/common/x86/idct_intrin_sse2.c
@@ -38,9 +38,7 @@
iadst4_sse2(in);
iadst4_sse2(in);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
// Final round and shift
@@ -110,9 +108,7 @@
iadst8_sse2(in);
iadst8_sse2(in);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
// Final rounding and shift
@@ -169,9 +165,7 @@
iadst16_sse2(in0, in1);
iadst16_sse2(in0, in1);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
write_buffer_8x16(dest, in0, stride);
diff --git a/vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h b/vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h
index 77f633e..eea9f97 100644
--- a/vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h
+++ b/vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h
@@ -23,42 +23,37 @@
#define SUB_EPI16 _mm_subs_epi16
#if FDCT32x32_HIGH_PRECISION
void vp10_fdct32x32_rows_c(const int16_t *intermediate, tran_low_t *out) {
- int i, j;
- for (i = 0; i < 32; ++i) {
- tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = intermediate[j * 32 + i];
- vp10_fdct32(temp_in, temp_out, 0);
- for (j = 0; j < 32; ++j)
- out[j + i * 32] =
- (tran_low_t)((temp_out[j] + 1 + (temp_out[j] < 0)) >> 2);
- }
+ int i, j;
+ for (i = 0; i < 32; ++i) {
+ tran_high_t temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j) temp_in[j] = intermediate[j * 32 + i];
+ vp10_fdct32(temp_in, temp_out, 0);
+ for (j = 0; j < 32; ++j)
+ out[j + i * 32] =
+ (tran_low_t)((temp_out[j] + 1 + (temp_out[j] < 0)) >> 2);
+ }
}
- #define HIGH_FDCT32x32_2D_C vp10_highbd_fdct32x32_c
- #define HIGH_FDCT32x32_2D_ROWS_C vp10_fdct32x32_rows_c
+#define HIGH_FDCT32x32_2D_C vp10_highbd_fdct32x32_c
+#define HIGH_FDCT32x32_2D_ROWS_C vp10_fdct32x32_rows_c
#else
void vp10_fdct32x32_rd_rows_c(const int16_t *intermediate, tran_low_t *out) {
- int i, j;
- for (i = 0; i < 32; ++i) {
- tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = intermediate[j * 32 + i];
- vp10_fdct32(temp_in, temp_out, 1);
- for (j = 0; j < 32; ++j)
- out[j + i * 32] = (tran_low_t)temp_out[j];
- }
+ int i, j;
+ for (i = 0; i < 32; ++i) {
+ tran_high_t temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j) temp_in[j] = intermediate[j * 32 + i];
+ vp10_fdct32(temp_in, temp_out, 1);
+ for (j = 0; j < 32; ++j) out[j + i * 32] = (tran_low_t)temp_out[j];
+ }
}
- #define HIGH_FDCT32x32_2D_C vp10_highbd_fdct32x32_rd_c
- #define HIGH_FDCT32x32_2D_ROWS_C vp10_fdct32x32_rd_rows_c
+#define HIGH_FDCT32x32_2D_C vp10_highbd_fdct32x32_rd_c
+#define HIGH_FDCT32x32_2D_ROWS_C vp10_fdct32x32_rd_rows_c
#endif // FDCT32x32_HIGH_PRECISION
#else
#define ADD_EPI16 _mm_add_epi16
#define SUB_EPI16 _mm_sub_epi16
#endif // DCT_HIGH_BIT_DEPTH
-
-void FDCT32x32_2D(const int16_t *input,
- tran_low_t *output_org, int stride) {
+void FDCT32x32_2D(const int16_t *input, tran_low_t *output_org, int stride) {
// Calculate pre-multiplied strides
const int str1 = stride;
const int str2 = 2 * stride;
@@ -71,42 +66,42 @@
// by constructing the 32 bit constant corresponding to that pair.
const __m128i k__cospi_p16_p16 = _mm_set1_epi16((int16_t)cospi_16_64);
const __m128i k__cospi_p16_m16 = pair_set_epi16(+cospi_16_64, -cospi_16_64);
- const __m128i k__cospi_m08_p24 = pair_set_epi16(-cospi_8_64, cospi_24_64);
+ const __m128i k__cospi_m08_p24 = pair_set_epi16(-cospi_8_64, cospi_24_64);
const __m128i k__cospi_m24_m08 = pair_set_epi16(-cospi_24_64, -cospi_8_64);
- const __m128i k__cospi_p24_p08 = pair_set_epi16(+cospi_24_64, cospi_8_64);
- const __m128i k__cospi_p12_p20 = pair_set_epi16(+cospi_12_64, cospi_20_64);
- const __m128i k__cospi_m20_p12 = pair_set_epi16(-cospi_20_64, cospi_12_64);
- const __m128i k__cospi_m04_p28 = pair_set_epi16(-cospi_4_64, cospi_28_64);
- const __m128i k__cospi_p28_p04 = pair_set_epi16(+cospi_28_64, cospi_4_64);
+ const __m128i k__cospi_p24_p08 = pair_set_epi16(+cospi_24_64, cospi_8_64);
+ const __m128i k__cospi_p12_p20 = pair_set_epi16(+cospi_12_64, cospi_20_64);
+ const __m128i k__cospi_m20_p12 = pair_set_epi16(-cospi_20_64, cospi_12_64);
+ const __m128i k__cospi_m04_p28 = pair_set_epi16(-cospi_4_64, cospi_28_64);
+ const __m128i k__cospi_p28_p04 = pair_set_epi16(+cospi_28_64, cospi_4_64);
const __m128i k__cospi_m28_m04 = pair_set_epi16(-cospi_28_64, -cospi_4_64);
const __m128i k__cospi_m12_m20 = pair_set_epi16(-cospi_12_64, -cospi_20_64);
- const __m128i k__cospi_p30_p02 = pair_set_epi16(+cospi_30_64, cospi_2_64);
- const __m128i k__cospi_p14_p18 = pair_set_epi16(+cospi_14_64, cospi_18_64);
- const __m128i k__cospi_p22_p10 = pair_set_epi16(+cospi_22_64, cospi_10_64);
- const __m128i k__cospi_p06_p26 = pair_set_epi16(+cospi_6_64, cospi_26_64);
- const __m128i k__cospi_m26_p06 = pair_set_epi16(-cospi_26_64, cospi_6_64);
- const __m128i k__cospi_m10_p22 = pair_set_epi16(-cospi_10_64, cospi_22_64);
- const __m128i k__cospi_m18_p14 = pair_set_epi16(-cospi_18_64, cospi_14_64);
- const __m128i k__cospi_m02_p30 = pair_set_epi16(-cospi_2_64, cospi_30_64);
- const __m128i k__cospi_p31_p01 = pair_set_epi16(+cospi_31_64, cospi_1_64);
- const __m128i k__cospi_p15_p17 = pair_set_epi16(+cospi_15_64, cospi_17_64);
- const __m128i k__cospi_p23_p09 = pair_set_epi16(+cospi_23_64, cospi_9_64);
- const __m128i k__cospi_p07_p25 = pair_set_epi16(+cospi_7_64, cospi_25_64);
- const __m128i k__cospi_m25_p07 = pair_set_epi16(-cospi_25_64, cospi_7_64);
- const __m128i k__cospi_m09_p23 = pair_set_epi16(-cospi_9_64, cospi_23_64);
- const __m128i k__cospi_m17_p15 = pair_set_epi16(-cospi_17_64, cospi_15_64);
- const __m128i k__cospi_m01_p31 = pair_set_epi16(-cospi_1_64, cospi_31_64);
- const __m128i k__cospi_p27_p05 = pair_set_epi16(+cospi_27_64, cospi_5_64);
- const __m128i k__cospi_p11_p21 = pair_set_epi16(+cospi_11_64, cospi_21_64);
- const __m128i k__cospi_p19_p13 = pair_set_epi16(+cospi_19_64, cospi_13_64);
- const __m128i k__cospi_p03_p29 = pair_set_epi16(+cospi_3_64, cospi_29_64);
- const __m128i k__cospi_m29_p03 = pair_set_epi16(-cospi_29_64, cospi_3_64);
- const __m128i k__cospi_m13_p19 = pair_set_epi16(-cospi_13_64, cospi_19_64);
- const __m128i k__cospi_m21_p11 = pair_set_epi16(-cospi_21_64, cospi_11_64);
- const __m128i k__cospi_m05_p27 = pair_set_epi16(-cospi_5_64, cospi_27_64);
+ const __m128i k__cospi_p30_p02 = pair_set_epi16(+cospi_30_64, cospi_2_64);
+ const __m128i k__cospi_p14_p18 = pair_set_epi16(+cospi_14_64, cospi_18_64);
+ const __m128i k__cospi_p22_p10 = pair_set_epi16(+cospi_22_64, cospi_10_64);
+ const __m128i k__cospi_p06_p26 = pair_set_epi16(+cospi_6_64, cospi_26_64);
+ const __m128i k__cospi_m26_p06 = pair_set_epi16(-cospi_26_64, cospi_6_64);
+ const __m128i k__cospi_m10_p22 = pair_set_epi16(-cospi_10_64, cospi_22_64);
+ const __m128i k__cospi_m18_p14 = pair_set_epi16(-cospi_18_64, cospi_14_64);
+ const __m128i k__cospi_m02_p30 = pair_set_epi16(-cospi_2_64, cospi_30_64);
+ const __m128i k__cospi_p31_p01 = pair_set_epi16(+cospi_31_64, cospi_1_64);
+ const __m128i k__cospi_p15_p17 = pair_set_epi16(+cospi_15_64, cospi_17_64);
+ const __m128i k__cospi_p23_p09 = pair_set_epi16(+cospi_23_64, cospi_9_64);
+ const __m128i k__cospi_p07_p25 = pair_set_epi16(+cospi_7_64, cospi_25_64);
+ const __m128i k__cospi_m25_p07 = pair_set_epi16(-cospi_25_64, cospi_7_64);
+ const __m128i k__cospi_m09_p23 = pair_set_epi16(-cospi_9_64, cospi_23_64);
+ const __m128i k__cospi_m17_p15 = pair_set_epi16(-cospi_17_64, cospi_15_64);
+ const __m128i k__cospi_m01_p31 = pair_set_epi16(-cospi_1_64, cospi_31_64);
+ const __m128i k__cospi_p27_p05 = pair_set_epi16(+cospi_27_64, cospi_5_64);
+ const __m128i k__cospi_p11_p21 = pair_set_epi16(+cospi_11_64, cospi_21_64);
+ const __m128i k__cospi_p19_p13 = pair_set_epi16(+cospi_19_64, cospi_13_64);
+ const __m128i k__cospi_p03_p29 = pair_set_epi16(+cospi_3_64, cospi_29_64);
+ const __m128i k__cospi_m29_p03 = pair_set_epi16(-cospi_29_64, cospi_3_64);
+ const __m128i k__cospi_m13_p19 = pair_set_epi16(-cospi_13_64, cospi_19_64);
+ const __m128i k__cospi_m21_p11 = pair_set_epi16(-cospi_21_64, cospi_11_64);
+ const __m128i k__cospi_m05_p27 = pair_set_epi16(-cospi_5_64, cospi_27_64);
const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i kZero = _mm_set1_epi16(0);
- const __m128i kOne = _mm_set1_epi16(1);
+ const __m128i kOne = _mm_set1_epi16(1);
// Do the two transform/transpose passes
int pass;
#if DCT_HIGH_BIT_DEPTH
@@ -124,125 +119,125 @@
// Note: even though all the loads below are aligned, using the aligned
// intrinsic make the code slightly slower.
if (0 == pass) {
- const int16_t *in = &input[column_start];
+ const int16_t *in = &input[column_start];
// step1[i] = (in[ 0 * stride] + in[(32 - 1) * stride]) << 2;
// Note: the next four blocks could be in a loop. That would help the
// instruction cache but is actually slower.
{
- const int16_t *ina = in + 0 * str1;
- const int16_t *inb = in + 31 * str1;
- __m128i *step1a = &step1[ 0];
+ const int16_t *ina = in + 0 * str1;
+ const int16_t *inb = in + 31 * str1;
+ __m128i *step1a = &step1[0];
__m128i *step1b = &step1[31];
- const __m128i ina0 = _mm_loadu_si128((const __m128i *)(ina));
- const __m128i ina1 = _mm_loadu_si128((const __m128i *)(ina + str1));
- const __m128i ina2 = _mm_loadu_si128((const __m128i *)(ina + str2));
- const __m128i ina3 = _mm_loadu_si128((const __m128i *)(ina + str3));
- const __m128i inb3 = _mm_loadu_si128((const __m128i *)(inb - str3));
- const __m128i inb2 = _mm_loadu_si128((const __m128i *)(inb - str2));
- const __m128i inb1 = _mm_loadu_si128((const __m128i *)(inb - str1));
- const __m128i inb0 = _mm_loadu_si128((const __m128i *)(inb));
- step1a[ 0] = _mm_add_epi16(ina0, inb0);
- step1a[ 1] = _mm_add_epi16(ina1, inb1);
- step1a[ 2] = _mm_add_epi16(ina2, inb2);
- step1a[ 3] = _mm_add_epi16(ina3, inb3);
+ const __m128i ina0 = _mm_loadu_si128((const __m128i *)(ina));
+ const __m128i ina1 = _mm_loadu_si128((const __m128i *)(ina + str1));
+ const __m128i ina2 = _mm_loadu_si128((const __m128i *)(ina + str2));
+ const __m128i ina3 = _mm_loadu_si128((const __m128i *)(ina + str3));
+ const __m128i inb3 = _mm_loadu_si128((const __m128i *)(inb - str3));
+ const __m128i inb2 = _mm_loadu_si128((const __m128i *)(inb - str2));
+ const __m128i inb1 = _mm_loadu_si128((const __m128i *)(inb - str1));
+ const __m128i inb0 = _mm_loadu_si128((const __m128i *)(inb));
+ step1a[0] = _mm_add_epi16(ina0, inb0);
+ step1a[1] = _mm_add_epi16(ina1, inb1);
+ step1a[2] = _mm_add_epi16(ina2, inb2);
+ step1a[3] = _mm_add_epi16(ina3, inb3);
step1b[-3] = _mm_sub_epi16(ina3, inb3);
step1b[-2] = _mm_sub_epi16(ina2, inb2);
step1b[-1] = _mm_sub_epi16(ina1, inb1);
step1b[-0] = _mm_sub_epi16(ina0, inb0);
- step1a[ 0] = _mm_slli_epi16(step1a[ 0], 2);
- step1a[ 1] = _mm_slli_epi16(step1a[ 1], 2);
- step1a[ 2] = _mm_slli_epi16(step1a[ 2], 2);
- step1a[ 3] = _mm_slli_epi16(step1a[ 3], 2);
+ step1a[0] = _mm_slli_epi16(step1a[0], 2);
+ step1a[1] = _mm_slli_epi16(step1a[1], 2);
+ step1a[2] = _mm_slli_epi16(step1a[2], 2);
+ step1a[3] = _mm_slli_epi16(step1a[3], 2);
step1b[-3] = _mm_slli_epi16(step1b[-3], 2);
step1b[-2] = _mm_slli_epi16(step1b[-2], 2);
step1b[-1] = _mm_slli_epi16(step1b[-1], 2);
step1b[-0] = _mm_slli_epi16(step1b[-0], 2);
}
{
- const int16_t *ina = in + 4 * str1;
- const int16_t *inb = in + 27 * str1;
- __m128i *step1a = &step1[ 4];
+ const int16_t *ina = in + 4 * str1;
+ const int16_t *inb = in + 27 * str1;
+ __m128i *step1a = &step1[4];
__m128i *step1b = &step1[27];
- const __m128i ina0 = _mm_loadu_si128((const __m128i *)(ina));
- const __m128i ina1 = _mm_loadu_si128((const __m128i *)(ina + str1));
- const __m128i ina2 = _mm_loadu_si128((const __m128i *)(ina + str2));
- const __m128i ina3 = _mm_loadu_si128((const __m128i *)(ina + str3));
- const __m128i inb3 = _mm_loadu_si128((const __m128i *)(inb - str3));
- const __m128i inb2 = _mm_loadu_si128((const __m128i *)(inb - str2));
- const __m128i inb1 = _mm_loadu_si128((const __m128i *)(inb - str1));
- const __m128i inb0 = _mm_loadu_si128((const __m128i *)(inb));
- step1a[ 0] = _mm_add_epi16(ina0, inb0);
- step1a[ 1] = _mm_add_epi16(ina1, inb1);
- step1a[ 2] = _mm_add_epi16(ina2, inb2);
- step1a[ 3] = _mm_add_epi16(ina3, inb3);
+ const __m128i ina0 = _mm_loadu_si128((const __m128i *)(ina));
+ const __m128i ina1 = _mm_loadu_si128((const __m128i *)(ina + str1));
+ const __m128i ina2 = _mm_loadu_si128((const __m128i *)(ina + str2));
+ const __m128i ina3 = _mm_loadu_si128((const __m128i *)(ina + str3));
+ const __m128i inb3 = _mm_loadu_si128((const __m128i *)(inb - str3));
+ const __m128i inb2 = _mm_loadu_si128((const __m128i *)(inb - str2));
+ const __m128i inb1 = _mm_loadu_si128((const __m128i *)(inb - str1));
+ const __m128i inb0 = _mm_loadu_si128((const __m128i *)(inb));
+ step1a[0] = _mm_add_epi16(ina0, inb0);
+ step1a[1] = _mm_add_epi16(ina1, inb1);
+ step1a[2] = _mm_add_epi16(ina2, inb2);
+ step1a[3] = _mm_add_epi16(ina3, inb3);
step1b[-3] = _mm_sub_epi16(ina3, inb3);
step1b[-2] = _mm_sub_epi16(ina2, inb2);
step1b[-1] = _mm_sub_epi16(ina1, inb1);
step1b[-0] = _mm_sub_epi16(ina0, inb0);
- step1a[ 0] = _mm_slli_epi16(step1a[ 0], 2);
- step1a[ 1] = _mm_slli_epi16(step1a[ 1], 2);
- step1a[ 2] = _mm_slli_epi16(step1a[ 2], 2);
- step1a[ 3] = _mm_slli_epi16(step1a[ 3], 2);
+ step1a[0] = _mm_slli_epi16(step1a[0], 2);
+ step1a[1] = _mm_slli_epi16(step1a[1], 2);
+ step1a[2] = _mm_slli_epi16(step1a[2], 2);
+ step1a[3] = _mm_slli_epi16(step1a[3], 2);
step1b[-3] = _mm_slli_epi16(step1b[-3], 2);
step1b[-2] = _mm_slli_epi16(step1b[-2], 2);
step1b[-1] = _mm_slli_epi16(step1b[-1], 2);
step1b[-0] = _mm_slli_epi16(step1b[-0], 2);
}
{
- const int16_t *ina = in + 8 * str1;
- const int16_t *inb = in + 23 * str1;
- __m128i *step1a = &step1[ 8];
+ const int16_t *ina = in + 8 * str1;
+ const int16_t *inb = in + 23 * str1;
+ __m128i *step1a = &step1[8];
__m128i *step1b = &step1[23];
- const __m128i ina0 = _mm_loadu_si128((const __m128i *)(ina));
- const __m128i ina1 = _mm_loadu_si128((const __m128i *)(ina + str1));
- const __m128i ina2 = _mm_loadu_si128((const __m128i *)(ina + str2));
- const __m128i ina3 = _mm_loadu_si128((const __m128i *)(ina + str3));
- const __m128i inb3 = _mm_loadu_si128((const __m128i *)(inb - str3));
- const __m128i inb2 = _mm_loadu_si128((const __m128i *)(inb - str2));
- const __m128i inb1 = _mm_loadu_si128((const __m128i *)(inb - str1));
- const __m128i inb0 = _mm_loadu_si128((const __m128i *)(inb));
- step1a[ 0] = _mm_add_epi16(ina0, inb0);
- step1a[ 1] = _mm_add_epi16(ina1, inb1);
- step1a[ 2] = _mm_add_epi16(ina2, inb2);
- step1a[ 3] = _mm_add_epi16(ina3, inb3);
+ const __m128i ina0 = _mm_loadu_si128((const __m128i *)(ina));
+ const __m128i ina1 = _mm_loadu_si128((const __m128i *)(ina + str1));
+ const __m128i ina2 = _mm_loadu_si128((const __m128i *)(ina + str2));
+ const __m128i ina3 = _mm_loadu_si128((const __m128i *)(ina + str3));
+ const __m128i inb3 = _mm_loadu_si128((const __m128i *)(inb - str3));
+ const __m128i inb2 = _mm_loadu_si128((const __m128i *)(inb - str2));
+ const __m128i inb1 = _mm_loadu_si128((const __m128i *)(inb - str1));
+ const __m128i inb0 = _mm_loadu_si128((const __m128i *)(inb));
+ step1a[0] = _mm_add_epi16(ina0, inb0);
+ step1a[1] = _mm_add_epi16(ina1, inb1);
+ step1a[2] = _mm_add_epi16(ina2, inb2);
+ step1a[3] = _mm_add_epi16(ina3, inb3);
step1b[-3] = _mm_sub_epi16(ina3, inb3);
step1b[-2] = _mm_sub_epi16(ina2, inb2);
step1b[-1] = _mm_sub_epi16(ina1, inb1);
step1b[-0] = _mm_sub_epi16(ina0, inb0);
- step1a[ 0] = _mm_slli_epi16(step1a[ 0], 2);
- step1a[ 1] = _mm_slli_epi16(step1a[ 1], 2);
- step1a[ 2] = _mm_slli_epi16(step1a[ 2], 2);
- step1a[ 3] = _mm_slli_epi16(step1a[ 3], 2);
+ step1a[0] = _mm_slli_epi16(step1a[0], 2);
+ step1a[1] = _mm_slli_epi16(step1a[1], 2);
+ step1a[2] = _mm_slli_epi16(step1a[2], 2);
+ step1a[3] = _mm_slli_epi16(step1a[3], 2);
step1b[-3] = _mm_slli_epi16(step1b[-3], 2);
step1b[-2] = _mm_slli_epi16(step1b[-2], 2);
step1b[-1] = _mm_slli_epi16(step1b[-1], 2);
step1b[-0] = _mm_slli_epi16(step1b[-0], 2);
}
{
- const int16_t *ina = in + 12 * str1;
- const int16_t *inb = in + 19 * str1;
+ const int16_t *ina = in + 12 * str1;
+ const int16_t *inb = in + 19 * str1;
__m128i *step1a = &step1[12];
__m128i *step1b = &step1[19];
- const __m128i ina0 = _mm_loadu_si128((const __m128i *)(ina));
- const __m128i ina1 = _mm_loadu_si128((const __m128i *)(ina + str1));
- const __m128i ina2 = _mm_loadu_si128((const __m128i *)(ina + str2));
- const __m128i ina3 = _mm_loadu_si128((const __m128i *)(ina + str3));
- const __m128i inb3 = _mm_loadu_si128((const __m128i *)(inb - str3));
- const __m128i inb2 = _mm_loadu_si128((const __m128i *)(inb - str2));
- const __m128i inb1 = _mm_loadu_si128((const __m128i *)(inb - str1));
- const __m128i inb0 = _mm_loadu_si128((const __m128i *)(inb));
- step1a[ 0] = _mm_add_epi16(ina0, inb0);
- step1a[ 1] = _mm_add_epi16(ina1, inb1);
- step1a[ 2] = _mm_add_epi16(ina2, inb2);
- step1a[ 3] = _mm_add_epi16(ina3, inb3);
+ const __m128i ina0 = _mm_loadu_si128((const __m128i *)(ina));
+ const __m128i ina1 = _mm_loadu_si128((const __m128i *)(ina + str1));
+ const __m128i ina2 = _mm_loadu_si128((const __m128i *)(ina + str2));
+ const __m128i ina3 = _mm_loadu_si128((const __m128i *)(ina + str3));
+ const __m128i inb3 = _mm_loadu_si128((const __m128i *)(inb - str3));
+ const __m128i inb2 = _mm_loadu_si128((const __m128i *)(inb - str2));
+ const __m128i inb1 = _mm_loadu_si128((const __m128i *)(inb - str1));
+ const __m128i inb0 = _mm_loadu_si128((const __m128i *)(inb));
+ step1a[0] = _mm_add_epi16(ina0, inb0);
+ step1a[1] = _mm_add_epi16(ina1, inb1);
+ step1a[2] = _mm_add_epi16(ina2, inb2);
+ step1a[3] = _mm_add_epi16(ina3, inb3);
step1b[-3] = _mm_sub_epi16(ina3, inb3);
step1b[-2] = _mm_sub_epi16(ina2, inb2);
step1b[-1] = _mm_sub_epi16(ina1, inb1);
step1b[-0] = _mm_sub_epi16(ina0, inb0);
- step1a[ 0] = _mm_slli_epi16(step1a[ 0], 2);
- step1a[ 1] = _mm_slli_epi16(step1a[ 1], 2);
- step1a[ 2] = _mm_slli_epi16(step1a[ 2], 2);
- step1a[ 3] = _mm_slli_epi16(step1a[ 3], 2);
+ step1a[0] = _mm_slli_epi16(step1a[0], 2);
+ step1a[1] = _mm_slli_epi16(step1a[1], 2);
+ step1a[2] = _mm_slli_epi16(step1a[2], 2);
+ step1a[3] = _mm_slli_epi16(step1a[3], 2);
step1b[-3] = _mm_slli_epi16(step1b[-3], 2);
step1b[-2] = _mm_slli_epi16(step1b[-2], 2);
step1b[-1] = _mm_slli_epi16(step1b[-1], 2);
@@ -257,14 +252,14 @@
// Note: the next four blocks could be in a loop. That would help the
// instruction cache but is actually slower.
{
- __m128i in00 = _mm_loadu_si128((const __m128i *)(in + 0 * 32));
- __m128i in01 = _mm_loadu_si128((const __m128i *)(in + 1 * 32));
- __m128i in02 = _mm_loadu_si128((const __m128i *)(in + 2 * 32));
- __m128i in03 = _mm_loadu_si128((const __m128i *)(in + 3 * 32));
- __m128i in28 = _mm_loadu_si128((const __m128i *)(in + 28 * 32));
- __m128i in29 = _mm_loadu_si128((const __m128i *)(in + 29 * 32));
- __m128i in30 = _mm_loadu_si128((const __m128i *)(in + 30 * 32));
- __m128i in31 = _mm_loadu_si128((const __m128i *)(in + 31 * 32));
+ __m128i in00 = _mm_loadu_si128((const __m128i *)(in + 0 * 32));
+ __m128i in01 = _mm_loadu_si128((const __m128i *)(in + 1 * 32));
+ __m128i in02 = _mm_loadu_si128((const __m128i *)(in + 2 * 32));
+ __m128i in03 = _mm_loadu_si128((const __m128i *)(in + 3 * 32));
+ __m128i in28 = _mm_loadu_si128((const __m128i *)(in + 28 * 32));
+ __m128i in29 = _mm_loadu_si128((const __m128i *)(in + 29 * 32));
+ __m128i in30 = _mm_loadu_si128((const __m128i *)(in + 30 * 32));
+ __m128i in31 = _mm_loadu_si128((const __m128i *)(in + 31 * 32));
step1[0] = ADD_EPI16(in00, in31);
step1[1] = ADD_EPI16(in01, in30);
step1[2] = ADD_EPI16(in02, in29);
@@ -284,14 +279,14 @@
#endif // DCT_HIGH_BIT_DEPTH
}
{
- __m128i in04 = _mm_loadu_si128((const __m128i *)(in + 4 * 32));
- __m128i in05 = _mm_loadu_si128((const __m128i *)(in + 5 * 32));
- __m128i in06 = _mm_loadu_si128((const __m128i *)(in + 6 * 32));
- __m128i in07 = _mm_loadu_si128((const __m128i *)(in + 7 * 32));
- __m128i in24 = _mm_loadu_si128((const __m128i *)(in + 24 * 32));
- __m128i in25 = _mm_loadu_si128((const __m128i *)(in + 25 * 32));
- __m128i in26 = _mm_loadu_si128((const __m128i *)(in + 26 * 32));
- __m128i in27 = _mm_loadu_si128((const __m128i *)(in + 27 * 32));
+ __m128i in04 = _mm_loadu_si128((const __m128i *)(in + 4 * 32));
+ __m128i in05 = _mm_loadu_si128((const __m128i *)(in + 5 * 32));
+ __m128i in06 = _mm_loadu_si128((const __m128i *)(in + 6 * 32));
+ __m128i in07 = _mm_loadu_si128((const __m128i *)(in + 7 * 32));
+ __m128i in24 = _mm_loadu_si128((const __m128i *)(in + 24 * 32));
+ __m128i in25 = _mm_loadu_si128((const __m128i *)(in + 25 * 32));
+ __m128i in26 = _mm_loadu_si128((const __m128i *)(in + 26 * 32));
+ __m128i in27 = _mm_loadu_si128((const __m128i *)(in + 27 * 32));
step1[4] = ADD_EPI16(in04, in27);
step1[5] = ADD_EPI16(in05, in26);
step1[6] = ADD_EPI16(in06, in25);
@@ -311,14 +306,14 @@
#endif // DCT_HIGH_BIT_DEPTH
}
{
- __m128i in08 = _mm_loadu_si128((const __m128i *)(in + 8 * 32));
- __m128i in09 = _mm_loadu_si128((const __m128i *)(in + 9 * 32));
- __m128i in10 = _mm_loadu_si128((const __m128i *)(in + 10 * 32));
- __m128i in11 = _mm_loadu_si128((const __m128i *)(in + 11 * 32));
- __m128i in20 = _mm_loadu_si128((const __m128i *)(in + 20 * 32));
- __m128i in21 = _mm_loadu_si128((const __m128i *)(in + 21 * 32));
- __m128i in22 = _mm_loadu_si128((const __m128i *)(in + 22 * 32));
- __m128i in23 = _mm_loadu_si128((const __m128i *)(in + 23 * 32));
+ __m128i in08 = _mm_loadu_si128((const __m128i *)(in + 8 * 32));
+ __m128i in09 = _mm_loadu_si128((const __m128i *)(in + 9 * 32));
+ __m128i in10 = _mm_loadu_si128((const __m128i *)(in + 10 * 32));
+ __m128i in11 = _mm_loadu_si128((const __m128i *)(in + 11 * 32));
+ __m128i in20 = _mm_loadu_si128((const __m128i *)(in + 20 * 32));
+ __m128i in21 = _mm_loadu_si128((const __m128i *)(in + 21 * 32));
+ __m128i in22 = _mm_loadu_si128((const __m128i *)(in + 22 * 32));
+ __m128i in23 = _mm_loadu_si128((const __m128i *)(in + 23 * 32));
step1[8] = ADD_EPI16(in08, in23);
step1[9] = ADD_EPI16(in09, in22);
step1[10] = ADD_EPI16(in10, in21);
@@ -338,14 +333,14 @@
#endif // DCT_HIGH_BIT_DEPTH
}
{
- __m128i in12 = _mm_loadu_si128((const __m128i *)(in + 12 * 32));
- __m128i in13 = _mm_loadu_si128((const __m128i *)(in + 13 * 32));
- __m128i in14 = _mm_loadu_si128((const __m128i *)(in + 14 * 32));
- __m128i in15 = _mm_loadu_si128((const __m128i *)(in + 15 * 32));
- __m128i in16 = _mm_loadu_si128((const __m128i *)(in + 16 * 32));
- __m128i in17 = _mm_loadu_si128((const __m128i *)(in + 17 * 32));
- __m128i in18 = _mm_loadu_si128((const __m128i *)(in + 18 * 32));
- __m128i in19 = _mm_loadu_si128((const __m128i *)(in + 19 * 32));
+ __m128i in12 = _mm_loadu_si128((const __m128i *)(in + 12 * 32));
+ __m128i in13 = _mm_loadu_si128((const __m128i *)(in + 13 * 32));
+ __m128i in14 = _mm_loadu_si128((const __m128i *)(in + 14 * 32));
+ __m128i in15 = _mm_loadu_si128((const __m128i *)(in + 15 * 32));
+ __m128i in16 = _mm_loadu_si128((const __m128i *)(in + 16 * 32));
+ __m128i in17 = _mm_loadu_si128((const __m128i *)(in + 17 * 32));
+ __m128i in18 = _mm_loadu_si128((const __m128i *)(in + 18 * 32));
+ __m128i in19 = _mm_loadu_si128((const __m128i *)(in + 19 * 32));
step1[12] = ADD_EPI16(in12, in19);
step1[13] = ADD_EPI16(in13, in18);
step1[14] = ADD_EPI16(in14, in17);
@@ -373,10 +368,10 @@
step2[3] = ADD_EPI16(step1[3], step1[12]);
step2[4] = ADD_EPI16(step1[4], step1[11]);
step2[5] = ADD_EPI16(step1[5], step1[10]);
- step2[6] = ADD_EPI16(step1[6], step1[ 9]);
- step2[7] = ADD_EPI16(step1[7], step1[ 8]);
- step2[8] = SUB_EPI16(step1[7], step1[ 8]);
- step2[9] = SUB_EPI16(step1[6], step1[ 9]);
+ step2[6] = ADD_EPI16(step1[6], step1[9]);
+ step2[7] = ADD_EPI16(step1[7], step1[8]);
+ step2[8] = SUB_EPI16(step1[7], step1[8]);
+ step2[9] = SUB_EPI16(step1[6], step1[9]);
step2[10] = SUB_EPI16(step1[5], step1[10]);
step2[11] = SUB_EPI16(step1[4], step1[11]);
step2[12] = SUB_EPI16(step1[3], step1[12]);
@@ -385,9 +380,8 @@
step2[15] = SUB_EPI16(step1[0], step1[15]);
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x16(
- &step2[0], &step2[1], &step2[2], &step2[3],
- &step2[4], &step2[5], &step2[6], &step2[7],
- &step2[8], &step2[9], &step2[10], &step2[11],
+ &step2[0], &step2[1], &step2[2], &step2[3], &step2[4], &step2[5],
+ &step2[6], &step2[7], &step2[8], &step2[9], &step2[10], &step2[11],
&step2[12], &step2[13], &step2[14], &step2[15]);
if (overflow) {
if (pass == 0)
@@ -483,16 +477,16 @@
// dump the magnitude by half, hence the intermediate values are within
// the range of 16 bits.
if (1 == pass) {
- __m128i s3_00_0 = _mm_cmplt_epi16(step2[ 0], kZero);
- __m128i s3_01_0 = _mm_cmplt_epi16(step2[ 1], kZero);
- __m128i s3_02_0 = _mm_cmplt_epi16(step2[ 2], kZero);
- __m128i s3_03_0 = _mm_cmplt_epi16(step2[ 3], kZero);
- __m128i s3_04_0 = _mm_cmplt_epi16(step2[ 4], kZero);
- __m128i s3_05_0 = _mm_cmplt_epi16(step2[ 5], kZero);
- __m128i s3_06_0 = _mm_cmplt_epi16(step2[ 6], kZero);
- __m128i s3_07_0 = _mm_cmplt_epi16(step2[ 7], kZero);
- __m128i s2_08_0 = _mm_cmplt_epi16(step2[ 8], kZero);
- __m128i s2_09_0 = _mm_cmplt_epi16(step2[ 9], kZero);
+ __m128i s3_00_0 = _mm_cmplt_epi16(step2[0], kZero);
+ __m128i s3_01_0 = _mm_cmplt_epi16(step2[1], kZero);
+ __m128i s3_02_0 = _mm_cmplt_epi16(step2[2], kZero);
+ __m128i s3_03_0 = _mm_cmplt_epi16(step2[3], kZero);
+ __m128i s3_04_0 = _mm_cmplt_epi16(step2[4], kZero);
+ __m128i s3_05_0 = _mm_cmplt_epi16(step2[5], kZero);
+ __m128i s3_06_0 = _mm_cmplt_epi16(step2[6], kZero);
+ __m128i s3_07_0 = _mm_cmplt_epi16(step2[7], kZero);
+ __m128i s2_08_0 = _mm_cmplt_epi16(step2[8], kZero);
+ __m128i s2_09_0 = _mm_cmplt_epi16(step2[9], kZero);
__m128i s3_10_0 = _mm_cmplt_epi16(step2[10], kZero);
__m128i s3_11_0 = _mm_cmplt_epi16(step2[11], kZero);
__m128i s3_12_0 = _mm_cmplt_epi16(step2[12], kZero);
@@ -516,16 +510,16 @@
__m128i s3_30_0 = _mm_cmplt_epi16(step1[30], kZero);
__m128i s3_31_0 = _mm_cmplt_epi16(step1[31], kZero);
- step2[0] = SUB_EPI16(step2[ 0], s3_00_0);
- step2[1] = SUB_EPI16(step2[ 1], s3_01_0);
- step2[2] = SUB_EPI16(step2[ 2], s3_02_0);
- step2[3] = SUB_EPI16(step2[ 3], s3_03_0);
- step2[4] = SUB_EPI16(step2[ 4], s3_04_0);
- step2[5] = SUB_EPI16(step2[ 5], s3_05_0);
- step2[6] = SUB_EPI16(step2[ 6], s3_06_0);
- step2[7] = SUB_EPI16(step2[ 7], s3_07_0);
- step2[8] = SUB_EPI16(step2[ 8], s2_08_0);
- step2[9] = SUB_EPI16(step2[ 9], s2_09_0);
+ step2[0] = SUB_EPI16(step2[0], s3_00_0);
+ step2[1] = SUB_EPI16(step2[1], s3_01_0);
+ step2[2] = SUB_EPI16(step2[2], s3_02_0);
+ step2[3] = SUB_EPI16(step2[3], s3_03_0);
+ step2[4] = SUB_EPI16(step2[4], s3_04_0);
+ step2[5] = SUB_EPI16(step2[5], s3_05_0);
+ step2[6] = SUB_EPI16(step2[6], s3_06_0);
+ step2[7] = SUB_EPI16(step2[7], s3_07_0);
+ step2[8] = SUB_EPI16(step2[8], s2_08_0);
+ step2[9] = SUB_EPI16(step2[9], s2_09_0);
step2[10] = SUB_EPI16(step2[10], s3_10_0);
step2[11] = SUB_EPI16(step2[11], s3_11_0);
step2[12] = SUB_EPI16(step2[12], s3_12_0);
@@ -550,29 +544,27 @@
step1[31] = SUB_EPI16(step1[31], s3_31_0);
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x32(
- &step2[0], &step2[1], &step2[2], &step2[3],
- &step2[4], &step2[5], &step2[6], &step2[7],
- &step2[8], &step2[9], &step2[10], &step2[11],
- &step2[12], &step2[13], &step2[14], &step2[15],
- &step1[16], &step1[17], &step1[18], &step1[19],
- &step2[20], &step2[21], &step2[22], &step2[23],
- &step2[24], &step2[25], &step2[26], &step2[27],
- &step1[28], &step1[29], &step1[30], &step1[31]);
+ &step2[0], &step2[1], &step2[2], &step2[3], &step2[4], &step2[5],
+ &step2[6], &step2[7], &step2[8], &step2[9], &step2[10], &step2[11],
+ &step2[12], &step2[13], &step2[14], &step2[15], &step1[16],
+ &step1[17], &step1[18], &step1[19], &step2[20], &step2[21],
+ &step2[22], &step2[23], &step2[24], &step2[25], &step2[26],
+ &step2[27], &step1[28], &step1[29], &step1[30], &step1[31]);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
- step2[0] = _mm_add_epi16(step2[ 0], kOne);
- step2[1] = _mm_add_epi16(step2[ 1], kOne);
- step2[2] = _mm_add_epi16(step2[ 2], kOne);
- step2[3] = _mm_add_epi16(step2[ 3], kOne);
- step2[4] = _mm_add_epi16(step2[ 4], kOne);
- step2[5] = _mm_add_epi16(step2[ 5], kOne);
- step2[6] = _mm_add_epi16(step2[ 6], kOne);
- step2[7] = _mm_add_epi16(step2[ 7], kOne);
- step2[8] = _mm_add_epi16(step2[ 8], kOne);
- step2[9] = _mm_add_epi16(step2[ 9], kOne);
+ step2[0] = _mm_add_epi16(step2[0], kOne);
+ step2[1] = _mm_add_epi16(step2[1], kOne);
+ step2[2] = _mm_add_epi16(step2[2], kOne);
+ step2[3] = _mm_add_epi16(step2[3], kOne);
+ step2[4] = _mm_add_epi16(step2[4], kOne);
+ step2[5] = _mm_add_epi16(step2[5], kOne);
+ step2[6] = _mm_add_epi16(step2[6], kOne);
+ step2[7] = _mm_add_epi16(step2[7], kOne);
+ step2[8] = _mm_add_epi16(step2[8], kOne);
+ step2[9] = _mm_add_epi16(step2[9], kOne);
step2[10] = _mm_add_epi16(step2[10], kOne);
step2[11] = _mm_add_epi16(step2[11], kOne);
step2[12] = _mm_add_epi16(step2[12], kOne);
@@ -596,16 +588,16 @@
step1[30] = _mm_add_epi16(step1[30], kOne);
step1[31] = _mm_add_epi16(step1[31], kOne);
- step2[0] = _mm_srai_epi16(step2[ 0], 2);
- step2[1] = _mm_srai_epi16(step2[ 1], 2);
- step2[2] = _mm_srai_epi16(step2[ 2], 2);
- step2[3] = _mm_srai_epi16(step2[ 3], 2);
- step2[4] = _mm_srai_epi16(step2[ 4], 2);
- step2[5] = _mm_srai_epi16(step2[ 5], 2);
- step2[6] = _mm_srai_epi16(step2[ 6], 2);
- step2[7] = _mm_srai_epi16(step2[ 7], 2);
- step2[8] = _mm_srai_epi16(step2[ 8], 2);
- step2[9] = _mm_srai_epi16(step2[ 9], 2);
+ step2[0] = _mm_srai_epi16(step2[0], 2);
+ step2[1] = _mm_srai_epi16(step2[1], 2);
+ step2[2] = _mm_srai_epi16(step2[2], 2);
+ step2[3] = _mm_srai_epi16(step2[3], 2);
+ step2[4] = _mm_srai_epi16(step2[4], 2);
+ step2[5] = _mm_srai_epi16(step2[5], 2);
+ step2[6] = _mm_srai_epi16(step2[6], 2);
+ step2[7] = _mm_srai_epi16(step2[7], 2);
+ step2[8] = _mm_srai_epi16(step2[8], 2);
+ step2[9] = _mm_srai_epi16(step2[9], 2);
step2[10] = _mm_srai_epi16(step2[10], 2);
step2[11] = _mm_srai_epi16(step2[11], 2);
step2[12] = _mm_srai_epi16(step2[12], 2);
@@ -634,821 +626,884 @@
#if FDCT32x32_HIGH_PRECISION
if (pass == 0) {
#endif
- // Stage 3
- {
- step3[0] = ADD_EPI16(step2[(8 - 1)], step2[0]);
- step3[1] = ADD_EPI16(step2[(8 - 2)], step2[1]);
- step3[2] = ADD_EPI16(step2[(8 - 3)], step2[2]);
- step3[3] = ADD_EPI16(step2[(8 - 4)], step2[3]);
- step3[4] = SUB_EPI16(step2[(8 - 5)], step2[4]);
- step3[5] = SUB_EPI16(step2[(8 - 6)], step2[5]);
- step3[6] = SUB_EPI16(step2[(8 - 7)], step2[6]);
- step3[7] = SUB_EPI16(step2[(8 - 8)], step2[7]);
+ // Stage 3
+ {
+ step3[0] = ADD_EPI16(step2[(8 - 1)], step2[0]);
+ step3[1] = ADD_EPI16(step2[(8 - 2)], step2[1]);
+ step3[2] = ADD_EPI16(step2[(8 - 3)], step2[2]);
+ step3[3] = ADD_EPI16(step2[(8 - 4)], step2[3]);
+ step3[4] = SUB_EPI16(step2[(8 - 5)], step2[4]);
+ step3[5] = SUB_EPI16(step2[(8 - 6)], step2[5]);
+ step3[6] = SUB_EPI16(step2[(8 - 7)], step2[6]);
+ step3[7] = SUB_EPI16(step2[(8 - 8)], step2[7]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&step3[0], &step3[1], &step3[2],
- &step3[3], &step3[4], &step3[5],
- &step3[6], &step3[7]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x8(&step3[0], &step3[1], &step3[2],
+ &step3[3], &step3[4], &step3[5],
+ &step3[6], &step3[7]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- const __m128i s3_10_0 = _mm_unpacklo_epi16(step2[13], step2[10]);
- const __m128i s3_10_1 = _mm_unpackhi_epi16(step2[13], step2[10]);
- const __m128i s3_11_0 = _mm_unpacklo_epi16(step2[12], step2[11]);
- const __m128i s3_11_1 = _mm_unpackhi_epi16(step2[12], step2[11]);
- const __m128i s3_10_2 = _mm_madd_epi16(s3_10_0, k__cospi_p16_m16);
- const __m128i s3_10_3 = _mm_madd_epi16(s3_10_1, k__cospi_p16_m16);
- const __m128i s3_11_2 = _mm_madd_epi16(s3_11_0, k__cospi_p16_m16);
- const __m128i s3_11_3 = _mm_madd_epi16(s3_11_1, k__cospi_p16_m16);
- const __m128i s3_12_2 = _mm_madd_epi16(s3_11_0, k__cospi_p16_p16);
- const __m128i s3_12_3 = _mm_madd_epi16(s3_11_1, k__cospi_p16_p16);
- const __m128i s3_13_2 = _mm_madd_epi16(s3_10_0, k__cospi_p16_p16);
- const __m128i s3_13_3 = _mm_madd_epi16(s3_10_1, k__cospi_p16_p16);
- // dct_const_round_shift
- const __m128i s3_10_4 = _mm_add_epi32(s3_10_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_10_5 = _mm_add_epi32(s3_10_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_11_4 = _mm_add_epi32(s3_11_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_11_5 = _mm_add_epi32(s3_11_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_12_4 = _mm_add_epi32(s3_12_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_12_5 = _mm_add_epi32(s3_12_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_13_4 = _mm_add_epi32(s3_13_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_13_5 = _mm_add_epi32(s3_13_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_10_6 = _mm_srai_epi32(s3_10_4, DCT_CONST_BITS);
- const __m128i s3_10_7 = _mm_srai_epi32(s3_10_5, DCT_CONST_BITS);
- const __m128i s3_11_6 = _mm_srai_epi32(s3_11_4, DCT_CONST_BITS);
- const __m128i s3_11_7 = _mm_srai_epi32(s3_11_5, DCT_CONST_BITS);
- const __m128i s3_12_6 = _mm_srai_epi32(s3_12_4, DCT_CONST_BITS);
- const __m128i s3_12_7 = _mm_srai_epi32(s3_12_5, DCT_CONST_BITS);
- const __m128i s3_13_6 = _mm_srai_epi32(s3_13_4, DCT_CONST_BITS);
- const __m128i s3_13_7 = _mm_srai_epi32(s3_13_5, DCT_CONST_BITS);
- // Combine
- step3[10] = _mm_packs_epi32(s3_10_6, s3_10_7);
- step3[11] = _mm_packs_epi32(s3_11_6, s3_11_7);
- step3[12] = _mm_packs_epi32(s3_12_6, s3_12_7);
- step3[13] = _mm_packs_epi32(s3_13_6, s3_13_7);
+ }
+ {
+ const __m128i s3_10_0 = _mm_unpacklo_epi16(step2[13], step2[10]);
+ const __m128i s3_10_1 = _mm_unpackhi_epi16(step2[13], step2[10]);
+ const __m128i s3_11_0 = _mm_unpacklo_epi16(step2[12], step2[11]);
+ const __m128i s3_11_1 = _mm_unpackhi_epi16(step2[12], step2[11]);
+ const __m128i s3_10_2 = _mm_madd_epi16(s3_10_0, k__cospi_p16_m16);
+ const __m128i s3_10_3 = _mm_madd_epi16(s3_10_1, k__cospi_p16_m16);
+ const __m128i s3_11_2 = _mm_madd_epi16(s3_11_0, k__cospi_p16_m16);
+ const __m128i s3_11_3 = _mm_madd_epi16(s3_11_1, k__cospi_p16_m16);
+ const __m128i s3_12_2 = _mm_madd_epi16(s3_11_0, k__cospi_p16_p16);
+ const __m128i s3_12_3 = _mm_madd_epi16(s3_11_1, k__cospi_p16_p16);
+ const __m128i s3_13_2 = _mm_madd_epi16(s3_10_0, k__cospi_p16_p16);
+ const __m128i s3_13_3 = _mm_madd_epi16(s3_10_1, k__cospi_p16_p16);
+ // dct_const_round_shift
+ const __m128i s3_10_4 = _mm_add_epi32(s3_10_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_10_5 = _mm_add_epi32(s3_10_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_11_4 = _mm_add_epi32(s3_11_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_11_5 = _mm_add_epi32(s3_11_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_12_4 = _mm_add_epi32(s3_12_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_12_5 = _mm_add_epi32(s3_12_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_13_4 = _mm_add_epi32(s3_13_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_13_5 = _mm_add_epi32(s3_13_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_10_6 = _mm_srai_epi32(s3_10_4, DCT_CONST_BITS);
+ const __m128i s3_10_7 = _mm_srai_epi32(s3_10_5, DCT_CONST_BITS);
+ const __m128i s3_11_6 = _mm_srai_epi32(s3_11_4, DCT_CONST_BITS);
+ const __m128i s3_11_7 = _mm_srai_epi32(s3_11_5, DCT_CONST_BITS);
+ const __m128i s3_12_6 = _mm_srai_epi32(s3_12_4, DCT_CONST_BITS);
+ const __m128i s3_12_7 = _mm_srai_epi32(s3_12_5, DCT_CONST_BITS);
+ const __m128i s3_13_6 = _mm_srai_epi32(s3_13_4, DCT_CONST_BITS);
+ const __m128i s3_13_7 = _mm_srai_epi32(s3_13_5, DCT_CONST_BITS);
+ // Combine
+ step3[10] = _mm_packs_epi32(s3_10_6, s3_10_7);
+ step3[11] = _mm_packs_epi32(s3_11_6, s3_11_7);
+ step3[12] = _mm_packs_epi32(s3_12_6, s3_12_7);
+ step3[13] = _mm_packs_epi32(s3_13_6, s3_13_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&step3[10], &step3[11],
- &step3[12], &step3[13]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x4(&step3[10], &step3[11], &step3[12],
+ &step3[13]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- step3[16] = ADD_EPI16(step2[23], step1[16]);
- step3[17] = ADD_EPI16(step2[22], step1[17]);
- step3[18] = ADD_EPI16(step2[21], step1[18]);
- step3[19] = ADD_EPI16(step2[20], step1[19]);
- step3[20] = SUB_EPI16(step1[19], step2[20]);
- step3[21] = SUB_EPI16(step1[18], step2[21]);
- step3[22] = SUB_EPI16(step1[17], step2[22]);
- step3[23] = SUB_EPI16(step1[16], step2[23]);
- step3[24] = SUB_EPI16(step1[31], step2[24]);
- step3[25] = SUB_EPI16(step1[30], step2[25]);
- step3[26] = SUB_EPI16(step1[29], step2[26]);
- step3[27] = SUB_EPI16(step1[28], step2[27]);
- step3[28] = ADD_EPI16(step2[27], step1[28]);
- step3[29] = ADD_EPI16(step2[26], step1[29]);
- step3[30] = ADD_EPI16(step2[25], step1[30]);
- step3[31] = ADD_EPI16(step2[24], step1[31]);
+ }
+ {
+ step3[16] = ADD_EPI16(step2[23], step1[16]);
+ step3[17] = ADD_EPI16(step2[22], step1[17]);
+ step3[18] = ADD_EPI16(step2[21], step1[18]);
+ step3[19] = ADD_EPI16(step2[20], step1[19]);
+ step3[20] = SUB_EPI16(step1[19], step2[20]);
+ step3[21] = SUB_EPI16(step1[18], step2[21]);
+ step3[22] = SUB_EPI16(step1[17], step2[22]);
+ step3[23] = SUB_EPI16(step1[16], step2[23]);
+ step3[24] = SUB_EPI16(step1[31], step2[24]);
+ step3[25] = SUB_EPI16(step1[30], step2[25]);
+ step3[26] = SUB_EPI16(step1[29], step2[26]);
+ step3[27] = SUB_EPI16(step1[28], step2[27]);
+ step3[28] = ADD_EPI16(step2[27], step1[28]);
+ step3[29] = ADD_EPI16(step2[26], step1[29]);
+ step3[30] = ADD_EPI16(step2[25], step1[30]);
+ step3[31] = ADD_EPI16(step2[24], step1[31]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x16(
- &step3[16], &step3[17], &step3[18], &step3[19],
- &step3[20], &step3[21], &step3[22], &step3[23],
- &step3[24], &step3[25], &step3[26], &step3[27],
- &step3[28], &step3[29], &step3[30], &step3[31]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x16(
+ &step3[16], &step3[17], &step3[18], &step3[19], &step3[20],
+ &step3[21], &step3[22], &step3[23], &step3[24], &step3[25],
+ &step3[26], &step3[27], &step3[28], &step3[29], &step3[30],
+ &step3[31]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
+ }
- // Stage 4
- {
- step1[0] = ADD_EPI16(step3[ 3], step3[ 0]);
- step1[1] = ADD_EPI16(step3[ 2], step3[ 1]);
- step1[2] = SUB_EPI16(step3[ 1], step3[ 2]);
- step1[3] = SUB_EPI16(step3[ 0], step3[ 3]);
- step1[8] = ADD_EPI16(step3[11], step2[ 8]);
- step1[9] = ADD_EPI16(step3[10], step2[ 9]);
- step1[10] = SUB_EPI16(step2[ 9], step3[10]);
- step1[11] = SUB_EPI16(step2[ 8], step3[11]);
- step1[12] = SUB_EPI16(step2[15], step3[12]);
- step1[13] = SUB_EPI16(step2[14], step3[13]);
- step1[14] = ADD_EPI16(step3[13], step2[14]);
- step1[15] = ADD_EPI16(step3[12], step2[15]);
+ // Stage 4
+ {
+ step1[0] = ADD_EPI16(step3[3], step3[0]);
+ step1[1] = ADD_EPI16(step3[2], step3[1]);
+ step1[2] = SUB_EPI16(step3[1], step3[2]);
+ step1[3] = SUB_EPI16(step3[0], step3[3]);
+ step1[8] = ADD_EPI16(step3[11], step2[8]);
+ step1[9] = ADD_EPI16(step3[10], step2[9]);
+ step1[10] = SUB_EPI16(step2[9], step3[10]);
+ step1[11] = SUB_EPI16(step2[8], step3[11]);
+ step1[12] = SUB_EPI16(step2[15], step3[12]);
+ step1[13] = SUB_EPI16(step2[14], step3[13]);
+ step1[14] = ADD_EPI16(step3[13], step2[14]);
+ step1[15] = ADD_EPI16(step3[12], step2[15]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x16(
- &step1[0], &step1[1], &step1[2], &step1[3],
- &step1[4], &step1[5], &step1[6], &step1[7],
- &step1[8], &step1[9], &step1[10], &step1[11],
- &step1[12], &step1[13], &step1[14], &step1[15]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x16(
+ &step1[0], &step1[1], &step1[2], &step1[3], &step1[4], &step1[5],
+ &step1[6], &step1[7], &step1[8], &step1[9], &step1[10],
+ &step1[11], &step1[12], &step1[13], &step1[14], &step1[15]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- const __m128i s1_05_0 = _mm_unpacklo_epi16(step3[6], step3[5]);
- const __m128i s1_05_1 = _mm_unpackhi_epi16(step3[6], step3[5]);
- const __m128i s1_05_2 = _mm_madd_epi16(s1_05_0, k__cospi_p16_m16);
- const __m128i s1_05_3 = _mm_madd_epi16(s1_05_1, k__cospi_p16_m16);
- const __m128i s1_06_2 = _mm_madd_epi16(s1_05_0, k__cospi_p16_p16);
- const __m128i s1_06_3 = _mm_madd_epi16(s1_05_1, k__cospi_p16_p16);
- // dct_const_round_shift
- const __m128i s1_05_4 = _mm_add_epi32(s1_05_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_05_5 = _mm_add_epi32(s1_05_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_06_4 = _mm_add_epi32(s1_06_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_06_5 = _mm_add_epi32(s1_06_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_05_6 = _mm_srai_epi32(s1_05_4, DCT_CONST_BITS);
- const __m128i s1_05_7 = _mm_srai_epi32(s1_05_5, DCT_CONST_BITS);
- const __m128i s1_06_6 = _mm_srai_epi32(s1_06_4, DCT_CONST_BITS);
- const __m128i s1_06_7 = _mm_srai_epi32(s1_06_5, DCT_CONST_BITS);
- // Combine
- step1[5] = _mm_packs_epi32(s1_05_6, s1_05_7);
- step1[6] = _mm_packs_epi32(s1_06_6, s1_06_7);
+ }
+ {
+ const __m128i s1_05_0 = _mm_unpacklo_epi16(step3[6], step3[5]);
+ const __m128i s1_05_1 = _mm_unpackhi_epi16(step3[6], step3[5]);
+ const __m128i s1_05_2 = _mm_madd_epi16(s1_05_0, k__cospi_p16_m16);
+ const __m128i s1_05_3 = _mm_madd_epi16(s1_05_1, k__cospi_p16_m16);
+ const __m128i s1_06_2 = _mm_madd_epi16(s1_05_0, k__cospi_p16_p16);
+ const __m128i s1_06_3 = _mm_madd_epi16(s1_05_1, k__cospi_p16_p16);
+ // dct_const_round_shift
+ const __m128i s1_05_4 = _mm_add_epi32(s1_05_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_05_5 = _mm_add_epi32(s1_05_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_06_4 = _mm_add_epi32(s1_06_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_06_5 = _mm_add_epi32(s1_06_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_05_6 = _mm_srai_epi32(s1_05_4, DCT_CONST_BITS);
+ const __m128i s1_05_7 = _mm_srai_epi32(s1_05_5, DCT_CONST_BITS);
+ const __m128i s1_06_6 = _mm_srai_epi32(s1_06_4, DCT_CONST_BITS);
+ const __m128i s1_06_7 = _mm_srai_epi32(s1_06_5, DCT_CONST_BITS);
+ // Combine
+ step1[5] = _mm_packs_epi32(s1_05_6, s1_05_7);
+ step1[6] = _mm_packs_epi32(s1_06_6, s1_06_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x2(&step1[5], &step1[6]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x2(&step1[5], &step1[6]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- const __m128i s1_18_0 = _mm_unpacklo_epi16(step3[18], step3[29]);
- const __m128i s1_18_1 = _mm_unpackhi_epi16(step3[18], step3[29]);
- const __m128i s1_19_0 = _mm_unpacklo_epi16(step3[19], step3[28]);
- const __m128i s1_19_1 = _mm_unpackhi_epi16(step3[19], step3[28]);
- const __m128i s1_20_0 = _mm_unpacklo_epi16(step3[20], step3[27]);
- const __m128i s1_20_1 = _mm_unpackhi_epi16(step3[20], step3[27]);
- const __m128i s1_21_0 = _mm_unpacklo_epi16(step3[21], step3[26]);
- const __m128i s1_21_1 = _mm_unpackhi_epi16(step3[21], step3[26]);
- const __m128i s1_18_2 = _mm_madd_epi16(s1_18_0, k__cospi_m08_p24);
- const __m128i s1_18_3 = _mm_madd_epi16(s1_18_1, k__cospi_m08_p24);
- const __m128i s1_19_2 = _mm_madd_epi16(s1_19_0, k__cospi_m08_p24);
- const __m128i s1_19_3 = _mm_madd_epi16(s1_19_1, k__cospi_m08_p24);
- const __m128i s1_20_2 = _mm_madd_epi16(s1_20_0, k__cospi_m24_m08);
- const __m128i s1_20_3 = _mm_madd_epi16(s1_20_1, k__cospi_m24_m08);
- const __m128i s1_21_2 = _mm_madd_epi16(s1_21_0, k__cospi_m24_m08);
- const __m128i s1_21_3 = _mm_madd_epi16(s1_21_1, k__cospi_m24_m08);
- const __m128i s1_26_2 = _mm_madd_epi16(s1_21_0, k__cospi_m08_p24);
- const __m128i s1_26_3 = _mm_madd_epi16(s1_21_1, k__cospi_m08_p24);
- const __m128i s1_27_2 = _mm_madd_epi16(s1_20_0, k__cospi_m08_p24);
- const __m128i s1_27_3 = _mm_madd_epi16(s1_20_1, k__cospi_m08_p24);
- const __m128i s1_28_2 = _mm_madd_epi16(s1_19_0, k__cospi_p24_p08);
- const __m128i s1_28_3 = _mm_madd_epi16(s1_19_1, k__cospi_p24_p08);
- const __m128i s1_29_2 = _mm_madd_epi16(s1_18_0, k__cospi_p24_p08);
- const __m128i s1_29_3 = _mm_madd_epi16(s1_18_1, k__cospi_p24_p08);
- // dct_const_round_shift
- const __m128i s1_18_4 = _mm_add_epi32(s1_18_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_18_5 = _mm_add_epi32(s1_18_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_19_4 = _mm_add_epi32(s1_19_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_19_5 = _mm_add_epi32(s1_19_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_20_4 = _mm_add_epi32(s1_20_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_20_5 = _mm_add_epi32(s1_20_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_21_4 = _mm_add_epi32(s1_21_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_21_5 = _mm_add_epi32(s1_21_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_26_4 = _mm_add_epi32(s1_26_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_26_5 = _mm_add_epi32(s1_26_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_27_4 = _mm_add_epi32(s1_27_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_27_5 = _mm_add_epi32(s1_27_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_28_4 = _mm_add_epi32(s1_28_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_28_5 = _mm_add_epi32(s1_28_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_29_4 = _mm_add_epi32(s1_29_2, k__DCT_CONST_ROUNDING);
- const __m128i s1_29_5 = _mm_add_epi32(s1_29_3, k__DCT_CONST_ROUNDING);
- const __m128i s1_18_6 = _mm_srai_epi32(s1_18_4, DCT_CONST_BITS);
- const __m128i s1_18_7 = _mm_srai_epi32(s1_18_5, DCT_CONST_BITS);
- const __m128i s1_19_6 = _mm_srai_epi32(s1_19_4, DCT_CONST_BITS);
- const __m128i s1_19_7 = _mm_srai_epi32(s1_19_5, DCT_CONST_BITS);
- const __m128i s1_20_6 = _mm_srai_epi32(s1_20_4, DCT_CONST_BITS);
- const __m128i s1_20_7 = _mm_srai_epi32(s1_20_5, DCT_CONST_BITS);
- const __m128i s1_21_6 = _mm_srai_epi32(s1_21_4, DCT_CONST_BITS);
- const __m128i s1_21_7 = _mm_srai_epi32(s1_21_5, DCT_CONST_BITS);
- const __m128i s1_26_6 = _mm_srai_epi32(s1_26_4, DCT_CONST_BITS);
- const __m128i s1_26_7 = _mm_srai_epi32(s1_26_5, DCT_CONST_BITS);
- const __m128i s1_27_6 = _mm_srai_epi32(s1_27_4, DCT_CONST_BITS);
- const __m128i s1_27_7 = _mm_srai_epi32(s1_27_5, DCT_CONST_BITS);
- const __m128i s1_28_6 = _mm_srai_epi32(s1_28_4, DCT_CONST_BITS);
- const __m128i s1_28_7 = _mm_srai_epi32(s1_28_5, DCT_CONST_BITS);
- const __m128i s1_29_6 = _mm_srai_epi32(s1_29_4, DCT_CONST_BITS);
- const __m128i s1_29_7 = _mm_srai_epi32(s1_29_5, DCT_CONST_BITS);
- // Combine
- step1[18] = _mm_packs_epi32(s1_18_6, s1_18_7);
- step1[19] = _mm_packs_epi32(s1_19_6, s1_19_7);
- step1[20] = _mm_packs_epi32(s1_20_6, s1_20_7);
- step1[21] = _mm_packs_epi32(s1_21_6, s1_21_7);
- step1[26] = _mm_packs_epi32(s1_26_6, s1_26_7);
- step1[27] = _mm_packs_epi32(s1_27_6, s1_27_7);
- step1[28] = _mm_packs_epi32(s1_28_6, s1_28_7);
- step1[29] = _mm_packs_epi32(s1_29_6, s1_29_7);
+ }
+ {
+ const __m128i s1_18_0 = _mm_unpacklo_epi16(step3[18], step3[29]);
+ const __m128i s1_18_1 = _mm_unpackhi_epi16(step3[18], step3[29]);
+ const __m128i s1_19_0 = _mm_unpacklo_epi16(step3[19], step3[28]);
+ const __m128i s1_19_1 = _mm_unpackhi_epi16(step3[19], step3[28]);
+ const __m128i s1_20_0 = _mm_unpacklo_epi16(step3[20], step3[27]);
+ const __m128i s1_20_1 = _mm_unpackhi_epi16(step3[20], step3[27]);
+ const __m128i s1_21_0 = _mm_unpacklo_epi16(step3[21], step3[26]);
+ const __m128i s1_21_1 = _mm_unpackhi_epi16(step3[21], step3[26]);
+ const __m128i s1_18_2 = _mm_madd_epi16(s1_18_0, k__cospi_m08_p24);
+ const __m128i s1_18_3 = _mm_madd_epi16(s1_18_1, k__cospi_m08_p24);
+ const __m128i s1_19_2 = _mm_madd_epi16(s1_19_0, k__cospi_m08_p24);
+ const __m128i s1_19_3 = _mm_madd_epi16(s1_19_1, k__cospi_m08_p24);
+ const __m128i s1_20_2 = _mm_madd_epi16(s1_20_0, k__cospi_m24_m08);
+ const __m128i s1_20_3 = _mm_madd_epi16(s1_20_1, k__cospi_m24_m08);
+ const __m128i s1_21_2 = _mm_madd_epi16(s1_21_0, k__cospi_m24_m08);
+ const __m128i s1_21_3 = _mm_madd_epi16(s1_21_1, k__cospi_m24_m08);
+ const __m128i s1_26_2 = _mm_madd_epi16(s1_21_0, k__cospi_m08_p24);
+ const __m128i s1_26_3 = _mm_madd_epi16(s1_21_1, k__cospi_m08_p24);
+ const __m128i s1_27_2 = _mm_madd_epi16(s1_20_0, k__cospi_m08_p24);
+ const __m128i s1_27_3 = _mm_madd_epi16(s1_20_1, k__cospi_m08_p24);
+ const __m128i s1_28_2 = _mm_madd_epi16(s1_19_0, k__cospi_p24_p08);
+ const __m128i s1_28_3 = _mm_madd_epi16(s1_19_1, k__cospi_p24_p08);
+ const __m128i s1_29_2 = _mm_madd_epi16(s1_18_0, k__cospi_p24_p08);
+ const __m128i s1_29_3 = _mm_madd_epi16(s1_18_1, k__cospi_p24_p08);
+ // dct_const_round_shift
+ const __m128i s1_18_4 = _mm_add_epi32(s1_18_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_18_5 = _mm_add_epi32(s1_18_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_19_4 = _mm_add_epi32(s1_19_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_19_5 = _mm_add_epi32(s1_19_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_20_4 = _mm_add_epi32(s1_20_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_20_5 = _mm_add_epi32(s1_20_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_21_4 = _mm_add_epi32(s1_21_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_21_5 = _mm_add_epi32(s1_21_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_26_4 = _mm_add_epi32(s1_26_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_26_5 = _mm_add_epi32(s1_26_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_27_4 = _mm_add_epi32(s1_27_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_27_5 = _mm_add_epi32(s1_27_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_28_4 = _mm_add_epi32(s1_28_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_28_5 = _mm_add_epi32(s1_28_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_29_4 = _mm_add_epi32(s1_29_2, k__DCT_CONST_ROUNDING);
+ const __m128i s1_29_5 = _mm_add_epi32(s1_29_3, k__DCT_CONST_ROUNDING);
+ const __m128i s1_18_6 = _mm_srai_epi32(s1_18_4, DCT_CONST_BITS);
+ const __m128i s1_18_7 = _mm_srai_epi32(s1_18_5, DCT_CONST_BITS);
+ const __m128i s1_19_6 = _mm_srai_epi32(s1_19_4, DCT_CONST_BITS);
+ const __m128i s1_19_7 = _mm_srai_epi32(s1_19_5, DCT_CONST_BITS);
+ const __m128i s1_20_6 = _mm_srai_epi32(s1_20_4, DCT_CONST_BITS);
+ const __m128i s1_20_7 = _mm_srai_epi32(s1_20_5, DCT_CONST_BITS);
+ const __m128i s1_21_6 = _mm_srai_epi32(s1_21_4, DCT_CONST_BITS);
+ const __m128i s1_21_7 = _mm_srai_epi32(s1_21_5, DCT_CONST_BITS);
+ const __m128i s1_26_6 = _mm_srai_epi32(s1_26_4, DCT_CONST_BITS);
+ const __m128i s1_26_7 = _mm_srai_epi32(s1_26_5, DCT_CONST_BITS);
+ const __m128i s1_27_6 = _mm_srai_epi32(s1_27_4, DCT_CONST_BITS);
+ const __m128i s1_27_7 = _mm_srai_epi32(s1_27_5, DCT_CONST_BITS);
+ const __m128i s1_28_6 = _mm_srai_epi32(s1_28_4, DCT_CONST_BITS);
+ const __m128i s1_28_7 = _mm_srai_epi32(s1_28_5, DCT_CONST_BITS);
+ const __m128i s1_29_6 = _mm_srai_epi32(s1_29_4, DCT_CONST_BITS);
+ const __m128i s1_29_7 = _mm_srai_epi32(s1_29_5, DCT_CONST_BITS);
+ // Combine
+ step1[18] = _mm_packs_epi32(s1_18_6, s1_18_7);
+ step1[19] = _mm_packs_epi32(s1_19_6, s1_19_7);
+ step1[20] = _mm_packs_epi32(s1_20_6, s1_20_7);
+ step1[21] = _mm_packs_epi32(s1_21_6, s1_21_7);
+ step1[26] = _mm_packs_epi32(s1_26_6, s1_26_7);
+ step1[27] = _mm_packs_epi32(s1_27_6, s1_27_7);
+ step1[28] = _mm_packs_epi32(s1_28_6, s1_28_7);
+ step1[29] = _mm_packs_epi32(s1_29_6, s1_29_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&step1[18], &step1[19], &step1[20],
- &step1[21], &step1[26], &step1[27],
- &step1[28], &step1[29]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x8(&step1[18], &step1[19], &step1[20],
+ &step1[21], &step1[26], &step1[27],
+ &step1[28], &step1[29]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- // Stage 5
- {
- step2[4] = ADD_EPI16(step1[5], step3[4]);
- step2[5] = SUB_EPI16(step3[4], step1[5]);
- step2[6] = SUB_EPI16(step3[7], step1[6]);
- step2[7] = ADD_EPI16(step1[6], step3[7]);
+ }
+ // Stage 5
+ {
+ step2[4] = ADD_EPI16(step1[5], step3[4]);
+ step2[5] = SUB_EPI16(step3[4], step1[5]);
+ step2[6] = SUB_EPI16(step3[7], step1[6]);
+ step2[7] = ADD_EPI16(step1[6], step3[7]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&step2[4], &step2[5],
- &step2[6], &step2[7]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x4(&step2[4], &step2[5], &step2[6],
+ &step2[7]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- const __m128i out_00_0 = _mm_unpacklo_epi16(step1[0], step1[1]);
- const __m128i out_00_1 = _mm_unpackhi_epi16(step1[0], step1[1]);
- const __m128i out_08_0 = _mm_unpacklo_epi16(step1[2], step1[3]);
- const __m128i out_08_1 = _mm_unpackhi_epi16(step1[2], step1[3]);
- const __m128i out_00_2 = _mm_madd_epi16(out_00_0, k__cospi_p16_p16);
- const __m128i out_00_3 = _mm_madd_epi16(out_00_1, k__cospi_p16_p16);
- const __m128i out_16_2 = _mm_madd_epi16(out_00_0, k__cospi_p16_m16);
- const __m128i out_16_3 = _mm_madd_epi16(out_00_1, k__cospi_p16_m16);
- const __m128i out_08_2 = _mm_madd_epi16(out_08_0, k__cospi_p24_p08);
- const __m128i out_08_3 = _mm_madd_epi16(out_08_1, k__cospi_p24_p08);
- const __m128i out_24_2 = _mm_madd_epi16(out_08_0, k__cospi_m08_p24);
- const __m128i out_24_3 = _mm_madd_epi16(out_08_1, k__cospi_m08_p24);
- // dct_const_round_shift
- const __m128i out_00_4 = _mm_add_epi32(out_00_2, k__DCT_CONST_ROUNDING);
- const __m128i out_00_5 = _mm_add_epi32(out_00_3, k__DCT_CONST_ROUNDING);
- const __m128i out_16_4 = _mm_add_epi32(out_16_2, k__DCT_CONST_ROUNDING);
- const __m128i out_16_5 = _mm_add_epi32(out_16_3, k__DCT_CONST_ROUNDING);
- const __m128i out_08_4 = _mm_add_epi32(out_08_2, k__DCT_CONST_ROUNDING);
- const __m128i out_08_5 = _mm_add_epi32(out_08_3, k__DCT_CONST_ROUNDING);
- const __m128i out_24_4 = _mm_add_epi32(out_24_2, k__DCT_CONST_ROUNDING);
- const __m128i out_24_5 = _mm_add_epi32(out_24_3, k__DCT_CONST_ROUNDING);
- const __m128i out_00_6 = _mm_srai_epi32(out_00_4, DCT_CONST_BITS);
- const __m128i out_00_7 = _mm_srai_epi32(out_00_5, DCT_CONST_BITS);
- const __m128i out_16_6 = _mm_srai_epi32(out_16_4, DCT_CONST_BITS);
- const __m128i out_16_7 = _mm_srai_epi32(out_16_5, DCT_CONST_BITS);
- const __m128i out_08_6 = _mm_srai_epi32(out_08_4, DCT_CONST_BITS);
- const __m128i out_08_7 = _mm_srai_epi32(out_08_5, DCT_CONST_BITS);
- const __m128i out_24_6 = _mm_srai_epi32(out_24_4, DCT_CONST_BITS);
- const __m128i out_24_7 = _mm_srai_epi32(out_24_5, DCT_CONST_BITS);
- // Combine
- out[ 0] = _mm_packs_epi32(out_00_6, out_00_7);
- out[16] = _mm_packs_epi32(out_16_6, out_16_7);
- out[ 8] = _mm_packs_epi32(out_08_6, out_08_7);
- out[24] = _mm_packs_epi32(out_24_6, out_24_7);
+ }
+ {
+ const __m128i out_00_0 = _mm_unpacklo_epi16(step1[0], step1[1]);
+ const __m128i out_00_1 = _mm_unpackhi_epi16(step1[0], step1[1]);
+ const __m128i out_08_0 = _mm_unpacklo_epi16(step1[2], step1[3]);
+ const __m128i out_08_1 = _mm_unpackhi_epi16(step1[2], step1[3]);
+ const __m128i out_00_2 = _mm_madd_epi16(out_00_0, k__cospi_p16_p16);
+ const __m128i out_00_3 = _mm_madd_epi16(out_00_1, k__cospi_p16_p16);
+ const __m128i out_16_2 = _mm_madd_epi16(out_00_0, k__cospi_p16_m16);
+ const __m128i out_16_3 = _mm_madd_epi16(out_00_1, k__cospi_p16_m16);
+ const __m128i out_08_2 = _mm_madd_epi16(out_08_0, k__cospi_p24_p08);
+ const __m128i out_08_3 = _mm_madd_epi16(out_08_1, k__cospi_p24_p08);
+ const __m128i out_24_2 = _mm_madd_epi16(out_08_0, k__cospi_m08_p24);
+ const __m128i out_24_3 = _mm_madd_epi16(out_08_1, k__cospi_m08_p24);
+ // dct_const_round_shift
+ const __m128i out_00_4 =
+ _mm_add_epi32(out_00_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_00_5 =
+ _mm_add_epi32(out_00_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_16_4 =
+ _mm_add_epi32(out_16_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_16_5 =
+ _mm_add_epi32(out_16_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_08_4 =
+ _mm_add_epi32(out_08_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_08_5 =
+ _mm_add_epi32(out_08_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_24_4 =
+ _mm_add_epi32(out_24_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_24_5 =
+ _mm_add_epi32(out_24_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_00_6 = _mm_srai_epi32(out_00_4, DCT_CONST_BITS);
+ const __m128i out_00_7 = _mm_srai_epi32(out_00_5, DCT_CONST_BITS);
+ const __m128i out_16_6 = _mm_srai_epi32(out_16_4, DCT_CONST_BITS);
+ const __m128i out_16_7 = _mm_srai_epi32(out_16_5, DCT_CONST_BITS);
+ const __m128i out_08_6 = _mm_srai_epi32(out_08_4, DCT_CONST_BITS);
+ const __m128i out_08_7 = _mm_srai_epi32(out_08_5, DCT_CONST_BITS);
+ const __m128i out_24_6 = _mm_srai_epi32(out_24_4, DCT_CONST_BITS);
+ const __m128i out_24_7 = _mm_srai_epi32(out_24_5, DCT_CONST_BITS);
+ // Combine
+ out[0] = _mm_packs_epi32(out_00_6, out_00_7);
+ out[16] = _mm_packs_epi32(out_16_6, out_16_7);
+ out[8] = _mm_packs_epi32(out_08_6, out_08_7);
+ out[24] = _mm_packs_epi32(out_24_6, out_24_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&out[0], &out[16],
- &out[8], &out[24]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow =
+ check_epi16_overflow_x4(&out[0], &out[16], &out[8], &out[24]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- const __m128i s2_09_0 = _mm_unpacklo_epi16(step1[ 9], step1[14]);
- const __m128i s2_09_1 = _mm_unpackhi_epi16(step1[ 9], step1[14]);
- const __m128i s2_10_0 = _mm_unpacklo_epi16(step1[10], step1[13]);
- const __m128i s2_10_1 = _mm_unpackhi_epi16(step1[10], step1[13]);
- const __m128i s2_09_2 = _mm_madd_epi16(s2_09_0, k__cospi_m08_p24);
- const __m128i s2_09_3 = _mm_madd_epi16(s2_09_1, k__cospi_m08_p24);
- const __m128i s2_10_2 = _mm_madd_epi16(s2_10_0, k__cospi_m24_m08);
- const __m128i s2_10_3 = _mm_madd_epi16(s2_10_1, k__cospi_m24_m08);
- const __m128i s2_13_2 = _mm_madd_epi16(s2_10_0, k__cospi_m08_p24);
- const __m128i s2_13_3 = _mm_madd_epi16(s2_10_1, k__cospi_m08_p24);
- const __m128i s2_14_2 = _mm_madd_epi16(s2_09_0, k__cospi_p24_p08);
- const __m128i s2_14_3 = _mm_madd_epi16(s2_09_1, k__cospi_p24_p08);
- // dct_const_round_shift
- const __m128i s2_09_4 = _mm_add_epi32(s2_09_2, k__DCT_CONST_ROUNDING);
- const __m128i s2_09_5 = _mm_add_epi32(s2_09_3, k__DCT_CONST_ROUNDING);
- const __m128i s2_10_4 = _mm_add_epi32(s2_10_2, k__DCT_CONST_ROUNDING);
- const __m128i s2_10_5 = _mm_add_epi32(s2_10_3, k__DCT_CONST_ROUNDING);
- const __m128i s2_13_4 = _mm_add_epi32(s2_13_2, k__DCT_CONST_ROUNDING);
- const __m128i s2_13_5 = _mm_add_epi32(s2_13_3, k__DCT_CONST_ROUNDING);
- const __m128i s2_14_4 = _mm_add_epi32(s2_14_2, k__DCT_CONST_ROUNDING);
- const __m128i s2_14_5 = _mm_add_epi32(s2_14_3, k__DCT_CONST_ROUNDING);
- const __m128i s2_09_6 = _mm_srai_epi32(s2_09_4, DCT_CONST_BITS);
- const __m128i s2_09_7 = _mm_srai_epi32(s2_09_5, DCT_CONST_BITS);
- const __m128i s2_10_6 = _mm_srai_epi32(s2_10_4, DCT_CONST_BITS);
- const __m128i s2_10_7 = _mm_srai_epi32(s2_10_5, DCT_CONST_BITS);
- const __m128i s2_13_6 = _mm_srai_epi32(s2_13_4, DCT_CONST_BITS);
- const __m128i s2_13_7 = _mm_srai_epi32(s2_13_5, DCT_CONST_BITS);
- const __m128i s2_14_6 = _mm_srai_epi32(s2_14_4, DCT_CONST_BITS);
- const __m128i s2_14_7 = _mm_srai_epi32(s2_14_5, DCT_CONST_BITS);
- // Combine
- step2[ 9] = _mm_packs_epi32(s2_09_6, s2_09_7);
- step2[10] = _mm_packs_epi32(s2_10_6, s2_10_7);
- step2[13] = _mm_packs_epi32(s2_13_6, s2_13_7);
- step2[14] = _mm_packs_epi32(s2_14_6, s2_14_7);
+ }
+ {
+ const __m128i s2_09_0 = _mm_unpacklo_epi16(step1[9], step1[14]);
+ const __m128i s2_09_1 = _mm_unpackhi_epi16(step1[9], step1[14]);
+ const __m128i s2_10_0 = _mm_unpacklo_epi16(step1[10], step1[13]);
+ const __m128i s2_10_1 = _mm_unpackhi_epi16(step1[10], step1[13]);
+ const __m128i s2_09_2 = _mm_madd_epi16(s2_09_0, k__cospi_m08_p24);
+ const __m128i s2_09_3 = _mm_madd_epi16(s2_09_1, k__cospi_m08_p24);
+ const __m128i s2_10_2 = _mm_madd_epi16(s2_10_0, k__cospi_m24_m08);
+ const __m128i s2_10_3 = _mm_madd_epi16(s2_10_1, k__cospi_m24_m08);
+ const __m128i s2_13_2 = _mm_madd_epi16(s2_10_0, k__cospi_m08_p24);
+ const __m128i s2_13_3 = _mm_madd_epi16(s2_10_1, k__cospi_m08_p24);
+ const __m128i s2_14_2 = _mm_madd_epi16(s2_09_0, k__cospi_p24_p08);
+ const __m128i s2_14_3 = _mm_madd_epi16(s2_09_1, k__cospi_p24_p08);
+ // dct_const_round_shift
+ const __m128i s2_09_4 = _mm_add_epi32(s2_09_2, k__DCT_CONST_ROUNDING);
+ const __m128i s2_09_5 = _mm_add_epi32(s2_09_3, k__DCT_CONST_ROUNDING);
+ const __m128i s2_10_4 = _mm_add_epi32(s2_10_2, k__DCT_CONST_ROUNDING);
+ const __m128i s2_10_5 = _mm_add_epi32(s2_10_3, k__DCT_CONST_ROUNDING);
+ const __m128i s2_13_4 = _mm_add_epi32(s2_13_2, k__DCT_CONST_ROUNDING);
+ const __m128i s2_13_5 = _mm_add_epi32(s2_13_3, k__DCT_CONST_ROUNDING);
+ const __m128i s2_14_4 = _mm_add_epi32(s2_14_2, k__DCT_CONST_ROUNDING);
+ const __m128i s2_14_5 = _mm_add_epi32(s2_14_3, k__DCT_CONST_ROUNDING);
+ const __m128i s2_09_6 = _mm_srai_epi32(s2_09_4, DCT_CONST_BITS);
+ const __m128i s2_09_7 = _mm_srai_epi32(s2_09_5, DCT_CONST_BITS);
+ const __m128i s2_10_6 = _mm_srai_epi32(s2_10_4, DCT_CONST_BITS);
+ const __m128i s2_10_7 = _mm_srai_epi32(s2_10_5, DCT_CONST_BITS);
+ const __m128i s2_13_6 = _mm_srai_epi32(s2_13_4, DCT_CONST_BITS);
+ const __m128i s2_13_7 = _mm_srai_epi32(s2_13_5, DCT_CONST_BITS);
+ const __m128i s2_14_6 = _mm_srai_epi32(s2_14_4, DCT_CONST_BITS);
+ const __m128i s2_14_7 = _mm_srai_epi32(s2_14_5, DCT_CONST_BITS);
+ // Combine
+ step2[9] = _mm_packs_epi32(s2_09_6, s2_09_7);
+ step2[10] = _mm_packs_epi32(s2_10_6, s2_10_7);
+ step2[13] = _mm_packs_epi32(s2_13_6, s2_13_7);
+ step2[14] = _mm_packs_epi32(s2_14_6, s2_14_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&step2[9], &step2[10],
- &step2[13], &step2[14]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x4(&step2[9], &step2[10], &step2[13],
+ &step2[14]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- step2[16] = ADD_EPI16(step1[19], step3[16]);
- step2[17] = ADD_EPI16(step1[18], step3[17]);
- step2[18] = SUB_EPI16(step3[17], step1[18]);
- step2[19] = SUB_EPI16(step3[16], step1[19]);
- step2[20] = SUB_EPI16(step3[23], step1[20]);
- step2[21] = SUB_EPI16(step3[22], step1[21]);
- step2[22] = ADD_EPI16(step1[21], step3[22]);
- step2[23] = ADD_EPI16(step1[20], step3[23]);
- step2[24] = ADD_EPI16(step1[27], step3[24]);
- step2[25] = ADD_EPI16(step1[26], step3[25]);
- step2[26] = SUB_EPI16(step3[25], step1[26]);
- step2[27] = SUB_EPI16(step3[24], step1[27]);
- step2[28] = SUB_EPI16(step3[31], step1[28]);
- step2[29] = SUB_EPI16(step3[30], step1[29]);
- step2[30] = ADD_EPI16(step1[29], step3[30]);
- step2[31] = ADD_EPI16(step1[28], step3[31]);
+ }
+ {
+ step2[16] = ADD_EPI16(step1[19], step3[16]);
+ step2[17] = ADD_EPI16(step1[18], step3[17]);
+ step2[18] = SUB_EPI16(step3[17], step1[18]);
+ step2[19] = SUB_EPI16(step3[16], step1[19]);
+ step2[20] = SUB_EPI16(step3[23], step1[20]);
+ step2[21] = SUB_EPI16(step3[22], step1[21]);
+ step2[22] = ADD_EPI16(step1[21], step3[22]);
+ step2[23] = ADD_EPI16(step1[20], step3[23]);
+ step2[24] = ADD_EPI16(step1[27], step3[24]);
+ step2[25] = ADD_EPI16(step1[26], step3[25]);
+ step2[26] = SUB_EPI16(step3[25], step1[26]);
+ step2[27] = SUB_EPI16(step3[24], step1[27]);
+ step2[28] = SUB_EPI16(step3[31], step1[28]);
+ step2[29] = SUB_EPI16(step3[30], step1[29]);
+ step2[30] = ADD_EPI16(step1[29], step3[30]);
+ step2[31] = ADD_EPI16(step1[28], step3[31]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x16(
- &step2[16], &step2[17], &step2[18], &step2[19],
- &step2[20], &step2[21], &step2[22], &step2[23],
- &step2[24], &step2[25], &step2[26], &step2[27],
- &step2[28], &step2[29], &step2[30], &step2[31]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x16(
+ &step2[16], &step2[17], &step2[18], &step2[19], &step2[20],
+ &step2[21], &step2[22], &step2[23], &step2[24], &step2[25],
+ &step2[26], &step2[27], &step2[28], &step2[29], &step2[30],
+ &step2[31]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- // Stage 6
- {
- const __m128i out_04_0 = _mm_unpacklo_epi16(step2[4], step2[7]);
- const __m128i out_04_1 = _mm_unpackhi_epi16(step2[4], step2[7]);
- const __m128i out_20_0 = _mm_unpacklo_epi16(step2[5], step2[6]);
- const __m128i out_20_1 = _mm_unpackhi_epi16(step2[5], step2[6]);
- const __m128i out_12_0 = _mm_unpacklo_epi16(step2[5], step2[6]);
- const __m128i out_12_1 = _mm_unpackhi_epi16(step2[5], step2[6]);
- const __m128i out_28_0 = _mm_unpacklo_epi16(step2[4], step2[7]);
- const __m128i out_28_1 = _mm_unpackhi_epi16(step2[4], step2[7]);
- const __m128i out_04_2 = _mm_madd_epi16(out_04_0, k__cospi_p28_p04);
- const __m128i out_04_3 = _mm_madd_epi16(out_04_1, k__cospi_p28_p04);
- const __m128i out_20_2 = _mm_madd_epi16(out_20_0, k__cospi_p12_p20);
- const __m128i out_20_3 = _mm_madd_epi16(out_20_1, k__cospi_p12_p20);
- const __m128i out_12_2 = _mm_madd_epi16(out_12_0, k__cospi_m20_p12);
- const __m128i out_12_3 = _mm_madd_epi16(out_12_1, k__cospi_m20_p12);
- const __m128i out_28_2 = _mm_madd_epi16(out_28_0, k__cospi_m04_p28);
- const __m128i out_28_3 = _mm_madd_epi16(out_28_1, k__cospi_m04_p28);
- // dct_const_round_shift
- const __m128i out_04_4 = _mm_add_epi32(out_04_2, k__DCT_CONST_ROUNDING);
- const __m128i out_04_5 = _mm_add_epi32(out_04_3, k__DCT_CONST_ROUNDING);
- const __m128i out_20_4 = _mm_add_epi32(out_20_2, k__DCT_CONST_ROUNDING);
- const __m128i out_20_5 = _mm_add_epi32(out_20_3, k__DCT_CONST_ROUNDING);
- const __m128i out_12_4 = _mm_add_epi32(out_12_2, k__DCT_CONST_ROUNDING);
- const __m128i out_12_5 = _mm_add_epi32(out_12_3, k__DCT_CONST_ROUNDING);
- const __m128i out_28_4 = _mm_add_epi32(out_28_2, k__DCT_CONST_ROUNDING);
- const __m128i out_28_5 = _mm_add_epi32(out_28_3, k__DCT_CONST_ROUNDING);
- const __m128i out_04_6 = _mm_srai_epi32(out_04_4, DCT_CONST_BITS);
- const __m128i out_04_7 = _mm_srai_epi32(out_04_5, DCT_CONST_BITS);
- const __m128i out_20_6 = _mm_srai_epi32(out_20_4, DCT_CONST_BITS);
- const __m128i out_20_7 = _mm_srai_epi32(out_20_5, DCT_CONST_BITS);
- const __m128i out_12_6 = _mm_srai_epi32(out_12_4, DCT_CONST_BITS);
- const __m128i out_12_7 = _mm_srai_epi32(out_12_5, DCT_CONST_BITS);
- const __m128i out_28_6 = _mm_srai_epi32(out_28_4, DCT_CONST_BITS);
- const __m128i out_28_7 = _mm_srai_epi32(out_28_5, DCT_CONST_BITS);
- // Combine
- out[4] = _mm_packs_epi32(out_04_6, out_04_7);
- out[20] = _mm_packs_epi32(out_20_6, out_20_7);
- out[12] = _mm_packs_epi32(out_12_6, out_12_7);
- out[28] = _mm_packs_epi32(out_28_6, out_28_7);
+ }
+ // Stage 6
+ {
+ const __m128i out_04_0 = _mm_unpacklo_epi16(step2[4], step2[7]);
+ const __m128i out_04_1 = _mm_unpackhi_epi16(step2[4], step2[7]);
+ const __m128i out_20_0 = _mm_unpacklo_epi16(step2[5], step2[6]);
+ const __m128i out_20_1 = _mm_unpackhi_epi16(step2[5], step2[6]);
+ const __m128i out_12_0 = _mm_unpacklo_epi16(step2[5], step2[6]);
+ const __m128i out_12_1 = _mm_unpackhi_epi16(step2[5], step2[6]);
+ const __m128i out_28_0 = _mm_unpacklo_epi16(step2[4], step2[7]);
+ const __m128i out_28_1 = _mm_unpackhi_epi16(step2[4], step2[7]);
+ const __m128i out_04_2 = _mm_madd_epi16(out_04_0, k__cospi_p28_p04);
+ const __m128i out_04_3 = _mm_madd_epi16(out_04_1, k__cospi_p28_p04);
+ const __m128i out_20_2 = _mm_madd_epi16(out_20_0, k__cospi_p12_p20);
+ const __m128i out_20_3 = _mm_madd_epi16(out_20_1, k__cospi_p12_p20);
+ const __m128i out_12_2 = _mm_madd_epi16(out_12_0, k__cospi_m20_p12);
+ const __m128i out_12_3 = _mm_madd_epi16(out_12_1, k__cospi_m20_p12);
+ const __m128i out_28_2 = _mm_madd_epi16(out_28_0, k__cospi_m04_p28);
+ const __m128i out_28_3 = _mm_madd_epi16(out_28_1, k__cospi_m04_p28);
+ // dct_const_round_shift
+ const __m128i out_04_4 =
+ _mm_add_epi32(out_04_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_04_5 =
+ _mm_add_epi32(out_04_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_20_4 =
+ _mm_add_epi32(out_20_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_20_5 =
+ _mm_add_epi32(out_20_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_12_4 =
+ _mm_add_epi32(out_12_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_12_5 =
+ _mm_add_epi32(out_12_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_28_4 =
+ _mm_add_epi32(out_28_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_28_5 =
+ _mm_add_epi32(out_28_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_04_6 = _mm_srai_epi32(out_04_4, DCT_CONST_BITS);
+ const __m128i out_04_7 = _mm_srai_epi32(out_04_5, DCT_CONST_BITS);
+ const __m128i out_20_6 = _mm_srai_epi32(out_20_4, DCT_CONST_BITS);
+ const __m128i out_20_7 = _mm_srai_epi32(out_20_5, DCT_CONST_BITS);
+ const __m128i out_12_6 = _mm_srai_epi32(out_12_4, DCT_CONST_BITS);
+ const __m128i out_12_7 = _mm_srai_epi32(out_12_5, DCT_CONST_BITS);
+ const __m128i out_28_6 = _mm_srai_epi32(out_28_4, DCT_CONST_BITS);
+ const __m128i out_28_7 = _mm_srai_epi32(out_28_5, DCT_CONST_BITS);
+ // Combine
+ out[4] = _mm_packs_epi32(out_04_6, out_04_7);
+ out[20] = _mm_packs_epi32(out_20_6, out_20_7);
+ out[12] = _mm_packs_epi32(out_12_6, out_12_7);
+ out[28] = _mm_packs_epi32(out_28_6, out_28_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&out[4], &out[20],
- &out[12], &out[28]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow =
+ check_epi16_overflow_x4(&out[4], &out[20], &out[12], &out[28]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- step3[8] = ADD_EPI16(step2[ 9], step1[ 8]);
- step3[9] = SUB_EPI16(step1[ 8], step2[ 9]);
- step3[10] = SUB_EPI16(step1[11], step2[10]);
- step3[11] = ADD_EPI16(step2[10], step1[11]);
- step3[12] = ADD_EPI16(step2[13], step1[12]);
- step3[13] = SUB_EPI16(step1[12], step2[13]);
- step3[14] = SUB_EPI16(step1[15], step2[14]);
- step3[15] = ADD_EPI16(step2[14], step1[15]);
+ }
+ {
+ step3[8] = ADD_EPI16(step2[9], step1[8]);
+ step3[9] = SUB_EPI16(step1[8], step2[9]);
+ step3[10] = SUB_EPI16(step1[11], step2[10]);
+ step3[11] = ADD_EPI16(step2[10], step1[11]);
+ step3[12] = ADD_EPI16(step2[13], step1[12]);
+ step3[13] = SUB_EPI16(step1[12], step2[13]);
+ step3[14] = SUB_EPI16(step1[15], step2[14]);
+ step3[15] = ADD_EPI16(step2[14], step1[15]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&step3[8], &step3[9], &step3[10],
- &step3[11], &step3[12], &step3[13],
- &step3[14], &step3[15]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x8(&step3[8], &step3[9], &step3[10],
+ &step3[11], &step3[12], &step3[13],
+ &step3[14], &step3[15]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- const __m128i s3_17_0 = _mm_unpacklo_epi16(step2[17], step2[30]);
- const __m128i s3_17_1 = _mm_unpackhi_epi16(step2[17], step2[30]);
- const __m128i s3_18_0 = _mm_unpacklo_epi16(step2[18], step2[29]);
- const __m128i s3_18_1 = _mm_unpackhi_epi16(step2[18], step2[29]);
- const __m128i s3_21_0 = _mm_unpacklo_epi16(step2[21], step2[26]);
- const __m128i s3_21_1 = _mm_unpackhi_epi16(step2[21], step2[26]);
- const __m128i s3_22_0 = _mm_unpacklo_epi16(step2[22], step2[25]);
- const __m128i s3_22_1 = _mm_unpackhi_epi16(step2[22], step2[25]);
- const __m128i s3_17_2 = _mm_madd_epi16(s3_17_0, k__cospi_m04_p28);
- const __m128i s3_17_3 = _mm_madd_epi16(s3_17_1, k__cospi_m04_p28);
- const __m128i s3_18_2 = _mm_madd_epi16(s3_18_0, k__cospi_m28_m04);
- const __m128i s3_18_3 = _mm_madd_epi16(s3_18_1, k__cospi_m28_m04);
- const __m128i s3_21_2 = _mm_madd_epi16(s3_21_0, k__cospi_m20_p12);
- const __m128i s3_21_3 = _mm_madd_epi16(s3_21_1, k__cospi_m20_p12);
- const __m128i s3_22_2 = _mm_madd_epi16(s3_22_0, k__cospi_m12_m20);
- const __m128i s3_22_3 = _mm_madd_epi16(s3_22_1, k__cospi_m12_m20);
- const __m128i s3_25_2 = _mm_madd_epi16(s3_22_0, k__cospi_m20_p12);
- const __m128i s3_25_3 = _mm_madd_epi16(s3_22_1, k__cospi_m20_p12);
- const __m128i s3_26_2 = _mm_madd_epi16(s3_21_0, k__cospi_p12_p20);
- const __m128i s3_26_3 = _mm_madd_epi16(s3_21_1, k__cospi_p12_p20);
- const __m128i s3_29_2 = _mm_madd_epi16(s3_18_0, k__cospi_m04_p28);
- const __m128i s3_29_3 = _mm_madd_epi16(s3_18_1, k__cospi_m04_p28);
- const __m128i s3_30_2 = _mm_madd_epi16(s3_17_0, k__cospi_p28_p04);
- const __m128i s3_30_3 = _mm_madd_epi16(s3_17_1, k__cospi_p28_p04);
- // dct_const_round_shift
- const __m128i s3_17_4 = _mm_add_epi32(s3_17_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_17_5 = _mm_add_epi32(s3_17_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_18_4 = _mm_add_epi32(s3_18_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_18_5 = _mm_add_epi32(s3_18_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_21_4 = _mm_add_epi32(s3_21_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_21_5 = _mm_add_epi32(s3_21_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_22_4 = _mm_add_epi32(s3_22_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_22_5 = _mm_add_epi32(s3_22_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_17_6 = _mm_srai_epi32(s3_17_4, DCT_CONST_BITS);
- const __m128i s3_17_7 = _mm_srai_epi32(s3_17_5, DCT_CONST_BITS);
- const __m128i s3_18_6 = _mm_srai_epi32(s3_18_4, DCT_CONST_BITS);
- const __m128i s3_18_7 = _mm_srai_epi32(s3_18_5, DCT_CONST_BITS);
- const __m128i s3_21_6 = _mm_srai_epi32(s3_21_4, DCT_CONST_BITS);
- const __m128i s3_21_7 = _mm_srai_epi32(s3_21_5, DCT_CONST_BITS);
- const __m128i s3_22_6 = _mm_srai_epi32(s3_22_4, DCT_CONST_BITS);
- const __m128i s3_22_7 = _mm_srai_epi32(s3_22_5, DCT_CONST_BITS);
- const __m128i s3_25_4 = _mm_add_epi32(s3_25_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_25_5 = _mm_add_epi32(s3_25_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_26_4 = _mm_add_epi32(s3_26_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_26_5 = _mm_add_epi32(s3_26_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_29_4 = _mm_add_epi32(s3_29_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_29_5 = _mm_add_epi32(s3_29_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_30_4 = _mm_add_epi32(s3_30_2, k__DCT_CONST_ROUNDING);
- const __m128i s3_30_5 = _mm_add_epi32(s3_30_3, k__DCT_CONST_ROUNDING);
- const __m128i s3_25_6 = _mm_srai_epi32(s3_25_4, DCT_CONST_BITS);
- const __m128i s3_25_7 = _mm_srai_epi32(s3_25_5, DCT_CONST_BITS);
- const __m128i s3_26_6 = _mm_srai_epi32(s3_26_4, DCT_CONST_BITS);
- const __m128i s3_26_7 = _mm_srai_epi32(s3_26_5, DCT_CONST_BITS);
- const __m128i s3_29_6 = _mm_srai_epi32(s3_29_4, DCT_CONST_BITS);
- const __m128i s3_29_7 = _mm_srai_epi32(s3_29_5, DCT_CONST_BITS);
- const __m128i s3_30_6 = _mm_srai_epi32(s3_30_4, DCT_CONST_BITS);
- const __m128i s3_30_7 = _mm_srai_epi32(s3_30_5, DCT_CONST_BITS);
- // Combine
- step3[17] = _mm_packs_epi32(s3_17_6, s3_17_7);
- step3[18] = _mm_packs_epi32(s3_18_6, s3_18_7);
- step3[21] = _mm_packs_epi32(s3_21_6, s3_21_7);
- step3[22] = _mm_packs_epi32(s3_22_6, s3_22_7);
- // Combine
- step3[25] = _mm_packs_epi32(s3_25_6, s3_25_7);
- step3[26] = _mm_packs_epi32(s3_26_6, s3_26_7);
- step3[29] = _mm_packs_epi32(s3_29_6, s3_29_7);
- step3[30] = _mm_packs_epi32(s3_30_6, s3_30_7);
+ }
+ {
+ const __m128i s3_17_0 = _mm_unpacklo_epi16(step2[17], step2[30]);
+ const __m128i s3_17_1 = _mm_unpackhi_epi16(step2[17], step2[30]);
+ const __m128i s3_18_0 = _mm_unpacklo_epi16(step2[18], step2[29]);
+ const __m128i s3_18_1 = _mm_unpackhi_epi16(step2[18], step2[29]);
+ const __m128i s3_21_0 = _mm_unpacklo_epi16(step2[21], step2[26]);
+ const __m128i s3_21_1 = _mm_unpackhi_epi16(step2[21], step2[26]);
+ const __m128i s3_22_0 = _mm_unpacklo_epi16(step2[22], step2[25]);
+ const __m128i s3_22_1 = _mm_unpackhi_epi16(step2[22], step2[25]);
+ const __m128i s3_17_2 = _mm_madd_epi16(s3_17_0, k__cospi_m04_p28);
+ const __m128i s3_17_3 = _mm_madd_epi16(s3_17_1, k__cospi_m04_p28);
+ const __m128i s3_18_2 = _mm_madd_epi16(s3_18_0, k__cospi_m28_m04);
+ const __m128i s3_18_3 = _mm_madd_epi16(s3_18_1, k__cospi_m28_m04);
+ const __m128i s3_21_2 = _mm_madd_epi16(s3_21_0, k__cospi_m20_p12);
+ const __m128i s3_21_3 = _mm_madd_epi16(s3_21_1, k__cospi_m20_p12);
+ const __m128i s3_22_2 = _mm_madd_epi16(s3_22_0, k__cospi_m12_m20);
+ const __m128i s3_22_3 = _mm_madd_epi16(s3_22_1, k__cospi_m12_m20);
+ const __m128i s3_25_2 = _mm_madd_epi16(s3_22_0, k__cospi_m20_p12);
+ const __m128i s3_25_3 = _mm_madd_epi16(s3_22_1, k__cospi_m20_p12);
+ const __m128i s3_26_2 = _mm_madd_epi16(s3_21_0, k__cospi_p12_p20);
+ const __m128i s3_26_3 = _mm_madd_epi16(s3_21_1, k__cospi_p12_p20);
+ const __m128i s3_29_2 = _mm_madd_epi16(s3_18_0, k__cospi_m04_p28);
+ const __m128i s3_29_3 = _mm_madd_epi16(s3_18_1, k__cospi_m04_p28);
+ const __m128i s3_30_2 = _mm_madd_epi16(s3_17_0, k__cospi_p28_p04);
+ const __m128i s3_30_3 = _mm_madd_epi16(s3_17_1, k__cospi_p28_p04);
+ // dct_const_round_shift
+ const __m128i s3_17_4 = _mm_add_epi32(s3_17_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_17_5 = _mm_add_epi32(s3_17_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_18_4 = _mm_add_epi32(s3_18_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_18_5 = _mm_add_epi32(s3_18_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_21_4 = _mm_add_epi32(s3_21_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_21_5 = _mm_add_epi32(s3_21_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_22_4 = _mm_add_epi32(s3_22_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_22_5 = _mm_add_epi32(s3_22_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_17_6 = _mm_srai_epi32(s3_17_4, DCT_CONST_BITS);
+ const __m128i s3_17_7 = _mm_srai_epi32(s3_17_5, DCT_CONST_BITS);
+ const __m128i s3_18_6 = _mm_srai_epi32(s3_18_4, DCT_CONST_BITS);
+ const __m128i s3_18_7 = _mm_srai_epi32(s3_18_5, DCT_CONST_BITS);
+ const __m128i s3_21_6 = _mm_srai_epi32(s3_21_4, DCT_CONST_BITS);
+ const __m128i s3_21_7 = _mm_srai_epi32(s3_21_5, DCT_CONST_BITS);
+ const __m128i s3_22_6 = _mm_srai_epi32(s3_22_4, DCT_CONST_BITS);
+ const __m128i s3_22_7 = _mm_srai_epi32(s3_22_5, DCT_CONST_BITS);
+ const __m128i s3_25_4 = _mm_add_epi32(s3_25_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_25_5 = _mm_add_epi32(s3_25_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_26_4 = _mm_add_epi32(s3_26_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_26_5 = _mm_add_epi32(s3_26_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_29_4 = _mm_add_epi32(s3_29_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_29_5 = _mm_add_epi32(s3_29_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_30_4 = _mm_add_epi32(s3_30_2, k__DCT_CONST_ROUNDING);
+ const __m128i s3_30_5 = _mm_add_epi32(s3_30_3, k__DCT_CONST_ROUNDING);
+ const __m128i s3_25_6 = _mm_srai_epi32(s3_25_4, DCT_CONST_BITS);
+ const __m128i s3_25_7 = _mm_srai_epi32(s3_25_5, DCT_CONST_BITS);
+ const __m128i s3_26_6 = _mm_srai_epi32(s3_26_4, DCT_CONST_BITS);
+ const __m128i s3_26_7 = _mm_srai_epi32(s3_26_5, DCT_CONST_BITS);
+ const __m128i s3_29_6 = _mm_srai_epi32(s3_29_4, DCT_CONST_BITS);
+ const __m128i s3_29_7 = _mm_srai_epi32(s3_29_5, DCT_CONST_BITS);
+ const __m128i s3_30_6 = _mm_srai_epi32(s3_30_4, DCT_CONST_BITS);
+ const __m128i s3_30_7 = _mm_srai_epi32(s3_30_5, DCT_CONST_BITS);
+ // Combine
+ step3[17] = _mm_packs_epi32(s3_17_6, s3_17_7);
+ step3[18] = _mm_packs_epi32(s3_18_6, s3_18_7);
+ step3[21] = _mm_packs_epi32(s3_21_6, s3_21_7);
+ step3[22] = _mm_packs_epi32(s3_22_6, s3_22_7);
+ // Combine
+ step3[25] = _mm_packs_epi32(s3_25_6, s3_25_7);
+ step3[26] = _mm_packs_epi32(s3_26_6, s3_26_7);
+ step3[29] = _mm_packs_epi32(s3_29_6, s3_29_7);
+ step3[30] = _mm_packs_epi32(s3_30_6, s3_30_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&step3[17], &step3[18], &step3[21],
- &step3[22], &step3[25], &step3[26],
- &step3[29], &step3[30]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x8(&step3[17], &step3[18], &step3[21],
+ &step3[22], &step3[25], &step3[26],
+ &step3[29], &step3[30]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- // Stage 7
- {
- const __m128i out_02_0 = _mm_unpacklo_epi16(step3[ 8], step3[15]);
- const __m128i out_02_1 = _mm_unpackhi_epi16(step3[ 8], step3[15]);
- const __m128i out_18_0 = _mm_unpacklo_epi16(step3[ 9], step3[14]);
- const __m128i out_18_1 = _mm_unpackhi_epi16(step3[ 9], step3[14]);
- const __m128i out_10_0 = _mm_unpacklo_epi16(step3[10], step3[13]);
- const __m128i out_10_1 = _mm_unpackhi_epi16(step3[10], step3[13]);
- const __m128i out_26_0 = _mm_unpacklo_epi16(step3[11], step3[12]);
- const __m128i out_26_1 = _mm_unpackhi_epi16(step3[11], step3[12]);
- const __m128i out_02_2 = _mm_madd_epi16(out_02_0, k__cospi_p30_p02);
- const __m128i out_02_3 = _mm_madd_epi16(out_02_1, k__cospi_p30_p02);
- const __m128i out_18_2 = _mm_madd_epi16(out_18_0, k__cospi_p14_p18);
- const __m128i out_18_3 = _mm_madd_epi16(out_18_1, k__cospi_p14_p18);
- const __m128i out_10_2 = _mm_madd_epi16(out_10_0, k__cospi_p22_p10);
- const __m128i out_10_3 = _mm_madd_epi16(out_10_1, k__cospi_p22_p10);
- const __m128i out_26_2 = _mm_madd_epi16(out_26_0, k__cospi_p06_p26);
- const __m128i out_26_3 = _mm_madd_epi16(out_26_1, k__cospi_p06_p26);
- const __m128i out_06_2 = _mm_madd_epi16(out_26_0, k__cospi_m26_p06);
- const __m128i out_06_3 = _mm_madd_epi16(out_26_1, k__cospi_m26_p06);
- const __m128i out_22_2 = _mm_madd_epi16(out_10_0, k__cospi_m10_p22);
- const __m128i out_22_3 = _mm_madd_epi16(out_10_1, k__cospi_m10_p22);
- const __m128i out_14_2 = _mm_madd_epi16(out_18_0, k__cospi_m18_p14);
- const __m128i out_14_3 = _mm_madd_epi16(out_18_1, k__cospi_m18_p14);
- const __m128i out_30_2 = _mm_madd_epi16(out_02_0, k__cospi_m02_p30);
- const __m128i out_30_3 = _mm_madd_epi16(out_02_1, k__cospi_m02_p30);
- // dct_const_round_shift
- const __m128i out_02_4 = _mm_add_epi32(out_02_2, k__DCT_CONST_ROUNDING);
- const __m128i out_02_5 = _mm_add_epi32(out_02_3, k__DCT_CONST_ROUNDING);
- const __m128i out_18_4 = _mm_add_epi32(out_18_2, k__DCT_CONST_ROUNDING);
- const __m128i out_18_5 = _mm_add_epi32(out_18_3, k__DCT_CONST_ROUNDING);
- const __m128i out_10_4 = _mm_add_epi32(out_10_2, k__DCT_CONST_ROUNDING);
- const __m128i out_10_5 = _mm_add_epi32(out_10_3, k__DCT_CONST_ROUNDING);
- const __m128i out_26_4 = _mm_add_epi32(out_26_2, k__DCT_CONST_ROUNDING);
- const __m128i out_26_5 = _mm_add_epi32(out_26_3, k__DCT_CONST_ROUNDING);
- const __m128i out_06_4 = _mm_add_epi32(out_06_2, k__DCT_CONST_ROUNDING);
- const __m128i out_06_5 = _mm_add_epi32(out_06_3, k__DCT_CONST_ROUNDING);
- const __m128i out_22_4 = _mm_add_epi32(out_22_2, k__DCT_CONST_ROUNDING);
- const __m128i out_22_5 = _mm_add_epi32(out_22_3, k__DCT_CONST_ROUNDING);
- const __m128i out_14_4 = _mm_add_epi32(out_14_2, k__DCT_CONST_ROUNDING);
- const __m128i out_14_5 = _mm_add_epi32(out_14_3, k__DCT_CONST_ROUNDING);
- const __m128i out_30_4 = _mm_add_epi32(out_30_2, k__DCT_CONST_ROUNDING);
- const __m128i out_30_5 = _mm_add_epi32(out_30_3, k__DCT_CONST_ROUNDING);
- const __m128i out_02_6 = _mm_srai_epi32(out_02_4, DCT_CONST_BITS);
- const __m128i out_02_7 = _mm_srai_epi32(out_02_5, DCT_CONST_BITS);
- const __m128i out_18_6 = _mm_srai_epi32(out_18_4, DCT_CONST_BITS);
- const __m128i out_18_7 = _mm_srai_epi32(out_18_5, DCT_CONST_BITS);
- const __m128i out_10_6 = _mm_srai_epi32(out_10_4, DCT_CONST_BITS);
- const __m128i out_10_7 = _mm_srai_epi32(out_10_5, DCT_CONST_BITS);
- const __m128i out_26_6 = _mm_srai_epi32(out_26_4, DCT_CONST_BITS);
- const __m128i out_26_7 = _mm_srai_epi32(out_26_5, DCT_CONST_BITS);
- const __m128i out_06_6 = _mm_srai_epi32(out_06_4, DCT_CONST_BITS);
- const __m128i out_06_7 = _mm_srai_epi32(out_06_5, DCT_CONST_BITS);
- const __m128i out_22_6 = _mm_srai_epi32(out_22_4, DCT_CONST_BITS);
- const __m128i out_22_7 = _mm_srai_epi32(out_22_5, DCT_CONST_BITS);
- const __m128i out_14_6 = _mm_srai_epi32(out_14_4, DCT_CONST_BITS);
- const __m128i out_14_7 = _mm_srai_epi32(out_14_5, DCT_CONST_BITS);
- const __m128i out_30_6 = _mm_srai_epi32(out_30_4, DCT_CONST_BITS);
- const __m128i out_30_7 = _mm_srai_epi32(out_30_5, DCT_CONST_BITS);
- // Combine
- out[ 2] = _mm_packs_epi32(out_02_6, out_02_7);
- out[18] = _mm_packs_epi32(out_18_6, out_18_7);
- out[10] = _mm_packs_epi32(out_10_6, out_10_7);
- out[26] = _mm_packs_epi32(out_26_6, out_26_7);
- out[ 6] = _mm_packs_epi32(out_06_6, out_06_7);
- out[22] = _mm_packs_epi32(out_22_6, out_22_7);
- out[14] = _mm_packs_epi32(out_14_6, out_14_7);
- out[30] = _mm_packs_epi32(out_30_6, out_30_7);
+ }
+ // Stage 7
+ {
+ const __m128i out_02_0 = _mm_unpacklo_epi16(step3[8], step3[15]);
+ const __m128i out_02_1 = _mm_unpackhi_epi16(step3[8], step3[15]);
+ const __m128i out_18_0 = _mm_unpacklo_epi16(step3[9], step3[14]);
+ const __m128i out_18_1 = _mm_unpackhi_epi16(step3[9], step3[14]);
+ const __m128i out_10_0 = _mm_unpacklo_epi16(step3[10], step3[13]);
+ const __m128i out_10_1 = _mm_unpackhi_epi16(step3[10], step3[13]);
+ const __m128i out_26_0 = _mm_unpacklo_epi16(step3[11], step3[12]);
+ const __m128i out_26_1 = _mm_unpackhi_epi16(step3[11], step3[12]);
+ const __m128i out_02_2 = _mm_madd_epi16(out_02_0, k__cospi_p30_p02);
+ const __m128i out_02_3 = _mm_madd_epi16(out_02_1, k__cospi_p30_p02);
+ const __m128i out_18_2 = _mm_madd_epi16(out_18_0, k__cospi_p14_p18);
+ const __m128i out_18_3 = _mm_madd_epi16(out_18_1, k__cospi_p14_p18);
+ const __m128i out_10_2 = _mm_madd_epi16(out_10_0, k__cospi_p22_p10);
+ const __m128i out_10_3 = _mm_madd_epi16(out_10_1, k__cospi_p22_p10);
+ const __m128i out_26_2 = _mm_madd_epi16(out_26_0, k__cospi_p06_p26);
+ const __m128i out_26_3 = _mm_madd_epi16(out_26_1, k__cospi_p06_p26);
+ const __m128i out_06_2 = _mm_madd_epi16(out_26_0, k__cospi_m26_p06);
+ const __m128i out_06_3 = _mm_madd_epi16(out_26_1, k__cospi_m26_p06);
+ const __m128i out_22_2 = _mm_madd_epi16(out_10_0, k__cospi_m10_p22);
+ const __m128i out_22_3 = _mm_madd_epi16(out_10_1, k__cospi_m10_p22);
+ const __m128i out_14_2 = _mm_madd_epi16(out_18_0, k__cospi_m18_p14);
+ const __m128i out_14_3 = _mm_madd_epi16(out_18_1, k__cospi_m18_p14);
+ const __m128i out_30_2 = _mm_madd_epi16(out_02_0, k__cospi_m02_p30);
+ const __m128i out_30_3 = _mm_madd_epi16(out_02_1, k__cospi_m02_p30);
+ // dct_const_round_shift
+ const __m128i out_02_4 =
+ _mm_add_epi32(out_02_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_02_5 =
+ _mm_add_epi32(out_02_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_18_4 =
+ _mm_add_epi32(out_18_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_18_5 =
+ _mm_add_epi32(out_18_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_10_4 =
+ _mm_add_epi32(out_10_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_10_5 =
+ _mm_add_epi32(out_10_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_26_4 =
+ _mm_add_epi32(out_26_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_26_5 =
+ _mm_add_epi32(out_26_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_06_4 =
+ _mm_add_epi32(out_06_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_06_5 =
+ _mm_add_epi32(out_06_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_22_4 =
+ _mm_add_epi32(out_22_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_22_5 =
+ _mm_add_epi32(out_22_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_14_4 =
+ _mm_add_epi32(out_14_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_14_5 =
+ _mm_add_epi32(out_14_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_30_4 =
+ _mm_add_epi32(out_30_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_30_5 =
+ _mm_add_epi32(out_30_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_02_6 = _mm_srai_epi32(out_02_4, DCT_CONST_BITS);
+ const __m128i out_02_7 = _mm_srai_epi32(out_02_5, DCT_CONST_BITS);
+ const __m128i out_18_6 = _mm_srai_epi32(out_18_4, DCT_CONST_BITS);
+ const __m128i out_18_7 = _mm_srai_epi32(out_18_5, DCT_CONST_BITS);
+ const __m128i out_10_6 = _mm_srai_epi32(out_10_4, DCT_CONST_BITS);
+ const __m128i out_10_7 = _mm_srai_epi32(out_10_5, DCT_CONST_BITS);
+ const __m128i out_26_6 = _mm_srai_epi32(out_26_4, DCT_CONST_BITS);
+ const __m128i out_26_7 = _mm_srai_epi32(out_26_5, DCT_CONST_BITS);
+ const __m128i out_06_6 = _mm_srai_epi32(out_06_4, DCT_CONST_BITS);
+ const __m128i out_06_7 = _mm_srai_epi32(out_06_5, DCT_CONST_BITS);
+ const __m128i out_22_6 = _mm_srai_epi32(out_22_4, DCT_CONST_BITS);
+ const __m128i out_22_7 = _mm_srai_epi32(out_22_5, DCT_CONST_BITS);
+ const __m128i out_14_6 = _mm_srai_epi32(out_14_4, DCT_CONST_BITS);
+ const __m128i out_14_7 = _mm_srai_epi32(out_14_5, DCT_CONST_BITS);
+ const __m128i out_30_6 = _mm_srai_epi32(out_30_4, DCT_CONST_BITS);
+ const __m128i out_30_7 = _mm_srai_epi32(out_30_5, DCT_CONST_BITS);
+ // Combine
+ out[2] = _mm_packs_epi32(out_02_6, out_02_7);
+ out[18] = _mm_packs_epi32(out_18_6, out_18_7);
+ out[10] = _mm_packs_epi32(out_10_6, out_10_7);
+ out[26] = _mm_packs_epi32(out_26_6, out_26_7);
+ out[6] = _mm_packs_epi32(out_06_6, out_06_7);
+ out[22] = _mm_packs_epi32(out_22_6, out_22_7);
+ out[14] = _mm_packs_epi32(out_14_6, out_14_7);
+ out[30] = _mm_packs_epi32(out_30_6, out_30_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&out[2], &out[18], &out[10],
- &out[26], &out[6], &out[22],
- &out[14], &out[30]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow =
+ check_epi16_overflow_x8(&out[2], &out[18], &out[10], &out[26],
+ &out[6], &out[22], &out[14], &out[30]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- step1[16] = ADD_EPI16(step3[17], step2[16]);
- step1[17] = SUB_EPI16(step2[16], step3[17]);
- step1[18] = SUB_EPI16(step2[19], step3[18]);
- step1[19] = ADD_EPI16(step3[18], step2[19]);
- step1[20] = ADD_EPI16(step3[21], step2[20]);
- step1[21] = SUB_EPI16(step2[20], step3[21]);
- step1[22] = SUB_EPI16(step2[23], step3[22]);
- step1[23] = ADD_EPI16(step3[22], step2[23]);
- step1[24] = ADD_EPI16(step3[25], step2[24]);
- step1[25] = SUB_EPI16(step2[24], step3[25]);
- step1[26] = SUB_EPI16(step2[27], step3[26]);
- step1[27] = ADD_EPI16(step3[26], step2[27]);
- step1[28] = ADD_EPI16(step3[29], step2[28]);
- step1[29] = SUB_EPI16(step2[28], step3[29]);
- step1[30] = SUB_EPI16(step2[31], step3[30]);
- step1[31] = ADD_EPI16(step3[30], step2[31]);
+ }
+ {
+ step1[16] = ADD_EPI16(step3[17], step2[16]);
+ step1[17] = SUB_EPI16(step2[16], step3[17]);
+ step1[18] = SUB_EPI16(step2[19], step3[18]);
+ step1[19] = ADD_EPI16(step3[18], step2[19]);
+ step1[20] = ADD_EPI16(step3[21], step2[20]);
+ step1[21] = SUB_EPI16(step2[20], step3[21]);
+ step1[22] = SUB_EPI16(step2[23], step3[22]);
+ step1[23] = ADD_EPI16(step3[22], step2[23]);
+ step1[24] = ADD_EPI16(step3[25], step2[24]);
+ step1[25] = SUB_EPI16(step2[24], step3[25]);
+ step1[26] = SUB_EPI16(step2[27], step3[26]);
+ step1[27] = ADD_EPI16(step3[26], step2[27]);
+ step1[28] = ADD_EPI16(step3[29], step2[28]);
+ step1[29] = SUB_EPI16(step2[28], step3[29]);
+ step1[30] = SUB_EPI16(step2[31], step3[30]);
+ step1[31] = ADD_EPI16(step3[30], step2[31]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x16(
- &step1[16], &step1[17], &step1[18], &step1[19],
- &step1[20], &step1[21], &step1[22], &step1[23],
- &step1[24], &step1[25], &step1[26], &step1[27],
- &step1[28], &step1[29], &step1[30], &step1[31]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = check_epi16_overflow_x16(
+ &step1[16], &step1[17], &step1[18], &step1[19], &step1[20],
+ &step1[21], &step1[22], &step1[23], &step1[24], &step1[25],
+ &step1[26], &step1[27], &step1[28], &step1[29], &step1[30],
+ &step1[31]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- // Final stage --- outputs indices are bit-reversed.
- {
- const __m128i out_01_0 = _mm_unpacklo_epi16(step1[16], step1[31]);
- const __m128i out_01_1 = _mm_unpackhi_epi16(step1[16], step1[31]);
- const __m128i out_17_0 = _mm_unpacklo_epi16(step1[17], step1[30]);
- const __m128i out_17_1 = _mm_unpackhi_epi16(step1[17], step1[30]);
- const __m128i out_09_0 = _mm_unpacklo_epi16(step1[18], step1[29]);
- const __m128i out_09_1 = _mm_unpackhi_epi16(step1[18], step1[29]);
- const __m128i out_25_0 = _mm_unpacklo_epi16(step1[19], step1[28]);
- const __m128i out_25_1 = _mm_unpackhi_epi16(step1[19], step1[28]);
- const __m128i out_01_2 = _mm_madd_epi16(out_01_0, k__cospi_p31_p01);
- const __m128i out_01_3 = _mm_madd_epi16(out_01_1, k__cospi_p31_p01);
- const __m128i out_17_2 = _mm_madd_epi16(out_17_0, k__cospi_p15_p17);
- const __m128i out_17_3 = _mm_madd_epi16(out_17_1, k__cospi_p15_p17);
- const __m128i out_09_2 = _mm_madd_epi16(out_09_0, k__cospi_p23_p09);
- const __m128i out_09_3 = _mm_madd_epi16(out_09_1, k__cospi_p23_p09);
- const __m128i out_25_2 = _mm_madd_epi16(out_25_0, k__cospi_p07_p25);
- const __m128i out_25_3 = _mm_madd_epi16(out_25_1, k__cospi_p07_p25);
- const __m128i out_07_2 = _mm_madd_epi16(out_25_0, k__cospi_m25_p07);
- const __m128i out_07_3 = _mm_madd_epi16(out_25_1, k__cospi_m25_p07);
- const __m128i out_23_2 = _mm_madd_epi16(out_09_0, k__cospi_m09_p23);
- const __m128i out_23_3 = _mm_madd_epi16(out_09_1, k__cospi_m09_p23);
- const __m128i out_15_2 = _mm_madd_epi16(out_17_0, k__cospi_m17_p15);
- const __m128i out_15_3 = _mm_madd_epi16(out_17_1, k__cospi_m17_p15);
- const __m128i out_31_2 = _mm_madd_epi16(out_01_0, k__cospi_m01_p31);
- const __m128i out_31_3 = _mm_madd_epi16(out_01_1, k__cospi_m01_p31);
- // dct_const_round_shift
- const __m128i out_01_4 = _mm_add_epi32(out_01_2, k__DCT_CONST_ROUNDING);
- const __m128i out_01_5 = _mm_add_epi32(out_01_3, k__DCT_CONST_ROUNDING);
- const __m128i out_17_4 = _mm_add_epi32(out_17_2, k__DCT_CONST_ROUNDING);
- const __m128i out_17_5 = _mm_add_epi32(out_17_3, k__DCT_CONST_ROUNDING);
- const __m128i out_09_4 = _mm_add_epi32(out_09_2, k__DCT_CONST_ROUNDING);
- const __m128i out_09_5 = _mm_add_epi32(out_09_3, k__DCT_CONST_ROUNDING);
- const __m128i out_25_4 = _mm_add_epi32(out_25_2, k__DCT_CONST_ROUNDING);
- const __m128i out_25_5 = _mm_add_epi32(out_25_3, k__DCT_CONST_ROUNDING);
- const __m128i out_07_4 = _mm_add_epi32(out_07_2, k__DCT_CONST_ROUNDING);
- const __m128i out_07_5 = _mm_add_epi32(out_07_3, k__DCT_CONST_ROUNDING);
- const __m128i out_23_4 = _mm_add_epi32(out_23_2, k__DCT_CONST_ROUNDING);
- const __m128i out_23_5 = _mm_add_epi32(out_23_3, k__DCT_CONST_ROUNDING);
- const __m128i out_15_4 = _mm_add_epi32(out_15_2, k__DCT_CONST_ROUNDING);
- const __m128i out_15_5 = _mm_add_epi32(out_15_3, k__DCT_CONST_ROUNDING);
- const __m128i out_31_4 = _mm_add_epi32(out_31_2, k__DCT_CONST_ROUNDING);
- const __m128i out_31_5 = _mm_add_epi32(out_31_3, k__DCT_CONST_ROUNDING);
- const __m128i out_01_6 = _mm_srai_epi32(out_01_4, DCT_CONST_BITS);
- const __m128i out_01_7 = _mm_srai_epi32(out_01_5, DCT_CONST_BITS);
- const __m128i out_17_6 = _mm_srai_epi32(out_17_4, DCT_CONST_BITS);
- const __m128i out_17_7 = _mm_srai_epi32(out_17_5, DCT_CONST_BITS);
- const __m128i out_09_6 = _mm_srai_epi32(out_09_4, DCT_CONST_BITS);
- const __m128i out_09_7 = _mm_srai_epi32(out_09_5, DCT_CONST_BITS);
- const __m128i out_25_6 = _mm_srai_epi32(out_25_4, DCT_CONST_BITS);
- const __m128i out_25_7 = _mm_srai_epi32(out_25_5, DCT_CONST_BITS);
- const __m128i out_07_6 = _mm_srai_epi32(out_07_4, DCT_CONST_BITS);
- const __m128i out_07_7 = _mm_srai_epi32(out_07_5, DCT_CONST_BITS);
- const __m128i out_23_6 = _mm_srai_epi32(out_23_4, DCT_CONST_BITS);
- const __m128i out_23_7 = _mm_srai_epi32(out_23_5, DCT_CONST_BITS);
- const __m128i out_15_6 = _mm_srai_epi32(out_15_4, DCT_CONST_BITS);
- const __m128i out_15_7 = _mm_srai_epi32(out_15_5, DCT_CONST_BITS);
- const __m128i out_31_6 = _mm_srai_epi32(out_31_4, DCT_CONST_BITS);
- const __m128i out_31_7 = _mm_srai_epi32(out_31_5, DCT_CONST_BITS);
- // Combine
- out[ 1] = _mm_packs_epi32(out_01_6, out_01_7);
- out[17] = _mm_packs_epi32(out_17_6, out_17_7);
- out[ 9] = _mm_packs_epi32(out_09_6, out_09_7);
- out[25] = _mm_packs_epi32(out_25_6, out_25_7);
- out[ 7] = _mm_packs_epi32(out_07_6, out_07_7);
- out[23] = _mm_packs_epi32(out_23_6, out_23_7);
- out[15] = _mm_packs_epi32(out_15_6, out_15_7);
- out[31] = _mm_packs_epi32(out_31_6, out_31_7);
+ }
+ // Final stage --- outputs indices are bit-reversed.
+ {
+ const __m128i out_01_0 = _mm_unpacklo_epi16(step1[16], step1[31]);
+ const __m128i out_01_1 = _mm_unpackhi_epi16(step1[16], step1[31]);
+ const __m128i out_17_0 = _mm_unpacklo_epi16(step1[17], step1[30]);
+ const __m128i out_17_1 = _mm_unpackhi_epi16(step1[17], step1[30]);
+ const __m128i out_09_0 = _mm_unpacklo_epi16(step1[18], step1[29]);
+ const __m128i out_09_1 = _mm_unpackhi_epi16(step1[18], step1[29]);
+ const __m128i out_25_0 = _mm_unpacklo_epi16(step1[19], step1[28]);
+ const __m128i out_25_1 = _mm_unpackhi_epi16(step1[19], step1[28]);
+ const __m128i out_01_2 = _mm_madd_epi16(out_01_0, k__cospi_p31_p01);
+ const __m128i out_01_3 = _mm_madd_epi16(out_01_1, k__cospi_p31_p01);
+ const __m128i out_17_2 = _mm_madd_epi16(out_17_0, k__cospi_p15_p17);
+ const __m128i out_17_3 = _mm_madd_epi16(out_17_1, k__cospi_p15_p17);
+ const __m128i out_09_2 = _mm_madd_epi16(out_09_0, k__cospi_p23_p09);
+ const __m128i out_09_3 = _mm_madd_epi16(out_09_1, k__cospi_p23_p09);
+ const __m128i out_25_2 = _mm_madd_epi16(out_25_0, k__cospi_p07_p25);
+ const __m128i out_25_3 = _mm_madd_epi16(out_25_1, k__cospi_p07_p25);
+ const __m128i out_07_2 = _mm_madd_epi16(out_25_0, k__cospi_m25_p07);
+ const __m128i out_07_3 = _mm_madd_epi16(out_25_1, k__cospi_m25_p07);
+ const __m128i out_23_2 = _mm_madd_epi16(out_09_0, k__cospi_m09_p23);
+ const __m128i out_23_3 = _mm_madd_epi16(out_09_1, k__cospi_m09_p23);
+ const __m128i out_15_2 = _mm_madd_epi16(out_17_0, k__cospi_m17_p15);
+ const __m128i out_15_3 = _mm_madd_epi16(out_17_1, k__cospi_m17_p15);
+ const __m128i out_31_2 = _mm_madd_epi16(out_01_0, k__cospi_m01_p31);
+ const __m128i out_31_3 = _mm_madd_epi16(out_01_1, k__cospi_m01_p31);
+ // dct_const_round_shift
+ const __m128i out_01_4 =
+ _mm_add_epi32(out_01_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_01_5 =
+ _mm_add_epi32(out_01_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_17_4 =
+ _mm_add_epi32(out_17_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_17_5 =
+ _mm_add_epi32(out_17_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_09_4 =
+ _mm_add_epi32(out_09_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_09_5 =
+ _mm_add_epi32(out_09_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_25_4 =
+ _mm_add_epi32(out_25_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_25_5 =
+ _mm_add_epi32(out_25_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_07_4 =
+ _mm_add_epi32(out_07_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_07_5 =
+ _mm_add_epi32(out_07_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_23_4 =
+ _mm_add_epi32(out_23_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_23_5 =
+ _mm_add_epi32(out_23_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_15_4 =
+ _mm_add_epi32(out_15_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_15_5 =
+ _mm_add_epi32(out_15_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_31_4 =
+ _mm_add_epi32(out_31_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_31_5 =
+ _mm_add_epi32(out_31_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_01_6 = _mm_srai_epi32(out_01_4, DCT_CONST_BITS);
+ const __m128i out_01_7 = _mm_srai_epi32(out_01_5, DCT_CONST_BITS);
+ const __m128i out_17_6 = _mm_srai_epi32(out_17_4, DCT_CONST_BITS);
+ const __m128i out_17_7 = _mm_srai_epi32(out_17_5, DCT_CONST_BITS);
+ const __m128i out_09_6 = _mm_srai_epi32(out_09_4, DCT_CONST_BITS);
+ const __m128i out_09_7 = _mm_srai_epi32(out_09_5, DCT_CONST_BITS);
+ const __m128i out_25_6 = _mm_srai_epi32(out_25_4, DCT_CONST_BITS);
+ const __m128i out_25_7 = _mm_srai_epi32(out_25_5, DCT_CONST_BITS);
+ const __m128i out_07_6 = _mm_srai_epi32(out_07_4, DCT_CONST_BITS);
+ const __m128i out_07_7 = _mm_srai_epi32(out_07_5, DCT_CONST_BITS);
+ const __m128i out_23_6 = _mm_srai_epi32(out_23_4, DCT_CONST_BITS);
+ const __m128i out_23_7 = _mm_srai_epi32(out_23_5, DCT_CONST_BITS);
+ const __m128i out_15_6 = _mm_srai_epi32(out_15_4, DCT_CONST_BITS);
+ const __m128i out_15_7 = _mm_srai_epi32(out_15_5, DCT_CONST_BITS);
+ const __m128i out_31_6 = _mm_srai_epi32(out_31_4, DCT_CONST_BITS);
+ const __m128i out_31_7 = _mm_srai_epi32(out_31_5, DCT_CONST_BITS);
+ // Combine
+ out[1] = _mm_packs_epi32(out_01_6, out_01_7);
+ out[17] = _mm_packs_epi32(out_17_6, out_17_7);
+ out[9] = _mm_packs_epi32(out_09_6, out_09_7);
+ out[25] = _mm_packs_epi32(out_25_6, out_25_7);
+ out[7] = _mm_packs_epi32(out_07_6, out_07_7);
+ out[23] = _mm_packs_epi32(out_23_6, out_23_7);
+ out[15] = _mm_packs_epi32(out_15_6, out_15_7);
+ out[31] = _mm_packs_epi32(out_31_6, out_31_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&out[1], &out[17], &out[9],
- &out[25], &out[7], &out[23],
- &out[15], &out[31]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow =
+ check_epi16_overflow_x8(&out[1], &out[17], &out[9], &out[25],
+ &out[7], &out[23], &out[15], &out[31]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
- {
- const __m128i out_05_0 = _mm_unpacklo_epi16(step1[20], step1[27]);
- const __m128i out_05_1 = _mm_unpackhi_epi16(step1[20], step1[27]);
- const __m128i out_21_0 = _mm_unpacklo_epi16(step1[21], step1[26]);
- const __m128i out_21_1 = _mm_unpackhi_epi16(step1[21], step1[26]);
- const __m128i out_13_0 = _mm_unpacklo_epi16(step1[22], step1[25]);
- const __m128i out_13_1 = _mm_unpackhi_epi16(step1[22], step1[25]);
- const __m128i out_29_0 = _mm_unpacklo_epi16(step1[23], step1[24]);
- const __m128i out_29_1 = _mm_unpackhi_epi16(step1[23], step1[24]);
- const __m128i out_05_2 = _mm_madd_epi16(out_05_0, k__cospi_p27_p05);
- const __m128i out_05_3 = _mm_madd_epi16(out_05_1, k__cospi_p27_p05);
- const __m128i out_21_2 = _mm_madd_epi16(out_21_0, k__cospi_p11_p21);
- const __m128i out_21_3 = _mm_madd_epi16(out_21_1, k__cospi_p11_p21);
- const __m128i out_13_2 = _mm_madd_epi16(out_13_0, k__cospi_p19_p13);
- const __m128i out_13_3 = _mm_madd_epi16(out_13_1, k__cospi_p19_p13);
- const __m128i out_29_2 = _mm_madd_epi16(out_29_0, k__cospi_p03_p29);
- const __m128i out_29_3 = _mm_madd_epi16(out_29_1, k__cospi_p03_p29);
- const __m128i out_03_2 = _mm_madd_epi16(out_29_0, k__cospi_m29_p03);
- const __m128i out_03_3 = _mm_madd_epi16(out_29_1, k__cospi_m29_p03);
- const __m128i out_19_2 = _mm_madd_epi16(out_13_0, k__cospi_m13_p19);
- const __m128i out_19_3 = _mm_madd_epi16(out_13_1, k__cospi_m13_p19);
- const __m128i out_11_2 = _mm_madd_epi16(out_21_0, k__cospi_m21_p11);
- const __m128i out_11_3 = _mm_madd_epi16(out_21_1, k__cospi_m21_p11);
- const __m128i out_27_2 = _mm_madd_epi16(out_05_0, k__cospi_m05_p27);
- const __m128i out_27_3 = _mm_madd_epi16(out_05_1, k__cospi_m05_p27);
- // dct_const_round_shift
- const __m128i out_05_4 = _mm_add_epi32(out_05_2, k__DCT_CONST_ROUNDING);
- const __m128i out_05_5 = _mm_add_epi32(out_05_3, k__DCT_CONST_ROUNDING);
- const __m128i out_21_4 = _mm_add_epi32(out_21_2, k__DCT_CONST_ROUNDING);
- const __m128i out_21_5 = _mm_add_epi32(out_21_3, k__DCT_CONST_ROUNDING);
- const __m128i out_13_4 = _mm_add_epi32(out_13_2, k__DCT_CONST_ROUNDING);
- const __m128i out_13_5 = _mm_add_epi32(out_13_3, k__DCT_CONST_ROUNDING);
- const __m128i out_29_4 = _mm_add_epi32(out_29_2, k__DCT_CONST_ROUNDING);
- const __m128i out_29_5 = _mm_add_epi32(out_29_3, k__DCT_CONST_ROUNDING);
- const __m128i out_03_4 = _mm_add_epi32(out_03_2, k__DCT_CONST_ROUNDING);
- const __m128i out_03_5 = _mm_add_epi32(out_03_3, k__DCT_CONST_ROUNDING);
- const __m128i out_19_4 = _mm_add_epi32(out_19_2, k__DCT_CONST_ROUNDING);
- const __m128i out_19_5 = _mm_add_epi32(out_19_3, k__DCT_CONST_ROUNDING);
- const __m128i out_11_4 = _mm_add_epi32(out_11_2, k__DCT_CONST_ROUNDING);
- const __m128i out_11_5 = _mm_add_epi32(out_11_3, k__DCT_CONST_ROUNDING);
- const __m128i out_27_4 = _mm_add_epi32(out_27_2, k__DCT_CONST_ROUNDING);
- const __m128i out_27_5 = _mm_add_epi32(out_27_3, k__DCT_CONST_ROUNDING);
- const __m128i out_05_6 = _mm_srai_epi32(out_05_4, DCT_CONST_BITS);
- const __m128i out_05_7 = _mm_srai_epi32(out_05_5, DCT_CONST_BITS);
- const __m128i out_21_6 = _mm_srai_epi32(out_21_4, DCT_CONST_BITS);
- const __m128i out_21_7 = _mm_srai_epi32(out_21_5, DCT_CONST_BITS);
- const __m128i out_13_6 = _mm_srai_epi32(out_13_4, DCT_CONST_BITS);
- const __m128i out_13_7 = _mm_srai_epi32(out_13_5, DCT_CONST_BITS);
- const __m128i out_29_6 = _mm_srai_epi32(out_29_4, DCT_CONST_BITS);
- const __m128i out_29_7 = _mm_srai_epi32(out_29_5, DCT_CONST_BITS);
- const __m128i out_03_6 = _mm_srai_epi32(out_03_4, DCT_CONST_BITS);
- const __m128i out_03_7 = _mm_srai_epi32(out_03_5, DCT_CONST_BITS);
- const __m128i out_19_6 = _mm_srai_epi32(out_19_4, DCT_CONST_BITS);
- const __m128i out_19_7 = _mm_srai_epi32(out_19_5, DCT_CONST_BITS);
- const __m128i out_11_6 = _mm_srai_epi32(out_11_4, DCT_CONST_BITS);
- const __m128i out_11_7 = _mm_srai_epi32(out_11_5, DCT_CONST_BITS);
- const __m128i out_27_6 = _mm_srai_epi32(out_27_4, DCT_CONST_BITS);
- const __m128i out_27_7 = _mm_srai_epi32(out_27_5, DCT_CONST_BITS);
- // Combine
- out[ 5] = _mm_packs_epi32(out_05_6, out_05_7);
- out[21] = _mm_packs_epi32(out_21_6, out_21_7);
- out[13] = _mm_packs_epi32(out_13_6, out_13_7);
- out[29] = _mm_packs_epi32(out_29_6, out_29_7);
- out[ 3] = _mm_packs_epi32(out_03_6, out_03_7);
- out[19] = _mm_packs_epi32(out_19_6, out_19_7);
- out[11] = _mm_packs_epi32(out_11_6, out_11_7);
- out[27] = _mm_packs_epi32(out_27_6, out_27_7);
+ }
+ {
+ const __m128i out_05_0 = _mm_unpacklo_epi16(step1[20], step1[27]);
+ const __m128i out_05_1 = _mm_unpackhi_epi16(step1[20], step1[27]);
+ const __m128i out_21_0 = _mm_unpacklo_epi16(step1[21], step1[26]);
+ const __m128i out_21_1 = _mm_unpackhi_epi16(step1[21], step1[26]);
+ const __m128i out_13_0 = _mm_unpacklo_epi16(step1[22], step1[25]);
+ const __m128i out_13_1 = _mm_unpackhi_epi16(step1[22], step1[25]);
+ const __m128i out_29_0 = _mm_unpacklo_epi16(step1[23], step1[24]);
+ const __m128i out_29_1 = _mm_unpackhi_epi16(step1[23], step1[24]);
+ const __m128i out_05_2 = _mm_madd_epi16(out_05_0, k__cospi_p27_p05);
+ const __m128i out_05_3 = _mm_madd_epi16(out_05_1, k__cospi_p27_p05);
+ const __m128i out_21_2 = _mm_madd_epi16(out_21_0, k__cospi_p11_p21);
+ const __m128i out_21_3 = _mm_madd_epi16(out_21_1, k__cospi_p11_p21);
+ const __m128i out_13_2 = _mm_madd_epi16(out_13_0, k__cospi_p19_p13);
+ const __m128i out_13_3 = _mm_madd_epi16(out_13_1, k__cospi_p19_p13);
+ const __m128i out_29_2 = _mm_madd_epi16(out_29_0, k__cospi_p03_p29);
+ const __m128i out_29_3 = _mm_madd_epi16(out_29_1, k__cospi_p03_p29);
+ const __m128i out_03_2 = _mm_madd_epi16(out_29_0, k__cospi_m29_p03);
+ const __m128i out_03_3 = _mm_madd_epi16(out_29_1, k__cospi_m29_p03);
+ const __m128i out_19_2 = _mm_madd_epi16(out_13_0, k__cospi_m13_p19);
+ const __m128i out_19_3 = _mm_madd_epi16(out_13_1, k__cospi_m13_p19);
+ const __m128i out_11_2 = _mm_madd_epi16(out_21_0, k__cospi_m21_p11);
+ const __m128i out_11_3 = _mm_madd_epi16(out_21_1, k__cospi_m21_p11);
+ const __m128i out_27_2 = _mm_madd_epi16(out_05_0, k__cospi_m05_p27);
+ const __m128i out_27_3 = _mm_madd_epi16(out_05_1, k__cospi_m05_p27);
+ // dct_const_round_shift
+ const __m128i out_05_4 =
+ _mm_add_epi32(out_05_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_05_5 =
+ _mm_add_epi32(out_05_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_21_4 =
+ _mm_add_epi32(out_21_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_21_5 =
+ _mm_add_epi32(out_21_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_13_4 =
+ _mm_add_epi32(out_13_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_13_5 =
+ _mm_add_epi32(out_13_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_29_4 =
+ _mm_add_epi32(out_29_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_29_5 =
+ _mm_add_epi32(out_29_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_03_4 =
+ _mm_add_epi32(out_03_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_03_5 =
+ _mm_add_epi32(out_03_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_19_4 =
+ _mm_add_epi32(out_19_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_19_5 =
+ _mm_add_epi32(out_19_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_11_4 =
+ _mm_add_epi32(out_11_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_11_5 =
+ _mm_add_epi32(out_11_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_27_4 =
+ _mm_add_epi32(out_27_2, k__DCT_CONST_ROUNDING);
+ const __m128i out_27_5 =
+ _mm_add_epi32(out_27_3, k__DCT_CONST_ROUNDING);
+ const __m128i out_05_6 = _mm_srai_epi32(out_05_4, DCT_CONST_BITS);
+ const __m128i out_05_7 = _mm_srai_epi32(out_05_5, DCT_CONST_BITS);
+ const __m128i out_21_6 = _mm_srai_epi32(out_21_4, DCT_CONST_BITS);
+ const __m128i out_21_7 = _mm_srai_epi32(out_21_5, DCT_CONST_BITS);
+ const __m128i out_13_6 = _mm_srai_epi32(out_13_4, DCT_CONST_BITS);
+ const __m128i out_13_7 = _mm_srai_epi32(out_13_5, DCT_CONST_BITS);
+ const __m128i out_29_6 = _mm_srai_epi32(out_29_4, DCT_CONST_BITS);
+ const __m128i out_29_7 = _mm_srai_epi32(out_29_5, DCT_CONST_BITS);
+ const __m128i out_03_6 = _mm_srai_epi32(out_03_4, DCT_CONST_BITS);
+ const __m128i out_03_7 = _mm_srai_epi32(out_03_5, DCT_CONST_BITS);
+ const __m128i out_19_6 = _mm_srai_epi32(out_19_4, DCT_CONST_BITS);
+ const __m128i out_19_7 = _mm_srai_epi32(out_19_5, DCT_CONST_BITS);
+ const __m128i out_11_6 = _mm_srai_epi32(out_11_4, DCT_CONST_BITS);
+ const __m128i out_11_7 = _mm_srai_epi32(out_11_5, DCT_CONST_BITS);
+ const __m128i out_27_6 = _mm_srai_epi32(out_27_4, DCT_CONST_BITS);
+ const __m128i out_27_7 = _mm_srai_epi32(out_27_5, DCT_CONST_BITS);
+ // Combine
+ out[5] = _mm_packs_epi32(out_05_6, out_05_7);
+ out[21] = _mm_packs_epi32(out_21_6, out_21_7);
+ out[13] = _mm_packs_epi32(out_13_6, out_13_7);
+ out[29] = _mm_packs_epi32(out_29_6, out_29_7);
+ out[3] = _mm_packs_epi32(out_03_6, out_03_7);
+ out[19] = _mm_packs_epi32(out_19_6, out_19_7);
+ out[11] = _mm_packs_epi32(out_11_6, out_11_7);
+ out[27] = _mm_packs_epi32(out_27_6, out_27_7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&out[5], &out[21], &out[13],
- &out[29], &out[3], &out[19],
- &out[11], &out[27]);
- if (overflow) {
- if (pass == 0)
- HIGH_FDCT32x32_2D_C(input, output_org, stride);
- else
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow =
+ check_epi16_overflow_x8(&out[5], &out[21], &out[13], &out[29],
+ &out[3], &out[19], &out[11], &out[27]);
+ if (overflow) {
+ if (pass == 0)
+ HIGH_FDCT32x32_2D_C(input, output_org, stride);
+ else
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- }
+ }
#if FDCT32x32_HIGH_PRECISION
} else {
__m128i lstep1[64], lstep2[64], lstep3[64];
@@ -1458,32 +1513,32 @@
// stage 3
{
// expanding to 32-bit length priori to addition operations
- lstep2[ 0] = _mm_unpacklo_epi16(step2[ 0], kZero);
- lstep2[ 1] = _mm_unpackhi_epi16(step2[ 0], kZero);
- lstep2[ 2] = _mm_unpacklo_epi16(step2[ 1], kZero);
- lstep2[ 3] = _mm_unpackhi_epi16(step2[ 1], kZero);
- lstep2[ 4] = _mm_unpacklo_epi16(step2[ 2], kZero);
- lstep2[ 5] = _mm_unpackhi_epi16(step2[ 2], kZero);
- lstep2[ 6] = _mm_unpacklo_epi16(step2[ 3], kZero);
- lstep2[ 7] = _mm_unpackhi_epi16(step2[ 3], kZero);
- lstep2[ 8] = _mm_unpacklo_epi16(step2[ 4], kZero);
- lstep2[ 9] = _mm_unpackhi_epi16(step2[ 4], kZero);
- lstep2[10] = _mm_unpacklo_epi16(step2[ 5], kZero);
- lstep2[11] = _mm_unpackhi_epi16(step2[ 5], kZero);
- lstep2[12] = _mm_unpacklo_epi16(step2[ 6], kZero);
- lstep2[13] = _mm_unpackhi_epi16(step2[ 6], kZero);
- lstep2[14] = _mm_unpacklo_epi16(step2[ 7], kZero);
- lstep2[15] = _mm_unpackhi_epi16(step2[ 7], kZero);
- lstep2[ 0] = _mm_madd_epi16(lstep2[ 0], kOne);
- lstep2[ 1] = _mm_madd_epi16(lstep2[ 1], kOne);
- lstep2[ 2] = _mm_madd_epi16(lstep2[ 2], kOne);
- lstep2[ 3] = _mm_madd_epi16(lstep2[ 3], kOne);
- lstep2[ 4] = _mm_madd_epi16(lstep2[ 4], kOne);
- lstep2[ 5] = _mm_madd_epi16(lstep2[ 5], kOne);
- lstep2[ 6] = _mm_madd_epi16(lstep2[ 6], kOne);
- lstep2[ 7] = _mm_madd_epi16(lstep2[ 7], kOne);
- lstep2[ 8] = _mm_madd_epi16(lstep2[ 8], kOne);
- lstep2[ 9] = _mm_madd_epi16(lstep2[ 9], kOne);
+ lstep2[0] = _mm_unpacklo_epi16(step2[0], kZero);
+ lstep2[1] = _mm_unpackhi_epi16(step2[0], kZero);
+ lstep2[2] = _mm_unpacklo_epi16(step2[1], kZero);
+ lstep2[3] = _mm_unpackhi_epi16(step2[1], kZero);
+ lstep2[4] = _mm_unpacklo_epi16(step2[2], kZero);
+ lstep2[5] = _mm_unpackhi_epi16(step2[2], kZero);
+ lstep2[6] = _mm_unpacklo_epi16(step2[3], kZero);
+ lstep2[7] = _mm_unpackhi_epi16(step2[3], kZero);
+ lstep2[8] = _mm_unpacklo_epi16(step2[4], kZero);
+ lstep2[9] = _mm_unpackhi_epi16(step2[4], kZero);
+ lstep2[10] = _mm_unpacklo_epi16(step2[5], kZero);
+ lstep2[11] = _mm_unpackhi_epi16(step2[5], kZero);
+ lstep2[12] = _mm_unpacklo_epi16(step2[6], kZero);
+ lstep2[13] = _mm_unpackhi_epi16(step2[6], kZero);
+ lstep2[14] = _mm_unpacklo_epi16(step2[7], kZero);
+ lstep2[15] = _mm_unpackhi_epi16(step2[7], kZero);
+ lstep2[0] = _mm_madd_epi16(lstep2[0], kOne);
+ lstep2[1] = _mm_madd_epi16(lstep2[1], kOne);
+ lstep2[2] = _mm_madd_epi16(lstep2[2], kOne);
+ lstep2[3] = _mm_madd_epi16(lstep2[3], kOne);
+ lstep2[4] = _mm_madd_epi16(lstep2[4], kOne);
+ lstep2[5] = _mm_madd_epi16(lstep2[5], kOne);
+ lstep2[6] = _mm_madd_epi16(lstep2[6], kOne);
+ lstep2[7] = _mm_madd_epi16(lstep2[7], kOne);
+ lstep2[8] = _mm_madd_epi16(lstep2[8], kOne);
+ lstep2[9] = _mm_madd_epi16(lstep2[9], kOne);
lstep2[10] = _mm_madd_epi16(lstep2[10], kOne);
lstep2[11] = _mm_madd_epi16(lstep2[11], kOne);
lstep2[12] = _mm_madd_epi16(lstep2[12], kOne);
@@ -1491,22 +1546,22 @@
lstep2[14] = _mm_madd_epi16(lstep2[14], kOne);
lstep2[15] = _mm_madd_epi16(lstep2[15], kOne);
- lstep3[ 0] = _mm_add_epi32(lstep2[14], lstep2[ 0]);
- lstep3[ 1] = _mm_add_epi32(lstep2[15], lstep2[ 1]);
- lstep3[ 2] = _mm_add_epi32(lstep2[12], lstep2[ 2]);
- lstep3[ 3] = _mm_add_epi32(lstep2[13], lstep2[ 3]);
- lstep3[ 4] = _mm_add_epi32(lstep2[10], lstep2[ 4]);
- lstep3[ 5] = _mm_add_epi32(lstep2[11], lstep2[ 5]);
- lstep3[ 6] = _mm_add_epi32(lstep2[ 8], lstep2[ 6]);
- lstep3[ 7] = _mm_add_epi32(lstep2[ 9], lstep2[ 7]);
- lstep3[ 8] = _mm_sub_epi32(lstep2[ 6], lstep2[ 8]);
- lstep3[ 9] = _mm_sub_epi32(lstep2[ 7], lstep2[ 9]);
- lstep3[10] = _mm_sub_epi32(lstep2[ 4], lstep2[10]);
- lstep3[11] = _mm_sub_epi32(lstep2[ 5], lstep2[11]);
- lstep3[12] = _mm_sub_epi32(lstep2[ 2], lstep2[12]);
- lstep3[13] = _mm_sub_epi32(lstep2[ 3], lstep2[13]);
- lstep3[14] = _mm_sub_epi32(lstep2[ 0], lstep2[14]);
- lstep3[15] = _mm_sub_epi32(lstep2[ 1], lstep2[15]);
+ lstep3[0] = _mm_add_epi32(lstep2[14], lstep2[0]);
+ lstep3[1] = _mm_add_epi32(lstep2[15], lstep2[1]);
+ lstep3[2] = _mm_add_epi32(lstep2[12], lstep2[2]);
+ lstep3[3] = _mm_add_epi32(lstep2[13], lstep2[3]);
+ lstep3[4] = _mm_add_epi32(lstep2[10], lstep2[4]);
+ lstep3[5] = _mm_add_epi32(lstep2[11], lstep2[5]);
+ lstep3[6] = _mm_add_epi32(lstep2[8], lstep2[6]);
+ lstep3[7] = _mm_add_epi32(lstep2[9], lstep2[7]);
+ lstep3[8] = _mm_sub_epi32(lstep2[6], lstep2[8]);
+ lstep3[9] = _mm_sub_epi32(lstep2[7], lstep2[9]);
+ lstep3[10] = _mm_sub_epi32(lstep2[4], lstep2[10]);
+ lstep3[11] = _mm_sub_epi32(lstep2[5], lstep2[11]);
+ lstep3[12] = _mm_sub_epi32(lstep2[2], lstep2[12]);
+ lstep3[13] = _mm_sub_epi32(lstep2[3], lstep2[13]);
+ lstep3[14] = _mm_sub_epi32(lstep2[0], lstep2[14]);
+ lstep3[15] = _mm_sub_epi32(lstep2[1], lstep2[15]);
}
{
const __m128i s3_10_0 = _mm_unpacklo_epi16(step2[13], step2[10]);
@@ -1644,10 +1699,10 @@
// stage 4
{
// expanding to 32-bit length priori to addition operations
- lstep2[16] = _mm_unpacklo_epi16(step2[ 8], kZero);
- lstep2[17] = _mm_unpackhi_epi16(step2[ 8], kZero);
- lstep2[18] = _mm_unpacklo_epi16(step2[ 9], kZero);
- lstep2[19] = _mm_unpackhi_epi16(step2[ 9], kZero);
+ lstep2[16] = _mm_unpacklo_epi16(step2[8], kZero);
+ lstep2[17] = _mm_unpackhi_epi16(step2[8], kZero);
+ lstep2[18] = _mm_unpacklo_epi16(step2[9], kZero);
+ lstep2[19] = _mm_unpackhi_epi16(step2[9], kZero);
lstep2[28] = _mm_unpacklo_epi16(step2[14], kZero);
lstep2[29] = _mm_unpackhi_epi16(step2[14], kZero);
lstep2[30] = _mm_unpacklo_epi16(step2[15], kZero);
@@ -1661,14 +1716,14 @@
lstep2[30] = _mm_madd_epi16(lstep2[30], kOne);
lstep2[31] = _mm_madd_epi16(lstep2[31], kOne);
- lstep1[ 0] = _mm_add_epi32(lstep3[ 6], lstep3[ 0]);
- lstep1[ 1] = _mm_add_epi32(lstep3[ 7], lstep3[ 1]);
- lstep1[ 2] = _mm_add_epi32(lstep3[ 4], lstep3[ 2]);
- lstep1[ 3] = _mm_add_epi32(lstep3[ 5], lstep3[ 3]);
- lstep1[ 4] = _mm_sub_epi32(lstep3[ 2], lstep3[ 4]);
- lstep1[ 5] = _mm_sub_epi32(lstep3[ 3], lstep3[ 5]);
- lstep1[ 6] = _mm_sub_epi32(lstep3[ 0], lstep3[ 6]);
- lstep1[ 7] = _mm_sub_epi32(lstep3[ 1], lstep3[ 7]);
+ lstep1[0] = _mm_add_epi32(lstep3[6], lstep3[0]);
+ lstep1[1] = _mm_add_epi32(lstep3[7], lstep3[1]);
+ lstep1[2] = _mm_add_epi32(lstep3[4], lstep3[2]);
+ lstep1[3] = _mm_add_epi32(lstep3[5], lstep3[3]);
+ lstep1[4] = _mm_sub_epi32(lstep3[2], lstep3[4]);
+ lstep1[5] = _mm_sub_epi32(lstep3[3], lstep3[5]);
+ lstep1[6] = _mm_sub_epi32(lstep3[0], lstep3[6]);
+ lstep1[7] = _mm_sub_epi32(lstep3[1], lstep3[7]);
lstep1[16] = _mm_add_epi32(lstep3[22], lstep2[16]);
lstep1[17] = _mm_add_epi32(lstep3[23], lstep2[17]);
lstep1[18] = _mm_add_epi32(lstep3[20], lstep2[18]);
@@ -1687,64 +1742,64 @@
lstep1[31] = _mm_add_epi32(lstep3[25], lstep2[31]);
}
{
- // to be continued...
- //
- const __m128i k32_p16_p16 = pair_set_epi32(cospi_16_64, cospi_16_64);
- const __m128i k32_p16_m16 = pair_set_epi32(cospi_16_64, -cospi_16_64);
+ // to be continued...
+ //
+ const __m128i k32_p16_p16 = pair_set_epi32(cospi_16_64, cospi_16_64);
+ const __m128i k32_p16_m16 = pair_set_epi32(cospi_16_64, -cospi_16_64);
- u[0] = _mm_unpacklo_epi32(lstep3[12], lstep3[10]);
- u[1] = _mm_unpackhi_epi32(lstep3[12], lstep3[10]);
- u[2] = _mm_unpacklo_epi32(lstep3[13], lstep3[11]);
- u[3] = _mm_unpackhi_epi32(lstep3[13], lstep3[11]);
+ u[0] = _mm_unpacklo_epi32(lstep3[12], lstep3[10]);
+ u[1] = _mm_unpackhi_epi32(lstep3[12], lstep3[10]);
+ u[2] = _mm_unpacklo_epi32(lstep3[13], lstep3[11]);
+ u[3] = _mm_unpackhi_epi32(lstep3[13], lstep3[11]);
- // TODO(jingning): manually inline k_madd_epi32_ to further hide
- // instruction latency.
- v[0] = k_madd_epi32(u[0], k32_p16_m16);
- v[1] = k_madd_epi32(u[1], k32_p16_m16);
- v[2] = k_madd_epi32(u[2], k32_p16_m16);
- v[3] = k_madd_epi32(u[3], k32_p16_m16);
- v[4] = k_madd_epi32(u[0], k32_p16_p16);
- v[5] = k_madd_epi32(u[1], k32_p16_p16);
- v[6] = k_madd_epi32(u[2], k32_p16_p16);
- v[7] = k_madd_epi32(u[3], k32_p16_p16);
+ // TODO(jingning): manually inline k_madd_epi32_ to further hide
+ // instruction latency.
+ v[0] = k_madd_epi32(u[0], k32_p16_m16);
+ v[1] = k_madd_epi32(u[1], k32_p16_m16);
+ v[2] = k_madd_epi32(u[2], k32_p16_m16);
+ v[3] = k_madd_epi32(u[3], k32_p16_m16);
+ v[4] = k_madd_epi32(u[0], k32_p16_p16);
+ v[5] = k_madd_epi32(u[1], k32_p16_p16);
+ v[6] = k_madd_epi32(u[2], k32_p16_p16);
+ v[7] = k_madd_epi32(u[3], k32_p16_p16);
#if DCT_HIGH_BIT_DEPTH
- overflow = k_check_epi32_overflow_8(&v[0], &v[1], &v[2], &v[3],
- &v[4], &v[5], &v[6], &v[7], &kZero);
- if (overflow) {
- HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
- return;
- }
+ overflow = k_check_epi32_overflow_8(&v[0], &v[1], &v[2], &v[3], &v[4],
+ &v[5], &v[6], &v[7], &kZero);
+ if (overflow) {
+ HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
+ return;
+ }
#endif // DCT_HIGH_BIT_DEPTH
- u[0] = k_packs_epi64(v[0], v[1]);
- u[1] = k_packs_epi64(v[2], v[3]);
- u[2] = k_packs_epi64(v[4], v[5]);
- u[3] = k_packs_epi64(v[6], v[7]);
+ u[0] = k_packs_epi64(v[0], v[1]);
+ u[1] = k_packs_epi64(v[2], v[3]);
+ u[2] = k_packs_epi64(v[4], v[5]);
+ u[3] = k_packs_epi64(v[6], v[7]);
- v[0] = _mm_add_epi32(u[0], k__DCT_CONST_ROUNDING);
- v[1] = _mm_add_epi32(u[1], k__DCT_CONST_ROUNDING);
- v[2] = _mm_add_epi32(u[2], k__DCT_CONST_ROUNDING);
- v[3] = _mm_add_epi32(u[3], k__DCT_CONST_ROUNDING);
+ v[0] = _mm_add_epi32(u[0], k__DCT_CONST_ROUNDING);
+ v[1] = _mm_add_epi32(u[1], k__DCT_CONST_ROUNDING);
+ v[2] = _mm_add_epi32(u[2], k__DCT_CONST_ROUNDING);
+ v[3] = _mm_add_epi32(u[3], k__DCT_CONST_ROUNDING);
- lstep1[10] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
- lstep1[11] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
- lstep1[12] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
- lstep1[13] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
+ lstep1[10] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
+ lstep1[11] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
+ lstep1[12] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
+ lstep1[13] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
}
{
const __m128i k32_m08_p24 = pair_set_epi32(-cospi_8_64, cospi_24_64);
const __m128i k32_m24_m08 = pair_set_epi32(-cospi_24_64, -cospi_8_64);
const __m128i k32_p24_p08 = pair_set_epi32(cospi_24_64, cospi_8_64);
- u[ 0] = _mm_unpacklo_epi32(lstep3[36], lstep3[58]);
- u[ 1] = _mm_unpackhi_epi32(lstep3[36], lstep3[58]);
- u[ 2] = _mm_unpacklo_epi32(lstep3[37], lstep3[59]);
- u[ 3] = _mm_unpackhi_epi32(lstep3[37], lstep3[59]);
- u[ 4] = _mm_unpacklo_epi32(lstep3[38], lstep3[56]);
- u[ 5] = _mm_unpackhi_epi32(lstep3[38], lstep3[56]);
- u[ 6] = _mm_unpacklo_epi32(lstep3[39], lstep3[57]);
- u[ 7] = _mm_unpackhi_epi32(lstep3[39], lstep3[57]);
- u[ 8] = _mm_unpacklo_epi32(lstep3[40], lstep3[54]);
- u[ 9] = _mm_unpackhi_epi32(lstep3[40], lstep3[54]);
+ u[0] = _mm_unpacklo_epi32(lstep3[36], lstep3[58]);
+ u[1] = _mm_unpackhi_epi32(lstep3[36], lstep3[58]);
+ u[2] = _mm_unpacklo_epi32(lstep3[37], lstep3[59]);
+ u[3] = _mm_unpackhi_epi32(lstep3[37], lstep3[59]);
+ u[4] = _mm_unpacklo_epi32(lstep3[38], lstep3[56]);
+ u[5] = _mm_unpackhi_epi32(lstep3[38], lstep3[56]);
+ u[6] = _mm_unpacklo_epi32(lstep3[39], lstep3[57]);
+ u[7] = _mm_unpackhi_epi32(lstep3[39], lstep3[57]);
+ u[8] = _mm_unpacklo_epi32(lstep3[40], lstep3[54]);
+ u[9] = _mm_unpackhi_epi32(lstep3[40], lstep3[54]);
u[10] = _mm_unpacklo_epi32(lstep3[41], lstep3[55]);
u[11] = _mm_unpackhi_epi32(lstep3[41], lstep3[55]);
u[12] = _mm_unpacklo_epi32(lstep3[42], lstep3[52]);
@@ -1752,16 +1807,16 @@
u[14] = _mm_unpacklo_epi32(lstep3[43], lstep3[53]);
u[15] = _mm_unpackhi_epi32(lstep3[43], lstep3[53]);
- v[ 0] = k_madd_epi32(u[ 0], k32_m08_p24);
- v[ 1] = k_madd_epi32(u[ 1], k32_m08_p24);
- v[ 2] = k_madd_epi32(u[ 2], k32_m08_p24);
- v[ 3] = k_madd_epi32(u[ 3], k32_m08_p24);
- v[ 4] = k_madd_epi32(u[ 4], k32_m08_p24);
- v[ 5] = k_madd_epi32(u[ 5], k32_m08_p24);
- v[ 6] = k_madd_epi32(u[ 6], k32_m08_p24);
- v[ 7] = k_madd_epi32(u[ 7], k32_m08_p24);
- v[ 8] = k_madd_epi32(u[ 8], k32_m24_m08);
- v[ 9] = k_madd_epi32(u[ 9], k32_m24_m08);
+ v[0] = k_madd_epi32(u[0], k32_m08_p24);
+ v[1] = k_madd_epi32(u[1], k32_m08_p24);
+ v[2] = k_madd_epi32(u[2], k32_m08_p24);
+ v[3] = k_madd_epi32(u[3], k32_m08_p24);
+ v[4] = k_madd_epi32(u[4], k32_m08_p24);
+ v[5] = k_madd_epi32(u[5], k32_m08_p24);
+ v[6] = k_madd_epi32(u[6], k32_m08_p24);
+ v[7] = k_madd_epi32(u[7], k32_m08_p24);
+ v[8] = k_madd_epi32(u[8], k32_m24_m08);
+ v[9] = k_madd_epi32(u[9], k32_m24_m08);
v[10] = k_madd_epi32(u[10], k32_m24_m08);
v[11] = k_madd_epi32(u[11], k32_m24_m08);
v[12] = k_madd_epi32(u[12], k32_m24_m08);
@@ -1772,41 +1827,40 @@
v[17] = k_madd_epi32(u[13], k32_m08_p24);
v[18] = k_madd_epi32(u[14], k32_m08_p24);
v[19] = k_madd_epi32(u[15], k32_m08_p24);
- v[20] = k_madd_epi32(u[ 8], k32_m08_p24);
- v[21] = k_madd_epi32(u[ 9], k32_m08_p24);
+ v[20] = k_madd_epi32(u[8], k32_m08_p24);
+ v[21] = k_madd_epi32(u[9], k32_m08_p24);
v[22] = k_madd_epi32(u[10], k32_m08_p24);
v[23] = k_madd_epi32(u[11], k32_m08_p24);
- v[24] = k_madd_epi32(u[ 4], k32_p24_p08);
- v[25] = k_madd_epi32(u[ 5], k32_p24_p08);
- v[26] = k_madd_epi32(u[ 6], k32_p24_p08);
- v[27] = k_madd_epi32(u[ 7], k32_p24_p08);
- v[28] = k_madd_epi32(u[ 0], k32_p24_p08);
- v[29] = k_madd_epi32(u[ 1], k32_p24_p08);
- v[30] = k_madd_epi32(u[ 2], k32_p24_p08);
- v[31] = k_madd_epi32(u[ 3], k32_p24_p08);
+ v[24] = k_madd_epi32(u[4], k32_p24_p08);
+ v[25] = k_madd_epi32(u[5], k32_p24_p08);
+ v[26] = k_madd_epi32(u[6], k32_p24_p08);
+ v[27] = k_madd_epi32(u[7], k32_p24_p08);
+ v[28] = k_madd_epi32(u[0], k32_p24_p08);
+ v[29] = k_madd_epi32(u[1], k32_p24_p08);
+ v[30] = k_madd_epi32(u[2], k32_p24_p08);
+ v[31] = k_madd_epi32(u[3], k32_p24_p08);
#if DCT_HIGH_BIT_DEPTH
overflow = k_check_epi32_overflow_32(
- &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7],
- &v[8], &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15],
- &v[16], &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23],
- &v[24], &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31],
- &kZero);
+ &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7], &v[8],
+ &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15], &v[16],
+ &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23], &v[24],
+ &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31], &kZero);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
- u[ 0] = k_packs_epi64(v[ 0], v[ 1]);
- u[ 1] = k_packs_epi64(v[ 2], v[ 3]);
- u[ 2] = k_packs_epi64(v[ 4], v[ 5]);
- u[ 3] = k_packs_epi64(v[ 6], v[ 7]);
- u[ 4] = k_packs_epi64(v[ 8], v[ 9]);
- u[ 5] = k_packs_epi64(v[10], v[11]);
- u[ 6] = k_packs_epi64(v[12], v[13]);
- u[ 7] = k_packs_epi64(v[14], v[15]);
- u[ 8] = k_packs_epi64(v[16], v[17]);
- u[ 9] = k_packs_epi64(v[18], v[19]);
+ u[0] = k_packs_epi64(v[0], v[1]);
+ u[1] = k_packs_epi64(v[2], v[3]);
+ u[2] = k_packs_epi64(v[4], v[5]);
+ u[3] = k_packs_epi64(v[6], v[7]);
+ u[4] = k_packs_epi64(v[8], v[9]);
+ u[5] = k_packs_epi64(v[10], v[11]);
+ u[6] = k_packs_epi64(v[12], v[13]);
+ u[7] = k_packs_epi64(v[14], v[15]);
+ u[8] = k_packs_epi64(v[16], v[17]);
+ u[9] = k_packs_epi64(v[18], v[19]);
u[10] = k_packs_epi64(v[20], v[21]);
u[11] = k_packs_epi64(v[22], v[23]);
u[12] = k_packs_epi64(v[24], v[25]);
@@ -1814,16 +1868,16 @@
u[14] = k_packs_epi64(v[28], v[29]);
u[15] = k_packs_epi64(v[30], v[31]);
- v[ 0] = _mm_add_epi32(u[ 0], k__DCT_CONST_ROUNDING);
- v[ 1] = _mm_add_epi32(u[ 1], k__DCT_CONST_ROUNDING);
- v[ 2] = _mm_add_epi32(u[ 2], k__DCT_CONST_ROUNDING);
- v[ 3] = _mm_add_epi32(u[ 3], k__DCT_CONST_ROUNDING);
- v[ 4] = _mm_add_epi32(u[ 4], k__DCT_CONST_ROUNDING);
- v[ 5] = _mm_add_epi32(u[ 5], k__DCT_CONST_ROUNDING);
- v[ 6] = _mm_add_epi32(u[ 6], k__DCT_CONST_ROUNDING);
- v[ 7] = _mm_add_epi32(u[ 7], k__DCT_CONST_ROUNDING);
- v[ 8] = _mm_add_epi32(u[ 8], k__DCT_CONST_ROUNDING);
- v[ 9] = _mm_add_epi32(u[ 9], k__DCT_CONST_ROUNDING);
+ v[0] = _mm_add_epi32(u[0], k__DCT_CONST_ROUNDING);
+ v[1] = _mm_add_epi32(u[1], k__DCT_CONST_ROUNDING);
+ v[2] = _mm_add_epi32(u[2], k__DCT_CONST_ROUNDING);
+ v[3] = _mm_add_epi32(u[3], k__DCT_CONST_ROUNDING);
+ v[4] = _mm_add_epi32(u[4], k__DCT_CONST_ROUNDING);
+ v[5] = _mm_add_epi32(u[5], k__DCT_CONST_ROUNDING);
+ v[6] = _mm_add_epi32(u[6], k__DCT_CONST_ROUNDING);
+ v[7] = _mm_add_epi32(u[7], k__DCT_CONST_ROUNDING);
+ v[8] = _mm_add_epi32(u[8], k__DCT_CONST_ROUNDING);
+ v[9] = _mm_add_epi32(u[9], k__DCT_CONST_ROUNDING);
v[10] = _mm_add_epi32(u[10], k__DCT_CONST_ROUNDING);
v[11] = _mm_add_epi32(u[11], k__DCT_CONST_ROUNDING);
v[12] = _mm_add_epi32(u[12], k__DCT_CONST_ROUNDING);
@@ -1831,16 +1885,16 @@
v[14] = _mm_add_epi32(u[14], k__DCT_CONST_ROUNDING);
v[15] = _mm_add_epi32(u[15], k__DCT_CONST_ROUNDING);
- lstep1[36] = _mm_srai_epi32(v[ 0], DCT_CONST_BITS);
- lstep1[37] = _mm_srai_epi32(v[ 1], DCT_CONST_BITS);
- lstep1[38] = _mm_srai_epi32(v[ 2], DCT_CONST_BITS);
- lstep1[39] = _mm_srai_epi32(v[ 3], DCT_CONST_BITS);
- lstep1[40] = _mm_srai_epi32(v[ 4], DCT_CONST_BITS);
- lstep1[41] = _mm_srai_epi32(v[ 5], DCT_CONST_BITS);
- lstep1[42] = _mm_srai_epi32(v[ 6], DCT_CONST_BITS);
- lstep1[43] = _mm_srai_epi32(v[ 7], DCT_CONST_BITS);
- lstep1[52] = _mm_srai_epi32(v[ 8], DCT_CONST_BITS);
- lstep1[53] = _mm_srai_epi32(v[ 9], DCT_CONST_BITS);
+ lstep1[36] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
+ lstep1[37] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
+ lstep1[38] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
+ lstep1[39] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
+ lstep1[40] = _mm_srai_epi32(v[4], DCT_CONST_BITS);
+ lstep1[41] = _mm_srai_epi32(v[5], DCT_CONST_BITS);
+ lstep1[42] = _mm_srai_epi32(v[6], DCT_CONST_BITS);
+ lstep1[43] = _mm_srai_epi32(v[7], DCT_CONST_BITS);
+ lstep1[52] = _mm_srai_epi32(v[8], DCT_CONST_BITS);
+ lstep1[53] = _mm_srai_epi32(v[9], DCT_CONST_BITS);
lstep1[54] = _mm_srai_epi32(v[10], DCT_CONST_BITS);
lstep1[55] = _mm_srai_epi32(v[11], DCT_CONST_BITS);
lstep1[56] = _mm_srai_epi32(v[12], DCT_CONST_BITS);
@@ -1850,10 +1904,10 @@
}
// stage 5
{
- lstep2[ 8] = _mm_add_epi32(lstep1[10], lstep3[ 8]);
- lstep2[ 9] = _mm_add_epi32(lstep1[11], lstep3[ 9]);
- lstep2[10] = _mm_sub_epi32(lstep3[ 8], lstep1[10]);
- lstep2[11] = _mm_sub_epi32(lstep3[ 9], lstep1[11]);
+ lstep2[8] = _mm_add_epi32(lstep1[10], lstep3[8]);
+ lstep2[9] = _mm_add_epi32(lstep1[11], lstep3[9]);
+ lstep2[10] = _mm_sub_epi32(lstep3[8], lstep1[10]);
+ lstep2[11] = _mm_sub_epi32(lstep3[9], lstep1[11]);
lstep2[12] = _mm_sub_epi32(lstep3[14], lstep1[12]);
lstep2[13] = _mm_sub_epi32(lstep3[15], lstep1[13]);
lstep2[14] = _mm_add_epi32(lstep1[12], lstep3[14]);
@@ -1876,16 +1930,16 @@
// TODO(jingning): manually inline k_madd_epi32_ to further hide
// instruction latency.
- v[ 0] = k_madd_epi32(u[0], k32_p16_p16);
- v[ 1] = k_madd_epi32(u[1], k32_p16_p16);
- v[ 2] = k_madd_epi32(u[2], k32_p16_p16);
- v[ 3] = k_madd_epi32(u[3], k32_p16_p16);
- v[ 4] = k_madd_epi32(u[0], k32_p16_m16);
- v[ 5] = k_madd_epi32(u[1], k32_p16_m16);
- v[ 6] = k_madd_epi32(u[2], k32_p16_m16);
- v[ 7] = k_madd_epi32(u[3], k32_p16_m16);
- v[ 8] = k_madd_epi32(u[4], k32_p24_p08);
- v[ 9] = k_madd_epi32(u[5], k32_p24_p08);
+ v[0] = k_madd_epi32(u[0], k32_p16_p16);
+ v[1] = k_madd_epi32(u[1], k32_p16_p16);
+ v[2] = k_madd_epi32(u[2], k32_p16_p16);
+ v[3] = k_madd_epi32(u[3], k32_p16_p16);
+ v[4] = k_madd_epi32(u[0], k32_p16_m16);
+ v[5] = k_madd_epi32(u[1], k32_p16_m16);
+ v[6] = k_madd_epi32(u[2], k32_p16_m16);
+ v[7] = k_madd_epi32(u[3], k32_p16_m16);
+ v[8] = k_madd_epi32(u[4], k32_p24_p08);
+ v[9] = k_madd_epi32(u[5], k32_p24_p08);
v[10] = k_madd_epi32(u[6], k32_p24_p08);
v[11] = k_madd_epi32(u[7], k32_p24_p08);
v[12] = k_madd_epi32(u[4], k32_m08_p24);
@@ -1895,9 +1949,8 @@
#if DCT_HIGH_BIT_DEPTH
overflow = k_check_epi32_overflow_16(
- &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7],
- &v[8], &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15],
- &kZero);
+ &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7], &v[8],
+ &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15], &kZero);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
@@ -1967,13 +2020,13 @@
u[7] = _mm_srai_epi32(u[7], 2);
// Combine
- out[ 0] = _mm_packs_epi32(u[0], u[1]);
+ out[0] = _mm_packs_epi32(u[0], u[1]);
out[16] = _mm_packs_epi32(u[2], u[3]);
- out[ 8] = _mm_packs_epi32(u[4], u[5]);
+ out[8] = _mm_packs_epi32(u[4], u[5]);
out[24] = _mm_packs_epi32(u[6], u[7]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&out[0], &out[16],
- &out[8], &out[24]);
+ overflow =
+ check_epi16_overflow_x4(&out[0], &out[16], &out[8], &out[24]);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
@@ -2002,8 +2055,8 @@
v[5] = k_madd_epi32(u[5], k32_m24_m08);
v[6] = k_madd_epi32(u[6], k32_m24_m08);
v[7] = k_madd_epi32(u[7], k32_m24_m08);
- v[ 8] = k_madd_epi32(u[4], k32_m08_p24);
- v[ 9] = k_madd_epi32(u[5], k32_m08_p24);
+ v[8] = k_madd_epi32(u[4], k32_m08_p24);
+ v[9] = k_madd_epi32(u[5], k32_m08_p24);
v[10] = k_madd_epi32(u[6], k32_m08_p24);
v[11] = k_madd_epi32(u[7], k32_m08_p24);
v[12] = k_madd_epi32(u[0], k32_p24_p08);
@@ -2013,9 +2066,8 @@
#if DCT_HIGH_BIT_DEPTH
overflow = k_check_epi32_overflow_16(
- &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7],
- &v[8], &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15],
- &kZero);
+ &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7], &v[8],
+ &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15], &kZero);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
@@ -2089,10 +2141,10 @@
const __m128i k32_m20_p12 = pair_set_epi32(-cospi_20_64, cospi_12_64);
const __m128i k32_m04_p28 = pair_set_epi32(-cospi_4_64, cospi_28_64);
- u[0] = _mm_unpacklo_epi32(lstep2[ 8], lstep2[14]);
- u[1] = _mm_unpackhi_epi32(lstep2[ 8], lstep2[14]);
- u[2] = _mm_unpacklo_epi32(lstep2[ 9], lstep2[15]);
- u[3] = _mm_unpackhi_epi32(lstep2[ 9], lstep2[15]);
+ u[0] = _mm_unpacklo_epi32(lstep2[8], lstep2[14]);
+ u[1] = _mm_unpackhi_epi32(lstep2[8], lstep2[14]);
+ u[2] = _mm_unpacklo_epi32(lstep2[9], lstep2[15]);
+ u[3] = _mm_unpackhi_epi32(lstep2[9], lstep2[15]);
u[4] = _mm_unpacklo_epi32(lstep2[10], lstep2[12]);
u[5] = _mm_unpackhi_epi32(lstep2[10], lstep2[12]);
u[6] = _mm_unpacklo_epi32(lstep2[11], lstep2[13]);
@@ -2101,10 +2153,10 @@
u[9] = _mm_unpackhi_epi32(lstep2[10], lstep2[12]);
u[10] = _mm_unpacklo_epi32(lstep2[11], lstep2[13]);
u[11] = _mm_unpackhi_epi32(lstep2[11], lstep2[13]);
- u[12] = _mm_unpacklo_epi32(lstep2[ 8], lstep2[14]);
- u[13] = _mm_unpackhi_epi32(lstep2[ 8], lstep2[14]);
- u[14] = _mm_unpacklo_epi32(lstep2[ 9], lstep2[15]);
- u[15] = _mm_unpackhi_epi32(lstep2[ 9], lstep2[15]);
+ u[12] = _mm_unpacklo_epi32(lstep2[8], lstep2[14]);
+ u[13] = _mm_unpackhi_epi32(lstep2[8], lstep2[14]);
+ u[14] = _mm_unpacklo_epi32(lstep2[9], lstep2[15]);
+ u[15] = _mm_unpackhi_epi32(lstep2[9], lstep2[15]);
v[0] = k_madd_epi32(u[0], k32_p28_p04);
v[1] = k_madd_epi32(u[1], k32_p28_p04);
@@ -2114,8 +2166,8 @@
v[5] = k_madd_epi32(u[5], k32_p12_p20);
v[6] = k_madd_epi32(u[6], k32_p12_p20);
v[7] = k_madd_epi32(u[7], k32_p12_p20);
- v[ 8] = k_madd_epi32(u[ 8], k32_m20_p12);
- v[ 9] = k_madd_epi32(u[ 9], k32_m20_p12);
+ v[8] = k_madd_epi32(u[8], k32_m20_p12);
+ v[9] = k_madd_epi32(u[9], k32_m20_p12);
v[10] = k_madd_epi32(u[10], k32_m20_p12);
v[11] = k_madd_epi32(u[11], k32_m20_p12);
v[12] = k_madd_epi32(u[12], k32_m04_p28);
@@ -2125,9 +2177,8 @@
#if DCT_HIGH_BIT_DEPTH
overflow = k_check_epi32_overflow_16(
- &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7],
- &v[8], &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15],
- &kZero);
+ &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7], &v[8],
+ &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15], &kZero);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
@@ -2196,13 +2247,13 @@
u[6] = _mm_srai_epi32(u[6], 2);
u[7] = _mm_srai_epi32(u[7], 2);
- out[ 4] = _mm_packs_epi32(u[0], u[1]);
+ out[4] = _mm_packs_epi32(u[0], u[1]);
out[20] = _mm_packs_epi32(u[2], u[3]);
out[12] = _mm_packs_epi32(u[4], u[5]);
out[28] = _mm_packs_epi32(u[6], u[7]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&out[4], &out[20],
- &out[12], &out[28]);
+ overflow =
+ check_epi16_overflow_x4(&out[4], &out[20], &out[12], &out[28]);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
@@ -2231,21 +2282,21 @@
const __m128i k32_m04_p28 = pair_set_epi32(-cospi_4_64, cospi_28_64);
const __m128i k32_m28_m04 = pair_set_epi32(-cospi_28_64, -cospi_4_64);
const __m128i k32_m20_p12 = pair_set_epi32(-cospi_20_64, cospi_12_64);
- const __m128i k32_m12_m20 = pair_set_epi32(-cospi_12_64,
- -cospi_20_64);
+ const __m128i k32_m12_m20 =
+ pair_set_epi32(-cospi_12_64, -cospi_20_64);
const __m128i k32_p12_p20 = pair_set_epi32(cospi_12_64, cospi_20_64);
const __m128i k32_p28_p04 = pair_set_epi32(cospi_28_64, cospi_4_64);
- u[ 0] = _mm_unpacklo_epi32(lstep2[34], lstep2[60]);
- u[ 1] = _mm_unpackhi_epi32(lstep2[34], lstep2[60]);
- u[ 2] = _mm_unpacklo_epi32(lstep2[35], lstep2[61]);
- u[ 3] = _mm_unpackhi_epi32(lstep2[35], lstep2[61]);
- u[ 4] = _mm_unpacklo_epi32(lstep2[36], lstep2[58]);
- u[ 5] = _mm_unpackhi_epi32(lstep2[36], lstep2[58]);
- u[ 6] = _mm_unpacklo_epi32(lstep2[37], lstep2[59]);
- u[ 7] = _mm_unpackhi_epi32(lstep2[37], lstep2[59]);
- u[ 8] = _mm_unpacklo_epi32(lstep2[42], lstep2[52]);
- u[ 9] = _mm_unpackhi_epi32(lstep2[42], lstep2[52]);
+ u[0] = _mm_unpacklo_epi32(lstep2[34], lstep2[60]);
+ u[1] = _mm_unpackhi_epi32(lstep2[34], lstep2[60]);
+ u[2] = _mm_unpacklo_epi32(lstep2[35], lstep2[61]);
+ u[3] = _mm_unpackhi_epi32(lstep2[35], lstep2[61]);
+ u[4] = _mm_unpacklo_epi32(lstep2[36], lstep2[58]);
+ u[5] = _mm_unpackhi_epi32(lstep2[36], lstep2[58]);
+ u[6] = _mm_unpacklo_epi32(lstep2[37], lstep2[59]);
+ u[7] = _mm_unpackhi_epi32(lstep2[37], lstep2[59]);
+ u[8] = _mm_unpacklo_epi32(lstep2[42], lstep2[52]);
+ u[9] = _mm_unpackhi_epi32(lstep2[42], lstep2[52]);
u[10] = _mm_unpacklo_epi32(lstep2[43], lstep2[53]);
u[11] = _mm_unpackhi_epi32(lstep2[43], lstep2[53]);
u[12] = _mm_unpacklo_epi32(lstep2[44], lstep2[50]);
@@ -2253,16 +2304,16 @@
u[14] = _mm_unpacklo_epi32(lstep2[45], lstep2[51]);
u[15] = _mm_unpackhi_epi32(lstep2[45], lstep2[51]);
- v[ 0] = k_madd_epi32(u[ 0], k32_m04_p28);
- v[ 1] = k_madd_epi32(u[ 1], k32_m04_p28);
- v[ 2] = k_madd_epi32(u[ 2], k32_m04_p28);
- v[ 3] = k_madd_epi32(u[ 3], k32_m04_p28);
- v[ 4] = k_madd_epi32(u[ 4], k32_m28_m04);
- v[ 5] = k_madd_epi32(u[ 5], k32_m28_m04);
- v[ 6] = k_madd_epi32(u[ 6], k32_m28_m04);
- v[ 7] = k_madd_epi32(u[ 7], k32_m28_m04);
- v[ 8] = k_madd_epi32(u[ 8], k32_m20_p12);
- v[ 9] = k_madd_epi32(u[ 9], k32_m20_p12);
+ v[0] = k_madd_epi32(u[0], k32_m04_p28);
+ v[1] = k_madd_epi32(u[1], k32_m04_p28);
+ v[2] = k_madd_epi32(u[2], k32_m04_p28);
+ v[3] = k_madd_epi32(u[3], k32_m04_p28);
+ v[4] = k_madd_epi32(u[4], k32_m28_m04);
+ v[5] = k_madd_epi32(u[5], k32_m28_m04);
+ v[6] = k_madd_epi32(u[6], k32_m28_m04);
+ v[7] = k_madd_epi32(u[7], k32_m28_m04);
+ v[8] = k_madd_epi32(u[8], k32_m20_p12);
+ v[9] = k_madd_epi32(u[9], k32_m20_p12);
v[10] = k_madd_epi32(u[10], k32_m20_p12);
v[11] = k_madd_epi32(u[11], k32_m20_p12);
v[12] = k_madd_epi32(u[12], k32_m12_m20);
@@ -2273,41 +2324,40 @@
v[17] = k_madd_epi32(u[13], k32_m20_p12);
v[18] = k_madd_epi32(u[14], k32_m20_p12);
v[19] = k_madd_epi32(u[15], k32_m20_p12);
- v[20] = k_madd_epi32(u[ 8], k32_p12_p20);
- v[21] = k_madd_epi32(u[ 9], k32_p12_p20);
+ v[20] = k_madd_epi32(u[8], k32_p12_p20);
+ v[21] = k_madd_epi32(u[9], k32_p12_p20);
v[22] = k_madd_epi32(u[10], k32_p12_p20);
v[23] = k_madd_epi32(u[11], k32_p12_p20);
- v[24] = k_madd_epi32(u[ 4], k32_m04_p28);
- v[25] = k_madd_epi32(u[ 5], k32_m04_p28);
- v[26] = k_madd_epi32(u[ 6], k32_m04_p28);
- v[27] = k_madd_epi32(u[ 7], k32_m04_p28);
- v[28] = k_madd_epi32(u[ 0], k32_p28_p04);
- v[29] = k_madd_epi32(u[ 1], k32_p28_p04);
- v[30] = k_madd_epi32(u[ 2], k32_p28_p04);
- v[31] = k_madd_epi32(u[ 3], k32_p28_p04);
+ v[24] = k_madd_epi32(u[4], k32_m04_p28);
+ v[25] = k_madd_epi32(u[5], k32_m04_p28);
+ v[26] = k_madd_epi32(u[6], k32_m04_p28);
+ v[27] = k_madd_epi32(u[7], k32_m04_p28);
+ v[28] = k_madd_epi32(u[0], k32_p28_p04);
+ v[29] = k_madd_epi32(u[1], k32_p28_p04);
+ v[30] = k_madd_epi32(u[2], k32_p28_p04);
+ v[31] = k_madd_epi32(u[3], k32_p28_p04);
#if DCT_HIGH_BIT_DEPTH
overflow = k_check_epi32_overflow_32(
- &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7],
- &v[8], &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15],
- &v[16], &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23],
- &v[24], &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31],
- &kZero);
+ &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7], &v[8],
+ &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15], &v[16],
+ &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23], &v[24],
+ &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31], &kZero);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
- u[ 0] = k_packs_epi64(v[ 0], v[ 1]);
- u[ 1] = k_packs_epi64(v[ 2], v[ 3]);
- u[ 2] = k_packs_epi64(v[ 4], v[ 5]);
- u[ 3] = k_packs_epi64(v[ 6], v[ 7]);
- u[ 4] = k_packs_epi64(v[ 8], v[ 9]);
- u[ 5] = k_packs_epi64(v[10], v[11]);
- u[ 6] = k_packs_epi64(v[12], v[13]);
- u[ 7] = k_packs_epi64(v[14], v[15]);
- u[ 8] = k_packs_epi64(v[16], v[17]);
- u[ 9] = k_packs_epi64(v[18], v[19]);
+ u[0] = k_packs_epi64(v[0], v[1]);
+ u[1] = k_packs_epi64(v[2], v[3]);
+ u[2] = k_packs_epi64(v[4], v[5]);
+ u[3] = k_packs_epi64(v[6], v[7]);
+ u[4] = k_packs_epi64(v[8], v[9]);
+ u[5] = k_packs_epi64(v[10], v[11]);
+ u[6] = k_packs_epi64(v[12], v[13]);
+ u[7] = k_packs_epi64(v[14], v[15]);
+ u[8] = k_packs_epi64(v[16], v[17]);
+ u[9] = k_packs_epi64(v[18], v[19]);
u[10] = k_packs_epi64(v[20], v[21]);
u[11] = k_packs_epi64(v[22], v[23]);
u[12] = k_packs_epi64(v[24], v[25]);
@@ -2315,16 +2365,16 @@
u[14] = k_packs_epi64(v[28], v[29]);
u[15] = k_packs_epi64(v[30], v[31]);
- v[ 0] = _mm_add_epi32(u[ 0], k__DCT_CONST_ROUNDING);
- v[ 1] = _mm_add_epi32(u[ 1], k__DCT_CONST_ROUNDING);
- v[ 2] = _mm_add_epi32(u[ 2], k__DCT_CONST_ROUNDING);
- v[ 3] = _mm_add_epi32(u[ 3], k__DCT_CONST_ROUNDING);
- v[ 4] = _mm_add_epi32(u[ 4], k__DCT_CONST_ROUNDING);
- v[ 5] = _mm_add_epi32(u[ 5], k__DCT_CONST_ROUNDING);
- v[ 6] = _mm_add_epi32(u[ 6], k__DCT_CONST_ROUNDING);
- v[ 7] = _mm_add_epi32(u[ 7], k__DCT_CONST_ROUNDING);
- v[ 8] = _mm_add_epi32(u[ 8], k__DCT_CONST_ROUNDING);
- v[ 9] = _mm_add_epi32(u[ 9], k__DCT_CONST_ROUNDING);
+ v[0] = _mm_add_epi32(u[0], k__DCT_CONST_ROUNDING);
+ v[1] = _mm_add_epi32(u[1], k__DCT_CONST_ROUNDING);
+ v[2] = _mm_add_epi32(u[2], k__DCT_CONST_ROUNDING);
+ v[3] = _mm_add_epi32(u[3], k__DCT_CONST_ROUNDING);
+ v[4] = _mm_add_epi32(u[4], k__DCT_CONST_ROUNDING);
+ v[5] = _mm_add_epi32(u[5], k__DCT_CONST_ROUNDING);
+ v[6] = _mm_add_epi32(u[6], k__DCT_CONST_ROUNDING);
+ v[7] = _mm_add_epi32(u[7], k__DCT_CONST_ROUNDING);
+ v[8] = _mm_add_epi32(u[8], k__DCT_CONST_ROUNDING);
+ v[9] = _mm_add_epi32(u[9], k__DCT_CONST_ROUNDING);
v[10] = _mm_add_epi32(u[10], k__DCT_CONST_ROUNDING);
v[11] = _mm_add_epi32(u[11], k__DCT_CONST_ROUNDING);
v[12] = _mm_add_epi32(u[12], k__DCT_CONST_ROUNDING);
@@ -2332,16 +2382,16 @@
v[14] = _mm_add_epi32(u[14], k__DCT_CONST_ROUNDING);
v[15] = _mm_add_epi32(u[15], k__DCT_CONST_ROUNDING);
- lstep3[34] = _mm_srai_epi32(v[ 0], DCT_CONST_BITS);
- lstep3[35] = _mm_srai_epi32(v[ 1], DCT_CONST_BITS);
- lstep3[36] = _mm_srai_epi32(v[ 2], DCT_CONST_BITS);
- lstep3[37] = _mm_srai_epi32(v[ 3], DCT_CONST_BITS);
- lstep3[42] = _mm_srai_epi32(v[ 4], DCT_CONST_BITS);
- lstep3[43] = _mm_srai_epi32(v[ 5], DCT_CONST_BITS);
- lstep3[44] = _mm_srai_epi32(v[ 6], DCT_CONST_BITS);
- lstep3[45] = _mm_srai_epi32(v[ 7], DCT_CONST_BITS);
- lstep3[50] = _mm_srai_epi32(v[ 8], DCT_CONST_BITS);
- lstep3[51] = _mm_srai_epi32(v[ 9], DCT_CONST_BITS);
+ lstep3[34] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
+ lstep3[35] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
+ lstep3[36] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
+ lstep3[37] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
+ lstep3[42] = _mm_srai_epi32(v[4], DCT_CONST_BITS);
+ lstep3[43] = _mm_srai_epi32(v[5], DCT_CONST_BITS);
+ lstep3[44] = _mm_srai_epi32(v[6], DCT_CONST_BITS);
+ lstep3[45] = _mm_srai_epi32(v[7], DCT_CONST_BITS);
+ lstep3[50] = _mm_srai_epi32(v[8], DCT_CONST_BITS);
+ lstep3[51] = _mm_srai_epi32(v[9], DCT_CONST_BITS);
lstep3[52] = _mm_srai_epi32(v[10], DCT_CONST_BITS);
lstep3[53] = _mm_srai_epi32(v[11], DCT_CONST_BITS);
lstep3[58] = _mm_srai_epi32(v[12], DCT_CONST_BITS);
@@ -2354,22 +2404,22 @@
const __m128i k32_p30_p02 = pair_set_epi32(cospi_30_64, cospi_2_64);
const __m128i k32_p14_p18 = pair_set_epi32(cospi_14_64, cospi_18_64);
const __m128i k32_p22_p10 = pair_set_epi32(cospi_22_64, cospi_10_64);
- const __m128i k32_p06_p26 = pair_set_epi32(cospi_6_64, cospi_26_64);
+ const __m128i k32_p06_p26 = pair_set_epi32(cospi_6_64, cospi_26_64);
const __m128i k32_m26_p06 = pair_set_epi32(-cospi_26_64, cospi_6_64);
const __m128i k32_m10_p22 = pair_set_epi32(-cospi_10_64, cospi_22_64);
const __m128i k32_m18_p14 = pair_set_epi32(-cospi_18_64, cospi_14_64);
const __m128i k32_m02_p30 = pair_set_epi32(-cospi_2_64, cospi_30_64);
- u[ 0] = _mm_unpacklo_epi32(lstep3[16], lstep3[30]);
- u[ 1] = _mm_unpackhi_epi32(lstep3[16], lstep3[30]);
- u[ 2] = _mm_unpacklo_epi32(lstep3[17], lstep3[31]);
- u[ 3] = _mm_unpackhi_epi32(lstep3[17], lstep3[31]);
- u[ 4] = _mm_unpacklo_epi32(lstep3[18], lstep3[28]);
- u[ 5] = _mm_unpackhi_epi32(lstep3[18], lstep3[28]);
- u[ 6] = _mm_unpacklo_epi32(lstep3[19], lstep3[29]);
- u[ 7] = _mm_unpackhi_epi32(lstep3[19], lstep3[29]);
- u[ 8] = _mm_unpacklo_epi32(lstep3[20], lstep3[26]);
- u[ 9] = _mm_unpackhi_epi32(lstep3[20], lstep3[26]);
+ u[0] = _mm_unpacklo_epi32(lstep3[16], lstep3[30]);
+ u[1] = _mm_unpackhi_epi32(lstep3[16], lstep3[30]);
+ u[2] = _mm_unpacklo_epi32(lstep3[17], lstep3[31]);
+ u[3] = _mm_unpackhi_epi32(lstep3[17], lstep3[31]);
+ u[4] = _mm_unpacklo_epi32(lstep3[18], lstep3[28]);
+ u[5] = _mm_unpackhi_epi32(lstep3[18], lstep3[28]);
+ u[6] = _mm_unpacklo_epi32(lstep3[19], lstep3[29]);
+ u[7] = _mm_unpackhi_epi32(lstep3[19], lstep3[29]);
+ u[8] = _mm_unpacklo_epi32(lstep3[20], lstep3[26]);
+ u[9] = _mm_unpackhi_epi32(lstep3[20], lstep3[26]);
u[10] = _mm_unpacklo_epi32(lstep3[21], lstep3[27]);
u[11] = _mm_unpackhi_epi32(lstep3[21], lstep3[27]);
u[12] = _mm_unpacklo_epi32(lstep3[22], lstep3[24]);
@@ -2377,16 +2427,16 @@
u[14] = _mm_unpacklo_epi32(lstep3[23], lstep3[25]);
u[15] = _mm_unpackhi_epi32(lstep3[23], lstep3[25]);
- v[ 0] = k_madd_epi32(u[ 0], k32_p30_p02);
- v[ 1] = k_madd_epi32(u[ 1], k32_p30_p02);
- v[ 2] = k_madd_epi32(u[ 2], k32_p30_p02);
- v[ 3] = k_madd_epi32(u[ 3], k32_p30_p02);
- v[ 4] = k_madd_epi32(u[ 4], k32_p14_p18);
- v[ 5] = k_madd_epi32(u[ 5], k32_p14_p18);
- v[ 6] = k_madd_epi32(u[ 6], k32_p14_p18);
- v[ 7] = k_madd_epi32(u[ 7], k32_p14_p18);
- v[ 8] = k_madd_epi32(u[ 8], k32_p22_p10);
- v[ 9] = k_madd_epi32(u[ 9], k32_p22_p10);
+ v[0] = k_madd_epi32(u[0], k32_p30_p02);
+ v[1] = k_madd_epi32(u[1], k32_p30_p02);
+ v[2] = k_madd_epi32(u[2], k32_p30_p02);
+ v[3] = k_madd_epi32(u[3], k32_p30_p02);
+ v[4] = k_madd_epi32(u[4], k32_p14_p18);
+ v[5] = k_madd_epi32(u[5], k32_p14_p18);
+ v[6] = k_madd_epi32(u[6], k32_p14_p18);
+ v[7] = k_madd_epi32(u[7], k32_p14_p18);
+ v[8] = k_madd_epi32(u[8], k32_p22_p10);
+ v[9] = k_madd_epi32(u[9], k32_p22_p10);
v[10] = k_madd_epi32(u[10], k32_p22_p10);
v[11] = k_madd_epi32(u[11], k32_p22_p10);
v[12] = k_madd_epi32(u[12], k32_p06_p26);
@@ -2397,41 +2447,40 @@
v[17] = k_madd_epi32(u[13], k32_m26_p06);
v[18] = k_madd_epi32(u[14], k32_m26_p06);
v[19] = k_madd_epi32(u[15], k32_m26_p06);
- v[20] = k_madd_epi32(u[ 8], k32_m10_p22);
- v[21] = k_madd_epi32(u[ 9], k32_m10_p22);
+ v[20] = k_madd_epi32(u[8], k32_m10_p22);
+ v[21] = k_madd_epi32(u[9], k32_m10_p22);
v[22] = k_madd_epi32(u[10], k32_m10_p22);
v[23] = k_madd_epi32(u[11], k32_m10_p22);
- v[24] = k_madd_epi32(u[ 4], k32_m18_p14);
- v[25] = k_madd_epi32(u[ 5], k32_m18_p14);
- v[26] = k_madd_epi32(u[ 6], k32_m18_p14);
- v[27] = k_madd_epi32(u[ 7], k32_m18_p14);
- v[28] = k_madd_epi32(u[ 0], k32_m02_p30);
- v[29] = k_madd_epi32(u[ 1], k32_m02_p30);
- v[30] = k_madd_epi32(u[ 2], k32_m02_p30);
- v[31] = k_madd_epi32(u[ 3], k32_m02_p30);
+ v[24] = k_madd_epi32(u[4], k32_m18_p14);
+ v[25] = k_madd_epi32(u[5], k32_m18_p14);
+ v[26] = k_madd_epi32(u[6], k32_m18_p14);
+ v[27] = k_madd_epi32(u[7], k32_m18_p14);
+ v[28] = k_madd_epi32(u[0], k32_m02_p30);
+ v[29] = k_madd_epi32(u[1], k32_m02_p30);
+ v[30] = k_madd_epi32(u[2], k32_m02_p30);
+ v[31] = k_madd_epi32(u[3], k32_m02_p30);
#if DCT_HIGH_BIT_DEPTH
overflow = k_check_epi32_overflow_32(
- &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7],
- &v[8], &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15],
- &v[16], &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23],
- &v[24], &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31],
- &kZero);
+ &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7], &v[8],
+ &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15], &v[16],
+ &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23], &v[24],
+ &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31], &kZero);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
- u[ 0] = k_packs_epi64(v[ 0], v[ 1]);
- u[ 1] = k_packs_epi64(v[ 2], v[ 3]);
- u[ 2] = k_packs_epi64(v[ 4], v[ 5]);
- u[ 3] = k_packs_epi64(v[ 6], v[ 7]);
- u[ 4] = k_packs_epi64(v[ 8], v[ 9]);
- u[ 5] = k_packs_epi64(v[10], v[11]);
- u[ 6] = k_packs_epi64(v[12], v[13]);
- u[ 7] = k_packs_epi64(v[14], v[15]);
- u[ 8] = k_packs_epi64(v[16], v[17]);
- u[ 9] = k_packs_epi64(v[18], v[19]);
+ u[0] = k_packs_epi64(v[0], v[1]);
+ u[1] = k_packs_epi64(v[2], v[3]);
+ u[2] = k_packs_epi64(v[4], v[5]);
+ u[3] = k_packs_epi64(v[6], v[7]);
+ u[4] = k_packs_epi64(v[8], v[9]);
+ u[5] = k_packs_epi64(v[10], v[11]);
+ u[6] = k_packs_epi64(v[12], v[13]);
+ u[7] = k_packs_epi64(v[14], v[15]);
+ u[8] = k_packs_epi64(v[16], v[17]);
+ u[9] = k_packs_epi64(v[18], v[19]);
u[10] = k_packs_epi64(v[20], v[21]);
u[11] = k_packs_epi64(v[22], v[23]);
u[12] = k_packs_epi64(v[24], v[25]);
@@ -2439,16 +2488,16 @@
u[14] = k_packs_epi64(v[28], v[29]);
u[15] = k_packs_epi64(v[30], v[31]);
- v[ 0] = _mm_add_epi32(u[ 0], k__DCT_CONST_ROUNDING);
- v[ 1] = _mm_add_epi32(u[ 1], k__DCT_CONST_ROUNDING);
- v[ 2] = _mm_add_epi32(u[ 2], k__DCT_CONST_ROUNDING);
- v[ 3] = _mm_add_epi32(u[ 3], k__DCT_CONST_ROUNDING);
- v[ 4] = _mm_add_epi32(u[ 4], k__DCT_CONST_ROUNDING);
- v[ 5] = _mm_add_epi32(u[ 5], k__DCT_CONST_ROUNDING);
- v[ 6] = _mm_add_epi32(u[ 6], k__DCT_CONST_ROUNDING);
- v[ 7] = _mm_add_epi32(u[ 7], k__DCT_CONST_ROUNDING);
- v[ 8] = _mm_add_epi32(u[ 8], k__DCT_CONST_ROUNDING);
- v[ 9] = _mm_add_epi32(u[ 9], k__DCT_CONST_ROUNDING);
+ v[0] = _mm_add_epi32(u[0], k__DCT_CONST_ROUNDING);
+ v[1] = _mm_add_epi32(u[1], k__DCT_CONST_ROUNDING);
+ v[2] = _mm_add_epi32(u[2], k__DCT_CONST_ROUNDING);
+ v[3] = _mm_add_epi32(u[3], k__DCT_CONST_ROUNDING);
+ v[4] = _mm_add_epi32(u[4], k__DCT_CONST_ROUNDING);
+ v[5] = _mm_add_epi32(u[5], k__DCT_CONST_ROUNDING);
+ v[6] = _mm_add_epi32(u[6], k__DCT_CONST_ROUNDING);
+ v[7] = _mm_add_epi32(u[7], k__DCT_CONST_ROUNDING);
+ v[8] = _mm_add_epi32(u[8], k__DCT_CONST_ROUNDING);
+ v[9] = _mm_add_epi32(u[9], k__DCT_CONST_ROUNDING);
v[10] = _mm_add_epi32(u[10], k__DCT_CONST_ROUNDING);
v[11] = _mm_add_epi32(u[11], k__DCT_CONST_ROUNDING);
v[12] = _mm_add_epi32(u[12], k__DCT_CONST_ROUNDING);
@@ -2456,16 +2505,16 @@
v[14] = _mm_add_epi32(u[14], k__DCT_CONST_ROUNDING);
v[15] = _mm_add_epi32(u[15], k__DCT_CONST_ROUNDING);
- u[ 0] = _mm_srai_epi32(v[ 0], DCT_CONST_BITS);
- u[ 1] = _mm_srai_epi32(v[ 1], DCT_CONST_BITS);
- u[ 2] = _mm_srai_epi32(v[ 2], DCT_CONST_BITS);
- u[ 3] = _mm_srai_epi32(v[ 3], DCT_CONST_BITS);
- u[ 4] = _mm_srai_epi32(v[ 4], DCT_CONST_BITS);
- u[ 5] = _mm_srai_epi32(v[ 5], DCT_CONST_BITS);
- u[ 6] = _mm_srai_epi32(v[ 6], DCT_CONST_BITS);
- u[ 7] = _mm_srai_epi32(v[ 7], DCT_CONST_BITS);
- u[ 8] = _mm_srai_epi32(v[ 8], DCT_CONST_BITS);
- u[ 9] = _mm_srai_epi32(v[ 9], DCT_CONST_BITS);
+ u[0] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
+ u[1] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
+ u[2] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
+ u[3] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
+ u[4] = _mm_srai_epi32(v[4], DCT_CONST_BITS);
+ u[5] = _mm_srai_epi32(v[5], DCT_CONST_BITS);
+ u[6] = _mm_srai_epi32(v[6], DCT_CONST_BITS);
+ u[7] = _mm_srai_epi32(v[7], DCT_CONST_BITS);
+ u[8] = _mm_srai_epi32(v[8], DCT_CONST_BITS);
+ u[9] = _mm_srai_epi32(v[9], DCT_CONST_BITS);
u[10] = _mm_srai_epi32(v[10], DCT_CONST_BITS);
u[11] = _mm_srai_epi32(v[11], DCT_CONST_BITS);
u[12] = _mm_srai_epi32(v[12], DCT_CONST_BITS);
@@ -2473,16 +2522,16 @@
u[14] = _mm_srai_epi32(v[14], DCT_CONST_BITS);
u[15] = _mm_srai_epi32(v[15], DCT_CONST_BITS);
- v[ 0] = _mm_cmplt_epi32(u[ 0], kZero);
- v[ 1] = _mm_cmplt_epi32(u[ 1], kZero);
- v[ 2] = _mm_cmplt_epi32(u[ 2], kZero);
- v[ 3] = _mm_cmplt_epi32(u[ 3], kZero);
- v[ 4] = _mm_cmplt_epi32(u[ 4], kZero);
- v[ 5] = _mm_cmplt_epi32(u[ 5], kZero);
- v[ 6] = _mm_cmplt_epi32(u[ 6], kZero);
- v[ 7] = _mm_cmplt_epi32(u[ 7], kZero);
- v[ 8] = _mm_cmplt_epi32(u[ 8], kZero);
- v[ 9] = _mm_cmplt_epi32(u[ 9], kZero);
+ v[0] = _mm_cmplt_epi32(u[0], kZero);
+ v[1] = _mm_cmplt_epi32(u[1], kZero);
+ v[2] = _mm_cmplt_epi32(u[2], kZero);
+ v[3] = _mm_cmplt_epi32(u[3], kZero);
+ v[4] = _mm_cmplt_epi32(u[4], kZero);
+ v[5] = _mm_cmplt_epi32(u[5], kZero);
+ v[6] = _mm_cmplt_epi32(u[6], kZero);
+ v[7] = _mm_cmplt_epi32(u[7], kZero);
+ v[8] = _mm_cmplt_epi32(u[8], kZero);
+ v[9] = _mm_cmplt_epi32(u[9], kZero);
v[10] = _mm_cmplt_epi32(u[10], kZero);
v[11] = _mm_cmplt_epi32(u[11], kZero);
v[12] = _mm_cmplt_epi32(u[12], kZero);
@@ -2490,16 +2539,16 @@
v[14] = _mm_cmplt_epi32(u[14], kZero);
v[15] = _mm_cmplt_epi32(u[15], kZero);
- u[ 0] = _mm_sub_epi32(u[ 0], v[ 0]);
- u[ 1] = _mm_sub_epi32(u[ 1], v[ 1]);
- u[ 2] = _mm_sub_epi32(u[ 2], v[ 2]);
- u[ 3] = _mm_sub_epi32(u[ 3], v[ 3]);
- u[ 4] = _mm_sub_epi32(u[ 4], v[ 4]);
- u[ 5] = _mm_sub_epi32(u[ 5], v[ 5]);
- u[ 6] = _mm_sub_epi32(u[ 6], v[ 6]);
- u[ 7] = _mm_sub_epi32(u[ 7], v[ 7]);
- u[ 8] = _mm_sub_epi32(u[ 8], v[ 8]);
- u[ 9] = _mm_sub_epi32(u[ 9], v[ 9]);
+ u[0] = _mm_sub_epi32(u[0], v[0]);
+ u[1] = _mm_sub_epi32(u[1], v[1]);
+ u[2] = _mm_sub_epi32(u[2], v[2]);
+ u[3] = _mm_sub_epi32(u[3], v[3]);
+ u[4] = _mm_sub_epi32(u[4], v[4]);
+ u[5] = _mm_sub_epi32(u[5], v[5]);
+ u[6] = _mm_sub_epi32(u[6], v[6]);
+ u[7] = _mm_sub_epi32(u[7], v[7]);
+ u[8] = _mm_sub_epi32(u[8], v[8]);
+ u[9] = _mm_sub_epi32(u[9], v[9]);
u[10] = _mm_sub_epi32(u[10], v[10]);
u[11] = _mm_sub_epi32(u[11], v[11]);
u[12] = _mm_sub_epi32(u[12], v[12]);
@@ -2507,16 +2556,16 @@
u[14] = _mm_sub_epi32(u[14], v[14]);
u[15] = _mm_sub_epi32(u[15], v[15]);
- v[ 0] = _mm_add_epi32(u[ 0], K32One);
- v[ 1] = _mm_add_epi32(u[ 1], K32One);
- v[ 2] = _mm_add_epi32(u[ 2], K32One);
- v[ 3] = _mm_add_epi32(u[ 3], K32One);
- v[ 4] = _mm_add_epi32(u[ 4], K32One);
- v[ 5] = _mm_add_epi32(u[ 5], K32One);
- v[ 6] = _mm_add_epi32(u[ 6], K32One);
- v[ 7] = _mm_add_epi32(u[ 7], K32One);
- v[ 8] = _mm_add_epi32(u[ 8], K32One);
- v[ 9] = _mm_add_epi32(u[ 9], K32One);
+ v[0] = _mm_add_epi32(u[0], K32One);
+ v[1] = _mm_add_epi32(u[1], K32One);
+ v[2] = _mm_add_epi32(u[2], K32One);
+ v[3] = _mm_add_epi32(u[3], K32One);
+ v[4] = _mm_add_epi32(u[4], K32One);
+ v[5] = _mm_add_epi32(u[5], K32One);
+ v[6] = _mm_add_epi32(u[6], K32One);
+ v[7] = _mm_add_epi32(u[7], K32One);
+ v[8] = _mm_add_epi32(u[8], K32One);
+ v[9] = _mm_add_epi32(u[9], K32One);
v[10] = _mm_add_epi32(u[10], K32One);
v[11] = _mm_add_epi32(u[11], K32One);
v[12] = _mm_add_epi32(u[12], K32One);
@@ -2524,16 +2573,16 @@
v[14] = _mm_add_epi32(u[14], K32One);
v[15] = _mm_add_epi32(u[15], K32One);
- u[ 0] = _mm_srai_epi32(v[ 0], 2);
- u[ 1] = _mm_srai_epi32(v[ 1], 2);
- u[ 2] = _mm_srai_epi32(v[ 2], 2);
- u[ 3] = _mm_srai_epi32(v[ 3], 2);
- u[ 4] = _mm_srai_epi32(v[ 4], 2);
- u[ 5] = _mm_srai_epi32(v[ 5], 2);
- u[ 6] = _mm_srai_epi32(v[ 6], 2);
- u[ 7] = _mm_srai_epi32(v[ 7], 2);
- u[ 8] = _mm_srai_epi32(v[ 8], 2);
- u[ 9] = _mm_srai_epi32(v[ 9], 2);
+ u[0] = _mm_srai_epi32(v[0], 2);
+ u[1] = _mm_srai_epi32(v[1], 2);
+ u[2] = _mm_srai_epi32(v[2], 2);
+ u[3] = _mm_srai_epi32(v[3], 2);
+ u[4] = _mm_srai_epi32(v[4], 2);
+ u[5] = _mm_srai_epi32(v[5], 2);
+ u[6] = _mm_srai_epi32(v[6], 2);
+ u[7] = _mm_srai_epi32(v[7], 2);
+ u[8] = _mm_srai_epi32(v[8], 2);
+ u[9] = _mm_srai_epi32(v[9], 2);
u[10] = _mm_srai_epi32(v[10], 2);
u[11] = _mm_srai_epi32(v[11], 2);
u[12] = _mm_srai_epi32(v[12], 2);
@@ -2541,18 +2590,18 @@
u[14] = _mm_srai_epi32(v[14], 2);
u[15] = _mm_srai_epi32(v[15], 2);
- out[ 2] = _mm_packs_epi32(u[0], u[1]);
+ out[2] = _mm_packs_epi32(u[0], u[1]);
out[18] = _mm_packs_epi32(u[2], u[3]);
out[10] = _mm_packs_epi32(u[4], u[5]);
out[26] = _mm_packs_epi32(u[6], u[7]);
- out[ 6] = _mm_packs_epi32(u[8], u[9]);
+ out[6] = _mm_packs_epi32(u[8], u[9]);
out[22] = _mm_packs_epi32(u[10], u[11]);
out[14] = _mm_packs_epi32(u[12], u[13]);
out[30] = _mm_packs_epi32(u[14], u[15]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&out[2], &out[18], &out[10],
- &out[26], &out[6], &out[22],
- &out[14], &out[30]);
+ overflow =
+ check_epi16_overflow_x8(&out[2], &out[18], &out[10], &out[26],
+ &out[6], &out[22], &out[14], &out[30]);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
@@ -2604,16 +2653,16 @@
const __m128i k32_m17_p15 = pair_set_epi32(-cospi_17_64, cospi_15_64);
const __m128i k32_m01_p31 = pair_set_epi32(-cospi_1_64, cospi_31_64);
- u[ 0] = _mm_unpacklo_epi32(lstep1[32], lstep1[62]);
- u[ 1] = _mm_unpackhi_epi32(lstep1[32], lstep1[62]);
- u[ 2] = _mm_unpacklo_epi32(lstep1[33], lstep1[63]);
- u[ 3] = _mm_unpackhi_epi32(lstep1[33], lstep1[63]);
- u[ 4] = _mm_unpacklo_epi32(lstep1[34], lstep1[60]);
- u[ 5] = _mm_unpackhi_epi32(lstep1[34], lstep1[60]);
- u[ 6] = _mm_unpacklo_epi32(lstep1[35], lstep1[61]);
- u[ 7] = _mm_unpackhi_epi32(lstep1[35], lstep1[61]);
- u[ 8] = _mm_unpacklo_epi32(lstep1[36], lstep1[58]);
- u[ 9] = _mm_unpackhi_epi32(lstep1[36], lstep1[58]);
+ u[0] = _mm_unpacklo_epi32(lstep1[32], lstep1[62]);
+ u[1] = _mm_unpackhi_epi32(lstep1[32], lstep1[62]);
+ u[2] = _mm_unpacklo_epi32(lstep1[33], lstep1[63]);
+ u[3] = _mm_unpackhi_epi32(lstep1[33], lstep1[63]);
+ u[4] = _mm_unpacklo_epi32(lstep1[34], lstep1[60]);
+ u[5] = _mm_unpackhi_epi32(lstep1[34], lstep1[60]);
+ u[6] = _mm_unpacklo_epi32(lstep1[35], lstep1[61]);
+ u[7] = _mm_unpackhi_epi32(lstep1[35], lstep1[61]);
+ u[8] = _mm_unpacklo_epi32(lstep1[36], lstep1[58]);
+ u[9] = _mm_unpackhi_epi32(lstep1[36], lstep1[58]);
u[10] = _mm_unpacklo_epi32(lstep1[37], lstep1[59]);
u[11] = _mm_unpackhi_epi32(lstep1[37], lstep1[59]);
u[12] = _mm_unpacklo_epi32(lstep1[38], lstep1[56]);
@@ -2621,16 +2670,16 @@
u[14] = _mm_unpacklo_epi32(lstep1[39], lstep1[57]);
u[15] = _mm_unpackhi_epi32(lstep1[39], lstep1[57]);
- v[ 0] = k_madd_epi32(u[ 0], k32_p31_p01);
- v[ 1] = k_madd_epi32(u[ 1], k32_p31_p01);
- v[ 2] = k_madd_epi32(u[ 2], k32_p31_p01);
- v[ 3] = k_madd_epi32(u[ 3], k32_p31_p01);
- v[ 4] = k_madd_epi32(u[ 4], k32_p15_p17);
- v[ 5] = k_madd_epi32(u[ 5], k32_p15_p17);
- v[ 6] = k_madd_epi32(u[ 6], k32_p15_p17);
- v[ 7] = k_madd_epi32(u[ 7], k32_p15_p17);
- v[ 8] = k_madd_epi32(u[ 8], k32_p23_p09);
- v[ 9] = k_madd_epi32(u[ 9], k32_p23_p09);
+ v[0] = k_madd_epi32(u[0], k32_p31_p01);
+ v[1] = k_madd_epi32(u[1], k32_p31_p01);
+ v[2] = k_madd_epi32(u[2], k32_p31_p01);
+ v[3] = k_madd_epi32(u[3], k32_p31_p01);
+ v[4] = k_madd_epi32(u[4], k32_p15_p17);
+ v[5] = k_madd_epi32(u[5], k32_p15_p17);
+ v[6] = k_madd_epi32(u[6], k32_p15_p17);
+ v[7] = k_madd_epi32(u[7], k32_p15_p17);
+ v[8] = k_madd_epi32(u[8], k32_p23_p09);
+ v[9] = k_madd_epi32(u[9], k32_p23_p09);
v[10] = k_madd_epi32(u[10], k32_p23_p09);
v[11] = k_madd_epi32(u[11], k32_p23_p09);
v[12] = k_madd_epi32(u[12], k32_p07_p25);
@@ -2641,41 +2690,40 @@
v[17] = k_madd_epi32(u[13], k32_m25_p07);
v[18] = k_madd_epi32(u[14], k32_m25_p07);
v[19] = k_madd_epi32(u[15], k32_m25_p07);
- v[20] = k_madd_epi32(u[ 8], k32_m09_p23);
- v[21] = k_madd_epi32(u[ 9], k32_m09_p23);
+ v[20] = k_madd_epi32(u[8], k32_m09_p23);
+ v[21] = k_madd_epi32(u[9], k32_m09_p23);
v[22] = k_madd_epi32(u[10], k32_m09_p23);
v[23] = k_madd_epi32(u[11], k32_m09_p23);
- v[24] = k_madd_epi32(u[ 4], k32_m17_p15);
- v[25] = k_madd_epi32(u[ 5], k32_m17_p15);
- v[26] = k_madd_epi32(u[ 6], k32_m17_p15);
- v[27] = k_madd_epi32(u[ 7], k32_m17_p15);
- v[28] = k_madd_epi32(u[ 0], k32_m01_p31);
- v[29] = k_madd_epi32(u[ 1], k32_m01_p31);
- v[30] = k_madd_epi32(u[ 2], k32_m01_p31);
- v[31] = k_madd_epi32(u[ 3], k32_m01_p31);
+ v[24] = k_madd_epi32(u[4], k32_m17_p15);
+ v[25] = k_madd_epi32(u[5], k32_m17_p15);
+ v[26] = k_madd_epi32(u[6], k32_m17_p15);
+ v[27] = k_madd_epi32(u[7], k32_m17_p15);
+ v[28] = k_madd_epi32(u[0], k32_m01_p31);
+ v[29] = k_madd_epi32(u[1], k32_m01_p31);
+ v[30] = k_madd_epi32(u[2], k32_m01_p31);
+ v[31] = k_madd_epi32(u[3], k32_m01_p31);
#if DCT_HIGH_BIT_DEPTH
overflow = k_check_epi32_overflow_32(
- &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7],
- &v[8], &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15],
- &v[16], &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23],
- &v[24], &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31],
- &kZero);
+ &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7], &v[8],
+ &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15], &v[16],
+ &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23], &v[24],
+ &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31], &kZero);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
- u[ 0] = k_packs_epi64(v[ 0], v[ 1]);
- u[ 1] = k_packs_epi64(v[ 2], v[ 3]);
- u[ 2] = k_packs_epi64(v[ 4], v[ 5]);
- u[ 3] = k_packs_epi64(v[ 6], v[ 7]);
- u[ 4] = k_packs_epi64(v[ 8], v[ 9]);
- u[ 5] = k_packs_epi64(v[10], v[11]);
- u[ 6] = k_packs_epi64(v[12], v[13]);
- u[ 7] = k_packs_epi64(v[14], v[15]);
- u[ 8] = k_packs_epi64(v[16], v[17]);
- u[ 9] = k_packs_epi64(v[18], v[19]);
+ u[0] = k_packs_epi64(v[0], v[1]);
+ u[1] = k_packs_epi64(v[2], v[3]);
+ u[2] = k_packs_epi64(v[4], v[5]);
+ u[3] = k_packs_epi64(v[6], v[7]);
+ u[4] = k_packs_epi64(v[8], v[9]);
+ u[5] = k_packs_epi64(v[10], v[11]);
+ u[6] = k_packs_epi64(v[12], v[13]);
+ u[7] = k_packs_epi64(v[14], v[15]);
+ u[8] = k_packs_epi64(v[16], v[17]);
+ u[9] = k_packs_epi64(v[18], v[19]);
u[10] = k_packs_epi64(v[20], v[21]);
u[11] = k_packs_epi64(v[22], v[23]);
u[12] = k_packs_epi64(v[24], v[25]);
@@ -2683,16 +2731,16 @@
u[14] = k_packs_epi64(v[28], v[29]);
u[15] = k_packs_epi64(v[30], v[31]);
- v[ 0] = _mm_add_epi32(u[ 0], k__DCT_CONST_ROUNDING);
- v[ 1] = _mm_add_epi32(u[ 1], k__DCT_CONST_ROUNDING);
- v[ 2] = _mm_add_epi32(u[ 2], k__DCT_CONST_ROUNDING);
- v[ 3] = _mm_add_epi32(u[ 3], k__DCT_CONST_ROUNDING);
- v[ 4] = _mm_add_epi32(u[ 4], k__DCT_CONST_ROUNDING);
- v[ 5] = _mm_add_epi32(u[ 5], k__DCT_CONST_ROUNDING);
- v[ 6] = _mm_add_epi32(u[ 6], k__DCT_CONST_ROUNDING);
- v[ 7] = _mm_add_epi32(u[ 7], k__DCT_CONST_ROUNDING);
- v[ 8] = _mm_add_epi32(u[ 8], k__DCT_CONST_ROUNDING);
- v[ 9] = _mm_add_epi32(u[ 9], k__DCT_CONST_ROUNDING);
+ v[0] = _mm_add_epi32(u[0], k__DCT_CONST_ROUNDING);
+ v[1] = _mm_add_epi32(u[1], k__DCT_CONST_ROUNDING);
+ v[2] = _mm_add_epi32(u[2], k__DCT_CONST_ROUNDING);
+ v[3] = _mm_add_epi32(u[3], k__DCT_CONST_ROUNDING);
+ v[4] = _mm_add_epi32(u[4], k__DCT_CONST_ROUNDING);
+ v[5] = _mm_add_epi32(u[5], k__DCT_CONST_ROUNDING);
+ v[6] = _mm_add_epi32(u[6], k__DCT_CONST_ROUNDING);
+ v[7] = _mm_add_epi32(u[7], k__DCT_CONST_ROUNDING);
+ v[8] = _mm_add_epi32(u[8], k__DCT_CONST_ROUNDING);
+ v[9] = _mm_add_epi32(u[9], k__DCT_CONST_ROUNDING);
v[10] = _mm_add_epi32(u[10], k__DCT_CONST_ROUNDING);
v[11] = _mm_add_epi32(u[11], k__DCT_CONST_ROUNDING);
v[12] = _mm_add_epi32(u[12], k__DCT_CONST_ROUNDING);
@@ -2700,16 +2748,16 @@
v[14] = _mm_add_epi32(u[14], k__DCT_CONST_ROUNDING);
v[15] = _mm_add_epi32(u[15], k__DCT_CONST_ROUNDING);
- u[ 0] = _mm_srai_epi32(v[ 0], DCT_CONST_BITS);
- u[ 1] = _mm_srai_epi32(v[ 1], DCT_CONST_BITS);
- u[ 2] = _mm_srai_epi32(v[ 2], DCT_CONST_BITS);
- u[ 3] = _mm_srai_epi32(v[ 3], DCT_CONST_BITS);
- u[ 4] = _mm_srai_epi32(v[ 4], DCT_CONST_BITS);
- u[ 5] = _mm_srai_epi32(v[ 5], DCT_CONST_BITS);
- u[ 6] = _mm_srai_epi32(v[ 6], DCT_CONST_BITS);
- u[ 7] = _mm_srai_epi32(v[ 7], DCT_CONST_BITS);
- u[ 8] = _mm_srai_epi32(v[ 8], DCT_CONST_BITS);
- u[ 9] = _mm_srai_epi32(v[ 9], DCT_CONST_BITS);
+ u[0] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
+ u[1] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
+ u[2] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
+ u[3] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
+ u[4] = _mm_srai_epi32(v[4], DCT_CONST_BITS);
+ u[5] = _mm_srai_epi32(v[5], DCT_CONST_BITS);
+ u[6] = _mm_srai_epi32(v[6], DCT_CONST_BITS);
+ u[7] = _mm_srai_epi32(v[7], DCT_CONST_BITS);
+ u[8] = _mm_srai_epi32(v[8], DCT_CONST_BITS);
+ u[9] = _mm_srai_epi32(v[9], DCT_CONST_BITS);
u[10] = _mm_srai_epi32(v[10], DCT_CONST_BITS);
u[11] = _mm_srai_epi32(v[11], DCT_CONST_BITS);
u[12] = _mm_srai_epi32(v[12], DCT_CONST_BITS);
@@ -2717,16 +2765,16 @@
u[14] = _mm_srai_epi32(v[14], DCT_CONST_BITS);
u[15] = _mm_srai_epi32(v[15], DCT_CONST_BITS);
- v[ 0] = _mm_cmplt_epi32(u[ 0], kZero);
- v[ 1] = _mm_cmplt_epi32(u[ 1], kZero);
- v[ 2] = _mm_cmplt_epi32(u[ 2], kZero);
- v[ 3] = _mm_cmplt_epi32(u[ 3], kZero);
- v[ 4] = _mm_cmplt_epi32(u[ 4], kZero);
- v[ 5] = _mm_cmplt_epi32(u[ 5], kZero);
- v[ 6] = _mm_cmplt_epi32(u[ 6], kZero);
- v[ 7] = _mm_cmplt_epi32(u[ 7], kZero);
- v[ 8] = _mm_cmplt_epi32(u[ 8], kZero);
- v[ 9] = _mm_cmplt_epi32(u[ 9], kZero);
+ v[0] = _mm_cmplt_epi32(u[0], kZero);
+ v[1] = _mm_cmplt_epi32(u[1], kZero);
+ v[2] = _mm_cmplt_epi32(u[2], kZero);
+ v[3] = _mm_cmplt_epi32(u[3], kZero);
+ v[4] = _mm_cmplt_epi32(u[4], kZero);
+ v[5] = _mm_cmplt_epi32(u[5], kZero);
+ v[6] = _mm_cmplt_epi32(u[6], kZero);
+ v[7] = _mm_cmplt_epi32(u[7], kZero);
+ v[8] = _mm_cmplt_epi32(u[8], kZero);
+ v[9] = _mm_cmplt_epi32(u[9], kZero);
v[10] = _mm_cmplt_epi32(u[10], kZero);
v[11] = _mm_cmplt_epi32(u[11], kZero);
v[12] = _mm_cmplt_epi32(u[12], kZero);
@@ -2734,16 +2782,16 @@
v[14] = _mm_cmplt_epi32(u[14], kZero);
v[15] = _mm_cmplt_epi32(u[15], kZero);
- u[ 0] = _mm_sub_epi32(u[ 0], v[ 0]);
- u[ 1] = _mm_sub_epi32(u[ 1], v[ 1]);
- u[ 2] = _mm_sub_epi32(u[ 2], v[ 2]);
- u[ 3] = _mm_sub_epi32(u[ 3], v[ 3]);
- u[ 4] = _mm_sub_epi32(u[ 4], v[ 4]);
- u[ 5] = _mm_sub_epi32(u[ 5], v[ 5]);
- u[ 6] = _mm_sub_epi32(u[ 6], v[ 6]);
- u[ 7] = _mm_sub_epi32(u[ 7], v[ 7]);
- u[ 8] = _mm_sub_epi32(u[ 8], v[ 8]);
- u[ 9] = _mm_sub_epi32(u[ 9], v[ 9]);
+ u[0] = _mm_sub_epi32(u[0], v[0]);
+ u[1] = _mm_sub_epi32(u[1], v[1]);
+ u[2] = _mm_sub_epi32(u[2], v[2]);
+ u[3] = _mm_sub_epi32(u[3], v[3]);
+ u[4] = _mm_sub_epi32(u[4], v[4]);
+ u[5] = _mm_sub_epi32(u[5], v[5]);
+ u[6] = _mm_sub_epi32(u[6], v[6]);
+ u[7] = _mm_sub_epi32(u[7], v[7]);
+ u[8] = _mm_sub_epi32(u[8], v[8]);
+ u[9] = _mm_sub_epi32(u[9], v[9]);
u[10] = _mm_sub_epi32(u[10], v[10]);
u[11] = _mm_sub_epi32(u[11], v[11]);
u[12] = _mm_sub_epi32(u[12], v[12]);
@@ -2785,18 +2833,18 @@
u[14] = _mm_srai_epi32(v[14], 2);
u[15] = _mm_srai_epi32(v[15], 2);
- out[ 1] = _mm_packs_epi32(u[0], u[1]);
+ out[1] = _mm_packs_epi32(u[0], u[1]);
out[17] = _mm_packs_epi32(u[2], u[3]);
- out[ 9] = _mm_packs_epi32(u[4], u[5]);
+ out[9] = _mm_packs_epi32(u[4], u[5]);
out[25] = _mm_packs_epi32(u[6], u[7]);
- out[ 7] = _mm_packs_epi32(u[8], u[9]);
+ out[7] = _mm_packs_epi32(u[8], u[9]);
out[23] = _mm_packs_epi32(u[10], u[11]);
out[15] = _mm_packs_epi32(u[12], u[13]);
out[31] = _mm_packs_epi32(u[14], u[15]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&out[1], &out[17], &out[9],
- &out[25], &out[7], &out[23],
- &out[15], &out[31]);
+ overflow =
+ check_epi16_overflow_x8(&out[1], &out[17], &out[9], &out[25],
+ &out[7], &out[23], &out[15], &out[31]);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
@@ -2813,16 +2861,16 @@
const __m128i k32_m21_p11 = pair_set_epi32(-cospi_21_64, cospi_11_64);
const __m128i k32_m05_p27 = pair_set_epi32(-cospi_5_64, cospi_27_64);
- u[ 0] = _mm_unpacklo_epi32(lstep1[40], lstep1[54]);
- u[ 1] = _mm_unpackhi_epi32(lstep1[40], lstep1[54]);
- u[ 2] = _mm_unpacklo_epi32(lstep1[41], lstep1[55]);
- u[ 3] = _mm_unpackhi_epi32(lstep1[41], lstep1[55]);
- u[ 4] = _mm_unpacklo_epi32(lstep1[42], lstep1[52]);
- u[ 5] = _mm_unpackhi_epi32(lstep1[42], lstep1[52]);
- u[ 6] = _mm_unpacklo_epi32(lstep1[43], lstep1[53]);
- u[ 7] = _mm_unpackhi_epi32(lstep1[43], lstep1[53]);
- u[ 8] = _mm_unpacklo_epi32(lstep1[44], lstep1[50]);
- u[ 9] = _mm_unpackhi_epi32(lstep1[44], lstep1[50]);
+ u[0] = _mm_unpacklo_epi32(lstep1[40], lstep1[54]);
+ u[1] = _mm_unpackhi_epi32(lstep1[40], lstep1[54]);
+ u[2] = _mm_unpacklo_epi32(lstep1[41], lstep1[55]);
+ u[3] = _mm_unpackhi_epi32(lstep1[41], lstep1[55]);
+ u[4] = _mm_unpacklo_epi32(lstep1[42], lstep1[52]);
+ u[5] = _mm_unpackhi_epi32(lstep1[42], lstep1[52]);
+ u[6] = _mm_unpacklo_epi32(lstep1[43], lstep1[53]);
+ u[7] = _mm_unpackhi_epi32(lstep1[43], lstep1[53]);
+ u[8] = _mm_unpacklo_epi32(lstep1[44], lstep1[50]);
+ u[9] = _mm_unpackhi_epi32(lstep1[44], lstep1[50]);
u[10] = _mm_unpacklo_epi32(lstep1[45], lstep1[51]);
u[11] = _mm_unpackhi_epi32(lstep1[45], lstep1[51]);
u[12] = _mm_unpacklo_epi32(lstep1[46], lstep1[48]);
@@ -2830,16 +2878,16 @@
u[14] = _mm_unpacklo_epi32(lstep1[47], lstep1[49]);
u[15] = _mm_unpackhi_epi32(lstep1[47], lstep1[49]);
- v[ 0] = k_madd_epi32(u[ 0], k32_p27_p05);
- v[ 1] = k_madd_epi32(u[ 1], k32_p27_p05);
- v[ 2] = k_madd_epi32(u[ 2], k32_p27_p05);
- v[ 3] = k_madd_epi32(u[ 3], k32_p27_p05);
- v[ 4] = k_madd_epi32(u[ 4], k32_p11_p21);
- v[ 5] = k_madd_epi32(u[ 5], k32_p11_p21);
- v[ 6] = k_madd_epi32(u[ 6], k32_p11_p21);
- v[ 7] = k_madd_epi32(u[ 7], k32_p11_p21);
- v[ 8] = k_madd_epi32(u[ 8], k32_p19_p13);
- v[ 9] = k_madd_epi32(u[ 9], k32_p19_p13);
+ v[0] = k_madd_epi32(u[0], k32_p27_p05);
+ v[1] = k_madd_epi32(u[1], k32_p27_p05);
+ v[2] = k_madd_epi32(u[2], k32_p27_p05);
+ v[3] = k_madd_epi32(u[3], k32_p27_p05);
+ v[4] = k_madd_epi32(u[4], k32_p11_p21);
+ v[5] = k_madd_epi32(u[5], k32_p11_p21);
+ v[6] = k_madd_epi32(u[6], k32_p11_p21);
+ v[7] = k_madd_epi32(u[7], k32_p11_p21);
+ v[8] = k_madd_epi32(u[8], k32_p19_p13);
+ v[9] = k_madd_epi32(u[9], k32_p19_p13);
v[10] = k_madd_epi32(u[10], k32_p19_p13);
v[11] = k_madd_epi32(u[11], k32_p19_p13);
v[12] = k_madd_epi32(u[12], k32_p03_p29);
@@ -2850,41 +2898,40 @@
v[17] = k_madd_epi32(u[13], k32_m29_p03);
v[18] = k_madd_epi32(u[14], k32_m29_p03);
v[19] = k_madd_epi32(u[15], k32_m29_p03);
- v[20] = k_madd_epi32(u[ 8], k32_m13_p19);
- v[21] = k_madd_epi32(u[ 9], k32_m13_p19);
+ v[20] = k_madd_epi32(u[8], k32_m13_p19);
+ v[21] = k_madd_epi32(u[9], k32_m13_p19);
v[22] = k_madd_epi32(u[10], k32_m13_p19);
v[23] = k_madd_epi32(u[11], k32_m13_p19);
- v[24] = k_madd_epi32(u[ 4], k32_m21_p11);
- v[25] = k_madd_epi32(u[ 5], k32_m21_p11);
- v[26] = k_madd_epi32(u[ 6], k32_m21_p11);
- v[27] = k_madd_epi32(u[ 7], k32_m21_p11);
- v[28] = k_madd_epi32(u[ 0], k32_m05_p27);
- v[29] = k_madd_epi32(u[ 1], k32_m05_p27);
- v[30] = k_madd_epi32(u[ 2], k32_m05_p27);
- v[31] = k_madd_epi32(u[ 3], k32_m05_p27);
+ v[24] = k_madd_epi32(u[4], k32_m21_p11);
+ v[25] = k_madd_epi32(u[5], k32_m21_p11);
+ v[26] = k_madd_epi32(u[6], k32_m21_p11);
+ v[27] = k_madd_epi32(u[7], k32_m21_p11);
+ v[28] = k_madd_epi32(u[0], k32_m05_p27);
+ v[29] = k_madd_epi32(u[1], k32_m05_p27);
+ v[30] = k_madd_epi32(u[2], k32_m05_p27);
+ v[31] = k_madd_epi32(u[3], k32_m05_p27);
#if DCT_HIGH_BIT_DEPTH
overflow = k_check_epi32_overflow_32(
- &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7],
- &v[8], &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15],
- &v[16], &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23],
- &v[24], &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31],
- &kZero);
+ &v[0], &v[1], &v[2], &v[3], &v[4], &v[5], &v[6], &v[7], &v[8],
+ &v[9], &v[10], &v[11], &v[12], &v[13], &v[14], &v[15], &v[16],
+ &v[17], &v[18], &v[19], &v[20], &v[21], &v[22], &v[23], &v[24],
+ &v[25], &v[26], &v[27], &v[28], &v[29], &v[30], &v[31], &kZero);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
- u[ 0] = k_packs_epi64(v[ 0], v[ 1]);
- u[ 1] = k_packs_epi64(v[ 2], v[ 3]);
- u[ 2] = k_packs_epi64(v[ 4], v[ 5]);
- u[ 3] = k_packs_epi64(v[ 6], v[ 7]);
- u[ 4] = k_packs_epi64(v[ 8], v[ 9]);
- u[ 5] = k_packs_epi64(v[10], v[11]);
- u[ 6] = k_packs_epi64(v[12], v[13]);
- u[ 7] = k_packs_epi64(v[14], v[15]);
- u[ 8] = k_packs_epi64(v[16], v[17]);
- u[ 9] = k_packs_epi64(v[18], v[19]);
+ u[0] = k_packs_epi64(v[0], v[1]);
+ u[1] = k_packs_epi64(v[2], v[3]);
+ u[2] = k_packs_epi64(v[4], v[5]);
+ u[3] = k_packs_epi64(v[6], v[7]);
+ u[4] = k_packs_epi64(v[8], v[9]);
+ u[5] = k_packs_epi64(v[10], v[11]);
+ u[6] = k_packs_epi64(v[12], v[13]);
+ u[7] = k_packs_epi64(v[14], v[15]);
+ u[8] = k_packs_epi64(v[16], v[17]);
+ u[9] = k_packs_epi64(v[18], v[19]);
u[10] = k_packs_epi64(v[20], v[21]);
u[11] = k_packs_epi64(v[22], v[23]);
u[12] = k_packs_epi64(v[24], v[25]);
@@ -2892,16 +2939,16 @@
u[14] = k_packs_epi64(v[28], v[29]);
u[15] = k_packs_epi64(v[30], v[31]);
- v[ 0] = _mm_add_epi32(u[ 0], k__DCT_CONST_ROUNDING);
- v[ 1] = _mm_add_epi32(u[ 1], k__DCT_CONST_ROUNDING);
- v[ 2] = _mm_add_epi32(u[ 2], k__DCT_CONST_ROUNDING);
- v[ 3] = _mm_add_epi32(u[ 3], k__DCT_CONST_ROUNDING);
- v[ 4] = _mm_add_epi32(u[ 4], k__DCT_CONST_ROUNDING);
- v[ 5] = _mm_add_epi32(u[ 5], k__DCT_CONST_ROUNDING);
- v[ 6] = _mm_add_epi32(u[ 6], k__DCT_CONST_ROUNDING);
- v[ 7] = _mm_add_epi32(u[ 7], k__DCT_CONST_ROUNDING);
- v[ 8] = _mm_add_epi32(u[ 8], k__DCT_CONST_ROUNDING);
- v[ 9] = _mm_add_epi32(u[ 9], k__DCT_CONST_ROUNDING);
+ v[0] = _mm_add_epi32(u[0], k__DCT_CONST_ROUNDING);
+ v[1] = _mm_add_epi32(u[1], k__DCT_CONST_ROUNDING);
+ v[2] = _mm_add_epi32(u[2], k__DCT_CONST_ROUNDING);
+ v[3] = _mm_add_epi32(u[3], k__DCT_CONST_ROUNDING);
+ v[4] = _mm_add_epi32(u[4], k__DCT_CONST_ROUNDING);
+ v[5] = _mm_add_epi32(u[5], k__DCT_CONST_ROUNDING);
+ v[6] = _mm_add_epi32(u[6], k__DCT_CONST_ROUNDING);
+ v[7] = _mm_add_epi32(u[7], k__DCT_CONST_ROUNDING);
+ v[8] = _mm_add_epi32(u[8], k__DCT_CONST_ROUNDING);
+ v[9] = _mm_add_epi32(u[9], k__DCT_CONST_ROUNDING);
v[10] = _mm_add_epi32(u[10], k__DCT_CONST_ROUNDING);
v[11] = _mm_add_epi32(u[11], k__DCT_CONST_ROUNDING);
v[12] = _mm_add_epi32(u[12], k__DCT_CONST_ROUNDING);
@@ -2909,16 +2956,16 @@
v[14] = _mm_add_epi32(u[14], k__DCT_CONST_ROUNDING);
v[15] = _mm_add_epi32(u[15], k__DCT_CONST_ROUNDING);
- u[ 0] = _mm_srai_epi32(v[ 0], DCT_CONST_BITS);
- u[ 1] = _mm_srai_epi32(v[ 1], DCT_CONST_BITS);
- u[ 2] = _mm_srai_epi32(v[ 2], DCT_CONST_BITS);
- u[ 3] = _mm_srai_epi32(v[ 3], DCT_CONST_BITS);
- u[ 4] = _mm_srai_epi32(v[ 4], DCT_CONST_BITS);
- u[ 5] = _mm_srai_epi32(v[ 5], DCT_CONST_BITS);
- u[ 6] = _mm_srai_epi32(v[ 6], DCT_CONST_BITS);
- u[ 7] = _mm_srai_epi32(v[ 7], DCT_CONST_BITS);
- u[ 8] = _mm_srai_epi32(v[ 8], DCT_CONST_BITS);
- u[ 9] = _mm_srai_epi32(v[ 9], DCT_CONST_BITS);
+ u[0] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
+ u[1] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
+ u[2] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
+ u[3] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
+ u[4] = _mm_srai_epi32(v[4], DCT_CONST_BITS);
+ u[5] = _mm_srai_epi32(v[5], DCT_CONST_BITS);
+ u[6] = _mm_srai_epi32(v[6], DCT_CONST_BITS);
+ u[7] = _mm_srai_epi32(v[7], DCT_CONST_BITS);
+ u[8] = _mm_srai_epi32(v[8], DCT_CONST_BITS);
+ u[9] = _mm_srai_epi32(v[9], DCT_CONST_BITS);
u[10] = _mm_srai_epi32(v[10], DCT_CONST_BITS);
u[11] = _mm_srai_epi32(v[11], DCT_CONST_BITS);
u[12] = _mm_srai_epi32(v[12], DCT_CONST_BITS);
@@ -2926,16 +2973,16 @@
u[14] = _mm_srai_epi32(v[14], DCT_CONST_BITS);
u[15] = _mm_srai_epi32(v[15], DCT_CONST_BITS);
- v[ 0] = _mm_cmplt_epi32(u[ 0], kZero);
- v[ 1] = _mm_cmplt_epi32(u[ 1], kZero);
- v[ 2] = _mm_cmplt_epi32(u[ 2], kZero);
- v[ 3] = _mm_cmplt_epi32(u[ 3], kZero);
- v[ 4] = _mm_cmplt_epi32(u[ 4], kZero);
- v[ 5] = _mm_cmplt_epi32(u[ 5], kZero);
- v[ 6] = _mm_cmplt_epi32(u[ 6], kZero);
- v[ 7] = _mm_cmplt_epi32(u[ 7], kZero);
- v[ 8] = _mm_cmplt_epi32(u[ 8], kZero);
- v[ 9] = _mm_cmplt_epi32(u[ 9], kZero);
+ v[0] = _mm_cmplt_epi32(u[0], kZero);
+ v[1] = _mm_cmplt_epi32(u[1], kZero);
+ v[2] = _mm_cmplt_epi32(u[2], kZero);
+ v[3] = _mm_cmplt_epi32(u[3], kZero);
+ v[4] = _mm_cmplt_epi32(u[4], kZero);
+ v[5] = _mm_cmplt_epi32(u[5], kZero);
+ v[6] = _mm_cmplt_epi32(u[6], kZero);
+ v[7] = _mm_cmplt_epi32(u[7], kZero);
+ v[8] = _mm_cmplt_epi32(u[8], kZero);
+ v[9] = _mm_cmplt_epi32(u[9], kZero);
v[10] = _mm_cmplt_epi32(u[10], kZero);
v[11] = _mm_cmplt_epi32(u[11], kZero);
v[12] = _mm_cmplt_epi32(u[12], kZero);
@@ -2943,16 +2990,16 @@
v[14] = _mm_cmplt_epi32(u[14], kZero);
v[15] = _mm_cmplt_epi32(u[15], kZero);
- u[ 0] = _mm_sub_epi32(u[ 0], v[ 0]);
- u[ 1] = _mm_sub_epi32(u[ 1], v[ 1]);
- u[ 2] = _mm_sub_epi32(u[ 2], v[ 2]);
- u[ 3] = _mm_sub_epi32(u[ 3], v[ 3]);
- u[ 4] = _mm_sub_epi32(u[ 4], v[ 4]);
- u[ 5] = _mm_sub_epi32(u[ 5], v[ 5]);
- u[ 6] = _mm_sub_epi32(u[ 6], v[ 6]);
- u[ 7] = _mm_sub_epi32(u[ 7], v[ 7]);
- u[ 8] = _mm_sub_epi32(u[ 8], v[ 8]);
- u[ 9] = _mm_sub_epi32(u[ 9], v[ 9]);
+ u[0] = _mm_sub_epi32(u[0], v[0]);
+ u[1] = _mm_sub_epi32(u[1], v[1]);
+ u[2] = _mm_sub_epi32(u[2], v[2]);
+ u[3] = _mm_sub_epi32(u[3], v[3]);
+ u[4] = _mm_sub_epi32(u[4], v[4]);
+ u[5] = _mm_sub_epi32(u[5], v[5]);
+ u[6] = _mm_sub_epi32(u[6], v[6]);
+ u[7] = _mm_sub_epi32(u[7], v[7]);
+ u[8] = _mm_sub_epi32(u[8], v[8]);
+ u[9] = _mm_sub_epi32(u[9], v[9]);
u[10] = _mm_sub_epi32(u[10], v[10]);
u[11] = _mm_sub_epi32(u[11], v[11]);
u[12] = _mm_sub_epi32(u[12], v[12]);
@@ -2994,18 +3041,18 @@
u[14] = _mm_srai_epi32(v[14], 2);
u[15] = _mm_srai_epi32(v[15], 2);
- out[ 5] = _mm_packs_epi32(u[0], u[1]);
+ out[5] = _mm_packs_epi32(u[0], u[1]);
out[21] = _mm_packs_epi32(u[2], u[3]);
out[13] = _mm_packs_epi32(u[4], u[5]);
out[29] = _mm_packs_epi32(u[6], u[7]);
- out[ 3] = _mm_packs_epi32(u[8], u[9]);
+ out[3] = _mm_packs_epi32(u[8], u[9]);
out[19] = _mm_packs_epi32(u[10], u[11]);
out[11] = _mm_packs_epi32(u[12], u[13]);
out[27] = _mm_packs_epi32(u[14], u[15]);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&out[5], &out[21], &out[13],
- &out[29], &out[3], &out[19],
- &out[11], &out[27]);
+ overflow =
+ check_epi16_overflow_x8(&out[5], &out[21], &out[13], &out[29],
+ &out[3], &out[19], &out[11], &out[27]);
if (overflow) {
HIGH_FDCT32x32_2D_ROWS_C(intermediate, output_org);
return;
diff --git a/vp10/common/x86/vp10_fwd_txfm_impl_sse2.h b/vp10/common/x86/vp10_fwd_txfm_impl_sse2.h
index 69889e2..b3bcc4a 100644
--- a/vp10/common/x86/vp10_fwd_txfm_impl_sse2.h
+++ b/vp10/common/x86/vp10_fwd_txfm_impl_sse2.h
@@ -43,44 +43,36 @@
// These are the coefficients used for the multiplies.
// In the comments, pN means cos(N pi /64) and mN is -cos(N pi /64),
// where cospi_N_64 = cos(N pi /64)
- const __m128i k__cospi_A = octa_set_epi16(cospi_16_64, cospi_16_64,
- cospi_16_64, cospi_16_64,
- cospi_16_64, -cospi_16_64,
- cospi_16_64, -cospi_16_64);
- const __m128i k__cospi_B = octa_set_epi16(cospi_16_64, -cospi_16_64,
- cospi_16_64, -cospi_16_64,
- cospi_16_64, cospi_16_64,
- cospi_16_64, cospi_16_64);
- const __m128i k__cospi_C = octa_set_epi16(cospi_8_64, cospi_24_64,
- cospi_8_64, cospi_24_64,
- cospi_24_64, -cospi_8_64,
- cospi_24_64, -cospi_8_64);
- const __m128i k__cospi_D = octa_set_epi16(cospi_24_64, -cospi_8_64,
- cospi_24_64, -cospi_8_64,
- cospi_8_64, cospi_24_64,
- cospi_8_64, cospi_24_64);
- const __m128i k__cospi_E = octa_set_epi16(cospi_16_64, cospi_16_64,
- cospi_16_64, cospi_16_64,
- cospi_16_64, cospi_16_64,
- cospi_16_64, cospi_16_64);
- const __m128i k__cospi_F = octa_set_epi16(cospi_16_64, -cospi_16_64,
- cospi_16_64, -cospi_16_64,
- cospi_16_64, -cospi_16_64,
- cospi_16_64, -cospi_16_64);
- const __m128i k__cospi_G = octa_set_epi16(cospi_8_64, cospi_24_64,
- cospi_8_64, cospi_24_64,
- -cospi_8_64, -cospi_24_64,
- -cospi_8_64, -cospi_24_64);
- const __m128i k__cospi_H = octa_set_epi16(cospi_24_64, -cospi_8_64,
- cospi_24_64, -cospi_8_64,
- -cospi_24_64, cospi_8_64,
- -cospi_24_64, cospi_8_64);
+ const __m128i k__cospi_A =
+ octa_set_epi16(cospi_16_64, cospi_16_64, cospi_16_64, cospi_16_64,
+ cospi_16_64, -cospi_16_64, cospi_16_64, -cospi_16_64);
+ const __m128i k__cospi_B =
+ octa_set_epi16(cospi_16_64, -cospi_16_64, cospi_16_64, -cospi_16_64,
+ cospi_16_64, cospi_16_64, cospi_16_64, cospi_16_64);
+ const __m128i k__cospi_C =
+ octa_set_epi16(cospi_8_64, cospi_24_64, cospi_8_64, cospi_24_64,
+ cospi_24_64, -cospi_8_64, cospi_24_64, -cospi_8_64);
+ const __m128i k__cospi_D =
+ octa_set_epi16(cospi_24_64, -cospi_8_64, cospi_24_64, -cospi_8_64,
+ cospi_8_64, cospi_24_64, cospi_8_64, cospi_24_64);
+ const __m128i k__cospi_E =
+ octa_set_epi16(cospi_16_64, cospi_16_64, cospi_16_64, cospi_16_64,
+ cospi_16_64, cospi_16_64, cospi_16_64, cospi_16_64);
+ const __m128i k__cospi_F =
+ octa_set_epi16(cospi_16_64, -cospi_16_64, cospi_16_64, -cospi_16_64,
+ cospi_16_64, -cospi_16_64, cospi_16_64, -cospi_16_64);
+ const __m128i k__cospi_G =
+ octa_set_epi16(cospi_8_64, cospi_24_64, cospi_8_64, cospi_24_64,
+ -cospi_8_64, -cospi_24_64, -cospi_8_64, -cospi_24_64);
+ const __m128i k__cospi_H =
+ octa_set_epi16(cospi_24_64, -cospi_8_64, cospi_24_64, -cospi_8_64,
+ -cospi_24_64, cospi_8_64, -cospi_24_64, cospi_8_64);
const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
// This second rounding constant saves doing some extra adds at the end
- const __m128i k__DCT_CONST_ROUNDING2 = _mm_set1_epi32(DCT_CONST_ROUNDING
- +(DCT_CONST_ROUNDING << 1));
- const int DCT_CONST_BITS2 = DCT_CONST_BITS + 2;
+ const __m128i k__DCT_CONST_ROUNDING2 =
+ _mm_set1_epi32(DCT_CONST_ROUNDING + (DCT_CONST_ROUNDING << 1));
+ const int DCT_CONST_BITS2 = DCT_CONST_BITS + 2;
const __m128i k__nonzero_bias_a = _mm_setr_epi16(0, 1, 1, 1, 1, 1, 1, 1);
const __m128i k__nonzero_bias_b = _mm_setr_epi16(1, 0, 0, 0, 0, 0, 0, 0);
__m128i in0, in1;
@@ -90,14 +82,14 @@
#endif
// Load inputs.
- in0 = _mm_loadl_epi64((const __m128i *)(input + 0 * stride));
- in1 = _mm_loadl_epi64((const __m128i *)(input + 1 * stride));
- in1 = _mm_unpacklo_epi64(in1, _mm_loadl_epi64((const __m128i *)
- (input + 2 * stride)));
- in0 = _mm_unpacklo_epi64(in0, _mm_loadl_epi64((const __m128i *)
- (input + 3 * stride)));
- // in0 = [i0 i1 i2 i3 iC iD iE iF]
- // in1 = [i4 i5 i6 i7 i8 i9 iA iB]
+ in0 = _mm_loadl_epi64((const __m128i *)(input + 0 * stride));
+ in1 = _mm_loadl_epi64((const __m128i *)(input + 1 * stride));
+ in1 = _mm_unpacklo_epi64(
+ in1, _mm_loadl_epi64((const __m128i *)(input + 2 * stride)));
+ in0 = _mm_unpacklo_epi64(
+ in0, _mm_loadl_epi64((const __m128i *)(input + 3 * stride)));
+// in0 = [i0 i1 i2 i3 iC iD iE iF]
+// in1 = [i4 i5 i6 i7 i8 i9 iA iB]
#if DCT_HIGH_BIT_DEPTH
// Check inputs small enough to use optimised code
cmp0 = _mm_xor_si128(_mm_cmpgt_epi16(in0, _mm_set1_epi16(0x3ff)),
@@ -192,10 +184,10 @@
// vertical DCTs finished. Now we do the horizontal DCTs.
// Stage 3: Add/subtract
- const __m128i t0 = ADD_EPI16(in0, in1);
- const __m128i t1 = SUB_EPI16(in0, in1);
// t0 = [c0 c1 c8 c9 c4 c5 cC cD]
// t1 = [c3 c2 cB cA -c7 -c6 -cF -cE]
+ const __m128i t0 = ADD_EPI16(in0, in1);
+ const __m128i t1 = SUB_EPI16(in0, in1);
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x2(&t0, &t1);
if (overflow) {
@@ -263,7 +255,6 @@
storeu_output(&in1, output + 2 * 4);
}
-
void FDCT8x8_2D(const int16_t *input, tran_low_t *output, int stride) {
int pass;
// Constants
@@ -283,14 +274,14 @@
int overflow;
#endif
// Load input
- __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
- __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
- __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
- __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
- __m128i in4 = _mm_load_si128((const __m128i *)(input + 4 * stride));
- __m128i in5 = _mm_load_si128((const __m128i *)(input + 5 * stride));
- __m128i in6 = _mm_load_si128((const __m128i *)(input + 6 * stride));
- __m128i in7 = _mm_load_si128((const __m128i *)(input + 7 * stride));
+ __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
+ __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
+ __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
+ __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
+ __m128i in4 = _mm_load_si128((const __m128i *)(input + 4 * stride));
+ __m128i in5 = _mm_load_si128((const __m128i *)(input + 5 * stride));
+ __m128i in6 = _mm_load_si128((const __m128i *)(input + 6 * stride));
+ __m128i in7 = _mm_load_si128((const __m128i *)(input + 7 * stride));
// Pre-condition input (shift by two)
in0 = _mm_slli_epi16(in0, 2);
in1 = _mm_slli_epi16(in1, 2);
@@ -319,8 +310,8 @@
const __m128i q7 = SUB_EPI16(in0, in7);
#if DCT_HIGH_BIT_DEPTH
if (pass == 1) {
- overflow = check_epi16_overflow_x8(&q0, &q1, &q2, &q3,
- &q4, &q5, &q6, &q7);
+ overflow =
+ check_epi16_overflow_x8(&q0, &q1, &q2, &q3, &q4, &q5, &q6, &q7);
if (overflow) {
vpx_highbd_fdct8x8_c(input, output, stride);
return;
@@ -630,22 +621,22 @@
__m128i res08, res09, res10, res11, res12, res13, res14, res15;
// Load and pre-condition input.
if (0 == pass) {
- in00 = _mm_load_si128((const __m128i *)(in + 0 * stride));
- in01 = _mm_load_si128((const __m128i *)(in + 1 * stride));
- in02 = _mm_load_si128((const __m128i *)(in + 2 * stride));
- in03 = _mm_load_si128((const __m128i *)(in + 3 * stride));
- in04 = _mm_load_si128((const __m128i *)(in + 4 * stride));
- in05 = _mm_load_si128((const __m128i *)(in + 5 * stride));
- in06 = _mm_load_si128((const __m128i *)(in + 6 * stride));
- in07 = _mm_load_si128((const __m128i *)(in + 7 * stride));
- in08 = _mm_load_si128((const __m128i *)(in + 8 * stride));
- in09 = _mm_load_si128((const __m128i *)(in + 9 * stride));
- in10 = _mm_load_si128((const __m128i *)(in + 10 * stride));
- in11 = _mm_load_si128((const __m128i *)(in + 11 * stride));
- in12 = _mm_load_si128((const __m128i *)(in + 12 * stride));
- in13 = _mm_load_si128((const __m128i *)(in + 13 * stride));
- in14 = _mm_load_si128((const __m128i *)(in + 14 * stride));
- in15 = _mm_load_si128((const __m128i *)(in + 15 * stride));
+ in00 = _mm_load_si128((const __m128i *)(in + 0 * stride));
+ in01 = _mm_load_si128((const __m128i *)(in + 1 * stride));
+ in02 = _mm_load_si128((const __m128i *)(in + 2 * stride));
+ in03 = _mm_load_si128((const __m128i *)(in + 3 * stride));
+ in04 = _mm_load_si128((const __m128i *)(in + 4 * stride));
+ in05 = _mm_load_si128((const __m128i *)(in + 5 * stride));
+ in06 = _mm_load_si128((const __m128i *)(in + 6 * stride));
+ in07 = _mm_load_si128((const __m128i *)(in + 7 * stride));
+ in08 = _mm_load_si128((const __m128i *)(in + 8 * stride));
+ in09 = _mm_load_si128((const __m128i *)(in + 9 * stride));
+ in10 = _mm_load_si128((const __m128i *)(in + 10 * stride));
+ in11 = _mm_load_si128((const __m128i *)(in + 11 * stride));
+ in12 = _mm_load_si128((const __m128i *)(in + 12 * stride));
+ in13 = _mm_load_si128((const __m128i *)(in + 13 * stride));
+ in14 = _mm_load_si128((const __m128i *)(in + 14 * stride));
+ in15 = _mm_load_si128((const __m128i *)(in + 15 * stride));
// x = x << 2
in00 = _mm_slli_epi16(in00, 2);
in01 = _mm_slli_epi16(in01, 2);
@@ -664,22 +655,22 @@
in14 = _mm_slli_epi16(in14, 2);
in15 = _mm_slli_epi16(in15, 2);
} else {
- in00 = _mm_load_si128((const __m128i *)(in + 0 * 16));
- in01 = _mm_load_si128((const __m128i *)(in + 1 * 16));
- in02 = _mm_load_si128((const __m128i *)(in + 2 * 16));
- in03 = _mm_load_si128((const __m128i *)(in + 3 * 16));
- in04 = _mm_load_si128((const __m128i *)(in + 4 * 16));
- in05 = _mm_load_si128((const __m128i *)(in + 5 * 16));
- in06 = _mm_load_si128((const __m128i *)(in + 6 * 16));
- in07 = _mm_load_si128((const __m128i *)(in + 7 * 16));
- in08 = _mm_load_si128((const __m128i *)(in + 8 * 16));
- in09 = _mm_load_si128((const __m128i *)(in + 9 * 16));
- in10 = _mm_load_si128((const __m128i *)(in + 10 * 16));
- in11 = _mm_load_si128((const __m128i *)(in + 11 * 16));
- in12 = _mm_load_si128((const __m128i *)(in + 12 * 16));
- in13 = _mm_load_si128((const __m128i *)(in + 13 * 16));
- in14 = _mm_load_si128((const __m128i *)(in + 14 * 16));
- in15 = _mm_load_si128((const __m128i *)(in + 15 * 16));
+ in00 = _mm_load_si128((const __m128i *)(in + 0 * 16));
+ in01 = _mm_load_si128((const __m128i *)(in + 1 * 16));
+ in02 = _mm_load_si128((const __m128i *)(in + 2 * 16));
+ in03 = _mm_load_si128((const __m128i *)(in + 3 * 16));
+ in04 = _mm_load_si128((const __m128i *)(in + 4 * 16));
+ in05 = _mm_load_si128((const __m128i *)(in + 5 * 16));
+ in06 = _mm_load_si128((const __m128i *)(in + 6 * 16));
+ in07 = _mm_load_si128((const __m128i *)(in + 7 * 16));
+ in08 = _mm_load_si128((const __m128i *)(in + 8 * 16));
+ in09 = _mm_load_si128((const __m128i *)(in + 9 * 16));
+ in10 = _mm_load_si128((const __m128i *)(in + 10 * 16));
+ in11 = _mm_load_si128((const __m128i *)(in + 11 * 16));
+ in12 = _mm_load_si128((const __m128i *)(in + 12 * 16));
+ in13 = _mm_load_si128((const __m128i *)(in + 13 * 16));
+ in14 = _mm_load_si128((const __m128i *)(in + 14 * 16));
+ in15 = _mm_load_si128((const __m128i *)(in + 15 * 16));
// x = (x + 1) >> 2
in00 = _mm_add_epi16(in00, kOne);
in01 = _mm_add_epi16(in01, kOne);
@@ -745,10 +736,9 @@
step1_6 = SUB_EPI16(in01, in14);
step1_7 = SUB_EPI16(in00, in15);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&step1_0, &step1_1,
- &step1_2, &step1_3,
- &step1_4, &step1_5,
- &step1_6, &step1_7);
+ overflow =
+ check_epi16_overflow_x8(&step1_0, &step1_1, &step1_2, &step1_3,
+ &step1_4, &step1_5, &step1_6, &step1_7);
if (overflow) {
vpx_highbd_fdct16x16_c(input, output, stride);
return;
@@ -767,8 +757,8 @@
const __m128i q6 = SUB_EPI16(input1, input6);
const __m128i q7 = SUB_EPI16(input0, input7);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&q0, &q1, &q2, &q3,
- &q4, &q5, &q6, &q7);
+ overflow =
+ check_epi16_overflow_x8(&q0, &q1, &q2, &q3, &q4, &q5, &q6, &q7);
if (overflow) {
vpx_highbd_fdct16x16_c(input, output, stride);
return;
@@ -818,12 +808,12 @@
// into 32 bits.
const __m128i d0 = _mm_unpacklo_epi16(q6, q5);
const __m128i d1 = _mm_unpackhi_epi16(q6, q5);
- const __m128i r0 = mult_round_shift(&d0, &d1, &k__cospi_p16_m16,
- &k__DCT_CONST_ROUNDING,
- DCT_CONST_BITS);
- const __m128i r1 = mult_round_shift(&d0, &d1, &k__cospi_p16_p16,
- &k__DCT_CONST_ROUNDING,
- DCT_CONST_BITS);
+ const __m128i r0 =
+ mult_round_shift(&d0, &d1, &k__cospi_p16_m16,
+ &k__DCT_CONST_ROUNDING, DCT_CONST_BITS);
+ const __m128i r1 =
+ mult_round_shift(&d0, &d1, &k__cospi_p16_p16,
+ &k__DCT_CONST_ROUNDING, DCT_CONST_BITS);
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x2(&r0, &r1);
if (overflow) {
@@ -860,8 +850,8 @@
res06 = mult_round_shift(&t2, &t3, &k__cospi_m20_p12,
&k__DCT_CONST_ROUNDING, DCT_CONST_BITS);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&res02, &res14,
- &res10, &res06);
+ overflow =
+ check_epi16_overflow_x4(&res02, &res14, &res10, &res06);
if (overflow) {
vpx_highbd_fdct16x16_c(input, output, stride);
return;
@@ -888,8 +878,8 @@
step2_4 = mult_round_shift(&t2, &t3, &k__cospi_p16_p16,
&k__DCT_CONST_ROUNDING, DCT_CONST_BITS);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&step2_2, &step2_3, &step2_5,
- &step2_4);
+ overflow =
+ check_epi16_overflow_x4(&step2_2, &step2_3, &step2_5, &step2_4);
if (overflow) {
vpx_highbd_fdct16x16_c(input, output, stride);
return;
@@ -907,10 +897,9 @@
step3_6 = ADD_EPI16(step1_6, step2_5);
step3_7 = ADD_EPI16(step1_7, step2_4);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&step3_0, &step3_1,
- &step3_2, &step3_3,
- &step3_4, &step3_5,
- &step3_6, &step3_7);
+ overflow =
+ check_epi16_overflow_x8(&step3_0, &step3_1, &step3_2, &step3_3,
+ &step3_4, &step3_5, &step3_6, &step3_7);
if (overflow) {
vpx_highbd_fdct16x16_c(input, output, stride);
return;
@@ -932,8 +921,8 @@
step2_5 = mult_round_shift(&t2, &t3, &k__cospi_p08_m24,
&k__DCT_CONST_ROUNDING, DCT_CONST_BITS);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x4(&step2_1, &step2_2, &step2_6,
- &step2_5);
+ overflow =
+ check_epi16_overflow_x4(&step2_1, &step2_2, &step2_6, &step2_5);
if (overflow) {
vpx_highbd_fdct16x16_c(input, output, stride);
return;
@@ -951,10 +940,9 @@
step1_6 = SUB_EPI16(step3_7, step2_6);
step1_7 = ADD_EPI16(step3_7, step2_6);
#if DCT_HIGH_BIT_DEPTH
- overflow = check_epi16_overflow_x8(&step1_0, &step1_1,
- &step1_2, &step1_3,
- &step1_4, &step1_5,
- &step1_6, &step1_7);
+ overflow =
+ check_epi16_overflow_x8(&step1_0, &step1_1, &step1_2, &step1_3,
+ &step1_4, &step1_5, &step1_6, &step1_7);
if (overflow) {
vpx_highbd_fdct16x16_c(input, output, stride);
return;
@@ -1006,16 +994,14 @@
}
}
// Transpose the results, do it as two 8x8 transposes.
- transpose_and_output8x8(&res00, &res01, &res02, &res03,
- &res04, &res05, &res06, &res07,
- pass, out0, out1);
- transpose_and_output8x8(&res08, &res09, &res10, &res11,
- &res12, &res13, &res14, &res15,
- pass, out0 + 8, out1 + 8);
+ transpose_and_output8x8(&res00, &res01, &res02, &res03, &res04, &res05,
+ &res06, &res07, pass, out0, out1);
+ transpose_and_output8x8(&res08, &res09, &res10, &res11, &res12, &res13,
+ &res14, &res15, pass, out0 + 8, out1 + 8);
if (pass == 0) {
- out0 += 8*16;
+ out0 += 8 * 16;
} else {
- out1 += 8*16;
+ out1 += 8 * 16;
}
}
// Setup in/out for next pass.
diff --git a/vp10/common/x86/vp10_fwd_txfm_sse2.c b/vp10/common/x86/vp10_fwd_txfm_sse2.c
index cf6aa8d..5b26a31 100644
--- a/vp10/common/x86/vp10_fwd_txfm_sse2.c
+++ b/vp10/common/x86/vp10_fwd_txfm_sse2.c
@@ -18,12 +18,12 @@
__m128i in0, in1;
__m128i tmp;
const __m128i zero = _mm_setzero_si128();
- in0 = _mm_loadl_epi64((const __m128i *)(input + 0 * stride));
- in1 = _mm_loadl_epi64((const __m128i *)(input + 1 * stride));
- in1 = _mm_unpacklo_epi64(in1, _mm_loadl_epi64((const __m128i *)
- (input + 2 * stride)));
- in0 = _mm_unpacklo_epi64(in0, _mm_loadl_epi64((const __m128i *)
- (input + 3 * stride)));
+ in0 = _mm_loadl_epi64((const __m128i *)(input + 0 * stride));
+ in1 = _mm_loadl_epi64((const __m128i *)(input + 1 * stride));
+ in1 = _mm_unpacklo_epi64(
+ in1, _mm_loadl_epi64((const __m128i *)(input + 2 * stride)));
+ in0 = _mm_unpacklo_epi64(
+ in0, _mm_loadl_epi64((const __m128i *)(input + 3 * stride)));
tmp = _mm_add_epi16(in0, in1);
in0 = _mm_unpacklo_epi16(zero, tmp);
@@ -44,19 +44,19 @@
}
void vp10_fdct8x8_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
- __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
- __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
- __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
- __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
+ __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
+ __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
+ __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
+ __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
__m128i u0, u1, sum;
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
- in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
+ in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
sum = _mm_add_epi16(u0, u1);
@@ -64,7 +64,7 @@
in2 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, in0);
- u0 = _mm_setzero_si128();
+ u0 = _mm_setzero_si128();
sum = _mm_add_epi16(sum, in2);
in0 = _mm_unpacklo_epi16(u0, sum);
@@ -84,7 +84,7 @@
}
void vp10_fdct16x16_1_sse2(const int16_t *input, tran_low_t *output,
- int stride) {
+ int stride) {
__m128i in0, in1, in2, in3;
__m128i u0, u1;
__m128i sum = _mm_setzero_si128();
@@ -92,49 +92,49 @@
for (i = 0; i < 2; ++i) {
input += 8 * i;
- in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
+ in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
- in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
+ in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
- in0 = _mm_load_si128((const __m128i *)(input + 8 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 9 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 10 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 11 * stride));
+ in0 = _mm_load_si128((const __m128i *)(input + 8 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 9 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 10 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 11 * stride));
sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
- in0 = _mm_load_si128((const __m128i *)(input + 12 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 13 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 14 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 15 * stride));
+ in0 = _mm_load_si128((const __m128i *)(input + 12 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 13 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 14 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 15 * stride));
sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
sum = _mm_add_epi16(sum, u1);
}
- u0 = _mm_setzero_si128();
+ u0 = _mm_setzero_si128();
in0 = _mm_unpacklo_epi16(u0, sum);
in1 = _mm_unpackhi_epi16(u0, sum);
in0 = _mm_srai_epi32(in0, 16);
@@ -153,60 +153,60 @@
}
void vp10_fdct32x32_1_sse2(const int16_t *input, tran_low_t *output,
- int stride) {
+ int stride) {
__m128i in0, in1, in2, in3;
__m128i u0, u1;
__m128i sum = _mm_setzero_si128();
int i;
for (i = 0; i < 8; ++i) {
- in0 = _mm_load_si128((const __m128i *)(input + 0));
- in1 = _mm_load_si128((const __m128i *)(input + 8));
- in2 = _mm_load_si128((const __m128i *)(input + 16));
- in3 = _mm_load_si128((const __m128i *)(input + 24));
+ in0 = _mm_load_si128((const __m128i *)(input + 0));
+ in1 = _mm_load_si128((const __m128i *)(input + 8));
+ in2 = _mm_load_si128((const __m128i *)(input + 16));
+ in3 = _mm_load_si128((const __m128i *)(input + 24));
input += stride;
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
- in0 = _mm_load_si128((const __m128i *)(input + 0));
- in1 = _mm_load_si128((const __m128i *)(input + 8));
- in2 = _mm_load_si128((const __m128i *)(input + 16));
- in3 = _mm_load_si128((const __m128i *)(input + 24));
+ in0 = _mm_load_si128((const __m128i *)(input + 0));
+ in1 = _mm_load_si128((const __m128i *)(input + 8));
+ in2 = _mm_load_si128((const __m128i *)(input + 16));
+ in3 = _mm_load_si128((const __m128i *)(input + 24));
input += stride;
sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
- in0 = _mm_load_si128((const __m128i *)(input + 0));
- in1 = _mm_load_si128((const __m128i *)(input + 8));
- in2 = _mm_load_si128((const __m128i *)(input + 16));
- in3 = _mm_load_si128((const __m128i *)(input + 24));
+ in0 = _mm_load_si128((const __m128i *)(input + 0));
+ in1 = _mm_load_si128((const __m128i *)(input + 8));
+ in2 = _mm_load_si128((const __m128i *)(input + 16));
+ in3 = _mm_load_si128((const __m128i *)(input + 24));
input += stride;
sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
- in0 = _mm_load_si128((const __m128i *)(input + 0));
- in1 = _mm_load_si128((const __m128i *)(input + 8));
- in2 = _mm_load_si128((const __m128i *)(input + 16));
- in3 = _mm_load_si128((const __m128i *)(input + 24));
+ in0 = _mm_load_si128((const __m128i *)(input + 0));
+ in1 = _mm_load_si128((const __m128i *)(input + 8));
+ in2 = _mm_load_si128((const __m128i *)(input + 16));
+ in3 = _mm_load_si128((const __m128i *)(input + 24));
input += stride;
sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
sum = _mm_add_epi16(sum, u1);
}
- u0 = _mm_setzero_si128();
+ u0 = _mm_setzero_si128();
in0 = _mm_unpacklo_epi16(u0, sum);
in1 = _mm_unpackhi_epi16(u0, sum);
in0 = _mm_srai_epi32(in0, 16);
@@ -229,43 +229,43 @@
#define FDCT8x8_2D vp10_fdct8x8_sse2
#define FDCT16x16_2D vp10_fdct16x16_sse2
#include "vp10/common/x86/vp10_fwd_txfm_impl_sse2.h"
-#undef FDCT4x4_2D
-#undef FDCT8x8_2D
-#undef FDCT16x16_2D
+#undef FDCT4x4_2D
+#undef FDCT8x8_2D
+#undef FDCT16x16_2D
#define FDCT32x32_2D vp10_fdct32x32_rd_sse2
#define FDCT32x32_HIGH_PRECISION 0
#include "vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h"
-#undef FDCT32x32_2D
-#undef FDCT32x32_HIGH_PRECISION
+#undef FDCT32x32_2D
+#undef FDCT32x32_HIGH_PRECISION
#define FDCT32x32_2D vp10_fdct32x32_sse2
#define FDCT32x32_HIGH_PRECISION 1
#include "vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h" // NOLINT
-#undef FDCT32x32_2D
-#undef FDCT32x32_HIGH_PRECISION
-#undef DCT_HIGH_BIT_DEPTH
+#undef FDCT32x32_2D
+#undef FDCT32x32_HIGH_PRECISION
+#undef DCT_HIGH_BIT_DEPTH
#if CONFIG_VPX_HIGHBITDEPTH
#define DCT_HIGH_BIT_DEPTH 1
#define FDCT4x4_2D vp10_highbd_fdct4x4_sse2
#define FDCT8x8_2D vp10_highbd_fdct8x8_sse2
#define FDCT16x16_2D vp10_highbd_fdct16x16_sse2
-#include "vp10/common/x86/vp10_fwd_txfm_impl_sse2.h" // NOLINT
-#undef FDCT4x4_2D
-#undef FDCT8x8_2D
-#undef FDCT16x16_2D
+#include "vp10/common/x86/vp10_fwd_txfm_impl_sse2.h" // NOLINT
+#undef FDCT4x4_2D
+#undef FDCT8x8_2D
+#undef FDCT16x16_2D
#define FDCT32x32_2D vp10_highbd_fdct32x32_rd_sse2
#define FDCT32x32_HIGH_PRECISION 0
-#include "vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h" // NOLINT
-#undef FDCT32x32_2D
-#undef FDCT32x32_HIGH_PRECISION
+#include "vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h" // NOLINT
+#undef FDCT32x32_2D
+#undef FDCT32x32_HIGH_PRECISION
#define FDCT32x32_2D vp10_highbd_fdct32x32_sse2
#define FDCT32x32_HIGH_PRECISION 1
-#include "vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h" // NOLINT
-#undef FDCT32x32_2D
-#undef FDCT32x32_HIGH_PRECISION
-#undef DCT_HIGH_BIT_DEPTH
+#include "vp10/common/x86/vp10_fwd_dct32x32_impl_sse2.h" // NOLINT
+#undef FDCT32x32_2D
+#undef FDCT32x32_HIGH_PRECISION
+#undef DCT_HIGH_BIT_DEPTH
#endif // CONFIG_VPX_HIGHBITDEPTH
diff --git a/vp10/common/x86/vp10_inv_txfm_sse2.c b/vp10/common/x86/vp10_inv_txfm_sse2.c
index 6c108a8..c200392 100644
--- a/vp10/common/x86/vp10_inv_txfm_sse2.c
+++ b/vp10/common/x86/vp10_inv_txfm_sse2.c
@@ -12,14 +12,14 @@
#include "vp10/common/x86/vp10_inv_txfm_sse2.h"
#include "vpx_dsp/x86/txfm_common_sse2.h"
-#define RECON_AND_STORE4X4(dest, in_x) \
-{ \
- __m128i d0 = _mm_cvtsi32_si128(*(const int *)(dest)); \
- d0 = _mm_unpacklo_epi8(d0, zero); \
- d0 = _mm_add_epi16(in_x, d0); \
- d0 = _mm_packus_epi16(d0, d0); \
- *(int *)(dest) = _mm_cvtsi128_si32(d0); \
-}
+#define RECON_AND_STORE4X4(dest, in_x) \
+ { \
+ __m128i d0 = _mm_cvtsi32_si128(*(const int *)(dest)); \
+ d0 = _mm_unpacklo_epi8(d0, zero); \
+ d0 = _mm_add_epi16(in_x, d0); \
+ d0 = _mm_packus_epi16(d0, d0); \
+ *(int *)(dest) = _mm_cvtsi128_si32(d0); \
+ }
void vp10_idct4x4_16_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
const __m128i zero = _mm_setzero_si128();
@@ -261,192 +261,189 @@
in[1] = _mm_packs_epi32(u[2], u[3]);
}
-#define TRANSPOSE_8X8(in0, in1, in2, in3, in4, in5, in6, in7, \
- out0, out1, out2, out3, out4, out5, out6, out7) \
- { \
- const __m128i tr0_0 = _mm_unpacklo_epi16(in0, in1); \
- const __m128i tr0_1 = _mm_unpacklo_epi16(in2, in3); \
- const __m128i tr0_2 = _mm_unpackhi_epi16(in0, in1); \
- const __m128i tr0_3 = _mm_unpackhi_epi16(in2, in3); \
- const __m128i tr0_4 = _mm_unpacklo_epi16(in4, in5); \
- const __m128i tr0_5 = _mm_unpacklo_epi16(in6, in7); \
- const __m128i tr0_6 = _mm_unpackhi_epi16(in4, in5); \
- const __m128i tr0_7 = _mm_unpackhi_epi16(in6, in7); \
- \
- const __m128i tr1_0 = _mm_unpacklo_epi32(tr0_0, tr0_1); \
- const __m128i tr1_1 = _mm_unpacklo_epi32(tr0_2, tr0_3); \
- const __m128i tr1_2 = _mm_unpackhi_epi32(tr0_0, tr0_1); \
- const __m128i tr1_3 = _mm_unpackhi_epi32(tr0_2, tr0_3); \
- const __m128i tr1_4 = _mm_unpacklo_epi32(tr0_4, tr0_5); \
- const __m128i tr1_5 = _mm_unpacklo_epi32(tr0_6, tr0_7); \
- const __m128i tr1_6 = _mm_unpackhi_epi32(tr0_4, tr0_5); \
- const __m128i tr1_7 = _mm_unpackhi_epi32(tr0_6, tr0_7); \
- \
- out0 = _mm_unpacklo_epi64(tr1_0, tr1_4); \
- out1 = _mm_unpackhi_epi64(tr1_0, tr1_4); \
- out2 = _mm_unpacklo_epi64(tr1_2, tr1_6); \
- out3 = _mm_unpackhi_epi64(tr1_2, tr1_6); \
- out4 = _mm_unpacklo_epi64(tr1_1, tr1_5); \
- out5 = _mm_unpackhi_epi64(tr1_1, tr1_5); \
- out6 = _mm_unpacklo_epi64(tr1_3, tr1_7); \
- out7 = _mm_unpackhi_epi64(tr1_3, tr1_7); \
+#define TRANSPOSE_8X8(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
+ out2, out3, out4, out5, out6, out7) \
+ { \
+ const __m128i tr0_0 = _mm_unpacklo_epi16(in0, in1); \
+ const __m128i tr0_1 = _mm_unpacklo_epi16(in2, in3); \
+ const __m128i tr0_2 = _mm_unpackhi_epi16(in0, in1); \
+ const __m128i tr0_3 = _mm_unpackhi_epi16(in2, in3); \
+ const __m128i tr0_4 = _mm_unpacklo_epi16(in4, in5); \
+ const __m128i tr0_5 = _mm_unpacklo_epi16(in6, in7); \
+ const __m128i tr0_6 = _mm_unpackhi_epi16(in4, in5); \
+ const __m128i tr0_7 = _mm_unpackhi_epi16(in6, in7); \
+ \
+ const __m128i tr1_0 = _mm_unpacklo_epi32(tr0_0, tr0_1); \
+ const __m128i tr1_1 = _mm_unpacklo_epi32(tr0_2, tr0_3); \
+ const __m128i tr1_2 = _mm_unpackhi_epi32(tr0_0, tr0_1); \
+ const __m128i tr1_3 = _mm_unpackhi_epi32(tr0_2, tr0_3); \
+ const __m128i tr1_4 = _mm_unpacklo_epi32(tr0_4, tr0_5); \
+ const __m128i tr1_5 = _mm_unpacklo_epi32(tr0_6, tr0_7); \
+ const __m128i tr1_6 = _mm_unpackhi_epi32(tr0_4, tr0_5); \
+ const __m128i tr1_7 = _mm_unpackhi_epi32(tr0_6, tr0_7); \
+ \
+ out0 = _mm_unpacklo_epi64(tr1_0, tr1_4); \
+ out1 = _mm_unpackhi_epi64(tr1_0, tr1_4); \
+ out2 = _mm_unpacklo_epi64(tr1_2, tr1_6); \
+ out3 = _mm_unpackhi_epi64(tr1_2, tr1_6); \
+ out4 = _mm_unpacklo_epi64(tr1_1, tr1_5); \
+ out5 = _mm_unpackhi_epi64(tr1_1, tr1_5); \
+ out6 = _mm_unpacklo_epi64(tr1_3, tr1_7); \
+ out7 = _mm_unpackhi_epi64(tr1_3, tr1_7); \
}
-#define TRANSPOSE_4X8_10(tmp0, tmp1, tmp2, tmp3, \
- out0, out1, out2, out3) \
- { \
- const __m128i tr0_0 = _mm_unpackhi_epi16(tmp0, tmp1); \
- const __m128i tr0_1 = _mm_unpacklo_epi16(tmp1, tmp0); \
- const __m128i tr0_4 = _mm_unpacklo_epi16(tmp2, tmp3); \
- const __m128i tr0_5 = _mm_unpackhi_epi16(tmp3, tmp2); \
- \
- const __m128i tr1_0 = _mm_unpacklo_epi32(tr0_0, tr0_1); \
- const __m128i tr1_2 = _mm_unpackhi_epi32(tr0_0, tr0_1); \
- const __m128i tr1_4 = _mm_unpacklo_epi32(tr0_4, tr0_5); \
- const __m128i tr1_6 = _mm_unpackhi_epi32(tr0_4, tr0_5); \
- \
- out0 = _mm_unpacklo_epi64(tr1_0, tr1_4); \
- out1 = _mm_unpackhi_epi64(tr1_0, tr1_4); \
- out2 = _mm_unpacklo_epi64(tr1_2, tr1_6); \
- out3 = _mm_unpackhi_epi64(tr1_2, tr1_6); \
+#define TRANSPOSE_4X8_10(tmp0, tmp1, tmp2, tmp3, out0, out1, out2, out3) \
+ { \
+ const __m128i tr0_0 = _mm_unpackhi_epi16(tmp0, tmp1); \
+ const __m128i tr0_1 = _mm_unpacklo_epi16(tmp1, tmp0); \
+ const __m128i tr0_4 = _mm_unpacklo_epi16(tmp2, tmp3); \
+ const __m128i tr0_5 = _mm_unpackhi_epi16(tmp3, tmp2); \
+ \
+ const __m128i tr1_0 = _mm_unpacklo_epi32(tr0_0, tr0_1); \
+ const __m128i tr1_2 = _mm_unpackhi_epi32(tr0_0, tr0_1); \
+ const __m128i tr1_4 = _mm_unpacklo_epi32(tr0_4, tr0_5); \
+ const __m128i tr1_6 = _mm_unpackhi_epi32(tr0_4, tr0_5); \
+ \
+ out0 = _mm_unpacklo_epi64(tr1_0, tr1_4); \
+ out1 = _mm_unpackhi_epi64(tr1_0, tr1_4); \
+ out2 = _mm_unpacklo_epi64(tr1_2, tr1_6); \
+ out3 = _mm_unpackhi_epi64(tr1_2, tr1_6); \
}
#define TRANSPOSE_8X8_10(in0, in1, in2, in3, out0, out1) \
- { \
- const __m128i tr0_0 = _mm_unpacklo_epi16(in0, in1); \
- const __m128i tr0_1 = _mm_unpacklo_epi16(in2, in3); \
- out0 = _mm_unpacklo_epi32(tr0_0, tr0_1); \
- out1 = _mm_unpackhi_epi32(tr0_0, tr0_1); \
+ { \
+ const __m128i tr0_0 = _mm_unpacklo_epi16(in0, in1); \
+ const __m128i tr0_1 = _mm_unpacklo_epi16(in2, in3); \
+ out0 = _mm_unpacklo_epi32(tr0_0, tr0_1); \
+ out1 = _mm_unpackhi_epi32(tr0_0, tr0_1); \
}
// Define Macro for multiplying elements by constants and adding them together.
-#define MULTIPLICATION_AND_ADD(lo_0, hi_0, lo_1, hi_1, \
- cst0, cst1, cst2, cst3, res0, res1, res2, res3) \
- { \
- tmp0 = _mm_madd_epi16(lo_0, cst0); \
- tmp1 = _mm_madd_epi16(hi_0, cst0); \
- tmp2 = _mm_madd_epi16(lo_0, cst1); \
- tmp3 = _mm_madd_epi16(hi_0, cst1); \
- tmp4 = _mm_madd_epi16(lo_1, cst2); \
- tmp5 = _mm_madd_epi16(hi_1, cst2); \
- tmp6 = _mm_madd_epi16(lo_1, cst3); \
- tmp7 = _mm_madd_epi16(hi_1, cst3); \
- \
- tmp0 = _mm_add_epi32(tmp0, rounding); \
- tmp1 = _mm_add_epi32(tmp1, rounding); \
- tmp2 = _mm_add_epi32(tmp2, rounding); \
- tmp3 = _mm_add_epi32(tmp3, rounding); \
- tmp4 = _mm_add_epi32(tmp4, rounding); \
- tmp5 = _mm_add_epi32(tmp5, rounding); \
- tmp6 = _mm_add_epi32(tmp6, rounding); \
- tmp7 = _mm_add_epi32(tmp7, rounding); \
- \
- tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
- tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
- tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
- tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
- tmp4 = _mm_srai_epi32(tmp4, DCT_CONST_BITS); \
- tmp5 = _mm_srai_epi32(tmp5, DCT_CONST_BITS); \
- tmp6 = _mm_srai_epi32(tmp6, DCT_CONST_BITS); \
- tmp7 = _mm_srai_epi32(tmp7, DCT_CONST_BITS); \
- \
- res0 = _mm_packs_epi32(tmp0, tmp1); \
- res1 = _mm_packs_epi32(tmp2, tmp3); \
- res2 = _mm_packs_epi32(tmp4, tmp5); \
- res3 = _mm_packs_epi32(tmp6, tmp7); \
+#define MULTIPLICATION_AND_ADD(lo_0, hi_0, lo_1, hi_1, cst0, cst1, cst2, cst3, \
+ res0, res1, res2, res3) \
+ { \
+ tmp0 = _mm_madd_epi16(lo_0, cst0); \
+ tmp1 = _mm_madd_epi16(hi_0, cst0); \
+ tmp2 = _mm_madd_epi16(lo_0, cst1); \
+ tmp3 = _mm_madd_epi16(hi_0, cst1); \
+ tmp4 = _mm_madd_epi16(lo_1, cst2); \
+ tmp5 = _mm_madd_epi16(hi_1, cst2); \
+ tmp6 = _mm_madd_epi16(lo_1, cst3); \
+ tmp7 = _mm_madd_epi16(hi_1, cst3); \
+ \
+ tmp0 = _mm_add_epi32(tmp0, rounding); \
+ tmp1 = _mm_add_epi32(tmp1, rounding); \
+ tmp2 = _mm_add_epi32(tmp2, rounding); \
+ tmp3 = _mm_add_epi32(tmp3, rounding); \
+ tmp4 = _mm_add_epi32(tmp4, rounding); \
+ tmp5 = _mm_add_epi32(tmp5, rounding); \
+ tmp6 = _mm_add_epi32(tmp6, rounding); \
+ tmp7 = _mm_add_epi32(tmp7, rounding); \
+ \
+ tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
+ tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
+ tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
+ tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
+ tmp4 = _mm_srai_epi32(tmp4, DCT_CONST_BITS); \
+ tmp5 = _mm_srai_epi32(tmp5, DCT_CONST_BITS); \
+ tmp6 = _mm_srai_epi32(tmp6, DCT_CONST_BITS); \
+ tmp7 = _mm_srai_epi32(tmp7, DCT_CONST_BITS); \
+ \
+ res0 = _mm_packs_epi32(tmp0, tmp1); \
+ res1 = _mm_packs_epi32(tmp2, tmp3); \
+ res2 = _mm_packs_epi32(tmp4, tmp5); \
+ res3 = _mm_packs_epi32(tmp6, tmp7); \
}
#define MULTIPLICATION_AND_ADD_2(lo_0, hi_0, cst0, cst1, res0, res1) \
- { \
- tmp0 = _mm_madd_epi16(lo_0, cst0); \
- tmp1 = _mm_madd_epi16(hi_0, cst0); \
- tmp2 = _mm_madd_epi16(lo_0, cst1); \
- tmp3 = _mm_madd_epi16(hi_0, cst1); \
- \
- tmp0 = _mm_add_epi32(tmp0, rounding); \
- tmp1 = _mm_add_epi32(tmp1, rounding); \
- tmp2 = _mm_add_epi32(tmp2, rounding); \
- tmp3 = _mm_add_epi32(tmp3, rounding); \
- \
- tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
- tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
- tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
- tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
- \
- res0 = _mm_packs_epi32(tmp0, tmp1); \
- res1 = _mm_packs_epi32(tmp2, tmp3); \
+ { \
+ tmp0 = _mm_madd_epi16(lo_0, cst0); \
+ tmp1 = _mm_madd_epi16(hi_0, cst0); \
+ tmp2 = _mm_madd_epi16(lo_0, cst1); \
+ tmp3 = _mm_madd_epi16(hi_0, cst1); \
+ \
+ tmp0 = _mm_add_epi32(tmp0, rounding); \
+ tmp1 = _mm_add_epi32(tmp1, rounding); \
+ tmp2 = _mm_add_epi32(tmp2, rounding); \
+ tmp3 = _mm_add_epi32(tmp3, rounding); \
+ \
+ tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
+ tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
+ tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
+ tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
+ \
+ res0 = _mm_packs_epi32(tmp0, tmp1); \
+ res1 = _mm_packs_epi32(tmp2, tmp3); \
}
-#define IDCT8(in0, in1, in2, in3, in4, in5, in6, in7, \
- out0, out1, out2, out3, out4, out5, out6, out7) \
- { \
- /* Stage1 */ \
- { \
- const __m128i lo_17 = _mm_unpacklo_epi16(in1, in7); \
- const __m128i hi_17 = _mm_unpackhi_epi16(in1, in7); \
- const __m128i lo_35 = _mm_unpacklo_epi16(in3, in5); \
- const __m128i hi_35 = _mm_unpackhi_epi16(in3, in5); \
- \
- MULTIPLICATION_AND_ADD(lo_17, hi_17, lo_35, hi_35, stg1_0, \
- stg1_1, stg1_2, stg1_3, stp1_4, \
- stp1_7, stp1_5, stp1_6) \
- } \
- \
- /* Stage2 */ \
- { \
- const __m128i lo_04 = _mm_unpacklo_epi16(in0, in4); \
- const __m128i hi_04 = _mm_unpackhi_epi16(in0, in4); \
- const __m128i lo_26 = _mm_unpacklo_epi16(in2, in6); \
- const __m128i hi_26 = _mm_unpackhi_epi16(in2, in6); \
- \
- MULTIPLICATION_AND_ADD(lo_04, hi_04, lo_26, hi_26, stg2_0, \
- stg2_1, stg2_2, stg2_3, stp2_0, \
- stp2_1, stp2_2, stp2_3) \
- \
- stp2_4 = _mm_adds_epi16(stp1_4, stp1_5); \
- stp2_5 = _mm_subs_epi16(stp1_4, stp1_5); \
- stp2_6 = _mm_subs_epi16(stp1_7, stp1_6); \
- stp2_7 = _mm_adds_epi16(stp1_7, stp1_6); \
- } \
- \
- /* Stage3 */ \
- { \
- const __m128i lo_56 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
- const __m128i hi_56 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
- \
- stp1_0 = _mm_adds_epi16(stp2_0, stp2_3); \
- stp1_1 = _mm_adds_epi16(stp2_1, stp2_2); \
- stp1_2 = _mm_subs_epi16(stp2_1, stp2_2); \
- stp1_3 = _mm_subs_epi16(stp2_0, stp2_3); \
- \
- tmp0 = _mm_madd_epi16(lo_56, stg2_1); \
- tmp1 = _mm_madd_epi16(hi_56, stg2_1); \
- tmp2 = _mm_madd_epi16(lo_56, stg2_0); \
- tmp3 = _mm_madd_epi16(hi_56, stg2_0); \
- \
- tmp0 = _mm_add_epi32(tmp0, rounding); \
- tmp1 = _mm_add_epi32(tmp1, rounding); \
- tmp2 = _mm_add_epi32(tmp2, rounding); \
- tmp3 = _mm_add_epi32(tmp3, rounding); \
- \
- tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
- tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
- tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
- tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
- \
- stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
- stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
- } \
- \
- /* Stage4 */ \
- out0 = _mm_adds_epi16(stp1_0, stp2_7); \
- out1 = _mm_adds_epi16(stp1_1, stp1_6); \
- out2 = _mm_adds_epi16(stp1_2, stp1_5); \
- out3 = _mm_adds_epi16(stp1_3, stp2_4); \
- out4 = _mm_subs_epi16(stp1_3, stp2_4); \
- out5 = _mm_subs_epi16(stp1_2, stp1_5); \
- out6 = _mm_subs_epi16(stp1_1, stp1_6); \
- out7 = _mm_subs_epi16(stp1_0, stp2_7); \
+#define IDCT8(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, out3, \
+ out4, out5, out6, out7) \
+ { \
+ /* Stage1 */ \
+ { \
+ const __m128i lo_17 = _mm_unpacklo_epi16(in1, in7); \
+ const __m128i hi_17 = _mm_unpackhi_epi16(in1, in7); \
+ const __m128i lo_35 = _mm_unpacklo_epi16(in3, in5); \
+ const __m128i hi_35 = _mm_unpackhi_epi16(in3, in5); \
+ \
+ MULTIPLICATION_AND_ADD(lo_17, hi_17, lo_35, hi_35, stg1_0, stg1_1, \
+ stg1_2, stg1_3, stp1_4, stp1_7, stp1_5, stp1_6) \
+ } \
+ \
+ /* Stage2 */ \
+ { \
+ const __m128i lo_04 = _mm_unpacklo_epi16(in0, in4); \
+ const __m128i hi_04 = _mm_unpackhi_epi16(in0, in4); \
+ const __m128i lo_26 = _mm_unpacklo_epi16(in2, in6); \
+ const __m128i hi_26 = _mm_unpackhi_epi16(in2, in6); \
+ \
+ MULTIPLICATION_AND_ADD(lo_04, hi_04, lo_26, hi_26, stg2_0, stg2_1, \
+ stg2_2, stg2_3, stp2_0, stp2_1, stp2_2, stp2_3) \
+ \
+ stp2_4 = _mm_adds_epi16(stp1_4, stp1_5); \
+ stp2_5 = _mm_subs_epi16(stp1_4, stp1_5); \
+ stp2_6 = _mm_subs_epi16(stp1_7, stp1_6); \
+ stp2_7 = _mm_adds_epi16(stp1_7, stp1_6); \
+ } \
+ \
+ /* Stage3 */ \
+ { \
+ const __m128i lo_56 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
+ const __m128i hi_56 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
+ \
+ stp1_0 = _mm_adds_epi16(stp2_0, stp2_3); \
+ stp1_1 = _mm_adds_epi16(stp2_1, stp2_2); \
+ stp1_2 = _mm_subs_epi16(stp2_1, stp2_2); \
+ stp1_3 = _mm_subs_epi16(stp2_0, stp2_3); \
+ \
+ tmp0 = _mm_madd_epi16(lo_56, stg2_1); \
+ tmp1 = _mm_madd_epi16(hi_56, stg2_1); \
+ tmp2 = _mm_madd_epi16(lo_56, stg2_0); \
+ tmp3 = _mm_madd_epi16(hi_56, stg2_0); \
+ \
+ tmp0 = _mm_add_epi32(tmp0, rounding); \
+ tmp1 = _mm_add_epi32(tmp1, rounding); \
+ tmp2 = _mm_add_epi32(tmp2, rounding); \
+ tmp3 = _mm_add_epi32(tmp3, rounding); \
+ \
+ tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
+ tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
+ tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
+ tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
+ \
+ stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
+ stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
+ } \
+ \
+ /* Stage4 */ \
+ out0 = _mm_adds_epi16(stp1_0, stp2_7); \
+ out1 = _mm_adds_epi16(stp1_1, stp1_6); \
+ out2 = _mm_adds_epi16(stp1_2, stp1_5); \
+ out3 = _mm_adds_epi16(stp1_3, stp2_4); \
+ out4 = _mm_subs_epi16(stp1_3, stp2_4); \
+ out5 = _mm_subs_epi16(stp1_2, stp1_5); \
+ out6 = _mm_subs_epi16(stp1_1, stp1_6); \
+ out7 = _mm_subs_epi16(stp1_0, stp2_7); \
}
void vp10_idct8x8_64_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
@@ -481,12 +478,12 @@
// 2-D
for (i = 0; i < 2; i++) {
// 8x8 Transpose is copied from vp10_fdct8x8_sse2()
- TRANSPOSE_8X8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ TRANSPOSE_8X8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
+ in4, in5, in6, in7);
// 4-stage 1D vp10_idct8x8
- IDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ IDCT8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4, in5,
+ in6, in7);
}
// Final rounding and shift
@@ -556,12 +553,12 @@
__m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
// 8x8 Transpose is copied from vp10_fdct8x8_sse2()
- TRANSPOSE_8X8(in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7],
- in0, in1, in2, in3, in4, in5, in6, in7);
+ TRANSPOSE_8X8(in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7], in0,
+ in1, in2, in3, in4, in5, in6, in7);
// 4-stage 1D vp10_idct8x8
- IDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
- in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7]);
+ IDCT8(in0, in1, in2, in3, in4, in5, in6, in7, in[0], in[1], in[2], in[3],
+ in[4], in[5], in[6], in[7]);
}
void vp10_iadst8_sse2(__m128i *in) {
@@ -901,8 +898,8 @@
TRANSPOSE_4X8_10(tmp0, tmp1, tmp2, tmp3, in0, in1, in2, in3)
- IDCT8(in0, in1, in2, in3, zero, zero, zero, zero,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ IDCT8(in0, in1, in2, in3, zero, zero, zero, zero, in0, in1, in2, in3, in4,
+ in5, in6, in7);
// Final rounding and shift
in0 = _mm_adds_epi16(in0, final_rounding);
in1 = _mm_adds_epi16(in1, final_rounding);
@@ -932,245 +929,237 @@
RECON_AND_STORE(dest + 7 * stride, in7);
}
-#define IDCT16 \
- /* Stage2 */ \
- { \
- const __m128i lo_1_15 = _mm_unpacklo_epi16(in[1], in[15]); \
- const __m128i hi_1_15 = _mm_unpackhi_epi16(in[1], in[15]); \
- const __m128i lo_9_7 = _mm_unpacklo_epi16(in[9], in[7]); \
- const __m128i hi_9_7 = _mm_unpackhi_epi16(in[9], in[7]); \
- const __m128i lo_5_11 = _mm_unpacklo_epi16(in[5], in[11]); \
- const __m128i hi_5_11 = _mm_unpackhi_epi16(in[5], in[11]); \
- const __m128i lo_13_3 = _mm_unpacklo_epi16(in[13], in[3]); \
- const __m128i hi_13_3 = _mm_unpackhi_epi16(in[13], in[3]); \
- \
- MULTIPLICATION_AND_ADD(lo_1_15, hi_1_15, lo_9_7, hi_9_7, \
- stg2_0, stg2_1, stg2_2, stg2_3, \
- stp2_8, stp2_15, stp2_9, stp2_14) \
- \
- MULTIPLICATION_AND_ADD(lo_5_11, hi_5_11, lo_13_3, hi_13_3, \
- stg2_4, stg2_5, stg2_6, stg2_7, \
- stp2_10, stp2_13, stp2_11, stp2_12) \
- } \
- \
- /* Stage3 */ \
- { \
- const __m128i lo_2_14 = _mm_unpacklo_epi16(in[2], in[14]); \
- const __m128i hi_2_14 = _mm_unpackhi_epi16(in[2], in[14]); \
- const __m128i lo_10_6 = _mm_unpacklo_epi16(in[10], in[6]); \
- const __m128i hi_10_6 = _mm_unpackhi_epi16(in[10], in[6]); \
- \
- MULTIPLICATION_AND_ADD(lo_2_14, hi_2_14, lo_10_6, hi_10_6, \
- stg3_0, stg3_1, stg3_2, stg3_3, \
- stp1_4, stp1_7, stp1_5, stp1_6) \
- \
- stp1_8_0 = _mm_add_epi16(stp2_8, stp2_9); \
- stp1_9 = _mm_sub_epi16(stp2_8, stp2_9); \
- stp1_10 = _mm_sub_epi16(stp2_11, stp2_10); \
- stp1_11 = _mm_add_epi16(stp2_11, stp2_10); \
- \
- stp1_12_0 = _mm_add_epi16(stp2_12, stp2_13); \
- stp1_13 = _mm_sub_epi16(stp2_12, stp2_13); \
- stp1_14 = _mm_sub_epi16(stp2_15, stp2_14); \
- stp1_15 = _mm_add_epi16(stp2_15, stp2_14); \
- } \
- \
- /* Stage4 */ \
- { \
- const __m128i lo_0_8 = _mm_unpacklo_epi16(in[0], in[8]); \
- const __m128i hi_0_8 = _mm_unpackhi_epi16(in[0], in[8]); \
- const __m128i lo_4_12 = _mm_unpacklo_epi16(in[4], in[12]); \
- const __m128i hi_4_12 = _mm_unpackhi_epi16(in[4], in[12]); \
- \
- const __m128i lo_9_14 = _mm_unpacklo_epi16(stp1_9, stp1_14); \
- const __m128i hi_9_14 = _mm_unpackhi_epi16(stp1_9, stp1_14); \
- const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
- const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
- \
- MULTIPLICATION_AND_ADD(lo_0_8, hi_0_8, lo_4_12, hi_4_12, \
- stg4_0, stg4_1, stg4_2, stg4_3, \
- stp2_0, stp2_1, stp2_2, stp2_3) \
- \
- stp2_4 = _mm_add_epi16(stp1_4, stp1_5); \
- stp2_5 = _mm_sub_epi16(stp1_4, stp1_5); \
- stp2_6 = _mm_sub_epi16(stp1_7, stp1_6); \
- stp2_7 = _mm_add_epi16(stp1_7, stp1_6); \
- \
- MULTIPLICATION_AND_ADD(lo_9_14, hi_9_14, lo_10_13, hi_10_13, \
- stg4_4, stg4_5, stg4_6, stg4_7, \
- stp2_9, stp2_14, stp2_10, stp2_13) \
- } \
- \
- /* Stage5 */ \
- { \
- const __m128i lo_6_5 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
- const __m128i hi_6_5 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
- \
- stp1_0 = _mm_add_epi16(stp2_0, stp2_3); \
- stp1_1 = _mm_add_epi16(stp2_1, stp2_2); \
- stp1_2 = _mm_sub_epi16(stp2_1, stp2_2); \
- stp1_3 = _mm_sub_epi16(stp2_0, stp2_3); \
- \
- tmp0 = _mm_madd_epi16(lo_6_5, stg4_1); \
- tmp1 = _mm_madd_epi16(hi_6_5, stg4_1); \
- tmp2 = _mm_madd_epi16(lo_6_5, stg4_0); \
- tmp3 = _mm_madd_epi16(hi_6_5, stg4_0); \
- \
- tmp0 = _mm_add_epi32(tmp0, rounding); \
- tmp1 = _mm_add_epi32(tmp1, rounding); \
- tmp2 = _mm_add_epi32(tmp2, rounding); \
- tmp3 = _mm_add_epi32(tmp3, rounding); \
- \
- tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
- tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
- tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
- tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
- \
- stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
- stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
- \
- stp1_8 = _mm_add_epi16(stp1_8_0, stp1_11); \
- stp1_9 = _mm_add_epi16(stp2_9, stp2_10); \
- stp1_10 = _mm_sub_epi16(stp2_9, stp2_10); \
- stp1_11 = _mm_sub_epi16(stp1_8_0, stp1_11); \
- \
- stp1_12 = _mm_sub_epi16(stp1_15, stp1_12_0); \
- stp1_13 = _mm_sub_epi16(stp2_14, stp2_13); \
- stp1_14 = _mm_add_epi16(stp2_14, stp2_13); \
- stp1_15 = _mm_add_epi16(stp1_15, stp1_12_0); \
- } \
- \
- /* Stage6 */ \
- { \
- const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
- const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
- const __m128i lo_11_12 = _mm_unpacklo_epi16(stp1_11, stp1_12); \
- const __m128i hi_11_12 = _mm_unpackhi_epi16(stp1_11, stp1_12); \
- \
- stp2_0 = _mm_add_epi16(stp1_0, stp2_7); \
- stp2_1 = _mm_add_epi16(stp1_1, stp1_6); \
- stp2_2 = _mm_add_epi16(stp1_2, stp1_5); \
- stp2_3 = _mm_add_epi16(stp1_3, stp2_4); \
- stp2_4 = _mm_sub_epi16(stp1_3, stp2_4); \
- stp2_5 = _mm_sub_epi16(stp1_2, stp1_5); \
- stp2_6 = _mm_sub_epi16(stp1_1, stp1_6); \
- stp2_7 = _mm_sub_epi16(stp1_0, stp2_7); \
- \
- MULTIPLICATION_AND_ADD(lo_10_13, hi_10_13, lo_11_12, hi_11_12, \
- stg6_0, stg4_0, stg6_0, stg4_0, \
- stp2_10, stp2_13, stp2_11, stp2_12) \
+#define IDCT16 \
+ /* Stage2 */ \
+ { \
+ const __m128i lo_1_15 = _mm_unpacklo_epi16(in[1], in[15]); \
+ const __m128i hi_1_15 = _mm_unpackhi_epi16(in[1], in[15]); \
+ const __m128i lo_9_7 = _mm_unpacklo_epi16(in[9], in[7]); \
+ const __m128i hi_9_7 = _mm_unpackhi_epi16(in[9], in[7]); \
+ const __m128i lo_5_11 = _mm_unpacklo_epi16(in[5], in[11]); \
+ const __m128i hi_5_11 = _mm_unpackhi_epi16(in[5], in[11]); \
+ const __m128i lo_13_3 = _mm_unpacklo_epi16(in[13], in[3]); \
+ const __m128i hi_13_3 = _mm_unpackhi_epi16(in[13], in[3]); \
+ \
+ MULTIPLICATION_AND_ADD(lo_1_15, hi_1_15, lo_9_7, hi_9_7, stg2_0, stg2_1, \
+ stg2_2, stg2_3, stp2_8, stp2_15, stp2_9, stp2_14) \
+ \
+ MULTIPLICATION_AND_ADD(lo_5_11, hi_5_11, lo_13_3, hi_13_3, stg2_4, stg2_5, \
+ stg2_6, stg2_7, stp2_10, stp2_13, stp2_11, stp2_12) \
+ } \
+ \
+ /* Stage3 */ \
+ { \
+ const __m128i lo_2_14 = _mm_unpacklo_epi16(in[2], in[14]); \
+ const __m128i hi_2_14 = _mm_unpackhi_epi16(in[2], in[14]); \
+ const __m128i lo_10_6 = _mm_unpacklo_epi16(in[10], in[6]); \
+ const __m128i hi_10_6 = _mm_unpackhi_epi16(in[10], in[6]); \
+ \
+ MULTIPLICATION_AND_ADD(lo_2_14, hi_2_14, lo_10_6, hi_10_6, stg3_0, stg3_1, \
+ stg3_2, stg3_3, stp1_4, stp1_7, stp1_5, stp1_6) \
+ \
+ stp1_8_0 = _mm_add_epi16(stp2_8, stp2_9); \
+ stp1_9 = _mm_sub_epi16(stp2_8, stp2_9); \
+ stp1_10 = _mm_sub_epi16(stp2_11, stp2_10); \
+ stp1_11 = _mm_add_epi16(stp2_11, stp2_10); \
+ \
+ stp1_12_0 = _mm_add_epi16(stp2_12, stp2_13); \
+ stp1_13 = _mm_sub_epi16(stp2_12, stp2_13); \
+ stp1_14 = _mm_sub_epi16(stp2_15, stp2_14); \
+ stp1_15 = _mm_add_epi16(stp2_15, stp2_14); \
+ } \
+ \
+ /* Stage4 */ \
+ { \
+ const __m128i lo_0_8 = _mm_unpacklo_epi16(in[0], in[8]); \
+ const __m128i hi_0_8 = _mm_unpackhi_epi16(in[0], in[8]); \
+ const __m128i lo_4_12 = _mm_unpacklo_epi16(in[4], in[12]); \
+ const __m128i hi_4_12 = _mm_unpackhi_epi16(in[4], in[12]); \
+ \
+ const __m128i lo_9_14 = _mm_unpacklo_epi16(stp1_9, stp1_14); \
+ const __m128i hi_9_14 = _mm_unpackhi_epi16(stp1_9, stp1_14); \
+ const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
+ const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
+ \
+ MULTIPLICATION_AND_ADD(lo_0_8, hi_0_8, lo_4_12, hi_4_12, stg4_0, stg4_1, \
+ stg4_2, stg4_3, stp2_0, stp2_1, stp2_2, stp2_3) \
+ \
+ stp2_4 = _mm_add_epi16(stp1_4, stp1_5); \
+ stp2_5 = _mm_sub_epi16(stp1_4, stp1_5); \
+ stp2_6 = _mm_sub_epi16(stp1_7, stp1_6); \
+ stp2_7 = _mm_add_epi16(stp1_7, stp1_6); \
+ \
+ MULTIPLICATION_AND_ADD(lo_9_14, hi_9_14, lo_10_13, hi_10_13, stg4_4, \
+ stg4_5, stg4_6, stg4_7, stp2_9, stp2_14, stp2_10, \
+ stp2_13) \
+ } \
+ \
+ /* Stage5 */ \
+ { \
+ const __m128i lo_6_5 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
+ const __m128i hi_6_5 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
+ \
+ stp1_0 = _mm_add_epi16(stp2_0, stp2_3); \
+ stp1_1 = _mm_add_epi16(stp2_1, stp2_2); \
+ stp1_2 = _mm_sub_epi16(stp2_1, stp2_2); \
+ stp1_3 = _mm_sub_epi16(stp2_0, stp2_3); \
+ \
+ tmp0 = _mm_madd_epi16(lo_6_5, stg4_1); \
+ tmp1 = _mm_madd_epi16(hi_6_5, stg4_1); \
+ tmp2 = _mm_madd_epi16(lo_6_5, stg4_0); \
+ tmp3 = _mm_madd_epi16(hi_6_5, stg4_0); \
+ \
+ tmp0 = _mm_add_epi32(tmp0, rounding); \
+ tmp1 = _mm_add_epi32(tmp1, rounding); \
+ tmp2 = _mm_add_epi32(tmp2, rounding); \
+ tmp3 = _mm_add_epi32(tmp3, rounding); \
+ \
+ tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
+ tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
+ tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
+ tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
+ \
+ stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
+ stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
+ \
+ stp1_8 = _mm_add_epi16(stp1_8_0, stp1_11); \
+ stp1_9 = _mm_add_epi16(stp2_9, stp2_10); \
+ stp1_10 = _mm_sub_epi16(stp2_9, stp2_10); \
+ stp1_11 = _mm_sub_epi16(stp1_8_0, stp1_11); \
+ \
+ stp1_12 = _mm_sub_epi16(stp1_15, stp1_12_0); \
+ stp1_13 = _mm_sub_epi16(stp2_14, stp2_13); \
+ stp1_14 = _mm_add_epi16(stp2_14, stp2_13); \
+ stp1_15 = _mm_add_epi16(stp1_15, stp1_12_0); \
+ } \
+ \
+ /* Stage6 */ \
+ { \
+ const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
+ const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
+ const __m128i lo_11_12 = _mm_unpacklo_epi16(stp1_11, stp1_12); \
+ const __m128i hi_11_12 = _mm_unpackhi_epi16(stp1_11, stp1_12); \
+ \
+ stp2_0 = _mm_add_epi16(stp1_0, stp2_7); \
+ stp2_1 = _mm_add_epi16(stp1_1, stp1_6); \
+ stp2_2 = _mm_add_epi16(stp1_2, stp1_5); \
+ stp2_3 = _mm_add_epi16(stp1_3, stp2_4); \
+ stp2_4 = _mm_sub_epi16(stp1_3, stp2_4); \
+ stp2_5 = _mm_sub_epi16(stp1_2, stp1_5); \
+ stp2_6 = _mm_sub_epi16(stp1_1, stp1_6); \
+ stp2_7 = _mm_sub_epi16(stp1_0, stp2_7); \
+ \
+ MULTIPLICATION_AND_ADD(lo_10_13, hi_10_13, lo_11_12, hi_11_12, stg6_0, \
+ stg4_0, stg6_0, stg4_0, stp2_10, stp2_13, stp2_11, \
+ stp2_12) \
}
-#define IDCT16_10 \
- /* Stage2 */ \
- { \
- const __m128i lo_1_15 = _mm_unpacklo_epi16(in[1], zero); \
- const __m128i hi_1_15 = _mm_unpackhi_epi16(in[1], zero); \
- const __m128i lo_13_3 = _mm_unpacklo_epi16(zero, in[3]); \
- const __m128i hi_13_3 = _mm_unpackhi_epi16(zero, in[3]); \
- \
- MULTIPLICATION_AND_ADD(lo_1_15, hi_1_15, lo_13_3, hi_13_3, \
- stg2_0, stg2_1, stg2_6, stg2_7, \
- stp1_8_0, stp1_15, stp1_11, stp1_12_0) \
- } \
- \
- /* Stage3 */ \
- { \
- const __m128i lo_2_14 = _mm_unpacklo_epi16(in[2], zero); \
- const __m128i hi_2_14 = _mm_unpackhi_epi16(in[2], zero); \
- \
- MULTIPLICATION_AND_ADD_2(lo_2_14, hi_2_14, \
- stg3_0, stg3_1, \
- stp2_4, stp2_7) \
- \
- stp1_9 = stp1_8_0; \
- stp1_10 = stp1_11; \
- \
- stp1_13 = stp1_12_0; \
- stp1_14 = stp1_15; \
- } \
- \
- /* Stage4 */ \
- { \
- const __m128i lo_0_8 = _mm_unpacklo_epi16(in[0], zero); \
- const __m128i hi_0_8 = _mm_unpackhi_epi16(in[0], zero); \
- \
- const __m128i lo_9_14 = _mm_unpacklo_epi16(stp1_9, stp1_14); \
- const __m128i hi_9_14 = _mm_unpackhi_epi16(stp1_9, stp1_14); \
- const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
- const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
- \
- MULTIPLICATION_AND_ADD_2(lo_0_8, hi_0_8, \
- stg4_0, stg4_1, \
- stp1_0, stp1_1) \
- stp2_5 = stp2_4; \
- stp2_6 = stp2_7; \
- \
- MULTIPLICATION_AND_ADD(lo_9_14, hi_9_14, lo_10_13, hi_10_13, \
- stg4_4, stg4_5, stg4_6, stg4_7, \
- stp2_9, stp2_14, stp2_10, stp2_13) \
- } \
- \
- /* Stage5 */ \
- { \
- const __m128i lo_6_5 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
- const __m128i hi_6_5 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
- \
- stp1_2 = stp1_1; \
- stp1_3 = stp1_0; \
- \
- tmp0 = _mm_madd_epi16(lo_6_5, stg4_1); \
- tmp1 = _mm_madd_epi16(hi_6_5, stg4_1); \
- tmp2 = _mm_madd_epi16(lo_6_5, stg4_0); \
- tmp3 = _mm_madd_epi16(hi_6_5, stg4_0); \
- \
- tmp0 = _mm_add_epi32(tmp0, rounding); \
- tmp1 = _mm_add_epi32(tmp1, rounding); \
- tmp2 = _mm_add_epi32(tmp2, rounding); \
- tmp3 = _mm_add_epi32(tmp3, rounding); \
- \
- tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
- tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
- tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
- tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
- \
- stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
- stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
- \
- stp1_8 = _mm_add_epi16(stp1_8_0, stp1_11); \
- stp1_9 = _mm_add_epi16(stp2_9, stp2_10); \
- stp1_10 = _mm_sub_epi16(stp2_9, stp2_10); \
- stp1_11 = _mm_sub_epi16(stp1_8_0, stp1_11); \
- \
- stp1_12 = _mm_sub_epi16(stp1_15, stp1_12_0); \
- stp1_13 = _mm_sub_epi16(stp2_14, stp2_13); \
- stp1_14 = _mm_add_epi16(stp2_14, stp2_13); \
- stp1_15 = _mm_add_epi16(stp1_15, stp1_12_0); \
- } \
- \
- /* Stage6 */ \
- { \
- const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
- const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
- const __m128i lo_11_12 = _mm_unpacklo_epi16(stp1_11, stp1_12); \
- const __m128i hi_11_12 = _mm_unpackhi_epi16(stp1_11, stp1_12); \
- \
- stp2_0 = _mm_add_epi16(stp1_0, stp2_7); \
- stp2_1 = _mm_add_epi16(stp1_1, stp1_6); \
- stp2_2 = _mm_add_epi16(stp1_2, stp1_5); \
- stp2_3 = _mm_add_epi16(stp1_3, stp2_4); \
- stp2_4 = _mm_sub_epi16(stp1_3, stp2_4); \
- stp2_5 = _mm_sub_epi16(stp1_2, stp1_5); \
- stp2_6 = _mm_sub_epi16(stp1_1, stp1_6); \
- stp2_7 = _mm_sub_epi16(stp1_0, stp2_7); \
- \
- MULTIPLICATION_AND_ADD(lo_10_13, hi_10_13, lo_11_12, hi_11_12, \
- stg6_0, stg4_0, stg6_0, stg4_0, \
- stp2_10, stp2_13, stp2_11, stp2_12) \
- }
+#define IDCT16_10 \
+ /* Stage2 */ \
+ { \
+ const __m128i lo_1_15 = _mm_unpacklo_epi16(in[1], zero); \
+ const __m128i hi_1_15 = _mm_unpackhi_epi16(in[1], zero); \
+ const __m128i lo_13_3 = _mm_unpacklo_epi16(zero, in[3]); \
+ const __m128i hi_13_3 = _mm_unpackhi_epi16(zero, in[3]); \
+ \
+ MULTIPLICATION_AND_ADD(lo_1_15, hi_1_15, lo_13_3, hi_13_3, stg2_0, stg2_1, \
+ stg2_6, stg2_7, stp1_8_0, stp1_15, stp1_11, \
+ stp1_12_0) \
+ } \
+ \
+ /* Stage3 */ \
+ { \
+ const __m128i lo_2_14 = _mm_unpacklo_epi16(in[2], zero); \
+ const __m128i hi_2_14 = _mm_unpackhi_epi16(in[2], zero); \
+ \
+ MULTIPLICATION_AND_ADD_2(lo_2_14, hi_2_14, stg3_0, stg3_1, stp2_4, stp2_7) \
+ \
+ stp1_9 = stp1_8_0; \
+ stp1_10 = stp1_11; \
+ \
+ stp1_13 = stp1_12_0; \
+ stp1_14 = stp1_15; \
+ } \
+ \
+ /* Stage4 */ \
+ { \
+ const __m128i lo_0_8 = _mm_unpacklo_epi16(in[0], zero); \
+ const __m128i hi_0_8 = _mm_unpackhi_epi16(in[0], zero); \
+ \
+ const __m128i lo_9_14 = _mm_unpacklo_epi16(stp1_9, stp1_14); \
+ const __m128i hi_9_14 = _mm_unpackhi_epi16(stp1_9, stp1_14); \
+ const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
+ const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
+ \
+ MULTIPLICATION_AND_ADD_2(lo_0_8, hi_0_8, stg4_0, stg4_1, stp1_0, stp1_1) \
+ stp2_5 = stp2_4; \
+ stp2_6 = stp2_7; \
+ \
+ MULTIPLICATION_AND_ADD(lo_9_14, hi_9_14, lo_10_13, hi_10_13, stg4_4, \
+ stg4_5, stg4_6, stg4_7, stp2_9, stp2_14, stp2_10, \
+ stp2_13) \
+ } \
+ \
+ /* Stage5 */ \
+ { \
+ const __m128i lo_6_5 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
+ const __m128i hi_6_5 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
+ \
+ stp1_2 = stp1_1; \
+ stp1_3 = stp1_0; \
+ \
+ tmp0 = _mm_madd_epi16(lo_6_5, stg4_1); \
+ tmp1 = _mm_madd_epi16(hi_6_5, stg4_1); \
+ tmp2 = _mm_madd_epi16(lo_6_5, stg4_0); \
+ tmp3 = _mm_madd_epi16(hi_6_5, stg4_0); \
+ \
+ tmp0 = _mm_add_epi32(tmp0, rounding); \
+ tmp1 = _mm_add_epi32(tmp1, rounding); \
+ tmp2 = _mm_add_epi32(tmp2, rounding); \
+ tmp3 = _mm_add_epi32(tmp3, rounding); \
+ \
+ tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
+ tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
+ tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
+ tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
+ \
+ stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
+ stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
+ \
+ stp1_8 = _mm_add_epi16(stp1_8_0, stp1_11); \
+ stp1_9 = _mm_add_epi16(stp2_9, stp2_10); \
+ stp1_10 = _mm_sub_epi16(stp2_9, stp2_10); \
+ stp1_11 = _mm_sub_epi16(stp1_8_0, stp1_11); \
+ \
+ stp1_12 = _mm_sub_epi16(stp1_15, stp1_12_0); \
+ stp1_13 = _mm_sub_epi16(stp2_14, stp2_13); \
+ stp1_14 = _mm_add_epi16(stp2_14, stp2_13); \
+ stp1_15 = _mm_add_epi16(stp1_15, stp1_12_0); \
+ } \
+ \
+ /* Stage6 */ \
+ { \
+ const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
+ const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
+ const __m128i lo_11_12 = _mm_unpacklo_epi16(stp1_11, stp1_12); \
+ const __m128i hi_11_12 = _mm_unpackhi_epi16(stp1_11, stp1_12); \
+ \
+ stp2_0 = _mm_add_epi16(stp1_0, stp2_7); \
+ stp2_1 = _mm_add_epi16(stp1_1, stp1_6); \
+ stp2_2 = _mm_add_epi16(stp1_2, stp1_5); \
+ stp2_3 = _mm_add_epi16(stp1_3, stp2_4); \
+ stp2_4 = _mm_sub_epi16(stp1_3, stp2_4); \
+ stp2_5 = _mm_sub_epi16(stp1_2, stp1_5); \
+ stp2_6 = _mm_sub_epi16(stp1_1, stp1_6); \
+ stp2_7 = _mm_sub_epi16(stp1_0, stp2_7); \
+ \
+ MULTIPLICATION_AND_ADD(lo_10_13, hi_10_13, lo_11_12, hi_11_12, stg6_0, \
+ stg4_0, stg6_0, stg4_0, stp2_10, stp2_13, stp2_11, \
+ stp2_12) \
+ }
void vp10_idct16x16_256_add_sse2(const int16_t *input, uint8_t *dest,
- int stride) {
+ int stride) {
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1 << 5);
const __m128i zero = _mm_setzero_si128();
@@ -1202,10 +1191,10 @@
__m128i in[16], l[16], r[16], *curr1;
__m128i stp1_0, stp1_1, stp1_2, stp1_3, stp1_4, stp1_5, stp1_6, stp1_7,
- stp1_8, stp1_9, stp1_10, stp1_11, stp1_12, stp1_13, stp1_14, stp1_15,
- stp1_8_0, stp1_12_0;
+ stp1_8, stp1_9, stp1_10, stp1_11, stp1_12, stp1_13, stp1_14, stp1_15,
+ stp1_8_0, stp1_12_0;
__m128i stp2_0, stp2_1, stp2_2, stp2_3, stp2_4, stp2_5, stp2_6, stp2_7,
- stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14, stp2_15;
+ stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14, stp2_15;
__m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
int i;
@@ -1294,8 +1283,7 @@
}
}
-void vp10_idct16x16_1_add_sse2(const int16_t *input,
- uint8_t *dest,
+void vp10_idct16x16_1_add_sse2(const int16_t *input, uint8_t *dest,
int stride) {
__m128i dc_value;
const __m128i zero = _mm_setzero_si128();
@@ -1308,16 +1296,16 @@
dc_value = _mm_set1_epi16(a);
for (i = 0; i < 2; ++i) {
- RECON_AND_STORE(dest + 0 * stride, dc_value);
- RECON_AND_STORE(dest + 1 * stride, dc_value);
- RECON_AND_STORE(dest + 2 * stride, dc_value);
- RECON_AND_STORE(dest + 3 * stride, dc_value);
- RECON_AND_STORE(dest + 4 * stride, dc_value);
- RECON_AND_STORE(dest + 5 * stride, dc_value);
- RECON_AND_STORE(dest + 6 * stride, dc_value);
- RECON_AND_STORE(dest + 7 * stride, dc_value);
- RECON_AND_STORE(dest + 8 * stride, dc_value);
- RECON_AND_STORE(dest + 9 * stride, dc_value);
+ RECON_AND_STORE(dest + 0 * stride, dc_value);
+ RECON_AND_STORE(dest + 1 * stride, dc_value);
+ RECON_AND_STORE(dest + 2 * stride, dc_value);
+ RECON_AND_STORE(dest + 3 * stride, dc_value);
+ RECON_AND_STORE(dest + 4 * stride, dc_value);
+ RECON_AND_STORE(dest + 5 * stride, dc_value);
+ RECON_AND_STORE(dest + 6 * stride, dc_value);
+ RECON_AND_STORE(dest + 7 * stride, dc_value);
+ RECON_AND_STORE(dest + 8 * stride, dc_value);
+ RECON_AND_STORE(dest + 9 * stride, dc_value);
RECON_AND_STORE(dest + 10 * stride, dc_value);
RECON_AND_STORE(dest + 11 * stride, dc_value);
RECON_AND_STORE(dest + 12 * stride, dc_value);
@@ -1901,9 +1889,9 @@
u[14] = _mm_srai_epi32(u[14], DCT_CONST_BITS);
u[15] = _mm_srai_epi32(u[15], DCT_CONST_BITS);
- s[8] = _mm_packs_epi32(u[0], u[1]);
+ s[8] = _mm_packs_epi32(u[0], u[1]);
s[15] = _mm_packs_epi32(u[2], u[3]);
- s[9] = _mm_packs_epi32(u[4], u[5]);
+ s[9] = _mm_packs_epi32(u[4], u[5]);
s[14] = _mm_packs_epi32(u[6], u[7]);
s[10] = _mm_packs_epi32(u[8], u[9]);
s[13] = _mm_packs_epi32(u[10], u[11]);
@@ -2031,7 +2019,7 @@
s[7] = _mm_add_epi16(t[6], t[7]);
s[8] = t[8];
s[15] = t[15];
- s[9] = _mm_packs_epi32(u[8], u[9]);
+ s[9] = _mm_packs_epi32(u[8], u[9]);
s[14] = _mm_packs_epi32(u[10], u[11]);
s[10] = _mm_packs_epi32(u[12], u[13]);
s[13] = _mm_packs_epi32(u[14], u[15]);
@@ -2155,7 +2143,7 @@
}
void vp10_idct16x16_10_add_sse2(const int16_t *input, uint8_t *dest,
- int stride) {
+ int stride) {
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1 << 5);
const __m128i zero = _mm_setzero_si128();
@@ -2177,11 +2165,11 @@
const __m128i stg6_0 = pair_set_epi16(-cospi_16_64, cospi_16_64);
__m128i in[16], l[16];
- __m128i stp1_0, stp1_1, stp1_2, stp1_3, stp1_4, stp1_5, stp1_6,
- stp1_8, stp1_9, stp1_10, stp1_11, stp1_12, stp1_13, stp1_14, stp1_15,
- stp1_8_0, stp1_12_0;
+ __m128i stp1_0, stp1_1, stp1_2, stp1_3, stp1_4, stp1_5, stp1_6, stp1_8,
+ stp1_9, stp1_10, stp1_11, stp1_12, stp1_13, stp1_14, stp1_15, stp1_8_0,
+ stp1_12_0;
__m128i stp2_0, stp2_1, stp2_2, stp2_3, stp2_4, stp2_5, stp2_6, stp2_7,
- stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14;
+ stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14;
__m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
int i;
// First 1-D inverse DCT
@@ -2213,7 +2201,7 @@
tmp5 = _mm_srai_epi32(tmp5, DCT_CONST_BITS);
tmp7 = _mm_srai_epi32(tmp7, DCT_CONST_BITS);
- stp2_8 = _mm_packs_epi32(tmp0, tmp2);
+ stp2_8 = _mm_packs_epi32(tmp0, tmp2);
stp2_11 = _mm_packs_epi32(tmp5, tmp7);
}
@@ -2277,9 +2265,9 @@
tmp2 = _mm_add_epi16(stp2_9, stp2_10);
tmp3 = _mm_sub_epi16(stp2_9, stp2_10);
- stp1_9 = _mm_unpacklo_epi64(tmp2, zero);
+ stp1_9 = _mm_unpacklo_epi64(tmp2, zero);
stp1_10 = _mm_unpacklo_epi64(tmp3, zero);
- stp1_8 = _mm_unpacklo_epi64(tmp0, zero);
+ stp1_8 = _mm_unpacklo_epi64(tmp0, zero);
stp1_11 = _mm_unpacklo_epi64(tmp1, zero);
stp1_13 = _mm_unpackhi_epi64(tmp3, zero);
@@ -2391,650 +2379,647 @@
}
}
-#define LOAD_DQCOEFF(reg, input) \
- { \
- reg = _mm_load_si128((const __m128i *) input); \
- input += 8; \
- } \
+#define LOAD_DQCOEFF(reg, input) \
+ { \
+ reg = _mm_load_si128((const __m128i *)input); \
+ input += 8; \
+ }
-#define IDCT32_34 \
-/* Stage1 */ \
-{ \
- const __m128i zero = _mm_setzero_si128();\
- const __m128i lo_1_31 = _mm_unpacklo_epi16(in[1], zero); \
- const __m128i hi_1_31 = _mm_unpackhi_epi16(in[1], zero); \
- \
- const __m128i lo_25_7= _mm_unpacklo_epi16(zero, in[7]); \
- const __m128i hi_25_7 = _mm_unpackhi_epi16(zero, in[7]); \
- \
- const __m128i lo_5_27 = _mm_unpacklo_epi16(in[5], zero); \
- const __m128i hi_5_27 = _mm_unpackhi_epi16(in[5], zero); \
- \
- const __m128i lo_29_3 = _mm_unpacklo_epi16(zero, in[3]); \
- const __m128i hi_29_3 = _mm_unpackhi_epi16(zero, in[3]); \
- \
- MULTIPLICATION_AND_ADD_2(lo_1_31, hi_1_31, stg1_0, \
- stg1_1, stp1_16, stp1_31); \
- MULTIPLICATION_AND_ADD_2(lo_25_7, hi_25_7, stg1_6, \
- stg1_7, stp1_19, stp1_28); \
- MULTIPLICATION_AND_ADD_2(lo_5_27, hi_5_27, stg1_8, \
- stg1_9, stp1_20, stp1_27); \
- MULTIPLICATION_AND_ADD_2(lo_29_3, hi_29_3, stg1_14, \
- stg1_15, stp1_23, stp1_24); \
-} \
-\
-/* Stage2 */ \
-{ \
- const __m128i zero = _mm_setzero_si128();\
- const __m128i lo_2_30 = _mm_unpacklo_epi16(in[2], zero); \
- const __m128i hi_2_30 = _mm_unpackhi_epi16(in[2], zero); \
- \
- const __m128i lo_26_6 = _mm_unpacklo_epi16(zero, in[6]); \
- const __m128i hi_26_6 = _mm_unpackhi_epi16(zero, in[6]); \
- \
- MULTIPLICATION_AND_ADD_2(lo_2_30, hi_2_30, stg2_0, \
- stg2_1, stp2_8, stp2_15); \
- MULTIPLICATION_AND_ADD_2(lo_26_6, hi_26_6, stg2_6, \
- stg2_7, stp2_11, stp2_12); \
- \
- stp2_16 = stp1_16; \
- stp2_19 = stp1_19; \
- \
- stp2_20 = stp1_20; \
- stp2_23 = stp1_23; \
- \
- stp2_24 = stp1_24; \
- stp2_27 = stp1_27; \
- \
- stp2_28 = stp1_28; \
- stp2_31 = stp1_31; \
-} \
-\
-/* Stage3 */ \
-{ \
- const __m128i zero = _mm_setzero_si128();\
- const __m128i lo_4_28 = _mm_unpacklo_epi16(in[4], zero); \
- const __m128i hi_4_28 = _mm_unpackhi_epi16(in[4], zero); \
- \
- const __m128i lo_17_30 = _mm_unpacklo_epi16(stp1_16, stp1_31); \
- const __m128i hi_17_30 = _mm_unpackhi_epi16(stp1_16, stp1_31); \
- const __m128i lo_18_29 = _mm_unpacklo_epi16(stp1_19, stp1_28); \
- const __m128i hi_18_29 = _mm_unpackhi_epi16(stp1_19, stp1_28); \
- \
- const __m128i lo_21_26 = _mm_unpacklo_epi16(stp1_20, stp1_27); \
- const __m128i hi_21_26 = _mm_unpackhi_epi16(stp1_20, stp1_27); \
- const __m128i lo_22_25 = _mm_unpacklo_epi16(stp1_23, stp1_24); \
- const __m128i hi_22_25 = _mm_unpackhi_epi16(stp1_23, stp2_24); \
- \
- MULTIPLICATION_AND_ADD_2(lo_4_28, hi_4_28, stg3_0, \
- stg3_1, stp1_4, stp1_7); \
- \
- stp1_8 = stp2_8; \
- stp1_11 = stp2_11; \
- stp1_12 = stp2_12; \
- stp1_15 = stp2_15; \
- \
- MULTIPLICATION_AND_ADD(lo_17_30, hi_17_30, lo_18_29, hi_18_29, stg3_4, \
- stg3_5, stg3_6, stg3_4, stp1_17, stp1_30, \
- stp1_18, stp1_29) \
- MULTIPLICATION_AND_ADD(lo_21_26, hi_21_26, lo_22_25, hi_22_25, stg3_8, \
- stg3_9, stg3_10, stg3_8, stp1_21, stp1_26, \
- stp1_22, stp1_25) \
- \
- stp1_16 = stp2_16; \
- stp1_31 = stp2_31; \
- stp1_19 = stp2_19; \
- stp1_20 = stp2_20; \
- stp1_23 = stp2_23; \
- stp1_24 = stp2_24; \
- stp1_27 = stp2_27; \
- stp1_28 = stp2_28; \
-} \
-\
-/* Stage4 */ \
-{ \
- const __m128i zero = _mm_setzero_si128();\
- const __m128i lo_0_16 = _mm_unpacklo_epi16(in[0], zero); \
- const __m128i hi_0_16 = _mm_unpackhi_epi16(in[0], zero); \
- \
- const __m128i lo_9_14 = _mm_unpacklo_epi16(stp2_8, stp2_15); \
- const __m128i hi_9_14 = _mm_unpackhi_epi16(stp2_8, stp2_15); \
- const __m128i lo_10_13 = _mm_unpacklo_epi16(stp2_11, stp2_12); \
- const __m128i hi_10_13 = _mm_unpackhi_epi16(stp2_11, stp2_12); \
- \
- MULTIPLICATION_AND_ADD_2(lo_0_16, hi_0_16, stg4_0, \
- stg4_1, stp2_0, stp2_1); \
- \
- stp2_4 = stp1_4; \
- stp2_5 = stp1_4; \
- stp2_6 = stp1_7; \
- stp2_7 = stp1_7; \
- \
- MULTIPLICATION_AND_ADD(lo_9_14, hi_9_14, lo_10_13, hi_10_13, stg4_4, \
- stg4_5, stg4_6, stg4_4, stp2_9, stp2_14, \
- stp2_10, stp2_13) \
- \
- stp2_8 = stp1_8; \
- stp2_15 = stp1_15; \
- stp2_11 = stp1_11; \
- stp2_12 = stp1_12; \
- \
- stp2_16 = _mm_add_epi16(stp1_16, stp1_19); \
- stp2_17 = _mm_add_epi16(stp1_17, stp1_18); \
- stp2_18 = _mm_sub_epi16(stp1_17, stp1_18); \
- stp2_19 = _mm_sub_epi16(stp1_16, stp1_19); \
- stp2_20 = _mm_sub_epi16(stp1_23, stp1_20); \
- stp2_21 = _mm_sub_epi16(stp1_22, stp1_21); \
- stp2_22 = _mm_add_epi16(stp1_22, stp1_21); \
- stp2_23 = _mm_add_epi16(stp1_23, stp1_20); \
- \
- stp2_24 = _mm_add_epi16(stp1_24, stp1_27); \
- stp2_25 = _mm_add_epi16(stp1_25, stp1_26); \
- stp2_26 = _mm_sub_epi16(stp1_25, stp1_26); \
- stp2_27 = _mm_sub_epi16(stp1_24, stp1_27); \
- stp2_28 = _mm_sub_epi16(stp1_31, stp1_28); \
- stp2_29 = _mm_sub_epi16(stp1_30, stp1_29); \
- stp2_30 = _mm_add_epi16(stp1_29, stp1_30); \
- stp2_31 = _mm_add_epi16(stp1_28, stp1_31); \
-} \
-\
-/* Stage5 */ \
-{ \
- const __m128i lo_6_5 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
- const __m128i hi_6_5 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
- const __m128i lo_18_29 = _mm_unpacklo_epi16(stp2_18, stp2_29); \
- const __m128i hi_18_29 = _mm_unpackhi_epi16(stp2_18, stp2_29); \
- \
- const __m128i lo_19_28 = _mm_unpacklo_epi16(stp2_19, stp2_28); \
- const __m128i hi_19_28 = _mm_unpackhi_epi16(stp2_19, stp2_28); \
- const __m128i lo_20_27 = _mm_unpacklo_epi16(stp2_20, stp2_27); \
- const __m128i hi_20_27 = _mm_unpackhi_epi16(stp2_20, stp2_27); \
- \
- const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
- const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
- \
- stp1_0 = stp2_0; \
- stp1_1 = stp2_1; \
- stp1_2 = stp2_1; \
- stp1_3 = stp2_0; \
- \
- tmp0 = _mm_madd_epi16(lo_6_5, stg4_1); \
- tmp1 = _mm_madd_epi16(hi_6_5, stg4_1); \
- tmp2 = _mm_madd_epi16(lo_6_5, stg4_0); \
- tmp3 = _mm_madd_epi16(hi_6_5, stg4_0); \
- \
- tmp0 = _mm_add_epi32(tmp0, rounding); \
- tmp1 = _mm_add_epi32(tmp1, rounding); \
- tmp2 = _mm_add_epi32(tmp2, rounding); \
- tmp3 = _mm_add_epi32(tmp3, rounding); \
- \
- tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
- tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
- tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
- tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
- \
- stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
- stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
- \
- stp1_4 = stp2_4; \
- stp1_7 = stp2_7; \
- \
- stp1_8 = _mm_add_epi16(stp2_8, stp2_11); \
- stp1_9 = _mm_add_epi16(stp2_9, stp2_10); \
- stp1_10 = _mm_sub_epi16(stp2_9, stp2_10); \
- stp1_11 = _mm_sub_epi16(stp2_8, stp2_11); \
- stp1_12 = _mm_sub_epi16(stp2_15, stp2_12); \
- stp1_13 = _mm_sub_epi16(stp2_14, stp2_13); \
- stp1_14 = _mm_add_epi16(stp2_14, stp2_13); \
- stp1_15 = _mm_add_epi16(stp2_15, stp2_12); \
- \
- stp1_16 = stp2_16; \
- stp1_17 = stp2_17; \
- \
- MULTIPLICATION_AND_ADD(lo_18_29, hi_18_29, lo_19_28, hi_19_28, stg4_4, \
- stg4_5, stg4_4, stg4_5, stp1_18, stp1_29, \
- stp1_19, stp1_28) \
- MULTIPLICATION_AND_ADD(lo_20_27, hi_20_27, lo_21_26, hi_21_26, stg4_6, \
- stg4_4, stg4_6, stg4_4, stp1_20, stp1_27, \
- stp1_21, stp1_26) \
- \
- stp1_22 = stp2_22; \
- stp1_23 = stp2_23; \
- stp1_24 = stp2_24; \
- stp1_25 = stp2_25; \
- stp1_30 = stp2_30; \
- stp1_31 = stp2_31; \
-} \
-\
-/* Stage6 */ \
-{ \
- const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
- const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
- const __m128i lo_11_12 = _mm_unpacklo_epi16(stp1_11, stp1_12); \
- const __m128i hi_11_12 = _mm_unpackhi_epi16(stp1_11, stp1_12); \
- \
- stp2_0 = _mm_add_epi16(stp1_0, stp1_7); \
- stp2_1 = _mm_add_epi16(stp1_1, stp1_6); \
- stp2_2 = _mm_add_epi16(stp1_2, stp1_5); \
- stp2_3 = _mm_add_epi16(stp1_3, stp1_4); \
- stp2_4 = _mm_sub_epi16(stp1_3, stp1_4); \
- stp2_5 = _mm_sub_epi16(stp1_2, stp1_5); \
- stp2_6 = _mm_sub_epi16(stp1_1, stp1_6); \
- stp2_7 = _mm_sub_epi16(stp1_0, stp1_7); \
- \
- stp2_8 = stp1_8; \
- stp2_9 = stp1_9; \
- stp2_14 = stp1_14; \
- stp2_15 = stp1_15; \
- \
- MULTIPLICATION_AND_ADD(lo_10_13, hi_10_13, lo_11_12, hi_11_12, \
- stg6_0, stg4_0, stg6_0, stg4_0, stp2_10, \
- stp2_13, stp2_11, stp2_12) \
- \
- stp2_16 = _mm_add_epi16(stp1_16, stp1_23); \
- stp2_17 = _mm_add_epi16(stp1_17, stp1_22); \
- stp2_18 = _mm_add_epi16(stp1_18, stp1_21); \
- stp2_19 = _mm_add_epi16(stp1_19, stp1_20); \
- stp2_20 = _mm_sub_epi16(stp1_19, stp1_20); \
- stp2_21 = _mm_sub_epi16(stp1_18, stp1_21); \
- stp2_22 = _mm_sub_epi16(stp1_17, stp1_22); \
- stp2_23 = _mm_sub_epi16(stp1_16, stp1_23); \
- \
- stp2_24 = _mm_sub_epi16(stp1_31, stp1_24); \
- stp2_25 = _mm_sub_epi16(stp1_30, stp1_25); \
- stp2_26 = _mm_sub_epi16(stp1_29, stp1_26); \
- stp2_27 = _mm_sub_epi16(stp1_28, stp1_27); \
- stp2_28 = _mm_add_epi16(stp1_27, stp1_28); \
- stp2_29 = _mm_add_epi16(stp1_26, stp1_29); \
- stp2_30 = _mm_add_epi16(stp1_25, stp1_30); \
- stp2_31 = _mm_add_epi16(stp1_24, stp1_31); \
-} \
-\
-/* Stage7 */ \
-{ \
- const __m128i lo_20_27 = _mm_unpacklo_epi16(stp2_20, stp2_27); \
- const __m128i hi_20_27 = _mm_unpackhi_epi16(stp2_20, stp2_27); \
- const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
- const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
- \
- const __m128i lo_22_25 = _mm_unpacklo_epi16(stp2_22, stp2_25); \
- const __m128i hi_22_25 = _mm_unpackhi_epi16(stp2_22, stp2_25); \
- const __m128i lo_23_24 = _mm_unpacklo_epi16(stp2_23, stp2_24); \
- const __m128i hi_23_24 = _mm_unpackhi_epi16(stp2_23, stp2_24); \
- \
- stp1_0 = _mm_add_epi16(stp2_0, stp2_15); \
- stp1_1 = _mm_add_epi16(stp2_1, stp2_14); \
- stp1_2 = _mm_add_epi16(stp2_2, stp2_13); \
- stp1_3 = _mm_add_epi16(stp2_3, stp2_12); \
- stp1_4 = _mm_add_epi16(stp2_4, stp2_11); \
- stp1_5 = _mm_add_epi16(stp2_5, stp2_10); \
- stp1_6 = _mm_add_epi16(stp2_6, stp2_9); \
- stp1_7 = _mm_add_epi16(stp2_7, stp2_8); \
- stp1_8 = _mm_sub_epi16(stp2_7, stp2_8); \
- stp1_9 = _mm_sub_epi16(stp2_6, stp2_9); \
- stp1_10 = _mm_sub_epi16(stp2_5, stp2_10); \
- stp1_11 = _mm_sub_epi16(stp2_4, stp2_11); \
- stp1_12 = _mm_sub_epi16(stp2_3, stp2_12); \
- stp1_13 = _mm_sub_epi16(stp2_2, stp2_13); \
- stp1_14 = _mm_sub_epi16(stp2_1, stp2_14); \
- stp1_15 = _mm_sub_epi16(stp2_0, stp2_15); \
- \
- stp1_16 = stp2_16; \
- stp1_17 = stp2_17; \
- stp1_18 = stp2_18; \
- stp1_19 = stp2_19; \
- \
- MULTIPLICATION_AND_ADD(lo_20_27, hi_20_27, lo_21_26, hi_21_26, stg6_0, \
- stg4_0, stg6_0, stg4_0, stp1_20, stp1_27, \
- stp1_21, stp1_26) \
- MULTIPLICATION_AND_ADD(lo_22_25, hi_22_25, lo_23_24, hi_23_24, stg6_0, \
- stg4_0, stg6_0, stg4_0, stp1_22, stp1_25, \
- stp1_23, stp1_24) \
- \
- stp1_28 = stp2_28; \
- stp1_29 = stp2_29; \
- stp1_30 = stp2_30; \
- stp1_31 = stp2_31; \
-}
+#define IDCT32_34 \
+ /* Stage1 */ \
+ { \
+ const __m128i zero = _mm_setzero_si128(); \
+ const __m128i lo_1_31 = _mm_unpacklo_epi16(in[1], zero); \
+ const __m128i hi_1_31 = _mm_unpackhi_epi16(in[1], zero); \
+ \
+ const __m128i lo_25_7 = _mm_unpacklo_epi16(zero, in[7]); \
+ const __m128i hi_25_7 = _mm_unpackhi_epi16(zero, in[7]); \
+ \
+ const __m128i lo_5_27 = _mm_unpacklo_epi16(in[5], zero); \
+ const __m128i hi_5_27 = _mm_unpackhi_epi16(in[5], zero); \
+ \
+ const __m128i lo_29_3 = _mm_unpacklo_epi16(zero, in[3]); \
+ const __m128i hi_29_3 = _mm_unpackhi_epi16(zero, in[3]); \
+ \
+ MULTIPLICATION_AND_ADD_2(lo_1_31, hi_1_31, stg1_0, stg1_1, stp1_16, \
+ stp1_31); \
+ MULTIPLICATION_AND_ADD_2(lo_25_7, hi_25_7, stg1_6, stg1_7, stp1_19, \
+ stp1_28); \
+ MULTIPLICATION_AND_ADD_2(lo_5_27, hi_5_27, stg1_8, stg1_9, stp1_20, \
+ stp1_27); \
+ MULTIPLICATION_AND_ADD_2(lo_29_3, hi_29_3, stg1_14, stg1_15, stp1_23, \
+ stp1_24); \
+ } \
+ \
+ /* Stage2 */ \
+ { \
+ const __m128i zero = _mm_setzero_si128(); \
+ const __m128i lo_2_30 = _mm_unpacklo_epi16(in[2], zero); \
+ const __m128i hi_2_30 = _mm_unpackhi_epi16(in[2], zero); \
+ \
+ const __m128i lo_26_6 = _mm_unpacklo_epi16(zero, in[6]); \
+ const __m128i hi_26_6 = _mm_unpackhi_epi16(zero, in[6]); \
+ \
+ MULTIPLICATION_AND_ADD_2(lo_2_30, hi_2_30, stg2_0, stg2_1, stp2_8, \
+ stp2_15); \
+ MULTIPLICATION_AND_ADD_2(lo_26_6, hi_26_6, stg2_6, stg2_7, stp2_11, \
+ stp2_12); \
+ \
+ stp2_16 = stp1_16; \
+ stp2_19 = stp1_19; \
+ \
+ stp2_20 = stp1_20; \
+ stp2_23 = stp1_23; \
+ \
+ stp2_24 = stp1_24; \
+ stp2_27 = stp1_27; \
+ \
+ stp2_28 = stp1_28; \
+ stp2_31 = stp1_31; \
+ } \
+ \
+ /* Stage3 */ \
+ { \
+ const __m128i zero = _mm_setzero_si128(); \
+ const __m128i lo_4_28 = _mm_unpacklo_epi16(in[4], zero); \
+ const __m128i hi_4_28 = _mm_unpackhi_epi16(in[4], zero); \
+ \
+ const __m128i lo_17_30 = _mm_unpacklo_epi16(stp1_16, stp1_31); \
+ const __m128i hi_17_30 = _mm_unpackhi_epi16(stp1_16, stp1_31); \
+ const __m128i lo_18_29 = _mm_unpacklo_epi16(stp1_19, stp1_28); \
+ const __m128i hi_18_29 = _mm_unpackhi_epi16(stp1_19, stp1_28); \
+ \
+ const __m128i lo_21_26 = _mm_unpacklo_epi16(stp1_20, stp1_27); \
+ const __m128i hi_21_26 = _mm_unpackhi_epi16(stp1_20, stp1_27); \
+ const __m128i lo_22_25 = _mm_unpacklo_epi16(stp1_23, stp1_24); \
+ const __m128i hi_22_25 = _mm_unpackhi_epi16(stp1_23, stp2_24); \
+ \
+ MULTIPLICATION_AND_ADD_2(lo_4_28, hi_4_28, stg3_0, stg3_1, stp1_4, \
+ stp1_7); \
+ \
+ stp1_8 = stp2_8; \
+ stp1_11 = stp2_11; \
+ stp1_12 = stp2_12; \
+ stp1_15 = stp2_15; \
+ \
+ MULTIPLICATION_AND_ADD(lo_17_30, hi_17_30, lo_18_29, hi_18_29, stg3_4, \
+ stg3_5, stg3_6, stg3_4, stp1_17, stp1_30, stp1_18, \
+ stp1_29) \
+ MULTIPLICATION_AND_ADD(lo_21_26, hi_21_26, lo_22_25, hi_22_25, stg3_8, \
+ stg3_9, stg3_10, stg3_8, stp1_21, stp1_26, stp1_22, \
+ stp1_25) \
+ \
+ stp1_16 = stp2_16; \
+ stp1_31 = stp2_31; \
+ stp1_19 = stp2_19; \
+ stp1_20 = stp2_20; \
+ stp1_23 = stp2_23; \
+ stp1_24 = stp2_24; \
+ stp1_27 = stp2_27; \
+ stp1_28 = stp2_28; \
+ } \
+ \
+ /* Stage4 */ \
+ { \
+ const __m128i zero = _mm_setzero_si128(); \
+ const __m128i lo_0_16 = _mm_unpacklo_epi16(in[0], zero); \
+ const __m128i hi_0_16 = _mm_unpackhi_epi16(in[0], zero); \
+ \
+ const __m128i lo_9_14 = _mm_unpacklo_epi16(stp2_8, stp2_15); \
+ const __m128i hi_9_14 = _mm_unpackhi_epi16(stp2_8, stp2_15); \
+ const __m128i lo_10_13 = _mm_unpacklo_epi16(stp2_11, stp2_12); \
+ const __m128i hi_10_13 = _mm_unpackhi_epi16(stp2_11, stp2_12); \
+ \
+ MULTIPLICATION_AND_ADD_2(lo_0_16, hi_0_16, stg4_0, stg4_1, stp2_0, \
+ stp2_1); \
+ \
+ stp2_4 = stp1_4; \
+ stp2_5 = stp1_4; \
+ stp2_6 = stp1_7; \
+ stp2_7 = stp1_7; \
+ \
+ MULTIPLICATION_AND_ADD(lo_9_14, hi_9_14, lo_10_13, hi_10_13, stg4_4, \
+ stg4_5, stg4_6, stg4_4, stp2_9, stp2_14, stp2_10, \
+ stp2_13) \
+ \
+ stp2_8 = stp1_8; \
+ stp2_15 = stp1_15; \
+ stp2_11 = stp1_11; \
+ stp2_12 = stp1_12; \
+ \
+ stp2_16 = _mm_add_epi16(stp1_16, stp1_19); \
+ stp2_17 = _mm_add_epi16(stp1_17, stp1_18); \
+ stp2_18 = _mm_sub_epi16(stp1_17, stp1_18); \
+ stp2_19 = _mm_sub_epi16(stp1_16, stp1_19); \
+ stp2_20 = _mm_sub_epi16(stp1_23, stp1_20); \
+ stp2_21 = _mm_sub_epi16(stp1_22, stp1_21); \
+ stp2_22 = _mm_add_epi16(stp1_22, stp1_21); \
+ stp2_23 = _mm_add_epi16(stp1_23, stp1_20); \
+ \
+ stp2_24 = _mm_add_epi16(stp1_24, stp1_27); \
+ stp2_25 = _mm_add_epi16(stp1_25, stp1_26); \
+ stp2_26 = _mm_sub_epi16(stp1_25, stp1_26); \
+ stp2_27 = _mm_sub_epi16(stp1_24, stp1_27); \
+ stp2_28 = _mm_sub_epi16(stp1_31, stp1_28); \
+ stp2_29 = _mm_sub_epi16(stp1_30, stp1_29); \
+ stp2_30 = _mm_add_epi16(stp1_29, stp1_30); \
+ stp2_31 = _mm_add_epi16(stp1_28, stp1_31); \
+ } \
+ \
+ /* Stage5 */ \
+ { \
+ const __m128i lo_6_5 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
+ const __m128i hi_6_5 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
+ const __m128i lo_18_29 = _mm_unpacklo_epi16(stp2_18, stp2_29); \
+ const __m128i hi_18_29 = _mm_unpackhi_epi16(stp2_18, stp2_29); \
+ \
+ const __m128i lo_19_28 = _mm_unpacklo_epi16(stp2_19, stp2_28); \
+ const __m128i hi_19_28 = _mm_unpackhi_epi16(stp2_19, stp2_28); \
+ const __m128i lo_20_27 = _mm_unpacklo_epi16(stp2_20, stp2_27); \
+ const __m128i hi_20_27 = _mm_unpackhi_epi16(stp2_20, stp2_27); \
+ \
+ const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
+ const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
+ \
+ stp1_0 = stp2_0; \
+ stp1_1 = stp2_1; \
+ stp1_2 = stp2_1; \
+ stp1_3 = stp2_0; \
+ \
+ tmp0 = _mm_madd_epi16(lo_6_5, stg4_1); \
+ tmp1 = _mm_madd_epi16(hi_6_5, stg4_1); \
+ tmp2 = _mm_madd_epi16(lo_6_5, stg4_0); \
+ tmp3 = _mm_madd_epi16(hi_6_5, stg4_0); \
+ \
+ tmp0 = _mm_add_epi32(tmp0, rounding); \
+ tmp1 = _mm_add_epi32(tmp1, rounding); \
+ tmp2 = _mm_add_epi32(tmp2, rounding); \
+ tmp3 = _mm_add_epi32(tmp3, rounding); \
+ \
+ tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
+ tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
+ tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
+ tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
+ \
+ stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
+ stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
+ \
+ stp1_4 = stp2_4; \
+ stp1_7 = stp2_7; \
+ \
+ stp1_8 = _mm_add_epi16(stp2_8, stp2_11); \
+ stp1_9 = _mm_add_epi16(stp2_9, stp2_10); \
+ stp1_10 = _mm_sub_epi16(stp2_9, stp2_10); \
+ stp1_11 = _mm_sub_epi16(stp2_8, stp2_11); \
+ stp1_12 = _mm_sub_epi16(stp2_15, stp2_12); \
+ stp1_13 = _mm_sub_epi16(stp2_14, stp2_13); \
+ stp1_14 = _mm_add_epi16(stp2_14, stp2_13); \
+ stp1_15 = _mm_add_epi16(stp2_15, stp2_12); \
+ \
+ stp1_16 = stp2_16; \
+ stp1_17 = stp2_17; \
+ \
+ MULTIPLICATION_AND_ADD(lo_18_29, hi_18_29, lo_19_28, hi_19_28, stg4_4, \
+ stg4_5, stg4_4, stg4_5, stp1_18, stp1_29, stp1_19, \
+ stp1_28) \
+ MULTIPLICATION_AND_ADD(lo_20_27, hi_20_27, lo_21_26, hi_21_26, stg4_6, \
+ stg4_4, stg4_6, stg4_4, stp1_20, stp1_27, stp1_21, \
+ stp1_26) \
+ \
+ stp1_22 = stp2_22; \
+ stp1_23 = stp2_23; \
+ stp1_24 = stp2_24; \
+ stp1_25 = stp2_25; \
+ stp1_30 = stp2_30; \
+ stp1_31 = stp2_31; \
+ } \
+ \
+ /* Stage6 */ \
+ { \
+ const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
+ const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
+ const __m128i lo_11_12 = _mm_unpacklo_epi16(stp1_11, stp1_12); \
+ const __m128i hi_11_12 = _mm_unpackhi_epi16(stp1_11, stp1_12); \
+ \
+ stp2_0 = _mm_add_epi16(stp1_0, stp1_7); \
+ stp2_1 = _mm_add_epi16(stp1_1, stp1_6); \
+ stp2_2 = _mm_add_epi16(stp1_2, stp1_5); \
+ stp2_3 = _mm_add_epi16(stp1_3, stp1_4); \
+ stp2_4 = _mm_sub_epi16(stp1_3, stp1_4); \
+ stp2_5 = _mm_sub_epi16(stp1_2, stp1_5); \
+ stp2_6 = _mm_sub_epi16(stp1_1, stp1_6); \
+ stp2_7 = _mm_sub_epi16(stp1_0, stp1_7); \
+ \
+ stp2_8 = stp1_8; \
+ stp2_9 = stp1_9; \
+ stp2_14 = stp1_14; \
+ stp2_15 = stp1_15; \
+ \
+ MULTIPLICATION_AND_ADD(lo_10_13, hi_10_13, lo_11_12, hi_11_12, stg6_0, \
+ stg4_0, stg6_0, stg4_0, stp2_10, stp2_13, stp2_11, \
+ stp2_12) \
+ \
+ stp2_16 = _mm_add_epi16(stp1_16, stp1_23); \
+ stp2_17 = _mm_add_epi16(stp1_17, stp1_22); \
+ stp2_18 = _mm_add_epi16(stp1_18, stp1_21); \
+ stp2_19 = _mm_add_epi16(stp1_19, stp1_20); \
+ stp2_20 = _mm_sub_epi16(stp1_19, stp1_20); \
+ stp2_21 = _mm_sub_epi16(stp1_18, stp1_21); \
+ stp2_22 = _mm_sub_epi16(stp1_17, stp1_22); \
+ stp2_23 = _mm_sub_epi16(stp1_16, stp1_23); \
+ \
+ stp2_24 = _mm_sub_epi16(stp1_31, stp1_24); \
+ stp2_25 = _mm_sub_epi16(stp1_30, stp1_25); \
+ stp2_26 = _mm_sub_epi16(stp1_29, stp1_26); \
+ stp2_27 = _mm_sub_epi16(stp1_28, stp1_27); \
+ stp2_28 = _mm_add_epi16(stp1_27, stp1_28); \
+ stp2_29 = _mm_add_epi16(stp1_26, stp1_29); \
+ stp2_30 = _mm_add_epi16(stp1_25, stp1_30); \
+ stp2_31 = _mm_add_epi16(stp1_24, stp1_31); \
+ } \
+ \
+ /* Stage7 */ \
+ { \
+ const __m128i lo_20_27 = _mm_unpacklo_epi16(stp2_20, stp2_27); \
+ const __m128i hi_20_27 = _mm_unpackhi_epi16(stp2_20, stp2_27); \
+ const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
+ const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
+ \
+ const __m128i lo_22_25 = _mm_unpacklo_epi16(stp2_22, stp2_25); \
+ const __m128i hi_22_25 = _mm_unpackhi_epi16(stp2_22, stp2_25); \
+ const __m128i lo_23_24 = _mm_unpacklo_epi16(stp2_23, stp2_24); \
+ const __m128i hi_23_24 = _mm_unpackhi_epi16(stp2_23, stp2_24); \
+ \
+ stp1_0 = _mm_add_epi16(stp2_0, stp2_15); \
+ stp1_1 = _mm_add_epi16(stp2_1, stp2_14); \
+ stp1_2 = _mm_add_epi16(stp2_2, stp2_13); \
+ stp1_3 = _mm_add_epi16(stp2_3, stp2_12); \
+ stp1_4 = _mm_add_epi16(stp2_4, stp2_11); \
+ stp1_5 = _mm_add_epi16(stp2_5, stp2_10); \
+ stp1_6 = _mm_add_epi16(stp2_6, stp2_9); \
+ stp1_7 = _mm_add_epi16(stp2_7, stp2_8); \
+ stp1_8 = _mm_sub_epi16(stp2_7, stp2_8); \
+ stp1_9 = _mm_sub_epi16(stp2_6, stp2_9); \
+ stp1_10 = _mm_sub_epi16(stp2_5, stp2_10); \
+ stp1_11 = _mm_sub_epi16(stp2_4, stp2_11); \
+ stp1_12 = _mm_sub_epi16(stp2_3, stp2_12); \
+ stp1_13 = _mm_sub_epi16(stp2_2, stp2_13); \
+ stp1_14 = _mm_sub_epi16(stp2_1, stp2_14); \
+ stp1_15 = _mm_sub_epi16(stp2_0, stp2_15); \
+ \
+ stp1_16 = stp2_16; \
+ stp1_17 = stp2_17; \
+ stp1_18 = stp2_18; \
+ stp1_19 = stp2_19; \
+ \
+ MULTIPLICATION_AND_ADD(lo_20_27, hi_20_27, lo_21_26, hi_21_26, stg6_0, \
+ stg4_0, stg6_0, stg4_0, stp1_20, stp1_27, stp1_21, \
+ stp1_26) \
+ MULTIPLICATION_AND_ADD(lo_22_25, hi_22_25, lo_23_24, hi_23_24, stg6_0, \
+ stg4_0, stg6_0, stg4_0, stp1_22, stp1_25, stp1_23, \
+ stp1_24) \
+ \
+ stp1_28 = stp2_28; \
+ stp1_29 = stp2_29; \
+ stp1_30 = stp2_30; \
+ stp1_31 = stp2_31; \
+ }
-
-#define IDCT32 \
-/* Stage1 */ \
-{ \
- const __m128i lo_1_31 = _mm_unpacklo_epi16(in[1], in[31]); \
- const __m128i hi_1_31 = _mm_unpackhi_epi16(in[1], in[31]); \
- const __m128i lo_17_15 = _mm_unpacklo_epi16(in[17], in[15]); \
- const __m128i hi_17_15 = _mm_unpackhi_epi16(in[17], in[15]); \
- \
- const __m128i lo_9_23 = _mm_unpacklo_epi16(in[9], in[23]); \
- const __m128i hi_9_23 = _mm_unpackhi_epi16(in[9], in[23]); \
- const __m128i lo_25_7= _mm_unpacklo_epi16(in[25], in[7]); \
- const __m128i hi_25_7 = _mm_unpackhi_epi16(in[25], in[7]); \
- \
- const __m128i lo_5_27 = _mm_unpacklo_epi16(in[5], in[27]); \
- const __m128i hi_5_27 = _mm_unpackhi_epi16(in[5], in[27]); \
- const __m128i lo_21_11 = _mm_unpacklo_epi16(in[21], in[11]); \
- const __m128i hi_21_11 = _mm_unpackhi_epi16(in[21], in[11]); \
- \
- const __m128i lo_13_19 = _mm_unpacklo_epi16(in[13], in[19]); \
- const __m128i hi_13_19 = _mm_unpackhi_epi16(in[13], in[19]); \
- const __m128i lo_29_3 = _mm_unpacklo_epi16(in[29], in[3]); \
- const __m128i hi_29_3 = _mm_unpackhi_epi16(in[29], in[3]); \
- \
- MULTIPLICATION_AND_ADD(lo_1_31, hi_1_31, lo_17_15, hi_17_15, stg1_0, \
- stg1_1, stg1_2, stg1_3, stp1_16, stp1_31, \
- stp1_17, stp1_30) \
- MULTIPLICATION_AND_ADD(lo_9_23, hi_9_23, lo_25_7, hi_25_7, stg1_4, \
- stg1_5, stg1_6, stg1_7, stp1_18, stp1_29, \
- stp1_19, stp1_28) \
- MULTIPLICATION_AND_ADD(lo_5_27, hi_5_27, lo_21_11, hi_21_11, stg1_8, \
- stg1_9, stg1_10, stg1_11, stp1_20, stp1_27, \
- stp1_21, stp1_26) \
- MULTIPLICATION_AND_ADD(lo_13_19, hi_13_19, lo_29_3, hi_29_3, stg1_12, \
- stg1_13, stg1_14, stg1_15, stp1_22, stp1_25, \
- stp1_23, stp1_24) \
-} \
-\
-/* Stage2 */ \
-{ \
- const __m128i lo_2_30 = _mm_unpacklo_epi16(in[2], in[30]); \
- const __m128i hi_2_30 = _mm_unpackhi_epi16(in[2], in[30]); \
- const __m128i lo_18_14 = _mm_unpacklo_epi16(in[18], in[14]); \
- const __m128i hi_18_14 = _mm_unpackhi_epi16(in[18], in[14]); \
- \
- const __m128i lo_10_22 = _mm_unpacklo_epi16(in[10], in[22]); \
- const __m128i hi_10_22 = _mm_unpackhi_epi16(in[10], in[22]); \
- const __m128i lo_26_6 = _mm_unpacklo_epi16(in[26], in[6]); \
- const __m128i hi_26_6 = _mm_unpackhi_epi16(in[26], in[6]); \
- \
- MULTIPLICATION_AND_ADD(lo_2_30, hi_2_30, lo_18_14, hi_18_14, stg2_0, \
- stg2_1, stg2_2, stg2_3, stp2_8, stp2_15, stp2_9, \
- stp2_14) \
- MULTIPLICATION_AND_ADD(lo_10_22, hi_10_22, lo_26_6, hi_26_6, stg2_4, \
- stg2_5, stg2_6, stg2_7, stp2_10, stp2_13, \
- stp2_11, stp2_12) \
- \
- stp2_16 = _mm_add_epi16(stp1_16, stp1_17); \
- stp2_17 = _mm_sub_epi16(stp1_16, stp1_17); \
- stp2_18 = _mm_sub_epi16(stp1_19, stp1_18); \
- stp2_19 = _mm_add_epi16(stp1_19, stp1_18); \
- \
- stp2_20 = _mm_add_epi16(stp1_20, stp1_21); \
- stp2_21 = _mm_sub_epi16(stp1_20, stp1_21); \
- stp2_22 = _mm_sub_epi16(stp1_23, stp1_22); \
- stp2_23 = _mm_add_epi16(stp1_23, stp1_22); \
- \
- stp2_24 = _mm_add_epi16(stp1_24, stp1_25); \
- stp2_25 = _mm_sub_epi16(stp1_24, stp1_25); \
- stp2_26 = _mm_sub_epi16(stp1_27, stp1_26); \
- stp2_27 = _mm_add_epi16(stp1_27, stp1_26); \
- \
- stp2_28 = _mm_add_epi16(stp1_28, stp1_29); \
- stp2_29 = _mm_sub_epi16(stp1_28, stp1_29); \
- stp2_30 = _mm_sub_epi16(stp1_31, stp1_30); \
- stp2_31 = _mm_add_epi16(stp1_31, stp1_30); \
-} \
-\
-/* Stage3 */ \
-{ \
- const __m128i lo_4_28 = _mm_unpacklo_epi16(in[4], in[28]); \
- const __m128i hi_4_28 = _mm_unpackhi_epi16(in[4], in[28]); \
- const __m128i lo_20_12 = _mm_unpacklo_epi16(in[20], in[12]); \
- const __m128i hi_20_12 = _mm_unpackhi_epi16(in[20], in[12]); \
- \
- const __m128i lo_17_30 = _mm_unpacklo_epi16(stp2_17, stp2_30); \
- const __m128i hi_17_30 = _mm_unpackhi_epi16(stp2_17, stp2_30); \
- const __m128i lo_18_29 = _mm_unpacklo_epi16(stp2_18, stp2_29); \
- const __m128i hi_18_29 = _mm_unpackhi_epi16(stp2_18, stp2_29); \
- \
- const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
- const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
- const __m128i lo_22_25 = _mm_unpacklo_epi16(stp2_22, stp2_25); \
- const __m128i hi_22_25 = _mm_unpackhi_epi16(stp2_22, stp2_25); \
- \
- MULTIPLICATION_AND_ADD(lo_4_28, hi_4_28, lo_20_12, hi_20_12, stg3_0, \
- stg3_1, stg3_2, stg3_3, stp1_4, stp1_7, stp1_5, \
- stp1_6) \
- \
- stp1_8 = _mm_add_epi16(stp2_8, stp2_9); \
- stp1_9 = _mm_sub_epi16(stp2_8, stp2_9); \
- stp1_10 = _mm_sub_epi16(stp2_11, stp2_10); \
- stp1_11 = _mm_add_epi16(stp2_11, stp2_10); \
- stp1_12 = _mm_add_epi16(stp2_12, stp2_13); \
- stp1_13 = _mm_sub_epi16(stp2_12, stp2_13); \
- stp1_14 = _mm_sub_epi16(stp2_15, stp2_14); \
- stp1_15 = _mm_add_epi16(stp2_15, stp2_14); \
- \
- MULTIPLICATION_AND_ADD(lo_17_30, hi_17_30, lo_18_29, hi_18_29, stg3_4, \
- stg3_5, stg3_6, stg3_4, stp1_17, stp1_30, \
- stp1_18, stp1_29) \
- MULTIPLICATION_AND_ADD(lo_21_26, hi_21_26, lo_22_25, hi_22_25, stg3_8, \
- stg3_9, stg3_10, stg3_8, stp1_21, stp1_26, \
- stp1_22, stp1_25) \
- \
- stp1_16 = stp2_16; \
- stp1_31 = stp2_31; \
- stp1_19 = stp2_19; \
- stp1_20 = stp2_20; \
- stp1_23 = stp2_23; \
- stp1_24 = stp2_24; \
- stp1_27 = stp2_27; \
- stp1_28 = stp2_28; \
-} \
-\
-/* Stage4 */ \
-{ \
- const __m128i lo_0_16 = _mm_unpacklo_epi16(in[0], in[16]); \
- const __m128i hi_0_16 = _mm_unpackhi_epi16(in[0], in[16]); \
- const __m128i lo_8_24 = _mm_unpacklo_epi16(in[8], in[24]); \
- const __m128i hi_8_24 = _mm_unpackhi_epi16(in[8], in[24]); \
- \
- const __m128i lo_9_14 = _mm_unpacklo_epi16(stp1_9, stp1_14); \
- const __m128i hi_9_14 = _mm_unpackhi_epi16(stp1_9, stp1_14); \
- const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
- const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
- \
- MULTIPLICATION_AND_ADD(lo_0_16, hi_0_16, lo_8_24, hi_8_24, stg4_0, \
- stg4_1, stg4_2, stg4_3, stp2_0, stp2_1, \
- stp2_2, stp2_3) \
- \
- stp2_4 = _mm_add_epi16(stp1_4, stp1_5); \
- stp2_5 = _mm_sub_epi16(stp1_4, stp1_5); \
- stp2_6 = _mm_sub_epi16(stp1_7, stp1_6); \
- stp2_7 = _mm_add_epi16(stp1_7, stp1_6); \
- \
- MULTIPLICATION_AND_ADD(lo_9_14, hi_9_14, lo_10_13, hi_10_13, stg4_4, \
- stg4_5, stg4_6, stg4_4, stp2_9, stp2_14, \
- stp2_10, stp2_13) \
- \
- stp2_8 = stp1_8; \
- stp2_15 = stp1_15; \
- stp2_11 = stp1_11; \
- stp2_12 = stp1_12; \
- \
- stp2_16 = _mm_add_epi16(stp1_16, stp1_19); \
- stp2_17 = _mm_add_epi16(stp1_17, stp1_18); \
- stp2_18 = _mm_sub_epi16(stp1_17, stp1_18); \
- stp2_19 = _mm_sub_epi16(stp1_16, stp1_19); \
- stp2_20 = _mm_sub_epi16(stp1_23, stp1_20); \
- stp2_21 = _mm_sub_epi16(stp1_22, stp1_21); \
- stp2_22 = _mm_add_epi16(stp1_22, stp1_21); \
- stp2_23 = _mm_add_epi16(stp1_23, stp1_20); \
- \
- stp2_24 = _mm_add_epi16(stp1_24, stp1_27); \
- stp2_25 = _mm_add_epi16(stp1_25, stp1_26); \
- stp2_26 = _mm_sub_epi16(stp1_25, stp1_26); \
- stp2_27 = _mm_sub_epi16(stp1_24, stp1_27); \
- stp2_28 = _mm_sub_epi16(stp1_31, stp1_28); \
- stp2_29 = _mm_sub_epi16(stp1_30, stp1_29); \
- stp2_30 = _mm_add_epi16(stp1_29, stp1_30); \
- stp2_31 = _mm_add_epi16(stp1_28, stp1_31); \
-} \
-\
-/* Stage5 */ \
-{ \
- const __m128i lo_6_5 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
- const __m128i hi_6_5 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
- const __m128i lo_18_29 = _mm_unpacklo_epi16(stp2_18, stp2_29); \
- const __m128i hi_18_29 = _mm_unpackhi_epi16(stp2_18, stp2_29); \
- \
- const __m128i lo_19_28 = _mm_unpacklo_epi16(stp2_19, stp2_28); \
- const __m128i hi_19_28 = _mm_unpackhi_epi16(stp2_19, stp2_28); \
- const __m128i lo_20_27 = _mm_unpacklo_epi16(stp2_20, stp2_27); \
- const __m128i hi_20_27 = _mm_unpackhi_epi16(stp2_20, stp2_27); \
- \
- const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
- const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
- \
- stp1_0 = _mm_add_epi16(stp2_0, stp2_3); \
- stp1_1 = _mm_add_epi16(stp2_1, stp2_2); \
- stp1_2 = _mm_sub_epi16(stp2_1, stp2_2); \
- stp1_3 = _mm_sub_epi16(stp2_0, stp2_3); \
- \
- tmp0 = _mm_madd_epi16(lo_6_5, stg4_1); \
- tmp1 = _mm_madd_epi16(hi_6_5, stg4_1); \
- tmp2 = _mm_madd_epi16(lo_6_5, stg4_0); \
- tmp3 = _mm_madd_epi16(hi_6_5, stg4_0); \
- \
- tmp0 = _mm_add_epi32(tmp0, rounding); \
- tmp1 = _mm_add_epi32(tmp1, rounding); \
- tmp2 = _mm_add_epi32(tmp2, rounding); \
- tmp3 = _mm_add_epi32(tmp3, rounding); \
- \
- tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
- tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
- tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
- tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
- \
- stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
- stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
- \
- stp1_4 = stp2_4; \
- stp1_7 = stp2_7; \
- \
- stp1_8 = _mm_add_epi16(stp2_8, stp2_11); \
- stp1_9 = _mm_add_epi16(stp2_9, stp2_10); \
- stp1_10 = _mm_sub_epi16(stp2_9, stp2_10); \
- stp1_11 = _mm_sub_epi16(stp2_8, stp2_11); \
- stp1_12 = _mm_sub_epi16(stp2_15, stp2_12); \
- stp1_13 = _mm_sub_epi16(stp2_14, stp2_13); \
- stp1_14 = _mm_add_epi16(stp2_14, stp2_13); \
- stp1_15 = _mm_add_epi16(stp2_15, stp2_12); \
- \
- stp1_16 = stp2_16; \
- stp1_17 = stp2_17; \
- \
- MULTIPLICATION_AND_ADD(lo_18_29, hi_18_29, lo_19_28, hi_19_28, stg4_4, \
- stg4_5, stg4_4, stg4_5, stp1_18, stp1_29, \
- stp1_19, stp1_28) \
- MULTIPLICATION_AND_ADD(lo_20_27, hi_20_27, lo_21_26, hi_21_26, stg4_6, \
- stg4_4, stg4_6, stg4_4, stp1_20, stp1_27, \
- stp1_21, stp1_26) \
- \
- stp1_22 = stp2_22; \
- stp1_23 = stp2_23; \
- stp1_24 = stp2_24; \
- stp1_25 = stp2_25; \
- stp1_30 = stp2_30; \
- stp1_31 = stp2_31; \
-} \
-\
-/* Stage6 */ \
-{ \
- const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
- const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
- const __m128i lo_11_12 = _mm_unpacklo_epi16(stp1_11, stp1_12); \
- const __m128i hi_11_12 = _mm_unpackhi_epi16(stp1_11, stp1_12); \
- \
- stp2_0 = _mm_add_epi16(stp1_0, stp1_7); \
- stp2_1 = _mm_add_epi16(stp1_1, stp1_6); \
- stp2_2 = _mm_add_epi16(stp1_2, stp1_5); \
- stp2_3 = _mm_add_epi16(stp1_3, stp1_4); \
- stp2_4 = _mm_sub_epi16(stp1_3, stp1_4); \
- stp2_5 = _mm_sub_epi16(stp1_2, stp1_5); \
- stp2_6 = _mm_sub_epi16(stp1_1, stp1_6); \
- stp2_7 = _mm_sub_epi16(stp1_0, stp1_7); \
- \
- stp2_8 = stp1_8; \
- stp2_9 = stp1_9; \
- stp2_14 = stp1_14; \
- stp2_15 = stp1_15; \
- \
- MULTIPLICATION_AND_ADD(lo_10_13, hi_10_13, lo_11_12, hi_11_12, \
- stg6_0, stg4_0, stg6_0, stg4_0, stp2_10, \
- stp2_13, stp2_11, stp2_12) \
- \
- stp2_16 = _mm_add_epi16(stp1_16, stp1_23); \
- stp2_17 = _mm_add_epi16(stp1_17, stp1_22); \
- stp2_18 = _mm_add_epi16(stp1_18, stp1_21); \
- stp2_19 = _mm_add_epi16(stp1_19, stp1_20); \
- stp2_20 = _mm_sub_epi16(stp1_19, stp1_20); \
- stp2_21 = _mm_sub_epi16(stp1_18, stp1_21); \
- stp2_22 = _mm_sub_epi16(stp1_17, stp1_22); \
- stp2_23 = _mm_sub_epi16(stp1_16, stp1_23); \
- \
- stp2_24 = _mm_sub_epi16(stp1_31, stp1_24); \
- stp2_25 = _mm_sub_epi16(stp1_30, stp1_25); \
- stp2_26 = _mm_sub_epi16(stp1_29, stp1_26); \
- stp2_27 = _mm_sub_epi16(stp1_28, stp1_27); \
- stp2_28 = _mm_add_epi16(stp1_27, stp1_28); \
- stp2_29 = _mm_add_epi16(stp1_26, stp1_29); \
- stp2_30 = _mm_add_epi16(stp1_25, stp1_30); \
- stp2_31 = _mm_add_epi16(stp1_24, stp1_31); \
-} \
-\
-/* Stage7 */ \
-{ \
- const __m128i lo_20_27 = _mm_unpacklo_epi16(stp2_20, stp2_27); \
- const __m128i hi_20_27 = _mm_unpackhi_epi16(stp2_20, stp2_27); \
- const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
- const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
- \
- const __m128i lo_22_25 = _mm_unpacklo_epi16(stp2_22, stp2_25); \
- const __m128i hi_22_25 = _mm_unpackhi_epi16(stp2_22, stp2_25); \
- const __m128i lo_23_24 = _mm_unpacklo_epi16(stp2_23, stp2_24); \
- const __m128i hi_23_24 = _mm_unpackhi_epi16(stp2_23, stp2_24); \
- \
- stp1_0 = _mm_add_epi16(stp2_0, stp2_15); \
- stp1_1 = _mm_add_epi16(stp2_1, stp2_14); \
- stp1_2 = _mm_add_epi16(stp2_2, stp2_13); \
- stp1_3 = _mm_add_epi16(stp2_3, stp2_12); \
- stp1_4 = _mm_add_epi16(stp2_4, stp2_11); \
- stp1_5 = _mm_add_epi16(stp2_5, stp2_10); \
- stp1_6 = _mm_add_epi16(stp2_6, stp2_9); \
- stp1_7 = _mm_add_epi16(stp2_7, stp2_8); \
- stp1_8 = _mm_sub_epi16(stp2_7, stp2_8); \
- stp1_9 = _mm_sub_epi16(stp2_6, stp2_9); \
- stp1_10 = _mm_sub_epi16(stp2_5, stp2_10); \
- stp1_11 = _mm_sub_epi16(stp2_4, stp2_11); \
- stp1_12 = _mm_sub_epi16(stp2_3, stp2_12); \
- stp1_13 = _mm_sub_epi16(stp2_2, stp2_13); \
- stp1_14 = _mm_sub_epi16(stp2_1, stp2_14); \
- stp1_15 = _mm_sub_epi16(stp2_0, stp2_15); \
- \
- stp1_16 = stp2_16; \
- stp1_17 = stp2_17; \
- stp1_18 = stp2_18; \
- stp1_19 = stp2_19; \
- \
- MULTIPLICATION_AND_ADD(lo_20_27, hi_20_27, lo_21_26, hi_21_26, stg6_0, \
- stg4_0, stg6_0, stg4_0, stp1_20, stp1_27, \
- stp1_21, stp1_26) \
- MULTIPLICATION_AND_ADD(lo_22_25, hi_22_25, lo_23_24, hi_23_24, stg6_0, \
- stg4_0, stg6_0, stg4_0, stp1_22, stp1_25, \
- stp1_23, stp1_24) \
- \
- stp1_28 = stp2_28; \
- stp1_29 = stp2_29; \
- stp1_30 = stp2_30; \
- stp1_31 = stp2_31; \
-}
+#define IDCT32 \
+ /* Stage1 */ \
+ { \
+ const __m128i lo_1_31 = _mm_unpacklo_epi16(in[1], in[31]); \
+ const __m128i hi_1_31 = _mm_unpackhi_epi16(in[1], in[31]); \
+ const __m128i lo_17_15 = _mm_unpacklo_epi16(in[17], in[15]); \
+ const __m128i hi_17_15 = _mm_unpackhi_epi16(in[17], in[15]); \
+ \
+ const __m128i lo_9_23 = _mm_unpacklo_epi16(in[9], in[23]); \
+ const __m128i hi_9_23 = _mm_unpackhi_epi16(in[9], in[23]); \
+ const __m128i lo_25_7 = _mm_unpacklo_epi16(in[25], in[7]); \
+ const __m128i hi_25_7 = _mm_unpackhi_epi16(in[25], in[7]); \
+ \
+ const __m128i lo_5_27 = _mm_unpacklo_epi16(in[5], in[27]); \
+ const __m128i hi_5_27 = _mm_unpackhi_epi16(in[5], in[27]); \
+ const __m128i lo_21_11 = _mm_unpacklo_epi16(in[21], in[11]); \
+ const __m128i hi_21_11 = _mm_unpackhi_epi16(in[21], in[11]); \
+ \
+ const __m128i lo_13_19 = _mm_unpacklo_epi16(in[13], in[19]); \
+ const __m128i hi_13_19 = _mm_unpackhi_epi16(in[13], in[19]); \
+ const __m128i lo_29_3 = _mm_unpacklo_epi16(in[29], in[3]); \
+ const __m128i hi_29_3 = _mm_unpackhi_epi16(in[29], in[3]); \
+ \
+ MULTIPLICATION_AND_ADD(lo_1_31, hi_1_31, lo_17_15, hi_17_15, stg1_0, \
+ stg1_1, stg1_2, stg1_3, stp1_16, stp1_31, stp1_17, \
+ stp1_30) \
+ MULTIPLICATION_AND_ADD(lo_9_23, hi_9_23, lo_25_7, hi_25_7, stg1_4, stg1_5, \
+ stg1_6, stg1_7, stp1_18, stp1_29, stp1_19, stp1_28) \
+ MULTIPLICATION_AND_ADD(lo_5_27, hi_5_27, lo_21_11, hi_21_11, stg1_8, \
+ stg1_9, stg1_10, stg1_11, stp1_20, stp1_27, \
+ stp1_21, stp1_26) \
+ MULTIPLICATION_AND_ADD(lo_13_19, hi_13_19, lo_29_3, hi_29_3, stg1_12, \
+ stg1_13, stg1_14, stg1_15, stp1_22, stp1_25, \
+ stp1_23, stp1_24) \
+ } \
+ \
+ /* Stage2 */ \
+ { \
+ const __m128i lo_2_30 = _mm_unpacklo_epi16(in[2], in[30]); \
+ const __m128i hi_2_30 = _mm_unpackhi_epi16(in[2], in[30]); \
+ const __m128i lo_18_14 = _mm_unpacklo_epi16(in[18], in[14]); \
+ const __m128i hi_18_14 = _mm_unpackhi_epi16(in[18], in[14]); \
+ \
+ const __m128i lo_10_22 = _mm_unpacklo_epi16(in[10], in[22]); \
+ const __m128i hi_10_22 = _mm_unpackhi_epi16(in[10], in[22]); \
+ const __m128i lo_26_6 = _mm_unpacklo_epi16(in[26], in[6]); \
+ const __m128i hi_26_6 = _mm_unpackhi_epi16(in[26], in[6]); \
+ \
+ MULTIPLICATION_AND_ADD(lo_2_30, hi_2_30, lo_18_14, hi_18_14, stg2_0, \
+ stg2_1, stg2_2, stg2_3, stp2_8, stp2_15, stp2_9, \
+ stp2_14) \
+ MULTIPLICATION_AND_ADD(lo_10_22, hi_10_22, lo_26_6, hi_26_6, stg2_4, \
+ stg2_5, stg2_6, stg2_7, stp2_10, stp2_13, stp2_11, \
+ stp2_12) \
+ \
+ stp2_16 = _mm_add_epi16(stp1_16, stp1_17); \
+ stp2_17 = _mm_sub_epi16(stp1_16, stp1_17); \
+ stp2_18 = _mm_sub_epi16(stp1_19, stp1_18); \
+ stp2_19 = _mm_add_epi16(stp1_19, stp1_18); \
+ \
+ stp2_20 = _mm_add_epi16(stp1_20, stp1_21); \
+ stp2_21 = _mm_sub_epi16(stp1_20, stp1_21); \
+ stp2_22 = _mm_sub_epi16(stp1_23, stp1_22); \
+ stp2_23 = _mm_add_epi16(stp1_23, stp1_22); \
+ \
+ stp2_24 = _mm_add_epi16(stp1_24, stp1_25); \
+ stp2_25 = _mm_sub_epi16(stp1_24, stp1_25); \
+ stp2_26 = _mm_sub_epi16(stp1_27, stp1_26); \
+ stp2_27 = _mm_add_epi16(stp1_27, stp1_26); \
+ \
+ stp2_28 = _mm_add_epi16(stp1_28, stp1_29); \
+ stp2_29 = _mm_sub_epi16(stp1_28, stp1_29); \
+ stp2_30 = _mm_sub_epi16(stp1_31, stp1_30); \
+ stp2_31 = _mm_add_epi16(stp1_31, stp1_30); \
+ } \
+ \
+ /* Stage3 */ \
+ { \
+ const __m128i lo_4_28 = _mm_unpacklo_epi16(in[4], in[28]); \
+ const __m128i hi_4_28 = _mm_unpackhi_epi16(in[4], in[28]); \
+ const __m128i lo_20_12 = _mm_unpacklo_epi16(in[20], in[12]); \
+ const __m128i hi_20_12 = _mm_unpackhi_epi16(in[20], in[12]); \
+ \
+ const __m128i lo_17_30 = _mm_unpacklo_epi16(stp2_17, stp2_30); \
+ const __m128i hi_17_30 = _mm_unpackhi_epi16(stp2_17, stp2_30); \
+ const __m128i lo_18_29 = _mm_unpacklo_epi16(stp2_18, stp2_29); \
+ const __m128i hi_18_29 = _mm_unpackhi_epi16(stp2_18, stp2_29); \
+ \
+ const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
+ const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
+ const __m128i lo_22_25 = _mm_unpacklo_epi16(stp2_22, stp2_25); \
+ const __m128i hi_22_25 = _mm_unpackhi_epi16(stp2_22, stp2_25); \
+ \
+ MULTIPLICATION_AND_ADD(lo_4_28, hi_4_28, lo_20_12, hi_20_12, stg3_0, \
+ stg3_1, stg3_2, stg3_3, stp1_4, stp1_7, stp1_5, \
+ stp1_6) \
+ \
+ stp1_8 = _mm_add_epi16(stp2_8, stp2_9); \
+ stp1_9 = _mm_sub_epi16(stp2_8, stp2_9); \
+ stp1_10 = _mm_sub_epi16(stp2_11, stp2_10); \
+ stp1_11 = _mm_add_epi16(stp2_11, stp2_10); \
+ stp1_12 = _mm_add_epi16(stp2_12, stp2_13); \
+ stp1_13 = _mm_sub_epi16(stp2_12, stp2_13); \
+ stp1_14 = _mm_sub_epi16(stp2_15, stp2_14); \
+ stp1_15 = _mm_add_epi16(stp2_15, stp2_14); \
+ \
+ MULTIPLICATION_AND_ADD(lo_17_30, hi_17_30, lo_18_29, hi_18_29, stg3_4, \
+ stg3_5, stg3_6, stg3_4, stp1_17, stp1_30, stp1_18, \
+ stp1_29) \
+ MULTIPLICATION_AND_ADD(lo_21_26, hi_21_26, lo_22_25, hi_22_25, stg3_8, \
+ stg3_9, stg3_10, stg3_8, stp1_21, stp1_26, stp1_22, \
+ stp1_25) \
+ \
+ stp1_16 = stp2_16; \
+ stp1_31 = stp2_31; \
+ stp1_19 = stp2_19; \
+ stp1_20 = stp2_20; \
+ stp1_23 = stp2_23; \
+ stp1_24 = stp2_24; \
+ stp1_27 = stp2_27; \
+ stp1_28 = stp2_28; \
+ } \
+ \
+ /* Stage4 */ \
+ { \
+ const __m128i lo_0_16 = _mm_unpacklo_epi16(in[0], in[16]); \
+ const __m128i hi_0_16 = _mm_unpackhi_epi16(in[0], in[16]); \
+ const __m128i lo_8_24 = _mm_unpacklo_epi16(in[8], in[24]); \
+ const __m128i hi_8_24 = _mm_unpackhi_epi16(in[8], in[24]); \
+ \
+ const __m128i lo_9_14 = _mm_unpacklo_epi16(stp1_9, stp1_14); \
+ const __m128i hi_9_14 = _mm_unpackhi_epi16(stp1_9, stp1_14); \
+ const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
+ const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
+ \
+ MULTIPLICATION_AND_ADD(lo_0_16, hi_0_16, lo_8_24, hi_8_24, stg4_0, stg4_1, \
+ stg4_2, stg4_3, stp2_0, stp2_1, stp2_2, stp2_3) \
+ \
+ stp2_4 = _mm_add_epi16(stp1_4, stp1_5); \
+ stp2_5 = _mm_sub_epi16(stp1_4, stp1_5); \
+ stp2_6 = _mm_sub_epi16(stp1_7, stp1_6); \
+ stp2_7 = _mm_add_epi16(stp1_7, stp1_6); \
+ \
+ MULTIPLICATION_AND_ADD(lo_9_14, hi_9_14, lo_10_13, hi_10_13, stg4_4, \
+ stg4_5, stg4_6, stg4_4, stp2_9, stp2_14, stp2_10, \
+ stp2_13) \
+ \
+ stp2_8 = stp1_8; \
+ stp2_15 = stp1_15; \
+ stp2_11 = stp1_11; \
+ stp2_12 = stp1_12; \
+ \
+ stp2_16 = _mm_add_epi16(stp1_16, stp1_19); \
+ stp2_17 = _mm_add_epi16(stp1_17, stp1_18); \
+ stp2_18 = _mm_sub_epi16(stp1_17, stp1_18); \
+ stp2_19 = _mm_sub_epi16(stp1_16, stp1_19); \
+ stp2_20 = _mm_sub_epi16(stp1_23, stp1_20); \
+ stp2_21 = _mm_sub_epi16(stp1_22, stp1_21); \
+ stp2_22 = _mm_add_epi16(stp1_22, stp1_21); \
+ stp2_23 = _mm_add_epi16(stp1_23, stp1_20); \
+ \
+ stp2_24 = _mm_add_epi16(stp1_24, stp1_27); \
+ stp2_25 = _mm_add_epi16(stp1_25, stp1_26); \
+ stp2_26 = _mm_sub_epi16(stp1_25, stp1_26); \
+ stp2_27 = _mm_sub_epi16(stp1_24, stp1_27); \
+ stp2_28 = _mm_sub_epi16(stp1_31, stp1_28); \
+ stp2_29 = _mm_sub_epi16(stp1_30, stp1_29); \
+ stp2_30 = _mm_add_epi16(stp1_29, stp1_30); \
+ stp2_31 = _mm_add_epi16(stp1_28, stp1_31); \
+ } \
+ \
+ /* Stage5 */ \
+ { \
+ const __m128i lo_6_5 = _mm_unpacklo_epi16(stp2_6, stp2_5); \
+ const __m128i hi_6_5 = _mm_unpackhi_epi16(stp2_6, stp2_5); \
+ const __m128i lo_18_29 = _mm_unpacklo_epi16(stp2_18, stp2_29); \
+ const __m128i hi_18_29 = _mm_unpackhi_epi16(stp2_18, stp2_29); \
+ \
+ const __m128i lo_19_28 = _mm_unpacklo_epi16(stp2_19, stp2_28); \
+ const __m128i hi_19_28 = _mm_unpackhi_epi16(stp2_19, stp2_28); \
+ const __m128i lo_20_27 = _mm_unpacklo_epi16(stp2_20, stp2_27); \
+ const __m128i hi_20_27 = _mm_unpackhi_epi16(stp2_20, stp2_27); \
+ \
+ const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
+ const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
+ \
+ stp1_0 = _mm_add_epi16(stp2_0, stp2_3); \
+ stp1_1 = _mm_add_epi16(stp2_1, stp2_2); \
+ stp1_2 = _mm_sub_epi16(stp2_1, stp2_2); \
+ stp1_3 = _mm_sub_epi16(stp2_0, stp2_3); \
+ \
+ tmp0 = _mm_madd_epi16(lo_6_5, stg4_1); \
+ tmp1 = _mm_madd_epi16(hi_6_5, stg4_1); \
+ tmp2 = _mm_madd_epi16(lo_6_5, stg4_0); \
+ tmp3 = _mm_madd_epi16(hi_6_5, stg4_0); \
+ \
+ tmp0 = _mm_add_epi32(tmp0, rounding); \
+ tmp1 = _mm_add_epi32(tmp1, rounding); \
+ tmp2 = _mm_add_epi32(tmp2, rounding); \
+ tmp3 = _mm_add_epi32(tmp3, rounding); \
+ \
+ tmp0 = _mm_srai_epi32(tmp0, DCT_CONST_BITS); \
+ tmp1 = _mm_srai_epi32(tmp1, DCT_CONST_BITS); \
+ tmp2 = _mm_srai_epi32(tmp2, DCT_CONST_BITS); \
+ tmp3 = _mm_srai_epi32(tmp3, DCT_CONST_BITS); \
+ \
+ stp1_5 = _mm_packs_epi32(tmp0, tmp1); \
+ stp1_6 = _mm_packs_epi32(tmp2, tmp3); \
+ \
+ stp1_4 = stp2_4; \
+ stp1_7 = stp2_7; \
+ \
+ stp1_8 = _mm_add_epi16(stp2_8, stp2_11); \
+ stp1_9 = _mm_add_epi16(stp2_9, stp2_10); \
+ stp1_10 = _mm_sub_epi16(stp2_9, stp2_10); \
+ stp1_11 = _mm_sub_epi16(stp2_8, stp2_11); \
+ stp1_12 = _mm_sub_epi16(stp2_15, stp2_12); \
+ stp1_13 = _mm_sub_epi16(stp2_14, stp2_13); \
+ stp1_14 = _mm_add_epi16(stp2_14, stp2_13); \
+ stp1_15 = _mm_add_epi16(stp2_15, stp2_12); \
+ \
+ stp1_16 = stp2_16; \
+ stp1_17 = stp2_17; \
+ \
+ MULTIPLICATION_AND_ADD(lo_18_29, hi_18_29, lo_19_28, hi_19_28, stg4_4, \
+ stg4_5, stg4_4, stg4_5, stp1_18, stp1_29, stp1_19, \
+ stp1_28) \
+ MULTIPLICATION_AND_ADD(lo_20_27, hi_20_27, lo_21_26, hi_21_26, stg4_6, \
+ stg4_4, stg4_6, stg4_4, stp1_20, stp1_27, stp1_21, \
+ stp1_26) \
+ \
+ stp1_22 = stp2_22; \
+ stp1_23 = stp2_23; \
+ stp1_24 = stp2_24; \
+ stp1_25 = stp2_25; \
+ stp1_30 = stp2_30; \
+ stp1_31 = stp2_31; \
+ } \
+ \
+ /* Stage6 */ \
+ { \
+ const __m128i lo_10_13 = _mm_unpacklo_epi16(stp1_10, stp1_13); \
+ const __m128i hi_10_13 = _mm_unpackhi_epi16(stp1_10, stp1_13); \
+ const __m128i lo_11_12 = _mm_unpacklo_epi16(stp1_11, stp1_12); \
+ const __m128i hi_11_12 = _mm_unpackhi_epi16(stp1_11, stp1_12); \
+ \
+ stp2_0 = _mm_add_epi16(stp1_0, stp1_7); \
+ stp2_1 = _mm_add_epi16(stp1_1, stp1_6); \
+ stp2_2 = _mm_add_epi16(stp1_2, stp1_5); \
+ stp2_3 = _mm_add_epi16(stp1_3, stp1_4); \
+ stp2_4 = _mm_sub_epi16(stp1_3, stp1_4); \
+ stp2_5 = _mm_sub_epi16(stp1_2, stp1_5); \
+ stp2_6 = _mm_sub_epi16(stp1_1, stp1_6); \
+ stp2_7 = _mm_sub_epi16(stp1_0, stp1_7); \
+ \
+ stp2_8 = stp1_8; \
+ stp2_9 = stp1_9; \
+ stp2_14 = stp1_14; \
+ stp2_15 = stp1_15; \
+ \
+ MULTIPLICATION_AND_ADD(lo_10_13, hi_10_13, lo_11_12, hi_11_12, stg6_0, \
+ stg4_0, stg6_0, stg4_0, stp2_10, stp2_13, stp2_11, \
+ stp2_12) \
+ \
+ stp2_16 = _mm_add_epi16(stp1_16, stp1_23); \
+ stp2_17 = _mm_add_epi16(stp1_17, stp1_22); \
+ stp2_18 = _mm_add_epi16(stp1_18, stp1_21); \
+ stp2_19 = _mm_add_epi16(stp1_19, stp1_20); \
+ stp2_20 = _mm_sub_epi16(stp1_19, stp1_20); \
+ stp2_21 = _mm_sub_epi16(stp1_18, stp1_21); \
+ stp2_22 = _mm_sub_epi16(stp1_17, stp1_22); \
+ stp2_23 = _mm_sub_epi16(stp1_16, stp1_23); \
+ \
+ stp2_24 = _mm_sub_epi16(stp1_31, stp1_24); \
+ stp2_25 = _mm_sub_epi16(stp1_30, stp1_25); \
+ stp2_26 = _mm_sub_epi16(stp1_29, stp1_26); \
+ stp2_27 = _mm_sub_epi16(stp1_28, stp1_27); \
+ stp2_28 = _mm_add_epi16(stp1_27, stp1_28); \
+ stp2_29 = _mm_add_epi16(stp1_26, stp1_29); \
+ stp2_30 = _mm_add_epi16(stp1_25, stp1_30); \
+ stp2_31 = _mm_add_epi16(stp1_24, stp1_31); \
+ } \
+ \
+ /* Stage7 */ \
+ { \
+ const __m128i lo_20_27 = _mm_unpacklo_epi16(stp2_20, stp2_27); \
+ const __m128i hi_20_27 = _mm_unpackhi_epi16(stp2_20, stp2_27); \
+ const __m128i lo_21_26 = _mm_unpacklo_epi16(stp2_21, stp2_26); \
+ const __m128i hi_21_26 = _mm_unpackhi_epi16(stp2_21, stp2_26); \
+ \
+ const __m128i lo_22_25 = _mm_unpacklo_epi16(stp2_22, stp2_25); \
+ const __m128i hi_22_25 = _mm_unpackhi_epi16(stp2_22, stp2_25); \
+ const __m128i lo_23_24 = _mm_unpacklo_epi16(stp2_23, stp2_24); \
+ const __m128i hi_23_24 = _mm_unpackhi_epi16(stp2_23, stp2_24); \
+ \
+ stp1_0 = _mm_add_epi16(stp2_0, stp2_15); \
+ stp1_1 = _mm_add_epi16(stp2_1, stp2_14); \
+ stp1_2 = _mm_add_epi16(stp2_2, stp2_13); \
+ stp1_3 = _mm_add_epi16(stp2_3, stp2_12); \
+ stp1_4 = _mm_add_epi16(stp2_4, stp2_11); \
+ stp1_5 = _mm_add_epi16(stp2_5, stp2_10); \
+ stp1_6 = _mm_add_epi16(stp2_6, stp2_9); \
+ stp1_7 = _mm_add_epi16(stp2_7, stp2_8); \
+ stp1_8 = _mm_sub_epi16(stp2_7, stp2_8); \
+ stp1_9 = _mm_sub_epi16(stp2_6, stp2_9); \
+ stp1_10 = _mm_sub_epi16(stp2_5, stp2_10); \
+ stp1_11 = _mm_sub_epi16(stp2_4, stp2_11); \
+ stp1_12 = _mm_sub_epi16(stp2_3, stp2_12); \
+ stp1_13 = _mm_sub_epi16(stp2_2, stp2_13); \
+ stp1_14 = _mm_sub_epi16(stp2_1, stp2_14); \
+ stp1_15 = _mm_sub_epi16(stp2_0, stp2_15); \
+ \
+ stp1_16 = stp2_16; \
+ stp1_17 = stp2_17; \
+ stp1_18 = stp2_18; \
+ stp1_19 = stp2_19; \
+ \
+ MULTIPLICATION_AND_ADD(lo_20_27, hi_20_27, lo_21_26, hi_21_26, stg6_0, \
+ stg4_0, stg6_0, stg4_0, stp1_20, stp1_27, stp1_21, \
+ stp1_26) \
+ MULTIPLICATION_AND_ADD(lo_22_25, hi_22_25, lo_23_24, hi_23_24, stg6_0, \
+ stg4_0, stg6_0, stg4_0, stp1_22, stp1_25, stp1_23, \
+ stp1_24) \
+ \
+ stp1_28 = stp2_28; \
+ stp1_29 = stp2_29; \
+ stp1_30 = stp2_30; \
+ stp1_31 = stp2_31; \
+ }
// Only upper-left 8x8 has non-zero coeff
void vp10_idct32x32_34_add_sse2(const int16_t *input, uint8_t *dest,
- int stride) {
+ int stride) {
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
- const __m128i final_rounding = _mm_set1_epi16(1<<5);
+ const __m128i final_rounding = _mm_set1_epi16(1 << 5);
// vp10_idct constants for each stage
const __m128i stg1_0 = pair_set_epi16(cospi_31_64, -cospi_1_64);
@@ -3070,15 +3055,13 @@
__m128i in[32], col[32];
__m128i stp1_0, stp1_1, stp1_2, stp1_3, stp1_4, stp1_5, stp1_6, stp1_7,
- stp1_8, stp1_9, stp1_10, stp1_11, stp1_12, stp1_13, stp1_14, stp1_15,
- stp1_16, stp1_17, stp1_18, stp1_19, stp1_20, stp1_21, stp1_22,
- stp1_23, stp1_24, stp1_25, stp1_26, stp1_27, stp1_28, stp1_29,
- stp1_30, stp1_31;
+ stp1_8, stp1_9, stp1_10, stp1_11, stp1_12, stp1_13, stp1_14, stp1_15,
+ stp1_16, stp1_17, stp1_18, stp1_19, stp1_20, stp1_21, stp1_22, stp1_23,
+ stp1_24, stp1_25, stp1_26, stp1_27, stp1_28, stp1_29, stp1_30, stp1_31;
__m128i stp2_0, stp2_1, stp2_2, stp2_3, stp2_4, stp2_5, stp2_6, stp2_7,
- stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14, stp2_15,
- stp2_16, stp2_17, stp2_18, stp2_19, stp2_20, stp2_21, stp2_22,
- stp2_23, stp2_24, stp2_25, stp2_26, stp2_27, stp2_28, stp2_29,
- stp2_30, stp2_31;
+ stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14, stp2_15,
+ stp2_16, stp2_17, stp2_18, stp2_19, stp2_20, stp2_21, stp2_22, stp2_23,
+ stp2_24, stp2_25, stp2_26, stp2_27, stp2_28, stp2_29, stp2_30, stp2_31;
__m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
int i;
@@ -3191,7 +3174,7 @@
}
void vp10_idct32x32_1024_add_sse2(const int16_t *input, uint8_t *dest,
- int stride) {
+ int stride) {
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1 << 5);
const __m128i zero = _mm_setzero_si128();
@@ -3246,15 +3229,13 @@
__m128i in[32], col[128], zero_idx[16];
__m128i stp1_0, stp1_1, stp1_2, stp1_3, stp1_4, stp1_5, stp1_6, stp1_7,
- stp1_8, stp1_9, stp1_10, stp1_11, stp1_12, stp1_13, stp1_14, stp1_15,
- stp1_16, stp1_17, stp1_18, stp1_19, stp1_20, stp1_21, stp1_22,
- stp1_23, stp1_24, stp1_25, stp1_26, stp1_27, stp1_28, stp1_29,
- stp1_30, stp1_31;
+ stp1_8, stp1_9, stp1_10, stp1_11, stp1_12, stp1_13, stp1_14, stp1_15,
+ stp1_16, stp1_17, stp1_18, stp1_19, stp1_20, stp1_21, stp1_22, stp1_23,
+ stp1_24, stp1_25, stp1_26, stp1_27, stp1_28, stp1_29, stp1_30, stp1_31;
__m128i stp2_0, stp2_1, stp2_2, stp2_3, stp2_4, stp2_5, stp2_6, stp2_7,
- stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14, stp2_15,
- stp2_16, stp2_17, stp2_18, stp2_19, stp2_20, stp2_21, stp2_22,
- stp2_23, stp2_24, stp2_25, stp2_26, stp2_27, stp2_28, stp2_29,
- stp2_30, stp2_31;
+ stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14, stp2_15,
+ stp2_16, stp2_17, stp2_18, stp2_19, stp2_20, stp2_21, stp2_22, stp2_23,
+ stp2_24, stp2_25, stp2_26, stp2_27, stp2_28, stp2_29, stp2_30, stp2_31;
__m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
int i, j, i32;
@@ -3466,8 +3447,7 @@
}
}
-void vp10_idct32x32_1_add_sse2(const int16_t *input,
- uint8_t *dest,
+void vp10_idct32x32_1_add_sse2(const int16_t *input, uint8_t *dest,
int stride) {
__m128i dc_value;
const __m128i zero = _mm_setzero_si128();
@@ -3503,7 +3483,7 @@
}
void vp10_highbd_idct4x4_16_add_sse2(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[4 * 4];
tran_low_t *outptr = out;
int i, j;
@@ -3607,8 +3587,7 @@
tran_low_t temp_in[4], temp_out[4];
// Columns
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j * 4 + i];
+ for (j = 0; j < 4; ++j) temp_in[j] = out[j * 4 + i];
vp10_highbd_idct4_c(temp_in, temp_out, bd);
for (j = 0; j < 4; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -3619,7 +3598,7 @@
}
void vp10_highbd_idct8x8_64_add_sse2(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
int i, j, test;
@@ -3697,19 +3676,18 @@
__m128i d[8];
for (i = 0; i < 8; i++) {
inptr[i] = _mm_add_epi16(inptr[i], sixteen);
- d[i] = _mm_loadu_si128((const __m128i *)(dest + stride*i));
+ d[i] = _mm_loadu_si128((const __m128i *)(dest + stride * i));
inptr[i] = _mm_srai_epi16(inptr[i], 5);
d[i] = clamp_high_sse2(_mm_adds_epi16(d[i], inptr[i]), bd);
// Store
- _mm_storeu_si128((__m128i *)(dest + stride*i), d[i]);
+ _mm_storeu_si128((__m128i *)(dest + stride * i), d[i]);
}
}
} else {
// Run the un-optimised column transform
tran_low_t temp_in[8], temp_out[8];
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
vp10_highbd_idct8_c(temp_in, temp_out, bd);
for (j = 0; j < 8; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -3720,7 +3698,7 @@
}
void vp10_highbd_idct8x8_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[8 * 8] = { 0 };
tran_low_t *outptr = out;
int i, j, test;
@@ -3801,19 +3779,18 @@
__m128i d[8];
for (i = 0; i < 8; i++) {
inptr[i] = _mm_add_epi16(inptr[i], sixteen);
- d[i] = _mm_loadu_si128((const __m128i *)(dest + stride*i));
+ d[i] = _mm_loadu_si128((const __m128i *)(dest + stride * i));
inptr[i] = _mm_srai_epi16(inptr[i], 5);
d[i] = clamp_high_sse2(_mm_adds_epi16(d[i], inptr[i]), bd);
// Store
- _mm_storeu_si128((__m128i *)(dest + stride*i), d[i]);
+ _mm_storeu_si128((__m128i *)(dest + stride * i), d[i]);
}
}
} else {
// Run the un-optimised column transform
tran_low_t temp_in[8], temp_out[8];
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
vp10_highbd_idct8_c(temp_in, temp_out, bd);
for (j = 0; j < 8; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -3824,7 +3801,7 @@
}
void vp10_highbd_idct16x16_256_add_sse2(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
int i, j, test;
@@ -3909,25 +3886,24 @@
{
__m128i d[2];
for (i = 0; i < 16; i++) {
- inptr[i ] = _mm_add_epi16(inptr[i ], rounding);
- inptr[i+16] = _mm_add_epi16(inptr[i+16], rounding);
- d[0] = _mm_loadu_si128((const __m128i *)(dest + stride*i));
- d[1] = _mm_loadu_si128((const __m128i *)(dest + stride*i + 8));
- inptr[i ] = _mm_srai_epi16(inptr[i ], 6);
- inptr[i+16] = _mm_srai_epi16(inptr[i+16], 6);
- d[0] = clamp_high_sse2(_mm_add_epi16(d[0], inptr[i ]), bd);
- d[1] = clamp_high_sse2(_mm_add_epi16(d[1], inptr[i+16]), bd);
+ inptr[i] = _mm_add_epi16(inptr[i], rounding);
+ inptr[i + 16] = _mm_add_epi16(inptr[i + 16], rounding);
+ d[0] = _mm_loadu_si128((const __m128i *)(dest + stride * i));
+ d[1] = _mm_loadu_si128((const __m128i *)(dest + stride * i + 8));
+ inptr[i] = _mm_srai_epi16(inptr[i], 6);
+ inptr[i + 16] = _mm_srai_epi16(inptr[i + 16], 6);
+ d[0] = clamp_high_sse2(_mm_add_epi16(d[0], inptr[i]), bd);
+ d[1] = clamp_high_sse2(_mm_add_epi16(d[1], inptr[i + 16]), bd);
// Store
- _mm_storeu_si128((__m128i *)(dest + stride*i), d[0]);
- _mm_storeu_si128((__m128i *)(dest + stride*i + 8), d[1]);
+ _mm_storeu_si128((__m128i *)(dest + stride * i), d[0]);
+ _mm_storeu_si128((__m128i *)(dest + stride * i + 8), d[1]);
}
}
} else {
// Run the un-optimised column transform
tran_low_t temp_in[16], temp_out[16];
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
vp10_highbd_idct16_c(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
@@ -3938,7 +3914,7 @@
}
void vp10_highbd_idct16x16_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+ int stride, int bd) {
tran_low_t out[16 * 16] = { 0 };
tran_low_t *outptr = out;
int i, j, test;
@@ -4028,25 +4004,24 @@
{
__m128i d[2];
for (i = 0; i < 16; i++) {
- inptr[i ] = _mm_add_epi16(inptr[i ], rounding);
- inptr[i+16] = _mm_add_epi16(inptr[i+16], rounding);
- d[0] = _mm_loadu_si128((const __m128i *)(dest + stride*i));
- d[1] = _mm_loadu_si128((const __m128i *)(dest + stride*i + 8));
- inptr[i ] = _mm_srai_epi16(inptr[i ], 6);
- inptr[i+16] = _mm_srai_epi16(inptr[i+16], 6);
- d[0] = clamp_high_sse2(_mm_add_epi16(d[0], inptr[i ]), bd);
- d[1] = clamp_high_sse2(_mm_add_epi16(d[1], inptr[i+16]), bd);
+ inptr[i] = _mm_add_epi16(inptr[i], rounding);
+ inptr[i + 16] = _mm_add_epi16(inptr[i + 16], rounding);
+ d[0] = _mm_loadu_si128((const __m128i *)(dest + stride * i));
+ d[1] = _mm_loadu_si128((const __m128i *)(dest + stride * i + 8));
+ inptr[i] = _mm_srai_epi16(inptr[i], 6);
+ inptr[i + 16] = _mm_srai_epi16(inptr[i + 16], 6);
+ d[0] = clamp_high_sse2(_mm_add_epi16(d[0], inptr[i]), bd);
+ d[1] = clamp_high_sse2(_mm_add_epi16(d[1], inptr[i + 16]), bd);
// Store
- _mm_storeu_si128((__m128i *)(dest + stride*i), d[0]);
- _mm_storeu_si128((__m128i *)(dest + stride*i + 8), d[1]);
+ _mm_storeu_si128((__m128i *)(dest + stride * i), d[0]);
+ _mm_storeu_si128((__m128i *)(dest + stride * i + 8), d[1]);
}
}
} else {
// Run the un-optimised column transform
tran_low_t temp_in[16], temp_out[16];
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
vp10_highbd_idct16_c(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
diff --git a/vp10/common/x86/vp10_inv_txfm_sse2.h b/vp10/common/x86/vp10_inv_txfm_sse2.h
index b79781a..9d8c46b 100644
--- a/vp10/common/x86/vp10_inv_txfm_sse2.h
+++ b/vp10/common/x86/vp10_inv_txfm_sse2.h
@@ -46,16 +46,16 @@
res[7] = _mm_unpackhi_epi64(tr1_6, tr1_7);
}
-#define TRANSPOSE_8X4(in0, in1, in2, in3, out0, out1) \
+#define TRANSPOSE_8X4(in0, in1, in2, in3, out0, out1) \
{ \
const __m128i tr0_0 = _mm_unpacklo_epi16(in0, in1); \
const __m128i tr0_1 = _mm_unpacklo_epi16(in2, in3); \
\
- in0 = _mm_unpacklo_epi32(tr0_0, tr0_1); /* i1 i0 */ \
- in1 = _mm_unpackhi_epi32(tr0_0, tr0_1); /* i3 i2 */ \
+ in0 = _mm_unpacklo_epi32(tr0_0, tr0_1); /* i1 i0 */ \
+ in1 = _mm_unpackhi_epi32(tr0_0, tr0_1); /* i3 i2 */ \
}
-static INLINE void array_transpose_4X8(__m128i *in, __m128i * out) {
+static INLINE void array_transpose_4X8(__m128i *in, __m128i *out) {
const __m128i tr0_0 = _mm_unpacklo_epi16(in[0], in[1]);
const __m128i tr0_1 = _mm_unpacklo_epi16(in[2], in[3]);
const __m128i tr0_4 = _mm_unpacklo_epi16(in[4], in[5]);
@@ -90,36 +90,36 @@
}
static INLINE void load_buffer_8x16(const int16_t *input, __m128i *in) {
- in[0] = _mm_load_si128((const __m128i *)(input + 0 * 16));
- in[1] = _mm_load_si128((const __m128i *)(input + 1 * 16));
- in[2] = _mm_load_si128((const __m128i *)(input + 2 * 16));
- in[3] = _mm_load_si128((const __m128i *)(input + 3 * 16));
- in[4] = _mm_load_si128((const __m128i *)(input + 4 * 16));
- in[5] = _mm_load_si128((const __m128i *)(input + 5 * 16));
- in[6] = _mm_load_si128((const __m128i *)(input + 6 * 16));
- in[7] = _mm_load_si128((const __m128i *)(input + 7 * 16));
+ in[0] = _mm_load_si128((const __m128i *)(input + 0 * 16));
+ in[1] = _mm_load_si128((const __m128i *)(input + 1 * 16));
+ in[2] = _mm_load_si128((const __m128i *)(input + 2 * 16));
+ in[3] = _mm_load_si128((const __m128i *)(input + 3 * 16));
+ in[4] = _mm_load_si128((const __m128i *)(input + 4 * 16));
+ in[5] = _mm_load_si128((const __m128i *)(input + 5 * 16));
+ in[6] = _mm_load_si128((const __m128i *)(input + 6 * 16));
+ in[7] = _mm_load_si128((const __m128i *)(input + 7 * 16));
- in[8] = _mm_load_si128((const __m128i *)(input + 8 * 16));
- in[9] = _mm_load_si128((const __m128i *)(input + 9 * 16));
- in[10] = _mm_load_si128((const __m128i *)(input + 10 * 16));
- in[11] = _mm_load_si128((const __m128i *)(input + 11 * 16));
- in[12] = _mm_load_si128((const __m128i *)(input + 12 * 16));
- in[13] = _mm_load_si128((const __m128i *)(input + 13 * 16));
- in[14] = _mm_load_si128((const __m128i *)(input + 14 * 16));
- in[15] = _mm_load_si128((const __m128i *)(input + 15 * 16));
+ in[8] = _mm_load_si128((const __m128i *)(input + 8 * 16));
+ in[9] = _mm_load_si128((const __m128i *)(input + 9 * 16));
+ in[10] = _mm_load_si128((const __m128i *)(input + 10 * 16));
+ in[11] = _mm_load_si128((const __m128i *)(input + 11 * 16));
+ in[12] = _mm_load_si128((const __m128i *)(input + 12 * 16));
+ in[13] = _mm_load_si128((const __m128i *)(input + 13 * 16));
+ in[14] = _mm_load_si128((const __m128i *)(input + 14 * 16));
+ in[15] = _mm_load_si128((const __m128i *)(input + 15 * 16));
}
-#define RECON_AND_STORE(dest, in_x) \
- { \
- __m128i d0 = _mm_loadl_epi64((__m128i *)(dest)); \
- d0 = _mm_unpacklo_epi8(d0, zero); \
- d0 = _mm_add_epi16(in_x, d0); \
- d0 = _mm_packus_epi16(d0, d0); \
- _mm_storel_epi64((__m128i *)(dest), d0); \
+#define RECON_AND_STORE(dest, in_x) \
+ { \
+ __m128i d0 = _mm_loadl_epi64((__m128i *)(dest)); \
+ d0 = _mm_unpacklo_epi8(d0, zero); \
+ d0 = _mm_add_epi16(in_x, d0); \
+ d0 = _mm_packus_epi16(d0, d0); \
+ _mm_storel_epi64((__m128i *)(dest), d0); \
}
static INLINE void write_buffer_8x16(uint8_t *dest, __m128i *in, int stride) {
- const __m128i final_rounding = _mm_set1_epi16(1<<5);
+ const __m128i final_rounding = _mm_set1_epi16(1 << 5);
const __m128i zero = _mm_setzero_si128();
// Final rounding and shift
in[0] = _mm_adds_epi16(in[0], final_rounding);
@@ -156,16 +156,16 @@
in[14] = _mm_srai_epi16(in[14], 6);
in[15] = _mm_srai_epi16(in[15], 6);
- RECON_AND_STORE(dest + 0 * stride, in[0]);
- RECON_AND_STORE(dest + 1 * stride, in[1]);
- RECON_AND_STORE(dest + 2 * stride, in[2]);
- RECON_AND_STORE(dest + 3 * stride, in[3]);
- RECON_AND_STORE(dest + 4 * stride, in[4]);
- RECON_AND_STORE(dest + 5 * stride, in[5]);
- RECON_AND_STORE(dest + 6 * stride, in[6]);
- RECON_AND_STORE(dest + 7 * stride, in[7]);
- RECON_AND_STORE(dest + 8 * stride, in[8]);
- RECON_AND_STORE(dest + 9 * stride, in[9]);
+ RECON_AND_STORE(dest + 0 * stride, in[0]);
+ RECON_AND_STORE(dest + 1 * stride, in[1]);
+ RECON_AND_STORE(dest + 2 * stride, in[2]);
+ RECON_AND_STORE(dest + 3 * stride, in[3]);
+ RECON_AND_STORE(dest + 4 * stride, in[4]);
+ RECON_AND_STORE(dest + 5 * stride, in[5]);
+ RECON_AND_STORE(dest + 6 * stride, in[6]);
+ RECON_AND_STORE(dest + 7 * stride, in[7]);
+ RECON_AND_STORE(dest + 8 * stride, in[8]);
+ RECON_AND_STORE(dest + 9 * stride, in[9]);
RECON_AND_STORE(dest + 10 * stride, in[10]);
RECON_AND_STORE(dest + 11 * stride, in[11]);
RECON_AND_STORE(dest + 12 * stride, in[12]);
diff --git a/vp10/decoder/decodeframe.c b/vp10/decoder/decodeframe.c
index 6b5d146..719c061 100644
--- a/vp10/decoder/decodeframe.c
+++ b/vp10/decoder/decodeframe.c
@@ -47,23 +47,21 @@
static int is_compound_reference_allowed(const VP10_COMMON *cm) {
int i;
- if (frame_is_intra_only(cm))
- return 0;
+ if (frame_is_intra_only(cm)) return 0;
for (i = 1; i < REFS_PER_FRAME; ++i)
- if (cm->ref_frame_sign_bias[i + 1] != cm->ref_frame_sign_bias[1])
- return 1;
+ if (cm->ref_frame_sign_bias[i + 1] != cm->ref_frame_sign_bias[1]) return 1;
return 0;
}
static void setup_compound_reference_mode(VP10_COMMON *cm) {
if (cm->ref_frame_sign_bias[LAST_FRAME] ==
- cm->ref_frame_sign_bias[GOLDEN_FRAME]) {
+ cm->ref_frame_sign_bias[GOLDEN_FRAME]) {
cm->comp_fixed_ref = ALTREF_FRAME;
cm->comp_var_ref[0] = LAST_FRAME;
cm->comp_var_ref[1] = GOLDEN_FRAME;
} else if (cm->ref_frame_sign_bias[LAST_FRAME] ==
- cm->ref_frame_sign_bias[ALTREF_FRAME]) {
+ cm->ref_frame_sign_bias[ALTREF_FRAME]) {
cm->comp_fixed_ref = GOLDEN_FRAME;
cm->comp_var_ref[0] = LAST_FRAME;
cm->comp_var_ref[1] = ALTREF_FRAME;
@@ -90,8 +88,7 @@
#else
static TX_MODE read_tx_mode(vpx_reader *r) {
TX_MODE tx_mode = vpx_read_literal(r, 2);
- if (tx_mode == ALLOW_32X32)
- tx_mode += vpx_read_bit(r);
+ if (tx_mode == ALLOW_32X32) tx_mode += vpx_read_bit(r);
return tx_mode;
}
#endif
@@ -127,12 +124,12 @@
}
#if CONFIG_MISC_FIXES
-static REFERENCE_MODE read_frame_reference_mode(const VP10_COMMON *cm,
- struct vpx_read_bit_buffer *rb) {
+static REFERENCE_MODE read_frame_reference_mode(
+ const VP10_COMMON *cm, struct vpx_read_bit_buffer *rb) {
if (is_compound_reference_allowed(cm)) {
- return vpx_rb_read_bit(rb) ? REFERENCE_MODE_SELECT
- : (vpx_rb_read_bit(rb) ? COMPOUND_REFERENCE
- : SINGLE_REFERENCE);
+ return vpx_rb_read_bit(rb)
+ ? REFERENCE_MODE_SELECT
+ : (vpx_rb_read_bit(rb) ? COMPOUND_REFERENCE : SINGLE_REFERENCE);
} else {
return SINGLE_REFERENCE;
}
@@ -141,9 +138,9 @@
static REFERENCE_MODE read_frame_reference_mode(const VP10_COMMON *cm,
vpx_reader *r) {
if (is_compound_reference_allowed(cm)) {
- return vpx_read_bit(r) ? (vpx_read_bit(r) ? REFERENCE_MODE_SELECT
- : COMPOUND_REFERENCE)
- : SINGLE_REFERENCE;
+ return vpx_read_bit(r)
+ ? (vpx_read_bit(r) ? REFERENCE_MODE_SELECT : COMPOUND_REFERENCE)
+ : SINGLE_REFERENCE;
} else {
return SINGLE_REFERENCE;
}
@@ -175,8 +172,7 @@
#if CONFIG_MISC_FIXES
vp10_diff_update_prob(r, &p[i]);
#else
- if (vpx_read(r, MV_UPDATE_PROB))
- p[i] = (vpx_read_literal(r, 7) << 1) | 1;
+ if (vpx_read(r, MV_UPDATE_PROB)) p[i] = (vpx_read_literal(r, 7) << 1) | 1;
#endif
}
@@ -209,10 +205,9 @@
}
}
-static void inverse_transform_block_inter(MACROBLOCKD* xd, int plane,
- const TX_SIZE tx_size,
- uint8_t *dst, int stride,
- int eob, int block) {
+static void inverse_transform_block_inter(MACROBLOCKD *xd, int plane,
+ const TX_SIZE tx_size, uint8_t *dst,
+ int stride, int eob, int block) {
struct macroblockd_plane *const pd = &xd->plane[plane];
TX_TYPE tx_type = get_tx_type(pd->plane_type, xd, block);
const int seg_id = xd->mi[0]->mbmi.segment_id;
@@ -237,9 +232,7 @@
vp10_highbd_inv_txfm_add_32x32(dqcoeff, dst, stride, eob, xd->bd,
tx_type);
break;
- default:
- assert(0 && "Invalid transform size");
- return;
+ default: assert(0 && "Invalid transform size"); return;
}
} else {
#endif // CONFIG_VPX_HIGHBITDEPTH
@@ -257,9 +250,7 @@
case TX_32X32:
vp10_inv_txfm_add_32x32(dqcoeff, dst, stride, eob, tx_type);
break;
- default:
- assert(0 && "Invalid transform size");
- return;
+ default: assert(0 && "Invalid transform size"); return;
}
#if CONFIG_VPX_HIGHBITDEPTH
}
@@ -278,11 +269,10 @@
}
}
-static void inverse_transform_block_intra(MACROBLOCKD* xd, int plane,
+static void inverse_transform_block_intra(MACROBLOCKD *xd, int plane,
const TX_TYPE tx_type,
- const TX_SIZE tx_size,
- uint8_t *dst, int stride,
- int eob) {
+ const TX_SIZE tx_size, uint8_t *dst,
+ int stride, int eob) {
struct macroblockd_plane *const pd = &xd->plane[plane];
const int seg_id = xd->mi[0]->mbmi.segment_id;
if (eob > 0) {
@@ -306,9 +296,7 @@
vp10_highbd_inv_txfm_add_32x32(dqcoeff, dst, stride, eob, xd->bd,
tx_type);
break;
- default:
- assert(0 && "Invalid transform size");
- return;
+ default: assert(0 && "Invalid transform size"); return;
}
} else {
#endif // CONFIG_VPX_HIGHBITDEPTH
@@ -326,9 +314,7 @@
case TX_32X32:
vp10_inv_txfm_add_32x32(dqcoeff, dst, stride, eob, tx_type);
break;
- default:
- assert(0 && "Invalid transform size");
- return;
+ default: assert(0 && "Invalid transform size"); return;
}
#if CONFIG_VPX_HIGHBITDEPTH
}
@@ -350,8 +336,7 @@
static void predict_and_reconstruct_intra_block(MACROBLOCKD *const xd,
vpx_reader *r,
MB_MODE_INFO *const mbmi,
- int plane,
- int row, int col,
+ int plane, int row, int col,
TX_SIZE tx_size) {
struct macroblockd_plane *const pd = &xd->plane[plane];
PREDICTION_MODE mode = (plane == 0) ? mbmi->mode : mbmi->uv_mode;
@@ -361,43 +346,42 @@
dst = &pd->dst.buf[4 * row * pd->dst.stride + 4 * col];
if (mbmi->sb_type < BLOCK_8X8)
- if (plane == 0)
- mode = xd->mi[0]->bmi[(row << 1) + col].as_mode;
+ if (plane == 0) mode = xd->mi[0]->bmi[(row << 1) + col].as_mode;
- vp10_predict_intra_block(xd, pd->n4_wl, pd->n4_hl, tx_size, mode,
- dst, pd->dst.stride, dst, pd->dst.stride,
- col, row, plane);
+ vp10_predict_intra_block(xd, pd->n4_wl, pd->n4_hl, tx_size, mode, dst,
+ pd->dst.stride, dst, pd->dst.stride, col, row,
+ plane);
if (!mbmi->skip) {
TX_TYPE tx_type = get_tx_type(plane_type, xd, block_idx);
const scan_order *sc = get_scan(tx_size, tx_type);
const int eob = vp10_decode_block_tokens(xd, plane, sc, col, row, tx_size,
r, mbmi->segment_id);
- inverse_transform_block_intra(xd, plane, tx_type, tx_size,
- dst, pd->dst.stride, eob);
+ inverse_transform_block_intra(xd, plane, tx_type, tx_size, dst,
+ pd->dst.stride, eob);
}
}
static int reconstruct_inter_block(MACROBLOCKD *const xd, vpx_reader *r,
- MB_MODE_INFO *const mbmi, int plane,
- int row, int col, TX_SIZE tx_size) {
+ MB_MODE_INFO *const mbmi, int plane, int row,
+ int col, TX_SIZE tx_size) {
struct macroblockd_plane *const pd = &xd->plane[plane];
PLANE_TYPE plane_type = (plane == 0) ? PLANE_TYPE_Y : PLANE_TYPE_UV;
int block_idx = (row << 1) + col;
TX_TYPE tx_type = get_tx_type(plane_type, xd, block_idx);
const scan_order *sc = get_scan(tx_size, tx_type);
const int eob = vp10_decode_block_tokens(xd, plane, sc, col, row, tx_size, r,
- mbmi->segment_id);
+ mbmi->segment_id);
- inverse_transform_block_inter(xd, plane, tx_size,
- &pd->dst.buf[4 * row * pd->dst.stride + 4 * col],
- pd->dst.stride, eob, block_idx);
+ inverse_transform_block_inter(
+ xd, plane, tx_size, &pd->dst.buf[4 * row * pd->dst.stride + 4 * col],
+ pd->dst.stride, eob, block_idx);
return eob;
}
-static void build_mc_border(const uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- int x, int y, int b_w, int b_h, int w, int h) {
+static void build_mc_border(const uint8_t *src, int src_stride, uint8_t *dst,
+ int dst_stride, int x, int y, int b_w, int b_h,
+ int w, int h) {
// Get a pointer to the start of the real data for this row.
const uint8_t *ref_row = src - x - y * src_stride;
@@ -410,39 +394,31 @@
int right = 0, copy;
int left = x < 0 ? -x : 0;
- if (left > b_w)
- left = b_w;
+ if (left > b_w) left = b_w;
- if (x + b_w > w)
- right = x + b_w - w;
+ if (x + b_w > w) right = x + b_w - w;
- if (right > b_w)
- right = b_w;
+ if (right > b_w) right = b_w;
copy = b_w - left - right;
- if (left)
- memset(dst, ref_row[0], left);
+ if (left) memset(dst, ref_row[0], left);
- if (copy)
- memcpy(dst + left, ref_row + x + left, copy);
+ if (copy) memcpy(dst + left, ref_row + x + left, copy);
- if (right)
- memset(dst + left + copy, ref_row[w - 1], right);
+ if (right) memset(dst + left + copy, ref_row[w - 1], right);
dst += dst_stride;
++y;
- if (y > 0 && y < h)
- ref_row += src_stride;
+ if (y > 0 && y < h) ref_row += src_stride;
} while (--b_h);
}
#if CONFIG_VPX_HIGHBITDEPTH
static void high_build_mc_border(const uint8_t *src8, int src_stride,
- uint16_t *dst, int dst_stride,
- int x, int y, int b_w, int b_h,
- int w, int h) {
+ uint16_t *dst, int dst_stride, int x, int y,
+ int b_w, int b_h, int w, int h) {
// Get a pointer to the start of the real data for this row.
const uint16_t *src = CONVERT_TO_SHORTPTR(src8);
const uint16_t *ref_row = src - x - y * src_stride;
@@ -456,31 +432,24 @@
int right = 0, copy;
int left = x < 0 ? -x : 0;
- if (left > b_w)
- left = b_w;
+ if (left > b_w) left = b_w;
- if (x + b_w > w)
- right = x + b_w - w;
+ if (x + b_w > w) right = x + b_w - w;
- if (right > b_w)
- right = b_w;
+ if (right > b_w) right = b_w;
copy = b_w - left - right;
- if (left)
- vpx_memset16(dst, ref_row[0], left);
+ if (left) vpx_memset16(dst, ref_row[0], left);
- if (copy)
- memcpy(dst + left, ref_row + x + left, copy * sizeof(uint16_t));
+ if (copy) memcpy(dst + left, ref_row + x + left, copy * sizeof(uint16_t));
- if (right)
- vpx_memset16(dst + left + copy, ref_row[w - 1], right);
+ if (right) vpx_memset16(dst + left + copy, ref_row[w - 1], right);
dst += dst_stride;
++y;
- if (y > 0 && y < h)
- ref_row += src_stride;
+ if (y > 0 && y < h) ref_row += src_stride;
} while (--b_h);
}
#endif // CONFIG_VPX_HIGHBITDEPTH
@@ -489,71 +458,65 @@
static void extend_and_predict(const uint8_t *buf_ptr1, int pre_buf_stride,
int x0, int y0, int b_w, int b_h,
int frame_width, int frame_height,
- int border_offset,
- uint8_t *const dst, int dst_buf_stride,
- int subpel_x, int subpel_y,
+ int border_offset, uint8_t *const dst,
+ int dst_buf_stride, int subpel_x, int subpel_y,
const InterpKernel *kernel,
- const struct scale_factors *sf,
- MACROBLOCKD *xd,
+ const struct scale_factors *sf, MACROBLOCKD *xd,
int w, int h, int ref, int xs, int ys) {
DECLARE_ALIGNED(16, uint16_t, mc_buf_high[80 * 2 * 80 * 2]);
const uint8_t *buf_ptr;
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- high_build_mc_border(buf_ptr1, pre_buf_stride, mc_buf_high, b_w,
- x0, y0, b_w, b_h, frame_width, frame_height);
+ high_build_mc_border(buf_ptr1, pre_buf_stride, mc_buf_high, b_w, x0, y0,
+ b_w, b_h, frame_width, frame_height);
buf_ptr = CONVERT_TO_BYTEPTR(mc_buf_high) + border_offset;
} else {
- build_mc_border(buf_ptr1, pre_buf_stride, (uint8_t *)mc_buf_high, b_w,
- x0, y0, b_w, b_h, frame_width, frame_height);
+ build_mc_border(buf_ptr1, pre_buf_stride, (uint8_t *)mc_buf_high, b_w, x0,
+ y0, b_w, b_h, frame_width, frame_height);
buf_ptr = ((uint8_t *)mc_buf_high) + border_offset;
}
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- high_inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x,
- subpel_y, sf, w, h, ref, kernel, xs, ys, xd->bd);
+ high_inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x, subpel_y,
+ sf, w, h, ref, kernel, xs, ys, xd->bd);
} else {
- inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x,
- subpel_y, sf, w, h, ref, kernel, xs, ys);
+ inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x, subpel_y, sf,
+ w, h, ref, kernel, xs, ys);
}
}
#else
static void extend_and_predict(const uint8_t *buf_ptr1, int pre_buf_stride,
int x0, int y0, int b_w, int b_h,
int frame_width, int frame_height,
- int border_offset,
- uint8_t *const dst, int dst_buf_stride,
- int subpel_x, int subpel_y,
+ int border_offset, uint8_t *const dst,
+ int dst_buf_stride, int subpel_x, int subpel_y,
const InterpKernel *kernel,
- const struct scale_factors *sf,
- int w, int h, int ref, int xs, int ys) {
+ const struct scale_factors *sf, int w, int h,
+ int ref, int xs, int ys) {
DECLARE_ALIGNED(16, uint8_t, mc_buf[80 * 2 * 80 * 2]);
const uint8_t *buf_ptr;
- build_mc_border(buf_ptr1, pre_buf_stride, mc_buf, b_w,
- x0, y0, b_w, b_h, frame_width, frame_height);
+ build_mc_border(buf_ptr1, pre_buf_stride, mc_buf, b_w, x0, y0, b_w, b_h,
+ frame_width, frame_height);
buf_ptr = mc_buf + border_offset;
- inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x,
- subpel_y, sf, w, h, ref, kernel, xs, ys);
+ inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x, subpel_y, sf, w,
+ h, ref, kernel, xs, ys);
}
#endif // CONFIG_VPX_HIGHBITDEPTH
-static void dec_build_inter_predictors(VP10Decoder *const pbi, MACROBLOCKD *xd,
- int plane, int bw, int bh, int x,
- int y, int w, int h, int mi_x, int mi_y,
- const InterpKernel *kernel,
- const struct scale_factors *sf,
- struct buf_2d *pre_buf,
- struct buf_2d *dst_buf, const MV* mv,
- RefCntBuffer *ref_frame_buf,
- int is_scaled, int ref) {
+static void dec_build_inter_predictors(
+ VP10Decoder *const pbi, MACROBLOCKD *xd, int plane, int bw, int bh, int x,
+ int y, int w, int h, int mi_x, int mi_y, const InterpKernel *kernel,
+ const struct scale_factors *sf, struct buf_2d *pre_buf,
+ struct buf_2d *dst_buf, const MV *mv, RefCntBuffer *ref_frame_buf,
+ int is_scaled, int ref) {
VP10_COMMON *const cm = &pbi->common;
struct macroblockd_plane *const pd = &xd->plane[plane];
uint8_t *const dst = dst_buf->buf + dst_buf->stride * y + x;
MV32 scaled_mv;
- int xs, ys, x0, y0, x0_16, y0_16, frame_width, frame_height,
- buf_stride, subpel_x, subpel_y;
+ int xs, ys, x0, y0, x0_16, y0_16, frame_width, frame_height, buf_stride,
+ subpel_x, subpel_y;
uint8_t *ref_frame, *buf_ptr;
// Get reference frame pointer, width and height.
@@ -564,14 +527,13 @@
} else {
frame_width = ref_frame_buf->buf.uv_crop_width;
frame_height = ref_frame_buf->buf.uv_crop_height;
- ref_frame = plane == 1 ? ref_frame_buf->buf.u_buffer
- : ref_frame_buf->buf.v_buffer;
+ ref_frame =
+ plane == 1 ? ref_frame_buf->buf.u_buffer : ref_frame_buf->buf.v_buffer;
}
if (is_scaled) {
- const MV mv_q4 = clamp_mv_to_umv_border_sb(xd, mv, bw, bh,
- pd->subsampling_x,
- pd->subsampling_y);
+ const MV mv_q4 = clamp_mv_to_umv_border_sb(
+ xd, mv, bw, bh, pd->subsampling_x, pd->subsampling_y);
// Co-ordinate of containing block to pixel precision.
int x_start = (-xd->mb_to_left_edge >> (3 + pd->subsampling_x));
int y_start = (-xd->mb_to_top_edge >> (3 + pd->subsampling_y));
@@ -623,8 +585,8 @@
// Do border extension if there is motion or the
// width/height is not a multiple of 8 pixels.
- if (is_scaled || scaled_mv.col || scaled_mv.row ||
- (frame_width & 0x7) || (frame_height & 0x7)) {
+ if (is_scaled || scaled_mv.col || scaled_mv.row || (frame_width & 0x7) ||
+ (frame_height & 0x7)) {
int y1 = ((y0_16 + (h - 1) * ys) >> SUBPEL_BITS) + 1;
// Get reference block bottom right horizontal coordinate.
@@ -658,11 +620,9 @@
const int b_h = y1 - y0 + 1;
const int border_offset = y_pad * 3 * b_w + x_pad * 3;
- extend_and_predict(buf_ptr1, buf_stride, x0, y0, b_w, b_h,
- frame_width, frame_height, border_offset,
- dst, dst_buf->stride,
- subpel_x, subpel_y,
- kernel, sf,
+ extend_and_predict(buf_ptr1, buf_stride, x0, y0, b_w, b_h, frame_width,
+ frame_height, border_offset, dst, dst_buf->stride,
+ subpel_x, subpel_y, kernel, sf,
#if CONFIG_VPX_HIGHBITDEPTH
xd,
#endif
@@ -672,11 +632,11 @@
} else {
// Wait until reference block is ready. Pad 7 more pixels as last 7
// pixels of each superblock row can be changed by next superblock row.
- if (cm->frame_parallel_decode) {
- const int y1 = (y0_16 + (h - 1) * ys) >> SUBPEL_BITS;
- vp10_frameworker_wait(pbi->frame_worker_owner, ref_frame_buf,
- VPXMAX(0, (y1 + 7)) << (plane == 0 ? 0 : 1));
- }
+ if (cm->frame_parallel_decode) {
+ const int y1 = (y0_16 + (h - 1) * ys) >> SUBPEL_BITS;
+ vp10_frameworker_wait(pbi->frame_worker_owner, ref_frame_buf,
+ VPXMAX(0, (y1 + 7)) << (plane == 0 ? 0 : 1));
+ }
}
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
@@ -687,14 +647,14 @@
subpel_y, sf, w, h, ref, kernel, xs, ys);
}
#else
- inter_predictor(buf_ptr, buf_stride, dst, dst_buf->stride, subpel_x,
- subpel_y, sf, w, h, ref, kernel, xs, ys);
+ inter_predictor(buf_ptr, buf_stride, dst, dst_buf->stride, subpel_x, subpel_y,
+ sf, w, h, ref, kernel, xs, ys);
#endif // CONFIG_VPX_HIGHBITDEPTH
}
static void dec_build_inter_predictors_sb(VP10Decoder *const pbi,
- MACROBLOCKD *xd,
- int mi_row, int mi_col) {
+ MACROBLOCKD *xd, int mi_row,
+ int mi_col) {
int plane;
const int mi_x = mi_col * MI_SIZE;
const int mi_y = mi_row * MI_SIZE;
@@ -733,28 +693,27 @@
for (y = 0; y < num_4x4_h; ++y) {
for (x = 0; x < num_4x4_w; ++x) {
const MV mv = average_split_mvs(pd, mi, ref, y * 2 + x);
- dec_build_inter_predictors(pbi, xd, plane, n4w_x4, n4h_x4,
- 4 * x, 4 * y, pw, ph, mi_x, mi_y, kernel,
- sf, pre_buf, dst_buf, &mv,
- ref_frame_buf, is_scaled, ref);
+ dec_build_inter_predictors(pbi, xd, plane, n4w_x4, n4h_x4, 4 * x,
+ 4 * y, pw, ph, mi_x, mi_y, kernel, sf,
+ pre_buf, dst_buf, &mv, ref_frame_buf,
+ is_scaled, ref);
}
}
} else {
const MV mv = mi->mbmi.mv[ref].as_mv;
- dec_build_inter_predictors(pbi, xd, plane, n4w_x4, n4h_x4,
- 0, 0, n4w_x4, n4h_x4, mi_x, mi_y, kernel,
- sf, pre_buf, dst_buf, &mv, ref_frame_buf,
- is_scaled, ref);
+ dec_build_inter_predictors(pbi, xd, plane, n4w_x4, n4h_x4, 0, 0, n4w_x4,
+ n4h_x4, mi_x, mi_y, kernel, sf, pre_buf,
+ dst_buf, &mv, ref_frame_buf, is_scaled, ref);
}
}
}
}
-static INLINE TX_SIZE dec_get_uv_tx_size(const MB_MODE_INFO *mbmi,
- int n4_wl, int n4_hl) {
+static INLINE TX_SIZE dec_get_uv_tx_size(const MB_MODE_INFO *mbmi, int n4_wl,
+ int n4_hl) {
// get minimum log2 num4x4s dimension
const int x = VPXMIN(n4_wl, n4_hl);
- return VPXMIN(mbmi->tx_size, x);
+ return VPXMIN(mbmi->tx_size, x);
}
static INLINE void dec_reset_skip_context(MACROBLOCKD *xd) {
@@ -779,8 +738,8 @@
static MB_MODE_INFO *set_offsets(VP10_COMMON *const cm, MACROBLOCKD *const xd,
BLOCK_SIZE bsize, int mi_row, int mi_col,
- int bw, int bh, int x_mis, int y_mis,
- int bwl, int bhl) {
+ int bw, int bh, int x_mis, int y_mis, int bwl,
+ int bhl) {
const int offset = mi_row * cm->mi_stride + mi_col;
int x, y;
const TileInfo *const tile = &xd->tile;
@@ -808,9 +767,8 @@
}
static void decode_block(VP10Decoder *const pbi, MACROBLOCKD *const xd,
- int mi_row, int mi_col,
- vpx_reader *r, BLOCK_SIZE bsize,
- int bwl, int bhl) {
+ int mi_row, int mi_col, vpx_reader *r,
+ BLOCK_SIZE bsize, int bwl, int bhl) {
VP10_COMMON *const cm = &pbi->common;
const int less8x8 = bsize < BLOCK_8X8;
const int bw = 1 << (bwl - 1);
@@ -818,15 +776,15 @@
const int x_mis = VPXMIN(bw, cm->mi_cols - mi_col);
const int y_mis = VPXMIN(bh, cm->mi_rows - mi_row);
- MB_MODE_INFO *mbmi = set_offsets(cm, xd, bsize, mi_row, mi_col,
- bw, bh, x_mis, y_mis, bwl, bhl);
+ MB_MODE_INFO *mbmi = set_offsets(cm, xd, bsize, mi_row, mi_col, bw, bh, x_mis,
+ y_mis, bwl, bhl);
if (bsize >= BLOCK_8X8 && (cm->subsampling_x || cm->subsampling_y)) {
const BLOCK_SIZE uv_subsize =
ss_size_lookup[bsize][cm->subsampling_x][cm->subsampling_y];
if (uv_subsize == BLOCK_INVALID)
- vpx_internal_error(xd->error_info,
- VPX_CODEC_CORRUPT_FRAME, "Invalid block size.");
+ vpx_internal_error(xd->error_info, VPX_CODEC_CORRUPT_FRAME,
+ "Invalid block size.");
}
vp10_read_mode_info(pbi, xd, mi_row, mi_col, r, x_mis, y_mis);
@@ -841,20 +799,24 @@
const struct macroblockd_plane *const pd = &xd->plane[plane];
const TX_SIZE tx_size =
plane ? dec_get_uv_tx_size(mbmi, pd->n4_wl, pd->n4_hl)
- : mbmi->tx_size;
+ : mbmi->tx_size;
const int num_4x4_w = pd->n4_w;
const int num_4x4_h = pd->n4_h;
const int step = (1 << tx_size);
int row, col;
- const int max_blocks_wide = num_4x4_w + (xd->mb_to_right_edge >= 0 ?
- 0 : xd->mb_to_right_edge >> (5 + pd->subsampling_x));
- const int max_blocks_high = num_4x4_h + (xd->mb_to_bottom_edge >= 0 ?
- 0 : xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
+ const int max_blocks_wide =
+ num_4x4_w + (xd->mb_to_right_edge >= 0
+ ? 0
+ : xd->mb_to_right_edge >> (5 + pd->subsampling_x));
+ const int max_blocks_high =
+ num_4x4_h + (xd->mb_to_bottom_edge >= 0
+ ? 0
+ : xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
for (row = 0; row < max_blocks_high; row += step)
for (col = 0; col < max_blocks_wide; col += step)
- predict_and_reconstruct_intra_block(xd, r, mbmi, plane,
- row, col, tx_size);
+ predict_and_reconstruct_intra_block(xd, r, mbmi, plane, row, col,
+ tx_size);
}
} else {
// Prediction
@@ -869,20 +831,24 @@
const struct macroblockd_plane *const pd = &xd->plane[plane];
const TX_SIZE tx_size =
plane ? dec_get_uv_tx_size(mbmi, pd->n4_wl, pd->n4_hl)
- : mbmi->tx_size;
+ : mbmi->tx_size;
const int num_4x4_w = pd->n4_w;
const int num_4x4_h = pd->n4_h;
const int step = (1 << tx_size);
int row, col;
- const int max_blocks_wide = num_4x4_w + (xd->mb_to_right_edge >= 0 ?
- 0 : xd->mb_to_right_edge >> (5 + pd->subsampling_x));
- const int max_blocks_high = num_4x4_h + (xd->mb_to_bottom_edge >= 0 ?
- 0 : xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
+ const int max_blocks_wide =
+ num_4x4_w + (xd->mb_to_right_edge >= 0
+ ? 0
+ : xd->mb_to_right_edge >> (5 + pd->subsampling_x));
+ const int max_blocks_high =
+ num_4x4_h +
+ (xd->mb_to_bottom_edge >= 0 ? 0 : xd->mb_to_bottom_edge >>
+ (5 + pd->subsampling_y));
for (row = 0; row < max_blocks_high; row += step)
for (col = 0; col < max_blocks_wide; col += step)
- eobtotal += reconstruct_inter_block(xd, r, mbmi, plane, row, col,
- tx_size);
+ eobtotal +=
+ reconstruct_inter_block(xd, r, mbmi, plane, row, col, tx_size);
}
if (!less8x8 && eobtotal == 0)
@@ -897,21 +863,19 @@
xd->corrupted |= vpx_reader_has_error(r);
}
-static INLINE int dec_partition_plane_context(const MACROBLOCKD *xd,
- int mi_row, int mi_col,
- int bsl) {
+static INLINE int dec_partition_plane_context(const MACROBLOCKD *xd, int mi_row,
+ int mi_col, int bsl) {
const PARTITION_CONTEXT *above_ctx = xd->above_seg_context + mi_col;
const PARTITION_CONTEXT *left_ctx = xd->left_seg_context + (mi_row & MI_MASK);
- int above = (*above_ctx >> bsl) & 1 , left = (*left_ctx >> bsl) & 1;
+ int above = (*above_ctx >> bsl) & 1, left = (*left_ctx >> bsl) & 1;
-// assert(bsl >= 0);
+ // assert(bsl >= 0);
return (left * 2 + above) + bsl * PARTITION_PLOFFSET;
}
-static INLINE void dec_update_partition_context(MACROBLOCKD *xd,
- int mi_row, int mi_col,
- BLOCK_SIZE subsize,
+static INLINE void dec_update_partition_context(MACROBLOCKD *xd, int mi_row,
+ int mi_col, BLOCK_SIZE subsize,
int bw) {
PARTITION_CONTEXT *const above_ctx = xd->above_seg_context + mi_col;
PARTITION_CONTEXT *const left_ctx = xd->left_seg_context + (mi_row & MI_MASK);
@@ -940,16 +904,15 @@
else
p = PARTITION_SPLIT;
- if (counts)
- ++counts->partition[ctx][p];
+ if (counts) ++counts->partition[ctx][p];
return p;
}
// TODO(slavarnway): eliminate bsize and subsize in future commits
static void decode_partition(VP10Decoder *const pbi, MACROBLOCKD *const xd,
- int mi_row, int mi_col,
- vpx_reader* r, BLOCK_SIZE bsize, int n4x4_l2) {
+ int mi_row, int mi_col, vpx_reader *r,
+ BLOCK_SIZE bsize, int n4x4_l2) {
VP10_COMMON *const cm = &pbi->common;
const int n8x8_l2 = n4x4_l2 - 1;
const int num_8x8_wh = 1 << n8x8_l2;
@@ -959,11 +922,10 @@
const int has_rows = (mi_row + hbs) < cm->mi_rows;
const int has_cols = (mi_col + hbs) < cm->mi_cols;
- if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
- return;
+ if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols) return;
- partition = read_partition(cm, xd, mi_row, mi_col, r, has_rows, has_cols,
- n8x8_l2);
+ partition =
+ read_partition(cm, xd, mi_row, mi_col, r, has_rows, has_cols, n8x8_l2);
subsize = subsize_lookup[partition][bsize]; // get_subsize(bsize, partition);
if (!hbs) {
// calculate bmode block dimensions (log 2)
@@ -994,8 +956,7 @@
decode_partition(pbi, xd, mi_row + hbs, mi_col + hbs, r, subsize,
n8x8_l2);
break;
- default:
- assert(0 && "Invalid partition type");
+ default: assert(0 && "Invalid partition type");
}
}
@@ -1005,12 +966,10 @@
dec_update_partition_context(xd, mi_row, mi_col, subsize, num_8x8_wh);
}
-static void setup_token_decoder(const uint8_t *data,
- const uint8_t *data_end,
+static void setup_token_decoder(const uint8_t *data, const uint8_t *data_end,
size_t read_size,
struct vpx_internal_error_info *error_info,
- vpx_reader *r,
- vpx_decrypt_cb decrypt_cb,
+ vpx_reader *r, vpx_decrypt_cb decrypt_cb,
void *decrypt_state) {
// Validate the calculated partition length. If the buffer
// described by the partition can't be fully read, then restrict
@@ -1037,12 +996,11 @@
vp10_diff_update_prob(r, &coef_probs[i][j][k][l][m]);
}
-static void read_coef_probs(FRAME_CONTEXT *fc, TX_MODE tx_mode,
- vpx_reader *r) {
- const TX_SIZE max_tx_size = tx_mode_to_biggest_tx_size[tx_mode];
- TX_SIZE tx_size;
- for (tx_size = TX_4X4; tx_size <= max_tx_size; ++tx_size)
- read_coef_probs_common(fc->coef_probs[tx_size], r);
+static void read_coef_probs(FRAME_CONTEXT *fc, TX_MODE tx_mode, vpx_reader *r) {
+ const TX_SIZE max_tx_size = tx_mode_to_biggest_tx_size[tx_mode];
+ TX_SIZE tx_size;
+ for (tx_size = TX_4X4; tx_size <= max_tx_size; ++tx_size)
+ read_coef_probs_common(fc->coef_probs[tx_size], r);
}
static void setup_segmentation(VP10_COMMON *const cm,
@@ -1057,8 +1015,7 @@
seg->update_data = 0;
seg->enabled = vpx_rb_read_bit(rb);
- if (!seg->enabled)
- return;
+ if (!seg->enabled) return;
// Segmentation map update
if (frame_is_intra_only(cm) || cm->error_resilient_mode) {
@@ -1069,8 +1026,8 @@
if (seg->update_map) {
#if !CONFIG_MISC_FIXES
for (i = 0; i < SEG_TREE_PROBS; i++)
- segp->tree_probs[i] = vpx_rb_read_bit(rb) ? vpx_rb_read_literal(rb, 8)
- : MAX_PROB;
+ segp->tree_probs[i] =
+ vpx_rb_read_bit(rb) ? vpx_rb_read_literal(rb, 8) : MAX_PROB;
#endif
if (frame_is_intra_only(cm) || cm->error_resilient_mode) {
seg->temporal_update = 0;
@@ -1080,11 +1037,10 @@
#if !CONFIG_MISC_FIXES
if (seg->temporal_update) {
for (i = 0; i < PREDICTION_PROBS; i++)
- segp->pred_probs[i] = vpx_rb_read_bit(rb) ? vpx_rb_read_literal(rb, 8)
- : MAX_PROB;
+ segp->pred_probs[i] =
+ vpx_rb_read_bit(rb) ? vpx_rb_read_literal(rb, 8) : MAX_PROB;
} else {
- for (i = 0; i < PREDICTION_PROBS; i++)
- segp->pred_probs[i] = MAX_PROB;
+ for (i = 0; i < PREDICTION_PROBS; i++) segp->pred_probs[i] = MAX_PROB;
}
#endif
}
@@ -1139,8 +1095,9 @@
}
static INLINE int read_delta_q(struct vpx_read_bit_buffer *rb) {
- return vpx_rb_read_bit(rb) ?
- vpx_rb_read_inv_signed_literal(rb, CONFIG_MISC_FIXES ? 6 : 4) : 0;
+ return vpx_rb_read_bit(rb)
+ ? vpx_rb_read_inv_signed_literal(rb, CONFIG_MISC_FIXES ? 6 : 4)
+ : 0;
}
static void setup_quantization(VP10_COMMON *const cm,
@@ -1158,24 +1115,25 @@
int i;
for (i = 0; i < MAX_SEGMENTS; ++i) {
const int qindex = vp10_get_qindex(&cm->seg, i, cm->base_qindex);
- cm->y_dequant[i][0] = vp10_dc_quant(qindex, cm->y_dc_delta_q,
- cm->bit_depth);
+ cm->y_dequant[i][0] =
+ vp10_dc_quant(qindex, cm->y_dc_delta_q, cm->bit_depth);
cm->y_dequant[i][1] = vp10_ac_quant(qindex, 0, cm->bit_depth);
- cm->uv_dequant[i][0] = vp10_dc_quant(qindex, cm->uv_dc_delta_q,
- cm->bit_depth);
- cm->uv_dequant[i][1] = vp10_ac_quant(qindex, cm->uv_ac_delta_q,
- cm->bit_depth);
+ cm->uv_dequant[i][0] =
+ vp10_dc_quant(qindex, cm->uv_dc_delta_q, cm->bit_depth);
+ cm->uv_dequant[i][1] =
+ vp10_ac_quant(qindex, cm->uv_ac_delta_q, cm->bit_depth);
}
} else {
const int qindex = cm->base_qindex;
// When segmentation is disabled, only the first value is used. The
// remaining are don't cares.
- cm->y_dequant[0][0] = vp10_dc_quant(qindex, cm->y_dc_delta_q, cm->bit_depth);
+ cm->y_dequant[0][0] =
+ vp10_dc_quant(qindex, cm->y_dc_delta_q, cm->bit_depth);
cm->y_dequant[0][1] = vp10_ac_quant(qindex, 0, cm->bit_depth);
- cm->uv_dequant[0][0] = vp10_dc_quant(qindex, cm->uv_dc_delta_q,
- cm->bit_depth);
- cm->uv_dequant[0][1] = vp10_ac_quant(qindex, cm->uv_ac_delta_q,
- cm->bit_depth);
+ cm->uv_dequant[0][0] =
+ vp10_dc_quant(qindex, cm->uv_dc_delta_q, cm->bit_depth);
+ cm->uv_dequant[0][1] =
+ vp10_ac_quant(qindex, cm->uv_ac_delta_q, cm->bit_depth);
}
}
@@ -1183,8 +1141,7 @@
return vpx_rb_read_bit(rb) ? SWITCHABLE : vpx_rb_read_literal(rb, 2);
}
-static void setup_render_size(VP10_COMMON *cm,
- struct vpx_read_bit_buffer *rb) {
+static void setup_render_size(VP10_COMMON *cm, struct vpx_read_bit_buffer *rb) {
cm->render_width = cm->width;
cm->render_height = cm->height;
if (vpx_rb_read_bit(rb))
@@ -1210,7 +1167,7 @@
const int new_mi_rows =
ALIGN_POWER_OF_TWO(height, MI_SIZE_LOG2) >> MI_SIZE_LOG2;
const int new_mi_cols =
- ALIGN_POWER_OF_TWO(width, MI_SIZE_LOG2) >> MI_SIZE_LOG2;
+ ALIGN_POWER_OF_TWO(width, MI_SIZE_LOG2) >> MI_SIZE_LOG2;
// Allocations in vp10_alloc_context_buffers() depend on individual
// dimensions as well as the overall size.
@@ -1240,13 +1197,12 @@
lock_buffer_pool(pool);
if (vpx_realloc_frame_buffer(
- get_frame_new_buffer(cm), cm->width, cm->height,
- cm->subsampling_x, cm->subsampling_y,
+ get_frame_new_buffer(cm), cm->width, cm->height, cm->subsampling_x,
+ cm->subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
cm->use_highbitdepth,
#endif
- VPX_DEC_BORDER_IN_PIXELS,
- cm->byte_alignment,
+ VPX_DEC_BORDER_IN_PIXELS, cm->byte_alignment,
&pool->frame_bufs[cm->new_fb_idx].raw_frame_buffer, pool->get_fb_cb,
pool->cb_priv)) {
unlock_buffer_pool(pool);
@@ -1260,7 +1216,7 @@
pool->frame_bufs[cm->new_fb_idx].buf.bit_depth = (unsigned int)cm->bit_depth;
pool->frame_bufs[cm->new_fb_idx].buf.color_space = cm->color_space;
pool->frame_bufs[cm->new_fb_idx].buf.color_range = cm->color_range;
- pool->frame_bufs[cm->new_fb_idx].buf.render_width = cm->render_width;
+ pool->frame_bufs[cm->new_fb_idx].buf.render_width = cm->render_width;
pool->frame_bufs[cm->new_fb_idx].buf.render_height = cm->render_height;
}
@@ -1307,22 +1263,19 @@
// has valid dimensions.
for (i = 0; i < REFS_PER_FRAME; ++i) {
RefBuffer *const ref_frame = &cm->frame_refs[i];
- has_valid_ref_frame |= valid_ref_frame_size(ref_frame->buf->y_crop_width,
- ref_frame->buf->y_crop_height,
- width, height);
+ has_valid_ref_frame |=
+ valid_ref_frame_size(ref_frame->buf->y_crop_width,
+ ref_frame->buf->y_crop_height, width, height);
}
if (!has_valid_ref_frame)
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
"Referenced frame has invalid size");
for (i = 0; i < REFS_PER_FRAME; ++i) {
RefBuffer *const ref_frame = &cm->frame_refs[i];
- if (!valid_ref_frame_img_fmt(
- ref_frame->buf->bit_depth,
- ref_frame->buf->subsampling_x,
- ref_frame->buf->subsampling_y,
- cm->bit_depth,
- cm->subsampling_x,
- cm->subsampling_y))
+ if (!valid_ref_frame_img_fmt(ref_frame->buf->bit_depth,
+ ref_frame->buf->subsampling_x,
+ ref_frame->buf->subsampling_y, cm->bit_depth,
+ cm->subsampling_x, cm->subsampling_y))
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
"Referenced frame has incompatible color format");
}
@@ -1334,13 +1287,12 @@
lock_buffer_pool(pool);
if (vpx_realloc_frame_buffer(
- get_frame_new_buffer(cm), cm->width, cm->height,
- cm->subsampling_x, cm->subsampling_y,
+ get_frame_new_buffer(cm), cm->width, cm->height, cm->subsampling_x,
+ cm->subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
cm->use_highbitdepth,
#endif
- VPX_DEC_BORDER_IN_PIXELS,
- cm->byte_alignment,
+ VPX_DEC_BORDER_IN_PIXELS, cm->byte_alignment,
&pool->frame_bufs[cm->new_fb_idx].raw_frame_buffer, pool->get_fb_cb,
pool->cb_priv)) {
unlock_buffer_pool(pool);
@@ -1354,7 +1306,7 @@
pool->frame_bufs[cm->new_fb_idx].buf.bit_depth = (unsigned int)cm->bit_depth;
pool->frame_bufs[cm->new_fb_idx].buf.color_space = cm->color_space;
pool->frame_bufs[cm->new_fb_idx].buf.color_range = cm->color_range;
- pool->frame_bufs[cm->new_fb_idx].buf.render_width = cm->render_width;
+ pool->frame_bufs[cm->new_fb_idx].buf.render_width = cm->render_width;
pool->frame_bufs[cm->new_fb_idx].buf.render_height = cm->render_height;
}
@@ -1365,8 +1317,7 @@
// columns
max_ones = max_log2_tile_cols - min_log2_tile_cols;
cm->log2_tile_cols = min_log2_tile_cols;
- while (max_ones-- && vpx_rb_read_bit(rb))
- cm->log2_tile_cols++;
+ while (max_ones-- && vpx_rb_read_bit(rb)) cm->log2_tile_cols++;
if (cm->log2_tile_cols > 6)
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
@@ -1374,8 +1325,7 @@
// rows
cm->log2_tile_rows = vpx_rb_read_bit(rb);
- if (cm->log2_tile_rows)
- cm->log2_tile_rows += vpx_rb_read_bit(rb);
+ if (cm->log2_tile_rows) cm->log2_tile_rows += vpx_rb_read_bit(rb);
#if CONFIG_MISC_FIXES
// tile size magnitude
@@ -1395,14 +1345,10 @@
static int mem_get_varsize(const uint8_t *data, const int mag) {
switch (mag) {
- case 0:
- return data[0];
- case 1:
- return mem_get_le16(data);
- case 2:
- return mem_get_le24(data);
- case 3:
- return mem_get_le32(data);
+ case 0: return data[0];
+ case 1: return mem_get_le16(data);
+ case 2: return mem_get_le24(data);
+ case 3: return mem_get_le32(data);
}
assert("Invalid tile size marker value" && 0);
@@ -1415,9 +1361,8 @@
static void get_tile_buffer(const uint8_t *const data_end,
const int tile_sz_mag, int is_last,
struct vpx_internal_error_info *error_info,
- const uint8_t **data,
- vpx_decrypt_cb decrypt_cb, void *decrypt_state,
- TileBuffer *buf) {
+ const uint8_t **data, vpx_decrypt_cb decrypt_cb,
+ void *decrypt_state, TileBuffer *buf) {
size_t size;
if (!is_last) {
@@ -1447,9 +1392,9 @@
*data += size;
}
-static void get_tile_buffers(VP10Decoder *pbi,
- const uint8_t *data, const uint8_t *data_end,
- int tile_cols, int tile_rows,
+static void get_tile_buffers(VP10Decoder *pbi, const uint8_t *data,
+ const uint8_t *data_end, int tile_cols,
+ int tile_rows,
TileBuffer (*tile_buffers)[1 << 6]) {
int r, c;
@@ -1458,15 +1403,14 @@
const int is_last = (r == tile_rows - 1) && (c == tile_cols - 1);
TileBuffer *const buf = &tile_buffers[r][c];
buf->col = c;
- get_tile_buffer(data_end, pbi->common.tile_sz_mag,
- is_last, &pbi->common.error, &data,
- pbi->decrypt_cb, pbi->decrypt_state, buf);
+ get_tile_buffer(data_end, pbi->common.tile_sz_mag, is_last,
+ &pbi->common.error, &data, pbi->decrypt_cb,
+ pbi->decrypt_state, buf);
}
}
}
-static const uint8_t *decode_tiles(VP10Decoder *pbi,
- const uint8_t *data,
+static const uint8_t *decode_tiles(VP10Decoder *pbi, const uint8_t *data,
const uint8_t *data_end) {
VP10_COMMON *const cm = &pbi->common;
const VPxWorkerInterface *const winterface = vpx_get_worker_interface();
@@ -1490,11 +1434,11 @@
}
if (cm->lf.filter_level && !cm->skip_loop_filter) {
- LFWorkerData *const lf_data = (LFWorkerData*)pbi->lf_worker.data1;
+ LFWorkerData *const lf_data = (LFWorkerData *)pbi->lf_worker.data1;
// Be sure to sync as we might be resuming after a failed frame decode.
winterface->sync(&pbi->lf_worker);
vp10_loop_filter_data_reset(lf_data, get_frame_new_buffer(cm), cm,
- pbi->mb.plane);
+ pbi->mb.plane);
}
assert(tile_rows <= 4);
@@ -1510,12 +1454,10 @@
get_tile_buffers(pbi, data, data_end, tile_cols, tile_rows, tile_buffers);
- if (pbi->tile_data == NULL ||
- (tile_cols * tile_rows) != pbi->total_tiles) {
+ if (pbi->tile_data == NULL || (tile_cols * tile_rows) != pbi->total_tiles) {
vpx_free(pbi->tile_data);
CHECK_MEM_ERROR(
- cm,
- pbi->tile_data,
+ cm, pbi->tile_data,
vpx_memalign(32, tile_cols * tile_rows * (sizeof(*pbi->tile_data))));
pbi->total_tiles = tile_rows * tile_cols;
}
@@ -1529,8 +1471,9 @@
tile_data->xd = pbi->mb;
tile_data->xd.corrupted = 0;
tile_data->xd.counts =
- cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_BACKWARD ?
- &cm->counts : NULL;
+ cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_BACKWARD
+ ? &cm->counts
+ : NULL;
vp10_zero(tile_data->dqcoeff);
vp10_tile_init(&tile_data->xd.tile, tile_data->cm, tile_row, tile_col);
setup_token_decoder(buf->data, data_end, buf->size, &cm->error,
@@ -1548,26 +1491,26 @@
for (mi_row = tile.mi_row_start; mi_row < tile.mi_row_end;
mi_row += MI_BLOCK_SIZE) {
for (tile_col = 0; tile_col < tile_cols; ++tile_col) {
- const int col = pbi->inv_tile_order ?
- tile_cols - tile_col - 1 : tile_col;
+ const int col =
+ pbi->inv_tile_order ? tile_cols - tile_col - 1 : tile_col;
tile_data = pbi->tile_data + tile_cols * tile_row + col;
vp10_tile_set_col(&tile, tile_data->cm, col);
vp10_zero(tile_data->xd.left_context);
vp10_zero(tile_data->xd.left_seg_context);
for (mi_col = tile.mi_col_start; mi_col < tile.mi_col_end;
mi_col += MI_BLOCK_SIZE) {
- decode_partition(pbi, &tile_data->xd, mi_row,
- mi_col, &tile_data->bit_reader, BLOCK_64X64, 4);
+ decode_partition(pbi, &tile_data->xd, mi_row, mi_col,
+ &tile_data->bit_reader, BLOCK_64X64, 4);
}
pbi->mb.corrupted |= tile_data->xd.corrupted;
if (pbi->mb.corrupted)
- vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
- "Failed to decode tile data");
+ vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
+ "Failed to decode tile data");
}
// Loopfilter one row.
if (cm->lf.filter_level && !cm->skip_loop_filter) {
const int lf_start = mi_row - MI_BLOCK_SIZE;
- LFWorkerData *const lf_data = (LFWorkerData*)pbi->lf_worker.data1;
+ LFWorkerData *const lf_data = (LFWorkerData *)pbi->lf_worker.data1;
// delay the loopfilter by 1 macroblock row.
if (lf_start < 0) continue;
@@ -1588,14 +1531,13 @@
// still be changed by the longest loopfilter of the next superblock
// row.
if (cm->frame_parallel_decode)
- vp10_frameworker_broadcast(pbi->cur_buf,
- mi_row << MI_BLOCK_SIZE_LOG2);
+ vp10_frameworker_broadcast(pbi->cur_buf, mi_row << MI_BLOCK_SIZE_LOG2);
}
}
// Loopfilter remaining rows in the frame.
if (cm->lf.filter_level && !cm->skip_loop_filter) {
- LFWorkerData *const lf_data = (LFWorkerData*)pbi->lf_worker.data1;
+ LFWorkerData *const lf_data = (LFWorkerData *)pbi->lf_worker.data1;
winterface->sync(&pbi->lf_worker);
lf_data->start = lf_data->stop;
lf_data->stop = cm->mi_rows;
@@ -1629,9 +1571,8 @@
vp10_zero(tile_data->xd.left_seg_context);
for (mi_col = tile->mi_col_start; mi_col < tile->mi_col_end;
mi_col += MI_BLOCK_SIZE) {
- decode_partition(tile_data->pbi, &tile_data->xd,
- mi_row, mi_col, &tile_data->bit_reader,
- BLOCK_64X64, 4);
+ decode_partition(tile_data->pbi, &tile_data->xd, mi_row, mi_col,
+ &tile_data->bit_reader, BLOCK_64X64, 4);
}
}
return !tile_data->xd.corrupted;
@@ -1639,13 +1580,12 @@
// sorts in descending order
static int compare_tile_buffers(const void *a, const void *b) {
- const TileBuffer *const buf1 = (const TileBuffer*)a;
- const TileBuffer *const buf2 = (const TileBuffer*)b;
+ const TileBuffer *const buf1 = (const TileBuffer *)a;
+ const TileBuffer *const buf2 = (const TileBuffer *)b;
return (int)(buf2->size - buf1->size);
}
-static const uint8_t *decode_tiles_mt(VP10Decoder *pbi,
- const uint8_t *data,
+static const uint8_t *decode_tiles_mt(VP10Decoder *pbi, const uint8_t *data,
const uint8_t *data_end) {
VP10_COMMON *const cm = &pbi->common;
const VPxWorkerInterface *const winterface = vpx_get_worker_interface();
@@ -1672,9 +1612,9 @@
// Ensure tile data offsets will be properly aligned. This may fail on
// platforms without DECLARE_ALIGNED().
assert((sizeof(*pbi->tile_worker_data) % 16) == 0);
- CHECK_MEM_ERROR(cm, pbi->tile_worker_data,
- vpx_memalign(32, num_threads *
- sizeof(*pbi->tile_worker_data)));
+ CHECK_MEM_ERROR(
+ cm, pbi->tile_worker_data,
+ vpx_memalign(32, num_threads * sizeof(*pbi->tile_worker_data)));
CHECK_MEM_ERROR(cm, pbi->tile_worker_info,
vpx_malloc(num_threads * sizeof(*pbi->tile_worker_info)));
for (i = 0; i < num_threads; ++i) {
@@ -1734,7 +1674,7 @@
for (i = 0; i < num_workers; ++i) {
TileWorkerData *const tile_data =
- (TileWorkerData*)pbi->tile_workers[i].data1;
+ (TileWorkerData *)pbi->tile_workers[i].data1;
vp10_zero(tile_data->counts);
}
}
@@ -1744,16 +1684,17 @@
int i;
for (i = 0; i < num_workers && n < tile_cols; ++i) {
VPxWorker *const worker = &pbi->tile_workers[i];
- TileWorkerData *const tile_data = (TileWorkerData*)worker->data1;
- TileInfo *const tile = (TileInfo*)worker->data2;
+ TileWorkerData *const tile_data = (TileWorkerData *)worker->data1;
+ TileInfo *const tile = (TileInfo *)worker->data2;
TileBuffer *const buf = &tile_buffers[0][n];
tile_data->pbi = pbi;
tile_data->xd = pbi->mb;
tile_data->xd.corrupted = 0;
tile_data->xd.counts =
- cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_BACKWARD ?
- &tile_data->counts : NULL;
+ cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_BACKWARD
+ ? &tile_data->counts
+ : NULL;
vp10_zero(tile_data->dqcoeff);
vp10_tile_init(tile, cm, 0, buf->col);
vp10_tile_init(&tile_data->xd.tile, cm, 0, buf->col);
@@ -1788,7 +1729,7 @@
}
if (final_worker > -1) {
TileWorkerData *const tile_data =
- (TileWorkerData*)pbi->tile_workers[final_worker].data1;
+ (TileWorkerData *)pbi->tile_workers[final_worker].data1;
bit_reader_end = vpx_reader_find_end(&tile_data->bit_reader);
final_worker = -1;
}
@@ -1798,7 +1739,7 @@
cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_BACKWARD) {
for (i = 0; i < num_workers; ++i) {
TileWorkerData *const tile_data =
- (TileWorkerData*)pbi->tile_workers[i].data1;
+ (TileWorkerData *)pbi->tile_workers[i].data1;
vp10_accumulate_frame_counts(cm, &tile_data->counts, 1);
}
}
@@ -1812,8 +1753,8 @@
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME, "Truncated packet");
}
-static void read_bitdepth_colorspace_sampling(
- VP10_COMMON *cm, struct vpx_read_bit_buffer *rb) {
+static void read_bitdepth_colorspace_sampling(VP10_COMMON *cm,
+ struct vpx_read_bit_buffer *rb) {
if (cm->profile >= PROFILE_2) {
cm->bit_depth = vpx_rb_read_bit(rb) ? VPX_BITS_12 : VPX_BITS_10;
#if CONFIG_VPX_HIGHBITDEPTH
@@ -1869,8 +1810,8 @@
cm->last_intra_only = cm->intra_only;
if (vpx_rb_read_literal(rb, 2) != VPX_FRAME_MARKER)
- vpx_internal_error(&cm->error, VPX_CODEC_UNSUP_BITSTREAM,
- "Invalid frame marker");
+ vpx_internal_error(&cm->error, VPX_CODEC_UNSUP_BITSTREAM,
+ "Invalid frame marker");
cm->profile = vp10_read_profile(rb);
#if CONFIG_VPX_HIGHBITDEPTH
@@ -1908,7 +1849,7 @@
return 0;
}
- cm->frame_type = (FRAME_TYPE) vpx_rb_read_bit(rb);
+ cm->frame_type = (FRAME_TYPE)vpx_rb_read_bit(rb);
cm->show_frame = vpx_rb_read_bit(rb);
cm->error_resilient_mode = vpx_rb_read_bit(rb);
@@ -1934,21 +1875,21 @@
cm->intra_only = cm->show_frame ? 0 : vpx_rb_read_bit(rb);
if (cm->error_resilient_mode) {
- cm->reset_frame_context = RESET_FRAME_CONTEXT_ALL;
+ cm->reset_frame_context = RESET_FRAME_CONTEXT_ALL;
} else {
#if CONFIG_MISC_FIXES
if (cm->intra_only) {
- cm->reset_frame_context =
- vpx_rb_read_bit(rb) ? RESET_FRAME_CONTEXT_ALL
- : RESET_FRAME_CONTEXT_CURRENT;
- } else {
- cm->reset_frame_context =
- vpx_rb_read_bit(rb) ? RESET_FRAME_CONTEXT_CURRENT
- : RESET_FRAME_CONTEXT_NONE;
- if (cm->reset_frame_context == RESET_FRAME_CONTEXT_CURRENT)
- cm->reset_frame_context =
- vpx_rb_read_bit(rb) ? RESET_FRAME_CONTEXT_ALL
+ cm->reset_frame_context = vpx_rb_read_bit(rb)
+ ? RESET_FRAME_CONTEXT_ALL
: RESET_FRAME_CONTEXT_CURRENT;
+ } else {
+ cm->reset_frame_context = vpx_rb_read_bit(rb)
+ ? RESET_FRAME_CONTEXT_CURRENT
+ : RESET_FRAME_CONTEXT_NONE;
+ if (cm->reset_frame_context == RESET_FRAME_CONTEXT_CURRENT)
+ cm->reset_frame_context = vpx_rb_read_bit(rb)
+ ? RESET_FRAME_CONTEXT_ALL
+ : RESET_FRAME_CONTEXT_CURRENT;
}
#else
static const RESET_FRAME_CONTEXT_MODE reset_frame_context_conv_tbl[4] = {
@@ -1991,7 +1932,7 @@
memset(&cm->ref_frame_map, -1, sizeof(cm->ref_frame_map));
pbi->need_resync = 0;
}
- } else if (pbi->need_resync != 1) { /* Skip if need resync */
+ } else if (pbi->need_resync != 1) { /* Skip if need resync */
pbi->refresh_frame_flags = vpx_rb_read_literal(rb, REF_FRAMES);
for (i = 0; i < REFS_PER_FRAME; ++i) {
const int ref = vpx_rb_read_literal(rb, REF_FRAMES_LOG2);
@@ -2010,16 +1951,14 @@
for (i = 0; i < REFS_PER_FRAME; ++i) {
RefBuffer *const ref_buf = &cm->frame_refs[i];
#if CONFIG_VPX_HIGHBITDEPTH
- vp10_setup_scale_factors_for_frame(&ref_buf->sf,
- ref_buf->buf->y_crop_width,
- ref_buf->buf->y_crop_height,
- cm->width, cm->height,
- cm->use_highbitdepth);
+ vp10_setup_scale_factors_for_frame(
+ &ref_buf->sf, ref_buf->buf->y_crop_width,
+ ref_buf->buf->y_crop_height, cm->width, cm->height,
+ cm->use_highbitdepth);
#else
- vp10_setup_scale_factors_for_frame(&ref_buf->sf,
- ref_buf->buf->y_crop_width,
- ref_buf->buf->y_crop_height,
- cm->width, cm->height);
+ vp10_setup_scale_factors_for_frame(
+ &ref_buf->sf, ref_buf->buf->y_crop_width,
+ ref_buf->buf->y_crop_height, cm->width, cm->height);
#endif
}
}
@@ -2029,7 +1968,7 @@
#endif
get_frame_new_buffer(cm)->color_space = cm->color_space;
get_frame_new_buffer(cm)->color_range = cm->color_range;
- get_frame_new_buffer(cm)->render_width = cm->render_width;
+ get_frame_new_buffer(cm)->render_width = cm->render_width;
get_frame_new_buffer(cm)->render_height = cm->render_height;
if (pbi->need_resync) {
@@ -2039,13 +1978,13 @@
}
if (!cm->error_resilient_mode) {
- cm->refresh_frame_context =
- vpx_rb_read_bit(rb) ? REFRESH_FRAME_CONTEXT_FORWARD
- : REFRESH_FRAME_CONTEXT_OFF;
+ cm->refresh_frame_context = vpx_rb_read_bit(rb)
+ ? REFRESH_FRAME_CONTEXT_FORWARD
+ : REFRESH_FRAME_CONTEXT_OFF;
if (cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_FORWARD) {
- cm->refresh_frame_context =
- vpx_rb_read_bit(rb) ? REFRESH_FRAME_CONTEXT_FORWARD
- : REFRESH_FRAME_CONTEXT_BACKWARD;
+ cm->refresh_frame_context = vpx_rb_read_bit(rb)
+ ? REFRESH_FRAME_CONTEXT_FORWARD
+ : REFRESH_FRAME_CONTEXT_BACKWARD;
#if !CONFIG_MISC_FIXES
} else {
vpx_rb_read_bit(rb); // parallel decoding mode flag
@@ -2097,20 +2036,18 @@
{
int i;
for (i = 0; i < MAX_SEGMENTS; ++i) {
- const int qindex = CONFIG_MISC_FIXES && cm->seg.enabled ?
- vp10_get_qindex(&cm->seg, i, cm->base_qindex) :
- cm->base_qindex;
- xd->lossless[i] = qindex == 0 &&
- cm->y_dc_delta_q == 0 &&
- cm->uv_dc_delta_q == 0 &&
- cm->uv_ac_delta_q == 0;
+ const int qindex = CONFIG_MISC_FIXES && cm->seg.enabled
+ ? vp10_get_qindex(&cm->seg, i, cm->base_qindex)
+ : cm->base_qindex;
+ xd->lossless[i] = qindex == 0 && cm->y_dc_delta_q == 0 &&
+ cm->uv_dc_delta_q == 0 && cm->uv_ac_delta_q == 0;
}
}
setup_segmentation_dequant(cm);
#if CONFIG_MISC_FIXES
- cm->tx_mode = (!cm->seg.enabled && xd->lossless[0]) ? ONLY_4X4
- : read_tx_mode(rb);
+ cm->tx_mode =
+ (!cm->seg.enabled && xd->lossless[0]) ? ONLY_4X4 : read_tx_mode(rb);
cm->reference_mode = read_frame_reference_mode(cm, rb);
#endif
@@ -2159,8 +2096,7 @@
#if !CONFIG_MISC_FIXES
cm->tx_mode = xd->lossless[0] ? ONLY_4X4 : read_tx_mode(&r);
#endif
- if (cm->tx_mode == TX_MODE_SELECT)
- read_tx_mode_probs(&fc->tx_probs, &r);
+ if (cm->tx_mode == TX_MODE_SELECT) read_tx_mode_probs(&fc->tx_probs, &r);
read_coef_probs(fc, cm->tx_mode, &r);
for (k = 0; k < SKIP_CONTEXTS; ++k)
@@ -2198,8 +2134,7 @@
read_inter_mode_probs(fc, &r);
- if (cm->interp_filter == SWITCHABLE)
- read_switchable_interp_probs(fc, &r);
+ if (cm->interp_filter == SWITCHABLE) read_switchable_interp_probs(fc, &r);
for (i = 0; i < INTRA_INTER_CONTEXTS; i++)
vp10_diff_update_prob(&r, &fc->intra_inter_prob[i]);
@@ -2230,7 +2165,7 @@
#ifdef NDEBUG
#define debug_check_frame_counts(cm) (void)0
-#else // !NDEBUG
+#else // !NDEBUG
// Counts should only be incremented when frame_parallel_decoding_mode and
// error_resilient_mode are disabled.
static void debug_check_frame_counts(const VP10_COMMON *const cm) {
@@ -2244,8 +2179,7 @@
sizeof(cm->counts.uv_mode)));
assert(!memcmp(cm->counts.partition, zero_counts.partition,
sizeof(cm->counts.partition)));
- assert(!memcmp(cm->counts.coef, zero_counts.coef,
- sizeof(cm->counts.coef)));
+ assert(!memcmp(cm->counts.coef, zero_counts.coef, sizeof(cm->counts.coef)));
assert(!memcmp(cm->counts.eob_branch, zero_counts.eob_branch,
sizeof(cm->counts.eob_branch)));
assert(!memcmp(cm->counts.switchable_interp, zero_counts.switchable_interp,
@@ -2271,11 +2205,8 @@
#endif // NDEBUG
static struct vpx_read_bit_buffer *init_read_bit_buffer(
- VP10Decoder *pbi,
- struct vpx_read_bit_buffer *rb,
- const uint8_t *data,
- const uint8_t *data_end,
- uint8_t clear_data[MAX_VP10_HEADER_SIZE]) {
+ VP10Decoder *pbi, struct vpx_read_bit_buffer *rb, const uint8_t *data,
+ const uint8_t *data_end, uint8_t clear_data[MAX_VP10_HEADER_SIZE]) {
rb->bit_offset = 0;
rb->error_handler = error_handler;
rb->error_handler_data = &pbi->common;
@@ -2299,8 +2230,8 @@
vpx_rb_read_literal(rb, 8) == VP10_SYNC_CODE_2;
}
-void vp10_read_frame_size(struct vpx_read_bit_buffer *rb,
- int *width, int *height) {
+void vp10_read_frame_size(struct vpx_read_bit_buffer *rb, int *width,
+ int *height) {
*width = vpx_rb_read_literal(rb, 16) + 1;
*height = vpx_rb_read_literal(rb, 16) + 1;
}
@@ -2308,21 +2239,19 @@
BITSTREAM_PROFILE vp10_read_profile(struct vpx_read_bit_buffer *rb) {
int profile = vpx_rb_read_bit(rb);
profile |= vpx_rb_read_bit(rb) << 1;
- if (profile > 2)
- profile += vpx_rb_read_bit(rb);
- return (BITSTREAM_PROFILE) profile;
+ if (profile > 2) profile += vpx_rb_read_bit(rb);
+ return (BITSTREAM_PROFILE)profile;
}
-void vp10_decode_frame(VP10Decoder *pbi,
- const uint8_t *data, const uint8_t *data_end,
- const uint8_t **p_data_end) {
+void vp10_decode_frame(VP10Decoder *pbi, const uint8_t *data,
+ const uint8_t *data_end, const uint8_t **p_data_end) {
VP10_COMMON *const cm = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
struct vpx_read_bit_buffer rb;
int context_updated = 0;
uint8_t clear_data[MAX_VP10_HEADER_SIZE];
- const size_t first_partition_size = read_uncompressed_header(pbi,
- init_read_bit_buffer(pbi, &rb, data, data_end, clear_data));
+ const size_t first_partition_size = read_uncompressed_header(
+ pbi, init_read_bit_buffer(pbi, &rb, data, data_end, clear_data));
const int tile_rows = 1 << cm->log2_tile_rows;
const int tile_cols = 1 << cm->log2_tile_cols;
YV12_BUFFER_CONFIG *const new_fb = get_frame_new_buffer(cm);
@@ -2339,12 +2268,10 @@
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
"Truncated packet or corrupt header length");
- cm->use_prev_frame_mvs = !cm->error_resilient_mode &&
- cm->width == cm->last_width &&
- cm->height == cm->last_height &&
- !cm->last_intra_only &&
- cm->last_show_frame &&
- (cm->last_frame_type != KEY_FRAME);
+ cm->use_prev_frame_mvs =
+ !cm->error_resilient_mode && cm->width == cm->last_width &&
+ cm->height == cm->last_height && !cm->last_intra_only &&
+ cm->last_show_frame && (cm->last_frame_type != KEY_FRAME);
vp10_setup_block_planes(xd, cm->subsampling_x, cm->subsampling_y);
@@ -2392,13 +2319,12 @@
// If multiple threads are used to decode tiles, then we use those
// threads to do parallel loopfiltering.
vp10_loop_filter_frame_mt(new_fb, cm, pbi->mb.plane,
- cm->lf.filter_level, 0, 0, pbi->tile_workers,
- pbi->num_tile_workers, &pbi->lf_row_sync);
+ cm->lf.filter_level, 0, 0, pbi->tile_workers,
+ pbi->num_tile_workers, &pbi->lf_row_sync);
}
} else {
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
"Decode failed. Frame data is corrupted.");
-
}
} else {
*p_data_end = decode_tiles(pbi, data + first_partition_size, data_end);
diff --git a/vp10/decoder/decodeframe.h b/vp10/decoder/decodeframe.h
index 770ae15..7fdff0b 100644
--- a/vp10/decoder/decodeframe.h
+++ b/vp10/decoder/decodeframe.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_DECODER_DECODEFRAME_H_
#define VP10_DECODER_DECODEFRAME_H_
@@ -20,13 +19,12 @@
struct vpx_read_bit_buffer;
int vp10_read_sync_code(struct vpx_read_bit_buffer *const rb);
-void vp10_read_frame_size(struct vpx_read_bit_buffer *rb,
- int *width, int *height);
+void vp10_read_frame_size(struct vpx_read_bit_buffer *rb, int *width,
+ int *height);
BITSTREAM_PROFILE vp10_read_profile(struct vpx_read_bit_buffer *rb);
-void vp10_decode_frame(struct VP10Decoder *pbi,
- const uint8_t *data, const uint8_t *data_end,
- const uint8_t **p_data_end);
+void vp10_decode_frame(struct VP10Decoder *pbi, const uint8_t *data,
+ const uint8_t *data_end, const uint8_t **p_data_end);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/decoder/decodemv.c b/vp10/decoder/decodemv.c
index a28ae55..4a34e52 100644
--- a/vp10/decoder/decodemv.c
+++ b/vp10/decoder/decodemv.c
@@ -27,7 +27,7 @@
static INLINE int read_uniform(vpx_reader *r, int n) {
int l = get_unsigned_bits(n);
int m = (1 << l) - n;
- int v = vpx_read_literal(r, l-1);
+ int v = vpx_read_literal(r, l - 1);
assert(l != 0);
@@ -46,35 +46,32 @@
const PREDICTION_MODE y_mode =
read_intra_mode(r, cm->fc->y_mode_prob[size_group]);
FRAME_COUNTS *counts = xd->counts;
- if (counts)
- ++counts->y_mode[size_group][y_mode];
+ if (counts) ++counts->y_mode[size_group][y_mode];
return y_mode;
}
static PREDICTION_MODE read_intra_mode_uv(VP10_COMMON *cm, MACROBLOCKD *xd,
vpx_reader *r,
PREDICTION_MODE y_mode) {
- const PREDICTION_MODE uv_mode = read_intra_mode(r,
- cm->fc->uv_mode_prob[y_mode]);
+ const PREDICTION_MODE uv_mode =
+ read_intra_mode(r, cm->fc->uv_mode_prob[y_mode]);
FRAME_COUNTS *counts = xd->counts;
- if (counts)
- ++counts->uv_mode[y_mode][uv_mode];
+ if (counts) ++counts->uv_mode[y_mode][uv_mode];
return uv_mode;
}
static PREDICTION_MODE read_inter_mode(VP10_COMMON *cm, MACROBLOCKD *xd,
vpx_reader *r, int ctx) {
- const int mode = vpx_read_tree(r, vp10_inter_mode_tree,
- cm->fc->inter_mode_probs[ctx]);
+ const int mode =
+ vpx_read_tree(r, vp10_inter_mode_tree, cm->fc->inter_mode_probs[ctx]);
FRAME_COUNTS *counts = xd->counts;
- if (counts)
- ++counts->inter_mode[ctx][mode];
+ if (counts) ++counts->inter_mode[ctx][mode];
return NEARESTMV + mode;
}
static int read_segment_id(vpx_reader *r,
- const struct segmentation_probs *segp) {
+ const struct segmentation_probs *segp) {
return vpx_read_tree(r, vp10_segment_tree, segp->tree_probs);
}
@@ -90,18 +87,16 @@
tx_size += vpx_read(r, tx_probs[2]);
}
- if (counts)
- ++get_tx_counts(max_tx_size, ctx, &counts->tx)[tx_size];
+ if (counts) ++get_tx_counts(max_tx_size, ctx, &counts->tx)[tx_size];
return (TX_SIZE)tx_size;
}
-static TX_SIZE read_tx_size(VP10_COMMON *cm, MACROBLOCKD *xd,
- int allow_select, vpx_reader *r) {
+static TX_SIZE read_tx_size(VP10_COMMON *cm, MACROBLOCKD *xd, int allow_select,
+ vpx_reader *r) {
TX_MODE tx_mode = cm->tx_mode;
BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
const TX_SIZE max_tx_size = max_txsize_lookup[bsize];
- if (xd->lossless[xd->mi[0]->mbmi.segment_id])
- return TX_4X4;
+ if (xd->lossless[xd->mi[0]->mbmi.segment_id]) return TX_4X4;
if (allow_select && tx_mode == TX_MODE_SELECT && bsize >= BLOCK_8X8)
return read_selected_tx_size(cm, xd, max_tx_size, r);
else
@@ -121,8 +116,8 @@
return segment_id;
}
-static void set_segment_id(VP10_COMMON *cm, int mi_offset,
- int x_mis, int y_mis, int segment_id) {
+static void set_segment_id(VP10_COMMON *cm, int mi_offset, int x_mis, int y_mis,
+ int segment_id) {
int x, y;
assert(segment_id >= 0 && segment_id < MAX_SEGMENTS);
@@ -145,33 +140,32 @@
int segment_id;
#if !CONFIG_MISC_FIXES
- (void) xd;
+ (void)xd;
#endif
- if (!seg->enabled)
- return 0; // Default for disabled segmentation
+ if (!seg->enabled) return 0; // Default for disabled segmentation
assert(seg->update_map && !seg->temporal_update);
segment_id = read_segment_id(r, segp);
#if CONFIG_MISC_FIXES
- if (counts)
- ++counts->seg.tree_total[segment_id];
+ if (counts) ++counts->seg.tree_total[segment_id];
#endif
set_segment_id(cm, mi_offset, x_mis, y_mis, segment_id);
return segment_id;
}
static void copy_segment_id(const VP10_COMMON *cm,
- const uint8_t *last_segment_ids,
- uint8_t *current_segment_ids,
- int mi_offset, int x_mis, int y_mis) {
+ const uint8_t *last_segment_ids,
+ uint8_t *current_segment_ids, int mi_offset,
+ int x_mis, int y_mis) {
int x, y;
for (y = 0; y < y_mis; y++)
for (x = 0; x < x_mis; x++)
- current_segment_ids[mi_offset + y * cm->mi_cols + x] = last_segment_ids ?
- last_segment_ids[mi_offset + y * cm->mi_cols + x] : 0;
+ current_segment_ids[mi_offset + y * cm->mi_cols + x] =
+ last_segment_ids ? last_segment_ids[mi_offset + y * cm->mi_cols + x]
+ : 0;
}
static int read_inter_segment_id(VP10_COMMON *const cm, MACROBLOCKD *const xd,
@@ -193,12 +187,12 @@
const int x_mis = VPXMIN(cm->mi_cols - mi_col, bw);
const int y_mis = VPXMIN(cm->mi_rows - mi_row, bh);
- if (!seg->enabled)
- return 0; // Default for disabled segmentation
+ if (!seg->enabled) return 0; // Default for disabled segmentation
- predicted_segment_id = cm->last_frame_seg_map ?
- dec_get_segment_id(cm, cm->last_frame_seg_map, mi_offset, x_mis, y_mis) :
- 0;
+ predicted_segment_id = cm->last_frame_seg_map
+ ? dec_get_segment_id(cm, cm->last_frame_seg_map,
+ mi_offset, x_mis, y_mis)
+ : 0;
if (!seg->update_map) {
copy_segment_id(cm, cm->last_frame_seg_map, cm->current_frame_seg_map,
@@ -211,50 +205,46 @@
const vpx_prob pred_prob = segp->pred_probs[ctx];
mbmi->seg_id_predicted = vpx_read(r, pred_prob);
#if CONFIG_MISC_FIXES
- if (counts)
- ++counts->seg.pred[ctx][mbmi->seg_id_predicted];
+ if (counts) ++counts->seg.pred[ctx][mbmi->seg_id_predicted];
#endif
if (mbmi->seg_id_predicted) {
segment_id = predicted_segment_id;
} else {
segment_id = read_segment_id(r, segp);
#if CONFIG_MISC_FIXES
- if (counts)
- ++counts->seg.tree_mispred[segment_id];
+ if (counts) ++counts->seg.tree_mispred[segment_id];
#endif
}
} else {
segment_id = read_segment_id(r, segp);
#if CONFIG_MISC_FIXES
- if (counts)
- ++counts->seg.tree_total[segment_id];
+ if (counts) ++counts->seg.tree_total[segment_id];
#endif
}
set_segment_id(cm, mi_offset, x_mis, y_mis, segment_id);
return segment_id;
}
-static int read_skip(VP10_COMMON *cm, const MACROBLOCKD *xd,
- int segment_id, vpx_reader *r) {
+static int read_skip(VP10_COMMON *cm, const MACROBLOCKD *xd, int segment_id,
+ vpx_reader *r) {
if (segfeature_active(&cm->seg, segment_id, SEG_LVL_SKIP)) {
return 1;
} else {
const int ctx = vp10_get_skip_context(xd);
const int skip = vpx_read(r, cm->fc->skip_probs[ctx]);
FRAME_COUNTS *counts = xd->counts;
- if (counts)
- ++counts->skip[ctx][skip];
+ if (counts) ++counts->skip[ctx][skip];
return skip;
}
}
static void read_intra_frame_mode_info(VP10_COMMON *const cm,
- MACROBLOCKD *const xd,
- int mi_row, int mi_col, vpx_reader *r) {
+ MACROBLOCKD *const xd, int mi_row,
+ int mi_col, vpx_reader *r) {
MODE_INFO *const mi = xd->mi[0];
MB_MODE_INFO *const mbmi = &mi->mbmi;
const MODE_INFO *above_mi = xd->above_mi;
- const MODE_INFO *left_mi = xd->left_mi;
+ const MODE_INFO *left_mi = xd->left_mi;
const BLOCK_SIZE bsize = mbmi->sb_type;
int i;
const int mi_offset = mi_row * cm->mi_cols + mi_col;
@@ -291,20 +281,19 @@
read_intra_mode(r, get_y_mode_probs(cm, mi, above_mi, left_mi, 2));
break;
default:
- mbmi->mode = read_intra_mode(r,
- get_y_mode_probs(cm, mi, above_mi, left_mi, 0));
+ mbmi->mode =
+ read_intra_mode(r, get_y_mode_probs(cm, mi, above_mi, left_mi, 0));
}
mbmi->uv_mode = read_intra_mode_uv(cm, xd, r, mbmi->mode);
- if (mbmi->tx_size < TX_32X32 &&
- cm->base_qindex > 0 && !mbmi->skip &&
+ if (mbmi->tx_size < TX_32X32 && cm->base_qindex > 0 && !mbmi->skip &&
!segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
FRAME_COUNTS *counts = xd->counts;
TX_TYPE tx_type_nom = intra_mode_to_tx_type_context[mbmi->mode];
- mbmi->tx_type = vpx_read_tree(
- r, vp10_ext_tx_tree,
- cm->fc->intra_ext_tx_prob[mbmi->tx_size][tx_type_nom]);
+ mbmi->tx_type =
+ vpx_read_tree(r, vp10_ext_tx_tree,
+ cm->fc->intra_ext_tx_prob[mbmi->tx_size][tx_type_nom]);
if (counts)
++counts->intra_ext_tx[mbmi->tx_size][tx_type_nom][mbmi->tx_type];
} else {
@@ -312,8 +301,8 @@
}
}
-static int read_mv_component(vpx_reader *r,
- const nmv_component *mvcomp, int usehp) {
+static int read_mv_component(vpx_reader *r, const nmv_component *mvcomp,
+ int usehp) {
int mag, d, fr, hp;
const int sign = vpx_read(r, mvcomp->sign);
const int mv_class = vpx_read_tree(r, vp10_mv_class_tree, mvcomp->classes);
@@ -328,18 +317,16 @@
const int n = mv_class + CLASS0_BITS - 1; // number of bits
d = 0;
- for (i = 0; i < n; ++i)
- d |= vpx_read(r, mvcomp->bits[i]) << i;
+ for (i = 0; i < n; ++i) d |= vpx_read(r, mvcomp->bits[i]) << i;
mag = CLASS0_SIZE << (mv_class + 2);
}
// Fractional part
- fr = vpx_read_tree(r, vp10_mv_fp_tree, class0 ? mvcomp->class0_fp[d]
- : mvcomp->fp);
+ fr = vpx_read_tree(r, vp10_mv_fp_tree,
+ class0 ? mvcomp->class0_fp[d] : mvcomp->fp);
// High precision part (if hp is not used, the default value of the hp is 1)
- hp = usehp ? vpx_read(r, class0 ? mvcomp->class0_hp : mvcomp->hp)
- : 1;
+ hp = usehp ? vpx_read(r, class0 ? mvcomp->class0_hp : mvcomp->hp) : 1;
// Result
mag += ((d << 3) | (fr << 1) | hp) + 1;
@@ -347,12 +334,12 @@
}
static INLINE void read_mv(vpx_reader *r, MV *mv, const MV *ref,
- const nmv_context *ctx,
- nmv_context_counts *counts, int allow_hp) {
+ const nmv_context *ctx, nmv_context_counts *counts,
+ int allow_hp) {
const MV_JOINT_TYPE joint_type =
(MV_JOINT_TYPE)vpx_read_tree(r, vp10_mv_joint_tree, ctx->joints);
const int use_hp = allow_hp && vp10_use_mv_hp(ref);
- MV diff = {0, 0};
+ MV diff = { 0, 0 };
if (mv_joint_vertical(joint_type))
diff.row = read_mv_component(r, &ctx->comps[0], use_hp);
@@ -374,8 +361,7 @@
const REFERENCE_MODE mode =
(REFERENCE_MODE)vpx_read(r, cm->fc->comp_inter_prob[ctx]);
FRAME_COUNTS *counts = xd->counts;
- if (counts)
- ++counts->comp_inter[ctx][mode];
+ if (counts) ++counts->comp_inter[ctx][mode];
return mode; // SINGLE_REFERENCE or COMPOUND_REFERENCE
} else {
return cm->reference_mode;
@@ -384,8 +370,8 @@
// Read the referncence frame
static void read_ref_frames(VP10_COMMON *const cm, MACROBLOCKD *const xd,
- vpx_reader *r,
- int segment_id, MV_REFERENCE_FRAME ref_frame[2]) {
+ vpx_reader *r, int segment_id,
+ MV_REFERENCE_FRAME ref_frame[2]) {
FRAME_CONTEXT *const fc = cm->fc;
FRAME_COUNTS *counts = xd->counts;
@@ -400,20 +386,17 @@
const int idx = cm->ref_frame_sign_bias[cm->comp_fixed_ref];
const int ctx = vp10_get_pred_context_comp_ref_p(cm, xd);
const int bit = vpx_read(r, fc->comp_ref_prob[ctx]);
- if (counts)
- ++counts->comp_ref[ctx][bit];
+ if (counts) ++counts->comp_ref[ctx][bit];
ref_frame[idx] = cm->comp_fixed_ref;
ref_frame[!idx] = cm->comp_var_ref[bit];
} else if (mode == SINGLE_REFERENCE) {
const int ctx0 = vp10_get_pred_context_single_ref_p1(xd);
const int bit0 = vpx_read(r, fc->single_ref_prob[ctx0][0]);
- if (counts)
- ++counts->single_ref[ctx0][0][bit0];
+ if (counts) ++counts->single_ref[ctx0][0][bit0];
if (bit0) {
const int ctx1 = vp10_get_pred_context_single_ref_p2(xd);
const int bit1 = vpx_read(r, fc->single_ref_prob[ctx1][1]);
- if (counts)
- ++counts->single_ref[ctx1][1][bit1];
+ if (counts) ++counts->single_ref[ctx1][1][bit1];
ref_frame[0] = bit1 ? ALTREF_FRAME : GOLDEN_FRAME;
} else {
ref_frame[0] = LAST_FRAME;
@@ -426,17 +409,14 @@
}
}
-
-static INLINE INTERP_FILTER read_switchable_interp_filter(
- VP10_COMMON *const cm, MACROBLOCKD *const xd,
- vpx_reader *r) {
+static INLINE INTERP_FILTER read_switchable_interp_filter(VP10_COMMON *const cm,
+ MACROBLOCKD *const xd,
+ vpx_reader *r) {
const int ctx = vp10_get_pred_context_switchable_interp(xd);
- const INTERP_FILTER type =
- (INTERP_FILTER)vpx_read_tree(r, vp10_switchable_interp_tree,
- cm->fc->switchable_interp_prob[ctx]);
+ const INTERP_FILTER type = (INTERP_FILTER)vpx_read_tree(
+ r, vp10_switchable_interp_tree, cm->fc->switchable_interp_prob[ctx]);
FRAME_COUNTS *counts = xd->counts;
- if (counts)
- ++counts->switchable_interp[ctx][type];
+ if (counts) ++counts->switchable_interp[ctx][type];
return type;
}
@@ -457,14 +437,12 @@
mbmi->mode = mi->bmi[3].as_mode;
break;
case BLOCK_4X8:
- mi->bmi[0].as_mode = mi->bmi[2].as_mode = read_intra_mode_y(cm, xd,
- r, 0);
+ mi->bmi[0].as_mode = mi->bmi[2].as_mode = read_intra_mode_y(cm, xd, r, 0);
mi->bmi[1].as_mode = mi->bmi[3].as_mode = mbmi->mode =
read_intra_mode_y(cm, xd, r, 0);
break;
case BLOCK_8X4:
- mi->bmi[0].as_mode = mi->bmi[1].as_mode = read_intra_mode_y(cm, xd,
- r, 0);
+ mi->bmi[0].as_mode = mi->bmi[1].as_mode = read_intra_mode_y(cm, xd, r, 0);
mi->bmi[2].as_mode = mi->bmi[3].as_mode = mbmi->mode =
read_intra_mode_y(cm, xd, r, 0);
break;
@@ -476,15 +454,15 @@
}
static INLINE int is_mv_valid(const MV *mv) {
- return mv->row > MV_LOW && mv->row < MV_UPP &&
- mv->col > MV_LOW && mv->col < MV_UPP;
+ return mv->row > MV_LOW && mv->row < MV_UPP && mv->col > MV_LOW &&
+ mv->col < MV_UPP;
}
static INLINE int assign_mv(VP10_COMMON *cm, MACROBLOCKD *xd,
- PREDICTION_MODE mode,
- int_mv mv[2], int_mv ref_mv[2],
- int_mv nearest_mv[2], int_mv near_mv[2],
- int is_compound, int allow_hp, vpx_reader *r) {
+ PREDICTION_MODE mode, int_mv mv[2],
+ int_mv ref_mv[2], int_mv nearest_mv[2],
+ int_mv near_mv[2], int is_compound, int allow_hp,
+ vpx_reader *r) {
int i;
int ret = 1;
@@ -501,25 +479,20 @@
}
case NEARESTMV: {
mv[0].as_int = nearest_mv[0].as_int;
- if (is_compound)
- mv[1].as_int = nearest_mv[1].as_int;
+ if (is_compound) mv[1].as_int = nearest_mv[1].as_int;
break;
}
case NEARMV: {
mv[0].as_int = near_mv[0].as_int;
- if (is_compound)
- mv[1].as_int = near_mv[1].as_int;
+ if (is_compound) mv[1].as_int = near_mv[1].as_int;
break;
}
case ZEROMV: {
mv[0].as_int = 0;
- if (is_compound)
- mv[1].as_int = 0;
+ if (is_compound) mv[1].as_int = 0;
break;
}
- default: {
- return 0;
- }
+ default: { return 0; }
}
return ret;
}
@@ -532,8 +505,7 @@
const int ctx = vp10_get_intra_inter_context(xd);
const int is_inter = vpx_read(r, cm->fc->intra_inter_prob[ctx]);
FRAME_COUNTS *counts = xd->counts;
- if (counts)
- ++counts->intra_inter[ctx][is_inter];
+ if (counts) ++counts->intra_inter[ctx][is_inter];
return is_inter;
}
}
@@ -541,13 +513,13 @@
static void fpm_sync(void *const data, int mi_row) {
VP10Decoder *const pbi = (VP10Decoder *)data;
vp10_frameworker_wait(pbi->frame_worker_owner, pbi->common.prev_frame,
- mi_row << MI_BLOCK_SIZE_LOG2);
+ mi_row << MI_BLOCK_SIZE_LOG2);
}
static void read_inter_block_mode_info(VP10Decoder *const pbi,
MACROBLOCKD *const xd,
- MODE_INFO *const mi,
- int mi_row, int mi_col, vpx_reader *r) {
+ MODE_INFO *const mi, int mi_row,
+ int mi_col, vpx_reader *r) {
VP10_COMMON *const cm = &pbi->common;
MB_MODE_INFO *const mbmi = &mi->mbmi;
const BLOCK_SIZE bsize = mbmi->sb_type;
@@ -568,23 +540,22 @@
if ((!vp10_is_valid_scale(&ref_buf->sf)))
vpx_internal_error(xd->error_info, VPX_CODEC_UNSUP_BITSTREAM,
"Reference frame has invalid dimensions");
- vp10_setup_pre_planes(xd, ref, ref_buf->buf, mi_row, mi_col,
- &ref_buf->sf);
- vp10_find_mv_refs(cm, xd, mi, frame, ref_mvs[frame],
- mi_row, mi_col, fpm_sync, (void *)pbi, inter_mode_ctx);
+ vp10_setup_pre_planes(xd, ref, ref_buf->buf, mi_row, mi_col, &ref_buf->sf);
+ vp10_find_mv_refs(cm, xd, mi, frame, ref_mvs[frame], mi_row, mi_col,
+ fpm_sync, (void *)pbi, inter_mode_ctx);
}
if (segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
mbmi->mode = ZEROMV;
if (bsize < BLOCK_8X8) {
- vpx_internal_error(xd->error_info, VPX_CODEC_UNSUP_BITSTREAM,
- "Invalid usage of segement feature on small blocks");
- return;
+ vpx_internal_error(xd->error_info, VPX_CODEC_UNSUP_BITSTREAM,
+ "Invalid usage of segement feature on small blocks");
+ return;
}
} else {
if (bsize >= BLOCK_8X8)
- mbmi->mode = read_inter_mode(cm, xd, r,
- inter_mode_ctx[mbmi->ref_frame[0]]);
+ mbmi->mode =
+ read_inter_mode(cm, xd, r, inter_mode_ctx[mbmi->ref_frame[0]]);
}
if (bsize < BLOCK_8X8 || mbmi->mode != ZEROMV) {
@@ -595,8 +566,8 @@
}
mbmi->interp_filter = (cm->interp_filter == SWITCHABLE)
- ? read_switchable_interp_filter(cm, xd, r)
- : cm->interp_filter;
+ ? read_switchable_interp_filter(cm, xd, r)
+ : cm->interp_filter;
if (bsize < BLOCK_8X8) {
const int num_4x4_w = 1 << xd->bmode_blocks_wl;
@@ -614,26 +585,21 @@
uint8_t dummy_mode_ctx[MAX_REF_FRAMES];
for (ref = 0; ref < 1 + is_compound; ++ref)
vp10_append_sub8x8_mvs_for_idx(cm, xd, j, ref, mi_row, mi_col,
- &nearest_sub8x8[ref],
- &near_sub8x8[ref],
- dummy_mode_ctx);
+ &nearest_sub8x8[ref],
+ &near_sub8x8[ref], dummy_mode_ctx);
}
- if (!assign_mv(cm, xd, b_mode, block, nearestmv,
- nearest_sub8x8, near_sub8x8,
- is_compound, allow_hp, r)) {
+ if (!assign_mv(cm, xd, b_mode, block, nearestmv, nearest_sub8x8,
+ near_sub8x8, is_compound, allow_hp, r)) {
xd->corrupted |= 1;
break;
};
mi->bmi[j].as_mv[0].as_int = block[0].as_int;
- if (is_compound)
- mi->bmi[j].as_mv[1].as_int = block[1].as_int;
+ if (is_compound) mi->bmi[j].as_mv[1].as_int = block[1].as_int;
- if (num_4x4_h == 2)
- mi->bmi[j + 2] = mi->bmi[j];
- if (num_4x4_w == 2)
- mi->bmi[j + 1] = mi->bmi[j];
+ if (num_4x4_h == 2) mi->bmi[j + 2] = mi->bmi[j];
+ if (num_4x4_w == 2) mi->bmi[j + 1] = mi->bmi[j];
}
}
@@ -648,8 +614,8 @@
}
static void read_inter_frame_mode_info(VP10Decoder *const pbi,
- MACROBLOCKD *const xd,
- int mi_row, int mi_col, vpx_reader *r) {
+ MACROBLOCKD *const xd, int mi_row,
+ int mi_col, vpx_reader *r) {
VP10_COMMON *const cm = &pbi->common;
MODE_INFO *const mi = xd->mi[0];
MB_MODE_INFO *const mbmi = &mi->mbmi;
@@ -667,21 +633,18 @@
else
read_intra_block_mode_info(cm, xd, mi, r);
- if (mbmi->tx_size < TX_32X32 &&
- cm->base_qindex > 0 && !mbmi->skip &&
+ if (mbmi->tx_size < TX_32X32 && cm->base_qindex > 0 && !mbmi->skip &&
!segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
FRAME_COUNTS *counts = xd->counts;
if (inter_block) {
- mbmi->tx_type = vpx_read_tree(
- r, vp10_ext_tx_tree,
- cm->fc->inter_ext_tx_prob[mbmi->tx_size]);
- if (counts)
- ++counts->inter_ext_tx[mbmi->tx_size][mbmi->tx_type];
+ mbmi->tx_type = vpx_read_tree(r, vp10_ext_tx_tree,
+ cm->fc->inter_ext_tx_prob[mbmi->tx_size]);
+ if (counts) ++counts->inter_ext_tx[mbmi->tx_size][mbmi->tx_type];
} else {
const TX_TYPE tx_type_nom = intra_mode_to_tx_type_context[mbmi->mode];
- mbmi->tx_type = vpx_read_tree(
- r, vp10_ext_tx_tree,
- cm->fc->intra_ext_tx_prob[mbmi->tx_size][tx_type_nom]);
+ mbmi->tx_type =
+ vpx_read_tree(r, vp10_ext_tx_tree,
+ cm->fc->intra_ext_tx_prob[mbmi->tx_size][tx_type_nom]);
if (counts)
++counts->intra_ext_tx[mbmi->tx_size][tx_type_nom][mbmi->tx_type];
}
@@ -690,12 +653,11 @@
}
}
-void vp10_read_mode_info(VP10Decoder *const pbi, MACROBLOCKD *xd,
- int mi_row, int mi_col, vpx_reader *r,
- int x_mis, int y_mis) {
+void vp10_read_mode_info(VP10Decoder *const pbi, MACROBLOCKD *xd, int mi_row,
+ int mi_col, vpx_reader *r, int x_mis, int y_mis) {
VP10_COMMON *const cm = &pbi->common;
MODE_INFO *const mi = xd->mi[0];
- MV_REF* frame_mvs = cm->cur_frame->mvs + mi_row * cm->mi_cols + mi_col;
+ MV_REF *frame_mvs = cm->cur_frame->mvs + mi_row * cm->mi_cols + mi_col;
int w, h;
if (frame_is_intra_only(cm)) {
diff --git a/vp10/decoder/decodemv.h b/vp10/decoder/decodemv.h
index 6653be5..f7bc738 100644
--- a/vp10/decoder/decodemv.h
+++ b/vp10/decoder/decodemv.h
@@ -19,9 +19,8 @@
extern "C" {
#endif
-void vp10_read_mode_info(VP10Decoder *const pbi, MACROBLOCKD *xd,
- int mi_row, int mi_col, vpx_reader *r,
- int x_mis, int y_mis);
+void vp10_read_mode_info(VP10Decoder *const pbi, MACROBLOCKD *xd, int mi_row,
+ int mi_col, vpx_reader *r, int x_mis, int y_mis);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/decoder/decoder.c b/vp10/decoder/decoder.c
index 8144ef5..f979abf 100644
--- a/vp10/decoder/decoder.c
+++ b/vp10/decoder/decoder.c
@@ -54,12 +54,10 @@
static int vp10_dec_alloc_mi(VP10_COMMON *cm, int mi_size) {
cm->mip = vpx_calloc(mi_size, sizeof(*cm->mip));
- if (!cm->mip)
- return 1;
+ if (!cm->mip) return 1;
cm->mi_alloc_size = mi_size;
- cm->mi_grid_base = (MODE_INFO **)vpx_calloc(mi_size, sizeof(MODE_INFO*));
- if (!cm->mi_grid_base)
- return 1;
+ cm->mi_grid_base = (MODE_INFO **)vpx_calloc(mi_size, sizeof(MODE_INFO *));
+ if (!cm->mi_grid_base) return 1;
return 0;
}
@@ -74,8 +72,7 @@
VP10Decoder *volatile const pbi = vpx_memalign(32, sizeof(*pbi));
VP10_COMMON *volatile const cm = pbi ? &pbi->common : NULL;
- if (!cm)
- return NULL;
+ if (!cm) return NULL;
vp10_zero(*pbi);
@@ -87,11 +84,10 @@
cm->error.setjmp = 1;
- CHECK_MEM_ERROR(cm, cm->fc,
- (FRAME_CONTEXT *)vpx_calloc(1, sizeof(*cm->fc)));
- CHECK_MEM_ERROR(cm, cm->frame_contexts,
- (FRAME_CONTEXT *)vpx_calloc(FRAME_CONTEXTS,
- sizeof(*cm->frame_contexts)));
+ CHECK_MEM_ERROR(cm, cm->fc, (FRAME_CONTEXT *)vpx_calloc(1, sizeof(*cm->fc)));
+ CHECK_MEM_ERROR(
+ cm, cm->frame_contexts,
+ (FRAME_CONTEXT *)vpx_calloc(FRAME_CONTEXTS, sizeof(*cm->frame_contexts)));
pbi->need_resync = 1;
once(initialize_dec);
@@ -123,8 +119,7 @@
void vp10_decoder_remove(VP10Decoder *pbi) {
int i;
- if (!pbi)
- return;
+ if (!pbi) return;
vpx_get_worker_interface()->end(&pbi->lf_worker);
vpx_free(pbi->lf_worker.data1);
@@ -146,8 +141,8 @@
static int equal_dimensions(const YV12_BUFFER_CONFIG *a,
const YV12_BUFFER_CONFIG *b) {
- return a->y_height == b->y_height && a->y_width == b->y_width &&
- a->uv_height == b->uv_height && a->uv_width == b->uv_width;
+ return a->y_height == b->y_height && a->y_width == b->y_width &&
+ a->uv_height == b->uv_height && a->uv_width == b->uv_width;
}
vpx_codec_err_t vp10_copy_reference_dec(VP10Decoder *pbi,
@@ -173,17 +168,15 @@
else
vpx_yv12_copy_frame(cfg, sd);
} else {
- vpx_internal_error(&cm->error, VPX_CODEC_ERROR,
- "Invalid reference frame");
+ vpx_internal_error(&cm->error, VPX_CODEC_ERROR, "Invalid reference frame");
}
return cm->error.error_code;
}
-
vpx_codec_err_t vp10_set_reference_dec(VP10_COMMON *cm,
- VPX_REFFRAME ref_frame_flag,
- YV12_BUFFER_CONFIG *sd) {
+ VPX_REFFRAME ref_frame_flag,
+ YV12_BUFFER_CONFIG *sd) {
RefBuffer *ref_buf = NULL;
RefCntBuffer *const frame_bufs = cm->buffer_pool->frame_bufs;
@@ -198,8 +191,7 @@
} else if (ref_frame_flag == VPX_ALT_FLAG) {
ref_buf = &cm->frame_refs[2];
} else {
- vpx_internal_error(&cm->error, VPX_CODEC_ERROR,
- "Invalid reference frame");
+ vpx_internal_error(&cm->error, VPX_CODEC_ERROR, "Invalid reference frame");
return cm->error.error_code;
}
@@ -211,8 +203,7 @@
// Find an empty frame buffer.
const int free_fb = get_free_fb(cm);
- if (cm->new_fb_idx == INVALID_IDX)
- return VPX_CODEC_MEM_ERROR;
+ if (cm->new_fb_idx == INVALID_IDX) return VPX_CODEC_MEM_ERROR;
// Decrease ref_count since it will be increased again in
// ref_cnt_fb() below.
@@ -269,8 +260,8 @@
cm->frame_refs[ref_index].idx = -1;
}
-int vp10_receive_compressed_data(VP10Decoder *pbi,
- size_t size, const uint8_t **psource) {
+int vp10_receive_compressed_data(VP10Decoder *pbi, size_t size,
+ const uint8_t **psource) {
VP10_COMMON *volatile const cm = &pbi->common;
BufferPool *volatile const pool = cm->buffer_pool;
RefCntBuffer *volatile const frame_bufs = cm->buffer_pool->frame_bufs;
@@ -297,14 +288,13 @@
// Check if the previous frame was a frame without any references to it.
// Release frame buffer if not decoding in frame parallel mode.
- if (!cm->frame_parallel_decode && cm->new_fb_idx >= 0
- && frame_bufs[cm->new_fb_idx].ref_count == 0)
+ if (!cm->frame_parallel_decode && cm->new_fb_idx >= 0 &&
+ frame_bufs[cm->new_fb_idx].ref_count == 0)
pool->release_fb_cb(pool->cb_priv,
&frame_bufs[cm->new_fb_idx].raw_frame_buffer);
// Find a free frame buffer. Return error if can not find any.
cm->new_fb_idx = get_free_fb(cm);
- if (cm->new_fb_idx == INVALID_IDX)
- return VPX_CODEC_MEM_ERROR;
+ if (cm->new_fb_idx == INVALID_IDX) return VPX_CODEC_MEM_ERROR;
// Assign a MV array to the frame buffer.
cm->cur_frame = &pool->frame_bufs[cm->new_fb_idx];
@@ -323,7 +313,6 @@
pbi->cur_buf = &frame_bufs[cm->new_fb_idx];
}
-
if (setjmp(cm->error.jmp)) {
const VPxWorkerInterface *const winterface = vpx_get_worker_interface();
int i;
@@ -414,14 +403,12 @@
VP10_COMMON *const cm = &pbi->common;
int ret = -1;
- if (pbi->ready_for_new_data == 1)
- return ret;
+ if (pbi->ready_for_new_data == 1) return ret;
pbi->ready_for_new_data = 1;
/* no raw frame to show!!! */
- if (!cm->show_frame)
- return ret;
+ if (!cm->show_frame) return ret;
pbi->ready_for_new_data = 1;
@@ -431,11 +418,10 @@
return ret;
}
-vpx_codec_err_t vp10_parse_superframe_index(const uint8_t *data,
- size_t data_sz,
- uint32_t sizes[8], int *count,
- vpx_decrypt_cb decrypt_cb,
- void *decrypt_state) {
+vpx_codec_err_t vp10_parse_superframe_index(const uint8_t *data, size_t data_sz,
+ uint32_t sizes[8], int *count,
+ vpx_decrypt_cb decrypt_cb,
+ void *decrypt_state) {
// A chunk ending with a byte matching 0xc0 is an invalid chunk unless
// it is a super frame index. If the last byte of real video compression
// data is 0xc0 the encoder must add a 0 byte. If we have the marker but
@@ -458,18 +444,16 @@
// This chunk is marked as having a superframe index but doesn't have
// enough data for it, thus it's an invalid superframe index.
- if (data_sz < index_sz)
- return VPX_CODEC_CORRUPT_FRAME;
+ if (data_sz < index_sz) return VPX_CODEC_CORRUPT_FRAME;
{
- const uint8_t marker2 = read_marker(decrypt_cb, decrypt_state,
- data + data_sz - index_sz);
+ const uint8_t marker2 =
+ read_marker(decrypt_cb, decrypt_state, data + data_sz - index_sz);
// This chunk is marked as having a superframe index but doesn't have
// the matching marker byte at the front of the index therefore it's an
// invalid chunk.
- if (marker != marker2)
- return VPX_CODEC_CORRUPT_FRAME;
+ if (marker != marker2) return VPX_CODEC_CORRUPT_FRAME;
}
{
@@ -488,8 +472,7 @@
for (i = 0; i < frames - CONFIG_MISC_FIXES; ++i) {
uint32_t this_sz = 0;
- for (j = 0; j < mag; ++j)
- this_sz |= (*x++) << (j * 8);
+ for (j = 0; j < mag; ++j) this_sz |= (*x++) << (j * 8);
this_sz += CONFIG_MISC_FIXES;
sizes[i] = this_sz;
#if CONFIG_MISC_FIXES
diff --git a/vp10/decoder/decoder.h b/vp10/decoder/decoder.h
index 2dec222..45d4f3d 100644
--- a/vp10/decoder/decoder.h
+++ b/vp10/decoder/decoder.h
@@ -58,9 +58,9 @@
// TODO(hkuang): Combine this with cur_buf in macroblockd as they are
// the same.
- RefCntBuffer *cur_buf; // Current decoding frame buffer.
+ RefCntBuffer *cur_buf; // Current decoding frame buffer.
- VPxWorker *frame_worker_owner; // frame_worker that owns this pbi.
+ VPxWorker *frame_worker_owner; // frame_worker that owns this pbi.
VPxWorker lf_worker;
VPxWorker *tile_workers;
TileWorkerData *tile_worker_data;
@@ -77,26 +77,25 @@
int max_threads;
int inv_tile_order;
- int need_resync; // wait for key/intra-only frame.
+ int need_resync; // wait for key/intra-only frame.
int hold_ref_buf; // hold the reference buffer.
} VP10Decoder;
-int vp10_receive_compressed_data(struct VP10Decoder *pbi,
- size_t size, const uint8_t **dest);
+int vp10_receive_compressed_data(struct VP10Decoder *pbi, size_t size,
+ const uint8_t **dest);
int vp10_get_raw_frame(struct VP10Decoder *pbi, YV12_BUFFER_CONFIG *sd);
vpx_codec_err_t vp10_copy_reference_dec(struct VP10Decoder *pbi,
+ VPX_REFFRAME ref_frame_flag,
+ YV12_BUFFER_CONFIG *sd);
+
+vpx_codec_err_t vp10_set_reference_dec(VP10_COMMON *cm,
VPX_REFFRAME ref_frame_flag,
YV12_BUFFER_CONFIG *sd);
-vpx_codec_err_t vp10_set_reference_dec(VP10_COMMON *cm,
- VPX_REFFRAME ref_frame_flag,
- YV12_BUFFER_CONFIG *sd);
-
static INLINE uint8_t read_marker(vpx_decrypt_cb decrypt_cb,
- void *decrypt_state,
- const uint8_t *data) {
+ void *decrypt_state, const uint8_t *data) {
if (decrypt_cb) {
uint8_t marker;
decrypt_cb(decrypt_state, data, &marker, 1);
@@ -107,11 +106,10 @@
// This function is exposed for use in tests, as well as the inlined function
// "read_marker".
-vpx_codec_err_t vp10_parse_superframe_index(const uint8_t *data,
- size_t data_sz,
- uint32_t sizes[8], int *count,
- vpx_decrypt_cb decrypt_cb,
- void *decrypt_state);
+vpx_codec_err_t vp10_parse_superframe_index(const uint8_t *data, size_t data_sz,
+ uint32_t sizes[8], int *count,
+ vpx_decrypt_cb decrypt_cb,
+ void *decrypt_state);
struct VP10Decoder *vp10_decoder_create(BufferPool *const pool);
diff --git a/vp10/decoder/detokenize.c b/vp10/decoder/detokenize.c
index 04bc0b8..e57d1ef 100644
--- a/vp10/decoder/detokenize.c
+++ b/vp10/decoder/detokenize.c
@@ -20,33 +20,30 @@
#include "vp10/decoder/detokenize.h"
-#define EOB_CONTEXT_NODE 0
-#define ZERO_CONTEXT_NODE 1
-#define ONE_CONTEXT_NODE 2
-#define LOW_VAL_CONTEXT_NODE 0
-#define TWO_CONTEXT_NODE 1
-#define THREE_CONTEXT_NODE 2
-#define HIGH_LOW_CONTEXT_NODE 3
-#define CAT_ONE_CONTEXT_NODE 4
-#define CAT_THREEFOUR_CONTEXT_NODE 5
-#define CAT_THREE_CONTEXT_NODE 6
-#define CAT_FIVE_CONTEXT_NODE 7
+#define EOB_CONTEXT_NODE 0
+#define ZERO_CONTEXT_NODE 1
+#define ONE_CONTEXT_NODE 2
+#define LOW_VAL_CONTEXT_NODE 0
+#define TWO_CONTEXT_NODE 1
+#define THREE_CONTEXT_NODE 2
+#define HIGH_LOW_CONTEXT_NODE 3
+#define CAT_ONE_CONTEXT_NODE 4
+#define CAT_THREEFOUR_CONTEXT_NODE 5
+#define CAT_THREE_CONTEXT_NODE 6
+#define CAT_FIVE_CONTEXT_NODE 7
-#define INCREMENT_COUNT(token) \
- do { \
- if (counts) \
- ++coef_counts[band][ctx][token]; \
+#define INCREMENT_COUNT(token) \
+ do { \
+ if (counts) ++coef_counts[band][ctx][token]; \
} while (0)
static INLINE int read_coeff(const vpx_prob *probs, int n, vpx_reader *r) {
int i, val = 0;
- for (i = 0; i < n; ++i)
- val = (val << 1) | vpx_read(r, probs[i]);
+ for (i = 0; i < n; ++i) val = (val << 1) | vpx_read(r, probs[i]);
return val;
}
-static int decode_coefs(const MACROBLOCKD *xd,
- PLANE_TYPE type,
+static int decode_coefs(const MACROBLOCKD *xd, PLANE_TYPE type,
tran_low_t *dqcoeff, TX_SIZE tx_size, const int16_t *dq,
int ctx, const int16_t *scan, const int16_t *nb,
vpx_reader *r) {
@@ -55,11 +52,11 @@
const FRAME_CONTEXT *const fc = xd->fc;
const int ref = is_inter_block(&xd->mi[0]->mbmi);
int band, c = 0;
- const vpx_prob (*coef_probs)[COEFF_CONTEXTS][UNCONSTRAINED_NODES] =
+ const vpx_prob(*coef_probs)[COEFF_CONTEXTS][UNCONSTRAINED_NODES] =
fc->coef_probs[tx_size][type][ref];
const vpx_prob *prob;
- unsigned int (*coef_counts)[COEFF_CONTEXTS][UNCONSTRAINED_NODES + 1];
- unsigned int (*eob_branch_count)[COEFF_CONTEXTS];
+ unsigned int(*coef_counts)[COEFF_CONTEXTS][UNCONSTRAINED_NODES + 1];
+ unsigned int(*eob_branch_count)[COEFF_CONTEXTS];
uint8_t token_cache[32 * 32];
const uint8_t *band_translate = get_band_translate(tx_size);
const int dq_shift = (tx_size == TX_32X32);
@@ -115,8 +112,7 @@
int val = -1;
band = *band_translate++;
prob = coef_probs[band][ctx];
- if (counts)
- ++eob_branch_count[band][ctx];
+ if (counts) ++eob_branch_count[band][ctx];
if (!vpx_read(r, prob[EOB_CONTEXT_NODE])) {
INCREMENT_COUNT(EOB_MODEL_TOKEN);
break;
@@ -127,8 +123,7 @@
dqv = dq[1];
token_cache[scan[c]] = 0;
++c;
- if (c >= max_eob)
- return c; // zero tokens at the end (no eob token)
+ if (c >= max_eob) return c; // zero tokens at the end (no eob token)
ctx = get_coef_context(nb, token_cache, c);
band = *band_translate++;
prob = coef_probs[band][ctx];
@@ -145,9 +140,7 @@
switch (token) {
case TWO_TOKEN:
case THREE_TOKEN:
- case FOUR_TOKEN:
- val = token;
- break;
+ case FOUR_TOKEN: val = token; break;
case CATEGORY1_TOKEN:
val = CAT1_MIN_VAL + read_coeff(cat1_prob, 1, r);
break;
@@ -181,9 +174,7 @@
case VPX_BITS_12:
val = CAT6_MIN_VAL + read_coeff(cat6p, 18 - skip_bits, r);
break;
- default:
- assert(0);
- return -1;
+ default: assert(0); return -1;
}
#else
val = CAT6_MIN_VAL + read_coeff(cat6p, 14 - skip_bits, r);
@@ -195,8 +186,7 @@
v = (val * dqv) >> dq_shift;
#if CONFIG_COEFFICIENT_RANGE_CHECKING
#if CONFIG_VPX_HIGHBITDEPTH
- dqcoeff[scan[c]] = highbd_check_range((vpx_read_bit(r) ? -v : v),
- xd->bd);
+ dqcoeff[scan[c]] = highbd_check_range((vpx_read_bit(r) ? -v : v), xd->bd);
#else
dqcoeff[scan[c]] = check_range(vpx_read_bit(r) ? -v : v);
#endif // CONFIG_VPX_HIGHBITDEPTH
@@ -212,12 +202,12 @@
return c;
}
-// TODO(slavarnway): Decode version of vp10_set_context. Modify vp10_set_context
+// TODO(slavarnway): Decode version of vp10_set_context. Modify
+// vp10_set_context
// after testing is complete, then delete this version.
-static
-void dec_set_contexts(const MACROBLOCKD *xd, struct macroblockd_plane *pd,
- TX_SIZE tx_size, int has_eob,
- int aoff, int loff) {
+static void dec_set_contexts(const MACROBLOCKD *xd,
+ struct macroblockd_plane *pd, TX_SIZE tx_size,
+ int has_eob, int aoff, int loff) {
ENTROPY_CONTEXT *const a = pd->above_context + aoff;
ENTROPY_CONTEXT *const l = pd->left_context + loff;
const int tx_size_in_blocks = 1 << tx_size;
@@ -225,16 +215,14 @@
// above
if (has_eob && xd->mb_to_right_edge < 0) {
int i;
- const int blocks_wide = pd->n4_w +
- (xd->mb_to_right_edge >> (5 + pd->subsampling_x));
+ const int blocks_wide =
+ pd->n4_w + (xd->mb_to_right_edge >> (5 + pd->subsampling_x));
int above_contexts = tx_size_in_blocks;
if (above_contexts + aoff > blocks_wide)
above_contexts = blocks_wide - aoff;
- for (i = 0; i < above_contexts; ++i)
- a[i] = has_eob;
- for (i = above_contexts; i < tx_size_in_blocks; ++i)
- a[i] = 0;
+ for (i = 0; i < above_contexts; ++i) a[i] = has_eob;
+ for (i = above_contexts; i < tx_size_in_blocks; ++i) a[i] = 0;
} else {
memset(a, has_eob, sizeof(ENTROPY_CONTEXT) * tx_size_in_blocks);
}
@@ -242,35 +230,27 @@
// left
if (has_eob && xd->mb_to_bottom_edge < 0) {
int i;
- const int blocks_high = pd->n4_h +
- (xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
+ const int blocks_high =
+ pd->n4_h + (xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
int left_contexts = tx_size_in_blocks;
- if (left_contexts + loff > blocks_high)
- left_contexts = blocks_high - loff;
+ if (left_contexts + loff > blocks_high) left_contexts = blocks_high - loff;
- for (i = 0; i < left_contexts; ++i)
- l[i] = has_eob;
- for (i = left_contexts; i < tx_size_in_blocks; ++i)
- l[i] = 0;
+ for (i = 0; i < left_contexts; ++i) l[i] = has_eob;
+ for (i = left_contexts; i < tx_size_in_blocks; ++i) l[i] = 0;
} else {
memset(l, has_eob, sizeof(ENTROPY_CONTEXT) * tx_size_in_blocks);
}
}
-int vp10_decode_block_tokens(MACROBLOCKD *xd,
- int plane, const scan_order *sc,
- int x, int y,
- TX_SIZE tx_size, vpx_reader *r,
- int seg_id) {
+int vp10_decode_block_tokens(MACROBLOCKD *xd, int plane, const scan_order *sc,
+ int x, int y, TX_SIZE tx_size, vpx_reader *r,
+ int seg_id) {
struct macroblockd_plane *const pd = &xd->plane[plane];
const int16_t *const dequant = pd->seg_dequant[seg_id];
- const int ctx = get_entropy_context(tx_size, pd->above_context + x,
- pd->left_context + y);
- const int eob = decode_coefs(xd, pd->plane_type,
- pd->dqcoeff, tx_size,
+ const int ctx =
+ get_entropy_context(tx_size, pd->above_context + x, pd->left_context + y);
+ const int eob = decode_coefs(xd, pd->plane_type, pd->dqcoeff, tx_size,
dequant, ctx, sc->scan, sc->neighbors, r);
dec_set_contexts(xd, pd, tx_size, eob > 0, x, y);
return eob;
}
-
-
diff --git a/vp10/decoder/detokenize.h b/vp10/decoder/detokenize.h
index c3fd90a..d7b459d 100644
--- a/vp10/decoder/detokenize.h
+++ b/vp10/decoder/detokenize.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_DECODER_DETOKENIZE_H_
#define VP10_DECODER_DETOKENIZE_H_
@@ -20,11 +19,9 @@
extern "C" {
#endif
-int vp10_decode_block_tokens(MACROBLOCKD *xd,
- int plane, const scan_order *sc,
- int x, int y,
- TX_SIZE tx_size, vpx_reader *r,
- int seg_id);
+int vp10_decode_block_tokens(MACROBLOCKD *xd, int plane, const scan_order *sc,
+ int x, int y, TX_SIZE tx_size, vpx_reader *r,
+ int seg_id);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/decoder/dsubexp.c b/vp10/decoder/dsubexp.c
index 36c1917..4e1ccc0 100644
--- a/vp10/decoder/dsubexp.c
+++ b/vp10/decoder/dsubexp.c
@@ -15,8 +15,7 @@
#include "vp10/decoder/dsubexp.h"
static int inv_recenter_nonneg(int v, int m) {
- if (v > 2 * m)
- return v;
+ if (v > 2 * m) return v;
return (v & 1) ? m - ((v + 1) >> 1) : m + (v >> 1);
}
@@ -25,10 +24,11 @@
const int l = 8;
const int m = (1 << l) - 191 + CONFIG_MISC_FIXES;
const int v = vpx_read_literal(r, l - 1);
- return v < m ? v : (v << 1) - m + vpx_read_bit(r);
+ return v < m ? v : (v << 1) - m + vpx_read_bit(r);
}
static int inv_remap_prob(int v, int m) {
+ /* clang-format off */
static uint8_t inv_map_table[MAX_PROB - CONFIG_MISC_FIXES] = {
7, 20, 33, 46, 59, 72, 85, 98, 111, 124, 137, 150, 163, 176, 189,
202, 215, 228, 241, 254, 1, 2, 3, 4, 5, 6, 8, 9, 10, 11,
@@ -50,7 +50,7 @@
#if !CONFIG_MISC_FIXES
253
#endif
- };
+ }; /* clang-format on */
assert(v < (int)(sizeof(inv_map_table) / sizeof(inv_map_table[0])));
v = inv_map_table[v];
m--;
@@ -62,16 +62,13 @@
}
static int decode_term_subexp(vpx_reader *r) {
- if (!vpx_read_bit(r))
- return vpx_read_literal(r, 4);
- if (!vpx_read_bit(r))
- return vpx_read_literal(r, 4) + 16;
- if (!vpx_read_bit(r))
- return vpx_read_literal(r, 5) + 32;
+ if (!vpx_read_bit(r)) return vpx_read_literal(r, 4);
+ if (!vpx_read_bit(r)) return vpx_read_literal(r, 4) + 16;
+ if (!vpx_read_bit(r)) return vpx_read_literal(r, 5) + 32;
return decode_uniform(r) + 64;
}
-void vp10_diff_update_prob(vpx_reader *r, vpx_prob* p) {
+void vp10_diff_update_prob(vpx_reader *r, vpx_prob *p) {
if (vpx_read(r, DIFF_UPDATE_PROB)) {
const int delp = decode_term_subexp(r);
*p = (vpx_prob)inv_remap_prob(delp, *p);
diff --git a/vp10/decoder/dsubexp.h b/vp10/decoder/dsubexp.h
index 1a7ed99..8d47df4 100644
--- a/vp10/decoder/dsubexp.h
+++ b/vp10/decoder/dsubexp.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_DECODER_DSUBEXP_H_
#define VP10_DECODER_DSUBEXP_H_
@@ -18,7 +17,7 @@
extern "C" {
#endif
-void vp10_diff_update_prob(vpx_reader *r, vpx_prob* p);
+void vp10_diff_update_prob(vpx_reader *r, vpx_prob *p);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/decoder/dthread.c b/vp10/decoder/dthread.c
index ac3cd2b..4dbcfbc 100644
--- a/vp10/decoder/dthread.c
+++ b/vp10/decoder/dthread.c
@@ -60,10 +60,9 @@
// TODO(hkuang): Remove worker parameter as it is only used in debug code.
void vp10_frameworker_wait(VPxWorker *const worker, RefCntBuffer *const ref_buf,
- int row) {
+ int row) {
#if CONFIG_MULTITHREAD
- if (!ref_buf)
- return;
+ if (!ref_buf) return;
#ifndef BUILDING_WITH_TSAN
// The following line of code will get harmless tsan error but it is the key
@@ -135,7 +134,7 @@
}
void vp10_frameworker_copy_context(VPxWorker *const dst_worker,
- VPxWorker *const src_worker) {
+ VPxWorker *const src_worker) {
#if CONFIG_MULTITHREAD
FrameWorkerData *const src_worker_data = (FrameWorkerData *)src_worker->data1;
FrameWorkerData *const dst_worker_data = (FrameWorkerData *)dst_worker->data1;
@@ -147,11 +146,12 @@
vp10_frameworker_lock_stats(src_worker);
while (!src_worker_data->frame_context_ready) {
pthread_cond_wait(&src_worker_data->stats_cond,
- &src_worker_data->stats_mutex);
+ &src_worker_data->stats_mutex);
}
- dst_cm->last_frame_seg_map = src_cm->seg.enabled ?
- src_cm->current_frame_seg_map : src_cm->last_frame_seg_map;
+ dst_cm->last_frame_seg_map = src_cm->seg.enabled
+ ? src_cm->current_frame_seg_map
+ : src_cm->last_frame_seg_map;
dst_worker_data->pbi->need_resync = src_worker_data->pbi->need_resync;
vp10_frameworker_unlock_stats(src_worker);
@@ -159,17 +159,18 @@
#if CONFIG_VPX_HIGHBITDEPTH
dst_cm->use_highbitdepth = src_cm->use_highbitdepth;
#endif
- dst_cm->prev_frame = src_cm->show_existing_frame ?
- src_cm->prev_frame : src_cm->cur_frame;
- dst_cm->last_width = !src_cm->show_existing_frame ?
- src_cm->width : src_cm->last_width;
- dst_cm->last_height = !src_cm->show_existing_frame ?
- src_cm->height : src_cm->last_height;
+ dst_cm->prev_frame =
+ src_cm->show_existing_frame ? src_cm->prev_frame : src_cm->cur_frame;
+ dst_cm->last_width =
+ !src_cm->show_existing_frame ? src_cm->width : src_cm->last_width;
+ dst_cm->last_height =
+ !src_cm->show_existing_frame ? src_cm->height : src_cm->last_height;
dst_cm->subsampling_x = src_cm->subsampling_x;
dst_cm->subsampling_y = src_cm->subsampling_y;
dst_cm->frame_type = src_cm->frame_type;
- dst_cm->last_show_frame = !src_cm->show_existing_frame ?
- src_cm->show_frame : src_cm->last_show_frame;
+ dst_cm->last_show_frame = !src_cm->show_existing_frame
+ ? src_cm->show_frame
+ : src_cm->last_show_frame;
for (i = 0; i < REF_FRAMES; ++i)
dst_cm->ref_frame_map[i] = src_cm->next_ref_frame_map[i];
@@ -183,7 +184,7 @@
memcpy(dst_cm->frame_contexts, src_cm->frame_contexts,
FRAME_CONTEXTS * sizeof(dst_cm->frame_contexts[0]));
#else
- (void) dst_worker;
- (void) src_worker;
+ (void)dst_worker;
+ (void)src_worker;
#endif // CONFIG_MULTITHREAD
}
diff --git a/vp10/decoder/dthread.h b/vp10/decoder/dthread.h
index 1b0dc01..cb78462 100644
--- a/vp10/decoder/dthread.h
+++ b/vp10/decoder/dthread.h
@@ -57,7 +57,7 @@
// start decoding next frame. So need to check whether worker is still decoding
// ref_buf.
void vp10_frameworker_wait(VPxWorker *const worker, RefCntBuffer *const ref_buf,
- int row);
+ int row);
// FrameWorker broadcasts its decoding progress so other workers that are
// waiting on it can resume decoding.
@@ -65,10 +65,10 @@
// Copy necessary decoding context from src worker to dst worker.
void vp10_frameworker_copy_context(VPxWorker *const dst_worker,
- VPxWorker *const src_worker);
+ VPxWorker *const src_worker);
#ifdef __cplusplus
-} // extern "C"
+} // extern "C"
#endif
#endif // VP10_DECODER_DTHREAD_H_
diff --git a/vp10/encoder/aq_complexity.c b/vp10/encoder/aq_complexity.c
index 2506a4e..912babc 100644
--- a/vp10/encoder/aq_complexity.c
+++ b/vp10/encoder/aq_complexity.c
@@ -19,25 +19,27 @@
#include "vpx_dsp/vpx_dsp_common.h"
#include "vpx_ports/system_state.h"
-#define AQ_C_SEGMENTS 5
-#define DEFAULT_AQ2_SEG 3 // Neutral Q segment
+#define AQ_C_SEGMENTS 5
+#define DEFAULT_AQ2_SEG 3 // Neutral Q segment
#define AQ_C_STRENGTHS 3
-static const double aq_c_q_adj_factor[AQ_C_STRENGTHS][AQ_C_SEGMENTS] =
- { {1.75, 1.25, 1.05, 1.00, 0.90},
- {2.00, 1.50, 1.15, 1.00, 0.85},
- {2.50, 1.75, 1.25, 1.00, 0.80} };
-static const double aq_c_transitions[AQ_C_STRENGTHS][AQ_C_SEGMENTS] =
- { {0.15, 0.30, 0.55, 2.00, 100.0},
- {0.20, 0.40, 0.65, 2.00, 100.0},
- {0.25, 0.50, 0.75, 2.00, 100.0} };
-static const double aq_c_var_thresholds[AQ_C_STRENGTHS][AQ_C_SEGMENTS] =
- { {-4.0, -3.0, -2.0, 100.00, 100.0},
- {-3.5, -2.5, -1.5, 100.00, 100.0},
- {-3.0, -2.0, -1.0, 100.00, 100.0} };
+static const double aq_c_q_adj_factor[AQ_C_STRENGTHS][AQ_C_SEGMENTS] = {
+ { 1.75, 1.25, 1.05, 1.00, 0.90 },
+ { 2.00, 1.50, 1.15, 1.00, 0.85 },
+ { 2.50, 1.75, 1.25, 1.00, 0.80 }
+};
+static const double aq_c_transitions[AQ_C_STRENGTHS][AQ_C_SEGMENTS] = {
+ { 0.15, 0.30, 0.55, 2.00, 100.0 },
+ { 0.20, 0.40, 0.65, 2.00, 100.0 },
+ { 0.25, 0.50, 0.75, 2.00, 100.0 }
+};
+static const double aq_c_var_thresholds[AQ_C_STRENGTHS][AQ_C_SEGMENTS] = {
+ { -4.0, -3.0, -2.0, 100.00, 100.0 },
+ { -3.5, -2.5, -1.5, 100.00, 100.0 },
+ { -3.0, -2.0, -1.0, 100.00, 100.0 }
+};
#define DEFAULT_COMPLEXITY 64
-
static int get_aq_c_strength(int q_index, vpx_bit_depth_t bit_depth) {
// Approximate base quatizer (truncated to int)
const int base_quant = vp10_ac_quant(q_index, 0, bit_depth) / 4;
@@ -81,14 +83,11 @@
for (segment = 0; segment < AQ_C_SEGMENTS; ++segment) {
int qindex_delta;
- if (segment == DEFAULT_AQ2_SEG)
- continue;
+ if (segment == DEFAULT_AQ2_SEG) continue;
- qindex_delta =
- vp10_compute_qdelta_by_rate(&cpi->rc, cm->frame_type, cm->base_qindex,
- aq_c_q_adj_factor[aq_strength][segment],
- cm->bit_depth);
-
+ qindex_delta = vp10_compute_qdelta_by_rate(
+ &cpi->rc, cm->frame_type, cm->base_qindex,
+ aq_c_q_adj_factor[aq_strength][segment], cm->bit_depth);
// For AQ complexity mode, we dont allow Q0 in a segment if the base
// Q is not 0. Q0 (lossless) implies 4x4 only and in AQ mode 2 a segment
@@ -112,7 +111,7 @@
// The choice of segment for a block depends on the ratio of the projected
// bits for the block vs a target average and its spatial complexity.
void vp10_caq_select_segment(VP10_COMP *cpi, MACROBLOCK *mb, BLOCK_SIZE bs,
- int mi_row, int mi_col, int projected_rate) {
+ int mi_row, int mi_col, int projected_rate) {
VP10_COMMON *const cm = &cpi->common;
const int mi_offset = mi_row * cm->mi_cols + mi_col;
@@ -129,26 +128,25 @@
} else {
// Rate depends on fraction of a SB64 in frame (xmis * ymis / bw * bh).
// It is converted to bits * 256 units.
- const int target_rate = (cpi->rc.sb64_target_rate * xmis * ymis * 256) /
- (bw * bh);
+ const int target_rate =
+ (cpi->rc.sb64_target_rate * xmis * ymis * 256) / (bw * bh);
double logvar;
double low_var_thresh;
const int aq_strength = get_aq_c_strength(cm->base_qindex, cm->bit_depth);
vpx_clear_system_state();
- low_var_thresh = (cpi->oxcf.pass == 2)
- ? VPXMAX(cpi->twopass.mb_av_energy, MIN_DEFAULT_LV_THRESH)
- : DEFAULT_LV_THRESH;
+ low_var_thresh = (cpi->oxcf.pass == 2) ? VPXMAX(cpi->twopass.mb_av_energy,
+ MIN_DEFAULT_LV_THRESH)
+ : DEFAULT_LV_THRESH;
vp10_setup_src_planes(mb, cpi->Source, mi_row, mi_col);
logvar = vp10_log_block_var(cpi, mb, bs);
- segment = AQ_C_SEGMENTS - 1; // Just in case no break out below.
+ segment = AQ_C_SEGMENTS - 1; // Just in case no break out below.
for (i = 0; i < AQ_C_SEGMENTS; ++i) {
// Test rate against a threshold value and variance against a threshold.
// Increasing segment number (higher variance and complexity) = higher Q.
- if ((projected_rate <
- target_rate * aq_c_transitions[aq_strength][i]) &&
+ if ((projected_rate < target_rate * aq_c_transitions[aq_strength][i]) &&
(logvar < (low_var_thresh + aq_c_var_thresholds[aq_strength][i]))) {
segment = i;
break;
diff --git a/vp10/encoder/aq_complexity.h b/vp10/encoder/aq_complexity.h
index f9de2ad..5ed461b 100644
--- a/vp10/encoder/aq_complexity.h
+++ b/vp10/encoder/aq_complexity.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_AQ_COMPLEXITY_H_
#define VP10_ENCODER_AQ_COMPLEXITY_H_
@@ -23,8 +22,8 @@
// Select a segment for the current Block.
void vp10_caq_select_segment(struct VP10_COMP *cpi, struct macroblock *,
- BLOCK_SIZE bs,
- int mi_row, int mi_col, int projected_rate);
+ BLOCK_SIZE bs, int mi_row, int mi_col,
+ int projected_rate);
// This function sets up a set of segments with delta Q values around
// the baseline frame quantizer.
diff --git a/vp10/encoder/aq_cyclicrefresh.c b/vp10/encoder/aq_cyclicrefresh.c
index 7a36d61..af49ee6 100644
--- a/vp10/encoder/aq_cyclicrefresh.c
+++ b/vp10/encoder/aq_cyclicrefresh.c
@@ -59,8 +59,7 @@
CYCLIC_REFRESH *vp10_cyclic_refresh_alloc(int mi_rows, int mi_cols) {
size_t last_coded_q_map_size;
CYCLIC_REFRESH *const cr = vpx_calloc(1, sizeof(*cr));
- if (cr == NULL)
- return NULL;
+ if (cr == NULL) return NULL;
cr->map = vpx_calloc(mi_rows * mi_cols, sizeof(*cr->map));
if (cr->map == NULL) {
@@ -94,7 +93,7 @@
// Average bits available per frame = avg_frame_bandwidth
// Number of (8x8) blocks in frame = mi_rows * mi_cols;
const float factor = 0.25;
- const int number_blocks = cm->mi_rows * cm->mi_cols;
+ const int number_blocks = cm->mi_rows * cm->mi_cols;
// The condition below corresponds to turning off at target bitrates:
// (at 30fps), ~12kbps for CIF, 36kbps for VGA, 100kps for HD/720p.
// Also turn off at very small frame sizes, to avoid too large fraction of
@@ -111,10 +110,8 @@
// size of the coding block (i.e., below min_block size rejected), coding
// mode, and rate/distortion.
static int candidate_refresh_aq(const CYCLIC_REFRESH *cr,
- const MB_MODE_INFO *mbmi,
- int64_t rate,
- int64_t dist,
- int bsize) {
+ const MB_MODE_INFO *mbmi, int64_t rate,
+ int64_t dist, int bsize) {
MV mv = mbmi->mv[0].as_mv;
// Reject the block for lower-qp coding if projected distortion
// is above the threshold, and any of the following is true:
@@ -126,11 +123,9 @@
mv.col > cr->motion_thresh || mv.col < -cr->motion_thresh ||
!is_inter_block(mbmi)))
return CR_SEGMENT_ID_BASE;
- else if (bsize >= BLOCK_16X16 &&
- rate < cr->thresh_rate_sb &&
- is_inter_block(mbmi) &&
- mbmi->mv[0].as_int == 0 &&
- cr->rate_boost_fac > 10)
+ else if (bsize >= BLOCK_16X16 && rate < cr->thresh_rate_sb &&
+ is_inter_block(mbmi) && mbmi->mv[0].as_int == 0 &&
+ cr->rate_boost_fac > 10)
// More aggressive delta-q for bigger blocks with zero motion.
return CR_SEGMENT_ID_BOOST2;
else
@@ -141,9 +136,8 @@
static int compute_deltaq(const VP10_COMP *cpi, int q, double rate_factor) {
const CYCLIC_REFRESH *const cr = cpi->cyclic_refresh;
const RATE_CONTROL *const rc = &cpi->rc;
- int deltaq = vp10_compute_qdelta_by_rate(rc, cpi->common.frame_type,
- q, rate_factor,
- cpi->common.bit_depth);
+ int deltaq = vp10_compute_qdelta_by_rate(rc, cpi->common.frame_type, q,
+ rate_factor, cpi->common.bit_depth);
if ((-deltaq) > cr->max_qdelta_perc * q / 100) {
deltaq = -cr->max_qdelta_perc * q / 100;
}
@@ -155,7 +149,7 @@
// (with different delta-q). Note this function is called in the postencode
// (called from rc_update_rate_correction_factors()).
int vp10_cyclic_refresh_estimate_bits_at_q(const VP10_COMP *cpi,
- double correction_factor) {
+ double correction_factor) {
const VP10_COMMON *const cm = &cpi->common;
const CYCLIC_REFRESH *const cr = cpi->cyclic_refresh;
int estimated_bits;
@@ -166,17 +160,18 @@
double weight_segment1 = (double)cr->actual_num_seg1_blocks / num8x8bl;
double weight_segment2 = (double)cr->actual_num_seg2_blocks / num8x8bl;
// Take segment weighted average for estimated bits.
- estimated_bits = (int)((1.0 - weight_segment1 - weight_segment2) *
- vp10_estimate_bits_at_q(cm->frame_type, cm->base_qindex, mbs,
- correction_factor, cm->bit_depth) +
- weight_segment1 *
- vp10_estimate_bits_at_q(cm->frame_type,
- cm->base_qindex + cr->qindex_delta[1], mbs,
- correction_factor, cm->bit_depth) +
- weight_segment2 *
- vp10_estimate_bits_at_q(cm->frame_type,
- cm->base_qindex + cr->qindex_delta[2], mbs,
- correction_factor, cm->bit_depth));
+ estimated_bits =
+ (int)((1.0 - weight_segment1 - weight_segment2) *
+ vp10_estimate_bits_at_q(cm->frame_type, cm->base_qindex, mbs,
+ correction_factor, cm->bit_depth) +
+ weight_segment1 *
+ vp10_estimate_bits_at_q(cm->frame_type,
+ cm->base_qindex + cr->qindex_delta[1],
+ mbs, correction_factor, cm->bit_depth) +
+ weight_segment2 *
+ vp10_estimate_bits_at_q(cm->frame_type,
+ cm->base_qindex + cr->qindex_delta[2],
+ mbs, correction_factor, cm->bit_depth));
return estimated_bits;
}
@@ -186,24 +181,28 @@
// Note: the segment map is set to either 0/CR_SEGMENT_ID_BASE (no refresh) or
// to 1/CR_SEGMENT_ID_BOOST1 (refresh) for each superblock, prior to encoding.
int vp10_cyclic_refresh_rc_bits_per_mb(const VP10_COMP *cpi, int i,
- double correction_factor) {
+ double correction_factor) {
const VP10_COMMON *const cm = &cpi->common;
CYCLIC_REFRESH *const cr = cpi->cyclic_refresh;
int bits_per_mb;
int num8x8bl = cm->MBs << 2;
// Weight for segment prior to encoding: take the average of the target
// number for the frame to be encoded and the actual from the previous frame.
- double weight_segment = (double)((cr->target_num_seg_blocks +
- cr->actual_num_seg1_blocks + cr->actual_num_seg2_blocks) >> 1) /
+ double weight_segment =
+ (double)((cr->target_num_seg_blocks + cr->actual_num_seg1_blocks +
+ cr->actual_num_seg2_blocks) >>
+ 1) /
num8x8bl;
// Compute delta-q corresponding to qindex i.
int deltaq = compute_deltaq(cpi, i, cr->rate_ratio_qdelta);
// Take segment weighted average for bits per mb.
- bits_per_mb = (int)((1.0 - weight_segment) *
- vp10_rc_bits_per_mb(cm->frame_type, i, correction_factor, cm->bit_depth) +
- weight_segment *
- vp10_rc_bits_per_mb(cm->frame_type, i + deltaq, correction_factor,
- cm->bit_depth));
+ bits_per_mb =
+ (int)((1.0 - weight_segment) * vp10_rc_bits_per_mb(cm->frame_type, i,
+ correction_factor,
+ cm->bit_depth) +
+ weight_segment * vp10_rc_bits_per_mb(cm->frame_type, i + deltaq,
+ correction_factor,
+ cm->bit_depth));
return bits_per_mb;
}
@@ -211,12 +210,9 @@
// check if we should reset the segment_id, and update the cyclic_refresh map
// and segmentation map.
void vp10_cyclic_refresh_update_segment(VP10_COMP *const cpi,
- MB_MODE_INFO *const mbmi,
- int mi_row, int mi_col,
- BLOCK_SIZE bsize,
- int64_t rate,
- int64_t dist,
- int skip) {
+ MB_MODE_INFO *const mbmi, int mi_row,
+ int mi_col, BLOCK_SIZE bsize,
+ int64_t rate, int64_t dist, int skip) {
const VP10_COMMON *const cm = &cpi->common;
CYCLIC_REFRESH *const cr = cpi->cyclic_refresh;
const int bw = num_8x8_blocks_wide_lookup[bsize];
@@ -224,19 +220,19 @@
const int xmis = VPXMIN(cm->mi_cols - mi_col, bw);
const int ymis = VPXMIN(cm->mi_rows - mi_row, bh);
const int block_index = mi_row * cm->mi_cols + mi_col;
- const int refresh_this_block = candidate_refresh_aq(cr, mbmi, rate, dist,
- bsize);
+ const int refresh_this_block =
+ candidate_refresh_aq(cr, mbmi, rate, dist, bsize);
// Default is to not update the refresh map.
int new_map_value = cr->map[block_index];
- int x = 0; int y = 0;
+ int x = 0;
+ int y = 0;
// If this block is labeled for refresh, check if we should reset the
// segment_id.
if (cyclic_refresh_segment_id_boosted(mbmi->segment_id)) {
mbmi->segment_id = refresh_this_block;
// Reset segment_id if will be skipped.
- if (skip)
- mbmi->segment_id = CR_SEGMENT_ID_BASE;
+ if (skip) mbmi->segment_id = CR_SEGMENT_ID_BASE;
}
// Update the cyclic refresh map, to be used for setting segmentation map
@@ -249,8 +245,7 @@
// Else if it is accepted as candidate for refresh, and has not already
// been refreshed (marked as 1) then mark it as a candidate for cleanup
// for future time (marked as 0), otherwise don't update it.
- if (cr->map[block_index] == 1)
- new_map_value = 0;
+ if (cr->map[block_index] == 1) new_map_value = 0;
} else {
// Leave it marked as block that is not candidate for refresh.
new_map_value = 1;
@@ -283,11 +278,12 @@
cr->actual_num_seg2_blocks = 0;
for (mi_row = 0; mi_row < cm->mi_rows; mi_row++)
for (mi_col = 0; mi_col < cm->mi_cols; mi_col++) {
- if (cyclic_refresh_segment_id(
- seg_map[mi_row * cm->mi_cols + mi_col]) == CR_SEGMENT_ID_BOOST1)
+ if (cyclic_refresh_segment_id(seg_map[mi_row * cm->mi_cols + mi_col]) ==
+ CR_SEGMENT_ID_BOOST1)
cr->actual_num_seg1_blocks++;
else if (cyclic_refresh_segment_id(
- seg_map[mi_row * cm->mi_cols + mi_col]) == CR_SEGMENT_ID_BOOST2)
+ seg_map[mi_row * cm->mi_cols + mi_col]) ==
+ CR_SEGMENT_ID_BOOST2)
cr->actual_num_seg2_blocks++;
}
}
@@ -324,22 +320,22 @@
for (mi_row = 0; mi_row < rows; mi_row++) {
for (mi_col = 0; mi_col < cols; mi_col++) {
- int16_t abs_mvr = mi[0]->mbmi.mv[0].as_mv.row >= 0 ?
- mi[0]->mbmi.mv[0].as_mv.row : -1 * mi[0]->mbmi.mv[0].as_mv.row;
- int16_t abs_mvc = mi[0]->mbmi.mv[0].as_mv.col >= 0 ?
- mi[0]->mbmi.mv[0].as_mv.col : -1 * mi[0]->mbmi.mv[0].as_mv.col;
+ int16_t abs_mvr = mi[0]->mbmi.mv[0].as_mv.row >= 0
+ ? mi[0]->mbmi.mv[0].as_mv.row
+ : -1 * mi[0]->mbmi.mv[0].as_mv.row;
+ int16_t abs_mvc = mi[0]->mbmi.mv[0].as_mv.col >= 0
+ ? mi[0]->mbmi.mv[0].as_mv.col
+ : -1 * mi[0]->mbmi.mv[0].as_mv.col;
// Calculate the motion of the background.
if (abs_mvr <= 16 && abs_mvc <= 16) {
cnt1++;
- if (abs_mvr == 0 && abs_mvc == 0)
- cnt2++;
+ if (abs_mvr == 0 && abs_mvc == 0) cnt2++;
}
mi++;
// Accumulate low_content_frame.
- if (cr->map[mi_row * cols + mi_col] < 1)
- low_content_frame++;
+ if (cr->map[mi_row * cols + mi_col] < 1) low_content_frame++;
}
mi += 8;
}
@@ -350,7 +346,7 @@
// Also, force this frame as a golden update frame if this frame will change
// the resolution (resize_pending != 0).
if (cpi->resize_pending != 0 ||
- (cnt1 * 10 > (70 * rows * cols) && cnt2 * 20 < cnt1)) {
+ (cnt1 * 10 > (70 * rows * cols) && cnt2 * 20 < cnt1)) {
vp10_cyclic_refresh_set_golden_update(cpi);
rc->frames_till_gf_update_due = rc->baseline_gf_interval;
@@ -360,8 +356,7 @@
force_gf_refresh = 1;
}
- fraction_low =
- (double)low_content_frame / (rows * cols);
+ fraction_low = (double)low_content_frame / (rows * cols);
// Update average.
cr->low_content_avg = (fraction_low + 3 * cr->low_content_avg) / 4;
if (!force_gf_refresh && cpi->refresh_golden_frame == 1) {
@@ -425,8 +420,7 @@
// for possible boost/refresh (segment 1). The segment id may get
// reset to 0 later if block gets coded anything other than ZEROMV.
if (cr->map[bl_index2] == 0) {
- if (cr->last_coded_q_map[bl_index2] > qindex_thresh)
- sum_map++;
+ if (cr->last_coded_q_map[bl_index2] > qindex_thresh) sum_map++;
} else if (cr->map[bl_index2] < 0) {
cr->map[bl_index2]++;
}
@@ -459,14 +453,12 @@
cr->time_for_refresh = 0;
// Use larger delta-qp (increase rate_ratio_qdelta) for first few (~4)
// periods of the refresh cycle, after a key frame.
- if (rc->frames_since_key < 4 * cr->percent_refresh)
+ if (rc->frames_since_key < 4 * cr->percent_refresh)
cr->rate_ratio_qdelta = 3.0;
else
cr->rate_ratio_qdelta = 2.0;
// Adjust some parameters for low resolutions at low bitrates.
- if (cm->width <= 352 &&
- cm->height <= 288 &&
- rc->avg_frame_bandwidth < 3400) {
+ if (cm->width <= 352 && cm->height <= 288 && rc->avg_frame_bandwidth < 3400) {
cr->motion_thresh = 4;
cr->rate_boost_fac = 10;
} else {
@@ -481,9 +473,8 @@
const RATE_CONTROL *const rc = &cpi->rc;
CYCLIC_REFRESH *const cr = cpi->cyclic_refresh;
struct segmentation *const seg = &cm->seg;
- const int apply_cyclic_refresh = apply_cyclic_refresh_bitrate(cm, rc);
- if (cm->current_video_frame == 0)
- cr->low_content_avg = 0.0;
+ const int apply_cyclic_refresh = apply_cyclic_refresh_bitrate(cm, rc);
+ if (cm->current_video_frame == 0) cr->low_content_avg = 0.0;
// Don't apply refresh on key frame or enhancement layer frames.
if (!apply_cyclic_refresh || cm->frame_type == KEY_FRAME) {
// Set segmentation map to 0 and disable.
@@ -517,7 +508,8 @@
seg->abs_delta = SEGMENT_DELTADATA;
// Note: setting temporal_update has no effect, as the seg-map coding method
- // (temporal or spatial) is determined in vp10_choose_segmap_coding_method(),
+ // (temporal or spatial) is determined in
+ // vp10_choose_segmap_coding_method(),
// based on the coding cost of each method. For error_resilient mode on the
// last_frame_seg_map is set to 0, so if temporal coding is used, it is
// relative to 0 previous map.
diff --git a/vp10/encoder/aq_cyclicrefresh.h b/vp10/encoder/aq_cyclicrefresh.h
index f6714c5..649d714 100644
--- a/vp10/encoder/aq_cyclicrefresh.h
+++ b/vp10/encoder/aq_cyclicrefresh.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_AQ_CYCLICREFRESH_H_
#define VP10_ENCODER_AQ_CYCLICREFRESH_H_
@@ -20,9 +19,9 @@
// The segment ids used in cyclic refresh: from base (no boost) to increasing
// boost (higher delta-qp).
-#define CR_SEGMENT_ID_BASE 0
-#define CR_SEGMENT_ID_BOOST1 1
-#define CR_SEGMENT_ID_BOOST2 2
+#define CR_SEGMENT_ID_BASE 0
+#define CR_SEGMENT_ID_BOOST1 1
+#define CR_SEGMENT_ID_BOOST2 2
// Maximum rate target ratio for setting segment delta-qp.
#define CR_MAX_RATE_TARGET_RATIO 4.0
@@ -39,20 +38,20 @@
// Estimate the bits, incorporating the delta-q from segment 1, after encoding
// the frame.
int vp10_cyclic_refresh_estimate_bits_at_q(const struct VP10_COMP *cpi,
- double correction_factor);
+ double correction_factor);
// Estimate the bits per mb, for a given q = i and a corresponding delta-q
// (for segment 1), prior to encoding the frame.
int vp10_cyclic_refresh_rc_bits_per_mb(const struct VP10_COMP *cpi, int i,
- double correction_factor);
+ double correction_factor);
// Prior to coding a given prediction block, of size bsize at (mi_row, mi_col),
// check if we should reset the segment_id, and update the cyclic_refresh map
// and segmentation map.
void vp10_cyclic_refresh_update_segment(struct VP10_COMP *const cpi,
- MB_MODE_INFO *const mbmi,
- int mi_row, int mi_col, BLOCK_SIZE bsize,
- int64_t rate, int64_t dist, int skip);
+ MB_MODE_INFO *const mbmi, int mi_row,
+ int mi_col, BLOCK_SIZE bsize,
+ int64_t rate, int64_t dist, int skip);
// Update the segmentation map, and related quantities: cyclic refresh map,
// refresh sb_index, and target number of blocks to be refreshed.
diff --git a/vp10/encoder/aq_variance.c b/vp10/encoder/aq_variance.c
index 15ab6b4..520640a 100644
--- a/vp10/encoder/aq_variance.c
+++ b/vp10/encoder/aq_variance.c
@@ -22,19 +22,19 @@
#define ENERGY_MIN (-4)
#define ENERGY_MAX (1)
-#define ENERGY_SPAN (ENERGY_MAX - ENERGY_MIN + 1)
-#define ENERGY_IN_BOUNDS(energy)\
+#define ENERGY_SPAN (ENERGY_MAX - ENERGY_MIN + 1)
+#define ENERGY_IN_BOUNDS(energy) \
assert((energy) >= ENERGY_MIN && (energy) <= ENERGY_MAX)
-static const double rate_ratio[MAX_SEGMENTS] =
- {2.5, 2.0, 1.5, 1.0, 0.75, 1.0, 1.0, 1.0};
-static const int segment_id[ENERGY_SPAN] = {0, 1, 1, 2, 3, 4};
+static const double rate_ratio[MAX_SEGMENTS] = { 2.5, 2.0, 1.5, 1.0,
+ 0.75, 1.0, 1.0, 1.0 };
+static const int segment_id[ENERGY_SPAN] = { 0, 1, 1, 2, 3, 4 };
-#define SEGMENT_ID(i) segment_id[(i) - ENERGY_MIN]
+#define SEGMENT_ID(i) segment_id[(i)-ENERGY_MIN]
-DECLARE_ALIGNED(16, static const uint8_t, vp10_64_zeros[64]) = {0};
+DECLARE_ALIGNED(16, static const uint8_t, vp10_64_zeros[64]) = { 0 };
#if CONFIG_VPX_HIGHBITDEPTH
-DECLARE_ALIGNED(16, static const uint16_t, vp10_highbd_64_zeros[64]) = {0};
+DECLARE_ALIGNED(16, static const uint16_t, vp10_highbd_64_zeros[64]) = { 0 };
#endif
unsigned int vp10_vaq_segment_id(int energy) {
@@ -60,7 +60,7 @@
for (i = 0; i < MAX_SEGMENTS; ++i) {
int qindex_delta =
vp10_compute_qdelta_by_rate(&cpi->rc, cm->frame_type, cm->base_qindex,
- rate_ratio[i], cm->bit_depth);
+ rate_ratio[i], cm->bit_depth);
// We don't allow qindex 0 in a segment if the base value is not 0.
// Q index 0 (lossless) implies 4x4 encoding only and in AQ mode a segment
@@ -84,9 +84,9 @@
/* TODO(agrange, paulwilkins): The block_variance calls the unoptimized versions
* of variance() and highbd_8_variance(). It should not.
*/
-static void aq_variance(const uint8_t *a, int a_stride,
- const uint8_t *b, int b_stride,
- int w, int h, unsigned int *sse, int *sum) {
+static void aq_variance(const uint8_t *a, int a_stride, const uint8_t *b,
+ int b_stride, int w, int h, unsigned int *sse,
+ int *sum) {
int i, j;
*sum = 0;
@@ -105,9 +105,9 @@
}
#if CONFIG_VPX_HIGHBITDEPTH
-static void aq_highbd_variance64(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h, uint64_t *sse, uint64_t *sum) {
+static void aq_highbd_variance64(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride, int w, int h,
+ uint64_t *sse, uint64_t *sum) {
int i, j;
uint16_t *a = CONVERT_TO_SHORTPTR(a8);
@@ -126,9 +126,9 @@
}
}
-static void aq_highbd_8_variance(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h, unsigned int *sse, int *sum) {
+static void aq_highbd_8_variance(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride, int w, int h,
+ unsigned int *sse, int *sum) {
uint64_t sse_long = 0;
uint64_t sum_long = 0;
aq_highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
@@ -141,10 +141,10 @@
BLOCK_SIZE bs) {
MACROBLOCKD *xd = &x->e_mbd;
unsigned int var, sse;
- int right_overflow = (xd->mb_to_right_edge < 0) ?
- ((-xd->mb_to_right_edge) >> 3) : 0;
- int bottom_overflow = (xd->mb_to_bottom_edge < 0) ?
- ((-xd->mb_to_bottom_edge) >> 3) : 0;
+ int right_overflow =
+ (xd->mb_to_right_edge < 0) ? ((-xd->mb_to_right_edge) >> 3) : 0;
+ int bottom_overflow =
+ (xd->mb_to_bottom_edge < 0) ? ((-xd->mb_to_bottom_edge) >> 3) : 0;
if (right_overflow || bottom_overflow) {
const int bw = 8 * num_8x8_blocks_wide_lookup[bs] - right_overflow;
@@ -158,30 +158,27 @@
sse >>= 2 * (xd->bd - 8);
avg >>= (xd->bd - 8);
} else {
- aq_variance(x->plane[0].src.buf, x->plane[0].src.stride,
- vp10_64_zeros, 0, bw, bh, &sse, &avg);
+ aq_variance(x->plane[0].src.buf, x->plane[0].src.stride, vp10_64_zeros, 0,
+ bw, bh, &sse, &avg);
}
#else
- aq_variance(x->plane[0].src.buf, x->plane[0].src.stride,
- vp10_64_zeros, 0, bw, bh, &sse, &avg);
+ aq_variance(x->plane[0].src.buf, x->plane[0].src.stride, vp10_64_zeros, 0,
+ bw, bh, &sse, &avg);
#endif // CONFIG_VPX_HIGHBITDEPTH
var = sse - (((int64_t)avg * avg) / (bw * bh));
return (256 * var) / (bw * bh);
} else {
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
- x->plane[0].src.stride,
- CONVERT_TO_BYTEPTR(vp10_highbd_64_zeros),
- 0, &sse);
+ var =
+ cpi->fn_ptr[bs].vf(x->plane[0].src.buf, x->plane[0].src.stride,
+ CONVERT_TO_BYTEPTR(vp10_highbd_64_zeros), 0, &sse);
} else {
- var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
- x->plane[0].src.stride,
+ var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf, x->plane[0].src.stride,
vp10_64_zeros, 0, &sse);
}
#else
- var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
- x->plane[0].src.stride,
+ var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf, x->plane[0].src.stride,
vp10_64_zeros, 0, &sse);
#endif // CONFIG_VPX_HIGHBITDEPTH
return (256 * var) >> num_pels_log2_lookup[bs];
@@ -200,7 +197,7 @@
double energy_midpoint;
vpx_clear_system_state();
energy_midpoint =
- (cpi->oxcf.pass == 2) ? cpi->twopass.mb_av_energy : DEFAULT_E_MIDPOINT;
+ (cpi->oxcf.pass == 2) ? cpi->twopass.mb_av_energy : DEFAULT_E_MIDPOINT;
energy = vp10_log_block_var(cpi, x, bs) - energy_midpoint;
return clamp((int)round(energy), ENERGY_MIN, ENERGY_MAX);
}
diff --git a/vp10/encoder/aq_variance.h b/vp10/encoder/aq_variance.h
index 318f5f2..cfb5b86 100644
--- a/vp10/encoder/aq_variance.h
+++ b/vp10/encoder/aq_variance.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_AQ_VARIANCE_H_
#define VP10_ENCODER_AQ_VARIANCE_H_
diff --git a/vp10/encoder/arm/neon/dct_neon.c b/vp10/encoder/arm/neon/dct_neon.c
index b37a2ff..e83853f 100644
--- a/vp10/encoder/arm/neon/dct_neon.c
+++ b/vp10/encoder/arm/neon/dct_neon.c
@@ -17,20 +17,17 @@
#include "vp10/common/blockd.h"
#include "vpx_dsp/txfm_common.h"
-void vp10_fdct8x8_quant_neon(const int16_t *input, int stride,
- int16_t* coeff_ptr, intptr_t n_coeffs,
- int skip_block, const int16_t* zbin_ptr,
- const int16_t* round_ptr, const int16_t* quant_ptr,
- const int16_t* quant_shift_ptr,
- int16_t* qcoeff_ptr, int16_t* dqcoeff_ptr,
- const int16_t* dequant_ptr, uint16_t* eob_ptr,
- const int16_t* scan_ptr,
- const int16_t* iscan_ptr) {
+void vp10_fdct8x8_quant_neon(
+ const int16_t* input, int stride, int16_t* coeff_ptr, intptr_t n_coeffs,
+ int skip_block, const int16_t* zbin_ptr, const int16_t* round_ptr,
+ const int16_t* quant_ptr, const int16_t* quant_shift_ptr,
+ int16_t* qcoeff_ptr, int16_t* dqcoeff_ptr, const int16_t* dequant_ptr,
+ uint16_t* eob_ptr, const int16_t* scan_ptr, const int16_t* iscan_ptr) {
int16_t temp_buffer[64];
(void)coeff_ptr;
vpx_fdct8x8_neon(input, temp_buffer, stride);
vp10_quantize_fp_neon(temp_buffer, n_coeffs, skip_block, zbin_ptr, round_ptr,
- quant_ptr, quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr,
- dequant_ptr, eob_ptr, scan_ptr, iscan_ptr);
+ quant_ptr, quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr,
+ dequant_ptr, eob_ptr, scan_ptr, iscan_ptr);
}
diff --git a/vp10/encoder/arm/neon/error_neon.c b/vp10/encoder/arm/neon/error_neon.c
index 009520a..34805d3 100644
--- a/vp10/encoder/arm/neon/error_neon.c
+++ b/vp10/encoder/arm/neon/error_neon.c
@@ -14,7 +14,7 @@
#include "./vp10_rtcd.h"
int64_t vp10_block_error_fp_neon(const int16_t *coeff, const int16_t *dqcoeff,
- int block_size) {
+ int block_size) {
int64x2_t error = vdupq_n_s64(0);
assert(block_size >= 8);
diff --git a/vp10/encoder/arm/neon/quantize_neon.c b/vp10/encoder/arm/neon/quantize_neon.c
index 9354ced..7b8e447 100644
--- a/vp10/encoder/arm/neon/quantize_neon.c
+++ b/vp10/encoder/arm/neon/quantize_neon.c
@@ -22,12 +22,12 @@
#include "vp10/encoder/rd.h"
void vp10_quantize_fp_neon(const int16_t *coeff_ptr, intptr_t count,
- int skip_block, const int16_t *zbin_ptr,
- const int16_t *round_ptr, const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr, int16_t *qcoeff_ptr,
- int16_t *dqcoeff_ptr, const int16_t *dequant_ptr,
- uint16_t *eob_ptr,
- const int16_t *scan, const int16_t *iscan) {
+ int skip_block, const int16_t *zbin_ptr,
+ const int16_t *round_ptr, const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr, int16_t *qcoeff_ptr,
+ int16_t *dqcoeff_ptr, const int16_t *dequant_ptr,
+ uint16_t *eob_ptr, const int16_t *scan,
+ const int16_t *iscan) {
// TODO(jingning) Decide the need of these arguments after the
// quantization process is completed.
(void)zbin_ptr;
@@ -54,12 +54,12 @@
const int16x8_t v_coeff = vld1q_s16(&coeff_ptr[0]);
const int16x8_t v_coeff_sign = vshrq_n_s16(v_coeff, 15);
const int16x8_t v_tmp = vabaq_s16(v_round, v_coeff, v_zero);
- const int32x4_t v_tmp_lo = vmull_s16(vget_low_s16(v_tmp),
- vget_low_s16(v_quant));
- const int32x4_t v_tmp_hi = vmull_s16(vget_high_s16(v_tmp),
- vget_high_s16(v_quant));
- const int16x8_t v_tmp2 = vcombine_s16(vshrn_n_s32(v_tmp_lo, 16),
- vshrn_n_s32(v_tmp_hi, 16));
+ const int32x4_t v_tmp_lo =
+ vmull_s16(vget_low_s16(v_tmp), vget_low_s16(v_quant));
+ const int32x4_t v_tmp_hi =
+ vmull_s16(vget_high_s16(v_tmp), vget_high_s16(v_quant));
+ const int16x8_t v_tmp2 =
+ vcombine_s16(vshrn_n_s32(v_tmp_lo, 16), vshrn_n_s32(v_tmp_hi, 16));
const uint16x8_t v_nz_mask = vceqq_s16(v_tmp2, v_zero);
const int16x8_t v_iscan_plus1 = vaddq_s16(v_iscan, v_one);
const int16x8_t v_nz_iscan = vbslq_s16(v_nz_mask, v_zero, v_iscan_plus1);
@@ -79,12 +79,12 @@
const int16x8_t v_coeff = vld1q_s16(&coeff_ptr[i]);
const int16x8_t v_coeff_sign = vshrq_n_s16(v_coeff, 15);
const int16x8_t v_tmp = vabaq_s16(v_round, v_coeff, v_zero);
- const int32x4_t v_tmp_lo = vmull_s16(vget_low_s16(v_tmp),
- vget_low_s16(v_quant));
- const int32x4_t v_tmp_hi = vmull_s16(vget_high_s16(v_tmp),
- vget_high_s16(v_quant));
- const int16x8_t v_tmp2 = vcombine_s16(vshrn_n_s32(v_tmp_lo, 16),
- vshrn_n_s32(v_tmp_hi, 16));
+ const int32x4_t v_tmp_lo =
+ vmull_s16(vget_low_s16(v_tmp), vget_low_s16(v_quant));
+ const int32x4_t v_tmp_hi =
+ vmull_s16(vget_high_s16(v_tmp), vget_high_s16(v_quant));
+ const int16x8_t v_tmp2 =
+ vcombine_s16(vshrn_n_s32(v_tmp_lo, 16), vshrn_n_s32(v_tmp_hi, 16));
const uint16x8_t v_nz_mask = vceqq_s16(v_tmp2, v_zero);
const int16x8_t v_iscan_plus1 = vaddq_s16(v_iscan, v_one);
const int16x8_t v_nz_iscan = vbslq_s16(v_nz_mask, v_zero, v_iscan_plus1);
@@ -96,9 +96,8 @@
vst1q_s16(&dqcoeff_ptr[i], v_dqcoeff);
}
{
- const int16x4_t v_eobmax_3210 =
- vmax_s16(vget_low_s16(v_eobmax_76543210),
- vget_high_s16(v_eobmax_76543210));
+ const int16x4_t v_eobmax_3210 = vmax_s16(
+ vget_low_s16(v_eobmax_76543210), vget_high_s16(v_eobmax_76543210));
const int64x1_t v_eobmax_xx32 =
vshr_n_s64(vreinterpret_s64_s16(v_eobmax_3210), 32);
const int16x4_t v_eobmax_tmp =
diff --git a/vp10/encoder/bitstream.c b/vp10/encoder/bitstream.c
index 0b26458..0fd3d5a 100644
--- a/vp10/encoder/bitstream.c
+++ b/vp10/encoder/bitstream.c
@@ -36,20 +36,22 @@
#include "vp10/encoder/tokenize.h"
static const struct vp10_token intra_mode_encodings[INTRA_MODES] = {
- {0, 1}, {6, 3}, {28, 5}, {30, 5}, {58, 6}, {59, 6}, {126, 7}, {127, 7},
- {62, 6}, {2, 2}};
+ { 0, 1 }, { 6, 3 }, { 28, 5 }, { 30, 5 }, { 58, 6 },
+ { 59, 6 }, { 126, 7 }, { 127, 7 }, { 62, 6 }, { 2, 2 }
+};
static const struct vp10_token switchable_interp_encodings[SWITCHABLE_FILTERS] =
- {{0, 1}, {2, 2}, {3, 2}};
-static const struct vp10_token partition_encodings[PARTITION_TYPES] =
- {{0, 1}, {2, 2}, {6, 3}, {7, 3}};
-static const struct vp10_token inter_mode_encodings[INTER_MODES] =
- {{2, 2}, {6, 3}, {0, 1}, {7, 3}};
+ { { 0, 1 }, { 2, 2 }, { 3, 2 } };
+static const struct vp10_token partition_encodings[PARTITION_TYPES] = {
+ { 0, 1 }, { 2, 2 }, { 6, 3 }, { 7, 3 }
+};
+static const struct vp10_token inter_mode_encodings[INTER_MODES] = {
+ { 2, 2 }, { 6, 3 }, { 0, 1 }, { 7, 3 }
+};
static INLINE void write_uniform(vpx_writer *w, int n, int v) {
int l = get_unsigned_bits(n);
int m = (1 << l) - n;
- if (l == 0)
- return;
+ if (l == 0) return;
if (v < m) {
vpx_write_literal(w, v, l - 1);
} else {
@@ -73,18 +75,18 @@
const vpx_prob *probs) {
assert(is_inter_mode(mode));
vp10_write_token(w, vp10_inter_mode_tree, probs,
- &inter_mode_encodings[INTER_OFFSET(mode)]);
+ &inter_mode_encodings[INTER_OFFSET(mode)]);
}
-static void encode_unsigned_max(struct vpx_write_bit_buffer *wb,
- int data, int max) {
+static void encode_unsigned_max(struct vpx_write_bit_buffer *wb, int data,
+ int max) {
vpx_wb_write_literal(wb, data, get_unsigned_bits(max));
}
static void prob_diff_update(const vpx_tree_index *tree,
vpx_prob probs[/*n - 1*/],
- const unsigned int counts[/*n - 1*/],
- int n, vpx_writer *w) {
+ const unsigned int counts[/*n - 1*/], int n,
+ vpx_writer *w) {
int i;
unsigned int branch_ct[32][2];
@@ -108,19 +110,18 @@
assert(n <= 32);
vp10_tree_probs_from_distribution(tree, branch_ct, counts);
for (i = 0; i < n - 1; ++i) {
- savings += vp10_cond_prob_diff_update_savings(&probs[i],
- branch_ct[i]);
+ savings += vp10_cond_prob_diff_update_savings(&probs[i], branch_ct[i]);
}
return savings;
}
-static void write_selected_tx_size(const VP10_COMMON *cm,
- const MACROBLOCKD *xd, vpx_writer *w) {
+static void write_selected_tx_size(const VP10_COMMON *cm, const MACROBLOCKD *xd,
+ vpx_writer *w) {
TX_SIZE tx_size = xd->mi[0]->mbmi.tx_size;
BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
const TX_SIZE max_tx_size = max_txsize_lookup[bsize];
- const vpx_prob *const tx_probs = get_tx_probs2(max_tx_size, xd,
- &cm->fc->tx_probs);
+ const vpx_prob *const tx_probs =
+ get_tx_probs2(max_tx_size, xd, &cm->fc->tx_probs);
vpx_write(w, tx_size != TX_4X4, tx_probs[0]);
if (tx_size != TX_4X4 && max_tx_size >= TX_16X16) {
vpx_write(w, tx_size != TX_8X8, tx_probs[1]);
@@ -175,36 +176,32 @@
if (do_update) {
for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
for (j = 0; j < TX_TYPES; ++j)
- prob_diff_update(vp10_ext_tx_tree,
- cm->fc->intra_ext_tx_prob[i][j],
- cm->counts.intra_ext_tx[i][j],
- TX_TYPES, w);
+ prob_diff_update(vp10_ext_tx_tree, cm->fc->intra_ext_tx_prob[i][j],
+ cm->counts.intra_ext_tx[i][j], TX_TYPES, w);
}
}
savings = 0;
for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
- savings += prob_diff_update_savings(
- vp10_ext_tx_tree, cm->fc->inter_ext_tx_prob[i],
- cm->counts.inter_ext_tx[i], TX_TYPES);
+ savings +=
+ prob_diff_update_savings(vp10_ext_tx_tree, cm->fc->inter_ext_tx_prob[i],
+ cm->counts.inter_ext_tx[i], TX_TYPES);
}
do_update = savings > savings_thresh;
vpx_write(w, do_update, GROUP_DIFF_UPDATE_PROB);
if (do_update) {
for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
- prob_diff_update(vp10_ext_tx_tree,
- cm->fc->inter_ext_tx_prob[i],
- cm->counts.inter_ext_tx[i],
- TX_TYPES, w);
+ prob_diff_update(vp10_ext_tx_tree, cm->fc->inter_ext_tx_prob[i],
+ cm->counts.inter_ext_tx[i], TX_TYPES, w);
}
}
}
-static void pack_mb_tokens(vpx_writer *w,
- TOKENEXTRA **tp, const TOKENEXTRA *const stop,
+static void pack_mb_tokens(vpx_writer *w, TOKENEXTRA **tp,
+ const TOKENEXTRA *const stop,
vpx_bit_depth_t bit_depth, const TX_SIZE tx) {
TOKENEXTRA *p = *tp;
#if !CONFIG_MISC_FIXES
- (void) tx;
+ (void)tx;
#endif
while (p < stop && p->token != EOSB_TOKEN) {
@@ -223,7 +220,7 @@
b = &vp10_extra_bits[t];
#else
const vp10_extra_bit *const b = &vp10_extra_bits[t];
- (void) bit_depth;
+ (void)bit_depth;
#endif // CONFIG_VPX_HIGHBITDEPTH
/* skip one or two nodes */
@@ -245,8 +242,8 @@
int bits = v >> (n - len);
vp10_write_tree(w, vp10_coef_tree, p->context_tree, bits, len, i);
vp10_write_tree(w, vp10_coef_con_tree,
- vp10_pareto8_full[p->context_tree[PIVOT_NODE] - 1],
- v, n - len, 0);
+ vp10_pareto8_full[p->context_tree[PIVOT_NODE] - 1], v,
+ n - len, 0);
} else {
vp10_write_tree(w, vp10_coef_tree, p->context_tree, v, n, i);
}
@@ -254,8 +251,7 @@
if (b->base_val) {
const int e = p->extra, l = b->len;
#if CONFIG_MISC_FIXES
- int skip_bits =
- (b->base_val == CAT6_MIN_VAL) ? TX_SIZES - 1 - tx : 0;
+ int skip_bits = (b->base_val == CAT6_MIN_VAL) ? TX_SIZES - 1 - tx : 0;
#else
int skip_bits = 0;
#endif
@@ -263,7 +259,7 @@
if (l) {
const unsigned char *pb = b->prob;
int v = e >> 1;
- int n = l; /* number of bits in v, assumed nonzero */
+ int n = l; /* number of bits in v, assumed nonzero */
int i = 0;
do {
@@ -305,7 +301,7 @@
if (segfeature_active(&cm->seg, segment_id, SEG_LVL_REF_FRAME)) {
assert(!is_compound);
assert(mbmi->ref_frame[0] ==
- get_segdata(&cm->seg, segment_id, SEG_LVL_REF_FRAME));
+ get_segdata(&cm->seg, segment_id, SEG_LVL_REF_FRAME));
} else {
// does the feature use compound prediction or not
// (if not specified at the frame/segment level)
@@ -356,8 +352,7 @@
const int pred_flag = mbmi->seg_id_predicted;
vpx_prob pred_prob = vp10_get_pred_prob_seg_id(segp, xd);
vpx_write(w, pred_flag, pred_prob);
- if (!pred_flag)
- write_segment_id(w, seg, segp, segment_id);
+ if (!pred_flag) write_segment_id(w, seg, segp, segment_id);
} else {
write_segment_id(w, seg, segp, segment_id);
}
@@ -403,8 +398,8 @@
if (cm->interp_filter == SWITCHABLE) {
const int ctx = vp10_get_pred_context_switchable_interp(xd);
vp10_write_token(w, vp10_switchable_interp_tree,
- cm->fc->switchable_interp_prob[ctx],
- &switchable_interp_encodings[mbmi->interp_filter]);
+ cm->fc->switchable_interp_prob[ctx],
+ &switchable_interp_encodings[mbmi->interp_filter]);
++cpi->interp_filter_selected[0][mbmi->interp_filter];
} else {
assert(mbmi->interp_filter == cm->interp_filter);
@@ -422,8 +417,8 @@
if (b_mode == NEWMV) {
for (ref = 0; ref < 1 + is_compound; ++ref)
vp10_encode_mv(cpi, w, &mi->bmi[j].as_mv[ref].as_mv,
- &mbmi_ext->ref_mvs[mbmi->ref_frame[ref]][0].as_mv,
- nmvc, allow_hp);
+ &mbmi_ext->ref_mvs[mbmi->ref_frame[ref]][0].as_mv,
+ nmvc, allow_hp);
}
}
}
@@ -431,19 +426,17 @@
if (mode == NEWMV) {
for (ref = 0; ref < 1 + is_compound; ++ref)
vp10_encode_mv(cpi, w, &mbmi->mv[ref].as_mv,
- &mbmi_ext->ref_mvs[mbmi->ref_frame[ref]][0].as_mv, nmvc,
- allow_hp);
+ &mbmi_ext->ref_mvs[mbmi->ref_frame[ref]][0].as_mv,
+ nmvc, allow_hp);
}
}
}
- if (mbmi->tx_size < TX_32X32 &&
- cm->base_qindex > 0 && !mbmi->skip &&
+ if (mbmi->tx_size < TX_32X32 && cm->base_qindex > 0 && !mbmi->skip &&
!segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
if (is_inter) {
- vp10_write_token(
- w, vp10_ext_tx_tree,
- cm->fc->inter_ext_tx_prob[mbmi->tx_size],
- &ext_tx_encodings[mbmi->tx_type]);
+ vp10_write_token(w, vp10_ext_tx_tree,
+ cm->fc->inter_ext_tx_prob[mbmi->tx_size],
+ &ext_tx_encodings[mbmi->tx_type]);
} else {
vp10_write_token(
w, vp10_ext_tx_tree,
@@ -452,8 +445,7 @@
&ext_tx_encodings[mbmi->tx_type]);
}
} else {
- if (!mbmi->skip)
- assert(mbmi->tx_type == DCT_DCT);
+ if (!mbmi->skip) assert(mbmi->tx_type == DCT_DCT);
}
}
@@ -471,8 +463,7 @@
const MB_MODE_INFO *const mbmi = &mi->mbmi;
const BLOCK_SIZE bsize = mbmi->sb_type;
- if (seg->update_map)
- write_segment_id(w, seg, segp, mbmi->segment_id);
+ if (seg->update_map) write_segment_id(w, seg, segp, mbmi->segment_id);
write_skip(cm, xd, mbmi->segment_id, mi, w);
@@ -499,8 +490,7 @@
write_intra_mode(w, mbmi->uv_mode, cm->fc->uv_mode_prob[mbmi->mode]);
- if (mbmi->tx_size < TX_32X32 &&
- cm->base_qindex > 0 && !mbmi->skip &&
+ if (mbmi->tx_size < TX_32X32 && cm->base_qindex > 0 && !mbmi->skip &&
!segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
vp10_write_token(
w, vp10_ext_tx_tree,
@@ -512,8 +502,8 @@
static void write_modes_b(VP10_COMP *cpi, const TileInfo *const tile,
vpx_writer *w, TOKENEXTRA **tok,
- const TOKENEXTRA *const tok_end,
- int mi_row, int mi_col) {
+ const TOKENEXTRA *const tok_end, int mi_row,
+ int mi_col) {
const VP10_COMMON *const cm = &cpi->common;
MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
MODE_INFO *m;
@@ -524,8 +514,7 @@
cpi->td.mb.mbmi_ext = cpi->mbmi_ext_base + (mi_row * cm->mi_cols + mi_col);
- set_mi_row_col(xd, tile,
- mi_row, num_8x8_blocks_high_lookup[m->mbmi.sb_type],
+ set_mi_row_col(xd, tile, mi_row, num_8x8_blocks_high_lookup[m->mbmi.sb_type],
mi_col, num_8x8_blocks_wide_lookup[m->mbmi.sb_type],
cm->mi_rows, cm->mi_cols);
if (frame_is_intra_only(cm)) {
@@ -537,8 +526,8 @@
if (!m->mbmi.skip) {
assert(*tok < tok_end);
for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
- TX_SIZE tx = plane ? get_uv_tx_size(&m->mbmi, &xd->plane[plane])
- : m->mbmi.tx_size;
+ TX_SIZE tx =
+ plane ? get_uv_tx_size(&m->mbmi, &xd->plane[plane]) : m->mbmi.tx_size;
pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx);
assert(*tok < tok_end && (*tok)->token == EOSB_TOKEN);
(*tok)++;
@@ -547,9 +536,9 @@
}
static void write_partition(const VP10_COMMON *const cm,
- const MACROBLOCKD *const xd,
- int hbs, int mi_row, int mi_col,
- PARTITION_TYPE p, BLOCK_SIZE bsize, vpx_writer *w) {
+ const MACROBLOCKD *const xd, int hbs, int mi_row,
+ int mi_col, PARTITION_TYPE p, BLOCK_SIZE bsize,
+ vpx_writer *w) {
const int ctx = partition_plane_context(xd, mi_row, mi_col, bsize);
const vpx_prob *const probs = cm->fc->partition_prob[ctx];
const int has_rows = (mi_row + hbs) < cm->mi_rows;
@@ -568,10 +557,10 @@
}
}
-static void write_modes_sb(VP10_COMP *cpi,
- const TileInfo *const tile, vpx_writer *w,
- TOKENEXTRA **tok, const TOKENEXTRA *const tok_end,
- int mi_row, int mi_col, BLOCK_SIZE bsize) {
+static void write_modes_sb(VP10_COMP *cpi, const TileInfo *const tile,
+ vpx_writer *w, TOKENEXTRA **tok,
+ const TOKENEXTRA *const tok_end, int mi_row,
+ int mi_col, BLOCK_SIZE bsize) {
const VP10_COMMON *const cm = &cpi->common;
MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
@@ -581,8 +570,7 @@
BLOCK_SIZE subsize;
const MODE_INFO *m = NULL;
- if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
- return;
+ if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols) return;
m = cm->mi_grid_visible[mi_row * cm->mi_stride + mi_col];
@@ -615,8 +603,7 @@
write_modes_sb(cpi, tile, w, tok, tok_end, mi_row + bs, mi_col + bs,
subsize);
break;
- default:
- assert(0);
+ default: assert(0);
}
}
@@ -626,9 +613,9 @@
update_partition_context(xd, mi_row, mi_col, subsize, bsize);
}
-static void write_modes(VP10_COMP *cpi,
- const TileInfo *const tile, vpx_writer *w,
- TOKENEXTRA **tok, const TOKENEXTRA *const tok_end) {
+static void write_modes(VP10_COMP *cpi, const TileInfo *const tile,
+ vpx_writer *w, TOKENEXTRA **tok,
+ const TOKENEXTRA *const tok_end) {
MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
int mi_row, mi_col;
@@ -637,8 +624,7 @@
vp10_zero(xd->left_seg_context);
for (mi_col = tile->mi_col_start; mi_col < tile->mi_col_end;
mi_col += MI_BLOCK_SIZE)
- write_modes_sb(cpi, tile, w, tok, tok_end, mi_row, mi_col,
- BLOCK_64X64);
+ write_modes_sb(cpi, tile, w, tok, tok_end, mi_row, mi_col, BLOCK_64X64);
}
}
@@ -646,7 +632,7 @@
vp10_coeff_stats *coef_branch_ct,
vp10_coeff_probs_model *coef_probs) {
vp10_coeff_count *coef_counts = cpi->td.rd_counts.coef_counts[tx_size];
- unsigned int (*eob_branch_ct)[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS] =
+ unsigned int(*eob_branch_ct)[REF_TYPES][COEF_BANDS][COEFF_CONTEXTS] =
cpi->common.counts.eob_branch[tx_size];
int i, j, k, l, m;
@@ -655,21 +641,21 @@
for (k = 0; k < COEF_BANDS; ++k) {
for (l = 0; l < BAND_COEFF_CONTEXTS(k); ++l) {
vp10_tree_probs_from_distribution(vp10_coef_tree,
- coef_branch_ct[i][j][k][l],
- coef_counts[i][j][k][l]);
- coef_branch_ct[i][j][k][l][0][1] = eob_branch_ct[i][j][k][l] -
- coef_branch_ct[i][j][k][l][0][0];
+ coef_branch_ct[i][j][k][l],
+ coef_counts[i][j][k][l]);
+ coef_branch_ct[i][j][k][l][0][1] =
+ eob_branch_ct[i][j][k][l] - coef_branch_ct[i][j][k][l][0][0];
for (m = 0; m < UNCONSTRAINED_NODES; ++m)
- coef_probs[i][j][k][l][m] = get_binary_prob(
- coef_branch_ct[i][j][k][l][m][0],
- coef_branch_ct[i][j][k][l][m][1]);
+ coef_probs[i][j][k][l][m] =
+ get_binary_prob(coef_branch_ct[i][j][k][l][m][0],
+ coef_branch_ct[i][j][k][l][m][1]);
}
}
}
}
}
-static void update_coef_probs_common(vpx_writer* const bc, VP10_COMP *cpi,
+static void update_coef_probs_common(vpx_writer *const bc, VP10_COMP *cpi,
TX_SIZE tx_size,
vp10_coeff_stats *frame_branch_ct,
vp10_coeff_probs_model *new_coef_probs) {
@@ -683,7 +669,7 @@
case TWO_LOOP: {
/* dry run to see if there is any update at all needed */
int savings = 0;
- int update[2] = {0, 0};
+ int update[2] = { 0, 0 };
for (i = 0; i < PLANE_TYPES; ++i) {
for (j = 0; j < REF_TYPES; ++j) {
for (k = 0; k < COEF_BANDS; ++k) {
@@ -700,8 +686,7 @@
else
s = vp10_prob_diff_update_savings_search(
frame_branch_ct[i][j][k][l][t], oldp, &newp, upd);
- if (s > 0 && newp != oldp)
- u = 1;
+ if (s > 0 && newp != oldp) u = 1;
if (u)
savings += s - (int)(vp10_cost_zero(upd));
else
@@ -737,10 +722,8 @@
old_coef_probs[i][j][k][l], &newp, upd, stepsize);
else
s = vp10_prob_diff_update_savings_search(
- frame_branch_ct[i][j][k][l][t],
- *oldp, &newp, upd);
- if (s > 0 && newp != *oldp)
- u = 1;
+ frame_branch_ct[i][j][k][l][t], *oldp, &newp, upd);
+ if (s > 0 && newp != *oldp) u = 1;
vpx_write(bc, u, upd);
if (u) {
/* send/use new probability */
@@ -775,12 +758,10 @@
old_coef_probs[i][j][k][l], &newp, upd, stepsize);
} else {
s = vp10_prob_diff_update_savings_search(
- frame_branch_ct[i][j][k][l][t],
- *oldp, &newp, upd);
+ frame_branch_ct[i][j][k][l][t], *oldp, &newp, upd);
}
- if (s > 0 && newp != *oldp)
- u = 1;
+ if (s > 0 && newp != *oldp) u = 1;
updates += u;
if (u == 0 && updates == 0) {
noupdates_before_first++;
@@ -809,12 +790,11 @@
}
return;
}
- default:
- assert(0);
+ default: assert(0);
}
}
-static void update_coef_probs(VP10_COMP *cpi, vpx_writer* w) {
+static void update_coef_probs(VP10_COMP *cpi, vpx_writer *w) {
const TX_MODE tx_mode = cpi->common.tx_mode;
const TX_SIZE max_tx_size = tx_mode_to_biggest_tx_size[tx_mode];
TX_SIZE tx_size;
@@ -825,8 +805,7 @@
(tx_size >= TX_16X16 && cpi->sf.tx_size_search_method == USE_TX_8X8)) {
vpx_write_bit(w, 0);
} else {
- build_tree_distribution(cpi, tx_size, frame_branch_ct,
- frame_coef_probs);
+ build_tree_distribution(cpi, tx_size, frame_branch_ct, frame_coef_probs);
update_coef_probs_common(w, cpi, tx_size, frame_branch_ct,
frame_coef_probs);
}
@@ -898,8 +877,7 @@
#endif
vpx_wb_write_bit(wb, seg->enabled);
- if (!seg->enabled)
- return;
+ if (!seg->enabled) return;
// Segmentation map
if (!frame_is_intra_only(cm) && !cm->error_resilient_mode) {
@@ -916,8 +894,7 @@
const int prob = segp->tree_probs[i];
const int update = prob != MAX_PROB;
vpx_wb_write_bit(wb, update);
- if (update)
- vpx_wb_write_literal(wb, prob, 8);
+ if (update) vpx_wb_write_literal(wb, prob, 8);
}
#endif
@@ -934,8 +911,7 @@
const int prob = segp->pred_probs[i];
const int update = prob != MAX_PROB;
vpx_wb_write_bit(wb, update);
- if (update)
- vpx_wb_write_literal(wb, prob, 8);
+ if (update) vpx_wb_write_literal(wb, prob, 8);
}
}
#endif
@@ -970,48 +946,42 @@
static void update_seg_probs(VP10_COMP *cpi, vpx_writer *w) {
VP10_COMMON *cm = &cpi->common;
- if (!cpi->common.seg.enabled)
- return;
+ if (!cpi->common.seg.enabled) return;
if (cpi->common.seg.temporal_update) {
int i;
for (i = 0; i < PREDICTION_PROBS; i++)
vp10_cond_prob_diff_update(w, &cm->fc->seg.pred_probs[i],
- cm->counts.seg.pred[i]);
+ cm->counts.seg.pred[i]);
prob_diff_update(vp10_segment_tree, cm->fc->seg.tree_probs,
- cm->counts.seg.tree_mispred, MAX_SEGMENTS, w);
+ cm->counts.seg.tree_mispred, MAX_SEGMENTS, w);
} else {
prob_diff_update(vp10_segment_tree, cm->fc->seg.tree_probs,
- cm->counts.seg.tree_total, MAX_SEGMENTS, w);
+ cm->counts.seg.tree_total, MAX_SEGMENTS, w);
}
}
static void write_txfm_mode(TX_MODE mode, struct vpx_write_bit_buffer *wb) {
vpx_wb_write_bit(wb, mode == TX_MODE_SELECT);
- if (mode != TX_MODE_SELECT)
- vpx_wb_write_literal(wb, mode, 2);
+ if (mode != TX_MODE_SELECT) vpx_wb_write_literal(wb, mode, 2);
}
#else
static void write_txfm_mode(TX_MODE mode, struct vpx_writer *wb) {
vpx_write_literal(wb, VPXMIN(mode, ALLOW_32X32), 2);
- if (mode >= ALLOW_32X32)
- vpx_write_bit(wb, mode == TX_MODE_SELECT);
+ if (mode >= ALLOW_32X32) vpx_write_bit(wb, mode == TX_MODE_SELECT);
}
#endif
-
static void update_txfm_probs(VP10_COMMON *cm, vpx_writer *w,
FRAME_COUNTS *counts) {
-
if (cm->tx_mode == TX_MODE_SELECT) {
int i, j;
unsigned int ct_8x8p[TX_SIZES - 3][2];
unsigned int ct_16x16p[TX_SIZES - 2][2];
unsigned int ct_32x32p[TX_SIZES - 1][2];
-
for (i = 0; i < TX_SIZE_CONTEXTS; i++) {
vp10_tx_counts_to_branch_counts_8x8(counts->tx.p8x8[i], ct_8x8p);
for (j = 0; j < TX_SIZES - 3; j++)
@@ -1022,14 +992,14 @@
vp10_tx_counts_to_branch_counts_16x16(counts->tx.p16x16[i], ct_16x16p);
for (j = 0; j < TX_SIZES - 2; j++)
vp10_cond_prob_diff_update(w, &cm->fc->tx_probs.p16x16[i][j],
- ct_16x16p[j]);
+ ct_16x16p[j]);
}
for (i = 0; i < TX_SIZE_CONTEXTS; i++) {
vp10_tx_counts_to_branch_counts_32x32(counts->tx.p32x32[i], ct_32x32p);
for (j = 0; j < TX_SIZES - 1; j++)
vp10_cond_prob_diff_update(w, &cm->fc->tx_probs.p32x32[i][j],
- ct_32x32p[j]);
+ ct_32x32p[j]);
}
}
}
@@ -1037,8 +1007,7 @@
static void write_interp_filter(INTERP_FILTER filter,
struct vpx_write_bit_buffer *wb) {
vpx_wb_write_bit(wb, filter == SWITCHABLE);
- if (filter != SWITCHABLE)
- vpx_wb_write_literal(wb, filter, 2);
+ if (filter != SWITCHABLE) vpx_wb_write_literal(wb, filter, 2);
}
static void fix_interp_filter(VP10_COMMON *cm, FRAME_COUNTS *counts) {
@@ -1071,16 +1040,13 @@
// columns
ones = cm->log2_tile_cols - min_log2_tile_cols;
- while (ones--)
- vpx_wb_write_bit(wb, 1);
+ while (ones--) vpx_wb_write_bit(wb, 1);
- if (cm->log2_tile_cols < max_log2_tile_cols)
- vpx_wb_write_bit(wb, 0);
+ if (cm->log2_tile_cols < max_log2_tile_cols) vpx_wb_write_bit(wb, 0);
// rows
vpx_wb_write_bit(wb, cm->log2_tile_rows != 0);
- if (cm->log2_tile_rows != 0)
- vpx_wb_write_bit(wb, cm->log2_tile_rows != 1);
+ if (cm->log2_tile_rows != 0) vpx_wb_write_bit(wb, cm->log2_tile_rows != 1);
}
static int get_refresh_mask(VP10_COMP *cpi) {
@@ -1129,15 +1095,15 @@
TOKENEXTRA *tok = cpi->tile_tok[tile_row][tile_col];
tok_end = cpi->tile_tok[tile_row][tile_col] +
- cpi->tok_count[tile_row][tile_col];
+ cpi->tok_count[tile_row][tile_col];
if (tile_col < tile_cols - 1 || tile_row < tile_rows - 1)
vpx_start_encode(&residual_bc, data_ptr + total_size + 4);
else
vpx_start_encode(&residual_bc, data_ptr + total_size);
- write_modes(cpi, &cpi->tile_data[tile_idx].tile_info,
- &residual_bc, &tok, tok_end);
+ write_modes(cpi, &cpi->tile_data[tile_idx].tile_info, &residual_bc, &tok,
+ tok_end);
assert(tok == tok_end);
vpx_stop_encode(&residual_bc);
if (tile_col < tile_cols - 1 || tile_row < tile_rows - 1) {
@@ -1161,8 +1127,8 @@
static void write_render_size(const VP10_COMMON *cm,
struct vpx_write_bit_buffer *wb) {
- const int scaling_active = cm->width != cm->render_width ||
- cm->height != cm->render_height;
+ const int scaling_active =
+ cm->width != cm->render_width || cm->height != cm->render_height;
vpx_wb_write_bit(wb, scaling_active);
if (scaling_active) {
vpx_wb_write_literal(wb, cm->render_width - 1, 16);
@@ -1188,8 +1154,8 @@
YV12_BUFFER_CONFIG *cfg = get_ref_frame_buffer(cpi, ref_frame);
if (cfg != NULL) {
- found = cm->width == cfg->y_crop_width &&
- cm->height == cfg->y_crop_height;
+ found =
+ cm->width == cfg->y_crop_width && cm->height == cfg->y_crop_height;
#if CONFIG_MISC_FIXES
found &= cm->render_width == cfg->render_width &&
cm->render_height == cfg->render_height;
@@ -1224,20 +1190,11 @@
static void write_profile(BITSTREAM_PROFILE profile,
struct vpx_write_bit_buffer *wb) {
switch (profile) {
- case PROFILE_0:
- vpx_wb_write_literal(wb, 0, 2);
- break;
- case PROFILE_1:
- vpx_wb_write_literal(wb, 2, 2);
- break;
- case PROFILE_2:
- vpx_wb_write_literal(wb, 1, 2);
- break;
- case PROFILE_3:
- vpx_wb_write_literal(wb, 6, 3);
- break;
- default:
- assert(0);
+ case PROFILE_0: vpx_wb_write_literal(wb, 0, 2); break;
+ case PROFILE_1: vpx_wb_write_literal(wb, 2, 2); break;
+ case PROFILE_2: vpx_wb_write_literal(wb, 1, 2); break;
+ case PROFILE_3: vpx_wb_write_literal(wb, 6, 3); break;
+ default: assert(0);
}
}
@@ -1284,8 +1241,7 @@
write_bitdepth_colorspace_sampling(cm, wb);
write_frame_size(cm, wb);
} else {
- if (!cm->show_frame)
- vpx_wb_write_bit(wb, cm->intra_only);
+ if (!cm->show_frame) vpx_wb_write_bit(wb, cm->intra_only);
if (!cm->error_resilient_mode) {
#if CONFIG_MISC_FIXES
@@ -1302,8 +1258,8 @@
#else
static const int reset_frame_context_conv_tbl[3] = { 0, 2, 3 };
- vpx_wb_write_literal(wb,
- reset_frame_context_conv_tbl[cm->reset_frame_context], 2);
+ vpx_wb_write_literal(
+ wb, reset_frame_context_conv_tbl[cm->reset_frame_context], 2);
#endif
}
@@ -1346,8 +1302,8 @@
#if CONFIG_MISC_FIXES
if (cm->refresh_frame_context != REFRESH_FRAME_CONTEXT_OFF)
#endif
- vpx_wb_write_bit(wb, cm->refresh_frame_context !=
- REFRESH_FRAME_CONTEXT_BACKWARD);
+ vpx_wb_write_bit(
+ wb, cm->refresh_frame_context != REFRESH_FRAME_CONTEXT_BACKWARD);
}
vpx_wb_write_literal(wb, cm->frame_context_idx, FRAME_CONTEXTS_LOG2);
@@ -1365,8 +1321,7 @@
const int use_compound_pred = cm->reference_mode != SINGLE_REFERENCE;
vpx_wb_write_bit(wb, use_hybrid_pred);
- if (!use_hybrid_pred)
- vpx_wb_write_bit(wb, use_compound_pred);
+ if (!use_hybrid_pred) vpx_wb_write_bit(wb, use_compound_pred);
}
#endif
@@ -1427,7 +1382,7 @@
for (i = 0; i < INTRA_INTER_CONTEXTS; i++)
vp10_cond_prob_diff_update(&header_bc, &fc->intra_inter_prob[i],
- counts->intra_inter[i]);
+ counts->intra_inter[i]);
if (cpi->allow_comp_inter_inter) {
const int use_hybrid_pred = cm->reference_mode == REFERENCE_MODE_SELECT;
@@ -1440,7 +1395,7 @@
if (use_hybrid_pred)
for (i = 0; i < COMP_INTER_CONTEXTS; i++)
vp10_cond_prob_diff_update(&header_bc, &fc->comp_inter_prob[i],
- counts->comp_inter[i]);
+ counts->comp_inter[i]);
}
#else
if (use_hybrid_pred)
@@ -1453,16 +1408,16 @@
if (cm->reference_mode != COMPOUND_REFERENCE) {
for (i = 0; i < REF_CONTEXTS; i++) {
vp10_cond_prob_diff_update(&header_bc, &fc->single_ref_prob[i][0],
- counts->single_ref[i][0]);
+ counts->single_ref[i][0]);
vp10_cond_prob_diff_update(&header_bc, &fc->single_ref_prob[i][1],
- counts->single_ref[i][1]);
+ counts->single_ref[i][1]);
}
}
if (cm->reference_mode != SINGLE_REFERENCE)
for (i = 0; i < REF_CONTEXTS; i++)
vp10_cond_prob_diff_update(&header_bc, &fc->comp_ref_prob[i],
- counts->comp_ref[i]);
+ counts->comp_ref[i]);
for (i = 0; i < BLOCK_SIZE_GROUPS; ++i)
prob_diff_update(vp10_intra_mode_tree, cm->fc->y_mode_prob[i],
@@ -1475,7 +1430,7 @@
#endif
vp10_write_nmv_probs(cm, cm->allow_high_precision_mv, &header_bc,
- &counts->mv);
+ &counts->mv);
update_ext_tx_probs(cm, &header_bc);
}
@@ -1486,8 +1441,8 @@
}
#if CONFIG_MISC_FIXES
-static int remux_tiles(uint8_t *dest, const int sz,
- const int n_tiles, const int mag) {
+static int remux_tiles(uint8_t *dest, const int sz, const int n_tiles,
+ const int mag) {
int rpos = 0, wpos = 0, n;
for (n = 0; n < n_tiles; n++) {
@@ -1499,18 +1454,11 @@
tile_sz = mem_get_le32(&dest[rpos]) + 1;
rpos += 4;
switch (mag) {
- case 0:
- dest[wpos] = tile_sz - 1;
- break;
- case 1:
- mem_put_le16(&dest[wpos], tile_sz - 1);
- break;
- case 2:
- mem_put_le24(&dest[wpos], tile_sz - 1);
- break;
+ case 0: dest[wpos] = tile_sz - 1; break;
+ case 1: mem_put_le16(&dest[wpos], tile_sz - 1); break;
+ case 2: mem_put_le24(&dest[wpos], tile_sz - 1); break;
case 3: // remuxing should only happen if mag < 3
- default:
- assert("Invalid value for tile size magnitude" && 0);
+ default: assert("Invalid value for tile size magnitude" && 0);
}
wpos += mag + 1;
}
@@ -1530,7 +1478,7 @@
void vp10_pack_bitstream(VP10_COMP *const cpi, uint8_t *dest, size_t *size) {
uint8_t *data = dest;
size_t first_part_size, uncompressed_hdr_size, data_sz;
- struct vpx_write_bit_buffer wb = {data, 0};
+ struct vpx_write_bit_buffer wb = { data, 0 };
struct vpx_write_bit_buffer saved_wb;
unsigned int max_tile;
#if CONFIG_MISC_FIXES
@@ -1563,8 +1511,7 @@
// Choose the (tile size) magnitude
for (mag = 0, mask = 0xff; mag < 4; mag++) {
- if (max_tile <= mask)
- break;
+ if (max_tile <= mask) break;
mask <<= 8;
mask |= 0xff;
}
diff --git a/vp10/encoder/bitstream.h b/vp10/encoder/bitstream.h
index b1da89f..ab692e8 100644
--- a/vp10/encoder/bitstream.h
+++ b/vp10/encoder/bitstream.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_BITSTREAM_H_
#define VP10_ENCODER_BITSTREAM_H_
diff --git a/vp10/encoder/block.h b/vp10/encoder/block.h
index ab0252b..3d35485 100644
--- a/vp10/encoder/block.h
+++ b/vp10/encoder/block.h
@@ -45,7 +45,7 @@
/* The [2] dimension is for whether we skip the EOB node (i.e. if previous
* coefficient in this block was zero) or not. */
typedef unsigned int vp10_coeff_cost[PLANE_TYPES][REF_TYPES][COEF_BANDS][2]
- [COEFF_CONTEXTS][ENTROPY_TOKENS];
+ [COEFF_CONTEXTS][ENTROPY_TOKENS];
typedef struct {
int_mv ref_mvs[MAX_REF_FRAMES][MAX_MV_REF_CANDIDATES];
@@ -70,8 +70,8 @@
int rddiv;
int rdmult;
int mb_energy;
- int * m_search_count_ptr;
- int * ex_search_count_ptr;
+ int *m_search_count_ptr;
+ int *ex_search_count_ptr;
// These are set to their default values at the beginning, and then adjusted
// further in the encoding process.
@@ -122,9 +122,9 @@
// skip forward transform and quantization
uint8_t skip_txfm[MAX_MB_PLANE << 2];
- #define SKIP_TXFM_NONE 0
- #define SKIP_TXFM_AC_DC 1
- #define SKIP_TXFM_AC_ONLY 2
+#define SKIP_TXFM_NONE 0
+#define SKIP_TXFM_AC_DC 1
+#define SKIP_TXFM_AC_ONLY 2
int64_t bsse[MAX_MB_PLANE << 2];
diff --git a/vp10/encoder/blockiness.c b/vp10/encoder/blockiness.c
index ede13e0..c57e4ef 100644
--- a/vp10/encoder/blockiness.c
+++ b/vp10/encoder/blockiness.c
@@ -70,9 +70,9 @@
s_blockiness += horizontal_filter(s);
r_blockiness += horizontal_filter(r);
sum_0 += s[0];
- sum_sq_0 += s[0]*s[0];
+ sum_sq_0 += s[0] * s[0];
sum_1 += s[-1];
- sum_sq_1 += s[-1]*s[-1];
+ sum_sq_1 += s[-1] * s[-1];
}
var_0 = variance(sum_0, sum_sq_0, size);
var_1 = variance(sum_1, sum_sq_1, size);
@@ -120,19 +120,19 @@
// This function returns the blockiness for the entire frame currently by
// looking at all borders in steps of 4.
double vp10_get_blockiness(const unsigned char *img1, int img1_pitch,
- const unsigned char *img2, int img2_pitch,
- int width, int height ) {
+ const unsigned char *img2, int img2_pitch, int width,
+ int height) {
double blockiness = 0;
int i, j;
vpx_clear_system_state();
- for (i = 0; i < height; i += 4, img1 += img1_pitch * 4,
- img2 += img2_pitch * 4) {
+ for (i = 0; i < height;
+ i += 4, img1 += img1_pitch * 4, img2 += img2_pitch * 4) {
for (j = 0; j < width; j += 4) {
if (i > 0 && i < height && j > 0 && j < width) {
- blockiness += blockiness_vertical(img1 + j, img1_pitch,
- img2 + j, img2_pitch, 4);
- blockiness += blockiness_horizontal(img1 + j, img1_pitch,
- img2 + j, img2_pitch, 4);
+ blockiness +=
+ blockiness_vertical(img1 + j, img1_pitch, img2 + j, img2_pitch, 4);
+ blockiness += blockiness_horizontal(img1 + j, img1_pitch, img2 + j,
+ img2_pitch, 4);
}
}
}
diff --git a/vp10/encoder/context_tree.c b/vp10/encoder/context_tree.c
index 6c056d2..07ae254 100644
--- a/vp10/encoder/context_tree.c
+++ b/vp10/encoder/context_tree.c
@@ -12,10 +12,7 @@
#include "vp10/encoder/encoder.h"
static const BLOCK_SIZE square[] = {
- BLOCK_8X8,
- BLOCK_16X16,
- BLOCK_32X32,
- BLOCK_64X64,
+ BLOCK_8X8, BLOCK_16X16, BLOCK_32X32, BLOCK_64X64,
};
static void alloc_mode_context(VP10_COMMON *cm, int num_4x4_blk,
@@ -25,8 +22,7 @@
int i, k;
ctx->num_4x4_blk = num_blk;
- CHECK_MEM_ERROR(cm, ctx->zcoeff_blk,
- vpx_calloc(num_blk, sizeof(uint8_t)));
+ CHECK_MEM_ERROR(cm, ctx->zcoeff_blk, vpx_calloc(num_blk, sizeof(uint8_t)));
for (i = 0; i < MAX_MB_PLANE; ++i) {
for (k = 0; k < 3; ++k) {
CHECK_MEM_ERROR(cm, ctx->coeff[i][k],
@@ -37,10 +33,10 @@
vpx_memalign(32, num_pix * sizeof(*ctx->dqcoeff[i][k])));
CHECK_MEM_ERROR(cm, ctx->eobs[i][k],
vpx_memalign(32, num_blk * sizeof(*ctx->eobs[i][k])));
- ctx->coeff_pbuf[i][k] = ctx->coeff[i][k];
- ctx->qcoeff_pbuf[i][k] = ctx->qcoeff[i][k];
+ ctx->coeff_pbuf[i][k] = ctx->coeff[i][k];
+ ctx->qcoeff_pbuf[i][k] = ctx->qcoeff[i][k];
ctx->dqcoeff_pbuf[i][k] = ctx->dqcoeff[i][k];
- ctx->eobs_pbuf[i][k] = ctx->eobs[i][k];
+ ctx->eobs_pbuf[i][k] = ctx->eobs[i][k];
}
}
}
@@ -71,12 +67,12 @@
static void alloc_tree_contexts(VP10_COMMON *cm, PC_TREE *tree,
int num_4x4_blk) {
alloc_mode_context(cm, num_4x4_blk, &tree->none);
- alloc_mode_context(cm, num_4x4_blk/2, &tree->horizontal[0]);
- alloc_mode_context(cm, num_4x4_blk/2, &tree->vertical[0]);
+ alloc_mode_context(cm, num_4x4_blk / 2, &tree->horizontal[0]);
+ alloc_mode_context(cm, num_4x4_blk / 2, &tree->vertical[0]);
if (num_4x4_blk > 4) {
- alloc_mode_context(cm, num_4x4_blk/2, &tree->horizontal[1]);
- alloc_mode_context(cm, num_4x4_blk/2, &tree->vertical[1]);
+ alloc_mode_context(cm, num_4x4_blk / 2, &tree->horizontal[1]);
+ alloc_mode_context(cm, num_4x4_blk / 2, &tree->vertical[1]);
} else {
memset(&tree->horizontal[1], 0, sizeof(tree->horizontal[1]));
memset(&tree->vertical[1], 0, sizeof(tree->vertical[1]));
@@ -106,19 +102,18 @@
int nodes;
vpx_free(td->leaf_tree);
- CHECK_MEM_ERROR(cm, td->leaf_tree, vpx_calloc(leaf_nodes,
- sizeof(*td->leaf_tree)));
+ CHECK_MEM_ERROR(cm, td->leaf_tree,
+ vpx_calloc(leaf_nodes, sizeof(*td->leaf_tree)));
vpx_free(td->pc_tree);
- CHECK_MEM_ERROR(cm, td->pc_tree, vpx_calloc(tree_nodes,
- sizeof(*td->pc_tree)));
+ CHECK_MEM_ERROR(cm, td->pc_tree,
+ vpx_calloc(tree_nodes, sizeof(*td->pc_tree)));
this_pc = &td->pc_tree[0];
this_leaf = &td->leaf_tree[0];
// 4x4 blocks smaller than 8x8 but in the same 8x8 block share the same
// context so we only need to allocate 1 for each 8x8 block.
- for (i = 0; i < leaf_nodes; ++i)
- alloc_mode_context(cm, 1, &td->leaf_tree[i]);
+ for (i = 0; i < leaf_nodes; ++i) alloc_mode_context(cm, 1, &td->leaf_tree[i]);
// Sets up all the leaf nodes in the tree.
for (pc_tree_index = 0; pc_tree_index < leaf_nodes; ++pc_tree_index) {
@@ -126,8 +121,7 @@
tree->block_size = square[0];
alloc_tree_contexts(cm, tree, 4);
tree->leaf_split[0] = this_leaf++;
- for (j = 1; j < 4; j++)
- tree->leaf_split[j] = tree->leaf_split[0];
+ for (j = 1; j < 4; j++) tree->leaf_split[j] = tree->leaf_split[0];
}
// Each node has 4 leaf nodes, fill each block_size level of the tree
@@ -137,8 +131,7 @@
PC_TREE *const tree = &td->pc_tree[pc_tree_index];
alloc_tree_contexts(cm, tree, 4 << (2 * square_index));
tree->block_size = square[square_index];
- for (j = 0; j < 4; j++)
- tree->split[j] = this_pc++;
+ for (j = 0; j < 4; j++) tree->split[j] = this_pc++;
++pc_tree_index;
}
++square_index;
@@ -152,12 +145,10 @@
int i;
// Set up all 4x4 mode contexts
- for (i = 0; i < 64; ++i)
- free_mode_context(&td->leaf_tree[i]);
+ for (i = 0; i < 64; ++i) free_mode_context(&td->leaf_tree[i]);
// Sets up all the leaf nodes in the tree.
- for (i = 0; i < tree_nodes; ++i)
- free_tree_contexts(&td->pc_tree[i]);
+ for (i = 0; i < tree_nodes; ++i) free_tree_contexts(&td->pc_tree[i]);
vpx_free(td->pc_tree);
td->pc_tree = NULL;
diff --git a/vp10/encoder/cost.c b/vp10/encoder/cost.c
index aab8263..a4cd43a 100644
--- a/vp10/encoder/cost.c
+++ b/vp10/encoder/cost.c
@@ -12,31 +12,30 @@
#include "vp10/encoder/cost.h"
const unsigned int vp10_prob_cost[256] = {
- 2047, 2047, 1791, 1641, 1535, 1452, 1385, 1328, 1279, 1235, 1196, 1161,
- 1129, 1099, 1072, 1046, 1023, 1000, 979, 959, 940, 922, 905, 889,
- 873, 858, 843, 829, 816, 803, 790, 778, 767, 755, 744, 733,
- 723, 713, 703, 693, 684, 675, 666, 657, 649, 641, 633, 625,
- 617, 609, 602, 594, 587, 580, 573, 567, 560, 553, 547, 541,
- 534, 528, 522, 516, 511, 505, 499, 494, 488, 483, 477, 472,
- 467, 462, 457, 452, 447, 442, 437, 433, 428, 424, 419, 415,
- 410, 406, 401, 397, 393, 389, 385, 381, 377, 373, 369, 365,
- 361, 357, 353, 349, 346, 342, 338, 335, 331, 328, 324, 321,
- 317, 314, 311, 307, 304, 301, 297, 294, 291, 288, 285, 281,
- 278, 275, 272, 269, 266, 263, 260, 257, 255, 252, 249, 246,
- 243, 240, 238, 235, 232, 229, 227, 224, 221, 219, 216, 214,
- 211, 208, 206, 203, 201, 198, 196, 194, 191, 189, 186, 184,
- 181, 179, 177, 174, 172, 170, 168, 165, 163, 161, 159, 156,
- 154, 152, 150, 148, 145, 143, 141, 139, 137, 135, 133, 131,
- 129, 127, 125, 123, 121, 119, 117, 115, 113, 111, 109, 107,
- 105, 103, 101, 99, 97, 95, 93, 92, 90, 88, 86, 84,
- 82, 81, 79, 77, 75, 73, 72, 70, 68, 66, 65, 63,
- 61, 60, 58, 56, 55, 53, 51, 50, 48, 46, 45, 43,
- 41, 40, 38, 37, 35, 33, 32, 30, 29, 27, 25, 24,
- 22, 21, 19, 18, 16, 15, 13, 12, 10, 9, 7, 6,
- 4, 3, 1, 1};
+ 2047, 2047, 1791, 1641, 1535, 1452, 1385, 1328, 1279, 1235, 1196, 1161, 1129,
+ 1099, 1072, 1046, 1023, 1000, 979, 959, 940, 922, 905, 889, 873, 858,
+ 843, 829, 816, 803, 790, 778, 767, 755, 744, 733, 723, 713, 703,
+ 693, 684, 675, 666, 657, 649, 641, 633, 625, 617, 609, 602, 594,
+ 587, 580, 573, 567, 560, 553, 547, 541, 534, 528, 522, 516, 511,
+ 505, 499, 494, 488, 483, 477, 472, 467, 462, 457, 452, 447, 442,
+ 437, 433, 428, 424, 419, 415, 410, 406, 401, 397, 393, 389, 385,
+ 381, 377, 373, 369, 365, 361, 357, 353, 349, 346, 342, 338, 335,
+ 331, 328, 324, 321, 317, 314, 311, 307, 304, 301, 297, 294, 291,
+ 288, 285, 281, 278, 275, 272, 269, 266, 263, 260, 257, 255, 252,
+ 249, 246, 243, 240, 238, 235, 232, 229, 227, 224, 221, 219, 216,
+ 214, 211, 208, 206, 203, 201, 198, 196, 194, 191, 189, 186, 184,
+ 181, 179, 177, 174, 172, 170, 168, 165, 163, 161, 159, 156, 154,
+ 152, 150, 148, 145, 143, 141, 139, 137, 135, 133, 131, 129, 127,
+ 125, 123, 121, 119, 117, 115, 113, 111, 109, 107, 105, 103, 101,
+ 99, 97, 95, 93, 92, 90, 88, 86, 84, 82, 81, 79, 77,
+ 75, 73, 72, 70, 68, 66, 65, 63, 61, 60, 58, 56, 55,
+ 53, 51, 50, 48, 46, 45, 43, 41, 40, 38, 37, 35, 33,
+ 32, 30, 29, 27, 25, 24, 22, 21, 19, 18, 16, 15, 13,
+ 12, 10, 9, 7, 6, 4, 3, 1, 1
+};
-static void cost(int *costs, vpx_tree tree, const vpx_prob *probs,
- int i, int c) {
+static void cost(int *costs, vpx_tree tree, const vpx_prob *probs, int i,
+ int c) {
const vpx_prob prob = probs[i / 2];
int b;
diff --git a/vp10/encoder/cost.h b/vp10/encoder/cost.h
index b9619c6..702f0a4 100644
--- a/vp10/encoder/cost.h
+++ b/vp10/encoder/cost.h
@@ -23,16 +23,16 @@
#define vp10_cost_one(prob) vp10_cost_zero(vpx_complement(prob))
-#define vp10_cost_bit(prob, bit) vp10_cost_zero((bit) ? vpx_complement(prob) \
- : (prob))
+#define vp10_cost_bit(prob, bit) \
+ vp10_cost_zero((bit) ? vpx_complement(prob) : (prob))
static INLINE unsigned int cost_branch256(const unsigned int ct[2],
vpx_prob p) {
return ct[0] * vp10_cost_zero(p) + ct[1] * vp10_cost_one(p);
}
-static INLINE int treed_cost(vpx_tree tree, const vpx_prob *probs,
- int bits, int len) {
+static INLINE int treed_cost(vpx_tree tree, const vpx_prob *probs, int bits,
+ int len) {
int cost = 0;
vpx_tree_index i = 0;
diff --git a/vp10/encoder/dct.c b/vp10/encoder/dct.c
index 6fb7870..2a7ba7e 100644
--- a/vp10/encoder/dct.c
+++ b/vp10/encoder/dct.c
@@ -165,11 +165,11 @@
input[3] = in[3] + in[12];
input[4] = in[4] + in[11];
input[5] = in[5] + in[10];
- input[6] = in[6] + in[ 9];
- input[7] = in[7] + in[ 8];
+ input[6] = in[6] + in[9];
+ input[7] = in[7] + in[8];
- step1[0] = in[7] - in[ 8];
- step1[1] = in[6] - in[ 9];
+ step1[0] = in[7] - in[8];
+ step1[1] = in[6] - in[9];
step1[2] = in[5] - in[10];
step1[3] = in[4] - in[11];
step1[4] = in[3] - in[12];
@@ -292,7 +292,6 @@
out[15] = (tran_low_t)fdct_round_shift(temp2);
}
-
/* TODO(angiebird): Unify this with vp10_fwd_txfm.c: vp10_fdct32
static void fdct32(const tran_low_t *input, tran_low_t *output) {
tran_high_t temp;
@@ -746,14 +745,14 @@
tran_high_t x7 = input[6];
// stage 1
- s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
- s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
+ s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
+ s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
s2 = cospi_10_64 * x2 + cospi_22_64 * x3;
s3 = cospi_22_64 * x2 - cospi_10_64 * x3;
s4 = cospi_18_64 * x4 + cospi_14_64 * x5;
s5 = cospi_14_64 * x4 - cospi_18_64 * x5;
- s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
- s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
+ s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
+ s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
x0 = fdct_round_shift(s0 + s4);
x1 = fdct_round_shift(s1 + s5);
@@ -769,10 +768,10 @@
s1 = x1;
s2 = x2;
s3 = x3;
- s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
- s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
- s6 = - cospi_24_64 * x6 + cospi_8_64 * x7;
- s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
+ s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
+ s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
+ s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
+ s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
x0 = s0 + s2;
x1 = s1 + s3;
@@ -826,11 +825,11 @@
tran_high_t x15 = input[14];
// stage 1
- s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
+ s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
- s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
+ s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
- s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
+ s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
@@ -839,9 +838,9 @@
s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
- s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
+ s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
- s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
+ s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
x0 = fdct_round_shift(s0 + s8);
x1 = fdct_round_shift(s1 + s9);
@@ -851,8 +850,8 @@
x5 = fdct_round_shift(s5 + s13);
x6 = fdct_round_shift(s6 + s14);
x7 = fdct_round_shift(s7 + s15);
- x8 = fdct_round_shift(s0 - s8);
- x9 = fdct_round_shift(s1 - s9);
+ x8 = fdct_round_shift(s0 - s8);
+ x9 = fdct_round_shift(s1 - s9);
x10 = fdct_round_shift(s2 - s10);
x11 = fdct_round_shift(s3 - s11);
x12 = fdct_round_shift(s4 - s12);
@@ -869,14 +868,14 @@
s5 = x5;
s6 = x6;
s7 = x7;
- s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
- s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
- s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
- s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
- s12 = - x12 * cospi_28_64 + x13 * cospi_4_64;
- s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
- s14 = - x14 * cospi_12_64 + x15 * cospi_20_64;
- s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
+ s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
+ s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
+ s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
+ s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
+ s12 = -x12 * cospi_28_64 + x13 * cospi_4_64;
+ s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
+ s14 = -x14 * cospi_12_64 + x15 * cospi_20_64;
+ s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
x0 = s0 + s4;
x1 = s1 + s5;
@@ -900,18 +899,18 @@
s1 = x1;
s2 = x2;
s3 = x3;
- s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
+ s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
s5 = x4 * cospi_24_64 - x5 * cospi_8_64;
- s6 = - x6 * cospi_24_64 + x7 * cospi_8_64;
- s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
+ s6 = -x6 * cospi_24_64 + x7 * cospi_8_64;
+ s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
s8 = x8;
s9 = x9;
s10 = x10;
s11 = x11;
- s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
+ s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
s13 = x12 * cospi_24_64 - x13 * cospi_8_64;
- s14 = - x14 * cospi_24_64 + x15 * cospi_8_64;
- s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
+ s14 = -x14 * cospi_24_64 + x15 * cospi_8_64;
+ s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
x0 = s0 + s2;
x1 = s1 + s3;
@@ -931,13 +930,13 @@
x15 = fdct_round_shift(s13 - s15);
// stage 4
- s2 = (- cospi_16_64) * (x2 + x3);
+ s2 = (-cospi_16_64) * (x2 + x3);
s3 = cospi_16_64 * (x2 - x3);
s6 = cospi_16_64 * (x6 + x7);
- s7 = cospi_16_64 * (- x6 + x7);
+ s7 = cospi_16_64 * (-x6 + x7);
s10 = cospi_16_64 * (x10 + x11);
- s11 = cospi_16_64 * (- x10 + x11);
- s14 = (- cospi_16_64) * (x14 + x15);
+ s11 = cospi_16_64 * (-x10 + x11);
+ s14 = (-cospi_16_64) * (x14 + x15);
s15 = cospi_16_64 * (x14 - x15);
x2 = fdct_round_shift(s2);
@@ -968,28 +967,28 @@
}
static const transform_2d FHT_4[] = {
- { fdct4, fdct4 }, // DCT_DCT = 0
- { fadst4, fdct4 }, // ADST_DCT = 1
- { fdct4, fadst4 }, // DCT_ADST = 2
- { fadst4, fadst4 } // ADST_ADST = 3
+ { fdct4, fdct4 }, // DCT_DCT = 0
+ { fadst4, fdct4 }, // ADST_DCT = 1
+ { fdct4, fadst4 }, // DCT_ADST = 2
+ { fadst4, fadst4 } // ADST_ADST = 3
};
static const transform_2d FHT_8[] = {
- { fdct8, fdct8 }, // DCT_DCT = 0
- { fadst8, fdct8 }, // ADST_DCT = 1
- { fdct8, fadst8 }, // DCT_ADST = 2
- { fadst8, fadst8 } // ADST_ADST = 3
+ { fdct8, fdct8 }, // DCT_DCT = 0
+ { fadst8, fdct8 }, // ADST_DCT = 1
+ { fdct8, fadst8 }, // DCT_ADST = 2
+ { fadst8, fadst8 } // ADST_ADST = 3
};
static const transform_2d FHT_16[] = {
- { fdct16, fdct16 }, // DCT_DCT = 0
- { fadst16, fdct16 }, // ADST_DCT = 1
- { fdct16, fadst16 }, // DCT_ADST = 2
- { fadst16, fadst16 } // ADST_ADST = 3
+ { fdct16, fdct16 }, // DCT_DCT = 0
+ { fadst16, fdct16 }, // ADST_DCT = 1
+ { fdct16, fadst16 }, // DCT_ADST = 2
+ { fadst16, fadst16 } // ADST_ADST = 3
};
-void vp10_fht4x4_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp10_fht4x4_c(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
if (tx_type == DCT_DCT) {
vpx_fdct4x4_c(input, output, stride);
} else {
@@ -1000,36 +999,29 @@
// Columns
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- temp_in[j] = input[j * stride + i] * 16;
- if (i == 0 && temp_in[0])
- temp_in[0] += 1;
+ for (j = 0; j < 4; ++j) temp_in[j] = input[j * stride + i] * 16;
+ if (i == 0 && temp_in[0]) temp_in[0] += 1;
ht.cols(temp_in, temp_out);
- for (j = 0; j < 4; ++j)
- out[j * 4 + i] = temp_out[j];
+ for (j = 0; j < 4; ++j) out[j * 4 + i] = temp_out[j];
}
// Rows
for (i = 0; i < 4; ++i) {
- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j + i * 4];
+ for (j = 0; j < 4; ++j) temp_in[j] = out[j + i * 4];
ht.rows(temp_in, temp_out);
- for (j = 0; j < 4; ++j)
- output[j + i * 4] = (temp_out[j] + 1) >> 2;
+ for (j = 0; j < 4; ++j) output[j + i * 4] = (temp_out[j] + 1) >> 2;
}
}
}
void vp10_fdct8x8_quant_c(const int16_t *input, int stride,
- tran_low_t *coeff_ptr, intptr_t n_coeffs,
- int skip_block,
- const int16_t *zbin_ptr, const int16_t *round_ptr,
- const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr,
- uint16_t *eob_ptr,
- const int16_t *scan, const int16_t *iscan) {
+ tran_low_t *coeff_ptr, intptr_t n_coeffs,
+ int skip_block, const int16_t *zbin_ptr,
+ const int16_t *round_ptr, const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr,
+ tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
+ const int16_t *dequant_ptr, uint16_t *eob_ptr,
+ const int16_t *scan, const int16_t *iscan) {
int eob = -1;
int i, j;
@@ -1061,8 +1053,8 @@
x3 = s0 - s3;
t0 = (x0 + x1) * cospi_16_64;
t1 = (x0 - x1) * cospi_16_64;
- t2 = x2 * cospi_24_64 + x3 * cospi_8_64;
- t3 = -x2 * cospi_8_64 + x3 * cospi_24_64;
+ t2 = x2 * cospi_24_64 + x3 * cospi_8_64;
+ t3 = -x2 * cospi_8_64 + x3 * cospi_24_64;
output[0 * 8] = (tran_low_t)fdct_round_shift(t0);
output[2 * 8] = (tran_low_t)fdct_round_shift(t2);
output[4 * 8] = (tran_low_t)fdct_round_shift(t1);
@@ -1081,10 +1073,10 @@
x3 = s7 + t3;
// stage 4
- t0 = x0 * cospi_28_64 + x3 * cospi_4_64;
- t1 = x1 * cospi_12_64 + x2 * cospi_20_64;
+ t0 = x0 * cospi_28_64 + x3 * cospi_4_64;
+ t1 = x1 * cospi_12_64 + x2 * cospi_20_64;
t2 = x2 * cospi_12_64 + x1 * -cospi_20_64;
- t3 = x3 * cospi_28_64 + x0 * -cospi_4_64;
+ t3 = x3 * cospi_28_64 + x0 * -cospi_4_64;
output[1 * 8] = (tran_low_t)fdct_round_shift(t0);
output[3 * 8] = (tran_low_t)fdct_round_shift(t2);
output[5 * 8] = (tran_low_t)fdct_round_shift(t1);
@@ -1097,8 +1089,7 @@
// Rows
for (i = 0; i < 8; ++i) {
fdct8(&intermediate[i * 8], &coeff_ptr[i * 8]);
- for (j = 0; j < 8; ++j)
- coeff_ptr[j + i * 8] /= 2;
+ for (j = 0; j < 8; ++j) coeff_ptr[j + i * 8] /= 2;
}
// TODO(jingning) Decide the need of these arguments after the
@@ -1125,15 +1116,14 @@
qcoeff_ptr[rc] = (tmp ^ coeff_sign) - coeff_sign;
dqcoeff_ptr[rc] = qcoeff_ptr[rc] * dequant_ptr[rc != 0];
- if (tmp)
- eob = i;
+ if (tmp) eob = i;
}
}
*eob_ptr = eob + 1;
}
-void vp10_fht8x8_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp10_fht8x8_c(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
if (tx_type == DCT_DCT) {
vpx_fdct8x8_c(input, output, stride);
} else {
@@ -1144,17 +1134,14 @@
// Columns
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = input[j * stride + i] * 4;
+ for (j = 0; j < 8; ++j) temp_in[j] = input[j * stride + i] * 4;
ht.cols(temp_in, temp_out);
- for (j = 0; j < 8; ++j)
- out[j * 8 + i] = temp_out[j];
+ for (j = 0; j < 8; ++j) out[j * 8 + i] = temp_out[j];
}
// Rows
for (i = 0; i < 8; ++i) {
- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j + i * 8];
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j + i * 8];
ht.rows(temp_in, temp_out);
for (j = 0; j < 8; ++j)
output[j + i * 8] = (temp_out[j] + (temp_out[j] < 0)) >> 1;
@@ -1218,8 +1205,8 @@
}
}
-void vp10_fht16x16_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp10_fht16x16_c(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
if (tx_type == DCT_DCT) {
vpx_fdct16x16_c(input, output, stride);
} else {
@@ -1230,8 +1217,7 @@
// Columns
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = input[j * stride + i] * 4;
+ for (j = 0; j < 16; ++j) temp_in[j] = input[j * stride + i] * 4;
ht.cols(temp_in, temp_out);
for (j = 0; j < 16; ++j)
out[j * 16 + i] = (temp_out[j] + 1 + (temp_out[j] < 0)) >> 2;
@@ -1239,33 +1225,31 @@
// Rows
for (i = 0; i < 16; ++i) {
- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j + i * 16];
+ for (j = 0; j < 16; ++j) temp_in[j] = out[j + i * 16];
ht.rows(temp_in, temp_out);
- for (j = 0; j < 16; ++j)
- output[j + i * 16] = temp_out[j];
+ for (j = 0; j < 16; ++j) output[j + i * 16] = temp_out[j];
}
}
}
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_highbd_fht4x4_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp10_highbd_fht4x4_c(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
vp10_fht4x4_c(input, output, stride, tx_type);
}
-void vp10_highbd_fht8x8_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp10_highbd_fht8x8_c(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
vp10_fht8x8_c(input, output, stride, tx_type);
}
void vp10_highbd_fwht4x4_c(const int16_t *input, tran_low_t *output,
- int stride) {
+ int stride) {
vp10_fwht4x4_c(input, output, stride);
}
void vp10_highbd_fht16x16_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+ int stride, int tx_type) {
vp10_fht16x16_c(input, output, stride, tx_type);
}
#endif // CONFIG_VPX_HIGHBITDEPTH
diff --git a/vp10/encoder/encodeframe.c b/vp10/encoder/encodeframe.c
index 99c8d79..f9fa104 100644
--- a/vp10/encoder/encodeframe.c
+++ b/vp10/encoder/encodeframe.c
@@ -46,90 +46,86 @@
#include "vp10/encoder/segmentation.h"
#include "vp10/encoder/tokenize.h"
-static void encode_superblock(VP10_COMP *cpi, ThreadData * td,
- TOKENEXTRA **t, int output_enabled,
- int mi_row, int mi_col, BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx);
+static void encode_superblock(VP10_COMP *cpi, ThreadData *td, TOKENEXTRA **t,
+ int output_enabled, int mi_row, int mi_col,
+ BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx);
// This is used as a reference when computing the source variance for the
// purposes of activity masking.
// Eventually this should be replaced by custom no-reference routines,
// which will be faster.
static const uint8_t VP9_VAR_OFFS[64] = {
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128
};
#if CONFIG_VPX_HIGHBITDEPTH
static const uint16_t VP9_HIGH_VAR_OFFS_8[64] = {
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128
};
static const uint16_t VP9_HIGH_VAR_OFFS_10[64] = {
- 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
- 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
- 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
- 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
- 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
- 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
- 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
- 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4
+ 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4,
+ 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4,
+ 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4,
+ 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4,
+ 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4,
+ 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4,
+ 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4,
+ 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4, 128 * 4
};
static const uint16_t VP9_HIGH_VAR_OFFS_12[64] = {
- 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
- 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
- 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
- 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
- 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
- 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
- 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
- 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16, 128 * 16,
+ 128 * 16
};
#endif // CONFIG_VPX_HIGHBITDEPTH
unsigned int vp10_get_sby_perpixel_variance(VP10_COMP *cpi,
- const struct buf_2d *ref,
- BLOCK_SIZE bs) {
+ const struct buf_2d *ref,
+ BLOCK_SIZE bs) {
unsigned int sse;
- const unsigned int var = cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
- VP9_VAR_OFFS, 0, &sse);
+ const unsigned int var =
+ cpi->fn_ptr[bs].vf(ref->buf, ref->stride, VP9_VAR_OFFS, 0, &sse);
return ROUND_POWER_OF_TWO(var, num_pels_log2_lookup[bs]);
}
#if CONFIG_VPX_HIGHBITDEPTH
-unsigned int vp10_high_get_sby_perpixel_variance(
- VP10_COMP *cpi, const struct buf_2d *ref, BLOCK_SIZE bs, int bd) {
+unsigned int vp10_high_get_sby_perpixel_variance(VP10_COMP *cpi,
+ const struct buf_2d *ref,
+ BLOCK_SIZE bs, int bd) {
unsigned int var, sse;
switch (bd) {
case 10:
- var = cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
- CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_10),
- 0, &sse);
+ var =
+ cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
+ CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_10), 0, &sse);
break;
case 12:
- var = cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
- CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_12),
- 0, &sse);
+ var =
+ cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
+ CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_12), 0, &sse);
break;
case 8:
default:
- var = cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
- CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_8),
- 0, &sse);
+ var =
+ cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
+ CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_8), 0, &sse);
break;
}
return ROUND_POWER_OF_TWO(var, num_pels_log2_lookup[bs]);
@@ -152,12 +148,10 @@
}
static BLOCK_SIZE get_rd_var_based_fixed_partition(VP10_COMP *cpi,
- MACROBLOCK *x,
- int mi_row,
+ MACROBLOCK *x, int mi_row,
int mi_col) {
- unsigned int var = get_sby_perpixel_diff_variance(cpi, &x->plane[0].src,
- mi_row, mi_col,
- BLOCK_64X64);
+ unsigned int var = get_sby_perpixel_diff_variance(
+ cpi, &x->plane[0].src, mi_row, mi_col, BLOCK_64X64);
if (var < 8)
return BLOCK_64X64;
else if (var < 128)
@@ -172,8 +166,7 @@
// pointers.
static INLINE void set_mode_info_offsets(VP10_COMP *const cpi,
MACROBLOCK *const x,
- MACROBLOCKD *const xd,
- int mi_row,
+ MACROBLOCKD *const xd, int mi_row,
int mi_col) {
VP10_COMMON *const cm = &cpi->common;
const int idx_str = xd->mi_stride * mi_row + mi_col;
@@ -210,8 +203,8 @@
// Set up distance of MB to edge of frame in 1/8th pel units.
assert(!(mi_col & (mi_width - 1)) && !(mi_row & (mi_height - 1)));
- set_mi_row_col(xd, tile, mi_row, mi_height, mi_col, mi_width,
- cm->mi_rows, cm->mi_cols);
+ set_mi_row_col(xd, tile, mi_row, mi_height, mi_col, mi_width, cm->mi_rows,
+ cm->mi_cols);
// Set up source buffers.
vp10_setup_src_planes(x, cpi->Source, mi_row, mi_col);
@@ -223,8 +216,8 @@
// Setup segment ID.
if (seg->enabled) {
if (cpi->oxcf.aq_mode != VARIANCE_AQ) {
- const uint8_t *const map = seg->update_map ? cpi->segmentation_map
- : cm->last_frame_seg_map;
+ const uint8_t *const map =
+ seg->update_map ? cpi->segmentation_map : cm->last_frame_seg_map;
mbmi->segment_id = get_segment_id(cm, map, bsize, mi_row, mi_col);
}
vp10_init_plane_quantizers(cpi, x);
@@ -239,10 +232,8 @@
xd->tile = *tile;
}
-static void set_block_size(VP10_COMP * const cpi,
- MACROBLOCK *const x,
- MACROBLOCKD *const xd,
- int mi_row, int mi_col,
+static void set_block_size(VP10_COMP *const cpi, MACROBLOCK *const x,
+ MACROBLOCKD *const xd, int mi_row, int mi_col,
BLOCK_SIZE bsize) {
if (cpi->common.mi_cols > mi_col && cpi->common.mi_rows > mi_row) {
set_mode_info_offsets(cpi, x, xd, mi_row, mi_col);
@@ -304,38 +295,37 @@
node->part_variances = NULL;
switch (bsize) {
case BLOCK_64X64: {
- v64x64 *vt = (v64x64 *) data;
+ v64x64 *vt = (v64x64 *)data;
node->part_variances = &vt->part_variances;
for (i = 0; i < 4; i++)
node->split[i] = &vt->split[i].part_variances.none;
break;
}
case BLOCK_32X32: {
- v32x32 *vt = (v32x32 *) data;
+ v32x32 *vt = (v32x32 *)data;
node->part_variances = &vt->part_variances;
for (i = 0; i < 4; i++)
node->split[i] = &vt->split[i].part_variances.none;
break;
}
case BLOCK_16X16: {
- v16x16 *vt = (v16x16 *) data;
+ v16x16 *vt = (v16x16 *)data;
node->part_variances = &vt->part_variances;
for (i = 0; i < 4; i++)
node->split[i] = &vt->split[i].part_variances.none;
break;
}
case BLOCK_8X8: {
- v8x8 *vt = (v8x8 *) data;
+ v8x8 *vt = (v8x8 *)data;
node->part_variances = &vt->part_variances;
for (i = 0; i < 4; i++)
node->split[i] = &vt->split[i].part_variances.none;
break;
}
case BLOCK_4X4: {
- v4x4 *vt = (v4x4 *) data;
+ v4x4 *vt = (v4x4 *)data;
node->part_variances = &vt->part_variances;
- for (i = 0; i < 4; i++)
- node->split[i] = &vt->split[i];
+ for (i = 0; i < 4; i++) node->split[i] = &vt->split[i];
break;
}
default: {
@@ -353,8 +343,10 @@
}
static void get_variance(var *v) {
- v->variance = (int)(256 * (v->sum_square_error -
- ((v->sum_error * v->sum_error) >> v->log2_count)) >> v->log2_count);
+ v->variance =
+ (int)(256 * (v->sum_square_error -
+ ((v->sum_error * v->sum_error) >> v->log2_count)) >>
+ v->log2_count);
}
static void sum_2_variances(const var *a, const var *b, var *r) {
@@ -375,17 +367,12 @@
&node.part_variances->none);
}
-static int set_vt_partitioning(VP10_COMP *cpi,
- MACROBLOCK *const x,
- MACROBLOCKD *const xd,
- void *data,
- BLOCK_SIZE bsize,
- int mi_row,
- int mi_col,
- int64_t threshold,
- BLOCK_SIZE bsize_min,
+static int set_vt_partitioning(VP10_COMP *cpi, MACROBLOCK *const x,
+ MACROBLOCKD *const xd, void *data,
+ BLOCK_SIZE bsize, int mi_row, int mi_col,
+ int64_t threshold, BLOCK_SIZE bsize_min,
int force_split) {
- VP10_COMMON * const cm = &cpi->common;
+ VP10_COMMON *const cm = &cpi->common;
variance_node vt;
const int block_width = num_8x8_blocks_wide_lookup[bsize];
const int block_height = num_8x8_blocks_high_lookup[bsize];
@@ -394,8 +381,7 @@
assert(block_height == block_width);
tree_to_node(data, bsize, &vt);
- if (force_split == 1)
- return 0;
+ if (force_split == 1) return 0;
// For bsize=bsize_min (16x16/8x8 for 8x8/4x4 downsampling), select if
// variance is below threshold, otherwise split will be selected.
@@ -418,7 +404,7 @@
// For key frame: take split for bsize above 32X32 or very high variance.
if (cm->frame_type == KEY_FRAME &&
(bsize > BLOCK_32X32 ||
- vt.part_variances->none.variance > (threshold << 4))) {
+ vt.part_variances->none.variance > (threshold << 4))) {
return 0;
}
// If variance is low, take the bsize (no split).
@@ -469,8 +455,8 @@
VP10_COMMON *const cm = &cpi->common;
const int is_key_frame = (cm->frame_type == KEY_FRAME);
const int threshold_multiplier = is_key_frame ? 20 : 1;
- const int64_t threshold_base = (int64_t)(threshold_multiplier *
- cpi->y_dequant[q][1]);
+ const int64_t threshold_base =
+ (int64_t)(threshold_multiplier * cpi->y_dequant[q][1]);
if (is_key_frame) {
thresholds[0] = threshold_base;
thresholds[1] = threshold_base >> 2;
@@ -508,8 +494,9 @@
if (cm->width <= 352 && cm->height <= 288)
cpi->vbp_threshold_sad = 100;
else
- cpi->vbp_threshold_sad = (cpi->y_dequant[q][1] << 1) > 1000 ?
- (cpi->y_dequant[q][1] << 1) : 1000;
+ cpi->vbp_threshold_sad = (cpi->y_dequant[q][1] << 1) > 1000
+ ? (cpi->y_dequant[q][1] << 1)
+ : 1000;
cpi->vbp_bsize_min = BLOCK_16X16;
}
cpi->vbp_threshold_minmax = 15 + (q >> 3);
@@ -522,8 +509,7 @@
#if CONFIG_VPX_HIGHBITDEPTH
int highbd_flag,
#endif
- int pixels_wide,
- int pixels_high) {
+ int pixels_wide, int pixels_high) {
int k;
int minmax_max = 0;
int minmax_min = 255;
@@ -537,22 +523,17 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (highbd_flag & YV12_FLAG_HIGHBITDEPTH) {
vpx_highbd_minmax_8x8(s + y8_idx * sp + x8_idx, sp,
- d + y8_idx * dp + x8_idx, dp,
- &min, &max);
+ d + y8_idx * dp + x8_idx, dp, &min, &max);
} else {
- vpx_minmax_8x8(s + y8_idx * sp + x8_idx, sp,
- d + y8_idx * dp + x8_idx, dp,
- &min, &max);
+ vpx_minmax_8x8(s + y8_idx * sp + x8_idx, sp, d + y8_idx * dp + x8_idx,
+ dp, &min, &max);
}
#else
- vpx_minmax_8x8(s + y8_idx * sp + x8_idx, sp,
- d + y8_idx * dp + x8_idx, dp,
+ vpx_minmax_8x8(s + y8_idx * sp + x8_idx, sp, d + y8_idx * dp + x8_idx, dp,
&min, &max);
#endif
- if ((max - min) > minmax_max)
- minmax_max = (max - min);
- if ((max - min) < minmax_min)
- minmax_min = (max - min);
+ if ((max - min) > minmax_max) minmax_max = (max - min);
+ if ((max - min) < minmax_min) minmax_min = (max - min);
}
}
return (minmax_max - minmax_min);
@@ -563,8 +544,7 @@
#if CONFIG_VPX_HIGHBITDEPTH
int highbd_flag,
#endif
- int pixels_wide,
- int pixels_high,
+ int pixels_wide, int pixels_high,
int is_key_frame) {
int k;
for (k = 0; k < 4; k++) {
@@ -582,13 +562,11 @@
d_avg = vpx_highbd_avg_4x4(d + y4_idx * dp + x4_idx, dp);
} else {
s_avg = vpx_avg_4x4(s + y4_idx * sp + x4_idx, sp);
- if (!is_key_frame)
- d_avg = vpx_avg_4x4(d + y4_idx * dp + x4_idx, dp);
+ if (!is_key_frame) d_avg = vpx_avg_4x4(d + y4_idx * dp + x4_idx, dp);
}
#else
s_avg = vpx_avg_4x4(s + y4_idx * sp + x4_idx, sp);
- if (!is_key_frame)
- d_avg = vpx_avg_4x4(d + y4_idx * dp + x4_idx, dp);
+ if (!is_key_frame) d_avg = vpx_avg_4x4(d + y4_idx * dp + x4_idx, dp);
#endif
sum = s_avg - d_avg;
sse = sum * sum;
@@ -602,8 +580,7 @@
#if CONFIG_VPX_HIGHBITDEPTH
int highbd_flag,
#endif
- int pixels_wide,
- int pixels_high,
+ int pixels_wide, int pixels_high,
int is_key_frame) {
int k;
for (k = 0; k < 4; k++) {
@@ -621,13 +598,11 @@
d_avg = vpx_highbd_avg_8x8(d + y8_idx * dp + x8_idx, dp);
} else {
s_avg = vpx_avg_8x8(s + y8_idx * sp + x8_idx, sp);
- if (!is_key_frame)
- d_avg = vpx_avg_8x8(d + y8_idx * dp + x8_idx, dp);
+ if (!is_key_frame) d_avg = vpx_avg_8x8(d + y8_idx * dp + x8_idx, dp);
}
#else
s_avg = vpx_avg_8x8(s + y8_idx * sp + x8_idx, sp);
- if (!is_key_frame)
- d_avg = vpx_avg_8x8(d + y8_idx * dp + x8_idx, dp);
+ if (!is_key_frame) d_avg = vpx_avg_8x8(d + y8_idx * dp + x8_idx, dp);
#endif
sum = s_avg - d_avg;
sse = sum * sum;
@@ -638,11 +613,9 @@
// This function chooses partitioning based on the variance between source and
// reconstructed last, where variance is computed for down-sampled inputs.
-static int choose_partitioning(VP10_COMP *cpi,
- const TileInfo *const tile,
- MACROBLOCK *x,
- int mi_row, int mi_col) {
- VP10_COMMON * const cm = &cpi->common;
+static int choose_partitioning(VP10_COMP *cpi, const TileInfo *const tile,
+ MACROBLOCK *x, int mi_row, int mi_col) {
+ VP10_COMMON *const cm = &cpi->common;
MACROBLOCKD *xd = &x->e_mbd;
int i, j, k, m;
v64x64 vt;
@@ -653,8 +626,8 @@
int sp;
int dp;
int pixels_wide = 64, pixels_high = 64;
- int64_t thresholds[4] = {cpi->vbp_thresholds[0], cpi->vbp_thresholds[1],
- cpi->vbp_thresholds[2], cpi->vbp_thresholds[3]};
+ int64_t thresholds[4] = { cpi->vbp_thresholds[0], cpi->vbp_thresholds[1],
+ cpi->vbp_thresholds[2], cpi->vbp_thresholds[3] };
// Always use 4x4 partition for key frame.
const int is_key_frame = (cm->frame_type == KEY_FRAME);
@@ -664,8 +637,8 @@
int segment_id = CR_SEGMENT_ID_BASE;
if (cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ && cm->seg.enabled) {
- const uint8_t *const map = cm->seg.update_map ? cpi->segmentation_map :
- cm->last_frame_seg_map;
+ const uint8_t *const map =
+ cm->seg.update_map ? cpi->segmentation_map : cm->last_frame_seg_map;
segment_id = get_segment_id(cm, map, BLOCK_64X64, mi_row, mi_col);
if (cyclic_refresh_segment_id_boosted(segment_id)) {
@@ -676,10 +649,8 @@
set_offsets(cpi, tile, x, mi_row, mi_col, BLOCK_64X64);
- if (xd->mb_to_right_edge < 0)
- pixels_wide += (xd->mb_to_right_edge >> 3);
- if (xd->mb_to_bottom_edge < 0)
- pixels_high += (xd->mb_to_bottom_edge >> 3);
+ if (xd->mb_to_right_edge < 0) pixels_wide += (xd->mb_to_right_edge >> 3);
+ if (xd->mb_to_bottom_edge < 0) pixels_high += (xd->mb_to_bottom_edge >> 3);
s = x->plane[0].src.buf;
sp = x->plane[0].src.stride;
@@ -691,25 +662,24 @@
const YV12_BUFFER_CONFIG *yv12_g = NULL;
unsigned int y_sad, y_sad_g;
- const BLOCK_SIZE bsize = BLOCK_32X32
- + (mi_col + 4 < cm->mi_cols) * 2 + (mi_row + 4 < cm->mi_rows);
+ const BLOCK_SIZE bsize = BLOCK_32X32 + (mi_col + 4 < cm->mi_cols) * 2 +
+ (mi_row + 4 < cm->mi_rows);
assert(yv12 != NULL);
yv12_g = get_ref_frame_buffer(cpi, GOLDEN_FRAME);
if (yv12_g && yv12_g != yv12) {
vp10_setup_pre_planes(xd, 0, yv12_g, mi_row, mi_col,
- &cm->frame_refs[GOLDEN_FRAME - 1].sf);
- y_sad_g = cpi->fn_ptr[bsize].sdf(x->plane[0].src.buf,
- x->plane[0].src.stride,
- xd->plane[0].pre[0].buf,
- xd->plane[0].pre[0].stride);
+ &cm->frame_refs[GOLDEN_FRAME - 1].sf);
+ y_sad_g = cpi->fn_ptr[bsize].sdf(
+ x->plane[0].src.buf, x->plane[0].src.stride, xd->plane[0].pre[0].buf,
+ xd->plane[0].pre[0].stride);
} else {
y_sad_g = UINT_MAX;
}
vp10_setup_pre_planes(xd, 0, yv12, mi_row, mi_col,
- &cm->frame_refs[LAST_FRAME - 1].sf);
+ &cm->frame_refs[LAST_FRAME - 1].sf);
mbmi->ref_frame[0] = LAST_FRAME;
mbmi->ref_frame[1] = NONE;
mbmi->sb_type = BLOCK_64X64;
@@ -719,7 +689,7 @@
y_sad = vp10_int_pro_motion_estimation(cpi, x, bsize, mi_row, mi_col);
if (y_sad_g < y_sad) {
vp10_setup_pre_planes(xd, 0, yv12_g, mi_row, mi_col,
- &cm->frame_refs[GOLDEN_FRAME - 1].sf);
+ &cm->frame_refs[GOLDEN_FRAME - 1].sf);
mbmi->ref_frame[0] = GOLDEN_FRAME;
mbmi->mv[0].as_int = 0;
y_sad = y_sad_g;
@@ -730,15 +700,15 @@
vp10_build_inter_predictors_sb(xd, mi_row, mi_col, BLOCK_64X64);
for (i = 1; i <= 2; ++i) {
- struct macroblock_plane *p = &x->plane[i];
+ struct macroblock_plane *p = &x->plane[i];
struct macroblockd_plane *pd = &xd->plane[i];
const BLOCK_SIZE bs = get_plane_block_size(bsize, pd);
if (bs == BLOCK_INVALID)
uv_sad = UINT_MAX;
else
- uv_sad = cpi->fn_ptr[bs].sdf(p->src.buf, p->src.stride,
- pd->dst.buf, pd->dst.stride);
+ uv_sad = cpi->fn_ptr[bs].sdf(p->src.buf, p->src.stride, pd->dst.buf,
+ pd->dst.stride);
x->color_sensitivity[i - 1] = uv_sad > (y_sad >> 2);
}
@@ -748,8 +718,7 @@
// If the y_sad is very small, take 64x64 as partition and exit.
// Don't check on boosted segment for now, as 64x64 is suppressed there.
- if (segment_id == CR_SEGMENT_ID_BASE &&
- y_sad < cpi->vbp_threshold_sad) {
+ if (segment_id == CR_SEGMENT_ID_BASE && y_sad < cpi->vbp_threshold_sad) {
const int block_width = num_8x8_blocks_wide_lookup[BLOCK_64X64];
const int block_height = num_8x8_blocks_high_lookup[BLOCK_64X64];
if (mi_col + block_width / 2 < cm->mi_cols &&
@@ -764,16 +733,10 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
switch (xd->bd) {
- case 10:
- d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_10);
- break;
- case 12:
- d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_12);
- break;
+ case 10: d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_10); break;
+ case 12: d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_12); break;
case 8:
- default:
- d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_8);
- break;
+ default: d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_8); break;
}
}
#endif // CONFIG_VPX_HIGHBITDEPTH
@@ -799,22 +762,19 @@
if (!is_key_frame) {
fill_variance_8x8avg(s, sp, d, dp, x16_idx, y16_idx, vst,
#if CONFIG_VPX_HIGHBITDEPTH
- xd->cur_buf->flags,
+ xd->cur_buf->flags,
#endif
- pixels_wide,
- pixels_high,
- is_key_frame);
+ pixels_wide, pixels_high, is_key_frame);
fill_variance_tree(&vt.split[i].split[j], BLOCK_16X16);
get_variance(&vt.split[i].split[j].part_variances.none);
- if (vt.split[i].split[j].part_variances.none.variance >
- thresholds[2]) {
+ if (vt.split[i].split[j].part_variances.none.variance > thresholds[2]) {
// 16X16 variance is above threshold for split, so force split to 8x8
// for this 16x16 block (this also forces splits for upper levels).
force_split[split_index] = 1;
force_split[i + 1] = 1;
force_split[0] = 1;
} else if (vt.split[i].split[j].part_variances.none.variance >
- thresholds[1] &&
+ thresholds[1] &&
!cyclic_refresh_segment_id_boosted(segment_id)) {
// We have some nominal amount of 16x16 variance (based on average),
// compute the minmax over the 8x8 sub-blocks, and if above threshold,
@@ -832,23 +792,20 @@
}
}
if (is_key_frame || (low_res &&
- vt.split[i].split[j].part_variances.none.variance >
- (thresholds[1] << 1))) {
+ vt.split[i].split[j].part_variances.none.variance >
+ (thresholds[1] << 1))) {
force_split[split_index] = 0;
// Go down to 4x4 down-sampling for variance.
variance4x4downsample[i2 + j] = 1;
for (k = 0; k < 4; k++) {
int x8_idx = x16_idx + ((k & 1) << 3);
int y8_idx = y16_idx + ((k >> 1) << 3);
- v8x8 *vst2 = is_key_frame ? &vst->split[k] :
- &vt2[i2 + j].split[k];
+ v8x8 *vst2 = is_key_frame ? &vst->split[k] : &vt2[i2 + j].split[k];
fill_variance_4x4avg(s, sp, d, dp, x8_idx, y8_idx, vst2,
#if CONFIG_VPX_HIGHBITDEPTH
xd->cur_buf->flags,
#endif
- pixels_wide,
- pixels_high,
- is_key_frame);
+ pixels_wide, pixels_high, is_key_frame);
}
}
}
@@ -859,10 +816,8 @@
const int i2 = i << 2;
for (j = 0; j < 4; j++) {
if (variance4x4downsample[i2 + j] == 1) {
- v16x16 *vtemp = (!is_key_frame) ? &vt2[i2 + j] :
- &vt.split[i].split[j];
- for (m = 0; m < 4; m++)
- fill_variance_tree(&vtemp->split[m], BLOCK_8X8);
+ v16x16 *vtemp = (!is_key_frame) ? &vt2[i2 + j] : &vt.split[i].split[j];
+ for (m = 0; m < 4; m++) fill_variance_tree(&vtemp->split[m], BLOCK_8X8);
fill_variance_tree(vtemp, BLOCK_16X16);
}
}
@@ -884,7 +839,7 @@
// Now go through the entire structure, splitting every block size until
// we get to one that's got a variance lower than our threshold.
- if ( mi_col + 8 > cm->mi_cols || mi_row + 8 > cm->mi_rows ||
+ if (mi_col + 8 > cm->mi_cols || mi_row + 8 > cm->mi_rows ||
!set_vt_partitioning(cpi, x, xd, &vt, BLOCK_64X64, mi_row, mi_col,
thresholds[0], BLOCK_16X16, force_split[0])) {
for (i = 0; i < 4; ++i) {
@@ -901,15 +856,13 @@
// For inter frames: if variance4x4downsample[] == 1 for this 16x16
// block, then the variance is based on 4x4 down-sampling, so use vt2
// in set_vt_partioning(), otherwise use vt.
- v16x16 *vtemp = (!is_key_frame &&
- variance4x4downsample[i2 + j] == 1) ?
- &vt2[i2 + j] : &vt.split[i].split[j];
- if (!set_vt_partitioning(cpi, x, xd, vtemp, BLOCK_16X16,
- mi_row + y32_idx + y16_idx,
- mi_col + x32_idx + x16_idx,
- thresholds[2],
- cpi->vbp_bsize_min,
- force_split[5 + i2 + j])) {
+ v16x16 *vtemp = (!is_key_frame && variance4x4downsample[i2 + j] == 1)
+ ? &vt2[i2 + j]
+ : &vt.split[i].split[j];
+ if (!set_vt_partitioning(
+ cpi, x, xd, vtemp, BLOCK_16X16, mi_row + y32_idx + y16_idx,
+ mi_col + x32_idx + x16_idx, thresholds[2], cpi->vbp_bsize_min,
+ force_split[5 + i2 + j])) {
for (k = 0; k < 4; ++k) {
const int x8_idx = (k & 1);
const int y8_idx = (k >> 1);
@@ -919,16 +872,14 @@
mi_row + y32_idx + y16_idx + y8_idx,
mi_col + x32_idx + x16_idx + x8_idx,
thresholds[3], BLOCK_8X8, 0)) {
- set_block_size(cpi, x, xd,
- (mi_row + y32_idx + y16_idx + y8_idx),
- (mi_col + x32_idx + x16_idx + x8_idx),
- BLOCK_4X4);
+ set_block_size(
+ cpi, x, xd, (mi_row + y32_idx + y16_idx + y8_idx),
+ (mi_col + x32_idx + x16_idx + x8_idx), BLOCK_4X4);
}
} else {
- set_block_size(cpi, x, xd,
- (mi_row + y32_idx + y16_idx + y8_idx),
- (mi_col + x32_idx + x16_idx + x8_idx),
- BLOCK_8X8);
+ set_block_size(
+ cpi, x, xd, (mi_row + y32_idx + y16_idx + y8_idx),
+ (mi_col + x32_idx + x16_idx + x8_idx), BLOCK_8X8);
}
}
}
@@ -939,8 +890,7 @@
return 0;
}
-static void update_state(VP10_COMP *cpi, ThreadData *td,
- PICK_MODE_CONTEXT *ctx,
+static void update_state(VP10_COMP *cpi, ThreadData *td, PICK_MODE_CONTEXT *ctx,
int mi_row, int mi_col, BLOCK_SIZE bsize,
int output_enabled) {
int i, x_idx, y;
@@ -958,8 +908,7 @@
const int bh = num_8x8_blocks_high_lookup[mi->mbmi.sb_type];
const int x_mis = VPXMIN(bw, cm->mi_cols - mi_col);
const int y_mis = VPXMIN(bh, cm->mi_rows - mi_row);
- MV_REF *const frame_mvs =
- cm->cur_frame->mvs + mi_row * cm->mi_cols + mi_col;
+ MV_REF *const frame_mvs = cm->cur_frame->mvs + mi_row * cm->mi_cols + mi_col;
int w, h;
const int mis = cm->mi_stride;
@@ -976,17 +925,15 @@
if (seg->enabled) {
// For in frame complexity AQ copy the segment id from the segment map.
if (cpi->oxcf.aq_mode == COMPLEXITY_AQ) {
- const uint8_t *const map = seg->update_map ? cpi->segmentation_map
- : cm->last_frame_seg_map;
- mi_addr->mbmi.segment_id =
- get_segment_id(cm, map, bsize, mi_row, mi_col);
+ const uint8_t *const map =
+ seg->update_map ? cpi->segmentation_map : cm->last_frame_seg_map;
+ mi_addr->mbmi.segment_id = get_segment_id(cm, map, bsize, mi_row, mi_col);
}
// Else for cyclic refresh mode update the segment map, set the segment id
// and then update the quantizer.
if (cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ) {
- vp10_cyclic_refresh_update_segment(cpi, &xd->mi[0]->mbmi, mi_row,
- mi_col, bsize, ctx->rate, ctx->dist,
- x->skip);
+ vp10_cyclic_refresh_update_segment(cpi, &xd->mi[0]->mbmi, mi_row, mi_col,
+ bsize, ctx->rate, ctx->dist, x->skip);
}
}
@@ -1005,20 +952,18 @@
p[i].eobs = ctx->eobs_pbuf[i][2];
}
- for (i = 0; i < 2; ++i)
- pd[i].color_index_map = ctx->color_index_map[i];
+ for (i = 0; i < 2; ++i) pd[i].color_index_map = ctx->color_index_map[i];
// Restore the coding context of the MB to that that was in place
// when the mode was picked for it
for (y = 0; y < mi_height; y++)
for (x_idx = 0; x_idx < mi_width; x_idx++)
- if ((xd->mb_to_right_edge >> (3 + MI_SIZE_LOG2)) + mi_width > x_idx
- && (xd->mb_to_bottom_edge >> (3 + MI_SIZE_LOG2)) + mi_height > y) {
+ if ((xd->mb_to_right_edge >> (3 + MI_SIZE_LOG2)) + mi_width > x_idx &&
+ (xd->mb_to_bottom_edge >> (3 + MI_SIZE_LOG2)) + mi_height > y) {
xd->mi[x_idx + y * mis] = mi_addr;
}
- if (cpi->oxcf.aq_mode)
- vp10_init_plane_quantizers(cpi, x);
+ if (cpi->oxcf.aq_mode) vp10_init_plane_quantizers(cpi, x);
if (is_inter_block(mbmi) && mbmi->sb_type < BLOCK_8X8) {
mbmi->mv[0].as_int = mi->bmi[3].as_mv[0].as_int;
@@ -1029,22 +974,16 @@
memcpy(x->zcoeff_blk[mbmi->tx_size], ctx->zcoeff_blk,
sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
- if (!output_enabled)
- return;
+ if (!output_enabled) return;
#if CONFIG_INTERNAL_STATS
if (frame_is_intra_only(cm)) {
static const int kf_mode_index[] = {
- THR_DC /*DC_PRED*/,
- THR_V_PRED /*V_PRED*/,
- THR_H_PRED /*H_PRED*/,
- THR_D45_PRED /*D45_PRED*/,
- THR_D135_PRED /*D135_PRED*/,
- THR_D117_PRED /*D117_PRED*/,
- THR_D153_PRED /*D153_PRED*/,
- THR_D207_PRED /*D207_PRED*/,
- THR_D63_PRED /*D63_PRED*/,
- THR_TM /*TM_PRED*/,
+ THR_DC /*DC_PRED*/, THR_V_PRED /*V_PRED*/,
+ THR_H_PRED /*H_PRED*/, THR_D45_PRED /*D45_PRED*/,
+ THR_D135_PRED /*D135_PRED*/, THR_D117_PRED /*D117_PRED*/,
+ THR_D153_PRED /*D153_PRED*/, THR_D207_PRED /*D207_PRED*/,
+ THR_D63_PRED /*D63_PRED*/, THR_TM /*TM_PRED*/,
};
++cpi->mode_chosen_counts[kf_mode_index[mbmi->mode]];
} else {
@@ -1083,9 +1022,9 @@
}
void vp10_setup_src_planes(MACROBLOCK *x, const YV12_BUFFER_CONFIG *src,
- int mi_row, int mi_col) {
- uint8_t *const buffers[3] = {src->y_buffer, src->u_buffer, src->v_buffer };
- const int strides[3] = {src->y_stride, src->uv_stride, src->uv_stride };
+ int mi_row, int mi_col) {
+ uint8_t *const buffers[3] = { src->y_buffer, src->u_buffer, src->v_buffer };
+ const int strides[3] = { src->y_stride, src->uv_stride, src->uv_stride };
int i;
// Set current frame pointer.
@@ -1097,24 +1036,20 @@
x->e_mbd.plane[i].subsampling_y);
}
-static int set_segment_rdmult(VP10_COMP *const cpi,
- MACROBLOCK *const x,
- int8_t segment_id) {
+static int set_segment_rdmult(VP10_COMP *const cpi, MACROBLOCK *const x,
+ int8_t segment_id) {
int segment_qindex;
VP10_COMMON *const cm = &cpi->common;
vp10_init_plane_quantizers(cpi, x);
vpx_clear_system_state();
- segment_qindex = vp10_get_qindex(&cm->seg, segment_id,
- cm->base_qindex);
+ segment_qindex = vp10_get_qindex(&cm->seg, segment_id, cm->base_qindex);
return vp10_compute_rd_mult(cpi, segment_qindex + cm->y_dc_delta_q);
}
-static void rd_pick_sb_modes(VP10_COMP *cpi,
- TileDataEnc *tile_data,
- MACROBLOCK *const x,
- int mi_row, int mi_col, RD_COST *rd_cost,
- BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
- int64_t best_rd) {
+static void rd_pick_sb_modes(VP10_COMP *cpi, TileDataEnc *tile_data,
+ MACROBLOCK *const x, int mi_row, int mi_col,
+ RD_COST *rd_cost, BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx, int64_t best_rd) {
VP10_COMMON *const cm = &cpi->common;
TileInfo *const tile_info = &tile_data->tile_info;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -1140,8 +1075,7 @@
p[i].eobs = ctx->eobs_pbuf[i][0];
}
- for (i = 0; i < 2; ++i)
- pd[i].color_index_map = ctx->color_index_map[i];
+ for (i = 0; i < 2; ++i) pd[i].color_index_map = ctx->color_index_map[i];
ctx->is_coded = 0;
ctx->skippable = 0;
@@ -1153,39 +1087,37 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- x->source_variance =
- vp10_high_get_sby_perpixel_variance(cpi, &x->plane[0].src,
- bsize, xd->bd);
+ x->source_variance = vp10_high_get_sby_perpixel_variance(
+ cpi, &x->plane[0].src, bsize, xd->bd);
} else {
x->source_variance =
- vp10_get_sby_perpixel_variance(cpi, &x->plane[0].src, bsize);
+ vp10_get_sby_perpixel_variance(cpi, &x->plane[0].src, bsize);
}
#else
x->source_variance =
- vp10_get_sby_perpixel_variance(cpi, &x->plane[0].src, bsize);
+ vp10_get_sby_perpixel_variance(cpi, &x->plane[0].src, bsize);
#endif // CONFIG_VPX_HIGHBITDEPTH
// Save rdmult before it might be changed, so it can be restored later.
orig_rdmult = x->rdmult;
if (aq_mode == VARIANCE_AQ) {
- const int energy = bsize <= BLOCK_16X16 ? x->mb_energy
- : vp10_block_energy(cpi, x, bsize);
- if (cm->frame_type == KEY_FRAME ||
- cpi->refresh_alt_ref_frame ||
+ const int energy =
+ bsize <= BLOCK_16X16 ? x->mb_energy : vp10_block_energy(cpi, x, bsize);
+ if (cm->frame_type == KEY_FRAME || cpi->refresh_alt_ref_frame ||
(cpi->refresh_golden_frame && !cpi->rc.is_src_frame_alt_ref)) {
mbmi->segment_id = vp10_vaq_segment_id(energy);
} else {
- const uint8_t *const map = cm->seg.update_map ? cpi->segmentation_map
- : cm->last_frame_seg_map;
+ const uint8_t *const map =
+ cm->seg.update_map ? cpi->segmentation_map : cm->last_frame_seg_map;
mbmi->segment_id = get_segment_id(cm, map, bsize, mi_row, mi_col);
}
x->rdmult = set_segment_rdmult(cpi, x, mbmi->segment_id);
} else if (aq_mode == COMPLEXITY_AQ) {
x->rdmult = set_segment_rdmult(cpi, x, mbmi->segment_id);
} else if (aq_mode == CYCLIC_REFRESH_AQ) {
- const uint8_t *const map = cm->seg.update_map ? cpi->segmentation_map
- : cm->last_frame_seg_map;
+ const uint8_t *const map =
+ cm->seg.update_map ? cpi->segmentation_map : cm->last_frame_seg_map;
// If segment is boosted, use rdmult for that segment.
if (cyclic_refresh_segment_id_boosted(
get_segment_id(cm, map, bsize, mi_row, mi_col)))
@@ -1200,22 +1132,20 @@
if (bsize >= BLOCK_8X8) {
if (segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP))
vp10_rd_pick_inter_mode_sb_seg_skip(cpi, tile_data, x, rd_cost, bsize,
- ctx, best_rd);
+ ctx, best_rd);
else
- vp10_rd_pick_inter_mode_sb(cpi, tile_data, x, mi_row, mi_col,
- rd_cost, bsize, ctx, best_rd);
+ vp10_rd_pick_inter_mode_sb(cpi, tile_data, x, mi_row, mi_col, rd_cost,
+ bsize, ctx, best_rd);
} else {
- vp10_rd_pick_inter_mode_sub8x8(cpi, tile_data, x, mi_row, mi_col,
- rd_cost, bsize, ctx, best_rd);
+ vp10_rd_pick_inter_mode_sub8x8(cpi, tile_data, x, mi_row, mi_col, rd_cost,
+ bsize, ctx, best_rd);
}
}
-
// Examine the resulting rate and for AQ mode 2 make a segment choice.
- if ((rd_cost->rate != INT_MAX) &&
- (aq_mode == COMPLEXITY_AQ) && (bsize >= BLOCK_16X16) &&
- (cm->frame_type == KEY_FRAME ||
- cpi->refresh_alt_ref_frame ||
+ if ((rd_cost->rate != INT_MAX) && (aq_mode == COMPLEXITY_AQ) &&
+ (bsize >= BLOCK_16X16) &&
+ (cm->frame_type == KEY_FRAME || cpi->refresh_alt_ref_frame ||
(cpi->refresh_golden_frame && !cpi->rc.is_src_frame_alt_ref))) {
vp10_caq_select_segment(cpi, x, bsize, mi_row, mi_col, rd_cost->rate);
}
@@ -1224,8 +1154,7 @@
// TODO(jingning) The rate-distortion optimization flow needs to be
// refactored to provide proper exit/return handle.
- if (rd_cost->rate == INT_MAX)
- rd_cost->rdcost = INT64_MAX;
+ if (rd_cost->rate == INT_MAX) rd_cost->rdcost = INT64_MAX;
ctx->rate = rd_cost->rate;
ctx->dist = rd_cost->dist;
@@ -1242,8 +1171,8 @@
if (!frame_is_intra_only(cm)) {
FRAME_COUNTS *const counts = td->counts;
const int inter_block = is_inter_block(mbmi);
- const int seg_ref_active = segfeature_active(&cm->seg, mbmi->segment_id,
- SEG_LVL_REF_FRAME);
+ const int seg_ref_active =
+ segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_REF_FRAME);
if (!seg_ref_active) {
counts->intra_inter[vp10_get_intra_inter_context(xd)][inter_block]++;
// If the segment reference feature is enabled we have only a single
@@ -1252,18 +1181,18 @@
if (inter_block) {
const MV_REFERENCE_FRAME ref0 = mbmi->ref_frame[0];
if (cm->reference_mode == REFERENCE_MODE_SELECT)
- counts->comp_inter[vp10_get_reference_mode_context(cm, xd)]
- [has_second_ref(mbmi)]++;
+ counts->comp_inter[vp10_get_reference_mode_context(
+ cm, xd)][has_second_ref(mbmi)]++;
if (has_second_ref(mbmi)) {
- counts->comp_ref[vp10_get_pred_context_comp_ref_p(cm, xd)]
- [ref0 == GOLDEN_FRAME]++;
+ counts->comp_ref[vp10_get_pred_context_comp_ref_p(
+ cm, xd)][ref0 == GOLDEN_FRAME]++;
} else {
- counts->single_ref[vp10_get_pred_context_single_ref_p1(xd)][0]
- [ref0 != LAST_FRAME]++;
+ counts->single_ref[vp10_get_pred_context_single_ref_p1(
+ xd)][0][ref0 != LAST_FRAME]++;
if (ref0 != LAST_FRAME)
- counts->single_ref[vp10_get_pred_context_single_ref_p2(xd)][1]
- [ref0 != GOLDEN_FRAME]++;
+ counts->single_ref[vp10_get_pred_context_single_ref_p2(
+ xd)][1][ref0 != GOLDEN_FRAME]++;
}
}
}
@@ -1301,17 +1230,15 @@
int mi_width = num_8x8_blocks_wide_lookup[bsize];
int mi_height = num_8x8_blocks_high_lookup[bsize];
for (p = 0; p < MAX_MB_PLANE; p++) {
- memcpy(
- xd->above_context[p] + ((mi_col * 2) >> xd->plane[p].subsampling_x),
- a + num_4x4_blocks_wide * p,
- (sizeof(ENTROPY_CONTEXT) * num_4x4_blocks_wide) >>
- xd->plane[p].subsampling_x);
- memcpy(
- xd->left_context[p]
- + ((mi_row & MI_MASK) * 2 >> xd->plane[p].subsampling_y),
- l + num_4x4_blocks_high * p,
- (sizeof(ENTROPY_CONTEXT) * num_4x4_blocks_high) >>
- xd->plane[p].subsampling_y);
+ memcpy(xd->above_context[p] + ((mi_col * 2) >> xd->plane[p].subsampling_x),
+ a + num_4x4_blocks_wide * p,
+ (sizeof(ENTROPY_CONTEXT) * num_4x4_blocks_wide) >>
+ xd->plane[p].subsampling_x);
+ memcpy(xd->left_context[p] +
+ ((mi_row & MI_MASK) * 2 >> xd->plane[p].subsampling_y),
+ l + num_4x4_blocks_high * p,
+ (sizeof(ENTROPY_CONTEXT) * num_4x4_blocks_high) >>
+ xd->plane[p].subsampling_y);
}
memcpy(xd->above_seg_context + mi_col, sa,
sizeof(*xd->above_seg_context) * mi_width);
@@ -1333,17 +1260,15 @@
// buffer the above/left context information of the block in search.
for (p = 0; p < MAX_MB_PLANE; ++p) {
- memcpy(
- a + num_4x4_blocks_wide * p,
- xd->above_context[p] + (mi_col * 2 >> xd->plane[p].subsampling_x),
- (sizeof(ENTROPY_CONTEXT) * num_4x4_blocks_wide) >>
- xd->plane[p].subsampling_x);
- memcpy(
- l + num_4x4_blocks_high * p,
- xd->left_context[p]
- + ((mi_row & MI_MASK) * 2 >> xd->plane[p].subsampling_y),
- (sizeof(ENTROPY_CONTEXT) * num_4x4_blocks_high) >>
- xd->plane[p].subsampling_y);
+ memcpy(a + num_4x4_blocks_wide * p,
+ xd->above_context[p] + (mi_col * 2 >> xd->plane[p].subsampling_x),
+ (sizeof(ENTROPY_CONTEXT) * num_4x4_blocks_wide) >>
+ xd->plane[p].subsampling_x);
+ memcpy(l + num_4x4_blocks_high * p,
+ xd->left_context[p] +
+ ((mi_row & MI_MASK) * 2 >> xd->plane[p].subsampling_y),
+ (sizeof(ENTROPY_CONTEXT) * num_4x4_blocks_high) >>
+ xd->plane[p].subsampling_y);
}
memcpy(sa, xd->above_seg_context + mi_col,
sizeof(*xd->above_seg_context) * mi_width);
@@ -1351,8 +1276,7 @@
sizeof(xd->left_seg_context[0]) * mi_height);
}
-static void encode_b(VP10_COMP *cpi, const TileInfo *const tile,
- ThreadData *td,
+static void encode_b(VP10_COMP *cpi, const TileInfo *const tile, ThreadData *td,
TOKENEXTRA **tp, int mi_row, int mi_col,
int output_enabled, BLOCK_SIZE bsize,
PICK_MODE_CONTEXT *ctx) {
@@ -1367,9 +1291,8 @@
}
static void encode_sb(VP10_COMP *cpi, ThreadData *td,
- const TileInfo *const tile,
- TOKENEXTRA **tp, int mi_row, int mi_col,
- int output_enabled, BLOCK_SIZE bsize,
+ const TileInfo *const tile, TOKENEXTRA **tp, int mi_row,
+ int mi_col, int output_enabled, BLOCK_SIZE bsize,
PC_TREE *pc_tree) {
VP10_COMMON *const cm = &cpi->common;
MACROBLOCK *const x = &td->mb;
@@ -1380,8 +1303,7 @@
PARTITION_TYPE partition;
BLOCK_SIZE subsize = bsize;
- if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
- return;
+ if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols) return;
if (bsize >= BLOCK_8X8) {
ctx = partition_plane_context(xd, mi_row, mi_col, bsize);
@@ -1431,9 +1353,7 @@
subsize, pc_tree->split[3]);
}
break;
- default:
- assert(0 && "Invalid partition type.");
- break;
+ default: assert(0 && "Invalid partition type."); break;
}
if (partition != PARTITION_SPLIT || bsize == BLOCK_8X8)
@@ -1443,9 +1363,8 @@
// Check to see if the given partition size is allowed for a specified number
// of 8x8 block rows and columns remaining in the image.
// If not then return the largest allowed partition size
-static BLOCK_SIZE find_partition_size(BLOCK_SIZE bsize,
- int rows_left, int cols_left,
- int *bh, int *bw) {
+static BLOCK_SIZE find_partition_size(BLOCK_SIZE bsize, int rows_left,
+ int cols_left, int *bh, int *bw) {
if (rows_left <= 0 || cols_left <= 0) {
return VPXMIN(bsize, BLOCK_8X8);
} else {
@@ -1460,9 +1379,10 @@
return bsize;
}
-static void set_partial_b64x64_partition(MODE_INFO *mi, int mis,
- int bh_in, int bw_in, int row8x8_remaining, int col8x8_remaining,
- BLOCK_SIZE bsize, MODE_INFO **mi_8x8) {
+static void set_partial_b64x64_partition(MODE_INFO *mi, int mis, int bh_in,
+ int bw_in, int row8x8_remaining,
+ int col8x8_remaining, BLOCK_SIZE bsize,
+ MODE_INFO **mi_8x8) {
int bh = bh_in;
int r, c;
for (r = 0; r < MI_BLOCK_SIZE; r += bh) {
@@ -1470,8 +1390,8 @@
for (c = 0; c < MI_BLOCK_SIZE; c += bw) {
const int index = r * mis + c;
mi_8x8[index] = mi + index;
- mi_8x8[index]->mbmi.sb_type = find_partition_size(bsize,
- row8x8_remaining - r, col8x8_remaining - c, &bh, &bw);
+ mi_8x8[index]->mbmi.sb_type = find_partition_size(
+ bsize, row8x8_remaining - r, col8x8_remaining - c, &bh, &bw);
}
}
}
@@ -1508,17 +1428,14 @@
} else {
// Else this is a partial SB64.
set_partial_b64x64_partition(mi_upper_left, mis, bh, bw, row8x8_remaining,
- col8x8_remaining, bsize, mi_8x8);
+ col8x8_remaining, bsize, mi_8x8);
}
}
-static void rd_use_partition(VP10_COMP *cpi,
- ThreadData *td,
- TileDataEnc *tile_data,
- MODE_INFO **mi_8x8, TOKENEXTRA **tp,
- int mi_row, int mi_col,
- BLOCK_SIZE bsize,
- int *rate, int64_t *dist,
+static void rd_use_partition(VP10_COMP *cpi, ThreadData *td,
+ TileDataEnc *tile_data, MODE_INFO **mi_8x8,
+ TOKENEXTRA **tp, int mi_row, int mi_col,
+ BLOCK_SIZE bsize, int *rate, int64_t *dist,
int do_recon, PC_TREE *pc_tree) {
VP10_COMMON *const cm = &cpi->common;
TileInfo *const tile_info = &tile_data->tile_info;
@@ -1540,8 +1457,7 @@
int do_partition_search = 1;
PICK_MODE_CONTEXT *ctx = &pc_tree->none;
- if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
- return;
+ if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols) return;
assert(num_4x4_blocks_wide_lookup[bsize] ==
num_4x4_blocks_high_lookup[bsize]);
@@ -1583,15 +1499,15 @@
mi_row + (mi_step >> 1) < cm->mi_rows &&
mi_col + (mi_step >> 1) < cm->mi_cols) {
pc_tree->partitioning = PARTITION_NONE;
- rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &none_rdc, bsize,
- ctx, INT64_MAX);
+ rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &none_rdc, bsize, ctx,
+ INT64_MAX);
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
if (none_rdc.rate < INT_MAX) {
none_rdc.rate += cpi->partition_cost[pl][PARTITION_NONE];
- none_rdc.rdcost = RDCOST(x->rdmult, x->rddiv, none_rdc.rate,
- none_rdc.dist);
+ none_rdc.rdcost =
+ RDCOST(x->rdmult, x->rddiv, none_rdc.rate, none_rdc.dist);
}
restore_context(x, mi_row, mi_col, a, l, sa, sl, bsize);
@@ -1602,23 +1518,21 @@
switch (partition) {
case PARTITION_NONE:
- rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &last_part_rdc,
- bsize, ctx, INT64_MAX);
+ rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &last_part_rdc, bsize,
+ ctx, INT64_MAX);
break;
case PARTITION_HORZ:
rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &last_part_rdc,
- subsize, &pc_tree->horizontal[0],
- INT64_MAX);
- if (last_part_rdc.rate != INT_MAX &&
- bsize >= BLOCK_8X8 && mi_row + (mi_step >> 1) < cm->mi_rows) {
+ subsize, &pc_tree->horizontal[0], INT64_MAX);
+ if (last_part_rdc.rate != INT_MAX && bsize >= BLOCK_8X8 &&
+ mi_row + (mi_step >> 1) < cm->mi_rows) {
RD_COST tmp_rdc;
PICK_MODE_CONTEXT *ctx = &pc_tree->horizontal[0];
vp10_rd_cost_init(&tmp_rdc);
update_state(cpi, td, ctx, mi_row, mi_col, subsize, 0);
encode_superblock(cpi, td, tp, 0, mi_row, mi_col, subsize, ctx);
- rd_pick_sb_modes(cpi, tile_data, x,
- mi_row + (mi_step >> 1), mi_col, &tmp_rdc,
- subsize, &pc_tree->horizontal[1], INT64_MAX);
+ rd_pick_sb_modes(cpi, tile_data, x, mi_row + (mi_step >> 1), mi_col,
+ &tmp_rdc, subsize, &pc_tree->horizontal[1], INT64_MAX);
if (tmp_rdc.rate == INT_MAX || tmp_rdc.dist == INT64_MAX) {
vp10_rd_cost_reset(&last_part_rdc);
break;
@@ -1631,17 +1545,16 @@
case PARTITION_VERT:
rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &last_part_rdc,
subsize, &pc_tree->vertical[0], INT64_MAX);
- if (last_part_rdc.rate != INT_MAX &&
- bsize >= BLOCK_8X8 && mi_col + (mi_step >> 1) < cm->mi_cols) {
+ if (last_part_rdc.rate != INT_MAX && bsize >= BLOCK_8X8 &&
+ mi_col + (mi_step >> 1) < cm->mi_cols) {
RD_COST tmp_rdc;
PICK_MODE_CONTEXT *ctx = &pc_tree->vertical[0];
vp10_rd_cost_init(&tmp_rdc);
update_state(cpi, td, ctx, mi_row, mi_col, subsize, 0);
encode_superblock(cpi, td, tp, 0, mi_row, mi_col, subsize, ctx);
- rd_pick_sb_modes(cpi, tile_data, x,
- mi_row, mi_col + (mi_step >> 1), &tmp_rdc,
- subsize, &pc_tree->vertical[bsize > BLOCK_8X8],
- INT64_MAX);
+ rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col + (mi_step >> 1),
+ &tmp_rdc, subsize,
+ &pc_tree->vertical[bsize > BLOCK_8X8], INT64_MAX);
if (tmp_rdc.rate == INT_MAX || tmp_rdc.dist == INT64_MAX) {
vp10_rd_cost_reset(&last_part_rdc);
break;
@@ -1669,11 +1582,10 @@
continue;
vp10_rd_cost_init(&tmp_rdc);
- rd_use_partition(cpi, td, tile_data,
- mi_8x8 + jj * bss * mis + ii * bss, tp,
- mi_row + y_idx, mi_col + x_idx, subsize,
- &tmp_rdc.rate, &tmp_rdc.dist,
- i != 3, pc_tree->split[i]);
+ rd_use_partition(cpi, td, tile_data, mi_8x8 + jj * bss * mis + ii * bss,
+ tp, mi_row + y_idx, mi_col + x_idx, subsize,
+ &tmp_rdc.rate, &tmp_rdc.dist, i != 3,
+ pc_tree->split[i]);
if (tmp_rdc.rate == INT_MAX || tmp_rdc.dist == INT64_MAX) {
vp10_rd_cost_reset(&last_part_rdc);
break;
@@ -1682,26 +1594,23 @@
last_part_rdc.dist += tmp_rdc.dist;
}
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
if (last_part_rdc.rate < INT_MAX) {
last_part_rdc.rate += cpi->partition_cost[pl][partition];
- last_part_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
- last_part_rdc.rate, last_part_rdc.dist);
+ last_part_rdc.rdcost =
+ RDCOST(x->rdmult, x->rddiv, last_part_rdc.rate, last_part_rdc.dist);
}
- if (do_partition_search
- && cpi->sf.adjust_partitioning_from_last_frame
- && cpi->sf.partition_search_type == SEARCH_PARTITION
- && partition != PARTITION_SPLIT && bsize > BLOCK_8X8
- && (mi_row + mi_step < cm->mi_rows ||
- mi_row + (mi_step >> 1) == cm->mi_rows)
- && (mi_col + mi_step < cm->mi_cols ||
- mi_col + (mi_step >> 1) == cm->mi_cols)) {
+ if (do_partition_search && cpi->sf.adjust_partitioning_from_last_frame &&
+ cpi->sf.partition_search_type == SEARCH_PARTITION &&
+ partition != PARTITION_SPLIT && bsize > BLOCK_8X8 &&
+ (mi_row + mi_step < cm->mi_rows ||
+ mi_row + (mi_step >> 1) == cm->mi_rows) &&
+ (mi_col + mi_step < cm->mi_cols ||
+ mi_col + (mi_step >> 1) == cm->mi_cols)) {
BLOCK_SIZE split_subsize = get_subsize(bsize, PARTITION_SPLIT);
chosen_rdc.rate = 0;
chosen_rdc.dist = 0;
@@ -1721,9 +1630,9 @@
save_context(x, mi_row, mi_col, a, l, sa, sl, bsize);
pc_tree->split[i]->partitioning = PARTITION_NONE;
- rd_pick_sb_modes(cpi, tile_data, x,
- mi_row + y_idx, mi_col + x_idx, &tmp_rdc,
- split_subsize, &pc_tree->split[i]->none, INT64_MAX);
+ rd_pick_sb_modes(cpi, tile_data, x, mi_row + y_idx, mi_col + x_idx,
+ &tmp_rdc, split_subsize, &pc_tree->split[i]->none,
+ INT64_MAX);
restore_context(x, mi_row, mi_col, a, l, sa, sl, bsize);
@@ -1736,7 +1645,7 @@
chosen_rdc.dist += tmp_rdc.dist;
if (i != 3)
- encode_sb(cpi, td, tile_info, tp, mi_row + y_idx, mi_col + x_idx, 0,
+ encode_sb(cpi, td, tile_info, tp, mi_row + y_idx, mi_col + x_idx, 0,
split_subsize, pc_tree->split[i]);
pl = partition_plane_context(xd, mi_row + y_idx, mi_col + x_idx,
@@ -1746,22 +1655,20 @@
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
if (chosen_rdc.rate < INT_MAX) {
chosen_rdc.rate += cpi->partition_cost[pl][PARTITION_SPLIT];
- chosen_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
- chosen_rdc.rate, chosen_rdc.dist);
+ chosen_rdc.rdcost =
+ RDCOST(x->rdmult, x->rddiv, chosen_rdc.rate, chosen_rdc.dist);
}
}
// If last_part is better set the partitioning to that.
if (last_part_rdc.rdcost < chosen_rdc.rdcost) {
mi_8x8[0]->mbmi.sb_type = bsize;
- if (bsize >= BLOCK_8X8)
- pc_tree->partitioning = partition;
+ if (bsize >= BLOCK_8X8) pc_tree->partitioning = partition;
chosen_rdc = last_part_rdc;
}
// If none was better set the partitioning to that.
if (none_rdc.rdcost < chosen_rdc.rdcost) {
- if (bsize >= BLOCK_8X8)
- pc_tree->partitioning = PARTITION_NONE;
+ if (bsize >= BLOCK_8X8) pc_tree->partitioning = PARTITION_NONE;
chosen_rdc = none_rdc;
}
@@ -1783,22 +1690,17 @@
}
static const BLOCK_SIZE min_partition_size[BLOCK_SIZES] = {
- BLOCK_4X4, BLOCK_4X4, BLOCK_4X4,
- BLOCK_4X4, BLOCK_4X4, BLOCK_4X4,
- BLOCK_8X8, BLOCK_8X8, BLOCK_8X8,
- BLOCK_16X16, BLOCK_16X16, BLOCK_16X16,
- BLOCK_16X16
+ BLOCK_4X4, BLOCK_4X4, BLOCK_4X4, BLOCK_4X4, BLOCK_4X4,
+ BLOCK_4X4, BLOCK_8X8, BLOCK_8X8, BLOCK_8X8, BLOCK_16X16,
+ BLOCK_16X16, BLOCK_16X16, BLOCK_16X16
};
static const BLOCK_SIZE max_partition_size[BLOCK_SIZES] = {
- BLOCK_8X8, BLOCK_16X16, BLOCK_16X16,
- BLOCK_16X16, BLOCK_32X32, BLOCK_32X32,
- BLOCK_32X32, BLOCK_64X64, BLOCK_64X64,
- BLOCK_64X64, BLOCK_64X64, BLOCK_64X64,
- BLOCK_64X64
+ BLOCK_8X8, BLOCK_16X16, BLOCK_16X16, BLOCK_16X16, BLOCK_32X32,
+ BLOCK_32X32, BLOCK_32X32, BLOCK_64X64, BLOCK_64X64, BLOCK_64X64,
+ BLOCK_64X64, BLOCK_64X64, BLOCK_64X64
};
-
// Look at all the mode_info entries for blocks that are part of this
// partition and find the min and max values for sb_type.
// At the moment this is designed to work on a 64x64 SB but could be
@@ -1811,14 +1713,14 @@
BLOCK_SIZE *max_block_size,
int bs_hist[BLOCK_SIZES]) {
int sb_width_in_blocks = MI_BLOCK_SIZE;
- int sb_height_in_blocks = MI_BLOCK_SIZE;
+ int sb_height_in_blocks = MI_BLOCK_SIZE;
int i, j;
int index = 0;
// Check the sb_type for each block that belongs to this region.
for (i = 0; i < sb_height_in_blocks; ++i) {
for (j = 0; j < sb_width_in_blocks; ++j) {
- MODE_INFO *mi = mi_8x8[index+j];
+ MODE_INFO *mi = mi_8x8[index + j];
BLOCK_SIZE sb_type = mi ? mi->mbmi.sb_type : 0;
bs_hist[sb_type]++;
*min_block_size = VPXMIN(*min_block_size, sb_type);
@@ -1830,19 +1732,16 @@
// Next square block size less or equal than current block size.
static const BLOCK_SIZE next_square_size[BLOCK_SIZES] = {
- BLOCK_4X4, BLOCK_4X4, BLOCK_4X4,
- BLOCK_8X8, BLOCK_8X8, BLOCK_8X8,
- BLOCK_16X16, BLOCK_16X16, BLOCK_16X16,
- BLOCK_32X32, BLOCK_32X32, BLOCK_32X32,
- BLOCK_64X64
+ BLOCK_4X4, BLOCK_4X4, BLOCK_4X4, BLOCK_8X8, BLOCK_8X8,
+ BLOCK_8X8, BLOCK_16X16, BLOCK_16X16, BLOCK_16X16, BLOCK_32X32,
+ BLOCK_32X32, BLOCK_32X32, BLOCK_64X64
};
// Look at neighboring blocks and set a min and max partition size based on
// what they chose.
static void rd_auto_partition_range(VP10_COMP *cpi, const TileInfo *const tile,
- MACROBLOCKD *const xd,
- int mi_row, int mi_col,
- BLOCK_SIZE *min_block_size,
+ MACROBLOCKD *const xd, int mi_row,
+ int mi_col, BLOCK_SIZE *min_block_size,
BLOCK_SIZE *max_block_size) {
VP10_COMMON *const cm = &cpi->common;
MODE_INFO **mi = xd->mi;
@@ -1853,7 +1752,7 @@
int bh, bw;
BLOCK_SIZE min_size = BLOCK_4X4;
BLOCK_SIZE max_size = BLOCK_64X64;
- int bs_hist[BLOCK_SIZES] = {0};
+ int bs_hist[BLOCK_SIZES] = { 0 };
// Trap case where we do not have a prediction.
if (left_in_image || above_in_image || cm->frame_type != KEY_FRAME) {
@@ -1890,8 +1789,7 @@
}
// Check border cases where max and min from neighbors may not be legal.
- max_size = find_partition_size(max_size,
- row8x8_remaining, col8x8_remaining,
+ max_size = find_partition_size(max_size, row8x8_remaining, col8x8_remaining,
&bh, &bw);
// Test for blocks at the edge of the active image.
// This may be the actual edge of the image or where there are formatting
@@ -1908,7 +1806,7 @@
// *min_block_size.
if (cpi->sf.use_square_partition_only &&
next_square_size[max_size] < min_size) {
- min_size = next_square_size[max_size];
+ min_size = next_square_size[max_size];
}
*min_block_size = min_size;
@@ -1916,10 +1814,10 @@
}
// TODO(jingning) refactor functions setting partition search range
-static void set_partition_range(VP10_COMMON *cm, MACROBLOCKD *xd,
- int mi_row, int mi_col, BLOCK_SIZE bsize,
+static void set_partition_range(VP10_COMMON *cm, MACROBLOCKD *xd, int mi_row,
+ int mi_col, BLOCK_SIZE bsize,
BLOCK_SIZE *min_bs, BLOCK_SIZE *max_bs) {
- int mi_width = num_8x8_blocks_wide_lookup[bsize];
+ int mi_width = num_8x8_blocks_wide_lookup[bsize];
int mi_height = num_8x8_blocks_high_lookup[bsize];
int idx, idy;
@@ -1978,16 +1876,19 @@
}
#if CONFIG_FP_MB_STATS
-const int num_16x16_blocks_wide_lookup[BLOCK_SIZES] =
- {1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 4, 4};
-const int num_16x16_blocks_high_lookup[BLOCK_SIZES] =
- {1, 1, 1, 1, 1, 1, 1, 2, 1, 2, 4, 2, 4};
-const int qindex_skip_threshold_lookup[BLOCK_SIZES] =
- {0, 10, 10, 30, 40, 40, 60, 80, 80, 90, 100, 100, 120};
-const int qindex_split_threshold_lookup[BLOCK_SIZES] =
- {0, 3, 3, 7, 15, 15, 30, 40, 40, 60, 80, 80, 120};
-const int complexity_16x16_blocks_threshold[BLOCK_SIZES] =
- {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 4, 4, 6};
+const int num_16x16_blocks_wide_lookup[BLOCK_SIZES] = { 1, 1, 1, 1, 1, 1, 1,
+ 1, 2, 2, 2, 4, 4 };
+const int num_16x16_blocks_high_lookup[BLOCK_SIZES] = { 1, 1, 1, 1, 1, 1, 1,
+ 2, 1, 2, 4, 2, 4 };
+const int qindex_skip_threshold_lookup[BLOCK_SIZES] = {
+ 0, 10, 10, 30, 40, 40, 60, 80, 80, 90, 100, 100, 120
+};
+const int qindex_split_threshold_lookup[BLOCK_SIZES] = {
+ 0, 3, 3, 7, 15, 15, 30, 40, 40, 60, 80, 80, 120
+};
+const int complexity_16x16_blocks_threshold[BLOCK_SIZES] = {
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 4, 4, 6
+};
typedef enum {
MV_ZERO = 0,
@@ -2026,10 +1927,10 @@
// unlikely to be selected depending on previous rate-distortion optimization
// results, for encoding speed-up.
static void rd_pick_partition(VP10_COMP *cpi, ThreadData *td,
- TileDataEnc *tile_data,
- TOKENEXTRA **tp, int mi_row, int mi_col,
- BLOCK_SIZE bsize, RD_COST *rd_cost,
- int64_t best_rd, PC_TREE *pc_tree) {
+ TileDataEnc *tile_data, TOKENEXTRA **tp,
+ int mi_row, int mi_col, BLOCK_SIZE bsize,
+ RD_COST *rd_cost, int64_t best_rd,
+ PC_TREE *pc_tree) {
VP10_COMMON *const cm = &cpi->common;
TileInfo *const tile_info = &tile_data->tile_info;
MACROBLOCK *const x = &td->mb;
@@ -2060,14 +1961,14 @@
#endif
int partition_none_allowed = !force_horz_split && !force_vert_split;
- int partition_horz_allowed = !force_vert_split && yss <= xss &&
- bsize >= BLOCK_8X8;
- int partition_vert_allowed = !force_horz_split && xss <= yss &&
- bsize >= BLOCK_8X8;
- (void) *tp_orig;
+ int partition_horz_allowed =
+ !force_vert_split && yss <= xss && bsize >= BLOCK_8X8;
+ int partition_vert_allowed =
+ !force_horz_split && xss <= yss && bsize >= BLOCK_8X8;
+ (void)*tp_orig;
assert(num_8x8_blocks_wide_lookup[bsize] ==
- num_8x8_blocks_high_lookup[bsize]);
+ num_8x8_blocks_high_lookup[bsize]);
vp10_rd_cost_init(&this_rdc);
vp10_rd_cost_init(&sum_rdc);
@@ -2080,8 +1981,10 @@
x->mb_energy = vp10_block_energy(cpi, x, bsize);
if (cpi->sf.cb_partition_search && bsize == BLOCK_16X16) {
- int cb_partition_search_ctrl = ((pc_tree->index == 0 || pc_tree->index == 3)
- + get_chessboard_index(cm->current_video_frame)) & 0x1;
+ int cb_partition_search_ctrl =
+ ((pc_tree->index == 0 || pc_tree->index == 3) +
+ get_chessboard_index(cm->current_video_frame)) &
+ 0x1;
if (cb_partition_search_ctrl && bsize > min_size && bsize < max_size)
set_partition_range(cm, xd, mi_row, mi_col, bsize, &min_size, &max_size);
@@ -2091,10 +1994,10 @@
// The threshold set here has to be of square block size.
if (cpi->sf.auto_min_max_partition_size) {
partition_none_allowed &= (bsize <= max_size && bsize >= min_size);
- partition_horz_allowed &= ((bsize <= max_size && bsize > min_size) ||
- force_horz_split);
- partition_vert_allowed &= ((bsize <= max_size && bsize > min_size) ||
- force_vert_split);
+ partition_horz_allowed &=
+ ((bsize <= max_size && bsize > min_size) || force_horz_split);
+ partition_vert_allowed &=
+ ((bsize <= max_size && bsize > min_size) || force_vert_split);
do_split &= bsize > min_size;
}
if (cpi->sf.use_square_partition_only) {
@@ -2107,8 +2010,8 @@
#if CONFIG_FP_MB_STATS
if (cpi->use_fp_mb_stats) {
set_offsets(cpi, tile_info, x, mi_row, mi_col, bsize);
- src_diff_var = get_sby_perpixel_diff_variance(cpi, &x->plane[0].src,
- mi_row, mi_col, bsize);
+ src_diff_var = get_sby_perpixel_diff_variance(cpi, &x->plane[0].src, mi_row,
+ mi_col, bsize);
}
#endif
@@ -2128,7 +2031,7 @@
// compute a complexity measure, basically measure inconsistency of motion
// vectors obtained from the first pass in the current block
- for (r = mb_row; r < mb_row_end ; r++) {
+ for (r = mb_row; r < mb_row_end; r++) {
for (c = mb_col; c < mb_col_end; c++) {
const int mb_index = r * cm->mb_cols + c;
@@ -2165,14 +2068,14 @@
// PARTITION_NONE
if (partition_none_allowed) {
- rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col,
- &this_rdc, bsize, ctx, best_rdc.rdcost);
+ rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &this_rdc, bsize, ctx,
+ best_rdc.rdcost);
if (this_rdc.rate != INT_MAX) {
if (bsize >= BLOCK_8X8) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
this_rdc.rate += cpi->partition_cost[pl][PARTITION_NONE];
- this_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
- this_rdc.rate, this_rdc.dist);
+ this_rdc.rdcost =
+ RDCOST(x->rdmult, x->rddiv, this_rdc.rate, this_rdc.dist);
}
if (this_rdc.rdcost < best_rdc.rdcost) {
@@ -2180,12 +2083,11 @@
int rate_breakout_thr = cpi->sf.partition_search_breakout_rate_thr;
best_rdc = this_rdc;
- if (bsize >= BLOCK_8X8)
- pc_tree->partitioning = PARTITION_NONE;
+ if (bsize >= BLOCK_8X8) pc_tree->partitioning = PARTITION_NONE;
// Adjust dist breakout threshold according to the partition size.
- dist_breakout_thr >>= 8 - (b_width_log2_lookup[bsize] +
- b_height_log2_lookup[bsize]);
+ dist_breakout_thr >>=
+ 8 - (b_width_log2_lookup[bsize] + b_height_log2_lookup[bsize]);
rate_breakout_thr *= num_pels_log2_lookup[bsize];
@@ -2196,7 +2098,7 @@
// early termination at that speed.
if (!x->e_mbd.lossless[xd->mi[0]->mbmi.segment_id] &&
(ctx->skippable && best_rdc.dist < dist_breakout_thr &&
- best_rdc.rate < rate_breakout_thr)) {
+ best_rdc.rate < rate_breakout_thr)) {
do_split = 0;
do_rect = 0;
}
@@ -2252,8 +2154,7 @@
}
// store estimated motion vector
- if (cpi->sf.adaptive_motion_search)
- store_pred_mv(x, ctx);
+ if (cpi->sf.adaptive_motion_search) store_pred_mv(x, ctx);
// PARTITION_SPLIT
// TODO(jingning): use the motion vectors given by the above search as
@@ -2267,23 +2168,20 @@
ctx->mic.mbmi.interp_filter;
rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &sum_rdc, subsize,
pc_tree->leaf_split[0], best_rdc.rdcost);
- if (sum_rdc.rate == INT_MAX)
- sum_rdc.rdcost = INT64_MAX;
+ if (sum_rdc.rate == INT_MAX) sum_rdc.rdcost = INT64_MAX;
} else {
for (i = 0; i < 4 && sum_rdc.rdcost < best_rdc.rdcost; ++i) {
- const int x_idx = (i & 1) * mi_step;
- const int y_idx = (i >> 1) * mi_step;
+ const int x_idx = (i & 1) * mi_step;
+ const int y_idx = (i >> 1) * mi_step;
if (mi_row + y_idx >= cm->mi_rows || mi_col + x_idx >= cm->mi_cols)
continue;
- if (cpi->sf.adaptive_motion_search)
- load_pred_mv(x, ctx);
+ if (cpi->sf.adaptive_motion_search) load_pred_mv(x, ctx);
pc_tree->split[i]->index = i;
- rd_pick_partition(cpi, td, tile_data, tp,
- mi_row + y_idx, mi_col + x_idx,
- subsize, &this_rdc,
+ rd_pick_partition(cpi, td, tile_data, tp, mi_row + y_idx,
+ mi_col + x_idx, subsize, &this_rdc,
best_rdc.rdcost - sum_rdc.rdcost, pc_tree->split[i]);
if (this_rdc.rate == INT_MAX) {
@@ -2300,8 +2198,7 @@
if (sum_rdc.rdcost < best_rdc.rdcost && i == 4) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
sum_rdc.rate += cpi->partition_cost[pl][PARTITION_SPLIT];
- sum_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
- sum_rdc.rate, sum_rdc.dist);
+ sum_rdc.rdcost = RDCOST(x->rdmult, x->rddiv, sum_rdc.rate, sum_rdc.dist);
if (sum_rdc.rdcost < best_rdc.rdcost) {
best_rdc = sum_rdc;
@@ -2310,8 +2207,7 @@
} else {
// skip rectangular partition test when larger block size
// gives better rd cost
- if (cpi->sf.less_rectangular_check)
- do_rect &= !partition_none_allowed;
+ if (cpi->sf.less_rectangular_check) do_rect &= !partition_none_allowed;
}
restore_context(x, mi_row, mi_col, a, l, sa, sl, bsize);
}
@@ -2319,13 +2215,11 @@
// PARTITION_HORZ
if (partition_horz_allowed &&
(do_rect || vp10_active_h_edge(cpi, mi_row, mi_step))) {
- subsize = get_subsize(bsize, PARTITION_HORZ);
- if (cpi->sf.adaptive_motion_search)
- load_pred_mv(x, ctx);
+ subsize = get_subsize(bsize, PARTITION_HORZ);
+ if (cpi->sf.adaptive_motion_search) load_pred_mv(x, ctx);
if (cpi->sf.adaptive_pred_interp_filter && bsize == BLOCK_8X8 &&
partition_none_allowed)
- pc_tree->horizontal[0].pred_interp_filter =
- ctx->mic.mbmi.interp_filter;
+ pc_tree->horizontal[0].pred_interp_filter = ctx->mic.mbmi.interp_filter;
rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &sum_rdc, subsize,
&pc_tree->horizontal[0], best_rdc.rdcost);
@@ -2335,14 +2229,12 @@
update_state(cpi, td, ctx, mi_row, mi_col, subsize, 0);
encode_superblock(cpi, td, tp, 0, mi_row, mi_col, subsize, ctx);
- if (cpi->sf.adaptive_motion_search)
- load_pred_mv(x, ctx);
+ if (cpi->sf.adaptive_motion_search) load_pred_mv(x, ctx);
if (cpi->sf.adaptive_pred_interp_filter && bsize == BLOCK_8X8 &&
partition_none_allowed)
- pc_tree->horizontal[1].pred_interp_filter =
- ctx->mic.mbmi.interp_filter;
- rd_pick_sb_modes(cpi, tile_data, x, mi_row + mi_step, mi_col,
- &this_rdc, subsize, &pc_tree->horizontal[1],
+ pc_tree->horizontal[1].pred_interp_filter = ctx->mic.mbmi.interp_filter;
+ rd_pick_sb_modes(cpi, tile_data, x, mi_row + mi_step, mi_col, &this_rdc,
+ subsize, &pc_tree->horizontal[1],
best_rdc.rdcost - sum_rdc.rdcost);
if (this_rdc.rate == INT_MAX) {
sum_rdc.rdcost = INT64_MAX;
@@ -2367,14 +2259,12 @@
// PARTITION_VERT
if (partition_vert_allowed &&
(do_rect || vp10_active_v_edge(cpi, mi_col, mi_step))) {
- subsize = get_subsize(bsize, PARTITION_VERT);
+ subsize = get_subsize(bsize, PARTITION_VERT);
- if (cpi->sf.adaptive_motion_search)
- load_pred_mv(x, ctx);
+ if (cpi->sf.adaptive_motion_search) load_pred_mv(x, ctx);
if (cpi->sf.adaptive_pred_interp_filter && bsize == BLOCK_8X8 &&
partition_none_allowed)
- pc_tree->vertical[0].pred_interp_filter =
- ctx->mic.mbmi.interp_filter;
+ pc_tree->vertical[0].pred_interp_filter = ctx->mic.mbmi.interp_filter;
rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col, &sum_rdc, subsize,
&pc_tree->vertical[0], best_rdc.rdcost);
if (sum_rdc.rdcost < best_rdc.rdcost && mi_col + mi_step < cm->mi_cols &&
@@ -2383,15 +2273,13 @@
encode_superblock(cpi, td, tp, 0, mi_row, mi_col, subsize,
&pc_tree->vertical[0]);
- if (cpi->sf.adaptive_motion_search)
- load_pred_mv(x, ctx);
+ if (cpi->sf.adaptive_motion_search) load_pred_mv(x, ctx);
if (cpi->sf.adaptive_pred_interp_filter && bsize == BLOCK_8X8 &&
partition_none_allowed)
- pc_tree->vertical[1].pred_interp_filter =
- ctx->mic.mbmi.interp_filter;
- rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col + mi_step,
- &this_rdc, subsize,
- &pc_tree->vertical[1], best_rdc.rdcost - sum_rdc.rdcost);
+ pc_tree->vertical[1].pred_interp_filter = ctx->mic.mbmi.interp_filter;
+ rd_pick_sb_modes(cpi, tile_data, x, mi_row, mi_col + mi_step, &this_rdc,
+ subsize, &pc_tree->vertical[1],
+ best_rdc.rdcost - sum_rdc.rdcost);
if (this_rdc.rate == INT_MAX) {
sum_rdc.rdcost = INT64_MAX;
} else {
@@ -2404,8 +2292,7 @@
if (sum_rdc.rdcost < best_rdc.rdcost) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
sum_rdc.rate += cpi->partition_cost[pl][PARTITION_VERT];
- sum_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
- sum_rdc.rate, sum_rdc.dist);
+ sum_rdc.rdcost = RDCOST(x->rdmult, x->rddiv, sum_rdc.rate, sum_rdc.dist);
if (sum_rdc.rdcost < best_rdc.rdcost) {
best_rdc = sum_rdc;
pc_tree->partitioning = PARTITION_VERT;
@@ -2418,15 +2305,14 @@
// warning related to the fact that best_rd isn't used after this
// point. This code should be refactored so that the duplicate
// checks occur in some sub function and thus are used...
- (void) best_rd;
+ (void)best_rd;
*rd_cost = best_rdc;
-
if (best_rdc.rate < INT_MAX && best_rdc.dist < INT64_MAX &&
pc_tree->index != 3) {
int output_enabled = (bsize == BLOCK_64X64);
- encode_sb(cpi, td, tile_info, tp, mi_row, mi_col, output_enabled,
- bsize, pc_tree);
+ encode_sb(cpi, td, tile_info, tp, mi_row, mi_col, output_enabled, bsize,
+ pc_tree);
}
if (bsize == BLOCK_64X64) {
@@ -2438,10 +2324,8 @@
}
}
-static void encode_rd_sb_row(VP10_COMP *cpi,
- ThreadData *td,
- TileDataEnc *tile_data,
- int mi_row,
+static void encode_rd_sb_row(VP10_COMP *cpi, ThreadData *td,
+ TileDataEnc *tile_data, int mi_row,
TOKENEXTRA **tp) {
VP10_COMMON *const cm = &cpi->common;
TileInfo *const tile_info = &tile_data->tile_info;
@@ -2468,8 +2352,7 @@
MODE_INFO **mi = cm->mi_grid_visible + idx_str;
if (sf->adaptive_pred_interp_filter) {
- for (i = 0; i < 64; ++i)
- td->leaf_tree[i].pred_interp_filter = SWITCHABLE;
+ for (i = 0; i < 64; ++i) td->leaf_tree[i].pred_interp_filter = SWITCHABLE;
for (i = 0; i < 64; ++i) {
td->pc_tree[i].vertical[0].pred_interp_filter = SWITCHABLE;
@@ -2483,8 +2366,8 @@
td->pc_root->index = 0;
if (seg->enabled) {
- const uint8_t *const map = seg->update_map ? cpi->segmentation_map
- : cm->last_frame_seg_map;
+ const uint8_t *const map =
+ seg->update_map ? cpi->segmentation_map : cm->last_frame_seg_map;
int segment_id = get_segment_id(cm, map, BLOCK_64X64, mi_row, mi_col);
seg_skip = segfeature_active(seg, segment_id, SEG_LVL_SKIP);
}
@@ -2495,27 +2378,26 @@
seg_skip ? BLOCK_64X64 : sf->always_this_block_size;
set_offsets(cpi, tile_info, x, mi_row, mi_col, BLOCK_64X64);
set_fixed_partitioning(cpi, tile_info, mi, mi_row, mi_col, bsize);
- rd_use_partition(cpi, td, tile_data, mi, tp, mi_row, mi_col,
- BLOCK_64X64, &dummy_rate, &dummy_dist, 1, td->pc_root);
+ rd_use_partition(cpi, td, tile_data, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ &dummy_rate, &dummy_dist, 1, td->pc_root);
} else if (cpi->partition_search_skippable_frame) {
BLOCK_SIZE bsize;
set_offsets(cpi, tile_info, x, mi_row, mi_col, BLOCK_64X64);
bsize = get_rd_var_based_fixed_partition(cpi, x, mi_row, mi_col);
set_fixed_partitioning(cpi, tile_info, mi, mi_row, mi_col, bsize);
- rd_use_partition(cpi, td, tile_data, mi, tp, mi_row, mi_col,
- BLOCK_64X64, &dummy_rate, &dummy_dist, 1, td->pc_root);
+ rd_use_partition(cpi, td, tile_data, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ &dummy_rate, &dummy_dist, 1, td->pc_root);
} else if (sf->partition_search_type == VAR_BASED_PARTITION &&
cm->frame_type != KEY_FRAME) {
choose_partitioning(cpi, tile_info, x, mi_row, mi_col);
- rd_use_partition(cpi, td, tile_data, mi, tp, mi_row, mi_col,
- BLOCK_64X64, &dummy_rate, &dummy_dist, 1, td->pc_root);
+ rd_use_partition(cpi, td, tile_data, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ &dummy_rate, &dummy_dist, 1, td->pc_root);
} else {
// If required set upper and lower partition size limits
if (sf->auto_min_max_partition_size) {
set_offsets(cpi, tile_info, x, mi_row, mi_col, BLOCK_64X64);
rd_auto_partition_range(cpi, tile_info, xd, mi_row, mi_col,
- &x->min_partition_size,
- &x->max_partition_size);
+ &x->min_partition_size, &x->max_partition_size);
}
rd_pick_partition(cpi, td, tile_data, tp, mi_row, mi_col, BLOCK_64X64,
&dummy_rdc, INT64_MAX, td->pc_root);
@@ -2537,8 +2419,7 @@
// Note: this memset assumes above_context[0], [1] and [2]
// are allocated as part of the same buffer.
memset(xd->above_context[0], 0,
- sizeof(*xd->above_context[0]) *
- 2 * aligned_mi_cols * MAX_MB_PLANE);
+ sizeof(*xd->above_context[0]) * 2 * aligned_mi_cols * MAX_MB_PLANE);
memset(xd->above_seg_context, 0,
sizeof(*xd->above_seg_context) * aligned_mi_cols);
}
@@ -2549,8 +2430,8 @@
if (segfeature_active(&cpi->common.seg, 1, SEG_LVL_REF_FRAME)) {
return 0;
} else {
- return (!!(ref_flags & VPX_GOLD_FLAG) + !!(ref_flags & VPX_LAST_FLAG)
- + !!(ref_flags & VPX_ALT_FLAG)) >= 2;
+ return (!!(ref_flags & VPX_GOLD_FLAG) + !!(ref_flags & VPX_LAST_FLAG) +
+ !!(ref_flags & VPX_ALT_FLAG)) >= 2;
}
}
@@ -2579,11 +2460,10 @@
}
static TX_MODE select_tx_mode(const VP10_COMP *cpi, MACROBLOCKD *const xd) {
- if (xd->lossless[0])
- return ONLY_4X4;
+ if (xd->lossless[0]) return ONLY_4X4;
if (cpi->sf.tx_size_search_method == USE_LARGESTALL)
return ALLOW_32X32;
- else if (cpi->sf.tx_size_search_method == USE_FULL_RD||
+ else if (cpi->sf.tx_size_search_method == USE_FULL_RD ||
cpi->sf.tx_size_search_method == USE_TX_8X8)
return TX_MODE_SELECT;
else
@@ -2599,10 +2479,9 @@
int tile_tok = 0;
if (cpi->tile_data == NULL || cpi->allocated_tiles < tile_cols * tile_rows) {
- if (cpi->tile_data != NULL)
- vpx_free(cpi->tile_data);
- CHECK_MEM_ERROR(cm, cpi->tile_data,
- vpx_malloc(tile_cols * tile_rows * sizeof(*cpi->tile_data)));
+ if (cpi->tile_data != NULL) vpx_free(cpi->tile_data);
+ CHECK_MEM_ERROR(cm, cpi->tile_data, vpx_malloc(tile_cols * tile_rows *
+ sizeof(*cpi->tile_data)));
cpi->allocated_tiles = tile_cols * tile_rows;
for (tile_row = 0; tile_row < tile_rows; ++tile_row)
@@ -2632,13 +2511,12 @@
}
}
-void vp10_encode_tile(VP10_COMP *cpi, ThreadData *td,
- int tile_row, int tile_col) {
+void vp10_encode_tile(VP10_COMP *cpi, ThreadData *td, int tile_row,
+ int tile_col) {
VP10_COMMON *const cm = &cpi->common;
const int tile_cols = 1 << cm->log2_tile_cols;
- TileDataEnc *this_tile =
- &cpi->tile_data[tile_row * tile_cols + tile_col];
- const TileInfo * const tile_info = &this_tile->tile_info;
+ TileDataEnc *this_tile = &cpi->tile_data[tile_row * tile_cols + tile_col];
+ const TileInfo *const tile_info = &this_tile->tile_info;
TOKENEXTRA *tok = cpi->tile_tok[tile_row][tile_col];
int mi_row;
@@ -2653,7 +2531,7 @@
cpi->tok_count[tile_row][tile_col] =
(unsigned int)(tok - cpi->tile_tok[tile_row][tile_col]);
assert(tok - cpi->tile_tok[tile_row][tile_col] <=
- allocated_tokens(*tile_info));
+ allocated_tokens(*tile_info));
}
static void encode_tiles(VP10_COMP *cpi) {
@@ -2673,10 +2551,9 @@
static int input_fpmb_stats(FIRSTPASS_MB_STATS *firstpass_mb_stats,
VP10_COMMON *cm, uint8_t **this_frame_mb_stats) {
uint8_t *mb_stats_in = firstpass_mb_stats->mb_stats_start +
- cm->current_video_frame * cm->MBs * sizeof(uint8_t);
+ cm->current_video_frame * cm->MBs * sizeof(uint8_t);
- if (mb_stats_in > firstpass_mb_stats->mb_stats_end)
- return EOF;
+ if (mb_stats_in > firstpass_mb_stats->mb_stats_end) return EOF;
*this_frame_mb_stats = mb_stats_in;
@@ -2703,17 +2580,14 @@
rdc->ex_search_count = 0; // Exhaustive mesh search hits.
for (i = 0; i < MAX_SEGMENTS; ++i) {
- const int qindex = CONFIG_MISC_FIXES && cm->seg.enabled ?
- vp10_get_qindex(&cm->seg, i, cm->base_qindex) :
- cm->base_qindex;
- xd->lossless[i] = qindex == 0 &&
- cm->y_dc_delta_q == 0 &&
- cm->uv_dc_delta_q == 0 &&
- cm->uv_ac_delta_q == 0;
+ const int qindex = CONFIG_MISC_FIXES && cm->seg.enabled
+ ? vp10_get_qindex(&cm->seg, i, cm->base_qindex)
+ : cm->base_qindex;
+ xd->lossless[i] = qindex == 0 && cm->y_dc_delta_q == 0 &&
+ cm->uv_dc_delta_q == 0 && cm->uv_ac_delta_q == 0;
}
- if (!cm->seg.enabled && xd->lossless[0])
- x->optimize = 0;
+ if (!cm->seg.enabled && xd->lossless[0]) x->optimize = 0;
cm->tx_mode = select_tx_mode(cpi, xd);
@@ -2722,15 +2596,13 @@
vp10_initialize_rd_consts(cpi);
vp10_initialize_me_consts(cpi, x, cm->base_qindex);
init_encode_frame_mb_context(cpi);
- cm->use_prev_frame_mvs = !cm->error_resilient_mode &&
- cm->width == cm->last_width &&
- cm->height == cm->last_height &&
- !cm->intra_only &&
- cm->last_show_frame;
+ cm->use_prev_frame_mvs =
+ !cm->error_resilient_mode && cm->width == cm->last_width &&
+ cm->height == cm->last_height && !cm->intra_only && cm->last_show_frame;
// Special case: set prev_mi to NULL when the previous mode info
// context cannot be used.
- cm->prev_mi = cm->use_prev_frame_mvs ?
- cm->prev_mip + cm->mi_stride + 1 : NULL;
+ cm->prev_mi =
+ cm->use_prev_frame_mvs ? cm->prev_mip + cm->mi_stride + 1 : NULL;
x->quant_fp = cpi->sf.use_quant_fp;
vp10_zero(x->skip_txfm);
@@ -2740,10 +2612,10 @@
vpx_usec_timer_start(&emr_timer);
#if CONFIG_FP_MB_STATS
- if (cpi->use_fp_mb_stats) {
- input_fpmb_stats(&cpi->twopass.firstpass_mb_stats, cm,
- &cpi->twopass.this_frame_mb_stats);
- }
+ if (cpi->use_fp_mb_stats) {
+ input_fpmb_stats(&cpi->twopass.firstpass_mb_stats, cm,
+ &cpi->twopass.this_frame_mb_stats);
+ }
#endif
// If allowed, encoding tiles in parallel with one thread handling one tile.
@@ -2764,8 +2636,7 @@
static INTERP_FILTER get_interp_filter(
const int64_t threshes[SWITCHABLE_FILTER_CONTEXTS], int is_alt_ref) {
- if (!is_alt_ref &&
- threshes[EIGHTTAP_SMOOTH] > threshes[EIGHTTAP] &&
+ if (!is_alt_ref && threshes[EIGHTTAP_SMOOTH] > threshes[EIGHTTAP] &&
threshes[EIGHTTAP_SMOOTH] > threshes[EIGHTTAP_SHARP] &&
threshes[EIGHTTAP_SMOOTH] > threshes[SWITCHABLE - 1]) {
return EIGHTTAP_SMOOTH;
@@ -2790,9 +2661,9 @@
// the other two.
if (!frame_is_intra_only(cm)) {
if ((cm->ref_frame_sign_bias[ALTREF_FRAME] ==
- cm->ref_frame_sign_bias[GOLDEN_FRAME]) ||
+ cm->ref_frame_sign_bias[GOLDEN_FRAME]) ||
(cm->ref_frame_sign_bias[ALTREF_FRAME] ==
- cm->ref_frame_sign_bias[LAST_FRAME])) {
+ cm->ref_frame_sign_bias[LAST_FRAME])) {
cpi->allow_comp_inter_inter = 0;
} else {
cpi->allow_comp_inter_inter = 1;
@@ -2826,10 +2697,8 @@
if (is_alt_ref || !cpi->allow_comp_inter_inter)
cm->reference_mode = SINGLE_REFERENCE;
else if (mode_thrs[COMPOUND_REFERENCE] > mode_thrs[SINGLE_REFERENCE] &&
- mode_thrs[COMPOUND_REFERENCE] >
- mode_thrs[REFERENCE_MODE_SELECT] &&
- check_dual_ref_flags(cpi) &&
- cpi->static_mb_pct == 100)
+ mode_thrs[COMPOUND_REFERENCE] > mode_thrs[REFERENCE_MODE_SELECT] &&
+ check_dual_ref_flags(cpi) && cpi->static_mb_pct == 100)
cm->reference_mode = COMPOUND_REFERENCE;
else if (mode_thrs[SINGLE_REFERENCE] > mode_thrs[REFERENCE_MODE_SELECT])
cm->reference_mode = SINGLE_REFERENCE;
@@ -2941,18 +2810,17 @@
++counts->uv_mode[y_mode][uv_mode];
}
-static void encode_superblock(VP10_COMP *cpi, ThreadData *td,
- TOKENEXTRA **t, int output_enabled,
- int mi_row, int mi_col, BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx) {
+static void encode_superblock(VP10_COMP *cpi, ThreadData *td, TOKENEXTRA **t,
+ int output_enabled, int mi_row, int mi_col,
+ BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx) {
VP10_COMMON *const cm = &cpi->common;
MACROBLOCK *const x = &td->mb;
MACROBLOCKD *const xd = &x->e_mbd;
MODE_INFO **mi_8x8 = xd->mi;
MODE_INFO *mi = mi_8x8[0];
MB_MODE_INFO *mbmi = &mi->mbmi;
- const int seg_skip = segfeature_active(&cm->seg, mbmi->segment_id,
- SEG_LVL_SKIP);
+ const int seg_skip =
+ segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP);
const int mis = cm->mi_stride;
const int mi_width = num_8x8_blocks_wide_lookup[bsize];
const int mi_height = num_8x8_blocks_high_lookup[bsize];
@@ -2962,8 +2830,7 @@
cpi->oxcf.aq_mode != CYCLIC_REFRESH_AQ &&
cpi->sf.allow_skip_recode;
- if (!x->skip_recode)
- memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
+ if (!x->skip_recode) memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
x->skip_optimize = ctx->is_coded;
ctx->is_coded = 1;
@@ -2983,11 +2850,10 @@
const int is_compound = has_second_ref(mbmi);
set_ref_ptrs(cm, xd, mbmi->ref_frame[0], mbmi->ref_frame[1]);
for (ref = 0; ref < 1 + is_compound; ++ref) {
- YV12_BUFFER_CONFIG *cfg = get_ref_frame_buffer(cpi,
- mbmi->ref_frame[ref]);
+ YV12_BUFFER_CONFIG *cfg = get_ref_frame_buffer(cpi, mbmi->ref_frame[ref]);
assert(cfg != NULL);
vp10_setup_pre_planes(xd, ref, cfg, mi_row, mi_col,
- &xd->block_refs[ref]->sf);
+ &xd->block_refs[ref]->sf);
}
if (!(cpi->sf.reuse_inter_pred_sby && ctx->pred_pixel_ready) || seg_skip)
vp10_build_inter_predictors_sby(xd, mi_row, mi_col,
@@ -3001,8 +2867,7 @@
}
if (output_enabled) {
- if (cm->tx_mode == TX_MODE_SELECT &&
- mbmi->sb_type >= BLOCK_8X8 &&
+ if (cm->tx_mode == TX_MODE_SELECT && mbmi->sb_type >= BLOCK_8X8 &&
!(is_inter_block(mbmi) && (mbmi->skip || seg_skip))) {
++get_tx_counts(max_txsize_lookup[bsize], get_tx_size_context(xd),
&td->counts->tx)[mbmi->tx_size];
@@ -3024,15 +2889,14 @@
}
++td->counts->tx.tx_totals[mbmi->tx_size];
++td->counts->tx.tx_totals[get_uv_tx_size(mbmi, &xd->plane[1])];
- if (mbmi->tx_size < TX_32X32 &&
- cm->base_qindex > 0 && !mbmi->skip &&
+ if (mbmi->tx_size < TX_32X32 && cm->base_qindex > 0 && !mbmi->skip &&
!segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
if (is_inter_block(mbmi)) {
++td->counts->inter_ext_tx[mbmi->tx_size][mbmi->tx_type];
} else {
- ++td->counts->intra_ext_tx[mbmi->tx_size]
- [intra_mode_to_tx_type_context[mbmi->mode]]
- [mbmi->tx_type];
+ ++td->counts
+ ->intra_ext_tx[mbmi->tx_size][intra_mode_to_tx_type_context
+ [mbmi->mode]][mbmi->tx_type];
}
}
}
diff --git a/vp10/encoder/encodeframe.h b/vp10/encoder/encodeframe.h
index fbb81f8..f24a572 100644
--- a/vp10/encoder/encodeframe.h
+++ b/vp10/encoder/encodeframe.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_ENCODEFRAME_H_
#define VP10_ENCODER_ENCODEFRAME_H_
@@ -31,14 +30,14 @@
#define VAR_HIST_SMALL_CUT_OFF 45
void vp10_setup_src_planes(struct macroblock *x,
- const struct yv12_buffer_config *src,
- int mi_row, int mi_col);
+ const struct yv12_buffer_config *src, int mi_row,
+ int mi_col);
void vp10_encode_frame(struct VP10_COMP *cpi);
void vp10_init_tile_data(struct VP10_COMP *cpi);
void vp10_encode_tile(struct VP10_COMP *cpi, struct ThreadData *td,
- int tile_row, int tile_col);
+ int tile_row, int tile_col);
void vp10_set_variance_partition_thresholds(struct VP10_COMP *cpi, int q);
diff --git a/vp10/encoder/encodemb.c b/vp10/encoder/encodemb.c
index 0a23d35..b2fbf13 100644
--- a/vp10/encoder/encodemb.c
+++ b/vp10/encoder/encodemb.c
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#include "./vp10_rtcd.h"
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"
@@ -53,32 +52,30 @@
#define RDTRUNC(RM, DM, R, D) ((128 + (R) * (RM)) & 0xFF)
typedef struct vp10_token_state {
- int rate;
- int error;
- int next;
- int16_t token;
- short qc;
+ int rate;
+ int error;
+ int next;
+ int16_t token;
+ short qc;
} vp10_token_state;
// TODO(jimbankoski): experiment to find optimal RD numbers.
static const int plane_rd_mult[PLANE_TYPES] = { 4, 2 };
-#define UPDATE_RD_COST()\
-{\
- rd_cost0 = RDCOST(rdmult, rddiv, rate0, error0);\
- rd_cost1 = RDCOST(rdmult, rddiv, rate1, error1);\
- if (rd_cost0 == rd_cost1) {\
- rd_cost0 = RDTRUNC(rdmult, rddiv, rate0, error0);\
- rd_cost1 = RDTRUNC(rdmult, rddiv, rate1, error1);\
- }\
-}
+#define UPDATE_RD_COST() \
+ { \
+ rd_cost0 = RDCOST(rdmult, rddiv, rate0, error0); \
+ rd_cost1 = RDCOST(rdmult, rddiv, rate1, error1); \
+ if (rd_cost0 == rd_cost1) { \
+ rd_cost0 = RDTRUNC(rdmult, rddiv, rate0, error0); \
+ rd_cost1 = RDTRUNC(rdmult, rddiv, rate1, error1); \
+ } \
+ }
// This function is a place holder for now but may ultimately need
// to scan previous tokens to work out the correct context.
-static int trellis_get_coeff_context(const int16_t *scan,
- const int16_t *nb,
- int idx, int token,
- uint8_t *token_cache) {
+static int trellis_get_coeff_context(const int16_t *scan, const int16_t *nb,
+ int idx, int token, uint8_t *token_cache) {
int bak = token_cache[scan[idx]], pt;
token_cache[scan[idx]] = vp10_pt_energy_class[token];
pt = get_coef_context(nb, token_cache, idx + 1);
@@ -86,8 +83,8 @@
return pt;
}
-static int optimize_b(MACROBLOCK *mb, int plane, int block,
- TX_SIZE tx_size, int ctx) {
+static int optimize_b(MACROBLOCK *mb, int plane, int block, TX_SIZE tx_size,
+ int ctx) {
MACROBLOCKD *const xd = &mb->e_mbd;
struct macroblock_plane *const p = &mb->plane[plane];
struct macroblockd_plane *const pd = &xd->plane[plane];
@@ -125,8 +122,7 @@
assert(eob <= default_eob);
/* Now set up a Viterbi trellis to evaluate alternative roundings. */
- if (!ref)
- rdmult = (rdmult * 9) >> 4;
+ if (!ref) rdmult = (rdmult * 9) >> 4;
/* Initialize the sentinel node of the trellis. */
tokens[eob][0].rate = 0;
@@ -157,10 +153,12 @@
if (next < default_eob) {
band = band_translate[i + 1];
pt = trellis_get_coeff_context(scan, nb, i, t0, token_cache);
- rate0 += mb->token_costs[tx_size][type][ref][band][0][pt]
- [tokens[next][0].token];
- rate1 += mb->token_costs[tx_size][type][ref][band][0][pt]
- [tokens[next][1].token];
+ rate0 +=
+ mb->token_costs[tx_size][type][ref][band][0][pt][tokens[next][0]
+ .token];
+ rate1 +=
+ mb->token_costs[tx_size][type][ref][band][0][pt][tokens[next][1]
+ .token];
}
UPDATE_RD_COST();
/* And pick the best. */
@@ -185,8 +183,8 @@
rate1 = tokens[next][1].rate;
if ((abs(x) * dequant_ptr[rc != 0] > abs(coeff[rc]) * mul) &&
- (abs(x) * dequant_ptr[rc != 0] < abs(coeff[rc]) * mul +
- dequant_ptr[rc != 0]))
+ (abs(x) * dequant_ptr[rc != 0] <
+ abs(coeff[rc]) * mul + dequant_ptr[rc != 0]))
shortcut = 1;
else
shortcut = 0;
@@ -212,13 +210,15 @@
band = band_translate[i + 1];
if (t0 != EOB_TOKEN) {
pt = trellis_get_coeff_context(scan, nb, i, t0, token_cache);
- rate0 += mb->token_costs[tx_size][type][ref][band][!x][pt]
- [tokens[next][0].token];
+ rate0 +=
+ mb->token_costs[tx_size][type][ref][band][!x][pt][tokens[next][0]
+ .token];
}
if (t1 != EOB_TOKEN) {
pt = trellis_get_coeff_context(scan, nb, i, t1, token_cache);
- rate1 += mb->token_costs[tx_size][type][ref][band][!x][pt]
- [tokens[next][1].token];
+ rate1 +=
+ mb->token_costs[tx_size][type][ref][band][!x][pt][tokens[next][1]
+ .token];
}
}
@@ -304,9 +304,8 @@
return final_eob;
}
-static INLINE void fdct32x32(int rd_transform,
- const int16_t *src, tran_low_t *dst,
- int src_stride) {
+static INLINE void fdct32x32(int rd_transform, const int16_t *src,
+ tran_low_t *dst, int src_stride) {
if (rd_transform)
vpx_fdct32x32_rd(src, dst, src_stride);
else
@@ -329,17 +328,11 @@
vp10_fwht4x4(src_diff, coeff, diff_stride);
} else {
switch (tx_type) {
- case DCT_DCT:
- vpx_fdct4x4(src_diff, coeff, diff_stride);
- break;
+ case DCT_DCT: vpx_fdct4x4(src_diff, coeff, diff_stride); break;
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- vp10_fht4x4(src_diff, coeff, diff_stride, tx_type);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: vp10_fht4x4(src_diff, coeff, diff_stride, tx_type); break;
+ default: assert(0); break;
}
}
}
@@ -350,12 +343,8 @@
case DCT_DCT:
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- vp10_fht8x8(src_diff, coeff, diff_stride, tx_type);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: vp10_fht8x8(src_diff, coeff, diff_stride, tx_type); break;
+ default: assert(0); break;
}
}
@@ -365,12 +354,8 @@
case DCT_DCT:
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- vp10_fht16x16(src_diff, coeff, diff_stride, tx_type);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: vp10_fht16x16(src_diff, coeff, diff_stride, tx_type); break;
+ default: assert(0); break;
}
}
@@ -378,17 +363,11 @@
tran_low_t *coeff, int diff_stride,
TX_TYPE tx_type) {
switch (tx_type) {
- case DCT_DCT:
- fdct32x32(rd_transform, src_diff, coeff, diff_stride);
- break;
+ case DCT_DCT: fdct32x32(rd_transform, src_diff, coeff, diff_stride); break;
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- assert(0);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: assert(0); break;
+ default: assert(0); break;
}
}
@@ -400,52 +379,40 @@
vp10_highbd_fwht4x4(src_diff, coeff, diff_stride);
} else {
switch (tx_type) {
- case DCT_DCT:
- vpx_highbd_fdct4x4(src_diff, coeff, diff_stride);
- break;
+ case DCT_DCT: vpx_highbd_fdct4x4(src_diff, coeff, diff_stride); break;
case ADST_DCT:
case DCT_ADST:
case ADST_ADST:
vp10_highbd_fht4x4(src_diff, coeff, diff_stride, tx_type);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
}
static void highbd_fwd_txfm_8x8(const int16_t *src_diff, tran_low_t *coeff,
- int diff_stride, TX_TYPE tx_type) {
+ int diff_stride, TX_TYPE tx_type) {
switch (tx_type) {
- case DCT_DCT:
- vpx_highbd_fdct8x8(src_diff, coeff, diff_stride);
- break;
+ case DCT_DCT: vpx_highbd_fdct8x8(src_diff, coeff, diff_stride); break;
case ADST_DCT:
case DCT_ADST:
case ADST_ADST:
vp10_highbd_fht8x8(src_diff, coeff, diff_stride, tx_type);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
static void highbd_fwd_txfm_16x16(const int16_t *src_diff, tran_low_t *coeff,
- int diff_stride, TX_TYPE tx_type) {
+ int diff_stride, TX_TYPE tx_type) {
switch (tx_type) {
- case DCT_DCT:
- vpx_highbd_fdct16x16(src_diff, coeff, diff_stride);
- break;
+ case DCT_DCT: vpx_highbd_fdct16x16(src_diff, coeff, diff_stride); break;
case ADST_DCT:
case DCT_ADST:
case ADST_ADST:
vp10_highbd_fht16x16(src_diff, coeff, diff_stride, tx_type);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
@@ -458,19 +425,14 @@
break;
case ADST_DCT:
case DCT_ADST:
- case ADST_ADST:
- assert(0);
- break;
- default:
- assert(0);
- break;
+ case ADST_ADST: assert(0); break;
+ default: assert(0); break;
}
}
#endif // CONFIG_VPX_HIGHBITDEPTH
-void vp10_xform_quant_fp(MACROBLOCK *x, int plane, int block,
- int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
+void vp10_xform_quant_fp(MACROBLOCK *x, int plane, int block, int blk_row,
+ int blk_col, BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
MACROBLOCKD *const xd = &x->e_mbd;
const struct macroblock_plane *const p = &x->plane[plane];
const struct macroblockd_plane *const pd = &xd->plane[plane];
@@ -491,24 +453,23 @@
case TX_32X32:
highbd_fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride);
vp10_highbd_quantize_fp_32x32(coeff, 1024, x->skip_block, p->zbin,
- p->round_fp, p->quant_fp, p->quant_shift,
- qcoeff, dqcoeff, pd->dequant,
- eob, scan_order->scan,
- scan_order->iscan);
+ p->round_fp, p->quant_fp, p->quant_shift,
+ qcoeff, dqcoeff, pd->dequant, eob,
+ scan_order->scan, scan_order->iscan);
break;
case TX_16X16:
vpx_highbd_fdct16x16(src_diff, coeff, diff_stride);
vp10_highbd_quantize_fp(coeff, 256, x->skip_block, p->zbin, p->round_fp,
- p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
case TX_8X8:
vpx_highbd_fdct8x8(src_diff, coeff, diff_stride);
vp10_highbd_quantize_fp(coeff, 64, x->skip_block, p->zbin, p->round_fp,
- p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
case TX_4X4:
if (xd->lossless[xd->mi[0]->mbmi.segment_id]) {
@@ -517,12 +478,11 @@
vpx_highbd_fdct4x4(src_diff, coeff, diff_stride);
}
vp10_highbd_quantize_fp(coeff, 16, x->skip_block, p->zbin, p->round_fp,
- p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
- default:
- assert(0);
+ default: assert(0);
}
return;
}
@@ -532,23 +492,21 @@
case TX_32X32:
fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride);
vp10_quantize_fp_32x32(coeff, 1024, x->skip_block, p->zbin, p->round_fp,
- p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob, scan_order->scan,
- scan_order->iscan);
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
case TX_16X16:
vpx_fdct16x16(src_diff, coeff, diff_stride);
vp10_quantize_fp(coeff, 256, x->skip_block, p->zbin, p->round_fp,
- p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, eob, scan_order->scan, scan_order->iscan);
break;
case TX_8X8:
- vp10_fdct8x8_quant(src_diff, diff_stride, coeff, 64,
- x->skip_block, p->zbin, p->round_fp,
- p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ vp10_fdct8x8_quant(src_diff, diff_stride, coeff, 64, x->skip_block,
+ p->zbin, p->round_fp, p->quant_fp, p->quant_shift,
+ qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
case TX_4X4:
if (xd->lossless[xd->mi[0]->mbmi.segment_id]) {
@@ -557,19 +515,15 @@
vpx_fdct4x4(src_diff, coeff, diff_stride);
}
vp10_quantize_fp(coeff, 16, x->skip_block, p->zbin, p->round_fp,
- p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, eob, scan_order->scan, scan_order->iscan);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
-void vp10_xform_quant_dc(MACROBLOCK *x, int plane, int block,
- int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
+void vp10_xform_quant_dc(MACROBLOCK *x, int plane, int block, int blk_row,
+ int blk_col, BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
MACROBLOCKD *const xd = &x->e_mbd;
const struct macroblock_plane *const p = &x->plane[plane];
const struct macroblockd_plane *const pd = &xd->plane[plane];
@@ -593,14 +547,14 @@
case TX_16X16:
vpx_highbd_fdct16x16_1(src_diff, coeff, diff_stride);
vpx_highbd_quantize_dc(coeff, 256, x->skip_block, p->round,
- p->quant_fp[0], qcoeff, dqcoeff,
- pd->dequant[0], eob);
+ p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0],
+ eob);
break;
case TX_8X8:
vpx_highbd_fdct8x8_1(src_diff, coeff, diff_stride);
vpx_highbd_quantize_dc(coeff, 64, x->skip_block, p->round,
- p->quant_fp[0], qcoeff, dqcoeff,
- pd->dequant[0], eob);
+ p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0],
+ eob);
break;
case TX_4X4:
if (xd->lossless[xd->mi[0]->mbmi.segment_id]) {
@@ -609,11 +563,10 @@
vpx_highbd_fdct4x4(src_diff, coeff, diff_stride);
}
vpx_highbd_quantize_dc(coeff, 16, x->skip_block, p->round,
- p->quant_fp[0], qcoeff, dqcoeff,
- pd->dequant[0], eob);
+ p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0],
+ eob);
break;
- default:
- assert(0);
+ default: assert(0);
}
return;
}
@@ -622,21 +575,18 @@
switch (tx_size) {
case TX_32X32:
vpx_fdct32x32_1(src_diff, coeff, diff_stride);
- vpx_quantize_dc_32x32(coeff, x->skip_block, p->round,
- p->quant_fp[0], qcoeff, dqcoeff,
- pd->dequant[0], eob);
+ vpx_quantize_dc_32x32(coeff, x->skip_block, p->round, p->quant_fp[0],
+ qcoeff, dqcoeff, pd->dequant[0], eob);
break;
case TX_16X16:
vpx_fdct16x16_1(src_diff, coeff, diff_stride);
- vpx_quantize_dc(coeff, 256, x->skip_block, p->round,
- p->quant_fp[0], qcoeff, dqcoeff,
- pd->dequant[0], eob);
+ vpx_quantize_dc(coeff, 256, x->skip_block, p->round, p->quant_fp[0],
+ qcoeff, dqcoeff, pd->dequant[0], eob);
break;
case TX_8X8:
vpx_fdct8x8_1(src_diff, coeff, diff_stride);
- vpx_quantize_dc(coeff, 64, x->skip_block, p->round,
- p->quant_fp[0], qcoeff, dqcoeff,
- pd->dequant[0], eob);
+ vpx_quantize_dc(coeff, 64, x->skip_block, p->round, p->quant_fp[0],
+ qcoeff, dqcoeff, pd->dequant[0], eob);
break;
case TX_4X4:
if (xd->lossless[xd->mi[0]->mbmi.segment_id]) {
@@ -644,21 +594,15 @@
} else {
vpx_fdct4x4(src_diff, coeff, diff_stride);
}
- vpx_quantize_dc(coeff, 16, x->skip_block, p->round,
- p->quant_fp[0], qcoeff, dqcoeff,
- pd->dequant[0], eob);
+ vpx_quantize_dc(coeff, 16, x->skip_block, p->round, p->quant_fp[0],
+ qcoeff, dqcoeff, pd->dequant[0], eob);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
-
-
-void vp10_xform_quant(MACROBLOCK *x, int plane, int block,
- int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
+void vp10_xform_quant(MACROBLOCK *x, int plane, int block, int blk_row,
+ int blk_col, BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
MACROBLOCKD *const xd = &x->e_mbd;
const struct macroblock_plane *const p = &x->plane[plane];
const struct macroblockd_plane *const pd = &xd->plane[plane];
@@ -675,39 +619,38 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- switch (tx_size) {
+ switch (tx_size) {
case TX_32X32:
highbd_fwd_txfm_32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride,
- tx_type);
+ tx_type);
vpx_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin,
p->round, p->quant, p->quant_shift, qcoeff,
- dqcoeff, pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ dqcoeff, pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
case TX_16X16:
highbd_fwd_txfm_16x16(src_diff, coeff, diff_stride, tx_type);
vpx_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
case TX_8X8:
highbd_fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type);
vpx_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
case TX_4X4:
vp10_highbd_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
xd->lossless[xd->mi[0]->mbmi.segment_id]);
vpx_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
break;
- default:
- assert(0);
+ default: assert(0);
}
return;
}
@@ -723,35 +666,29 @@
break;
case TX_16X16:
fwd_txfm_16x16(src_diff, coeff, diff_stride, tx_type);
- vpx_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
- p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
+ vpx_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, p->quant,
+ p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob,
scan_order->scan, scan_order->iscan);
break;
case TX_8X8:
fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type);
- vpx_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
- p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
+ vpx_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, p->quant,
+ p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob,
scan_order->scan, scan_order->iscan);
break;
case TX_4X4:
vp10_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
xd->lossless[xd->mi[0]->mbmi.segment_id]);
- vpx_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
- p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
+ vpx_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, p->quant,
+ p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob,
scan_order->scan, scan_order->iscan);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
static void encode_block(int plane, int block, int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize,
- TX_SIZE tx_size, void *arg) {
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size, void *arg) {
struct encode_b_args *const args = arg;
MACROBLOCK *const x = args->x;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -783,20 +720,20 @@
*a = *l = 0;
return;
} else {
- vp10_xform_quant_fp(x, plane, block, blk_row, blk_col,
- plane_bsize, tx_size);
+ vp10_xform_quant_fp(x, plane, block, blk_row, blk_col, plane_bsize,
+ tx_size);
}
} else {
if (max_txsize_lookup[plane_bsize] == tx_size) {
int txfm_blk_index = (plane << 2) + (block >> (tx_size << 1));
if (x->skip_txfm[txfm_blk_index] == SKIP_TXFM_NONE) {
// full forward transform and quantization
- vp10_xform_quant(x, plane, block, blk_row, blk_col,
- plane_bsize, tx_size);
+ vp10_xform_quant(x, plane, block, blk_row, blk_col, plane_bsize,
+ tx_size);
} else if (x->skip_txfm[txfm_blk_index] == SKIP_TXFM_AC_ONLY) {
// fast path forward transform and quantization
- vp10_xform_quant_dc(x, plane, block, blk_row, blk_col,
- plane_bsize, tx_size);
+ vp10_xform_quant_dc(x, plane, block, blk_row, blk_col, plane_bsize,
+ tx_size);
} else {
// skip forward transform
p->eobs[block] = 0;
@@ -804,8 +741,8 @@
return;
}
} else {
- vp10_xform_quant(x, plane, block, blk_row, blk_col,
- plane_bsize, tx_size);
+ vp10_xform_quant(x, plane, block, blk_row, blk_col, plane_bsize,
+ tx_size);
}
}
}
@@ -817,11 +754,9 @@
*a = *l = p->eobs[block] > 0;
}
- if (p->eobs[block])
- *(args->skip) = 0;
+ if (p->eobs[block]) *(args->skip) = 0;
- if (p->eobs[block] == 0)
- return;
+ if (p->eobs[block] == 0) return;
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
switch (tx_size) {
@@ -845,9 +780,7 @@
p->eobs[block], xd->bd, tx_type,
xd->lossless[xd->mi[0]->mbmi.segment_id]);
break;
- default:
- assert(0 && "Invalid transform size");
- break;
+ default: assert(0 && "Invalid transform size"); break;
}
return;
@@ -874,15 +807,13 @@
vp10_inv_txfm_add_4x4(dqcoeff, dst, pd->dst.stride, p->eobs[block],
tx_type, xd->lossless[xd->mi[0]->mbmi.segment_id]);
break;
- default:
- assert(0 && "Invalid transform size");
- break;
+ default: assert(0 && "Invalid transform size"); break;
}
}
static void encode_block_pass1(int plane, int block, int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize,
- TX_SIZE tx_size, void *arg) {
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
+ void *arg) {
MACROBLOCK *const x = (MACROBLOCK *)arg;
MACROBLOCKD *const xd = &x->e_mbd;
struct macroblock_plane *const p = &x->plane[plane];
@@ -897,11 +828,11 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
if (xd->lossless[0]) {
- vp10_highbd_iwht4x4_add(dqcoeff, dst, pd->dst.stride,
- p->eobs[block], xd->bd);
+ vp10_highbd_iwht4x4_add(dqcoeff, dst, pd->dst.stride, p->eobs[block],
+ xd->bd);
} else {
- vp10_highbd_idct4x4_add(dqcoeff, dst, pd->dst.stride,
- p->eobs[block], xd->bd);
+ vp10_highbd_idct4x4_add(dqcoeff, dst, pd->dst.stride, p->eobs[block],
+ xd->bd);
}
return;
}
@@ -917,41 +848,39 @@
void vp10_encode_sby_pass1(MACROBLOCK *x, BLOCK_SIZE bsize) {
vp10_subtract_plane(x, bsize, 0);
vp10_foreach_transformed_block_in_plane(&x->e_mbd, bsize, 0,
- encode_block_pass1, x);
+ encode_block_pass1, x);
}
void vp10_encode_sb(MACROBLOCK *x, BLOCK_SIZE bsize) {
MACROBLOCKD *const xd = &x->e_mbd;
struct optimize_ctx ctx;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
- struct encode_b_args arg = {x, &ctx, &mbmi->skip};
+ struct encode_b_args arg = { x, &ctx, &mbmi->skip };
int plane;
mbmi->skip = 1;
- if (x->skip)
- return;
+ if (x->skip) return;
for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
- if (!x->skip_recode)
- vp10_subtract_plane(x, bsize, plane);
+ if (!x->skip_recode) vp10_subtract_plane(x, bsize, plane);
if (x->optimize && (!x->skip_recode || !x->skip_optimize)) {
- const struct macroblockd_plane* const pd = &xd->plane[plane];
+ const struct macroblockd_plane *const pd = &xd->plane[plane];
const TX_SIZE tx_size = plane ? get_uv_tx_size(mbmi, pd) : mbmi->tx_size;
- vp10_get_entropy_contexts(bsize, tx_size, pd,
- ctx.ta[plane], ctx.tl[plane]);
+ vp10_get_entropy_contexts(bsize, tx_size, pd, ctx.ta[plane],
+ ctx.tl[plane]);
}
vp10_foreach_transformed_block_in_plane(xd, bsize, plane, encode_block,
- &arg);
+ &arg);
}
}
void vp10_encode_block_intra(int plane, int block, int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize,
- TX_SIZE tx_size, void *arg) {
- struct encode_b_args* const args = arg;
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
+ void *arg) {
+ struct encode_b_args *const args = arg;
MACROBLOCK *const x = args->x;
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
@@ -977,16 +906,16 @@
src_diff = &p->src_diff[4 * (blk_row * diff_stride + blk_col)];
mode = plane == 0 ? get_y_mode(xd->mi[0], block) : mbmi->uv_mode;
- vp10_predict_intra_block(xd, bwl, bhl, tx_size, mode, dst, dst_stride,
- dst, dst_stride, blk_col, blk_row, plane);
+ vp10_predict_intra_block(xd, bwl, bhl, tx_size, mode, dst, dst_stride, dst,
+ dst_stride, blk_col, blk_row, plane);
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
switch (tx_size) {
case TX_32X32:
if (!x->skip_recode) {
- vpx_highbd_subtract_block(32, 32, src_diff, diff_stride,
- src, src_stride, dst, dst_stride, xd->bd);
+ vpx_highbd_subtract_block(32, 32, src_diff, diff_stride, src,
+ src_stride, dst, dst_stride, xd->bd);
highbd_fwd_txfm_32x32(x->use_lp32x32fdct, src_diff, coeff,
diff_stride, tx_type);
vpx_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin,
@@ -1000,13 +929,13 @@
break;
case TX_16X16:
if (!x->skip_recode) {
- vpx_highbd_subtract_block(16, 16, src_diff, diff_stride,
- src, src_stride, dst, dst_stride, xd->bd);
+ vpx_highbd_subtract_block(16, 16, src_diff, diff_stride, src,
+ src_stride, dst, dst_stride, xd->bd);
highbd_fwd_txfm_16x16(src_diff, coeff, diff_stride, tx_type);
vpx_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
}
if (*eob)
vp10_highbd_inv_txfm_add_16x16(dqcoeff, dst, dst_stride, *eob, xd->bd,
@@ -1014,13 +943,13 @@
break;
case TX_8X8:
if (!x->skip_recode) {
- vpx_highbd_subtract_block(8, 8, src_diff, diff_stride,
- src, src_stride, dst, dst_stride, xd->bd);
+ vpx_highbd_subtract_block(8, 8, src_diff, diff_stride, src,
+ src_stride, dst, dst_stride, xd->bd);
highbd_fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type);
vpx_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
}
if (*eob)
vp10_highbd_inv_txfm_add_8x8(dqcoeff, dst, dst_stride, *eob, xd->bd,
@@ -1028,14 +957,14 @@
break;
case TX_4X4:
if (!x->skip_recode) {
- vpx_highbd_subtract_block(4, 4, src_diff, diff_stride,
- src, src_stride, dst, dst_stride, xd->bd);
+ vpx_highbd_subtract_block(4, 4, src_diff, diff_stride, src,
+ src_stride, dst, dst_stride, xd->bd);
vp10_highbd_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
xd->lossless[mbmi->segment_id]);
vpx_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob,
- scan_order->scan, scan_order->iscan);
+ pd->dequant, eob, scan_order->scan,
+ scan_order->iscan);
}
if (*eob)
@@ -1045,12 +974,9 @@
vp10_highbd_inv_txfm_add_4x4(dqcoeff, dst, dst_stride, *eob, xd->bd,
tx_type, xd->lossless[mbmi->segment_id]);
break;
- default:
- assert(0);
- return;
+ default: assert(0); return;
}
- if (*eob)
- *(args->skip) = 0;
+ if (*eob) *(args->skip) = 0;
return;
}
#endif // CONFIG_VPX_HIGHBITDEPTH
@@ -1058,8 +984,8 @@
switch (tx_size) {
case TX_32X32:
if (!x->skip_recode) {
- vpx_subtract_block(32, 32, src_diff, diff_stride,
- src, src_stride, dst, dst_stride);
+ vpx_subtract_block(32, 32, src_diff, diff_stride, src, src_stride, dst,
+ dst_stride);
fwd_txfm_32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride,
tx_type);
vpx_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round,
@@ -1072,40 +998,36 @@
break;
case TX_16X16:
if (!x->skip_recode) {
- vpx_subtract_block(16, 16, src_diff, diff_stride,
- src, src_stride, dst, dst_stride);
+ vpx_subtract_block(16, 16, src_diff, diff_stride, src, src_stride, dst,
+ dst_stride);
fwd_txfm_16x16(src_diff, coeff, diff_stride, tx_type);
- vpx_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
- p->quant, p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob, scan_order->scan,
- scan_order->iscan);
+ vpx_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, p->quant,
+ p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob,
+ scan_order->scan, scan_order->iscan);
}
if (*eob)
vp10_inv_txfm_add_16x16(dqcoeff, dst, dst_stride, *eob, tx_type);
break;
case TX_8X8:
if (!x->skip_recode) {
- vpx_subtract_block(8, 8, src_diff, diff_stride,
- src, src_stride, dst, dst_stride);
+ vpx_subtract_block(8, 8, src_diff, diff_stride, src, src_stride, dst,
+ dst_stride);
fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type);
vpx_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, p->quant,
- p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob, scan_order->scan,
- scan_order->iscan);
+ p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob,
+ scan_order->scan, scan_order->iscan);
}
- if (*eob)
- vp10_inv_txfm_add_8x8(dqcoeff, dst, dst_stride, *eob, tx_type);
+ if (*eob) vp10_inv_txfm_add_8x8(dqcoeff, dst, dst_stride, *eob, tx_type);
break;
case TX_4X4:
if (!x->skip_recode) {
- vpx_subtract_block(4, 4, src_diff, diff_stride,
- src, src_stride, dst, dst_stride);
+ vpx_subtract_block(4, 4, src_diff, diff_stride, src, src_stride, dst,
+ dst_stride);
vp10_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
xd->lossless[mbmi->segment_id]);
vpx_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, p->quant,
- p->quant_shift, qcoeff, dqcoeff,
- pd->dequant, eob, scan_order->scan,
- scan_order->iscan);
+ p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob,
+ scan_order->scan, scan_order->iscan);
}
if (*eob) {
@@ -1116,17 +1038,14 @@
xd->lossless[mbmi->segment_id]);
}
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
- if (*eob)
- *(args->skip) = 0;
+ if (*eob) *(args->skip) = 0;
}
void vp10_encode_intra_block_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
const MACROBLOCKD *const xd = &x->e_mbd;
- struct encode_b_args arg = {x, NULL, &xd->mi[0]->mbmi.skip};
+ struct encode_b_args arg = { x, NULL, &xd->mi[0]->mbmi.skip };
vp10_foreach_transformed_block_in_plane(xd, bsize, plane,
vp10_encode_block_intra, &arg);
diff --git a/vp10/encoder/encodemb.h b/vp10/encoder/encodemb.h
index b3e6f63..4d1c826 100644
--- a/vp10/encoder/encodemb.h
+++ b/vp10/encoder/encodemb.h
@@ -25,21 +25,18 @@
};
void vp10_encode_sb(MACROBLOCK *x, BLOCK_SIZE bsize);
void vp10_encode_sby_pass1(MACROBLOCK *x, BLOCK_SIZE bsize);
-void vp10_xform_quant_fp(MACROBLOCK *x, int plane, int block,
- int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size);
-void vp10_xform_quant_dc(MACROBLOCK *x, int plane, int block,
- int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size);
-void vp10_xform_quant(MACROBLOCK *x, int plane, int block,
- int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size);
+void vp10_xform_quant_fp(MACROBLOCK *x, int plane, int block, int blk_row,
+ int blk_col, BLOCK_SIZE plane_bsize, TX_SIZE tx_size);
+void vp10_xform_quant_dc(MACROBLOCK *x, int plane, int block, int blk_row,
+ int blk_col, BLOCK_SIZE plane_bsize, TX_SIZE tx_size);
+void vp10_xform_quant(MACROBLOCK *x, int plane, int block, int blk_row,
+ int blk_col, BLOCK_SIZE plane_bsize, TX_SIZE tx_size);
void vp10_subtract_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane);
void vp10_encode_block_intra(int plane, int block, int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize,
- TX_SIZE tx_size, void *arg);
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
+ void *arg);
void vp10_encode_intra_block_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane);
diff --git a/vp10/encoder/encodemv.c b/vp10/encoder/encodemv.c
index 0736c65..ef670f5 100644
--- a/vp10/encoder/encodemv.c
+++ b/vp10/encoder/encodemv.c
@@ -31,15 +31,15 @@
vp10_tokens_from_tree(mv_fp_encodings, vp10_mv_fp_tree);
}
-static void encode_mv_component(vpx_writer* w, int comp,
- const nmv_component* mvcomp, int usehp) {
+static void encode_mv_component(vpx_writer *w, int comp,
+ const nmv_component *mvcomp, int usehp) {
int offset;
const int sign = comp < 0;
const int mag = sign ? -comp : comp;
const int mv_class = vp10_get_mv_class(mag - 1, &offset);
- const int d = offset >> 3; // int mv data
- const int fr = (offset >> 1) & 3; // fractional mv data
- const int hp = offset & 1; // high precision mv data
+ const int d = offset >> 3; // int mv data
+ const int fr = (offset >> 1) & 3; // fractional mv data
+ const int hp = offset & 1; // high precision mv data
assert(comp != 0);
@@ -48,33 +48,30 @@
// Class
vp10_write_token(w, vp10_mv_class_tree, mvcomp->classes,
- &mv_class_encodings[mv_class]);
+ &mv_class_encodings[mv_class]);
// Integer bits
if (mv_class == MV_CLASS_0) {
vp10_write_token(w, vp10_mv_class0_tree, mvcomp->class0,
- &mv_class0_encodings[d]);
+ &mv_class0_encodings[d]);
} else {
int i;
const int n = mv_class + CLASS0_BITS - 1; // number of bits
- for (i = 0; i < n; ++i)
- vpx_write(w, (d >> i) & 1, mvcomp->bits[i]);
+ for (i = 0; i < n; ++i) vpx_write(w, (d >> i) & 1, mvcomp->bits[i]);
}
// Fractional bits
vp10_write_token(w, vp10_mv_fp_tree,
- mv_class == MV_CLASS_0 ? mvcomp->class0_fp[d] : mvcomp->fp,
- &mv_fp_encodings[fr]);
+ mv_class == MV_CLASS_0 ? mvcomp->class0_fp[d] : mvcomp->fp,
+ &mv_fp_encodings[fr]);
// High precision bit
if (usehp)
- vpx_write(w, hp,
- mv_class == MV_CLASS_0 ? mvcomp->class0_hp : mvcomp->hp);
+ vpx_write(w, hp, mv_class == MV_CLASS_0 ? mvcomp->class0_hp : mvcomp->hp);
}
-
static void build_nmv_component_cost_table(int *mvcost,
- const nmv_component* const mvcomp,
+ const nmv_component *const mvcomp,
int usehp) {
int i, v;
int sign_cost[2], class_cost[MV_CLASSES], class0_cost[CLASS0_SIZE];
@@ -107,16 +104,15 @@
z = v - 1;
c = vp10_get_mv_class(z, &o);
cost += class_cost[c];
- d = (o >> 3); /* int mv data */
- f = (o >> 1) & 3; /* fractional pel mv data */
- e = (o & 1); /* high precision mv data */
+ d = (o >> 3); /* int mv data */
+ f = (o >> 1) & 3; /* fractional pel mv data */
+ e = (o & 1); /* high precision mv data */
if (c == MV_CLASS_0) {
cost += class0_cost[d];
} else {
int i, b;
- b = c + CLASS0_BITS - 1; /* number of bits */
- for (i = 0; i < b; ++i)
- cost += bits_cost[i][((d >> i) & 1)];
+ b = c + CLASS0_BITS - 1; /* number of bits */
+ for (i = 0; i < b; ++i) cost += bits_cost[i][((d >> i) & 1)];
}
if (c == MV_CLASS_0) {
cost += class0_fp_cost[d][f];
@@ -138,7 +134,7 @@
static void update_mv(vpx_writer *w, const unsigned int ct[2], vpx_prob *cur_p,
vpx_prob upd_p) {
#if CONFIG_MISC_FIXES
- (void) upd_p;
+ (void)upd_p;
vp10_cond_prob_diff_update(w, cur_p, ct);
#else
const vpx_prob new_p = get_binary_prob(ct[0], ct[1]) | 1;
@@ -154,8 +150,8 @@
static void write_mv_update(const vpx_tree_index *tree,
vpx_prob probs[/*n - 1*/],
- const unsigned int counts[/*n - 1*/],
- int n, vpx_writer *w) {
+ const unsigned int counts[/*n - 1*/], int n,
+ vpx_writer *w) {
int i;
unsigned int branch_ct[32][2];
@@ -168,11 +164,12 @@
}
void vp10_write_nmv_probs(VP10_COMMON *cm, int usehp, vpx_writer *w,
- nmv_context_counts *const counts) {
+ nmv_context_counts *const counts) {
int i, j;
nmv_context *const mvc = &cm->fc->nmvc;
- write_mv_update(vp10_mv_joint_tree, mvc->joints, counts->joints, MV_JOINTS, w);
+ write_mv_update(vp10_mv_joint_tree, mvc->joints, counts->joints, MV_JOINTS,
+ w);
for (i = 0; i < 2; ++i) {
nmv_component *comp = &mvc->comps[i];
@@ -205,15 +202,14 @@
}
}
-void vp10_encode_mv(VP10_COMP* cpi, vpx_writer* w,
- const MV* mv, const MV* ref,
- const nmv_context* mvctx, int usehp) {
- const MV diff = {mv->row - ref->row,
- mv->col - ref->col};
+void vp10_encode_mv(VP10_COMP *cpi, vpx_writer *w, const MV *mv, const MV *ref,
+ const nmv_context *mvctx, int usehp) {
+ const MV diff = { mv->row - ref->row, mv->col - ref->col };
const MV_JOINT_TYPE j = vp10_get_mv_joint(&diff);
usehp = usehp && vp10_use_mv_hp(ref);
- vp10_write_token(w, vp10_mv_joint_tree, mvctx->joints, &mv_joint_encodings[j]);
+ vp10_write_token(w, vp10_mv_joint_tree, mvctx->joints,
+ &mv_joint_encodings[j]);
if (mv_joint_vertical(j))
encode_mv_component(w, diff.row, &mvctx->comps[0], usehp);
@@ -229,21 +225,20 @@
}
void vp10_build_nmv_cost_table(int *mvjoint, int *mvcost[2],
- const nmv_context* ctx, int usehp) {
+ const nmv_context *ctx, int usehp) {
vp10_cost_tokens(mvjoint, ctx->joints, vp10_mv_joint_tree);
build_nmv_component_cost_table(mvcost[0], &ctx->comps[0], usehp);
build_nmv_component_cost_table(mvcost[1], &ctx->comps[1], usehp);
}
static void inc_mvs(const MB_MODE_INFO *mbmi, const MB_MODE_INFO_EXT *mbmi_ext,
- const int_mv mvs[2],
- nmv_context_counts *counts) {
+ const int_mv mvs[2], nmv_context_counts *counts) {
int i;
for (i = 0; i < 1 + has_second_ref(mbmi); ++i) {
const MV *ref = &mbmi_ext->ref_mvs[mbmi->ref_frame[i]][0].as_mv;
- const MV diff = {mvs[i].as_mv.row - ref->row,
- mvs[i].as_mv.col - ref->col};
+ const MV diff = { mvs[i].as_mv.row - ref->row,
+ mvs[i].as_mv.col - ref->col };
vp10_inc_mv(&diff, counts, vp10_use_mv_hp(ref));
}
}
@@ -267,8 +262,6 @@
}
}
} else {
- if (mbmi->mode == NEWMV)
- inc_mvs(mbmi, mbmi_ext, mbmi->mv, &td->counts->mv);
+ if (mbmi->mode == NEWMV) inc_mvs(mbmi, mbmi_ext, mbmi->mv, &td->counts->mv);
}
}
-
diff --git a/vp10/encoder/encodemv.h b/vp10/encoder/encodemv.h
index 006f6d7..6187488 100644
--- a/vp10/encoder/encodemv.h
+++ b/vp10/encoder/encodemv.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_ENCODEMV_H_
#define VP10_ENCODER_ENCODEMV_H_
@@ -21,13 +20,13 @@
void vp10_entropy_mv_init(void);
void vp10_write_nmv_probs(VP10_COMMON *cm, int usehp, vpx_writer *w,
- nmv_context_counts *const counts);
+ nmv_context_counts *const counts);
-void vp10_encode_mv(VP10_COMP *cpi, vpx_writer* w, const MV* mv, const MV* ref,
- const nmv_context* mvctx, int usehp);
+void vp10_encode_mv(VP10_COMP *cpi, vpx_writer *w, const MV *mv, const MV *ref,
+ const nmv_context *mvctx, int usehp);
void vp10_build_nmv_cost_table(int *mvjoint, int *mvcost[2],
- const nmv_context* mvctx, int usehp);
+ const nmv_context *mvctx, int usehp);
void vp10_update_mv_count(ThreadData *td);
diff --git a/vp10/encoder/encoder.c b/vp10/encoder/encoder.c
index 91c7e43..af68409 100644
--- a/vp10/encoder/encoder.c
+++ b/vp10/encoder/encoder.c
@@ -58,14 +58,14 @@
#define AM_SEGMENT_ID_INACTIVE 7
#define AM_SEGMENT_ID_ACTIVE 0
-#define SHARP_FILTER_QTHRESH 0 /* Q threshold for 8-tap sharp filter */
+#define SHARP_FILTER_QTHRESH 0 /* Q threshold for 8-tap sharp filter */
-#define ALTREF_HIGH_PRECISION_MV 1 // Whether to use high precision mv
- // for altref computation.
-#define HIGH_PRECISION_MV_QTHRESH 200 // Q threshold for high precision
- // mv. Choose a very high value for
- // now so that HIGH_PRECISION is always
- // chosen.
+#define ALTREF_HIGH_PRECISION_MV 1 // Whether to use high precision mv
+ // for altref computation.
+#define HIGH_PRECISION_MV_QTHRESH 200 // Q threshold for high precision
+ // mv. Choose a very high value for
+ // now so that HIGH_PRECISION is always
+ // chosen.
// #define OUTPUT_YUV_REC
#ifdef OUTPUT_YUV_DENOISED
@@ -97,15 +97,15 @@
case THREEFIVE:
*hr = 3;
*hs = 5;
- break;
+ break;
case ONETWO:
*hr = 1;
*hs = 2;
- break;
+ break;
default:
*hr = 1;
*hs = 1;
- assert(0);
+ assert(0);
break;
}
}
@@ -143,8 +143,8 @@
vp10_enable_segfeature(seg, AM_SEGMENT_ID_INACTIVE, SEG_LVL_ALT_LF);
// Setting the data to -MAX_LOOP_FILTER will result in the computed loop
// filter level being zero regardless of the value of seg->abs_delta.
- vp10_set_segdata(seg, AM_SEGMENT_ID_INACTIVE,
- SEG_LVL_ALT_LF, -MAX_LOOP_FILTER);
+ vp10_set_segdata(seg, AM_SEGMENT_ID_INACTIVE, SEG_LVL_ALT_LF,
+ -MAX_LOOP_FILTER);
} else {
vp10_disable_segfeature(seg, AM_SEGMENT_ID_INACTIVE, SEG_LVL_SKIP);
vp10_disable_segfeature(seg, AM_SEGMENT_ID_INACTIVE, SEG_LVL_ALT_LF);
@@ -157,10 +157,8 @@
}
}
-int vp10_set_active_map(VP10_COMP* cpi,
- unsigned char* new_map_16x16,
- int rows,
- int cols) {
+int vp10_set_active_map(VP10_COMP *cpi, unsigned char *new_map_16x16, int rows,
+ int cols) {
if (rows == cpi->common.mb_rows && cols == cpi->common.mb_cols) {
unsigned char *const active_map_8x8 = cpi->active_map.map;
const int mi_rows = cpi->common.mi_rows;
@@ -186,13 +184,11 @@
}
}
-int vp10_get_active_map(VP10_COMP* cpi,
- unsigned char* new_map_16x16,
- int rows,
- int cols) {
+int vp10_get_active_map(VP10_COMP *cpi, unsigned char *new_map_16x16, int rows,
+ int cols) {
if (rows == cpi->common.mb_rows && cols == cpi->common.mb_cols &&
new_map_16x16) {
- unsigned char* const seg_map_8x8 = cpi->segmentation_map;
+ unsigned char *const seg_map_8x8 = cpi->segmentation_map;
const int mi_rows = cpi->common.mi_rows;
const int mi_cols = cpi->common.mi_cols;
memset(new_map_16x16, !cpi->active_map.enabled, rows * cols);
@@ -268,19 +264,16 @@
static int vp10_enc_alloc_mi(VP10_COMMON *cm, int mi_size) {
cm->mip = vpx_calloc(mi_size, sizeof(*cm->mip));
- if (!cm->mip)
- return 1;
+ if (!cm->mip) return 1;
cm->prev_mip = vpx_calloc(mi_size, sizeof(*cm->prev_mip));
- if (!cm->prev_mip)
- return 1;
+ if (!cm->prev_mip) return 1;
cm->mi_alloc_size = mi_size;
- cm->mi_grid_base = (MODE_INFO **)vpx_calloc(mi_size, sizeof(MODE_INFO*));
- if (!cm->mi_grid_base)
- return 1;
- cm->prev_mi_grid_base = (MODE_INFO **)vpx_calloc(mi_size, sizeof(MODE_INFO*));
- if (!cm->prev_mi_grid_base)
- return 1;
+ cm->mi_grid_base = (MODE_INFO **)vpx_calloc(mi_size, sizeof(MODE_INFO *));
+ if (!cm->mi_grid_base) return 1;
+ cm->prev_mi_grid_base =
+ (MODE_INFO **)vpx_calloc(mi_size, sizeof(MODE_INFO *));
+ if (!cm->prev_mi_grid_base) return 1;
return 0;
}
@@ -399,7 +392,7 @@
// restored with a call to vp10_restore_coding_context. These functions are
// intended for use in a re-code loop in vp10_compress_frame where the
// quantizer value is adjusted between loop iterations.
- vp10_copy(cc->nmvjointcost, cpi->td.mb.nmvjointcost);
+ vp10_copy(cc->nmvjointcost, cpi->td.mb.nmvjointcost);
memcpy(cc->nmvcosts[0], cpi->nmvcosts[0],
MV_VALS * sizeof(*cpi->nmvcosts[0]));
@@ -414,8 +407,8 @@
vp10_copy(cc->segment_pred_probs, cm->segp.pred_probs);
#endif
- memcpy(cpi->coding_context.last_frame_seg_map_copy,
- cm->last_frame_seg_map, (cm->mi_rows * cm->mi_cols));
+ memcpy(cpi->coding_context.last_frame_seg_map_copy, cm->last_frame_seg_map,
+ (cm->mi_rows * cm->mi_cols));
vp10_copy(cc->last_ref_lf_deltas, cm->lf.last_ref_deltas);
vp10_copy(cc->last_mode_lf_deltas, cm->lf.last_mode_deltas);
@@ -442,8 +435,7 @@
vp10_copy(cm->segp.pred_probs, cc->segment_pred_probs);
#endif
- memcpy(cm->last_frame_seg_map,
- cpi->coding_context.last_frame_seg_map_copy,
+ memcpy(cm->last_frame_seg_map, cpi->coding_context.last_frame_seg_map_copy,
(cm->mi_rows * cm->mi_cols));
vp10_copy(cm->lf.last_ref_deltas, cc->last_ref_lf_deltas);
@@ -495,8 +487,8 @@
seg->update_map = 1;
seg->update_data = 1;
- qi_delta = vp10_compute_qdelta(rc, rc->avg_q, rc->avg_q * 0.875,
- cm->bit_depth);
+ qi_delta =
+ vp10_compute_qdelta(rc, rc->avg_q, rc->avg_q * 0.875, cm->bit_depth);
vp10_set_segdata(seg, 1, SEG_LVL_ALT_Q, qi_delta - 2);
vp10_set_segdata(seg, 1, SEG_LVL_ALT_LF, -2);
@@ -518,7 +510,7 @@
seg->abs_delta = SEGMENT_DELTADATA;
qi_delta = vp10_compute_qdelta(rc, rc->avg_q, rc->avg_q * 1.125,
- cm->bit_depth);
+ cm->bit_depth);
vp10_set_segdata(seg, 1, SEG_LVL_ALT_Q, qi_delta + 2);
vp10_enable_segfeature(seg, 1, SEG_LVL_ALT_Q);
@@ -598,18 +590,17 @@
if (!cpi->lookahead)
cpi->lookahead = vp10_lookahead_init(oxcf->width, oxcf->height,
- cm->subsampling_x, cm->subsampling_y,
+ cm->subsampling_x, cm->subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
- cm->use_highbitdepth,
+ cm->use_highbitdepth,
#endif
- oxcf->lag_in_frames);
+ oxcf->lag_in_frames);
if (!cpi->lookahead)
vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate lag buffers");
// TODO(agrange) Check if ARF is enabled and skip allocation if not.
- if (vpx_realloc_frame_buffer(&cpi->alt_ref_buffer,
- oxcf->width, oxcf->height,
+ if (vpx_realloc_frame_buffer(&cpi->alt_ref_buffer, oxcf->width, oxcf->height,
cm->subsampling_x, cm->subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
cm->use_highbitdepth,
@@ -622,8 +613,7 @@
static void alloc_util_frame_buffers(VP10_COMP *cpi) {
VP10_COMMON *const cm = &cpi->common;
- if (vpx_realloc_frame_buffer(&cpi->last_frame_uf,
- cm->width, cm->height,
+ if (vpx_realloc_frame_buffer(&cpi->last_frame_uf, cm->width, cm->height,
cm->subsampling_x, cm->subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
cm->use_highbitdepth,
@@ -633,8 +623,7 @@
vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate last frame buffer");
- if (vpx_realloc_frame_buffer(&cpi->scaled_source,
- cm->width, cm->height,
+ if (vpx_realloc_frame_buffer(&cpi->scaled_source, cm->width, cm->height,
cm->subsampling_x, cm->subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
cm->use_highbitdepth,
@@ -644,8 +633,7 @@
vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate scaled source buffer");
- if (vpx_realloc_frame_buffer(&cpi->scaled_last_source,
- cm->width, cm->height,
+ if (vpx_realloc_frame_buffer(&cpi->scaled_last_source, cm->width, cm->height,
cm->subsampling_x, cm->subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
cm->use_highbitdepth,
@@ -656,14 +644,12 @@
"Failed to allocate scaled last source buffer");
}
-
static int alloc_context_buffers_ext(VP10_COMP *cpi) {
VP10_COMMON *cm = &cpi->common;
int mi_size = cm->mi_cols * cm->mi_rows;
cpi->mbmi_ext_base = vpx_calloc(mi_size, sizeof(*cpi->mbmi_ext_base));
- if (!cpi->mbmi_ext_base)
- return 1;
+ if (!cpi->mbmi_ext_base) return 1;
return 0;
}
@@ -680,7 +666,7 @@
{
unsigned int tokens = get_token_alloc(cm->mb_rows, cm->mb_cols);
CHECK_MEM_ERROR(cm, cpi->tile_tok[0][0],
- vpx_calloc(tokens, sizeof(*cpi->tile_tok[0][0])));
+ vpx_calloc(tokens, sizeof(*cpi->tile_tok[0][0])));
}
vp10_setup_pc_tree(&cpi->common, &cpi->td);
@@ -697,8 +683,8 @@
int min_log2_tile_cols, max_log2_tile_cols;
vp10_get_tile_n_bits(cm->mi_cols, &min_log2_tile_cols, &max_log2_tile_cols);
- cm->log2_tile_cols = clamp(cpi->oxcf.tile_columns,
- min_log2_tile_cols, max_log2_tile_cols);
+ cm->log2_tile_cols =
+ clamp(cpi->oxcf.tile_columns, min_log2_tile_cols, max_log2_tile_cols);
cm->log2_tile_rows = cpi->oxcf.tile_rows;
}
@@ -759,154 +745,122 @@
const int64_t maximum = oxcf->maximum_buffer_size_ms;
rc->starting_buffer_level = starting * bandwidth / 1000;
- rc->optimal_buffer_level = (optimal == 0) ? bandwidth / 8
- : optimal * bandwidth / 1000;
- rc->maximum_buffer_size = (maximum == 0) ? bandwidth / 8
- : maximum * bandwidth / 1000;
+ rc->optimal_buffer_level =
+ (optimal == 0) ? bandwidth / 8 : optimal * bandwidth / 1000;
+ rc->maximum_buffer_size =
+ (maximum == 0) ? bandwidth / 8 : maximum * bandwidth / 1000;
}
#if CONFIG_VPX_HIGHBITDEPTH
#define HIGHBD_BFP(BT, SDF, SDAF, VF, SVF, SVAF, SDX3F, SDX8F, SDX4DF) \
- cpi->fn_ptr[BT].sdf = SDF; \
- cpi->fn_ptr[BT].sdaf = SDAF; \
- cpi->fn_ptr[BT].vf = VF; \
- cpi->fn_ptr[BT].svf = SVF; \
- cpi->fn_ptr[BT].svaf = SVAF; \
- cpi->fn_ptr[BT].sdx3f = SDX3F; \
- cpi->fn_ptr[BT].sdx8f = SDX8F; \
- cpi->fn_ptr[BT].sdx4df = SDX4DF;
+ cpi->fn_ptr[BT].sdf = SDF; \
+ cpi->fn_ptr[BT].sdaf = SDAF; \
+ cpi->fn_ptr[BT].vf = VF; \
+ cpi->fn_ptr[BT].svf = SVF; \
+ cpi->fn_ptr[BT].svaf = SVAF; \
+ cpi->fn_ptr[BT].sdx3f = SDX3F; \
+ cpi->fn_ptr[BT].sdx8f = SDX8F; \
+ cpi->fn_ptr[BT].sdx4df = SDX4DF;
-#define MAKE_BFP_SAD_WRAPPER(fnname) \
-static unsigned int fnname##_bits8(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride) { \
- return fnname(src_ptr, source_stride, ref_ptr, ref_stride); \
-} \
-static unsigned int fnname##_bits10(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride) { \
- return fnname(src_ptr, source_stride, ref_ptr, ref_stride) >> 2; \
-} \
-static unsigned int fnname##_bits12(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride) { \
- return fnname(src_ptr, source_stride, ref_ptr, ref_stride) >> 4; \
-}
+#define MAKE_BFP_SAD_WRAPPER(fnname) \
+ static unsigned int fnname##_bits8(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, int ref_stride) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride); \
+ } \
+ static unsigned int fnname##_bits10( \
+ const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, \
+ int ref_stride) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride) >> 2; \
+ } \
+ static unsigned int fnname##_bits12( \
+ const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, \
+ int ref_stride) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride) >> 4; \
+ }
-#define MAKE_BFP_SADAVG_WRAPPER(fnname) static unsigned int \
-fnname##_bits8(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- const uint8_t *second_pred) { \
- return fnname(src_ptr, source_stride, ref_ptr, ref_stride, second_pred); \
-} \
-static unsigned int fnname##_bits10(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- const uint8_t *second_pred) { \
- return fnname(src_ptr, source_stride, ref_ptr, ref_stride, \
- second_pred) >> 2; \
-} \
-static unsigned int fnname##_bits12(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- const uint8_t *second_pred) { \
- return fnname(src_ptr, source_stride, ref_ptr, ref_stride, \
- second_pred) >> 4; \
-}
+#define MAKE_BFP_SADAVG_WRAPPER(fnname) \
+ static unsigned int fnname##_bits8( \
+ const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, \
+ int ref_stride, const uint8_t *second_pred) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride, second_pred); \
+ } \
+ static unsigned int fnname##_bits10( \
+ const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, \
+ int ref_stride, const uint8_t *second_pred) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride, second_pred) >> \
+ 2; \
+ } \
+ static unsigned int fnname##_bits12( \
+ const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, \
+ int ref_stride, const uint8_t *second_pred) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride, second_pred) >> \
+ 4; \
+ }
-#define MAKE_BFP_SAD3_WRAPPER(fnname) \
-static void fnname##_bits8(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- unsigned int *sad_array) { \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
-} \
-static void fnname##_bits10(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- unsigned int *sad_array) { \
- int i; \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
- for (i = 0; i < 3; i++) \
- sad_array[i] >>= 2; \
-} \
-static void fnname##_bits12(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- unsigned int *sad_array) { \
- int i; \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
- for (i = 0; i < 3; i++) \
- sad_array[i] >>= 4; \
-}
+#define MAKE_BFP_SAD3_WRAPPER(fnname) \
+ static void fnname##_bits8(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *ref_ptr, int ref_stride, \
+ unsigned int *sad_array) { \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ } \
+ static void fnname##_bits10(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *ref_ptr, int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 3; i++) sad_array[i] >>= 2; \
+ } \
+ static void fnname##_bits12(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *ref_ptr, int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 3; i++) sad_array[i] >>= 4; \
+ }
-#define MAKE_BFP_SAD8_WRAPPER(fnname) \
-static void fnname##_bits8(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- unsigned int *sad_array) { \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
-} \
-static void fnname##_bits10(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- unsigned int *sad_array) { \
- int i; \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
- for (i = 0; i < 8; i++) \
- sad_array[i] >>= 2; \
-} \
-static void fnname##_bits12(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t *ref_ptr, \
- int ref_stride, \
- unsigned int *sad_array) { \
- int i; \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
- for (i = 0; i < 8; i++) \
- sad_array[i] >>= 4; \
-}
-#define MAKE_BFP_SAD4D_WRAPPER(fnname) \
-static void fnname##_bits8(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t* const ref_ptr[], \
- int ref_stride, \
- unsigned int *sad_array) { \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
-} \
-static void fnname##_bits10(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t* const ref_ptr[], \
- int ref_stride, \
- unsigned int *sad_array) { \
- int i; \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
- for (i = 0; i < 4; i++) \
- sad_array[i] >>= 2; \
-} \
-static void fnname##_bits12(const uint8_t *src_ptr, \
- int source_stride, \
- const uint8_t* const ref_ptr[], \
- int ref_stride, \
- unsigned int *sad_array) { \
- int i; \
- fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
- for (i = 0; i < 4; i++) \
- sad_array[i] >>= 4; \
-}
+#define MAKE_BFP_SAD8_WRAPPER(fnname) \
+ static void fnname##_bits8(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *ref_ptr, int ref_stride, \
+ unsigned int *sad_array) { \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ } \
+ static void fnname##_bits10(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *ref_ptr, int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 8; i++) sad_array[i] >>= 2; \
+ } \
+ static void fnname##_bits12(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *ref_ptr, int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 8; i++) sad_array[i] >>= 4; \
+ }
+#define MAKE_BFP_SAD4D_WRAPPER(fnname) \
+ static void fnname##_bits8(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *const ref_ptr[], int ref_stride, \
+ unsigned int *sad_array) { \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ } \
+ static void fnname##_bits10(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *const ref_ptr[], int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 4; i++) sad_array[i] >>= 2; \
+ } \
+ static void fnname##_bits12(const uint8_t *src_ptr, int source_stride, \
+ const uint8_t *const ref_ptr[], int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 4; i++) sad_array[i] >>= 4; \
+ }
+/* clang-format off */
MAKE_BFP_SAD_WRAPPER(vpx_highbd_sad32x16)
MAKE_BFP_SADAVG_WRAPPER(vpx_highbd_sad32x16_avg)
MAKE_BFP_SAD4D_WRAPPER(vpx_highbd_sad32x16x4d)
@@ -962,410 +916,269 @@
MAKE_BFP_SAD3_WRAPPER(vpx_highbd_sad4x4x3)
MAKE_BFP_SAD8_WRAPPER(vpx_highbd_sad4x4x8)
MAKE_BFP_SAD4D_WRAPPER(vpx_highbd_sad4x4x4d)
+/* clang-format on */
-static void highbd_set_var_fns(VP10_COMP *const cpi) {
+static void highbd_set_var_fns(VP10_COMP *const cpi) {
VP10_COMMON *const cm = &cpi->common;
if (cm->use_highbitdepth) {
switch (cm->bit_depth) {
case VPX_BITS_8:
- HIGHBD_BFP(BLOCK_32X16,
- vpx_highbd_sad32x16_bits8,
- vpx_highbd_sad32x16_avg_bits8,
- vpx_highbd_8_variance32x16,
+ HIGHBD_BFP(BLOCK_32X16, vpx_highbd_sad32x16_bits8,
+ vpx_highbd_sad32x16_avg_bits8, vpx_highbd_8_variance32x16,
vpx_highbd_8_sub_pixel_variance32x16,
- vpx_highbd_8_sub_pixel_avg_variance32x16,
- NULL,
- NULL,
+ vpx_highbd_8_sub_pixel_avg_variance32x16, NULL, NULL,
vpx_highbd_sad32x16x4d_bits8)
- HIGHBD_BFP(BLOCK_16X32,
- vpx_highbd_sad16x32_bits8,
- vpx_highbd_sad16x32_avg_bits8,
- vpx_highbd_8_variance16x32,
+ HIGHBD_BFP(BLOCK_16X32, vpx_highbd_sad16x32_bits8,
+ vpx_highbd_sad16x32_avg_bits8, vpx_highbd_8_variance16x32,
vpx_highbd_8_sub_pixel_variance16x32,
- vpx_highbd_8_sub_pixel_avg_variance16x32,
- NULL,
- NULL,
+ vpx_highbd_8_sub_pixel_avg_variance16x32, NULL, NULL,
vpx_highbd_sad16x32x4d_bits8)
- HIGHBD_BFP(BLOCK_64X32,
- vpx_highbd_sad64x32_bits8,
- vpx_highbd_sad64x32_avg_bits8,
- vpx_highbd_8_variance64x32,
+ HIGHBD_BFP(BLOCK_64X32, vpx_highbd_sad64x32_bits8,
+ vpx_highbd_sad64x32_avg_bits8, vpx_highbd_8_variance64x32,
vpx_highbd_8_sub_pixel_variance64x32,
- vpx_highbd_8_sub_pixel_avg_variance64x32,
- NULL,
- NULL,
+ vpx_highbd_8_sub_pixel_avg_variance64x32, NULL, NULL,
vpx_highbd_sad64x32x4d_bits8)
- HIGHBD_BFP(BLOCK_32X64,
- vpx_highbd_sad32x64_bits8,
- vpx_highbd_sad32x64_avg_bits8,
- vpx_highbd_8_variance32x64,
+ HIGHBD_BFP(BLOCK_32X64, vpx_highbd_sad32x64_bits8,
+ vpx_highbd_sad32x64_avg_bits8, vpx_highbd_8_variance32x64,
vpx_highbd_8_sub_pixel_variance32x64,
- vpx_highbd_8_sub_pixel_avg_variance32x64,
- NULL,
- NULL,
+ vpx_highbd_8_sub_pixel_avg_variance32x64, NULL, NULL,
vpx_highbd_sad32x64x4d_bits8)
- HIGHBD_BFP(BLOCK_32X32,
- vpx_highbd_sad32x32_bits8,
- vpx_highbd_sad32x32_avg_bits8,
- vpx_highbd_8_variance32x32,
+ HIGHBD_BFP(BLOCK_32X32, vpx_highbd_sad32x32_bits8,
+ vpx_highbd_sad32x32_avg_bits8, vpx_highbd_8_variance32x32,
vpx_highbd_8_sub_pixel_variance32x32,
vpx_highbd_8_sub_pixel_avg_variance32x32,
- vpx_highbd_sad32x32x3_bits8,
- vpx_highbd_sad32x32x8_bits8,
+ vpx_highbd_sad32x32x3_bits8, vpx_highbd_sad32x32x8_bits8,
vpx_highbd_sad32x32x4d_bits8)
- HIGHBD_BFP(BLOCK_64X64,
- vpx_highbd_sad64x64_bits8,
- vpx_highbd_sad64x64_avg_bits8,
- vpx_highbd_8_variance64x64,
+ HIGHBD_BFP(BLOCK_64X64, vpx_highbd_sad64x64_bits8,
+ vpx_highbd_sad64x64_avg_bits8, vpx_highbd_8_variance64x64,
vpx_highbd_8_sub_pixel_variance64x64,
vpx_highbd_8_sub_pixel_avg_variance64x64,
- vpx_highbd_sad64x64x3_bits8,
- vpx_highbd_sad64x64x8_bits8,
+ vpx_highbd_sad64x64x3_bits8, vpx_highbd_sad64x64x8_bits8,
vpx_highbd_sad64x64x4d_bits8)
- HIGHBD_BFP(BLOCK_16X16,
- vpx_highbd_sad16x16_bits8,
- vpx_highbd_sad16x16_avg_bits8,
- vpx_highbd_8_variance16x16,
+ HIGHBD_BFP(BLOCK_16X16, vpx_highbd_sad16x16_bits8,
+ vpx_highbd_sad16x16_avg_bits8, vpx_highbd_8_variance16x16,
vpx_highbd_8_sub_pixel_variance16x16,
vpx_highbd_8_sub_pixel_avg_variance16x16,
- vpx_highbd_sad16x16x3_bits8,
- vpx_highbd_sad16x16x8_bits8,
+ vpx_highbd_sad16x16x3_bits8, vpx_highbd_sad16x16x8_bits8,
vpx_highbd_sad16x16x4d_bits8)
- HIGHBD_BFP(BLOCK_16X8,
- vpx_highbd_sad16x8_bits8,
- vpx_highbd_sad16x8_avg_bits8,
- vpx_highbd_8_variance16x8,
- vpx_highbd_8_sub_pixel_variance16x8,
- vpx_highbd_8_sub_pixel_avg_variance16x8,
- vpx_highbd_sad16x8x3_bits8,
- vpx_highbd_sad16x8x8_bits8,
- vpx_highbd_sad16x8x4d_bits8)
+ HIGHBD_BFP(
+ BLOCK_16X8, vpx_highbd_sad16x8_bits8, vpx_highbd_sad16x8_avg_bits8,
+ vpx_highbd_8_variance16x8, vpx_highbd_8_sub_pixel_variance16x8,
+ vpx_highbd_8_sub_pixel_avg_variance16x8, vpx_highbd_sad16x8x3_bits8,
+ vpx_highbd_sad16x8x8_bits8, vpx_highbd_sad16x8x4d_bits8)
- HIGHBD_BFP(BLOCK_8X16,
- vpx_highbd_sad8x16_bits8,
- vpx_highbd_sad8x16_avg_bits8,
- vpx_highbd_8_variance8x16,
- vpx_highbd_8_sub_pixel_variance8x16,
- vpx_highbd_8_sub_pixel_avg_variance8x16,
- vpx_highbd_sad8x16x3_bits8,
- vpx_highbd_sad8x16x8_bits8,
- vpx_highbd_sad8x16x4d_bits8)
+ HIGHBD_BFP(
+ BLOCK_8X16, vpx_highbd_sad8x16_bits8, vpx_highbd_sad8x16_avg_bits8,
+ vpx_highbd_8_variance8x16, vpx_highbd_8_sub_pixel_variance8x16,
+ vpx_highbd_8_sub_pixel_avg_variance8x16, vpx_highbd_sad8x16x3_bits8,
+ vpx_highbd_sad8x16x8_bits8, vpx_highbd_sad8x16x4d_bits8)
- HIGHBD_BFP(BLOCK_8X8,
- vpx_highbd_sad8x8_bits8,
- vpx_highbd_sad8x8_avg_bits8,
- vpx_highbd_8_variance8x8,
- vpx_highbd_8_sub_pixel_variance8x8,
- vpx_highbd_8_sub_pixel_avg_variance8x8,
- vpx_highbd_sad8x8x3_bits8,
- vpx_highbd_sad8x8x8_bits8,
- vpx_highbd_sad8x8x4d_bits8)
+ HIGHBD_BFP(
+ BLOCK_8X8, vpx_highbd_sad8x8_bits8, vpx_highbd_sad8x8_avg_bits8,
+ vpx_highbd_8_variance8x8, vpx_highbd_8_sub_pixel_variance8x8,
+ vpx_highbd_8_sub_pixel_avg_variance8x8, vpx_highbd_sad8x8x3_bits8,
+ vpx_highbd_sad8x8x8_bits8, vpx_highbd_sad8x8x4d_bits8)
- HIGHBD_BFP(BLOCK_8X4,
- vpx_highbd_sad8x4_bits8,
- vpx_highbd_sad8x4_avg_bits8,
- vpx_highbd_8_variance8x4,
+ HIGHBD_BFP(BLOCK_8X4, vpx_highbd_sad8x4_bits8,
+ vpx_highbd_sad8x4_avg_bits8, vpx_highbd_8_variance8x4,
vpx_highbd_8_sub_pixel_variance8x4,
- vpx_highbd_8_sub_pixel_avg_variance8x4,
- NULL,
- vpx_highbd_sad8x4x8_bits8,
- vpx_highbd_sad8x4x4d_bits8)
+ vpx_highbd_8_sub_pixel_avg_variance8x4, NULL,
+ vpx_highbd_sad8x4x8_bits8, vpx_highbd_sad8x4x4d_bits8)
- HIGHBD_BFP(BLOCK_4X8,
- vpx_highbd_sad4x8_bits8,
- vpx_highbd_sad4x8_avg_bits8,
- vpx_highbd_8_variance4x8,
+ HIGHBD_BFP(BLOCK_4X8, vpx_highbd_sad4x8_bits8,
+ vpx_highbd_sad4x8_avg_bits8, vpx_highbd_8_variance4x8,
vpx_highbd_8_sub_pixel_variance4x8,
- vpx_highbd_8_sub_pixel_avg_variance4x8,
- NULL,
- vpx_highbd_sad4x8x8_bits8,
- vpx_highbd_sad4x8x4d_bits8)
+ vpx_highbd_8_sub_pixel_avg_variance4x8, NULL,
+ vpx_highbd_sad4x8x8_bits8, vpx_highbd_sad4x8x4d_bits8)
- HIGHBD_BFP(BLOCK_4X4,
- vpx_highbd_sad4x4_bits8,
- vpx_highbd_sad4x4_avg_bits8,
- vpx_highbd_8_variance4x4,
- vpx_highbd_8_sub_pixel_variance4x4,
- vpx_highbd_8_sub_pixel_avg_variance4x4,
- vpx_highbd_sad4x4x3_bits8,
- vpx_highbd_sad4x4x8_bits8,
- vpx_highbd_sad4x4x4d_bits8)
+ HIGHBD_BFP(
+ BLOCK_4X4, vpx_highbd_sad4x4_bits8, vpx_highbd_sad4x4_avg_bits8,
+ vpx_highbd_8_variance4x4, vpx_highbd_8_sub_pixel_variance4x4,
+ vpx_highbd_8_sub_pixel_avg_variance4x4, vpx_highbd_sad4x4x3_bits8,
+ vpx_highbd_sad4x4x8_bits8, vpx_highbd_sad4x4x4d_bits8)
break;
case VPX_BITS_10:
- HIGHBD_BFP(BLOCK_32X16,
- vpx_highbd_sad32x16_bits10,
- vpx_highbd_sad32x16_avg_bits10,
- vpx_highbd_10_variance32x16,
+ HIGHBD_BFP(BLOCK_32X16, vpx_highbd_sad32x16_bits10,
+ vpx_highbd_sad32x16_avg_bits10, vpx_highbd_10_variance32x16,
vpx_highbd_10_sub_pixel_variance32x16,
- vpx_highbd_10_sub_pixel_avg_variance32x16,
- NULL,
- NULL,
+ vpx_highbd_10_sub_pixel_avg_variance32x16, NULL, NULL,
vpx_highbd_sad32x16x4d_bits10)
- HIGHBD_BFP(BLOCK_16X32,
- vpx_highbd_sad16x32_bits10,
- vpx_highbd_sad16x32_avg_bits10,
- vpx_highbd_10_variance16x32,
+ HIGHBD_BFP(BLOCK_16X32, vpx_highbd_sad16x32_bits10,
+ vpx_highbd_sad16x32_avg_bits10, vpx_highbd_10_variance16x32,
vpx_highbd_10_sub_pixel_variance16x32,
- vpx_highbd_10_sub_pixel_avg_variance16x32,
- NULL,
- NULL,
+ vpx_highbd_10_sub_pixel_avg_variance16x32, NULL, NULL,
vpx_highbd_sad16x32x4d_bits10)
- HIGHBD_BFP(BLOCK_64X32,
- vpx_highbd_sad64x32_bits10,
- vpx_highbd_sad64x32_avg_bits10,
- vpx_highbd_10_variance64x32,
+ HIGHBD_BFP(BLOCK_64X32, vpx_highbd_sad64x32_bits10,
+ vpx_highbd_sad64x32_avg_bits10, vpx_highbd_10_variance64x32,
vpx_highbd_10_sub_pixel_variance64x32,
- vpx_highbd_10_sub_pixel_avg_variance64x32,
- NULL,
- NULL,
+ vpx_highbd_10_sub_pixel_avg_variance64x32, NULL, NULL,
vpx_highbd_sad64x32x4d_bits10)
- HIGHBD_BFP(BLOCK_32X64,
- vpx_highbd_sad32x64_bits10,
- vpx_highbd_sad32x64_avg_bits10,
- vpx_highbd_10_variance32x64,
+ HIGHBD_BFP(BLOCK_32X64, vpx_highbd_sad32x64_bits10,
+ vpx_highbd_sad32x64_avg_bits10, vpx_highbd_10_variance32x64,
vpx_highbd_10_sub_pixel_variance32x64,
- vpx_highbd_10_sub_pixel_avg_variance32x64,
- NULL,
- NULL,
+ vpx_highbd_10_sub_pixel_avg_variance32x64, NULL, NULL,
vpx_highbd_sad32x64x4d_bits10)
- HIGHBD_BFP(BLOCK_32X32,
- vpx_highbd_sad32x32_bits10,
- vpx_highbd_sad32x32_avg_bits10,
- vpx_highbd_10_variance32x32,
+ HIGHBD_BFP(BLOCK_32X32, vpx_highbd_sad32x32_bits10,
+ vpx_highbd_sad32x32_avg_bits10, vpx_highbd_10_variance32x32,
vpx_highbd_10_sub_pixel_variance32x32,
vpx_highbd_10_sub_pixel_avg_variance32x32,
- vpx_highbd_sad32x32x3_bits10,
- vpx_highbd_sad32x32x8_bits10,
+ vpx_highbd_sad32x32x3_bits10, vpx_highbd_sad32x32x8_bits10,
vpx_highbd_sad32x32x4d_bits10)
- HIGHBD_BFP(BLOCK_64X64,
- vpx_highbd_sad64x64_bits10,
- vpx_highbd_sad64x64_avg_bits10,
- vpx_highbd_10_variance64x64,
+ HIGHBD_BFP(BLOCK_64X64, vpx_highbd_sad64x64_bits10,
+ vpx_highbd_sad64x64_avg_bits10, vpx_highbd_10_variance64x64,
vpx_highbd_10_sub_pixel_variance64x64,
vpx_highbd_10_sub_pixel_avg_variance64x64,
- vpx_highbd_sad64x64x3_bits10,
- vpx_highbd_sad64x64x8_bits10,
+ vpx_highbd_sad64x64x3_bits10, vpx_highbd_sad64x64x8_bits10,
vpx_highbd_sad64x64x4d_bits10)
- HIGHBD_BFP(BLOCK_16X16,
- vpx_highbd_sad16x16_bits10,
- vpx_highbd_sad16x16_avg_bits10,
- vpx_highbd_10_variance16x16,
+ HIGHBD_BFP(BLOCK_16X16, vpx_highbd_sad16x16_bits10,
+ vpx_highbd_sad16x16_avg_bits10, vpx_highbd_10_variance16x16,
vpx_highbd_10_sub_pixel_variance16x16,
vpx_highbd_10_sub_pixel_avg_variance16x16,
- vpx_highbd_sad16x16x3_bits10,
- vpx_highbd_sad16x16x8_bits10,
+ vpx_highbd_sad16x16x3_bits10, vpx_highbd_sad16x16x8_bits10,
vpx_highbd_sad16x16x4d_bits10)
- HIGHBD_BFP(BLOCK_16X8,
- vpx_highbd_sad16x8_bits10,
- vpx_highbd_sad16x8_avg_bits10,
- vpx_highbd_10_variance16x8,
+ HIGHBD_BFP(BLOCK_16X8, vpx_highbd_sad16x8_bits10,
+ vpx_highbd_sad16x8_avg_bits10, vpx_highbd_10_variance16x8,
vpx_highbd_10_sub_pixel_variance16x8,
vpx_highbd_10_sub_pixel_avg_variance16x8,
- vpx_highbd_sad16x8x3_bits10,
- vpx_highbd_sad16x8x8_bits10,
+ vpx_highbd_sad16x8x3_bits10, vpx_highbd_sad16x8x8_bits10,
vpx_highbd_sad16x8x4d_bits10)
- HIGHBD_BFP(BLOCK_8X16,
- vpx_highbd_sad8x16_bits10,
- vpx_highbd_sad8x16_avg_bits10,
- vpx_highbd_10_variance8x16,
+ HIGHBD_BFP(BLOCK_8X16, vpx_highbd_sad8x16_bits10,
+ vpx_highbd_sad8x16_avg_bits10, vpx_highbd_10_variance8x16,
vpx_highbd_10_sub_pixel_variance8x16,
vpx_highbd_10_sub_pixel_avg_variance8x16,
- vpx_highbd_sad8x16x3_bits10,
- vpx_highbd_sad8x16x8_bits10,
+ vpx_highbd_sad8x16x3_bits10, vpx_highbd_sad8x16x8_bits10,
vpx_highbd_sad8x16x4d_bits10)
- HIGHBD_BFP(BLOCK_8X8,
- vpx_highbd_sad8x8_bits10,
- vpx_highbd_sad8x8_avg_bits10,
- vpx_highbd_10_variance8x8,
- vpx_highbd_10_sub_pixel_variance8x8,
- vpx_highbd_10_sub_pixel_avg_variance8x8,
- vpx_highbd_sad8x8x3_bits10,
- vpx_highbd_sad8x8x8_bits10,
- vpx_highbd_sad8x8x4d_bits10)
+ HIGHBD_BFP(
+ BLOCK_8X8, vpx_highbd_sad8x8_bits10, vpx_highbd_sad8x8_avg_bits10,
+ vpx_highbd_10_variance8x8, vpx_highbd_10_sub_pixel_variance8x8,
+ vpx_highbd_10_sub_pixel_avg_variance8x8, vpx_highbd_sad8x8x3_bits10,
+ vpx_highbd_sad8x8x8_bits10, vpx_highbd_sad8x8x4d_bits10)
- HIGHBD_BFP(BLOCK_8X4,
- vpx_highbd_sad8x4_bits10,
- vpx_highbd_sad8x4_avg_bits10,
- vpx_highbd_10_variance8x4,
+ HIGHBD_BFP(BLOCK_8X4, vpx_highbd_sad8x4_bits10,
+ vpx_highbd_sad8x4_avg_bits10, vpx_highbd_10_variance8x4,
vpx_highbd_10_sub_pixel_variance8x4,
- vpx_highbd_10_sub_pixel_avg_variance8x4,
- NULL,
- vpx_highbd_sad8x4x8_bits10,
- vpx_highbd_sad8x4x4d_bits10)
+ vpx_highbd_10_sub_pixel_avg_variance8x4, NULL,
+ vpx_highbd_sad8x4x8_bits10, vpx_highbd_sad8x4x4d_bits10)
- HIGHBD_BFP(BLOCK_4X8,
- vpx_highbd_sad4x8_bits10,
- vpx_highbd_sad4x8_avg_bits10,
- vpx_highbd_10_variance4x8,
+ HIGHBD_BFP(BLOCK_4X8, vpx_highbd_sad4x8_bits10,
+ vpx_highbd_sad4x8_avg_bits10, vpx_highbd_10_variance4x8,
vpx_highbd_10_sub_pixel_variance4x8,
- vpx_highbd_10_sub_pixel_avg_variance4x8,
- NULL,
- vpx_highbd_sad4x8x8_bits10,
- vpx_highbd_sad4x8x4d_bits10)
+ vpx_highbd_10_sub_pixel_avg_variance4x8, NULL,
+ vpx_highbd_sad4x8x8_bits10, vpx_highbd_sad4x8x4d_bits10)
- HIGHBD_BFP(BLOCK_4X4,
- vpx_highbd_sad4x4_bits10,
- vpx_highbd_sad4x4_avg_bits10,
- vpx_highbd_10_variance4x4,
- vpx_highbd_10_sub_pixel_variance4x4,
- vpx_highbd_10_sub_pixel_avg_variance4x4,
- vpx_highbd_sad4x4x3_bits10,
- vpx_highbd_sad4x4x8_bits10,
- vpx_highbd_sad4x4x4d_bits10)
+ HIGHBD_BFP(
+ BLOCK_4X4, vpx_highbd_sad4x4_bits10, vpx_highbd_sad4x4_avg_bits10,
+ vpx_highbd_10_variance4x4, vpx_highbd_10_sub_pixel_variance4x4,
+ vpx_highbd_10_sub_pixel_avg_variance4x4, vpx_highbd_sad4x4x3_bits10,
+ vpx_highbd_sad4x4x8_bits10, vpx_highbd_sad4x4x4d_bits10)
break;
case VPX_BITS_12:
- HIGHBD_BFP(BLOCK_32X16,
- vpx_highbd_sad32x16_bits12,
- vpx_highbd_sad32x16_avg_bits12,
- vpx_highbd_12_variance32x16,
+ HIGHBD_BFP(BLOCK_32X16, vpx_highbd_sad32x16_bits12,
+ vpx_highbd_sad32x16_avg_bits12, vpx_highbd_12_variance32x16,
vpx_highbd_12_sub_pixel_variance32x16,
- vpx_highbd_12_sub_pixel_avg_variance32x16,
- NULL,
- NULL,
+ vpx_highbd_12_sub_pixel_avg_variance32x16, NULL, NULL,
vpx_highbd_sad32x16x4d_bits12)
- HIGHBD_BFP(BLOCK_16X32,
- vpx_highbd_sad16x32_bits12,
- vpx_highbd_sad16x32_avg_bits12,
- vpx_highbd_12_variance16x32,
+ HIGHBD_BFP(BLOCK_16X32, vpx_highbd_sad16x32_bits12,
+ vpx_highbd_sad16x32_avg_bits12, vpx_highbd_12_variance16x32,
vpx_highbd_12_sub_pixel_variance16x32,
- vpx_highbd_12_sub_pixel_avg_variance16x32,
- NULL,
- NULL,
+ vpx_highbd_12_sub_pixel_avg_variance16x32, NULL, NULL,
vpx_highbd_sad16x32x4d_bits12)
- HIGHBD_BFP(BLOCK_64X32,
- vpx_highbd_sad64x32_bits12,
- vpx_highbd_sad64x32_avg_bits12,
- vpx_highbd_12_variance64x32,
+ HIGHBD_BFP(BLOCK_64X32, vpx_highbd_sad64x32_bits12,
+ vpx_highbd_sad64x32_avg_bits12, vpx_highbd_12_variance64x32,
vpx_highbd_12_sub_pixel_variance64x32,
- vpx_highbd_12_sub_pixel_avg_variance64x32,
- NULL,
- NULL,
+ vpx_highbd_12_sub_pixel_avg_variance64x32, NULL, NULL,
vpx_highbd_sad64x32x4d_bits12)
- HIGHBD_BFP(BLOCK_32X64,
- vpx_highbd_sad32x64_bits12,
- vpx_highbd_sad32x64_avg_bits12,
- vpx_highbd_12_variance32x64,
+ HIGHBD_BFP(BLOCK_32X64, vpx_highbd_sad32x64_bits12,
+ vpx_highbd_sad32x64_avg_bits12, vpx_highbd_12_variance32x64,
vpx_highbd_12_sub_pixel_variance32x64,
- vpx_highbd_12_sub_pixel_avg_variance32x64,
- NULL,
- NULL,
+ vpx_highbd_12_sub_pixel_avg_variance32x64, NULL, NULL,
vpx_highbd_sad32x64x4d_bits12)
- HIGHBD_BFP(BLOCK_32X32,
- vpx_highbd_sad32x32_bits12,
- vpx_highbd_sad32x32_avg_bits12,
- vpx_highbd_12_variance32x32,
+ HIGHBD_BFP(BLOCK_32X32, vpx_highbd_sad32x32_bits12,
+ vpx_highbd_sad32x32_avg_bits12, vpx_highbd_12_variance32x32,
vpx_highbd_12_sub_pixel_variance32x32,
vpx_highbd_12_sub_pixel_avg_variance32x32,
- vpx_highbd_sad32x32x3_bits12,
- vpx_highbd_sad32x32x8_bits12,
+ vpx_highbd_sad32x32x3_bits12, vpx_highbd_sad32x32x8_bits12,
vpx_highbd_sad32x32x4d_bits12)
- HIGHBD_BFP(BLOCK_64X64,
- vpx_highbd_sad64x64_bits12,
- vpx_highbd_sad64x64_avg_bits12,
- vpx_highbd_12_variance64x64,
+ HIGHBD_BFP(BLOCK_64X64, vpx_highbd_sad64x64_bits12,
+ vpx_highbd_sad64x64_avg_bits12, vpx_highbd_12_variance64x64,
vpx_highbd_12_sub_pixel_variance64x64,
vpx_highbd_12_sub_pixel_avg_variance64x64,
- vpx_highbd_sad64x64x3_bits12,
- vpx_highbd_sad64x64x8_bits12,
+ vpx_highbd_sad64x64x3_bits12, vpx_highbd_sad64x64x8_bits12,
vpx_highbd_sad64x64x4d_bits12)
- HIGHBD_BFP(BLOCK_16X16,
- vpx_highbd_sad16x16_bits12,
- vpx_highbd_sad16x16_avg_bits12,
- vpx_highbd_12_variance16x16,
+ HIGHBD_BFP(BLOCK_16X16, vpx_highbd_sad16x16_bits12,
+ vpx_highbd_sad16x16_avg_bits12, vpx_highbd_12_variance16x16,
vpx_highbd_12_sub_pixel_variance16x16,
vpx_highbd_12_sub_pixel_avg_variance16x16,
- vpx_highbd_sad16x16x3_bits12,
- vpx_highbd_sad16x16x8_bits12,
+ vpx_highbd_sad16x16x3_bits12, vpx_highbd_sad16x16x8_bits12,
vpx_highbd_sad16x16x4d_bits12)
- HIGHBD_BFP(BLOCK_16X8,
- vpx_highbd_sad16x8_bits12,
- vpx_highbd_sad16x8_avg_bits12,
- vpx_highbd_12_variance16x8,
+ HIGHBD_BFP(BLOCK_16X8, vpx_highbd_sad16x8_bits12,
+ vpx_highbd_sad16x8_avg_bits12, vpx_highbd_12_variance16x8,
vpx_highbd_12_sub_pixel_variance16x8,
vpx_highbd_12_sub_pixel_avg_variance16x8,
- vpx_highbd_sad16x8x3_bits12,
- vpx_highbd_sad16x8x8_bits12,
+ vpx_highbd_sad16x8x3_bits12, vpx_highbd_sad16x8x8_bits12,
vpx_highbd_sad16x8x4d_bits12)
- HIGHBD_BFP(BLOCK_8X16,
- vpx_highbd_sad8x16_bits12,
- vpx_highbd_sad8x16_avg_bits12,
- vpx_highbd_12_variance8x16,
+ HIGHBD_BFP(BLOCK_8X16, vpx_highbd_sad8x16_bits12,
+ vpx_highbd_sad8x16_avg_bits12, vpx_highbd_12_variance8x16,
vpx_highbd_12_sub_pixel_variance8x16,
vpx_highbd_12_sub_pixel_avg_variance8x16,
- vpx_highbd_sad8x16x3_bits12,
- vpx_highbd_sad8x16x8_bits12,
+ vpx_highbd_sad8x16x3_bits12, vpx_highbd_sad8x16x8_bits12,
vpx_highbd_sad8x16x4d_bits12)
- HIGHBD_BFP(BLOCK_8X8,
- vpx_highbd_sad8x8_bits12,
- vpx_highbd_sad8x8_avg_bits12,
- vpx_highbd_12_variance8x8,
- vpx_highbd_12_sub_pixel_variance8x8,
- vpx_highbd_12_sub_pixel_avg_variance8x8,
- vpx_highbd_sad8x8x3_bits12,
- vpx_highbd_sad8x8x8_bits12,
- vpx_highbd_sad8x8x4d_bits12)
+ HIGHBD_BFP(
+ BLOCK_8X8, vpx_highbd_sad8x8_bits12, vpx_highbd_sad8x8_avg_bits12,
+ vpx_highbd_12_variance8x8, vpx_highbd_12_sub_pixel_variance8x8,
+ vpx_highbd_12_sub_pixel_avg_variance8x8, vpx_highbd_sad8x8x3_bits12,
+ vpx_highbd_sad8x8x8_bits12, vpx_highbd_sad8x8x4d_bits12)
- HIGHBD_BFP(BLOCK_8X4,
- vpx_highbd_sad8x4_bits12,
- vpx_highbd_sad8x4_avg_bits12,
- vpx_highbd_12_variance8x4,
+ HIGHBD_BFP(BLOCK_8X4, vpx_highbd_sad8x4_bits12,
+ vpx_highbd_sad8x4_avg_bits12, vpx_highbd_12_variance8x4,
vpx_highbd_12_sub_pixel_variance8x4,
- vpx_highbd_12_sub_pixel_avg_variance8x4,
- NULL,
- vpx_highbd_sad8x4x8_bits12,
- vpx_highbd_sad8x4x4d_bits12)
+ vpx_highbd_12_sub_pixel_avg_variance8x4, NULL,
+ vpx_highbd_sad8x4x8_bits12, vpx_highbd_sad8x4x4d_bits12)
- HIGHBD_BFP(BLOCK_4X8,
- vpx_highbd_sad4x8_bits12,
- vpx_highbd_sad4x8_avg_bits12,
- vpx_highbd_12_variance4x8,
+ HIGHBD_BFP(BLOCK_4X8, vpx_highbd_sad4x8_bits12,
+ vpx_highbd_sad4x8_avg_bits12, vpx_highbd_12_variance4x8,
vpx_highbd_12_sub_pixel_variance4x8,
- vpx_highbd_12_sub_pixel_avg_variance4x8,
- NULL,
- vpx_highbd_sad4x8x8_bits12,
- vpx_highbd_sad4x8x4d_bits12)
+ vpx_highbd_12_sub_pixel_avg_variance4x8, NULL,
+ vpx_highbd_sad4x8x8_bits12, vpx_highbd_sad4x8x4d_bits12)
- HIGHBD_BFP(BLOCK_4X4,
- vpx_highbd_sad4x4_bits12,
- vpx_highbd_sad4x4_avg_bits12,
- vpx_highbd_12_variance4x4,
- vpx_highbd_12_sub_pixel_variance4x4,
- vpx_highbd_12_sub_pixel_avg_variance4x4,
- vpx_highbd_sad4x4x3_bits12,
- vpx_highbd_sad4x4x8_bits12,
- vpx_highbd_sad4x4x4d_bits12)
+ HIGHBD_BFP(
+ BLOCK_4X4, vpx_highbd_sad4x4_bits12, vpx_highbd_sad4x4_avg_bits12,
+ vpx_highbd_12_variance4x4, vpx_highbd_12_sub_pixel_variance4x4,
+ vpx_highbd_12_sub_pixel_avg_variance4x4, vpx_highbd_sad4x4x3_bits12,
+ vpx_highbd_sad4x4x8_bits12, vpx_highbd_sad4x4x4d_bits12)
break;
default:
- assert(0 && "cm->bit_depth should be VPX_BITS_8, "
- "VPX_BITS_10 or VPX_BITS_12");
+ assert(0 &&
+ "cm->bit_depth should be VPX_BITS_8, "
+ "VPX_BITS_10 or VPX_BITS_12");
}
}
}
@@ -1380,8 +1193,7 @@
vpx_calloc(cm->mi_rows * cm->mi_cols, 1));
// Create a map used for cyclic background refresh.
- if (cpi->cyclic_refresh)
- vp10_cyclic_refresh_free(cpi->cyclic_refresh);
+ if (cpi->cyclic_refresh) vp10_cyclic_refresh_free(cpi->cyclic_refresh);
CHECK_MEM_ERROR(cm, cpi->cyclic_refresh,
vp10_cyclic_refresh_alloc(cm->mi_rows, cm->mi_cols));
@@ -1401,8 +1213,7 @@
VP10_COMMON *const cm = &cpi->common;
RATE_CONTROL *const rc = &cpi->rc;
- if (cm->profile != oxcf->profile)
- cm->profile = oxcf->profile;
+ if (cm->profile != oxcf->profile) cm->profile = oxcf->profile;
cm->bit_depth = oxcf->bit_depth;
cm->color_space = oxcf->color_space;
cm->color_range = oxcf->color_range;
@@ -1425,10 +1236,11 @@
cpi->refresh_golden_frame = 0;
cpi->refresh_last_frame = 1;
- cm->refresh_frame_context =
- oxcf->error_resilient_mode ? REFRESH_FRAME_CONTEXT_OFF :
- oxcf->frame_parallel_decoding_mode ? REFRESH_FRAME_CONTEXT_FORWARD
- : REFRESH_FRAME_CONTEXT_BACKWARD;
+ cm->refresh_frame_context = oxcf->error_resilient_mode
+ ? REFRESH_FRAME_CONTEXT_OFF
+ : oxcf->frame_parallel_decoding_mode
+ ? REFRESH_FRAME_CONTEXT_FORWARD
+ : REFRESH_FRAME_CONTEXT_BACKWARD;
cm->reset_frame_context = RESET_FRAME_CONTEXT_NONE;
vp10_reset_segment_features(cm);
@@ -1500,7 +1312,7 @@
#ifndef M_LOG2_E
#define M_LOG2_E 0.693147180559945309417
#endif
-#define log2f(x) (log (x) / (float) M_LOG2_E)
+#define log2f(x) (log(x) / (float)M_LOG2_E)
static void cal_nmvjointsadcost(int *mvjointsadcost) {
mvjointsadcost[0] = 600;
@@ -1539,15 +1351,13 @@
} while (++i <= MV_MAX);
}
-
VP10_COMP *vp10_create_compressor(VP10EncoderConfig *oxcf,
- BufferPool *const pool) {
+ BufferPool *const pool) {
unsigned int i;
VP10_COMP *volatile const cpi = vpx_memalign(32, sizeof(VP10_COMP));
VP10_COMMON *volatile const cm = cpi != NULL ? &cpi->common : NULL;
- if (!cm)
- return NULL;
+ if (!cm) return NULL;
vp10_zero(*cpi);
@@ -1562,11 +1372,10 @@
cm->free_mi = vp10_enc_free_mi;
cm->setup_mi = vp10_enc_setup_mi;
- CHECK_MEM_ERROR(cm, cm->fc,
- (FRAME_CONTEXT *)vpx_calloc(1, sizeof(*cm->fc)));
- CHECK_MEM_ERROR(cm, cm->frame_contexts,
- (FRAME_CONTEXT *)vpx_calloc(FRAME_CONTEXTS,
- sizeof(*cm->frame_contexts)));
+ CHECK_MEM_ERROR(cm, cm->fc, (FRAME_CONTEXT *)vpx_calloc(1, sizeof(*cm->fc)));
+ CHECK_MEM_ERROR(
+ cm, cm->frame_contexts,
+ (FRAME_CONTEXT *)vpx_calloc(FRAME_CONTEXTS, sizeof(*cm->frame_contexts)));
cpi->resize_state = 0;
cpi->resize_avg_qp = 0;
@@ -1599,11 +1408,11 @@
CHECK_MEM_ERROR(cm, cpi->nmvsadcosts_hp[1],
vpx_calloc(MV_VALS, sizeof(*cpi->nmvsadcosts_hp[1])));
- for (i = 0; i < (sizeof(cpi->mbgraph_stats) /
- sizeof(cpi->mbgraph_stats[0])); i++) {
- CHECK_MEM_ERROR(cm, cpi->mbgraph_stats[i].mb_stats,
- vpx_calloc(cm->MBs *
- sizeof(*cpi->mbgraph_stats[i].mb_stats), 1));
+ for (i = 0; i < (sizeof(cpi->mbgraph_stats) / sizeof(cpi->mbgraph_stats[0]));
+ i++) {
+ CHECK_MEM_ERROR(
+ cm, cpi->mbgraph_stats[i].mb_stats,
+ vpx_calloc(cm->MBs * sizeof(*cpi->mbgraph_stats[i].mb_stats), 1));
}
#if CONFIG_FP_MB_STATS
@@ -1647,7 +1456,7 @@
}
if (cpi->b_calculate_ssimg) {
- cpi->ssimg.worst= 100.0;
+ cpi->ssimg.worst = 100.0;
}
cpi->fastssim.worst = 100.0;
@@ -1659,8 +1468,8 @@
}
if (cpi->b_calculate_consistency) {
- cpi->ssim_vars = vpx_malloc(sizeof(*cpi->ssim_vars) *
- 4 * cpi->common.mi_rows * cpi->common.mi_cols);
+ cpi->ssim_vars = vpx_malloc(sizeof(*cpi->ssim_vars) * 4 *
+ cpi->common.mi_rows * cpi->common.mi_cols);
cpi->worst_consistency = 100.0;
}
@@ -1725,79 +1534,71 @@
vp10_set_speed_features_framesize_dependent(cpi);
// Allocate memory to store variances for a frame.
- CHECK_MEM_ERROR(cm, cpi->source_diff_var,
- vpx_calloc(cm->MBs, sizeof(diff)));
+ CHECK_MEM_ERROR(cm, cpi->source_diff_var, vpx_calloc(cm->MBs, sizeof(diff)));
cpi->source_var_thresh = 0;
cpi->frames_till_next_var_check = 0;
-#define BFP(BT, SDF, SDAF, VF, SVF, SVAF, SDX3F, SDX8F, SDX4DF)\
- cpi->fn_ptr[BT].sdf = SDF; \
- cpi->fn_ptr[BT].sdaf = SDAF; \
- cpi->fn_ptr[BT].vf = VF; \
- cpi->fn_ptr[BT].svf = SVF; \
- cpi->fn_ptr[BT].svaf = SVAF; \
- cpi->fn_ptr[BT].sdx3f = SDX3F; \
- cpi->fn_ptr[BT].sdx8f = SDX8F; \
- cpi->fn_ptr[BT].sdx4df = SDX4DF;
+#define BFP(BT, SDF, SDAF, VF, SVF, SVAF, SDX3F, SDX8F, SDX4DF) \
+ cpi->fn_ptr[BT].sdf = SDF; \
+ cpi->fn_ptr[BT].sdaf = SDAF; \
+ cpi->fn_ptr[BT].vf = VF; \
+ cpi->fn_ptr[BT].svf = SVF; \
+ cpi->fn_ptr[BT].svaf = SVAF; \
+ cpi->fn_ptr[BT].sdx3f = SDX3F; \
+ cpi->fn_ptr[BT].sdx8f = SDX8F; \
+ cpi->fn_ptr[BT].sdx4df = SDX4DF;
- BFP(BLOCK_32X16, vpx_sad32x16, vpx_sad32x16_avg,
- vpx_variance32x16, vpx_sub_pixel_variance32x16,
- vpx_sub_pixel_avg_variance32x16, NULL, NULL, vpx_sad32x16x4d)
+ BFP(BLOCK_32X16, vpx_sad32x16, vpx_sad32x16_avg, vpx_variance32x16,
+ vpx_sub_pixel_variance32x16, vpx_sub_pixel_avg_variance32x16, NULL, NULL,
+ vpx_sad32x16x4d)
- BFP(BLOCK_16X32, vpx_sad16x32, vpx_sad16x32_avg,
- vpx_variance16x32, vpx_sub_pixel_variance16x32,
- vpx_sub_pixel_avg_variance16x32, NULL, NULL, vpx_sad16x32x4d)
+ BFP(BLOCK_16X32, vpx_sad16x32, vpx_sad16x32_avg, vpx_variance16x32,
+ vpx_sub_pixel_variance16x32, vpx_sub_pixel_avg_variance16x32, NULL, NULL,
+ vpx_sad16x32x4d)
- BFP(BLOCK_64X32, vpx_sad64x32, vpx_sad64x32_avg,
- vpx_variance64x32, vpx_sub_pixel_variance64x32,
- vpx_sub_pixel_avg_variance64x32, NULL, NULL, vpx_sad64x32x4d)
+ BFP(BLOCK_64X32, vpx_sad64x32, vpx_sad64x32_avg, vpx_variance64x32,
+ vpx_sub_pixel_variance64x32, vpx_sub_pixel_avg_variance64x32, NULL, NULL,
+ vpx_sad64x32x4d)
- BFP(BLOCK_32X64, vpx_sad32x64, vpx_sad32x64_avg,
- vpx_variance32x64, vpx_sub_pixel_variance32x64,
- vpx_sub_pixel_avg_variance32x64, NULL, NULL, vpx_sad32x64x4d)
+ BFP(BLOCK_32X64, vpx_sad32x64, vpx_sad32x64_avg, vpx_variance32x64,
+ vpx_sub_pixel_variance32x64, vpx_sub_pixel_avg_variance32x64, NULL, NULL,
+ vpx_sad32x64x4d)
- BFP(BLOCK_32X32, vpx_sad32x32, vpx_sad32x32_avg,
- vpx_variance32x32, vpx_sub_pixel_variance32x32,
- vpx_sub_pixel_avg_variance32x32, vpx_sad32x32x3, vpx_sad32x32x8,
- vpx_sad32x32x4d)
+ BFP(BLOCK_32X32, vpx_sad32x32, vpx_sad32x32_avg, vpx_variance32x32,
+ vpx_sub_pixel_variance32x32, vpx_sub_pixel_avg_variance32x32,
+ vpx_sad32x32x3, vpx_sad32x32x8, vpx_sad32x32x4d)
- BFP(BLOCK_64X64, vpx_sad64x64, vpx_sad64x64_avg,
- vpx_variance64x64, vpx_sub_pixel_variance64x64,
- vpx_sub_pixel_avg_variance64x64, vpx_sad64x64x3, vpx_sad64x64x8,
- vpx_sad64x64x4d)
+ BFP(BLOCK_64X64, vpx_sad64x64, vpx_sad64x64_avg, vpx_variance64x64,
+ vpx_sub_pixel_variance64x64, vpx_sub_pixel_avg_variance64x64,
+ vpx_sad64x64x3, vpx_sad64x64x8, vpx_sad64x64x4d)
- BFP(BLOCK_16X16, vpx_sad16x16, vpx_sad16x16_avg,
- vpx_variance16x16, vpx_sub_pixel_variance16x16,
- vpx_sub_pixel_avg_variance16x16, vpx_sad16x16x3, vpx_sad16x16x8,
- vpx_sad16x16x4d)
+ BFP(BLOCK_16X16, vpx_sad16x16, vpx_sad16x16_avg, vpx_variance16x16,
+ vpx_sub_pixel_variance16x16, vpx_sub_pixel_avg_variance16x16,
+ vpx_sad16x16x3, vpx_sad16x16x8, vpx_sad16x16x4d)
- BFP(BLOCK_16X8, vpx_sad16x8, vpx_sad16x8_avg,
- vpx_variance16x8, vpx_sub_pixel_variance16x8,
- vpx_sub_pixel_avg_variance16x8,
- vpx_sad16x8x3, vpx_sad16x8x8, vpx_sad16x8x4d)
+ BFP(BLOCK_16X8, vpx_sad16x8, vpx_sad16x8_avg, vpx_variance16x8,
+ vpx_sub_pixel_variance16x8, vpx_sub_pixel_avg_variance16x8, vpx_sad16x8x3,
+ vpx_sad16x8x8, vpx_sad16x8x4d)
- BFP(BLOCK_8X16, vpx_sad8x16, vpx_sad8x16_avg,
- vpx_variance8x16, vpx_sub_pixel_variance8x16,
- vpx_sub_pixel_avg_variance8x16,
- vpx_sad8x16x3, vpx_sad8x16x8, vpx_sad8x16x4d)
+ BFP(BLOCK_8X16, vpx_sad8x16, vpx_sad8x16_avg, vpx_variance8x16,
+ vpx_sub_pixel_variance8x16, vpx_sub_pixel_avg_variance8x16, vpx_sad8x16x3,
+ vpx_sad8x16x8, vpx_sad8x16x4d)
- BFP(BLOCK_8X8, vpx_sad8x8, vpx_sad8x8_avg,
- vpx_variance8x8, vpx_sub_pixel_variance8x8,
- vpx_sub_pixel_avg_variance8x8,
- vpx_sad8x8x3, vpx_sad8x8x8, vpx_sad8x8x4d)
+ BFP(BLOCK_8X8, vpx_sad8x8, vpx_sad8x8_avg, vpx_variance8x8,
+ vpx_sub_pixel_variance8x8, vpx_sub_pixel_avg_variance8x8, vpx_sad8x8x3,
+ vpx_sad8x8x8, vpx_sad8x8x4d)
- BFP(BLOCK_8X4, vpx_sad8x4, vpx_sad8x4_avg,
- vpx_variance8x4, vpx_sub_pixel_variance8x4,
- vpx_sub_pixel_avg_variance8x4, NULL, vpx_sad8x4x8, vpx_sad8x4x4d)
+ BFP(BLOCK_8X4, vpx_sad8x4, vpx_sad8x4_avg, vpx_variance8x4,
+ vpx_sub_pixel_variance8x4, vpx_sub_pixel_avg_variance8x4, NULL,
+ vpx_sad8x4x8, vpx_sad8x4x4d)
- BFP(BLOCK_4X8, vpx_sad4x8, vpx_sad4x8_avg,
- vpx_variance4x8, vpx_sub_pixel_variance4x8,
- vpx_sub_pixel_avg_variance4x8, NULL, vpx_sad4x8x8, vpx_sad4x8x4d)
+ BFP(BLOCK_4X8, vpx_sad4x8, vpx_sad4x8_avg, vpx_variance4x8,
+ vpx_sub_pixel_variance4x8, vpx_sub_pixel_avg_variance4x8, NULL,
+ vpx_sad4x8x8, vpx_sad4x8x4d)
- BFP(BLOCK_4X4, vpx_sad4x4, vpx_sad4x4_avg,
- vpx_variance4x4, vpx_sub_pixel_variance4x4,
- vpx_sub_pixel_avg_variance4x4,
- vpx_sad4x4x3, vpx_sad4x4x8, vpx_sad4x4x4d)
+ BFP(BLOCK_4X4, vpx_sad4x4, vpx_sad4x4_avg, vpx_variance4x4,
+ vpx_sub_pixel_variance4x4, vpx_sub_pixel_avg_variance4x4, vpx_sad4x4x3,
+ vpx_sad4x4x8, vpx_sad4x4x4d)
#if CONFIG_VPX_HIGHBITDEPTH
highbd_set_var_fns(cpi);
@@ -1816,8 +1617,7 @@
return cpi;
}
-#define SNPRINT(H, T) \
- snprintf((H) + strlen(H), sizeof(H) - strlen(H), (T))
+#define SNPRINT(H, T) snprintf((H) + strlen(H), sizeof(H) - strlen(H), (T))
#define SNPRINT2(H, T, V) \
snprintf((H) + strlen(H), sizeof(H) - strlen(H), (T), (V))
@@ -1827,8 +1627,7 @@
unsigned int i;
int t;
- if (!cpi)
- return;
+ if (!cpi) return;
cm = &cpi->common;
if (cm->current_video_frame > 0) {
@@ -1836,28 +1635,27 @@
vpx_clear_system_state();
if (cpi->oxcf.pass != 1) {
- char headings[512] = {0};
- char results[512] = {0};
+ char headings[512] = { 0 };
+ char results[512] = { 0 };
FILE *f = fopen("opsnr.stt", "a");
- double time_encoded = (cpi->last_end_time_stamp_seen
- - cpi->first_time_stamp_ever) / 10000000.000;
- double total_encode_time = (cpi->time_receive_data +
- cpi->time_compress_data) / 1000.000;
+ double time_encoded =
+ (cpi->last_end_time_stamp_seen - cpi->first_time_stamp_ever) /
+ 10000000.000;
+ double total_encode_time =
+ (cpi->time_receive_data + cpi->time_compress_data) / 1000.000;
const double dr =
- (double)cpi->bytes * (double) 8 / (double)1000 / time_encoded;
+ (double)cpi->bytes * (double)8 / (double)1000 / time_encoded;
const double peak = (double)((1 << cpi->oxcf.input_bit_depth) - 1);
if (cpi->b_calculate_psnr) {
- const double total_psnr =
- vpx_sse_to_psnr((double)cpi->total_samples, peak,
- (double)cpi->total_sq_error);
- const double totalp_psnr =
- vpx_sse_to_psnr((double)cpi->totalp_samples, peak,
- (double)cpi->totalp_sq_error);
- const double total_ssim = 100 * pow(cpi->summed_quality /
- cpi->summed_weights, 8.0);
- const double totalp_ssim = 100 * pow(cpi->summedp_quality /
- cpi->summedp_weights, 8.0);
+ const double total_psnr = vpx_sse_to_psnr(
+ (double)cpi->total_samples, peak, (double)cpi->total_sq_error);
+ const double totalp_psnr = vpx_sse_to_psnr(
+ (double)cpi->totalp_samples, peak, (double)cpi->totalp_sq_error);
+ const double total_ssim =
+ 100 * pow(cpi->summed_quality / cpi->summed_weights, 8.0);
+ const double totalp_ssim =
+ 100 * pow(cpi->summedp_quality / cpi->summedp_weights, 8.0);
snprintf(headings, sizeof(headings),
"Bitrate\tAVGPsnr\tGLBPsnr\tAVPsnrP\tGLPsnrP\t"
@@ -1868,12 +1666,10 @@
"%7.3f\t%7.3f\t%7.3f\t%7.3f\t"
"%7.3f\t%7.3f\t%7.3f\t%7.3f",
dr, cpi->psnr.stat[ALL] / cpi->count, total_psnr,
- cpi->psnrp.stat[ALL] / cpi->count, totalp_psnr,
- total_ssim, totalp_ssim,
- cpi->fastssim.stat[ALL] / cpi->count,
- cpi->psnrhvs.stat[ALL] / cpi->count,
- cpi->psnr.worst, cpi->worst_ssim, cpi->fastssim.worst,
- cpi->psnrhvs.worst);
+ cpi->psnrp.stat[ALL] / cpi->count, totalp_psnr, total_ssim,
+ totalp_ssim, cpi->fastssim.stat[ALL] / cpi->count,
+ cpi->psnrhvs.stat[ALL] / cpi->count, cpi->psnr.worst,
+ cpi->worst_ssim, cpi->fastssim.worst, cpi->psnrhvs.worst);
if (cpi->b_calculate_blockiness) {
SNPRINT(headings, "\t Block\tWstBlck");
@@ -1935,13 +1731,12 @@
vpx_free(cpi->tile_thr_data);
vpx_free(cpi->workers);
- if (cpi->num_workers > 1)
- vp10_loop_filter_dealloc(&cpi->lf_row_sync);
+ if (cpi->num_workers > 1) vp10_loop_filter_dealloc(&cpi->lf_row_sync);
dealloc_compressor_data(cpi);
- for (i = 0; i < sizeof(cpi->mbgraph_stats) /
- sizeof(cpi->mbgraph_stats[0]); ++i) {
+ for (i = 0; i < sizeof(cpi->mbgraph_stats) / sizeof(cpi->mbgraph_stats[0]);
+ ++i) {
vpx_free(cpi->mbgraph_stats[i].mb_stats);
}
@@ -1980,9 +1775,9 @@
/* TODO(yaowu): The block_variance calls the unoptimized versions of variance()
* and highbd_8_variance(). It should not.
*/
-static void encoder_variance(const uint8_t *a, int a_stride,
- const uint8_t *b, int b_stride,
- int w, int h, unsigned int *sse, int *sum) {
+static void encoder_variance(const uint8_t *a, int a_stride, const uint8_t *b,
+ int b_stride, int w, int h, unsigned int *sse,
+ int *sum) {
int i, j;
*sum = 0;
@@ -2001,10 +1796,9 @@
}
#if CONFIG_VPX_HIGHBITDEPTH
-static void encoder_highbd_variance64(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h, uint64_t *sse,
- uint64_t *sum) {
+static void encoder_highbd_variance64(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride, int w,
+ int h, uint64_t *sse, uint64_t *sum) {
int i, j;
uint16_t *a = CONVERT_TO_SHORTPTR(a8);
@@ -2023,22 +1817,20 @@
}
}
-static void encoder_highbd_8_variance(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h,
- unsigned int *sse, int *sum) {
+static void encoder_highbd_8_variance(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride, int w,
+ int h, unsigned int *sse, int *sum) {
uint64_t sse_long = 0;
uint64_t sum_long = 0;
- encoder_highbd_variance64(a8, a_stride, b8, b_stride, w, h,
- &sse_long, &sum_long);
+ encoder_highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long,
+ &sum_long);
*sse = (unsigned int)sse_long;
*sum = (int)sum_long;
}
#endif // CONFIG_VPX_HIGHBITDEPTH
-static int64_t get_sse(const uint8_t *a, int a_stride,
- const uint8_t *b, int b_stride,
- int width, int height) {
+static int64_t get_sse(const uint8_t *a, int a_stride, const uint8_t *b,
+ int b_stride, int width, int height) {
const int dw = width % 16;
const int dh = height % 16;
int64_t total_sse = 0;
@@ -2047,15 +1839,15 @@
int x, y;
if (dw > 0) {
- encoder_variance(&a[width - dw], a_stride, &b[width - dw], b_stride,
- dw, height, &sse, &sum);
+ encoder_variance(&a[width - dw], a_stride, &b[width - dw], b_stride, dw,
+ height, &sse, &sum);
total_sse += sse;
}
if (dh > 0) {
encoder_variance(&a[(height - dh) * a_stride], a_stride,
- &b[(height - dh) * b_stride], b_stride,
- width - dw, dh, &sse, &sum);
+ &b[(height - dh) * b_stride], b_stride, width - dw, dh,
+ &sse, &sum);
total_sse += sse;
}
@@ -2079,9 +1871,8 @@
#if CONFIG_VPX_HIGHBITDEPTH
static int64_t highbd_get_sse_shift(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int width, int height,
- unsigned int input_shift) {
+ const uint8_t *b8, int b_stride, int width,
+ int height, unsigned int input_shift) {
const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
const uint16_t *b = CONVERT_TO_SHORTPTR(b8);
int64_t total_sse = 0;
@@ -2098,9 +1889,8 @@
return total_sse;
}
-static int64_t highbd_get_sse(const uint8_t *a, int a_stride,
- const uint8_t *b, int b_stride,
- int width, int height) {
+static int64_t highbd_get_sse(const uint8_t *a, int a_stride, const uint8_t *b,
+ int b_stride, int width, int height) {
int64_t total_sse = 0;
int x, y;
const int dw = width % 16;
@@ -2108,9 +1898,8 @@
unsigned int sse = 0;
int sum = 0;
if (dw > 0) {
- encoder_highbd_8_variance(&a[width - dw], a_stride,
- &b[width - dw], b_stride,
- dw, height, &sse, &sum);
+ encoder_highbd_8_variance(&a[width - dw], a_stride, &b[width - dw],
+ b_stride, dw, height, &sse, &sum);
total_sse += sse;
}
if (dh > 0) {
@@ -2143,18 +1932,16 @@
#if CONFIG_VPX_HIGHBITDEPTH
static void calc_highbd_psnr(const YV12_BUFFER_CONFIG *a,
- const YV12_BUFFER_CONFIG *b,
- PSNR_STATS *psnr,
+ const YV12_BUFFER_CONFIG *b, PSNR_STATS *psnr,
unsigned int bit_depth,
unsigned int in_bit_depth) {
- const int widths[3] =
- {a->y_crop_width, a->uv_crop_width, a->uv_crop_width };
- const int heights[3] =
- {a->y_crop_height, a->uv_crop_height, a->uv_crop_height};
- const uint8_t *a_planes[3] = {a->y_buffer, a->u_buffer, a->v_buffer };
- const int a_strides[3] = {a->y_stride, a->uv_stride, a->uv_stride};
- const uint8_t *b_planes[3] = {b->y_buffer, b->u_buffer, b->v_buffer };
- const int b_strides[3] = {b->y_stride, b->uv_stride, b->uv_stride};
+ const int widths[3] = { a->y_crop_width, a->uv_crop_width, a->uv_crop_width };
+ const int heights[3] = { a->y_crop_height, a->uv_crop_height,
+ a->uv_crop_height };
+ const uint8_t *a_planes[3] = { a->y_buffer, a->u_buffer, a->v_buffer };
+ const int a_strides[3] = { a->y_stride, a->uv_stride, a->uv_stride };
+ const uint8_t *b_planes[3] = { b->y_buffer, b->u_buffer, b->v_buffer };
+ const int b_strides[3] = { b->y_stride, b->uv_stride, b->uv_stride };
int i;
uint64_t total_sse = 0;
uint32_t total_samples = 0;
@@ -2168,17 +1955,14 @@
uint64_t sse;
if (a->flags & YV12_FLAG_HIGHBITDEPTH) {
if (input_shift) {
- sse = highbd_get_sse_shift(a_planes[i], a_strides[i],
- b_planes[i], b_strides[i], w, h,
- input_shift);
+ sse = highbd_get_sse_shift(a_planes[i], a_strides[i], b_planes[i],
+ b_strides[i], w, h, input_shift);
} else {
- sse = highbd_get_sse(a_planes[i], a_strides[i],
- b_planes[i], b_strides[i], w, h);
+ sse = highbd_get_sse(a_planes[i], a_strides[i], b_planes[i],
+ b_strides[i], w, h);
}
} else {
- sse = get_sse(a_planes[i], a_strides[i],
- b_planes[i], b_strides[i],
- w, h);
+ sse = get_sse(a_planes[i], a_strides[i], b_planes[i], b_strides[i], w, h);
}
psnr->sse[1 + i] = sse;
psnr->samples[1 + i] = samples;
@@ -2190,23 +1974,22 @@
psnr->sse[0] = total_sse;
psnr->samples[0] = total_samples;
- psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
- (double)total_sse);
+ psnr->psnr[0] =
+ vpx_sse_to_psnr((double)total_samples, peak, (double)total_sse);
}
-#else // !CONFIG_VPX_HIGHBITDEPTH
+#else // !CONFIG_VPX_HIGHBITDEPTH
static void calc_psnr(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b,
PSNR_STATS *psnr) {
static const double peak = 255.0;
- const int widths[3] = {
- a->y_crop_width, a->uv_crop_width, a->uv_crop_width};
- const int heights[3] = {
- a->y_crop_height, a->uv_crop_height, a->uv_crop_height};
- const uint8_t *a_planes[3] = {a->y_buffer, a->u_buffer, a->v_buffer};
- const int a_strides[3] = {a->y_stride, a->uv_stride, a->uv_stride};
- const uint8_t *b_planes[3] = {b->y_buffer, b->u_buffer, b->v_buffer};
- const int b_strides[3] = {b->y_stride, b->uv_stride, b->uv_stride};
+ const int widths[3] = { a->y_crop_width, a->uv_crop_width, a->uv_crop_width };
+ const int heights[3] = { a->y_crop_height, a->uv_crop_height,
+ a->uv_crop_height };
+ const uint8_t *a_planes[3] = { a->y_buffer, a->u_buffer, a->v_buffer };
+ const int a_strides[3] = { a->y_stride, a->uv_stride, a->uv_stride };
+ const uint8_t *b_planes[3] = { b->y_buffer, b->u_buffer, b->v_buffer };
+ const int b_strides[3] = { b->y_stride, b->uv_stride, b->uv_stride };
int i;
uint64_t total_sse = 0;
uint32_t total_samples = 0;
@@ -2215,9 +1998,8 @@
const int w = widths[i];
const int h = heights[i];
const uint32_t samples = w * h;
- const uint64_t sse = get_sse(a_planes[i], a_strides[i],
- b_planes[i], b_strides[i],
- w, h);
+ const uint64_t sse =
+ get_sse(a_planes[i], a_strides[i], b_planes[i], b_strides[i], w, h);
psnr->sse[1 + i] = sse;
psnr->samples[1 + i] = samples;
psnr->psnr[1 + i] = vpx_sse_to_psnr(samples, peak, (double)sse);
@@ -2228,8 +2010,8 @@
psnr->sse[0] = total_sse;
psnr->samples[0] = total_samples;
- psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
- (double)total_sse);
+ psnr->psnr[0] =
+ vpx_sse_to_psnr((double)total_samples, peak, (double)total_sse);
}
#endif // CONFIG_VPX_HIGHBITDEPTH
@@ -2254,8 +2036,7 @@
}
int vp10_use_as_reference(VP10_COMP *cpi, int ref_frame_flags) {
- if (ref_frame_flags > 7)
- return -1;
+ if (ref_frame_flags > 7) return -1;
cpi->ref_frame_flags = ref_frame_flags;
return 0;
@@ -2268,8 +2049,8 @@
cpi->ext_refresh_frame_flags_pending = 1;
}
-static YV12_BUFFER_CONFIG *get_vp10_ref_frame_buffer(VP10_COMP *cpi,
- VPX_REFFRAME ref_frame_flag) {
+static YV12_BUFFER_CONFIG *get_vp10_ref_frame_buffer(
+ VP10_COMP *cpi, VPX_REFFRAME ref_frame_flag) {
MV_REFERENCE_FRAME ref_frame = NONE;
if (ref_frame_flag == VPX_LAST_FLAG)
ref_frame = LAST_FRAME;
@@ -2282,7 +2063,7 @@
}
int vp10_copy_reference_enc(VP10_COMP *cpi, VPX_REFFRAME ref_frame_flag,
- YV12_BUFFER_CONFIG *sd) {
+ YV12_BUFFER_CONFIG *sd) {
YV12_BUFFER_CONFIG *cfg = get_vp10_ref_frame_buffer(cpi, ref_frame_flag);
if (cfg) {
vpx_yv12_copy_frame(cfg, sd);
@@ -2293,7 +2074,7 @@
}
int vp10_set_reference_enc(VP10_COMP *cpi, VPX_REFFRAME ref_frame_flag,
- YV12_BUFFER_CONFIG *sd) {
+ YV12_BUFFER_CONFIG *sd) {
YV12_BUFFER_CONFIG *cfg = get_vp10_ref_frame_buffer(cpi, ref_frame_flag);
if (cfg) {
vpx_yv12_copy_frame(sd, cfg);
@@ -2303,7 +2084,7 @@
}
}
-int vp10_update_entropy(VP10_COMP * cpi, int update) {
+int vp10_update_entropy(VP10_COMP *cpi, int update) {
cpi->ext_refresh_frame_context = update;
cpi->ext_refresh_frame_context_pending = 1;
return 0;
@@ -2352,7 +2133,7 @@
uint16_t *src16 = CONVERT_TO_SHORTPTR(s->y_buffer);
do {
- fwrite(src16, s->y_width, 2, yuv_rec_file);
+ fwrite(src16, s->y_width, 2, yuv_rec_file);
src16 += s->y_stride;
} while (--h);
@@ -2360,7 +2141,7 @@
h = s->uv_height;
do {
- fwrite(src16, s->uv_width, 2, yuv_rec_file);
+ fwrite(src16, s->uv_width, 2, yuv_rec_file);
src16 += s->uv_stride;
} while (--h);
@@ -2378,7 +2159,7 @@
#endif // CONFIG_VPX_HIGHBITDEPTH
do {
- fwrite(src, s->y_width, 1, yuv_rec_file);
+ fwrite(src, s->y_width, 1, yuv_rec_file);
src += s->y_stride;
} while (--h);
@@ -2386,7 +2167,7 @@
h = s->uv_height;
do {
- fwrite(src, s->uv_width, 1, yuv_rec_file);
+ fwrite(src, s->uv_width, 1, yuv_rec_file);
src += s->uv_stride;
} while (--h);
@@ -2412,32 +2193,33 @@
#endif // CONFIG_VPX_HIGHBITDEPTH
// TODO(dkovalev): replace YV12_BUFFER_CONFIG with vpx_image_t
int i;
- const uint8_t *const srcs[3] = {src->y_buffer, src->u_buffer, src->v_buffer};
- const int src_strides[3] = {src->y_stride, src->uv_stride, src->uv_stride};
- const int src_widths[3] = {src->y_crop_width, src->uv_crop_width,
- src->uv_crop_width };
- const int src_heights[3] = {src->y_crop_height, src->uv_crop_height,
- src->uv_crop_height};
- uint8_t *const dsts[3] = {dst->y_buffer, dst->u_buffer, dst->v_buffer};
- const int dst_strides[3] = {dst->y_stride, dst->uv_stride, dst->uv_stride};
- const int dst_widths[3] = {dst->y_crop_width, dst->uv_crop_width,
- dst->uv_crop_width};
- const int dst_heights[3] = {dst->y_crop_height, dst->uv_crop_height,
- dst->uv_crop_height};
+ const uint8_t *const srcs[3] = { src->y_buffer, src->u_buffer,
+ src->v_buffer };
+ const int src_strides[3] = { src->y_stride, src->uv_stride, src->uv_stride };
+ const int src_widths[3] = { src->y_crop_width, src->uv_crop_width,
+ src->uv_crop_width };
+ const int src_heights[3] = { src->y_crop_height, src->uv_crop_height,
+ src->uv_crop_height };
+ uint8_t *const dsts[3] = { dst->y_buffer, dst->u_buffer, dst->v_buffer };
+ const int dst_strides[3] = { dst->y_stride, dst->uv_stride, dst->uv_stride };
+ const int dst_widths[3] = { dst->y_crop_width, dst->uv_crop_width,
+ dst->uv_crop_width };
+ const int dst_heights[3] = { dst->y_crop_height, dst->uv_crop_height,
+ dst->uv_crop_height };
for (i = 0; i < MAX_MB_PLANE; ++i) {
#if CONFIG_VPX_HIGHBITDEPTH
if (src->flags & YV12_FLAG_HIGHBITDEPTH) {
vp10_highbd_resize_plane(srcs[i], src_heights[i], src_widths[i],
- src_strides[i], dsts[i], dst_heights[i],
- dst_widths[i], dst_strides[i], bd);
+ src_strides[i], dsts[i], dst_heights[i],
+ dst_widths[i], dst_strides[i], bd);
} else {
vp10_resize_plane(srcs[i], src_heights[i], src_widths[i], src_strides[i],
- dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]);
+ dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]);
}
#else
vp10_resize_plane(srcs[i], src_heights[i], src_widths[i], src_strides[i],
- dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]);
+ dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]);
#endif // CONFIG_VPX_HIGHBITDEPTH
}
vpx_extend_frame_borders(dst);
@@ -2454,10 +2236,11 @@
const int src_h = src->y_crop_height;
const int dst_w = dst->y_crop_width;
const int dst_h = dst->y_crop_height;
- const uint8_t *const srcs[3] = {src->y_buffer, src->u_buffer, src->v_buffer};
- const int src_strides[3] = {src->y_stride, src->uv_stride, src->uv_stride};
- uint8_t *const dsts[3] = {dst->y_buffer, dst->u_buffer, dst->v_buffer};
- const int dst_strides[3] = {dst->y_stride, dst->uv_stride, dst->uv_stride};
+ const uint8_t *const srcs[3] = { src->y_buffer, src->u_buffer,
+ src->v_buffer };
+ const int src_strides[3] = { src->y_stride, src->uv_stride, src->uv_stride };
+ uint8_t *const dsts[3] = { dst->y_buffer, dst->u_buffer, dst->v_buffer };
+ const int dst_strides[3] = { dst->y_stride, dst->uv_stride, dst->uv_stride };
const InterpKernel *const kernel = vp10_filter_kernels[EIGHTTAP];
int x, y, i;
@@ -2469,8 +2252,9 @@
const int y_q4 = y * (16 / factor) * src_h / dst_h;
const int src_stride = src_strides[i];
const int dst_stride = dst_strides[i];
- const uint8_t *src_ptr = srcs[i] + (y / factor) * src_h / dst_h *
- src_stride + (x / factor) * src_w / dst_w;
+ const uint8_t *src_ptr = srcs[i] +
+ (y / factor) * src_h / dst_h * src_stride +
+ (x / factor) * src_w / dst_w;
uint8_t *dst_ptr = dsts[i] + (y / factor) * dst_stride + (x / factor);
#if CONFIG_VPX_HIGHBITDEPTH
@@ -2482,14 +2266,14 @@
} else {
vpx_convolve8(src_ptr, src_stride, dst_ptr, dst_stride,
kernel[x_q4 & 0xf], 16 * src_w / dst_w,
- kernel[y_q4 & 0xf], 16 * src_h / dst_h,
- 16 / factor, 16 / factor);
+ kernel[y_q4 & 0xf], 16 * src_h / dst_h, 16 / factor,
+ 16 / factor);
}
#else
vpx_convolve8(src_ptr, src_stride, dst_ptr, dst_stride,
kernel[x_q4 & 0xf], 16 * src_w / dst_w,
- kernel[y_q4 & 0xf], 16 * src_h / dst_h,
- 16 / factor, 16 / factor);
+ kernel[y_q4 & 0xf], 16 * src_h / dst_h, 16 / factor,
+ 16 / factor);
#endif // CONFIG_VPX_HIGHBITDEPTH
}
}
@@ -2506,8 +2290,9 @@
if (rc->frame_size_selector == UNSCALED &&
q >= rc->rf_level_maxq[gf_group->rf_level[gf_group->index]]) {
- const int max_size_thresh = (int)(rate_thresh_mult[SCALE_STEP1]
- * VPXMAX(rc->this_frame_target, rc->avg_frame_bandwidth));
+ const int max_size_thresh =
+ (int)(rate_thresh_mult[SCALE_STEP1] *
+ VPXMAX(rc->this_frame_target, rc->avg_frame_bandwidth));
scale = rc->projected_frame_size > max_size_thresh ? 1 : 0;
}
return scale;
@@ -2515,8 +2300,7 @@
// Function to test for conditions that indicate we should loop
// back and recode a frame.
-static int recode_loop_test(VP10_COMP *cpi,
- int high_limit, int low_limit,
+static int recode_loop_test(VP10_COMP *cpi, int high_limit, int low_limit,
int q, int maxq, int minq) {
const RATE_CONTROL *const rc = &cpi->rc;
const VP10EncoderConfig *const oxcf = &cpi->oxcf;
@@ -2525,14 +2309,12 @@
if ((rc->projected_frame_size >= rc->max_frame_bandwidth) ||
(cpi->sf.recode_loop == ALLOW_RECODE) ||
- (frame_is_kfgfarf &&
- (cpi->sf.recode_loop == ALLOW_RECODE_KFARFGF))) {
- if (frame_is_kfgfarf &&
- (oxcf->resize_mode == RESIZE_DYNAMIC) &&
+ (frame_is_kfgfarf && (cpi->sf.recode_loop == ALLOW_RECODE_KFARFGF))) {
+ if (frame_is_kfgfarf && (oxcf->resize_mode == RESIZE_DYNAMIC) &&
scale_down(cpi, q)) {
- // Code this group at a lower resolution.
- cpi->resize_pending = 1;
- return 1;
+ // Code this group at a lower resolution.
+ cpi->resize_pending = 1;
+ return 1;
}
// TODO(agrange) high_limit could be greater than the scale-down threshold.
@@ -2552,16 +2334,16 @@
}
void vp10_update_reference_frames(VP10_COMP *cpi) {
- VP10_COMMON * const cm = &cpi->common;
+ VP10_COMMON *const cm = &cpi->common;
BufferPool *const pool = cm->buffer_pool;
// At this point the new frame has been encoded.
// If any buffer copy / swapping is signaled it should be done here.
if (cm->frame_type == KEY_FRAME) {
- ref_cnt_fb(pool->frame_bufs,
- &cm->ref_frame_map[cpi->gld_fb_idx], cm->new_fb_idx);
- ref_cnt_fb(pool->frame_bufs,
- &cm->ref_frame_map[cpi->alt_fb_idx], cm->new_fb_idx);
+ ref_cnt_fb(pool->frame_bufs, &cm->ref_frame_map[cpi->gld_fb_idx],
+ cm->new_fb_idx);
+ ref_cnt_fb(pool->frame_bufs, &cm->ref_frame_map[cpi->alt_fb_idx],
+ cm->new_fb_idx);
} else if (vp10_preserve_existing_gf(cpi)) {
// We have decided to preserve the previously existing golden frame as our
// new ARF frame. However, in the short term in function
@@ -2573,8 +2355,8 @@
// slot and, if we're updating the GF, the current frame becomes the new GF.
int tmp;
- ref_cnt_fb(pool->frame_bufs,
- &cm->ref_frame_map[cpi->alt_fb_idx], cm->new_fb_idx);
+ ref_cnt_fb(pool->frame_bufs, &cm->ref_frame_map[cpi->alt_fb_idx],
+ cm->new_fb_idx);
tmp = cpi->alt_fb_idx;
cpi->alt_fb_idx = cpi->gld_fb_idx;
@@ -2587,16 +2369,15 @@
arf_idx = gf_group->arf_update_idx[gf_group->index];
}
- ref_cnt_fb(pool->frame_bufs,
- &cm->ref_frame_map[arf_idx], cm->new_fb_idx);
+ ref_cnt_fb(pool->frame_bufs, &cm->ref_frame_map[arf_idx], cm->new_fb_idx);
memcpy(cpi->interp_filter_selected[ALTREF_FRAME],
cpi->interp_filter_selected[0],
sizeof(cpi->interp_filter_selected[0]));
}
if (cpi->refresh_golden_frame) {
- ref_cnt_fb(pool->frame_bufs,
- &cm->ref_frame_map[cpi->gld_fb_idx], cm->new_fb_idx);
+ ref_cnt_fb(pool->frame_bufs, &cm->ref_frame_map[cpi->gld_fb_idx],
+ cm->new_fb_idx);
if (!cpi->rc.is_src_frame_alt_ref)
memcpy(cpi->interp_filter_selected[GOLDEN_FRAME],
cpi->interp_filter_selected[0],
@@ -2609,8 +2390,8 @@
}
if (cpi->refresh_last_frame) {
- ref_cnt_fb(pool->frame_bufs,
- &cm->ref_frame_map[cpi->lst_fb_idx], cm->new_fb_idx);
+ ref_cnt_fb(pool->frame_bufs, &cm->ref_frame_map[cpi->lst_fb_idx],
+ cm->new_fb_idx);
if (!cpi->rc.is_src_frame_alt_ref)
memcpy(cpi->interp_filter_selected[LAST_FRAME],
cpi->interp_filter_selected[0],
@@ -2639,9 +2420,8 @@
if (lf->filter_level > 0) {
if (cpi->num_workers > 1)
vp10_loop_filter_frame_mt(cm->frame_to_show, cm, xd->plane,
- lf->filter_level, 0, 0,
- cpi->workers, cpi->num_workers,
- &cpi->lf_row_sync);
+ lf->filter_level, 0, 0, cpi->workers,
+ cpi->num_workers, &cpi->lf_row_sync);
else
vp10_loop_filter_frame(cm->frame_to_show, cm, xd, lf->filter_level, 0, 0);
}
@@ -2649,16 +2429,13 @@
vpx_extend_frame_inner_borders(cm->frame_to_show);
}
-static INLINE void alloc_frame_mvs(const VP10_COMMON *cm,
- int buffer_idx) {
+static INLINE void alloc_frame_mvs(const VP10_COMMON *cm, int buffer_idx) {
RefCntBuffer *const new_fb_ptr = &cm->buffer_pool->frame_bufs[buffer_idx];
- if (new_fb_ptr->mvs == NULL ||
- new_fb_ptr->mi_rows < cm->mi_rows ||
+ if (new_fb_ptr->mvs == NULL || new_fb_ptr->mi_rows < cm->mi_rows ||
new_fb_ptr->mi_cols < cm->mi_cols) {
vpx_free(new_fb_ptr->mvs);
- new_fb_ptr->mvs =
- (MV_REF *)vpx_calloc(cm->mi_rows * cm->mi_cols,
- sizeof(*new_fb_ptr->mvs));
+ new_fb_ptr->mvs = (MV_REF *)vpx_calloc(cm->mi_rows * cm->mi_cols,
+ sizeof(*new_fb_ptr->mvs));
new_fb_ptr->mi_rows = cm->mi_rows;
new_fb_ptr->mi_cols = cm->mi_cols;
}
@@ -2667,14 +2444,15 @@
void vp10_scale_references(VP10_COMP *cpi) {
VP10_COMMON *cm = &cpi->common;
MV_REFERENCE_FRAME ref_frame;
- const VPX_REFFRAME ref_mask[3] = {VPX_LAST_FLAG, VPX_GOLD_FLAG, VPX_ALT_FLAG};
+ const VPX_REFFRAME ref_mask[3] = { VPX_LAST_FLAG, VPX_GOLD_FLAG,
+ VPX_ALT_FLAG };
for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ++ref_frame) {
// Need to convert from VPX_REFFRAME to index into ref_mask (subtract 1).
if (cpi->ref_frame_flags & ref_mask[ref_frame - 1]) {
BufferPool *const pool = cm->buffer_pool;
- const YV12_BUFFER_CONFIG *const ref = get_ref_frame_buffer(cpi,
- ref_frame);
+ const YV12_BUFFER_CONFIG *const ref =
+ get_ref_frame_buffer(cpi, ref_frame);
if (ref == NULL) {
cpi->scaled_ref_idx[ref_frame - 1] = INVALID_IDX;
@@ -2690,18 +2468,14 @@
new_fb = get_free_fb(cm);
force_scaling = 1;
}
- if (new_fb == INVALID_IDX)
- return;
+ if (new_fb == INVALID_IDX) return;
new_fb_ptr = &pool->frame_bufs[new_fb];
- if (force_scaling ||
- new_fb_ptr->buf.y_crop_width != cm->width ||
+ if (force_scaling || new_fb_ptr->buf.y_crop_width != cm->width ||
new_fb_ptr->buf.y_crop_height != cm->height) {
- vpx_realloc_frame_buffer(&new_fb_ptr->buf,
- cm->width, cm->height,
- cm->subsampling_x, cm->subsampling_y,
- cm->use_highbitdepth,
- VPX_ENC_BORDER_IN_PIXELS, cm->byte_alignment,
- NULL, NULL, NULL);
+ vpx_realloc_frame_buffer(
+ &new_fb_ptr->buf, cm->width, cm->height, cm->subsampling_x,
+ cm->subsampling_y, cm->use_highbitdepth, VPX_ENC_BORDER_IN_PIXELS,
+ cm->byte_alignment, NULL, NULL, NULL);
scale_and_extend_frame(ref, &new_fb_ptr->buf, (int)cm->bit_depth);
cpi->scaled_ref_idx[ref_frame - 1] = new_fb;
alloc_frame_mvs(cm, new_fb);
@@ -2715,14 +2489,11 @@
new_fb = get_free_fb(cm);
force_scaling = 1;
}
- if (new_fb == INVALID_IDX)
- return;
+ if (new_fb == INVALID_IDX) return;
new_fb_ptr = &pool->frame_bufs[new_fb];
- if (force_scaling ||
- new_fb_ptr->buf.y_crop_width != cm->width ||
+ if (force_scaling || new_fb_ptr->buf.y_crop_width != cm->width ||
new_fb_ptr->buf.y_crop_height != cm->height) {
- vpx_realloc_frame_buffer(&new_fb_ptr->buf,
- cm->width, cm->height,
+ vpx_realloc_frame_buffer(&new_fb_ptr->buf, cm->width, cm->height,
cm->subsampling_x, cm->subsampling_y,
VPX_ENC_BORDER_IN_PIXELS, cm->byte_alignment,
NULL, NULL, NULL);
@@ -2740,8 +2511,7 @@
++buf->ref_count;
}
} else {
- if (cpi->oxcf.pass != 0)
- cpi->scaled_ref_idx[ref_frame - 1] = INVALID_IDX;
+ if (cpi->oxcf.pass != 0) cpi->scaled_ref_idx[ref_frame - 1] = INVALID_IDX;
}
}
}
@@ -2758,22 +2528,21 @@
refresh[2] = (cpi->refresh_alt_ref_frame) ? 1 : 0;
for (i = LAST_FRAME; i <= ALTREF_FRAME; ++i) {
const int idx = cpi->scaled_ref_idx[i - 1];
- RefCntBuffer *const buf = idx != INVALID_IDX ?
- &cm->buffer_pool->frame_bufs[idx] : NULL;
+ RefCntBuffer *const buf =
+ idx != INVALID_IDX ? &cm->buffer_pool->frame_bufs[idx] : NULL;
const YV12_BUFFER_CONFIG *const ref = get_ref_frame_buffer(cpi, i);
if (buf != NULL &&
- (refresh[i - 1] ||
- (buf->buf.y_crop_width == ref->y_crop_width &&
- buf->buf.y_crop_height == ref->y_crop_height))) {
+ (refresh[i - 1] || (buf->buf.y_crop_width == ref->y_crop_width &&
+ buf->buf.y_crop_height == ref->y_crop_height))) {
--buf->ref_count;
- cpi->scaled_ref_idx[i -1] = INVALID_IDX;
+ cpi->scaled_ref_idx[i - 1] = INVALID_IDX;
}
}
} else {
for (i = 0; i < MAX_REF_FRAMES; ++i) {
const int idx = cpi->scaled_ref_idx[i];
- RefCntBuffer *const buf = idx != INVALID_IDX ?
- &cm->buffer_pool->frame_bufs[idx] : NULL;
+ RefCntBuffer *const buf =
+ idx != INVALID_IDX ? &cm->buffer_pool->frame_bufs[idx] : NULL;
if (buf != NULL) {
--buf->ref_count;
cpi->scaled_ref_idx[i] = INVALID_IDX;
@@ -2907,8 +2676,8 @@
cpi->common.interp_filter = cpi->sf.default_interp_filter;
}
-static void set_size_dependent_vars(VP10_COMP *cpi, int *q,
- int *bottom_index, int *top_index) {
+static void set_size_dependent_vars(VP10_COMP *cpi, int *q, int *bottom_index,
+ int *top_index) {
VP10_COMMON *const cm = &cpi->common;
const VP10EncoderConfig *const oxcf = &cpi->oxcf;
@@ -2946,40 +2715,37 @@
VP10EncoderConfig *const oxcf = &cpi->oxcf;
MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
- if (oxcf->pass == 2 &&
- oxcf->rc_mode == VPX_VBR &&
+ if (oxcf->pass == 2 && oxcf->rc_mode == VPX_VBR &&
((oxcf->resize_mode == RESIZE_FIXED && cm->current_video_frame == 0) ||
- (oxcf->resize_mode == RESIZE_DYNAMIC && cpi->resize_pending))) {
- vp10_calculate_coded_size(
- cpi, &oxcf->scaled_frame_width, &oxcf->scaled_frame_height);
+ (oxcf->resize_mode == RESIZE_DYNAMIC && cpi->resize_pending))) {
+ vp10_calculate_coded_size(cpi, &oxcf->scaled_frame_width,
+ &oxcf->scaled_frame_height);
// There has been a change in frame size.
vp10_set_size_literal(cpi, oxcf->scaled_frame_width,
- oxcf->scaled_frame_height);
+ oxcf->scaled_frame_height);
}
- if (oxcf->pass == 0 &&
- oxcf->rc_mode == VPX_CBR &&
+ if (oxcf->pass == 0 && oxcf->rc_mode == VPX_CBR &&
oxcf->resize_mode == RESIZE_DYNAMIC) {
- if (cpi->resize_pending == 1) {
- oxcf->scaled_frame_width =
- (cm->width * cpi->resize_scale_num) / cpi->resize_scale_den;
- oxcf->scaled_frame_height =
- (cm->height * cpi->resize_scale_num) /cpi->resize_scale_den;
- } else if (cpi->resize_pending == -1) {
- // Go back up to original size.
- oxcf->scaled_frame_width = oxcf->width;
- oxcf->scaled_frame_height = oxcf->height;
- }
- if (cpi->resize_pending != 0) {
- // There has been a change in frame size.
- vp10_set_size_literal(cpi,
- oxcf->scaled_frame_width,
- oxcf->scaled_frame_height);
+ if (cpi->resize_pending == 1) {
+ oxcf->scaled_frame_width =
+ (cm->width * cpi->resize_scale_num) / cpi->resize_scale_den;
+ oxcf->scaled_frame_height =
+ (cm->height * cpi->resize_scale_num) / cpi->resize_scale_den;
+ } else if (cpi->resize_pending == -1) {
+ // Go back up to original size.
+ oxcf->scaled_frame_width = oxcf->width;
+ oxcf->scaled_frame_height = oxcf->height;
+ }
+ if (cpi->resize_pending != 0) {
+ // There has been a change in frame size.
+ vp10_set_size_literal(cpi, oxcf->scaled_frame_width,
+ oxcf->scaled_frame_height);
- // TODO(agrange) Scale cpi->max_mv_magnitude if frame-size has changed.
- set_mv_search_params(cpi);
- }
+ // TODO(agrange) Scale cpi->max_mv_magnitude if frame-size has changed.
+ set_mv_search_params(cpi);
+ }
}
if (oxcf->pass == 2) {
@@ -2989,14 +2755,13 @@
alloc_frame_mvs(cm, cm->new_fb_idx);
// Reset the frame pointers to the current frame size.
- vpx_realloc_frame_buffer(get_frame_new_buffer(cm),
- cm->width, cm->height,
+ vpx_realloc_frame_buffer(get_frame_new_buffer(cm), cm->width, cm->height,
cm->subsampling_x, cm->subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
cm->use_highbitdepth,
#endif
- VPX_ENC_BORDER_IN_PIXELS, cm->byte_alignment,
- NULL, NULL, NULL);
+ VPX_ENC_BORDER_IN_PIXELS, cm->byte_alignment, NULL,
+ NULL, NULL);
alloc_util_frame_buffers(cpi);
init_motion_estimation(cpi);
@@ -3011,18 +2776,15 @@
YV12_BUFFER_CONFIG *const buf = &cm->buffer_pool->frame_bufs[buf_idx].buf;
ref_buf->buf = buf;
#if CONFIG_VPX_HIGHBITDEPTH
- vp10_setup_scale_factors_for_frame(&ref_buf->sf,
- buf->y_crop_width, buf->y_crop_height,
- cm->width, cm->height,
- (buf->flags & YV12_FLAG_HIGHBITDEPTH) ?
- 1 : 0);
+ vp10_setup_scale_factors_for_frame(
+ &ref_buf->sf, buf->y_crop_width, buf->y_crop_height, cm->width,
+ cm->height, (buf->flags & YV12_FLAG_HIGHBITDEPTH) ? 1 : 0);
#else
- vp10_setup_scale_factors_for_frame(&ref_buf->sf,
- buf->y_crop_width, buf->y_crop_height,
- cm->width, cm->height);
+ vp10_setup_scale_factors_for_frame(&ref_buf->sf, buf->y_crop_width,
+ buf->y_crop_height, cm->width,
+ cm->height);
#endif // CONFIG_VPX_HIGHBITDEPTH
- if (vp10_is_scaled(&ref_buf->sf))
- vpx_extend_frame_borders(buf);
+ if (vp10_is_scaled(&ref_buf->sf)) vpx_extend_frame_borders(buf);
} else {
ref_buf->buf = NULL;
}
@@ -3041,24 +2803,21 @@
// For 1 pass CBR under dynamic resize mode: use faster scaling for source.
// Only for 2x2 scaling for now.
- if (cpi->oxcf.pass == 0 &&
- cpi->oxcf.rc_mode == VPX_CBR &&
+ if (cpi->oxcf.pass == 0 && cpi->oxcf.rc_mode == VPX_CBR &&
cpi->oxcf.resize_mode == RESIZE_DYNAMIC &&
cpi->un_scaled_source->y_width == (cm->width << 1) &&
cpi->un_scaled_source->y_height == (cm->height << 1)) {
- cpi->Source = vp10_scale_if_required_fast(cm,
- cpi->un_scaled_source,
- &cpi->scaled_source);
+ cpi->Source = vp10_scale_if_required_fast(cm, cpi->un_scaled_source,
+ &cpi->scaled_source);
if (cpi->unscaled_last_source != NULL)
- cpi->Last_Source = vp10_scale_if_required_fast(cm,
- cpi->unscaled_last_source,
- &cpi->scaled_last_source);
+ cpi->Last_Source = vp10_scale_if_required_fast(
+ cm, cpi->unscaled_last_source, &cpi->scaled_last_source);
} else {
- cpi->Source = vp10_scale_if_required(cm, cpi->un_scaled_source,
- &cpi->scaled_source);
+ cpi->Source =
+ vp10_scale_if_required(cm, cpi->un_scaled_source, &cpi->scaled_source);
if (cpi->unscaled_last_source != NULL)
cpi->Last_Source = vp10_scale_if_required(cm, cpi->unscaled_last_source,
- &cpi->scaled_last_source);
+ &cpi->scaled_last_source);
}
if (frame_is_intra_only(cm) == 0) {
@@ -3090,8 +2849,7 @@
// Update some stats from cyclic refresh, and check if we should not update
// golden reference, for 1 pass CBR.
- if (cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ &&
- cm->frame_type != KEY_FRAME &&
+ if (cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ && cm->frame_type != KEY_FRAME &&
(cpi->oxcf.pass == 0 && cpi->oxcf.rc_mode == VPX_CBR))
vp10_cyclic_refresh_check_golden_update(cpi);
@@ -3101,8 +2859,7 @@
vpx_clear_system_state();
}
-static void encode_with_recode_loop(VP10_COMP *cpi,
- size_t *size,
+static void encode_with_recode_loop(VP10_COMP *cpi, size_t *size,
uint8_t *dest) {
VP10_COMMON *const cm = &cpi->common;
RATE_CONTROL *const rc = &cpi->rc;
@@ -3145,16 +2902,16 @@
// Decide frame size bounds first time through.
if (loop_count == 0) {
vp10_rc_compute_frame_size_bounds(cpi, rc->this_frame_target,
- &frame_under_shoot_limit,
- &frame_over_shoot_limit);
+ &frame_under_shoot_limit,
+ &frame_over_shoot_limit);
}
- cpi->Source = vp10_scale_if_required(cm, cpi->un_scaled_source,
- &cpi->scaled_source);
+ cpi->Source =
+ vp10_scale_if_required(cm, cpi->un_scaled_source, &cpi->scaled_source);
if (cpi->unscaled_last_source != NULL)
cpi->Last_Source = vp10_scale_if_required(cm, cpi->unscaled_last_source,
- &cpi->scaled_last_source);
+ &cpi->scaled_last_source);
if (frame_is_intra_only(cm) == 0) {
if (loop_count > 0) {
@@ -3165,8 +2922,7 @@
vp10_set_quantizer(cm, q);
- if (loop_count == 0)
- setup_frame(cpi);
+ if (loop_count == 0) setup_frame(cpi);
// Variance adaptive and in frame q adjustment experiments are mutually
// exclusive.
@@ -3195,16 +2951,14 @@
rc->projected_frame_size = (int)(*size) << 3;
restore_coding_context(cpi);
- if (frame_over_shoot_limit == 0)
- frame_over_shoot_limit = 1;
+ if (frame_over_shoot_limit == 0) frame_over_shoot_limit = 1;
}
if (cpi->oxcf.rc_mode == VPX_Q) {
loop = 0;
} else {
- if ((cm->frame_type == KEY_FRAME) &&
- rc->this_key_frame_forced &&
- (rc->projected_frame_size < rc->max_frame_bandwidth)) {
+ if ((cm->frame_type == KEY_FRAME) && rc->this_key_frame_forced &&
+ (rc->projected_frame_size < rc->max_frame_bandwidth)) {
int last_q = q;
int64_t kf_err;
@@ -3251,9 +3005,9 @@
q = clamp(q, q_low, q_high);
loop = q != last_q;
- } else if (recode_loop_test(
- cpi, frame_over_shoot_limit, frame_under_shoot_limit,
- q, VPXMAX(q_high, top_index), bottom_index)) {
+ } else if (recode_loop_test(cpi, frame_over_shoot_limit,
+ frame_under_shoot_limit, q,
+ VPXMAX(q_high, top_index), bottom_index)) {
// Is the projected frame size out of range and are we allowed
// to attempt to recode.
int last_q = q;
@@ -3294,13 +3048,13 @@
// Update rate_correction_factor unless
vp10_rc_update_rate_correction_factors(cpi);
- q = vp10_rc_regulate_q(cpi, rc->this_frame_target,
- bottom_index, VPXMAX(q_high, top_index));
+ q = vp10_rc_regulate_q(cpi, rc->this_frame_target, bottom_index,
+ VPXMAX(q_high, top_index));
while (q < q_low && retries < 10) {
vp10_rc_update_rate_correction_factors(cpi);
- q = vp10_rc_regulate_q(cpi, rc->this_frame_target,
- bottom_index, VPXMAX(q_high, top_index));
+ q = vp10_rc_regulate_q(cpi, rc->this_frame_target, bottom_index,
+ VPXMAX(q_high, top_index));
retries++;
}
}
@@ -3315,21 +3069,20 @@
q = (q_high + q_low) / 2;
} else {
vp10_rc_update_rate_correction_factors(cpi);
- q = vp10_rc_regulate_q(cpi, rc->this_frame_target,
- bottom_index, top_index);
+ q = vp10_rc_regulate_q(cpi, rc->this_frame_target, bottom_index,
+ top_index);
// Special case reset for qlow for constrained quality.
// This should only trigger where there is very substantial
// undershoot on a frame and the auto cq level is above
// the user passsed in value.
- if (cpi->oxcf.rc_mode == VPX_CQ &&
- q < q_low) {
+ if (cpi->oxcf.rc_mode == VPX_CQ && q < q_low) {
q_low = q;
}
while (q > q_high && retries < 10) {
vp10_rc_update_rate_correction_factors(cpi);
- q = vp10_rc_regulate_q(cpi, rc->this_frame_target,
- bottom_index, top_index);
+ q = vp10_rc_regulate_q(cpi, rc->this_frame_target, bottom_index,
+ top_index);
retries++;
}
}
@@ -3369,17 +3122,13 @@
const int gold_is_alt = map[cpi->gld_fb_idx] == map[cpi->alt_fb_idx];
int flags = VPX_ALT_FLAG | VPX_GOLD_FLAG | VPX_LAST_FLAG;
- if (gold_is_last)
- flags &= ~VPX_GOLD_FLAG;
+ if (gold_is_last) flags &= ~VPX_GOLD_FLAG;
- if (cpi->rc.frames_till_gf_update_due == INT_MAX)
- flags &= ~VPX_GOLD_FLAG;
+ if (cpi->rc.frames_till_gf_update_due == INT_MAX) flags &= ~VPX_GOLD_FLAG;
- if (alt_is_last)
- flags &= ~VPX_ALT_FLAG;
+ if (alt_is_last) flags &= ~VPX_ALT_FLAG;
- if (gold_is_alt)
- flags &= ~VPX_ALT_FLAG;
+ if (gold_is_alt) flags &= ~VPX_ALT_FLAG;
return flags;
}
@@ -3402,13 +3151,12 @@
}
YV12_BUFFER_CONFIG *vp10_scale_if_required_fast(VP10_COMMON *cm,
- YV12_BUFFER_CONFIG *unscaled,
- YV12_BUFFER_CONFIG *scaled) {
+ YV12_BUFFER_CONFIG *unscaled,
+ YV12_BUFFER_CONFIG *scaled) {
if (cm->mi_cols * MI_SIZE != unscaled->y_width ||
cm->mi_rows * MI_SIZE != unscaled->y_height) {
// For 2x2 scaling down.
- vpx_scale_frame(unscaled, scaled, unscaled->y_buffer, 9, 2, 1,
- 2, 1, 0);
+ vpx_scale_frame(unscaled, scaled, unscaled->y_buffer, 9, 2, 1, 2, 1, 0);
vpx_extend_frame_borders(scaled);
return scaled;
} else {
@@ -3417,8 +3165,8 @@
}
YV12_BUFFER_CONFIG *vp10_scale_if_required(VP10_COMMON *cm,
- YV12_BUFFER_CONFIG *unscaled,
- YV12_BUFFER_CONFIG *scaled) {
+ YV12_BUFFER_CONFIG *unscaled,
+ YV12_BUFFER_CONFIG *scaled) {
if (cm->mi_cols * MI_SIZE != unscaled->y_width ||
cm->mi_rows * MI_SIZE != unscaled->y_height) {
#if CONFIG_VPX_HIGHBITDEPTH
@@ -3443,18 +3191,17 @@
(gf_group->rf_level[gf_group->index] == GF_ARF_LOW));
} else {
arf_sign_bias =
- (cpi->rc.source_alt_ref_active && !cpi->refresh_alt_ref_frame);
+ (cpi->rc.source_alt_ref_active && !cpi->refresh_alt_ref_frame);
}
cm->ref_frame_sign_bias[ALTREF_FRAME] = arf_sign_bias;
}
static int setup_interp_filter_search_mask(VP10_COMP *cpi) {
INTERP_FILTER ifilter;
- int ref_total[MAX_REF_FRAMES] = {0};
+ int ref_total[MAX_REF_FRAMES] = { 0 };
MV_REFERENCE_FRAME ref;
int mask = 0;
- if (cpi->common.last_frame_type == KEY_FRAME ||
- cpi->refresh_alt_ref_frame)
+ if (cpi->common.last_frame_type == KEY_FRAME || cpi->refresh_alt_ref_frame)
return mask;
for (ref = LAST_FRAME; ref <= ALTREF_FRAME; ++ref)
for (ifilter = EIGHTTAP; ifilter <= EIGHTTAP_SHARP; ++ifilter)
@@ -3462,20 +3209,19 @@
for (ifilter = EIGHTTAP; ifilter <= EIGHTTAP_SHARP; ++ifilter) {
if ((ref_total[LAST_FRAME] &&
- cpi->interp_filter_selected[LAST_FRAME][ifilter] == 0) &&
+ cpi->interp_filter_selected[LAST_FRAME][ifilter] == 0) &&
(ref_total[GOLDEN_FRAME] == 0 ||
- cpi->interp_filter_selected[GOLDEN_FRAME][ifilter] * 50
- < ref_total[GOLDEN_FRAME]) &&
+ cpi->interp_filter_selected[GOLDEN_FRAME][ifilter] * 50 <
+ ref_total[GOLDEN_FRAME]) &&
(ref_total[ALTREF_FRAME] == 0 ||
- cpi->interp_filter_selected[ALTREF_FRAME][ifilter] * 50
- < ref_total[ALTREF_FRAME]))
+ cpi->interp_filter_selected[ALTREF_FRAME][ifilter] * 50 <
+ ref_total[ALTREF_FRAME]))
mask |= 1 << ifilter;
}
return mask;
}
-static void encode_frame_to_data_rate(VP10_COMP *cpi,
- size_t *size,
+static void encode_frame_to_data_rate(VP10_COMP *cpi, size_t *size,
uint8_t *dest,
unsigned int *frame_flags) {
VP10_COMMON *const cm = &cpi->common;
@@ -3492,10 +3238,8 @@
// Set default state for segment based loop filter update flags.
cm->lf.mode_ref_delta_update = 0;
- if (cpi->oxcf.pass == 2 &&
- cpi->sf.adaptive_interp_filter_search)
- cpi->sf.interp_filter_search_mask =
- setup_interp_filter_search_mask(cpi);
+ if (cpi->oxcf.pass == 2 && cpi->sf.adaptive_interp_filter_search)
+ cpi->sf.interp_filter_search_mask = setup_interp_filter_search_mask(cpi);
// Set various flags etc to special state if it is a key frame.
if (frame_is_intra_only(cm)) {
@@ -3525,8 +3269,7 @@
// For 1 pass CBR, check if we are dropping this frame.
// Never drop on key frame.
- if (oxcf->pass == 0 &&
- oxcf->rc_mode == VPX_CBR &&
+ if (oxcf->pass == 0 && oxcf->rc_mode == VPX_CBR &&
cm->frame_type != KEY_FRAME) {
if (vp10_rc_drop_frame(cpi)) {
vp10_rc_postencode_update_drop_frame(cpi);
@@ -3560,8 +3303,8 @@
if (cpi->rc.next_key_frame_forced && cpi->rc.frames_to_key == 1) {
#if CONFIG_VPX_HIGHBITDEPTH
if (cm->use_highbitdepth) {
- cpi->ambient_err = vp10_highbd_get_y_sse(cpi->Source,
- get_frame_new_buffer(cm));
+ cpi->ambient_err =
+ vp10_highbd_get_y_sse(cpi->Source, get_frame_new_buffer(cm));
} else {
cpi->ambient_err = vp10_get_y_sse(cpi->Source, get_frame_new_buffer(cm));
}
@@ -3571,13 +3314,12 @@
}
// If the encoder forced a KEY_FRAME decision
- if (cm->frame_type == KEY_FRAME)
- cpi->refresh_last_frame = 1;
+ if (cm->frame_type == KEY_FRAME) cpi->refresh_last_frame = 1;
cm->frame_to_show = get_frame_new_buffer(cm);
cm->frame_to_show->color_space = cm->color_space;
cm->frame_to_show->color_range = cm->color_range;
- cm->frame_to_show->render_width = cm->render_width;
+ cm->frame_to_show->render_width = cm->render_width;
cm->frame_to_show->render_height = cm->render_height;
// Pick the loop filter level for the frame.
@@ -3586,8 +3328,7 @@
// build the bitstream
vp10_pack_bitstream(cpi, dest, size);
- if (cm->seg.update_map)
- update_reference_segmentation_map(cpi);
+ if (cm->seg.update_map) update_reference_segmentation_map(cpi);
if (frame_is_intra_only(cm) == 0) {
release_scaled_references(cpi);
@@ -3603,8 +3344,7 @@
#if CONFIG_MISC_FIXES
vp10_adapt_intra_frame_probs(cm);
#else
- if (!frame_is_intra_only(cm))
- vp10_adapt_intra_frame_probs(cm);
+ if (!frame_is_intra_only(cm)) vp10_adapt_intra_frame_probs(cm);
#endif
}
@@ -3653,8 +3393,7 @@
cm->last_height = cm->height;
// reset to normal state now that we are done.
- if (!cm->show_existing_frame)
- cm->last_show_frame = cm->show_frame;
+ if (!cm->show_existing_frame) cm->last_show_frame = cm->show_frame;
if (cm->show_frame) {
vp10_swap_mi_and_prev_mi(cm);
@@ -3675,8 +3414,8 @@
encode_frame_to_data_rate(cpi, size, dest, frame_flags);
}
-static void Pass2Encode(VP10_COMP *cpi, size_t *size,
- uint8_t *dest, unsigned int *frame_flags) {
+static void Pass2Encode(VP10_COMP *cpi, size_t *size, uint8_t *dest,
+ unsigned int *frame_flags) {
cpi->allow_encode_breakout = ENCODE_BREAKOUT_ENABLED;
encode_frame_to_data_rate(cpi, size, dest, frame_flags);
@@ -3725,8 +3464,8 @@
}
int vp10_receive_raw_frame(VP10_COMP *cpi, unsigned int frame_flags,
- YV12_BUFFER_CONFIG *sd, int64_t time_stamp,
- int64_t end_time) {
+ YV12_BUFFER_CONFIG *sd, int64_t time_stamp,
+ int64_t end_time) {
VP10_COMMON *cm = &cpi->common;
struct vpx_usec_timer timer;
int res = 0;
@@ -3743,9 +3482,9 @@
if (vp10_lookahead_push(cpi->lookahead, sd, time_stamp, end_time,
#if CONFIG_VPX_HIGHBITDEPTH
- use_highbitdepth,
+ use_highbitdepth,
#endif // CONFIG_VPX_HIGHBITDEPTH
- frame_flags))
+ frame_flags))
res = -1;
vpx_usec_timer_mark(&timer);
cpi->time_receive_data += vpx_usec_timer_elapsed(&timer);
@@ -3766,17 +3505,13 @@
return res;
}
-
static int frame_is_reference(const VP10_COMP *cpi) {
const VP10_COMMON *cm = &cpi->common;
- return cm->frame_type == KEY_FRAME ||
- cpi->refresh_last_frame ||
- cpi->refresh_golden_frame ||
- cpi->refresh_alt_ref_frame ||
+ return cm->frame_type == KEY_FRAME || cpi->refresh_last_frame ||
+ cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame ||
cm->refresh_frame_context != REFRESH_FRAME_CONTEXT_OFF ||
- cm->lf.mode_ref_delta_update ||
- cm->seg.update_map ||
+ cm->lf.mode_ref_delta_update || cm->seg.update_map ||
cm->seg.update_data;
}
@@ -3789,8 +3524,8 @@
this_duration = source->ts_end - source->ts_start;
step = 1;
} else {
- int64_t last_duration = cpi->last_end_time_stamp_seen
- - cpi->last_time_stamp_seen;
+ int64_t last_duration =
+ cpi->last_end_time_stamp_seen - cpi->last_time_stamp_seen;
this_duration = source->ts_end - cpi->last_end_time_stamp_seen;
@@ -3844,10 +3579,10 @@
if (cpi->oxcf.pass == 2) {
const GF_GROUP *const gf_group = &cpi->twopass.gf_group;
rc->is_src_frame_alt_ref =
- (gf_group->update_type[gf_group->index] == OVERLAY_UPDATE);
+ (gf_group->update_type[gf_group->index] == OVERLAY_UPDATE);
} else {
- rc->is_src_frame_alt_ref = cpi->alt_ref_source &&
- (source == cpi->alt_ref_source);
+ rc->is_src_frame_alt_ref =
+ cpi->alt_ref_source && (source == cpi->alt_ref_source);
}
if (rc->is_src_frame_alt_ref) {
@@ -3862,8 +3597,8 @@
#if CONFIG_INTERNAL_STATS
extern double vp10_get_blockiness(const unsigned char *img1, int img1_pitch,
- const unsigned char *img2, int img2_pitch,
- int width, int height);
+ const unsigned char *img2, int img2_pitch,
+ int width, int height);
static void adjust_image_stat(double y, double u, double v, double all,
ImageStat *s) {
@@ -3876,13 +3611,13 @@
#endif // CONFIG_INTERNAL_STATS
int vp10_get_compressed_data(VP10_COMP *cpi, unsigned int *frame_flags,
- size_t *size, uint8_t *dest,
- int64_t *time_stamp, int64_t *time_end, int flush) {
+ size_t *size, uint8_t *dest, int64_t *time_stamp,
+ int64_t *time_end, int flush) {
const VP10EncoderConfig *const oxcf = &cpi->oxcf;
VP10_COMMON *const cm = &cpi->common;
BufferPool *const pool = cm->buffer_pool;
RATE_CONTROL *const rc = &cpi->rc;
- struct vpx_usec_timer cmptimer;
+ struct vpx_usec_timer cmptimer;
YV12_BUFFER_CONFIG *force_src_buffer = NULL;
struct lookahead_entry *last_source = NULL;
struct lookahead_entry *source = NULL;
@@ -3902,10 +3637,11 @@
// Normal defaults
cm->reset_frame_context = RESET_FRAME_CONTEXT_NONE;
- cm->refresh_frame_context =
- oxcf->error_resilient_mode ? REFRESH_FRAME_CONTEXT_OFF :
- oxcf->frame_parallel_decoding_mode ? REFRESH_FRAME_CONTEXT_FORWARD
- : REFRESH_FRAME_CONTEXT_BACKWARD;
+ cm->refresh_frame_context = oxcf->error_resilient_mode
+ ? REFRESH_FRAME_CONTEXT_OFF
+ : oxcf->frame_parallel_decoding_mode
+ ? REFRESH_FRAME_CONTEXT_FORWARD
+ : REFRESH_FRAME_CONTEXT_BACKWARD;
cpi->refresh_last_frame = 1;
cpi->refresh_golden_frame = 0;
@@ -3959,8 +3695,8 @@
}
if (source) {
- cpi->un_scaled_source = cpi->Source = force_src_buffer ? force_src_buffer
- : &source->img;
+ cpi->un_scaled_source = cpi->Source =
+ force_src_buffer ? force_src_buffer : &source->img;
cpi->unscaled_last_source = last_source != NULL ? &last_source->img : NULL;
@@ -3971,7 +3707,7 @@
} else {
*size = 0;
if (flush && oxcf->pass == 1 && !cpi->twopass.first_pass_done) {
- vp10_end_first_pass(cpi); /* get last stats packet */
+ vp10_end_first_pass(cpi); /* get last stats packet */
cpi->twopass.first_pass_done = 1;
}
return -1;
@@ -3997,8 +3733,7 @@
}
cm->new_fb_idx = get_free_fb(cm);
- if (cm->new_fb_idx == INVALID_IDX)
- return -1;
+ if (cm->new_fb_idx == INVALID_IDX) return -1;
cm->cur_frame = &pool->frame_bufs[cm->new_fb_idx];
@@ -4023,8 +3758,7 @@
}
if (cpi->oxcf.pass != 0 || frame_is_intra_only(cm) == 1) {
- for (i = 0; i < MAX_REF_FRAMES; ++i)
- cpi->scaled_ref_idx[i] = INVALID_IDX;
+ for (i = 0; i < MAX_REF_FRAMES; ++i) cpi->scaled_ref_idx[i] = INVALID_IDX;
}
if (oxcf->pass == 1) {
@@ -4087,8 +3821,8 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (cm->use_highbitdepth) {
- frame_ssim2 = vpx_highbd_calc_ssim(orig, recon, &weight,
- (int)cm->bit_depth);
+ frame_ssim2 =
+ vpx_highbd_calc_ssim(orig, recon, &weight, (int)cm->bit_depth);
} else {
frame_ssim2 = vpx_calc_ssim(orig, recon, &weight);
}
@@ -4096,7 +3830,7 @@
frame_ssim2 = vpx_calc_ssim(orig, recon, &weight);
#endif // CONFIG_VPX_HIGHBITDEPTH
- cpi->worst_ssim= VPXMIN(cpi->worst_ssim, frame_ssim2);
+ cpi->worst_ssim = VPXMIN(cpi->worst_ssim, frame_ssim2);
cpi->summed_quality += frame_ssim2 * weight;
cpi->summed_weights += weight;
@@ -4140,8 +3874,8 @@
&cpi->metrics, 1);
const double peak = (double)((1 << cpi->oxcf.input_bit_depth) - 1);
- double consistency = vpx_sse_to_psnr(samples, peak,
- (double)cpi->total_inconsistency);
+ double consistency =
+ vpx_sse_to_psnr(samples, peak, (double)cpi->total_inconsistency);
if (consistency > 0.0)
cpi->worst_consistency =
VPXMIN(cpi->worst_consistency, consistency);
@@ -4156,8 +3890,8 @@
frame_all = vpx_highbd_calc_ssimg(cpi->Source, cm->frame_to_show, &y,
&u, &v, (int)cm->bit_depth);
} else {
- frame_all = vpx_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u,
- &v);
+ frame_all =
+ vpx_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u, &v);
}
#else
frame_all = vpx_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u, &v);
@@ -4169,8 +3903,8 @@
#endif
{
double y, u, v, frame_all;
- frame_all = vpx_calc_fastssim(cpi->Source, cm->frame_to_show, &y, &u,
- &v);
+ frame_all =
+ vpx_calc_fastssim(cpi->Source, cm->frame_to_show, &y, &u, &v);
adjust_image_stat(y, u, v, frame_all, &cpi->fastssim);
/* TODO(JBB): add 10/12 bit support */
}
@@ -4212,13 +3946,12 @@
}
}
-int vp10_set_internal_size(VP10_COMP *cpi,
- VPX_SCALING horiz_mode, VPX_SCALING vert_mode) {
+int vp10_set_internal_size(VP10_COMP *cpi, VPX_SCALING horiz_mode,
+ VPX_SCALING vert_mode) {
VP10_COMMON *cm = &cpi->common;
int hr = 0, hs = 0, vr = 0, vs = 0;
- if (horiz_mode > ONETWO || vert_mode > ONETWO)
- return -1;
+ if (horiz_mode > ONETWO || vert_mode > ONETWO) return -1;
Scale2Ratio(horiz_mode, &hr, &hs);
Scale2Ratio(vert_mode, &vr, &vs);
@@ -4235,7 +3968,7 @@
}
int vp10_set_size_literal(VP10_COMP *cpi, unsigned int width,
- unsigned int height) {
+ unsigned int height) {
VP10_COMMON *cm = &cpi->common;
#if CONFIG_VPX_HIGHBITDEPTH
check_initial_width(cpi, cm->use_highbitdepth, 1, 1);
@@ -4267,7 +4000,7 @@
}
int64_t vp10_get_y_sse(const YV12_BUFFER_CONFIG *a,
- const YV12_BUFFER_CONFIG *b) {
+ const YV12_BUFFER_CONFIG *b) {
assert(a->y_crop_width == b->y_crop_width);
assert(a->y_crop_height == b->y_crop_height);
@@ -4277,7 +4010,7 @@
#if CONFIG_VPX_HIGHBITDEPTH
int64_t vp10_highbd_get_y_sse(const YV12_BUFFER_CONFIG *a,
- const YV12_BUFFER_CONFIG *b) {
+ const YV12_BUFFER_CONFIG *b) {
assert(a->y_crop_width == b->y_crop_width);
assert(a->y_crop_height == b->y_crop_height);
assert((a->flags & YV12_FLAG_HIGHBITDEPTH) != 0);
@@ -4288,40 +4021,32 @@
}
#endif // CONFIG_VPX_HIGHBITDEPTH
-int vp10_get_quantizer(VP10_COMP *cpi) {
- return cpi->common.base_qindex;
-}
+int vp10_get_quantizer(VP10_COMP *cpi) { return cpi->common.base_qindex; }
void vp10_apply_encoding_flags(VP10_COMP *cpi, vpx_enc_frame_flags_t flags) {
- if (flags & (VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_GF |
- VP8_EFLAG_NO_REF_ARF)) {
+ if (flags &
+ (VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF)) {
int ref = 7;
- if (flags & VP8_EFLAG_NO_REF_LAST)
- ref ^= VPX_LAST_FLAG;
+ if (flags & VP8_EFLAG_NO_REF_LAST) ref ^= VPX_LAST_FLAG;
- if (flags & VP8_EFLAG_NO_REF_GF)
- ref ^= VPX_GOLD_FLAG;
+ if (flags & VP8_EFLAG_NO_REF_GF) ref ^= VPX_GOLD_FLAG;
- if (flags & VP8_EFLAG_NO_REF_ARF)
- ref ^= VPX_ALT_FLAG;
+ if (flags & VP8_EFLAG_NO_REF_ARF) ref ^= VPX_ALT_FLAG;
vp10_use_as_reference(cpi, ref);
}
- if (flags & (VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
- VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_FORCE_GF |
- VP8_EFLAG_FORCE_ARF)) {
+ if (flags &
+ (VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_FORCE_GF | VP8_EFLAG_FORCE_ARF)) {
int upd = 7;
- if (flags & VP8_EFLAG_NO_UPD_LAST)
- upd ^= VPX_LAST_FLAG;
+ if (flags & VP8_EFLAG_NO_UPD_LAST) upd ^= VPX_LAST_FLAG;
- if (flags & VP8_EFLAG_NO_UPD_GF)
- upd ^= VPX_GOLD_FLAG;
+ if (flags & VP8_EFLAG_NO_UPD_GF) upd ^= VPX_GOLD_FLAG;
- if (flags & VP8_EFLAG_NO_UPD_ARF)
- upd ^= VPX_ALT_FLAG;
+ if (flags & VP8_EFLAG_NO_UPD_ARF) upd ^= VPX_ALT_FLAG;
vp10_update_reference(cpi, upd);
}
diff --git a/vp10/encoder/encoder.h b/vp10/encoder/encoder.h
index 87219ed..3547cba 100644
--- a/vp10/encoder/encoder.h
+++ b/vp10/encoder/encoder.h
@@ -64,7 +64,6 @@
FRAME_CONTEXT fc;
} CODING_CONTEXT;
-
typedef enum {
// encode_breakout is disabled.
ENCODE_BREAKOUT_DISABLED = 0,
@@ -75,10 +74,10 @@
} ENCODE_BREAKOUT_TYPE;
typedef enum {
- NORMAL = 0,
- FOURFIVE = 1,
- THREEFIVE = 2,
- ONETWO = 3
+ NORMAL = 0,
+ FOURFIVE = 1,
+ THREEFIVE = 2,
+ ONETWO = 3
} VPX_SCALING;
typedef enum {
@@ -98,7 +97,7 @@
} MODE;
typedef enum {
- FRAMEFLAGS_KEY = 1 << 0,
+ FRAMEFLAGS_KEY = 1 << 0,
FRAMEFLAGS_GOLDEN = 1 << 1,
FRAMEFLAGS_ALTREF = 1 << 2,
} FRAMETYPE_FLAGS;
@@ -120,14 +119,14 @@
typedef struct VP10EncoderConfig {
BITSTREAM_PROFILE profile;
vpx_bit_depth_t bit_depth; // Codec bit-depth.
- int width; // width of data passed to the compressor
- int height; // height of data passed to the compressor
+ int width; // width of data passed to the compressor
+ int height; // height of data passed to the compressor
unsigned int input_bit_depth; // Input bit depth.
- double init_framerate; // set to passed in framerate
- int64_t target_bandwidth; // bandwidth to be used in kilobits per second
+ double init_framerate; // set to passed in framerate
+ int64_t target_bandwidth; // bandwidth to be used in kilobits per second
int noise_sensitivity; // pre processing blur: recommendation 0
- int sharpness; // sharpening output: recommendation 0:
+ int sharpness; // sharpening output: recommendation 0:
int speed;
// maximum allowed bitrate for any intra frame in % of bitrate target.
unsigned int rc_max_intra_bitrate_pct;
@@ -179,7 +178,7 @@
int frame_periodic_boost;
// two pass datarate control
- int two_pass_vbrbias; // two pass datarate control tweaks
+ int two_pass_vbrbias; // two pass datarate control tweaks
int two_pass_vbrmin_section;
int two_pass_vbrmax_section;
// END DATARATE CONTROL OPTIONS
@@ -267,15 +266,10 @@
unsigned char *map;
} ActiveMap;
-typedef enum {
- Y,
- U,
- V,
- ALL
-} STAT_TYPE;
+typedef enum { Y, U, V, ALL } STAT_TYPE;
typedef struct IMAGE_STAT {
- double stat[ALL+1];
+ double stat[ALL + 1];
double worst;
} ImageStat;
@@ -287,8 +281,8 @@
DECLARE_ALIGNED(16, int16_t, uv_dequant[QINDEX_RANGE][8]);
VP10_COMMON common;
VP10EncoderConfig oxcf;
- struct lookahead_ctx *lookahead;
- struct lookahead_entry *alt_ref_source;
+ struct lookahead_ctx *lookahead;
+ struct lookahead_entry *alt_ref_source;
YV12_BUFFER_CONFIG *Source;
YV12_BUFFER_CONFIG *Last_Source; // NULL for first frame and alt_ref frames
@@ -346,11 +340,11 @@
int interp_filter_selected[MAX_REF_FRAMES][SWITCHABLE];
- struct vpx_codec_pkt_list *output_pkt_list;
+ struct vpx_codec_pkt_list *output_pkt_list;
MBGRAPH_FRAME_STATS mbgraph_stats[MAX_LAG_BUFFERS];
- int mbgraph_n_frames; // number of frames filled in the above
- int static_mb_pct; // % forced skip mbs by segmentation
+ int mbgraph_n_frames; // number of frames filled in the above
+ int static_mb_pct; // % forced skip mbs by segmentation
int ref_frame_flags;
SPEED_FEATURES sf;
@@ -370,7 +364,7 @@
unsigned char *segmentation_map;
// segment threashold for encode breakout
- int segment_encode_breakout[MAX_SEGMENTS];
+ int segment_encode_breakout[MAX_SEGMENTS];
CYCLIC_REFRESH *cyclic_refresh;
ActiveMap active_map;
@@ -392,11 +386,10 @@
YV12_BUFFER_CONFIG alt_ref_buffer;
-
#if CONFIG_INTERNAL_STATS
unsigned int mode_chosen_counts[MAX_MODES];
- int count;
+ int count;
uint64_t total_sq_error;
uint64_t total_samples;
ImageStat psnr;
@@ -408,7 +401,7 @@
double total_blockiness;
double worst_blockiness;
- int bytes;
+ int bytes;
double summed_quality;
double summed_weights;
double summedp_quality;
@@ -491,20 +484,20 @@
void vp10_initialize_enc(void);
struct VP10_COMP *vp10_create_compressor(VP10EncoderConfig *oxcf,
- BufferPool *const pool);
+ BufferPool *const pool);
void vp10_remove_compressor(VP10_COMP *cpi);
void vp10_change_config(VP10_COMP *cpi, const VP10EncoderConfig *oxcf);
- // receive a frames worth of data. caller can assume that a copy of this
- // frame is made and not just a copy of the pointer..
+// receive a frames worth of data. caller can assume that a copy of this
+// frame is made and not just a copy of the pointer..
int vp10_receive_raw_frame(VP10_COMP *cpi, unsigned int frame_flags,
- YV12_BUFFER_CONFIG *sd, int64_t time_stamp,
- int64_t end_time_stamp);
+ YV12_BUFFER_CONFIG *sd, int64_t time_stamp,
+ int64_t end_time_stamp);
int vp10_get_compressed_data(VP10_COMP *cpi, unsigned int *frame_flags,
- size_t *size, uint8_t *dest,
- int64_t *time_stamp, int64_t *time_end, int flush);
+ size_t *size, uint8_t *dest, int64_t *time_stamp,
+ int64_t *time_end, int flush);
int vp10_get_preview_raw_frame(VP10_COMP *cpi, YV12_BUFFER_CONFIG *dest);
@@ -513,10 +506,10 @@
void vp10_update_reference(VP10_COMP *cpi, int ref_frame_flags);
int vp10_copy_reference_enc(VP10_COMP *cpi, VPX_REFFRAME ref_frame_flag,
- YV12_BUFFER_CONFIG *sd);
+ YV12_BUFFER_CONFIG *sd);
int vp10_set_reference_enc(VP10_COMP *cpi, VPX_REFFRAME ref_frame_flag,
- YV12_BUFFER_CONFIG *sd);
+ YV12_BUFFER_CONFIG *sd);
int vp10_update_entropy(VP10_COMP *cpi, int update);
@@ -524,17 +517,16 @@
int vp10_get_active_map(VP10_COMP *cpi, unsigned char *map, int rows, int cols);
-int vp10_set_internal_size(VP10_COMP *cpi,
- VPX_SCALING horiz_mode, VPX_SCALING vert_mode);
+int vp10_set_internal_size(VP10_COMP *cpi, VPX_SCALING horiz_mode,
+ VPX_SCALING vert_mode);
int vp10_set_size_literal(VP10_COMP *cpi, unsigned int width,
- unsigned int height);
+ unsigned int height);
int vp10_get_quantizer(struct VP10_COMP *cpi);
static INLINE int frame_is_kf_gf_arf(const VP10_COMP *cpi) {
- return frame_is_intra_only(&cpi->common) ||
- cpi->refresh_alt_ref_frame ||
+ return frame_is_intra_only(&cpi->common) || cpi->refresh_alt_ref_frame ||
(cpi->refresh_golden_frame && !cpi->rc.is_src_frame_alt_ref);
}
@@ -560,8 +552,8 @@
VP10_COMP *cpi, MV_REFERENCE_FRAME ref_frame) {
VP10_COMMON *const cm = &cpi->common;
const int buf_idx = get_ref_frame_buf_idx(cpi, ref_frame);
- return
- buf_idx != INVALID_IDX ? &cm->buffer_pool->frame_bufs[buf_idx].buf : NULL;
+ return buf_idx != INVALID_IDX ? &cm->buffer_pool->frame_bufs[buf_idx].buf
+ : NULL;
}
static INLINE int get_token_alloc(int mb_rows, int mb_cols) {
@@ -582,10 +574,11 @@
return get_token_alloc(tile_mb_rows, tile_mb_cols);
}
-int64_t vp10_get_y_sse(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b);
+int64_t vp10_get_y_sse(const YV12_BUFFER_CONFIG *a,
+ const YV12_BUFFER_CONFIG *b);
#if CONFIG_VPX_HIGHBITDEPTH
int64_t vp10_highbd_get_y_sse(const YV12_BUFFER_CONFIG *a,
- const YV12_BUFFER_CONFIG *b);
+ const YV12_BUFFER_CONFIG *b);
#endif // CONFIG_VPX_HIGHBITDEPTH
void vp10_alloc_compressor_data(VP10_COMP *cpi);
@@ -597,12 +590,12 @@
void vp10_set_high_precision_mv(VP10_COMP *cpi, int allow_high_precision_mv);
YV12_BUFFER_CONFIG *vp10_scale_if_required_fast(VP10_COMMON *cm,
- YV12_BUFFER_CONFIG *unscaled,
- YV12_BUFFER_CONFIG *scaled);
+ YV12_BUFFER_CONFIG *unscaled,
+ YV12_BUFFER_CONFIG *scaled);
YV12_BUFFER_CONFIG *vp10_scale_if_required(VP10_COMMON *cm,
- YV12_BUFFER_CONFIG *unscaled,
- YV12_BUFFER_CONFIG *scaled);
+ YV12_BUFFER_CONFIG *unscaled,
+ YV12_BUFFER_CONFIG *scaled);
void vp10_apply_encoding_flags(VP10_COMP *cpi, vpx_enc_frame_flags_t flags);
@@ -614,10 +607,10 @@
static INLINE void set_ref_ptrs(VP10_COMMON *cm, MACROBLOCKD *xd,
MV_REFERENCE_FRAME ref0,
MV_REFERENCE_FRAME ref1) {
- xd->block_refs[0] = &cm->frame_refs[ref0 >= LAST_FRAME ? ref0 - LAST_FRAME
- : 0];
- xd->block_refs[1] = &cm->frame_refs[ref1 >= LAST_FRAME ? ref1 - LAST_FRAME
- : 0];
+ xd->block_refs[0] =
+ &cm->frame_refs[ref0 >= LAST_FRAME ? ref0 - LAST_FRAME : 0];
+ xd->block_refs[1] =
+ &cm->frame_refs[ref1 >= LAST_FRAME ? ref1 - LAST_FRAME : 0];
}
static INLINE int get_chessboard_index(const int frame_index) {
diff --git a/vp10/encoder/ethread.c b/vp10/encoder/ethread.c
index ad47ccf..21f9d71 100644
--- a/vp10/encoder/ethread.c
+++ b/vp10/encoder/ethread.c
@@ -31,7 +31,6 @@
td->rd_counts.coef_counts[i][j][k][l][m][n] +=
td_t->rd_counts.coef_counts[i][j][k][l][m][n];
-
// Counts of all motion searches and exhuastive mesh searches.
td->rd_counts.m_search_count += td_t->rd_counts.m_search_count;
td->rd_counts.ex_search_count += td_t->rd_counts.ex_search_count;
@@ -44,10 +43,10 @@
const int tile_rows = 1 << cm->log2_tile_rows;
int t;
- (void) unused;
+ (void)unused;
for (t = thread_data->start; t < tile_rows * tile_cols;
- t += cpi->num_workers) {
+ t += cpi->num_workers) {
int tile_row = t / tile_cols;
int tile_col = t % tile_cols;
@@ -74,8 +73,7 @@
vpx_malloc(allocated_workers * sizeof(*cpi->workers)));
CHECK_MEM_ERROR(cm, cpi->tile_thr_data,
- vpx_calloc(allocated_workers,
- sizeof(*cpi->tile_thr_data)));
+ vpx_calloc(allocated_workers, sizeof(*cpi->tile_thr_data)));
for (i = 0; i < allocated_workers; i++) {
VPxWorker *const worker = &cpi->workers[i];
@@ -122,7 +120,7 @@
worker->hook = (VPxWorkerHook)enc_worker_hook;
worker->data1 = &cpi->tile_thr_data[i];
worker->data2 = NULL;
- thread_data = (EncWorkerData*)worker->data1;
+ thread_data = (EncWorkerData *)worker->data1;
// Before encoding a frame, copy the thread data from cpi.
if (thread_data->td != &cpi->td) {
@@ -138,7 +136,7 @@
// Encode a frame
for (i = 0; i < num_workers; i++) {
VPxWorker *const worker = &cpi->workers[i];
- EncWorkerData *const thread_data = (EncWorkerData*)worker->data1;
+ EncWorkerData *const thread_data = (EncWorkerData *)worker->data1;
// Set the starting tile for each thread.
thread_data->start = i;
@@ -157,7 +155,7 @@
for (i = 0; i < num_workers; i++) {
VPxWorker *const worker = &cpi->workers[i];
- EncWorkerData *const thread_data = (EncWorkerData*)worker->data1;
+ EncWorkerData *const thread_data = (EncWorkerData *)worker->data1;
// Accumulate counters.
if (i < cpi->num_workers - 1) {
diff --git a/vp10/encoder/extend.c b/vp10/encoder/extend.c
index b6d32c6..3b0c20d 100644
--- a/vp10/encoder/extend.c
+++ b/vp10/encoder/extend.c
@@ -16,8 +16,7 @@
#include "vp10/encoder/extend.h"
static void copy_and_extend_plane(const uint8_t *src, int src_pitch,
- uint8_t *dst, int dst_pitch,
- int w, int h,
+ uint8_t *dst, int dst_pitch, int w, int h,
int extend_top, int extend_left,
int extend_bottom, int extend_right) {
int i, linesize;
@@ -43,7 +42,7 @@
src_ptr1 = dst - extend_left;
src_ptr2 = dst + dst_pitch * (h - 1) - extend_left;
dst_ptr1 = dst + dst_pitch * (-extend_top) - extend_left;
- dst_ptr2 = dst + dst_pitch * (h) - extend_left;
+ dst_ptr2 = dst + dst_pitch * (h)-extend_left;
linesize = extend_left + extend_right + w;
for (i = 0; i < extend_top; i++) {
@@ -59,9 +58,8 @@
#if CONFIG_VPX_HIGHBITDEPTH
static void highbd_copy_and_extend_plane(const uint8_t *src8, int src_pitch,
- uint8_t *dst8, int dst_pitch,
- int w, int h,
- int extend_top, int extend_left,
+ uint8_t *dst8, int dst_pitch, int w,
+ int h, int extend_top, int extend_left,
int extend_bottom, int extend_right) {
int i, linesize;
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
@@ -88,7 +86,7 @@
src_ptr1 = dst - extend_left;
src_ptr2 = dst + dst_pitch * (h - 1) - extend_left;
dst_ptr1 = dst + dst_pitch * (-extend_top) - extend_left;
- dst_ptr2 = dst + dst_pitch * (h) - extend_left;
+ dst_ptr2 = dst + dst_pitch * (h)-extend_left;
linesize = extend_left + extend_right + w;
for (i = 0; i < extend_top; i++) {
@@ -104,7 +102,7 @@
#endif // CONFIG_VPX_HIGHBITDEPTH
void vp10_copy_and_extend_frame(const YV12_BUFFER_CONFIG *src,
- YV12_BUFFER_CONFIG *dst) {
+ YV12_BUFFER_CONFIG *dst) {
// Extend src frame in buffer
// Altref filtering assumes 16 pixel extension
const int et_y = 16;
@@ -127,51 +125,46 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (src->flags & YV12_FLAG_HIGHBITDEPTH) {
- highbd_copy_and_extend_plane(src->y_buffer, src->y_stride,
- dst->y_buffer, dst->y_stride,
- src->y_crop_width, src->y_crop_height,
- et_y, el_y, eb_y, er_y);
+ highbd_copy_and_extend_plane(src->y_buffer, src->y_stride, dst->y_buffer,
+ dst->y_stride, src->y_crop_width,
+ src->y_crop_height, et_y, el_y, eb_y, er_y);
- highbd_copy_and_extend_plane(src->u_buffer, src->uv_stride,
- dst->u_buffer, dst->uv_stride,
- src->uv_crop_width, src->uv_crop_height,
- et_uv, el_uv, eb_uv, er_uv);
+ highbd_copy_and_extend_plane(
+ src->u_buffer, src->uv_stride, dst->u_buffer, dst->uv_stride,
+ src->uv_crop_width, src->uv_crop_height, et_uv, el_uv, eb_uv, er_uv);
- highbd_copy_and_extend_plane(src->v_buffer, src->uv_stride,
- dst->v_buffer, dst->uv_stride,
- src->uv_crop_width, src->uv_crop_height,
- et_uv, el_uv, eb_uv, er_uv);
+ highbd_copy_and_extend_plane(
+ src->v_buffer, src->uv_stride, dst->v_buffer, dst->uv_stride,
+ src->uv_crop_width, src->uv_crop_height, et_uv, el_uv, eb_uv, er_uv);
return;
}
#endif // CONFIG_VPX_HIGHBITDEPTH
- copy_and_extend_plane(src->y_buffer, src->y_stride,
- dst->y_buffer, dst->y_stride,
- src->y_crop_width, src->y_crop_height,
+ copy_and_extend_plane(src->y_buffer, src->y_stride, dst->y_buffer,
+ dst->y_stride, src->y_crop_width, src->y_crop_height,
et_y, el_y, eb_y, er_y);
- copy_and_extend_plane(src->u_buffer, src->uv_stride,
- dst->u_buffer, dst->uv_stride,
- src->uv_crop_width, src->uv_crop_height,
+ copy_and_extend_plane(src->u_buffer, src->uv_stride, dst->u_buffer,
+ dst->uv_stride, src->uv_crop_width, src->uv_crop_height,
et_uv, el_uv, eb_uv, er_uv);
- copy_and_extend_plane(src->v_buffer, src->uv_stride,
- dst->v_buffer, dst->uv_stride,
- src->uv_crop_width, src->uv_crop_height,
+ copy_and_extend_plane(src->v_buffer, src->uv_stride, dst->v_buffer,
+ dst->uv_stride, src->uv_crop_width, src->uv_crop_height,
et_uv, el_uv, eb_uv, er_uv);
}
void vp10_copy_and_extend_frame_with_rect(const YV12_BUFFER_CONFIG *src,
- YV12_BUFFER_CONFIG *dst,
- int srcy, int srcx,
- int srch, int srcw) {
+ YV12_BUFFER_CONFIG *dst, int srcy,
+ int srcx, int srch, int srcw) {
// If the side is not touching the bounder then don't extend.
const int et_y = srcy ? 0 : dst->border;
const int el_y = srcx ? 0 : dst->border;
- const int eb_y = srcy + srch != src->y_height ? 0 :
- dst->border + dst->y_height - src->y_height;
- const int er_y = srcx + srcw != src->y_width ? 0 :
- dst->border + dst->y_width - src->y_width;
+ const int eb_y = srcy + srch != src->y_height
+ ? 0
+ : dst->border + dst->y_height - src->y_height;
+ const int er_y = srcx + srcw != src->y_width
+ ? 0
+ : dst->border + dst->y_width - src->y_width;
const int src_y_offset = srcy * src->y_stride + srcx;
const int dst_y_offset = srcy * dst->y_stride + srcx;
@@ -185,17 +178,14 @@
const int srcw_uv = ROUND_POWER_OF_TWO(srcw, 1);
copy_and_extend_plane(src->y_buffer + src_y_offset, src->y_stride,
- dst->y_buffer + dst_y_offset, dst->y_stride,
- srcw, srch,
+ dst->y_buffer + dst_y_offset, dst->y_stride, srcw, srch,
et_y, el_y, eb_y, er_y);
copy_and_extend_plane(src->u_buffer + src_uv_offset, src->uv_stride,
- dst->u_buffer + dst_uv_offset, dst->uv_stride,
- srcw_uv, srch_uv,
- et_uv, el_uv, eb_uv, er_uv);
+ dst->u_buffer + dst_uv_offset, dst->uv_stride, srcw_uv,
+ srch_uv, et_uv, el_uv, eb_uv, er_uv);
copy_and_extend_plane(src->v_buffer + src_uv_offset, src->uv_stride,
- dst->v_buffer + dst_uv_offset, dst->uv_stride,
- srcw_uv, srch_uv,
- et_uv, el_uv, eb_uv, er_uv);
+ dst->v_buffer + dst_uv_offset, dst->uv_stride, srcw_uv,
+ srch_uv, et_uv, el_uv, eb_uv, er_uv);
}
diff --git a/vp10/encoder/extend.h b/vp10/encoder/extend.h
index 6f502ef..f7967fd 100644
--- a/vp10/encoder/extend.h
+++ b/vp10/encoder/extend.h
@@ -18,14 +18,12 @@
extern "C" {
#endif
-
void vp10_copy_and_extend_frame(const YV12_BUFFER_CONFIG *src,
- YV12_BUFFER_CONFIG *dst);
+ YV12_BUFFER_CONFIG *dst);
void vp10_copy_and_extend_frame_with_rect(const YV12_BUFFER_CONFIG *src,
- YV12_BUFFER_CONFIG *dst,
- int srcy, int srcx,
- int srch, int srcw);
+ YV12_BUFFER_CONFIG *dst, int srcy,
+ int srcx, int srch, int srcw);
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/vp10/encoder/firstpass.c b/vp10/encoder/firstpass.c
index 450c939..b5c728c 100644
--- a/vp10/encoder/firstpass.c
+++ b/vp10/encoder/firstpass.c
@@ -38,35 +38,34 @@
#include "vp10/encoder/rd.h"
#include "vpx_dsp/variance.h"
-#define OUTPUT_FPF 0
-#define ARF_STATS_OUTPUT 0
+#define OUTPUT_FPF 0
+#define ARF_STATS_OUTPUT 0
#define GROUP_ADAPTIVE_MAXQ 1
-#define BOOST_BREAKOUT 12.5
-#define BOOST_FACTOR 12.5
-#define ERR_DIVISOR 128.0
-#define FACTOR_PT_LOW 0.70
-#define FACTOR_PT_HIGH 0.90
-#define FIRST_PASS_Q 10.0
-#define GF_MAX_BOOST 96.0
-#define INTRA_MODE_PENALTY 1024
-#define KF_MAX_BOOST 128.0
-#define MIN_ARF_GF_BOOST 240
-#define MIN_DECAY_FACTOR 0.01
-#define MIN_KF_BOOST 300
+#define BOOST_BREAKOUT 12.5
+#define BOOST_FACTOR 12.5
+#define ERR_DIVISOR 128.0
+#define FACTOR_PT_LOW 0.70
+#define FACTOR_PT_HIGH 0.90
+#define FIRST_PASS_Q 10.0
+#define GF_MAX_BOOST 96.0
+#define INTRA_MODE_PENALTY 1024
+#define KF_MAX_BOOST 128.0
+#define MIN_ARF_GF_BOOST 240
+#define MIN_DECAY_FACTOR 0.01
+#define MIN_KF_BOOST 300
#define NEW_MV_MODE_PENALTY 32
-#define DARK_THRESH 64
-#define DEFAULT_GRP_WEIGHT 1.0
-#define RC_FACTOR_MIN 0.75
-#define RC_FACTOR_MAX 1.75
-
+#define DARK_THRESH 64
+#define DEFAULT_GRP_WEIGHT 1.0
+#define RC_FACTOR_MIN 0.75
+#define RC_FACTOR_MAX 1.75
#define NCOUNT_INTRA_THRESH 8192
#define NCOUNT_INTRA_FACTOR 3
#define NCOUNT_FRAME_II_THRESH 5.0
-#define DOUBLE_DIVIDE_CHECK(x) ((x) < 0 ? (x) - 0.000001 : (x) + 0.000001)
+#define DOUBLE_DIVIDE_CHECK(x) ((x) < 0 ? (x)-0.000001 : (x) + 0.000001)
#if ARF_STATS_OUTPUT
unsigned int arf_count = 0;
@@ -74,8 +73,7 @@
// Resets the first pass file to the given position using a relative seek from
// the current position.
-static void reset_fpf_position(TWO_PASS *p,
- const FIRSTPASS_STATS *position) {
+static void reset_fpf_position(TWO_PASS *p, const FIRSTPASS_STATS *position) {
p->stats_in = position;
}
@@ -90,8 +88,7 @@
}
static int input_stats(TWO_PASS *p, FIRSTPASS_STATS *fps) {
- if (p->stats_in >= p->stats_in_end)
- return EOF;
+ if (p->stats_in >= p->stats_in_end) return EOF;
*fps = *p->stats_in;
++p->stats_in;
@@ -112,39 +109,24 @@
FILE *fpfile;
fpfile = fopen("firstpass.stt", "a");
- fprintf(fpfile, "%12.0lf %12.4lf %12.0lf %12.0lf %12.0lf %12.4lf %12.4lf"
+ fprintf(fpfile,
+ "%12.0lf %12.4lf %12.0lf %12.0lf %12.0lf %12.4lf %12.4lf"
"%12.4lf %12.4lf %12.4lf %12.4lf %12.4lf %12.4lf %12.4lf %12.4lf"
"%12.4lf %12.4lf %12.0lf %12.0lf %12.0lf %12.4lf\n",
- stats->frame,
- stats->weight,
- stats->intra_error,
- stats->coded_error,
- stats->sr_coded_error,
- stats->pcnt_inter,
- stats->pcnt_motion,
- stats->pcnt_second_ref,
- stats->pcnt_neutral,
- stats->intra_skip_pct,
- stats->inactive_zone_rows,
- stats->inactive_zone_cols,
- stats->MVr,
- stats->mvr_abs,
- stats->MVc,
- stats->mvc_abs,
- stats->MVrv,
- stats->MVcv,
- stats->mv_in_out_count,
- stats->new_mv_count,
- stats->count,
- stats->duration);
+ stats->frame, stats->weight, stats->intra_error, stats->coded_error,
+ stats->sr_coded_error, stats->pcnt_inter, stats->pcnt_motion,
+ stats->pcnt_second_ref, stats->pcnt_neutral, stats->intra_skip_pct,
+ stats->inactive_zone_rows, stats->inactive_zone_cols, stats->MVr,
+ stats->mvr_abs, stats->MVc, stats->mvc_abs, stats->MVrv,
+ stats->MVcv, stats->mv_in_out_count, stats->new_mv_count,
+ stats->count, stats->duration);
fclose(fpfile);
}
#endif
}
#if CONFIG_FP_MB_STATS
-static void output_fpmb_stats(uint8_t *this_frame_mb_stats,
- VP10_COMMON *cm,
+static void output_fpmb_stats(uint8_t *this_frame_mb_stats, VP10_COMMON *cm,
struct vpx_codec_pkt_list *pktlist) {
struct vpx_codec_cx_pkt pkt;
pkt.kind = VPX_CODEC_FPMB_STATS_PKT;
@@ -160,23 +142,23 @@
section->intra_error = 0.0;
section->coded_error = 0.0;
section->sr_coded_error = 0.0;
- section->pcnt_inter = 0.0;
- section->pcnt_motion = 0.0;
+ section->pcnt_inter = 0.0;
+ section->pcnt_motion = 0.0;
section->pcnt_second_ref = 0.0;
section->pcnt_neutral = 0.0;
section->intra_skip_pct = 0.0;
section->inactive_zone_rows = 0.0;
section->inactive_zone_cols = 0.0;
section->MVr = 0.0;
- section->mvr_abs = 0.0;
- section->MVc = 0.0;
- section->mvc_abs = 0.0;
- section->MVrv = 0.0;
- section->MVcv = 0.0;
- section->mv_in_out_count = 0.0;
+ section->mvr_abs = 0.0;
+ section->MVc = 0.0;
+ section->mvc_abs = 0.0;
+ section->MVrv = 0.0;
+ section->MVcv = 0.0;
+ section->mv_in_out_count = 0.0;
section->new_mv_count = 0.0;
- section->count = 0.0;
- section->duration = 1.0;
+ section->count = 0.0;
+ section->duration = 1.0;
}
static void accumulate_stats(FIRSTPASS_STATS *section,
@@ -186,7 +168,7 @@
section->intra_error += frame->intra_error;
section->coded_error += frame->coded_error;
section->sr_coded_error += frame->sr_coded_error;
- section->pcnt_inter += frame->pcnt_inter;
+ section->pcnt_inter += frame->pcnt_inter;
section->pcnt_motion += frame->pcnt_motion;
section->pcnt_second_ref += frame->pcnt_second_ref;
section->pcnt_neutral += frame->pcnt_neutral;
@@ -194,15 +176,15 @@
section->inactive_zone_rows += frame->inactive_zone_rows;
section->inactive_zone_cols += frame->inactive_zone_cols;
section->MVr += frame->MVr;
- section->mvr_abs += frame->mvr_abs;
- section->MVc += frame->MVc;
- section->mvc_abs += frame->mvc_abs;
- section->MVrv += frame->MVrv;
- section->MVcv += frame->MVcv;
- section->mv_in_out_count += frame->mv_in_out_count;
+ section->mvr_abs += frame->mvr_abs;
+ section->MVc += frame->MVc;
+ section->mvc_abs += frame->mvc_abs;
+ section->MVrv += frame->MVrv;
+ section->MVcv += frame->MVcv;
+ section->mv_in_out_count += frame->mv_in_out_count;
section->new_mv_count += frame->new_mv_count;
- section->count += frame->count;
- section->duration += frame->duration;
+ section->count += frame->count;
+ section->duration += frame->duration;
}
static void subtract_stats(FIRSTPASS_STATS *section,
@@ -212,7 +194,7 @@
section->intra_error -= frame->intra_error;
section->coded_error -= frame->coded_error;
section->sr_coded_error -= frame->sr_coded_error;
- section->pcnt_inter -= frame->pcnt_inter;
+ section->pcnt_inter -= frame->pcnt_inter;
section->pcnt_motion -= frame->pcnt_motion;
section->pcnt_second_ref -= frame->pcnt_second_ref;
section->pcnt_neutral -= frame->pcnt_neutral;
@@ -220,15 +202,15 @@
section->inactive_zone_rows -= frame->inactive_zone_rows;
section->inactive_zone_cols -= frame->inactive_zone_cols;
section->MVr -= frame->MVr;
- section->mvr_abs -= frame->mvr_abs;
- section->MVc -= frame->MVc;
- section->mvc_abs -= frame->mvc_abs;
- section->MVrv -= frame->MVrv;
- section->MVcv -= frame->MVcv;
- section->mv_in_out_count -= frame->mv_in_out_count;
+ section->mvr_abs -= frame->mvr_abs;
+ section->MVc -= frame->MVc;
+ section->mvc_abs -= frame->mvc_abs;
+ section->MVrv -= frame->MVrv;
+ section->MVcv -= frame->MVcv;
+ section->mv_in_out_count -= frame->mv_in_out_count;
section->new_mv_count -= frame->new_mv_count;
- section->count -= frame->count;
- section->duration -= frame->duration;
+ section->count -= frame->count;
+ section->duration -= frame->duration;
}
// Calculate an active area of the image that discounts formatting
@@ -236,13 +218,13 @@
#define MIN_ACTIVE_AREA 0.5
#define MAX_ACTIVE_AREA 1.0
static double calculate_active_area(const VP10_COMP *cpi,
- const FIRSTPASS_STATS *this_frame)
-{
+ const FIRSTPASS_STATS *this_frame) {
double active_pct;
- active_pct = 1.0 -
- ((this_frame->intra_skip_pct / 2) +
- ((this_frame->inactive_zone_rows * 2) / (double)cpi->common.mb_rows));
+ active_pct =
+ 1.0 -
+ ((this_frame->intra_skip_pct / 2) +
+ ((this_frame->inactive_zone_rows * 2) / (double)cpi->common.mb_rows));
return fclamp(active_pct, MIN_ACTIVE_AREA, MAX_ACTIVE_AREA);
}
@@ -257,8 +239,9 @@
const double av_weight = stats->weight / stats->count;
const double av_err = (stats->coded_error * av_weight) / stats->count;
double modified_error =
- av_err * pow(this_frame->coded_error * this_frame->weight /
- DOUBLE_DIVIDE_CHECK(av_err), oxcf->two_pass_vbrbias / 100.0);
+ av_err * pow(this_frame->coded_error * this_frame->weight /
+ DOUBLE_DIVIDE_CHECK(av_err),
+ oxcf->two_pass_vbrbias / 100.0);
// Correction for active area. Frames with a reduced active area
// (eg due to formatting bars) have a higher error per mb for the
@@ -266,17 +249,18 @@
// 0.5N blocks of complexity 2X is a little easier than coding N
// blocks of complexity X.
modified_error *=
- pow(calculate_active_area(cpi, this_frame), ACT_AREA_CORRECTION);
+ pow(calculate_active_area(cpi, this_frame), ACT_AREA_CORRECTION);
- return fclamp(modified_error,
- twopass->modified_error_min, twopass->modified_error_max);
+ return fclamp(modified_error, twopass->modified_error_min,
+ twopass->modified_error_max);
}
// This function returns the maximum target rate per frame.
static int frame_max_bits(const RATE_CONTROL *rc,
const VP10EncoderConfig *oxcf) {
int64_t max_bits = ((int64_t)rc->avg_frame_bandwidth *
- (int64_t)oxcf->two_pass_vbrmax_section) / 100;
+ (int64_t)oxcf->two_pass_vbrmax_section) /
+ 100;
if (max_bits < 0)
max_bits = 0;
else if (max_bits > rc->max_frame_bandwidth)
@@ -295,14 +279,10 @@
static vpx_variance_fn_t get_block_variance_fn(BLOCK_SIZE bsize) {
switch (bsize) {
- case BLOCK_8X8:
- return vpx_mse8x8;
- case BLOCK_16X8:
- return vpx_mse16x8;
- case BLOCK_8X16:
- return vpx_mse8x16;
- default:
- return vpx_mse16x16;
+ case BLOCK_8X8: return vpx_mse8x8;
+ case BLOCK_16X8: return vpx_mse16x8;
+ case BLOCK_8X16: return vpx_mse8x16;
+ default: return vpx_mse16x16;
}
}
@@ -321,38 +301,26 @@
switch (bd) {
default:
switch (bsize) {
- case BLOCK_8X8:
- return vpx_highbd_8_mse8x8;
- case BLOCK_16X8:
- return vpx_highbd_8_mse16x8;
- case BLOCK_8X16:
- return vpx_highbd_8_mse8x16;
- default:
- return vpx_highbd_8_mse16x16;
+ case BLOCK_8X8: return vpx_highbd_8_mse8x8;
+ case BLOCK_16X8: return vpx_highbd_8_mse16x8;
+ case BLOCK_8X16: return vpx_highbd_8_mse8x16;
+ default: return vpx_highbd_8_mse16x16;
}
break;
case 10:
switch (bsize) {
- case BLOCK_8X8:
- return vpx_highbd_10_mse8x8;
- case BLOCK_16X8:
- return vpx_highbd_10_mse16x8;
- case BLOCK_8X16:
- return vpx_highbd_10_mse8x16;
- default:
- return vpx_highbd_10_mse16x16;
+ case BLOCK_8X8: return vpx_highbd_10_mse8x8;
+ case BLOCK_16X8: return vpx_highbd_10_mse16x8;
+ case BLOCK_8X16: return vpx_highbd_10_mse8x16;
+ default: return vpx_highbd_10_mse16x16;
}
break;
case 12:
switch (bsize) {
- case BLOCK_8X8:
- return vpx_highbd_12_mse8x8;
- case BLOCK_16X8:
- return vpx_highbd_12_mse16x8;
- case BLOCK_8X16:
- return vpx_highbd_12_mse8x16;
- default:
- return vpx_highbd_12_mse16x16;
+ case BLOCK_8X8: return vpx_highbd_12_mse8x8;
+ case BLOCK_16X8: return vpx_highbd_12_mse16x8;
+ case BLOCK_8X16: return vpx_highbd_12_mse8x16;
+ default: return vpx_highbd_12_mse16x16;
}
break;
}
@@ -375,8 +343,7 @@
int sr = 0;
const int dim = VPXMIN(cpi->initial_width, cpi->initial_height);
- while ((dim << sr) < MAX_FULL_PEL_VAL)
- ++sr;
+ while ((dim << sr) < MAX_FULL_PEL_VAL) ++sr;
return sr;
}
@@ -384,8 +351,8 @@
const MV *ref_mv, MV *best_mv,
int *best_motion_err) {
MACROBLOCKD *const xd = &x->e_mbd;
- MV tmp_mv = {0, 0};
- MV ref_mv_full = {ref_mv->row >> 3, ref_mv->col >> 3};
+ MV tmp_mv = { 0, 0 };
+ MV ref_mv_full = { ref_mv->row >> 3, ref_mv->col >> 3 };
int num00, tmp_err, n;
const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
vpx_variance_fn_ptr_t v_fn_ptr = cpi->fn_ptr[bsize];
@@ -407,12 +374,11 @@
// Center the initial step/diamond search on best mv.
tmp_err = cpi->diamond_search_sad(x, &cpi->ss_cfg, &ref_mv_full, &tmp_mv,
- step_param,
- x->sadperbit16, &num00, &v_fn_ptr, ref_mv);
+ step_param, x->sadperbit16, &num00,
+ &v_fn_ptr, ref_mv);
if (tmp_err < INT_MAX)
tmp_err = vp10_get_mvpred_var(x, &tmp_mv, ref_mv, &v_fn_ptr, 1);
- if (tmp_err < INT_MAX - new_mv_mode_penalty)
- tmp_err += new_mv_mode_penalty;
+ if (tmp_err < INT_MAX - new_mv_mode_penalty) tmp_err += new_mv_mode_penalty;
if (tmp_err < *best_motion_err) {
*best_motion_err = tmp_err;
@@ -430,8 +396,8 @@
--num00;
} else {
tmp_err = cpi->diamond_search_sad(x, &cpi->ss_cfg, &ref_mv_full, &tmp_mv,
- step_param + n, x->sadperbit16,
- &num00, &v_fn_ptr, ref_mv);
+ step_param + n, x->sadperbit16, &num00,
+ &v_fn_ptr, ref_mv);
if (tmp_err < INT_MAX)
tmp_err = vp10_get_mvpred_var(x, &tmp_mv, ref_mv, &v_fn_ptr, 1);
if (tmp_err < INT_MAX - new_mv_mode_penalty)
@@ -447,11 +413,9 @@
static BLOCK_SIZE get_bsize(const VP10_COMMON *cm, int mb_row, int mb_col) {
if (2 * mb_col + 1 < cm->mi_cols) {
- return 2 * mb_row + 1 < cm->mi_rows ? BLOCK_16X16
- : BLOCK_16X8;
+ return 2 * mb_row + 1 < cm->mi_rows ? BLOCK_16X16 : BLOCK_16X8;
} else {
- return 2 * mb_row + 1 < cm->mi_rows ? BLOCK_8X16
- : BLOCK_8X8;
+ return 2 * mb_row + 1 < cm->mi_rows ? BLOCK_8X16 : BLOCK_8X8;
}
}
@@ -459,11 +423,9 @@
int i;
for (i = 0; i < QINDEX_RANGE; ++i)
- if (vp10_convert_qindex_to_q(i, bit_depth) >= FIRST_PASS_Q)
- break;
+ if (vp10_convert_qindex_to_q(i, bit_depth) >= FIRST_PASS_Q) break;
- if (i == QINDEX_RANGE)
- i--;
+ if (i == QINDEX_RANGE) i--;
return i;
}
@@ -471,8 +433,7 @@
static void set_first_pass_params(VP10_COMP *cpi) {
VP10_COMMON *const cm = &cpi->common;
if (!cpi->refresh_alt_ref_frame &&
- (cm->current_video_frame == 0 ||
- (cpi->frame_flags & FRAMEFLAGS_KEY))) {
+ (cm->current_video_frame == 0 || (cpi->frame_flags & FRAMEFLAGS_KEY))) {
cm->frame_type = KEY_FRAME;
} else {
cm->frame_type = INTER_FRAME;
@@ -511,9 +472,9 @@
int image_data_start_row = INVALID_ROW;
int new_mv_count = 0;
int sum_in_vectors = 0;
- MV lastmv = {0, 0};
+ MV lastmv = { 0, 0 };
TWO_PASS *twopass = &cpi->twopass;
- const MV zero_mv = {0, 0};
+ const MV zero_mv = { 0, 0 };
int recon_y_stride, recon_uv_stride, uv_mb_height;
YV12_BUFFER_CONFIG *const lst_yv12 = get_ref_frame_buffer(cpi, LAST_FRAME);
@@ -576,7 +537,7 @@
uv_mb_height = 16 >> (new_yv12->y_height > new_yv12->uv_height);
for (mb_row = 0; mb_row < cm->mb_rows; ++mb_row) {
- MV best_ref_mv = {0, 0};
+ MV best_ref_mv = { 0, 0 };
// Reset above block coeffs.
xd->up_available = (mb_row != 0);
@@ -586,8 +547,7 @@
// Set up limit values for motion vectors to prevent them extending
// outside the UMV borders.
x->mv_row_min = -((mb_row * 16) + BORDER_MV_PIXELS_B16);
- x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16)
- + BORDER_MV_PIXELS_B16;
+ x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + BORDER_MV_PIXELS_B16;
for (mb_col = 0; mb_col < cm->mb_cols; ++mb_col) {
int this_error;
@@ -608,16 +568,15 @@
xd->left_available = (mb_col != 0);
xd->mi[0]->mbmi.sb_type = bsize;
xd->mi[0]->mbmi.ref_frame[0] = INTRA_FRAME;
- set_mi_row_col(xd, &tile,
- mb_row << 1, num_8x8_blocks_high_lookup[bsize],
+ set_mi_row_col(xd, &tile, mb_row << 1, num_8x8_blocks_high_lookup[bsize],
mb_col << 1, num_8x8_blocks_wide_lookup[bsize],
cm->mi_rows, cm->mi_cols);
// Do intra 16x16 prediction.
xd->mi[0]->mbmi.segment_id = 0;
xd->mi[0]->mbmi.mode = DC_PRED;
- xd->mi[0]->mbmi.tx_size = use_dc_pred ?
- (bsize >= BLOCK_16X16 ? TX_16X16 : TX_8X8) : TX_4X4;
+ xd->mi[0]->mbmi.tx_size =
+ use_dc_pred ? (bsize >= BLOCK_16X16 ? TX_16X16 : TX_8X8) : TX_4X4;
vp10_encode_intra_block_plane(x, bsize, 0);
this_error = vpx_get_mb_ss(x->plane[0].src_diff);
@@ -635,17 +594,13 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (cm->use_highbitdepth) {
switch (cm->bit_depth) {
- case VPX_BITS_8:
- break;
- case VPX_BITS_10:
- this_error >>= 4;
- break;
- case VPX_BITS_12:
- this_error >>= 8;
- break;
+ case VPX_BITS_8: break;
+ case VPX_BITS_10: this_error >>= 4; break;
+ case VPX_BITS_12: this_error >>= 8; break;
default:
- assert(0 && "cm->bit_depth should be VPX_BITS_8, "
- "VPX_BITS_10 or VPX_BITS_12");
+ assert(0 &&
+ "cm->bit_depth should be VPX_BITS_8, "
+ "VPX_BITS_10 or VPX_BITS_12");
return;
}
}
@@ -699,7 +654,7 @@
if (cm->current_video_frame > 0) {
int tmp_err, motion_error, raw_motion_error;
// Assume 0,0 motion with no mv overhead.
- MV mv = {0, 0} , tmp_mv = {0, 0};
+ MV mv = { 0, 0 }, tmp_mv = { 0, 0 };
struct buf_2d unscaled_last_source_buf_2d;
xd->plane[0].pre[0].buf = first_ref_buf->y_buffer + recon_yoffset;
@@ -708,12 +663,12 @@
motion_error = highbd_get_prediction_error(
bsize, &x->plane[0].src, &xd->plane[0].pre[0], xd->bd);
} else {
- motion_error = get_prediction_error(
- bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
+ motion_error = get_prediction_error(bsize, &x->plane[0].src,
+ &xd->plane[0].pre[0]);
}
#else
- motion_error = get_prediction_error(
- bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
+ motion_error =
+ get_prediction_error(bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
#endif // CONFIG_VPX_HIGHBITDEPTH
// Compute the motion error of the 0,0 motion using the last source
@@ -728,12 +683,12 @@
raw_motion_error = highbd_get_prediction_error(
bsize, &x->plane[0].src, &unscaled_last_source_buf_2d, xd->bd);
} else {
- raw_motion_error = get_prediction_error(
- bsize, &x->plane[0].src, &unscaled_last_source_buf_2d);
+ raw_motion_error = get_prediction_error(bsize, &x->plane[0].src,
+ &unscaled_last_source_buf_2d);
}
#else
- raw_motion_error = get_prediction_error(
- bsize, &x->plane[0].src, &unscaled_last_source_buf_2d);
+ raw_motion_error = get_prediction_error(bsize, &x->plane[0].src,
+ &unscaled_last_source_buf_2d);
#endif // CONFIG_VPX_HIGHBITDEPTH
// TODO(pengchong): Replace the hard-coded threshold
@@ -765,12 +720,12 @@
gf_motion_error = highbd_get_prediction_error(
bsize, &x->plane[0].src, &xd->plane[0].pre[0], xd->bd);
} else {
- gf_motion_error = get_prediction_error(
- bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
+ gf_motion_error = get_prediction_error(bsize, &x->plane[0].src,
+ &xd->plane[0].pre[0]);
}
#else
- gf_motion_error = get_prediction_error(
- bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
+ gf_motion_error = get_prediction_error(bsize, &x->plane[0].src,
+ &xd->plane[0].pre[0]);
#endif // CONFIG_VPX_HIGHBITDEPTH
first_pass_motion_search(cpi, x, &zero_mv, &tmp_mv,
@@ -826,12 +781,12 @@
if (((this_error - intrapenalty) * 9 <= motion_error * 10) &&
(this_error < (2 * intrapenalty))) {
neutral_count += 1.0;
- // Also track cases where the intra is not much worse than the inter
- // and use this in limiting the GF/arf group length.
+ // Also track cases where the intra is not much worse than the inter
+ // and use this in limiting the GF/arf group length.
} else if ((this_error > NCOUNT_INTRA_THRESH) &&
(this_error < (NCOUNT_INTRA_FACTOR * motion_error))) {
- neutral_count += (double)motion_error /
- DOUBLE_DIVIDE_CHECK((double)this_error);
+ neutral_count +=
+ (double)motion_error / DOUBLE_DIVIDE_CHECK((double)this_error);
}
mv.row *= 8;
@@ -901,8 +856,7 @@
#endif
// Non-zero vector, was it different from the last non zero vector?
- if (!is_equal_mv(&mv, &lastmv))
- ++new_mv_count;
+ if (!is_equal_mv(&mv, &lastmv)) ++new_mv_count;
lastmv = mv;
// Does the row vector point inwards or outwards?
@@ -948,10 +902,10 @@
// Adjust to the next row of MBs.
x->plane[0].src.buf += 16 * x->plane[0].src.stride - 16 * cm->mb_cols;
- x->plane[1].src.buf += uv_mb_height * x->plane[1].src.stride -
- uv_mb_height * cm->mb_cols;
- x->plane[2].src.buf += uv_mb_height * x->plane[1].src.stride -
- uv_mb_height * cm->mb_cols;
+ x->plane[1].src.buf +=
+ uv_mb_height * x->plane[1].src.stride - uv_mb_height * cm->mb_cols;
+ x->plane[2].src.buf +=
+ uv_mb_height * x->plane[1].src.stride - uv_mb_height * cm->mb_cols;
vpx_clear_system_state();
}
@@ -976,7 +930,8 @@
// Initial estimate here uses sqrt(mbs) to define the min_err, where the
// number of mbs is proportional to the image area.
const int num_mbs = (cpi->oxcf.resize_mode != RESIZE_NONE)
- ? cpi->initial_mbs : cpi->common.MBs;
+ ? cpi->initial_mbs
+ : cpi->common.MBs;
const double min_err = 200 * sqrt(num_mbs);
intra_factor = intra_factor / (double)num_mbs;
@@ -1000,10 +955,10 @@
fps.mvr_abs = (double)sum_mvr_abs / mvcount;
fps.MVc = (double)sum_mvc / mvcount;
fps.mvc_abs = (double)sum_mvc_abs / mvcount;
- fps.MVrv = ((double)sum_mvrs -
- ((double)sum_mvr * sum_mvr / mvcount)) / mvcount;
- fps.MVcv = ((double)sum_mvcs -
- ((double)sum_mvc * sum_mvc / mvcount)) / mvcount;
+ fps.MVrv =
+ ((double)sum_mvrs - ((double)sum_mvr * sum_mvr / mvcount)) / mvcount;
+ fps.MVcv =
+ ((double)sum_mvcs - ((double)sum_mvc * sum_mvc / mvcount)) / mvcount;
fps.mv_in_out_count = (double)sum_in_vectors / (mvcount * 2);
fps.new_mv_count = new_mv_count;
fps.pcnt_motion = (double)mvcount / num_mbs;
@@ -1084,11 +1039,8 @@
++cm->current_video_frame;
}
-static double calc_correction_factor(double err_per_mb,
- double err_divisor,
- double pt_low,
- double pt_high,
- int q,
+static double calc_correction_factor(double err_per_mb, double err_divisor,
+ double pt_low, double pt_high, int q,
vpx_bit_depth_t bit_depth) {
const double error_term = err_per_mb / err_divisor;
@@ -1097,8 +1049,7 @@
VPXMIN(vp10_convert_qindex_to_q(q, bit_depth) * 0.01 + pt_low, pt_high);
// Calculate correction factor.
- if (power_term < 1.0)
- assert(error_term >= 0.0);
+ if (power_term < 1.0) assert(error_term >= 0.0);
return fclamp(pow(error_term, power_term), 0.05, 5.0);
}
@@ -1122,35 +1073,31 @@
return rc->worst_quality; // Highest value allowed
} else {
const int num_mbs = (cpi->oxcf.resize_mode != RESIZE_NONE)
- ? cpi->initial_mbs : cpi->common.MBs;
+ ? cpi->initial_mbs
+ : cpi->common.MBs;
const int active_mbs = VPXMAX(1, num_mbs - (int)(num_mbs * inactive_zone));
const double av_err_per_mb = section_err / active_mbs;
const double speed_term = 1.0 + 0.04 * oxcf->speed;
const double ediv_size_correction = (double)num_mbs / EDIV_SIZE_FACTOR;
- const int target_norm_bits_per_mb = ((uint64_t)section_target_bandwidth <<
- BPER_MB_NORMBITS) / active_mbs;
+ const int target_norm_bits_per_mb =
+ ((uint64_t)section_target_bandwidth << BPER_MB_NORMBITS) / active_mbs;
int q;
// Try and pick a max Q that will be high enough to encode the
// content at the given rate.
for (q = rc->best_quality; q < rc->worst_quality; ++q) {
- const double factor =
- calc_correction_factor(av_err_per_mb,
- ERR_DIVISOR - ediv_size_correction,
- FACTOR_PT_LOW, FACTOR_PT_HIGH, q,
- cpi->common.bit_depth);
- const int bits_per_mb =
- vp10_rc_bits_per_mb(INTER_FRAME, q,
- factor * speed_term * group_weight_factor,
- cpi->common.bit_depth);
- if (bits_per_mb <= target_norm_bits_per_mb)
- break;
+ const double factor = calc_correction_factor(
+ av_err_per_mb, ERR_DIVISOR - ediv_size_correction, FACTOR_PT_LOW,
+ FACTOR_PT_HIGH, q, cpi->common.bit_depth);
+ const int bits_per_mb = vp10_rc_bits_per_mb(
+ INTER_FRAME, q, factor * speed_term * group_weight_factor,
+ cpi->common.bit_depth);
+ if (bits_per_mb <= target_norm_bits_per_mb) break;
}
// Restriction on active max q for constrained quality mode.
- if (cpi->oxcf.rc_mode == VPX_CQ)
- q = VPXMAX(q, oxcf->cq_level);
+ if (cpi->oxcf.rc_mode == VPX_CQ) q = VPXMAX(q, oxcf->cq_level);
return q;
}
}
@@ -1180,9 +1127,8 @@
setup_rf_level_maxq(cpi);
}
-void vp10_calculate_coded_size(VP10_COMP *cpi,
- int *scaled_frame_width,
- int *scaled_frame_height) {
+void vp10_calculate_coded_size(VP10_COMP *cpi, int *scaled_frame_width,
+ int *scaled_frame_height) {
RATE_CONTROL *const rc = &cpi->rc;
*scaled_frame_width = rc->frame_width[rc->frame_size_selector];
*scaled_frame_height = rc->frame_height[rc->frame_size_selector];
@@ -1197,8 +1143,7 @@
zero_stats(&twopass->total_stats);
zero_stats(&twopass->total_left_stats);
- if (!twopass->stats_in_end)
- return;
+ if (!twopass->stats_in_end) return;
stats = &twopass->total_stats;
@@ -1212,8 +1157,8 @@
// It is calculated based on the actual durations of all frames from the
// first pass.
vp10_new_framerate(cpi, frame_rate);
- twopass->bits_left = (int64_t)(stats->duration * oxcf->target_bandwidth /
- 10000000.0);
+ twopass->bits_left =
+ (int64_t)(stats->duration * oxcf->target_bandwidth / 10000000.0);
// This variable monitors how far behind the second ref update is lagging.
twopass->sr_update_lag = 1;
@@ -1221,14 +1166,14 @@
// Scan the first pass file and calculate a modified total error based upon
// the bias/power function used to allocate bits.
{
- const double avg_error = stats->coded_error /
- DOUBLE_DIVIDE_CHECK(stats->count);
+ const double avg_error =
+ stats->coded_error / DOUBLE_DIVIDE_CHECK(stats->count);
const FIRSTPASS_STATS *s = twopass->stats_in;
double modified_error_total = 0.0;
- twopass->modified_error_min = (avg_error *
- oxcf->two_pass_vbrmin_section) / 100;
- twopass->modified_error_max = (avg_error *
- oxcf->two_pass_vbrmax_section) / 100;
+ twopass->modified_error_min =
+ (avg_error * oxcf->two_pass_vbrmin_section) / 100;
+ twopass->modified_error_max =
+ (avg_error * oxcf->two_pass_vbrmax_section) / 100;
while (s < twopass->stats_in_end) {
modified_error_total += calculate_modified_err(cpi, twopass, oxcf, s);
++s;
@@ -1260,15 +1205,14 @@
static double get_sr_decay_rate(const VP10_COMP *cpi,
const FIRSTPASS_STATS *frame) {
- const int num_mbs = (cpi->oxcf.resize_mode != RESIZE_NONE)
- ? cpi->initial_mbs : cpi->common.MBs;
- double sr_diff =
- (frame->sr_coded_error - frame->coded_error) / num_mbs;
+ const int num_mbs = (cpi->oxcf.resize_mode != RESIZE_NONE) ? cpi->initial_mbs
+ : cpi->common.MBs;
+ double sr_diff = (frame->sr_coded_error - frame->coded_error) / num_mbs;
double sr_decay = 1.0;
double modified_pct_inter;
double modified_pcnt_intra;
const double motion_amplitude_factor =
- frame->pcnt_motion * ((frame->mvc_abs + frame->mvr_abs) / 2);
+ frame->pcnt_motion * ((frame->mvc_abs + frame->mvr_abs) / 2);
modified_pct_inter = frame->pcnt_inter;
if ((frame->intra_error / DOUBLE_DIVIDE_CHECK(frame->coded_error)) <
@@ -1277,7 +1221,6 @@
}
modified_pcnt_intra = 100 * (1.0 - modified_pct_inter);
-
if ((sr_diff > LOW_SR_DIFF_TRHESH)) {
sr_diff = VPXMIN(sr_diff, SR_DIFF_MAX);
sr_decay = 1.0 - (SR_DIFF_PART * sr_diff) -
@@ -1291,8 +1234,7 @@
// quality is decaying from frame to frame.
static double get_zero_motion_factor(const VP10_COMP *cpi,
const FIRSTPASS_STATS *frame) {
- const double zero_motion_pct = frame->pcnt_inter -
- frame->pcnt_motion;
+ const double zero_motion_pct = frame->pcnt_inter - frame->pcnt_motion;
double sr_decay = get_sr_decay_rate(cpi, frame);
return VPXMIN(sr_decay, zero_motion_pct);
}
@@ -1303,8 +1245,8 @@
const FIRSTPASS_STATS *next_frame) {
const double sr_decay_rate = get_sr_decay_rate(cpi, next_frame);
const double zero_motion_factor =
- (0.95 * pow((next_frame->pcnt_inter - next_frame->pcnt_motion),
- ZM_POWER_FACTOR));
+ (0.95 * pow((next_frame->pcnt_inter - next_frame->pcnt_motion),
+ ZM_POWER_FACTOR));
return VPXMAX(zero_motion_factor,
(sr_decay_rate + ((1.0 - sr_decay_rate) * zero_motion_factor)));
@@ -1313,8 +1255,8 @@
// Function to test for a condition where a complex transition is followed
// by a static section. For example in slide shows where there is a fade
// between slides. This is to help with more optimal kf and gf positioning.
-static int detect_transition_to_still(VP10_COMP *cpi,
- int frame_interval, int still_interval,
+static int detect_transition_to_still(VP10_COMP *cpi, int frame_interval,
+ int still_interval,
double loop_decay_rate,
double last_decay_rate) {
TWO_PASS *const twopass = &cpi->twopass;
@@ -1323,19 +1265,16 @@
// Break clause to detect very still sections after motion
// For example a static image after a fade or other transition
// instead of a clean scene cut.
- if (frame_interval > rc->min_gf_interval &&
- loop_decay_rate >= 0.999 &&
+ if (frame_interval > rc->min_gf_interval && loop_decay_rate >= 0.999 &&
last_decay_rate < 0.9) {
int j;
// Look ahead a few frames to see if static condition persists...
for (j = 0; j < still_interval; ++j) {
const FIRSTPASS_STATS *stats = &twopass->stats_in[j];
- if (stats >= twopass->stats_in_end)
- break;
+ if (stats >= twopass->stats_in_end) break;
- if (stats->pcnt_inter - stats->pcnt_motion < 0.999)
- break;
+ if (stats->pcnt_inter - stats->pcnt_motion < 0.999) break;
}
// Only if it does do we signal a transition to still.
@@ -1377,30 +1316,28 @@
// Accumulate a measure of how uniform (or conversely how random) the motion
// field is (a ratio of abs(mv) / mv).
if (pct > 0.05) {
- const double mvr_ratio = fabs(stats->mvr_abs) /
- DOUBLE_DIVIDE_CHECK(fabs(stats->MVr));
- const double mvc_ratio = fabs(stats->mvc_abs) /
- DOUBLE_DIVIDE_CHECK(fabs(stats->MVc));
+ const double mvr_ratio =
+ fabs(stats->mvr_abs) / DOUBLE_DIVIDE_CHECK(fabs(stats->MVr));
+ const double mvc_ratio =
+ fabs(stats->mvc_abs) / DOUBLE_DIVIDE_CHECK(fabs(stats->MVc));
- *mv_ratio_accumulator += pct * (mvr_ratio < stats->mvr_abs ?
- mvr_ratio : stats->mvr_abs);
- *mv_ratio_accumulator += pct * (mvc_ratio < stats->mvc_abs ?
- mvc_ratio : stats->mvc_abs);
+ *mv_ratio_accumulator +=
+ pct * (mvr_ratio < stats->mvr_abs ? mvr_ratio : stats->mvr_abs);
+ *mv_ratio_accumulator +=
+ pct * (mvc_ratio < stats->mvc_abs ? mvc_ratio : stats->mvc_abs);
}
}
#define BASELINE_ERR_PER_MB 1000.0
static double calc_frame_boost(VP10_COMP *cpi,
const FIRSTPASS_STATS *this_frame,
- double this_frame_mv_in_out,
- double max_boost) {
+ double this_frame_mv_in_out, double max_boost) {
double frame_boost;
- const double lq =
- vp10_convert_qindex_to_q(cpi->rc.avg_frame_qindex[INTER_FRAME],
- cpi->common.bit_depth);
+ const double lq = vp10_convert_qindex_to_q(
+ cpi->rc.avg_frame_qindex[INTER_FRAME], cpi->common.bit_depth);
const double boost_q_correction = VPXMIN((0.5 + (lq * 0.015)), 1.5);
- int num_mbs = (cpi->oxcf.resize_mode != RESIZE_NONE)
- ? cpi->initial_mbs : cpi->common.MBs;
+ int num_mbs = (cpi->oxcf.resize_mode != RESIZE_NONE) ? cpi->initial_mbs
+ : cpi->common.MBs;
// Correct for any inactive region in the image
num_mbs = (int)VPXMAX(1, num_mbs * calculate_active_area(cpi, this_frame));
@@ -1422,9 +1359,8 @@
return VPXMIN(frame_boost, max_boost * boost_q_correction);
}
-static int calc_arf_boost(VP10_COMP *cpi, int offset,
- int f_frames, int b_frames,
- int *f_boost, int *b_boost) {
+static int calc_arf_boost(VP10_COMP *cpi, int offset, int f_frames,
+ int b_frames, int *f_boost, int *b_boost) {
TWO_PASS *const twopass = &cpi->twopass;
int i;
double boost_score = 0.0;
@@ -1439,14 +1375,12 @@
// Search forward from the proposed arf/next gf position.
for (i = 0; i < f_frames; ++i) {
const FIRSTPASS_STATS *this_frame = read_frame_stats(twopass, i + offset);
- if (this_frame == NULL)
- break;
+ if (this_frame == NULL) break;
// Update the motion related elements to the boost calculation.
- accumulate_frame_motion_stats(this_frame,
- &this_frame_mv_in_out, &mv_in_out_accumulator,
- &abs_mv_in_out_accumulator,
- &mv_ratio_accumulator);
+ accumulate_frame_motion_stats(
+ this_frame, &this_frame_mv_in_out, &mv_in_out_accumulator,
+ &abs_mv_in_out_accumulator, &mv_ratio_accumulator);
// We want to discount the flash frame itself and the recovery
// frame that follows as both will have poor scores.
@@ -1457,12 +1391,13 @@
if (!flash_detected) {
decay_accumulator *= get_prediction_decay_rate(cpi, this_frame);
decay_accumulator = decay_accumulator < MIN_DECAY_FACTOR
- ? MIN_DECAY_FACTOR : decay_accumulator;
+ ? MIN_DECAY_FACTOR
+ : decay_accumulator;
}
- boost_score += decay_accumulator * calc_frame_boost(cpi, this_frame,
- this_frame_mv_in_out,
- GF_MAX_BOOST);
+ boost_score +=
+ decay_accumulator *
+ calc_frame_boost(cpi, this_frame, this_frame_mv_in_out, GF_MAX_BOOST);
}
*f_boost = (int)boost_score;
@@ -1478,14 +1413,12 @@
// Search backward towards last gf position.
for (i = -1; i >= -b_frames; --i) {
const FIRSTPASS_STATS *this_frame = read_frame_stats(twopass, i + offset);
- if (this_frame == NULL)
- break;
+ if (this_frame == NULL) break;
// Update the motion related elements to the boost calculation.
- accumulate_frame_motion_stats(this_frame,
- &this_frame_mv_in_out, &mv_in_out_accumulator,
- &abs_mv_in_out_accumulator,
- &mv_ratio_accumulator);
+ accumulate_frame_motion_stats(
+ this_frame, &this_frame_mv_in_out, &mv_in_out_accumulator,
+ &abs_mv_in_out_accumulator, &mv_ratio_accumulator);
// We want to discount the the flash frame itself and the recovery
// frame that follows as both will have poor scores.
@@ -1496,12 +1429,13 @@
if (!flash_detected) {
decay_accumulator *= get_prediction_decay_rate(cpi, this_frame);
decay_accumulator = decay_accumulator < MIN_DECAY_FACTOR
- ? MIN_DECAY_FACTOR : decay_accumulator;
+ ? MIN_DECAY_FACTOR
+ : decay_accumulator;
}
- boost_score += decay_accumulator * calc_frame_boost(cpi, this_frame,
- this_frame_mv_in_out,
- GF_MAX_BOOST);
+ boost_score +=
+ decay_accumulator *
+ calc_frame_boost(cpi, this_frame, this_frame_mv_in_out, GF_MAX_BOOST);
}
*b_boost = (int)boost_score;
@@ -1549,9 +1483,10 @@
}
// Clamp odd edge cases.
- total_group_bits = (total_group_bits < 0) ?
- 0 : (total_group_bits > twopass->kf_group_bits) ?
- twopass->kf_group_bits : total_group_bits;
+ total_group_bits =
+ (total_group_bits < 0) ? 0 : (total_group_bits > twopass->kf_group_bits)
+ ? twopass->kf_group_bits
+ : total_group_bits;
// Clip based on user supplied data rate variability limit.
if (total_group_bits > (int64_t)max_bits * rc->baseline_gf_interval)
@@ -1561,13 +1496,12 @@
}
// Calculate the number bits extra to assign to boosted frames in a group.
-static int calculate_boost_bits(int frame_count,
- int boost, int64_t total_group_bits) {
+static int calculate_boost_bits(int frame_count, int boost,
+ int64_t total_group_bits) {
int allocation_chunks;
// return 0 for invalid inputs (could arise e.g. through rounding errors)
- if (!boost || (total_group_bits <= 0) || (frame_count <= 0) )
- return 0;
+ if (!boost || (total_group_bits <= 0) || (frame_count <= 0)) return 0;
allocation_chunks = (frame_count * 100) + boost;
@@ -1637,14 +1571,12 @@
}
// Step over the golden frame / overlay frame
- if (EOF == input_stats(twopass, &frame_stats))
- return;
+ if (EOF == input_stats(twopass, &frame_stats)) return;
}
// Deduct the boost bits for arf (or gf if it is not a key frame)
// from the group total.
- if (rc->source_alt_ref_pending || !key_frame)
- total_group_bits -= gf_arf_bits;
+ if (rc->source_alt_ref_pending || !key_frame) total_group_bits -= gf_arf_bits;
// Store the bits to spend on the ARF if there is one.
if (rc->source_alt_ref_pending) {
@@ -1657,8 +1589,8 @@
gf_group->arf_update_idx[alt_frame_index] = arf_buffer_indices[0];
gf_group->arf_ref_idx[alt_frame_index] =
- arf_buffer_indices[cpi->multi_arf_last_grp_enabled &&
- rc->source_alt_ref_active];
+ arf_buffer_indices[cpi->multi_arf_last_grp_enabled &&
+ rc->source_alt_ref_active];
++frame_index;
if (cpi->multi_arf_enabled) {
@@ -1666,7 +1598,7 @@
gf_group->update_type[frame_index] = ARF_UPDATE;
gf_group->rf_level[frame_index] = GF_ARF_LOW;
gf_group->arf_src_offset[frame_index] =
- (unsigned char)((rc->baseline_gf_interval >> 1) - 1);
+ (unsigned char)((rc->baseline_gf_interval >> 1) - 1);
gf_group->arf_update_idx[frame_index] = arf_buffer_indices[1];
gf_group->arf_ref_idx[frame_index] = arf_buffer_indices[0];
++frame_index;
@@ -1679,8 +1611,7 @@
// Allocate bits to the other frames in the group.
for (i = 0; i < rc->baseline_gf_interval - rc->source_alt_ref_pending; ++i) {
int arf_idx = 0;
- if (EOF == input_stats(twopass, &frame_stats))
- break;
+ if (EOF == input_stats(twopass, &frame_stats)) break;
modified_err = calculate_modified_err(cpi, twopass, oxcf, &frame_stats);
@@ -1695,14 +1626,13 @@
mid_boost_bits += (target_frame_size >> 4);
target_frame_size -= (target_frame_size >> 4);
- if (frame_index <= mid_frame_idx)
- arf_idx = 1;
+ if (frame_index <= mid_frame_idx) arf_idx = 1;
}
gf_group->arf_update_idx[frame_index] = arf_buffer_indices[arf_idx];
gf_group->arf_ref_idx[frame_index] = arf_buffer_indices[arf_idx];
- target_frame_size = clamp(target_frame_size, 0,
- VPXMIN(max_bits, (int)total_group_bits));
+ target_frame_size =
+ clamp(target_frame_size, 0, VPXMIN(max_bits, (int)total_group_bits));
gf_group->update_type[frame_index] = LF_UPDATE;
gf_group->rf_level[frame_index] = INTER_NORMAL;
@@ -1817,12 +1747,10 @@
// Set a maximum and minimum interval for the GF group.
// If the image appears almost completely static we can extend beyond this.
{
- int int_max_q =
- (int)(vp10_convert_qindex_to_q(twopass->active_worst_quality,
- cpi->common.bit_depth));
- int int_lbq =
- (int)(vp10_convert_qindex_to_q(rc->last_boosted_qindex,
- cpi->common.bit_depth));
+ int int_max_q = (int)(vp10_convert_qindex_to_q(
+ twopass->active_worst_quality, cpi->common.bit_depth));
+ int int_lbq = (int)(vp10_convert_qindex_to_q(rc->last_boosted_qindex,
+ cpi->common.bit_depth));
active_min_gf_interval = rc->min_gf_interval + VPXMIN(2, int_max_q / 200);
if (active_min_gf_interval > rc->max_gf_interval)
active_min_gf_interval = rc->max_gf_interval;
@@ -1858,18 +1786,16 @@
gf_group_skip_pct += this_frame->intra_skip_pct;
gf_group_inactive_zone_rows += this_frame->inactive_zone_rows;
- if (EOF == input_stats(twopass, &next_frame))
- break;
+ if (EOF == input_stats(twopass, &next_frame)) break;
// Test for the case where there is a brief flash but the prediction
// quality back to an earlier frame is then restored.
flash_detected = detect_flash(twopass, 0);
// Update the motion related elements to the boost calculation.
- accumulate_frame_motion_stats(&next_frame,
- &this_frame_mv_in_out, &mv_in_out_accumulator,
- &abs_mv_in_out_accumulator,
- &mv_ratio_accumulator);
+ accumulate_frame_motion_stats(
+ &next_frame, &this_frame_mv_in_out, &mv_in_out_accumulator,
+ &abs_mv_in_out_accumulator, &mv_ratio_accumulator);
// Accumulate the effect of prediction quality decay.
if (!flash_detected) {
@@ -1892,23 +1818,23 @@
}
// Calculate a boost number for this frame.
- boost_score += decay_accumulator * calc_frame_boost(cpi, &next_frame,
- this_frame_mv_in_out,
- GF_MAX_BOOST);
+ boost_score +=
+ decay_accumulator *
+ calc_frame_boost(cpi, &next_frame, this_frame_mv_in_out, GF_MAX_BOOST);
// Break out conditions.
if (
- // Break at active_max_gf_interval unless almost totally static.
- (i >= (active_max_gf_interval + arf_active_or_kf) &&
- zero_motion_accumulator < 0.995) ||
- (
- // Don't break out with a very short interval.
- (i >= active_min_gf_interval + arf_active_or_kf) &&
- (!flash_detected) &&
- ((mv_ratio_accumulator > mv_ratio_accumulator_thresh) ||
- (abs_mv_in_out_accumulator > 3.0) ||
- (mv_in_out_accumulator < -2.0) ||
- ((boost_score - old_boost_score) < BOOST_BREAKOUT)))) {
+ // Break at active_max_gf_interval unless almost totally static.
+ (i >= (active_max_gf_interval + arf_active_or_kf) &&
+ zero_motion_accumulator < 0.995) ||
+ (
+ // Don't break out with a very short interval.
+ (i >= active_min_gf_interval + arf_active_or_kf) &&
+ (!flash_detected) &&
+ ((mv_ratio_accumulator > mv_ratio_accumulator_thresh) ||
+ (abs_mv_in_out_accumulator > 3.0) ||
+ (mv_in_out_accumulator < -2.0) ||
+ ((boost_score - old_boost_score) < BOOST_BREAKOUT)))) {
boost_score = old_boost_score;
break;
}
@@ -1923,18 +1849,19 @@
rc->constrained_gf_group = (i >= rc->frames_to_key) ? 1 : 0;
// Should we use the alternate reference frame.
- if (allow_alt_ref &&
- (i < cpi->oxcf.lag_in_frames) &&
- (i >= rc->min_gf_interval)) {
+ if (allow_alt_ref && (i < cpi->oxcf.lag_in_frames) &&
+ (i >= rc->min_gf_interval)) {
// Calculate the boost for alt ref.
- rc->gfu_boost = calc_arf_boost(cpi, 0, (i - 1), (i - 1), &f_boost,
- &b_boost);
+ rc->gfu_boost =
+ calc_arf_boost(cpi, 0, (i - 1), (i - 1), &f_boost, &b_boost);
rc->source_alt_ref_pending = 1;
// Test to see if multi arf is appropriate.
cpi->multi_arf_enabled =
- (cpi->multi_arf_allowed && (rc->baseline_gf_interval >= 6) &&
- (zero_motion_accumulator < 0.995)) ? 1 : 0;
+ (cpi->multi_arf_allowed && (rc->baseline_gf_interval >= 6) &&
+ (zero_motion_accumulator < 0.995))
+ ? 1
+ : 0;
} else {
rc->gfu_boost = VPXMAX((int)boost_score, MIN_ARF_GF_BOOST);
rc->source_alt_ref_pending = 0;
@@ -1959,13 +1886,13 @@
// of the allocated bit budget.
if ((cpi->oxcf.rc_mode != VPX_Q) && (rc->baseline_gf_interval > 1)) {
const int vbr_group_bits_per_frame =
- (int)(gf_group_bits / rc->baseline_gf_interval);
- const double group_av_err = gf_group_raw_error / rc->baseline_gf_interval;
+ (int)(gf_group_bits / rc->baseline_gf_interval);
+ const double group_av_err = gf_group_raw_error / rc->baseline_gf_interval;
const double group_av_skip_pct =
- gf_group_skip_pct / rc->baseline_gf_interval;
+ gf_group_skip_pct / rc->baseline_gf_interval;
const double group_av_inactive_zone =
- ((gf_group_inactive_zone_rows * 2) /
- (rc->baseline_gf_interval * (double)cm->mb_rows));
+ ((gf_group_inactive_zone_rows * 2) /
+ (rc->baseline_gf_interval * (double)cm->mb_rows));
int tmp_q;
// rc factor is a weight factor that corrects for local rate control drift.
@@ -1977,19 +1904,17 @@
rc_factor = VPXMIN(RC_FACTOR_MAX,
(double)(100 - rc->rate_error_estimate) / 100.0);
}
- tmp_q =
- get_twopass_worst_quality(cpi, group_av_err,
- (group_av_skip_pct + group_av_inactive_zone),
- vbr_group_bits_per_frame,
- twopass->kfgroup_inter_fraction * rc_factor);
+ tmp_q = get_twopass_worst_quality(
+ cpi, group_av_err, (group_av_skip_pct + group_av_inactive_zone),
+ vbr_group_bits_per_frame, twopass->kfgroup_inter_fraction * rc_factor);
twopass->active_worst_quality =
- VPXMAX(tmp_q, twopass->active_worst_quality >> 1);
+ VPXMAX(tmp_q, twopass->active_worst_quality >> 1);
}
#endif
// Calculate the extra bits to be used for boosted frame(s)
- gf_arf_bits = calculate_boost_bits(rc->baseline_gf_interval,
- rc->gfu_boost, gf_group_bits);
+ gf_arf_bits = calculate_boost_bits(rc->baseline_gf_interval, rc->gfu_boost,
+ gf_group_bits);
// Adjust KF group bits and error remaining.
twopass->kf_group_error_left -= (int64_t)gf_group_err;
@@ -2016,9 +1941,8 @@
// Calculate a section intra ratio used in setting max loop filter.
if (cpi->common.frame_type != KEY_FRAME) {
- twopass->section_intra_rating =
- calculate_section_intra_ratio(start_pos, twopass->stats_in_end,
- rc->baseline_gf_interval);
+ twopass->section_intra_rating = calculate_section_intra_ratio(
+ start_pos, twopass->stats_in_end, rc->baseline_gf_interval);
}
if (oxcf->resize_mode == RESIZE_DYNAMIC) {
@@ -2060,7 +1984,7 @@
int is_viable_kf = 0;
double pcnt_intra = 1.0 - this_frame->pcnt_inter;
double modified_pcnt_inter =
- this_frame->pcnt_inter - this_frame->pcnt_neutral;
+ this_frame->pcnt_inter - this_frame->pcnt_neutral;
// Does the frame satisfy the primary criteria of a key frame?
// See above for an explanation of the test criteria.
@@ -2072,15 +1996,15 @@
(pcnt_intra > (INTRA_VS_INTER_THRESH * modified_pcnt_inter)) &&
((this_frame->intra_error /
DOUBLE_DIVIDE_CHECK(this_frame->coded_error)) <
- KF_II_ERR_THRESHOLD) &&
+ KF_II_ERR_THRESHOLD) &&
((fabs(last_frame->coded_error - this_frame->coded_error) /
- DOUBLE_DIVIDE_CHECK(this_frame->coded_error) >
+ DOUBLE_DIVIDE_CHECK(this_frame->coded_error) >
ERR_CHANGE_THRESHOLD) ||
(fabs(last_frame->intra_error - this_frame->intra_error) /
- DOUBLE_DIVIDE_CHECK(this_frame->intra_error) >
+ DOUBLE_DIVIDE_CHECK(this_frame->intra_error) >
ERR_CHANGE_THRESHOLD) ||
((next_frame->intra_error /
- DOUBLE_DIVIDE_CHECK(next_frame->coded_error)) >
+ DOUBLE_DIVIDE_CHECK(next_frame->coded_error)) >
II_IMPROVEMENT_THRESHOLD))))) {
int i;
const FIRSTPASS_STATS *start_pos = twopass->stats_in;
@@ -2094,8 +2018,7 @@
double next_iiratio = (BOOST_FACTOR * local_next_frame.intra_error /
DOUBLE_DIVIDE_CHECK(local_next_frame.coded_error));
- if (next_iiratio > KF_II_MAX)
- next_iiratio = KF_II_MAX;
+ if (next_iiratio > KF_II_MAX) next_iiratio = KF_II_MAX;
// Cumulative effect of decay in prediction quality.
if (local_next_frame.pcnt_inter > 0.85)
@@ -2107,10 +2030,9 @@
boost_score += (decay_accumulator * next_iiratio);
// Test various breakout clauses.
- if ((local_next_frame.pcnt_inter < 0.05) ||
- (next_iiratio < 1.5) ||
- (((local_next_frame.pcnt_inter -
- local_next_frame.pcnt_neutral) < 0.20) &&
+ if ((local_next_frame.pcnt_inter < 0.05) || (next_iiratio < 1.5) ||
+ (((local_next_frame.pcnt_inter - local_next_frame.pcnt_neutral) <
+ 0.20) &&
(next_iiratio < 3.0)) ||
((boost_score - old_boost_score) < 3.0) ||
(local_next_frame.intra_error < 200)) {
@@ -2120,8 +2042,7 @@
old_boost_score = boost_score;
// Get the next frame details
- if (EOF == input_stats(twopass, &local_next_frame))
- break;
+ if (EOF == input_stats(twopass, &local_next_frame)) break;
}
// If there is tolerable prediction for at least the next 3 frames then
@@ -2157,7 +2078,7 @@
double boost_score = 0.0;
double kf_mod_err = 0.0;
double kf_group_err = 0.0;
- double recent_loop_decay[8] = {1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0};
+ double recent_loop_decay[8] = { 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0 };
vp10_zero(next_frame);
@@ -2212,8 +2133,7 @@
// quality since the last GF or KF.
recent_loop_decay[i % 8] = loop_decay_rate;
decay_accumulator = 1.0;
- for (j = 0; j < 8; ++j)
- decay_accumulator *= recent_loop_decay[j];
+ for (j = 0; j < 8; ++j) decay_accumulator *= recent_loop_decay[j];
// Special check for transition or high motion followed by a
// static scene.
@@ -2226,8 +2146,7 @@
// If we don't have a real key frame within the next two
// key_freq intervals then break out of the loop.
- if (rc->frames_to_key >= 2 * cpi->oxcf.key_freq)
- break;
+ if (rc->frames_to_key >= 2 * cpi->oxcf.key_freq) break;
} else {
++rc->frames_to_key;
}
@@ -2238,8 +2157,7 @@
// We already breakout of the loop above at 2x max.
// This code centers the extra kf if the actual natural interval
// is between 1x and 2x.
- if (cpi->oxcf.auto_key &&
- rc->frames_to_key > cpi->oxcf.key_freq) {
+ if (cpi->oxcf.auto_key && rc->frames_to_key > cpi->oxcf.key_freq) {
FIRSTPASS_STATS tmp_frame = first_frame;
rc->frames_to_key /= 2;
@@ -2278,8 +2196,8 @@
// Default allocation based on bits left and relative
// complexity of the section.
- twopass->kf_group_bits = (int64_t)(twopass->bits_left *
- (kf_group_err / twopass->modified_error_left));
+ twopass->kf_group_bits = (int64_t)(
+ twopass->bits_left * (kf_group_err / twopass->modified_error_left));
// Clip based on maximum per frame rate defined by the user.
max_grp_bits = (int64_t)max_bits * (int64_t)rc->frames_to_key;
@@ -2298,23 +2216,22 @@
decay_accumulator = 1.0;
boost_score = 0.0;
for (i = 0; i < (rc->frames_to_key - 1); ++i) {
- if (EOF == input_stats(twopass, &next_frame))
- break;
+ if (EOF == input_stats(twopass, &next_frame)) break;
// Monitor for static sections.
- zero_motion_accumulator = VPXMIN(
- zero_motion_accumulator, get_zero_motion_factor(cpi, &next_frame));
+ zero_motion_accumulator = VPXMIN(zero_motion_accumulator,
+ get_zero_motion_factor(cpi, &next_frame));
// Not all frames in the group are necessarily used in calculating boost.
if ((i <= rc->max_gf_interval) ||
((i <= (rc->max_gf_interval * 4)) && (decay_accumulator > 0.5))) {
const double frame_boost =
- calc_frame_boost(cpi, this_frame, 0, KF_MAX_BOOST);
+ calc_frame_boost(cpi, this_frame, 0, KF_MAX_BOOST);
// How fast is prediction quality decaying.
if (!detect_flash(twopass, 0)) {
const double loop_decay_rate =
- get_prediction_decay_rate(cpi, &next_frame);
+ get_prediction_decay_rate(cpi, &next_frame);
decay_accumulator *= loop_decay_rate;
decay_accumulator = VPXMAX(decay_accumulator, MIN_DECAY_FACTOR);
av_decay_accumulator += decay_accumulator;
@@ -2331,9 +2248,8 @@
twopass->kf_zeromotion_pct = (int)(zero_motion_accumulator * 100.0);
// Calculate a section intra ratio used in setting max loop filter.
- twopass->section_intra_rating =
- calculate_section_intra_ratio(start_position, twopass->stats_in_end,
- rc->frames_to_key);
+ twopass->section_intra_rating = calculate_section_intra_ratio(
+ start_position, twopass->stats_in_end, rc->frames_to_key);
// Apply various clamps for min and max boost
rc->kf_boost = (int)(av_decay_accumulator * boost_score);
@@ -2341,15 +2257,15 @@
rc->kf_boost = VPXMAX(rc->kf_boost, MIN_KF_BOOST);
// Work out how many bits to allocate for the key frame itself.
- kf_bits = calculate_boost_bits((rc->frames_to_key - 1),
- rc->kf_boost, twopass->kf_group_bits);
+ kf_bits = calculate_boost_bits((rc->frames_to_key - 1), rc->kf_boost,
+ twopass->kf_group_bits);
// Work out the fraction of the kf group bits reserved for the inter frames
// within the group after discounting the bits for the kf itself.
if (twopass->kf_group_bits) {
twopass->kfgroup_inter_fraction =
- (double)(twopass->kf_group_bits - kf_bits) /
- (double)twopass->kf_group_bits;
+ (double)(twopass->kf_group_bits - kf_bits) /
+ (double)twopass->kf_group_bits;
} else {
twopass->kfgroup_inter_fraction = 1.0;
}
@@ -2407,9 +2323,7 @@
cpi->refresh_golden_frame = 0;
cpi->refresh_alt_ref_frame = 1;
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
@@ -2421,13 +2335,15 @@
const TWO_PASS *const twopass = &cpi->twopass;
return (!frame_is_intra_only(&cpi->common) &&
- twopass->stats_in - 2 > twopass->stats_in_start &&
- twopass->stats_in < twopass->stats_in_end &&
- (twopass->stats_in - 1)->pcnt_inter - (twopass->stats_in - 1)->pcnt_motion
- == 1 &&
- (twopass->stats_in - 2)->pcnt_inter - (twopass->stats_in - 2)->pcnt_motion
- == 1 &&
- twopass->stats_in->pcnt_inter - twopass->stats_in->pcnt_motion == 1);
+ twopass->stats_in - 2 > twopass->stats_in_start &&
+ twopass->stats_in < twopass->stats_in_end &&
+ (twopass->stats_in - 1)->pcnt_inter -
+ (twopass->stats_in - 1)->pcnt_motion ==
+ 1 &&
+ (twopass->stats_in - 2)->pcnt_inter -
+ (twopass->stats_in - 2)->pcnt_motion ==
+ 1 &&
+ twopass->stats_in->pcnt_inter - twopass->stats_in->pcnt_motion == 1);
}
void vp10_rc_get_second_pass_params(VP10_COMP *cpi) {
@@ -2440,11 +2356,9 @@
int target_rate;
- frames_left = (int)(twopass->total_stats.count -
- cm->current_video_frame);
+ frames_left = (int)(twopass->total_stats.count - cm->current_video_frame);
- if (!twopass->stats_in)
- return;
+ if (!twopass->stats_in) return;
// If this is an arf frame then we dont want to read the stats file or
// advance the input pointer as we already have what we need.
@@ -2472,20 +2386,19 @@
twopass->active_worst_quality = cpi->oxcf.cq_level;
} else if (cm->current_video_frame == 0) {
// Special case code for first frame.
- const int section_target_bandwidth = (int)(twopass->bits_left /
- frames_left);
+ const int section_target_bandwidth =
+ (int)(twopass->bits_left / frames_left);
const double section_length = twopass->total_left_stats.count;
const double section_error =
- twopass->total_left_stats.coded_error / section_length;
+ twopass->total_left_stats.coded_error / section_length;
const double section_intra_skip =
- twopass->total_left_stats.intra_skip_pct / section_length;
+ twopass->total_left_stats.intra_skip_pct / section_length;
const double section_inactive_zone =
- (twopass->total_left_stats.inactive_zone_rows * 2) /
- ((double)cm->mb_rows * section_length);
- const int tmp_q =
- get_twopass_worst_quality(cpi, section_error,
- section_intra_skip + section_inactive_zone,
- section_target_bandwidth, DEFAULT_GRP_WEIGHT);
+ (twopass->total_left_stats.inactive_zone_rows * 2) /
+ ((double)cm->mb_rows * section_length);
+ const int tmp_q = get_twopass_worst_quality(
+ cpi, section_error, section_intra_skip + section_inactive_zone,
+ section_target_bandwidth, DEFAULT_GRP_WEIGHT);
twopass->active_worst_quality = tmp_q;
twopass->baseline_active_worst_quality = tmp_q;
@@ -2497,8 +2410,7 @@
rc->avg_frame_qindex[KEY_FRAME] = rc->last_q[KEY_FRAME];
}
vp10_zero(this_frame);
- if (EOF == input_stats(twopass, &this_frame))
- return;
+ if (EOF == input_stats(twopass, &this_frame)) return;
// Set the frame content type flag.
if (this_frame.intra_skip_pct >= FC_ANIMATION_THRESH)
@@ -2528,9 +2440,9 @@
FILE *fpfile;
fpfile = fopen("arf.stt", "a");
++arf_count;
- fprintf(fpfile, "%10d %10ld %10d %10d %10ld\n",
- cm->current_video_frame, rc->frames_till_gf_update_due,
- rc->kf_boost, arf_count, rc->gfu_boost);
+ fprintf(fpfile, "%10d %10ld %10d %10d %10ld\n", cm->current_video_frame,
+ rc->frames_till_gf_update_due, rc->kf_boost, arf_count,
+ rc->gfu_boost);
fclose(fpfile);
}
@@ -2555,11 +2467,12 @@
{
const int num_mbs = (cpi->oxcf.resize_mode != RESIZE_NONE)
- ? cpi->initial_mbs : cpi->common.MBs;
+ ? cpi->initial_mbs
+ : cpi->common.MBs;
// The multiplication by 256 reverses a scaling factor of (>> 8)
// applied when combining MB error values for the frame.
twopass->mb_av_energy =
- log(((this_frame.intra_error * 256.0) / num_mbs) + 1.0);
+ log(((this_frame.intra_error * 256.0) / num_mbs) + 1.0);
}
// Update the total stats remaining structure.
@@ -2585,7 +2498,7 @@
// Calculate the pct rc error.
if (rc->total_actual_bits) {
rc->rate_error_estimate =
- (int)((rc->vbr_bits_off_target * 100) / rc->total_actual_bits);
+ (int)((rc->vbr_bits_off_target * 100) / rc->total_actual_bits);
rc->rate_error_estimate = clamp(rc->rate_error_estimate, -100, 100);
} else {
rc->rate_error_estimate = 0;
@@ -2605,7 +2518,7 @@
(cpi->twopass.gf_zeromotion_pct < VLOW_MOTION_THRESHOLD) &&
!cpi->rc.is_src_frame_alt_ref) {
const int maxq_adj_limit =
- rc->worst_quality - twopass->active_worst_quality;
+ rc->worst_quality - twopass->active_worst_quality;
const int minq_adj_limit =
(cpi->oxcf.rc_mode == VPX_CQ ? MINQ_ADJ_LIMIT_CQ : MINQ_ADJ_LIMIT);
@@ -2614,7 +2527,7 @@
--twopass->extend_maxq;
if (rc->rolling_target_bits >= rc->rolling_actual_bits)
++twopass->extend_minq;
- // Overshoot.
+ // Overshoot.
} else if (rc->rate_error_estimate < -cpi->oxcf.over_shoot_pct) {
--twopass->extend_minq;
if (rc->rolling_target_bits < rc->rolling_actual_bits)
@@ -2643,14 +2556,14 @@
int fast_extra_thresh = rc->base_frame_target / HIGH_UNDERSHOOT_RATIO;
if (rc->projected_frame_size < fast_extra_thresh) {
rc->vbr_bits_off_target_fast +=
- fast_extra_thresh - rc->projected_frame_size;
+ fast_extra_thresh - rc->projected_frame_size;
rc->vbr_bits_off_target_fast =
- VPXMIN(rc->vbr_bits_off_target_fast, (4 * rc->avg_frame_bandwidth));
+ VPXMIN(rc->vbr_bits_off_target_fast, (4 * rc->avg_frame_bandwidth));
// Fast adaptation of minQ if necessary to use up the extra bits.
if (rc->avg_frame_bandwidth) {
twopass->extend_minq_fast =
- (int)(rc->vbr_bits_off_target_fast * 8 / rc->avg_frame_bandwidth);
+ (int)(rc->vbr_bits_off_target_fast * 8 / rc->avg_frame_bandwidth);
}
twopass->extend_minq_fast = VPXMIN(
twopass->extend_minq_fast, minq_adj_limit - twopass->extend_minq);
diff --git a/vp10/encoder/firstpass.h b/vp10/encoder/firstpass.h
index 68a8887..2fbd3f7 100644
--- a/vp10/encoder/firstpass.h
+++ b/vp10/encoder/firstpass.h
@@ -155,9 +155,8 @@
void vp10_init_subsampling(struct VP10_COMP *cpi);
-void vp10_calculate_coded_size(struct VP10_COMP *cpi,
- int *scaled_frame_width,
- int *scaled_frame_height);
+void vp10_calculate_coded_size(struct VP10_COMP *cpi, int *scaled_frame_width,
+ int *scaled_frame_height);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/encoder/lookahead.c b/vp10/encoder/lookahead.c
index 6289fbc..19beee3 100644
--- a/vp10/encoder/lookahead.c
+++ b/vp10/encoder/lookahead.c
@@ -25,35 +25,31 @@
struct lookahead_entry *buf = ctx->buf + index;
assert(index < ctx->max_sz);
- if (++index >= ctx->max_sz)
- index -= ctx->max_sz;
+ if (++index >= ctx->max_sz) index -= ctx->max_sz;
*idx = index;
return buf;
}
-
void vp10_lookahead_destroy(struct lookahead_ctx *ctx) {
if (ctx) {
if (ctx->buf) {
unsigned int i;
- for (i = 0; i < ctx->max_sz; i++)
- vpx_free_frame_buffer(&ctx->buf[i].img);
+ for (i = 0; i < ctx->max_sz; i++) vpx_free_frame_buffer(&ctx->buf[i].img);
free(ctx->buf);
}
free(ctx);
}
}
-
struct lookahead_ctx *vp10_lookahead_init(unsigned int width,
- unsigned int height,
- unsigned int subsampling_x,
- unsigned int subsampling_y,
+ unsigned int height,
+ unsigned int subsampling_x,
+ unsigned int subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
- int use_highbitdepth,
+ int use_highbitdepth,
#endif
- unsigned int depth) {
+ unsigned int depth) {
struct lookahead_ctx *ctx = NULL;
// Clamp the lookahead queue depth
@@ -69,32 +65,30 @@
unsigned int i;
ctx->max_sz = depth;
ctx->buf = calloc(depth, sizeof(*ctx->buf));
- if (!ctx->buf)
- goto bail;
+ if (!ctx->buf) goto bail;
for (i = 0; i < depth; i++)
- if (vpx_alloc_frame_buffer(&ctx->buf[i].img,
- width, height, subsampling_x, subsampling_y,
+ if (vpx_alloc_frame_buffer(
+ &ctx->buf[i].img, width, height, subsampling_x, subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
- use_highbitdepth,
+ use_highbitdepth,
#endif
- VPX_ENC_BORDER_IN_PIXELS,
- legacy_byte_alignment))
+ VPX_ENC_BORDER_IN_PIXELS, legacy_byte_alignment))
goto bail;
}
return ctx;
- bail:
+bail:
vp10_lookahead_destroy(ctx);
return NULL;
}
#define USE_PARTIAL_COPY 0
-int vp10_lookahead_push(struct lookahead_ctx *ctx, YV12_BUFFER_CONFIG *src,
- int64_t ts_start, int64_t ts_end,
+int vp10_lookahead_push(struct lookahead_ctx *ctx, YV12_BUFFER_CONFIG *src,
+ int64_t ts_start, int64_t ts_end,
#if CONFIG_VPX_HIGHBITDEPTH
- int use_highbitdepth,
+ int use_highbitdepth,
#endif
- unsigned int flags) {
+ unsigned int flags) {
struct lookahead_entry *buf;
#if USE_PARTIAL_COPY
int row, col, active_end;
@@ -109,8 +103,7 @@
int subsampling_y = src->subsampling_y;
int larger_dimensions, new_dimensions;
- if (ctx->sz + 1 + MAX_PRE_FRAMES > ctx->max_sz)
- return 1;
+ if (ctx->sz + 1 + MAX_PRE_FRAMES > ctx->max_sz) return 1;
ctx->sz++;
buf = pop(ctx, &ctx->write_idx);
@@ -118,8 +111,7 @@
height != buf->img.y_crop_height ||
uv_width != buf->img.uv_crop_width ||
uv_height != buf->img.uv_crop_height;
- larger_dimensions = width > buf->img.y_width ||
- height > buf->img.y_height ||
+ larger_dimensions = width > buf->img.y_width || height > buf->img.y_height ||
uv_width > buf->img.uv_width ||
uv_height > buf->img.uv_height;
assert(!larger_dimensions || new_dimensions);
@@ -139,27 +131,22 @@
while (1) {
// Find the first active macroblock in this row.
for (; col < mb_cols; ++col) {
- if (active_map[col])
- break;
+ if (active_map[col]) break;
}
// No more active macroblock in this row.
- if (col == mb_cols)
- break;
+ if (col == mb_cols) break;
// Find the end of active region in this row.
active_end = col;
for (; active_end < mb_cols; ++active_end) {
- if (!active_map[active_end])
- break;
+ if (!active_map[active_end]) break;
}
// Only copy this active region.
- vp10_copy_and_extend_frame_with_rect(src, &buf->img,
- row << 4,
- col << 4, 16,
- (active_end - col) << 4);
+ vp10_copy_and_extend_frame_with_rect(src, &buf->img, row << 4, col << 4,
+ 16, (active_end - col) << 4);
// Start again from the end of this active region.
col = active_end;
@@ -172,14 +159,13 @@
if (larger_dimensions) {
YV12_BUFFER_CONFIG new_img;
memset(&new_img, 0, sizeof(new_img));
- if (vpx_alloc_frame_buffer(&new_img,
- width, height, subsampling_x, subsampling_y,
+ if (vpx_alloc_frame_buffer(&new_img, width, height, subsampling_x,
+ subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
use_highbitdepth,
#endif
- VPX_ENC_BORDER_IN_PIXELS,
- 0))
- return 1;
+ VPX_ENC_BORDER_IN_PIXELS, 0))
+ return 1;
vpx_free_frame_buffer(&buf->img);
buf->img = new_img;
} else if (new_dimensions) {
@@ -202,9 +188,8 @@
return 0;
}
-
struct lookahead_entry *vp10_lookahead_pop(struct lookahead_ctx *ctx,
- int drain) {
+ int drain) {
struct lookahead_entry *buf = NULL;
if (ctx->sz && (drain || ctx->sz == ctx->max_sz - MAX_PRE_FRAMES)) {
@@ -214,25 +199,22 @@
return buf;
}
-
struct lookahead_entry *vp10_lookahead_peek(struct lookahead_ctx *ctx,
- int index) {
+ int index) {
struct lookahead_entry *buf = NULL;
if (index >= 0) {
// Forward peek
if (index < (int)ctx->sz) {
index += ctx->read_idx;
- if (index >= (int)ctx->max_sz)
- index -= ctx->max_sz;
+ if (index >= (int)ctx->max_sz) index -= ctx->max_sz;
buf = ctx->buf + index;
}
} else if (index < 0) {
// Backward peek
if (-index <= MAX_PRE_FRAMES) {
index += ctx->read_idx;
- if (index < 0)
- index += ctx->max_sz;
+ if (index < 0) index += ctx->max_sz;
buf = ctx->buf + index;
}
}
@@ -240,6 +222,4 @@
return buf;
}
-unsigned int vp10_lookahead_depth(struct lookahead_ctx *ctx) {
- return ctx->sz;
-}
+unsigned int vp10_lookahead_depth(struct lookahead_ctx *ctx) { return ctx->sz; }
diff --git a/vp10/encoder/lookahead.h b/vp10/encoder/lookahead.h
index fb28c32..4b5fe7f 100644
--- a/vp10/encoder/lookahead.h
+++ b/vp10/encoder/lookahead.h
@@ -21,10 +21,10 @@
#define MAX_LAG_BUFFERS 25
struct lookahead_entry {
- YV12_BUFFER_CONFIG img;
- int64_t ts_start;
- int64_t ts_end;
- unsigned int flags;
+ YV12_BUFFER_CONFIG img;
+ int64_t ts_start;
+ int64_t ts_end;
+ unsigned int flags;
};
// The max of past frames we want to keep in the queue.
@@ -44,20 +44,18 @@
* may be done when buffers are enqueued.
*/
struct lookahead_ctx *vp10_lookahead_init(unsigned int width,
- unsigned int height,
- unsigned int subsampling_x,
- unsigned int subsampling_y,
+ unsigned int height,
+ unsigned int subsampling_x,
+ unsigned int subsampling_y,
#if CONFIG_VPX_HIGHBITDEPTH
- int use_highbitdepth,
+ int use_highbitdepth,
#endif
- unsigned int depth);
-
+ unsigned int depth);
/**\brief Destroys the lookahead stage
*/
void vp10_lookahead_destroy(struct lookahead_ctx *ctx);
-
/**\brief Enqueue a source buffer
*
* This function will copy the source image into a new framebuffer with
@@ -74,12 +72,11 @@
* \param[in] active_map Map that specifies which macroblock is active
*/
int vp10_lookahead_push(struct lookahead_ctx *ctx, YV12_BUFFER_CONFIG *src,
- int64_t ts_start, int64_t ts_end,
+ int64_t ts_start, int64_t ts_end,
#if CONFIG_VPX_HIGHBITDEPTH
- int use_highbitdepth,
+ int use_highbitdepth,
#endif
- unsigned int flags);
-
+ unsigned int flags);
/**\brief Get the next source buffer to encode
*
@@ -92,8 +89,7 @@
* \retval NULL, if drain not set and queue not of the configured depth
*/
struct lookahead_entry *vp10_lookahead_pop(struct lookahead_ctx *ctx,
- int drain);
-
+ int drain);
/**\brief Get a future source buffer to encode
*
@@ -103,8 +99,7 @@
* \retval NULL, if no buffer exists at the specified index
*/
struct lookahead_entry *vp10_lookahead_peek(struct lookahead_ctx *ctx,
- int index);
-
+ int index);
/**\brief Get the number of frames currently in the lookahead queue
*
diff --git a/vp10/encoder/mbgraph.c b/vp10/encoder/mbgraph.c
index 98c532e..1579bf4 100644
--- a/vp10/encoder/mbgraph.c
+++ b/vp10/encoder/mbgraph.c
@@ -22,11 +22,8 @@
#include "vp10/common/reconinter.h"
#include "vp10/common/reconintra.h"
-
-static unsigned int do_16x16_motion_iteration(VP10_COMP *cpi,
- const MV *ref_mv,
- MV *dst_mv,
- int mb_row,
+static unsigned int do_16x16_motion_iteration(VP10_COMP *cpi, const MV *ref_mv,
+ MV *dst_mv, int mb_row,
int mb_col) {
MACROBLOCK *const x = &cpi->td.mb;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -51,8 +48,7 @@
/*cpi->sf.search_method == HEX*/
vp10_hex_search(x, &ref_full, step_param, x->errorperbit, 0,
- cond_cost_list(cpi, cost_list),
- &v_fn_ptr, 0, ref_mv, dst_mv);
+ cond_cost_list(cpi, cost_list), &v_fn_ptr, 0, ref_mv, dst_mv);
// Try sub-pixel MC
// if (bestsme > error_thresh && bestsme < INT_MAX)
@@ -62,9 +58,8 @@
cpi->find_fractional_mv_step(
x, dst_mv, ref_mv, cpi->common.allow_high_precision_mv, x->errorperbit,
&v_fn_ptr, 0, mv_sf->subpel_iters_per_step,
- cond_cost_list(cpi, cost_list),
- NULL, NULL,
- &distortion, &sse, NULL, 0, 0);
+ cond_cost_list(cpi, cost_list), NULL, NULL, &distortion, &sse, NULL, 0,
+ 0);
}
xd->mi[0]->mbmi.mode = NEWMV;
@@ -107,10 +102,10 @@
// based search as well.
if (ref_mv->row != 0 || ref_mv->col != 0) {
unsigned int tmp_err;
- MV zero_ref_mv = {0, 0}, tmp_mv;
+ MV zero_ref_mv = { 0, 0 }, tmp_mv;
- tmp_err = do_16x16_motion_iteration(cpi, &zero_ref_mv, &tmp_mv,
- mb_row, mb_col);
+ tmp_err =
+ do_16x16_motion_iteration(cpi, &zero_ref_mv, &tmp_mv, mb_row, mb_col);
if (tmp_err < err) {
dst_mv->as_mv = tmp_mv;
err = tmp_err;
@@ -135,7 +130,7 @@
return err;
}
static int find_best_16x16_intra(VP10_COMP *cpi, PREDICTION_MODE *pbest_mode) {
- MACROBLOCK *const x = &cpi->td.mb;
+ MACROBLOCK *const x = &cpi->td.mb;
MACROBLOCKD *const xd = &x->e_mbd;
PREDICTION_MODE best_mode = -1, mode;
unsigned int best_err = INT_MAX;
@@ -146,38 +141,30 @@
unsigned int err;
xd->mi[0]->mbmi.mode = mode;
- vp10_predict_intra_block(xd, 2, 2, TX_16X16, mode,
- x->plane[0].src.buf, x->plane[0].src.stride,
- xd->plane[0].dst.buf, xd->plane[0].dst.stride,
- 0, 0, 0);
+ vp10_predict_intra_block(xd, 2, 2, TX_16X16, mode, x->plane[0].src.buf,
+ x->plane[0].src.stride, xd->plane[0].dst.buf,
+ xd->plane[0].dst.stride, 0, 0, 0);
err = vpx_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride,
xd->plane[0].dst.buf, xd->plane[0].dst.stride);
// find best
if (err < best_err) {
- best_err = err;
+ best_err = err;
best_mode = mode;
}
}
- if (pbest_mode)
- *pbest_mode = best_mode;
+ if (pbest_mode) *pbest_mode = best_mode;
return best_err;
}
-static void update_mbgraph_mb_stats
-(
- VP10_COMP *cpi,
- MBGRAPH_MB_STATS *stats,
- YV12_BUFFER_CONFIG *buf,
- int mb_y_offset,
- YV12_BUFFER_CONFIG *golden_ref,
- const MV *prev_golden_ref_mv,
- YV12_BUFFER_CONFIG *alt_ref,
- int mb_row,
- int mb_col
-) {
+static void update_mbgraph_mb_stats(VP10_COMP *cpi, MBGRAPH_MB_STATS *stats,
+ YV12_BUFFER_CONFIG *buf, int mb_y_offset,
+ YV12_BUFFER_CONFIG *golden_ref,
+ const MV *prev_golden_ref_mv,
+ YV12_BUFFER_CONFIG *alt_ref, int mb_row,
+ int mb_col) {
MACROBLOCK *const x = &cpi->td.mb;
MACROBLOCKD *const xd = &x->e_mbd;
int intra_error;
@@ -191,10 +178,8 @@
xd->plane[0].dst.stride = get_frame_new_buffer(cm)->y_stride;
// do intra 16x16 prediction
- intra_error = find_best_16x16_intra(cpi,
- &stats->ref[INTRA_FRAME].m.mode);
- if (intra_error <= 0)
- intra_error = 1;
+ intra_error = find_best_16x16_intra(cpi, &stats->ref[INTRA_FRAME].m.mode);
+ if (intra_error <= 0) intra_error = 1;
stats->ref[INTRA_FRAME].err = intra_error;
// Golden frame MV search, if it exists and is different than last frame
@@ -202,10 +187,9 @@
int g_motion_error;
xd->plane[0].pre[0].buf = golden_ref->y_buffer + mb_y_offset;
xd->plane[0].pre[0].stride = golden_ref->y_stride;
- g_motion_error = do_16x16_motion_search(cpi,
- prev_golden_ref_mv,
- &stats->ref[GOLDEN_FRAME].m.mv,
- mb_row, mb_col);
+ g_motion_error =
+ do_16x16_motion_search(cpi, prev_golden_ref_mv,
+ &stats->ref[GOLDEN_FRAME].m.mv, mb_row, mb_col);
stats->ref[GOLDEN_FRAME].err = g_motion_error;
} else {
stats->ref[GOLDEN_FRAME].err = INT_MAX;
@@ -218,8 +202,8 @@
int a_motion_error;
xd->plane[0].pre[0].buf = alt_ref->y_buffer + mb_y_offset;
xd->plane[0].pre[0].stride = alt_ref->y_stride;
- a_motion_error = do_16x16_zerozero_search(cpi,
- &stats->ref[ALTREF_FRAME].m.mv);
+ a_motion_error =
+ do_16x16_zerozero_search(cpi, &stats->ref[ALTREF_FRAME].m.mv);
stats->ref[ALTREF_FRAME].err = a_motion_error;
} else {
@@ -239,17 +223,17 @@
int mb_col, mb_row, offset = 0;
int mb_y_offset = 0, arf_y_offset = 0, gld_y_offset = 0;
- MV gld_top_mv = {0, 0};
+ MV gld_top_mv = { 0, 0 };
MODE_INFO mi_local;
vp10_zero(mi_local);
// Set up limit values for motion vectors to prevent them extending outside
// the UMV borders.
- x->mv_row_min = -BORDER_MV_PIXELS_B16;
- x->mv_row_max = (cm->mb_rows - 1) * 8 + BORDER_MV_PIXELS_B16;
- xd->up_available = 0;
- xd->plane[0].dst.stride = buf->y_stride;
- xd->plane[0].pre[0].stride = buf->y_stride;
+ x->mv_row_min = -BORDER_MV_PIXELS_B16;
+ x->mv_row_max = (cm->mb_rows - 1) * 8 + BORDER_MV_PIXELS_B16;
+ xd->up_available = 0;
+ xd->plane[0].dst.stride = buf->y_stride;
+ xd->plane[0].pre[0].stride = buf->y_stride;
xd->plane[1].dst.stride = buf->uv_stride;
xd->mi[0] = &mi_local;
mi_local.mbmi.sb_type = BLOCK_16X16;
@@ -258,41 +242,39 @@
for (mb_row = 0; mb_row < cm->mb_rows; mb_row++) {
MV gld_left_mv = gld_top_mv;
- int mb_y_in_offset = mb_y_offset;
+ int mb_y_in_offset = mb_y_offset;
int arf_y_in_offset = arf_y_offset;
int gld_y_in_offset = gld_y_offset;
// Set up limit values for motion vectors to prevent them extending outside
// the UMV borders.
- x->mv_col_min = -BORDER_MV_PIXELS_B16;
- x->mv_col_max = (cm->mb_cols - 1) * 8 + BORDER_MV_PIXELS_B16;
+ x->mv_col_min = -BORDER_MV_PIXELS_B16;
+ x->mv_col_max = (cm->mb_cols - 1) * 8 + BORDER_MV_PIXELS_B16;
xd->left_available = 0;
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++) {
MBGRAPH_MB_STATS *mb_stats = &stats->mb_stats[offset + mb_col];
- update_mbgraph_mb_stats(cpi, mb_stats, buf, mb_y_in_offset,
- golden_ref, &gld_left_mv, alt_ref,
- mb_row, mb_col);
+ update_mbgraph_mb_stats(cpi, mb_stats, buf, mb_y_in_offset, golden_ref,
+ &gld_left_mv, alt_ref, mb_row, mb_col);
gld_left_mv = mb_stats->ref[GOLDEN_FRAME].m.mv.as_mv;
if (mb_col == 0) {
gld_top_mv = gld_left_mv;
}
xd->left_available = 1;
- mb_y_in_offset += 16;
- gld_y_in_offset += 16;
- arf_y_in_offset += 16;
- x->mv_col_min -= 16;
- x->mv_col_max -= 16;
+ mb_y_in_offset += 16;
+ gld_y_in_offset += 16;
+ arf_y_in_offset += 16;
+ x->mv_col_min -= 16;
+ x->mv_col_max -= 16;
}
xd->up_available = 1;
- mb_y_offset += buf->y_stride * 16;
- gld_y_offset += golden_ref->y_stride * 16;
- if (alt_ref)
- arf_y_offset += alt_ref->y_stride * 16;
- x->mv_row_min -= 16;
- x->mv_row_max -= 16;
- offset += cm->mb_cols;
+ mb_y_offset += buf->y_stride * 16;
+ gld_y_offset += golden_ref->y_stride * 16;
+ if (alt_ref) arf_y_offset += alt_ref->y_stride * 16;
+ x->mv_row_min -= 16;
+ x->mv_row_max -= 16;
+ offset += cm->mb_cols;
}
}
@@ -306,9 +288,9 @@
int *arf_not_zz;
- CHECK_MEM_ERROR(cm, arf_not_zz,
- vpx_calloc(cm->mb_rows * cm->mb_cols * sizeof(*arf_not_zz),
- 1));
+ CHECK_MEM_ERROR(
+ cm, arf_not_zz,
+ vpx_calloc(cm->mb_rows * cm->mb_cols * sizeof(*arf_not_zz), 1));
// We are not interested in results beyond the alt ref itself.
if (n_frames > cpi->rc.frames_till_gf_update_due)
@@ -324,12 +306,11 @@
MBGRAPH_MB_STATS *mb_stats = &frame_stats->mb_stats[offset + mb_col];
int altref_err = mb_stats->ref[ALTREF_FRAME].err;
- int intra_err = mb_stats->ref[INTRA_FRAME ].err;
+ int intra_err = mb_stats->ref[INTRA_FRAME].err;
int golden_err = mb_stats->ref[GOLDEN_FRAME].err;
// Test for altref vs intra and gf and that its mv was 0,0.
- if (altref_err > 1000 ||
- altref_err > intra_err ||
+ if (altref_err > 1000 || altref_err > intra_err ||
altref_err > golden_err) {
arf_not_zz[offset + mb_col]++;
}
@@ -384,11 +365,9 @@
// we need to look ahead beyond where the ARF transitions into
// being a GF - so exit if we don't look ahead beyond that
- if (n_frames <= cpi->rc.frames_till_gf_update_due)
- return;
+ if (n_frames <= cpi->rc.frames_till_gf_update_due) return;
- if (n_frames > MAX_LAG_BUFFERS)
- n_frames = MAX_LAG_BUFFERS;
+ if (n_frames > MAX_LAG_BUFFERS) n_frames = MAX_LAG_BUFFERS;
cpi->mbgraph_n_frames = n_frames;
for (i = 0; i < n_frames; i++) {
@@ -407,8 +386,8 @@
assert(q_cur != NULL);
- update_mbgraph_frame_stats(cpi, frame_stats, &q_cur->img,
- golden_ref, cpi->Source);
+ update_mbgraph_frame_stats(cpi, frame_stats, &q_cur->img, golden_ref,
+ cpi->Source);
}
vpx_clear_system_state();
diff --git a/vp10/encoder/mbgraph.h b/vp10/encoder/mbgraph.h
index 3408464..e2daae5 100644
--- a/vp10/encoder/mbgraph.h
+++ b/vp10/encoder/mbgraph.h
@@ -25,9 +25,7 @@
} ref[MAX_REF_FRAMES];
} MBGRAPH_MB_STATS;
-typedef struct {
- MBGRAPH_MB_STATS *mb_stats;
-} MBGRAPH_FRAME_STATS;
+typedef struct { MBGRAPH_MB_STATS *mb_stats; } MBGRAPH_FRAME_STATS;
struct VP10_COMP;
diff --git a/vp10/encoder/mcomp.c b/vp10/encoder/mcomp.c
index 8d5b662..17cc2f3 100644
--- a/vp10/encoder/mcomp.c
+++ b/vp10/encoder/mcomp.c
@@ -45,14 +45,10 @@
// Get intersection of UMV window and valid MV window to reduce # of checks
// in diamond search.
- if (x->mv_col_min < col_min)
- x->mv_col_min = col_min;
- if (x->mv_col_max > col_max)
- x->mv_col_max = col_max;
- if (x->mv_row_min < row_min)
- x->mv_row_min = row_min;
- if (x->mv_row_max > row_max)
- x->mv_row_max = row_max;
+ if (x->mv_col_min < col_min) x->mv_col_min = col_min;
+ if (x->mv_col_max > col_max) x->mv_col_max = col_max;
+ if (x->mv_row_min < row_min) x->mv_row_min = row_min;
+ if (x->mv_row_max > row_max) x->mv_row_max = row_max;
}
int vp10_init_search_range(int size) {
@@ -60,44 +56,39 @@
// Minimum search size no matter what the passed in value.
size = VPXMAX(16, size);
- while ((size << sr) < MAX_FULL_PEL_VAL)
- sr++;
+ while ((size << sr) < MAX_FULL_PEL_VAL) sr++;
sr = VPXMIN(sr, MAX_MVSEARCH_STEPS - 2);
return sr;
}
-static INLINE int mv_cost(const MV *mv,
- const int *joint_cost, int *const comp_cost[2]) {
- return joint_cost[vp10_get_mv_joint(mv)] +
- comp_cost[0][mv->row] + comp_cost[1][mv->col];
+static INLINE int mv_cost(const MV *mv, const int *joint_cost,
+ int *const comp_cost[2]) {
+ return joint_cost[vp10_get_mv_joint(mv)] + comp_cost[0][mv->row] +
+ comp_cost[1][mv->col];
}
-int vp10_mv_bit_cost(const MV *mv, const MV *ref,
- const int *mvjcost, int *mvcost[2], int weight) {
- const MV diff = { mv->row - ref->row,
- mv->col - ref->col };
+int vp10_mv_bit_cost(const MV *mv, const MV *ref, const int *mvjcost,
+ int *mvcost[2], int weight) {
+ const MV diff = { mv->row - ref->row, mv->col - ref->col };
return ROUND_POWER_OF_TWO(mv_cost(&diff, mvjcost, mvcost) * weight, 7);
}
-static int mv_err_cost(const MV *mv, const MV *ref,
- const int *mvjcost, int *mvcost[2],
- int error_per_bit) {
+static int mv_err_cost(const MV *mv, const MV *ref, const int *mvjcost,
+ int *mvcost[2], int error_per_bit) {
if (mvcost) {
- const MV diff = { mv->row - ref->row,
- mv->col - ref->col };
- return ROUND_POWER_OF_TWO(mv_cost(&diff, mvjcost, mvcost) *
- error_per_bit, 13);
+ const MV diff = { mv->row - ref->row, mv->col - ref->col };
+ return ROUND_POWER_OF_TWO(mv_cost(&diff, mvjcost, mvcost) * error_per_bit,
+ 13);
}
return 0;
}
static int mvsad_err_cost(const MACROBLOCK *x, const MV *mv, const MV *ref,
int error_per_bit) {
- const MV diff = { mv->row - ref->row,
- mv->col - ref->col };
- return ROUND_POWER_OF_TWO(mv_cost(&diff, x->nmvjointsadcost,
- x->nmvsadcost) * error_per_bit, 8);
+ const MV diff = { mv->row - ref->row, mv->col - ref->col };
+ return ROUND_POWER_OF_TWO(
+ mv_cost(&diff, x->nmvjointsadcost, x->nmvsadcost) * error_per_bit, 8);
}
void vp10_init_dsmotion_compensation(search_site_config *cfg, int stride) {
@@ -108,7 +99,7 @@
for (len = MAX_FIRST_STEP; len > 0; len /= 2) {
// Generate offsets for 4 search sites per step.
- const MV ss_mvs[] = {{-len, 0}, {len, 0}, {0, -len}, {0, len}};
+ const MV ss_mvs[] = { { -len, 0 }, { len, 0 }, { 0, -len }, { 0, len } };
int i;
for (i = 0; i < 4; ++i) {
search_site *const ss = &cfg->ss[ss_count++];
@@ -129,10 +120,9 @@
for (len = MAX_FIRST_STEP; len > 0; len /= 2) {
// Generate offsets for 8 search sites per step.
- const MV ss_mvs[8] = {
- {-len, 0 }, {len, 0 }, { 0, -len}, {0, len},
- {-len, -len}, {-len, len}, {len, -len}, {len, len}
- };
+ const MV ss_mvs[8] = { { -len, 0 }, { len, 0 }, { 0, -len },
+ { 0, len }, { -len, -len }, { -len, len },
+ { len, -len }, { len, len } };
int i;
for (i = 0; i < 8; ++i) {
search_site *const ss = &cfg->ss[ss_count++];
@@ -156,172 +146,147 @@
*/
/* estimated cost of a motion vector (r,c) */
-#define MVC(r, c) \
- (mvcost ? \
- ((mvjcost[((r) != rr) * 2 + ((c) != rc)] + \
- mvcost[0][((r) - rr)] + mvcost[1][((c) - rc)]) * \
- error_per_bit + 4096) >> 13 : 0)
-
+#define MVC(r, c) \
+ (mvcost \
+ ? ((mvjcost[((r) != rr) * 2 + ((c) != rc)] + mvcost[0][((r)-rr)] + \
+ mvcost[1][((c)-rc)]) * \
+ error_per_bit + \
+ 4096) >> \
+ 13 \
+ : 0)
// convert motion vector component to offset for sv[a]f calc
-static INLINE int sp(int x) {
- return x & 7;
-}
+static INLINE int sp(int x) { return x & 7; }
static INLINE const uint8_t *pre(const uint8_t *buf, int stride, int r, int c) {
return &buf[(r >> 3) * stride + (c >> 3)];
}
/* checks if (r, c) has better score than previous best */
-#define CHECK_BETTER(v, r, c) \
- if (c >= minc && c <= maxc && r >= minr && r <= maxr) { \
- if (second_pred == NULL) \
- thismse = vfp->svf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), z, \
- src_stride, &sse); \
- else \
- thismse = vfp->svaf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), \
- z, src_stride, &sse, second_pred); \
- if ((v = MVC(r, c) + thismse) < besterr) { \
- besterr = v; \
- br = r; \
- bc = c; \
- *distortion = thismse; \
- *sse1 = sse; \
- } \
- } else { \
- v = INT_MAX; \
+#define CHECK_BETTER(v, r, c) \
+ if (c >= minc && c <= maxc && r >= minr && r <= maxr) { \
+ if (second_pred == NULL) \
+ thismse = vfp->svf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), z, \
+ src_stride, &sse); \
+ else \
+ thismse = vfp->svaf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), z, \
+ src_stride, &sse, second_pred); \
+ if ((v = MVC(r, c) + thismse) < besterr) { \
+ besterr = v; \
+ br = r; \
+ bc = c; \
+ *distortion = thismse; \
+ *sse1 = sse; \
+ } \
+ } else { \
+ v = INT_MAX; \
}
-#define FIRST_LEVEL_CHECKS \
- { \
- unsigned int left, right, up, down, diag; \
- CHECK_BETTER(left, tr, tc - hstep); \
- CHECK_BETTER(right, tr, tc + hstep); \
- CHECK_BETTER(up, tr - hstep, tc); \
- CHECK_BETTER(down, tr + hstep, tc); \
- whichdir = (left < right ? 0 : 1) + \
- (up < down ? 0 : 2); \
- switch (whichdir) { \
- case 0: \
- CHECK_BETTER(diag, tr - hstep, tc - hstep); \
- break; \
- case 1: \
- CHECK_BETTER(diag, tr - hstep, tc + hstep); \
- break; \
- case 2: \
- CHECK_BETTER(diag, tr + hstep, tc - hstep); \
- break; \
- case 3: \
- CHECK_BETTER(diag, tr + hstep, tc + hstep); \
- break; \
- } \
+#define FIRST_LEVEL_CHECKS \
+ { \
+ unsigned int left, right, up, down, diag; \
+ CHECK_BETTER(left, tr, tc - hstep); \
+ CHECK_BETTER(right, tr, tc + hstep); \
+ CHECK_BETTER(up, tr - hstep, tc); \
+ CHECK_BETTER(down, tr + hstep, tc); \
+ whichdir = (left < right ? 0 : 1) + (up < down ? 0 : 2); \
+ switch (whichdir) { \
+ case 0: CHECK_BETTER(diag, tr - hstep, tc - hstep); break; \
+ case 1: CHECK_BETTER(diag, tr - hstep, tc + hstep); break; \
+ case 2: CHECK_BETTER(diag, tr + hstep, tc - hstep); break; \
+ case 3: CHECK_BETTER(diag, tr + hstep, tc + hstep); break; \
+ } \
}
-#define SECOND_LEVEL_CHECKS \
- { \
- int kr, kc; \
- unsigned int second; \
- if (tr != br && tc != bc) { \
- kr = br - tr; \
- kc = bc - tc; \
- CHECK_BETTER(second, tr + kr, tc + 2 * kc); \
- CHECK_BETTER(second, tr + 2 * kr, tc + kc); \
- } else if (tr == br && tc != bc) { \
- kc = bc - tc; \
- CHECK_BETTER(second, tr + hstep, tc + 2 * kc); \
- CHECK_BETTER(second, tr - hstep, tc + 2 * kc); \
- switch (whichdir) { \
- case 0: \
- case 1: \
- CHECK_BETTER(second, tr + hstep, tc + kc); \
- break; \
- case 2: \
- case 3: \
- CHECK_BETTER(second, tr - hstep, tc + kc); \
- break; \
- } \
- } else if (tr != br && tc == bc) { \
- kr = br - tr; \
- CHECK_BETTER(second, tr + 2 * kr, tc + hstep); \
- CHECK_BETTER(second, tr + 2 * kr, tc - hstep); \
- switch (whichdir) { \
- case 0: \
- case 2: \
- CHECK_BETTER(second, tr + kr, tc + hstep); \
- break; \
- case 1: \
- case 3: \
- CHECK_BETTER(second, tr + kr, tc - hstep); \
- break; \
- } \
- } \
+#define SECOND_LEVEL_CHECKS \
+ { \
+ int kr, kc; \
+ unsigned int second; \
+ if (tr != br && tc != bc) { \
+ kr = br - tr; \
+ kc = bc - tc; \
+ CHECK_BETTER(second, tr + kr, tc + 2 * kc); \
+ CHECK_BETTER(second, tr + 2 * kr, tc + kc); \
+ } else if (tr == br && tc != bc) { \
+ kc = bc - tc; \
+ CHECK_BETTER(second, tr + hstep, tc + 2 * kc); \
+ CHECK_BETTER(second, tr - hstep, tc + 2 * kc); \
+ switch (whichdir) { \
+ case 0: \
+ case 1: CHECK_BETTER(second, tr + hstep, tc + kc); break; \
+ case 2: \
+ case 3: CHECK_BETTER(second, tr - hstep, tc + kc); break; \
+ } \
+ } else if (tr != br && tc == bc) { \
+ kr = br - tr; \
+ CHECK_BETTER(second, tr + 2 * kr, tc + hstep); \
+ CHECK_BETTER(second, tr + 2 * kr, tc - hstep); \
+ switch (whichdir) { \
+ case 0: \
+ case 2: CHECK_BETTER(second, tr + kr, tc + hstep); break; \
+ case 1: \
+ case 3: CHECK_BETTER(second, tr + kr, tc - hstep); break; \
+ } \
+ } \
}
// TODO(yunqingwang): SECOND_LEVEL_CHECKS_BEST was a rewrote of
// SECOND_LEVEL_CHECKS, and SECOND_LEVEL_CHECKS should be rewritten
// later in the same way.
-#define SECOND_LEVEL_CHECKS_BEST \
- { \
- unsigned int second; \
- int br0 = br; \
- int bc0 = bc; \
- assert(tr == br || tc == bc); \
- if (tr == br && tc != bc) { \
- kc = bc - tc; \
- } else if (tr != br && tc == bc) { \
- kr = br - tr; \
- } \
- CHECK_BETTER(second, br0 + kr, bc0); \
- CHECK_BETTER(second, br0, bc0 + kc); \
- if (br0 != br || bc0 != bc) { \
- CHECK_BETTER(second, br0 + kr, bc0 + kc); \
- } \
+#define SECOND_LEVEL_CHECKS_BEST \
+ { \
+ unsigned int second; \
+ int br0 = br; \
+ int bc0 = bc; \
+ assert(tr == br || tc == bc); \
+ if (tr == br && tc != bc) { \
+ kc = bc - tc; \
+ } else if (tr != br && tc == bc) { \
+ kr = br - tr; \
+ } \
+ CHECK_BETTER(second, br0 + kr, bc0); \
+ CHECK_BETTER(second, br0, bc0 + kc); \
+ if (br0 != br || bc0 != bc) { \
+ CHECK_BETTER(second, br0 + kr, bc0 + kc); \
+ } \
}
-#define SETUP_SUBPEL_SEARCH \
- const uint8_t *const z = x->plane[0].src.buf; \
- const int src_stride = x->plane[0].src.stride; \
- const MACROBLOCKD *xd = &x->e_mbd; \
- unsigned int besterr = INT_MAX; \
- unsigned int sse; \
- unsigned int whichdir; \
- int thismse; \
- const unsigned int halfiters = iters_per_step; \
- const unsigned int quarteriters = iters_per_step; \
- const unsigned int eighthiters = iters_per_step; \
- const int y_stride = xd->plane[0].pre[0].stride; \
- const int offset = bestmv->row * y_stride + bestmv->col; \
- const uint8_t *const y = xd->plane[0].pre[0].buf; \
- \
- int rr = ref_mv->row; \
- int rc = ref_mv->col; \
- int br = bestmv->row * 8; \
- int bc = bestmv->col * 8; \
- int hstep = 4; \
- const int minc = VPXMAX(x->mv_col_min * 8, ref_mv->col - MV_MAX); \
- const int maxc = VPXMIN(x->mv_col_max * 8, ref_mv->col + MV_MAX); \
- const int minr = VPXMAX(x->mv_row_min * 8, ref_mv->row - MV_MAX); \
- const int maxr = VPXMIN(x->mv_row_max * 8, ref_mv->row + MV_MAX); \
- int tr = br; \
- int tc = bc; \
- \
- bestmv->row *= 8; \
+#define SETUP_SUBPEL_SEARCH \
+ const uint8_t *const z = x->plane[0].src.buf; \
+ const int src_stride = x->plane[0].src.stride; \
+ const MACROBLOCKD *xd = &x->e_mbd; \
+ unsigned int besterr = INT_MAX; \
+ unsigned int sse; \
+ unsigned int whichdir; \
+ int thismse; \
+ const unsigned int halfiters = iters_per_step; \
+ const unsigned int quarteriters = iters_per_step; \
+ const unsigned int eighthiters = iters_per_step; \
+ const int y_stride = xd->plane[0].pre[0].stride; \
+ const int offset = bestmv->row * y_stride + bestmv->col; \
+ const uint8_t *const y = xd->plane[0].pre[0].buf; \
+ \
+ int rr = ref_mv->row; \
+ int rc = ref_mv->col; \
+ int br = bestmv->row * 8; \
+ int bc = bestmv->col * 8; \
+ int hstep = 4; \
+ const int minc = VPXMAX(x->mv_col_min * 8, ref_mv->col - MV_MAX); \
+ const int maxc = VPXMIN(x->mv_col_max * 8, ref_mv->col + MV_MAX); \
+ const int minr = VPXMAX(x->mv_row_min * 8, ref_mv->row - MV_MAX); \
+ const int maxr = VPXMIN(x->mv_row_max * 8, ref_mv->row + MV_MAX); \
+ int tr = br; \
+ int tc = bc; \
+ \
+ bestmv->row *= 8; \
bestmv->col *= 8;
-static unsigned int setup_center_error(const MACROBLOCKD *xd,
- const MV *bestmv,
- const MV *ref_mv,
- int error_per_bit,
- const vpx_variance_fn_ptr_t *vfp,
- const uint8_t *const src,
- const int src_stride,
- const uint8_t *const y,
- int y_stride,
- const uint8_t *second_pred,
- int w, int h, int offset,
- int *mvjcost, int *mvcost[2],
- unsigned int *sse1,
- int *distortion) {
+static unsigned int setup_center_error(
+ const MACROBLOCKD *xd, const MV *bestmv, const MV *ref_mv,
+ int error_per_bit, const vpx_variance_fn_ptr_t *vfp,
+ const uint8_t *const src, const int src_stride, const uint8_t *const y,
+ int y_stride, const uint8_t *second_pred, int w, int h, int offset,
+ int *mvjcost, int *mvcost[2], unsigned int *sse1, int *distortion) {
unsigned int besterr;
#if CONFIG_VPX_HIGHBITDEPTH
if (second_pred != NULL) {
@@ -329,8 +294,8 @@
DECLARE_ALIGNED(16, uint16_t, comp_pred16[64 * 64]);
vpx_highbd_comp_avg_pred(comp_pred16, second_pred, w, h, y + offset,
y_stride);
- besterr = vfp->vf(CONVERT_TO_BYTEPTR(comp_pred16), w, src, src_stride,
- sse1);
+ besterr =
+ vfp->vf(CONVERT_TO_BYTEPTR(comp_pred16), w, src, src_stride, sse1);
} else {
DECLARE_ALIGNED(16, uint8_t, comp_pred[64 * 64]);
vpx_comp_avg_pred(comp_pred, second_pred, w, h, y + offset, y_stride);
@@ -342,7 +307,7 @@
*distortion = besterr;
besterr += mv_err_cost(bestmv, ref_mv, mvjcost, mvcost, error_per_bit);
#else
- (void) xd;
+ (void)xd;
if (second_pred != NULL) {
DECLARE_ALIGNED(16, uint8_t, comp_pred[64 * 64]);
vpx_comp_avg_pred(comp_pred, second_pred, w, h, y + offset, y_stride);
@@ -361,10 +326,8 @@
}
static INLINE int is_cost_list_wellbehaved(int *cost_list) {
- return cost_list[0] < cost_list[1] &&
- cost_list[0] < cost_list[2] &&
- cost_list[0] < cost_list[3] &&
- cost_list[0] < cost_list[4];
+ return cost_list[0] < cost_list[1] && cost_list[0] < cost_list[2] &&
+ cost_list[0] < cost_list[3] && cost_list[0] < cost_list[4];
}
// Returns surface minima estimate at given precision in 1/2^n bits.
@@ -375,8 +338,7 @@
// x0 = 1/2 (S1 - S3)/(S1 + S3 - 2*S0),
// y0 = 1/2 (S4 - S2)/(S4 + S2 - 2*S0).
// The code below is an integerized version of that.
-static void get_cost_surf_min(int *cost_list, int *ir, int *ic,
- int bits) {
+static void get_cost_surf_min(int *cost_list, int *ir, int *ic, int bits) {
*ic = divide_and_round((cost_list[1] - cost_list[3]) * (1 << (bits - 1)),
(cost_list[1] - 2 * cost_list[0] + cost_list[3]));
*ir = divide_and_round((cost_list[4] - cost_list[2]) * (1 << (bits - 1)),
@@ -384,37 +346,26 @@
}
int vp10_find_best_sub_pixel_tree_pruned_evenmore(
- const MACROBLOCK *x,
- MV *bestmv, const MV *ref_mv,
- int allow_hp,
- int error_per_bit,
- const vpx_variance_fn_ptr_t *vfp,
- int forced_stop,
- int iters_per_step,
- int *cost_list,
- int *mvjcost, int *mvcost[2],
- int *distortion,
- unsigned int *sse1,
- const uint8_t *second_pred,
- int w, int h) {
+ const MACROBLOCK *x, MV *bestmv, const MV *ref_mv, int allow_hp,
+ int error_per_bit, const vpx_variance_fn_ptr_t *vfp, int forced_stop,
+ int iters_per_step, int *cost_list, int *mvjcost, int *mvcost[2],
+ int *distortion, unsigned int *sse1, const uint8_t *second_pred, int w,
+ int h) {
SETUP_SUBPEL_SEARCH;
- besterr = setup_center_error(xd, bestmv, ref_mv, error_per_bit, vfp,
- z, src_stride, y, y_stride, second_pred,
- w, h, offset, mvjcost, mvcost,
- sse1, distortion);
- (void) halfiters;
- (void) quarteriters;
- (void) eighthiters;
- (void) whichdir;
- (void) allow_hp;
- (void) forced_stop;
- (void) hstep;
+ besterr = setup_center_error(xd, bestmv, ref_mv, error_per_bit, vfp, z,
+ src_stride, y, y_stride, second_pred, w, h,
+ offset, mvjcost, mvcost, sse1, distortion);
+ (void)halfiters;
+ (void)quarteriters;
+ (void)eighthiters;
+ (void)whichdir;
+ (void)allow_hp;
+ (void)forced_stop;
+ (void)hstep;
- if (cost_list &&
- cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
+ if (cost_list && cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
cost_list[2] != INT_MAX && cost_list[3] != INT_MAX &&
- cost_list[4] != INT_MAX &&
- is_cost_list_wellbehaved(cost_list)) {
+ cost_list[4] != INT_MAX && is_cost_list_wellbehaved(cost_list)) {
int ir, ic;
unsigned int minpt;
get_cost_surf_min(cost_list, &ir, &ic, 2);
@@ -463,29 +414,19 @@
return besterr;
}
-int vp10_find_best_sub_pixel_tree_pruned_more(const MACROBLOCK *x,
- MV *bestmv, const MV *ref_mv,
- int allow_hp,
- int error_per_bit,
- const vpx_variance_fn_ptr_t *vfp,
- int forced_stop,
- int iters_per_step,
- int *cost_list,
- int *mvjcost, int *mvcost[2],
- int *distortion,
- unsigned int *sse1,
- const uint8_t *second_pred,
- int w, int h) {
+int vp10_find_best_sub_pixel_tree_pruned_more(
+ const MACROBLOCK *x, MV *bestmv, const MV *ref_mv, int allow_hp,
+ int error_per_bit, const vpx_variance_fn_ptr_t *vfp, int forced_stop,
+ int iters_per_step, int *cost_list, int *mvjcost, int *mvcost[2],
+ int *distortion, unsigned int *sse1, const uint8_t *second_pred, int w,
+ int h) {
SETUP_SUBPEL_SEARCH;
- besterr = setup_center_error(xd, bestmv, ref_mv, error_per_bit, vfp,
- z, src_stride, y, y_stride, second_pred,
- w, h, offset, mvjcost, mvcost,
- sse1, distortion);
- if (cost_list &&
- cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
+ besterr = setup_center_error(xd, bestmv, ref_mv, error_per_bit, vfp, z,
+ src_stride, y, y_stride, second_pred, w, h,
+ offset, mvjcost, mvcost, sse1, distortion);
+ if (cost_list && cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
cost_list[2] != INT_MAX && cost_list[3] != INT_MAX &&
- cost_list[4] != INT_MAX &&
- is_cost_list_wellbehaved(cost_list)) {
+ cost_list[4] != INT_MAX && is_cost_list_wellbehaved(cost_list)) {
unsigned int minpt;
int ir, ic;
get_cost_surf_min(cost_list, &ir, &ic, 1);
@@ -524,8 +465,8 @@
}
// These lines insure static analysis doesn't warn that
// tr and tc aren't used after the above point.
- (void) tr;
- (void) tc;
+ (void)tr;
+ (void)tc;
bestmv->row = br;
bestmv->col = bc;
@@ -537,26 +478,17 @@
return besterr;
}
-int vp10_find_best_sub_pixel_tree_pruned(const MACROBLOCK *x,
- MV *bestmv, const MV *ref_mv,
- int allow_hp,
- int error_per_bit,
- const vpx_variance_fn_ptr_t *vfp,
- int forced_stop,
- int iters_per_step,
- int *cost_list,
- int *mvjcost, int *mvcost[2],
- int *distortion,
- unsigned int *sse1,
- const uint8_t *second_pred,
- int w, int h) {
+int vp10_find_best_sub_pixel_tree_pruned(
+ const MACROBLOCK *x, MV *bestmv, const MV *ref_mv, int allow_hp,
+ int error_per_bit, const vpx_variance_fn_ptr_t *vfp, int forced_stop,
+ int iters_per_step, int *cost_list, int *mvjcost, int *mvcost[2],
+ int *distortion, unsigned int *sse1, const uint8_t *second_pred, int w,
+ int h) {
SETUP_SUBPEL_SEARCH;
- besterr = setup_center_error(xd, bestmv, ref_mv, error_per_bit, vfp,
- z, src_stride, y, y_stride, second_pred,
- w, h, offset, mvjcost, mvcost,
- sse1, distortion);
- if (cost_list &&
- cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
+ besterr = setup_center_error(xd, bestmv, ref_mv, error_per_bit, vfp, z,
+ src_stride, y, y_stride, second_pred, w, h,
+ offset, mvjcost, mvcost, sse1, distortion);
+ if (cost_list && cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
cost_list[2] != INT_MAX && cost_list[3] != INT_MAX &&
cost_list[4] != INT_MAX) {
unsigned int left, right, up, down, diag;
@@ -619,8 +551,8 @@
}
// These lines insure static analysis doesn't warn that
// tr and tc aren't used after the above point.
- (void) tr;
- (void) tc;
+ (void)tr;
+ (void)tc;
bestmv->row = br;
bestmv->col = bc;
@@ -633,25 +565,19 @@
}
static const MV search_step_table[12] = {
- // left, right, up, down
- {0, -4}, {0, 4}, {-4, 0}, {4, 0},
- {0, -2}, {0, 2}, {-2, 0}, {2, 0},
- {0, -1}, {0, 1}, {-1, 0}, {1, 0}
+ // left, right, up, down
+ { 0, -4 }, { 0, 4 }, { -4, 0 }, { 4, 0 }, { 0, -2 }, { 0, 2 },
+ { -2, 0 }, { 2, 0 }, { 0, -1 }, { 0, 1 }, { -1, 0 }, { 1, 0 }
};
-int vp10_find_best_sub_pixel_tree(const MACROBLOCK *x,
- MV *bestmv, const MV *ref_mv,
- int allow_hp,
- int error_per_bit,
- const vpx_variance_fn_ptr_t *vfp,
- int forced_stop,
- int iters_per_step,
- int *cost_list,
- int *mvjcost, int *mvcost[2],
- int *distortion,
- unsigned int *sse1,
- const uint8_t *second_pred,
- int w, int h) {
+int vp10_find_best_sub_pixel_tree(const MACROBLOCK *x, MV *bestmv,
+ const MV *ref_mv, int allow_hp,
+ int error_per_bit,
+ const vpx_variance_fn_ptr_t *vfp,
+ int forced_stop, int iters_per_step,
+ int *cost_list, int *mvjcost, int *mvcost[2],
+ int *distortion, unsigned int *sse1,
+ const uint8_t *second_pred, int w, int h) {
const uint8_t *const z = x->plane[0].src.buf;
const uint8_t *const src_address = z;
const int src_stride = x->plane[0].src.stride;
@@ -681,18 +607,16 @@
int kr, kc;
if (!(allow_hp && vp10_use_mv_hp(ref_mv)))
- if (round == 3)
- round = 2;
+ if (round == 3) round = 2;
bestmv->row *= 8;
bestmv->col *= 8;
- besterr = setup_center_error(xd, bestmv, ref_mv, error_per_bit, vfp,
- z, src_stride, y, y_stride, second_pred,
- w, h, offset, mvjcost, mvcost,
- sse1, distortion);
+ besterr = setup_center_error(xd, bestmv, ref_mv, error_per_bit, vfp, z,
+ src_stride, y, y_stride, second_pred, w, h,
+ offset, mvjcost, mvcost, sse1, distortion);
- (void) cost_list; // to silence compiler warning
+ (void)cost_list; // to silence compiler warning
for (iter = 0; iter < round; ++iter) {
// Check vertical and horizontal sub-pixel positions.
@@ -705,13 +629,13 @@
this_mv.row = tr;
this_mv.col = tc;
if (second_pred == NULL)
- thismse = vfp->svf(pre_address, y_stride, sp(tc), sp(tr),
- src_address, src_stride, &sse);
+ thismse = vfp->svf(pre_address, y_stride, sp(tc), sp(tr), src_address,
+ src_stride, &sse);
else
thismse = vfp->svaf(pre_address, y_stride, sp(tc), sp(tr),
src_address, src_stride, &sse, second_pred);
- cost_array[idx] = thismse +
- mv_err_cost(&this_mv, ref_mv, mvjcost, mvcost, error_per_bit);
+ cost_array[idx] = thismse + mv_err_cost(&this_mv, ref_mv, mvjcost,
+ mvcost, error_per_bit);
if (cost_array[idx] < besterr) {
best_idx = idx;
@@ -732,15 +656,15 @@
tr = br + kr;
if (tc >= minc && tc <= maxc && tr >= minr && tr <= maxr) {
const uint8_t *const pre_address = y + (tr >> 3) * y_stride + (tc >> 3);
- MV this_mv = {tr, tc};
+ MV this_mv = { tr, tc };
if (second_pred == NULL)
- thismse = vfp->svf(pre_address, y_stride, sp(tc), sp(tr),
- src_address, src_stride, &sse);
+ thismse = vfp->svf(pre_address, y_stride, sp(tc), sp(tr), src_address,
+ src_stride, &sse);
else
- thismse = vfp->svaf(pre_address, y_stride, sp(tc), sp(tr),
- src_address, src_stride, &sse, second_pred);
- cost_array[4] = thismse +
- mv_err_cost(&this_mv, ref_mv, mvjcost, mvcost, error_per_bit);
+ thismse = vfp->svaf(pre_address, y_stride, sp(tc), sp(tr), src_address,
+ src_stride, &sse, second_pred);
+ cost_array[4] = thismse + mv_err_cost(&this_mv, ref_mv, mvjcost, mvcost,
+ error_per_bit);
if (cost_array[4] < besterr) {
best_idx = 4;
@@ -760,8 +684,7 @@
bc = tc;
}
- if (iters_per_step > 1 && best_idx != -1)
- SECOND_LEVEL_CHECKS_BEST;
+ if (iters_per_step > 1 && best_idx != -1) SECOND_LEVEL_CHECKS_BEST;
tr = br;
tc = bc;
@@ -776,8 +699,8 @@
// These lines insure static analysis doesn't warn that
// tr and tc aren't used after the above point.
- (void) tr;
- (void) tc;
+ (void)tr;
+ (void)tc;
bestmv->row = br;
bestmv->col = bc;
@@ -795,10 +718,8 @@
static INLINE int check_bounds(const MACROBLOCK *x, int row, int col,
int range) {
- return ((row - range) >= x->mv_row_min) &
- ((row + range) <= x->mv_row_max) &
- ((col - range) >= x->mv_col_min) &
- ((col + range) <= x->mv_col_max);
+ return ((row - range) >= x->mv_row_min) & ((row + range) <= x->mv_row_max) &
+ ((col - range) >= x->mv_col_min) & ((col + range) <= x->mv_col_max);
}
static INLINE int is_mv_in(const MACROBLOCK *x, const MV *mv) {
@@ -806,33 +727,31 @@
(mv->row >= x->mv_row_min) && (mv->row <= x->mv_row_max);
}
-#define CHECK_BETTER \
- {\
- if (thissad < bestsad) {\
- if (use_mvcost) \
- thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);\
- if (thissad < bestsad) {\
- bestsad = thissad;\
- best_site = i;\
- }\
- }\
+#define CHECK_BETTER \
+ { \
+ if (thissad < bestsad) { \
+ if (use_mvcost) \
+ thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit); \
+ if (thissad < bestsad) { \
+ bestsad = thissad; \
+ best_site = i; \
+ } \
+ } \
}
-#define MAX_PATTERN_SCALES 11
-#define MAX_PATTERN_CANDIDATES 8 // max number of canddiates per scale
-#define PATTERN_CANDIDATES_REF 3 // number of refinement candidates
+#define MAX_PATTERN_SCALES 11
+#define MAX_PATTERN_CANDIDATES 8 // max number of canddiates per scale
+#define PATTERN_CANDIDATES_REF 3 // number of refinement candidates
// Calculate and return a sad+mvcost list around an integer best pel.
-static INLINE void calc_int_cost_list(const MACROBLOCK *x,
- const MV *ref_mv,
+static INLINE void calc_int_cost_list(const MACROBLOCK *x, const MV *ref_mv,
int sadpb,
const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *best_mv,
- int *cost_list) {
- static const MV neighbors[4] = {{0, -1}, {1, 0}, {0, 1}, {-1, 0}};
+ const MV *best_mv, int *cost_list) {
+ static const MV neighbors[4] = { { 0, -1 }, { 1, 0 }, { 0, 1 }, { -1, 0 } };
const struct buf_2d *const what = &x->plane[0].src;
const struct buf_2d *const in_what = &x->e_mbd.plane[0].pre[0];
- const MV fcenter_mv = {ref_mv->row >> 3, ref_mv->col >> 3};
+ const MV fcenter_mv = { ref_mv->row >> 3, ref_mv->col >> 3 };
int br = best_mv->row;
int bc = best_mv->col;
MV this_mv;
@@ -841,34 +760,32 @@
this_mv.row = br;
this_mv.col = bc;
- cost_list[0] = fn_ptr->vf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride, &sse) +
+ cost_list[0] =
+ fn_ptr->vf(what->buf, what->stride, get_buf_from_mv(in_what, &this_mv),
+ in_what->stride, &sse) +
mvsad_err_cost(x, &this_mv, &fcenter_mv, sadpb);
if (check_bounds(x, br, bc, 1)) {
for (i = 0; i < 4; i++) {
- const MV this_mv = {br + neighbors[i].row,
- bc + neighbors[i].col};
+ const MV this_mv = { br + neighbors[i].row, bc + neighbors[i].col };
cost_list[i + 1] = fn_ptr->vf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
in_what->stride, &sse) +
- // mvsad_err_cost(x, &this_mv, &fcenter_mv, sadpb);
- mv_err_cost(&this_mv, &fcenter_mv, x->nmvjointcost, x->mvcost,
- x->errorperbit);
+ // mvsad_err_cost(x, &this_mv, &fcenter_mv, sadpb);
+ mv_err_cost(&this_mv, &fcenter_mv, x->nmvjointcost,
+ x->mvcost, x->errorperbit);
}
} else {
for (i = 0; i < 4; i++) {
- const MV this_mv = {br + neighbors[i].row,
- bc + neighbors[i].col};
+ const MV this_mv = { br + neighbors[i].row, bc + neighbors[i].col };
if (!is_mv_in(x, &this_mv))
cost_list[i + 1] = INT_MAX;
else
cost_list[i + 1] = fn_ptr->vf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
in_what->stride, &sse) +
- // mvsad_err_cost(x, &this_mv, &fcenter_mv, sadpb);
- mv_err_cost(&this_mv, &fcenter_mv, x->nmvjointcost, x->mvcost,
- x->errorperbit);
+ // mvsad_err_cost(x, &this_mv, &fcenter_mv, sadpb);
+ mv_err_cost(&this_mv, &fcenter_mv, x->nmvjointcost,
+ x->mvcost, x->errorperbit);
}
}
}
@@ -878,19 +795,12 @@
// candidates as indicated in the num_candidates and candidates arrays
// passed into this function
//
-static int vp10_pattern_search(const MACROBLOCK *x,
- MV *ref_mv,
- int search_param,
- int sad_per_bit,
- int do_init_search,
- int *cost_list,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost,
- const MV *center_mv,
- MV *best_mv,
- const int num_candidates[MAX_PATTERN_SCALES],
- const MV candidates[MAX_PATTERN_SCALES]
- [MAX_PATTERN_CANDIDATES]) {
+static int vp10_pattern_search(
+ const MACROBLOCK *x, MV *ref_mv, int search_param, int sad_per_bit,
+ int do_init_search, int *cost_list, const vpx_variance_fn_ptr_t *vfp,
+ int use_mvcost, const MV *center_mv, MV *best_mv,
+ const int num_candidates[MAX_PATTERN_SCALES],
+ const MV candidates[MAX_PATTERN_SCALES][MAX_PATTERN_CANDIDATES]) {
const MACROBLOCKD *const xd = &x->e_mbd;
static const int search_param_to_steps[MAX_MVSEARCH_STEPS] = {
10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0,
@@ -902,7 +812,7 @@
int bestsad = INT_MAX;
int thissad;
int k = -1;
- const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
+ const MV fcenter_mv = { center_mv->row >> 3, center_mv->col >> 3 };
int best_init_s = search_param_to_steps[search_param];
// adjust ref_mv to make sure it is within MV range
clamp_mv(ref_mv, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max);
@@ -910,9 +820,9 @@
bc = ref_mv->col;
// Work out the start point for the search
- bestsad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, ref_mv), in_what->stride) +
- mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
+ bestsad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, ref_mv),
+ in_what->stride) +
+ mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
// Search all possible scales upto the search param around the center point
// pick the scale of the point that is best as the starting scale of
@@ -924,22 +834,21 @@
int best_site = -1;
if (check_bounds(x, br, bc, 1 << t)) {
for (i = 0; i < num_candidates[t]; i++) {
- const MV this_mv = {br + candidates[t][i].row,
- bc + candidates[t][i].col};
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[t][i].row,
+ bc + candidates[t][i].col };
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
} else {
for (i = 0; i < num_candidates[t]; i++) {
- const MV this_mv = {br + candidates[t][i].row,
- bc + candidates[t][i].col};
- if (!is_mv_in(x, &this_mv))
- continue;
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[t][i].row,
+ bc + candidates[t][i].col };
+ if (!is_mv_in(x, &this_mv)) continue;
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
}
@@ -967,22 +876,21 @@
if (!do_init_search || s != best_init_s) {
if (check_bounds(x, br, bc, 1 << s)) {
for (i = 0; i < num_candidates[s]; i++) {
- const MV this_mv = {br + candidates[s][i].row,
- bc + candidates[s][i].col};
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[s][i].row,
+ bc + candidates[s][i].col };
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
} else {
for (i = 0; i < num_candidates[s]; i++) {
- const MV this_mv = {br + candidates[s][i].row,
- bc + candidates[s][i].col};
- if (!is_mv_in(x, &this_mv))
- continue;
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[s][i].row,
+ bc + candidates[s][i].col };
+ if (!is_mv_in(x, &this_mv)) continue;
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
}
@@ -1005,22 +913,25 @@
if (check_bounds(x, br, bc, 1 << s)) {
for (i = 0; i < PATTERN_CANDIDATES_REF; i++) {
- const MV this_mv = {br + candidates[s][next_chkpts_indices[i]].row,
- bc + candidates[s][next_chkpts_indices[i]].col};
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = {
+ br + candidates[s][next_chkpts_indices[i]].row,
+ bc + candidates[s][next_chkpts_indices[i]].col
+ };
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
} else {
for (i = 0; i < PATTERN_CANDIDATES_REF; i++) {
- const MV this_mv = {br + candidates[s][next_chkpts_indices[i]].row,
- bc + candidates[s][next_chkpts_indices[i]].col};
- if (!is_mv_in(x, &this_mv))
- continue;
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = {
+ br + candidates[s][next_chkpts_indices[i]].row,
+ bc + candidates[s][next_chkpts_indices[i]].col
+ };
+ if (!is_mv_in(x, &this_mv)) continue;
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
}
@@ -1053,19 +964,12 @@
// are 4 1-away neighbors, and cost_list is non-null
// TODO(debargha): Merge this function with the one above. Also remove
// use_mvcost option since it is always 1, to save unnecessary branches.
-static int vp10_pattern_search_sad(const MACROBLOCK *x,
- MV *ref_mv,
- int search_param,
- int sad_per_bit,
- int do_init_search,
- int *cost_list,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost,
- const MV *center_mv,
- MV *best_mv,
- const int num_candidates[MAX_PATTERN_SCALES],
- const MV candidates[MAX_PATTERN_SCALES]
- [MAX_PATTERN_CANDIDATES]) {
+static int vp10_pattern_search_sad(
+ const MACROBLOCK *x, MV *ref_mv, int search_param, int sad_per_bit,
+ int do_init_search, int *cost_list, const vpx_variance_fn_ptr_t *vfp,
+ int use_mvcost, const MV *center_mv, MV *best_mv,
+ const int num_candidates[MAX_PATTERN_SCALES],
+ const MV candidates[MAX_PATTERN_SCALES][MAX_PATTERN_CANDIDATES]) {
const MACROBLOCKD *const xd = &x->e_mbd;
static const int search_param_to_steps[MAX_MVSEARCH_STEPS] = {
10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0,
@@ -1077,7 +981,7 @@
int bestsad = INT_MAX;
int thissad;
int k = -1;
- const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
+ const MV fcenter_mv = { center_mv->row >> 3, center_mv->col >> 3 };
int best_init_s = search_param_to_steps[search_param];
// adjust ref_mv to make sure it is within MV range
clamp_mv(ref_mv, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max);
@@ -1089,9 +993,9 @@
}
// Work out the start point for the search
- bestsad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, ref_mv), in_what->stride) +
- mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
+ bestsad = vfp->sdf(what->buf, what->stride, get_buf_from_mv(in_what, ref_mv),
+ in_what->stride) +
+ mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
// Search all possible scales upto the search param around the center point
// pick the scale of the point that is best as the starting scale of
@@ -1103,22 +1007,21 @@
int best_site = -1;
if (check_bounds(x, br, bc, 1 << t)) {
for (i = 0; i < num_candidates[t]; i++) {
- const MV this_mv = {br + candidates[t][i].row,
- bc + candidates[t][i].col};
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[t][i].row,
+ bc + candidates[t][i].col };
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
} else {
for (i = 0; i < num_candidates[t]; i++) {
- const MV this_mv = {br + candidates[t][i].row,
- bc + candidates[t][i].col};
- if (!is_mv_in(x, &this_mv))
- continue;
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[t][i].row,
+ bc + candidates[t][i].col };
+ if (!is_mv_in(x, &this_mv)) continue;
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
}
@@ -1146,22 +1049,21 @@
if (!do_init_search || s != best_init_s) {
if (check_bounds(x, br, bc, 1 << s)) {
for (i = 0; i < num_candidates[s]; i++) {
- const MV this_mv = {br + candidates[s][i].row,
- bc + candidates[s][i].col};
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[s][i].row,
+ bc + candidates[s][i].col };
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
} else {
for (i = 0; i < num_candidates[s]; i++) {
- const MV this_mv = {br + candidates[s][i].row,
- bc + candidates[s][i].col};
- if (!is_mv_in(x, &this_mv))
- continue;
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[s][i].row,
+ bc + candidates[s][i].col };
+ if (!is_mv_in(x, &this_mv)) continue;
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
}
@@ -1184,22 +1086,25 @@
if (check_bounds(x, br, bc, 1 << s)) {
for (i = 0; i < PATTERN_CANDIDATES_REF; i++) {
- const MV this_mv = {br + candidates[s][next_chkpts_indices[i]].row,
- bc + candidates[s][next_chkpts_indices[i]].col};
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = {
+ br + candidates[s][next_chkpts_indices[i]].row,
+ bc + candidates[s][next_chkpts_indices[i]].col
+ };
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
} else {
for (i = 0; i < PATTERN_CANDIDATES_REF; i++) {
- const MV this_mv = {br + candidates[s][next_chkpts_indices[i]].row,
- bc + candidates[s][next_chkpts_indices[i]].col};
- if (!is_mv_in(x, &this_mv))
- continue;
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = {
+ br + candidates[s][next_chkpts_indices[i]].row,
+ bc + candidates[s][next_chkpts_indices[i]].col
+ };
+ if (!is_mv_in(x, &this_mv)) continue;
+ thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
}
@@ -1218,24 +1123,21 @@
if (!do_init_search || s != best_init_s) {
if (check_bounds(x, br, bc, 1 << s)) {
for (i = 0; i < num_candidates[s]; i++) {
- const MV this_mv = {br + candidates[s][i].row,
- bc + candidates[s][i].col};
- cost_list[i + 1] =
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[s][i].row,
+ bc + candidates[s][i].col };
+ cost_list[i + 1] = thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
} else {
for (i = 0; i < num_candidates[s]; i++) {
- const MV this_mv = {br + candidates[s][i].row,
- bc + candidates[s][i].col};
- if (!is_mv_in(x, &this_mv))
- continue;
- cost_list[i + 1] =
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + candidates[s][i].row,
+ bc + candidates[s][i].col };
+ if (!is_mv_in(x, &this_mv)) continue;
+ cost_list[i + 1] = thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
}
@@ -1258,26 +1160,28 @@
if (check_bounds(x, br, bc, 1 << s)) {
for (i = 0; i < PATTERN_CANDIDATES_REF; i++) {
- const MV this_mv = {br + candidates[s][next_chkpts_indices[i]].row,
- bc + candidates[s][next_chkpts_indices[i]].col};
- cost_list[next_chkpts_indices[i] + 1] =
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = {
+ br + candidates[s][next_chkpts_indices[i]].row,
+ bc + candidates[s][next_chkpts_indices[i]].col
+ };
+ cost_list[next_chkpts_indices[i] + 1] = thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
} else {
for (i = 0; i < PATTERN_CANDIDATES_REF; i++) {
- const MV this_mv = {br + candidates[s][next_chkpts_indices[i]].row,
- bc + candidates[s][next_chkpts_indices[i]].col};
+ const MV this_mv = {
+ br + candidates[s][next_chkpts_indices[i]].row,
+ bc + candidates[s][next_chkpts_indices[i]].col
+ };
if (!is_mv_in(x, &this_mv)) {
cost_list[next_chkpts_indices[i] + 1] = INT_MAX;
continue;
}
- cost_list[next_chkpts_indices[i] + 1] =
- thissad = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ cost_list[next_chkpts_indices[i] + 1] = thissad =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
CHECK_BETTER
}
}
@@ -1298,34 +1202,31 @@
// cost_list[3]: sad at delta { 0, 1} (right) from the best integer pel
// cost_list[4]: sad at delta {-1, 0} (top) from the best integer pel
if (cost_list) {
- static const MV neighbors[4] = {{0, -1}, {1, 0}, {0, 1}, {-1, 0}};
+ static const MV neighbors[4] = { { 0, -1 }, { 1, 0 }, { 0, 1 }, { -1, 0 } };
if (cost_list[0] == INT_MAX) {
cost_list[0] = bestsad;
if (check_bounds(x, br, bc, 1)) {
for (i = 0; i < 4; i++) {
- const MV this_mv = { br + neighbors[i].row,
- bc + neighbors[i].col };
- cost_list[i + 1] = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ const MV this_mv = { br + neighbors[i].row, bc + neighbors[i].col };
+ cost_list[i + 1] =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
}
} else {
for (i = 0; i < 4; i++) {
- const MV this_mv = {br + neighbors[i].row,
- bc + neighbors[i].col};
+ const MV this_mv = { br + neighbors[i].row, bc + neighbors[i].col };
if (!is_mv_in(x, &this_mv))
cost_list[i + 1] = INT_MAX;
else
- cost_list[i + 1] = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ cost_list[i + 1] =
+ vfp->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv), in_what->stride);
}
}
} else {
if (use_mvcost) {
for (i = 0; i < 4; i++) {
- const MV this_mv = {br + neighbors[i].row,
- bc + neighbors[i].col};
+ const MV this_mv = { br + neighbors[i].row, bc + neighbors[i].col };
if (cost_list[i + 1] != INT_MAX) {
cost_list[i + 1] +=
mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
@@ -1339,183 +1240,320 @@
return bestsad;
}
-int vp10_get_mvpred_var(const MACROBLOCK *x,
- const MV *best_mv, const MV *center_mv,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost) {
+int vp10_get_mvpred_var(const MACROBLOCK *x, const MV *best_mv,
+ const MV *center_mv, const vpx_variance_fn_ptr_t *vfp,
+ int use_mvcost) {
const MACROBLOCKD *const xd = &x->e_mbd;
const struct buf_2d *const what = &x->plane[0].src;
const struct buf_2d *const in_what = &xd->plane[0].pre[0];
- const MV mv = {best_mv->row * 8, best_mv->col * 8};
+ const MV mv = { best_mv->row * 8, best_mv->col * 8 };
unsigned int unused;
- return vfp->vf(what->buf, what->stride,
- get_buf_from_mv(in_what, best_mv), in_what->stride, &unused) +
- (use_mvcost ? mv_err_cost(&mv, center_mv, x->nmvjointcost,
- x->mvcost, x->errorperbit) : 0);
+ return vfp->vf(what->buf, what->stride, get_buf_from_mv(in_what, best_mv),
+ in_what->stride, &unused) +
+ (use_mvcost ? mv_err_cost(&mv, center_mv, x->nmvjointcost, x->mvcost,
+ x->errorperbit)
+ : 0);
}
-int vp10_get_mvpred_av_var(const MACROBLOCK *x,
- const MV *best_mv, const MV *center_mv,
- const uint8_t *second_pred,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost) {
+int vp10_get_mvpred_av_var(const MACROBLOCK *x, const MV *best_mv,
+ const MV *center_mv, const uint8_t *second_pred,
+ const vpx_variance_fn_ptr_t *vfp, int use_mvcost) {
const MACROBLOCKD *const xd = &x->e_mbd;
const struct buf_2d *const what = &x->plane[0].src;
const struct buf_2d *const in_what = &xd->plane[0].pre[0];
- const MV mv = {best_mv->row * 8, best_mv->col * 8};
+ const MV mv = { best_mv->row * 8, best_mv->col * 8 };
unsigned int unused;
return vfp->svaf(get_buf_from_mv(in_what, best_mv), in_what->stride, 0, 0,
what->buf, what->stride, &unused, second_pred) +
- (use_mvcost ? mv_err_cost(&mv, center_mv, x->nmvjointcost,
- x->mvcost, x->errorperbit) : 0);
+ (use_mvcost ? mv_err_cost(&mv, center_mv, x->nmvjointcost, x->mvcost,
+ x->errorperbit)
+ : 0);
}
-int vp10_hex_search(const MACROBLOCK *x,
- MV *ref_mv,
- int search_param,
- int sad_per_bit,
- int do_init_search,
- int *cost_list,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost,
- const MV *center_mv, MV *best_mv) {
+int vp10_hex_search(const MACROBLOCK *x, MV *ref_mv, int search_param,
+ int sad_per_bit, int do_init_search, int *cost_list,
+ const vpx_variance_fn_ptr_t *vfp, int use_mvcost,
+ const MV *center_mv, MV *best_mv) {
// First scale has 8-closest points, the rest have 6 points in hex shape
// at increasing scales
- static const int hex_num_candidates[MAX_PATTERN_SCALES] = {
- 8, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6
- };
+ static const int hex_num_candidates[MAX_PATTERN_SCALES] = { 8, 6, 6, 6, 6, 6,
+ 6, 6, 6, 6, 6 };
// Note that the largest candidate step at each scale is 2^scale
static const MV hex_candidates[MAX_PATTERN_SCALES][MAX_PATTERN_CANDIDATES] = {
- {{-1, -1}, {0, -1}, {1, -1}, {1, 0}, {1, 1}, { 0, 1}, { -1, 1}, {-1, 0}},
- {{-1, -2}, {1, -2}, {2, 0}, {1, 2}, { -1, 2}, { -2, 0}},
- {{-2, -4}, {2, -4}, {4, 0}, {2, 4}, { -2, 4}, { -4, 0}},
- {{-4, -8}, {4, -8}, {8, 0}, {4, 8}, { -4, 8}, { -8, 0}},
- {{-8, -16}, {8, -16}, {16, 0}, {8, 16}, { -8, 16}, { -16, 0}},
- {{-16, -32}, {16, -32}, {32, 0}, {16, 32}, { -16, 32}, { -32, 0}},
- {{-32, -64}, {32, -64}, {64, 0}, {32, 64}, { -32, 64}, { -64, 0}},
- {{-64, -128}, {64, -128}, {128, 0}, {64, 128}, { -64, 128}, { -128, 0}},
- {{-128, -256}, {128, -256}, {256, 0}, {128, 256}, { -128, 256}, { -256, 0}},
- {{-256, -512}, {256, -512}, {512, 0}, {256, 512}, { -256, 512}, { -512, 0}},
- {{-512, -1024}, {512, -1024}, {1024, 0}, {512, 1024}, { -512, 1024},
- { -1024, 0}},
+ { { -1, -1 },
+ { 0, -1 },
+ { 1, -1 },
+ { 1, 0 },
+ { 1, 1 },
+ { 0, 1 },
+ { -1, 1 },
+ { -1, 0 } },
+ { { -1, -2 }, { 1, -2 }, { 2, 0 }, { 1, 2 }, { -1, 2 }, { -2, 0 } },
+ { { -2, -4 }, { 2, -4 }, { 4, 0 }, { 2, 4 }, { -2, 4 }, { -4, 0 } },
+ { { -4, -8 }, { 4, -8 }, { 8, 0 }, { 4, 8 }, { -4, 8 }, { -8, 0 } },
+ { { -8, -16 }, { 8, -16 }, { 16, 0 }, { 8, 16 }, { -8, 16 }, { -16, 0 } },
+ { { -16, -32 },
+ { 16, -32 },
+ { 32, 0 },
+ { 16, 32 },
+ { -16, 32 },
+ { -32, 0 } },
+ { { -32, -64 },
+ { 32, -64 },
+ { 64, 0 },
+ { 32, 64 },
+ { -32, 64 },
+ { -64, 0 } },
+ { { -64, -128 },
+ { 64, -128 },
+ { 128, 0 },
+ { 64, 128 },
+ { -64, 128 },
+ { -128, 0 } },
+ { { -128, -256 },
+ { 128, -256 },
+ { 256, 0 },
+ { 128, 256 },
+ { -128, 256 },
+ { -256, 0 } },
+ { { -256, -512 },
+ { 256, -512 },
+ { 512, 0 },
+ { 256, 512 },
+ { -256, 512 },
+ { -512, 0 } },
+ { { -512, -1024 },
+ { 512, -1024 },
+ { 1024, 0 },
+ { 512, 1024 },
+ { -512, 1024 },
+ { -1024, 0 } },
};
- return vp10_pattern_search(x, ref_mv, search_param, sad_per_bit,
- do_init_search, cost_list, vfp, use_mvcost,
- center_mv, best_mv,
- hex_num_candidates, hex_candidates);
+ return vp10_pattern_search(
+ x, ref_mv, search_param, sad_per_bit, do_init_search, cost_list, vfp,
+ use_mvcost, center_mv, best_mv, hex_num_candidates, hex_candidates);
}
-int vp10_bigdia_search(const MACROBLOCK *x,
- MV *ref_mv,
- int search_param,
- int sad_per_bit,
- int do_init_search,
- int *cost_list,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost,
- const MV *center_mv,
- MV *best_mv) {
+int vp10_bigdia_search(const MACROBLOCK *x, MV *ref_mv, int search_param,
+ int sad_per_bit, int do_init_search, int *cost_list,
+ const vpx_variance_fn_ptr_t *vfp, int use_mvcost,
+ const MV *center_mv, MV *best_mv) {
// First scale has 4-closest points, the rest have 8 points in diamond
// shape at increasing scales
static const int bigdia_num_candidates[MAX_PATTERN_SCALES] = {
4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
};
// Note that the largest candidate step at each scale is 2^scale
- static const MV bigdia_candidates[MAX_PATTERN_SCALES]
- [MAX_PATTERN_CANDIDATES] = {
- {{0, -1}, {1, 0}, { 0, 1}, {-1, 0}},
- {{-1, -1}, {0, -2}, {1, -1}, {2, 0}, {1, 1}, {0, 2}, {-1, 1}, {-2, 0}},
- {{-2, -2}, {0, -4}, {2, -2}, {4, 0}, {2, 2}, {0, 4}, {-2, 2}, {-4, 0}},
- {{-4, -4}, {0, -8}, {4, -4}, {8, 0}, {4, 4}, {0, 8}, {-4, 4}, {-8, 0}},
- {{-8, -8}, {0, -16}, {8, -8}, {16, 0}, {8, 8}, {0, 16}, {-8, 8}, {-16, 0}},
- {{-16, -16}, {0, -32}, {16, -16}, {32, 0}, {16, 16}, {0, 32},
- {-16, 16}, {-32, 0}},
- {{-32, -32}, {0, -64}, {32, -32}, {64, 0}, {32, 32}, {0, 64},
- {-32, 32}, {-64, 0}},
- {{-64, -64}, {0, -128}, {64, -64}, {128, 0}, {64, 64}, {0, 128},
- {-64, 64}, {-128, 0}},
- {{-128, -128}, {0, -256}, {128, -128}, {256, 0}, {128, 128}, {0, 256},
- {-128, 128}, {-256, 0}},
- {{-256, -256}, {0, -512}, {256, -256}, {512, 0}, {256, 256}, {0, 512},
- {-256, 256}, {-512, 0}},
- {{-512, -512}, {0, -1024}, {512, -512}, {1024, 0}, {512, 512}, {0, 1024},
- {-512, 512}, {-1024, 0}},
- };
- return vp10_pattern_search_sad(x, ref_mv, search_param, sad_per_bit,
- do_init_search, cost_list, vfp, use_mvcost,
- center_mv, best_mv,
- bigdia_num_candidates, bigdia_candidates);
+ static const MV
+ bigdia_candidates[MAX_PATTERN_SCALES][MAX_PATTERN_CANDIDATES] = {
+ { { 0, -1 }, { 1, 0 }, { 0, 1 }, { -1, 0 } },
+ { { -1, -1 },
+ { 0, -2 },
+ { 1, -1 },
+ { 2, 0 },
+ { 1, 1 },
+ { 0, 2 },
+ { -1, 1 },
+ { -2, 0 } },
+ { { -2, -2 },
+ { 0, -4 },
+ { 2, -2 },
+ { 4, 0 },
+ { 2, 2 },
+ { 0, 4 },
+ { -2, 2 },
+ { -4, 0 } },
+ { { -4, -4 },
+ { 0, -8 },
+ { 4, -4 },
+ { 8, 0 },
+ { 4, 4 },
+ { 0, 8 },
+ { -4, 4 },
+ { -8, 0 } },
+ { { -8, -8 },
+ { 0, -16 },
+ { 8, -8 },
+ { 16, 0 },
+ { 8, 8 },
+ { 0, 16 },
+ { -8, 8 },
+ { -16, 0 } },
+ { { -16, -16 },
+ { 0, -32 },
+ { 16, -16 },
+ { 32, 0 },
+ { 16, 16 },
+ { 0, 32 },
+ { -16, 16 },
+ { -32, 0 } },
+ { { -32, -32 },
+ { 0, -64 },
+ { 32, -32 },
+ { 64, 0 },
+ { 32, 32 },
+ { 0, 64 },
+ { -32, 32 },
+ { -64, 0 } },
+ { { -64, -64 },
+ { 0, -128 },
+ { 64, -64 },
+ { 128, 0 },
+ { 64, 64 },
+ { 0, 128 },
+ { -64, 64 },
+ { -128, 0 } },
+ { { -128, -128 },
+ { 0, -256 },
+ { 128, -128 },
+ { 256, 0 },
+ { 128, 128 },
+ { 0, 256 },
+ { -128, 128 },
+ { -256, 0 } },
+ { { -256, -256 },
+ { 0, -512 },
+ { 256, -256 },
+ { 512, 0 },
+ { 256, 256 },
+ { 0, 512 },
+ { -256, 256 },
+ { -512, 0 } },
+ { { -512, -512 },
+ { 0, -1024 },
+ { 512, -512 },
+ { 1024, 0 },
+ { 512, 512 },
+ { 0, 1024 },
+ { -512, 512 },
+ { -1024, 0 } },
+ };
+ return vp10_pattern_search_sad(
+ x, ref_mv, search_param, sad_per_bit, do_init_search, cost_list, vfp,
+ use_mvcost, center_mv, best_mv, bigdia_num_candidates, bigdia_candidates);
}
-int vp10_square_search(const MACROBLOCK *x,
- MV *ref_mv,
- int search_param,
- int sad_per_bit,
- int do_init_search,
- int *cost_list,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost,
- const MV *center_mv,
- MV *best_mv) {
+int vp10_square_search(const MACROBLOCK *x, MV *ref_mv, int search_param,
+ int sad_per_bit, int do_init_search, int *cost_list,
+ const vpx_variance_fn_ptr_t *vfp, int use_mvcost,
+ const MV *center_mv, MV *best_mv) {
// All scales have 8 closest points in square shape
static const int square_num_candidates[MAX_PATTERN_SCALES] = {
8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
};
// Note that the largest candidate step at each scale is 2^scale
- static const MV square_candidates[MAX_PATTERN_SCALES]
- [MAX_PATTERN_CANDIDATES] = {
- {{-1, -1}, {0, -1}, {1, -1}, {1, 0}, {1, 1}, {0, 1}, {-1, 1}, {-1, 0}},
- {{-2, -2}, {0, -2}, {2, -2}, {2, 0}, {2, 2}, {0, 2}, {-2, 2}, {-2, 0}},
- {{-4, -4}, {0, -4}, {4, -4}, {4, 0}, {4, 4}, {0, 4}, {-4, 4}, {-4, 0}},
- {{-8, -8}, {0, -8}, {8, -8}, {8, 0}, {8, 8}, {0, 8}, {-8, 8}, {-8, 0}},
- {{-16, -16}, {0, -16}, {16, -16}, {16, 0}, {16, 16}, {0, 16},
- {-16, 16}, {-16, 0}},
- {{-32, -32}, {0, -32}, {32, -32}, {32, 0}, {32, 32}, {0, 32},
- {-32, 32}, {-32, 0}},
- {{-64, -64}, {0, -64}, {64, -64}, {64, 0}, {64, 64}, {0, 64},
- {-64, 64}, {-64, 0}},
- {{-128, -128}, {0, -128}, {128, -128}, {128, 0}, {128, 128}, {0, 128},
- {-128, 128}, {-128, 0}},
- {{-256, -256}, {0, -256}, {256, -256}, {256, 0}, {256, 256}, {0, 256},
- {-256, 256}, {-256, 0}},
- {{-512, -512}, {0, -512}, {512, -512}, {512, 0}, {512, 512}, {0, 512},
- {-512, 512}, {-512, 0}},
- {{-1024, -1024}, {0, -1024}, {1024, -1024}, {1024, 0}, {1024, 1024},
- {0, 1024}, {-1024, 1024}, {-1024, 0}},
- };
- return vp10_pattern_search(x, ref_mv, search_param, sad_per_bit,
- do_init_search, cost_list, vfp, use_mvcost,
- center_mv, best_mv,
- square_num_candidates, square_candidates);
+ static const MV
+ square_candidates[MAX_PATTERN_SCALES][MAX_PATTERN_CANDIDATES] = {
+ { { -1, -1 },
+ { 0, -1 },
+ { 1, -1 },
+ { 1, 0 },
+ { 1, 1 },
+ { 0, 1 },
+ { -1, 1 },
+ { -1, 0 } },
+ { { -2, -2 },
+ { 0, -2 },
+ { 2, -2 },
+ { 2, 0 },
+ { 2, 2 },
+ { 0, 2 },
+ { -2, 2 },
+ { -2, 0 } },
+ { { -4, -4 },
+ { 0, -4 },
+ { 4, -4 },
+ { 4, 0 },
+ { 4, 4 },
+ { 0, 4 },
+ { -4, 4 },
+ { -4, 0 } },
+ { { -8, -8 },
+ { 0, -8 },
+ { 8, -8 },
+ { 8, 0 },
+ { 8, 8 },
+ { 0, 8 },
+ { -8, 8 },
+ { -8, 0 } },
+ { { -16, -16 },
+ { 0, -16 },
+ { 16, -16 },
+ { 16, 0 },
+ { 16, 16 },
+ { 0, 16 },
+ { -16, 16 },
+ { -16, 0 } },
+ { { -32, -32 },
+ { 0, -32 },
+ { 32, -32 },
+ { 32, 0 },
+ { 32, 32 },
+ { 0, 32 },
+ { -32, 32 },
+ { -32, 0 } },
+ { { -64, -64 },
+ { 0, -64 },
+ { 64, -64 },
+ { 64, 0 },
+ { 64, 64 },
+ { 0, 64 },
+ { -64, 64 },
+ { -64, 0 } },
+ { { -128, -128 },
+ { 0, -128 },
+ { 128, -128 },
+ { 128, 0 },
+ { 128, 128 },
+ { 0, 128 },
+ { -128, 128 },
+ { -128, 0 } },
+ { { -256, -256 },
+ { 0, -256 },
+ { 256, -256 },
+ { 256, 0 },
+ { 256, 256 },
+ { 0, 256 },
+ { -256, 256 },
+ { -256, 0 } },
+ { { -512, -512 },
+ { 0, -512 },
+ { 512, -512 },
+ { 512, 0 },
+ { 512, 512 },
+ { 0, 512 },
+ { -512, 512 },
+ { -512, 0 } },
+ { { -1024, -1024 },
+ { 0, -1024 },
+ { 1024, -1024 },
+ { 1024, 0 },
+ { 1024, 1024 },
+ { 0, 1024 },
+ { -1024, 1024 },
+ { -1024, 0 } },
+ };
+ return vp10_pattern_search(
+ x, ref_mv, search_param, sad_per_bit, do_init_search, cost_list, vfp,
+ use_mvcost, center_mv, best_mv, square_num_candidates, square_candidates);
}
-int vp10_fast_hex_search(const MACROBLOCK *x,
- MV *ref_mv,
- int search_param,
- int sad_per_bit,
- int do_init_search, // must be zero for fast_hex
- int *cost_list,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost,
- const MV *center_mv,
- MV *best_mv) {
+int vp10_fast_hex_search(const MACROBLOCK *x, MV *ref_mv, int search_param,
+ int sad_per_bit,
+ int do_init_search, // must be zero for fast_hex
+ int *cost_list, const vpx_variance_fn_ptr_t *vfp,
+ int use_mvcost, const MV *center_mv, MV *best_mv) {
return vp10_hex_search(
x, ref_mv, VPXMAX(MAX_MVSEARCH_STEPS - 2, search_param), sad_per_bit,
do_init_search, cost_list, vfp, use_mvcost, center_mv, best_mv);
}
-int vp10_fast_dia_search(const MACROBLOCK *x,
- MV *ref_mv,
- int search_param,
- int sad_per_bit,
- int do_init_search,
- int *cost_list,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost,
- const MV *center_mv,
- MV *best_mv) {
+int vp10_fast_dia_search(const MACROBLOCK *x, MV *ref_mv, int search_param,
+ int sad_per_bit, int do_init_search, int *cost_list,
+ const vpx_variance_fn_ptr_t *vfp, int use_mvcost,
+ const MV *center_mv, MV *best_mv) {
return vp10_bigdia_search(
x, ref_mv, VPXMAX(MAX_MVSEARCH_STEPS - 2, search_param), sad_per_bit,
do_init_search, cost_list, vfp, use_mvcost, center_mv, best_mv);
@@ -1525,15 +1563,14 @@
// Exhuastive motion search around a given centre position with a given
// step size.
-static int exhuastive_mesh_search(const MACROBLOCK *x,
- MV *ref_mv, MV *best_mv,
+static int exhuastive_mesh_search(const MACROBLOCK *x, MV *ref_mv, MV *best_mv,
int range, int step, int sad_per_bit,
const vpx_variance_fn_ptr_t *fn_ptr,
const MV *center_mv) {
const MACROBLOCKD *const xd = &x->e_mbd;
const struct buf_2d *const what = &x->plane[0].src;
const struct buf_2d *const in_what = &xd->plane[0].pre[0];
- MV fcenter_mv = {center_mv->row, center_mv->col};
+ MV fcenter_mv = { center_mv->row, center_mv->col };
unsigned int best_sad = INT_MAX;
int r, c, i;
int start_col, end_col, start_row, end_row;
@@ -1541,12 +1578,13 @@
assert(step >= 1);
- clamp_mv(&fcenter_mv, x->mv_col_min, x->mv_col_max,
- x->mv_row_min, x->mv_row_max);
+ clamp_mv(&fcenter_mv, x->mv_col_min, x->mv_col_max, x->mv_row_min,
+ x->mv_row_max);
*best_mv = fcenter_mv;
- best_sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &fcenter_mv), in_what->stride) +
- mvsad_err_cost(x, &fcenter_mv, ref_mv, sad_per_bit);
+ best_sad =
+ fn_ptr->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &fcenter_mv), in_what->stride) +
+ mvsad_err_cost(x, &fcenter_mv, ref_mv, sad_per_bit);
start_row = VPXMAX(-range, x->mv_row_min - fcenter_mv.row);
start_col = VPXMAX(-range, x->mv_col_min - fcenter_mv.col);
end_row = VPXMIN(range, x->mv_row_max - fcenter_mv.row);
@@ -1556,9 +1594,10 @@
for (c = start_col; c <= end_col; c += col_step) {
// Step > 1 means we are not checking every location in this pass.
if (step > 1) {
- const MV mv = {fcenter_mv.row + r, fcenter_mv.col + c};
- unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &mv), in_what->stride);
+ const MV mv = { fcenter_mv.row + r, fcenter_mv.col + c };
+ unsigned int sad =
+ fn_ptr->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &mv),
+ in_what->stride);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, ref_mv, sad_per_bit);
if (sad < best_sad) {
@@ -1572,17 +1611,16 @@
unsigned int sads[4];
const uint8_t *addrs[4];
for (i = 0; i < 4; ++i) {
- const MV mv = {fcenter_mv.row + r, fcenter_mv.col + c + i};
+ const MV mv = { fcenter_mv.row + r, fcenter_mv.col + c + i };
addrs[i] = get_buf_from_mv(in_what, &mv);
}
- fn_ptr->sdx4df(what->buf, what->stride, addrs,
- in_what->stride, sads);
+ fn_ptr->sdx4df(what->buf, what->stride, addrs, in_what->stride, sads);
for (i = 0; i < 4; ++i) {
if (sads[i] < best_sad) {
- const MV mv = {fcenter_mv.row + r, fcenter_mv.col + c + i};
- const unsigned int sad = sads[i] +
- mvsad_err_cost(x, &mv, ref_mv, sad_per_bit);
+ const MV mv = { fcenter_mv.row + r, fcenter_mv.col + c + i };
+ const unsigned int sad =
+ sads[i] + mvsad_err_cost(x, &mv, ref_mv, sad_per_bit);
if (sad < best_sad) {
best_sad = sad;
*best_mv = mv;
@@ -1591,9 +1629,10 @@
}
} else {
for (i = 0; i < end_col - c; ++i) {
- const MV mv = {fcenter_mv.row + r, fcenter_mv.col + c + i};
- unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &mv), in_what->stride);
+ const MV mv = { fcenter_mv.row + r, fcenter_mv.col + c + i };
+ unsigned int sad =
+ fn_ptr->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &mv), in_what->stride);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, ref_mv, sad_per_bit);
if (sad < best_sad) {
@@ -1611,11 +1650,10 @@
}
int vp10_diamond_search_sad_c(const MACROBLOCK *x,
- const search_site_config *cfg,
- MV *ref_mv, MV *best_mv, int search_param,
- int sad_per_bit, int *num00,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv) {
+ const search_site_config *cfg, MV *ref_mv,
+ MV *best_mv, int search_param, int sad_per_bit,
+ int *num00, const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv) {
int i, j, step;
const MACROBLOCKD *const xd = &x->e_mbd;
@@ -1640,7 +1678,7 @@
const search_site *ss = &cfg->ss[search_param * cfg->searches_per_step];
const int tot_steps = (cfg->ss_count / cfg->searches_per_step) - search_param;
- const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
+ const MV fcenter_mv = { center_mv->row >> 3, center_mv->col >> 3 };
clamp_mv(ref_mv, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max);
ref_row = ref_mv->row;
ref_col = ref_mv->col;
@@ -1653,8 +1691,8 @@
best_address = in_what;
// Check the starting position
- bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride)
- + mvsad_err_cost(x, best_mv, &fcenter_mv, sad_per_bit);
+ bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride) +
+ mvsad_err_cost(x, best_mv, &fcenter_mv, sad_per_bit);
i = 1;
@@ -1685,10 +1723,10 @@
for (t = 0; t < 4; t++, i++) {
if (sad_array[t] < bestsad) {
- const MV this_mv = {best_mv->row + ss[i].mv.row,
- best_mv->col + ss[i].mv.col};
- sad_array[t] += mvsad_err_cost(x, &this_mv, &fcenter_mv,
- sad_per_bit);
+ const MV this_mv = { best_mv->row + ss[i].mv.row,
+ best_mv->col + ss[i].mv.col };
+ sad_array[t] +=
+ mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
if (sad_array[t] < bestsad) {
bestsad = sad_array[t];
best_site = i;
@@ -1699,13 +1737,13 @@
} else {
for (j = 0; j < cfg->searches_per_step; j++) {
// Trap illegal vectors
- const MV this_mv = {best_mv->row + ss[i].mv.row,
- best_mv->col + ss[i].mv.col};
+ const MV this_mv = { best_mv->row + ss[i].mv.row,
+ best_mv->col + ss[i].mv.col };
if (is_mv_in(x, &this_mv)) {
const uint8_t *const check_here = ss[i].offset + best_address;
- unsigned int thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ unsigned int thissad =
+ fn_ptr->sdf(what, what_stride, check_here, in_what_stride);
if (thissad < bestsad) {
thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
@@ -1725,12 +1763,12 @@
last_site = best_site;
#if defined(NEW_DIAMOND_SEARCH)
while (1) {
- const MV this_mv = {best_mv->row + ss[best_site].mv.row,
- best_mv->col + ss[best_site].mv.col};
+ const MV this_mv = { best_mv->row + ss[best_site].mv.row,
+ best_mv->col + ss[best_site].mv.col };
if (is_mv_in(x, &this_mv)) {
const uint8_t *const check_here = ss[best_site].offset + best_address;
- unsigned int thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ unsigned int thissad =
+ fn_ptr->sdf(what, what_stride, check_here, in_what_stride);
if (thissad < bestsad) {
thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
if (thissad < bestsad) {
@@ -1770,8 +1808,7 @@
for (d = -8; d <= 8; d += 16) {
int this_pos = offset + d;
// check limit
- if (this_pos < 0 || this_pos > bw)
- continue;
+ if (this_pos < 0 || this_pos > bw) continue;
this_sad = vpx_vector_var(&ref[this_pos], src, bwl);
if (this_sad < best_sad) {
best_sad = this_sad;
@@ -1783,8 +1820,7 @@
for (d = -4; d <= 4; d += 8) {
int this_pos = offset + d;
// check limit
- if (this_pos < 0 || this_pos > bw)
- continue;
+ if (this_pos < 0 || this_pos > bw) continue;
this_sad = vpx_vector_var(&ref[this_pos], src, bwl);
if (this_sad < best_sad) {
best_sad = this_sad;
@@ -1796,8 +1832,7 @@
for (d = -2; d <= 2; d += 4) {
int this_pos = offset + d;
// check limit
- if (this_pos < 0 || this_pos > bw)
- continue;
+ if (this_pos < 0 || this_pos > bw) continue;
this_sad = vpx_vector_var(&ref[this_pos], src, bwl);
if (this_sad < best_sad) {
best_sad = this_sad;
@@ -1809,8 +1844,7 @@
for (d = -1; d <= 1; d += 2) {
int this_pos = offset + d;
// check limit
- if (this_pos < 0 || this_pos > bw)
- continue;
+ if (this_pos < 0 || this_pos > bw) continue;
this_sad = vpx_vector_var(&ref[this_pos], src, bwl);
if (this_sad < best_sad) {
best_sad = this_sad;
@@ -1822,15 +1856,15 @@
}
static const MV search_pos[4] = {
- {-1, 0}, {0, -1}, {0, 1}, {1, 0},
+ { -1, 0 }, { 0, -1 }, { 0, 1 }, { 1, 0 },
};
unsigned int vp10_int_pro_motion_estimation(const VP10_COMP *cpi, MACROBLOCK *x,
- BLOCK_SIZE bsize,
- int mi_row, int mi_col) {
+ BLOCK_SIZE bsize, int mi_row,
+ int mi_col) {
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
- struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0, 0}};
+ struct buf_2d backup_yv12[MAX_MB_PLANE] = { { 0, 0 } };
DECLARE_ALIGNED(16, int16_t, hbuf[128]);
DECLARE_ALIGNED(16, int16_t, vbuf[128]);
DECLARE_ALIGNED(16, int16_t, src_hbuf[64]);
@@ -1855,8 +1889,7 @@
// Swap out the reference frame for a version that's been scaled to
// match the resolution of the current frame, allowing the existing
// motion search code to be used without additional modifications.
- for (i = 0; i < MAX_MB_PLANE; i++)
- backup_yv12[i] = xd->plane[i].pre[0];
+ for (i = 0; i < MAX_MB_PLANE; i++) backup_yv12[i] = xd->plane[i].pre[0];
vp10_setup_pre_planes(xd, 0, scaled_ref_frame, mi_row, mi_col, NULL);
}
@@ -1870,8 +1903,7 @@
if (scaled_ref_frame) {
int i;
- for (i = 0; i < MAX_MB_PLANE; i++)
- xd->plane[i].pre[0] = backup_yv12[i];
+ for (i = 0; i < MAX_MB_PLANE; i++) xd->plane[i].pre[0] = backup_yv12[i];
}
return this_sad;
}
@@ -1912,11 +1944,8 @@
best_sad = cpi->fn_ptr[bsize].sdf(src_buf, src_stride, ref_buf, ref_stride);
{
- const uint8_t * const pos[4] = {
- ref_buf - ref_stride,
- ref_buf - 1,
- ref_buf + 1,
- ref_buf + ref_stride,
+ const uint8_t *const pos[4] = {
+ ref_buf - ref_stride, ref_buf - 1, ref_buf + 1, ref_buf + ref_stride,
};
cpi->fn_ptr[bsize].sdx4df(src_buf, src_stride, pos, ref_stride, this_sad);
@@ -1942,8 +1971,7 @@
ref_buf = xd->plane[0].pre[0].buf + this_mv.row * ref_stride + this_mv.col;
- tmp_sad = cpi->fn_ptr[bsize].sdf(src_buf, src_stride,
- ref_buf, ref_stride);
+ tmp_sad = cpi->fn_ptr[bsize].sdf(src_buf, src_stride, ref_buf, ref_stride);
if (best_sad > tmp_sad) {
*tmp_mv = this_mv;
best_sad = tmp_sad;
@@ -1954,8 +1982,7 @@
if (scaled_ref_frame) {
int i;
- for (i = 0; i < MAX_MB_PLANE; i++)
- xd->plane[i].pre[0] = backup_yv12[i];
+ for (i = 0; i < MAX_MB_PLANE; i++) xd->plane[i].pre[0] = backup_yv12[i];
}
return best_sad;
@@ -1964,25 +1991,22 @@
/* do_refine: If last step (1-away) of n-step search doesn't pick the center
point as the best match, we will do a final 1-away diamond
refining search */
-int vp10_full_pixel_diamond(const VP10_COMP *cpi, MACROBLOCK *x,
- MV *mvp_full, int step_param,
- int sadpb, int further_steps, int do_refine,
- int *cost_list,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *ref_mv, MV *dst_mv) {
+int vp10_full_pixel_diamond(const VP10_COMP *cpi, MACROBLOCK *x, MV *mvp_full,
+ int step_param, int sadpb, int further_steps,
+ int do_refine, int *cost_list,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *ref_mv, MV *dst_mv) {
MV temp_mv;
int thissme, n, num00 = 0;
int bestsme = cpi->diamond_search_sad(x, &cpi->ss_cfg, mvp_full, &temp_mv,
- step_param, sadpb, &n,
- fn_ptr, ref_mv);
+ step_param, sadpb, &n, fn_ptr, ref_mv);
if (bestsme < INT_MAX)
bestsme = vp10_get_mvpred_var(x, &temp_mv, ref_mv, fn_ptr, 1);
*dst_mv = temp_mv;
// If there won't be more n-step search, check to see if refining search is
// needed.
- if (n > further_steps)
- do_refine = 0;
+ if (n > further_steps) do_refine = 0;
while (n < further_steps) {
++n;
@@ -1991,14 +2015,13 @@
num00--;
} else {
thissme = cpi->diamond_search_sad(x, &cpi->ss_cfg, mvp_full, &temp_mv,
- step_param + n, sadpb, &num00,
- fn_ptr, ref_mv);
+ step_param + n, sadpb, &num00, fn_ptr,
+ ref_mv);
if (thissme < INT_MAX)
thissme = vp10_get_mvpred_var(x, &temp_mv, ref_mv, fn_ptr, 1);
// check to see if refining search is needed.
- if (num00 > further_steps - n)
- do_refine = 0;
+ if (num00 > further_steps - n) do_refine = 0;
if (thissme < bestsme) {
bestsme = thissme;
@@ -2011,8 +2034,8 @@
if (do_refine) {
const int search_range = 8;
MV best_mv = *dst_mv;
- thissme = vp10_refining_search_sad(x, &best_mv, sadpb, search_range,
- fn_ptr, ref_mv);
+ thissme = vp10_refining_search_sad(x, &best_mv, sadpb, search_range, fn_ptr,
+ ref_mv);
if (thissme < INT_MAX)
thissme = vp10_get_mvpred_var(x, &best_mv, ref_mv, fn_ptr, 1);
if (thissme < bestsme) {
@@ -2034,12 +2057,12 @@
// Runs an limited range exhaustive mesh search using a pattern set
// according to the encode speed profile.
static int full_pixel_exhaustive(VP10_COMP *cpi, MACROBLOCK *x,
- MV *centre_mv_full, int sadpb, int *cost_list,
+ MV *centre_mv_full, int sadpb, int *cost_list,
const vpx_variance_fn_ptr_t *fn_ptr,
const MV *ref_mv, MV *dst_mv) {
const SPEED_FEATURES *const sf = &cpi->sf;
- MV temp_mv = {centre_mv_full->row, centre_mv_full->col};
- MV f_ref_mv = {ref_mv->row >> 3, ref_mv->col >> 3};
+ MV temp_mv = { centre_mv_full->row, centre_mv_full->col };
+ MV f_ref_mv = { ref_mv->row >> 3, ref_mv->col >> 3 };
int bestsme;
int i;
int interval = sf->mesh_patterns[0].interval;
@@ -2050,8 +2073,8 @@
++(*x->ex_search_count_ptr);
// Trap illegal values for interval and range for this function.
- if ((range < MIN_RANGE) || (range > MAX_RANGE) ||
- (interval < MIN_INTERVAL) || (interval > range))
+ if ((range < MIN_RANGE) || (range > MAX_RANGE) || (interval < MIN_INTERVAL) ||
+ (interval > range))
return INT_MAX;
baseline_interval_divisor = range / interval;
@@ -2063,21 +2086,19 @@
interval = VPXMAX(interval, range / baseline_interval_divisor);
// initial search
- bestsme = exhuastive_mesh_search(x, &f_ref_mv, &temp_mv, range,
- interval, sadpb, fn_ptr, &temp_mv);
+ bestsme = exhuastive_mesh_search(x, &f_ref_mv, &temp_mv, range, interval,
+ sadpb, fn_ptr, &temp_mv);
if ((interval > MIN_INTERVAL) && (range > MIN_RANGE)) {
// Progressive searches with range and step size decreasing each time
// till we reach a step size of 1. Then break out.
for (i = 1; i < MAX_MESH_STEP; ++i) {
// First pass with coarser step and longer range
- bestsme = exhuastive_mesh_search(x, &f_ref_mv, &temp_mv,
- sf->mesh_patterns[i].range,
- sf->mesh_patterns[i].interval,
- sadpb, fn_ptr, &temp_mv);
+ bestsme = exhuastive_mesh_search(
+ x, &f_ref_mv, &temp_mv, sf->mesh_patterns[i].range,
+ sf->mesh_patterns[i].interval, sadpb, fn_ptr, &temp_mv);
- if (sf->mesh_patterns[i].interval == 1)
- break;
+ if (sf->mesh_patterns[i].interval == 1) break;
}
}
@@ -2093,9 +2114,9 @@
}
int vp10_full_search_sad_c(const MACROBLOCK *x, const MV *ref_mv,
- int sad_per_bit, int distance,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv, MV *best_mv) {
+ int sad_per_bit, int distance,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv, MV *best_mv) {
int r, c;
const MACROBLOCKD *const xd = &x->e_mbd;
const struct buf_2d *const what = &x->plane[0].src;
@@ -2104,18 +2125,20 @@
const int row_max = VPXMIN(ref_mv->row + distance, x->mv_row_max);
const int col_min = VPXMAX(ref_mv->col - distance, x->mv_col_min);
const int col_max = VPXMIN(ref_mv->col + distance, x->mv_col_max);
- const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
- int best_sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, ref_mv), in_what->stride) +
+ const MV fcenter_mv = { center_mv->row >> 3, center_mv->col >> 3 };
+ int best_sad =
+ fn_ptr->sdf(what->buf, what->stride, get_buf_from_mv(in_what, ref_mv),
+ in_what->stride) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
*best_mv = *ref_mv;
for (r = row_min; r < row_max; ++r) {
for (c = col_min; c < col_max; ++c) {
- const MV mv = {r, c};
- const int sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &mv), in_what->stride) +
- mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
+ const MV mv = { r, c };
+ const int sad =
+ fn_ptr->sdf(what->buf, what->stride, get_buf_from_mv(in_what, &mv),
+ in_what->stride) +
+ mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
best_sad = sad;
*best_mv = mv;
@@ -2126,9 +2149,9 @@
}
int vp10_full_search_sadx3(const MACROBLOCK *x, const MV *ref_mv,
- int sad_per_bit, int distance,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv, MV *best_mv) {
+ int sad_per_bit, int distance,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv, MV *best_mv) {
int r;
const MACROBLOCKD *const xd = &x->e_mbd;
const struct buf_2d *const what = &x->plane[0].src;
@@ -2137,9 +2160,10 @@
const int row_max = VPXMIN(ref_mv->row + distance, x->mv_row_max);
const int col_min = VPXMAX(ref_mv->col - distance, x->mv_col_min);
const int col_max = VPXMIN(ref_mv->col + distance, x->mv_col_max);
- const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
- unsigned int best_sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, ref_mv), in_what->stride) +
+ const MV fcenter_mv = { center_mv->row >> 3, center_mv->col >> 3 };
+ unsigned int best_sad =
+ fn_ptr->sdf(what->buf, what->stride, get_buf_from_mv(in_what, ref_mv),
+ in_what->stride) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
*best_mv = *ref_mv;
@@ -2158,7 +2182,7 @@
for (i = 0; i < 3; ++i) {
unsigned int sad = sads[i];
if (sad < best_sad) {
- const MV mv = {r, c};
+ const MV mv = { r, c };
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
best_sad = sad;
@@ -2172,10 +2196,10 @@
}
while (c < col_max) {
- unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
- check_here, in_what->stride);
+ unsigned int sad =
+ fn_ptr->sdf(what->buf, what->stride, check_here, in_what->stride);
if (sad < best_sad) {
- const MV mv = {r, c};
+ const MV mv = { r, c };
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
best_sad = sad;
@@ -2191,9 +2215,9 @@
}
int vp10_full_search_sadx8(const MACROBLOCK *x, const MV *ref_mv,
- int sad_per_bit, int distance,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv, MV *best_mv) {
+ int sad_per_bit, int distance,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv, MV *best_mv) {
int r;
const MACROBLOCKD *const xd = &x->e_mbd;
const struct buf_2d *const what = &x->plane[0].src;
@@ -2202,9 +2226,10 @@
const int row_max = VPXMIN(ref_mv->row + distance, x->mv_row_max);
const int col_min = VPXMAX(ref_mv->col - distance, x->mv_col_min);
const int col_max = VPXMIN(ref_mv->col + distance, x->mv_col_max);
- const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
- unsigned int best_sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, ref_mv), in_what->stride) +
+ const MV fcenter_mv = { center_mv->row >> 3, center_mv->col >> 3 };
+ unsigned int best_sad =
+ fn_ptr->sdf(what->buf, what->stride, get_buf_from_mv(in_what, ref_mv),
+ in_what->stride) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
*best_mv = *ref_mv;
@@ -2223,7 +2248,7 @@
for (i = 0; i < 8; ++i) {
unsigned int sad = sads[i];
if (sad < best_sad) {
- const MV mv = {r, c};
+ const MV mv = { r, c };
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
best_sad = sad;
@@ -2247,7 +2272,7 @@
for (i = 0; i < 3; ++i) {
unsigned int sad = sads[i];
if (sad < best_sad) {
- const MV mv = {r, c};
+ const MV mv = { r, c };
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
best_sad = sad;
@@ -2261,10 +2286,10 @@
}
while (c < col_max) {
- unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
- check_here, in_what->stride);
+ unsigned int sad =
+ fn_ptr->sdf(what->buf, what->stride, check_here, in_what->stride);
if (sad < best_sad) {
- const MV mv = {r, c};
+ const MV mv = { r, c };
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
best_sad = sad;
@@ -2279,19 +2304,18 @@
return best_sad;
}
-int vp10_refining_search_sad(const MACROBLOCK *x,
- MV *ref_mv, int error_per_bit,
- int search_range,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv) {
+int vp10_refining_search_sad(const MACROBLOCK *x, MV *ref_mv, int error_per_bit,
+ int search_range,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv) {
const MACROBLOCKD *const xd = &x->e_mbd;
- const MV neighbors[4] = {{ -1, 0}, {0, -1}, {0, 1}, {1, 0}};
+ const MV neighbors[4] = { { -1, 0 }, { 0, -1 }, { 0, 1 }, { 1, 0 } };
const struct buf_2d *const what = &x->plane[0].src;
const struct buf_2d *const in_what = &xd->plane[0].pre[0];
- const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
+ const MV fcenter_mv = { center_mv->row >> 3, center_mv->col >> 3 };
const uint8_t *best_address = get_buf_from_mv(in_what, ref_mv);
- unsigned int best_sad = fn_ptr->sdf(what->buf, what->stride, best_address,
- in_what->stride) +
+ unsigned int best_sad =
+ fn_ptr->sdf(what->buf, what->stride, best_address, in_what->stride) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, error_per_bit);
int i, j;
@@ -2304,19 +2328,16 @@
if (all_in) {
unsigned int sads[4];
- const uint8_t *const positions[4] = {
- best_address - in_what->stride,
- best_address - 1,
- best_address + 1,
- best_address + in_what->stride
- };
+ const uint8_t *const positions[4] = { best_address - in_what->stride,
+ best_address - 1, best_address + 1,
+ best_address + in_what->stride };
fn_ptr->sdx4df(what->buf, what->stride, positions, in_what->stride, sads);
for (j = 0; j < 4; ++j) {
if (sads[j] < best_sad) {
- const MV mv = {ref_mv->row + neighbors[j].row,
- ref_mv->col + neighbors[j].col};
+ const MV mv = { ref_mv->row + neighbors[j].row,
+ ref_mv->col + neighbors[j].col };
sads[j] += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit);
if (sads[j] < best_sad) {
best_sad = sads[j];
@@ -2326,13 +2347,13 @@
}
} else {
for (j = 0; j < 4; ++j) {
- const MV mv = {ref_mv->row + neighbors[j].row,
- ref_mv->col + neighbors[j].col};
+ const MV mv = { ref_mv->row + neighbors[j].row,
+ ref_mv->col + neighbors[j].col };
if (is_mv_in(x, &mv)) {
- unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &mv),
- in_what->stride);
+ unsigned int sad =
+ fn_ptr->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &mv), in_what->stride);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit);
if (sad < best_sad) {
@@ -2358,20 +2379,19 @@
// This function is called when we do joint motion search in comp_inter_inter
// mode.
-int vp10_refining_search_8p_c(const MACROBLOCK *x,
- MV *ref_mv, int error_per_bit,
- int search_range,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv,
- const uint8_t *second_pred) {
- const MV neighbors[8] = {{-1, 0}, {0, -1}, {0, 1}, {1, 0},
- {-1, -1}, {1, -1}, {-1, 1}, {1, 1}};
+int vp10_refining_search_8p_c(const MACROBLOCK *x, MV *ref_mv,
+ int error_per_bit, int search_range,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv, const uint8_t *second_pred) {
+ const MV neighbors[8] = { { -1, 0 }, { 0, -1 }, { 0, 1 }, { 1, 0 },
+ { -1, -1 }, { 1, -1 }, { -1, 1 }, { 1, 1 } };
const MACROBLOCKD *const xd = &x->e_mbd;
const struct buf_2d *const what = &x->plane[0].src;
const struct buf_2d *const in_what = &xd->plane[0].pre[0];
- const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
- unsigned int best_sad = fn_ptr->sdaf(what->buf, what->stride,
- get_buf_from_mv(in_what, ref_mv), in_what->stride, second_pred) +
+ const MV fcenter_mv = { center_mv->row >> 3, center_mv->col >> 3 };
+ unsigned int best_sad =
+ fn_ptr->sdaf(what->buf, what->stride, get_buf_from_mv(in_what, ref_mv),
+ in_what->stride, second_pred) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, error_per_bit);
int i, j;
@@ -2379,12 +2399,13 @@
int best_site = -1;
for (j = 0; j < 8; ++j) {
- const MV mv = {ref_mv->row + neighbors[j].row,
- ref_mv->col + neighbors[j].col};
+ const MV mv = { ref_mv->row + neighbors[j].row,
+ ref_mv->col + neighbors[j].col };
if (is_mv_in(x, &mv)) {
- unsigned int sad = fn_ptr->sdaf(what->buf, what->stride,
- get_buf_from_mv(in_what, &mv), in_what->stride, second_pred);
+ unsigned int sad =
+ fn_ptr->sdaf(what->buf, what->stride, get_buf_from_mv(in_what, &mv),
+ in_what->stride, second_pred);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit);
if (sad < best_sad) {
@@ -2408,21 +2429,19 @@
#define MIN_EX_SEARCH_LIMIT 128
static int is_exhaustive_allowed(VP10_COMP *cpi, MACROBLOCK *x) {
const SPEED_FEATURES *const sf = &cpi->sf;
- const int max_ex = VPXMAX(MIN_EX_SEARCH_LIMIT,
- (*x->m_search_count_ptr * sf->max_exaustive_pct) / 100);
+ const int max_ex =
+ VPXMAX(MIN_EX_SEARCH_LIMIT,
+ (*x->m_search_count_ptr * sf->max_exaustive_pct) / 100);
return sf->allow_exhaustive_searches &&
- (sf->exhaustive_searches_thresh < INT_MAX) &&
- (*x->ex_search_count_ptr <= max_ex) &&
- !cpi->rc.is_src_frame_alt_ref;
+ (sf->exhaustive_searches_thresh < INT_MAX) &&
+ (*x->ex_search_count_ptr <= max_ex) && !cpi->rc.is_src_frame_alt_ref;
}
-int vp10_full_pixel_search(VP10_COMP *cpi, MACROBLOCK *x,
- BLOCK_SIZE bsize, MV *mvp_full,
- int step_param, int error_per_bit,
- int *cost_list,
- const MV *ref_mv, MV *tmp_mv,
- int var_max, int rd) {
+int vp10_full_pixel_search(VP10_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bsize,
+ MV *mvp_full, int step_param, int error_per_bit,
+ int *cost_list, const MV *ref_mv, MV *tmp_mv,
+ int var_max, int rd) {
const SPEED_FEATURES *const sf = &cpi->sf;
const SEARCH_METHODS method = sf->mv.search_method;
vpx_variance_fn_ptr_t *fn_ptr = &cpi->fn_ptr[bsize];
@@ -2441,42 +2460,41 @@
switch (method) {
case FAST_DIAMOND:
var = vp10_fast_dia_search(x, mvp_full, step_param, error_per_bit, 0,
- cost_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case FAST_HEX:
var = vp10_fast_hex_search(x, mvp_full, step_param, error_per_bit, 0,
- cost_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case HEX:
var = vp10_hex_search(x, mvp_full, step_param, error_per_bit, 1,
- cost_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case SQUARE:
var = vp10_square_search(x, mvp_full, step_param, error_per_bit, 1,
- cost_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case BIGDIA:
var = vp10_bigdia_search(x, mvp_full, step_param, error_per_bit, 1,
- cost_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case NSTEP:
var = vp10_full_pixel_diamond(cpi, x, mvp_full, step_param, error_per_bit,
- MAX_MVSEARCH_STEPS - 1 - step_param,
- 1, cost_list, fn_ptr, ref_mv, tmp_mv);
+ MAX_MVSEARCH_STEPS - 1 - step_param, 1,
+ cost_list, fn_ptr, ref_mv, tmp_mv);
// Should we allow a follow on exhaustive search?
if (is_exhaustive_allowed(cpi, x)) {
int64_t exhuastive_thr = sf->exhaustive_searches_thresh;
- exhuastive_thr >>= 8 - (b_width_log2_lookup[bsize] +
- b_height_log2_lookup[bsize]);
+ exhuastive_thr >>=
+ 8 - (b_width_log2_lookup[bsize] + b_height_log2_lookup[bsize]);
// Threshold variance for an exhaustive full search.
if (var > exhuastive_thr) {
- int var_ex;
+ int var_ex;
MV tmp_mv_ex;
- var_ex = full_pixel_exhaustive(cpi, x, tmp_mv,
- error_per_bit, cost_list, fn_ptr,
- ref_mv, &tmp_mv_ex);
+ var_ex = full_pixel_exhaustive(cpi, x, tmp_mv, error_per_bit,
+ cost_list, fn_ptr, ref_mv, &tmp_mv_ex);
if (var_ex < var) {
var = var_ex;
@@ -2487,8 +2505,7 @@
break;
break;
- default:
- assert(0 && "Invalid search method.");
+ default: assert(0 && "Invalid search method.");
}
if (method != NSTEP && rd && var < var_max)
diff --git a/vp10/encoder/mcomp.h b/vp10/encoder/mcomp.h
index 7597981..e17264b 100644
--- a/vp10/encoder/mcomp.h
+++ b/vp10/encoder/mcomp.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_MCOMP_H_
#define VP10_ENCODER_MCOMP_H_
@@ -26,7 +25,7 @@
// Enable the use of motion vector in range [-1023, 1023].
#define MAX_FULL_PEL_VAL ((1 << (MAX_MVSEARCH_STEPS - 1)) - 1)
// Maximum size of the first step in full pel units
-#define MAX_FIRST_STEP (1 << (MAX_MVSEARCH_STEPS-1))
+#define MAX_FIRST_STEP (1 << (MAX_MVSEARCH_STEPS - 1))
// Allowed motion vector pixel distance outside image border
// for Block_16x16
#define BORDER_MV_PIXELS_B16 (16 + VPX_INTERP_EXTEND)
@@ -44,59 +43,48 @@
} search_site_config;
void vp10_init_dsmotion_compensation(search_site_config *cfg, int stride);
-void vp10_init3smotion_compensation(search_site_config *cfg, int stride);
+void vp10_init3smotion_compensation(search_site_config *cfg, int stride);
void vp10_set_mv_search_range(MACROBLOCK *x, const MV *mv);
-int vp10_mv_bit_cost(const MV *mv, const MV *ref,
- const int *mvjcost, int *mvcost[2], int weight);
+int vp10_mv_bit_cost(const MV *mv, const MV *ref, const int *mvjcost,
+ int *mvcost[2], int weight);
// Utility to compute variance + MV rate cost for a given MV
-int vp10_get_mvpred_var(const MACROBLOCK *x,
- const MV *best_mv, const MV *center_mv,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost);
-int vp10_get_mvpred_av_var(const MACROBLOCK *x,
- const MV *best_mv, const MV *center_mv,
- const uint8_t *second_pred,
- const vpx_variance_fn_ptr_t *vfp,
- int use_mvcost);
+int vp10_get_mvpred_var(const MACROBLOCK *x, const MV *best_mv,
+ const MV *center_mv, const vpx_variance_fn_ptr_t *vfp,
+ int use_mvcost);
+int vp10_get_mvpred_av_var(const MACROBLOCK *x, const MV *best_mv,
+ const MV *center_mv, const uint8_t *second_pred,
+ const vpx_variance_fn_ptr_t *vfp, int use_mvcost);
struct VP10_COMP;
struct SPEED_FEATURES;
int vp10_init_search_range(int size);
-int vp10_refining_search_sad(const struct macroblock *x,
- struct mv *ref_mv,
- int sad_per_bit, int distance,
- const struct vpx_variance_vtable *fn_ptr,
- const struct mv *center_mv);
+int vp10_refining_search_sad(const struct macroblock *x, struct mv *ref_mv,
+ int sad_per_bit, int distance,
+ const struct vpx_variance_vtable *fn_ptr,
+ const struct mv *center_mv);
// Runs sequence of diamond searches in smaller steps for RD.
int vp10_full_pixel_diamond(const struct VP10_COMP *cpi, MACROBLOCK *x,
- MV *mvp_full, int step_param,
- int sadpb, int further_steps, int do_refine,
- int *cost_list,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *ref_mv, MV *dst_mv);
+ MV *mvp_full, int step_param, int sadpb,
+ int further_steps, int do_refine, int *cost_list,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *ref_mv, MV *dst_mv);
// Perform integral projection based motion estimation.
unsigned int vp10_int_pro_motion_estimation(const struct VP10_COMP *cpi,
- MACROBLOCK *x,
- BLOCK_SIZE bsize,
- int mi_row, int mi_col);
+ MACROBLOCK *x, BLOCK_SIZE bsize,
+ int mi_row, int mi_col);
-typedef int (integer_mv_pattern_search_fn) (
- const MACROBLOCK *x,
- MV *ref_mv,
- int search_param,
- int error_per_bit,
- int do_init_search,
- int *cost_list,
- const vpx_variance_fn_ptr_t *vf,
- int use_mvcost,
- const MV *center_mv,
- MV *best_mv);
+typedef int(integer_mv_pattern_search_fn)(const MACROBLOCK *x, MV *ref_mv,
+ int search_param, int error_per_bit,
+ int do_init_search, int *cost_list,
+ const vpx_variance_fn_ptr_t *vf,
+ int use_mvcost, const MV *center_mv,
+ MV *best_mv);
integer_mv_pattern_search_fn vp10_hex_search;
integer_mv_pattern_search_fn vp10_bigdia_search;
@@ -104,59 +92,45 @@
integer_mv_pattern_search_fn vp10_fast_hex_search;
integer_mv_pattern_search_fn vp10_fast_dia_search;
-typedef int (fractional_mv_step_fp) (
- const MACROBLOCK *x,
- MV *bestmv, const MV *ref_mv,
- int allow_hp,
- int error_per_bit,
- const vpx_variance_fn_ptr_t *vfp,
+typedef int(fractional_mv_step_fp)(
+ const MACROBLOCK *x, MV *bestmv, const MV *ref_mv, int allow_hp,
+ int error_per_bit, const vpx_variance_fn_ptr_t *vfp,
int forced_stop, // 0 - full, 1 - qtr only, 2 - half only
- int iters_per_step,
- int *cost_list,
- int *mvjcost, int *mvcost[2],
- int *distortion, unsigned int *sse1,
- const uint8_t *second_pred,
- int w, int h);
+ int iters_per_step, int *cost_list, int *mvjcost, int *mvcost[2],
+ int *distortion, unsigned int *sse1, const uint8_t *second_pred, int w,
+ int h);
extern fractional_mv_step_fp vp10_find_best_sub_pixel_tree;
extern fractional_mv_step_fp vp10_find_best_sub_pixel_tree_pruned;
extern fractional_mv_step_fp vp10_find_best_sub_pixel_tree_pruned_more;
extern fractional_mv_step_fp vp10_find_best_sub_pixel_tree_pruned_evenmore;
-typedef int (*vp10_full_search_fn_t)(const MACROBLOCK *x,
- const MV *ref_mv, int sad_per_bit,
- int distance,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv, MV *best_mv);
+typedef int (*vp10_full_search_fn_t)(const MACROBLOCK *x, const MV *ref_mv,
+ int sad_per_bit, int distance,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv, MV *best_mv);
-typedef int (*vp10_refining_search_fn_t)(const MACROBLOCK *x,
- MV *ref_mv, int sad_per_bit,
- int distance,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv);
+typedef int (*vp10_refining_search_fn_t)(const MACROBLOCK *x, MV *ref_mv,
+ int sad_per_bit, int distance,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv);
-typedef int (*vp10_diamond_search_fn_t)(const MACROBLOCK *x,
- const search_site_config *cfg,
- MV *ref_mv, MV *best_mv,
- int search_param, int sad_per_bit,
- int *num00,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv);
+typedef int (*vp10_diamond_search_fn_t)(
+ const MACROBLOCK *x, const search_site_config *cfg, MV *ref_mv, MV *best_mv,
+ int search_param, int sad_per_bit, int *num00,
+ const vpx_variance_fn_ptr_t *fn_ptr, const MV *center_mv);
-int vp10_refining_search_8p_c(const MACROBLOCK *x,
- MV *ref_mv, int error_per_bit,
- int search_range,
- const vpx_variance_fn_ptr_t *fn_ptr,
- const MV *center_mv, const uint8_t *second_pred);
+int vp10_refining_search_8p_c(const MACROBLOCK *x, MV *ref_mv,
+ int error_per_bit, int search_range,
+ const vpx_variance_fn_ptr_t *fn_ptr,
+ const MV *center_mv, const uint8_t *second_pred);
struct VP10_COMP;
int vp10_full_pixel_search(struct VP10_COMP *cpi, MACROBLOCK *x,
- BLOCK_SIZE bsize, MV *mvp_full,
- int step_param, int error_per_bit,
- int *cost_list,
- const MV *ref_mv, MV *tmp_mv,
- int var_max, int rd);
+ BLOCK_SIZE bsize, MV *mvp_full, int step_param,
+ int error_per_bit, int *cost_list, const MV *ref_mv,
+ MV *tmp_mv, int var_max, int rd);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/encoder/mips/msa/error_msa.c b/vp10/encoder/mips/msa/error_msa.c
index dacca32..9a098ea 100644
--- a/vp10/encoder/mips/msa/error_msa.c
+++ b/vp10/encoder/mips/msa/error_msa.c
@@ -11,100 +11,93 @@
#include "./vp10_rtcd.h"
#include "vpx_dsp/mips/macros_msa.h"
-#define BLOCK_ERROR_BLOCKSIZE_MSA(BSize) \
-static int64_t block_error_##BSize##size_msa(const int16_t *coeff_ptr, \
- const int16_t *dq_coeff_ptr, \
- int64_t *ssz) { \
- int64_t err = 0; \
- uint32_t loop_cnt; \
- v8i16 coeff, dq_coeff, coeff_r_h, coeff_l_h; \
- v4i32 diff_r, diff_l, coeff_r_w, coeff_l_w; \
- v2i64 sq_coeff_r, sq_coeff_l; \
- v2i64 err0, err_dup0, err1, err_dup1; \
- \
- coeff = LD_SH(coeff_ptr); \
- dq_coeff = LD_SH(dq_coeff_ptr); \
- UNPCK_SH_SW(coeff, coeff_r_w, coeff_l_w); \
- ILVRL_H2_SH(coeff, dq_coeff, coeff_r_h, coeff_l_h); \
- HSUB_UH2_SW(coeff_r_h, coeff_l_h, diff_r, diff_l); \
- DOTP_SW2_SD(coeff_r_w, coeff_l_w, coeff_r_w, coeff_l_w, \
- sq_coeff_r, sq_coeff_l); \
- DOTP_SW2_SD(diff_r, diff_l, diff_r, diff_l, err0, err1); \
- \
- coeff = LD_SH(coeff_ptr + 8); \
- dq_coeff = LD_SH(dq_coeff_ptr + 8); \
- UNPCK_SH_SW(coeff, coeff_r_w, coeff_l_w); \
- ILVRL_H2_SH(coeff, dq_coeff, coeff_r_h, coeff_l_h); \
- HSUB_UH2_SW(coeff_r_h, coeff_l_h, diff_r, diff_l); \
- DPADD_SD2_SD(coeff_r_w, coeff_l_w, sq_coeff_r, sq_coeff_l); \
- DPADD_SD2_SD(diff_r, diff_l, err0, err1); \
- \
- coeff_ptr += 16; \
- dq_coeff_ptr += 16; \
- \
- for (loop_cnt = ((BSize >> 4) - 1); loop_cnt--;) { \
- coeff = LD_SH(coeff_ptr); \
- dq_coeff = LD_SH(dq_coeff_ptr); \
- UNPCK_SH_SW(coeff, coeff_r_w, coeff_l_w); \
- ILVRL_H2_SH(coeff, dq_coeff, coeff_r_h, coeff_l_h); \
- HSUB_UH2_SW(coeff_r_h, coeff_l_h, diff_r, diff_l); \
- DPADD_SD2_SD(coeff_r_w, coeff_l_w, sq_coeff_r, sq_coeff_l); \
- DPADD_SD2_SD(diff_r, diff_l, err0, err1); \
- \
- coeff = LD_SH(coeff_ptr + 8); \
- dq_coeff = LD_SH(dq_coeff_ptr + 8); \
- UNPCK_SH_SW(coeff, coeff_r_w, coeff_l_w); \
- ILVRL_H2_SH(coeff, dq_coeff, coeff_r_h, coeff_l_h); \
- HSUB_UH2_SW(coeff_r_h, coeff_l_h, diff_r, diff_l); \
- DPADD_SD2_SD(coeff_r_w, coeff_l_w, sq_coeff_r, sq_coeff_l); \
- DPADD_SD2_SD(diff_r, diff_l, err0, err1); \
- \
- coeff_ptr += 16; \
- dq_coeff_ptr += 16; \
- } \
- \
- err_dup0 = __msa_splati_d(sq_coeff_r, 1); \
- err_dup1 = __msa_splati_d(sq_coeff_l, 1); \
- sq_coeff_r += err_dup0; \
- sq_coeff_l += err_dup1; \
- *ssz = __msa_copy_s_d(sq_coeff_r, 0); \
- *ssz += __msa_copy_s_d(sq_coeff_l, 0); \
- \
- err_dup0 = __msa_splati_d(err0, 1); \
- err_dup1 = __msa_splati_d(err1, 1); \
- err0 += err_dup0; \
- err1 += err_dup1; \
- err = __msa_copy_s_d(err0, 0); \
- err += __msa_copy_s_d(err1, 0); \
- \
- return err; \
-}
+#define BLOCK_ERROR_BLOCKSIZE_MSA(BSize) \
+ static int64_t block_error_##BSize##size_msa( \
+ const int16_t *coeff_ptr, const int16_t *dq_coeff_ptr, int64_t *ssz) { \
+ int64_t err = 0; \
+ uint32_t loop_cnt; \
+ v8i16 coeff, dq_coeff, coeff_r_h, coeff_l_h; \
+ v4i32 diff_r, diff_l, coeff_r_w, coeff_l_w; \
+ v2i64 sq_coeff_r, sq_coeff_l; \
+ v2i64 err0, err_dup0, err1, err_dup1; \
+ \
+ coeff = LD_SH(coeff_ptr); \
+ dq_coeff = LD_SH(dq_coeff_ptr); \
+ UNPCK_SH_SW(coeff, coeff_r_w, coeff_l_w); \
+ ILVRL_H2_SH(coeff, dq_coeff, coeff_r_h, coeff_l_h); \
+ HSUB_UH2_SW(coeff_r_h, coeff_l_h, diff_r, diff_l); \
+ DOTP_SW2_SD(coeff_r_w, coeff_l_w, coeff_r_w, coeff_l_w, sq_coeff_r, \
+ sq_coeff_l); \
+ DOTP_SW2_SD(diff_r, diff_l, diff_r, diff_l, err0, err1); \
+ \
+ coeff = LD_SH(coeff_ptr + 8); \
+ dq_coeff = LD_SH(dq_coeff_ptr + 8); \
+ UNPCK_SH_SW(coeff, coeff_r_w, coeff_l_w); \
+ ILVRL_H2_SH(coeff, dq_coeff, coeff_r_h, coeff_l_h); \
+ HSUB_UH2_SW(coeff_r_h, coeff_l_h, diff_r, diff_l); \
+ DPADD_SD2_SD(coeff_r_w, coeff_l_w, sq_coeff_r, sq_coeff_l); \
+ DPADD_SD2_SD(diff_r, diff_l, err0, err1); \
+ \
+ coeff_ptr += 16; \
+ dq_coeff_ptr += 16; \
+ \
+ for (loop_cnt = ((BSize >> 4) - 1); loop_cnt--;) { \
+ coeff = LD_SH(coeff_ptr); \
+ dq_coeff = LD_SH(dq_coeff_ptr); \
+ UNPCK_SH_SW(coeff, coeff_r_w, coeff_l_w); \
+ ILVRL_H2_SH(coeff, dq_coeff, coeff_r_h, coeff_l_h); \
+ HSUB_UH2_SW(coeff_r_h, coeff_l_h, diff_r, diff_l); \
+ DPADD_SD2_SD(coeff_r_w, coeff_l_w, sq_coeff_r, sq_coeff_l); \
+ DPADD_SD2_SD(diff_r, diff_l, err0, err1); \
+ \
+ coeff = LD_SH(coeff_ptr + 8); \
+ dq_coeff = LD_SH(dq_coeff_ptr + 8); \
+ UNPCK_SH_SW(coeff, coeff_r_w, coeff_l_w); \
+ ILVRL_H2_SH(coeff, dq_coeff, coeff_r_h, coeff_l_h); \
+ HSUB_UH2_SW(coeff_r_h, coeff_l_h, diff_r, diff_l); \
+ DPADD_SD2_SD(coeff_r_w, coeff_l_w, sq_coeff_r, sq_coeff_l); \
+ DPADD_SD2_SD(diff_r, diff_l, err0, err1); \
+ \
+ coeff_ptr += 16; \
+ dq_coeff_ptr += 16; \
+ } \
+ \
+ err_dup0 = __msa_splati_d(sq_coeff_r, 1); \
+ err_dup1 = __msa_splati_d(sq_coeff_l, 1); \
+ sq_coeff_r += err_dup0; \
+ sq_coeff_l += err_dup1; \
+ *ssz = __msa_copy_s_d(sq_coeff_r, 0); \
+ *ssz += __msa_copy_s_d(sq_coeff_l, 0); \
+ \
+ err_dup0 = __msa_splati_d(err0, 1); \
+ err_dup1 = __msa_splati_d(err1, 1); \
+ err0 += err_dup0; \
+ err1 += err_dup1; \
+ err = __msa_copy_s_d(err0, 0); \
+ err += __msa_copy_s_d(err1, 0); \
+ \
+ return err; \
+ }
-BLOCK_ERROR_BLOCKSIZE_MSA(16);
-BLOCK_ERROR_BLOCKSIZE_MSA(64);
-BLOCK_ERROR_BLOCKSIZE_MSA(256);
-BLOCK_ERROR_BLOCKSIZE_MSA(1024);
+/* clang-format off */
+BLOCK_ERROR_BLOCKSIZE_MSA(16)
+BLOCK_ERROR_BLOCKSIZE_MSA(64)
+BLOCK_ERROR_BLOCKSIZE_MSA(256)
+BLOCK_ERROR_BLOCKSIZE_MSA(1024)
+/* clang-format on */
int64_t vp10_block_error_msa(const tran_low_t *coeff_ptr,
- const tran_low_t *dq_coeff_ptr,
- intptr_t blk_size, int64_t *ssz) {
+ const tran_low_t *dq_coeff_ptr, intptr_t blk_size,
+ int64_t *ssz) {
int64_t err;
const int16_t *coeff = (const int16_t *)coeff_ptr;
const int16_t *dq_coeff = (const int16_t *)dq_coeff_ptr;
switch (blk_size) {
- case 16:
- err = block_error_16size_msa(coeff, dq_coeff, ssz);
- break;
- case 64:
- err = block_error_64size_msa(coeff, dq_coeff, ssz);
- break;
- case 256:
- err = block_error_256size_msa(coeff, dq_coeff, ssz);
- break;
- case 1024:
- err = block_error_1024size_msa(coeff, dq_coeff, ssz);
- break;
+ case 16: err = block_error_16size_msa(coeff, dq_coeff, ssz); break;
+ case 64: err = block_error_64size_msa(coeff, dq_coeff, ssz); break;
+ case 256: err = block_error_256size_msa(coeff, dq_coeff, ssz); break;
+ case 1024: err = block_error_1024size_msa(coeff, dq_coeff, ssz); break;
default:
err = vp10_block_error_c(coeff_ptr, dq_coeff_ptr, blk_size, ssz);
break;
diff --git a/vp10/encoder/mips/msa/fdct16x16_msa.c b/vp10/encoder/mips/msa/fdct16x16_msa.c
index d78fc64..c39a292 100644
--- a/vp10/encoder/mips/msa/fdct16x16_msa.c
+++ b/vp10/encoder/mips/msa/fdct16x16_msa.c
@@ -159,8 +159,8 @@
/* load input data */
LD_SH8(input, 16, l0, l1, l2, l3, l4, l5, l6, l7);
- TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7,
- r0, r1, r2, r3, r4, r5, r6, r7);
+ TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7, r0, r1, r2, r3, r4, r5, r6,
+ r7);
FDCT_POSTPROC_2V_NEG_H(r0, r1);
FDCT_POSTPROC_2V_NEG_H(r2, r3);
FDCT_POSTPROC_2V_NEG_H(r4, r5);
@@ -169,8 +169,8 @@
out += 64;
LD_SH8(input + 8, 16, l8, l9, l10, l11, l12, l13, l14, l15);
- TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15,
- r8, r9, r10, r11, r12, r13, r14, r15);
+ TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15, r8, r9, r10, r11,
+ r12, r13, r14, r15);
FDCT_POSTPROC_2V_NEG_H(r8, r9);
FDCT_POSTPROC_2V_NEG_H(r10, r11);
FDCT_POSTPROC_2V_NEG_H(r12, r13);
@@ -181,8 +181,8 @@
/* load input data */
input += 128;
LD_SH8(input, 16, l0, l1, l2, l3, l4, l5, l6, l7);
- TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7,
- r0, r1, r2, r3, r4, r5, r6, r7);
+ TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7, r0, r1, r2, r3, r4, r5, r6,
+ r7);
FDCT_POSTPROC_2V_NEG_H(r0, r1);
FDCT_POSTPROC_2V_NEG_H(r2, r3);
FDCT_POSTPROC_2V_NEG_H(r4, r5);
@@ -191,8 +191,8 @@
out += 64;
LD_SH8(input + 8, 16, l8, l9, l10, l11, l12, l13, l14, l15);
- TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15,
- r8, r9, r10, r11, r12, r13, r14, r15);
+ TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15, r8, r9, r10, r11,
+ r12, r13, r14, r15);
FDCT_POSTPROC_2V_NEG_H(r8, r9);
FDCT_POSTPROC_2V_NEG_H(r10, r11);
FDCT_POSTPROC_2V_NEG_H(r12, r13);
@@ -339,24 +339,24 @@
v8i16 l0, l1, l2, l3, l4, l5, l6, l7, l8, l9, l10, l11, l12, l13, l14, l15;
/* load input data */
- LD_SH16(input, 8, l0, l8, l1, l9, l2, l10, l3, l11,
- l4, l12, l5, l13, l6, l14, l7, l15);
- TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7,
- r0, r1, r2, r3, r4, r5, r6, r7);
- TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15,
- r8, r9, r10, r11, r12, r13, r14, r15);
+ LD_SH16(input, 8, l0, l8, l1, l9, l2, l10, l3, l11, l4, l12, l5, l13, l6, l14,
+ l7, l15);
+ TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7, r0, r1, r2, r3, r4, r5, r6,
+ r7);
+ TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15, r8, r9, r10, r11,
+ r12, r13, r14, r15);
ST_SH8(r0, r8, r1, r9, r2, r10, r3, r11, out, 8);
ST_SH8(r4, r12, r5, r13, r6, r14, r7, r15, (out + 64), 8);
out += 16 * 8;
/* load input data */
input += 128;
- LD_SH16(input, 8, l0, l8, l1, l9, l2, l10, l3, l11,
- l4, l12, l5, l13, l6, l14, l7, l15);
- TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7,
- r0, r1, r2, r3, r4, r5, r6, r7);
- TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15,
- r8, r9, r10, r11, r12, r13, r14, r15);
+ LD_SH16(input, 8, l0, l8, l1, l9, l2, l10, l3, l11, l4, l12, l5, l13, l6, l14,
+ l7, l15);
+ TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7, r0, r1, r2, r3, r4, r5, r6,
+ r7);
+ TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15, r8, r9, r10, r11,
+ r12, r13, r14, r15);
ST_SH8(r0, r8, r1, r9, r2, r10, r3, r11, out, 8);
ST_SH8(r4, r12, r5, r13, r6, r14, r7, r15, (out + 64), 8);
}
@@ -371,10 +371,10 @@
LD_SH8(temp, 16, in0, in1, in2, in3, in4, in5, in6, in7);
temp = intermediate + 8;
LD_SH8(temp, 16, in8, in9, in10, in11, in12, in13, in14, in15);
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- TRANSPOSE8x8_SH_SH(in8, in9, in10, in11, in12, in13, in14, in15,
- in8, in9, in10, in11, in12, in13, in14, in15);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
+ in4, in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in8, in9, in10, in11, in12, in13, in14, in15, in8, in9,
+ in10, in11, in12, in13, in14, in15);
FDCT_POSTPROC_2V_NEG_H(in0, in1);
FDCT_POSTPROC_2V_NEG_H(in2, in3);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
@@ -383,29 +383,28 @@
FDCT_POSTPROC_2V_NEG_H(in10, in11);
FDCT_POSTPROC_2V_NEG_H(in12, in13);
FDCT_POSTPROC_2V_NEG_H(in14, in15);
- BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7,
- in8, in9, in10, in11, in12, in13, in14, in15,
- tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7,
- in8, in9, in10, in11, in12, in13, in14, in15);
+ BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7, in8, in9, in10, in11,
+ in12, in13, in14, in15, tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6,
+ tmp7, in8, in9, in10, in11, in12, in13, in14, in15);
temp = intermediate;
ST_SH8(in8, in9, in10, in11, in12, in13, in14, in15, temp, 16);
- FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7,
- tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
+ FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp0, tmp1,
+ tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
temp = intermediate;
LD_SH8(temp, 16, in8, in9, in10, in11, in12, in13, in14, in15);
- FDCT8x16_ODD(in8, in9, in10, in11, in12, in13, in14, in15,
- in0, in1, in2, in3, in4, in5, in6, in7);
- TRANSPOSE8x8_SH_SH(tmp0, in0, tmp1, in1, tmp2, in2, tmp3, in3,
- tmp0, in0, tmp1, in1, tmp2, in2, tmp3, in3);
+ FDCT8x16_ODD(in8, in9, in10, in11, in12, in13, in14, in15, in0, in1, in2, in3,
+ in4, in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(tmp0, in0, tmp1, in1, tmp2, in2, tmp3, in3, tmp0, in0,
+ tmp1, in1, tmp2, in2, tmp3, in3);
ST_SH8(tmp0, in0, tmp1, in1, tmp2, in2, tmp3, in3, out, 16);
- TRANSPOSE8x8_SH_SH(tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7,
- tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7);
+ TRANSPOSE8x8_SH_SH(tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7, tmp4, in4,
+ tmp5, in5, tmp6, in6, tmp7, in7);
out = output + 8;
ST_SH8(tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7, out, 16);
}
-void vp10_fht16x16_msa(const int16_t *input, int16_t *output,
- int32_t stride, int32_t tx_type) {
+void vp10_fht16x16_msa(const int16_t *input, int16_t *output, int32_t stride,
+ int32_t tx_type) {
DECLARE_ALIGNED(32, int16_t, tmp[256]);
DECLARE_ALIGNED(32, int16_t, trans_buf[256]);
DECLARE_ALIGNED(32, int16_t, tmp_buf[128]);
@@ -413,35 +412,31 @@
int16_t *ptmpbuf = &tmp_buf[0];
int16_t *trans = &trans_buf[0];
const int32_t const_arr[29 * 4] = {
- 52707308, 52707308, 52707308, 52707308,
- -1072430300, -1072430300, -1072430300, -1072430300,
- 795618043, 795618043, 795618043, 795618043,
- -721080468, -721080468, -721080468, -721080468,
- 459094491, 459094491, 459094491, 459094491,
- -970646691, -970646691, -970646691, -970646691,
- 1010963856, 1010963856, 1010963856, 1010963856,
- -361743294, -361743294, -361743294, -361743294,
- 209469125, 209469125, 209469125, 209469125,
- -1053094788, -1053094788, -1053094788, -1053094788,
- 1053160324, 1053160324, 1053160324, 1053160324,
- 639644520, 639644520, 639644520, 639644520,
- -862444000, -862444000, -862444000, -862444000,
- 1062144356, 1062144356, 1062144356, 1062144356,
- -157532337, -157532337, -157532337, -157532337,
- 260914709, 260914709, 260914709, 260914709,
- -1041559667, -1041559667, -1041559667, -1041559667,
- 920985831, 920985831, 920985831, 920985831,
- -551995675, -551995675, -551995675, -551995675,
- 596522295, 596522295, 596522295, 596522295,
- 892853362, 892853362, 892853362, 892853362,
- -892787826, -892787826, -892787826, -892787826,
- 410925857, 410925857, 410925857, 410925857,
- -992012162, -992012162, -992012162, -992012162,
- 992077698, 992077698, 992077698, 992077698,
- 759246145, 759246145, 759246145, 759246145,
- -759180609, -759180609, -759180609, -759180609,
- -759222975, -759222975, -759222975, -759222975,
- 759288511, 759288511, 759288511, 759288511 };
+ 52707308, 52707308, 52707308, 52707308, -1072430300,
+ -1072430300, -1072430300, -1072430300, 795618043, 795618043,
+ 795618043, 795618043, -721080468, -721080468, -721080468,
+ -721080468, 459094491, 459094491, 459094491, 459094491,
+ -970646691, -970646691, -970646691, -970646691, 1010963856,
+ 1010963856, 1010963856, 1010963856, -361743294, -361743294,
+ -361743294, -361743294, 209469125, 209469125, 209469125,
+ 209469125, -1053094788, -1053094788, -1053094788, -1053094788,
+ 1053160324, 1053160324, 1053160324, 1053160324, 639644520,
+ 639644520, 639644520, 639644520, -862444000, -862444000,
+ -862444000, -862444000, 1062144356, 1062144356, 1062144356,
+ 1062144356, -157532337, -157532337, -157532337, -157532337,
+ 260914709, 260914709, 260914709, 260914709, -1041559667,
+ -1041559667, -1041559667, -1041559667, 920985831, 920985831,
+ 920985831, 920985831, -551995675, -551995675, -551995675,
+ -551995675, 596522295, 596522295, 596522295, 596522295,
+ 892853362, 892853362, 892853362, 892853362, -892787826,
+ -892787826, -892787826, -892787826, 410925857, 410925857,
+ 410925857, 410925857, -992012162, -992012162, -992012162,
+ -992012162, 992077698, 992077698, 992077698, 992077698,
+ 759246145, 759246145, 759246145, 759246145, -759180609,
+ -759180609, -759180609, -759180609, -759222975, -759222975,
+ -759222975, -759222975, 759288511, 759288511, 759288511,
+ 759288511
+ };
switch (tx_type) {
case DCT_DCT:
@@ -500,8 +495,6 @@
fadst16_transpose_msa(tmp, output);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
diff --git a/vp10/encoder/mips/msa/fdct4x4_msa.c b/vp10/encoder/mips/msa/fdct4x4_msa.c
index f9028f0..def9b45 100644
--- a/vp10/encoder/mips/msa/fdct4x4_msa.c
+++ b/vp10/encoder/mips/msa/fdct4x4_msa.c
@@ -14,7 +14,7 @@
#include "vp10/encoder/mips/msa/fdct_msa.h"
void vp10_fwht4x4_msa(const int16_t *input, int16_t *output,
- int32_t src_stride) {
+ int32_t src_stride) {
v8i16 in0, in1, in2, in3, in4;
LD_SH4(input, src_stride, in0, in1, in2, in3);
@@ -46,7 +46,7 @@
}
void vp10_fht4x4_msa(const int16_t *input, int16_t *output, int32_t stride,
- int32_t tx_type) {
+ int32_t tx_type) {
v8i16 in0, in1, in2, in3;
LD_SH4(input, stride, in0, in1, in2, in3);
@@ -86,9 +86,7 @@
TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
VPX_FADST4(in0, in1, in2, in3, in0, in1, in2, in3);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
diff --git a/vp10/encoder/mips/msa/fdct8x8_msa.c b/vp10/encoder/mips/msa/fdct8x8_msa.c
index 8e1e111..7843a62 100644
--- a/vp10/encoder/mips/msa/fdct8x8_msa.c
+++ b/vp10/encoder/mips/msa/fdct8x8_msa.c
@@ -14,7 +14,7 @@
#include "vp10/encoder/mips/msa/fdct_msa.h"
void vp10_fht8x8_msa(const int16_t *input, int16_t *output, int32_t stride,
- int32_t tx_type) {
+ int32_t tx_type) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
LD_SH8(input, stride, in0, in1, in2, in3, in4, in5, in6, in7);
@@ -23,44 +23,42 @@
switch (tx_type) {
case DCT_DCT:
- VPX_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- VPX_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ VPX_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2,
+ in3, in4, in5, in6, in7);
+ VPX_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
break;
case ADST_DCT:
- VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- VPX_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2,
+ in3, in4, in5, in6, in7);
+ VPX_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
break;
case DCT_ADST:
- VPX_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ VPX_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2,
+ in3, in4, in5, in6, in7);
+ VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
break;
case ADST_ADST:
- VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
- VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2,
+ in3, in4, in5, in6, in7);
+ VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
+ in5, in6, in7);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
- TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
- in0, in1, in2, in3, in4, in5, in6, in7);
+ TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
+ in4, in5, in6, in7);
SRLI_AVE_S_4V_H(in0, in1, in2, in3, in4, in5, in6, in7);
ST_SH8(in0, in1, in2, in3, in4, in5, in6, in7, output, 8);
}
diff --git a/vp10/encoder/mips/msa/fdct_msa.h b/vp10/encoder/mips/msa/fdct_msa.h
index bf56251..9a14625 100644
--- a/vp10/encoder/mips/msa/fdct_msa.h
+++ b/vp10/encoder/mips/msa/fdct_msa.h
@@ -15,103 +15,102 @@
#include "vpx_dsp/mips/txfm_macros_msa.h"
#include "vpx_ports/mem.h"
-#define VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, \
- out0, out1, out2, out3, out4, out5, out6, out7) { \
- v8i16 cnst0_m, cnst1_m, cnst2_m, cnst3_m, cnst4_m; \
- v8i16 vec0_m, vec1_m, vec2_m, vec3_m, s0_m, s1_m; \
- v8i16 coeff0_m = { cospi_2_64, cospi_6_64, cospi_10_64, cospi_14_64, \
- cospi_18_64, cospi_22_64, cospi_26_64, cospi_30_64 }; \
- v8i16 coeff1_m = { cospi_8_64, -cospi_8_64, cospi_16_64, -cospi_16_64, \
- cospi_24_64, -cospi_24_64, 0, 0 }; \
- \
- SPLATI_H2_SH(coeff0_m, 0, 7, cnst0_m, cnst1_m); \
- cnst2_m = -cnst0_m; \
- ILVEV_H2_SH(cnst0_m, cnst1_m, cnst1_m, cnst2_m, cnst0_m, cnst1_m); \
- SPLATI_H2_SH(coeff0_m, 4, 3, cnst2_m, cnst3_m); \
- cnst4_m = -cnst2_m; \
- ILVEV_H2_SH(cnst2_m, cnst3_m, cnst3_m, cnst4_m, cnst2_m, cnst3_m); \
- \
- ILVRL_H2_SH(in0, in7, vec1_m, vec0_m); \
- ILVRL_H2_SH(in4, in3, vec3_m, vec2_m); \
- DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
- cnst1_m, cnst2_m, cnst3_m, in7, in0, \
- in4, in3); \
- \
- SPLATI_H2_SH(coeff0_m, 2, 5, cnst0_m, cnst1_m); \
- cnst2_m = -cnst0_m; \
- ILVEV_H2_SH(cnst0_m, cnst1_m, cnst1_m, cnst2_m, cnst0_m, cnst1_m); \
- SPLATI_H2_SH(coeff0_m, 6, 1, cnst2_m, cnst3_m); \
- cnst4_m = -cnst2_m; \
- ILVEV_H2_SH(cnst2_m, cnst3_m, cnst3_m, cnst4_m, cnst2_m, cnst3_m); \
- \
- ILVRL_H2_SH(in2, in5, vec1_m, vec0_m); \
- ILVRL_H2_SH(in6, in1, vec3_m, vec2_m); \
- \
- DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
- cnst1_m, cnst2_m, cnst3_m, in5, in2, \
- in6, in1); \
- BUTTERFLY_4(in7, in0, in2, in5, s1_m, s0_m, in2, in5); \
- out7 = -s0_m; \
- out0 = s1_m; \
- \
- SPLATI_H4_SH(coeff1_m, 0, 4, 1, 5, cnst0_m, cnst1_m, cnst2_m, cnst3_m); \
- \
- ILVEV_H2_SH(cnst3_m, cnst0_m, cnst1_m, cnst2_m, cnst3_m, cnst2_m); \
- cnst0_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
- cnst1_m = cnst0_m; \
- \
- ILVRL_H2_SH(in4, in3, vec1_m, vec0_m); \
- ILVRL_H2_SH(in6, in1, vec3_m, vec2_m); \
- DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
- cnst2_m, cnst3_m, cnst1_m, out1, out6, \
- s0_m, s1_m); \
- \
- SPLATI_H2_SH(coeff1_m, 2, 3, cnst0_m, cnst1_m); \
- cnst1_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
- \
- ILVRL_H2_SH(in2, in5, vec1_m, vec0_m); \
- ILVRL_H2_SH(s0_m, s1_m, vec3_m, vec2_m); \
- out3 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
- out4 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m); \
- out2 = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst0_m); \
- out5 = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst1_m); \
- \
- out1 = -out1; \
- out3 = -out3; \
- out5 = -out5; \
-}
+#define VPX_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, \
+ out3, out4, out5, out6, out7) \
+ { \
+ v8i16 cnst0_m, cnst1_m, cnst2_m, cnst3_m, cnst4_m; \
+ v8i16 vec0_m, vec1_m, vec2_m, vec3_m, s0_m, s1_m; \
+ v8i16 coeff0_m = { cospi_2_64, cospi_6_64, cospi_10_64, cospi_14_64, \
+ cospi_18_64, cospi_22_64, cospi_26_64, cospi_30_64 }; \
+ v8i16 coeff1_m = { cospi_8_64, -cospi_8_64, cospi_16_64, -cospi_16_64, \
+ cospi_24_64, -cospi_24_64, 0, 0 }; \
+ \
+ SPLATI_H2_SH(coeff0_m, 0, 7, cnst0_m, cnst1_m); \
+ cnst2_m = -cnst0_m; \
+ ILVEV_H2_SH(cnst0_m, cnst1_m, cnst1_m, cnst2_m, cnst0_m, cnst1_m); \
+ SPLATI_H2_SH(coeff0_m, 4, 3, cnst2_m, cnst3_m); \
+ cnst4_m = -cnst2_m; \
+ ILVEV_H2_SH(cnst2_m, cnst3_m, cnst3_m, cnst4_m, cnst2_m, cnst3_m); \
+ \
+ ILVRL_H2_SH(in0, in7, vec1_m, vec0_m); \
+ ILVRL_H2_SH(in4, in3, vec3_m, vec2_m); \
+ DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, cnst1_m, \
+ cnst2_m, cnst3_m, in7, in0, in4, in3); \
+ \
+ SPLATI_H2_SH(coeff0_m, 2, 5, cnst0_m, cnst1_m); \
+ cnst2_m = -cnst0_m; \
+ ILVEV_H2_SH(cnst0_m, cnst1_m, cnst1_m, cnst2_m, cnst0_m, cnst1_m); \
+ SPLATI_H2_SH(coeff0_m, 6, 1, cnst2_m, cnst3_m); \
+ cnst4_m = -cnst2_m; \
+ ILVEV_H2_SH(cnst2_m, cnst3_m, cnst3_m, cnst4_m, cnst2_m, cnst3_m); \
+ \
+ ILVRL_H2_SH(in2, in5, vec1_m, vec0_m); \
+ ILVRL_H2_SH(in6, in1, vec3_m, vec2_m); \
+ \
+ DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, cnst1_m, \
+ cnst2_m, cnst3_m, in5, in2, in6, in1); \
+ BUTTERFLY_4(in7, in0, in2, in5, s1_m, s0_m, in2, in5); \
+ out7 = -s0_m; \
+ out0 = s1_m; \
+ \
+ SPLATI_H4_SH(coeff1_m, 0, 4, 1, 5, cnst0_m, cnst1_m, cnst2_m, cnst3_m); \
+ \
+ ILVEV_H2_SH(cnst3_m, cnst0_m, cnst1_m, cnst2_m, cnst3_m, cnst2_m); \
+ cnst0_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
+ cnst1_m = cnst0_m; \
+ \
+ ILVRL_H2_SH(in4, in3, vec1_m, vec0_m); \
+ ILVRL_H2_SH(in6, in1, vec3_m, vec2_m); \
+ DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, cnst2_m, \
+ cnst3_m, cnst1_m, out1, out6, s0_m, s1_m); \
+ \
+ SPLATI_H2_SH(coeff1_m, 2, 3, cnst0_m, cnst1_m); \
+ cnst1_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
+ \
+ ILVRL_H2_SH(in2, in5, vec1_m, vec0_m); \
+ ILVRL_H2_SH(s0_m, s1_m, vec3_m, vec2_m); \
+ out3 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
+ out4 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m); \
+ out2 = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst0_m); \
+ out5 = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst1_m); \
+ \
+ out1 = -out1; \
+ out3 = -out3; \
+ out5 = -out5; \
+ }
-#define VPX_FADST4(in0, in1, in2, in3, out0, out1, out2, out3) { \
- v4i32 s0_m, s1_m, s2_m, s3_m, constant_m; \
- v4i32 in0_r_m, in1_r_m, in2_r_m, in3_r_m; \
- \
- UNPCK_R_SH_SW(in0, in0_r_m); \
- UNPCK_R_SH_SW(in1, in1_r_m); \
- UNPCK_R_SH_SW(in2, in2_r_m); \
- UNPCK_R_SH_SW(in3, in3_r_m); \
- \
- constant_m = __msa_fill_w(sinpi_4_9); \
- MUL2(in0_r_m, constant_m, in3_r_m, constant_m, s1_m, s0_m); \
- \
- constant_m = __msa_fill_w(sinpi_1_9); \
- s0_m += in0_r_m * constant_m; \
- s1_m -= in1_r_m * constant_m; \
- \
- constant_m = __msa_fill_w(sinpi_2_9); \
- s0_m += in1_r_m * constant_m; \
- s1_m += in3_r_m * constant_m; \
- \
- s2_m = in0_r_m + in1_r_m - in3_r_m; \
- \
- constant_m = __msa_fill_w(sinpi_3_9); \
- MUL2(in2_r_m, constant_m, s2_m, constant_m, s3_m, in1_r_m); \
- \
- in0_r_m = s0_m + s3_m; \
- s2_m = s1_m - s3_m; \
- s3_m = s1_m - s0_m + s3_m; \
- \
- SRARI_W4_SW(in0_r_m, in1_r_m, s2_m, s3_m, DCT_CONST_BITS); \
- PCKEV_H4_SH(in0_r_m, in0_r_m, in1_r_m, in1_r_m, s2_m, s2_m, \
- s3_m, s3_m, out0, out1, out2, out3); \
-}
-#endif // VP10_ENCODER_MIPS_MSA_VP10_FDCT_MSA_H_
+#define VPX_FADST4(in0, in1, in2, in3, out0, out1, out2, out3) \
+ { \
+ v4i32 s0_m, s1_m, s2_m, s3_m, constant_m; \
+ v4i32 in0_r_m, in1_r_m, in2_r_m, in3_r_m; \
+ \
+ UNPCK_R_SH_SW(in0, in0_r_m); \
+ UNPCK_R_SH_SW(in1, in1_r_m); \
+ UNPCK_R_SH_SW(in2, in2_r_m); \
+ UNPCK_R_SH_SW(in3, in3_r_m); \
+ \
+ constant_m = __msa_fill_w(sinpi_4_9); \
+ MUL2(in0_r_m, constant_m, in3_r_m, constant_m, s1_m, s0_m); \
+ \
+ constant_m = __msa_fill_w(sinpi_1_9); \
+ s0_m += in0_r_m * constant_m; \
+ s1_m -= in1_r_m * constant_m; \
+ \
+ constant_m = __msa_fill_w(sinpi_2_9); \
+ s0_m += in1_r_m * constant_m; \
+ s1_m += in3_r_m * constant_m; \
+ \
+ s2_m = in0_r_m + in1_r_m - in3_r_m; \
+ \
+ constant_m = __msa_fill_w(sinpi_3_9); \
+ MUL2(in2_r_m, constant_m, s2_m, constant_m, s3_m, in1_r_m); \
+ \
+ in0_r_m = s0_m + s3_m; \
+ s2_m = s1_m - s3_m; \
+ s3_m = s1_m - s0_m + s3_m; \
+ \
+ SRARI_W4_SW(in0_r_m, in1_r_m, s2_m, s3_m, DCT_CONST_BITS); \
+ PCKEV_H4_SH(in0_r_m, in0_r_m, in1_r_m, in1_r_m, s2_m, s2_m, s3_m, s3_m, \
+ out0, out1, out2, out3); \
+ }
+#endif // VP10_ENCODER_MIPS_MSA_VP10_FDCT_MSA_H_
diff --git a/vp10/encoder/mips/msa/temporal_filter_msa.c b/vp10/encoder/mips/msa/temporal_filter_msa.c
index 5d4558b..5feeab3 100644
--- a/vp10/encoder/mips/msa/temporal_filter_msa.c
+++ b/vp10/encoder/mips/msa/temporal_filter_msa.c
@@ -11,12 +11,9 @@
#include "./vp10_rtcd.h"
#include "vpx_dsp/mips/macros_msa.h"
-static void temporal_filter_apply_8size_msa(uint8_t *frm1_ptr,
- uint32_t stride,
- uint8_t *frm2_ptr,
- int32_t filt_sth,
- int32_t filt_wgt,
- uint32_t *acc,
+static void temporal_filter_apply_8size_msa(uint8_t *frm1_ptr, uint32_t stride,
+ uint8_t *frm2_ptr, int32_t filt_sth,
+ int32_t filt_wgt, uint32_t *acc,
uint16_t *cnt) {
uint32_t row;
uint64_t f0, f1, f2, f3;
@@ -54,10 +51,10 @@
HSUB_UB2_SH(frm_r, frm_l, diff0, diff1);
UNPCK_SH_SW(diff0, diff0_r, diff0_l);
UNPCK_SH_SW(diff1, diff1_r, diff1_l);
- MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l,
- diff1_l, mod0_w, mod1_w, mod2_w, mod3_w);
- MUL4(mod0_w, cnst3, mod1_w, cnst3, mod2_w, cnst3, mod3_w, cnst3,
+ MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l, diff1_l,
mod0_w, mod1_w, mod2_w, mod3_w);
+ MUL4(mod0_w, cnst3, mod1_w, cnst3, mod2_w, cnst3, mod3_w, cnst3, mod0_w,
+ mod1_w, mod2_w, mod3_w);
SRAR_W4_SW(mod0_w, mod1_w, mod2_w, mod3_w, strength);
diff0_r = (mod0_w < cnst16);
@@ -65,8 +62,8 @@
diff1_r = (mod2_w < cnst16);
diff1_l = (mod3_w < cnst16);
- SUB4(cnst16, mod0_w, cnst16, mod1_w, cnst16, mod2_w, cnst16, mod3_w,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ SUB4(cnst16, mod0_w, cnst16, mod1_w, cnst16, mod2_w, cnst16, mod3_w, mod0_w,
+ mod1_w, mod2_w, mod3_w);
mod0_w = diff0_r & mod0_w;
mod1_w = diff0_l & mod1_w;
@@ -85,8 +82,8 @@
UNPCK_SH_SW(frm2_l, frm2_lr, frm2_ll);
MUL4(mod0_w, frm2_rr, mod1_w, frm2_rl, mod2_w, frm2_lr, mod3_w, frm2_ll,
mod0_w, mod1_w, mod2_w, mod3_w);
- ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3, mod0_w, mod1_w,
+ mod2_w, mod3_w);
ST_SW2(mod0_w, mod1_w, acc, 4);
acc += 8;
@@ -101,10 +98,10 @@
HSUB_UB2_SH(frm_r, frm_l, diff0, diff1);
UNPCK_SH_SW(diff0, diff0_r, diff0_l);
UNPCK_SH_SW(diff1, diff1_r, diff1_l);
- MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l,
- diff1_l, mod0_w, mod1_w, mod2_w, mod3_w);
- MUL4(mod0_w, cnst3, mod1_w, cnst3, mod2_w, cnst3, mod3_w, cnst3,
+ MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l, diff1_l,
mod0_w, mod1_w, mod2_w, mod3_w);
+ MUL4(mod0_w, cnst3, mod1_w, cnst3, mod2_w, cnst3, mod3_w, cnst3, mod0_w,
+ mod1_w, mod2_w, mod3_w);
SRAR_W4_SW(mod0_w, mod1_w, mod2_w, mod3_w, strength);
diff0_r = (mod0_w < cnst16);
@@ -112,8 +109,8 @@
diff1_r = (mod2_w < cnst16);
diff1_l = (mod3_w < cnst16);
- SUB4(cnst16, mod0_w, cnst16, mod1_w, cnst16, mod2_w, cnst16, mod3_w,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ SUB4(cnst16, mod0_w, cnst16, mod1_w, cnst16, mod2_w, cnst16, mod3_w, mod0_w,
+ mod1_w, mod2_w, mod3_w);
mod0_w = diff0_r & mod0_w;
mod1_w = diff0_l & mod1_w;
@@ -131,8 +128,8 @@
UNPCK_SH_SW(frm2_l, frm2_lr, frm2_ll);
MUL4(mod0_w, frm2_rr, mod1_w, frm2_rl, mod2_w, frm2_lr, mod3_w, frm2_ll,
mod0_w, mod1_w, mod2_w, mod3_w);
- ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3, mod0_w, mod1_w,
+ mod2_w, mod3_w);
ST_SW2(mod0_w, mod1_w, acc, 4);
acc += 8;
@@ -141,13 +138,10 @@
}
}
-static void temporal_filter_apply_16size_msa(uint8_t *frm1_ptr,
- uint32_t stride,
+static void temporal_filter_apply_16size_msa(uint8_t *frm1_ptr, uint32_t stride,
uint8_t *frm2_ptr,
- int32_t filt_sth,
- int32_t filt_wgt,
- uint32_t *acc,
- uint16_t *cnt) {
+ int32_t filt_sth, int32_t filt_wgt,
+ uint32_t *acc, uint16_t *cnt) {
uint32_t row;
v16i8 frm1, frm2, frm3, frm4;
v16u8 frm_r, frm_l;
@@ -183,8 +177,8 @@
UNPCK_SH_SW(diff1, diff1_r, diff1_l);
MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l, diff1_l,
mod0_w, mod1_w, mod2_w, mod3_w);
- MUL4(mod0_w, cnst3, mod1_w, cnst3, mod2_w, cnst3, mod3_w, cnst3,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ MUL4(mod0_w, cnst3, mod1_w, cnst3, mod2_w, cnst3, mod3_w, cnst3, mod0_w,
+ mod1_w, mod2_w, mod3_w);
SRAR_W4_SW(mod0_w, mod1_w, mod2_w, mod3_w, strength);
diff0_r = (mod0_w < cnst16);
@@ -192,8 +186,8 @@
diff1_r = (mod2_w < cnst16);
diff1_l = (mod3_w < cnst16);
- SUB4(cnst16, mod0_w, cnst16, mod1_w, cnst16, mod2_w, cnst16, mod3_w,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ SUB4(cnst16, mod0_w, cnst16, mod1_w, cnst16, mod2_w, cnst16, mod3_w, mod0_w,
+ mod1_w, mod2_w, mod3_w);
mod0_w = diff0_r & mod0_w;
mod1_w = diff0_l & mod1_w;
@@ -212,8 +206,8 @@
UNPCK_SH_SW(frm2_l, frm2_lr, frm2_ll);
MUL4(mod0_w, frm2_rr, mod1_w, frm2_rl, mod2_w, frm2_lr, mod3_w, frm2_ll,
mod0_w, mod1_w, mod2_w, mod3_w);
- ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3, mod0_w, mod1_w,
+ mod2_w, mod3_w);
ST_SW2(mod0_w, mod1_w, acc, 4);
acc += 8;
@@ -230,8 +224,8 @@
UNPCK_SH_SW(diff1, diff1_r, diff1_l);
MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l, diff1_l,
mod0_w, mod1_w, mod2_w, mod3_w);
- MUL4(mod0_w, cnst3, mod1_w, cnst3, mod2_w, cnst3, mod3_w, cnst3,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ MUL4(mod0_w, cnst3, mod1_w, cnst3, mod2_w, cnst3, mod3_w, cnst3, mod0_w,
+ mod1_w, mod2_w, mod3_w);
SRAR_W4_SW(mod0_w, mod1_w, mod2_w, mod3_w, strength);
diff0_r = (mod0_w < cnst16);
@@ -239,8 +233,8 @@
diff1_r = (mod2_w < cnst16);
diff1_l = (mod3_w < cnst16);
- SUB4(cnst16, mod0_w, cnst16, mod1_w, cnst16, mod2_w, cnst16, mod3_w,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ SUB4(cnst16, mod0_w, cnst16, mod1_w, cnst16, mod2_w, cnst16, mod3_w, mod0_w,
+ mod1_w, mod2_w, mod3_w);
mod0_w = diff0_r & mod0_w;
mod1_w = diff0_l & mod1_w;
@@ -259,8 +253,8 @@
UNPCK_SH_SW(frm2_l, frm2_lr, frm2_ll);
MUL4(mod0_w, frm2_rr, mod1_w, frm2_rl, mod2_w, frm2_lr, mod3_w, frm2_ll,
mod0_w, mod1_w, mod2_w, mod3_w);
- ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3,
- mod0_w, mod1_w, mod2_w, mod3_w);
+ ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3, mod0_w, mod1_w,
+ mod2_w, mod3_w);
ST_SW2(mod0_w, mod1_w, acc, 4);
acc += 8;
ST_SW2(mod2_w, mod3_w, acc, 4);
@@ -272,18 +266,18 @@
}
void vp10_temporal_filter_apply_msa(uint8_t *frame1_ptr, uint32_t stride,
- uint8_t *frame2_ptr, uint32_t blk_w,
- uint32_t blk_h, int32_t strength,
- int32_t filt_wgt, uint32_t *accu,
- uint16_t *cnt) {
+ uint8_t *frame2_ptr, uint32_t blk_w,
+ uint32_t blk_h, int32_t strength,
+ int32_t filt_wgt, uint32_t *accu,
+ uint16_t *cnt) {
if (8 == (blk_w * blk_h)) {
- temporal_filter_apply_8size_msa(frame1_ptr, stride, frame2_ptr,
- strength, filt_wgt, accu, cnt);
+ temporal_filter_apply_8size_msa(frame1_ptr, stride, frame2_ptr, strength,
+ filt_wgt, accu, cnt);
} else if (16 == (blk_w * blk_h)) {
- temporal_filter_apply_16size_msa(frame1_ptr, stride, frame2_ptr,
- strength, filt_wgt, accu, cnt);
+ temporal_filter_apply_16size_msa(frame1_ptr, stride, frame2_ptr, strength,
+ filt_wgt, accu, cnt);
} else {
vp10_temporal_filter_apply_c(frame1_ptr, stride, frame2_ptr, blk_w, blk_h,
- strength, filt_wgt, accu, cnt);
+ strength, filt_wgt, accu, cnt);
}
}
diff --git a/vp10/encoder/picklpf.c b/vp10/encoder/picklpf.c
index b8e692e..0550df8 100644
--- a/vp10/encoder/picklpf.c
+++ b/vp10/encoder/picklpf.c
@@ -34,20 +34,19 @@
}
}
-
static int64_t try_filter_frame(const YV12_BUFFER_CONFIG *sd,
- VP10_COMP *const cpi,
- int filt_level, int partial_frame) {
+ VP10_COMP *const cpi, int filt_level,
+ int partial_frame) {
VP10_COMMON *const cm = &cpi->common;
int64_t filt_err;
if (cpi->num_workers > 1)
vp10_loop_filter_frame_mt(cm->frame_to_show, cm, cpi->td.mb.e_mbd.plane,
- filt_level, 1, partial_frame,
- cpi->workers, cpi->num_workers, &cpi->lf_row_sync);
+ filt_level, 1, partial_frame, cpi->workers,
+ cpi->num_workers, &cpi->lf_row_sync);
else
vp10_loop_filter_frame(cm->frame_to_show, cm, &cpi->td.mb.e_mbd, filt_level,
- 1, partial_frame);
+ 1, partial_frame);
#if CONFIG_VPX_HIGHBITDEPTH
if (cm->use_highbitdepth) {
@@ -103,8 +102,7 @@
bias = (bias * cpi->twopass.section_intra_rating) / 20;
// yx, bias less for large block size
- if (cm->tx_mode != ONLY_4X4)
- bias >>= 1;
+ if (cm->tx_mode != ONLY_4X4) bias >>= 1;
if (filt_direction <= 0 && filt_low != filt_mid) {
// Get Low filter error score
@@ -115,8 +113,7 @@
// filter value.
if ((ss_err[filt_low] - bias) < best_err) {
// Was it actually better than the previous best?
- if (ss_err[filt_low] < best_err)
- best_err = ss_err[filt_low];
+ if (ss_err[filt_low] < best_err) best_err = ss_err[filt_low];
filt_best = filt_low;
}
@@ -148,21 +145,20 @@
}
void vp10_pick_filter_level(const YV12_BUFFER_CONFIG *sd, VP10_COMP *cpi,
- LPF_PICK_METHOD method) {
+ LPF_PICK_METHOD method) {
VP10_COMMON *const cm = &cpi->common;
struct loopfilter *const lf = &cm->lf;
- lf->sharpness_level = cm->frame_type == KEY_FRAME ? 0
- : cpi->oxcf.sharpness;
+ lf->sharpness_level = cm->frame_type == KEY_FRAME ? 0 : cpi->oxcf.sharpness;
if (method == LPF_PICK_MINIMAL_LPF && lf->filter_level) {
- lf->filter_level = 0;
+ lf->filter_level = 0;
} else if (method >= LPF_PICK_FROM_Q) {
const int min_filter_level = 0;
const int max_filter_level = get_max_filter_level(cpi);
const int q = vp10_ac_quant(cm->base_qindex, 0, cm->bit_depth);
- // These values were determined by linear fitting the result of the
- // searched level, filt_guess = q * 0.316206 + 3.87252
+// These values were determined by linear fitting the result of the
+// searched level, filt_guess = q * 0.316206 + 3.87252
#if CONFIG_VPX_HIGHBITDEPTH
int filt_guess;
switch (cm->bit_depth) {
@@ -176,18 +172,18 @@
filt_guess = ROUND_POWER_OF_TWO(q * 20723 + 16242526, 22);
break;
default:
- assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 "
- "or VPX_BITS_12");
+ assert(0 &&
+ "bit_depth should be VPX_BITS_8, VPX_BITS_10 "
+ "or VPX_BITS_12");
return;
}
#else
int filt_guess = ROUND_POWER_OF_TWO(q * 20723 + 1015158, 18);
#endif // CONFIG_VPX_HIGHBITDEPTH
- if (cm->frame_type == KEY_FRAME)
- filt_guess -= 4;
+ if (cm->frame_type == KEY_FRAME) filt_guess -= 4;
lf->filter_level = clamp(filt_guess, min_filter_level, max_filter_level);
} else {
- lf->filter_level = search_filter_level(sd, cpi,
- method == LPF_PICK_FROM_SUBIMAGE);
+ lf->filter_level =
+ search_filter_level(sd, cpi, method == LPF_PICK_FROM_SUBIMAGE);
}
}
diff --git a/vp10/encoder/picklpf.h b/vp10/encoder/picklpf.h
index 21a8758..7d5f167 100644
--- a/vp10/encoder/picklpf.h
+++ b/vp10/encoder/picklpf.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_PICKLPF_H_
#define VP10_ENCODER_PICKLPF_H_
@@ -22,7 +21,7 @@
struct VP10_COMP;
void vp10_pick_filter_level(const struct yv12_buffer_config *sd,
- struct VP10_COMP *cpi, LPF_PICK_METHOD method);
+ struct VP10_COMP *cpi, LPF_PICK_METHOD method);
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/vp10/encoder/quantize.c b/vp10/encoder/quantize.c
index 66a8e5f..0688a69 100644
--- a/vp10/encoder/quantize.c
+++ b/vp10/encoder/quantize.c
@@ -21,13 +21,12 @@
#include "vp10/encoder/rd.h"
void vp10_quantize_fp_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs,
- int skip_block,
- const int16_t *zbin_ptr, const int16_t *round_ptr,
- const int16_t *quant_ptr, const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr,
- uint16_t *eob_ptr,
- const int16_t *scan, const int16_t *iscan) {
+ int skip_block, const int16_t *zbin_ptr,
+ const int16_t *round_ptr, const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr,
+ uint16_t *eob_ptr, const int16_t *scan,
+ const int16_t *iscan) {
int i, eob = -1;
// TODO(jingning) Decide the need of these arguments after the
// quantization process is completed.
@@ -53,27 +52,21 @@
qcoeff_ptr[rc] = (tmp ^ coeff_sign) - coeff_sign;
dqcoeff_ptr[rc] = qcoeff_ptr[rc] * dequant_ptr[rc != 0];
- if (tmp)
- eob = i;
+ if (tmp) eob = i;
}
}
*eob_ptr = eob + 1;
}
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_highbd_quantize_fp_c(const tran_low_t *coeff_ptr,
- intptr_t count,
- int skip_block,
- const int16_t *zbin_ptr,
- const int16_t *round_ptr,
- const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr,
- tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr,
- uint16_t *eob_ptr,
- const int16_t *scan,
- const int16_t *iscan) {
+void vp10_highbd_quantize_fp_c(const tran_low_t *coeff_ptr, intptr_t count,
+ int skip_block, const int16_t *zbin_ptr,
+ const int16_t *round_ptr,
+ const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr,
+ tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
+ const int16_t *dequant_ptr, uint16_t *eob_ptr,
+ const int16_t *scan, const int16_t *iscan) {
int i;
int eob = -1;
// TODO(jingning) Decide the need of these arguments after the
@@ -97,8 +90,7 @@
const uint32_t abs_qcoeff = (uint32_t)((tmp * quant_ptr[rc != 0]) >> 16);
qcoeff_ptr[rc] = (tran_low_t)((abs_qcoeff ^ coeff_sign) - coeff_sign);
dqcoeff_ptr[rc] = qcoeff_ptr[rc] * dequant_ptr[rc != 0];
- if (abs_qcoeff)
- eob = i;
+ if (abs_qcoeff) eob = i;
}
}
*eob_ptr = eob + 1;
@@ -108,14 +100,13 @@
// TODO(jingning) Refactor this file and combine functions with similar
// operations.
void vp10_quantize_fp_32x32_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs,
- int skip_block,
- const int16_t *zbin_ptr, const int16_t *round_ptr,
- const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr,
- uint16_t *eob_ptr,
- const int16_t *scan, const int16_t *iscan) {
+ int skip_block, const int16_t *zbin_ptr,
+ const int16_t *round_ptr,
+ const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr,
+ tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
+ const int16_t *dequant_ptr, uint16_t *eob_ptr,
+ const int16_t *scan, const int16_t *iscan) {
int i, eob = -1;
(void)zbin_ptr;
(void)quant_shift_ptr;
@@ -140,25 +131,19 @@
dqcoeff_ptr[rc] = qcoeff_ptr[rc] * dequant_ptr[rc != 0] / 2;
}
- if (tmp)
- eob = i;
+ if (tmp) eob = i;
}
}
*eob_ptr = eob + 1;
}
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_highbd_quantize_fp_32x32_c(const tran_low_t *coeff_ptr,
- intptr_t n_coeffs, int skip_block,
- const int16_t *zbin_ptr,
- const int16_t *round_ptr,
- const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr,
- tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr,
- uint16_t *eob_ptr,
- const int16_t *scan, const int16_t *iscan) {
+void vp10_highbd_quantize_fp_32x32_c(
+ const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block,
+ const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr,
+ const int16_t *scan, const int16_t *iscan) {
int i, eob = -1;
(void)zbin_ptr;
(void)quant_shift_ptr;
@@ -176,15 +161,14 @@
const int abs_coeff = (coeff ^ coeff_sign) - coeff_sign;
if (abs_coeff >= (dequant_ptr[rc != 0] >> 2)) {
- const int64_t tmp = abs_coeff
- + ROUND_POWER_OF_TWO(round_ptr[rc != 0], 1);
- abs_qcoeff = (uint32_t) ((tmp * quant_ptr[rc != 0]) >> 15);
+ const int64_t tmp =
+ abs_coeff + ROUND_POWER_OF_TWO(round_ptr[rc != 0], 1);
+ abs_qcoeff = (uint32_t)((tmp * quant_ptr[rc != 0]) >> 15);
qcoeff_ptr[rc] = (tran_low_t)((abs_qcoeff ^ coeff_sign) - coeff_sign);
dqcoeff_ptr[rc] = qcoeff_ptr[rc] * dequant_ptr[rc != 0] / 2;
}
- if (abs_qcoeff)
- eob = i;
+ if (abs_qcoeff) eob = i;
}
}
*eob_ptr = eob + 1;
@@ -192,37 +176,33 @@
#endif
void vp10_regular_quantize_b_4x4(MACROBLOCK *x, int plane, int block,
- const int16_t *scan, const int16_t *iscan) {
+ const int16_t *scan, const int16_t *iscan) {
MACROBLOCKD *const xd = &x->e_mbd;
struct macroblock_plane *p = &x->plane[plane];
struct macroblockd_plane *pd = &xd->plane[plane];
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- vpx_highbd_quantize_b(BLOCK_OFFSET(p->coeff, block),
- 16, x->skip_block,
+ vpx_highbd_quantize_b(BLOCK_OFFSET(p->coeff, block), 16, x->skip_block,
p->zbin, p->round, p->quant, p->quant_shift,
BLOCK_OFFSET(p->qcoeff, block),
- BLOCK_OFFSET(pd->dqcoeff, block),
- pd->dequant, &p->eobs[block],
- scan, iscan);
+ BLOCK_OFFSET(pd->dqcoeff, block), pd->dequant,
+ &p->eobs[block], scan, iscan);
return;
}
#endif
- vpx_quantize_b(BLOCK_OFFSET(p->coeff, block),
- 16, x->skip_block,
- p->zbin, p->round, p->quant, p->quant_shift,
+ vpx_quantize_b(BLOCK_OFFSET(p->coeff, block), 16, x->skip_block, p->zbin,
+ p->round, p->quant, p->quant_shift,
BLOCK_OFFSET(p->qcoeff, block),
- BLOCK_OFFSET(pd->dqcoeff, block),
- pd->dequant, &p->eobs[block], scan, iscan);
+ BLOCK_OFFSET(pd->dqcoeff, block), pd->dequant, &p->eobs[block],
+ scan, iscan);
}
static void invert_quant(int16_t *quant, int16_t *shift, int d) {
unsigned t;
int l;
t = d;
- for (l = 0; t > 1; l++)
- t >>= 1;
+ for (l = 0; t > 1; l++) t >>= 1;
t = 1 + (1 << (16 + l)) / d;
*quant = (int16_t)(t - (1 << 16));
*shift = 1 << (16 - l);
@@ -232,18 +212,15 @@
const int quant = vp10_dc_quant(q, 0, bit_depth);
#if CONFIG_VPX_HIGHBITDEPTH
switch (bit_depth) {
- case VPX_BITS_8:
- return q == 0 ? 64 : (quant < 148 ? 84 : 80);
- case VPX_BITS_10:
- return q == 0 ? 64 : (quant < 592 ? 84 : 80);
- case VPX_BITS_12:
- return q == 0 ? 64 : (quant < 2368 ? 84 : 80);
+ case VPX_BITS_8: return q == 0 ? 64 : (quant < 148 ? 84 : 80);
+ case VPX_BITS_10: return q == 0 ? 64 : (quant < 592 ? 84 : 80);
+ case VPX_BITS_12: return q == 0 ? 64 : (quant < 2368 ? 84 : 80);
default:
assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
return -1;
}
#else
- (void) bit_depth;
+ (void)bit_depth;
return q == 0 ? 64 : (quant < 148 ? 84 : 80);
#endif
}
@@ -259,8 +236,7 @@
for (i = 0; i < 2; ++i) {
int qrounding_factor_fp = i == 0 ? 48 : 42;
- if (q == 0)
- qrounding_factor_fp = 64;
+ if (q == 0) qrounding_factor_fp = 64;
// y
quant = i == 0 ? vp10_dc_quant(q, cm->y_dc_delta_q, cm->bit_depth)
@@ -275,8 +251,8 @@
// uv
quant = i == 0 ? vp10_dc_quant(q, cm->uv_dc_delta_q, cm->bit_depth)
: vp10_ac_quant(q, cm->uv_ac_delta_q, cm->bit_depth);
- invert_quant(&quants->uv_quant[q][i],
- &quants->uv_quant_shift[q][i], quant);
+ invert_quant(&quants->uv_quant[q][i], &quants->uv_quant_shift[q][i],
+ quant);
quants->uv_quant_fp[q][i] = (1 << 16) / quant;
quants->uv_round_fp[q][i] = (qrounding_factor_fp * quant) >> 7;
quants->uv_zbin[q][i] = ROUND_POWER_OF_TWO(qzbin_factor * quant, 7);
@@ -364,14 +340,11 @@
// Table that converts 0-63 Q-range values passed in outside to the Qindex
// range used internally.
static const int quantizer_to_qindex[] = {
- 0, 4, 8, 12, 16, 20, 24, 28,
- 32, 36, 40, 44, 48, 52, 56, 60,
- 64, 68, 72, 76, 80, 84, 88, 92,
- 96, 100, 104, 108, 112, 116, 120, 124,
- 128, 132, 136, 140, 144, 148, 152, 156,
- 160, 164, 168, 172, 176, 180, 184, 188,
- 192, 196, 200, 204, 208, 212, 216, 220,
- 224, 228, 232, 236, 240, 244, 249, 255,
+ 0, 4, 8, 12, 16, 20, 24, 28, 32, 36, 40, 44, 48,
+ 52, 56, 60, 64, 68, 72, 76, 80, 84, 88, 92, 96, 100,
+ 104, 108, 112, 116, 120, 124, 128, 132, 136, 140, 144, 148, 152,
+ 156, 160, 164, 168, 172, 176, 180, 184, 188, 192, 196, 200, 204,
+ 208, 212, 216, 220, 224, 228, 232, 236, 240, 244, 249, 255,
};
int vp10_quantizer_to_qindex(int quantizer) {
@@ -382,8 +355,7 @@
int quantizer;
for (quantizer = 0; quantizer < 64; ++quantizer)
- if (quantizer_to_qindex[quantizer] >= qindex)
- return quantizer;
+ if (quantizer_to_qindex[quantizer] >= qindex) return quantizer;
return 63;
}
diff --git a/vp10/encoder/quantize.h b/vp10/encoder/quantize.h
index b44088e..29f044c 100644
--- a/vp10/encoder/quantize.h
+++ b/vp10/encoder/quantize.h
@@ -38,7 +38,7 @@
} QUANTS;
void vp10_regular_quantize_b_4x4(MACROBLOCK *x, int plane, int block,
- const int16_t *scan, const int16_t *iscan);
+ const int16_t *scan, const int16_t *iscan);
struct VP10_COMP;
struct VP10Common;
diff --git a/vp10/encoder/ratectrl.c b/vp10/encoder/ratectrl.c
index 1666b2e..73593cb 100644
--- a/vp10/encoder/ratectrl.c
+++ b/vp10/encoder/ratectrl.c
@@ -46,29 +46,24 @@
#define FRAME_OVERHEAD_BITS 200
#if CONFIG_VPX_HIGHBITDEPTH
-#define ASSIGN_MINQ_TABLE(bit_depth, name) \
- do { \
- switch (bit_depth) { \
- case VPX_BITS_8: \
- name = name##_8; \
- break; \
- case VPX_BITS_10: \
- name = name##_10; \
- break; \
- case VPX_BITS_12: \
- name = name##_12; \
- break; \
- default: \
- assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10" \
- " or VPX_BITS_12"); \
- name = NULL; \
- } \
+#define ASSIGN_MINQ_TABLE(bit_depth, name) \
+ do { \
+ switch (bit_depth) { \
+ case VPX_BITS_8: name = name##_8; break; \
+ case VPX_BITS_10: name = name##_10; break; \
+ case VPX_BITS_12: name = name##_12; break; \
+ default: \
+ assert(0 && \
+ "bit_depth should be VPX_BITS_8, VPX_BITS_10" \
+ " or VPX_BITS_12"); \
+ name = NULL; \
+ } \
} while (0)
#else
#define ASSIGN_MINQ_TABLE(bit_depth, name) \
- do { \
- (void) bit_depth; \
- name = name##_8; \
+ do { \
+ (void) bit_depth; \
+ name = name##_8; \
} while (0)
#endif
@@ -111,20 +106,18 @@
// Special case handling to deal with the step from q2.0
// down to lossless mode represented by q 1.0.
- if (minqtarget <= 2.0)
- return 0;
+ if (minqtarget <= 2.0) return 0;
for (i = 0; i < QINDEX_RANGE; i++) {
- if (minqtarget <= vp10_convert_qindex_to_q(i, bit_depth))
- return i;
+ if (minqtarget <= vp10_convert_qindex_to_q(i, bit_depth)) return i;
}
return QINDEX_RANGE - 1;
}
-static void init_minq_luts(int *kf_low_m, int *kf_high_m,
- int *arfgf_low, int *arfgf_high,
- int *inter, int *rtc, vpx_bit_depth_t bit_depth) {
+static void init_minq_luts(int *kf_low_m, int *kf_high_m, int *arfgf_low,
+ int *arfgf_high, int *inter, int *rtc,
+ vpx_bit_depth_t bit_depth) {
int i;
for (i = 0; i < QINDEX_RANGE; i++) {
const double maxq = vp10_convert_qindex_to_q(i, bit_depth);
@@ -155,15 +148,12 @@
// quantizer tables easier. If necessary they can be replaced by lookup
// tables if and when things settle down in the experimental bitstream
double vp10_convert_qindex_to_q(int qindex, vpx_bit_depth_t bit_depth) {
- // Convert the index to a real Q value (scaled down to match old Q values)
+// Convert the index to a real Q value (scaled down to match old Q values)
#if CONFIG_VPX_HIGHBITDEPTH
switch (bit_depth) {
- case VPX_BITS_8:
- return vp10_ac_quant(qindex, 0, bit_depth) / 4.0;
- case VPX_BITS_10:
- return vp10_ac_quant(qindex, 0, bit_depth) / 16.0;
- case VPX_BITS_12:
- return vp10_ac_quant(qindex, 0, bit_depth) / 64.0;
+ case VPX_BITS_8: return vp10_ac_quant(qindex, 0, bit_depth) / 4.0;
+ case VPX_BITS_10: return vp10_ac_quant(qindex, 0, bit_depth) / 16.0;
+ case VPX_BITS_12: return vp10_ac_quant(qindex, 0, bit_depth) / 64.0;
default:
assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
return -1.0;
@@ -174,8 +164,7 @@
}
int vp10_rc_bits_per_mb(FRAME_TYPE frame_type, int qindex,
- double correction_factor,
- vpx_bit_depth_t bit_depth) {
+ double correction_factor, vpx_bit_depth_t bit_depth) {
const double q = vp10_convert_qindex_to_q(qindex, bit_depth);
int enumerator = frame_type == KEY_FRAME ? 2700000 : 1800000;
@@ -188,10 +177,10 @@
}
int vp10_estimate_bits_at_q(FRAME_TYPE frame_type, int q, int mbs,
- double correction_factor,
- vpx_bit_depth_t bit_depth) {
- const int bpm = (int)(vp10_rc_bits_per_mb(frame_type, q, correction_factor,
- bit_depth));
+ double correction_factor,
+ vpx_bit_depth_t bit_depth) {
+ const int bpm =
+ (int)(vp10_rc_bits_per_mb(frame_type, q, correction_factor, bit_depth));
return VPXMAX(FRAME_OVERHEAD_BITS,
(int)((uint64_t)bpm * mbs) >> BPER_MB_NORMBITS);
}
@@ -199,10 +188,9 @@
int vp10_rc_clamp_pframe_target_size(const VP10_COMP *const cpi, int target) {
const RATE_CONTROL *rc = &cpi->rc;
const VP10EncoderConfig *oxcf = &cpi->oxcf;
- const int min_frame_target = VPXMAX(rc->min_frame_bandwidth,
- rc->avg_frame_bandwidth >> 5);
- if (target < min_frame_target)
- target = min_frame_target;
+ const int min_frame_target =
+ VPXMAX(rc->min_frame_bandwidth, rc->avg_frame_bandwidth >> 5);
+ if (target < min_frame_target) target = min_frame_target;
if (cpi->refresh_golden_frame && rc->is_src_frame_alt_ref) {
// If there is an active ARF at this location use the minimum
// bits on this frame even if it is a constructed arf.
@@ -211,11 +199,10 @@
target = min_frame_target;
}
// Clip the frame target to the maximum allowed value.
- if (target > rc->max_frame_bandwidth)
- target = rc->max_frame_bandwidth;
+ if (target > rc->max_frame_bandwidth) target = rc->max_frame_bandwidth;
if (oxcf->rc_max_inter_bitrate_pct) {
- const int max_rate = rc->avg_frame_bandwidth *
- oxcf->rc_max_inter_bitrate_pct / 100;
+ const int max_rate =
+ rc->avg_frame_bandwidth * oxcf->rc_max_inter_bitrate_pct / 100;
target = VPXMIN(target, max_rate);
}
return target;
@@ -225,12 +212,11 @@
const RATE_CONTROL *rc = &cpi->rc;
const VP10EncoderConfig *oxcf = &cpi->oxcf;
if (oxcf->rc_max_intra_bitrate_pct) {
- const int max_rate = rc->avg_frame_bandwidth *
- oxcf->rc_max_intra_bitrate_pct / 100;
+ const int max_rate =
+ rc->avg_frame_bandwidth * oxcf->rc_max_intra_bitrate_pct / 100;
target = VPXMIN(target, max_rate);
}
- if (target > rc->max_frame_bandwidth)
- target = rc->max_frame_bandwidth;
+ if (target > rc->max_frame_bandwidth) target = rc->max_frame_bandwidth;
return target;
}
@@ -251,8 +237,8 @@
rc->buffer_level = rc->bits_off_target;
}
-int vp10_rc_get_default_min_gf_interval(
- int width, int height, double framerate) {
+int vp10_rc_get_default_min_gf_interval(int width, int height,
+ double framerate) {
// Assume we do not need any constraint lower than 4K 20 fps
static const double factor_safe = 3840 * 2160 * 20.0;
const double factor = width * height * framerate;
@@ -283,20 +269,20 @@
rc->avg_frame_qindex[KEY_FRAME] = oxcf->worst_allowed_q;
rc->avg_frame_qindex[INTER_FRAME] = oxcf->worst_allowed_q;
} else {
- rc->avg_frame_qindex[KEY_FRAME] = (oxcf->worst_allowed_q +
- oxcf->best_allowed_q) / 2;
- rc->avg_frame_qindex[INTER_FRAME] = (oxcf->worst_allowed_q +
- oxcf->best_allowed_q) / 2;
+ rc->avg_frame_qindex[KEY_FRAME] =
+ (oxcf->worst_allowed_q + oxcf->best_allowed_q) / 2;
+ rc->avg_frame_qindex[INTER_FRAME] =
+ (oxcf->worst_allowed_q + oxcf->best_allowed_q) / 2;
}
rc->last_q[KEY_FRAME] = oxcf->best_allowed_q;
rc->last_q[INTER_FRAME] = oxcf->worst_allowed_q;
- rc->buffer_level = rc->starting_buffer_level;
+ rc->buffer_level = rc->starting_buffer_level;
rc->bits_off_target = rc->starting_buffer_level;
- rc->rolling_target_bits = rc->avg_frame_bandwidth;
- rc->rolling_actual_bits = rc->avg_frame_bandwidth;
+ rc->rolling_target_bits = rc->avg_frame_bandwidth;
+ rc->rolling_actual_bits = rc->avg_frame_bandwidth;
rc->long_rolling_target_bits = rc->avg_frame_bandwidth;
rc->long_rolling_actual_bits = rc->avg_frame_bandwidth;
@@ -346,13 +332,11 @@
} else {
// If buffer is below drop_mark, for now just drop every other frame
// (starting with the next frame) until it increases back over drop_mark.
- int drop_mark = (int)(oxcf->drop_frames_water_mark *
- rc->optimal_buffer_level / 100);
- if ((rc->buffer_level > drop_mark) &&
- (rc->decimation_factor > 0)) {
+ int drop_mark =
+ (int)(oxcf->drop_frames_water_mark * rc->optimal_buffer_level / 100);
+ if ((rc->buffer_level > drop_mark) && (rc->decimation_factor > 0)) {
--rc->decimation_factor;
- } else if (rc->buffer_level <= drop_mark &&
- rc->decimation_factor == 0) {
+ } else if (rc->buffer_level <= drop_mark && rc->decimation_factor == 0) {
rc->decimation_factor = 1;
}
if (rc->decimation_factor > 0) {
@@ -379,7 +363,7 @@
rcf = rc->rate_correction_factors[KF_STD];
} else if (cpi->oxcf.pass == 2) {
RATE_FACTOR_LEVEL rf_lvl =
- cpi->twopass.gf_group.rf_level[cpi->twopass.gf_group.index];
+ cpi->twopass.gf_group.rf_level[cpi->twopass.gf_group.index];
rcf = rc->rate_correction_factors[rf_lvl];
} else {
if ((cpi->refresh_alt_ref_frame || cpi->refresh_golden_frame) &&
@@ -405,7 +389,7 @@
rc->rate_correction_factors[KF_STD] = factor;
} else if (cpi->oxcf.pass == 2) {
RATE_FACTOR_LEVEL rf_lvl =
- cpi->twopass.gf_group.rf_level[cpi->twopass.gf_group.index];
+ cpi->twopass.gf_group.rf_level[cpi->twopass.gf_group.index];
rc->rate_correction_factors[rf_lvl] = factor;
} else {
if ((cpi->refresh_alt_ref_frame || cpi->refresh_golden_frame) &&
@@ -426,8 +410,7 @@
int projected_size_based_on_q = 0;
// Do not update the rate factors for arf overlay frames.
- if (cpi->rc.is_src_frame_alt_ref)
- return;
+ if (cpi->rc.is_src_frame_alt_ref) return;
// Clear down mmx registers to allow floating point in what follows
vpx_clear_system_state();
@@ -439,21 +422,19 @@
projected_size_based_on_q =
vp10_cyclic_refresh_estimate_bits_at_q(cpi, rate_correction_factor);
} else {
- projected_size_based_on_q = vp10_estimate_bits_at_q(cpi->common.frame_type,
- cm->base_qindex,
- cm->MBs,
- rate_correction_factor,
- cm->bit_depth);
+ projected_size_based_on_q =
+ vp10_estimate_bits_at_q(cpi->common.frame_type, cm->base_qindex,
+ cm->MBs, rate_correction_factor, cm->bit_depth);
}
// Work out a size correction factor.
if (projected_size_based_on_q > FRAME_OVERHEAD_BITS)
correction_factor = (int)((100 * (int64_t)cpi->rc.projected_frame_size) /
- projected_size_based_on_q);
+ projected_size_based_on_q);
// More heavily damped adjustment used if we have been oscillating either side
// of target.
- adjustment_limit = 0.25 +
- 0.5 * VPXMIN(1, fabs(log10(0.01 * correction_factor)));
+ adjustment_limit =
+ 0.25 + 0.5 * VPXMIN(1, fabs(log10(0.01 * correction_factor)));
cpi->rc.q_2_frame = cpi->rc.q_1_frame;
cpi->rc.q_1_frame = cm->base_qindex;
@@ -467,16 +448,16 @@
if (correction_factor > 102) {
// We are not already at the worst allowable quality
- correction_factor = (int)(100 + ((correction_factor - 100) *
- adjustment_limit));
+ correction_factor =
+ (int)(100 + ((correction_factor - 100) * adjustment_limit));
rate_correction_factor = (rate_correction_factor * correction_factor) / 100;
// Keep rate_correction_factor within limits
if (rate_correction_factor > MAX_BPB_FACTOR)
rate_correction_factor = MAX_BPB_FACTOR;
} else if (correction_factor < 99) {
// We are not already at the best allowable quality
- correction_factor = (int)(100 - ((100 - correction_factor) *
- adjustment_limit));
+ correction_factor =
+ (int)(100 - ((100 - correction_factor) * adjustment_limit));
rate_correction_factor = (rate_correction_factor * correction_factor) / 100;
// Keep rate_correction_factor within limits
@@ -487,9 +468,8 @@
set_rate_correction_factor(cpi, rate_correction_factor);
}
-
int vp10_rc_regulate_q(const VP10_COMP *cpi, int target_bits_per_frame,
- int active_best_quality, int active_worst_quality) {
+ int active_best_quality, int active_worst_quality) {
const VP10_COMMON *const cm = &cpi->common;
int q = active_worst_quality;
int last_error = INT_MAX;
@@ -508,9 +488,8 @@
bits_per_mb_at_this_q =
(int)vp10_cyclic_refresh_rc_bits_per_mb(cpi, i, correction_factor);
} else {
- bits_per_mb_at_this_q = (int)vp10_rc_bits_per_mb(cm->frame_type, i,
- correction_factor,
- cm->bit_depth);
+ bits_per_mb_at_this_q = (int)vp10_rc_bits_per_mb(
+ cm->frame_type, i, correction_factor, cm->bit_depth);
}
if (bits_per_mb_at_this_q <= target_bits_per_mb) {
@@ -577,13 +556,13 @@
int active_worst_quality;
if (cpi->common.frame_type == KEY_FRAME) {
- active_worst_quality = curr_frame == 0 ? rc->worst_quality
- : rc->last_q[KEY_FRAME] * 2;
+ active_worst_quality =
+ curr_frame == 0 ? rc->worst_quality : rc->last_q[KEY_FRAME] * 2;
} else {
if (!rc->is_src_frame_alt_ref &&
(cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame)) {
- active_worst_quality = curr_frame == 1 ? rc->last_q[KEY_FRAME] * 5 / 4
- : rc->last_q[INTER_FRAME];
+ active_worst_quality = curr_frame == 1 ? rc->last_q[KEY_FRAME] * 5 / 4
+ : rc->last_q[INTER_FRAME];
} else {
active_worst_quality = curr_frame == 1 ? rc->last_q[KEY_FRAME] * 2
: rc->last_q[INTER_FRAME] * 2;
@@ -607,28 +586,27 @@
int adjustment = 0;
int active_worst_quality;
int ambient_qp;
- if (cm->frame_type == KEY_FRAME)
- return rc->worst_quality;
+ if (cm->frame_type == KEY_FRAME) return rc->worst_quality;
// For ambient_qp we use minimum of avg_frame_qindex[KEY_FRAME/INTER_FRAME]
// for the first few frames following key frame. These are both initialized
// to worst_quality and updated with (3/4, 1/4) average in postencode_update.
// So for first few frames following key, the qp of that key frame is weighted
// into the active_worst_quality setting.
- ambient_qp = (cm->current_video_frame < 5) ?
- VPXMIN(rc->avg_frame_qindex[INTER_FRAME],
- rc->avg_frame_qindex[KEY_FRAME]) :
- rc->avg_frame_qindex[INTER_FRAME];
+ ambient_qp = (cm->current_video_frame < 5)
+ ? VPXMIN(rc->avg_frame_qindex[INTER_FRAME],
+ rc->avg_frame_qindex[KEY_FRAME])
+ : rc->avg_frame_qindex[INTER_FRAME];
active_worst_quality = VPXMIN(rc->worst_quality, ambient_qp * 5 / 4);
if (rc->buffer_level > rc->optimal_buffer_level) {
// Adjust down.
// Maximum limit for down adjustment, ~30%.
int max_adjustment_down = active_worst_quality / 3;
if (max_adjustment_down) {
- buff_lvl_step = ((rc->maximum_buffer_size -
- rc->optimal_buffer_level) / max_adjustment_down);
+ buff_lvl_step = ((rc->maximum_buffer_size - rc->optimal_buffer_level) /
+ max_adjustment_down);
if (buff_lvl_step)
adjustment = (int)((rc->buffer_level - rc->optimal_buffer_level) /
- buff_lvl_step);
+ buff_lvl_step);
active_worst_quality -= adjustment;
}
} else if (rc->buffer_level > critical_level) {
@@ -668,18 +646,16 @@
if (rc->this_key_frame_forced) {
int qindex = rc->last_boosted_qindex;
double last_boosted_q = vp10_convert_qindex_to_q(qindex, cm->bit_depth);
- int delta_qindex = vp10_compute_qdelta(rc, last_boosted_q,
- (last_boosted_q * 0.75),
- cm->bit_depth);
+ int delta_qindex = vp10_compute_qdelta(
+ rc, last_boosted_q, (last_boosted_q * 0.75), cm->bit_depth);
active_best_quality = VPXMAX(qindex + delta_qindex, rc->best_quality);
} else if (cm->current_video_frame > 0) {
// not first frame of one pass and kf_boost is set
double q_adj_factor = 1.0;
double q_val;
- active_best_quality =
- get_kf_active_quality(rc, rc->avg_frame_qindex[KEY_FRAME],
- cm->bit_depth);
+ active_best_quality = get_kf_active_quality(
+ rc, rc->avg_frame_qindex[KEY_FRAME], cm->bit_depth);
// Allow somewhat lower kf minq with small image formats.
if ((cm->width * cm->height) <= (352 * 288)) {
@@ -689,9 +665,8 @@
// Convert the adjustment factor to a qindex delta
// on active_best_quality.
q_val = vp10_convert_qindex_to_q(active_best_quality, cm->bit_depth);
- active_best_quality += vp10_compute_qdelta(rc, q_val,
- q_val * q_adj_factor,
- cm->bit_depth);
+ active_best_quality +=
+ vp10_compute_qdelta(rc, q_val, q_val * q_adj_factor, cm->bit_depth);
}
} else if (!rc->is_src_frame_alt_ref &&
(cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame)) {
@@ -721,24 +696,22 @@
}
// Clip the active best and worst quality values to limits
- active_best_quality = clamp(active_best_quality,
- rc->best_quality, rc->worst_quality);
- active_worst_quality = clamp(active_worst_quality,
- active_best_quality, rc->worst_quality);
+ active_best_quality =
+ clamp(active_best_quality, rc->best_quality, rc->worst_quality);
+ active_worst_quality =
+ clamp(active_worst_quality, active_best_quality, rc->worst_quality);
*top_index = active_worst_quality;
*bottom_index = active_best_quality;
#if LIMIT_QRANGE_FOR_ALTREF_AND_KEY
// Limit Q range for the adaptive loop.
- if (cm->frame_type == KEY_FRAME &&
- !rc->this_key_frame_forced &&
+ if (cm->frame_type == KEY_FRAME && !rc->this_key_frame_forced &&
!(cm->current_video_frame == 0)) {
int qdelta = 0;
vpx_clear_system_state();
- qdelta = vp10_compute_qdelta_by_rate(&cpi->rc, cm->frame_type,
- active_worst_quality, 2.0,
- cm->bit_depth);
+ qdelta = vp10_compute_qdelta_by_rate(
+ &cpi->rc, cm->frame_type, active_worst_quality, 2.0, cm->bit_depth);
*top_index = active_worst_quality + qdelta;
*top_index = (*top_index > *bottom_index) ? *top_index : *bottom_index;
}
@@ -748,8 +721,8 @@
if (cm->frame_type == KEY_FRAME && rc->this_key_frame_forced) {
q = rc->last_boosted_qindex;
} else {
- q = vp10_rc_regulate_q(cpi, rc->this_frame_target,
- active_best_quality, active_worst_quality);
+ q = vp10_rc_regulate_q(cpi, rc->this_frame_target, active_best_quality,
+ active_worst_quality);
if (q > *top_index) {
// Special case when we are targeting the max allowed rate
if (rc->this_frame_target >= rc->max_frame_bandwidth)
@@ -758,8 +731,7 @@
q = *top_index;
}
}
- assert(*top_index <= rc->worst_quality &&
- *top_index >= rc->best_quality);
+ assert(*top_index <= rc->worst_quality && *top_index >= rc->best_quality);
assert(*bottom_index <= rc->worst_quality &&
*bottom_index >= rc->best_quality);
assert(q <= rc->worst_quality && q >= rc->best_quality);
@@ -770,8 +742,7 @@
const VP10EncoderConfig *const oxcf) {
static const double cq_adjust_threshold = 0.1;
int active_cq_level = oxcf->cq_level;
- if (oxcf->rc_mode == VPX_CQ &&
- rc->total_target_bits > 0) {
+ if (oxcf->rc_mode == VPX_CQ && rc->total_target_bits > 0) {
const double x = (double)rc->total_actual_bits / rc->total_target_bits;
if (x < cq_adjust_threshold) {
active_cq_level = (int)(active_cq_level * x / cq_adjust_threshold);
@@ -797,24 +768,21 @@
if (oxcf->rc_mode == VPX_Q) {
int qindex = cq_level;
double q = vp10_convert_qindex_to_q(qindex, cm->bit_depth);
- int delta_qindex = vp10_compute_qdelta(rc, q, q * 0.25,
- cm->bit_depth);
+ int delta_qindex = vp10_compute_qdelta(rc, q, q * 0.25, cm->bit_depth);
active_best_quality = VPXMAX(qindex + delta_qindex, rc->best_quality);
} else if (rc->this_key_frame_forced) {
int qindex = rc->last_boosted_qindex;
double last_boosted_q = vp10_convert_qindex_to_q(qindex, cm->bit_depth);
- int delta_qindex = vp10_compute_qdelta(rc, last_boosted_q,
- last_boosted_q * 0.75,
- cm->bit_depth);
+ int delta_qindex = vp10_compute_qdelta(
+ rc, last_boosted_q, last_boosted_q * 0.75, cm->bit_depth);
active_best_quality = VPXMAX(qindex + delta_qindex, rc->best_quality);
} else {
// not first frame of one pass and kf_boost is set
double q_adj_factor = 1.0;
double q_val;
- active_best_quality =
- get_kf_active_quality(rc, rc->avg_frame_qindex[KEY_FRAME],
- cm->bit_depth);
+ active_best_quality = get_kf_active_quality(
+ rc, rc->avg_frame_qindex[KEY_FRAME], cm->bit_depth);
// Allow somewhat lower kf minq with small image formats.
if ((cm->width * cm->height) <= (352 * 288)) {
@@ -824,9 +792,8 @@
// Convert the adjustment factor to a qindex delta
// on active_best_quality.
q_val = vp10_convert_qindex_to_q(active_best_quality, cm->bit_depth);
- active_best_quality += vp10_compute_qdelta(rc, q_val,
- q_val * q_adj_factor,
- cm->bit_depth);
+ active_best_quality +=
+ vp10_compute_qdelta(rc, q_val, q_val * q_adj_factor, cm->bit_depth);
}
} else if (!rc->is_src_frame_alt_ref &&
(cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame)) {
@@ -841,8 +808,7 @@
}
// For constrained quality dont allow Q less than the cq level
if (oxcf->rc_mode == VPX_CQ) {
- if (q < cq_level)
- q = cq_level;
+ if (q < cq_level) q = cq_level;
active_best_quality = get_gf_active_quality(rc, q, cm->bit_depth);
@@ -865,12 +831,11 @@
if (oxcf->rc_mode == VPX_Q) {
int qindex = cq_level;
double q = vp10_convert_qindex_to_q(qindex, cm->bit_depth);
- double delta_rate[FIXED_GF_INTERVAL] =
- {0.50, 1.0, 0.85, 1.0, 0.70, 1.0, 0.85, 1.0};
- int delta_qindex =
- vp10_compute_qdelta(rc, q,
- q * delta_rate[cm->current_video_frame %
- FIXED_GF_INTERVAL], cm->bit_depth);
+ double delta_rate[FIXED_GF_INTERVAL] = { 0.50, 1.0, 0.85, 1.0,
+ 0.70, 1.0, 0.85, 1.0 };
+ int delta_qindex = vp10_compute_qdelta(
+ rc, q, q * delta_rate[cm->current_video_frame % FIXED_GF_INTERVAL],
+ cm->bit_depth);
active_best_quality = VPXMAX(qindex + delta_qindex, rc->best_quality);
} else {
// Use the lower of active_worst_quality and recent/average Q.
@@ -880,18 +845,17 @@
active_best_quality = inter_minq[rc->avg_frame_qindex[KEY_FRAME]];
// For the constrained quality mode we don't want
// q to fall below the cq level.
- if ((oxcf->rc_mode == VPX_CQ) &&
- (active_best_quality < cq_level)) {
+ if ((oxcf->rc_mode == VPX_CQ) && (active_best_quality < cq_level)) {
active_best_quality = cq_level;
}
}
}
// Clip the active best and worst quality values to limits
- active_best_quality = clamp(active_best_quality,
- rc->best_quality, rc->worst_quality);
- active_worst_quality = clamp(active_worst_quality,
- active_best_quality, rc->worst_quality);
+ active_best_quality =
+ clamp(active_best_quality, rc->best_quality, rc->worst_quality);
+ active_worst_quality =
+ clamp(active_worst_quality, active_best_quality, rc->worst_quality);
*top_index = active_worst_quality;
*bottom_index = active_best_quality;
@@ -902,17 +866,14 @@
vpx_clear_system_state();
// Limit Q range for the adaptive loop.
- if (cm->frame_type == KEY_FRAME &&
- !rc->this_key_frame_forced &&
+ if (cm->frame_type == KEY_FRAME && !rc->this_key_frame_forced &&
!(cm->current_video_frame == 0)) {
- qdelta = vp10_compute_qdelta_by_rate(&cpi->rc, cm->frame_type,
- active_worst_quality, 2.0,
- cm->bit_depth);
+ qdelta = vp10_compute_qdelta_by_rate(
+ &cpi->rc, cm->frame_type, active_worst_quality, 2.0, cm->bit_depth);
} else if (!rc->is_src_frame_alt_ref &&
(cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame)) {
- qdelta = vp10_compute_qdelta_by_rate(&cpi->rc, cm->frame_type,
- active_worst_quality, 1.75,
- cm->bit_depth);
+ qdelta = vp10_compute_qdelta_by_rate(
+ &cpi->rc, cm->frame_type, active_worst_quality, 1.75, cm->bit_depth);
}
*top_index = active_worst_quality + qdelta;
*top_index = (*top_index > *bottom_index) ? *top_index : *bottom_index;
@@ -921,12 +882,12 @@
if (oxcf->rc_mode == VPX_Q) {
q = active_best_quality;
- // Special case code to try and match quality with forced key frames
+ // Special case code to try and match quality with forced key frames
} else if ((cm->frame_type == KEY_FRAME) && rc->this_key_frame_forced) {
q = rc->last_boosted_qindex;
} else {
- q = vp10_rc_regulate_q(cpi, rc->this_frame_target,
- active_best_quality, active_worst_quality);
+ q = vp10_rc_regulate_q(cpi, rc->this_frame_target, active_best_quality,
+ active_worst_quality);
if (q > *top_index) {
// Special case when we are targeting the max allowed rate
if (rc->this_frame_target >= rc->max_frame_bandwidth)
@@ -936,8 +897,7 @@
}
}
- assert(*top_index <= rc->worst_quality &&
- *top_index >= rc->best_quality);
+ assert(*top_index <= rc->worst_quality && *top_index >= rc->best_quality);
assert(*bottom_index <= rc->worst_quality &&
*bottom_index >= rc->best_quality);
assert(q <= rc->worst_quality && q >= rc->best_quality);
@@ -952,19 +912,19 @@
1.75, // GF_ARF_STD
2.00, // KF_STD
};
- static const FRAME_TYPE frame_type[RATE_FACTOR_LEVELS] =
- {INTER_FRAME, INTER_FRAME, INTER_FRAME, INTER_FRAME, KEY_FRAME};
+ static const FRAME_TYPE frame_type[RATE_FACTOR_LEVELS] = {
+ INTER_FRAME, INTER_FRAME, INTER_FRAME, INTER_FRAME, KEY_FRAME
+ };
const VP10_COMMON *const cm = &cpi->common;
- int qdelta = vp10_compute_qdelta_by_rate(&cpi->rc, frame_type[rf_level],
- q, rate_factor_deltas[rf_level],
- cm->bit_depth);
+ int qdelta =
+ vp10_compute_qdelta_by_rate(&cpi->rc, frame_type[rf_level], q,
+ rate_factor_deltas[rf_level], cm->bit_depth);
return qdelta;
}
#define STATIC_MOTION_THRESH 95
static int rc_pick_q_and_bounds_two_pass(const VP10_COMP *cpi,
- int *bottom_index,
- int *top_index) {
+ int *bottom_index, int *top_index) {
const VP10_COMMON *const cm = &cpi->common;
const RATE_CONTROL *const rc = &cpi->rc;
const VP10EncoderConfig *const oxcf = &cpi->oxcf;
@@ -989,17 +949,15 @@
qindex = VPXMIN(rc->last_kf_qindex, rc->last_boosted_qindex);
active_best_quality = qindex;
last_boosted_q = vp10_convert_qindex_to_q(qindex, cm->bit_depth);
- delta_qindex = vp10_compute_qdelta(rc, last_boosted_q,
- last_boosted_q * 1.25,
- cm->bit_depth);
+ delta_qindex = vp10_compute_qdelta(
+ rc, last_boosted_q, last_boosted_q * 1.25, cm->bit_depth);
active_worst_quality =
VPXMIN(qindex + delta_qindex, active_worst_quality);
} else {
qindex = rc->last_boosted_qindex;
last_boosted_q = vp10_convert_qindex_to_q(qindex, cm->bit_depth);
- delta_qindex = vp10_compute_qdelta(rc, last_boosted_q,
- last_boosted_q * 0.75,
- cm->bit_depth);
+ delta_qindex = vp10_compute_qdelta(
+ rc, last_boosted_q, last_boosted_q * 0.75, cm->bit_depth);
active_best_quality = VPXMAX(qindex + delta_qindex, rc->best_quality);
}
} else {
@@ -1007,8 +965,8 @@
double q_adj_factor = 1.0;
double q_val;
// Baseline value derived from cpi->active_worst_quality and kf boost.
- active_best_quality = get_kf_active_quality(rc, active_worst_quality,
- cm->bit_depth);
+ active_best_quality =
+ get_kf_active_quality(rc, active_worst_quality, cm->bit_depth);
// Allow somewhat lower kf minq with small image formats.
if ((cm->width * cm->height) <= (352 * 288)) {
@@ -1021,9 +979,8 @@
// Convert the adjustment factor to a qindex delta
// on active_best_quality.
q_val = vp10_convert_qindex_to_q(active_best_quality, cm->bit_depth);
- active_best_quality += vp10_compute_qdelta(rc, q_val,
- q_val * q_adj_factor,
- cm->bit_depth);
+ active_best_quality +=
+ vp10_compute_qdelta(rc, q_val, q_val * q_adj_factor, cm->bit_depth);
}
} else if (!rc->is_src_frame_alt_ref &&
(cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame)) {
@@ -1038,8 +995,7 @@
}
// For constrained quality dont allow Q less than the cq level
if (oxcf->rc_mode == VPX_CQ) {
- if (q < cq_level)
- q = cq_level;
+ if (q < cq_level) q = cq_level;
active_best_quality = get_gf_active_quality(rc, q, cm->bit_depth);
@@ -1050,7 +1006,7 @@
if (!cpi->refresh_alt_ref_frame) {
active_best_quality = cq_level;
} else {
- active_best_quality = get_gf_active_quality(rc, q, cm->bit_depth);
+ active_best_quality = get_gf_active_quality(rc, q, cm->bit_depth);
// Modify best quality for second level arfs. For mode VPX_Q this
// becomes the baseline frame q.
@@ -1068,8 +1024,7 @@
// For the constrained quality mode we don't want
// q to fall below the cq level.
- if ((oxcf->rc_mode == VPX_CQ) &&
- (active_best_quality < cq_level)) {
+ if ((oxcf->rc_mode == VPX_CQ) && (active_best_quality < cq_level)) {
active_best_quality = cq_level;
}
}
@@ -1083,11 +1038,11 @@
(!rc->is_src_frame_alt_ref &&
(cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame))) {
active_best_quality -=
- (cpi->twopass.extend_minq + cpi->twopass.extend_minq_fast);
+ (cpi->twopass.extend_minq + cpi->twopass.extend_minq_fast);
active_worst_quality += (cpi->twopass.extend_maxq / 2);
} else {
active_best_quality -=
- (cpi->twopass.extend_minq + cpi->twopass.extend_minq_fast) / 2;
+ (cpi->twopass.extend_minq + cpi->twopass.extend_minq_fast) / 2;
active_worst_quality += cpi->twopass.extend_maxq;
}
}
@@ -1095,33 +1050,31 @@
#if LIMIT_QRANGE_FOR_ALTREF_AND_KEY
vpx_clear_system_state();
// Static forced key frames Q restrictions dealt with elsewhere.
- if (!(frame_is_intra_only(cm)) ||
- !rc->this_key_frame_forced ||
+ if (!(frame_is_intra_only(cm)) || !rc->this_key_frame_forced ||
(cpi->twopass.last_kfgroup_zeromotion_pct < STATIC_MOTION_THRESH)) {
- int qdelta = vp10_frame_type_qdelta(cpi, gf_group->rf_level[gf_group->index],
- active_worst_quality);
- active_worst_quality = VPXMAX(active_worst_quality + qdelta,
- active_best_quality);
+ int qdelta = vp10_frame_type_qdelta(
+ cpi, gf_group->rf_level[gf_group->index], active_worst_quality);
+ active_worst_quality =
+ VPXMAX(active_worst_quality + qdelta, active_best_quality);
}
#endif
// Modify active_best_quality for downscaled normal frames.
if (rc->frame_size_selector != UNSCALED && !frame_is_kf_gf_arf(cpi)) {
- int qdelta = vp10_compute_qdelta_by_rate(rc, cm->frame_type,
- active_best_quality, 2.0,
- cm->bit_depth);
+ int qdelta = vp10_compute_qdelta_by_rate(
+ rc, cm->frame_type, active_best_quality, 2.0, cm->bit_depth);
active_best_quality =
VPXMAX(active_best_quality + qdelta, rc->best_quality);
}
- active_best_quality = clamp(active_best_quality,
- rc->best_quality, rc->worst_quality);
- active_worst_quality = clamp(active_worst_quality,
- active_best_quality, rc->worst_quality);
+ active_best_quality =
+ clamp(active_best_quality, rc->best_quality, rc->worst_quality);
+ active_worst_quality =
+ clamp(active_worst_quality, active_best_quality, rc->worst_quality);
if (oxcf->rc_mode == VPX_Q) {
q = active_best_quality;
- // Special case code to try and match quality with forced key frames.
+ // Special case code to try and match quality with forced key frames.
} else if (frame_is_intra_only(cm) && rc->this_key_frame_forced) {
// If static since last kf use better of last boosted and last kf q.
if (cpi->twopass.last_kfgroup_zeromotion_pct >= STATIC_MOTION_THRESH) {
@@ -1130,8 +1083,8 @@
q = rc->last_boosted_qindex;
}
} else {
- q = vp10_rc_regulate_q(cpi, rc->this_frame_target,
- active_best_quality, active_worst_quality);
+ q = vp10_rc_regulate_q(cpi, rc->this_frame_target, active_best_quality,
+ active_worst_quality);
if (q > active_worst_quality) {
// Special case when we are targeting the max allowed rate.
if (rc->this_frame_target >= rc->max_frame_bandwidth)
@@ -1145,16 +1098,15 @@
*top_index = active_worst_quality;
*bottom_index = active_best_quality;
- assert(*top_index <= rc->worst_quality &&
- *top_index >= rc->best_quality);
+ assert(*top_index <= rc->worst_quality && *top_index >= rc->best_quality);
assert(*bottom_index <= rc->worst_quality &&
*bottom_index >= rc->best_quality);
assert(q <= rc->worst_quality && q >= rc->best_quality);
return q;
}
-int vp10_rc_pick_q_and_bounds(const VP10_COMP *cpi,
- int *bottom_index, int *top_index) {
+int vp10_rc_pick_q_and_bounds(const VP10_COMP *cpi, int *bottom_index,
+ int *top_index) {
int q;
if (cpi->oxcf.pass == 0) {
if (cpi->oxcf.rc_mode == VPX_CBR)
@@ -1168,20 +1120,19 @@
return q;
}
-void vp10_rc_compute_frame_size_bounds(const VP10_COMP *cpi,
- int frame_target,
- int *frame_under_shoot_limit,
- int *frame_over_shoot_limit) {
+void vp10_rc_compute_frame_size_bounds(const VP10_COMP *cpi, int frame_target,
+ int *frame_under_shoot_limit,
+ int *frame_over_shoot_limit) {
if (cpi->oxcf.rc_mode == VPX_Q) {
*frame_under_shoot_limit = 0;
- *frame_over_shoot_limit = INT_MAX;
+ *frame_over_shoot_limit = INT_MAX;
} else {
// For very small rate targets where the fractional adjustment
// may be tiny make sure there is at least a minimum range.
const int tolerance = (cpi->sf.recode_tolerance * frame_target) / 100;
*frame_under_shoot_limit = VPXMAX(frame_target - tolerance - 200, 0);
- *frame_over_shoot_limit = VPXMIN(frame_target + tolerance + 200,
- cpi->rc.max_frame_bandwidth);
+ *frame_over_shoot_limit =
+ VPXMIN(frame_target + tolerance + 200, cpi->rc.max_frame_bandwidth);
}
}
@@ -1194,12 +1145,12 @@
// Modify frame size target when down-scaling.
if (cpi->oxcf.resize_mode == RESIZE_DYNAMIC &&
rc->frame_size_selector != UNSCALED)
- rc->this_frame_target = (int)(rc->this_frame_target
- * rate_thresh_mult[rc->frame_size_selector]);
+ rc->this_frame_target = (int)(rc->this_frame_target *
+ rate_thresh_mult[rc->frame_size_selector]);
// Target rate per SB64 (including partial SB64s.
- rc->sb64_target_rate = ((int64_t)rc->this_frame_target * 64 * 64) /
- (cm->width * cm->height);
+ rc->sb64_target_rate =
+ ((int64_t)rc->this_frame_target * 64 * 64) / (cm->width * cm->height);
}
static void update_alt_ref_frame_stats(VP10_COMP *cpi) {
@@ -1233,13 +1184,11 @@
}
// Decrement count down till next gf
- if (rc->frames_till_gf_update_due > 0)
- rc->frames_till_gf_update_due--;
+ if (rc->frames_till_gf_update_due > 0) rc->frames_till_gf_update_due--;
} else if (!cpi->refresh_alt_ref_frame) {
// Decrement count down till next gf
- if (rc->frames_till_gf_update_due > 0)
- rc->frames_till_gf_update_due--;
+ if (rc->frames_till_gf_update_due > 0) rc->frames_till_gf_update_due--;
rc->frames_since_golden++;
}
@@ -1271,7 +1220,7 @@
!(cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame)) {
rc->last_q[INTER_FRAME] = qindex;
rc->avg_frame_qindex[INTER_FRAME] =
- ROUND_POWER_OF_TWO(3 * rc->avg_frame_qindex[INTER_FRAME] + qindex, 2);
+ ROUND_POWER_OF_TWO(3 * rc->avg_frame_qindex[INTER_FRAME] + qindex, 2);
rc->ni_frames++;
rc->tot_q += vp10_convert_qindex_to_q(qindex, cm->bit_depth);
rc->avg_q = rc->tot_q / rc->ni_frames;
@@ -1287,15 +1236,13 @@
// If all mbs in this group are skipped only update if the Q value is
// better than that already stored.
// This is used to help set quality in forced key frames to reduce popping
- if ((qindex < rc->last_boosted_qindex) ||
- (cm->frame_type == KEY_FRAME) ||
+ if ((qindex < rc->last_boosted_qindex) || (cm->frame_type == KEY_FRAME) ||
(!rc->constrained_gf_group &&
(cpi->refresh_alt_ref_frame ||
(cpi->refresh_golden_frame && !rc->is_src_frame_alt_ref)))) {
rc->last_boosted_qindex = qindex;
}
- if (cm->frame_type == KEY_FRAME)
- rc->last_kf_qindex = qindex;
+ if (cm->frame_type == KEY_FRAME) rc->last_kf_qindex = qindex;
update_buffer_level(cpi, rc->projected_frame_size);
@@ -1326,8 +1273,7 @@
// Update the Golden frame stats as appropriate.
update_golden_frame_stats(cpi);
- if (cm->frame_type == KEY_FRAME)
- rc->frames_since_key = 0;
+ if (cm->frame_type == KEY_FRAME) rc->frames_since_key = 0;
if (cm->show_frame) {
rc->frames_since_key++;
rc->frames_to_key--;
@@ -1351,19 +1297,20 @@
}
// Use this macro to turn on/off use of alt-refs in one-pass mode.
-#define USE_ALTREF_FOR_ONE_PASS 1
+#define USE_ALTREF_FOR_ONE_PASS 1
static int calc_pframe_target_size_one_pass_vbr(const VP10_COMP *const cpi) {
static const int af_ratio = 10;
const RATE_CONTROL *const rc = &cpi->rc;
int target;
#if USE_ALTREF_FOR_ONE_PASS
- target = (!rc->is_src_frame_alt_ref &&
- (cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame)) ?
- (rc->avg_frame_bandwidth * rc->baseline_gf_interval * af_ratio) /
- (rc->baseline_gf_interval + af_ratio - 1) :
- (rc->avg_frame_bandwidth * rc->baseline_gf_interval) /
- (rc->baseline_gf_interval + af_ratio - 1);
+ target =
+ (!rc->is_src_frame_alt_ref &&
+ (cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame))
+ ? (rc->avg_frame_bandwidth * rc->baseline_gf_interval * af_ratio) /
+ (rc->baseline_gf_interval + af_ratio - 1)
+ : (rc->avg_frame_bandwidth * rc->baseline_gf_interval) /
+ (rc->baseline_gf_interval + af_ratio - 1);
#else
target = rc->avg_frame_bandwidth;
#endif
@@ -1383,13 +1330,11 @@
int target;
// TODO(yaowu): replace the "auto_key && 0" below with proper decision logic.
if (!cpi->refresh_alt_ref_frame &&
- (cm->current_video_frame == 0 ||
- (cpi->frame_flags & FRAMEFLAGS_KEY) ||
- rc->frames_to_key == 0 ||
- (cpi->oxcf.auto_key && 0))) {
+ (cm->current_video_frame == 0 || (cpi->frame_flags & FRAMEFLAGS_KEY) ||
+ rc->frames_to_key == 0 || (cpi->oxcf.auto_key && 0))) {
cm->frame_type = KEY_FRAME;
- rc->this_key_frame_forced = cm->current_video_frame != 0 &&
- rc->frames_to_key == 0;
+ rc->this_key_frame_forced =
+ cm->current_video_frame != 0 && rc->frames_to_key == 0;
rc->frames_to_key = cpi->oxcf.key_freq;
rc->kf_boost = DEFAULT_KF_BOOST;
rc->source_alt_ref_active = 0;
@@ -1428,11 +1373,12 @@
if (oxcf->gf_cbr_boost_pct) {
const int af_ratio_pct = oxcf->gf_cbr_boost_pct + 100;
- target = cpi->refresh_golden_frame ?
- (rc->avg_frame_bandwidth * rc->baseline_gf_interval * af_ratio_pct) /
- (rc->baseline_gf_interval * 100 + af_ratio_pct - 100) :
- (rc->avg_frame_bandwidth * rc->baseline_gf_interval * 100) /
- (rc->baseline_gf_interval * 100 + af_ratio_pct - 100);
+ target = cpi->refresh_golden_frame
+ ? (rc->avg_frame_bandwidth * rc->baseline_gf_interval *
+ af_ratio_pct) /
+ (rc->baseline_gf_interval * 100 + af_ratio_pct - 100)
+ : (rc->avg_frame_bandwidth * rc->baseline_gf_interval * 100) /
+ (rc->baseline_gf_interval * 100 + af_ratio_pct - 100);
} else {
target = rc->avg_frame_bandwidth;
}
@@ -1448,8 +1394,8 @@
target += (target * pct_high) / 200;
}
if (oxcf->rc_max_inter_bitrate_pct) {
- const int max_rate = rc->avg_frame_bandwidth *
- oxcf->rc_max_inter_bitrate_pct / 100;
+ const int max_rate =
+ rc->avg_frame_bandwidth * oxcf->rc_max_inter_bitrate_pct / 100;
target = VPXMIN(target, max_rate);
}
return VPXMAX(min_frame_target, target);
@@ -1460,15 +1406,15 @@
int target;
if (cpi->common.current_video_frame == 0) {
target = ((rc->starting_buffer_level / 2) > INT_MAX)
- ? INT_MAX : (int)(rc->starting_buffer_level / 2);
+ ? INT_MAX
+ : (int)(rc->starting_buffer_level / 2);
} else {
int kf_boost = 32;
double framerate = cpi->framerate;
kf_boost = VPXMAX(kf_boost, (int)(2 * framerate - 16));
- if (rc->frames_since_key < framerate / 2) {
- kf_boost = (int)(kf_boost * rc->frames_since_key /
- (framerate / 2));
+ if (rc->frames_since_key < framerate / 2) {
+ kf_boost = (int)(kf_boost * rc->frames_since_key / (framerate / 2));
}
target = ((16 + kf_boost) * rc->avg_frame_bandwidth) >> 4;
}
@@ -1480,13 +1426,11 @@
RATE_CONTROL *const rc = &cpi->rc;
int target;
// TODO(yaowu): replace the "auto_key && 0" below with proper decision logic.
- if ((cm->current_video_frame == 0 ||
- (cpi->frame_flags & FRAMEFLAGS_KEY) ||
- rc->frames_to_key == 0 ||
- (cpi->oxcf.auto_key && 0))) {
+ if ((cm->current_video_frame == 0 || (cpi->frame_flags & FRAMEFLAGS_KEY) ||
+ rc->frames_to_key == 0 || (cpi->oxcf.auto_key && 0))) {
cm->frame_type = KEY_FRAME;
- rc->this_key_frame_forced = cm->current_video_frame != 0 &&
- rc->frames_to_key == 0;
+ rc->this_key_frame_forced =
+ cm->current_video_frame != 0 && rc->frames_to_key == 0;
rc->frames_to_key = cpi->oxcf.key_freq;
rc->kf_boost = DEFAULT_KF_BOOST;
rc->source_alt_ref_active = 0;
@@ -1525,7 +1469,7 @@
}
int vp10_compute_qdelta(const RATE_CONTROL *rc, double qstart, double qtarget,
- vpx_bit_depth_t bit_depth) {
+ vpx_bit_depth_t bit_depth) {
int start_index = rc->worst_quality;
int target_index = rc->worst_quality;
int i;
@@ -1533,29 +1477,27 @@
// Convert the average q value to an index.
for (i = rc->best_quality; i < rc->worst_quality; ++i) {
start_index = i;
- if (vp10_convert_qindex_to_q(i, bit_depth) >= qstart)
- break;
+ if (vp10_convert_qindex_to_q(i, bit_depth) >= qstart) break;
}
// Convert the q target to an index
for (i = rc->best_quality; i < rc->worst_quality; ++i) {
target_index = i;
- if (vp10_convert_qindex_to_q(i, bit_depth) >= qtarget)
- break;
+ if (vp10_convert_qindex_to_q(i, bit_depth) >= qtarget) break;
}
return target_index - start_index;
}
int vp10_compute_qdelta_by_rate(const RATE_CONTROL *rc, FRAME_TYPE frame_type,
- int qindex, double rate_target_ratio,
- vpx_bit_depth_t bit_depth) {
+ int qindex, double rate_target_ratio,
+ vpx_bit_depth_t bit_depth) {
int target_index = rc->worst_quality;
int i;
// Look up the current projected bits per block for the base index
- const int base_bits_per_mb = vp10_rc_bits_per_mb(frame_type, qindex, 1.0,
- bit_depth);
+ const int base_bits_per_mb =
+ vp10_rc_bits_per_mb(frame_type, qindex, 1.0, bit_depth);
// Find the target bits per mb based on the base value and given ratio.
const int target_bits_per_mb = (int)(rate_target_ratio * base_bits_per_mb);
@@ -1572,7 +1514,7 @@
}
void vp10_rc_set_gf_interval_range(const VP10_COMP *const cpi,
- RATE_CONTROL *const rc) {
+ RATE_CONTROL *const rc) {
const VP10EncoderConfig *const oxcf = &cpi->oxcf;
// Special case code for 1 pass fixed Q mode tests
@@ -1614,8 +1556,8 @@
int vbr_max_bits;
rc->avg_frame_bandwidth = (int)(oxcf->target_bandwidth / cpi->framerate);
- rc->min_frame_bandwidth = (int)(rc->avg_frame_bandwidth *
- oxcf->two_pass_vbrmin_section / 100);
+ rc->min_frame_bandwidth =
+ (int)(rc->avg_frame_bandwidth * oxcf->two_pass_vbrmin_section / 100);
rc->min_frame_bandwidth =
VPXMAX(rc->min_frame_bandwidth, FRAME_OVERHEAD_BITS);
@@ -1627,8 +1569,9 @@
// a very high rate is given on the command line or the the rate cannnot
// be acheived because of a user specificed max q (e.g. when the user
// specifies lossless encode.
- vbr_max_bits = (int)(((int64_t)rc->avg_frame_bandwidth *
- oxcf->two_pass_vbrmax_section) / 100);
+ vbr_max_bits =
+ (int)(((int64_t)rc->avg_frame_bandwidth * oxcf->two_pass_vbrmax_section) /
+ 100);
rc->max_frame_bandwidth =
VPXMAX(VPXMAX((cm->MBs * MAX_MB_RATE), MAXRATE_1080P), vbr_max_bits);
@@ -1655,13 +1598,13 @@
// vbr_bits_off_target > 0 means we have extra bits to spend
if (vbr_bits_off_target > 0) {
- *this_frame_target +=
- (vbr_bits_off_target > max_delta) ? max_delta
- : (int)vbr_bits_off_target;
+ *this_frame_target += (vbr_bits_off_target > max_delta)
+ ? max_delta
+ : (int)vbr_bits_off_target;
} else {
- *this_frame_target -=
- (vbr_bits_off_target < -max_delta) ? max_delta
- : (int)-vbr_bits_off_target;
+ *this_frame_target -= (vbr_bits_off_target < -max_delta)
+ ? max_delta
+ : (int)-vbr_bits_off_target;
}
// Fast redistribution of bits arising from massive local undershoot.
@@ -1744,7 +1687,7 @@
cpi->resize_scale_num = 1;
cpi->resize_scale_den = 2;
tot_scale_change = (cpi->resize_scale_den * cpi->resize_scale_den) /
- (cpi->resize_scale_num * cpi->resize_scale_num);
+ (cpi->resize_scale_num * cpi->resize_scale_num);
// Reset buffer level to optimal, update target size.
rc->buffer_level = rc->optimal_buffer_level;
rc->bits_off_target = rc->optimal_buffer_level;
@@ -1754,26 +1697,22 @@
vp10_cyclic_refresh_reset_resize(cpi);
// Get the projected qindex, based on the scaled target frame size (scaled
// so target_bits_per_mb in vp10_rc_regulate_q will be correct target).
- target_bits_per_frame = (resize_now == 1) ?
- rc->this_frame_target * tot_scale_change :
- rc->this_frame_target / tot_scale_change;
+ target_bits_per_frame = (resize_now == 1)
+ ? rc->this_frame_target * tot_scale_change
+ : rc->this_frame_target / tot_scale_change;
active_worst_quality = calc_active_worst_quality_one_pass_cbr(cpi);
- qindex = vp10_rc_regulate_q(cpi,
- target_bits_per_frame,
- rc->best_quality,
- active_worst_quality);
+ qindex = vp10_rc_regulate_q(cpi, target_bits_per_frame, rc->best_quality,
+ active_worst_quality);
// If resize is down, check if projected q index is close to worst_quality,
// and if so, reduce the rate correction factor (since likely can afford
// lower q for resized frame).
- if (resize_now == 1 &&
- qindex > 90 * cpi->rc.worst_quality / 100) {
+ if (resize_now == 1 && qindex > 90 * cpi->rc.worst_quality / 100) {
rc->rate_correction_factors[INTER_NORMAL] *= 0.85;
}
// If resize is back up, check if projected q index is too much above the
// current base_qindex, and if so, reduce the rate correction factor
// (since prefer to keep q for resized frame at least close to previous q).
- if (resize_now == -1 &&
- qindex > 130 * cm->base_qindex / 100) {
+ if (resize_now == -1 && qindex > 130 * cm->base_qindex / 100) {
rc->rate_correction_factors[INTER_NORMAL] *= 0.9;
}
}
diff --git a/vp10/encoder/ratectrl.h b/vp10/encoder/ratectrl.h
index 0b9fd45..f84df83 100644
--- a/vp10/encoder/ratectrl.h
+++ b/vp10/encoder/ratectrl.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_RATECTRL_H_
#define VP10_ENCODER_RATECTRL_H_
@@ -22,11 +21,11 @@
#endif
// Bits Per MB at different Q (Multiplied by 512)
-#define BPER_MB_NORMBITS 9
+#define BPER_MB_NORMBITS 9
-#define MIN_GF_INTERVAL 4
-#define MAX_GF_INTERVAL 16
-#define FIXED_GF_INTERVAL 8 // Used in some testing modes only
+#define MIN_GF_INTERVAL 4
+#define MAX_GF_INTERVAL 16
+#define FIXED_GF_INTERVAL 8 // Used in some testing modes only
typedef enum {
INTER_NORMAL = 0,
@@ -49,25 +48,25 @@
// e.g. 24 => 16/24 = 2/3 of native size. The restriction to 1/16th is
// intended to match the capabilities of the normative scaling filters,
// giving precedence to the up-scaling accuracy.
-static const int frame_scale_factor[FRAME_SCALE_STEPS] = {16, 24};
+static const int frame_scale_factor[FRAME_SCALE_STEPS] = { 16, 24 };
// Multiplier of the target rate to be used as threshold for triggering scaling.
-static const double rate_thresh_mult[FRAME_SCALE_STEPS] = {1.0, 2.0};
+static const double rate_thresh_mult[FRAME_SCALE_STEPS] = { 1.0, 2.0 };
// Scale dependent Rate Correction Factor multipliers. Compensates for the
// greater number of bits per pixel generated in down-scaled frames.
-static const double rcf_mult[FRAME_SCALE_STEPS] = {1.0, 2.0};
+static const double rcf_mult[FRAME_SCALE_STEPS] = { 1.0, 2.0 };
typedef struct {
// Rate targetting variables
- int base_frame_target; // A baseline frame target before adjustment
- // for previous under or over shoot.
- int this_frame_target; // Actual frame target after rc adjustment.
+ int base_frame_target; // A baseline frame target before adjustment
+ // for previous under or over shoot.
+ int this_frame_target; // Actual frame target after rc adjustment.
int projected_frame_size;
int sb64_target_rate;
- int last_q[FRAME_TYPES]; // Separate values for Intra/Inter
- int last_boosted_qindex; // Last boosted GF/KF/ARF q
- int last_kf_qindex; // Q index of the last key frame coded.
+ int last_q[FRAME_TYPES]; // Separate values for Intra/Inter
+ int last_boosted_qindex; // Last boosted GF/KF/ARF q
+ int last_kf_qindex; // Q index of the last key frame coded.
int gfu_boost;
int last_boost;
@@ -149,17 +148,18 @@
struct VP10EncoderConfig;
void vp10_rc_init(const struct VP10EncoderConfig *oxcf, int pass,
- RATE_CONTROL *rc);
+ RATE_CONTROL *rc);
int vp10_estimate_bits_at_q(FRAME_TYPE frame_kind, int q, int mbs,
- double correction_factor,
- vpx_bit_depth_t bit_depth);
+ double correction_factor,
+ vpx_bit_depth_t bit_depth);
double vp10_convert_qindex_to_q(int qindex, vpx_bit_depth_t bit_depth);
void vp10_rc_init_minq_luts(void);
-int vp10_rc_get_default_min_gf_interval(int width, int height, double framerate);
+int vp10_rc_get_default_min_gf_interval(int width, int height,
+ double framerate);
// Note vp10_rc_get_default_max_gf_interval() requires the min_gf_interval to
// be passed in to ensure that the max_gf_interval returned is at least as bis
// as that.
@@ -207,28 +207,27 @@
// Computes frame size bounds.
void vp10_rc_compute_frame_size_bounds(const struct VP10_COMP *cpi,
- int this_frame_target,
- int *frame_under_shoot_limit,
- int *frame_over_shoot_limit);
+ int this_frame_target,
+ int *frame_under_shoot_limit,
+ int *frame_over_shoot_limit);
// Picks q and q bounds given the target for bits
-int vp10_rc_pick_q_and_bounds(const struct VP10_COMP *cpi,
- int *bottom_index,
- int *top_index);
+int vp10_rc_pick_q_and_bounds(const struct VP10_COMP *cpi, int *bottom_index,
+ int *top_index);
// Estimates q to achieve a target bits per frame
int vp10_rc_regulate_q(const struct VP10_COMP *cpi, int target_bits_per_frame,
- int active_best_quality, int active_worst_quality);
+ int active_best_quality, int active_worst_quality);
// Estimates bits per mb for a given qindex and correction factor.
int vp10_rc_bits_per_mb(FRAME_TYPE frame_type, int qindex,
- double correction_factor, vpx_bit_depth_t bit_depth);
+ double correction_factor, vpx_bit_depth_t bit_depth);
// Clamping utilities for bitrate targets for iframes and pframes.
int vp10_rc_clamp_iframe_target_size(const struct VP10_COMP *const cpi,
- int target);
+ int target);
int vp10_rc_clamp_pframe_target_size(const struct VP10_COMP *const cpi,
- int target);
+ int target);
// Utility to set frame_target into the RATE_CONTROL structure
// This function is called only from the vp10_rc_get_..._params() functions.
void vp10_rc_set_frame_target(struct VP10_COMP *cpi, int target);
@@ -236,20 +235,20 @@
// Computes a q delta (in "q index" terms) to get from a starting q value
// to a target q value
int vp10_compute_qdelta(const RATE_CONTROL *rc, double qstart, double qtarget,
- vpx_bit_depth_t bit_depth);
+ vpx_bit_depth_t bit_depth);
// Computes a q delta (in "q index" terms) to get from a starting q value
// to a value that should equate to the given rate ratio.
int vp10_compute_qdelta_by_rate(const RATE_CONTROL *rc, FRAME_TYPE frame_type,
- int qindex, double rate_target_ratio,
- vpx_bit_depth_t bit_depth);
+ int qindex, double rate_target_ratio,
+ vpx_bit_depth_t bit_depth);
int vp10_frame_type_qdelta(const struct VP10_COMP *cpi, int rf_level, int q);
void vp10_rc_update_framerate(struct VP10_COMP *cpi);
void vp10_rc_set_gf_interval_range(const struct VP10_COMP *const cpi,
- RATE_CONTROL *const rc);
+ RATE_CONTROL *const rc);
void vp10_set_target_rate(struct VP10_COMP *cpi);
diff --git a/vp10/encoder/rd.c b/vp10/encoder/rd.c
index acc58e2..f72ebdb 100644
--- a/vp10/encoder/rd.c
+++ b/vp10/encoder/rd.c
@@ -40,8 +40,8 @@
#include "vp10/encoder/rd.h"
#include "vp10/encoder/tokenize.h"
-#define RD_THRESH_POW 1.25
-#define RD_MULT_EPB_RATIO 64
+#define RD_THRESH_POW 1.25
+#define RD_MULT_EPB_RATIO 64
// Factor to weigh the rate for switchable interp filters.
#define SWITCHABLE_INTERP_RATE_FACTOR 1
@@ -73,26 +73,25 @@
for (i = 0; i < INTRA_MODES; ++i)
for (j = 0; j < INTRA_MODES; ++j)
vp10_cost_tokens(cpi->y_mode_costs[i][j], vp10_kf_y_mode_prob[i][j],
- vp10_intra_mode_tree);
+ vp10_intra_mode_tree);
vp10_cost_tokens(cpi->mbmode_cost, fc->y_mode_prob[1], vp10_intra_mode_tree);
for (i = 0; i < INTRA_MODES; ++i)
- vp10_cost_tokens(cpi->intra_uv_mode_cost[i],
- fc->uv_mode_prob[i], vp10_intra_mode_tree);
+ vp10_cost_tokens(cpi->intra_uv_mode_cost[i], fc->uv_mode_prob[i],
+ vp10_intra_mode_tree);
for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i)
vp10_cost_tokens(cpi->switchable_interp_costs[i],
- fc->switchable_interp_prob[i], vp10_switchable_interp_tree);
+ fc->switchable_interp_prob[i],
+ vp10_switchable_interp_tree);
for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
for (j = 0; j < TX_TYPES; ++j)
vp10_cost_tokens(cpi->intra_tx_type_costs[i][j],
- fc->intra_ext_tx_prob[i][j],
- vp10_ext_tx_tree);
+ fc->intra_ext_tx_prob[i][j], vp10_ext_tx_tree);
}
for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
- vp10_cost_tokens(cpi->inter_tx_type_costs[i],
- fc->inter_ext_tx_prob[i],
+ vp10_cost_tokens(cpi->inter_tx_type_costs[i], fc->inter_ext_tx_prob[i],
vp10_ext_tx_tree);
}
}
@@ -108,10 +107,9 @@
for (l = 0; l < BAND_COEFF_CONTEXTS(k); ++l) {
vpx_prob probs[ENTROPY_NODES];
vp10_model_to_full_probs(p[t][i][j][k][l], probs);
- vp10_cost_tokens((int *)c[t][i][j][k][0][l], probs,
- vp10_coef_tree);
+ vp10_cost_tokens((int *)c[t][i][j][k][0][l], probs, vp10_coef_tree);
vp10_cost_tokens_skip((int *)c[t][i][j][k][1][l], probs,
- vp10_coef_tree);
+ vp10_coef_tree);
assert(c[t][i][j][k][0][l][EOB_TOKEN] ==
c[t][i][j][k][1][l][EOB_TOKEN]);
}
@@ -152,28 +150,19 @@
#endif
}
-static const int rd_boost_factor[16] = {
- 64, 32, 32, 32, 24, 16, 12, 12,
- 8, 8, 4, 4, 2, 2, 1, 0
-};
-static const int rd_frame_type_factor[FRAME_UPDATE_TYPES] = {
- 128, 144, 128, 128, 144
-};
+static const int rd_boost_factor[16] = { 64, 32, 32, 32, 24, 16, 12, 12,
+ 8, 8, 4, 4, 2, 2, 1, 0 };
+static const int rd_frame_type_factor[FRAME_UPDATE_TYPES] = { 128, 144, 128,
+ 128, 144 };
int vp10_compute_rd_mult(const VP10_COMP *cpi, int qindex) {
const int64_t q = vp10_dc_quant(qindex, 0, cpi->common.bit_depth);
#if CONFIG_VPX_HIGHBITDEPTH
int64_t rdmult = 0;
switch (cpi->common.bit_depth) {
- case VPX_BITS_8:
- rdmult = 88 * q * q / 24;
- break;
- case VPX_BITS_10:
- rdmult = ROUND_POWER_OF_TWO(88 * q * q / 24, 4);
- break;
- case VPX_BITS_12:
- rdmult = ROUND_POWER_OF_TWO(88 * q * q / 24, 8);
- break;
+ case VPX_BITS_8: rdmult = 88 * q * q / 24; break;
+ case VPX_BITS_10: rdmult = ROUND_POWER_OF_TWO(88 * q * q / 24, 4); break;
+ case VPX_BITS_12: rdmult = ROUND_POWER_OF_TWO(88 * q * q / 24, 8); break;
default:
assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
return -1;
@@ -189,8 +178,7 @@
rdmult = (rdmult * rd_frame_type_factor[frame_type]) >> 7;
rdmult += ((rdmult * rd_boost_factor[boost_index]) >> 7);
}
- if (rdmult < 1)
- rdmult = 1;
+ if (rdmult < 1) rdmult = 1;
return (int)rdmult;
}
@@ -198,21 +186,15 @@
double q;
#if CONFIG_VPX_HIGHBITDEPTH
switch (bit_depth) {
- case VPX_BITS_8:
- q = vp10_dc_quant(qindex, 0, VPX_BITS_8) / 4.0;
- break;
- case VPX_BITS_10:
- q = vp10_dc_quant(qindex, 0, VPX_BITS_10) / 16.0;
- break;
- case VPX_BITS_12:
- q = vp10_dc_quant(qindex, 0, VPX_BITS_12) / 64.0;
- break;
+ case VPX_BITS_8: q = vp10_dc_quant(qindex, 0, VPX_BITS_8) / 4.0; break;
+ case VPX_BITS_10: q = vp10_dc_quant(qindex, 0, VPX_BITS_10) / 16.0; break;
+ case VPX_BITS_12: q = vp10_dc_quant(qindex, 0, VPX_BITS_12) / 64.0; break;
default:
assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
return -1;
}
#else
- (void) bit_depth;
+ (void)bit_depth;
q = vp10_dc_quant(qindex, 0, VPX_BITS_8) / 4.0;
#endif // CONFIG_VPX_HIGHBITDEPTH
// TODO(debargha): Adjust the function below.
@@ -250,7 +232,8 @@
for (segment_id = 0; segment_id < MAX_SEGMENTS; ++segment_id) {
const int qindex =
clamp(vp10_get_qindex(&cm->seg, segment_id, cm->base_qindex) +
- cm->y_dc_delta_q, 0, MAXQ);
+ cm->y_dc_delta_q,
+ 0, MAXQ);
const int q = compute_rd_thresh_factor(qindex, cm->bit_depth);
for (bsize = 0; bsize < BLOCK_SIZES; ++bsize) {
@@ -261,10 +244,9 @@
if (bsize >= BLOCK_8X8) {
for (i = 0; i < MAX_MODES; ++i)
- rd->threshes[segment_id][bsize][i] =
- rd->thresh_mult[i] < thresh_max
- ? rd->thresh_mult[i] * t / 4
- : INT_MAX;
+ rd->threshes[segment_id][bsize][i] = rd->thresh_mult[i] < thresh_max
+ ? rd->thresh_mult[i] * t / 4
+ : INT_MAX;
} else {
for (i = 0; i < MAX_REFS; ++i)
rd->threshes[segment_id][bsize][i] =
@@ -291,7 +273,9 @@
x->errorperbit += (x->errorperbit == 0);
x->select_tx_size = (cpi->sf.tx_size_search_method == USE_LARGESTALL &&
- cm->frame_type != KEY_FRAME) ? 0 : 1;
+ cm->frame_type != KEY_FRAME)
+ ? 0
+ : 1;
set_block_thresholds(cm, rd);
@@ -301,20 +285,20 @@
cm->frame_type == KEY_FRAME) {
for (i = 0; i < PARTITION_CONTEXTS; ++i)
vp10_cost_tokens(cpi->partition_cost[i], cm->fc->partition_prob[i],
- vp10_partition_tree);
+ vp10_partition_tree);
}
fill_mode_costs(cpi);
if (!frame_is_intra_only(cm)) {
- vp10_build_nmv_cost_table(x->nmvjointcost,
- cm->allow_high_precision_mv ? x->nmvcost_hp
- : x->nmvcost,
- &cm->fc->nmvc, cm->allow_high_precision_mv);
+ vp10_build_nmv_cost_table(
+ x->nmvjointcost,
+ cm->allow_high_precision_mv ? x->nmvcost_hp : x->nmvcost, &cm->fc->nmvc,
+ cm->allow_high_precision_mv);
for (i = 0; i < INTER_MODE_CONTEXTS; ++i)
vp10_cost_tokens((int *)cpi->inter_mode_cost[i],
- cm->fc->inter_mode_probs[i], vp10_inter_mode_tree);
+ cm->fc->inter_mode_probs[i], vp10_inter_mode_tree);
}
}
@@ -332,19 +316,15 @@
// where r = exp(-sqrt(2) * x) and x = qpstep / sqrt(variance),
// and H(x) is the binary entropy function.
static const int rate_tab_q10[] = {
- 65536, 6086, 5574, 5275, 5063, 4899, 4764, 4651,
- 4553, 4389, 4255, 4142, 4044, 3958, 3881, 3811,
- 3748, 3635, 3538, 3453, 3376, 3307, 3244, 3186,
- 3133, 3037, 2952, 2877, 2809, 2747, 2690, 2638,
- 2589, 2501, 2423, 2353, 2290, 2232, 2179, 2130,
- 2084, 2001, 1928, 1862, 1802, 1748, 1698, 1651,
- 1608, 1530, 1460, 1398, 1342, 1290, 1243, 1199,
- 1159, 1086, 1021, 963, 911, 864, 821, 781,
- 745, 680, 623, 574, 530, 490, 455, 424,
- 395, 345, 304, 269, 239, 213, 190, 171,
- 154, 126, 104, 87, 73, 61, 52, 44,
- 38, 28, 21, 16, 12, 10, 8, 6,
- 5, 3, 2, 1, 1, 1, 0, 0,
+ 65536, 6086, 5574, 5275, 5063, 4899, 4764, 4651, 4553, 4389, 4255, 4142,
+ 4044, 3958, 3881, 3811, 3748, 3635, 3538, 3453, 3376, 3307, 3244, 3186,
+ 3133, 3037, 2952, 2877, 2809, 2747, 2690, 2638, 2589, 2501, 2423, 2353,
+ 2290, 2232, 2179, 2130, 2084, 2001, 1928, 1862, 1802, 1748, 1698, 1651,
+ 1608, 1530, 1460, 1398, 1342, 1290, 1243, 1199, 1159, 1086, 1021, 963,
+ 911, 864, 821, 781, 745, 680, 623, 574, 530, 490, 455, 424,
+ 395, 345, 304, 269, 239, 213, 190, 171, 154, 126, 104, 87,
+ 73, 61, 52, 44, 38, 28, 21, 16, 12, 10, 8, 6,
+ 5, 3, 2, 1, 1, 1, 0, 0,
};
// Normalized distortion:
// This table models the normalized distortion for a Laplacian source
@@ -354,34 +334,29 @@
// where x = qpstep / sqrt(variance).
// Note the actual distortion is Dn * variance.
static const int dist_tab_q10[] = {
- 0, 0, 1, 1, 1, 2, 2, 2,
- 3, 3, 4, 5, 5, 6, 7, 7,
- 8, 9, 11, 12, 13, 15, 16, 17,
- 18, 21, 24, 26, 29, 31, 34, 36,
- 39, 44, 49, 54, 59, 64, 69, 73,
- 78, 88, 97, 106, 115, 124, 133, 142,
- 151, 167, 184, 200, 215, 231, 245, 260,
- 274, 301, 327, 351, 375, 397, 418, 439,
- 458, 495, 528, 559, 587, 613, 637, 659,
- 680, 717, 749, 777, 801, 823, 842, 859,
- 874, 899, 919, 936, 949, 960, 969, 977,
- 983, 994, 1001, 1006, 1010, 1013, 1015, 1017,
- 1018, 1020, 1022, 1022, 1023, 1023, 1023, 1024,
+ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 4, 5,
+ 5, 6, 7, 7, 8, 9, 11, 12, 13, 15, 16, 17,
+ 18, 21, 24, 26, 29, 31, 34, 36, 39, 44, 49, 54,
+ 59, 64, 69, 73, 78, 88, 97, 106, 115, 124, 133, 142,
+ 151, 167, 184, 200, 215, 231, 245, 260, 274, 301, 327, 351,
+ 375, 397, 418, 439, 458, 495, 528, 559, 587, 613, 637, 659,
+ 680, 717, 749, 777, 801, 823, 842, 859, 874, 899, 919, 936,
+ 949, 960, 969, 977, 983, 994, 1001, 1006, 1010, 1013, 1015, 1017,
+ 1018, 1020, 1022, 1022, 1023, 1023, 1023, 1024,
};
static const int xsq_iq_q10[] = {
- 0, 4, 8, 12, 16, 20, 24, 28,
- 32, 40, 48, 56, 64, 72, 80, 88,
- 96, 112, 128, 144, 160, 176, 192, 208,
- 224, 256, 288, 320, 352, 384, 416, 448,
- 480, 544, 608, 672, 736, 800, 864, 928,
- 992, 1120, 1248, 1376, 1504, 1632, 1760, 1888,
- 2016, 2272, 2528, 2784, 3040, 3296, 3552, 3808,
- 4064, 4576, 5088, 5600, 6112, 6624, 7136, 7648,
- 8160, 9184, 10208, 11232, 12256, 13280, 14304, 15328,
- 16352, 18400, 20448, 22496, 24544, 26592, 28640, 30688,
- 32736, 36832, 40928, 45024, 49120, 53216, 57312, 61408,
- 65504, 73696, 81888, 90080, 98272, 106464, 114656, 122848,
- 131040, 147424, 163808, 180192, 196576, 212960, 229344, 245728,
+ 0, 4, 8, 12, 16, 20, 24, 28, 32,
+ 40, 48, 56, 64, 72, 80, 88, 96, 112,
+ 128, 144, 160, 176, 192, 208, 224, 256, 288,
+ 320, 352, 384, 416, 448, 480, 544, 608, 672,
+ 736, 800, 864, 928, 992, 1120, 1248, 1376, 1504,
+ 1632, 1760, 1888, 2016, 2272, 2528, 2784, 3040, 3296,
+ 3552, 3808, 4064, 4576, 5088, 5600, 6112, 6624, 7136,
+ 7648, 8160, 9184, 10208, 11232, 12256, 13280, 14304, 15328,
+ 16352, 18400, 20448, 22496, 24544, 26592, 28640, 30688, 32736,
+ 36832, 40928, 45024, 49120, 53216, 57312, 61408, 65504, 73696,
+ 81888, 90080, 98272, 106464, 114656, 122848, 131040, 147424, 163808,
+ 180192, 196576, 212960, 229344, 245728,
};
const int tmp = (xsq_q10 >> 2) + 8;
const int k = get_msb(tmp) - 3;
@@ -394,8 +369,8 @@
}
void vp10_model_rd_from_var_lapndz(unsigned int var, unsigned int n_log2,
- unsigned int qstep, int *rate,
- int64_t *dist) {
+ unsigned int qstep, int *rate,
+ int64_t *dist) {
// This function models the rate and distortion for a Laplacian
// source with given variance when quantized with a uniform quantizer
// with given stepsize. The closed form expressions are in:
@@ -418,9 +393,9 @@
}
void vp10_get_entropy_contexts(BLOCK_SIZE bsize, TX_SIZE tx_size,
- const struct macroblockd_plane *pd,
- ENTROPY_CONTEXT t_above[16],
- ENTROPY_CONTEXT t_left[16]) {
+ const struct macroblockd_plane *pd,
+ ENTROPY_CONTEXT t_above[16],
+ ENTROPY_CONTEXT t_left[16]) {
const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
const int num_4x4_w = num_4x4_blocks_wide_lookup[plane_bsize];
const int num_4x4_h = num_4x4_blocks_high_lookup[plane_bsize];
@@ -451,15 +426,12 @@
for (i = 0; i < num_4x4_h; i += 8)
t_left[i] = !!*(const uint64_t *)&left[i];
break;
- default:
- assert(0 && "Invalid transform size.");
- break;
+ default: assert(0 && "Invalid transform size."); break;
}
}
-void vp10_mv_pred(VP10_COMP *cpi, MACROBLOCK *x,
- uint8_t *ref_y_buffer, int ref_y_stride,
- int ref_frame, BLOCK_SIZE block_size) {
+void vp10_mv_pred(VP10_COMP *cpi, MACROBLOCK *x, uint8_t *ref_y_buffer,
+ int ref_y_stride, int ref_frame, BLOCK_SIZE block_size) {
int i;
int zero_seen = 0;
int best_index = 0;
@@ -469,9 +441,9 @@
int near_same_nearest;
uint8_t *src_y_ptr = x->plane[0].src.buf;
uint8_t *ref_y_ptr;
- const int num_mv_refs = MAX_MV_REF_CANDIDATES +
- (cpi->sf.adaptive_motion_search &&
- block_size < x->max_partition_size);
+ const int num_mv_refs =
+ MAX_MV_REF_CANDIDATES +
+ (cpi->sf.adaptive_motion_search && block_size < x->max_partition_size);
MV pred_mv[3];
pred_mv[0] = x->mbmi_ext->ref_mvs[ref_frame][0].as_mv;
@@ -479,25 +451,22 @@
pred_mv[2] = x->pred_mv[ref_frame];
assert(num_mv_refs <= (int)(sizeof(pred_mv) / sizeof(pred_mv[0])));
- near_same_nearest =
- x->mbmi_ext->ref_mvs[ref_frame][0].as_int ==
- x->mbmi_ext->ref_mvs[ref_frame][1].as_int;
+ near_same_nearest = x->mbmi_ext->ref_mvs[ref_frame][0].as_int ==
+ x->mbmi_ext->ref_mvs[ref_frame][1].as_int;
// Get the sad for each candidate reference mv.
for (i = 0; i < num_mv_refs; ++i) {
const MV *this_mv = &pred_mv[i];
int fp_row, fp_col;
- if (i == 1 && near_same_nearest)
- continue;
+ if (i == 1 && near_same_nearest) continue;
fp_row = (this_mv->row + 3 + (this_mv->row >= 0)) >> 3;
fp_col = (this_mv->col + 3 + (this_mv->col >= 0)) >> 3;
max_mv = VPXMAX(max_mv, VPXMAX(abs(this_mv->row), abs(this_mv->col)) >> 3);
- if (fp_row ==0 && fp_col == 0 && zero_seen)
- continue;
- zero_seen |= (fp_row ==0 && fp_col == 0);
+ if (fp_row == 0 && fp_col == 0 && zero_seen) continue;
+ zero_seen |= (fp_row == 0 && fp_col == 0);
- ref_y_ptr =&ref_y_buffer[ref_y_stride * fp_row + fp_col];
+ ref_y_ptr = &ref_y_buffer[ref_y_stride * fp_row + fp_col];
// Find sad for current vector.
this_sad = cpi->fn_ptr[block_size].sdf(src_y_ptr, x->plane[0].src.stride,
ref_y_ptr, ref_y_stride);
@@ -515,11 +484,10 @@
}
void vp10_setup_pred_block(const MACROBLOCKD *xd,
- struct buf_2d dst[MAX_MB_PLANE],
- const YV12_BUFFER_CONFIG *src,
- int mi_row, int mi_col,
- const struct scale_factors *scale,
- const struct scale_factors *scale_uv) {
+ struct buf_2d dst[MAX_MB_PLANE],
+ const YV12_BUFFER_CONFIG *src, int mi_row,
+ int mi_col, const struct scale_factors *scale,
+ const struct scale_factors *scale_uv) {
int i;
dst[0].buf = src->y_buffer;
@@ -530,33 +498,33 @@
for (i = 0; i < MAX_MB_PLANE; ++i) {
setup_pred_plane(dst + i, dst[i].buf, dst[i].stride, mi_row, mi_col,
- i ? scale_uv : scale,
- xd->plane[i].subsampling_x, xd->plane[i].subsampling_y);
+ i ? scale_uv : scale, xd->plane[i].subsampling_x,
+ xd->plane[i].subsampling_y);
}
}
-int vp10_raster_block_offset(BLOCK_SIZE plane_bsize,
- int raster_block, int stride) {
+int vp10_raster_block_offset(BLOCK_SIZE plane_bsize, int raster_block,
+ int stride) {
const int bw = b_width_log2_lookup[plane_bsize];
const int y = 4 * (raster_block >> bw);
const int x = 4 * (raster_block & ((1 << bw) - 1));
return y * stride + x;
}
-int16_t* vp10_raster_block_offset_int16(BLOCK_SIZE plane_bsize,
- int raster_block, int16_t *base) {
+int16_t *vp10_raster_block_offset_int16(BLOCK_SIZE plane_bsize,
+ int raster_block, int16_t *base) {
const int stride = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
return base + vp10_raster_block_offset(plane_bsize, raster_block, stride);
}
YV12_BUFFER_CONFIG *vp10_get_scaled_ref_frame(const VP10_COMP *cpi,
- int ref_frame) {
+ int ref_frame) {
const VP10_COMMON *const cm = &cpi->common;
const int scaled_idx = cpi->scaled_ref_idx[ref_frame - 1];
const int ref_idx = get_ref_frame_buf_idx(cpi, ref_frame);
- return
- (scaled_idx != ref_idx && scaled_idx != INVALID_IDX) ?
- &cm->buffer_pool->frame_bufs[scaled_idx].buf : NULL;
+ return (scaled_idx != ref_idx && scaled_idx != INVALID_IDX)
+ ? &cm->buffer_pool->frame_bufs[scaled_idx].buf
+ : NULL;
}
int vp10_get_switchable_rate(const VP10_COMP *cpi,
@@ -564,7 +532,7 @@
const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
const int ctx = vp10_get_pred_context_switchable_interp(xd);
return SWITCHABLE_INTERP_RATE_FACTOR *
- cpi->switchable_interp_costs[ctx][mbmi->interp_filter];
+ cpi->switchable_interp_costs[ctx][mbmi->interp_filter];
}
void vp10_set_rd_speed_thresholds(VP10_COMP *cpi) {
@@ -613,7 +581,7 @@
rd->thresh_mult[THR_H_PRED] += 2000;
rd->thresh_mult[THR_V_PRED] += 2000;
- rd->thresh_mult[THR_D45_PRED ] += 2500;
+ rd->thresh_mult[THR_D45_PRED] += 2500;
rd->thresh_mult[THR_D135_PRED] += 2500;
rd->thresh_mult[THR_D117_PRED] += 2500;
rd->thresh_mult[THR_D153_PRED] += 2500;
@@ -622,16 +590,17 @@
}
void vp10_set_rd_speed_thresholds_sub8x8(VP10_COMP *cpi) {
- static const int thresh_mult[2][MAX_REFS] =
- {{2500, 2500, 2500, 4500, 4500, 2500},
- {2000, 2000, 2000, 4000, 4000, 2000}};
+ static const int thresh_mult[2][MAX_REFS] = {
+ { 2500, 2500, 2500, 4500, 4500, 2500 },
+ { 2000, 2000, 2000, 4000, 4000, 2000 }
+ };
RD_OPT *const rd = &cpi->rd;
const int idx = cpi->oxcf.mode == BEST;
memcpy(rd->thresh_mult_sub8x8, thresh_mult[idx], sizeof(thresh_mult[idx]));
}
void vp10_update_rd_thresh_fact(int (*factor_buf)[MAX_MODES], int rd_thresh,
- int bsize, int best_mode_index) {
+ int bsize, int best_mode_index) {
if (rd_thresh > 0) {
const int top_mode = bsize < BLOCK_8X8 ? MAX_REFS : MAX_MODES;
int mode;
@@ -652,16 +621,13 @@
}
int vp10_get_intra_cost_penalty(int qindex, int qdelta,
- vpx_bit_depth_t bit_depth) {
+ vpx_bit_depth_t bit_depth) {
const int q = vp10_dc_quant(qindex, qdelta, bit_depth);
#if CONFIG_VPX_HIGHBITDEPTH
switch (bit_depth) {
- case VPX_BITS_8:
- return 20 * q;
- case VPX_BITS_10:
- return 5 * q;
- case VPX_BITS_12:
- return ROUND_POWER_OF_TWO(5 * q, 2);
+ case VPX_BITS_8: return 20 * q;
+ case VPX_BITS_10: return 5 * q;
+ case VPX_BITS_12: return ROUND_POWER_OF_TWO(5 * q, 2);
default:
assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
return -1;
@@ -670,4 +636,3 @@
return 20 * q;
#endif // CONFIG_VPX_HIGHBITDEPTH
}
-
diff --git a/vp10/encoder/rd.h b/vp10/encoder/rd.h
index cd58bf8..f9f811c 100644
--- a/vp10/encoder/rd.h
+++ b/vp10/encoder/rd.h
@@ -22,22 +22,21 @@
extern "C" {
#endif
-#define RDDIV_BITS 7
+#define RDDIV_BITS 7
-#define RDCOST(RM, DM, R, D) \
- (((128 + ((int64_t)R) * (RM)) >> 8) + (D << DM))
-#define QIDX_SKIP_THRESH 115
+#define RDCOST(RM, DM, R, D) (((128 + ((int64_t)R) * (RM)) >> 8) + (D << DM))
+#define QIDX_SKIP_THRESH 115
-#define MV_COST_WEIGHT 108
-#define MV_COST_WEIGHT_SUB 120
+#define MV_COST_WEIGHT 108
+#define MV_COST_WEIGHT_SUB 120
#define INVALID_MV 0x80008000
#define MAX_MODES 30
-#define MAX_REFS 6
+#define MAX_REFS 6
#define RD_THRESH_MAX_FACT 64
-#define RD_THRESH_INC 1
+#define RD_THRESH_INC 1
// This enumerator type needs to be kept aligned with the mode order in
// const MODE_DEFINITION vp10_mode_order[MAX_MODES] used in the rd code.
@@ -130,57 +129,55 @@
void vp10_initialize_rd_consts(struct VP10_COMP *cpi);
-void vp10_initialize_me_consts(struct VP10_COMP *cpi,
- MACROBLOCK *x, int qindex);
+void vp10_initialize_me_consts(struct VP10_COMP *cpi, MACROBLOCK *x,
+ int qindex);
void vp10_model_rd_from_var_lapndz(unsigned int var, unsigned int n,
- unsigned int qstep, int *rate,
- int64_t *dist);
+ unsigned int qstep, int *rate,
+ int64_t *dist);
int vp10_get_switchable_rate(const struct VP10_COMP *cpi,
- const MACROBLOCKD *const xd);
+ const MACROBLOCKD *const xd);
-int vp10_raster_block_offset(BLOCK_SIZE plane_bsize,
- int raster_block, int stride);
+int vp10_raster_block_offset(BLOCK_SIZE plane_bsize, int raster_block,
+ int stride);
-int16_t* vp10_raster_block_offset_int16(BLOCK_SIZE plane_bsize,
- int raster_block, int16_t *base);
+int16_t *vp10_raster_block_offset_int16(BLOCK_SIZE plane_bsize,
+ int raster_block, int16_t *base);
YV12_BUFFER_CONFIG *vp10_get_scaled_ref_frame(const struct VP10_COMP *cpi,
- int ref_frame);
+ int ref_frame);
void vp10_init_me_luts(void);
void vp10_get_entropy_contexts(BLOCK_SIZE bsize, TX_SIZE tx_size,
- const struct macroblockd_plane *pd,
- ENTROPY_CONTEXT t_above[16],
- ENTROPY_CONTEXT t_left[16]);
+ const struct macroblockd_plane *pd,
+ ENTROPY_CONTEXT t_above[16],
+ ENTROPY_CONTEXT t_left[16]);
void vp10_set_rd_speed_thresholds(struct VP10_COMP *cpi);
void vp10_set_rd_speed_thresholds_sub8x8(struct VP10_COMP *cpi);
void vp10_update_rd_thresh_fact(int (*fact)[MAX_MODES], int rd_thresh,
- int bsize, int best_mode_index);
+ int bsize, int best_mode_index);
static INLINE int rd_less_than_thresh(int64_t best_rd, int thresh,
int thresh_fact) {
- return best_rd < ((int64_t)thresh * thresh_fact >> 5) || thresh == INT_MAX;
+ return best_rd < ((int64_t)thresh * thresh_fact >> 5) || thresh == INT_MAX;
}
-void vp10_mv_pred(struct VP10_COMP *cpi, MACROBLOCK *x,
- uint8_t *ref_y_buffer, int ref_y_stride,
- int ref_frame, BLOCK_SIZE block_size);
+void vp10_mv_pred(struct VP10_COMP *cpi, MACROBLOCK *x, uint8_t *ref_y_buffer,
+ int ref_y_stride, int ref_frame, BLOCK_SIZE block_size);
void vp10_setup_pred_block(const MACROBLOCKD *xd,
- struct buf_2d dst[MAX_MB_PLANE],
- const YV12_BUFFER_CONFIG *src,
- int mi_row, int mi_col,
- const struct scale_factors *scale,
- const struct scale_factors *scale_uv);
+ struct buf_2d dst[MAX_MB_PLANE],
+ const YV12_BUFFER_CONFIG *src, int mi_row,
+ int mi_col, const struct scale_factors *scale,
+ const struct scale_factors *scale_uv);
int vp10_get_intra_cost_penalty(int qindex, int qdelta,
- vpx_bit_depth_t bit_depth);
+ vpx_bit_depth_t bit_depth);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/encoder/rdopt.c b/vp10/encoder/rdopt.c
index 6b59161..ca71429 100644
--- a/vp10/encoder/rdopt.c
+++ b/vp10/encoder/rdopt.c
@@ -42,17 +42,17 @@
#include "vp10/encoder/rdopt.h"
#include "vp10/encoder/aq_variance.h"
-#define LAST_FRAME_MODE_MASK ((1 << GOLDEN_FRAME) | (1 << ALTREF_FRAME) | \
- (1 << INTRA_FRAME))
-#define GOLDEN_FRAME_MODE_MASK ((1 << LAST_FRAME) | (1 << ALTREF_FRAME) | \
- (1 << INTRA_FRAME))
-#define ALT_REF_MODE_MASK ((1 << LAST_FRAME) | (1 << GOLDEN_FRAME) | \
- (1 << INTRA_FRAME))
+#define LAST_FRAME_MODE_MASK \
+ ((1 << GOLDEN_FRAME) | (1 << ALTREF_FRAME) | (1 << INTRA_FRAME))
+#define GOLDEN_FRAME_MODE_MASK \
+ ((1 << LAST_FRAME) | (1 << ALTREF_FRAME) | (1 << INTRA_FRAME))
+#define ALT_REF_MODE_MASK \
+ ((1 << LAST_FRAME) | (1 << GOLDEN_FRAME) | (1 << INTRA_FRAME))
-#define SECOND_REF_FRAME_MASK ((1 << ALTREF_FRAME) | 0x01)
+#define SECOND_REF_FRAME_MASK ((1 << ALTREF_FRAME) | 0x01)
-#define MIN_EARLY_TERM_INDEX 3
-#define NEW_MV_DISCOUNT_FACTOR 8
+#define MIN_EARLY_TERM_INDEX 3
+#define NEW_MV_DISCOUNT_FACTOR 8
const double ext_tx_th = 0.99;
@@ -61,9 +61,7 @@
MV_REFERENCE_FRAME ref_frame[2];
} MODE_DEFINITION;
-typedef struct {
- MV_REFERENCE_FRAME ref_frame[2];
-} REF_DEFINITION;
+typedef struct { MV_REFERENCE_FRAME ref_frame[2]; } REF_DEFINITION;
struct rdcost_block_args {
MACROBLOCK *x;
@@ -82,95 +80,91 @@
#define LAST_NEW_MV_INDEX 6
static const MODE_DEFINITION vp10_mode_order[MAX_MODES] = {
- {NEARESTMV, {LAST_FRAME, NONE}},
- {NEARESTMV, {ALTREF_FRAME, NONE}},
- {NEARESTMV, {GOLDEN_FRAME, NONE}},
+ { NEARESTMV, { LAST_FRAME, NONE } },
+ { NEARESTMV, { ALTREF_FRAME, NONE } },
+ { NEARESTMV, { GOLDEN_FRAME, NONE } },
- {DC_PRED, {INTRA_FRAME, NONE}},
+ { DC_PRED, { INTRA_FRAME, NONE } },
- {NEWMV, {LAST_FRAME, NONE}},
- {NEWMV, {ALTREF_FRAME, NONE}},
- {NEWMV, {GOLDEN_FRAME, NONE}},
+ { NEWMV, { LAST_FRAME, NONE } },
+ { NEWMV, { ALTREF_FRAME, NONE } },
+ { NEWMV, { GOLDEN_FRAME, NONE } },
- {NEARMV, {LAST_FRAME, NONE}},
- {NEARMV, {ALTREF_FRAME, NONE}},
- {NEARMV, {GOLDEN_FRAME, NONE}},
+ { NEARMV, { LAST_FRAME, NONE } },
+ { NEARMV, { ALTREF_FRAME, NONE } },
+ { NEARMV, { GOLDEN_FRAME, NONE } },
- {ZEROMV, {LAST_FRAME, NONE}},
- {ZEROMV, {GOLDEN_FRAME, NONE}},
- {ZEROMV, {ALTREF_FRAME, NONE}},
+ { ZEROMV, { LAST_FRAME, NONE } },
+ { ZEROMV, { GOLDEN_FRAME, NONE } },
+ { ZEROMV, { ALTREF_FRAME, NONE } },
- {NEARESTMV, {LAST_FRAME, ALTREF_FRAME}},
- {NEARESTMV, {GOLDEN_FRAME, ALTREF_FRAME}},
+ { NEARESTMV, { LAST_FRAME, ALTREF_FRAME } },
+ { NEARESTMV, { GOLDEN_FRAME, ALTREF_FRAME } },
- {TM_PRED, {INTRA_FRAME, NONE}},
+ { TM_PRED, { INTRA_FRAME, NONE } },
- {NEARMV, {LAST_FRAME, ALTREF_FRAME}},
- {NEWMV, {LAST_FRAME, ALTREF_FRAME}},
- {NEARMV, {GOLDEN_FRAME, ALTREF_FRAME}},
- {NEWMV, {GOLDEN_FRAME, ALTREF_FRAME}},
+ { NEARMV, { LAST_FRAME, ALTREF_FRAME } },
+ { NEWMV, { LAST_FRAME, ALTREF_FRAME } },
+ { NEARMV, { GOLDEN_FRAME, ALTREF_FRAME } },
+ { NEWMV, { GOLDEN_FRAME, ALTREF_FRAME } },
- {ZEROMV, {LAST_FRAME, ALTREF_FRAME}},
- {ZEROMV, {GOLDEN_FRAME, ALTREF_FRAME}},
+ { ZEROMV, { LAST_FRAME, ALTREF_FRAME } },
+ { ZEROMV, { GOLDEN_FRAME, ALTREF_FRAME } },
- {H_PRED, {INTRA_FRAME, NONE}},
- {V_PRED, {INTRA_FRAME, NONE}},
- {D135_PRED, {INTRA_FRAME, NONE}},
- {D207_PRED, {INTRA_FRAME, NONE}},
- {D153_PRED, {INTRA_FRAME, NONE}},
- {D63_PRED, {INTRA_FRAME, NONE}},
- {D117_PRED, {INTRA_FRAME, NONE}},
- {D45_PRED, {INTRA_FRAME, NONE}},
+ { H_PRED, { INTRA_FRAME, NONE } },
+ { V_PRED, { INTRA_FRAME, NONE } },
+ { D135_PRED, { INTRA_FRAME, NONE } },
+ { D207_PRED, { INTRA_FRAME, NONE } },
+ { D153_PRED, { INTRA_FRAME, NONE } },
+ { D63_PRED, { INTRA_FRAME, NONE } },
+ { D117_PRED, { INTRA_FRAME, NONE } },
+ { D45_PRED, { INTRA_FRAME, NONE } },
};
static const REF_DEFINITION vp10_ref_order[MAX_REFS] = {
- {{LAST_FRAME, NONE}},
- {{GOLDEN_FRAME, NONE}},
- {{ALTREF_FRAME, NONE}},
- {{LAST_FRAME, ALTREF_FRAME}},
- {{GOLDEN_FRAME, ALTREF_FRAME}},
- {{INTRA_FRAME, NONE}},
+ { { LAST_FRAME, NONE } }, { { GOLDEN_FRAME, NONE } },
+ { { ALTREF_FRAME, NONE } }, { { LAST_FRAME, ALTREF_FRAME } },
+ { { GOLDEN_FRAME, ALTREF_FRAME } }, { { INTRA_FRAME, NONE } },
};
static INLINE int write_uniform_cost(int n, int v) {
int l = get_unsigned_bits(n), m = (1 << l) - n;
- if (l == 0)
- return 0;
+ if (l == 0) return 0;
if (v < m)
return (l - 1) * vp10_cost_bit(128, 0);
else
return l * vp10_cost_bit(128, 0);
}
-static void swap_block_ptr(MACROBLOCK *x, PICK_MODE_CONTEXT *ctx,
- int m, int n, int min_plane, int max_plane) {
+static void swap_block_ptr(MACROBLOCK *x, PICK_MODE_CONTEXT *ctx, int m, int n,
+ int min_plane, int max_plane) {
int i;
for (i = min_plane; i < max_plane; ++i) {
struct macroblock_plane *const p = &x->plane[i];
struct macroblockd_plane *const pd = &x->e_mbd.plane[i];
- p->coeff = ctx->coeff_pbuf[i][m];
- p->qcoeff = ctx->qcoeff_pbuf[i][m];
+ p->coeff = ctx->coeff_pbuf[i][m];
+ p->qcoeff = ctx->qcoeff_pbuf[i][m];
pd->dqcoeff = ctx->dqcoeff_pbuf[i][m];
- p->eobs = ctx->eobs_pbuf[i][m];
+ p->eobs = ctx->eobs_pbuf[i][m];
- ctx->coeff_pbuf[i][m] = ctx->coeff_pbuf[i][n];
- ctx->qcoeff_pbuf[i][m] = ctx->qcoeff_pbuf[i][n];
+ ctx->coeff_pbuf[i][m] = ctx->coeff_pbuf[i][n];
+ ctx->qcoeff_pbuf[i][m] = ctx->qcoeff_pbuf[i][n];
ctx->dqcoeff_pbuf[i][m] = ctx->dqcoeff_pbuf[i][n];
- ctx->eobs_pbuf[i][m] = ctx->eobs_pbuf[i][n];
+ ctx->eobs_pbuf[i][m] = ctx->eobs_pbuf[i][n];
- ctx->coeff_pbuf[i][n] = p->coeff;
- ctx->qcoeff_pbuf[i][n] = p->qcoeff;
+ ctx->coeff_pbuf[i][n] = p->coeff;
+ ctx->qcoeff_pbuf[i][n] = p->qcoeff;
ctx->dqcoeff_pbuf[i][n] = pd->dqcoeff;
- ctx->eobs_pbuf[i][n] = p->eobs;
+ ctx->eobs_pbuf[i][n] = p->eobs;
}
}
-static void model_rd_for_sb(VP10_COMP *cpi, BLOCK_SIZE bsize,
- MACROBLOCK *x, MACROBLOCKD *xd,
- int *out_rate_sum, int64_t *out_dist_sum,
- int *skip_txfm_sb, int64_t *skip_sse_sb) {
+static void model_rd_for_sb(VP10_COMP *cpi, BLOCK_SIZE bsize, MACROBLOCK *x,
+ MACROBLOCKD *xd, int *out_rate_sum,
+ int64_t *out_dist_sum, int *skip_txfm_sb,
+ int64_t *skip_sse_sb) {
// Note our transform coeffs are 8 times an orthogonal transform.
// Hence quantizer step is also 8 times. To get effective quantizer
// we need to divide by 8 before sending to modeling function.
@@ -188,10 +182,9 @@
int64_t dist;
const int dequant_shift =
#if CONFIG_VPX_HIGHBITDEPTH
- (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ?
- xd->bd - 5 :
+ (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ? xd->bd - 5 :
#endif // CONFIG_VPX_HIGHBITDEPTH
- 3;
+ 3;
x->pred_sse[ref] = 0;
@@ -222,8 +215,8 @@
int block_idx = (idy << 1) + idx;
int low_err_skip = 0;
- var = cpi->fn_ptr[unit_size].vf(src, p->src.stride,
- dst, pd->dst.stride, &sse);
+ var = cpi->fn_ptr[unit_size].vf(src, p->src.stride, dst, pd->dst.stride,
+ &sse);
x->bsse[(i << 2) + block_idx] = sse;
sum_sse += sse;
@@ -243,11 +236,9 @@
}
}
- if (skip_flag && !low_err_skip)
- skip_flag = 0;
+ if (skip_flag && !low_err_skip) skip_flag = 0;
- if (i == 0)
- x->pred_sse[ref] += sse;
+ if (i == 0) x->pred_sse[ref] += sse;
}
}
@@ -268,8 +259,8 @@
dist_sum += dist;
} else {
vp10_model_rd_from_var_lapndz(sum_sse, num_pels_log2_lookup[bs],
- pd->dequant[1] >> dequant_shift,
- &rate, &dist);
+ pd->dequant[1] >> dequant_shift, &rate,
+ &dist);
rate_sum += rate;
dist_sum += dist;
}
@@ -282,13 +273,13 @@
}
int64_t vp10_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff,
- intptr_t block_size, int64_t *ssz) {
+ intptr_t block_size, int64_t *ssz) {
int i;
int64_t error = 0, sqcoeff = 0;
for (i = 0; i < block_size; i++) {
const int diff = coeff[i] - dqcoeff[i];
- error += diff * diff;
+ error += diff * diff;
sqcoeff += coeff[i] * coeff[i];
}
@@ -297,13 +288,13 @@
}
int64_t vp10_block_error_fp_c(const int16_t *coeff, const int16_t *dqcoeff,
- int block_size) {
+ int block_size) {
int i;
int64_t error = 0;
for (i = 0; i < block_size; i++) {
const int diff = coeff[i] - dqcoeff[i];
- error += diff * diff;
+ error += diff * diff;
}
return error;
@@ -311,9 +302,8 @@
#if CONFIG_VPX_HIGHBITDEPTH
int64_t vp10_highbd_block_error_c(const tran_low_t *coeff,
- const tran_low_t *dqcoeff,
- intptr_t block_size,
- int64_t *ssz, int bd) {
+ const tran_low_t *dqcoeff,
+ intptr_t block_size, int64_t *ssz, int bd) {
int i;
int64_t error = 0, sqcoeff = 0;
int shift = 2 * (bd - 8);
@@ -321,7 +311,7 @@
for (i = 0; i < block_size; i++) {
const int64_t diff = coeff[i] - dqcoeff[i];
- error += diff * diff;
+ error += diff * diff;
sqcoeff += (int64_t)coeff[i] * (int64_t)coeff[i];
}
assert(error >= 0 && sqcoeff >= 0);
@@ -339,17 +329,14 @@
* 16th coefficient in a 4x4 block or the 64th coefficient in a 8x8 block,
* were non-zero). */
static const int16_t band_counts[TX_SIZES][8] = {
- { 1, 2, 3, 4, 3, 16 - 13, 0 },
- { 1, 2, 3, 4, 11, 64 - 21, 0 },
- { 1, 2, 3, 4, 11, 256 - 21, 0 },
+ { 1, 2, 3, 4, 3, 16 - 13, 0 },
+ { 1, 2, 3, 4, 11, 64 - 21, 0 },
+ { 1, 2, 3, 4, 11, 256 - 21, 0 },
{ 1, 2, 3, 4, 11, 1024 - 21, 0 },
};
-static int cost_coeffs(MACROBLOCK *x,
- int plane, int block,
- ENTROPY_CONTEXT *A, ENTROPY_CONTEXT *L,
- TX_SIZE tx_size,
- const int16_t *scan, const int16_t *nb,
- int use_fast_coef_costing) {
+static int cost_coeffs(MACROBLOCK *x, int plane, int block, ENTROPY_CONTEXT *A,
+ ENTROPY_CONTEXT *L, TX_SIZE tx_size, const int16_t *scan,
+ const int16_t *nb, int use_fast_coef_costing) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
const struct macroblock_plane *p = &x->plane[plane];
@@ -358,8 +345,8 @@
const int16_t *band_count = &band_counts[tx_size][1];
const int eob = p->eobs[block];
const tran_low_t *const qcoeff = BLOCK_OFFSET(p->qcoeff, block);
- unsigned int (*token_costs)[2][COEFF_CONTEXTS][ENTROPY_TOKENS] =
- x->token_costs[tx_size][type][is_inter_block(mbmi)];
+ unsigned int(*token_costs)[2][COEFF_CONTEXTS][ENTROPY_TOKENS] =
+ x->token_costs[tx_size][type][is_inter_block(mbmi)];
uint8_t token_cache[32 * 32];
int pt = combine_entropy_contexts(*A, *L);
int c, cost;
@@ -386,7 +373,7 @@
EXTRABIT e;
vp10_get_token_extra(v, &prev_t, &e);
cost = (*token_costs)[0][pt][prev_t] +
- vp10_get_cost(prev_t, e, cat6_high_cost);
+ vp10_get_cost(prev_t, e, cat6_high_cost);
token_cache[0] = vp10_pt_energy_class[prev_t];
++token_costs;
@@ -400,11 +387,11 @@
vp10_get_token_extra(v, &t, &e);
if (use_fast_coef_costing) {
cost += (*token_costs)[!prev_t][!prev_t][t] +
- vp10_get_cost(t, e, cat6_high_cost);
+ vp10_get_cost(t, e, cat6_high_cost);
} else {
pt = get_coef_context(nb, token_cache, c);
cost += (*token_costs)[!prev_t][pt][t] +
- vp10_get_cost(t, e, cat6_high_cost);
+ vp10_get_cost(t, e, cat6_high_cost);
token_cache[rc] = vp10_pt_energy_class[t];
}
prev_t = t;
@@ -434,7 +421,7 @@
static void dist_block(MACROBLOCK *x, int plane, int block, TX_SIZE tx_size,
int64_t *out_dist, int64_t *out_sse) {
const int ss_txfrm_size = tx_size << 1;
- MACROBLOCKD* const xd = &x->e_mbd;
+ MACROBLOCKD *const xd = &x->e_mbd;
const struct macroblock_plane *const p = &x->plane[plane];
const struct macroblockd_plane *const pd = &xd->plane[plane];
int64_t this_sse;
@@ -444,25 +431,24 @@
#if CONFIG_VPX_HIGHBITDEPTH
const int bd = (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ? xd->bd : 8;
*out_dist = vp10_highbd_block_error(coeff, dqcoeff, 16 << ss_txfrm_size,
- &this_sse, bd) >> shift;
+ &this_sse, bd) >>
+ shift;
#else
- *out_dist = vp10_block_error(coeff, dqcoeff, 16 << ss_txfrm_size,
- &this_sse) >> shift;
+ *out_dist =
+ vp10_block_error(coeff, dqcoeff, 16 << ss_txfrm_size, &this_sse) >> shift;
#endif // CONFIG_VPX_HIGHBITDEPTH
*out_sse = this_sse >> shift;
}
static int rate_block(int plane, int block, int blk_row, int blk_col,
- TX_SIZE tx_size, struct rdcost_block_args* args) {
+ TX_SIZE tx_size, struct rdcost_block_args *args) {
return cost_coeffs(args->x, plane, block, args->t_above + blk_col,
- args->t_left + blk_row, tx_size,
- args->so->scan, args->so->neighbors,
- args->use_fast_coef_costing);
+ args->t_left + blk_row, tx_size, args->so->scan,
+ args->so->neighbors, args->use_fast_coef_costing);
}
static void block_rd_txfm(int plane, int block, int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize,
- TX_SIZE tx_size, void *arg) {
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size, void *arg) {
struct rdcost_block_args *args = arg;
MACROBLOCK *const x = args->x;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -472,29 +458,27 @@
int64_t dist;
int64_t sse;
- if (args->exit_early)
- return;
+ if (args->exit_early) return;
if (!is_inter_block(mbmi)) {
- struct encode_b_args arg = {x, NULL, &mbmi->skip};
- vp10_encode_block_intra(plane, block, blk_row, blk_col,
- plane_bsize, tx_size, &arg);
+ struct encode_b_args arg = { x, NULL, &mbmi->skip };
+ vp10_encode_block_intra(plane, block, blk_row, blk_col, plane_bsize,
+ tx_size, &arg);
dist_block(x, plane, block, tx_size, &dist, &sse);
} else if (max_txsize_lookup[plane_bsize] == tx_size) {
if (x->skip_txfm[(plane << 2) + (block >> (tx_size << 1))] ==
SKIP_TXFM_NONE) {
// full forward transform and quantization
- vp10_xform_quant(x, plane, block, blk_row, blk_col,
- plane_bsize, tx_size);
+ vp10_xform_quant(x, plane, block, blk_row, blk_col, plane_bsize, tx_size);
dist_block(x, plane, block, tx_size, &dist, &sse);
} else if (x->skip_txfm[(plane << 2) + (block >> (tx_size << 1))] ==
SKIP_TXFM_AC_ONLY) {
// compute DC coefficient
- tran_low_t *const coeff = BLOCK_OFFSET(x->plane[plane].coeff, block);
+ tran_low_t *const coeff = BLOCK_OFFSET(x->plane[plane].coeff, block);
tran_low_t *const dqcoeff = BLOCK_OFFSET(xd->plane[plane].dqcoeff, block);
- vp10_xform_quant_dc(x, plane, block, blk_row, blk_col,
- plane_bsize, tx_size);
- sse = x->bsse[(plane << 2) + (block >> (tx_size << 1))] << 4;
+ vp10_xform_quant_dc(x, plane, block, blk_row, blk_col, plane_bsize,
+ tx_size);
+ sse = x->bsse[(plane << 2) + (block >> (tx_size << 1))] << 4;
dist = sse;
if (x->plane[plane].eobs[block]) {
const int64_t orig_sse = (int64_t)coeff[0] * coeff[0];
@@ -503,8 +487,7 @@
#if CONFIG_VPX_HIGHBITDEPTH
dc_correct >>= ((xd->bd - 8) * 2);
#endif
- if (tx_size != TX_32X32)
- dc_correct >>= 2;
+ if (tx_size != TX_32X32) dc_correct >>= 2;
dist = VPXMAX(0, sse - dc_correct);
}
@@ -512,7 +495,7 @@
// SKIP_TXFM_AC_DC
// skip forward transform
x->plane[plane].eobs[block] = 0;
- sse = x->bsse[(plane << 2) + (block >> (tx_size << 1))] << 4;
+ sse = x->bsse[(plane << 2) + (block >> (tx_size << 1))] << 4;
dist = sse;
}
} else {
@@ -534,7 +517,8 @@
// TODO(jingning): temporarily enabled only for luma component
rd = VPXMIN(rd1, rd2);
if (plane == 0)
- x->zcoeff_blk[tx_size][block] = !x->plane[plane].eobs[block] ||
+ x->zcoeff_blk[tx_size][block] =
+ !x->plane[plane].eobs[block] ||
(rd1 > rd2 && !xd->lossless[mbmi->segment_id]);
args->this_rate += rate;
@@ -550,11 +534,9 @@
args->skippable &= !x->plane[plane].eobs[block];
}
-static void txfm_rd_in_plane(MACROBLOCK *x,
- int *rate, int64_t *distortion,
- int *skippable, int64_t *sse,
- int64_t ref_best_rd, int plane,
- BLOCK_SIZE bsize, TX_SIZE tx_size,
+static void txfm_rd_in_plane(MACROBLOCK *x, int *rate, int64_t *distortion,
+ int *skippable, int64_t *sse, int64_t ref_best_rd,
+ int plane, BLOCK_SIZE bsize, TX_SIZE tx_size,
int use_fast_coef_casting) {
MACROBLOCKD *const xd = &x->e_mbd;
const struct macroblockd_plane *const pd = &xd->plane[plane];
@@ -566,34 +548,31 @@
args.use_fast_coef_costing = use_fast_coef_casting;
args.skippable = 1;
- if (plane == 0)
- xd->mi[0]->mbmi.tx_size = tx_size;
+ if (plane == 0) xd->mi[0]->mbmi.tx_size = tx_size;
vp10_get_entropy_contexts(bsize, tx_size, pd, args.t_above, args.t_left);
tx_type = get_tx_type(pd->plane_type, xd, 0);
args.so = get_scan(tx_size, tx_type);
- vp10_foreach_transformed_block_in_plane(xd, bsize, plane,
- block_rd_txfm, &args);
+ vp10_foreach_transformed_block_in_plane(xd, bsize, plane, block_rd_txfm,
+ &args);
if (args.exit_early) {
- *rate = INT_MAX;
+ *rate = INT_MAX;
*distortion = INT64_MAX;
- *sse = INT64_MAX;
- *skippable = 0;
+ *sse = INT64_MAX;
+ *skippable = 0;
} else {
*distortion = args.this_dist;
- *rate = args.this_rate;
- *sse = args.this_sse;
- *skippable = args.skippable;
+ *rate = args.this_rate;
+ *sse = args.this_sse;
+ *skippable = args.skippable;
}
}
-static void choose_largest_tx_size(VP10_COMP *cpi, MACROBLOCK *x,
- int *rate, int64_t *distortion,
- int *skip, int64_t *sse,
- int64_t ref_best_rd,
- BLOCK_SIZE bs) {
+static void choose_largest_tx_size(VP10_COMP *cpi, MACROBLOCK *x, int *rate,
+ int64_t *distortion, int *skip, int64_t *sse,
+ int64_t ref_best_rd, BLOCK_SIZE bs) {
const TX_SIZE max_tx_size = max_txsize_lookup[bs];
VP10_COMMON *const cm = &cpi->common;
const TX_SIZE largest_tx_size = tx_mode_to_biggest_tx_size[cm->tx_mode];
@@ -604,20 +583,17 @@
int r, s;
int64_t d, psse, this_rd, best_rd = INT64_MAX;
vpx_prob skip_prob = vp10_get_skip_prob(cm, xd);
- int s0 = vp10_cost_bit(skip_prob, 0);
- int s1 = vp10_cost_bit(skip_prob, 1);
+ int s0 = vp10_cost_bit(skip_prob, 0);
+ int s1 = vp10_cost_bit(skip_prob, 1);
const int is_inter = is_inter_block(mbmi);
mbmi->tx_size = VPXMIN(max_tx_size, largest_tx_size);
- if (mbmi->tx_size < TX_32X32 &&
- !xd->lossless[mbmi->segment_id]) {
+ if (mbmi->tx_size < TX_32X32 && !xd->lossless[mbmi->segment_id]) {
for (tx_type = 0; tx_type < TX_TYPES; ++tx_type) {
mbmi->tx_type = tx_type;
- txfm_rd_in_plane(x, &r, &d, &s,
- &psse, ref_best_rd, 0, bs, mbmi->tx_size,
+ txfm_rd_in_plane(x, &r, &d, &s, &psse, ref_best_rd, 0, bs, mbmi->tx_size,
cpi->sf.use_fast_coef_costing);
- if (r == INT_MAX)
- continue;
+ if (r == INT_MAX) continue;
if (is_inter)
r += cpi->inter_tx_type_costs[mbmi->tx_size][mbmi->tx_type];
else
@@ -638,41 +614,35 @@
}
}
mbmi->tx_type = best_tx_type;
- txfm_rd_in_plane(x, rate, distortion, skip,
- sse, ref_best_rd, 0, bs,
+ txfm_rd_in_plane(x, rate, distortion, skip, sse, ref_best_rd, 0, bs,
mbmi->tx_size, cpi->sf.use_fast_coef_costing);
if (mbmi->tx_size < TX_32X32 && !xd->lossless[mbmi->segment_id] &&
*rate != INT_MAX) {
if (is_inter)
*rate += cpi->inter_tx_type_costs[mbmi->tx_size][mbmi->tx_type];
else
- *rate += cpi->intra_tx_type_costs[mbmi->tx_size]
- [intra_mode_to_tx_type_context[mbmi->mode]]
- [mbmi->tx_type];
+ *rate += cpi->intra_tx_type_costs
+ [mbmi->tx_size][intra_mode_to_tx_type_context[mbmi->mode]]
+ [mbmi->tx_type];
}
}
-static void choose_smallest_tx_size(VP10_COMP *cpi, MACROBLOCK *x,
- int *rate, int64_t *distortion,
- int *skip, int64_t *sse,
- int64_t ref_best_rd,
+static void choose_smallest_tx_size(VP10_COMP *cpi, MACROBLOCK *x, int *rate,
+ int64_t *distortion, int *skip,
+ int64_t *sse, int64_t ref_best_rd,
BLOCK_SIZE bs) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
mbmi->tx_size = TX_4X4;
- txfm_rd_in_plane(x, rate, distortion, skip,
- sse, ref_best_rd, 0, bs,
+ txfm_rd_in_plane(x, rate, distortion, skip, sse, ref_best_rd, 0, bs,
mbmi->tx_size, cpi->sf.use_fast_coef_costing);
}
-static void choose_tx_size_from_rd(VP10_COMP *cpi, MACROBLOCK *x,
- int *rate,
- int64_t *distortion,
- int *skip,
- int64_t *psse,
- int64_t ref_best_rd,
+static void choose_tx_size_from_rd(VP10_COMP *cpi, MACROBLOCK *x, int *rate,
+ int64_t *distortion, int *skip,
+ int64_t *psse, int64_t ref_best_rd,
BLOCK_SIZE bs) {
const TX_SIZE max_tx_size = max_txsize_lookup[bs];
VP10_COMMON *const cm = &cpi->common;
@@ -707,15 +677,15 @@
}
*distortion = INT64_MAX;
- *rate = INT_MAX;
- *skip = 0;
- *psse = INT64_MAX;
+ *rate = INT_MAX;
+ *skip = 0;
+ *psse = INT64_MAX;
for (tx_type = DCT_DCT; tx_type < TX_TYPES; ++tx_type) {
last_rd = INT64_MAX;
for (n = start_tx; n >= end_tx; --n) {
int r_tx_size = 0;
- for (m = 0; m <= n - (n == (int) max_tx_size); ++m) {
+ for (m = 0; m <= n - (n == (int)max_tx_size); ++m) {
if (m == n)
r_tx_size += vp10_cost_zero(tx_probs[m]);
else
@@ -726,56 +696,50 @@
continue;
}
mbmi->tx_type = tx_type;
- txfm_rd_in_plane(x, &r, &d, &s,
- &sse, ref_best_rd, 0, bs, n,
+ txfm_rd_in_plane(x, &r, &d, &s, &sse, ref_best_rd, 0, bs, n,
cpi->sf.use_fast_coef_costing);
- if (n < TX_32X32 &&
- !xd->lossless[xd->mi[0]->mbmi.segment_id] &&
+ if (n < TX_32X32 && !xd->lossless[xd->mi[0]->mbmi.segment_id] &&
r != INT_MAX) {
if (is_inter)
r += cpi->inter_tx_type_costs[mbmi->tx_size][mbmi->tx_type];
else
- r += cpi->intra_tx_type_costs[mbmi->tx_size]
- [intra_mode_to_tx_type_context[mbmi->mode]]
- [mbmi->tx_type];
+ r += cpi->intra_tx_type_costs
+ [mbmi->tx_size][intra_mode_to_tx_type_context[mbmi->mode]]
+ [mbmi->tx_type];
}
- if (r == INT_MAX)
- continue;
+ if (r == INT_MAX) continue;
if (s) {
if (is_inter) {
rd = RDCOST(x->rdmult, x->rddiv, s1, sse);
} else {
- rd = RDCOST(x->rdmult, x->rddiv, s1 + r_tx_size * tx_select, sse);
+ rd = RDCOST(x->rdmult, x->rddiv, s1 + r_tx_size * tx_select, sse);
}
} else {
rd = RDCOST(x->rdmult, x->rddiv, r + s0 + r_tx_size * tx_select, d);
}
- if (tx_select && !(s && is_inter))
- r += r_tx_size;
+ if (tx_select && !(s && is_inter)) r += r_tx_size;
if (is_inter && !xd->lossless[xd->mi[0]->mbmi.segment_id] && !s)
rd = VPXMIN(rd, RDCOST(x->rdmult, x->rddiv, s1, sse));
// Early termination in transform size search.
if (cpi->sf.tx_size_search_breakout &&
- (rd == INT64_MAX ||
- (s == 1 && tx_type != DCT_DCT && n < start_tx) ||
- (n < (int) max_tx_size && rd > last_rd)))
+ (rd == INT64_MAX || (s == 1 && tx_type != DCT_DCT && n < start_tx) ||
+ (n < (int)max_tx_size && rd > last_rd)))
break;
last_rd = rd;
if (rd <
- (is_inter && best_tx_type == DCT_DCT ? ext_tx_th : 1) *
- best_rd) {
+ (is_inter && best_tx_type == DCT_DCT ? ext_tx_th : 1) * best_rd) {
best_tx = n;
best_rd = rd;
*distortion = d;
- *rate = r;
- *skip = s;
- *psse = sse;
+ *rate = r;
+ *skip = s;
+ *psse = sse;
best_tx_type = mbmi->tx_type;
}
}
@@ -783,17 +747,14 @@
mbmi->tx_size = best_tx;
mbmi->tx_type = best_tx_type;
- if (mbmi->tx_size >= TX_32X32)
- assert(mbmi->tx_type == DCT_DCT);
- txfm_rd_in_plane(x, &r, &d, &s,
- &sse, ref_best_rd, 0, bs, best_tx,
+ if (mbmi->tx_size >= TX_32X32) assert(mbmi->tx_type == DCT_DCT);
+ txfm_rd_in_plane(x, &r, &d, &s, &sse, ref_best_rd, 0, bs, best_tx,
cpi->sf.use_fast_coef_costing);
}
static void super_block_yrd(VP10_COMP *cpi, MACROBLOCK *x, int *rate,
- int64_t *distortion, int *skip,
- int64_t *psse, BLOCK_SIZE bs,
- int64_t ref_best_rd) {
+ int64_t *distortion, int *skip, int64_t *psse,
+ BLOCK_SIZE bs, int64_t ref_best_rd) {
MACROBLOCKD *xd = &x->e_mbd;
int64_t sse;
int64_t *ret_sse = psse ? psse : &sse;
@@ -808,39 +769,33 @@
choose_largest_tx_size(cpi, x, rate, distortion, skip, ret_sse, ref_best_rd,
bs);
} else {
- choose_tx_size_from_rd(cpi, x, rate, distortion, skip, ret_sse,
- ref_best_rd, bs);
+ choose_tx_size_from_rd(cpi, x, rate, distortion, skip, ret_sse, ref_best_rd,
+ bs);
}
}
static int conditional_skipintra(PREDICTION_MODE mode,
PREDICTION_MODE best_intra_mode) {
- if (mode == D117_PRED &&
- best_intra_mode != V_PRED &&
+ if (mode == D117_PRED && best_intra_mode != V_PRED &&
best_intra_mode != D135_PRED)
return 1;
- if (mode == D63_PRED &&
- best_intra_mode != V_PRED &&
+ if (mode == D63_PRED && best_intra_mode != V_PRED &&
best_intra_mode != D45_PRED)
return 1;
- if (mode == D207_PRED &&
- best_intra_mode != H_PRED &&
+ if (mode == D207_PRED && best_intra_mode != H_PRED &&
best_intra_mode != D45_PRED)
return 1;
- if (mode == D153_PRED &&
- best_intra_mode != H_PRED &&
+ if (mode == D153_PRED && best_intra_mode != H_PRED &&
best_intra_mode != D135_PRED)
return 1;
return 0;
}
-static int64_t rd_pick_intra4x4block(VP10_COMP *cpi, MACROBLOCK *x,
- int row, int col,
- PREDICTION_MODE *best_mode,
- const int *bmode_costs,
- ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l,
- int *bestrate, int *bestratey,
- int64_t *bestdistortion,
+static int64_t rd_pick_intra4x4block(VP10_COMP *cpi, MACROBLOCK *x, int row,
+ int col, PREDICTION_MODE *best_mode,
+ const int *bmode_costs, ENTROPY_CONTEXT *a,
+ ENTROPY_CONTEXT *l, int *bestrate,
+ int *bestratey, int64_t *bestdistortion,
BLOCK_SIZE bsize, int64_t rd_thresh) {
PREDICTION_MODE mode;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -873,14 +828,12 @@
int64_t distortion = 0;
int rate = bmode_costs[mode];
- if (!(cpi->sf.intra_y_mode_mask[TX_4X4] & (1 << mode)))
- continue;
+ if (!(cpi->sf.intra_y_mode_mask[TX_4X4] & (1 << mode))) continue;
// Only do the oblique modes if the best so far is
// one of the neighboring directional modes
if (cpi->sf.mode_search_skip_flags & FLAG_SKIP_INTRA_DIRMISMATCH) {
- if (conditional_skipintra(mode, *best_mode))
- continue;
+ if (conditional_skipintra(mode, *best_mode)) continue;
}
memcpy(tempa, ta, sizeof(ta));
@@ -891,16 +844,14 @@
const int block = (row + idy) * 2 + (col + idx);
const uint8_t *const src = &src_init[idx * 4 + idy * 4 * src_stride];
uint8_t *const dst = &dst_init[idx * 4 + idy * 4 * dst_stride];
- int16_t *const src_diff = vp10_raster_block_offset_int16(BLOCK_8X8,
- block,
- p->src_diff);
+ int16_t *const src_diff =
+ vp10_raster_block_offset_int16(BLOCK_8X8, block, p->src_diff);
tran_low_t *const coeff = BLOCK_OFFSET(x->plane[0].coeff, block);
xd->mi[0]->bmi[block].as_mode = mode;
- vp10_predict_intra_block(xd, 1, 1, TX_4X4, mode, dst, dst_stride,
- dst, dst_stride,
- col + idx, row + idy, 0);
- vpx_highbd_subtract_block(4, 4, src_diff, 8, src, src_stride,
- dst, dst_stride, xd->bd);
+ vp10_predict_intra_block(xd, 1, 1, TX_4X4, mode, dst, dst_stride, dst,
+ dst_stride, col + idx, row + idy, 0);
+ vpx_highbd_subtract_block(4, 4, src_diff, 8, src, src_stride, dst,
+ dst_stride, xd->bd);
if (xd->lossless[xd->mi[0]->mbmi.segment_id]) {
TX_TYPE tx_type = get_tx_type(PLANE_TYPE_Y, xd, block);
const scan_order *so = get_scan(TX_4X4, tx_type);
@@ -911,9 +862,9 @@
cpi->sf.use_fast_coef_costing);
if (RDCOST(x->rdmult, x->rddiv, ratey, distortion) >= best_rd)
goto next_highbd;
- vp10_highbd_inv_txfm_add_4x4(BLOCK_OFFSET(pd->dqcoeff, block),
- dst, dst_stride, p->eobs[block],
- xd->bd, DCT_DCT, 1);
+ vp10_highbd_inv_txfm_add_4x4(BLOCK_OFFSET(pd->dqcoeff, block), dst,
+ dst_stride, p->eobs[block], xd->bd,
+ DCT_DCT, 1);
} else {
int64_t unused;
TX_TYPE tx_type = get_tx_type(PLANE_TYPE_Y, xd, block);
@@ -923,14 +874,15 @@
ratey += cost_coeffs(x, 0, block, tempa + idx, templ + idy, TX_4X4,
so->scan, so->neighbors,
cpi->sf.use_fast_coef_costing);
- distortion += vp10_highbd_block_error(
- coeff, BLOCK_OFFSET(pd->dqcoeff, block),
- 16, &unused, xd->bd) >> 2;
+ distortion +=
+ vp10_highbd_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, block),
+ 16, &unused, xd->bd) >>
+ 2;
if (RDCOST(x->rdmult, x->rddiv, ratey, distortion) >= best_rd)
goto next_highbd;
- vp10_highbd_inv_txfm_add_4x4(BLOCK_OFFSET(pd->dqcoeff, block),
- dst, dst_stride, p->eobs[block],
- xd->bd, tx_type, 0);
+ vp10_highbd_inv_txfm_add_4x4(BLOCK_OFFSET(pd->dqcoeff, block), dst,
+ dst_stride, p->eobs[block], xd->bd,
+ tx_type, 0);
}
}
}
@@ -952,16 +904,13 @@
num_4x4_blocks_wide * 4 * sizeof(uint16_t));
}
}
- next_highbd:
- {}
+ next_highbd : {}
}
- if (best_rd >= rd_thresh)
- return best_rd;
+ if (best_rd >= rd_thresh) return best_rd;
for (idy = 0; idy < num_4x4_blocks_high * 4; ++idy) {
memcpy(CONVERT_TO_SHORTPTR(dst_init + idy * dst_stride),
- best_dst16 + idy * 8,
- num_4x4_blocks_wide * 4 * sizeof(uint16_t));
+ best_dst16 + idy * 8, num_4x4_blocks_wide * 4 * sizeof(uint16_t));
}
return best_rd;
@@ -974,14 +923,12 @@
int64_t distortion = 0;
int rate = bmode_costs[mode];
- if (!(cpi->sf.intra_y_mode_mask[TX_4X4] & (1 << mode)))
- continue;
+ if (!(cpi->sf.intra_y_mode_mask[TX_4X4] & (1 << mode))) continue;
// Only do the oblique modes if the best so far is
// one of the neighboring directional modes
if (cpi->sf.mode_search_skip_flags & FLAG_SKIP_INTRA_DIRMISMATCH) {
- if (conditional_skipintra(mode, *best_mode))
- continue;
+ if (conditional_skipintra(mode, *best_mode)) continue;
}
memcpy(tempa, ta, sizeof(ta));
@@ -996,8 +943,8 @@
vp10_raster_block_offset_int16(BLOCK_8X8, block, p->src_diff);
tran_low_t *const coeff = BLOCK_OFFSET(x->plane[0].coeff, block);
xd->mi[0]->bmi[block].as_mode = mode;
- vp10_predict_intra_block(xd, 1, 1, TX_4X4, mode, dst, dst_stride,
- dst, dst_stride, col + idx, row + idy, 0);
+ vp10_predict_intra_block(xd, 1, 1, TX_4X4, mode, dst, dst_stride, dst,
+ dst_stride, col + idx, row + idy, 0);
vpx_subtract_block(4, 4, src_diff, 8, src, src_stride, dst, dst_stride);
if (xd->lossless[xd->mi[0]->mbmi.segment_id]) {
@@ -1010,8 +957,8 @@
cpi->sf.use_fast_coef_costing);
if (RDCOST(x->rdmult, x->rddiv, ratey, distortion) >= best_rd)
goto next;
- vp10_inv_txfm_add_4x4(BLOCK_OFFSET(pd->dqcoeff, block),
- dst, dst_stride, p->eobs[block], DCT_DCT, 1);
+ vp10_inv_txfm_add_4x4(BLOCK_OFFSET(pd->dqcoeff, block), dst,
+ dst_stride, p->eobs[block], DCT_DCT, 1);
} else {
int64_t unused;
TX_TYPE tx_type = get_tx_type(PLANE_TYPE_Y, xd, block);
@@ -1019,14 +966,16 @@
vp10_fwd_txfm_4x4(src_diff, coeff, 8, tx_type, 0);
vp10_regular_quantize_b_4x4(x, 0, block, so->scan, so->iscan);
ratey += cost_coeffs(x, 0, block, tempa + idx, templ + idy, TX_4X4,
- so->scan, so->neighbors,
- cpi->sf.use_fast_coef_costing);
- distortion += vp10_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, block),
- 16, &unused) >> 2;
+ so->scan, so->neighbors,
+ cpi->sf.use_fast_coef_costing);
+ distortion +=
+ vp10_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, block), 16,
+ &unused) >>
+ 2;
if (RDCOST(x->rdmult, x->rddiv, ratey, distortion) >= best_rd)
goto next;
- vp10_inv_txfm_add_4x4(BLOCK_OFFSET(pd->dqcoeff, block),
- dst, dst_stride, p->eobs[block], tx_type, 0);
+ vp10_inv_txfm_add_4x4(BLOCK_OFFSET(pd->dqcoeff, block), dst,
+ dst_stride, p->eobs[block], tx_type, 0);
}
}
}
@@ -1046,12 +995,10 @@
memcpy(best_dst + idy * 8, dst_init + idy * dst_stride,
num_4x4_blocks_wide * 4);
}
- next:
- {}
+ next : {}
}
- if (best_rd >= rd_thresh)
- return best_rd;
+ if (best_rd >= rd_thresh) return best_rd;
for (idy = 0; idy < num_4x4_blocks_high * 4; ++idy)
memcpy(dst_init + idy * dst_stride, best_dst + idy * 8,
@@ -1094,14 +1041,13 @@
const PREDICTION_MODE A = vp10_above_block_mode(mic, above_mi, i);
const PREDICTION_MODE L = vp10_left_block_mode(mic, left_mi, i);
- bmode_costs = cpi->y_mode_costs[A][L];
+ bmode_costs = cpi->y_mode_costs[A][L];
}
this_rd = rd_pick_intra4x4block(cpi, mb, idy, idx, &best_mode,
bmode_costs, t_above + idx, t_left + idy,
&r, &ry, &d, bsize, best_rd - total_rd);
- if (this_rd >= best_rd - total_rd)
- return INT64_MAX;
+ if (this_rd >= best_rd - total_rd) return INT64_MAX;
total_rd += this_rd;
cost += r;
@@ -1114,8 +1060,7 @@
for (j = 1; j < num_4x4_blocks_wide; ++j)
mic->bmi[i + j].as_mode = best_mode;
- if (total_rd >= best_rd)
- return INT64_MAX;
+ if (total_rd >= best_rd) return INT64_MAX;
}
}
@@ -1128,10 +1073,9 @@
}
// This function is used only for intra_only frames
-static int64_t rd_pick_intra_sby_mode(VP10_COMP *cpi, MACROBLOCK *x,
- int *rate, int *rate_tokenonly,
- int64_t *distortion, int *skippable,
- BLOCK_SIZE bsize,
+static int64_t rd_pick_intra_sby_mode(VP10_COMP *cpi, MACROBLOCK *x, int *rate,
+ int *rate_tokenonly, int64_t *distortion,
+ int *skippable, BLOCK_SIZE bsize,
int64_t best_rd) {
PREDICTION_MODE mode;
PREDICTION_MODE mode_selected = DC_PRED;
@@ -1154,24 +1098,23 @@
for (mode = DC_PRED; mode <= TM_PRED; mode++) {
mic->mbmi.mode = mode;
- super_block_yrd(cpi, x, &this_rate_tokenonly, &this_distortion,
- &s, NULL, bsize, best_rd);
+ super_block_yrd(cpi, x, &this_rate_tokenonly, &this_distortion, &s, NULL,
+ bsize, best_rd);
- if (this_rate_tokenonly == INT_MAX)
- continue;
+ if (this_rate_tokenonly == INT_MAX) continue;
this_rate = this_rate_tokenonly + bmode_costs[mode];
this_rd = RDCOST(x->rdmult, x->rddiv, this_rate, this_distortion);
if (this_rd < best_rd) {
- mode_selected = mode;
- best_rd = this_rd;
- best_tx = mic->mbmi.tx_size;
- best_tx_type = mic->mbmi.tx_type;
- *rate = this_rate;
+ mode_selected = mode;
+ best_rd = this_rd;
+ best_tx = mic->mbmi.tx_size;
+ best_tx_type = mic->mbmi.tx_type;
+ *rate = this_rate;
*rate_tokenonly = this_rate_tokenonly;
- *distortion = this_distortion;
- *skippable = s;
+ *distortion = this_distortion;
+ *skippable = s;
}
}
@@ -1184,10 +1127,9 @@
// Return value 0: early termination triggered, no valid rd cost available;
// 1: rd cost values are valid.
-static int super_block_uvrd(const VP10_COMP *cpi, MACROBLOCK *x,
- int *rate, int64_t *distortion, int *skippable,
- int64_t *sse, BLOCK_SIZE bsize,
- int64_t ref_best_rd) {
+static int super_block_uvrd(const VP10_COMP *cpi, MACROBLOCK *x, int *rate,
+ int64_t *distortion, int *skippable, int64_t *sse,
+ BLOCK_SIZE bsize, int64_t ref_best_rd) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
const TX_SIZE uv_tx_size = get_uv_tx_size(mbmi, &xd->plane[1]);
@@ -1196,8 +1138,7 @@
int64_t pndist = 0, pnsse = 0;
int is_cost_valid = 1;
- if (ref_best_rd < 0)
- is_cost_valid = 0;
+ if (ref_best_rd < 0) is_cost_valid = 0;
if (is_inter_block(mbmi) && is_cost_valid) {
int plane;
@@ -1211,9 +1152,8 @@
*skippable = 1;
for (plane = 1; plane < MAX_MB_PLANE; ++plane) {
- txfm_rd_in_plane(x, &pnrate, &pndist, &pnskip, &pnsse,
- ref_best_rd, plane, bsize, uv_tx_size,
- cpi->sf.use_fast_coef_costing);
+ txfm_rd_in_plane(x, &pnrate, &pndist, &pnskip, &pnsse, ref_best_rd, plane,
+ bsize, uv_tx_size, cpi->sf.use_fast_coef_costing);
if (pnrate == INT_MAX) {
is_cost_valid = 0;
break;
@@ -1236,10 +1176,10 @@
}
static int64_t rd_pick_intra_sbuv_mode(VP10_COMP *cpi, MACROBLOCK *x,
- PICK_MODE_CONTEXT *ctx,
- int *rate, int *rate_tokenonly,
- int64_t *distortion, int *skippable,
- BLOCK_SIZE bsize, TX_SIZE max_tx_size) {
+ PICK_MODE_CONTEXT *ctx, int *rate,
+ int *rate_tokenonly, int64_t *distortion,
+ int *skippable, BLOCK_SIZE bsize,
+ TX_SIZE max_tx_size) {
MACROBLOCKD *xd = &x->e_mbd;
PREDICTION_MODE mode;
PREDICTION_MODE mode_selected = DC_PRED;
@@ -1249,27 +1189,25 @@
memset(x->skip_txfm, SKIP_TXFM_NONE, sizeof(x->skip_txfm));
for (mode = DC_PRED; mode <= TM_PRED; ++mode) {
- if (!(cpi->sf.intra_uv_mode_mask[max_tx_size] & (1 << mode)))
- continue;
+ if (!(cpi->sf.intra_uv_mode_mask[max_tx_size] & (1 << mode))) continue;
xd->mi[0]->mbmi.uv_mode = mode;
- if (!super_block_uvrd(cpi, x, &this_rate_tokenonly,
- &this_distortion, &s, &this_sse, bsize, best_rd))
+ if (!super_block_uvrd(cpi, x, &this_rate_tokenonly, &this_distortion, &s,
+ &this_sse, bsize, best_rd))
continue;
this_rate = this_rate_tokenonly +
- cpi->intra_uv_mode_cost[xd->mi[0]->mbmi.mode][mode];
+ cpi->intra_uv_mode_cost[xd->mi[0]->mbmi.mode][mode];
this_rd = RDCOST(x->rdmult, x->rddiv, this_rate, this_distortion);
if (this_rd < best_rd) {
- mode_selected = mode;
- best_rd = this_rd;
- *rate = this_rate;
+ mode_selected = mode;
+ best_rd = this_rd;
+ *rate = this_rate;
*rate_tokenonly = this_rate_tokenonly;
- *distortion = this_distortion;
- *skippable = s;
- if (!x->select_tx_size)
- swap_block_ptr(x, ctx, 2, 0, 1, MAX_MB_PLANE);
+ *distortion = this_distortion;
+ *skippable = s;
+ if (!x->select_tx_size) swap_block_ptr(x, ctx, 2, 0, 1, MAX_MB_PLANE);
}
}
@@ -1277,38 +1215,36 @@
return best_rd;
}
-static int64_t rd_sbuv_dcpred(const VP10_COMP *cpi, MACROBLOCK *x,
- int *rate, int *rate_tokenonly,
- int64_t *distortion, int *skippable,
- BLOCK_SIZE bsize) {
+static int64_t rd_sbuv_dcpred(const VP10_COMP *cpi, MACROBLOCK *x, int *rate,
+ int *rate_tokenonly, int64_t *distortion,
+ int *skippable, BLOCK_SIZE bsize) {
int64_t unused;
x->e_mbd.mi[0]->mbmi.uv_mode = DC_PRED;
memset(x->skip_txfm, SKIP_TXFM_NONE, sizeof(x->skip_txfm));
- super_block_uvrd(cpi, x, rate_tokenonly, distortion,
- skippable, &unused, bsize, INT64_MAX);
+ super_block_uvrd(cpi, x, rate_tokenonly, distortion, skippable, &unused,
+ bsize, INT64_MAX);
*rate = *rate_tokenonly +
- cpi->intra_uv_mode_cost[x->e_mbd.mi[0]->mbmi.mode][DC_PRED];
+ cpi->intra_uv_mode_cost[x->e_mbd.mi[0]->mbmi.mode][DC_PRED];
return RDCOST(x->rdmult, x->rddiv, *rate, *distortion);
}
static void choose_intra_uv_mode(VP10_COMP *cpi, MACROBLOCK *const x,
- PICK_MODE_CONTEXT *ctx,
- BLOCK_SIZE bsize, TX_SIZE max_tx_size,
- int *rate_uv, int *rate_uv_tokenonly,
- int64_t *dist_uv, int *skip_uv,
- PREDICTION_MODE *mode_uv) {
+ PICK_MODE_CONTEXT *ctx, BLOCK_SIZE bsize,
+ TX_SIZE max_tx_size, int *rate_uv,
+ int *rate_uv_tokenonly, int64_t *dist_uv,
+ int *skip_uv, PREDICTION_MODE *mode_uv) {
// Use an estimated rd for uv_intra based on DC_PRED if the
// appropriate speed flag is set.
if (cpi->sf.use_uv_intra_rd_estimate) {
- rd_sbuv_dcpred(cpi, x, rate_uv, rate_uv_tokenonly, dist_uv,
- skip_uv, bsize < BLOCK_8X8 ? BLOCK_8X8 : bsize);
- // Else do a proper rd search for each possible transform size that may
- // be considered in the main rd loop.
+ rd_sbuv_dcpred(cpi, x, rate_uv, rate_uv_tokenonly, dist_uv, skip_uv,
+ bsize < BLOCK_8X8 ? BLOCK_8X8 : bsize);
+ // Else do a proper rd search for each possible transform size that may
+ // be considered in the main rd loop.
} else {
- rd_pick_intra_sbuv_mode(cpi, x, ctx,
- rate_uv, rate_uv_tokenonly, dist_uv, skip_uv,
- bsize < BLOCK_8X8 ? BLOCK_8X8 : bsize, max_tx_size);
+ rd_pick_intra_sbuv_mode(cpi, x, ctx, rate_uv, rate_uv_tokenonly, dist_uv,
+ skip_uv, bsize < BLOCK_8X8 ? BLOCK_8X8 : bsize,
+ max_tx_size);
}
*mode_uv = x->e_mbd.mi[0]->mbmi.uv_mode;
}
@@ -1320,8 +1256,7 @@
}
static int set_and_cost_bmi_mvs(VP10_COMP *cpi, MACROBLOCK *x, MACROBLOCKD *xd,
- int i,
- PREDICTION_MODE mode, int_mv this_mv[2],
+ int i, PREDICTION_MODE mode, int_mv this_mv[2],
int_mv frame_mv[MB_MODE_COUNT][MAX_REF_FRAMES],
int_mv seg_mvs[MAX_REF_FRAMES],
int_mv *best_ref_mv[2], const int *mvjcost,
@@ -1339,11 +1274,12 @@
case NEWMV:
this_mv[0].as_int = seg_mvs[mbmi->ref_frame[0]].as_int;
thismvcost += vp10_mv_bit_cost(&this_mv[0].as_mv, &best_ref_mv[0]->as_mv,
- mvjcost, mvcost, MV_COST_WEIGHT_SUB);
+ mvjcost, mvcost, MV_COST_WEIGHT_SUB);
if (is_compound) {
this_mv[1].as_int = seg_mvs[mbmi->ref_frame[1]].as_int;
- thismvcost += vp10_mv_bit_cost(&this_mv[1].as_mv, &best_ref_mv[1]->as_mv,
- mvjcost, mvcost, MV_COST_WEIGHT_SUB);
+ thismvcost +=
+ vp10_mv_bit_cost(&this_mv[1].as_mv, &best_ref_mv[1]->as_mv, mvjcost,
+ mvcost, MV_COST_WEIGHT_SUB);
}
break;
case NEARMV:
@@ -1354,16 +1290,13 @@
break;
case ZEROMV:
this_mv[0].as_int = 0;
- if (is_compound)
- this_mv[1].as_int = 0;
+ if (is_compound) this_mv[1].as_int = 0;
break;
- default:
- break;
+ default: break;
}
mic->bmi[i].as_mv[0].as_int = this_mv[0].as_int;
- if (is_compound)
- mic->bmi[i].as_mv[1].as_int = this_mv[1].as_int;
+ if (is_compound) mic->bmi[i].as_mv[1].as_int = this_mv[1].as_int;
mic->bmi[i].as_mode = mode;
@@ -1372,19 +1305,14 @@
memmove(&mic->bmi[i + idy * 2 + idx], &mic->bmi[i], sizeof(mic->bmi[i]));
return cost_mv_ref(cpi, mode, mbmi_ext->mode_context[mbmi->ref_frame[0]]) +
- thismvcost;
+ thismvcost;
}
-static int64_t encode_inter_mb_segment(VP10_COMP *cpi,
- MACROBLOCK *x,
- int64_t best_yrd,
- int i,
- int *labelyrate,
+static int64_t encode_inter_mb_segment(VP10_COMP *cpi, MACROBLOCK *x,
+ int64_t best_yrd, int i, int *labelyrate,
int64_t *distortion, int64_t *sse,
- ENTROPY_CONTEXT *ta,
- ENTROPY_CONTEXT *tl,
- int ir, int ic,
- int mi_row, int mi_col) {
+ ENTROPY_CONTEXT *ta, ENTROPY_CONTEXT *tl,
+ int ir, int ic, int mi_row, int mi_col) {
int k;
MACROBLOCKD *xd = &x->e_mbd;
struct macroblockd_plane *const pd = &xd->plane[0];
@@ -1398,8 +1326,8 @@
const uint8_t *const src =
&p->src.buf[vp10_raster_block_offset(BLOCK_8X8, i, p->src.stride)];
- uint8_t *const dst = &pd->dst.buf[vp10_raster_block_offset(BLOCK_8X8, i,
- pd->dst.stride)];
+ uint8_t *const dst =
+ &pd->dst.buf[vp10_raster_block_offset(BLOCK_8X8, i, pd->dst.stride)];
int64_t thisdistortion = 0, thissse = 0;
int thisrate = 0;
TX_TYPE tx_type = get_tx_type(PLANE_TYPE_Y, xd, i);
@@ -1421,12 +1349,13 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
vpx_highbd_subtract_block(
- height, width, vp10_raster_block_offset_int16(BLOCK_8X8, i, p->src_diff),
- 8, src, p->src.stride, dst, pd->dst.stride, xd->bd);
+ height, width,
+ vp10_raster_block_offset_int16(BLOCK_8X8, i, p->src_diff), 8, src,
+ p->src.stride, dst, pd->dst.stride, xd->bd);
} else {
- vpx_subtract_block(
- height, width, vp10_raster_block_offset_int16(BLOCK_8X8, i, p->src_diff),
- 8, src, p->src.stride, dst, pd->dst.stride);
+ vpx_subtract_block(height, width, vp10_raster_block_offset_int16(
+ BLOCK_8X8, i, p->src_diff),
+ 8, src, p->src.stride, dst, pd->dst.stride);
}
#else
vpx_subtract_block(height, width,
@@ -1438,7 +1367,7 @@
for (idy = 0; idy < height / 4; ++idy) {
for (idx = 0; idx < width / 4; ++idx) {
int64_t ssz, rd, rd1, rd2;
- tran_low_t* coeff;
+ tran_low_t *coeff;
k += (idy * 2 + idx);
coeff = BLOCK_OFFSET(p->coeff, k);
@@ -1447,26 +1376,24 @@
vp10_regular_quantize_b_4x4(x, 0, k, so->scan, so->iscan);
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- thisdistortion += vp10_highbd_block_error(coeff,
- BLOCK_OFFSET(pd->dqcoeff, k),
- 16, &ssz, xd->bd);
+ thisdistortion += vp10_highbd_block_error(
+ coeff, BLOCK_OFFSET(pd->dqcoeff, k), 16, &ssz, xd->bd);
} else {
- thisdistortion += vp10_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, k),
- 16, &ssz);
+ thisdistortion +=
+ vp10_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, k), 16, &ssz);
}
#else
- thisdistortion += vp10_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, k),
- 16, &ssz);
+ thisdistortion +=
+ vp10_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, k), 16, &ssz);
#endif // CONFIG_VPX_HIGHBITDEPTH
thissse += ssz;
- thisrate += cost_coeffs(x, 0, k, ta + (k & 1), tl + (k >> 1), TX_4X4,
- so->scan, so->neighbors,
- cpi->sf.use_fast_coef_costing);
+ thisrate +=
+ cost_coeffs(x, 0, k, ta + (k & 1), tl + (k >> 1), TX_4X4, so->scan,
+ so->neighbors, cpi->sf.use_fast_coef_costing);
rd1 = RDCOST(x->rdmult, x->rddiv, thisrate, thisdistortion >> 2);
rd2 = RDCOST(x->rdmult, x->rddiv, 0, thissse >> 2);
rd = VPXMIN(rd1, rd2);
- if (rd >= best_yrd)
- return INT64_MAX;
+ if (rd >= best_yrd) return INT64_MAX;
}
}
@@ -1504,10 +1431,8 @@
} BEST_SEG_INFO;
static INLINE int mv_check_bounds(const MACROBLOCK *x, const MV *mv) {
- return (mv->row >> 3) < x->mv_row_min ||
- (mv->row >> 3) > x->mv_row_max ||
- (mv->col >> 3) < x->mv_col_min ||
- (mv->col >> 3) > x->mv_col_max;
+ return (mv->row >> 3) < x->mv_row_min || (mv->row >> 3) > x->mv_row_max ||
+ (mv->col >> 3) < x->mv_col_min || (mv->col >> 3) > x->mv_col_max;
}
static INLINE void mi_buf_shift(MACROBLOCK *x, int i) {
@@ -1515,14 +1440,16 @@
struct macroblock_plane *const p = &x->plane[0];
struct macroblockd_plane *const pd = &x->e_mbd.plane[0];
- p->src.buf = &p->src.buf[vp10_raster_block_offset(BLOCK_8X8, i,
- p->src.stride)];
+ p->src.buf =
+ &p->src.buf[vp10_raster_block_offset(BLOCK_8X8, i, p->src.stride)];
assert(((intptr_t)pd->pre[0].buf & 0x7) == 0);
- pd->pre[0].buf = &pd->pre[0].buf[vp10_raster_block_offset(BLOCK_8X8, i,
- pd->pre[0].stride)];
+ pd->pre[0].buf =
+ &pd->pre[0]
+ .buf[vp10_raster_block_offset(BLOCK_8X8, i, pd->pre[0].stride)];
if (has_second_ref(mbmi))
- pd->pre[1].buf = &pd->pre[1].buf[vp10_raster_block_offset(BLOCK_8X8, i,
- pd->pre[1].stride)];
+ pd->pre[1].buf =
+ &pd->pre[1]
+ .buf[vp10_raster_block_offset(BLOCK_8X8, i, pd->pre[1].stride)];
}
static INLINE void mi_buf_restore(MACROBLOCK *x, struct buf_2d orig_src,
@@ -1530,8 +1457,7 @@
MB_MODE_INFO *mbmi = &x->e_mbd.mi[0]->mbmi;
x->plane[0].src = orig_src;
x->e_mbd.plane[0].pre[0] = orig_pre[0];
- if (has_second_ref(mbmi))
- x->e_mbd.plane[0].pre[1] = orig_pre[1];
+ if (has_second_ref(mbmi)) x->e_mbd.plane[0].pre[1] = orig_pre[1];
}
static INLINE int mv_has_subpel(const MV *mv) {
@@ -1540,10 +1466,11 @@
// Check if NEARESTMV/NEARMV/ZEROMV is the cheapest way encode zero motion.
// TODO(aconverse): Find out if this is still productive then clean up or remove
-static int check_best_zero_mv(
- const VP10_COMP *cpi, const uint8_t mode_context[MAX_REF_FRAMES],
- int_mv frame_mv[MB_MODE_COUNT][MAX_REF_FRAMES], int this_mode,
- const MV_REFERENCE_FRAME ref_frames[2]) {
+static int check_best_zero_mv(const VP10_COMP *cpi,
+ const uint8_t mode_context[MAX_REF_FRAMES],
+ int_mv frame_mv[MB_MODE_COUNT][MAX_REF_FRAMES],
+ int this_mode,
+ const MV_REFERENCE_FRAME ref_frames[2]) {
if ((this_mode == NEARMV || this_mode == NEARESTMV || this_mode == ZEROMV) &&
frame_mv[this_mode][ref_frames[0]].as_int == 0 &&
(ref_frames[1] == NONE ||
@@ -1575,10 +1502,8 @@
return 1;
}
-static void joint_motion_search(VP10_COMP *cpi, MACROBLOCK *x,
- BLOCK_SIZE bsize,
- int_mv *frame_mv,
- int mi_row, int mi_col,
+static void joint_motion_search(VP10_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bsize,
+ int_mv *frame_mv, int mi_row, int mi_col,
int_mv single_newmv[MAX_REF_FRAMES],
int *rate_mv) {
const VP10_COMMON *const cm = &cpi->common;
@@ -1586,8 +1511,8 @@
const int ph = 4 * num_4x4_blocks_high_lookup[bsize];
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
- const int refs[2] = {mbmi->ref_frame[0],
- mbmi->ref_frame[1] < 0 ? 0 : mbmi->ref_frame[1]};
+ const int refs[2] = { mbmi->ref_frame[0],
+ mbmi->ref_frame[1] < 0 ? 0 : mbmi->ref_frame[1] };
int_mv ref_mv[2];
int ite, ref;
const InterpKernel *kernel = vp10_filter_kernels[mbmi->interp_filter];
@@ -1595,13 +1520,13 @@
// Do joint motion search in compound mode to get more accurate mv.
struct buf_2d backup_yv12[2][MAX_MB_PLANE];
- int last_besterr[2] = {INT_MAX, INT_MAX};
+ int last_besterr[2] = { INT_MAX, INT_MAX };
const YV12_BUFFER_CONFIG *const scaled_ref_frame[2] = {
vp10_get_scaled_ref_frame(cpi, mbmi->ref_frame[0]),
vp10_get_scaled_ref_frame(cpi, mbmi->ref_frame[1])
};
- // Prediction buffer from second frame.
+// Prediction buffer from second frame.
#if CONFIG_VPX_HIGHBITDEPTH
DECLARE_ALIGNED(16, uint16_t, second_pred_alloc_16[64 * 64]);
uint8_t *second_pred;
@@ -1620,21 +1545,20 @@
for (i = 0; i < MAX_MB_PLANE; i++)
backup_yv12[ref][i] = xd->plane[i].pre[ref];
vp10_setup_pre_planes(xd, ref, scaled_ref_frame[ref], mi_row, mi_col,
- NULL);
+ NULL);
}
frame_mv[refs[ref]].as_int = single_newmv[refs[ref]].as_int;
}
- // Since we have scaled the reference frames to match the size of the current
- // frame we must use a unit scaling factor during mode selection.
+// Since we have scaled the reference frames to match the size of the current
+// frame we must use a unit scaling factor during mode selection.
#if CONFIG_VPX_HIGHBITDEPTH
- vp10_setup_scale_factors_for_frame(&sf, cm->width, cm->height,
- cm->width, cm->height,
- cm->use_highbitdepth);
+ vp10_setup_scale_factors_for_frame(&sf, cm->width, cm->height, cm->width,
+ cm->height, cm->use_highbitdepth);
#else
- vp10_setup_scale_factors_for_frame(&sf, cm->width, cm->height,
- cm->width, cm->height);
+ vp10_setup_scale_factors_for_frame(&sf, cm->width, cm->height, cm->width,
+ cm->height);
#endif // CONFIG_VPX_HIGHBITDEPTH
// Allow joint search multiple times iteratively for each reference frame
@@ -1658,41 +1582,30 @@
ref_yv12[0] = xd->plane[0].pre[0];
ref_yv12[1] = xd->plane[0].pre[1];
- // Get the prediction block from the 'other' reference frame.
+// Get the prediction block from the 'other' reference frame.
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
second_pred = CONVERT_TO_BYTEPTR(second_pred_alloc_16);
- vp10_highbd_build_inter_predictor(ref_yv12[!id].buf,
- ref_yv12[!id].stride,
- second_pred, pw,
- &frame_mv[refs[!id]].as_mv,
- &sf, pw, ph, 0,
- kernel, MV_PRECISION_Q3,
- mi_col * MI_SIZE, mi_row * MI_SIZE,
- xd->bd);
+ vp10_highbd_build_inter_predictor(
+ ref_yv12[!id].buf, ref_yv12[!id].stride, second_pred, pw,
+ &frame_mv[refs[!id]].as_mv, &sf, pw, ph, 0, kernel, MV_PRECISION_Q3,
+ mi_col * MI_SIZE, mi_row * MI_SIZE, xd->bd);
} else {
second_pred = (uint8_t *)second_pred_alloc_16;
- vp10_build_inter_predictor(ref_yv12[!id].buf,
- ref_yv12[!id].stride,
- second_pred, pw,
- &frame_mv[refs[!id]].as_mv,
- &sf, pw, ph, 0,
- kernel, MV_PRECISION_Q3,
- mi_col * MI_SIZE, mi_row * MI_SIZE);
+ vp10_build_inter_predictor(ref_yv12[!id].buf, ref_yv12[!id].stride,
+ second_pred, pw, &frame_mv[refs[!id]].as_mv,
+ &sf, pw, ph, 0, kernel, MV_PRECISION_Q3,
+ mi_col * MI_SIZE, mi_row * MI_SIZE);
}
#else
- vp10_build_inter_predictor(ref_yv12[!id].buf,
- ref_yv12[!id].stride,
- second_pred, pw,
- &frame_mv[refs[!id]].as_mv,
- &sf, pw, ph, 0,
- kernel, MV_PRECISION_Q3,
- mi_col * MI_SIZE, mi_row * MI_SIZE);
+ vp10_build_inter_predictor(ref_yv12[!id].buf, ref_yv12[!id].stride,
+ second_pred, pw, &frame_mv[refs[!id]].as_mv, &sf,
+ pw, ph, 0, kernel, MV_PRECISION_Q3,
+ mi_col * MI_SIZE, mi_row * MI_SIZE);
#endif // CONFIG_VPX_HIGHBITDEPTH
// Do compound motion search on the current reference frame.
- if (id)
- xd->plane[0].pre[0] = ref_yv12[id];
+ if (id) xd->plane[0].pre[0] = ref_yv12[id];
vp10_set_mv_search_range(x, &ref_mv[id].as_mv);
// Use the mv result from the single mode as mv predictor.
@@ -1702,13 +1615,12 @@
tmp_mv.row >>= 3;
// Small-range full-pixel motion search.
- bestsme = vp10_refining_search_8p_c(x, &tmp_mv, sadpb,
- search_range,
- &cpi->fn_ptr[bsize],
- &ref_mv[id].as_mv, second_pred);
+ bestsme = vp10_refining_search_8p_c(x, &tmp_mv, sadpb, search_range,
+ &cpi->fn_ptr[bsize], &ref_mv[id].as_mv,
+ second_pred);
if (bestsme < INT_MAX)
bestsme = vp10_get_mvpred_av_var(x, &tmp_mv, &ref_mv[id].as_mv,
- second_pred, &cpi->fn_ptr[bsize], 1);
+ second_pred, &cpi->fn_ptr[bsize], 1);
x->mv_col_min = tmp_col_min;
x->mv_col_max = tmp_col_max;
@@ -1719,21 +1631,14 @@
int dis; /* TODO: use dis in distortion calculation later. */
unsigned int sse;
bestsme = cpi->find_fractional_mv_step(
- x, &tmp_mv,
- &ref_mv[id].as_mv,
- cpi->common.allow_high_precision_mv,
- x->errorperbit,
- &cpi->fn_ptr[bsize],
- 0, cpi->sf.mv.subpel_iters_per_step,
- NULL,
- x->nmvjointcost, x->mvcost,
- &dis, &sse, second_pred,
- pw, ph);
+ x, &tmp_mv, &ref_mv[id].as_mv, cpi->common.allow_high_precision_mv,
+ x->errorperbit, &cpi->fn_ptr[bsize], 0,
+ cpi->sf.mv.subpel_iters_per_step, NULL, x->nmvjointcost, x->mvcost,
+ &dis, &sse, second_pred, pw, ph);
}
// Restore the pointer to the first (possibly scaled) prediction buffer.
- if (id)
- xd->plane[0].pre[0] = ref_yv12[0];
+ if (id) xd->plane[0].pre[0] = ref_yv12[0];
if (bestsme < last_besterr[id]) {
frame_mv[refs[id]].as_mv = tmp_mv;
@@ -1754,22 +1659,17 @@
}
*rate_mv += vp10_mv_bit_cost(&frame_mv[refs[ref]].as_mv,
- &x->mbmi_ext->ref_mvs[refs[ref]][0].as_mv,
- x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
+ &x->mbmi_ext->ref_mvs[refs[ref]][0].as_mv,
+ x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
}
}
-static int64_t rd_pick_best_sub8x8_mode(VP10_COMP *cpi, MACROBLOCK *x,
- int_mv *best_ref_mv,
- int_mv *second_best_ref_mv,
- int64_t best_rd, int *returntotrate,
- int *returnyrate,
- int64_t *returndistortion,
- int *skippable, int64_t *psse,
- int mvthresh,
- int_mv seg_mvs[4][MAX_REF_FRAMES],
- BEST_SEG_INFO *bsi_buf, int filter_idx,
- int mi_row, int mi_col) {
+static int64_t rd_pick_best_sub8x8_mode(
+ VP10_COMP *cpi, MACROBLOCK *x, int_mv *best_ref_mv,
+ int_mv *second_best_ref_mv, int64_t best_rd, int *returntotrate,
+ int *returnyrate, int64_t *returndistortion, int *skippable, int64_t *psse,
+ int mvthresh, int_mv seg_mvs[4][MAX_REF_FRAMES], BEST_SEG_INFO *bsi_buf,
+ int filter_idx, int mi_row, int mi_col) {
int i;
BEST_SEG_INFO *bsi = bsi_buf + filter_idx;
MACROBLOCKD *xd = &x->e_mbd;
@@ -1803,8 +1703,7 @@
bsi->mvp.as_int = best_ref_mv->as_int;
bsi->mvthresh = mvthresh;
- for (i = 0; i < 4; i++)
- bsi->modes[i] = ZEROMV;
+ for (i = 0; i < 4; i++) bsi->modes[i] = ZEROMV;
memcpy(t_above, pd->above_context, sizeof(t_above));
memcpy(t_left, pd->left_context, sizeof(t_left));
@@ -1830,10 +1729,9 @@
for (ref = 0; ref < 1 + has_second_rf; ++ref) {
const MV_REFERENCE_FRAME frame = mbmi->ref_frame[ref];
frame_mv[ZEROMV][frame].as_int = 0;
- vp10_append_sub8x8_mvs_for_idx(cm, xd, i, ref, mi_row, mi_col,
- &frame_mv[NEARESTMV][frame],
- &frame_mv[NEARMV][frame],
- mbmi_ext->mode_context);
+ vp10_append_sub8x8_mvs_for_idx(
+ cm, xd, i, ref, mi_row, mi_col, &frame_mv[NEARESTMV][frame],
+ &frame_mv[NEARMV][frame], mbmi_ext->mode_context);
}
// search for the best motion vector on this segment
@@ -1843,8 +1741,7 @@
mode_idx = INTER_OFFSET(this_mode);
bsi->rdstat[i][mode_idx].brdcost = INT64_MAX;
- if (!(inter_mode_mask & (1 << this_mode)))
- continue;
+ if (!(inter_mode_mask & (1 << this_mode))) continue;
if (!check_best_zero_mv(cpi, mbmi_ext->mode_context, frame_mv,
this_mode, mbmi->ref_frame))
@@ -1869,15 +1766,13 @@
/* Is the best so far sufficiently good that we cant justify doing
* and new motion search. */
- if (best_rd < label_mv_thresh)
- break;
+ if (best_rd < label_mv_thresh) break;
if (cpi->oxcf.mode != BEST) {
// use previous block's result as next block's MV predictor.
if (i > 0) {
bsi->mvp.as_int = mi->bmi[i - 1].as_mv[0].as_int;
- if (i == 2)
- bsi->mvp.as_int = mi->bmi[i - 2].as_mv[0].as_int;
+ if (i == 2) bsi->mvp.as_int = mi->bmi[i - 2].as_mv[0].as_int;
}
}
if (i == 0)
@@ -1890,8 +1785,8 @@
// Take wtd average of the step_params based on the last frame's
// max mv magnitude and the best ref mvs of the current block for
// the given reference.
- step_param = (vp10_init_search_range(max_mv) +
- cpi->mv_step_param) / 2;
+ step_param =
+ (vp10_init_search_range(max_mv) + cpi->mv_step_param) / 2;
} else {
step_param = cpi->mv_step_param;
}
@@ -1913,24 +1808,16 @@
bestsme = vp10_full_pixel_search(
cpi, x, bsize, &mvp_full, step_param, sadpb,
cpi->sf.mv.subpel_search_method != SUBPEL_TREE ? cost_list : NULL,
- &bsi->ref_mv[0]->as_mv, new_mv,
- INT_MAX, 1);
+ &bsi->ref_mv[0]->as_mv, new_mv, INT_MAX, 1);
if (bestsme < INT_MAX) {
int distortion;
cpi->find_fractional_mv_step(
- x,
- new_mv,
- &bsi->ref_mv[0]->as_mv,
- cm->allow_high_precision_mv,
+ x, new_mv, &bsi->ref_mv[0]->as_mv, cm->allow_high_precision_mv,
x->errorperbit, &cpi->fn_ptr[bsize],
- cpi->sf.mv.subpel_force_stop,
- cpi->sf.mv.subpel_iters_per_step,
- cond_cost_list(cpi, cost_list),
- x->nmvjointcost, x->mvcost,
- &distortion,
- &x->pred_sse[mbmi->ref_frame[0]],
- NULL, 0, 0);
+ cpi->sf.mv.subpel_force_stop, cpi->sf.mv.subpel_iters_per_step,
+ cond_cost_list(cpi, cost_list), x->nmvjointcost, x->mvcost,
+ &distortion, &x->pred_sse[mbmi->ref_frame[0]], NULL, 0, 0);
// save motion search result for use in compound prediction
seg_mvs[i][mbmi->ref_frame[0]].as_mv = *new_mv;
@@ -1955,9 +1842,8 @@
mi_buf_shift(x, i);
if (cpi->sf.comp_inter_joint_search_thresh <= bsize) {
int rate_mv;
- joint_motion_search(cpi, x, bsize, frame_mv[this_mode],
- mi_row, mi_col, seg_mvs[i],
- &rate_mv);
+ joint_motion_search(cpi, x, bsize, frame_mv[this_mode], mi_row,
+ mi_col, seg_mvs[i], &rate_mv);
seg_mvs[i][mbmi->ref_frame[0]].as_int =
frame_mv[this_mode][mbmi->ref_frame[0]].as_int;
seg_mvs[i][mbmi->ref_frame[1]].as_int =
@@ -1967,10 +1853,9 @@
mi_buf_restore(x, orig_src, orig_pre);
}
- bsi->rdstat[i][mode_idx].brate =
- set_and_cost_bmi_mvs(cpi, x, xd, i, this_mode, mode_mv[this_mode],
- frame_mv, seg_mvs[i], bsi->ref_mv,
- x->nmvjointcost, x->mvcost);
+ bsi->rdstat[i][mode_idx].brate = set_and_cost_bmi_mvs(
+ cpi, x, xd, i, this_mode, mode_mv[this_mode], frame_mv, seg_mvs[i],
+ bsi->ref_mv, x->nmvjointcost, x->mvcost);
for (ref = 0; ref < 1 + has_second_rf; ++ref) {
bsi->rdstat[i][mode_idx].mvs[ref].as_int =
@@ -1985,8 +1870,7 @@
// Trap vectors that reach beyond the UMV borders
if (mv_check_bounds(x, &mode_mv[this_mode][0].as_mv) ||
- (has_second_rf &&
- mv_check_bounds(x, &mode_mv[this_mode][1].as_mv)))
+ (has_second_rf && mv_check_bounds(x, &mode_mv[this_mode][1].as_mv)))
continue;
if (filter_idx > 0) {
@@ -1997,7 +1881,7 @@
for (ref = 0; ref < 1 + has_second_rf; ++ref) {
subpelmv |= mv_has_subpel(&mode_mv[this_mode][ref].as_mv);
have_ref &= mode_mv[this_mode][ref].as_int ==
- ref_bsi->rdstat[i][mode_idx].mvs[ref].as_int;
+ ref_bsi->rdstat[i][mode_idx].mvs[ref].as_int;
}
if (filter_idx > 1 && !subpelmv && !have_ref) {
@@ -2005,7 +1889,7 @@
have_ref = 1;
for (ref = 0; ref < 1 + has_second_rf; ++ref)
have_ref &= mode_mv[this_mode][ref].as_int ==
- ref_bsi->rdstat[i][mode_idx].mvs[ref].as_int;
+ ref_bsi->rdstat[i][mode_idx].mvs[ref].as_int;
}
if (!subpelmv && have_ref &&
@@ -2027,19 +1911,14 @@
}
}
- bsi->rdstat[i][mode_idx].brdcost =
- encode_inter_mb_segment(cpi, x,
- bsi->segment_rd - this_segment_rd, i,
- &bsi->rdstat[i][mode_idx].byrate,
- &bsi->rdstat[i][mode_idx].bdist,
- &bsi->rdstat[i][mode_idx].bsse,
- bsi->rdstat[i][mode_idx].ta,
- bsi->rdstat[i][mode_idx].tl,
- idy, idx,
- mi_row, mi_col);
+ bsi->rdstat[i][mode_idx].brdcost = encode_inter_mb_segment(
+ cpi, x, bsi->segment_rd - this_segment_rd, i,
+ &bsi->rdstat[i][mode_idx].byrate, &bsi->rdstat[i][mode_idx].bdist,
+ &bsi->rdstat[i][mode_idx].bsse, bsi->rdstat[i][mode_idx].ta,
+ bsi->rdstat[i][mode_idx].tl, idy, idx, mi_row, mi_col);
if (bsi->rdstat[i][mode_idx].brdcost < INT64_MAX) {
- bsi->rdstat[i][mode_idx].brdcost += RDCOST(x->rdmult, x->rddiv,
- bsi->rdstat[i][mode_idx].brate, 0);
+ bsi->rdstat[i][mode_idx].brdcost +=
+ RDCOST(x->rdmult, x->rddiv, bsi->rdstat[i][mode_idx].brate, 0);
bsi->rdstat[i][mode_idx].brate += bsi->rdstat[i][mode_idx].byrate;
bsi->rdstat[i][mode_idx].eobs = p->eobs[i];
if (num_4x4_blocks_wide > 1)
@@ -2095,11 +1974,9 @@
bsi->sse = block_sse;
// update the coding decisions
- for (k = 0; k < 4; ++k)
- bsi->modes[k] = mi->bmi[k].as_mode;
+ for (k = 0; k < 4; ++k) bsi->modes[k] = mi->bmi[k].as_mode;
- if (bsi->segment_rd > best_rd)
- return INT64_MAX;
+ if (bsi->segment_rd > best_rd) return INT64_MAX;
/* set it to the best */
for (i = 0; i < 4; i++) {
mode_idx = INTER_OFFSET(bsi->modes[i]);
@@ -2124,16 +2001,15 @@
}
static void estimate_ref_frame_costs(const VP10_COMMON *cm,
- const MACROBLOCKD *xd,
- int segment_id,
+ const MACROBLOCKD *xd, int segment_id,
unsigned int *ref_costs_single,
unsigned int *ref_costs_comp,
vpx_prob *comp_mode_p) {
- int seg_ref_active = segfeature_active(&cm->seg, segment_id,
- SEG_LVL_REF_FRAME);
+ int seg_ref_active =
+ segfeature_active(&cm->seg, segment_id, SEG_LVL_REF_FRAME);
if (seg_ref_active) {
memset(ref_costs_single, 0, MAX_REF_FRAMES * sizeof(*ref_costs_single));
- memset(ref_costs_comp, 0, MAX_REF_FRAMES * sizeof(*ref_costs_comp));
+ memset(ref_costs_comp, 0, MAX_REF_FRAMES * sizeof(*ref_costs_comp));
*comp_mode_p = 128;
} else {
vpx_prob intra_inter_p = vp10_get_intra_inter_prob(cm, xd);
@@ -2158,13 +2034,13 @@
ref_costs_single[LAST_FRAME] = ref_costs_single[GOLDEN_FRAME] =
ref_costs_single[ALTREF_FRAME] = base_cost;
- ref_costs_single[LAST_FRAME] += vp10_cost_bit(ref_single_p1, 0);
+ ref_costs_single[LAST_FRAME] += vp10_cost_bit(ref_single_p1, 0);
ref_costs_single[GOLDEN_FRAME] += vp10_cost_bit(ref_single_p1, 1);
ref_costs_single[ALTREF_FRAME] += vp10_cost_bit(ref_single_p1, 1);
ref_costs_single[GOLDEN_FRAME] += vp10_cost_bit(ref_single_p2, 0);
ref_costs_single[ALTREF_FRAME] += vp10_cost_bit(ref_single_p2, 1);
} else {
- ref_costs_single[LAST_FRAME] = 512;
+ ref_costs_single[LAST_FRAME] = 512;
ref_costs_single[GOLDEN_FRAME] = 512;
ref_costs_single[ALTREF_FRAME] = 512;
}
@@ -2175,20 +2051,19 @@
if (cm->reference_mode == REFERENCE_MODE_SELECT)
base_cost += vp10_cost_bit(comp_inter_p, 1);
- ref_costs_comp[LAST_FRAME] = base_cost + vp10_cost_bit(ref_comp_p, 0);
+ ref_costs_comp[LAST_FRAME] = base_cost + vp10_cost_bit(ref_comp_p, 0);
ref_costs_comp[GOLDEN_FRAME] = base_cost + vp10_cost_bit(ref_comp_p, 1);
} else {
- ref_costs_comp[LAST_FRAME] = 512;
+ ref_costs_comp[LAST_FRAME] = 512;
ref_costs_comp[GOLDEN_FRAME] = 512;
}
}
}
-static void store_coding_context(MACROBLOCK *x, PICK_MODE_CONTEXT *ctx,
- int mode_index,
- int64_t comp_pred_diff[REFERENCE_MODES],
- int64_t best_filter_diff[SWITCHABLE_FILTER_CONTEXTS],
- int skippable) {
+static void store_coding_context(
+ MACROBLOCK *x, PICK_MODE_CONTEXT *ctx, int mode_index,
+ int64_t comp_pred_diff[REFERENCE_MODES],
+ int64_t best_filter_diff[SWITCHABLE_FILTER_CONTEXTS], int skippable) {
MACROBLOCKD *const xd = &x->e_mbd;
// Take a snapshot of the coding context so it can be
@@ -2199,7 +2074,7 @@
ctx->mic = *xd->mi[0];
ctx->mbmi_ext = *x->mbmi_ext;
ctx->single_pred_diff = (int)comp_pred_diff[SINGLE_REFERENCE];
- ctx->comp_pred_diff = (int)comp_pred_diff[COMPOUND_REFERENCE];
+ ctx->comp_pred_diff = (int)comp_pred_diff[COMPOUND_REFERENCE];
ctx->hybrid_pred_diff = (int)comp_pred_diff[REFERENCE_MODE_SELECT];
memcpy(ctx->best_filter_diff, best_filter_diff,
@@ -2208,8 +2083,7 @@
static void setup_buffer_inter(VP10_COMP *cpi, MACROBLOCK *x,
MV_REFERENCE_FRAME ref_frame,
- BLOCK_SIZE block_size,
- int mi_row, int mi_col,
+ BLOCK_SIZE block_size, int mi_row, int mi_col,
int_mv frame_nearest_mv[MAX_REF_FRAMES],
int_mv frame_near_mv[MAX_REF_FRAMES],
struct buf_2d yv12_mb[4][MAX_MB_PLANE]) {
@@ -2228,8 +2102,8 @@
vp10_setup_pred_block(xd, yv12_mb[ref_frame], yv12, mi_row, mi_col, sf, sf);
// Gets an initial list of candidate vectors from neighbours and orders them
- vp10_find_mv_refs(cm, xd, mi, ref_frame, candidates, mi_row, mi_col,
- NULL, NULL, mbmi_ext->mode_context);
+ vp10_find_mv_refs(cm, xd, mi, ref_frame, candidates, mi_row, mi_col, NULL,
+ NULL, mbmi_ext->mode_context);
// Candidate refinement carried out at encoder and decoder
vp10_find_best_ref_mvs(cm->allow_high_precision_mv, candidates,
@@ -2240,18 +2114,17 @@
// in full and choose the best as the centre point for subsequent searches.
// The current implementation doesn't support scaling.
if (!vp10_is_scaled(sf) && block_size >= BLOCK_8X8)
- vp10_mv_pred(cpi, x, yv12_mb[ref_frame][0].buf, yv12->y_stride,
- ref_frame, block_size);
+ vp10_mv_pred(cpi, x, yv12_mb[ref_frame][0].buf, yv12->y_stride, ref_frame,
+ block_size);
}
static void single_motion_search(VP10_COMP *cpi, MACROBLOCK *x,
- BLOCK_SIZE bsize,
- int mi_row, int mi_col,
+ BLOCK_SIZE bsize, int mi_row, int mi_col,
int_mv *tmp_mv, int *rate_mv) {
MACROBLOCKD *xd = &x->e_mbd;
const VP10_COMMON *cm = &cpi->common;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
- struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0, 0}};
+ struct buf_2d backup_yv12[MAX_MB_PLANE] = { { 0, 0 } };
int bestsme = INT_MAX;
int step_param;
int sadpb = x->sadperbit16;
@@ -2265,8 +2138,8 @@
int tmp_row_max = x->mv_row_max;
int cost_list[5];
- const YV12_BUFFER_CONFIG *scaled_ref_frame = vp10_get_scaled_ref_frame(cpi,
- ref);
+ const YV12_BUFFER_CONFIG *scaled_ref_frame =
+ vp10_get_scaled_ref_frame(cpi, ref);
MV pred_mv[3];
pred_mv[0] = x->mbmi_ext->ref_mvs[ref][0].as_mv;
@@ -2278,8 +2151,7 @@
// Swap out the reference frame for a version that's been scaled to
// match the resolution of the current frame, allowing the existing
// motion search code to be used without additional modifications.
- for (i = 0; i < MAX_MB_PLANE; i++)
- backup_yv12[i] = xd->plane[i].pre[0];
+ for (i = 0; i < MAX_MB_PLANE; i++) backup_yv12[i] = xd->plane[i].pre[0];
vp10_setup_pre_planes(xd, 0, scaled_ref_frame, mi_row, mi_col, NULL);
}
@@ -2292,8 +2164,9 @@
// Take wtd average of the step_params based on the last frame's
// max mv magnitude and that based on the best ref mvs of the current
// block for the given reference.
- step_param = (vp10_init_search_range(x->max_mv_context[ref]) +
- cpi->mv_step_param) / 2;
+ step_param =
+ (vp10_init_search_range(x->max_mv_context[ref]) + cpi->mv_step_param) /
+ 2;
} else {
step_param = cpi->mv_step_param;
}
@@ -2310,8 +2183,7 @@
int bhl = b_height_log2_lookup[bsize];
int tlevel = x->pred_mv_sad[ref] >> (bwl + bhl + 4);
- if (tlevel < 5)
- step_param += 2;
+ if (tlevel < 5) step_param += 2;
// prev_mv_sad is not setup for dynamically scaled frames.
if (cpi->oxcf.resize_mode != RESIZE_DYNAMIC) {
@@ -2339,8 +2211,8 @@
mvp_full.row >>= 3;
bestsme = vp10_full_pixel_search(cpi, x, bsize, &mvp_full, step_param, sadpb,
- cond_cost_list(cpi, cost_list),
- &ref_mv, &tmp_mv->as_mv, INT_MAX, 1);
+ cond_cost_list(cpi, cost_list), &ref_mv,
+ &tmp_mv->as_mv, INT_MAX, 1);
x->mv_col_min = tmp_col_min;
x->mv_col_max = tmp_col_max;
@@ -2348,32 +2220,24 @@
x->mv_row_max = tmp_row_max;
if (bestsme < INT_MAX) {
- int dis; /* TODO: use dis in distortion calculation later. */
- cpi->find_fractional_mv_step(x, &tmp_mv->as_mv, &ref_mv,
- cm->allow_high_precision_mv,
- x->errorperbit,
- &cpi->fn_ptr[bsize],
- cpi->sf.mv.subpel_force_stop,
- cpi->sf.mv.subpel_iters_per_step,
- cond_cost_list(cpi, cost_list),
- x->nmvjointcost, x->mvcost,
- &dis, &x->pred_sse[ref], NULL, 0, 0);
+ int dis; /* TODO: use dis in distortion calculation later. */
+ cpi->find_fractional_mv_step(
+ x, &tmp_mv->as_mv, &ref_mv, cm->allow_high_precision_mv, x->errorperbit,
+ &cpi->fn_ptr[bsize], cpi->sf.mv.subpel_force_stop,
+ cpi->sf.mv.subpel_iters_per_step, cond_cost_list(cpi, cost_list),
+ x->nmvjointcost, x->mvcost, &dis, &x->pred_sse[ref], NULL, 0, 0);
}
- *rate_mv = vp10_mv_bit_cost(&tmp_mv->as_mv, &ref_mv,
- x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
+ *rate_mv = vp10_mv_bit_cost(&tmp_mv->as_mv, &ref_mv, x->nmvjointcost,
+ x->mvcost, MV_COST_WEIGHT);
- if (cpi->sf.adaptive_motion_search)
- x->pred_mv[ref] = tmp_mv->as_mv;
+ if (cpi->sf.adaptive_motion_search) x->pred_mv[ref] = tmp_mv->as_mv;
if (scaled_ref_frame) {
int i;
- for (i = 0; i < MAX_MB_PLANE; i++)
- xd->plane[i].pre[0] = backup_yv12[i];
+ for (i = 0; i < MAX_MB_PLANE; i++) xd->plane[i].pre[0] = backup_yv12[i];
}
}
-
-
static INLINE void restore_dst_buf(MACROBLOCKD *xd,
uint8_t *orig_dst[MAX_MB_PLANE],
int orig_dst_stride[MAX_MB_PLANE]) {
@@ -2391,13 +2255,11 @@
// However, once established that vector may be usable through the nearest and
// near mv modes to reduce distortion in subsequent blocks and also improve
// visual quality.
-static int discount_newmv_test(const VP10_COMP *cpi,
- int this_mode,
+static int discount_newmv_test(const VP10_COMP *cpi, int this_mode,
int_mv this_mv,
int_mv (*mode_mv)[MAX_REF_FRAMES],
int ref_frame) {
- return (!cpi->rc.is_src_frame_alt_ref &&
- (this_mode == NEWMV) &&
+ return (!cpi->rc.is_src_frame_alt_ref && (this_mode == NEWMV) &&
(this_mv.as_int != 0) &&
((mode_mv[NEARESTMV][ref_frame].as_int == 0) ||
(mode_mv[NEARESTMV][ref_frame].as_int == INVALID_MV)) &&
@@ -2406,32 +2268,25 @@
}
#define LEFT_TOP_MARGIN ((VPX_ENC_BORDER_IN_PIXELS - VPX_INTERP_EXTEND) << 3)
-#define RIGHT_BOTTOM_MARGIN ((VPX_ENC_BORDER_IN_PIXELS -\
- VPX_INTERP_EXTEND) << 3)
+#define RIGHT_BOTTOM_MARGIN \
+ ((VPX_ENC_BORDER_IN_PIXELS - VPX_INTERP_EXTEND) << 3)
// TODO(jingning): this mv clamping function should be block size dependent.
static INLINE void clamp_mv2(MV *mv, const MACROBLOCKD *xd) {
clamp_mv(mv, xd->mb_to_left_edge - LEFT_TOP_MARGIN,
- xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN,
- xd->mb_to_top_edge - LEFT_TOP_MARGIN,
- xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN);
+ xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN,
+ xd->mb_to_top_edge - LEFT_TOP_MARGIN,
+ xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN);
}
-static int64_t handle_inter_mode(VP10_COMP *cpi, MACROBLOCK *x,
- BLOCK_SIZE bsize,
- int *rate2, int64_t *distortion,
- int *skippable,
- int *rate_y, int *rate_uv,
- int *disable_skip,
- int_mv (*mode_mv)[MAX_REF_FRAMES],
- int mi_row, int mi_col,
- int_mv single_newmv[MAX_REF_FRAMES],
- INTERP_FILTER (*single_filter)[MAX_REF_FRAMES],
- int (*single_skippable)[MAX_REF_FRAMES],
- int64_t *psse,
- const int64_t ref_best_rd,
- int64_t *mask_filter,
- int64_t filter_cache[]) {
+static int64_t handle_inter_mode(
+ VP10_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bsize, int *rate2,
+ int64_t *distortion, int *skippable, int *rate_y, int *rate_uv,
+ int *disable_skip, int_mv (*mode_mv)[MAX_REF_FRAMES], int mi_row,
+ int mi_col, int_mv single_newmv[MAX_REF_FRAMES],
+ INTERP_FILTER (*single_filter)[MAX_REF_FRAMES],
+ int (*single_skippable)[MAX_REF_FRAMES], int64_t *psse,
+ const int64_t ref_best_rd, int64_t *mask_filter, int64_t filter_cache[]) {
VP10_COMMON *cm = &cpi->common;
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
@@ -2441,7 +2296,7 @@
int_mv *frame_mv = mode_mv[this_mode];
int i;
int refs[2] = { mbmi->ref_frame[0],
- (mbmi->ref_frame[1] < 0 ? 0 : mbmi->ref_frame[1]) };
+ (mbmi->ref_frame[1] < 0 ? 0 : mbmi->ref_frame[1]) };
int_mv cur_mv[2];
#if CONFIG_VPX_HIGHBITDEPTH
DECLARE_ALIGNED(16, uint16_t, tmp_buf16[MAX_MB_PLANE * 64 * 64]);
@@ -2457,13 +2312,16 @@
int orig_dst_stride[MAX_MB_PLANE];
int rs = 0;
INTERP_FILTER best_filter = SWITCHABLE;
- uint8_t skip_txfm[MAX_MB_PLANE << 2] = {0};
- int64_t bsse[MAX_MB_PLANE << 2] = {0};
+ uint8_t skip_txfm[MAX_MB_PLANE << 2] = { 0 };
+ int64_t bsse[MAX_MB_PLANE << 2] = { 0 };
int bsl = mi_width_log2_lookup[bsize];
- int pred_filter_search = cpi->sf.cb_pred_filter_search ?
- (((mi_row + mi_col) >> bsl) +
- get_chessboard_index(cm->current_video_frame)) & 0x1 : 0;
+ int pred_filter_search =
+ cpi->sf.cb_pred_filter_search
+ ? (((mi_row + mi_col) >> bsl) +
+ get_chessboard_index(cm->current_video_frame)) &
+ 0x1
+ : 0;
int skip_txfm_sb = 0;
int64_t skip_sse_sb = INT64_MAX;
@@ -2479,13 +2337,10 @@
if (pred_filter_search) {
INTERP_FILTER af = SWITCHABLE, lf = SWITCHABLE;
- if (xd->up_available)
- af = xd->mi[-xd->mi_stride]->mbmi.interp_filter;
- if (xd->left_available)
- lf = xd->mi[-1]->mbmi.interp_filter;
+ if (xd->up_available) af = xd->mi[-xd->mi_stride]->mbmi.interp_filter;
+ if (xd->left_available) lf = xd->mi[-1]->mbmi.interp_filter;
- if ((this_mode != NEWMV) || (af == lf))
- best_filter = af;
+ if ((this_mode != NEWMV) || (af == lf)) best_filter = af;
}
if (is_comp_pred) {
@@ -2508,26 +2363,24 @@
frame_mv[refs[1]].as_int = single_newmv[refs[1]].as_int;
if (cpi->sf.comp_inter_joint_search_thresh <= bsize) {
- joint_motion_search(cpi, x, bsize, frame_mv,
- mi_row, mi_col, single_newmv, &rate_mv);
+ joint_motion_search(cpi, x, bsize, frame_mv, mi_row, mi_col,
+ single_newmv, &rate_mv);
} else {
- rate_mv = vp10_mv_bit_cost(&frame_mv[refs[0]].as_mv,
+ rate_mv = vp10_mv_bit_cost(&frame_mv[refs[0]].as_mv,
&x->mbmi_ext->ref_mvs[refs[0]][0].as_mv,
x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
rate_mv += vp10_mv_bit_cost(&frame_mv[refs[1]].as_mv,
- &x->mbmi_ext->ref_mvs[refs[1]][0].as_mv,
- x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
+ &x->mbmi_ext->ref_mvs[refs[1]][0].as_mv,
+ x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
}
*rate2 += rate_mv;
} else {
int_mv tmp_mv;
- single_motion_search(cpi, x, bsize, mi_row, mi_col,
- &tmp_mv, &rate_mv);
- if (tmp_mv.as_int == INVALID_MV)
- return INT64_MAX;
+ single_motion_search(cpi, x, bsize, mi_row, mi_col, &tmp_mv, &rate_mv);
+ if (tmp_mv.as_int == INVALID_MV) return INT64_MAX;
- frame_mv[refs[0]].as_int =
- xd->mi[0]->bmi[0].as_mv[0].as_int = tmp_mv.as_int;
+ frame_mv[refs[0]].as_int = xd->mi[0]->bmi[0].as_mv[0].as_int =
+ tmp_mv.as_int;
single_newmv[refs[0]].as_int = tmp_mv.as_int;
// Estimate the rate implications of a new mv but discount this
@@ -2545,11 +2398,9 @@
for (i = 0; i < is_comp_pred + 1; ++i) {
cur_mv[i] = frame_mv[refs[i]];
// Clip "next_nearest" so that it does not extend to far out of image
- if (this_mode != NEWMV)
- clamp_mv2(&cur_mv[i].as_mv, xd);
+ if (this_mode != NEWMV) clamp_mv2(&cur_mv[i].as_mv, xd);
- if (mv_check_bounds(x, &cur_mv[i].as_mv))
- return INT64_MAX;
+ if (mv_check_bounds(x, &cur_mv[i].as_mv)) return INT64_MAX;
mbmi->mv[i].as_int = cur_mv[i].as_int;
}
@@ -2570,12 +2421,11 @@
//
// Under some circumstances we discount the cost of new mv mode to encourage
// initiation of a motion field.
- if (discount_newmv_test(cpi, this_mode, frame_mv[refs[0]],
- mode_mv, refs[0])) {
- *rate2 += VPXMIN(cost_mv_ref(cpi, this_mode,
- mbmi_ext->mode_context[refs[0]]),
- cost_mv_ref(cpi, NEARESTMV,
- mbmi_ext->mode_context[refs[0]]));
+ if (discount_newmv_test(cpi, this_mode, frame_mv[refs[0]], mode_mv,
+ refs[0])) {
+ *rate2 +=
+ VPXMIN(cost_mv_ref(cpi, this_mode, mbmi_ext->mode_context[refs[0]]),
+ cost_mv_ref(cpi, NEARESTMV, mbmi_ext->mode_context[refs[0]]));
} else {
*rate2 += cost_mv_ref(cpi, this_mode, mbmi_ext->mode_context[refs[0]]);
}
@@ -2587,13 +2437,11 @@
pred_exists = 0;
// Are all MVs integer pel for Y and UV
intpel_mv = !mv_has_subpel(&mbmi->mv[0].as_mv);
- if (is_comp_pred)
- intpel_mv &= !mv_has_subpel(&mbmi->mv[1].as_mv);
+ if (is_comp_pred) intpel_mv &= !mv_has_subpel(&mbmi->mv[1].as_mv);
// Search for best switchable filter by checking the variance of
// pred error irrespective of whether the filter will be used
- for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i)
- filter_cache[i] = INT64_MAX;
+ for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i) filter_cache[i] = INT64_MAX;
if (cm->interp_filter != BILINEAR) {
if (x->source_variance < cpi->sf.disable_filter_search_var_thresh) {
@@ -2618,8 +2466,7 @@
filter_cache[i] = rd;
filter_cache[SWITCHABLE_FILTERS] =
VPXMIN(filter_cache[SWITCHABLE_FILTERS], rd + rs_rd);
- if (cm->interp_filter == SWITCHABLE)
- rd += rs_rd;
+ if (cm->interp_filter == SWITCHABLE) rd += rs_rd;
*mask_filter = VPXMAX(*mask_filter, rd);
} else {
int rate_sum = 0;
@@ -2631,8 +2478,7 @@
continue;
}
- if ((cm->interp_filter == SWITCHABLE &&
- (!i || best_needs_copy)) ||
+ if ((cm->interp_filter == SWITCHABLE && (!i || best_needs_copy)) ||
(cm->interp_filter != SWITCHABLE &&
(cm->interp_filter == mbmi->interp_filter ||
(i == 0 && intpel_mv)))) {
@@ -2644,15 +2490,14 @@
}
}
vp10_build_inter_predictors_sb(xd, mi_row, mi_col, bsize);
- model_rd_for_sb(cpi, bsize, x, xd, &rate_sum, &dist_sum,
- &tmp_skip_sb, &tmp_skip_sse);
+ model_rd_for_sb(cpi, bsize, x, xd, &rate_sum, &dist_sum, &tmp_skip_sb,
+ &tmp_skip_sse);
rd = RDCOST(x->rdmult, x->rddiv, rate_sum, dist_sum);
filter_cache[i] = rd;
filter_cache[SWITCHABLE_FILTERS] =
VPXMIN(filter_cache[SWITCHABLE_FILTERS], rd + rs_rd);
- if (cm->interp_filter == SWITCHABLE)
- rd += rs_rd;
+ if (cm->interp_filter == SWITCHABLE) rd += rs_rd;
*mask_filter = VPXMAX(*mask_filter, rd);
if (i == 0 && intpel_mv) {
@@ -2692,8 +2537,8 @@
}
}
// Set the appropriate filter
- mbmi->interp_filter = cm->interp_filter != SWITCHABLE ?
- cm->interp_filter : best_filter;
+ mbmi->interp_filter =
+ cm->interp_filter != SWITCHABLE ? cm->interp_filter : best_filter;
rs = cm->interp_filter == SWITCHABLE ? vp10_get_switchable_rate(cpi, xd) : 0;
if (pred_exists) {
@@ -2712,15 +2557,14 @@
// switchable list (ex. bilinear) is indicated at the frame level, or
// skip condition holds.
vp10_build_inter_predictors_sb(xd, mi_row, mi_col, bsize);
- model_rd_for_sb(cpi, bsize, x, xd, &tmp_rate, &tmp_dist,
- &skip_txfm_sb, &skip_sse_sb);
+ model_rd_for_sb(cpi, bsize, x, xd, &tmp_rate, &tmp_dist, &skip_txfm_sb,
+ &skip_sse_sb);
rd = RDCOST(x->rdmult, x->rddiv, rs + tmp_rate, tmp_dist);
memcpy(skip_txfm, x->skip_txfm, sizeof(skip_txfm));
memcpy(bsse, x->bsse, sizeof(bsse));
}
- if (!is_comp_pred)
- single_filter[this_mode][refs[0]] = mbmi->interp_filter;
+ if (!is_comp_pred) single_filter[this_mode][refs[0]] = mbmi->interp_filter;
if (cpi->sf.adaptive_mode_search)
if (is_comp_pred)
@@ -2737,8 +2581,7 @@
}
}
- if (cm->interp_filter == SWITCHABLE)
- *rate2 += rs;
+ if (cm->interp_filter == SWITCHABLE) *rate2 += rs;
memcpy(x->skip_txfm, skip_txfm, sizeof(skip_txfm));
memcpy(x->bsse, bsse, sizeof(bsse));
@@ -2750,8 +2593,8 @@
// Y cost and distortion
vp10_subtract_plane(x, bsize, 0);
- super_block_yrd(cpi, x, rate_y, &distortion_y, &skippable_y, psse,
- bsize, ref_best_rd);
+ super_block_yrd(cpi, x, rate_y, &distortion_y, &skippable_y, psse, bsize,
+ ref_best_rd);
if (*rate_y == INT_MAX) {
*rate2 = INT_MAX;
@@ -2788,16 +2631,15 @@
*distortion = skip_sse_sb;
}
- if (!is_comp_pred)
- single_skippable[this_mode][refs[0]] = *skippable;
+ if (!is_comp_pred) single_skippable[this_mode][refs[0]] = *skippable;
restore_dst_buf(xd, orig_dst, orig_dst_stride);
return 0; // The rate-distortion cost will be re-calculated by caller.
}
-void vp10_rd_pick_intra_mode_sb(VP10_COMP *cpi, MACROBLOCK *x,
- RD_COST *rd_cost, BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx, int64_t best_rd) {
+void vp10_rd_pick_intra_mode_sb(VP10_COMP *cpi, MACROBLOCK *x, RD_COST *rd_cost,
+ BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd) {
VP10_COMMON *const cm = &cpi->common;
MACROBLOCKD *const xd = &x->e_mbd;
struct macroblockd_plane *const pd = xd->plane;
@@ -2810,9 +2652,8 @@
xd->mi[0]->mbmi.ref_frame[1] = NONE;
if (bsize >= BLOCK_8X8) {
- if (rd_pick_intra_sby_mode(cpi, x, &rate_y, &rate_y_tokenonly,
- &dist_y, &y_skip, bsize,
- best_rd) >= best_rd) {
+ if (rd_pick_intra_sby_mode(cpi, x, &rate_y, &rate_y_tokenonly, &dist_y,
+ &y_skip, bsize, best_rd) >= best_rd) {
rd_cost->rate = INT_MAX;
return;
}
@@ -2824,20 +2665,18 @@
return;
}
}
- max_uv_tx_size = get_uv_tx_size_impl(xd->mi[0]->mbmi.tx_size, bsize,
- pd[1].subsampling_x,
- pd[1].subsampling_y);
- rd_pick_intra_sbuv_mode(cpi, x, ctx, &rate_uv, &rate_uv_tokenonly,
- &dist_uv, &uv_skip, VPXMAX(BLOCK_8X8, bsize),
- max_uv_tx_size);
+ max_uv_tx_size = get_uv_tx_size_impl(
+ xd->mi[0]->mbmi.tx_size, bsize, pd[1].subsampling_x, pd[1].subsampling_y);
+ rd_pick_intra_sbuv_mode(cpi, x, ctx, &rate_uv, &rate_uv_tokenonly, &dist_uv,
+ &uv_skip, VPXMAX(BLOCK_8X8, bsize), max_uv_tx_size);
if (y_skip && uv_skip) {
rd_cost->rate = rate_y + rate_uv - rate_y_tokenonly - rate_uv_tokenonly +
vp10_cost_bit(vp10_get_skip_prob(cm, xd), 1);
rd_cost->dist = dist_y + dist_uv;
} else {
- rd_cost->rate = rate_y + rate_uv +
- vp10_cost_bit(vp10_get_skip_prob(cm, xd), 0);
+ rd_cost->rate =
+ rate_y + rate_uv + vp10_cost_bit(vp10_get_skip_prob(cm, xd), 0);
rd_cost->dist = dist_y + dist_uv;
}
@@ -2851,10 +2690,8 @@
#define LOW_VAR_THRESH 16
#define VLOW_ADJ_MAX 25
#define VHIGH_ADJ_MAX 8
-static void rd_variance_adjustment(VP10_COMP *cpi,
- MACROBLOCK *x,
- BLOCK_SIZE bsize,
- int64_t *this_rd,
+static void rd_variance_adjustment(VP10_COMP *cpi, MACROBLOCK *x,
+ BLOCK_SIZE bsize, int64_t *this_rd,
MV_REFERENCE_FRAME ref_frame,
unsigned int source_variance) {
MACROBLOCKD *const xd = &x->e_mbd;
@@ -2863,30 +2700,29 @@
int64_t var_error = 0;
int64_t var_factor = 0;
- if (*this_rd == INT64_MAX)
- return;
+ if (*this_rd == INT64_MAX) return;
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- recon_variance =
- vp10_high_get_sby_perpixel_variance(cpi, &xd->plane[0].dst, bsize, xd->bd);
+ recon_variance = vp10_high_get_sby_perpixel_variance(cpi, &xd->plane[0].dst,
+ bsize, xd->bd);
} else {
recon_variance =
- vp10_get_sby_perpixel_variance(cpi, &xd->plane[0].dst, bsize);
+ vp10_get_sby_perpixel_variance(cpi, &xd->plane[0].dst, bsize);
}
#else
recon_variance =
- vp10_get_sby_perpixel_variance(cpi, &xd->plane[0].dst, bsize);
+ vp10_get_sby_perpixel_variance(cpi, &xd->plane[0].dst, bsize);
#endif // CONFIG_VPX_HIGHBITDEPTH
if ((source_variance + recon_variance) > LOW_VAR_THRESH) {
absvar_diff = (source_variance > recon_variance)
- ? (source_variance - recon_variance)
- : (recon_variance - source_variance);
+ ? (source_variance - recon_variance)
+ : (recon_variance - source_variance);
var_error = ((int64_t)200 * source_variance * recon_variance) /
- (((int64_t)source_variance * source_variance) +
- ((int64_t)recon_variance * recon_variance));
+ (((int64_t)source_variance * source_variance) +
+ ((int64_t)recon_variance * recon_variance));
var_error = 100 - var_error;
}
@@ -2896,8 +2732,8 @@
if ((source_variance > LOW_VAR_THRESH) && (ref_frame == INTRA_FRAME) &&
(source_variance > recon_variance)) {
var_factor = VPXMIN(absvar_diff, VPXMIN(VLOW_ADJ_MAX, var_error));
- // A second possible case of interest is where the source variance
- // is very low and we wish to discourage false texture or motion trails.
+ // A second possible case of interest is where the source variance
+ // is very low and we wish to discourage false texture or motion trails.
} else if ((source_variance < (LOW_VAR_THRESH >> 1)) &&
(recon_variance > source_variance)) {
var_factor = VPXMIN(absvar_diff, VPXMIN(VHIGH_ADJ_MAX, var_error));
@@ -2905,12 +2741,11 @@
*this_rd += (*this_rd * var_factor) / 100;
}
-
// Do we have an internal image edge (e.g. formatting bars).
int vp10_internal_image_edge(VP10_COMP *cpi) {
return (cpi->oxcf.pass == 2) &&
- ((cpi->twopass.this_frame_stats.inactive_zone_rows > 0) ||
- (cpi->twopass.this_frame_stats.inactive_zone_cols > 0));
+ ((cpi->twopass.this_frame_stats.inactive_zone_rows > 0) ||
+ (cpi->twopass.this_frame_stats.inactive_zone_cols > 0));
}
// Checks to see if a super block is on a horizontal image edge.
@@ -2970,16 +2805,13 @@
// Checks to see if a super block is at the edge of the active image.
// In most cases this is the "real" edge unless there are formatting
// bars embedded in the stream.
-int vp10_active_edge_sb(VP10_COMP *cpi,
- int mi_row, int mi_col) {
+int vp10_active_edge_sb(VP10_COMP *cpi, int mi_row, int mi_col) {
return vp10_active_h_edge(cpi, mi_row, MI_BLOCK_SIZE) ||
vp10_active_v_edge(cpi, mi_col, MI_BLOCK_SIZE);
}
-void vp10_rd_pick_inter_mode_sb(VP10_COMP *cpi,
- TileDataEnc *tile_data,
- MACROBLOCK *x,
- int mi_row, int mi_col,
+void vp10_rd_pick_inter_mode_sb(VP10_COMP *cpi, TileDataEnc *tile_data,
+ MACROBLOCK *x, int mi_row, int mi_col,
RD_COST *rd_cost, BLOCK_SIZE bsize,
PICK_MODE_CONTEXT *ctx,
int64_t best_rd_so_far) {
@@ -3034,20 +2866,16 @@
vp10_zero(best_mbmode);
- for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i)
- filter_cache[i] = INT64_MAX;
+ for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i) filter_cache[i] = INT64_MAX;
estimate_ref_frame_costs(cm, xd, segment_id, ref_costs_single, ref_costs_comp,
&comp_mode_p);
- for (i = 0; i < REFERENCE_MODES; ++i)
- best_pred_rd[i] = INT64_MAX;
+ for (i = 0; i < REFERENCE_MODES; ++i) best_pred_rd[i] = INT64_MAX;
for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++)
best_filter_rd[i] = INT64_MAX;
- for (i = 0; i < TX_SIZES; i++)
- rate_uv_intra[i] = INT_MAX;
- for (i = 0; i < MAX_REF_FRAMES; ++i)
- x->pred_sse[i] = INT_MAX;
+ for (i = 0; i < TX_SIZES; i++) rate_uv_intra[i] = INT_MAX;
+ for (i = 0; i < MAX_REF_FRAMES; ++i) x->pred_sse[i] = INT_MAX;
for (i = 0; i < MB_MODE_COUNT; ++i) {
for (k = 0; k < MAX_REF_FRAMES; ++k) {
single_inter_filter[i][k] = SWITCHABLE;
@@ -3140,12 +2968,11 @@
mode_skip_mask[INTRA_FRAME] |=
~(sf->intra_y_mode_mask[max_txsize_lookup[bsize]]);
- for (i = 0; i <= LAST_NEW_MV_INDEX; ++i)
- mode_threshold[i] = 0;
+ for (i = 0; i <= LAST_NEW_MV_INDEX; ++i) mode_threshold[i] = 0;
for (i = LAST_NEW_MV_INDEX + 1; i < MAX_MODES; ++i)
mode_threshold[i] = ((int64_t)rd_threshes[i] * rd_thresh_freq_fact[i]) >> 5;
- midx = sf->schedule_mode_search ? mode_skip_start : 0;
+ midx = sf->schedule_mode_search ? mode_skip_start : 0;
while (midx > 4) {
uint8_t end_pos = 0;
for (i = 5; i < midx; ++i) {
@@ -3180,8 +3007,7 @@
// skip mask to look at a subset of the remaining modes.
if (midx == mode_skip_start && best_mode_index >= 0) {
switch (best_mbmode.ref_frame[0]) {
- case INTRA_FRAME:
- break;
+ case INTRA_FRAME: break;
case LAST_FRAME:
ref_frame_skip_mask[0] |= LAST_FRAME_MODE_MASK;
ref_frame_skip_mask[1] |= SECOND_REF_FRAME_MASK;
@@ -3190,13 +3016,9 @@
ref_frame_skip_mask[0] |= GOLDEN_FRAME_MODE_MASK;
ref_frame_skip_mask[1] |= SECOND_REF_FRAME_MASK;
break;
- case ALTREF_FRAME:
- ref_frame_skip_mask[0] |= ALT_REF_MODE_MASK;
- break;
+ case ALTREF_FRAME: ref_frame_skip_mask[0] |= ALT_REF_MODE_MASK; break;
case NONE:
- case MAX_REF_FRAMES:
- assert(0 && "Invalid Reference frame");
- break;
+ case MAX_REF_FRAMES: assert(0 && "Invalid Reference frame"); break;
}
}
@@ -3204,29 +3026,24 @@
(ref_frame_skip_mask[1] & (1 << VPXMAX(0, second_ref_frame))))
continue;
- if (mode_skip_mask[ref_frame] & (1 << this_mode))
- continue;
+ if (mode_skip_mask[ref_frame] & (1 << this_mode)) continue;
// Test best rd so far against threshold for trying this mode.
if (best_mode_skippable && sf->schedule_mode_search)
mode_threshold[mode_index] <<= 1;
- if (best_rd < mode_threshold[mode_index])
- continue;
+ if (best_rd < mode_threshold[mode_index]) continue;
comp_pred = second_ref_frame > INTRA_FRAME;
if (comp_pred) {
- if (!cpi->allow_comp_inter_inter)
- continue;
+ if (!cpi->allow_comp_inter_inter) continue;
// Skip compound inter modes if ARF is not available.
- if (!(cpi->ref_frame_flags & flag_list[second_ref_frame]))
- continue;
+ if (!(cpi->ref_frame_flags & flag_list[second_ref_frame])) continue;
// Do not allow compound prediction if the segment level reference frame
// feature is in use as in this case there can only be one reference.
- if (segfeature_active(seg, segment_id, SEG_LVL_REF_FRAME))
- continue;
+ if (segfeature_active(seg, segment_id, SEG_LVL_REF_FRAME)) continue;
if ((mode_search_skip_flags & FLAG_SKIP_COMP_BESTINTRA) &&
best_mode_index >= 0 && best_mbmode.ref_frame[0] == INTRA_FRAME)
@@ -3255,19 +3072,17 @@
// one of the neighboring directional modes
if ((mode_search_skip_flags & FLAG_SKIP_INTRA_BESTINTER) &&
(this_mode >= D45_PRED && this_mode <= TM_PRED)) {
- if (best_mode_index >= 0 &&
- best_mbmode.ref_frame[0] > INTRA_FRAME)
+ if (best_mode_index >= 0 && best_mbmode.ref_frame[0] > INTRA_FRAME)
continue;
}
if (mode_search_skip_flags & FLAG_SKIP_INTRA_DIRMISMATCH) {
- if (conditional_skipintra(this_mode, best_intra_mode))
- continue;
+ if (conditional_skipintra(this_mode, best_intra_mode)) continue;
}
}
} else {
- const MV_REFERENCE_FRAME ref_frames[2] = {ref_frame, second_ref_frame};
- if (!check_best_zero_mv(cpi, mbmi_ext->mode_context, frame_mv,
- this_mode, ref_frames))
+ const MV_REFERENCE_FRAME ref_frames[2] = { ref_frame, second_ref_frame };
+ if (!check_best_zero_mv(cpi, mbmi_ext->mode_context, frame_mv, this_mode,
+ ref_frames))
continue;
}
@@ -3277,8 +3092,8 @@
mbmi->ref_frame[1] = second_ref_frame;
// Evaluate all sub-pel filters irrespective of whether we can use
// them for this frame.
- mbmi->interp_filter = cm->interp_filter == SWITCHABLE ? EIGHTTAP
- : cm->interp_filter;
+ mbmi->interp_filter =
+ cm->interp_filter == SWITCHABLE ? EIGHTTAP : cm->interp_filter;
mbmi->mv[0].as_int = mbmi->mv[1].as_int = 0;
x->skip = 0;
@@ -3287,25 +3102,23 @@
// Select prediction reference frames.
for (i = 0; i < MAX_MB_PLANE; i++) {
xd->plane[i].pre[0] = yv12_mb[ref_frame][i];
- if (comp_pred)
- xd->plane[i].pre[1] = yv12_mb[second_ref_frame][i];
+ if (comp_pred) xd->plane[i].pre[1] = yv12_mb[second_ref_frame][i];
}
if (ref_frame == INTRA_FRAME) {
TX_SIZE uv_tx;
struct macroblockd_plane *const pd = &xd->plane[1];
memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
- super_block_yrd(cpi, x, &rate_y, &distortion_y, &skippable,
- NULL, bsize, best_rd);
- if (rate_y == INT_MAX)
- continue;
+ super_block_yrd(cpi, x, &rate_y, &distortion_y, &skippable, NULL, bsize,
+ best_rd);
+ if (rate_y == INT_MAX) continue;
uv_tx = get_uv_tx_size_impl(mbmi->tx_size, bsize, pd->subsampling_x,
pd->subsampling_y);
if (rate_uv_intra[uv_tx] == INT_MAX) {
- choose_intra_uv_mode(cpi, x, ctx, bsize, uv_tx,
- &rate_uv_intra[uv_tx], &rate_uv_tokenonly[uv_tx],
- &dist_uv[uv_tx], &skip_uv[uv_tx], &mode_uv[uv_tx]);
+ choose_intra_uv_mode(cpi, x, ctx, bsize, uv_tx, &rate_uv_intra[uv_tx],
+ &rate_uv_tokenonly[uv_tx], &dist_uv[uv_tx],
+ &skip_uv[uv_tx], &mode_uv[uv_tx]);
}
rate_uv = rate_uv_tokenonly[uv_tx];
@@ -3318,21 +3131,16 @@
rate2 += intra_cost_penalty;
distortion2 = distortion_y + distortion_uv;
} else {
- this_rd = handle_inter_mode(cpi, x, bsize,
- &rate2, &distortion2, &skippable,
- &rate_y, &rate_uv,
- &disable_skip, frame_mv,
- mi_row, mi_col,
- single_newmv, single_inter_filter,
- single_skippable, &total_sse, best_rd,
- &mask_filter, filter_cache);
- if (this_rd == INT64_MAX)
- continue;
+ this_rd = handle_inter_mode(
+ cpi, x, bsize, &rate2, &distortion2, &skippable, &rate_y, &rate_uv,
+ &disable_skip, frame_mv, mi_row, mi_col, single_newmv,
+ single_inter_filter, single_skippable, &total_sse, best_rd,
+ &mask_filter, filter_cache);
+ if (this_rd == INT64_MAX) continue;
compmode_cost = vp10_cost_bit(comp_mode_p, comp_pred);
- if (cm->reference_mode == REFERENCE_MODE_SELECT)
- rate2 += compmode_cost;
+ if (cm->reference_mode == REFERENCE_MODE_SELECT) rate2 += compmode_cost;
}
// Estimate the reference frame signaling cost and add it
@@ -3374,11 +3182,11 @@
// Apply an adjustment to the rd value based on the similarity of the
// source variance and reconstructed variance.
- rd_variance_adjustment(cpi, x, bsize, &this_rd,
- ref_frame, x->source_variance);
+ rd_variance_adjustment(cpi, x, bsize, &this_rd, ref_frame,
+ x->source_variance);
if (ref_frame == INTRA_FRAME) {
- // Keep record of best intra rd
+ // Keep record of best intra rd
if (this_rd < best_intra_rd) {
best_intra_rd = this_rd;
best_intra_mode = mbmi->mode;
@@ -3415,8 +3223,7 @@
best_skip2 = this_skip2;
best_mode_skippable = skippable;
- if (!x->select_tx_size)
- swap_block_ptr(x, ctx, 1, 0, 0, max_plane);
+ if (!x->select_tx_size) swap_block_ptr(x, ctx, 1, 0, 0, max_plane);
memcpy(ctx->zcoeff_blk, x->zcoeff_blk[mbmi->tx_size],
sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
@@ -3436,8 +3243,7 @@
const int var_adjust = (x->source_variance < 16);
scale -= var_adjust;
}
- if (ref_frame > INTRA_FRAME &&
- distortion2 * scale < qstep * qstep) {
+ if (ref_frame > INTRA_FRAME && distortion2 * scale < qstep * qstep) {
early_term = 1;
}
}
@@ -3471,8 +3277,9 @@
/* keep record of best filter type */
if (!mode_excluded && cm->interp_filter != BILINEAR) {
- int64_t ref = filter_cache[cm->interp_filter == SWITCHABLE ?
- SWITCHABLE_FILTERS : cm->interp_filter];
+ int64_t ref =
+ filter_cache[cm->interp_filter == SWITCHABLE ? SWITCHABLE_FILTERS
+ : cm->interp_filter];
for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++) {
int64_t adj_rd;
@@ -3493,11 +3300,9 @@
}
}
- if (early_term)
- break;
+ if (early_term) break;
- if (x->skip && !comp_pred)
- break;
+ if (x->skip && !comp_pred) break;
}
// The inter modes' rate costs are not calculated precisely in some cases.
@@ -3505,20 +3310,23 @@
// ZEROMV. Here, checks are added for those cases, and the mode decisions
// are corrected.
if (best_mbmode.mode == NEWMV) {
- const MV_REFERENCE_FRAME refs[2] = {best_mbmode.ref_frame[0],
- best_mbmode.ref_frame[1]};
+ const MV_REFERENCE_FRAME refs[2] = { best_mbmode.ref_frame[0],
+ best_mbmode.ref_frame[1] };
int comp_pred_mode = refs[1] > INTRA_FRAME;
if (frame_mv[NEARESTMV][refs[0]].as_int == best_mbmode.mv[0].as_int &&
- ((comp_pred_mode && frame_mv[NEARESTMV][refs[1]].as_int ==
- best_mbmode.mv[1].as_int) || !comp_pred_mode))
+ ((comp_pred_mode &&
+ frame_mv[NEARESTMV][refs[1]].as_int == best_mbmode.mv[1].as_int) ||
+ !comp_pred_mode))
best_mbmode.mode = NEARESTMV;
else if (frame_mv[NEARMV][refs[0]].as_int == best_mbmode.mv[0].as_int &&
- ((comp_pred_mode && frame_mv[NEARMV][refs[1]].as_int ==
- best_mbmode.mv[1].as_int) || !comp_pred_mode))
+ ((comp_pred_mode &&
+ frame_mv[NEARMV][refs[1]].as_int == best_mbmode.mv[1].as_int) ||
+ !comp_pred_mode))
best_mbmode.mode = NEARMV;
else if (best_mbmode.mv[0].as_int == 0 &&
- ((comp_pred_mode && best_mbmode.mv[1].as_int == 0) || !comp_pred_mode))
+ ((comp_pred_mode && best_mbmode.mv[1].as_int == 0) ||
+ !comp_pred_mode))
best_mbmode.mode = ZEROMV;
}
@@ -3537,8 +3345,7 @@
uv_tx_size = get_uv_tx_size(mbmi, &xd->plane[1]);
rd_pick_intra_sbuv_mode(cpi, x, ctx, &rate_uv_intra[uv_tx_size],
&rate_uv_tokenonly[uv_tx_size],
- &dist_uv[uv_tx_size],
- &skip_uv[uv_tx_size],
+ &dist_uv[uv_tx_size], &skip_uv[uv_tx_size],
bsize < BLOCK_8X8 ? BLOCK_8X8 : bsize,
uv_tx_size);
}
@@ -3550,7 +3357,7 @@
if (!cpi->rc.is_src_frame_alt_ref)
vp10_update_rd_thresh_fact(tile_data->thresh_freq_fact,
- sf->adaptive_rd_thresh, bsize, best_mode_index);
+ sf->adaptive_rd_thresh, bsize, best_mode_index);
// macroblock modes
*mbmi = best_mbmode;
@@ -3583,8 +3390,7 @@
if (!x->skip && !x->select_tx_size) {
int has_high_freq_coeff = 0;
int plane;
- int max_plane = is_inter_block(&xd->mi[0]->mbmi)
- ? MAX_MB_PLANE : 1;
+ int max_plane = is_inter_block(&xd->mi[0]->mbmi) ? MAX_MB_PLANE : 1;
for (plane = 0; plane < max_plane; ++plane) {
x->plane[plane].eobs = ctx->eobs_pbuf[plane][1];
has_high_freq_coeff |= vp10_has_high_freq_in_plane(x, bsize, plane);
@@ -3604,13 +3410,11 @@
best_filter_diff, best_mode_skippable);
}
-void vp10_rd_pick_inter_mode_sb_seg_skip(VP10_COMP *cpi,
- TileDataEnc *tile_data,
- MACROBLOCK *x,
- RD_COST *rd_cost,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far) {
+void vp10_rd_pick_inter_mode_sb_seg_skip(VP10_COMP *cpi, TileDataEnc *tile_data,
+ MACROBLOCK *x, RD_COST *rd_cost,
+ BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far) {
VP10_COMMON *const cm = &cpi->common;
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
@@ -3629,10 +3433,8 @@
estimate_ref_frame_costs(cm, xd, segment_id, ref_costs_single, ref_costs_comp,
&comp_mode_p);
- for (i = 0; i < MAX_REF_FRAMES; ++i)
- x->pred_sse[i] = INT_MAX;
- for (i = LAST_FRAME; i < MAX_REF_FRAMES; ++i)
- x->pred_mv_sad[i] = INT_MAX;
+ for (i = 0; i < MAX_REF_FRAMES; ++i) x->pred_sse[i] = INT_MAX;
+ for (i = LAST_FRAME; i < MAX_REF_FRAMES; ++i) x->pred_mv_sad[i] = INT_MAX;
rd_cost->rate = INT_MAX;
@@ -3691,25 +3493,20 @@
(cm->interp_filter == mbmi->interp_filter));
vp10_update_rd_thresh_fact(tile_data->thresh_freq_fact,
- cpi->sf.adaptive_rd_thresh, bsize, THR_ZEROMV);
+ cpi->sf.adaptive_rd_thresh, bsize, THR_ZEROMV);
vp10_zero(best_pred_diff);
vp10_zero(best_filter_diff);
- if (!x->select_tx_size)
- swap_block_ptr(x, ctx, 1, 0, 0, MAX_MB_PLANE);
- store_coding_context(x, ctx, THR_ZEROMV,
- best_pred_diff, best_filter_diff, 0);
+ if (!x->select_tx_size) swap_block_ptr(x, ctx, 1, 0, 0, MAX_MB_PLANE);
+ store_coding_context(x, ctx, THR_ZEROMV, best_pred_diff, best_filter_diff, 0);
}
-void vp10_rd_pick_inter_mode_sub8x8(VP10_COMP *cpi,
- TileDataEnc *tile_data,
- MACROBLOCK *x,
- int mi_row, int mi_col,
- RD_COST *rd_cost,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far) {
+void vp10_rd_pick_inter_mode_sub8x8(VP10_COMP *cpi, TileDataEnc *tile_data,
+ MACROBLOCK *x, int mi_row, int mi_col,
+ RD_COST *rd_cost, BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far) {
VP10_COMMON *const cm = &cpi->common;
RD_OPT *const rd_opt = &cpi->rd;
SPEED_FEATURES *const sf = &cpi->sf;
@@ -3739,7 +3536,7 @@
int skip_uv;
PREDICTION_MODE mode_uv = DC_PRED;
const int intra_cost_penalty = vp10_get_intra_cost_penalty(
- cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth);
+ cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth);
int_mv seg_mvs[4][MAX_REF_FRAMES];
b_mode_info best_bmodes[4];
int best_skip2 = 0;
@@ -3747,25 +3544,22 @@
int64_t mask_filter = 0;
int64_t filter_cache[SWITCHABLE_FILTER_CONTEXTS];
int internal_active_edge =
- vp10_active_edge_sb(cpi, mi_row, mi_col) && vp10_internal_image_edge(cpi);
+ vp10_active_edge_sb(cpi, mi_row, mi_col) && vp10_internal_image_edge(cpi);
memset(x->zcoeff_blk[TX_4X4], 0, 4);
vp10_zero(best_mbmode);
- for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i)
- filter_cache[i] = INT64_MAX;
+ for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i) filter_cache[i] = INT64_MAX;
for (i = 0; i < 4; i++) {
int j;
- for (j = 0; j < MAX_REF_FRAMES; j++)
- seg_mvs[i][j].as_int = INVALID_MV;
+ for (j = 0; j < MAX_REF_FRAMES; j++) seg_mvs[i][j].as_int = INVALID_MV;
}
estimate_ref_frame_costs(cm, xd, segment_id, ref_costs_single, ref_costs_comp,
&comp_mode_p);
- for (i = 0; i < REFERENCE_MODES; ++i)
- best_pred_rd[i] = INT64_MAX;
+ for (i = 0; i < REFERENCE_MODES; ++i) best_pred_rd[i] = INT64_MAX;
for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++)
best_filter_rd[i] = INT64_MAX;
rate_uv_intra = INT_MAX;
@@ -3775,8 +3569,7 @@
for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ref_frame++) {
if (cpi->ref_frame_flags & flag_list[ref_frame]) {
setup_buffer_inter(cpi, x, ref_frame, bsize, mi_row, mi_col,
- frame_mv[NEARESTMV], frame_mv[NEARMV],
- yv12_mb);
+ frame_mv[NEARESTMV], frame_mv[NEARMV], yv12_mb);
} else {
ref_frame_skip_mask[0] |= (1 << ref_frame);
ref_frame_skip_mask[1] |= SECOND_REF_FRAME_MASK;
@@ -3806,8 +3599,7 @@
if (ref_index > 2 && sf->mode_skip_start < MAX_MODES) {
if (ref_index == 3) {
switch (best_mbmode.ref_frame[0]) {
- case INTRA_FRAME:
- break;
+ case INTRA_FRAME: break;
case LAST_FRAME:
ref_frame_skip_mask[0] |= (1 << GOLDEN_FRAME) | (1 << ALTREF_FRAME);
ref_frame_skip_mask[1] |= SECOND_REF_FRAME_MASK;
@@ -3820,9 +3612,7 @@
ref_frame_skip_mask[0] |= (1 << GOLDEN_FRAME) | (1 << LAST_FRAME);
break;
case NONE:
- case MAX_REF_FRAMES:
- assert(0 && "Invalid Reference frame");
- break;
+ case MAX_REF_FRAMES: assert(0 && "Invalid Reference frame"); break;
}
}
}
@@ -3840,14 +3630,11 @@
comp_pred = second_ref_frame > INTRA_FRAME;
if (comp_pred) {
- if (!cpi->allow_comp_inter_inter)
- continue;
- if (!(cpi->ref_frame_flags & flag_list[second_ref_frame]))
- continue;
+ if (!cpi->allow_comp_inter_inter) continue;
+ if (!(cpi->ref_frame_flags & flag_list[second_ref_frame])) continue;
// Do not allow compound prediction if the segment level reference frame
// feature is in use as in this case there can only be one reference.
- if (segfeature_active(seg, segment_id, SEG_LVL_REF_FRAME))
- continue;
+ if (segfeature_active(seg, segment_id, SEG_LVL_REF_FRAME)) continue;
if ((sf->mode_search_skip_flags & FLAG_SKIP_COMP_BESTINTRA) &&
best_mbmode.ref_frame[0] == INTRA_FRAME)
@@ -3874,9 +3661,9 @@
if (segfeature_active(seg, segment_id, SEG_LVL_REF_FRAME) &&
get_segdata(seg, segment_id, SEG_LVL_REF_FRAME) != (int)ref_frame) {
continue;
- // Disable this drop out case if the ref frame
- // segment level feature is enabled for this segment. This is to
- // prevent the possibility that we end up unable to pick any mode.
+ // Disable this drop out case if the ref frame
+ // segment level feature is enabled for this segment. This is to
+ // prevent the possibility that we end up unable to pick any mode.
} else if (!segfeature_active(seg, segment_id, SEG_LVL_REF_FRAME)) {
// Only consider ZEROMV/ALTREF_FRAME for alt ref frame,
// unless ARNR filtering is enabled in which case we want
@@ -3892,33 +3679,29 @@
mbmi->ref_frame[1] = second_ref_frame;
// Evaluate all sub-pel filters irrespective of whether we can use
// them for this frame.
- mbmi->interp_filter = cm->interp_filter == SWITCHABLE ? EIGHTTAP
- : cm->interp_filter;
+ mbmi->interp_filter =
+ cm->interp_filter == SWITCHABLE ? EIGHTTAP : cm->interp_filter;
x->skip = 0;
set_ref_ptrs(cm, xd, ref_frame, second_ref_frame);
// Select prediction reference frames.
for (i = 0; i < MAX_MB_PLANE; i++) {
xd->plane[i].pre[0] = yv12_mb[ref_frame][i];
- if (comp_pred)
- xd->plane[i].pre[1] = yv12_mb[second_ref_frame][i];
+ if (comp_pred) xd->plane[i].pre[1] = yv12_mb[second_ref_frame][i];
}
if (ref_frame == INTRA_FRAME) {
int rate;
- if (rd_pick_intra_sub_8x8_y_mode(cpi, x, &rate, &rate_y,
- &distortion_y, best_rd) >= best_rd)
+ if (rd_pick_intra_sub_8x8_y_mode(cpi, x, &rate, &rate_y, &distortion_y,
+ best_rd) >= best_rd)
continue;
rate2 += rate;
rate2 += intra_cost_penalty;
distortion2 += distortion_y;
if (rate_uv_intra == INT_MAX) {
- choose_intra_uv_mode(cpi, x, ctx, bsize, TX_4X4,
- &rate_uv_intra,
- &rate_uv_tokenonly,
- &dist_uv, &skip_uv,
- &mode_uv);
+ choose_intra_uv_mode(cpi, x, ctx, bsize, TX_4X4, &rate_uv_intra,
+ &rate_uv_tokenonly, &dist_uv, &skip_uv, &mode_uv);
}
rate2 += rate_uv_intra;
rate_uv = rate_uv_tokenonly;
@@ -3934,19 +3717,20 @@
int64_t tmp_best_distortion = INT_MAX, tmp_best_sse, uv_sse;
int tmp_best_skippable = 0;
int switchable_filter_index;
- int_mv *second_ref = comp_pred ?
- &x->mbmi_ext->ref_mvs[second_ref_frame][0] : NULL;
+ int_mv *second_ref =
+ comp_pred ? &x->mbmi_ext->ref_mvs[second_ref_frame][0] : NULL;
b_mode_info tmp_best_bmodes[16];
MB_MODE_INFO tmp_best_mbmode;
BEST_SEG_INFO bsi[SWITCHABLE_FILTERS];
int pred_exists = 0;
int uv_skippable;
- this_rd_thresh = (ref_frame == LAST_FRAME) ?
- rd_opt->threshes[segment_id][bsize][THR_LAST] :
- rd_opt->threshes[segment_id][bsize][THR_ALTR];
- this_rd_thresh = (ref_frame == GOLDEN_FRAME) ?
- rd_opt->threshes[segment_id][bsize][THR_GOLD] : this_rd_thresh;
+ this_rd_thresh = (ref_frame == LAST_FRAME)
+ ? rd_opt->threshes[segment_id][bsize][THR_LAST]
+ : rd_opt->threshes[segment_id][bsize][THR_ALTR];
+ this_rd_thresh = (ref_frame == GOLDEN_FRAME)
+ ? rd_opt->threshes[segment_id][bsize][THR_GOLD]
+ : this_rd_thresh;
for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i)
filter_cache[i] = INT64_MAX;
@@ -3958,8 +3742,9 @@
ctx->pred_interp_filter < SWITCHABLE) {
tmp_best_filter = ctx->pred_interp_filter;
} else if (sf->adaptive_pred_interp_filter == 2) {
- tmp_best_filter = ctx->pred_interp_filter < SWITCHABLE ?
- ctx->pred_interp_filter : 0;
+ tmp_best_filter = ctx->pred_interp_filter < SWITCHABLE
+ ? ctx->pred_interp_filter
+ : 0;
} else {
for (switchable_filter_index = 0;
switchable_filter_index < SWITCHABLE_FILTERS;
@@ -3968,24 +3753,19 @@
int64_t rs_rd;
MB_MODE_INFO_EXT *mbmi_ext = x->mbmi_ext;
mbmi->interp_filter = switchable_filter_index;
- tmp_rd = rd_pick_best_sub8x8_mode(cpi, x,
- &mbmi_ext->ref_mvs[ref_frame][0],
- second_ref, best_yrd, &rate,
- &rate_y, &distortion,
- &skippable, &total_sse,
- (int) this_rd_thresh, seg_mvs,
- bsi, switchable_filter_index,
- mi_row, mi_col);
+ tmp_rd = rd_pick_best_sub8x8_mode(
+ cpi, x, &mbmi_ext->ref_mvs[ref_frame][0], second_ref, best_yrd,
+ &rate, &rate_y, &distortion, &skippable, &total_sse,
+ (int)this_rd_thresh, seg_mvs, bsi, switchable_filter_index,
+ mi_row, mi_col);
- if (tmp_rd == INT64_MAX)
- continue;
+ if (tmp_rd == INT64_MAX) continue;
rs = vp10_get_switchable_rate(cpi, xd);
rs_rd = RDCOST(x->rdmult, x->rddiv, rs, 0);
filter_cache[switchable_filter_index] = tmp_rd;
filter_cache[SWITCHABLE_FILTERS] =
VPXMIN(filter_cache[SWITCHABLE_FILTERS], tmp_rd + rs_rd);
- if (cm->interp_filter == SWITCHABLE)
- tmp_rd += rs_rd;
+ if (cm->interp_filter == SWITCHABLE) tmp_rd += rs_rd;
mask_filter = VPXMAX(mask_filter, tmp_rd);
@@ -4009,8 +3789,7 @@
x->zcoeff_blk[TX_4X4][i] = !x->plane[0].eobs[i];
}
pred_exists = 1;
- if (switchable_filter_index == 0 &&
- sf->use_rd_breakout &&
+ if (switchable_filter_index == 0 && sf->use_rd_breakout &&
best_rd < INT64_MAX) {
if (tmp_best_rdu / 2 > best_rd) {
// skip searching the other filters if the first is
@@ -4025,22 +3804,19 @@
}
}
- if (tmp_best_rdu == INT64_MAX && pred_exists)
- continue;
+ if (tmp_best_rdu == INT64_MAX && pred_exists) continue;
- mbmi->interp_filter = (cm->interp_filter == SWITCHABLE ?
- tmp_best_filter : cm->interp_filter);
+ mbmi->interp_filter =
+ (cm->interp_filter == SWITCHABLE ? tmp_best_filter
+ : cm->interp_filter);
if (!pred_exists) {
// Handles the special case when a filter that is not in the
// switchable list (bilinear, 6-tap) is indicated at the frame level
- tmp_rd = rd_pick_best_sub8x8_mode(cpi, x,
- &x->mbmi_ext->ref_mvs[ref_frame][0],
- second_ref, best_yrd, &rate, &rate_y,
- &distortion, &skippable, &total_sse,
- (int) this_rd_thresh, seg_mvs, bsi, 0,
- mi_row, mi_col);
- if (tmp_rd == INT64_MAX)
- continue;
+ tmp_rd = rd_pick_best_sub8x8_mode(
+ cpi, x, &x->mbmi_ext->ref_mvs[ref_frame][0], second_ref, best_yrd,
+ &rate, &rate_y, &distortion, &skippable, &total_sse,
+ (int)this_rd_thresh, seg_mvs, bsi, 0, mi_row, mi_col);
+ if (tmp_rd == INT64_MAX) continue;
} else {
total_sse = tmp_best_sse;
rate = tmp_best_rate;
@@ -4048,8 +3824,7 @@
distortion = tmp_best_distortion;
skippable = tmp_best_skippable;
*mbmi = tmp_best_mbmode;
- for (i = 0; i < 4; i++)
- xd->mi[0]->bmi[i] = tmp_best_bmodes[i];
+ for (i = 0; i < 4; i++) xd->mi[0]->bmi[i] = tmp_best_bmodes[i];
}
rate2 += rate;
@@ -4064,15 +3839,14 @@
compmode_cost = vp10_cost_bit(comp_mode_p, comp_pred);
- tmp_best_rdu = best_rd -
- VPXMIN(RDCOST(x->rdmult, x->rddiv, rate2, distortion2),
- RDCOST(x->rdmult, x->rddiv, 0, total_sse));
+ tmp_best_rdu =
+ best_rd - VPXMIN(RDCOST(x->rdmult, x->rddiv, rate2, distortion2),
+ RDCOST(x->rdmult, x->rddiv, 0, total_sse));
if (tmp_best_rdu > 0) {
// If even the 'Y' rd value of split is higher than best so far
// then dont bother looking at UV
- vp10_build_inter_predictors_sbuv(&x->e_mbd, mi_row, mi_col,
- BLOCK_8X8);
+ vp10_build_inter_predictors_sbuv(&x->e_mbd, mi_row, mi_col, BLOCK_8X8);
memset(x->skip_txfm, SKIP_TXFM_NONE, sizeof(x->skip_txfm));
if (!super_block_uvrd(cpi, x, &rate_uv, &distortion_uv, &uv_skippable,
&uv_sse, BLOCK_8X8, tmp_best_rdu))
@@ -4085,8 +3859,7 @@
}
}
- if (cm->reference_mode == REFERENCE_MODE_SELECT)
- rate2 += compmode_cost;
+ if (cm->reference_mode == REFERENCE_MODE_SELECT) rate2 += compmode_cost;
// Estimate the reference frame signaling cost and add it
// to the rolling cost variable.
@@ -4148,17 +3921,15 @@
rd_cost->dist = distortion2;
rd_cost->rdcost = this_rd;
best_rd = this_rd;
- best_yrd = best_rd -
- RDCOST(x->rdmult, x->rddiv, rate_uv, distortion_uv);
+ best_yrd =
+ best_rd - RDCOST(x->rdmult, x->rddiv, rate_uv, distortion_uv);
best_mbmode = *mbmi;
best_skip2 = this_skip2;
- if (!x->select_tx_size)
- swap_block_ptr(x, ctx, 1, 0, 0, max_plane);
+ if (!x->select_tx_size) swap_block_ptr(x, ctx, 1, 0, 0, max_plane);
memcpy(ctx->zcoeff_blk, x->zcoeff_blk[TX_4X4],
sizeof(ctx->zcoeff_blk[0]) * ctx->num_4x4_blk);
- for (i = 0; i < 4; i++)
- best_bmodes[i] = xd->mi[0]->bmi[i];
+ for (i = 0; i < 4; i++) best_bmodes[i] = xd->mi[0]->bmi[i];
// TODO(debargha): enhance this test with a better distortion prediction
// based on qp, activity mask and history
@@ -4176,8 +3947,7 @@
const int var_adjust = (x->source_variance < 16);
scale -= var_adjust;
}
- if (ref_frame > INTRA_FRAME &&
- distortion2 * scale < qstep * qstep) {
+ if (ref_frame > INTRA_FRAME && distortion2 * scale < qstep * qstep) {
early_term = 1;
}
}
@@ -4211,8 +3981,9 @@
/* keep record of best filter type */
if (!mode_excluded && !disable_skip && ref_frame != INTRA_FRAME &&
cm->interp_filter != BILINEAR) {
- int64_t ref = filter_cache[cm->interp_filter == SWITCHABLE ?
- SWITCHABLE_FILTERS : cm->interp_filter];
+ int64_t ref =
+ filter_cache[cm->interp_filter == SWITCHABLE ? SWITCHABLE_FILTERS
+ : cm->interp_filter];
int64_t adj_rd;
for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++) {
if (ref == INT64_MAX)
@@ -4231,11 +4002,9 @@
}
}
- if (early_term)
- break;
+ if (early_term) break;
- if (x->skip && !comp_pred)
- break;
+ if (x->skip && !comp_pred) break;
}
if (best_rd >= best_rd_so_far) {
@@ -4249,11 +4018,8 @@
// Do Intra UV best rd mode selection if best mode choice above was intra.
if (best_mbmode.ref_frame[0] == INTRA_FRAME) {
*mbmi = best_mbmode;
- rd_pick_intra_sbuv_mode(cpi, x, ctx, &rate_uv_intra,
- &rate_uv_tokenonly,
- &dist_uv,
- &skip_uv,
- BLOCK_8X8, TX_4X4);
+ rd_pick_intra_sbuv_mode(cpi, x, ctx, &rate_uv_intra, &rate_uv_tokenonly,
+ &dist_uv, &skip_uv, BLOCK_8X8, TX_4X4);
}
}
@@ -4269,14 +4035,13 @@
!is_inter_block(&best_mbmode));
vp10_update_rd_thresh_fact(tile_data->thresh_freq_fact,
- sf->adaptive_rd_thresh, bsize, best_ref_index);
+ sf->adaptive_rd_thresh, bsize, best_ref_index);
// macroblock modes
*mbmi = best_mbmode;
x->skip |= best_skip2;
if (!is_inter_block(&best_mbmode)) {
- for (i = 0; i < 4; i++)
- xd->mi[0]->bmi[i].as_mode = best_bmodes[i].as_mode;
+ for (i = 0; i < 4; i++) xd->mi[0]->bmi[i].as_mode = best_bmodes[i].as_mode;
} else {
for (i = 0; i < 4; ++i)
memcpy(&xd->mi[0]->bmi[i], &best_bmodes[i], sizeof(b_mode_info));
@@ -4305,6 +4070,6 @@
vp10_zero(best_filter_diff);
}
- store_coding_context(x, ctx, best_ref_index,
- best_pred_diff, best_filter_diff, 0);
+ store_coding_context(x, ctx, best_ref_index, best_pred_diff, best_filter_diff,
+ 0);
}
diff --git a/vp10/encoder/rdopt.h b/vp10/encoder/rdopt.h
index 6526d2b..b887b8a 100644
--- a/vp10/encoder/rdopt.h
+++ b/vp10/encoder/rdopt.h
@@ -26,33 +26,28 @@
struct RD_COST;
void vp10_rd_pick_intra_mode_sb(struct VP10_COMP *cpi, struct macroblock *x,
- struct RD_COST *rd_cost, BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx, int64_t best_rd);
+ struct RD_COST *rd_cost, BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx, int64_t best_rd);
unsigned int vp10_get_sby_perpixel_variance(VP10_COMP *cpi,
- const struct buf_2d *ref,
- BLOCK_SIZE bs);
+ const struct buf_2d *ref,
+ BLOCK_SIZE bs);
#if CONFIG_VPX_HIGHBITDEPTH
unsigned int vp10_high_get_sby_perpixel_variance(VP10_COMP *cpi,
- const struct buf_2d *ref,
- BLOCK_SIZE bs, int bd);
+ const struct buf_2d *ref,
+ BLOCK_SIZE bs, int bd);
#endif
void vp10_rd_pick_inter_mode_sb(struct VP10_COMP *cpi,
- struct TileDataEnc *tile_data,
- struct macroblock *x,
- int mi_row, int mi_col,
- struct RD_COST *rd_cost,
- BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far);
+ struct TileDataEnc *tile_data,
+ struct macroblock *x, int mi_row, int mi_col,
+ struct RD_COST *rd_cost, BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx, int64_t best_rd_so_far);
-void vp10_rd_pick_inter_mode_sb_seg_skip(struct VP10_COMP *cpi,
- struct TileDataEnc *tile_data,
- struct macroblock *x,
- struct RD_COST *rd_cost,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far);
+void vp10_rd_pick_inter_mode_sb_seg_skip(
+ struct VP10_COMP *cpi, struct TileDataEnc *tile_data, struct macroblock *x,
+ struct RD_COST *rd_cost, BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far);
int vp10_internal_image_edge(struct VP10_COMP *cpi);
int vp10_active_h_edge(struct VP10_COMP *cpi, int mi_row, int mi_step);
@@ -60,12 +55,11 @@
int vp10_active_edge_sb(struct VP10_COMP *cpi, int mi_row, int mi_col);
void vp10_rd_pick_inter_mode_sub8x8(struct VP10_COMP *cpi,
- struct TileDataEnc *tile_data,
- struct macroblock *x,
- int mi_row, int mi_col,
- struct RD_COST *rd_cost,
- BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far);
+ struct TileDataEnc *tile_data,
+ struct macroblock *x, int mi_row,
+ int mi_col, struct RD_COST *rd_cost,
+ BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far);
#ifdef __cplusplus
} // extern "C"
diff --git a/vp10/encoder/resize.c b/vp10/encoder/resize.c
index 862ab02..b75ceb4 100644
--- a/vp10/encoder/resize.c
+++ b/vp10/encoder/resize.c
@@ -22,198 +22,118 @@
#include "vp10/common/common.h"
#include "vp10/encoder/resize.h"
-#define FILTER_BITS 7
+#define FILTER_BITS 7
-#define INTERP_TAPS 8
-#define SUBPEL_BITS 5
-#define SUBPEL_MASK ((1 << SUBPEL_BITS) - 1)
-#define INTERP_PRECISION_BITS 32
+#define INTERP_TAPS 8
+#define SUBPEL_BITS 5
+#define SUBPEL_MASK ((1 << SUBPEL_BITS) - 1)
+#define INTERP_PRECISION_BITS 32
typedef int16_t interp_kernel[INTERP_TAPS];
// Filters for interpolation (0.5-band) - note this also filters integer pels.
static const interp_kernel filteredinterp_filters500[(1 << SUBPEL_BITS)] = {
- {-3, 0, 35, 64, 35, 0, -3, 0},
- {-3, -1, 34, 64, 36, 1, -3, 0},
- {-3, -1, 32, 64, 38, 1, -3, 0},
- {-2, -2, 31, 63, 39, 2, -3, 0},
- {-2, -2, 29, 63, 41, 2, -3, 0},
- {-2, -2, 28, 63, 42, 3, -4, 0},
- {-2, -3, 27, 63, 43, 4, -4, 0},
- {-2, -3, 25, 62, 45, 5, -4, 0},
- {-2, -3, 24, 62, 46, 5, -4, 0},
- {-2, -3, 23, 61, 47, 6, -4, 0},
- {-2, -3, 21, 60, 49, 7, -4, 0},
- {-1, -4, 20, 60, 50, 8, -4, -1},
- {-1, -4, 19, 59, 51, 9, -4, -1},
- {-1, -4, 17, 58, 52, 10, -4, 0},
- {-1, -4, 16, 57, 53, 12, -4, -1},
- {-1, -4, 15, 56, 54, 13, -4, -1},
- {-1, -4, 14, 55, 55, 14, -4, -1},
- {-1, -4, 13, 54, 56, 15, -4, -1},
- {-1, -4, 12, 53, 57, 16, -4, -1},
- {0, -4, 10, 52, 58, 17, -4, -1},
- {-1, -4, 9, 51, 59, 19, -4, -1},
- {-1, -4, 8, 50, 60, 20, -4, -1},
- {0, -4, 7, 49, 60, 21, -3, -2},
- {0, -4, 6, 47, 61, 23, -3, -2},
- {0, -4, 5, 46, 62, 24, -3, -2},
- {0, -4, 5, 45, 62, 25, -3, -2},
- {0, -4, 4, 43, 63, 27, -3, -2},
- {0, -4, 3, 42, 63, 28, -2, -2},
- {0, -3, 2, 41, 63, 29, -2, -2},
- {0, -3, 2, 39, 63, 31, -2, -2},
- {0, -3, 1, 38, 64, 32, -1, -3},
- {0, -3, 1, 36, 64, 34, -1, -3}
+ { -3, 0, 35, 64, 35, 0, -3, 0 }, { -3, -1, 34, 64, 36, 1, -3, 0 },
+ { -3, -1, 32, 64, 38, 1, -3, 0 }, { -2, -2, 31, 63, 39, 2, -3, 0 },
+ { -2, -2, 29, 63, 41, 2, -3, 0 }, { -2, -2, 28, 63, 42, 3, -4, 0 },
+ { -2, -3, 27, 63, 43, 4, -4, 0 }, { -2, -3, 25, 62, 45, 5, -4, 0 },
+ { -2, -3, 24, 62, 46, 5, -4, 0 }, { -2, -3, 23, 61, 47, 6, -4, 0 },
+ { -2, -3, 21, 60, 49, 7, -4, 0 }, { -1, -4, 20, 60, 50, 8, -4, -1 },
+ { -1, -4, 19, 59, 51, 9, -4, -1 }, { -1, -4, 17, 58, 52, 10, -4, 0 },
+ { -1, -4, 16, 57, 53, 12, -4, -1 }, { -1, -4, 15, 56, 54, 13, -4, -1 },
+ { -1, -4, 14, 55, 55, 14, -4, -1 }, { -1, -4, 13, 54, 56, 15, -4, -1 },
+ { -1, -4, 12, 53, 57, 16, -4, -1 }, { 0, -4, 10, 52, 58, 17, -4, -1 },
+ { -1, -4, 9, 51, 59, 19, -4, -1 }, { -1, -4, 8, 50, 60, 20, -4, -1 },
+ { 0, -4, 7, 49, 60, 21, -3, -2 }, { 0, -4, 6, 47, 61, 23, -3, -2 },
+ { 0, -4, 5, 46, 62, 24, -3, -2 }, { 0, -4, 5, 45, 62, 25, -3, -2 },
+ { 0, -4, 4, 43, 63, 27, -3, -2 }, { 0, -4, 3, 42, 63, 28, -2, -2 },
+ { 0, -3, 2, 41, 63, 29, -2, -2 }, { 0, -3, 2, 39, 63, 31, -2, -2 },
+ { 0, -3, 1, 38, 64, 32, -1, -3 }, { 0, -3, 1, 36, 64, 34, -1, -3 }
};
// Filters for interpolation (0.625-band) - note this also filters integer pels.
static const interp_kernel filteredinterp_filters625[(1 << SUBPEL_BITS)] = {
- {-1, -8, 33, 80, 33, -8, -1, 0},
- {-1, -8, 30, 80, 35, -8, -1, 1},
- {-1, -8, 28, 80, 37, -7, -2, 1},
- {0, -8, 26, 79, 39, -7, -2, 1},
- {0, -8, 24, 79, 41, -7, -2, 1},
- {0, -8, 22, 78, 43, -6, -2, 1},
- {0, -8, 20, 78, 45, -5, -3, 1},
- {0, -8, 18, 77, 48, -5, -3, 1},
- {0, -8, 16, 76, 50, -4, -3, 1},
- {0, -8, 15, 75, 52, -3, -4, 1},
- {0, -7, 13, 74, 54, -3, -4, 1},
- {0, -7, 11, 73, 56, -2, -4, 1},
- {0, -7, 10, 71, 58, -1, -4, 1},
- {1, -7, 8, 70, 60, 0, -5, 1},
- {1, -6, 6, 68, 62, 1, -5, 1},
- {1, -6, 5, 67, 63, 2, -5, 1},
- {1, -6, 4, 65, 65, 4, -6, 1},
- {1, -5, 2, 63, 67, 5, -6, 1},
- {1, -5, 1, 62, 68, 6, -6, 1},
- {1, -5, 0, 60, 70, 8, -7, 1},
- {1, -4, -1, 58, 71, 10, -7, 0},
- {1, -4, -2, 56, 73, 11, -7, 0},
- {1, -4, -3, 54, 74, 13, -7, 0},
- {1, -4, -3, 52, 75, 15, -8, 0},
- {1, -3, -4, 50, 76, 16, -8, 0},
- {1, -3, -5, 48, 77, 18, -8, 0},
- {1, -3, -5, 45, 78, 20, -8, 0},
- {1, -2, -6, 43, 78, 22, -8, 0},
- {1, -2, -7, 41, 79, 24, -8, 0},
- {1, -2, -7, 39, 79, 26, -8, 0},
- {1, -2, -7, 37, 80, 28, -8, -1},
- {1, -1, -8, 35, 80, 30, -8, -1},
+ { -1, -8, 33, 80, 33, -8, -1, 0 }, { -1, -8, 30, 80, 35, -8, -1, 1 },
+ { -1, -8, 28, 80, 37, -7, -2, 1 }, { 0, -8, 26, 79, 39, -7, -2, 1 },
+ { 0, -8, 24, 79, 41, -7, -2, 1 }, { 0, -8, 22, 78, 43, -6, -2, 1 },
+ { 0, -8, 20, 78, 45, -5, -3, 1 }, { 0, -8, 18, 77, 48, -5, -3, 1 },
+ { 0, -8, 16, 76, 50, -4, -3, 1 }, { 0, -8, 15, 75, 52, -3, -4, 1 },
+ { 0, -7, 13, 74, 54, -3, -4, 1 }, { 0, -7, 11, 73, 56, -2, -4, 1 },
+ { 0, -7, 10, 71, 58, -1, -4, 1 }, { 1, -7, 8, 70, 60, 0, -5, 1 },
+ { 1, -6, 6, 68, 62, 1, -5, 1 }, { 1, -6, 5, 67, 63, 2, -5, 1 },
+ { 1, -6, 4, 65, 65, 4, -6, 1 }, { 1, -5, 2, 63, 67, 5, -6, 1 },
+ { 1, -5, 1, 62, 68, 6, -6, 1 }, { 1, -5, 0, 60, 70, 8, -7, 1 },
+ { 1, -4, -1, 58, 71, 10, -7, 0 }, { 1, -4, -2, 56, 73, 11, -7, 0 },
+ { 1, -4, -3, 54, 74, 13, -7, 0 }, { 1, -4, -3, 52, 75, 15, -8, 0 },
+ { 1, -3, -4, 50, 76, 16, -8, 0 }, { 1, -3, -5, 48, 77, 18, -8, 0 },
+ { 1, -3, -5, 45, 78, 20, -8, 0 }, { 1, -2, -6, 43, 78, 22, -8, 0 },
+ { 1, -2, -7, 41, 79, 24, -8, 0 }, { 1, -2, -7, 39, 79, 26, -8, 0 },
+ { 1, -2, -7, 37, 80, 28, -8, -1 }, { 1, -1, -8, 35, 80, 30, -8, -1 },
};
// Filters for interpolation (0.75-band) - note this also filters integer pels.
static const interp_kernel filteredinterp_filters750[(1 << SUBPEL_BITS)] = {
- {2, -11, 25, 96, 25, -11, 2, 0},
- {2, -11, 22, 96, 28, -11, 2, 0},
- {2, -10, 19, 95, 31, -11, 2, 0},
- {2, -10, 17, 95, 34, -12, 2, 0},
- {2, -9, 14, 94, 37, -12, 2, 0},
- {2, -8, 12, 93, 40, -12, 1, 0},
- {2, -8, 9, 92, 43, -12, 1, 1},
- {2, -7, 7, 91, 46, -12, 1, 0},
- {2, -7, 5, 90, 49, -12, 1, 0},
- {2, -6, 3, 88, 52, -12, 0, 1},
- {2, -5, 1, 86, 55, -12, 0, 1},
- {2, -5, -1, 84, 58, -11, 0, 1},
- {2, -4, -2, 82, 61, -11, -1, 1},
- {2, -4, -4, 80, 64, -10, -1, 1},
- {1, -3, -5, 77, 67, -9, -1, 1},
- {1, -3, -6, 75, 70, -8, -2, 1},
- {1, -2, -7, 72, 72, -7, -2, 1},
- {1, -2, -8, 70, 75, -6, -3, 1},
- {1, -1, -9, 67, 77, -5, -3, 1},
- {1, -1, -10, 64, 80, -4, -4, 2},
- {1, -1, -11, 61, 82, -2, -4, 2},
- {1, 0, -11, 58, 84, -1, -5, 2},
- {1, 0, -12, 55, 86, 1, -5, 2},
- {1, 0, -12, 52, 88, 3, -6, 2},
- {0, 1, -12, 49, 90, 5, -7, 2},
- {0, 1, -12, 46, 91, 7, -7, 2},
- {1, 1, -12, 43, 92, 9, -8, 2},
- {0, 1, -12, 40, 93, 12, -8, 2},
- {0, 2, -12, 37, 94, 14, -9, 2},
- {0, 2, -12, 34, 95, 17, -10, 2},
- {0, 2, -11, 31, 95, 19, -10, 2},
- {0, 2, -11, 28, 96, 22, -11, 2}
+ { 2, -11, 25, 96, 25, -11, 2, 0 }, { 2, -11, 22, 96, 28, -11, 2, 0 },
+ { 2, -10, 19, 95, 31, -11, 2, 0 }, { 2, -10, 17, 95, 34, -12, 2, 0 },
+ { 2, -9, 14, 94, 37, -12, 2, 0 }, { 2, -8, 12, 93, 40, -12, 1, 0 },
+ { 2, -8, 9, 92, 43, -12, 1, 1 }, { 2, -7, 7, 91, 46, -12, 1, 0 },
+ { 2, -7, 5, 90, 49, -12, 1, 0 }, { 2, -6, 3, 88, 52, -12, 0, 1 },
+ { 2, -5, 1, 86, 55, -12, 0, 1 }, { 2, -5, -1, 84, 58, -11, 0, 1 },
+ { 2, -4, -2, 82, 61, -11, -1, 1 }, { 2, -4, -4, 80, 64, -10, -1, 1 },
+ { 1, -3, -5, 77, 67, -9, -1, 1 }, { 1, -3, -6, 75, 70, -8, -2, 1 },
+ { 1, -2, -7, 72, 72, -7, -2, 1 }, { 1, -2, -8, 70, 75, -6, -3, 1 },
+ { 1, -1, -9, 67, 77, -5, -3, 1 }, { 1, -1, -10, 64, 80, -4, -4, 2 },
+ { 1, -1, -11, 61, 82, -2, -4, 2 }, { 1, 0, -11, 58, 84, -1, -5, 2 },
+ { 1, 0, -12, 55, 86, 1, -5, 2 }, { 1, 0, -12, 52, 88, 3, -6, 2 },
+ { 0, 1, -12, 49, 90, 5, -7, 2 }, { 0, 1, -12, 46, 91, 7, -7, 2 },
+ { 1, 1, -12, 43, 92, 9, -8, 2 }, { 0, 1, -12, 40, 93, 12, -8, 2 },
+ { 0, 2, -12, 37, 94, 14, -9, 2 }, { 0, 2, -12, 34, 95, 17, -10, 2 },
+ { 0, 2, -11, 31, 95, 19, -10, 2 }, { 0, 2, -11, 28, 96, 22, -11, 2 }
};
// Filters for interpolation (0.875-band) - note this also filters integer pels.
static const interp_kernel filteredinterp_filters875[(1 << SUBPEL_BITS)] = {
- {3, -8, 13, 112, 13, -8, 3, 0},
- {3, -7, 10, 112, 17, -9, 3, -1},
- {2, -6, 7, 111, 21, -9, 3, -1},
- {2, -5, 4, 111, 24, -10, 3, -1},
- {2, -4, 1, 110, 28, -11, 3, -1},
- {1, -3, -1, 108, 32, -12, 4, -1},
- {1, -2, -3, 106, 36, -13, 4, -1},
- {1, -1, -6, 105, 40, -14, 4, -1},
- {1, -1, -7, 102, 44, -14, 4, -1},
- {1, 0, -9, 100, 48, -15, 4, -1},
- {1, 1, -11, 97, 53, -16, 4, -1},
- {0, 1, -12, 95, 57, -16, 4, -1},
- {0, 2, -13, 91, 61, -16, 4, -1},
- {0, 2, -14, 88, 65, -16, 4, -1},
- {0, 3, -15, 84, 69, -17, 4, 0},
- {0, 3, -16, 81, 73, -16, 3, 0},
- {0, 3, -16, 77, 77, -16, 3, 0},
- {0, 3, -16, 73, 81, -16, 3, 0},
- {0, 4, -17, 69, 84, -15, 3, 0},
- {-1, 4, -16, 65, 88, -14, 2, 0},
- {-1, 4, -16, 61, 91, -13, 2, 0},
- {-1, 4, -16, 57, 95, -12, 1, 0},
- {-1, 4, -16, 53, 97, -11, 1, 1},
- {-1, 4, -15, 48, 100, -9, 0, 1},
- {-1, 4, -14, 44, 102, -7, -1, 1},
- {-1, 4, -14, 40, 105, -6, -1, 1},
- {-1, 4, -13, 36, 106, -3, -2, 1},
- {-1, 4, -12, 32, 108, -1, -3, 1},
- {-1, 3, -11, 28, 110, 1, -4, 2},
- {-1, 3, -10, 24, 111, 4, -5, 2},
- {-1, 3, -9, 21, 111, 7, -6, 2},
- {-1, 3, -9, 17, 112, 10, -7, 3}
+ { 3, -8, 13, 112, 13, -8, 3, 0 }, { 3, -7, 10, 112, 17, -9, 3, -1 },
+ { 2, -6, 7, 111, 21, -9, 3, -1 }, { 2, -5, 4, 111, 24, -10, 3, -1 },
+ { 2, -4, 1, 110, 28, -11, 3, -1 }, { 1, -3, -1, 108, 32, -12, 4, -1 },
+ { 1, -2, -3, 106, 36, -13, 4, -1 }, { 1, -1, -6, 105, 40, -14, 4, -1 },
+ { 1, -1, -7, 102, 44, -14, 4, -1 }, { 1, 0, -9, 100, 48, -15, 4, -1 },
+ { 1, 1, -11, 97, 53, -16, 4, -1 }, { 0, 1, -12, 95, 57, -16, 4, -1 },
+ { 0, 2, -13, 91, 61, -16, 4, -1 }, { 0, 2, -14, 88, 65, -16, 4, -1 },
+ { 0, 3, -15, 84, 69, -17, 4, 0 }, { 0, 3, -16, 81, 73, -16, 3, 0 },
+ { 0, 3, -16, 77, 77, -16, 3, 0 }, { 0, 3, -16, 73, 81, -16, 3, 0 },
+ { 0, 4, -17, 69, 84, -15, 3, 0 }, { -1, 4, -16, 65, 88, -14, 2, 0 },
+ { -1, 4, -16, 61, 91, -13, 2, 0 }, { -1, 4, -16, 57, 95, -12, 1, 0 },
+ { -1, 4, -16, 53, 97, -11, 1, 1 }, { -1, 4, -15, 48, 100, -9, 0, 1 },
+ { -1, 4, -14, 44, 102, -7, -1, 1 }, { -1, 4, -14, 40, 105, -6, -1, 1 },
+ { -1, 4, -13, 36, 106, -3, -2, 1 }, { -1, 4, -12, 32, 108, -1, -3, 1 },
+ { -1, 3, -11, 28, 110, 1, -4, 2 }, { -1, 3, -10, 24, 111, 4, -5, 2 },
+ { -1, 3, -9, 21, 111, 7, -6, 2 }, { -1, 3, -9, 17, 112, 10, -7, 3 }
};
// Filters for interpolation (full-band) - no filtering for integer pixels
static const interp_kernel filteredinterp_filters1000[(1 << SUBPEL_BITS)] = {
- {0, 0, 0, 128, 0, 0, 0, 0},
- {0, 1, -3, 128, 3, -1, 0, 0},
- {-1, 2, -6, 127, 7, -2, 1, 0},
- {-1, 3, -9, 126, 12, -4, 1, 0},
- {-1, 4, -12, 125, 16, -5, 1, 0},
- {-1, 4, -14, 123, 20, -6, 2, 0},
- {-1, 5, -15, 120, 25, -8, 2, 0},
- {-1, 5, -17, 118, 30, -9, 3, -1},
- {-1, 6, -18, 114, 35, -10, 3, -1},
- {-1, 6, -19, 111, 41, -12, 3, -1},
- {-1, 6, -20, 107, 46, -13, 4, -1},
- {-1, 6, -21, 103, 52, -14, 4, -1},
- {-1, 6, -21, 99, 57, -16, 5, -1},
- {-1, 6, -21, 94, 63, -17, 5, -1},
- {-1, 6, -20, 89, 68, -18, 5, -1},
- {-1, 6, -20, 84, 73, -19, 6, -1},
- {-1, 6, -20, 79, 79, -20, 6, -1},
- {-1, 6, -19, 73, 84, -20, 6, -1},
- {-1, 5, -18, 68, 89, -20, 6, -1},
- {-1, 5, -17, 63, 94, -21, 6, -1},
- {-1, 5, -16, 57, 99, -21, 6, -1},
- {-1, 4, -14, 52, 103, -21, 6, -1},
- {-1, 4, -13, 46, 107, -20, 6, -1},
- {-1, 3, -12, 41, 111, -19, 6, -1},
- {-1, 3, -10, 35, 114, -18, 6, -1},
- {-1, 3, -9, 30, 118, -17, 5, -1},
- {0, 2, -8, 25, 120, -15, 5, -1},
- {0, 2, -6, 20, 123, -14, 4, -1},
- {0, 1, -5, 16, 125, -12, 4, -1},
- {0, 1, -4, 12, 126, -9, 3, -1},
- {0, 1, -2, 7, 127, -6, 2, -1},
- {0, 0, -1, 3, 128, -3, 1, 0}
+ { 0, 0, 0, 128, 0, 0, 0, 0 }, { 0, 1, -3, 128, 3, -1, 0, 0 },
+ { -1, 2, -6, 127, 7, -2, 1, 0 }, { -1, 3, -9, 126, 12, -4, 1, 0 },
+ { -1, 4, -12, 125, 16, -5, 1, 0 }, { -1, 4, -14, 123, 20, -6, 2, 0 },
+ { -1, 5, -15, 120, 25, -8, 2, 0 }, { -1, 5, -17, 118, 30, -9, 3, -1 },
+ { -1, 6, -18, 114, 35, -10, 3, -1 }, { -1, 6, -19, 111, 41, -12, 3, -1 },
+ { -1, 6, -20, 107, 46, -13, 4, -1 }, { -1, 6, -21, 103, 52, -14, 4, -1 },
+ { -1, 6, -21, 99, 57, -16, 5, -1 }, { -1, 6, -21, 94, 63, -17, 5, -1 },
+ { -1, 6, -20, 89, 68, -18, 5, -1 }, { -1, 6, -20, 84, 73, -19, 6, -1 },
+ { -1, 6, -20, 79, 79, -20, 6, -1 }, { -1, 6, -19, 73, 84, -20, 6, -1 },
+ { -1, 5, -18, 68, 89, -20, 6, -1 }, { -1, 5, -17, 63, 94, -21, 6, -1 },
+ { -1, 5, -16, 57, 99, -21, 6, -1 }, { -1, 4, -14, 52, 103, -21, 6, -1 },
+ { -1, 4, -13, 46, 107, -20, 6, -1 }, { -1, 3, -12, 41, 111, -19, 6, -1 },
+ { -1, 3, -10, 35, 114, -18, 6, -1 }, { -1, 3, -9, 30, 118, -17, 5, -1 },
+ { 0, 2, -8, 25, 120, -15, 5, -1 }, { 0, 2, -6, 20, 123, -14, 4, -1 },
+ { 0, 1, -5, 16, 125, -12, 4, -1 }, { 0, 1, -4, 12, 126, -9, 3, -1 },
+ { 0, 1, -2, 7, 127, -6, 2, -1 }, { 0, 0, -1, 3, 128, -3, 1, 0 }
};
// Filters for factor of 2 downsampling.
-static const int16_t vp10_down2_symeven_half_filter[] = {56, 12, -3, -1};
-static const int16_t vp10_down2_symodd_half_filter[] = {64, 35, 0, -3};
+static const int16_t vp10_down2_symeven_half_filter[] = { 56, 12, -3, -1 };
+static const int16_t vp10_down2_symodd_half_filter[] = { 64, 35, 0, -3 };
static const interp_kernel *choose_interp_filter(int inlength, int outlength) {
int outlength16 = outlength * 16;
@@ -231,11 +151,14 @@
static void interpolate(const uint8_t *const input, int inlength,
uint8_t *output, int outlength) {
- const int64_t delta = (((uint64_t)inlength << 32) + outlength / 2) /
- outlength;
- const int64_t offset = inlength > outlength ?
- (((int64_t)(inlength - outlength) << 31) + outlength / 2) / outlength :
- -(((int64_t)(outlength - inlength) << 31) + outlength / 2) / outlength;
+ const int64_t delta =
+ (((uint64_t)inlength << 32) + outlength / 2) / outlength;
+ const int64_t offset =
+ inlength > outlength
+ ? (((int64_t)(inlength - outlength) << 31) + outlength / 2) /
+ outlength
+ : -(((int64_t)(outlength - inlength) << 31) + outlength / 2) /
+ outlength;
uint8_t *optr = output;
int x, x1, x2, sum, k, int_pel, sub_pel;
int64_t y;
@@ -252,8 +175,8 @@
x1 = x;
x = outlength - 1;
y = delta * x + offset;
- while ((y >> INTERP_PRECISION_BITS) +
- (int64_t)(INTERP_TAPS / 2) >= inlength) {
+ while ((y >> INTERP_PRECISION_BITS) + (int64_t)(INTERP_TAPS / 2) >=
+ inlength) {
x--;
y -= delta;
}
@@ -267,8 +190,8 @@
sum = 0;
for (k = 0; k < INTERP_TAPS; ++k) {
const int pk = int_pel - INTERP_TAPS / 2 + 1 + k;
- sum += filter[k] * input[(pk < 0 ? 0 :
- (pk >= inlength ? inlength - 1 : pk))];
+ sum += filter[k] *
+ input[(pk < 0 ? 0 : (pk >= inlength ? inlength - 1 : pk))];
}
*optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
}
@@ -281,9 +204,9 @@
filter = interp_filters[sub_pel];
sum = 0;
for (k = 0; k < INTERP_TAPS; ++k)
- sum += filter[k] * input[(int_pel - INTERP_TAPS / 2 + 1 + k < 0 ?
- 0 :
- int_pel - INTERP_TAPS / 2 + 1 + k)];
+ sum += filter[k] * input[(int_pel - INTERP_TAPS / 2 + 1 + k < 0
+ ? 0
+ : int_pel - INTERP_TAPS / 2 + 1 + k)];
*optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
}
// Middle part.
@@ -305,9 +228,9 @@
filter = interp_filters[sub_pel];
sum = 0;
for (k = 0; k < INTERP_TAPS; ++k)
- sum += filter[k] * input[(int_pel - INTERP_TAPS / 2 + 1 + k >=
- inlength ? inlength - 1 :
- int_pel - INTERP_TAPS / 2 + 1 + k)];
+ sum += filter[k] * input[(int_pel - INTERP_TAPS / 2 + 1 + k >= inlength
+ ? inlength - 1
+ : int_pel - INTERP_TAPS / 2 + 1 + k)];
*optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
}
}
@@ -331,7 +254,7 @@
for (j = 0; j < filter_len_half; ++j) {
sum += (input[(i - j < 0 ? 0 : i - j)] +
input[(i + 1 + j >= length ? length - 1 : i + 1 + j)]) *
- filter[j];
+ filter[j];
}
sum >>= FILTER_BITS;
*optr++ = clip_pixel(sum);
@@ -361,7 +284,7 @@
for (j = 0; j < filter_len_half; ++j) {
sum += (input[i - j] +
input[(i + 1 + j >= length ? length - 1 : i + 1 + j)]) *
- filter[j];
+ filter[j];
}
sum >>= FILTER_BITS;
*optr++ = clip_pixel(sum);
@@ -387,7 +310,7 @@
for (j = 1; j < filter_len_half; ++j) {
sum += (input[(i - j < 0 ? 0 : i - j)] +
input[(i + j >= length ? length - 1 : i + j)]) *
- filter[j];
+ filter[j];
}
sum >>= FILTER_BITS;
*optr++ = clip_pixel(sum);
@@ -416,7 +339,7 @@
int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
for (j = 1; j < filter_len_half; ++j) {
sum += (input[i - j] + input[(i + j >= length ? length - 1 : i + j)]) *
- filter[j];
+ filter[j];
}
sum >>= FILTER_BITS;
*optr++ = clip_pixel(sum);
@@ -426,8 +349,7 @@
static int get_down2_length(int length, int steps) {
int s;
- for (s = 0; s < steps; ++s)
- length = (length + 1) >> 1;
+ for (s = 0; s < steps; ++s) length = (length + 1) >> 1;
return length;
}
@@ -441,11 +363,8 @@
return steps;
}
-static void resize_multistep(const uint8_t *const input,
- int length,
- uint8_t *output,
- int olength,
- uint8_t *buf) {
+static void resize_multistep(const uint8_t *const input, int length,
+ uint8_t *output, int olength, uint8_t *buf) {
int steps;
if (length == olength) {
memcpy(output, input, sizeof(output[0]) * length);
@@ -482,8 +401,7 @@
if (filteredlength != olength) {
interpolate(out, filteredlength, output, olength);
}
- if (tmpbuf)
- free(tmpbuf);
+ if (tmpbuf) free(tmpbuf);
} else {
interpolate(input, length, output, olength);
}
@@ -507,26 +425,21 @@
}
}
-void vp10_resize_plane(const uint8_t *const input,
- int height,
- int width,
- int in_stride,
- uint8_t *output,
- int height2,
- int width2,
- int out_stride) {
+void vp10_resize_plane(const uint8_t *const input, int height, int width,
+ int in_stride, uint8_t *output, int height2, int width2,
+ int out_stride) {
int i;
uint8_t *intbuf = (uint8_t *)malloc(sizeof(uint8_t) * width2 * height);
- uint8_t *tmpbuf = (uint8_t *)malloc(sizeof(uint8_t) *
- (width < height ? height : width));
+ uint8_t *tmpbuf =
+ (uint8_t *)malloc(sizeof(uint8_t) * (width < height ? height : width));
uint8_t *arrbuf = (uint8_t *)malloc(sizeof(uint8_t) * (height + height2));
assert(width > 0);
assert(height > 0);
assert(width2 > 0);
assert(height2 > 0);
for (i = 0; i < height; ++i)
- resize_multistep(input + in_stride * i, width,
- intbuf + width2 * i, width2, tmpbuf);
+ resize_multistep(input + in_stride * i, width, intbuf + width2 * i, width2,
+ tmpbuf);
for (i = 0; i < width2; ++i) {
fill_col_to_arr(intbuf + i, width2, height, arrbuf);
resize_multistep(arrbuf, height, arrbuf + height, height2, tmpbuf);
@@ -542,9 +455,12 @@
uint16_t *output, int outlength, int bd) {
const int64_t delta =
(((uint64_t)inlength << 32) + outlength / 2) / outlength;
- const int64_t offset = inlength > outlength ?
- (((int64_t)(inlength - outlength) << 31) + outlength / 2) / outlength :
- -(((int64_t)(outlength - inlength) << 31) + outlength / 2) / outlength;
+ const int64_t offset =
+ inlength > outlength
+ ? (((int64_t)(inlength - outlength) << 31) + outlength / 2) /
+ outlength
+ : -(((int64_t)(outlength - inlength) << 31) + outlength / 2) /
+ outlength;
uint16_t *optr = output;
int x, x1, x2, sum, k, int_pel, sub_pel;
int64_t y;
@@ -561,8 +477,8 @@
x1 = x;
x = outlength - 1;
y = delta * x + offset;
- while ((y >> INTERP_PRECISION_BITS) +
- (int64_t)(INTERP_TAPS / 2) >= inlength) {
+ while ((y >> INTERP_PRECISION_BITS) + (int64_t)(INTERP_TAPS / 2) >=
+ inlength) {
x--;
y -= delta;
}
@@ -577,7 +493,7 @@
for (k = 0; k < INTERP_TAPS; ++k) {
const int pk = int_pel - INTERP_TAPS / 2 + 1 + k;
sum += filter[k] *
- input[(pk < 0 ? 0 : (pk >= inlength ? inlength - 1 : pk))];
+ input[(pk < 0 ? 0 : (pk >= inlength ? inlength - 1 : pk))];
}
*optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
}
@@ -590,9 +506,9 @@
filter = interp_filters[sub_pel];
sum = 0;
for (k = 0; k < INTERP_TAPS; ++k)
- sum += filter[k] *
- input[(int_pel - INTERP_TAPS / 2 + 1 + k < 0 ?
- 0 : int_pel - INTERP_TAPS / 2 + 1 + k)];
+ sum += filter[k] * input[(int_pel - INTERP_TAPS / 2 + 1 + k < 0
+ ? 0
+ : int_pel - INTERP_TAPS / 2 + 1 + k)];
*optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
}
// Middle part.
@@ -614,9 +530,9 @@
filter = interp_filters[sub_pel];
sum = 0;
for (k = 0; k < INTERP_TAPS; ++k)
- sum += filter[k] * input[(int_pel - INTERP_TAPS / 2 + 1 + k >=
- inlength ? inlength - 1 :
- int_pel - INTERP_TAPS / 2 + 1 + k)];
+ sum += filter[k] * input[(int_pel - INTERP_TAPS / 2 + 1 + k >= inlength
+ ? inlength - 1
+ : int_pel - INTERP_TAPS / 2 + 1 + k)];
*optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
}
}
@@ -640,7 +556,7 @@
for (j = 0; j < filter_len_half; ++j) {
sum += (input[(i - j < 0 ? 0 : i - j)] +
input[(i + 1 + j >= length ? length - 1 : i + 1 + j)]) *
- filter[j];
+ filter[j];
}
sum >>= FILTER_BITS;
*optr++ = clip_pixel_highbd(sum, bd);
@@ -670,7 +586,7 @@
for (j = 0; j < filter_len_half; ++j) {
sum += (input[i - j] +
input[(i + 1 + j >= length ? length - 1 : i + 1 + j)]) *
- filter[j];
+ filter[j];
}
sum >>= FILTER_BITS;
*optr++ = clip_pixel_highbd(sum, bd);
@@ -679,7 +595,7 @@
}
static void highbd_down2_symodd(const uint16_t *const input, int length,
- uint16_t *output, int bd) {
+ uint16_t *output, int bd) {
// Actual filter len = 2 * filter_len_half - 1.
static const int16_t *filter = vp10_down2_symodd_half_filter;
const int filter_len_half = sizeof(vp10_down2_symodd_half_filter) / 2;
@@ -696,7 +612,7 @@
for (j = 1; j < filter_len_half; ++j) {
sum += (input[(i - j < 0 ? 0 : i - j)] +
input[(i + j >= length ? length - 1 : i + j)]) *
- filter[j];
+ filter[j];
}
sum >>= FILTER_BITS;
*optr++ = clip_pixel_highbd(sum, bd);
@@ -725,7 +641,7 @@
int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
for (j = 1; j < filter_len_half; ++j) {
sum += (input[i - j] + input[(i + j >= length ? length - 1 : i + j)]) *
- filter[j];
+ filter[j];
}
sum >>= FILTER_BITS;
*optr++ = clip_pixel_highbd(sum, bd);
@@ -733,12 +649,9 @@
}
}
-static void highbd_resize_multistep(const uint16_t *const input,
- int length,
- uint16_t *output,
- int olength,
- uint16_t *buf,
- int bd) {
+static void highbd_resize_multistep(const uint16_t *const input, int length,
+ uint16_t *output, int olength,
+ uint16_t *buf, int bd) {
int steps;
if (length == olength) {
memcpy(output, input, sizeof(output[0]) * length);
@@ -775,8 +688,7 @@
if (filteredlength != olength) {
highbd_interpolate(out, filteredlength, output, olength, bd);
}
- if (tmpbuf)
- free(tmpbuf);
+ if (tmpbuf) free(tmpbuf);
} else {
highbd_interpolate(input, length, output, olength, bd);
}
@@ -802,19 +714,13 @@
}
}
-void vp10_highbd_resize_plane(const uint8_t *const input,
- int height,
- int width,
- int in_stride,
- uint8_t *output,
- int height2,
- int width2,
- int out_stride,
- int bd) {
+void vp10_highbd_resize_plane(const uint8_t *const input, int height, int width,
+ int in_stride, uint8_t *output, int height2,
+ int width2, int out_stride, int bd) {
int i;
uint16_t *intbuf = (uint16_t *)malloc(sizeof(uint16_t) * width2 * height);
- uint16_t *tmpbuf = (uint16_t *)malloc(sizeof(uint16_t) *
- (width < height ? height : width));
+ uint16_t *tmpbuf =
+ (uint16_t *)malloc(sizeof(uint16_t) * (width < height ? height : width));
uint16_t *arrbuf = (uint16_t *)malloc(sizeof(uint16_t) * (height + height2));
for (i = 0; i < height; ++i) {
highbd_resize_multistep(CONVERT_TO_SHORTPTR(input + in_stride * i), width,
@@ -833,96 +739,82 @@
}
#endif // CONFIG_VPX_HIGHBITDEPTH
-void vp10_resize_frame420(const uint8_t *const y,
- int y_stride,
- const uint8_t *const u, const uint8_t *const v,
- int uv_stride,
- int height, int width,
- uint8_t *oy, int oy_stride,
- uint8_t *ou, uint8_t *ov, int ouv_stride,
- int oheight, int owidth) {
- vp10_resize_plane(y, height, width, y_stride,
- oy, oheight, owidth, oy_stride);
- vp10_resize_plane(u, height / 2, width / 2, uv_stride,
- ou, oheight / 2, owidth / 2, ouv_stride);
- vp10_resize_plane(v, height / 2, width / 2, uv_stride,
- ov, oheight / 2, owidth / 2, ouv_stride);
+void vp10_resize_frame420(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width, uint8_t *oy,
+ int oy_stride, uint8_t *ou, uint8_t *ov,
+ int ouv_stride, int oheight, int owidth) {
+ vp10_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride);
+ vp10_resize_plane(u, height / 2, width / 2, uv_stride, ou, oheight / 2,
+ owidth / 2, ouv_stride);
+ vp10_resize_plane(v, height / 2, width / 2, uv_stride, ov, oheight / 2,
+ owidth / 2, ouv_stride);
}
void vp10_resize_frame422(const uint8_t *const y, int y_stride,
- const uint8_t *const u, const uint8_t *const v,
- int uv_stride,
- int height, int width,
- uint8_t *oy, int oy_stride,
- uint8_t *ou, uint8_t *ov, int ouv_stride,
- int oheight, int owidth) {
- vp10_resize_plane(y, height, width, y_stride,
- oy, oheight, owidth, oy_stride);
- vp10_resize_plane(u, height, width / 2, uv_stride,
- ou, oheight, owidth / 2, ouv_stride);
- vp10_resize_plane(v, height, width / 2, uv_stride,
- ov, oheight, owidth / 2, ouv_stride);
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width, uint8_t *oy,
+ int oy_stride, uint8_t *ou, uint8_t *ov,
+ int ouv_stride, int oheight, int owidth) {
+ vp10_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride);
+ vp10_resize_plane(u, height, width / 2, uv_stride, ou, oheight, owidth / 2,
+ ouv_stride);
+ vp10_resize_plane(v, height, width / 2, uv_stride, ov, oheight, owidth / 2,
+ ouv_stride);
}
void vp10_resize_frame444(const uint8_t *const y, int y_stride,
- const uint8_t *const u, const uint8_t *const v,
- int uv_stride,
- int height, int width,
- uint8_t *oy, int oy_stride,
- uint8_t *ou, uint8_t *ov, int ouv_stride,
- int oheight, int owidth) {
- vp10_resize_plane(y, height, width, y_stride,
- oy, oheight, owidth, oy_stride);
- vp10_resize_plane(u, height, width, uv_stride,
- ou, oheight, owidth, ouv_stride);
- vp10_resize_plane(v, height, width, uv_stride,
- ov, oheight, owidth, ouv_stride);
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width, uint8_t *oy,
+ int oy_stride, uint8_t *ou, uint8_t *ov,
+ int ouv_stride, int oheight, int owidth) {
+ vp10_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride);
+ vp10_resize_plane(u, height, width, uv_stride, ou, oheight, owidth,
+ ouv_stride);
+ vp10_resize_plane(v, height, width, uv_stride, ov, oheight, owidth,
+ ouv_stride);
}
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_highbd_resize_frame420(const uint8_t *const y,
- int y_stride,
- const uint8_t *const u, const uint8_t *const v,
- int uv_stride,
- int height, int width,
- uint8_t *oy, int oy_stride,
- uint8_t *ou, uint8_t *ov, int ouv_stride,
- int oheight, int owidth, int bd) {
- vp10_highbd_resize_plane(y, height, width, y_stride,
- oy, oheight, owidth, oy_stride, bd);
- vp10_highbd_resize_plane(u, height / 2, width / 2, uv_stride,
- ou, oheight / 2, owidth / 2, ouv_stride, bd);
- vp10_highbd_resize_plane(v, height / 2, width / 2, uv_stride,
- ov, oheight / 2, owidth / 2, ouv_stride, bd);
+void vp10_highbd_resize_frame420(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width,
+ uint8_t *oy, int oy_stride, uint8_t *ou,
+ uint8_t *ov, int ouv_stride, int oheight,
+ int owidth, int bd) {
+ vp10_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth,
+ oy_stride, bd);
+ vp10_highbd_resize_plane(u, height / 2, width / 2, uv_stride, ou, oheight / 2,
+ owidth / 2, ouv_stride, bd);
+ vp10_highbd_resize_plane(v, height / 2, width / 2, uv_stride, ov, oheight / 2,
+ owidth / 2, ouv_stride, bd);
}
void vp10_highbd_resize_frame422(const uint8_t *const y, int y_stride,
- const uint8_t *const u, const uint8_t *const v,
- int uv_stride,
- int height, int width,
- uint8_t *oy, int oy_stride,
- uint8_t *ou, uint8_t *ov, int ouv_stride,
- int oheight, int owidth, int bd) {
- vp10_highbd_resize_plane(y, height, width, y_stride,
- oy, oheight, owidth, oy_stride, bd);
- vp10_highbd_resize_plane(u, height, width / 2, uv_stride,
- ou, oheight, owidth / 2, ouv_stride, bd);
- vp10_highbd_resize_plane(v, height, width / 2, uv_stride,
- ov, oheight, owidth / 2, ouv_stride, bd);
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width,
+ uint8_t *oy, int oy_stride, uint8_t *ou,
+ uint8_t *ov, int ouv_stride, int oheight,
+ int owidth, int bd) {
+ vp10_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth,
+ oy_stride, bd);
+ vp10_highbd_resize_plane(u, height, width / 2, uv_stride, ou, oheight,
+ owidth / 2, ouv_stride, bd);
+ vp10_highbd_resize_plane(v, height, width / 2, uv_stride, ov, oheight,
+ owidth / 2, ouv_stride, bd);
}
void vp10_highbd_resize_frame444(const uint8_t *const y, int y_stride,
- const uint8_t *const u, const uint8_t *const v,
- int uv_stride,
- int height, int width,
- uint8_t *oy, int oy_stride,
- uint8_t *ou, uint8_t *ov, int ouv_stride,
- int oheight, int owidth, int bd) {
- vp10_highbd_resize_plane(y, height, width, y_stride,
- oy, oheight, owidth, oy_stride, bd);
- vp10_highbd_resize_plane(u, height, width, uv_stride,
- ou, oheight, owidth, ouv_stride, bd);
- vp10_highbd_resize_plane(v, height, width, uv_stride,
- ov, oheight, owidth, ouv_stride, bd);
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width,
+ uint8_t *oy, int oy_stride, uint8_t *ou,
+ uint8_t *ov, int ouv_stride, int oheight,
+ int owidth, int bd) {
+ vp10_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth,
+ oy_stride, bd);
+ vp10_highbd_resize_plane(u, height, width, uv_stride, ou, oheight, owidth,
+ ouv_stride, bd);
+ vp10_highbd_resize_plane(v, height, width, uv_stride, ov, oheight, owidth,
+ ouv_stride, bd);
}
#endif // CONFIG_VPX_HIGHBITDEPTH
diff --git a/vp10/encoder/resize.h b/vp10/encoder/resize.h
index a85735b..9dc34ef 100644
--- a/vp10/encoder/resize.h
+++ b/vp10/encoder/resize.h
@@ -18,116 +18,51 @@
extern "C" {
#endif
-void vp10_resize_plane(const uint8_t *const input,
- int height,
- int width,
- int in_stride,
- uint8_t *output,
- int height2,
- int width2,
- int out_stride);
-void vp10_resize_frame420(const uint8_t *const y,
- int y_stride,
- const uint8_t *const u,
- const uint8_t *const v,
- int uv_stride,
- int height,
- int width,
- uint8_t *oy,
- int oy_stride,
- uint8_t *ou,
- uint8_t *ov,
- int ouv_stride,
- int oheight,
- int owidth);
-void vp10_resize_frame422(const uint8_t *const y,
- int y_stride,
- const uint8_t *const u,
- const uint8_t *const v,
- int uv_stride,
- int height,
- int width,
- uint8_t *oy,
- int oy_stride,
- uint8_t *ou,
- uint8_t *ov,
- int ouv_stride,
- int oheight,
- int owidth);
-void vp10_resize_frame444(const uint8_t *const y,
- int y_stride,
- const uint8_t *const u,
- const uint8_t *const v,
- int uv_stride,
- int height,
- int width,
- uint8_t *oy,
- int oy_stride,
- uint8_t *ou,
- uint8_t *ov,
- int ouv_stride,
- int oheight,
- int owidth);
+void vp10_resize_plane(const uint8_t *const input, int height, int width,
+ int in_stride, uint8_t *output, int height2, int width2,
+ int out_stride);
+void vp10_resize_frame420(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width, uint8_t *oy,
+ int oy_stride, uint8_t *ou, uint8_t *ov,
+ int ouv_stride, int oheight, int owidth);
+void vp10_resize_frame422(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width, uint8_t *oy,
+ int oy_stride, uint8_t *ou, uint8_t *ov,
+ int ouv_stride, int oheight, int owidth);
+void vp10_resize_frame444(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width, uint8_t *oy,
+ int oy_stride, uint8_t *ou, uint8_t *ov,
+ int ouv_stride, int oheight, int owidth);
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_highbd_resize_plane(const uint8_t *const input,
- int height,
- int width,
- int in_stride,
- uint8_t *output,
- int height2,
- int width2,
- int out_stride,
- int bd);
-void vp10_highbd_resize_frame420(const uint8_t *const y,
- int y_stride,
- const uint8_t *const u,
- const uint8_t *const v,
- int uv_stride,
- int height,
- int width,
- uint8_t *oy,
- int oy_stride,
- uint8_t *ou,
- uint8_t *ov,
- int ouv_stride,
- int oheight,
- int owidth,
- int bd);
-void vp10_highbd_resize_frame422(const uint8_t *const y,
- int y_stride,
- const uint8_t *const u,
- const uint8_t *const v,
- int uv_stride,
- int height,
- int width,
- uint8_t *oy,
- int oy_stride,
- uint8_t *ou,
- uint8_t *ov,
- int ouv_stride,
- int oheight,
- int owidth,
- int bd);
-void vp10_highbd_resize_frame444(const uint8_t *const y,
- int y_stride,
- const uint8_t *const u,
- const uint8_t *const v,
- int uv_stride,
- int height,
- int width,
- uint8_t *oy,
- int oy_stride,
- uint8_t *ou,
- uint8_t *ov,
- int ouv_stride,
- int oheight,
- int owidth,
- int bd);
-#endif // CONFIG_VPX_HIGHBITDEPTH
+void vp10_highbd_resize_plane(const uint8_t *const input, int height, int width,
+ int in_stride, uint8_t *output, int height2,
+ int width2, int out_stride, int bd);
+void vp10_highbd_resize_frame420(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width,
+ uint8_t *oy, int oy_stride, uint8_t *ou,
+ uint8_t *ov, int ouv_stride, int oheight,
+ int owidth, int bd);
+void vp10_highbd_resize_frame422(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width,
+ uint8_t *oy, int oy_stride, uint8_t *ou,
+ uint8_t *ov, int ouv_stride, int oheight,
+ int owidth, int bd);
+void vp10_highbd_resize_frame444(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride, int height, int width,
+ uint8_t *oy, int oy_stride, uint8_t *ou,
+ uint8_t *ov, int ouv_stride, int oheight,
+ int owidth, int bd);
+#endif // CONFIG_VPX_HIGHBITDEPTH
#ifdef __cplusplus
} // extern "C"
#endif
-#endif // VP10_ENCODER_RESIZE_H_
+#endif // VP10_ENCODER_RESIZE_H_
diff --git a/vp10/encoder/segmentation.c b/vp10/encoder/segmentation.c
index 677910f..1b29366 100644
--- a/vp10/encoder/segmentation.c
+++ b/vp10/encoder/segmentation.c
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#include <limits.h>
#include "vpx_mem/vpx_mem.h"
@@ -32,31 +31,31 @@
seg->update_data = 0;
}
-void vp10_set_segment_data(struct segmentation *seg,
- signed char *feature_data,
- unsigned char abs_delta) {
+void vp10_set_segment_data(struct segmentation *seg, signed char *feature_data,
+ unsigned char abs_delta) {
seg->abs_delta = abs_delta;
memcpy(seg->feature_data, feature_data, sizeof(seg->feature_data));
}
void vp10_disable_segfeature(struct segmentation *seg, int segment_id,
- SEG_LVL_FEATURES feature_id) {
+ SEG_LVL_FEATURES feature_id) {
seg->feature_mask[segment_id] &= ~(1 << feature_id);
}
void vp10_clear_segdata(struct segmentation *seg, int segment_id,
- SEG_LVL_FEATURES feature_id) {
+ SEG_LVL_FEATURES feature_id) {
seg->feature_data[segment_id][feature_id] = 0;
}
// Based on set of segment counts calculate a probability tree
static void calc_segtree_probs(unsigned *segcounts,
- vpx_prob *segment_tree_probs, const vpx_prob *cur_tree_probs) {
+ vpx_prob *segment_tree_probs,
+ const vpx_prob *cur_tree_probs) {
// Work out probabilities of each segment
- const unsigned cc[4] = {
- segcounts[0] + segcounts[1], segcounts[2] + segcounts[3],
- segcounts[4] + segcounts[5], segcounts[6] + segcounts[7]
- };
+ const unsigned cc[4] = { segcounts[0] + segcounts[1],
+ segcounts[2] + segcounts[3],
+ segcounts[4] + segcounts[5],
+ segcounts[6] + segcounts[7] };
const unsigned ccc[2] = { cc[0] + cc[1], cc[2] + cc[3] };
#if CONFIG_MISC_FIXES
int i;
@@ -72,13 +71,13 @@
#if CONFIG_MISC_FIXES
for (i = 0; i < 7; i++) {
- const unsigned *ct = i == 0 ? ccc : i < 3 ? cc + (i & 2)
- : segcounts + (i - 3) * 2;
- vp10_prob_diff_update_savings_search(ct,
- cur_tree_probs[i], &segment_tree_probs[i], DIFF_UPDATE_PROB);
+ const unsigned *ct =
+ i == 0 ? ccc : i < 3 ? cc + (i & 2) : segcounts + (i - 3) * 2;
+ vp10_prob_diff_update_savings_search(
+ ct, cur_tree_probs[i], &segment_tree_probs[i], DIFF_UPDATE_PROB);
}
#else
- (void) cur_tree_probs;
+ (void)cur_tree_probs;
#endif
}
@@ -92,13 +91,11 @@
const int c4567 = c45 + c67;
// Cost the top node of the tree
- int cost = c0123 * vp10_cost_zero(probs[0]) +
- c4567 * vp10_cost_one(probs[0]);
+ int cost = c0123 * vp10_cost_zero(probs[0]) + c4567 * vp10_cost_one(probs[0]);
// Cost subsequent levels
if (c0123 > 0) {
- cost += c01 * vp10_cost_zero(probs[1]) +
- c23 * vp10_cost_one(probs[1]);
+ cost += c01 * vp10_cost_zero(probs[1]) + c23 * vp10_cost_one(probs[1]);
if (c01 > 0)
cost += segcounts[0] * vp10_cost_zero(probs[3]) +
@@ -109,8 +106,7 @@
}
if (c4567 > 0) {
- cost += c45 * vp10_cost_zero(probs[2]) +
- c67 * vp10_cost_one(probs[2]);
+ cost += c45 * vp10_cost_zero(probs[2]) + c67 * vp10_cost_one(probs[2]);
if (c45 > 0)
cost += segcounts[4] * vp10_cost_zero(probs[5]) +
@@ -127,12 +123,11 @@
const TileInfo *tile, MODE_INFO **mi,
unsigned *no_pred_segcounts,
unsigned (*temporal_predictor_count)[2],
- unsigned *t_unpred_seg_counts,
- int bw, int bh, int mi_row, int mi_col) {
+ unsigned *t_unpred_seg_counts, int bw, int bh,
+ int mi_row, int mi_col) {
int segment_id;
- if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
- return;
+ if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols) return;
xd->mi = mi;
segment_id = xd->mi[0]->mbmi.segment_id;
@@ -146,8 +141,8 @@
if (cm->frame_type != KEY_FRAME) {
const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
// Test to see if the segment id matches the predicted value.
- const int pred_segment_id = get_segment_id(cm, cm->last_frame_seg_map,
- bsize, mi_row, mi_col);
+ const int pred_segment_id =
+ get_segment_id(cm, cm->last_frame_seg_map, bsize, mi_row, mi_col);
const int pred_flag = pred_segment_id == segment_id;
const int pred_context = vp10_get_pred_context_seg_id(xd);
@@ -157,8 +152,7 @@
temporal_predictor_count[pred_context][pred_flag]++;
// Update the "unpredicted" segment count
- if (!pred_flag)
- t_unpred_seg_counts[segment_id]++;
+ if (!pred_flag) t_unpred_seg_counts[segment_id]++;
}
}
@@ -166,15 +160,13 @@
const TileInfo *tile, MODE_INFO **mi,
unsigned *no_pred_segcounts,
unsigned (*temporal_predictor_count)[2],
- unsigned *t_unpred_seg_counts,
- int mi_row, int mi_col,
+ unsigned *t_unpred_seg_counts, int mi_row, int mi_col,
BLOCK_SIZE bsize) {
const int mis = cm->mi_stride;
int bw, bh;
const int bs = num_8x8_blocks_wide_lookup[bsize], hbs = bs / 2;
- if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
- return;
+ if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols) return;
bw = num_8x8_blocks_wide_lookup[mi[0]->mbmi.sb_type];
bh = num_8x8_blocks_high_lookup[mi[0]->mbmi.sb_type];
@@ -191,9 +183,9 @@
} else if (bw < bs && bh == bs) {
count_segs(cm, xd, tile, mi, no_pred_segcounts, temporal_predictor_count,
t_unpred_seg_counts, hbs, bs, mi_row, mi_col);
- count_segs(cm, xd, tile, mi + hbs,
- no_pred_segcounts, temporal_predictor_count, t_unpred_seg_counts,
- hbs, bs, mi_row, mi_col + hbs);
+ count_segs(cm, xd, tile, mi + hbs, no_pred_segcounts,
+ temporal_predictor_count, t_unpred_seg_counts, hbs, bs, mi_row,
+ mi_col + hbs);
} else {
const BLOCK_SIZE subsize = subsize_lookup[PARTITION_SPLIT][bsize];
int n;
@@ -204,9 +196,8 @@
const int mi_dc = hbs * (n & 1);
const int mi_dr = hbs * (n >> 1);
- count_segs_sb(cm, xd, tile, &mi[mi_dr * mis + mi_dc],
- no_pred_segcounts, temporal_predictor_count,
- t_unpred_seg_counts,
+ count_segs_sb(cm, xd, tile, &mi[mi_dr * mis + mi_dc], no_pred_segcounts,
+ temporal_predictor_count, t_unpred_seg_counts,
mi_row + mi_dr, mi_col + mi_dc, subsize);
}
}
@@ -226,7 +217,7 @@
int i, tile_col, mi_row, mi_col;
#if CONFIG_MISC_FIXES
- unsigned (*temporal_predictor_count)[2] = cm->counts.seg.pred;
+ unsigned(*temporal_predictor_count)[2] = cm->counts.seg.pred;
unsigned *no_pred_segcounts = cm->counts.seg.tree_total;
unsigned *t_unpred_seg_counts = cm->counts.seg.tree_mispred;
#else
@@ -240,7 +231,7 @@
vpx_prob t_nopred_prob[PREDICTION_PROBS];
#if CONFIG_MISC_FIXES
- (void) xd;
+ (void)xd;
#else
// Set default state for the segment tree probabilities and the
// temporal coding probabilities
@@ -262,8 +253,8 @@
for (mi_col = tile.mi_col_start; mi_col < tile.mi_col_end;
mi_col += 8, mi += 8)
count_segs_sb(cm, xd, &tile, mi, no_pred_segcounts,
- temporal_predictor_count, t_unpred_seg_counts,
- mi_row, mi_col, BLOCK_64X64);
+ temporal_predictor_count, t_unpred_seg_counts, mi_row,
+ mi_col, BLOCK_64X64);
}
}
diff --git a/vp10/encoder/segmentation.h b/vp10/encoder/segmentation.h
index b8e6c06..c07f37f 100644
--- a/vp10/encoder/segmentation.h
+++ b/vp10/encoder/segmentation.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_SEGMENTATION_H_
#define VP10_ENCODER_SEGMENTATION_H_
@@ -22,12 +21,10 @@
void vp10_enable_segmentation(struct segmentation *seg);
void vp10_disable_segmentation(struct segmentation *seg);
-void vp10_disable_segfeature(struct segmentation *seg,
- int segment_id,
- SEG_LVL_FEATURES feature_id);
-void vp10_clear_segdata(struct segmentation *seg,
- int segment_id,
- SEG_LVL_FEATURES feature_id);
+void vp10_disable_segfeature(struct segmentation *seg, int segment_id,
+ SEG_LVL_FEATURES feature_id);
+void vp10_clear_segdata(struct segmentation *seg, int segment_id,
+ SEG_LVL_FEATURES feature_id);
// The values given for each segment can be either deltas (from the default
// value chosen for the frame) or absolute values.
@@ -40,7 +37,7 @@
// abs_delta = SEGMENT_DELTADATA (deltas) abs_delta = SEGMENT_ABSDATA (use
// the absolute values given).
void vp10_set_segment_data(struct segmentation *seg, signed char *feature_data,
- unsigned char abs_delta);
+ unsigned char abs_delta);
void vp10_choose_segmap_coding_method(VP10_COMMON *cm, MACROBLOCKD *xd);
diff --git a/vp10/encoder/skin_detection.c b/vp10/encoder/skin_detection.c
index 1714cd3..324ac59 100644
--- a/vp10/encoder/skin_detection.c
+++ b/vp10/encoder/skin_detection.c
@@ -16,9 +16,9 @@
#include "vp10/encoder/skin_detection.h"
// Fixed-point skin color model parameters.
-static const int skin_mean[2] = {7463, 9614}; // q6
-static const int skin_inv_cov[4] = {4107, 1663, 1663, 2157}; // q16
-static const int skin_threshold = 1570636; // q18
+static const int skin_mean[2] = { 7463, 9614 }; // q6
+static const int skin_inv_cov[4] = { 4107, 1663, 1663, 2157 }; // q16
+static const int skin_threshold = 1570636; // q18
// Thresholds on luminance.
static const int y_low = 20;
@@ -34,10 +34,9 @@
const int cb_diff_q2 = (cb_diff_q12 + (1 << 9)) >> 10;
const int cbcr_diff_q2 = (cbcr_diff_q12 + (1 << 9)) >> 10;
const int cr_diff_q2 = (cr_diff_q12 + (1 << 9)) >> 10;
- const int skin_diff = skin_inv_cov[0] * cb_diff_q2 +
- skin_inv_cov[1] * cbcr_diff_q2 +
- skin_inv_cov[2] * cbcr_diff_q2 +
- skin_inv_cov[3] * cr_diff_q2;
+ const int skin_diff =
+ skin_inv_cov[0] * cb_diff_q2 + skin_inv_cov[1] * cbcr_diff_q2 +
+ skin_inv_cov[2] * cbcr_diff_q2 + skin_inv_cov[3] * cr_diff_q2;
return skin_diff;
}
@@ -61,11 +60,11 @@
const int src_uvstride = cpi->Source->uv_stride;
YV12_BUFFER_CONFIG skinmap;
memset(&skinmap, 0, sizeof(YV12_BUFFER_CONFIG));
- if (vpx_alloc_frame_buffer(&skinmap, cm->width, cm->height,
- cm->subsampling_x, cm->subsampling_y,
- VPX_ENC_BORDER_IN_PIXELS, cm->byte_alignment)) {
- vpx_free_frame_buffer(&skinmap);
- return;
+ if (vpx_alloc_frame_buffer(&skinmap, cm->width, cm->height, cm->subsampling_x,
+ cm->subsampling_y, VPX_ENC_BORDER_IN_PIXELS,
+ cm->byte_alignment)) {
+ vpx_free_frame_buffer(&skinmap);
+ return;
}
memset(skinmap.buffer_alloc, 128, skinmap.frame_size);
y = skinmap.y_buffer;
diff --git a/vp10/encoder/speed_features.c b/vp10/encoder/speed_features.c
index b82dec0..d95dc65 100644
--- a/vp10/encoder/speed_features.c
+++ b/vp10/encoder/speed_features.c
@@ -17,21 +17,23 @@
#include "vpx_dsp/vpx_dsp_common.h"
// Mesh search patters for various speed settings
-static MESH_PATTERN best_quality_mesh_pattern[MAX_MESH_STEP] =
- {{64, 4}, {28, 2}, {15, 1}, {7, 1}};
+static MESH_PATTERN best_quality_mesh_pattern[MAX_MESH_STEP] = {
+ { 64, 4 }, { 28, 2 }, { 15, 1 }, { 7, 1 }
+};
#define MAX_MESH_SPEED 5 // Max speed setting for mesh motion method
-static MESH_PATTERN good_quality_mesh_patterns[MAX_MESH_SPEED + 1]
- [MAX_MESH_STEP] =
- {{{64, 8}, {28, 4}, {15, 1}, {7, 1}},
- {{64, 8}, {28, 4}, {15, 1}, {7, 1}},
- {{64, 8}, {14, 2}, {7, 1}, {7, 1}},
- {{64, 16}, {24, 8}, {12, 4}, {7, 1}},
- {{64, 16}, {24, 8}, {12, 4}, {7, 1}},
- {{64, 16}, {24, 8}, {12, 4}, {7, 1}},
+static MESH_PATTERN
+ good_quality_mesh_patterns[MAX_MESH_SPEED + 1][MAX_MESH_STEP] = {
+ { { 64, 8 }, { 28, 4 }, { 15, 1 }, { 7, 1 } },
+ { { 64, 8 }, { 28, 4 }, { 15, 1 }, { 7, 1 } },
+ { { 64, 8 }, { 14, 2 }, { 7, 1 }, { 7, 1 } },
+ { { 64, 16 }, { 24, 8 }, { 12, 4 }, { 7, 1 } },
+ { { 64, 16 }, { 24, 8 }, { 12, 4 }, { 7, 1 } },
+ { { 64, 16 }, { 24, 8 }, { 12, 4 }, { 7, 1 } },
};
-static unsigned char good_quality_max_mesh_pct[MAX_MESH_SPEED + 1] =
- {50, 25, 15, 5, 1, 1};
+static unsigned char good_quality_max_mesh_pct[MAX_MESH_SPEED + 1] = {
+ 50, 25, 15, 5, 1, 1
+};
// Intra only frames, golden frames (except alt ref overlays) and
// alt ref frames tend to be coded at a higher than ambient quality
@@ -68,8 +70,8 @@
if (speed >= 1) {
if (VPXMIN(cm->width, cm->height) >= 720) {
- sf->disable_split_mask = cm->show_frame ? DISABLE_ALL_SPLIT
- : DISABLE_ALL_INTER_SPLIT;
+ sf->disable_split_mask =
+ cm->show_frame ? DISABLE_ALL_SPLIT : DISABLE_ALL_INTER_SPLIT;
sf->partition_search_breakout_dist_thr = (1 << 23);
} else {
sf->disable_split_mask = DISABLE_COMPOUND_SPLIT;
@@ -79,8 +81,8 @@
if (speed >= 2) {
if (VPXMIN(cm->width, cm->height) >= 720) {
- sf->disable_split_mask = cm->show_frame ? DISABLE_ALL_SPLIT
- : DISABLE_ALL_INTER_SPLIT;
+ sf->disable_split_mask =
+ cm->show_frame ? DISABLE_ALL_SPLIT : DISABLE_ALL_INTER_SPLIT;
sf->adaptive_pred_interp_filter = 0;
sf->partition_search_breakout_dist_thr = (1 << 24);
sf->partition_search_breakout_rate_thr = 120;
@@ -141,7 +143,7 @@
sf->use_square_partition_only = !frame_is_intra_only(cm);
}
- sf->less_rectangular_check = 1;
+ sf->less_rectangular_check = 1;
sf->use_rd_breakout = 1;
sf->adaptive_motion_search = 1;
@@ -162,14 +164,14 @@
}
if (speed >= 2) {
- sf->tx_size_search_method = frame_is_boosted(cpi) ? USE_FULL_RD
- : USE_LARGESTALL;
+ sf->tx_size_search_method =
+ frame_is_boosted(cpi) ? USE_FULL_RD : USE_LARGESTALL;
- sf->mode_search_skip_flags = (cm->frame_type == KEY_FRAME) ? 0 :
- FLAG_SKIP_INTRA_DIRMISMATCH |
- FLAG_SKIP_INTRA_BESTINTER |
- FLAG_SKIP_COMP_BESTINTRA |
- FLAG_SKIP_INTRA_LOWVAR;
+ sf->mode_search_skip_flags =
+ (cm->frame_type == KEY_FRAME) ? 0 : FLAG_SKIP_INTRA_DIRMISMATCH |
+ FLAG_SKIP_INTRA_BESTINTER |
+ FLAG_SKIP_COMP_BESTINTRA |
+ FLAG_SKIP_INTRA_LOWVAR;
sf->disable_filter_search_var_thresh = 100;
sf->comp_inter_joint_search_thresh = BLOCK_SIZES;
sf->auto_min_max_partition_size = RELAXED_NEIGHBORING_MIN_MAX;
@@ -178,8 +180,8 @@
if (speed >= 3) {
sf->use_square_partition_only = !frame_is_intra_only(cm);
- sf->tx_size_search_method = frame_is_intra_only(cm) ? USE_FULL_RD
- : USE_LARGESTALL;
+ sf->tx_size_search_method =
+ frame_is_intra_only(cm) ? USE_FULL_RD : USE_LARGESTALL;
sf->mv.subpel_search_method = SUBPEL_TREE_PRUNED;
sf->adaptive_pred_interp_filter = 0;
sf->adaptive_mode_search = 1;
@@ -225,13 +227,14 @@
}
static void set_rt_speed_feature_framesize_dependent(VP10_COMP *cpi,
- SPEED_FEATURES *sf, int speed) {
+ SPEED_FEATURES *sf,
+ int speed) {
VP10_COMMON *const cm = &cpi->common;
if (speed >= 1) {
if (VPXMIN(cm->width, cm->height) >= 720) {
- sf->disable_split_mask = cm->show_frame ? DISABLE_ALL_SPLIT
- : DISABLE_ALL_INTER_SPLIT;
+ sf->disable_split_mask =
+ cm->show_frame ? DISABLE_ALL_SPLIT : DISABLE_ALL_INTER_SPLIT;
} else {
sf->disable_split_mask = DISABLE_COMPOUND_SPLIT;
}
@@ -239,8 +242,8 @@
if (speed >= 2) {
if (VPXMIN(cm->width, cm->height) >= 720) {
- sf->disable_split_mask = cm->show_frame ? DISABLE_ALL_SPLIT
- : DISABLE_ALL_INTER_SPLIT;
+ sf->disable_split_mask =
+ cm->show_frame ? DISABLE_ALL_SPLIT : DISABLE_ALL_INTER_SPLIT;
} else {
sf->disable_split_mask = LAST_AND_INTRA_SPLIT_ONLY;
}
@@ -255,13 +258,13 @@
}
if (speed >= 7) {
- sf->encode_breakout_thresh = (VPXMIN(cm->width, cm->height) >= 720) ?
- 800 : 300;
+ sf->encode_breakout_thresh =
+ (VPXMIN(cm->width, cm->height) >= 720) ? 800 : 300;
}
}
-static void set_rt_speed_feature(VP10_COMP *cpi, SPEED_FEATURES *sf,
- int speed, vpx_tune_content content) {
+static void set_rt_speed_feature(VP10_COMP *cpi, SPEED_FEATURES *sf, int speed,
+ vpx_tune_content content) {
VP10_COMMON *const cm = &cpi->common;
const int is_keyframe = cm->frame_type == KEY_FRAME;
const int frames_since_key = is_keyframe ? 0 : cpi->rc.frames_since_key;
@@ -274,8 +277,8 @@
if (speed >= 1) {
sf->use_square_partition_only = !frame_is_intra_only(cm);
sf->less_rectangular_check = 1;
- sf->tx_size_search_method = frame_is_intra_only(cm) ? USE_FULL_RD
- : USE_LARGESTALL;
+ sf->tx_size_search_method =
+ frame_is_intra_only(cm) ? USE_FULL_RD : USE_LARGESTALL;
sf->use_rd_breakout = 1;
@@ -289,11 +292,11 @@
}
if (speed >= 2) {
- sf->mode_search_skip_flags = (cm->frame_type == KEY_FRAME) ? 0 :
- FLAG_SKIP_INTRA_DIRMISMATCH |
- FLAG_SKIP_INTRA_BESTINTER |
- FLAG_SKIP_COMP_BESTINTRA |
- FLAG_SKIP_INTRA_LOWVAR;
+ sf->mode_search_skip_flags =
+ (cm->frame_type == KEY_FRAME) ? 0 : FLAG_SKIP_INTRA_DIRMISMATCH |
+ FLAG_SKIP_INTRA_BESTINTER |
+ FLAG_SKIP_COMP_BESTINTRA |
+ FLAG_SKIP_INTRA_LOWVAR;
sf->adaptive_pred_interp_filter = 2;
sf->disable_filter_search_var_thresh = 50;
sf->comp_inter_joint_search_thresh = BLOCK_SIZES;
@@ -326,8 +329,8 @@
sf->use_fast_coef_costing = 0;
sf->auto_min_max_partition_size = STRICT_NEIGHBORING_MIN_MAX;
sf->adjust_partitioning_from_last_frame =
- cm->last_frame_type != cm->frame_type || (0 ==
- (frames_since_key + 1) % sf->last_partitioning_redo_frequency);
+ cm->last_frame_type != cm->frame_type ||
+ (0 == (frames_since_key + 1) % sf->last_partitioning_redo_frequency);
sf->mv.subpel_force_stop = 1;
for (i = 0; i < TX_SIZES; i++) {
sf->intra_y_mode_mask[i] = INTRA_DC_H_V;
@@ -347,11 +350,12 @@
if (speed >= 5) {
sf->use_quant_fp = !is_keyframe;
- sf->auto_min_max_partition_size = is_keyframe ? RELAXED_NEIGHBORING_MIN_MAX
- : STRICT_NEIGHBORING_MIN_MAX;
+ sf->auto_min_max_partition_size =
+ is_keyframe ? RELAXED_NEIGHBORING_MIN_MAX : STRICT_NEIGHBORING_MIN_MAX;
sf->default_max_partition_size = BLOCK_32X32;
sf->default_min_partition_size = BLOCK_8X8;
- sf->force_frame_boost = is_keyframe ||
+ sf->force_frame_boost =
+ is_keyframe ||
(frames_since_key % (sf->last_partitioning_redo_frequency << 1) == 1);
sf->max_delta_qindex = is_keyframe ? 20 : 15;
sf->partition_search_type = REFERENCE_PARTITION;
@@ -494,8 +498,7 @@
sf->use_fast_coef_costing = 0;
sf->mode_skip_start = MAX_MODES; // Mode index at which mode skip mask set
sf->schedule_mode_search = 0;
- for (i = 0; i < BLOCK_SIZES; ++i)
- sf->inter_mode_mask[i] = INTER_ALL;
+ for (i = 0; i < BLOCK_SIZES; ++i) sf->inter_mode_mask[i] = INTER_ALL;
sf->max_intra_bsize = BLOCK_64X64;
sf->reuse_inter_pred_sby = 0;
// This setting only takes effect when partition_search_type is set
@@ -541,8 +544,7 @@
sf->exhaustive_searches_thresh = sf->exhaustive_searches_thresh << 1;
for (i = 0; i < MAX_MESH_STEP; ++i) {
- sf->mesh_patterns[i].range =
- good_quality_mesh_patterns[speed][i].range;
+ sf->mesh_patterns[i].range = good_quality_mesh_patterns[speed][i].range;
sf->mesh_patterns[i].interval =
good_quality_mesh_patterns[speed][i].interval;
}
@@ -550,8 +552,7 @@
// Slow quant, dct and trellis not worthwhile for first pass
// so make sure they are always turned off.
- if (oxcf->pass == 1)
- sf->optimize_coefficients = 0;
+ if (oxcf->pass == 1) sf->optimize_coefficients = 0;
// No recode for 1 pass.
if (oxcf->pass == 0) {
@@ -566,7 +567,8 @@
} else if (sf->mv.subpel_search_method == SUBPEL_TREE_PRUNED_MORE) {
cpi->find_fractional_mv_step = vp10_find_best_sub_pixel_tree_pruned_more;
} else if (sf->mv.subpel_search_method == SUBPEL_TREE_PRUNED_EVENMORE) {
- cpi->find_fractional_mv_step = vp10_find_best_sub_pixel_tree_pruned_evenmore;
+ cpi->find_fractional_mv_step =
+ vp10_find_best_sub_pixel_tree_pruned_evenmore;
}
x->optimize = sf->optimize_coefficients == 1 && oxcf->pass != 1;
diff --git a/vp10/encoder/speed_features.h b/vp10/encoder/speed_features.h
index 3b91999..72b5847 100644
--- a/vp10/encoder/speed_features.h
+++ b/vp10/encoder/speed_features.h
@@ -18,17 +18,14 @@
#endif
enum {
- INTRA_ALL = (1 << DC_PRED) |
- (1 << V_PRED) | (1 << H_PRED) |
- (1 << D45_PRED) | (1 << D135_PRED) |
- (1 << D117_PRED) | (1 << D153_PRED) |
- (1 << D207_PRED) | (1 << D63_PRED) |
- (1 << TM_PRED),
- INTRA_DC = (1 << DC_PRED),
- INTRA_DC_TM = (1 << DC_PRED) | (1 << TM_PRED),
- INTRA_DC_H_V = (1 << DC_PRED) | (1 << V_PRED) | (1 << H_PRED),
- INTRA_DC_TM_H_V = (1 << DC_PRED) | (1 << TM_PRED) | (1 << V_PRED) |
- (1 << H_PRED)
+ INTRA_ALL = (1 << DC_PRED) | (1 << V_PRED) | (1 << H_PRED) | (1 << D45_PRED) |
+ (1 << D135_PRED) | (1 << D117_PRED) | (1 << D153_PRED) |
+ (1 << D207_PRED) | (1 << D63_PRED) | (1 << TM_PRED),
+ INTRA_DC = (1 << DC_PRED),
+ INTRA_DC_TM = (1 << DC_PRED) | (1 << TM_PRED),
+ INTRA_DC_H_V = (1 << DC_PRED) | (1 << V_PRED) | (1 << H_PRED),
+ INTRA_DC_TM_H_V =
+ (1 << DC_PRED) | (1 << TM_PRED) | (1 << V_PRED) | (1 << H_PRED)
};
enum {
@@ -42,20 +39,15 @@
};
enum {
- DISABLE_ALL_INTER_SPLIT = (1 << THR_COMP_GA) |
- (1 << THR_COMP_LA) |
- (1 << THR_ALTR) |
- (1 << THR_GOLD) |
- (1 << THR_LAST),
+ DISABLE_ALL_INTER_SPLIT = (1 << THR_COMP_GA) | (1 << THR_COMP_LA) |
+ (1 << THR_ALTR) | (1 << THR_GOLD) | (1 << THR_LAST),
- DISABLE_ALL_SPLIT = (1 << THR_INTRA) | DISABLE_ALL_INTER_SPLIT,
+ DISABLE_ALL_SPLIT = (1 << THR_INTRA) | DISABLE_ALL_INTER_SPLIT,
- DISABLE_COMPOUND_SPLIT = (1 << THR_COMP_GA) | (1 << THR_COMP_LA),
+ DISABLE_COMPOUND_SPLIT = (1 << THR_COMP_GA) | (1 << THR_COMP_LA),
- LAST_AND_INTRA_SPLIT_ONLY = (1 << THR_COMP_GA) |
- (1 << THR_COMP_LA) |
- (1 << THR_ALTR) |
- (1 << THR_GOLD)
+ LAST_AND_INTRA_SPLIT_ONLY = (1 << THR_COMP_GA) | (1 << THR_COMP_LA) |
+ (1 << THR_ALTR) | (1 << THR_GOLD)
};
typedef enum {
diff --git a/vp10/encoder/subexp.c b/vp10/encoder/subexp.c
index d407477..db76a27 100644
--- a/vp10/encoder/subexp.c
+++ b/vp10/encoder/subexp.c
@@ -14,26 +14,29 @@
#include "vp10/encoder/cost.h"
#include "vp10/encoder/subexp.h"
-#define vp10_cost_upd256 ((int)(vp10_cost_one(upd) - vp10_cost_zero(upd)))
+#define vp10_cost_upd256 ((int)(vp10_cost_one(upd) - vp10_cost_zero(upd)))
static const uint8_t update_bits[255] = {
- 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
- 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
- 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
- 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
- 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
- 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
- 10, 11 - CONFIG_MISC_FIXES,
- 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
- 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
- 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
- 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
- 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
- 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
- 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
- 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 0,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+ 6, 6, 6, 6, 6, 6, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+ 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 11 - CONFIG_MISC_FIXES,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+ 11, 11, 11, 11, 11, 11, 11, 0,
};
static int recenter_nonneg(int v, int m) {
@@ -50,23 +53,23 @@
static const uint8_t map_table[MAX_PROB - 1] = {
// generated by:
// map_table[j] = split_index(j, MAX_PROB - 1, MODULUS_PARAM);
- 20, 21, 22, 23, 24, 25, 0, 26, 27, 28, 29, 30, 31, 32, 33,
- 34, 35, 36, 37, 1, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
- 48, 49, 2, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61,
- 3, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 4, 74,
- 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 5, 86, 87, 88,
- 89, 90, 91, 92, 93, 94, 95, 96, 97, 6, 98, 99, 100, 101, 102,
- 103, 104, 105, 106, 107, 108, 109, 7, 110, 111, 112, 113, 114, 115, 116,
- 117, 118, 119, 120, 121, 8, 122, 123, 124, 125, 126, 127, 128, 129, 130,
- 131, 132, 133, 9, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144,
- 145, 10, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 11,
- 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 12, 170, 171,
- 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 13, 182, 183, 184, 185,
- 186, 187, 188, 189, 190, 191, 192, 193, 14, 194, 195, 196, 197, 198, 199,
- 200, 201, 202, 203, 204, 205, 15, 206, 207, 208, 209, 210, 211, 212, 213,
- 214, 215, 216, 217, 16, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227,
- 228, 229, 17, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241,
- 18, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 19,
+ 20, 21, 22, 23, 24, 25, 0, 26, 27, 28, 29, 30, 31, 32, 33,
+ 34, 35, 36, 37, 1, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
+ 48, 49, 2, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61,
+ 3, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 4, 74,
+ 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 5, 86, 87, 88,
+ 89, 90, 91, 92, 93, 94, 95, 96, 97, 6, 98, 99, 100, 101, 102,
+ 103, 104, 105, 106, 107, 108, 109, 7, 110, 111, 112, 113, 114, 115, 116,
+ 117, 118, 119, 120, 121, 8, 122, 123, 124, 125, 126, 127, 128, 129, 130,
+ 131, 132, 133, 9, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144,
+ 145, 10, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 11,
+ 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 12, 170, 171,
+ 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 13, 182, 183, 184, 185,
+ 186, 187, 188, 189, 190, 191, 192, 193, 14, 194, 195, 196, 197, 198, 199,
+ 200, 201, 202, 203, 204, 205, 15, 206, 207, 208, 209, 210, 211, 212, 213,
+ 214, 215, 216, 217, 16, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227,
+ 228, 229, 17, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241,
+ 18, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 19,
};
v--;
m--;
@@ -117,9 +120,8 @@
encode_term_subexp(w, delp);
}
-int vp10_prob_diff_update_savings_search(const unsigned int *ct,
- vpx_prob oldp, vpx_prob *bestp,
- vpx_prob upd) {
+int vp10_prob_diff_update_savings_search(const unsigned int *ct, vpx_prob oldp,
+ vpx_prob *bestp, vpx_prob upd) {
const int old_b = cost_branch256(ct, oldp);
int bestsavings = 0;
vpx_prob newp, bestnewp = oldp;
@@ -139,10 +141,9 @@
}
int vp10_prob_diff_update_savings_search_model(const unsigned int *ct,
- const vpx_prob *oldp,
- vpx_prob *bestp,
- vpx_prob upd,
- int stepsize) {
+ const vpx_prob *oldp,
+ vpx_prob *bestp, vpx_prob upd,
+ int stepsize) {
int i, old_b, new_b, update_b, savings, bestsavings, step;
int newp;
vpx_prob bestnewp, newplist[ENTROPY_NODES], oldplist[ENTROPY_NODES];
@@ -158,15 +159,14 @@
if (*bestp > oldp[PIVOT_NODE]) {
step = -stepsize;
for (newp = *bestp; newp > oldp[PIVOT_NODE]; newp += step) {
- if (newp < 1 || newp > 255)
- continue;
+ if (newp < 1 || newp > 255) continue;
newplist[PIVOT_NODE] = newp;
vp10_model_to_full_probs(newplist, newplist);
for (i = UNCONSTRAINED_NODES, new_b = 0; i < ENTROPY_NODES; ++i)
new_b += cost_branch256(ct + 2 * i, newplist[i]);
new_b += cost_branch256(ct + 2 * PIVOT_NODE, newplist[PIVOT_NODE]);
- update_b = prob_diff_update_cost(newp, oldp[PIVOT_NODE]) +
- vp10_cost_upd256;
+ update_b =
+ prob_diff_update_cost(newp, oldp[PIVOT_NODE]) + vp10_cost_upd256;
savings = old_b - new_b - update_b;
if (savings > bestsavings) {
bestsavings = savings;
@@ -176,15 +176,14 @@
} else {
step = stepsize;
for (newp = *bestp; newp < oldp[PIVOT_NODE]; newp += step) {
- if (newp < 1 || newp > 255)
- continue;
+ if (newp < 1 || newp > 255) continue;
newplist[PIVOT_NODE] = newp;
vp10_model_to_full_probs(newplist, newplist);
for (i = UNCONSTRAINED_NODES, new_b = 0; i < ENTROPY_NODES; ++i)
new_b += cost_branch256(ct + 2 * i, newplist[i]);
new_b += cost_branch256(ct + 2 * PIVOT_NODE, newplist[PIVOT_NODE]);
- update_b = prob_diff_update_cost(newp, oldp[PIVOT_NODE]) +
- vp10_cost_upd256;
+ update_b =
+ prob_diff_update_cost(newp, oldp[PIVOT_NODE]) + vp10_cost_upd256;
savings = old_b - new_b - update_b;
if (savings > bestsavings) {
bestsavings = savings;
@@ -198,11 +197,11 @@
}
void vp10_cond_prob_diff_update(vpx_writer *w, vpx_prob *oldp,
- const unsigned int ct[2]) {
+ const unsigned int ct[2]) {
const vpx_prob upd = DIFF_UPDATE_PROB;
vpx_prob newp = get_binary_prob(ct[0], ct[1]);
- const int savings = vp10_prob_diff_update_savings_search(ct, *oldp, &newp,
- upd);
+ const int savings =
+ vp10_prob_diff_update_savings_search(ct, *oldp, &newp, upd);
assert(newp >= 1);
if (savings > 0) {
vpx_write(w, 1, upd);
@@ -217,7 +216,7 @@
const unsigned int ct[2]) {
const vpx_prob upd = DIFF_UPDATE_PROB;
vpx_prob newp = get_binary_prob(ct[0], ct[1]);
- const int savings = vp10_prob_diff_update_savings_search(ct, *oldp, &newp,
- upd);
+ const int savings =
+ vp10_prob_diff_update_savings_search(ct, *oldp, &newp, upd);
return savings;
}
diff --git a/vp10/encoder/subexp.h b/vp10/encoder/subexp.h
index 091334f..5adc6e5 100644
--- a/vp10/encoder/subexp.h
+++ b/vp10/encoder/subexp.h
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#ifndef VP10_ENCODER_SUBEXP_H_
#define VP10_ENCODER_SUBEXP_H_
@@ -20,22 +19,19 @@
struct vpx_writer;
-void vp10_write_prob_diff_update(struct vpx_writer *w,
- vpx_prob newp, vpx_prob oldp);
+void vp10_write_prob_diff_update(struct vpx_writer *w, vpx_prob newp,
+ vpx_prob oldp);
void vp10_cond_prob_diff_update(struct vpx_writer *w, vpx_prob *oldp,
- const unsigned int ct[2]);
+ const unsigned int ct[2]);
-int vp10_prob_diff_update_savings_search(const unsigned int *ct,
- vpx_prob oldp, vpx_prob *bestp,
- vpx_prob upd);
-
+int vp10_prob_diff_update_savings_search(const unsigned int *ct, vpx_prob oldp,
+ vpx_prob *bestp, vpx_prob upd);
int vp10_prob_diff_update_savings_search_model(const unsigned int *ct,
- const vpx_prob *oldp,
- vpx_prob *bestp,
- vpx_prob upd,
- int stepsize);
+ const vpx_prob *oldp,
+ vpx_prob *bestp, vpx_prob upd,
+ int stepsize);
int vp10_cond_prob_diff_update_savings(vpx_prob *oldp,
const unsigned int ct[2]);
diff --git a/vp10/encoder/temporal_filter.c b/vp10/encoder/temporal_filter.c
index 51c7324..a88564d 100644
--- a/vp10/encoder/temporal_filter.c
+++ b/vp10/encoder/temporal_filter.c
@@ -31,22 +31,14 @@
static int fixed_divide[512];
-static void temporal_filter_predictors_mb_c(MACROBLOCKD *xd,
- uint8_t *y_mb_ptr,
- uint8_t *u_mb_ptr,
- uint8_t *v_mb_ptr,
- int stride,
- int uv_block_width,
- int uv_block_height,
- int mv_row,
- int mv_col,
- uint8_t *pred,
- struct scale_factors *scale,
- int x, int y) {
+static void temporal_filter_predictors_mb_c(
+ MACROBLOCKD *xd, uint8_t *y_mb_ptr, uint8_t *u_mb_ptr, uint8_t *v_mb_ptr,
+ int stride, int uv_block_width, int uv_block_height, int mv_row, int mv_col,
+ uint8_t *pred, struct scale_factors *scale, int x, int y) {
const int which_mv = 0;
const MV mv = { mv_row, mv_col };
const InterpKernel *const kernel =
- vp10_filter_kernels[xd->mi[0]->mbmi.interp_filter];
+ vp10_filter_kernels[xd->mi[0]->mbmi.interp_filter];
enum mv_precision mv_precision_uv;
int uv_stride;
@@ -60,74 +52,46 @@
#if CONFIG_VPX_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- vp10_highbd_build_inter_predictor(y_mb_ptr, stride,
- &pred[0], 16,
- &mv,
- scale,
- 16, 16,
- which_mv,
- kernel, MV_PRECISION_Q3, x, y, xd->bd);
+ vp10_highbd_build_inter_predictor(y_mb_ptr, stride, &pred[0], 16, &mv,
+ scale, 16, 16, which_mv, kernel,
+ MV_PRECISION_Q3, x, y, xd->bd);
- vp10_highbd_build_inter_predictor(u_mb_ptr, uv_stride,
- &pred[256], uv_block_width,
- &mv,
- scale,
- uv_block_width, uv_block_height,
- which_mv,
- kernel, mv_precision_uv, x, y, xd->bd);
+ vp10_highbd_build_inter_predictor(u_mb_ptr, uv_stride, &pred[256],
+ uv_block_width, &mv, scale,
+ uv_block_width, uv_block_height, which_mv,
+ kernel, mv_precision_uv, x, y, xd->bd);
- vp10_highbd_build_inter_predictor(v_mb_ptr, uv_stride,
- &pred[512], uv_block_width,
- &mv,
- scale,
- uv_block_width, uv_block_height,
- which_mv,
- kernel, mv_precision_uv, x, y, xd->bd);
+ vp10_highbd_build_inter_predictor(v_mb_ptr, uv_stride, &pred[512],
+ uv_block_width, &mv, scale,
+ uv_block_width, uv_block_height, which_mv,
+ kernel, mv_precision_uv, x, y, xd->bd);
return;
}
#endif // CONFIG_VPX_HIGHBITDEPTH
- vp10_build_inter_predictor(y_mb_ptr, stride,
- &pred[0], 16,
- &mv,
- scale,
- 16, 16,
- which_mv,
- kernel, MV_PRECISION_Q3, x, y);
+ vp10_build_inter_predictor(y_mb_ptr, stride, &pred[0], 16, &mv, scale, 16, 16,
+ which_mv, kernel, MV_PRECISION_Q3, x, y);
- vp10_build_inter_predictor(u_mb_ptr, uv_stride,
- &pred[256], uv_block_width,
- &mv,
- scale,
- uv_block_width, uv_block_height,
- which_mv,
- kernel, mv_precision_uv, x, y);
+ vp10_build_inter_predictor(u_mb_ptr, uv_stride, &pred[256], uv_block_width,
+ &mv, scale, uv_block_width, uv_block_height,
+ which_mv, kernel, mv_precision_uv, x, y);
- vp10_build_inter_predictor(v_mb_ptr, uv_stride,
- &pred[512], uv_block_width,
- &mv,
- scale,
- uv_block_width, uv_block_height,
- which_mv,
- kernel, mv_precision_uv, x, y);
+ vp10_build_inter_predictor(v_mb_ptr, uv_stride, &pred[512], uv_block_width,
+ &mv, scale, uv_block_width, uv_block_height,
+ which_mv, kernel, mv_precision_uv, x, y);
}
void vp10_temporal_filter_init(void) {
int i;
fixed_divide[0] = 0;
- for (i = 1; i < 512; ++i)
- fixed_divide[i] = 0x80000 / i;
+ for (i = 1; i < 512; ++i) fixed_divide[i] = 0x80000 / i;
}
-void vp10_temporal_filter_apply_c(uint8_t *frame1,
- unsigned int stride,
- uint8_t *frame2,
- unsigned int block_width,
- unsigned int block_height,
- int strength,
- int filter_weight,
- unsigned int *accumulator,
- uint16_t *count) {
+void vp10_temporal_filter_apply_c(uint8_t *frame1, unsigned int stride,
+ uint8_t *frame2, unsigned int block_width,
+ unsigned int block_height, int strength,
+ int filter_weight, unsigned int *accumulator,
+ uint16_t *count) {
unsigned int i, j, k;
int modifier;
int byte = 0;
@@ -138,17 +102,16 @@
int src_byte = frame1[byte];
int pixel_value = *frame2++;
- modifier = src_byte - pixel_value;
+ modifier = src_byte - pixel_value;
// This is an integer approximation of:
// float coeff = (3.0 * modifer * modifier) / pow(2, strength);
// modifier = (int)roundf(coeff > 16 ? 0 : 16-coeff);
- modifier *= modifier;
- modifier *= 3;
- modifier += rounding;
+ modifier *= modifier;
+ modifier *= 3;
+ modifier += rounding;
modifier >>= strength;
- if (modifier > 16)
- modifier = 16;
+ if (modifier > 16) modifier = 16;
modifier = 16 - modifier;
modifier *= filter_weight;
@@ -164,15 +127,10 @@
}
#if CONFIG_VPX_HIGHBITDEPTH
-void vp10_highbd_temporal_filter_apply_c(uint8_t *frame1_8,
- unsigned int stride,
- uint8_t *frame2_8,
- unsigned int block_width,
- unsigned int block_height,
- int strength,
- int filter_weight,
- unsigned int *accumulator,
- uint16_t *count) {
+void vp10_highbd_temporal_filter_apply_c(
+ uint8_t *frame1_8, unsigned int stride, uint8_t *frame2_8,
+ unsigned int block_width, unsigned int block_height, int strength,
+ int filter_weight, unsigned int *accumulator, uint16_t *count) {
uint16_t *frame1 = CONVERT_TO_SHORTPTR(frame1_8);
uint16_t *frame2 = CONVERT_TO_SHORTPTR(frame2_8);
unsigned int i, j, k;
@@ -185,7 +143,7 @@
int src_byte = frame1[byte];
int pixel_value = *frame2++;
- modifier = src_byte - pixel_value;
+ modifier = src_byte - pixel_value;
// This is an integer approximation of:
// float coeff = (3.0 * modifer * modifier) / pow(2, strength);
// modifier = (int)roundf(coeff > 16 ? 0 : 16-coeff);
@@ -194,8 +152,7 @@
modifier += rounding;
modifier >>= strength;
- if (modifier > 16)
- modifier = 16;
+ if (modifier > 16) modifier = 16;
modifier = 16 - modifier;
modifier *= filter_weight;
@@ -225,7 +182,7 @@
unsigned int sse;
int cost_list[5];
- MV best_ref_mv1 = {0, 0};
+ MV best_ref_mv1 = { 0, 0 };
MV best_ref_mv1_full; /* full-pixel value of best_ref_mv1 */
MV *ref_mv = &x->e_mbd.mi[0]->bmi[0].as_mv[0].as_mv;
@@ -247,19 +204,15 @@
// Ignore mv costing by sending NULL pointer instead of cost arrays
vp10_hex_search(x, &best_ref_mv1_full, step_param, sadpb, 1,
- cond_cost_list(cpi, cost_list),
- &cpi->fn_ptr[BLOCK_16X16], 0, &best_ref_mv1, ref_mv);
+ cond_cost_list(cpi, cost_list), &cpi->fn_ptr[BLOCK_16X16], 0,
+ &best_ref_mv1, ref_mv);
// Ignore mv costing by sending NULL pointer instead of cost array
- bestsme = cpi->find_fractional_mv_step(x, ref_mv,
- &best_ref_mv1,
- cpi->common.allow_high_precision_mv,
- x->errorperbit,
- &cpi->fn_ptr[BLOCK_16X16],
- 0, mv_sf->subpel_iters_per_step,
- cond_cost_list(cpi, cost_list),
- NULL, NULL,
- &distortion, &sse, NULL, 0, 0);
+ bestsme = cpi->find_fractional_mv_step(
+ x, ref_mv, &best_ref_mv1, cpi->common.allow_high_precision_mv,
+ x->errorperbit, &cpi->fn_ptr[BLOCK_16X16], 0,
+ mv_sf->subpel_iters_per_step, cond_cost_list(cpi, cost_list), NULL, NULL,
+ &distortion, &sse, NULL, 0, 0);
// Restore input state
x->plane[0].src = src;
@@ -270,8 +223,7 @@
static void temporal_filter_iterate_c(VP10_COMP *cpi,
YV12_BUFFER_CONFIG **frames,
- int frame_count,
- int alt_ref_index,
+ int frame_count, int alt_ref_index,
int strength,
struct scale_factors *scale) {
int byte;
@@ -288,17 +240,17 @@
YV12_BUFFER_CONFIG *f = frames[alt_ref_index];
uint8_t *dst1, *dst2;
#if CONFIG_VPX_HIGHBITDEPTH
- DECLARE_ALIGNED(16, uint16_t, predictor16[16 * 16 * 3]);
- DECLARE_ALIGNED(16, uint8_t, predictor8[16 * 16 * 3]);
+ DECLARE_ALIGNED(16, uint16_t, predictor16[16 * 16 * 3]);
+ DECLARE_ALIGNED(16, uint8_t, predictor8[16 * 16 * 3]);
uint8_t *predictor;
#else
- DECLARE_ALIGNED(16, uint8_t, predictor[16 * 16 * 3]);
+ DECLARE_ALIGNED(16, uint8_t, predictor[16 * 16 * 3]);
#endif
const int mb_uv_height = 16 >> mbd->plane[1].subsampling_y;
- const int mb_uv_width = 16 >> mbd->plane[1].subsampling_x;
+ const int mb_uv_width = 16 >> mbd->plane[1].subsampling_x;
// Save input state
- uint8_t* input_buffer[MAX_MB_PLANE];
+ uint8_t *input_buffer[MAX_MB_PLANE];
int i;
#if CONFIG_VPX_HIGHBITDEPTH
if (mbd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
@@ -308,8 +260,7 @@
}
#endif
- for (i = 0; i < MAX_MB_PLANE; i++)
- input_buffer[i] = mbd->plane[i].pre[0].buf;
+ for (i = 0; i < MAX_MB_PLANE; i++) input_buffer[i] = mbd->plane[i].pre[0].buf;
for (mb_row = 0; mb_row < mb_rows; mb_row++) {
// Source frames are extended to 16 pixels. This is different than
@@ -324,8 +275,8 @@
// To keep the mv in play for both Y and UV planes the max that it
// can be on a border is therefore 16 - (2*VPX_INTERP_EXTEND+1).
cpi->td.mb.mv_row_min = -((mb_row * 16) + (17 - 2 * VPX_INTERP_EXTEND));
- cpi->td.mb.mv_row_max = ((mb_rows - 1 - mb_row) * 16)
- + (17 - 2 * VPX_INTERP_EXTEND);
+ cpi->td.mb.mv_row_max =
+ ((mb_rows - 1 - mb_row) * 16) + (17 - 2 * VPX_INTERP_EXTEND);
for (mb_col = 0; mb_col < mb_cols; mb_col++) {
int i, j, k;
@@ -335,15 +286,14 @@
memset(count, 0, 16 * 16 * 3 * sizeof(count[0]));
cpi->td.mb.mv_col_min = -((mb_col * 16) + (17 - 2 * VPX_INTERP_EXTEND));
- cpi->td.mb.mv_col_max = ((mb_cols - 1 - mb_col) * 16)
- + (17 - 2 * VPX_INTERP_EXTEND);
+ cpi->td.mb.mv_col_max =
+ ((mb_cols - 1 - mb_col) * 16) + (17 - 2 * VPX_INTERP_EXTEND);
for (frame = 0; frame < frame_count; frame++) {
- const int thresh_low = 10000;
+ const int thresh_low = 10000;
const int thresh_high = 20000;
- if (frames[frame] == NULL)
- continue;
+ if (frames[frame] == NULL) continue;
mbd->mi[0]->bmi[0].as_mv[0].as_mv.row = 0;
mbd->mi[0]->bmi[0].as_mv[0].as_mv.col = 0;
@@ -352,84 +302,68 @@
filter_weight = 2;
} else {
// Find best match in this frame by MC
- int err = temporal_filter_find_matching_mb_c(cpi,
- frames[alt_ref_index]->y_buffer + mb_y_offset,
- frames[frame]->y_buffer + mb_y_offset,
- frames[frame]->y_stride);
+ int err = temporal_filter_find_matching_mb_c(
+ cpi, frames[alt_ref_index]->y_buffer + mb_y_offset,
+ frames[frame]->y_buffer + mb_y_offset, frames[frame]->y_stride);
// Assign higher weight to matching MB if it's error
// score is lower. If not applying MC default behavior
// is to weight all MBs equal.
- filter_weight = err < thresh_low
- ? 2 : err < thresh_high ? 1 : 0;
+ filter_weight = err < thresh_low ? 2 : err < thresh_high ? 1 : 0;
}
if (filter_weight != 0) {
// Construct the predictors
- temporal_filter_predictors_mb_c(mbd,
- frames[frame]->y_buffer + mb_y_offset,
+ temporal_filter_predictors_mb_c(
+ mbd, frames[frame]->y_buffer + mb_y_offset,
frames[frame]->u_buffer + mb_uv_offset,
- frames[frame]->v_buffer + mb_uv_offset,
- frames[frame]->y_stride,
- mb_uv_width, mb_uv_height,
- mbd->mi[0]->bmi[0].as_mv[0].as_mv.row,
- mbd->mi[0]->bmi[0].as_mv[0].as_mv.col,
- predictor, scale,
+ frames[frame]->v_buffer + mb_uv_offset, frames[frame]->y_stride,
+ mb_uv_width, mb_uv_height, mbd->mi[0]->bmi[0].as_mv[0].as_mv.row,
+ mbd->mi[0]->bmi[0].as_mv[0].as_mv.col, predictor, scale,
mb_col * 16, mb_row * 16);
#if CONFIG_VPX_HIGHBITDEPTH
if (mbd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
int adj_strength = strength + 2 * (mbd->bd - 8);
// Apply the filter (YUV)
- vp10_highbd_temporal_filter_apply(f->y_buffer + mb_y_offset,
- f->y_stride,
- predictor, 16, 16, adj_strength,
- filter_weight,
- accumulator, count);
- vp10_highbd_temporal_filter_apply(f->u_buffer + mb_uv_offset,
- f->uv_stride, predictor + 256,
- mb_uv_width, mb_uv_height,
- adj_strength,
- filter_weight, accumulator + 256,
- count + 256);
- vp10_highbd_temporal_filter_apply(f->v_buffer + mb_uv_offset,
- f->uv_stride, predictor + 512,
- mb_uv_width, mb_uv_height,
- adj_strength, filter_weight,
- accumulator + 512, count + 512);
+ vp10_highbd_temporal_filter_apply(
+ f->y_buffer + mb_y_offset, f->y_stride, predictor, 16, 16,
+ adj_strength, filter_weight, accumulator, count);
+ vp10_highbd_temporal_filter_apply(
+ f->u_buffer + mb_uv_offset, f->uv_stride, predictor + 256,
+ mb_uv_width, mb_uv_height, adj_strength, filter_weight,
+ accumulator + 256, count + 256);
+ vp10_highbd_temporal_filter_apply(
+ f->v_buffer + mb_uv_offset, f->uv_stride, predictor + 512,
+ mb_uv_width, mb_uv_height, adj_strength, filter_weight,
+ accumulator + 512, count + 512);
} else {
// Apply the filter (YUV)
vp10_temporal_filter_apply(f->y_buffer + mb_y_offset, f->y_stride,
- predictor, 16, 16,
- strength, filter_weight,
- accumulator, count);
+ predictor, 16, 16, strength,
+ filter_weight, accumulator, count);
vp10_temporal_filter_apply(f->u_buffer + mb_uv_offset, f->uv_stride,
- predictor + 256,
- mb_uv_width, mb_uv_height, strength,
- filter_weight, accumulator + 256,
- count + 256);
+ predictor + 256, mb_uv_width,
+ mb_uv_height, strength, filter_weight,
+ accumulator + 256, count + 256);
vp10_temporal_filter_apply(f->v_buffer + mb_uv_offset, f->uv_stride,
- predictor + 512,
- mb_uv_width, mb_uv_height, strength,
- filter_weight, accumulator + 512,
- count + 512);
+ predictor + 512, mb_uv_width,
+ mb_uv_height, strength, filter_weight,
+ accumulator + 512, count + 512);
}
#else
// Apply the filter (YUV)
vp10_temporal_filter_apply(f->y_buffer + mb_y_offset, f->y_stride,
- predictor, 16, 16,
- strength, filter_weight,
- accumulator, count);
+ predictor, 16, 16, strength, filter_weight,
+ accumulator, count);
vp10_temporal_filter_apply(f->u_buffer + mb_uv_offset, f->uv_stride,
- predictor + 256,
- mb_uv_width, mb_uv_height, strength,
- filter_weight, accumulator + 256,
- count + 256);
+ predictor + 256, mb_uv_width, mb_uv_height,
+ strength, filter_weight, accumulator + 256,
+ count + 256);
vp10_temporal_filter_apply(f->v_buffer + mb_uv_offset, f->uv_stride,
- predictor + 512,
- mb_uv_width, mb_uv_height, strength,
- filter_weight, accumulator + 512,
- count + 512);
+ predictor + 512, mb_uv_width, mb_uv_height,
+ strength, filter_weight, accumulator + 512,
+ count + 512);
#endif // CONFIG_VPX_HIGHBITDEPTH
}
}
@@ -584,13 +518,11 @@
}
// Restore input state
- for (i = 0; i < MAX_MB_PLANE; i++)
- mbd->plane[i].pre[0].buf = input_buffer[i];
+ for (i = 0; i < MAX_MB_PLANE; i++) mbd->plane[i].pre[0].buf = input_buffer[i];
}
// Apply buffer limits and context specific adjustments to arnr filter.
-static void adjust_arnr_filter(VP10_COMP *cpi,
- int distance, int group_boost,
+static void adjust_arnr_filter(VP10_COMP *cpi, int distance, int group_boost,
int *arnr_frames, int *arnr_strength) {
const VP10EncoderConfig *const oxcf = &cpi->oxcf;
const int frames_after_arf =
@@ -600,34 +532,30 @@
int q, frames, strength;
// Define the forward and backwards filter limits for this arnr group.
- if (frames_fwd > frames_after_arf)
- frames_fwd = frames_after_arf;
- if (frames_fwd > distance)
- frames_fwd = distance;
+ if (frames_fwd > frames_after_arf) frames_fwd = frames_after_arf;
+ if (frames_fwd > distance) frames_fwd = distance;
frames_bwd = frames_fwd;
// For even length filter there is one more frame backward
// than forward: e.g. len=6 ==> bbbAff, len=7 ==> bbbAfff.
- if (frames_bwd < distance)
- frames_bwd += (oxcf->arnr_max_frames + 1) & 0x1;
+ if (frames_bwd < distance) frames_bwd += (oxcf->arnr_max_frames + 1) & 0x1;
// Set the baseline active filter size.
frames = frames_bwd + 1 + frames_fwd;
// Adjust the strength based on active max q.
if (cpi->common.current_video_frame > 1)
- q = ((int)vp10_convert_qindex_to_q(
- cpi->rc.avg_frame_qindex[INTER_FRAME], cpi->common.bit_depth));
+ q = ((int)vp10_convert_qindex_to_q(cpi->rc.avg_frame_qindex[INTER_FRAME],
+ cpi->common.bit_depth));
else
- q = ((int)vp10_convert_qindex_to_q(
- cpi->rc.avg_frame_qindex[KEY_FRAME], cpi->common.bit_depth));
+ q = ((int)vp10_convert_qindex_to_q(cpi->rc.avg_frame_qindex[KEY_FRAME],
+ cpi->common.bit_depth));
if (q > 16) {
strength = oxcf->arnr_strength;
} else {
strength = oxcf->arnr_strength - ((16 - q) / 2);
- if (strength < 0)
- strength = 0;
+ if (strength < 0) strength = 0;
}
// Adjust number of frames in filter and strength based on gf boost level.
@@ -661,7 +589,7 @@
int frames_to_blur_backward;
int frames_to_blur_forward;
struct scale_factors sf;
- YV12_BUFFER_CONFIG *frames[MAX_LAG_BUFFERS] = {NULL};
+ YV12_BUFFER_CONFIG *frames[MAX_LAG_BUFFERS] = { NULL };
// Apply context specific adjustments to the arnr filter parameters.
adjust_arnr_filter(cpi, distance, rc->gfu_boost, &frames_to_blur, &strength);
@@ -672,28 +600,24 @@
// Setup frame pointers, NULL indicates frame not included in filter.
for (frame = 0; frame < frames_to_blur; ++frame) {
const int which_buffer = start_frame - frame;
- struct lookahead_entry *buf = vp10_lookahead_peek(cpi->lookahead,
- which_buffer);
+ struct lookahead_entry *buf =
+ vp10_lookahead_peek(cpi->lookahead, which_buffer);
frames[frames_to_blur - 1 - frame] = &buf->img;
}
if (frames_to_blur > 0) {
- // Setup scaling factors. Scaling on each of the arnr frames is not
- // supported.
- // ARF is produced at the native frame size and resized when coded.
+// Setup scaling factors. Scaling on each of the arnr frames is not
+// supported.
+// ARF is produced at the native frame size and resized when coded.
#if CONFIG_VPX_HIGHBITDEPTH
- vp10_setup_scale_factors_for_frame(&sf,
- frames[0]->y_crop_width,
- frames[0]->y_crop_height,
- frames[0]->y_crop_width,
- frames[0]->y_crop_height,
- cpi->common.use_highbitdepth);
+ vp10_setup_scale_factors_for_frame(
+ &sf, frames[0]->y_crop_width, frames[0]->y_crop_height,
+ frames[0]->y_crop_width, frames[0]->y_crop_height,
+ cpi->common.use_highbitdepth);
#else
- vp10_setup_scale_factors_for_frame(&sf,
- frames[0]->y_crop_width,
- frames[0]->y_crop_height,
- frames[0]->y_crop_width,
- frames[0]->y_crop_height);
+ vp10_setup_scale_factors_for_frame(
+ &sf, frames[0]->y_crop_width, frames[0]->y_crop_height,
+ frames[0]->y_crop_width, frames[0]->y_crop_height);
#endif // CONFIG_VPX_HIGHBITDEPTH
}
diff --git a/vp10/encoder/tokenize.c b/vp10/encoder/tokenize.c
index 46f3b65..60960c6 100644
--- a/vp10/encoder/tokenize.c
+++ b/vp10/encoder/tokenize.c
@@ -25,439 +25,456 @@
#include "vp10/encoder/tokenize.h"
static const TOKENVALUE dct_cat_lt_10_value_tokens[] = {
- {9, 63}, {9, 61}, {9, 59}, {9, 57}, {9, 55}, {9, 53}, {9, 51}, {9, 49},
- {9, 47}, {9, 45}, {9, 43}, {9, 41}, {9, 39}, {9, 37}, {9, 35}, {9, 33},
- {9, 31}, {9, 29}, {9, 27}, {9, 25}, {9, 23}, {9, 21}, {9, 19}, {9, 17},
- {9, 15}, {9, 13}, {9, 11}, {9, 9}, {9, 7}, {9, 5}, {9, 3}, {9, 1},
- {8, 31}, {8, 29}, {8, 27}, {8, 25}, {8, 23}, {8, 21},
- {8, 19}, {8, 17}, {8, 15}, {8, 13}, {8, 11}, {8, 9},
- {8, 7}, {8, 5}, {8, 3}, {8, 1},
- {7, 15}, {7, 13}, {7, 11}, {7, 9}, {7, 7}, {7, 5}, {7, 3}, {7, 1},
- {6, 7}, {6, 5}, {6, 3}, {6, 1}, {5, 3}, {5, 1},
- {4, 1}, {3, 1}, {2, 1}, {1, 1}, {0, 0},
- {1, 0}, {2, 0}, {3, 0}, {4, 0},
- {5, 0}, {5, 2}, {6, 0}, {6, 2}, {6, 4}, {6, 6},
- {7, 0}, {7, 2}, {7, 4}, {7, 6}, {7, 8}, {7, 10}, {7, 12}, {7, 14},
- {8, 0}, {8, 2}, {8, 4}, {8, 6}, {8, 8}, {8, 10}, {8, 12},
- {8, 14}, {8, 16}, {8, 18}, {8, 20}, {8, 22}, {8, 24},
- {8, 26}, {8, 28}, {8, 30}, {9, 0}, {9, 2},
- {9, 4}, {9, 6}, {9, 8}, {9, 10}, {9, 12}, {9, 14}, {9, 16},
- {9, 18}, {9, 20}, {9, 22}, {9, 24}, {9, 26}, {9, 28},
- {9, 30}, {9, 32}, {9, 34}, {9, 36}, {9, 38}, {9, 40},
- {9, 42}, {9, 44}, {9, 46}, {9, 48}, {9, 50}, {9, 52},
- {9, 54}, {9, 56}, {9, 58}, {9, 60}, {9, 62}
+ { 9, 63 }, { 9, 61 }, { 9, 59 }, { 9, 57 }, { 9, 55 }, { 9, 53 }, { 9, 51 },
+ { 9, 49 }, { 9, 47 }, { 9, 45 }, { 9, 43 }, { 9, 41 }, { 9, 39 }, { 9, 37 },
+ { 9, 35 }, { 9, 33 }, { 9, 31 }, { 9, 29 }, { 9, 27 }, { 9, 25 }, { 9, 23 },
+ { 9, 21 }, { 9, 19 }, { 9, 17 }, { 9, 15 }, { 9, 13 }, { 9, 11 }, { 9, 9 },
+ { 9, 7 }, { 9, 5 }, { 9, 3 }, { 9, 1 }, { 8, 31 }, { 8, 29 }, { 8, 27 },
+ { 8, 25 }, { 8, 23 }, { 8, 21 }, { 8, 19 }, { 8, 17 }, { 8, 15 }, { 8, 13 },
+ { 8, 11 }, { 8, 9 }, { 8, 7 }, { 8, 5 }, { 8, 3 }, { 8, 1 }, { 7, 15 },
+ { 7, 13 }, { 7, 11 }, { 7, 9 }, { 7, 7 }, { 7, 5 }, { 7, 3 }, { 7, 1 },
+ { 6, 7 }, { 6, 5 }, { 6, 3 }, { 6, 1 }, { 5, 3 }, { 5, 1 }, { 4, 1 },
+ { 3, 1 }, { 2, 1 }, { 1, 1 }, { 0, 0 }, { 1, 0 }, { 2, 0 }, { 3, 0 },
+ { 4, 0 }, { 5, 0 }, { 5, 2 }, { 6, 0 }, { 6, 2 }, { 6, 4 }, { 6, 6 },
+ { 7, 0 }, { 7, 2 }, { 7, 4 }, { 7, 6 }, { 7, 8 }, { 7, 10 }, { 7, 12 },
+ { 7, 14 }, { 8, 0 }, { 8, 2 }, { 8, 4 }, { 8, 6 }, { 8, 8 }, { 8, 10 },
+ { 8, 12 }, { 8, 14 }, { 8, 16 }, { 8, 18 }, { 8, 20 }, { 8, 22 }, { 8, 24 },
+ { 8, 26 }, { 8, 28 }, { 8, 30 }, { 9, 0 }, { 9, 2 }, { 9, 4 }, { 9, 6 },
+ { 9, 8 }, { 9, 10 }, { 9, 12 }, { 9, 14 }, { 9, 16 }, { 9, 18 }, { 9, 20 },
+ { 9, 22 }, { 9, 24 }, { 9, 26 }, { 9, 28 }, { 9, 30 }, { 9, 32 }, { 9, 34 },
+ { 9, 36 }, { 9, 38 }, { 9, 40 }, { 9, 42 }, { 9, 44 }, { 9, 46 }, { 9, 48 },
+ { 9, 50 }, { 9, 52 }, { 9, 54 }, { 9, 56 }, { 9, 58 }, { 9, 60 }, { 9, 62 }
};
-const TOKENVALUE *vp10_dct_cat_lt_10_value_tokens = dct_cat_lt_10_value_tokens +
- (sizeof(dct_cat_lt_10_value_tokens) / sizeof(*dct_cat_lt_10_value_tokens))
- / 2;
+const TOKENVALUE *vp10_dct_cat_lt_10_value_tokens =
+ dct_cat_lt_10_value_tokens +
+ (sizeof(dct_cat_lt_10_value_tokens) / sizeof(*dct_cat_lt_10_value_tokens)) /
+ 2;
// Array indices are identical to previously-existing CONTEXT_NODE indices
const vpx_tree_index vp10_coef_tree[TREE_SIZE(ENTROPY_TOKENS)] = {
- -EOB_TOKEN, 2, // 0 = EOB
- -ZERO_TOKEN, 4, // 1 = ZERO
- -ONE_TOKEN, 6, // 2 = ONE
- 8, 12, // 3 = LOW_VAL
- -TWO_TOKEN, 10, // 4 = TWO
- -THREE_TOKEN, -FOUR_TOKEN, // 5 = THREE
- 14, 16, // 6 = HIGH_LOW
- -CATEGORY1_TOKEN, -CATEGORY2_TOKEN, // 7 = CAT_ONE
- 18, 20, // 8 = CAT_THREEFOUR
- -CATEGORY3_TOKEN, -CATEGORY4_TOKEN, // 9 = CAT_THREE
- -CATEGORY5_TOKEN, -CATEGORY6_TOKEN // 10 = CAT_FIVE
+ -EOB_TOKEN,
+ 2, // 0 = EOB
+ -ZERO_TOKEN,
+ 4, // 1 = ZERO
+ -ONE_TOKEN,
+ 6, // 2 = ONE
+ 8,
+ 12, // 3 = LOW_VAL
+ -TWO_TOKEN,
+ 10, // 4 = TWO
+ -THREE_TOKEN,
+ -FOUR_TOKEN, // 5 = THREE
+ 14,
+ 16, // 6 = HIGH_LOW
+ -CATEGORY1_TOKEN,
+ -CATEGORY2_TOKEN, // 7 = CAT_ONE
+ 18,
+ 20, // 8 = CAT_THREEFOUR
+ -CATEGORY3_TOKEN,
+ -CATEGORY4_TOKEN, // 9 = CAT_THREE
+ -CATEGORY5_TOKEN,
+ -CATEGORY6_TOKEN // 10 = CAT_FIVE
};
-static const vpx_tree_index cat1[2] = {0, 0};
-static const vpx_tree_index cat2[4] = {2, 2, 0, 0};
-static const vpx_tree_index cat3[6] = {2, 2, 4, 4, 0, 0};
-static const vpx_tree_index cat4[8] = {2, 2, 4, 4, 6, 6, 0, 0};
-static const vpx_tree_index cat5[10] = {2, 2, 4, 4, 6, 6, 8, 8, 0, 0};
-static const vpx_tree_index cat6[28] = {2, 2, 4, 4, 6, 6, 8, 8, 10, 10, 12, 12,
- 14, 14, 16, 16, 18, 18, 20, 20, 22, 22, 24, 24, 26, 26, 0, 0};
+static const vpx_tree_index cat1[2] = { 0, 0 };
+static const vpx_tree_index cat2[4] = { 2, 2, 0, 0 };
+static const vpx_tree_index cat3[6] = { 2, 2, 4, 4, 0, 0 };
+static const vpx_tree_index cat4[8] = { 2, 2, 4, 4, 6, 6, 0, 0 };
+static const vpx_tree_index cat5[10] = { 2, 2, 4, 4, 6, 6, 8, 8, 0, 0 };
+static const vpx_tree_index cat6[28] = { 2, 2, 4, 4, 6, 6, 8, 8, 10, 10,
+ 12, 12, 14, 14, 16, 16, 18, 18, 20, 20,
+ 22, 22, 24, 24, 26, 26, 0, 0 };
-static const int16_t zero_cost[] = {0};
-static const int16_t sign_cost[] = {255, 257};
-static const int16_t cat1_cost[] = {429, 431, 616, 618};
-static const int16_t cat2_cost[] = {624, 626, 727, 729, 848, 850, 951, 953};
-static const int16_t cat3_cost[] = {
- 820, 822, 893, 895, 940, 942, 1013, 1015, 1096, 1098, 1169, 1171, 1216, 1218,
- 1289, 1291
-};
-static const int16_t cat4_cost[] = {
- 1032, 1034, 1075, 1077, 1105, 1107, 1148, 1150, 1194, 1196, 1237, 1239,
- 1267, 1269, 1310, 1312, 1328, 1330, 1371, 1373, 1401, 1403, 1444, 1446,
- 1490, 1492, 1533, 1535, 1563, 1565, 1606, 1608
-};
+static const int16_t zero_cost[] = { 0 };
+static const int16_t sign_cost[] = { 255, 257 };
+static const int16_t cat1_cost[] = { 429, 431, 616, 618 };
+static const int16_t cat2_cost[] = { 624, 626, 727, 729, 848, 850, 951, 953 };
+static const int16_t cat3_cost[] = { 820, 822, 893, 895, 940, 942,
+ 1013, 1015, 1096, 1098, 1169, 1171,
+ 1216, 1218, 1289, 1291 };
+static const int16_t cat4_cost[] = { 1032, 1034, 1075, 1077, 1105, 1107, 1148,
+ 1150, 1194, 1196, 1237, 1239, 1267, 1269,
+ 1310, 1312, 1328, 1330, 1371, 1373, 1401,
+ 1403, 1444, 1446, 1490, 1492, 1533, 1535,
+ 1563, 1565, 1606, 1608 };
static const int16_t cat5_cost[] = {
- 1269, 1271, 1283, 1285, 1306, 1308, 1320,
- 1322, 1347, 1349, 1361, 1363, 1384, 1386, 1398, 1400, 1443, 1445, 1457,
- 1459, 1480, 1482, 1494, 1496, 1521, 1523, 1535, 1537, 1558, 1560, 1572,
- 1574, 1592, 1594, 1606, 1608, 1629, 1631, 1643, 1645, 1670, 1672, 1684,
- 1686, 1707, 1709, 1721, 1723, 1766, 1768, 1780, 1782, 1803, 1805, 1817,
- 1819, 1844, 1846, 1858, 1860, 1881, 1883, 1895, 1897
+ 1269, 1271, 1283, 1285, 1306, 1308, 1320, 1322, 1347, 1349, 1361, 1363, 1384,
+ 1386, 1398, 1400, 1443, 1445, 1457, 1459, 1480, 1482, 1494, 1496, 1521, 1523,
+ 1535, 1537, 1558, 1560, 1572, 1574, 1592, 1594, 1606, 1608, 1629, 1631, 1643,
+ 1645, 1670, 1672, 1684, 1686, 1707, 1709, 1721, 1723, 1766, 1768, 1780, 1782,
+ 1803, 1805, 1817, 1819, 1844, 1846, 1858, 1860, 1881, 1883, 1895, 1897
};
const int16_t vp10_cat6_low_cost[256] = {
- 1638, 1640, 1646, 1648, 1652, 1654, 1660, 1662,
- 1670, 1672, 1678, 1680, 1684, 1686, 1692, 1694, 1711, 1713, 1719, 1721,
- 1725, 1727, 1733, 1735, 1743, 1745, 1751, 1753, 1757, 1759, 1765, 1767,
- 1787, 1789, 1795, 1797, 1801, 1803, 1809, 1811, 1819, 1821, 1827, 1829,
- 1833, 1835, 1841, 1843, 1860, 1862, 1868, 1870, 1874, 1876, 1882, 1884,
- 1892, 1894, 1900, 1902, 1906, 1908, 1914, 1916, 1940, 1942, 1948, 1950,
- 1954, 1956, 1962, 1964, 1972, 1974, 1980, 1982, 1986, 1988, 1994, 1996,
- 2013, 2015, 2021, 2023, 2027, 2029, 2035, 2037, 2045, 2047, 2053, 2055,
- 2059, 2061, 2067, 2069, 2089, 2091, 2097, 2099, 2103, 2105, 2111, 2113,
- 2121, 2123, 2129, 2131, 2135, 2137, 2143, 2145, 2162, 2164, 2170, 2172,
- 2176, 2178, 2184, 2186, 2194, 2196, 2202, 2204, 2208, 2210, 2216, 2218,
- 2082, 2084, 2090, 2092, 2096, 2098, 2104, 2106, 2114, 2116, 2122, 2124,
- 2128, 2130, 2136, 2138, 2155, 2157, 2163, 2165, 2169, 2171, 2177, 2179,
- 2187, 2189, 2195, 2197, 2201, 2203, 2209, 2211, 2231, 2233, 2239, 2241,
- 2245, 2247, 2253, 2255, 2263, 2265, 2271, 2273, 2277, 2279, 2285, 2287,
- 2304, 2306, 2312, 2314, 2318, 2320, 2326, 2328, 2336, 2338, 2344, 2346,
- 2350, 2352, 2358, 2360, 2384, 2386, 2392, 2394, 2398, 2400, 2406, 2408,
- 2416, 2418, 2424, 2426, 2430, 2432, 2438, 2440, 2457, 2459, 2465, 2467,
- 2471, 2473, 2479, 2481, 2489, 2491, 2497, 2499, 2503, 2505, 2511, 2513,
- 2533, 2535, 2541, 2543, 2547, 2549, 2555, 2557, 2565, 2567, 2573, 2575,
- 2579, 2581, 2587, 2589, 2606, 2608, 2614, 2616, 2620, 2622, 2628, 2630,
- 2638, 2640, 2646, 2648, 2652, 2654, 2660, 2662
+ 1638, 1640, 1646, 1648, 1652, 1654, 1660, 1662, 1670, 1672, 1678, 1680, 1684,
+ 1686, 1692, 1694, 1711, 1713, 1719, 1721, 1725, 1727, 1733, 1735, 1743, 1745,
+ 1751, 1753, 1757, 1759, 1765, 1767, 1787, 1789, 1795, 1797, 1801, 1803, 1809,
+ 1811, 1819, 1821, 1827, 1829, 1833, 1835, 1841, 1843, 1860, 1862, 1868, 1870,
+ 1874, 1876, 1882, 1884, 1892, 1894, 1900, 1902, 1906, 1908, 1914, 1916, 1940,
+ 1942, 1948, 1950, 1954, 1956, 1962, 1964, 1972, 1974, 1980, 1982, 1986, 1988,
+ 1994, 1996, 2013, 2015, 2021, 2023, 2027, 2029, 2035, 2037, 2045, 2047, 2053,
+ 2055, 2059, 2061, 2067, 2069, 2089, 2091, 2097, 2099, 2103, 2105, 2111, 2113,
+ 2121, 2123, 2129, 2131, 2135, 2137, 2143, 2145, 2162, 2164, 2170, 2172, 2176,
+ 2178, 2184, 2186, 2194, 2196, 2202, 2204, 2208, 2210, 2216, 2218, 2082, 2084,
+ 2090, 2092, 2096, 2098, 2104, 2106, 2114, 2116, 2122, 2124, 2128, 2130, 2136,
+ 2138, 2155, 2157, 2163, 2165, 2169, 2171, 2177, 2179, 2187, 2189, 2195, 2197,
+ 2201, 2203, 2209, 2211, 2231, 2233, 2239, 2241, 2245, 2247, 2253, 2255, 2263,
+ 2265, 2271, 2273, 2277, 2279, 2285, 2287, 2304, 2306, 2312, 2314, 2318, 2320,
+ 2326, 2328, 2336, 2338, 2344, 2346, 2350, 2352, 2358, 2360, 2384, 2386, 2392,
+ 2394, 2398, 2400, 2406, 2408, 2416, 2418, 2424, 2426, 2430, 2432, 2438, 2440,
+ 2457, 2459, 2465, 2467, 2471, 2473, 2479, 2481, 2489, 2491, 2497, 2499, 2503,
+ 2505, 2511, 2513, 2533, 2535, 2541, 2543, 2547, 2549, 2555, 2557, 2565, 2567,
+ 2573, 2575, 2579, 2581, 2587, 2589, 2606, 2608, 2614, 2616, 2620, 2622, 2628,
+ 2630, 2638, 2640, 2646, 2648, 2652, 2654, 2660, 2662
};
const int16_t vp10_cat6_high_cost[128] = {
- 72, 892, 1183, 2003, 1448, 2268, 2559, 3379,
- 1709, 2529, 2820, 3640, 3085, 3905, 4196, 5016, 2118, 2938, 3229, 4049,
- 3494, 4314, 4605, 5425, 3755, 4575, 4866, 5686, 5131, 5951, 6242, 7062,
- 2118, 2938, 3229, 4049, 3494, 4314, 4605, 5425, 3755, 4575, 4866, 5686,
- 5131, 5951, 6242, 7062, 4164, 4984, 5275, 6095, 5540, 6360, 6651, 7471,
- 5801, 6621, 6912, 7732, 7177, 7997, 8288, 9108, 2118, 2938, 3229, 4049,
- 3494, 4314, 4605, 5425, 3755, 4575, 4866, 5686, 5131, 5951, 6242, 7062,
- 4164, 4984, 5275, 6095, 5540, 6360, 6651, 7471, 5801, 6621, 6912, 7732,
- 7177, 7997, 8288, 9108, 4164, 4984, 5275, 6095, 5540, 6360, 6651, 7471,
- 5801, 6621, 6912, 7732, 7177, 7997, 8288, 9108, 6210, 7030, 7321, 8141,
- 7586, 8406, 8697, 9517, 7847, 8667, 8958, 9778, 9223, 10043, 10334, 11154
+ 72, 892, 1183, 2003, 1448, 2268, 2559, 3379, 1709, 2529, 2820, 3640,
+ 3085, 3905, 4196, 5016, 2118, 2938, 3229, 4049, 3494, 4314, 4605, 5425,
+ 3755, 4575, 4866, 5686, 5131, 5951, 6242, 7062, 2118, 2938, 3229, 4049,
+ 3494, 4314, 4605, 5425, 3755, 4575, 4866, 5686, 5131, 5951, 6242, 7062,
+ 4164, 4984, 5275, 6095, 5540, 6360, 6651, 7471, 5801, 6621, 6912, 7732,
+ 7177, 7997, 8288, 9108, 2118, 2938, 3229, 4049, 3494, 4314, 4605, 5425,
+ 3755, 4575, 4866, 5686, 5131, 5951, 6242, 7062, 4164, 4984, 5275, 6095,
+ 5540, 6360, 6651, 7471, 5801, 6621, 6912, 7732, 7177, 7997, 8288, 9108,
+ 4164, 4984, 5275, 6095, 5540, 6360, 6651, 7471, 5801, 6621, 6912, 7732,
+ 7177, 7997, 8288, 9108, 6210, 7030, 7321, 8141, 7586, 8406, 8697, 9517,
+ 7847, 8667, 8958, 9778, 9223, 10043, 10334, 11154
};
#if CONFIG_VPX_HIGHBITDEPTH
const int16_t vp10_cat6_high10_high_cost[512] = {
- 74, 894, 1185, 2005, 1450, 2270, 2561,
- 3381, 1711, 2531, 2822, 3642, 3087, 3907, 4198, 5018, 2120, 2940, 3231,
- 4051, 3496, 4316, 4607, 5427, 3757, 4577, 4868, 5688, 5133, 5953, 6244,
- 7064, 2120, 2940, 3231, 4051, 3496, 4316, 4607, 5427, 3757, 4577, 4868,
- 5688, 5133, 5953, 6244, 7064, 4166, 4986, 5277, 6097, 5542, 6362, 6653,
- 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110, 2120, 2940, 3231,
- 4051, 3496, 4316, 4607, 5427, 3757, 4577, 4868, 5688, 5133, 5953, 6244,
- 7064, 4166, 4986, 5277, 6097, 5542, 6362, 6653, 7473, 5803, 6623, 6914,
- 7734, 7179, 7999, 8290, 9110, 4166, 4986, 5277, 6097, 5542, 6362, 6653,
- 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110, 6212, 7032, 7323,
- 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225, 10045, 10336,
- 11156, 2120, 2940, 3231, 4051, 3496, 4316, 4607, 5427, 3757, 4577, 4868,
- 5688, 5133, 5953, 6244, 7064, 4166, 4986, 5277, 6097, 5542, 6362, 6653,
- 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110, 4166, 4986, 5277,
- 6097, 5542, 6362, 6653, 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290,
- 9110, 6212, 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960,
- 9780, 9225, 10045, 10336, 11156, 4166, 4986, 5277, 6097, 5542, 6362, 6653,
- 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110, 6212, 7032, 7323,
- 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225, 10045, 10336,
- 11156, 6212, 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960,
- 9780, 9225, 10045, 10336, 11156, 8258, 9078, 9369, 10189, 9634, 10454,
- 10745, 11565, 9895, 10715, 11006, 11826, 11271, 12091, 12382, 13202, 2120,
- 2940, 3231, 4051, 3496, 4316, 4607, 5427, 3757, 4577, 4868, 5688, 5133,
- 5953, 6244, 7064, 4166, 4986, 5277, 6097, 5542, 6362, 6653, 7473, 5803,
- 6623, 6914, 7734, 7179, 7999, 8290, 9110, 4166, 4986, 5277, 6097, 5542,
- 6362, 6653, 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110, 6212,
- 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225,
- 10045, 10336, 11156, 4166, 4986, 5277, 6097, 5542, 6362, 6653, 7473, 5803,
- 6623, 6914, 7734, 7179, 7999, 8290, 9110, 6212, 7032, 7323, 8143, 7588,
- 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225, 10045, 10336, 11156, 6212,
- 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225,
- 10045, 10336, 11156, 8258, 9078, 9369, 10189, 9634, 10454, 10745, 11565,
- 9895, 10715, 11006, 11826, 11271, 12091, 12382, 13202, 4166, 4986, 5277,
- 6097, 5542, 6362, 6653, 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290,
- 9110, 6212, 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960,
- 9780, 9225, 10045, 10336, 11156, 6212, 7032, 7323, 8143, 7588, 8408, 8699,
- 9519, 7849, 8669, 8960, 9780, 9225, 10045, 10336, 11156, 8258, 9078, 9369,
- 10189, 9634, 10454, 10745, 11565, 9895, 10715, 11006, 11826, 11271, 12091,
- 12382, 13202, 6212, 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669,
- 8960, 9780, 9225, 10045, 10336, 11156, 8258, 9078, 9369, 10189, 9634, 10454,
- 10745, 11565, 9895, 10715, 11006, 11826, 11271, 12091, 12382, 13202, 8258,
- 9078, 9369, 10189, 9634, 10454, 10745, 11565, 9895, 10715, 11006, 11826,
- 11271, 12091, 12382, 13202, 10304, 11124, 11415, 12235, 11680, 12500, 12791,
- 13611, 11941, 12761, 13052, 13872, 13317, 14137, 14428, 15248,
+ 74, 894, 1185, 2005, 1450, 2270, 2561, 3381, 1711, 2531, 2822,
+ 3642, 3087, 3907, 4198, 5018, 2120, 2940, 3231, 4051, 3496, 4316,
+ 4607, 5427, 3757, 4577, 4868, 5688, 5133, 5953, 6244, 7064, 2120,
+ 2940, 3231, 4051, 3496, 4316, 4607, 5427, 3757, 4577, 4868, 5688,
+ 5133, 5953, 6244, 7064, 4166, 4986, 5277, 6097, 5542, 6362, 6653,
+ 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110, 2120, 2940,
+ 3231, 4051, 3496, 4316, 4607, 5427, 3757, 4577, 4868, 5688, 5133,
+ 5953, 6244, 7064, 4166, 4986, 5277, 6097, 5542, 6362, 6653, 7473,
+ 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110, 4166, 4986, 5277,
+ 6097, 5542, 6362, 6653, 7473, 5803, 6623, 6914, 7734, 7179, 7999,
+ 8290, 9110, 6212, 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849,
+ 8669, 8960, 9780, 9225, 10045, 10336, 11156, 2120, 2940, 3231, 4051,
+ 3496, 4316, 4607, 5427, 3757, 4577, 4868, 5688, 5133, 5953, 6244,
+ 7064, 4166, 4986, 5277, 6097, 5542, 6362, 6653, 7473, 5803, 6623,
+ 6914, 7734, 7179, 7999, 8290, 9110, 4166, 4986, 5277, 6097, 5542,
+ 6362, 6653, 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110,
+ 6212, 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960,
+ 9780, 9225, 10045, 10336, 11156, 4166, 4986, 5277, 6097, 5542, 6362,
+ 6653, 7473, 5803, 6623, 6914, 7734, 7179, 7999, 8290, 9110, 6212,
+ 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960, 9780,
+ 9225, 10045, 10336, 11156, 6212, 7032, 7323, 8143, 7588, 8408, 8699,
+ 9519, 7849, 8669, 8960, 9780, 9225, 10045, 10336, 11156, 8258, 9078,
+ 9369, 10189, 9634, 10454, 10745, 11565, 9895, 10715, 11006, 11826, 11271,
+ 12091, 12382, 13202, 2120, 2940, 3231, 4051, 3496, 4316, 4607, 5427,
+ 3757, 4577, 4868, 5688, 5133, 5953, 6244, 7064, 4166, 4986, 5277,
+ 6097, 5542, 6362, 6653, 7473, 5803, 6623, 6914, 7734, 7179, 7999,
+ 8290, 9110, 4166, 4986, 5277, 6097, 5542, 6362, 6653, 7473, 5803,
+ 6623, 6914, 7734, 7179, 7999, 8290, 9110, 6212, 7032, 7323, 8143,
+ 7588, 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225, 10045, 10336,
+ 11156, 4166, 4986, 5277, 6097, 5542, 6362, 6653, 7473, 5803, 6623,
+ 6914, 7734, 7179, 7999, 8290, 9110, 6212, 7032, 7323, 8143, 7588,
+ 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225, 10045, 10336, 11156,
+ 6212, 7032, 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960,
+ 9780, 9225, 10045, 10336, 11156, 8258, 9078, 9369, 10189, 9634, 10454,
+ 10745, 11565, 9895, 10715, 11006, 11826, 11271, 12091, 12382, 13202, 4166,
+ 4986, 5277, 6097, 5542, 6362, 6653, 7473, 5803, 6623, 6914, 7734,
+ 7179, 7999, 8290, 9110, 6212, 7032, 7323, 8143, 7588, 8408, 8699,
+ 9519, 7849, 8669, 8960, 9780, 9225, 10045, 10336, 11156, 6212, 7032,
+ 7323, 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225,
+ 10045, 10336, 11156, 8258, 9078, 9369, 10189, 9634, 10454, 10745, 11565,
+ 9895, 10715, 11006, 11826, 11271, 12091, 12382, 13202, 6212, 7032, 7323,
+ 8143, 7588, 8408, 8699, 9519, 7849, 8669, 8960, 9780, 9225, 10045,
+ 10336, 11156, 8258, 9078, 9369, 10189, 9634, 10454, 10745, 11565, 9895,
+ 10715, 11006, 11826, 11271, 12091, 12382, 13202, 8258, 9078, 9369, 10189,
+ 9634, 10454, 10745, 11565, 9895, 10715, 11006, 11826, 11271, 12091, 12382,
+ 13202, 10304, 11124, 11415, 12235, 11680, 12500, 12791, 13611, 11941, 12761,
+ 13052, 13872, 13317, 14137, 14428, 15248,
};
const int16_t vp10_cat6_high12_high_cost[2048] = {
- 76, 896, 1187, 2007, 1452, 2272, 2563,
- 3383, 1713, 2533, 2824, 3644, 3089, 3909, 4200, 5020, 2122, 2942, 3233,
- 4053, 3498, 4318, 4609, 5429, 3759, 4579, 4870, 5690, 5135, 5955, 6246,
- 7066, 2122, 2942, 3233, 4053, 3498, 4318, 4609, 5429, 3759, 4579, 4870,
- 5690, 5135, 5955, 6246, 7066, 4168, 4988, 5279, 6099, 5544, 6364, 6655,
- 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 2122, 2942, 3233,
- 4053, 3498, 4318, 4609, 5429, 3759, 4579, 4870, 5690, 5135, 5955, 6246,
- 7066, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916,
- 7736, 7181, 8001, 8292, 9112, 4168, 4988, 5279, 6099, 5544, 6364, 6655,
- 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325,
- 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338,
- 11158, 2122, 2942, 3233, 4053, 3498, 4318, 4609, 5429, 3759, 4579, 4870,
- 5690, 5135, 5955, 6246, 7066, 4168, 4988, 5279, 6099, 5544, 6364, 6655,
- 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 4168, 4988, 5279,
- 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292,
- 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
- 9782, 9227, 10047, 10338, 11158, 4168, 4988, 5279, 6099, 5544, 6364, 6655,
- 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325,
- 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338,
- 11158, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
- 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456,
- 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 2122,
- 2942, 3233, 4053, 3498, 4318, 4609, 5429, 3759, 4579, 4870, 5690, 5135,
- 5955, 6246, 7066, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805,
- 6625, 6916, 7736, 7181, 8001, 8292, 9112, 4168, 4988, 5279, 6099, 5544,
- 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214,
- 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227,
- 10047, 10338, 11158, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805,
- 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590,
- 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 6214,
- 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227,
- 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567,
- 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 4168, 4988, 5279,
- 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292,
- 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
- 9782, 9227, 10047, 10338, 11158, 6214, 7034, 7325, 8145, 7590, 8410, 8701,
- 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371,
- 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093,
- 12384, 13204, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671,
- 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456,
- 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 8260,
- 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
- 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793,
- 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 2122, 2942,
- 3233, 4053, 3498, 4318, 4609, 5429, 3759, 4579, 4870, 5690, 5135, 5955,
- 6246, 7066, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625,
- 6916, 7736, 7181, 8001, 8292, 9112, 4168, 4988, 5279, 6099, 5544, 6364,
- 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034,
- 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047,
- 10338, 11158, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625,
- 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410,
- 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034,
- 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047,
- 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897,
- 10717, 11008, 11828, 11273, 12093, 12384, 13204, 4168, 4988, 5279, 6099,
- 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112,
- 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782,
- 9227, 10047, 10338, 11158, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521,
- 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191,
- 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384,
- 13204, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
- 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456,
- 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 8260,
- 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
- 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793,
- 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 4168, 4988,
- 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001,
- 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671,
- 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034, 7325, 8145, 7590, 8410,
- 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080,
- 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273,
- 12093, 12384, 13204, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851,
- 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636,
- 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204,
- 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008,
- 11828, 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502,
- 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 6214,
- 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227,
- 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567,
- 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 8260, 9080, 9371,
- 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093,
- 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943,
- 12763, 13054, 13874, 13319, 14139, 14430, 15250, 8260, 9080, 9371, 10191,
- 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384,
+ 76, 896, 1187, 2007, 1452, 2272, 2563, 3383, 1713, 2533, 2824,
+ 3644, 3089, 3909, 4200, 5020, 2122, 2942, 3233, 4053, 3498, 4318,
+ 4609, 5429, 3759, 4579, 4870, 5690, 5135, 5955, 6246, 7066, 2122,
+ 2942, 3233, 4053, 3498, 4318, 4609, 5429, 3759, 4579, 4870, 5690,
+ 5135, 5955, 6246, 7066, 4168, 4988, 5279, 6099, 5544, 6364, 6655,
+ 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 2122, 2942,
+ 3233, 4053, 3498, 4318, 4609, 5429, 3759, 4579, 4870, 5690, 5135,
+ 5955, 6246, 7066, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475,
+ 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 4168, 4988, 5279,
+ 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001,
+ 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851,
+ 8671, 8962, 9782, 9227, 10047, 10338, 11158, 2122, 2942, 3233, 4053,
+ 3498, 4318, 4609, 5429, 3759, 4579, 4870, 5690, 5135, 5955, 6246,
+ 7066, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625,
+ 6916, 7736, 7181, 8001, 8292, 9112, 4168, 4988, 5279, 6099, 5544,
+ 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112,
+ 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
+ 9782, 9227, 10047, 10338, 11158, 4168, 4988, 5279, 6099, 5544, 6364,
+ 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214,
+ 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782,
+ 9227, 10047, 10338, 11158, 6214, 7034, 7325, 8145, 7590, 8410, 8701,
+ 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080,
+ 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273,
+ 12093, 12384, 13204, 2122, 2942, 3233, 4053, 3498, 4318, 4609, 5429,
+ 3759, 4579, 4870, 5690, 5135, 5955, 6246, 7066, 4168, 4988, 5279,
+ 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001,
+ 8292, 9112, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805,
+ 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145,
+ 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338,
+ 11158, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625,
+ 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590,
+ 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158,
+ 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
+ 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456,
+ 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 4168,
+ 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736,
+ 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701,
+ 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034,
+ 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227,
+ 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567,
+ 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 6214, 7034, 7325,
+ 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047,
+ 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897,
+ 10717, 11008, 11828, 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191,
+ 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384,
13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763,
- 13054, 13874, 13319, 14139, 14430, 15250, 10306, 11126, 11417, 12237, 11682,
- 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250,
- 12352, 13172, 13463, 14283, 13728, 14548, 14839, 15659, 13989, 14809, 15100,
- 15920, 15365, 16185, 16476, 17296, 2122, 2942, 3233, 4053, 3498, 4318, 4609,
- 5429, 3759, 4579, 4870, 5690, 5135, 5955, 6246, 7066, 4168, 4988, 5279,
- 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292,
- 9112, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916,
- 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701,
- 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 4168, 4988, 5279,
- 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292,
- 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
- 9782, 9227, 10047, 10338, 11158, 6214, 7034, 7325, 8145, 7590, 8410, 8701,
- 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371,
- 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093,
- 12384, 13204, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625,
- 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410,
- 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034,
- 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047,
- 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897,
- 10717, 11008, 11828, 11273, 12093, 12384, 13204, 6214, 7034, 7325, 8145,
- 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158,
- 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008,
- 11828, 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191, 9636, 10456,
- 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306,
- 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874,
- 13319, 14139, 14430, 15250, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475,
- 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145,
- 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158,
- 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782,
- 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747,
- 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 6214, 7034,
- 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047,
- 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897,
- 10717, 11008, 11828, 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191,
- 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384,
- 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763,
- 13054, 13874, 13319, 14139, 14430, 15250, 6214, 7034, 7325, 8145, 7590,
- 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260,
- 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
- 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191, 9636, 10456, 10747,
- 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126,
+ 13054, 13874, 13319, 14139, 14430, 15250, 2122, 2942, 3233, 4053, 3498,
+ 4318, 4609, 5429, 3759, 4579, 4870, 5690, 5135, 5955, 6246, 7066,
+ 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916,
+ 7736, 7181, 8001, 8292, 9112, 4168, 4988, 5279, 6099, 5544, 6364,
+ 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214,
+ 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782,
+ 9227, 10047, 10338, 11158, 4168, 4988, 5279, 6099, 5544, 6364, 6655,
+ 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034,
+ 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227,
+ 10047, 10338, 11158, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521,
+ 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371,
+ 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093,
+ 12384, 13204, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805,
+ 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145,
+ 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338,
+ 11158, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671,
+ 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636,
+ 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204,
+ 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
+ 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456,
+ 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 8260,
+ 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
+ 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793,
+ 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 4168, 4988,
+ 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181,
+ 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521,
+ 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034, 7325,
+ 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047,
+ 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897,
+ 10717, 11008, 11828, 11273, 12093, 12384, 13204, 6214, 7034, 7325, 8145,
+ 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338,
+ 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717,
+ 11008, 11828, 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191, 9636,
+ 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204,
+ 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054,
+ 13874, 13319, 14139, 14430, 15250, 6214, 7034, 7325, 8145, 7590, 8410,
+ 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260,
+ 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
+ 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191, 9636, 10456, 10747,
+ 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126,
11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319,
- 14139, 14430, 15250, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567,
- 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126, 11417,
+ 14139, 14430, 15250, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567,
+ 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126, 11417,
12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139,
14430, 15250, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943,
12763, 13054, 13874, 13319, 14139, 14430, 15250, 12352, 13172, 13463, 14283,
13728, 14548, 14839, 15659, 13989, 14809, 15100, 15920, 15365, 16185, 16476,
- 17296, 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916,
- 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701,
- 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034, 7325,
- 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338,
- 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717,
- 11008, 11828, 11273, 12093, 12384, 13204, 6214, 7034, 7325, 8145, 7590,
- 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260,
- 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
- 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191, 9636, 10456, 10747,
- 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126,
- 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319,
- 14139, 14430, 15250, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851,
- 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636,
- 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204,
- 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008,
- 11828, 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502,
- 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 8260,
- 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
- 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793,
- 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 10306, 11126,
- 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319,
- 14139, 14430, 15250, 12352, 13172, 13463, 14283, 13728, 14548, 14839, 15659,
- 13989, 14809, 15100, 15920, 15365, 16185, 16476, 17296, 6214, 7034, 7325,
- 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338,
- 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717,
- 11008, 11828, 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191, 9636,
- 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204,
- 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054,
- 13874, 13319, 14139, 14430, 15250, 8260, 9080, 9371, 10191, 9636, 10456,
- 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306,
+ 17296, 2122, 2942, 3233, 4053, 3498, 4318, 4609, 5429, 3759, 4579,
+ 4870, 5690, 5135, 5955, 6246, 7066, 4168, 4988, 5279, 6099, 5544,
+ 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112,
+ 4168, 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916,
+ 7736, 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410,
+ 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 4168,
+ 4988, 5279, 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736,
+ 7181, 8001, 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701,
+ 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034,
+ 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227,
+ 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567,
+ 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 4168, 4988, 5279,
+ 6099, 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001,
+ 8292, 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851,
+ 8671, 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034, 7325, 8145,
+ 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338,
+ 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717,
+ 11008, 11828, 11273, 12093, 12384, 13204, 6214, 7034, 7325, 8145, 7590,
+ 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158,
+ 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008,
+ 11828, 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191, 9636, 10456,
+ 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306,
11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874,
- 13319, 14139, 14430, 15250, 10306, 11126, 11417, 12237, 11682, 12502, 12793,
+ 13319, 14139, 14430, 15250, 4168, 4988, 5279, 6099, 5544, 6364, 6655,
+ 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292, 9112, 6214, 7034,
+ 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227,
+ 10047, 10338, 11158, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521,
+ 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371,
+ 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093,
+ 12384, 13204, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851,
+ 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191,
+ 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384,
+ 13204, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717,
+ 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682,
+ 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250,
+ 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962,
+ 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456,
+ 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 8260,
+ 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
+ 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793,
+ 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 8260, 9080,
+ 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273,
+ 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613,
+ 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 10306, 11126, 11417,
+ 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139,
+ 14430, 15250, 12352, 13172, 13463, 14283, 13728, 14548, 14839, 15659, 13989,
+ 14809, 15100, 15920, 15365, 16185, 16476, 17296, 4168, 4988, 5279, 6099,
+ 5544, 6364, 6655, 7475, 5805, 6625, 6916, 7736, 7181, 8001, 8292,
+ 9112, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671,
+ 8962, 9782, 9227, 10047, 10338, 11158, 6214, 7034, 7325, 8145, 7590,
+ 8410, 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158,
+ 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008,
+ 11828, 11273, 12093, 12384, 13204, 6214, 7034, 7325, 8145, 7590, 8410,
+ 8701, 9521, 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260,
+ 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828,
+ 11273, 12093, 12384, 13204, 8260, 9080, 9371, 10191, 9636, 10456, 10747,
+ 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126,
+ 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319,
+ 14139, 14430, 15250, 6214, 7034, 7325, 8145, 7590, 8410, 8701, 9521,
+ 7851, 8671, 8962, 9782, 9227, 10047, 10338, 11158, 8260, 9080, 9371,
+ 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093,
+ 12384, 13204, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897,
+ 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237,
+ 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430,
+ 15250, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717,
+ 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682,
+ 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250,
+ 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054,
+ 13874, 13319, 14139, 14430, 15250, 12352, 13172, 13463, 14283, 13728, 14548,
+ 14839, 15659, 13989, 14809, 15100, 15920, 15365, 16185, 16476, 17296, 6214,
+ 7034, 7325, 8145, 7590, 8410, 8701, 9521, 7851, 8671, 8962, 9782,
+ 9227, 10047, 10338, 11158, 8260, 9080, 9371, 10191, 9636, 10456, 10747,
+ 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 8260, 9080,
+ 9371, 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273,
+ 12093, 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613,
+ 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 8260, 9080, 9371,
+ 10191, 9636, 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093,
+ 12384, 13204, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943,
+ 12763, 13054, 13874, 13319, 14139, 14430, 15250, 10306, 11126, 11417, 12237,
+ 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430,
+ 15250, 12352, 13172, 13463, 14283, 13728, 14548, 14839, 15659, 13989, 14809,
+ 15100, 15920, 15365, 16185, 16476, 17296, 8260, 9080, 9371, 10191, 9636,
+ 10456, 10747, 11567, 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204,
+ 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054,
+ 13874, 13319, 14139, 14430, 15250, 10306, 11126, 11417, 12237, 11682, 12502,
+ 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 12352,
+ 13172, 13463, 14283, 13728, 14548, 14839, 15659, 13989, 14809, 15100, 15920,
+ 15365, 16185, 16476, 17296, 10306, 11126, 11417, 12237, 11682, 12502, 12793,
13613, 11943, 12763, 13054, 13874, 13319, 14139, 14430, 15250, 12352, 13172,
13463, 14283, 13728, 14548, 14839, 15659, 13989, 14809, 15100, 15920, 15365,
- 16185, 16476, 17296, 8260, 9080, 9371, 10191, 9636, 10456, 10747, 11567,
- 9897, 10717, 11008, 11828, 11273, 12093, 12384, 13204, 10306, 11126, 11417,
- 12237, 11682, 12502, 12793, 13613, 11943, 12763, 13054, 13874, 13319, 14139,
- 14430, 15250, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943,
- 12763, 13054, 13874, 13319, 14139, 14430, 15250, 12352, 13172, 13463, 14283,
- 13728, 14548, 14839, 15659, 13989, 14809, 15100, 15920, 15365, 16185, 16476,
- 17296, 10306, 11126, 11417, 12237, 11682, 12502, 12793, 13613, 11943, 12763,
- 13054, 13874, 13319, 14139, 14430, 15250, 12352, 13172, 13463, 14283, 13728,
- 14548, 14839, 15659, 13989, 14809, 15100, 15920, 15365, 16185, 16476, 17296,
- 12352, 13172, 13463, 14283, 13728, 14548, 14839, 15659, 13989, 14809, 15100,
- 15920, 15365, 16185, 16476, 17296, 14398, 15218, 15509, 16329, 15774, 16594,
- 16885, 17705, 16035, 16855, 17146, 17966, 17411, 18231, 18522, 19342
+ 16185, 16476, 17296, 12352, 13172, 13463, 14283, 13728, 14548, 14839, 15659,
+ 13989, 14809, 15100, 15920, 15365, 16185, 16476, 17296, 14398, 15218, 15509,
+ 16329, 15774, 16594, 16885, 17705, 16035, 16855, 17146, 17966, 17411, 18231,
+ 18522, 19342
};
#endif
#if CONFIG_VPX_HIGHBITDEPTH
-static const vpx_tree_index cat1_high10[2] = {0, 0};
-static const vpx_tree_index cat2_high10[4] = {2, 2, 0, 0};
-static const vpx_tree_index cat3_high10[6] = {2, 2, 4, 4, 0, 0};
-static const vpx_tree_index cat4_high10[8] = {2, 2, 4, 4, 6, 6, 0, 0};
-static const vpx_tree_index cat5_high10[10] = {2, 2, 4, 4, 6, 6, 8, 8, 0, 0};
-static const vpx_tree_index cat6_high10[32] = {2, 2, 4, 4, 6, 6, 8, 8, 10, 10,
- 12, 12, 14, 14, 16, 16, 18, 18, 20, 20, 22, 22, 24, 24, 26, 26, 28, 28,
- 30, 30, 0, 0};
-static const vpx_tree_index cat1_high12[2] = {0, 0};
-static const vpx_tree_index cat2_high12[4] = {2, 2, 0, 0};
-static const vpx_tree_index cat3_high12[6] = {2, 2, 4, 4, 0, 0};
-static const vpx_tree_index cat4_high12[8] = {2, 2, 4, 4, 6, 6, 0, 0};
-static const vpx_tree_index cat5_high12[10] = {2, 2, 4, 4, 6, 6, 8, 8, 0, 0};
-static const vpx_tree_index cat6_high12[36] = {2, 2, 4, 4, 6, 6, 8, 8, 10, 10,
- 12, 12, 14, 14, 16, 16, 18, 18, 20, 20, 22, 22, 24, 24, 26, 26, 28, 28,
- 30, 30, 32, 32, 34, 34, 0, 0};
+static const vpx_tree_index cat1_high10[2] = { 0, 0 };
+static const vpx_tree_index cat2_high10[4] = { 2, 2, 0, 0 };
+static const vpx_tree_index cat3_high10[6] = { 2, 2, 4, 4, 0, 0 };
+static const vpx_tree_index cat4_high10[8] = { 2, 2, 4, 4, 6, 6, 0, 0 };
+static const vpx_tree_index cat5_high10[10] = { 2, 2, 4, 4, 6, 6, 8, 8, 0, 0 };
+static const vpx_tree_index cat6_high10[32] = { 2, 2, 4, 4, 6, 6, 8, 8,
+ 10, 10, 12, 12, 14, 14, 16, 16,
+ 18, 18, 20, 20, 22, 22, 24, 24,
+ 26, 26, 28, 28, 30, 30, 0, 0 };
+static const vpx_tree_index cat1_high12[2] = { 0, 0 };
+static const vpx_tree_index cat2_high12[4] = { 2, 2, 0, 0 };
+static const vpx_tree_index cat3_high12[6] = { 2, 2, 4, 4, 0, 0 };
+static const vpx_tree_index cat4_high12[8] = { 2, 2, 4, 4, 6, 6, 0, 0 };
+static const vpx_tree_index cat5_high12[10] = { 2, 2, 4, 4, 6, 6, 8, 8, 0, 0 };
+static const vpx_tree_index cat6_high12[36] = {
+ 2, 2, 4, 4, 6, 6, 8, 8, 10, 10, 12, 12, 14, 14, 16, 16, 18, 18,
+ 20, 20, 22, 22, 24, 24, 26, 26, 28, 28, 30, 30, 32, 32, 34, 34, 0, 0
+};
#endif
const vp10_extra_bit vp10_extra_bits[ENTROPY_TOKENS] = {
- {0, 0, 0, 0, zero_cost}, // ZERO_TOKEN
- {0, 0, 0, 1, sign_cost}, // ONE_TOKEN
- {0, 0, 0, 2, sign_cost}, // TWO_TOKEN
- {0, 0, 0, 3, sign_cost}, // THREE_TOKEN
- {0, 0, 0, 4, sign_cost}, // FOUR_TOKEN
- {cat1, vp10_cat1_prob, 1, CAT1_MIN_VAL, cat1_cost}, // CATEGORY1_TOKEN
- {cat2, vp10_cat2_prob, 2, CAT2_MIN_VAL, cat2_cost}, // CATEGORY2_TOKEN
- {cat3, vp10_cat3_prob, 3, CAT3_MIN_VAL, cat3_cost}, // CATEGORY3_TOKEN
- {cat4, vp10_cat4_prob, 4, CAT4_MIN_VAL, cat4_cost}, // CATEGORY4_TOKEN
- {cat5, vp10_cat5_prob, 5, CAT5_MIN_VAL, cat5_cost}, // CATEGORY5_TOKEN
- {cat6, vp10_cat6_prob, 14, CAT6_MIN_VAL, 0}, // CATEGORY6_TOKEN
- {0, 0, 0, 0, zero_cost} // EOB_TOKEN
+ { 0, 0, 0, 0, zero_cost }, // ZERO_TOKEN
+ { 0, 0, 0, 1, sign_cost }, // ONE_TOKEN
+ { 0, 0, 0, 2, sign_cost }, // TWO_TOKEN
+ { 0, 0, 0, 3, sign_cost }, // THREE_TOKEN
+ { 0, 0, 0, 4, sign_cost }, // FOUR_TOKEN
+ { cat1, vp10_cat1_prob, 1, CAT1_MIN_VAL, cat1_cost }, // CATEGORY1_TOKEN
+ { cat2, vp10_cat2_prob, 2, CAT2_MIN_VAL, cat2_cost }, // CATEGORY2_TOKEN
+ { cat3, vp10_cat3_prob, 3, CAT3_MIN_VAL, cat3_cost }, // CATEGORY3_TOKEN
+ { cat4, vp10_cat4_prob, 4, CAT4_MIN_VAL, cat4_cost }, // CATEGORY4_TOKEN
+ { cat5, vp10_cat5_prob, 5, CAT5_MIN_VAL, cat5_cost }, // CATEGORY5_TOKEN
+ { cat6, vp10_cat6_prob, 14, CAT6_MIN_VAL, 0 }, // CATEGORY6_TOKEN
+ { 0, 0, 0, 0, zero_cost } // EOB_TOKEN
};
#if CONFIG_VPX_HIGHBITDEPTH
const vp10_extra_bit vp10_extra_bits_high10[ENTROPY_TOKENS] = {
- {0, 0, 0, 0, zero_cost}, // ZERO
- {0, 0, 0, 1, sign_cost}, // ONE
- {0, 0, 0, 2, sign_cost}, // TWO
- {0, 0, 0, 3, sign_cost}, // THREE
- {0, 0, 0, 4, sign_cost}, // FOUR
- {cat1_high10, vp10_cat1_prob_high10, 1, CAT1_MIN_VAL, cat1_cost}, // CAT1
- {cat2_high10, vp10_cat2_prob_high10, 2, CAT2_MIN_VAL, cat2_cost}, // CAT2
- {cat3_high10, vp10_cat3_prob_high10, 3, CAT3_MIN_VAL, cat3_cost}, // CAT3
- {cat4_high10, vp10_cat4_prob_high10, 4, CAT4_MIN_VAL, cat4_cost}, // CAT4
- {cat5_high10, vp10_cat5_prob_high10, 5, CAT5_MIN_VAL, cat5_cost}, // CAT5
- {cat6_high10, vp10_cat6_prob_high10, 16, CAT6_MIN_VAL, 0}, // CAT6
- {0, 0, 0, 0, zero_cost} // EOB
+ { 0, 0, 0, 0, zero_cost }, // ZERO
+ { 0, 0, 0, 1, sign_cost }, // ONE
+ { 0, 0, 0, 2, sign_cost }, // TWO
+ { 0, 0, 0, 3, sign_cost }, // THREE
+ { 0, 0, 0, 4, sign_cost }, // FOUR
+ { cat1_high10, vp10_cat1_prob_high10, 1, CAT1_MIN_VAL, cat1_cost }, // CAT1
+ { cat2_high10, vp10_cat2_prob_high10, 2, CAT2_MIN_VAL, cat2_cost }, // CAT2
+ { cat3_high10, vp10_cat3_prob_high10, 3, CAT3_MIN_VAL, cat3_cost }, // CAT3
+ { cat4_high10, vp10_cat4_prob_high10, 4, CAT4_MIN_VAL, cat4_cost }, // CAT4
+ { cat5_high10, vp10_cat5_prob_high10, 5, CAT5_MIN_VAL, cat5_cost }, // CAT5
+ { cat6_high10, vp10_cat6_prob_high10, 16, CAT6_MIN_VAL, 0 }, // CAT6
+ { 0, 0, 0, 0, zero_cost } // EOB
};
const vp10_extra_bit vp10_extra_bits_high12[ENTROPY_TOKENS] = {
- {0, 0, 0, 0, zero_cost}, // ZERO
- {0, 0, 0, 1, sign_cost}, // ONE
- {0, 0, 0, 2, sign_cost}, // TWO
- {0, 0, 0, 3, sign_cost}, // THREE
- {0, 0, 0, 4, sign_cost}, // FOUR
- {cat1_high12, vp10_cat1_prob_high12, 1, CAT1_MIN_VAL, cat1_cost}, // CAT1
- {cat2_high12, vp10_cat2_prob_high12, 2, CAT2_MIN_VAL, cat2_cost}, // CAT2
- {cat3_high12, vp10_cat3_prob_high12, 3, CAT3_MIN_VAL, cat3_cost}, // CAT3
- {cat4_high12, vp10_cat4_prob_high12, 4, CAT4_MIN_VAL, cat4_cost}, // CAT4
- {cat5_high12, vp10_cat5_prob_high12, 5, CAT5_MIN_VAL, cat5_cost}, // CAT5
- {cat6_high12, vp10_cat6_prob_high12, 18, CAT6_MIN_VAL, 0}, // CAT6
- {0, 0, 0, 0, zero_cost} // EOB
+ { 0, 0, 0, 0, zero_cost }, // ZERO
+ { 0, 0, 0, 1, sign_cost }, // ONE
+ { 0, 0, 0, 2, sign_cost }, // TWO
+ { 0, 0, 0, 3, sign_cost }, // THREE
+ { 0, 0, 0, 4, sign_cost }, // FOUR
+ { cat1_high12, vp10_cat1_prob_high12, 1, CAT1_MIN_VAL, cat1_cost }, // CAT1
+ { cat2_high12, vp10_cat2_prob_high12, 2, CAT2_MIN_VAL, cat2_cost }, // CAT2
+ { cat3_high12, vp10_cat3_prob_high12, 3, CAT3_MIN_VAL, cat3_cost }, // CAT3
+ { cat4_high12, vp10_cat4_prob_high12, 4, CAT4_MIN_VAL, cat4_cost }, // CAT4
+ { cat5_high12, vp10_cat5_prob_high12, 5, CAT5_MIN_VAL, cat5_cost }, // CAT5
+ { cat6_high12, vp10_cat6_prob_high12, 18, CAT6_MIN_VAL, 0 }, // CAT6
+ { 0, 0, 0, 0, zero_cost } // EOB
};
#endif
const struct vp10_token vp10_coef_encodings[ENTROPY_TOKENS] = {
- {2, 2}, {6, 3}, {28, 5}, {58, 6}, {59, 6}, {60, 6}, {61, 6}, {124, 7},
- {125, 7}, {126, 7}, {127, 7}, {0, 1}
+ { 2, 2 }, { 6, 3 }, { 28, 5 }, { 58, 6 }, { 59, 6 }, { 60, 6 },
+ { 61, 6 }, { 124, 7 }, { 125, 7 }, { 126, 7 }, { 127, 7 }, { 0, 1 }
};
-
struct tokenize_b_args {
VP10_COMP *cpi;
ThreadData *td;
TOKENEXTRA **tp;
};
-static void set_entropy_context_b(int plane, int block,
- int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize,
+static void set_entropy_context_b(int plane, int block, int blk_row,
+ int blk_col, BLOCK_SIZE plane_bsize,
TX_SIZE tx_size, void *arg) {
- struct tokenize_b_args* const args = arg;
+ struct tokenize_b_args *const args = arg;
ThreadData *const td = args->td;
MACROBLOCK *const x = &td->mb;
MACROBLOCKD *const xd = &x->e_mbd;
struct macroblock_plane *p = &x->plane[plane];
struct macroblockd_plane *pd = &xd->plane[plane];
- vp10_set_contexts(xd, pd, plane_bsize, tx_size, p->eobs[block] > 0,
- blk_col, blk_row);
+ vp10_set_contexts(xd, pd, plane_bsize, tx_size, p->eobs[block] > 0, blk_col,
+ blk_row);
}
static INLINE void add_token(TOKENEXTRA **t, const vpx_prob *context_tree,
int32_t extra, uint8_t token,
- uint8_t skip_eob_node,
- unsigned int *counts) {
+ uint8_t skip_eob_node, unsigned int *counts) {
(*t)->token = token;
(*t)->extra = extra;
(*t)->context_tree = context_tree;
@@ -468,8 +485,7 @@
static INLINE void add_token_no_extra(TOKENEXTRA **t,
const vpx_prob *context_tree,
- uint8_t token,
- uint8_t skip_eob_node,
+ uint8_t token, uint8_t skip_eob_node,
unsigned int *counts) {
(*t)->token = token;
(*t)->context_tree = context_tree;
@@ -485,9 +501,8 @@
}
static void tokenize_b(int plane, int block, int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize,
- TX_SIZE tx_size, void *arg) {
- struct tokenize_b_args* const args = arg;
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size, void *arg) {
+ struct tokenize_b_args *const args = arg;
VP10_COMP *cpi = args->cpi;
ThreadData *const td = args->td;
MACROBLOCK *const x = &td->mb;
@@ -499,7 +514,7 @@
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
int pt; /* near block/prev token context index */
int c;
- TOKENEXTRA *t = *tp; /* store tokens starting here */
+ TOKENEXTRA *t = *tp; /* store tokens starting here */
int eob = p->eobs[block];
const PLANE_TYPE type = pd->plane_type;
const tran_low_t *qcoeff = BLOCK_OFFSET(p->qcoeff, block);
@@ -567,8 +582,7 @@
int *skippable;
};
static void is_skippable(int plane, int block, int blk_row, int blk_col,
- BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
- void *argv) {
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size, void *argv) {
struct is_skippable_args *args = argv;
(void)plane;
(void)plane_bsize;
@@ -582,9 +596,9 @@
// vp10_foreach_transform_block() and simplify is_skippable().
int vp10_is_skippable_in_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
int result = 1;
- struct is_skippable_args args = {x->plane[plane].eobs, &result};
+ struct is_skippable_args args = { x->plane[plane].eobs, &result };
vp10_foreach_transformed_block_in_plane(&x->e_mbd, bsize, plane, is_skippable,
- &args);
+ &args);
return result;
}
@@ -593,35 +607,34 @@
void *argv) {
struct is_skippable_args *args = argv;
int eobs = (tx_size == TX_4X4) ? 3 : 10;
- (void) plane;
- (void) plane_bsize;
- (void) blk_row;
- (void) blk_col;
+ (void)plane;
+ (void)plane_bsize;
+ (void)blk_row;
+ (void)blk_col;
*(args->skippable) |= (args->eobs[block] > eobs);
}
int vp10_has_high_freq_in_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
int result = 0;
- struct is_skippable_args args = {x->plane[plane].eobs, &result};
+ struct is_skippable_args args = { x->plane[plane].eobs, &result };
vp10_foreach_transformed_block_in_plane(&x->e_mbd, bsize, plane,
- has_high_freq_coeff, &args);
+ has_high_freq_coeff, &args);
return result;
}
void vp10_tokenize_sb(VP10_COMP *cpi, ThreadData *td, TOKENEXTRA **t,
- int dry_run, BLOCK_SIZE bsize) {
+ int dry_run, BLOCK_SIZE bsize) {
VP10_COMMON *const cm = &cpi->common;
MACROBLOCK *const x = &td->mb;
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
const int ctx = vp10_get_skip_context(xd);
- const int skip_inc = !segfeature_active(&cm->seg, mbmi->segment_id,
- SEG_LVL_SKIP);
- struct tokenize_b_args arg = {cpi, td, t};
+ const int skip_inc =
+ !segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP);
+ struct tokenize_b_args arg = { cpi, td, t };
if (mbmi->skip) {
- if (!dry_run)
- td->counts->skip[ctx][1] += skip_inc;
+ if (!dry_run) td->counts->skip[ctx][1] += skip_inc;
reset_skip_context(xd, bsize);
return;
}
diff --git a/vp10/encoder/tokenize.h b/vp10/encoder/tokenize.h
index 536a55c..0a02048 100644
--- a/vp10/encoder/tokenize.h
+++ b/vp10/encoder/tokenize.h
@@ -20,15 +20,14 @@
extern "C" {
#endif
-#define EOSB_TOKEN 127 // Not signalled, encoder only
+#define EOSB_TOKEN 127 // Not signalled, encoder only
#if CONFIG_VPX_HIGHBITDEPTH
- typedef int32_t EXTRABIT;
+typedef int32_t EXTRABIT;
#else
- typedef int16_t EXTRABIT;
+typedef int16_t EXTRABIT;
#endif
-
typedef struct {
int16_t token;
EXTRABIT extra;
@@ -52,7 +51,7 @@
struct ThreadData;
void vp10_tokenize_sb(struct VP10_COMP *cpi, struct ThreadData *td,
- TOKENEXTRA **t, int dry_run, BLOCK_SIZE bsize);
+ TOKENEXTRA **t, int dry_run, BLOCK_SIZE bsize);
extern const int16_t *vp10_dct_value_cost_ptr;
/* TODO: The Token field should be broken out into a separate char array to
@@ -66,27 +65,26 @@
extern const int16_t vp10_cat6_high10_high_cost[512];
extern const int16_t vp10_cat6_high12_high_cost[2048];
static INLINE int16_t vp10_get_cost(int16_t token, EXTRABIT extrabits,
- const int16_t *cat6_high_table) {
- if (token != CATEGORY6_TOKEN)
- return vp10_extra_bits[token].cost[extrabits];
- return vp10_cat6_low_cost[extrabits & 0xff]
- + cat6_high_table[extrabits >> 8];
+ const int16_t *cat6_high_table) {
+ if (token != CATEGORY6_TOKEN) return vp10_extra_bits[token].cost[extrabits];
+ return vp10_cat6_low_cost[extrabits & 0xff] + cat6_high_table[extrabits >> 8];
}
#if CONFIG_VPX_HIGHBITDEPTH
-static INLINE const int16_t* vp10_get_high_cost_table(int bit_depth) {
+static INLINE const int16_t *vp10_get_high_cost_table(int bit_depth) {
return bit_depth == 8 ? vp10_cat6_high_cost
- : (bit_depth == 10 ? vp10_cat6_high10_high_cost :
- vp10_cat6_high12_high_cost);
+ : (bit_depth == 10 ? vp10_cat6_high10_high_cost
+ : vp10_cat6_high12_high_cost);
}
#else
-static INLINE const int16_t* vp10_get_high_cost_table(int bit_depth) {
- (void) bit_depth;
+static INLINE const int16_t *vp10_get_high_cost_table(int bit_depth) {
+ (void)bit_depth;
return vp10_cat6_high_cost;
}
#endif // CONFIG_VPX_HIGHBITDEPTH
-static INLINE void vp10_get_token_extra(int v, int16_t *token, EXTRABIT *extra) {
+static INLINE void vp10_get_token_extra(int v, int16_t *token,
+ EXTRABIT *extra) {
if (v >= CAT6_MIN_VAL || v <= -CAT6_MIN_VAL) {
*token = CATEGORY6_TOKEN;
if (v >= CAT6_MIN_VAL)
@@ -99,12 +97,10 @@
*extra = vp10_dct_cat_lt_10_value_tokens[v].extra;
}
static INLINE int16_t vp10_get_token(int v) {
- if (v >= CAT6_MIN_VAL || v <= -CAT6_MIN_VAL)
- return 10;
+ if (v >= CAT6_MIN_VAL || v <= -CAT6_MIN_VAL) return 10;
return vp10_dct_cat_lt_10_value_tokens[v].token;
}
-
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/vp10/encoder/treewriter.c b/vp10/encoder/treewriter.c
index 1f42f32..152bf40 100644
--- a/vp10/encoder/treewriter.c
+++ b/vp10/encoder/treewriter.c
@@ -27,7 +27,7 @@
}
void vp10_tokens_from_tree(struct vp10_token *tokens,
- const vpx_tree_index *tree) {
+ const vpx_tree_index *tree) {
tree2tok(tokens, tree, 0, 0, 0);
}
@@ -52,7 +52,7 @@
}
void vp10_tree_probs_from_distribution(vpx_tree tree,
- unsigned int branch_ct[/* n-1 */][2],
- const unsigned int num_events[/* n */]) {
+ unsigned int branch_ct[/* n-1 */][2],
+ const unsigned int num_events[/* n */]) {
convert_distribution(0, tree, branch_ct, num_events);
}
diff --git a/vp10/encoder/treewriter.h b/vp10/encoder/treewriter.h
index 6b76a03..05de002 100644
--- a/vp10/encoder/treewriter.h
+++ b/vp10/encoder/treewriter.h
@@ -18,19 +18,19 @@
#endif
void vp10_tree_probs_from_distribution(vpx_tree tree,
- unsigned int branch_ct[ /* n - 1 */ ][2],
- const unsigned int num_events[ /* n */ ]);
+ unsigned int branch_ct[/* n - 1 */][2],
+ const unsigned int num_events[/* n */]);
struct vp10_token {
int value;
int len;
};
-void vp10_tokens_from_tree(struct vp10_token*, const vpx_tree_index *);
+void vp10_tokens_from_tree(struct vp10_token *, const vpx_tree_index *);
static INLINE void vp10_write_tree(vpx_writer *w, const vpx_tree_index *tree,
- const vpx_prob *probs, int bits, int len,
- vpx_tree_index i) {
+ const vpx_prob *probs, int bits, int len,
+ vpx_tree_index i) {
do {
const int bit = (bits >> --len) & 1;
vpx_write(w, bit, probs[i >> 1]);
@@ -39,8 +39,8 @@
}
static INLINE void vp10_write_token(vpx_writer *w, const vpx_tree_index *tree,
- const vpx_prob *probs,
- const struct vp10_token *token) {
+ const vpx_prob *probs,
+ const struct vp10_token *token) {
vp10_write_tree(w, tree, probs, token->value, token->len, 0);
}
diff --git a/vp10/encoder/x86/dct_sse2.c b/vp10/encoder/x86/dct_sse2.c
index e111157..7a61ada 100644
--- a/vp10/encoder/x86/dct_sse2.c
+++ b/vp10/encoder/x86/dct_sse2.c
@@ -78,8 +78,8 @@
const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
__m128i u[4], v[4];
- u[0]=_mm_unpacklo_epi16(in[0], in[1]);
- u[1]=_mm_unpacklo_epi16(in[3], in[2]);
+ u[0] = _mm_unpacklo_epi16(in[0], in[1]);
+ u[1] = _mm_unpacklo_epi16(in[3], in[2]);
v[0] = _mm_add_epi16(u[0], u[1]);
v[1] = _mm_sub_epi16(u[0], u[1]);
@@ -151,14 +151,12 @@
transpose_4x4(in);
}
-void vp10_fht4x4_sse2(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp10_fht4x4_sse2(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
__m128i in[4];
switch (tx_type) {
- case DCT_DCT:
- vpx_fdct4x4_sse2(input, output, stride);
- break;
+ case DCT_DCT: vpx_fdct4x4_sse2(input, output, stride); break;
case ADST_DCT:
load_buffer_4x4(input, in, stride);
fadst4_sse2(in);
@@ -177,21 +175,16 @@
fadst4_sse2(in);
write_buffer_4x4(output, in);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
-void vp10_fdct8x8_quant_sse2(const int16_t *input, int stride,
- int16_t* coeff_ptr, intptr_t n_coeffs,
- int skip_block, const int16_t* zbin_ptr,
- const int16_t* round_ptr, const int16_t* quant_ptr,
- const int16_t* quant_shift_ptr, int16_t* qcoeff_ptr,
- int16_t* dqcoeff_ptr, const int16_t* dequant_ptr,
- uint16_t* eob_ptr,
- const int16_t* scan_ptr,
- const int16_t* iscan_ptr) {
+void vp10_fdct8x8_quant_sse2(
+ const int16_t *input, int stride, int16_t *coeff_ptr, intptr_t n_coeffs,
+ int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr,
+ const int16_t *quant_ptr, const int16_t *quant_shift_ptr,
+ int16_t *qcoeff_ptr, int16_t *dqcoeff_ptr, const int16_t *dequant_ptr,
+ uint16_t *eob_ptr, const int16_t *scan_ptr, const int16_t *iscan_ptr) {
__m128i zero;
int pass;
// Constants
@@ -208,14 +201,14 @@
const __m128i k__cospi_m20_p12 = pair_set_epi16(-cospi_20_64, cospi_12_64);
const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
// Load input
- __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
- __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
- __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
- __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
- __m128i in4 = _mm_load_si128((const __m128i *)(input + 4 * stride));
- __m128i in5 = _mm_load_si128((const __m128i *)(input + 5 * stride));
- __m128i in6 = _mm_load_si128((const __m128i *)(input + 6 * stride));
- __m128i in7 = _mm_load_si128((const __m128i *)(input + 7 * stride));
+ __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
+ __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
+ __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
+ __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
+ __m128i in4 = _mm_load_si128((const __m128i *)(input + 4 * stride));
+ __m128i in5 = _mm_load_si128((const __m128i *)(input + 5 * stride));
+ __m128i in6 = _mm_load_si128((const __m128i *)(input + 6 * stride));
+ __m128i in7 = _mm_load_si128((const __m128i *)(input + 7 * stride));
__m128i *in[8];
int index = 0;
@@ -469,9 +462,9 @@
// Setup global values
{
- round = _mm_load_si128((const __m128i*)round_ptr);
- quant = _mm_load_si128((const __m128i*)quant_ptr);
- dequant = _mm_load_si128((const __m128i*)dequant_ptr);
+ round = _mm_load_si128((const __m128i *)round_ptr);
+ quant = _mm_load_si128((const __m128i *)quant_ptr);
+ dequant = _mm_load_si128((const __m128i *)dequant_ptr);
}
{
@@ -503,15 +496,15 @@
qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
- _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), qcoeff0);
- _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, qcoeff1);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs), qcoeff0);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs) + 1, qcoeff1);
coeff0 = _mm_mullo_epi16(qcoeff0, dequant);
dequant = _mm_unpackhi_epi64(dequant, dequant);
coeff1 = _mm_mullo_epi16(qcoeff1, dequant);
- _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), coeff0);
- _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, coeff1);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs), coeff0);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs) + 1, coeff1);
}
{
@@ -524,8 +517,8 @@
zero_coeff1 = _mm_cmpeq_epi16(coeff1, zero);
nzero_coeff0 = _mm_cmpeq_epi16(zero_coeff0, zero);
nzero_coeff1 = _mm_cmpeq_epi16(zero_coeff1, zero);
- iscan0 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs));
- iscan1 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs) + 1);
+ iscan0 = _mm_load_si128((const __m128i *)(iscan_ptr + n_coeffs));
+ iscan1 = _mm_load_si128((const __m128i *)(iscan_ptr + n_coeffs) + 1);
// Add one to convert from indices to counts
iscan0 = _mm_sub_epi16(iscan0, nzero_coeff0);
iscan1 = _mm_sub_epi16(iscan1, nzero_coeff1);
@@ -568,14 +561,14 @@
qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
- _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), qcoeff0);
- _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, qcoeff1);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs), qcoeff0);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs) + 1, qcoeff1);
coeff0 = _mm_mullo_epi16(qcoeff0, dequant);
coeff1 = _mm_mullo_epi16(qcoeff1, dequant);
- _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), coeff0);
- _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, coeff1);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs), coeff0);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs) + 1, coeff1);
}
{
@@ -588,8 +581,8 @@
zero_coeff1 = _mm_cmpeq_epi16(coeff1, zero);
nzero_coeff0 = _mm_cmpeq_epi16(zero_coeff0, zero);
nzero_coeff1 = _mm_cmpeq_epi16(zero_coeff1, zero);
- iscan0 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs));
- iscan1 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs) + 1);
+ iscan0 = _mm_load_si128((const __m128i *)(iscan_ptr + n_coeffs));
+ iscan1 = _mm_load_si128((const __m128i *)(iscan_ptr + n_coeffs) + 1);
// Add one to convert from indices to counts
iscan0 = _mm_sub_epi16(iscan0, nzero_coeff0);
iscan1 = _mm_sub_epi16(iscan1, nzero_coeff1);
@@ -615,10 +608,10 @@
}
} else {
do {
- _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), zero);
- _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, zero);
- _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), zero);
- _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, zero);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs), zero);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs) + 1, zero);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs), zero);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs) + 1, zero);
n_coeffs += 8 * 2;
} while (n_coeffs < 0);
*eob_ptr = 0;
@@ -628,14 +621,14 @@
// load 8x8 array
static INLINE void load_buffer_8x8(const int16_t *input, __m128i *in,
int stride) {
- in[0] = _mm_load_si128((const __m128i *)(input + 0 * stride));
- in[1] = _mm_load_si128((const __m128i *)(input + 1 * stride));
- in[2] = _mm_load_si128((const __m128i *)(input + 2 * stride));
- in[3] = _mm_load_si128((const __m128i *)(input + 3 * stride));
- in[4] = _mm_load_si128((const __m128i *)(input + 4 * stride));
- in[5] = _mm_load_si128((const __m128i *)(input + 5 * stride));
- in[6] = _mm_load_si128((const __m128i *)(input + 6 * stride));
- in[7] = _mm_load_si128((const __m128i *)(input + 7 * stride));
+ in[0] = _mm_load_si128((const __m128i *)(input + 0 * stride));
+ in[1] = _mm_load_si128((const __m128i *)(input + 1 * stride));
+ in[2] = _mm_load_si128((const __m128i *)(input + 2 * stride));
+ in[3] = _mm_load_si128((const __m128i *)(input + 3 * stride));
+ in[4] = _mm_load_si128((const __m128i *)(input + 4 * stride));
+ in[5] = _mm_load_si128((const __m128i *)(input + 5 * stride));
+ in[6] = _mm_load_si128((const __m128i *)(input + 6 * stride));
+ in[7] = _mm_load_si128((const __m128i *)(input + 7 * stride));
in[0] = _mm_slli_epi16(in[0], 2);
in[1] = _mm_slli_epi16(in[1], 2);
@@ -930,14 +923,14 @@
__m128i in0, in1, in2, in3, in4, in5, in6, in7;
// properly aligned for butterfly input
- in0 = in[7];
- in1 = in[0];
- in2 = in[5];
- in3 = in[2];
- in4 = in[3];
- in5 = in[4];
- in6 = in[1];
- in7 = in[6];
+ in0 = in[7];
+ in1 = in[0];
+ in2 = in[5];
+ in3 = in[2];
+ in4 = in[3];
+ in5 = in[4];
+ in6 = in[1];
+ in7 = in[6];
// column transformation
// stage 1
@@ -1135,14 +1128,12 @@
array_transpose_8x8(in, in);
}
-void vp10_fht8x8_sse2(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp10_fht8x8_sse2(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
__m128i in[8];
switch (tx_type) {
- case DCT_DCT:
- vpx_fdct8x8_sse2(input, output, stride);
- break;
+ case DCT_DCT: vpx_fdct8x8_sse2(input, output, stride); break;
case ADST_DCT:
load_buffer_8x8(input, in, stride);
fadst8_sse2(in);
@@ -1164,13 +1155,11 @@
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
-static INLINE void load_buffer_16x16(const int16_t* input, __m128i *in0,
+static INLINE void load_buffer_16x16(const int16_t *input, __m128i *in0,
__m128i *in1, int stride) {
// load first 8 columns
load_buffer_8x8(input, in0, stride);
@@ -1530,13 +1519,13 @@
v[14] = _mm_srai_epi32(u[14], DCT_CONST_BITS);
v[15] = _mm_srai_epi32(u[15], DCT_CONST_BITS);
- in[1] = _mm_packs_epi32(v[0], v[1]);
- in[9] = _mm_packs_epi32(v[2], v[3]);
- in[5] = _mm_packs_epi32(v[4], v[5]);
+ in[1] = _mm_packs_epi32(v[0], v[1]);
+ in[9] = _mm_packs_epi32(v[2], v[3]);
+ in[5] = _mm_packs_epi32(v[4], v[5]);
in[13] = _mm_packs_epi32(v[6], v[7]);
- in[3] = _mm_packs_epi32(v[8], v[9]);
+ in[3] = _mm_packs_epi32(v[8], v[9]);
in[11] = _mm_packs_epi32(v[10], v[11]);
- in[7] = _mm_packs_epi32(v[12], v[13]);
+ in[7] = _mm_packs_epi32(v[12], v[13]);
in[15] = _mm_packs_epi32(v[14], v[15]);
}
@@ -2022,14 +2011,12 @@
array_transpose_16x16(in0, in1);
}
-void vp10_fht16x16_sse2(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp10_fht16x16_sse2(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
__m128i in0[16], in1[16];
switch (tx_type) {
- case DCT_DCT:
- vpx_fdct16x16_sse2(input, output, stride);
- break;
+ case DCT_DCT: vpx_fdct16x16_sse2(input, output, stride); break;
case ADST_DCT:
load_buffer_16x16(input, in0, in1, stride);
fadst16_sse2(in0, in1);
@@ -2051,8 +2038,6 @@
fadst16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
- default:
- assert(0);
- break;
+ default: assert(0); break;
}
}
diff --git a/vp10/encoder/x86/dct_ssse3.c b/vp10/encoder/x86/dct_ssse3.c
index df298d8..74794b1 100644
--- a/vp10/encoder/x86/dct_ssse3.c
+++ b/vp10/encoder/x86/dct_ssse3.c
@@ -20,16 +20,12 @@
#include "vpx_dsp/x86/inv_txfm_sse2.h"
#include "vpx_dsp/x86/txfm_common_sse2.h"
-void vp10_fdct8x8_quant_ssse3(const int16_t *input, int stride,
- int16_t* coeff_ptr, intptr_t n_coeffs,
- int skip_block, const int16_t* zbin_ptr,
- const int16_t* round_ptr, const int16_t* quant_ptr,
- const int16_t* quant_shift_ptr,
- int16_t* qcoeff_ptr,
- int16_t* dqcoeff_ptr, const int16_t* dequant_ptr,
- uint16_t* eob_ptr,
- const int16_t* scan_ptr,
- const int16_t* iscan_ptr) {
+void vp10_fdct8x8_quant_ssse3(
+ const int16_t* input, int stride, int16_t* coeff_ptr, intptr_t n_coeffs,
+ int skip_block, const int16_t* zbin_ptr, const int16_t* round_ptr,
+ const int16_t* quant_ptr, const int16_t* quant_shift_ptr,
+ int16_t* qcoeff_ptr, int16_t* dqcoeff_ptr, const int16_t* dequant_ptr,
+ uint16_t* eob_ptr, const int16_t* scan_ptr, const int16_t* iscan_ptr) {
__m128i zero;
int pass;
// Constants
@@ -47,15 +43,15 @@
const __m128i k__cospi_m20_p12 = pair_set_epi16(-cospi_20_64, cospi_12_64);
const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
// Load input
- __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
- __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
- __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
- __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
- __m128i in4 = _mm_load_si128((const __m128i *)(input + 4 * stride));
- __m128i in5 = _mm_load_si128((const __m128i *)(input + 5 * stride));
- __m128i in6 = _mm_load_si128((const __m128i *)(input + 6 * stride));
- __m128i in7 = _mm_load_si128((const __m128i *)(input + 7 * stride));
- __m128i *in[8];
+ __m128i in0 = _mm_load_si128((const __m128i*)(input + 0 * stride));
+ __m128i in1 = _mm_load_si128((const __m128i*)(input + 1 * stride));
+ __m128i in2 = _mm_load_si128((const __m128i*)(input + 2 * stride));
+ __m128i in3 = _mm_load_si128((const __m128i*)(input + 3 * stride));
+ __m128i in4 = _mm_load_si128((const __m128i*)(input + 4 * stride));
+ __m128i in5 = _mm_load_si128((const __m128i*)(input + 5 * stride));
+ __m128i in6 = _mm_load_si128((const __m128i*)(input + 6 * stride));
+ __m128i in7 = _mm_load_si128((const __m128i*)(input + 7 * stride));
+ __m128i* in[8];
int index = 0;
(void)scan_ptr;
@@ -393,7 +389,7 @@
qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
nzflag = _mm_movemask_epi8(_mm_cmpgt_epi16(qcoeff0, thr)) |
- _mm_movemask_epi8(_mm_cmpgt_epi16(qcoeff1, thr));
+ _mm_movemask_epi8(_mm_cmpgt_epi16(qcoeff1, thr));
if (nzflag) {
qcoeff0 = _mm_adds_epi16(qcoeff0, round);
diff --git a/vp10/encoder/x86/error_intrin_avx2.c b/vp10/encoder/x86/error_intrin_avx2.c
index 9766be2..560cd4a 100644
--- a/vp10/encoder/x86/error_intrin_avx2.c
+++ b/vp10/encoder/x86/error_intrin_avx2.c
@@ -13,10 +13,8 @@
#include "./vp10_rtcd.h"
#include "vpx/vpx_integer.h"
-int64_t vp10_block_error_avx2(const int16_t *coeff,
- const int16_t *dqcoeff,
- intptr_t block_size,
- int64_t *ssz) {
+int64_t vp10_block_error_avx2(const int16_t *coeff, const int16_t *dqcoeff,
+ intptr_t block_size, int64_t *ssz) {
__m256i sse_reg, ssz_reg, coeff_reg, dqcoeff_reg;
__m256i exp_dqcoeff_lo, exp_dqcoeff_hi, exp_coeff_lo, exp_coeff_hi;
__m256i sse_reg_64hi, ssz_reg_64hi;
@@ -29,7 +27,7 @@
sse_reg = _mm256_set1_epi16(0);
ssz_reg = _mm256_set1_epi16(0);
- for (i = 0 ; i < block_size ; i+= 16) {
+ for (i = 0; i < block_size; i += 16) {
// load 32 bytes from coeff and dqcoeff
coeff_reg = _mm256_loadu_si256((const __m256i *)(coeff + i));
dqcoeff_reg = _mm256_loadu_si256((const __m256i *)(dqcoeff + i));
@@ -66,8 +64,8 @@
_mm256_extractf128_si256(ssz_reg, 1));
// store the results
- _mm_storel_epi64((__m128i*)(&sse), sse_reg128);
+ _mm_storel_epi64((__m128i *)(&sse), sse_reg128);
- _mm_storel_epi64((__m128i*)(ssz), ssz_reg128);
+ _mm_storel_epi64((__m128i *)(ssz), ssz_reg128);
return sse;
}
diff --git a/vp10/encoder/x86/highbd_block_error_intrin_sse2.c b/vp10/encoder/x86/highbd_block_error_intrin_sse2.c
index 6b4cf50..eefca1c 100644
--- a/vp10/encoder/x86/highbd_block_error_intrin_sse2.c
+++ b/vp10/encoder/x86/highbd_block_error_intrin_sse2.c
@@ -14,8 +14,8 @@
#include "vp10/common/common.h"
int64_t vp10_highbd_block_error_sse2(tran_low_t *coeff, tran_low_t *dqcoeff,
- intptr_t block_size, int64_t *ssz,
- int bps) {
+ intptr_t block_size, int64_t *ssz,
+ int bps) {
int i, j, test;
uint32_t temp[4];
__m128i max, min, cmp0, cmp1, cmp2, cmp3;
@@ -23,41 +23,41 @@
const int shift = 2 * (bps - 8);
const int rounding = shift > 0 ? 1 << (shift - 1) : 0;
- for (i = 0; i < block_size; i+=8) {
+ for (i = 0; i < block_size; i += 8) {
// Load the data into xmm registers
- __m128i mm_coeff = _mm_load_si128((__m128i*) (coeff + i));
- __m128i mm_coeff2 = _mm_load_si128((__m128i*) (coeff + i + 4));
- __m128i mm_dqcoeff = _mm_load_si128((__m128i*) (dqcoeff + i));
- __m128i mm_dqcoeff2 = _mm_load_si128((__m128i*) (dqcoeff + i + 4));
+ __m128i mm_coeff = _mm_load_si128((__m128i *)(coeff + i));
+ __m128i mm_coeff2 = _mm_load_si128((__m128i *)(coeff + i + 4));
+ __m128i mm_dqcoeff = _mm_load_si128((__m128i *)(dqcoeff + i));
+ __m128i mm_dqcoeff2 = _mm_load_si128((__m128i *)(dqcoeff + i + 4));
// Check if any values require more than 15 bit
max = _mm_set1_epi32(0x3fff);
min = _mm_set1_epi32(0xffffc000);
cmp0 = _mm_xor_si128(_mm_cmpgt_epi32(mm_coeff, max),
- _mm_cmplt_epi32(mm_coeff, min));
+ _mm_cmplt_epi32(mm_coeff, min));
cmp1 = _mm_xor_si128(_mm_cmpgt_epi32(mm_coeff2, max),
- _mm_cmplt_epi32(mm_coeff2, min));
+ _mm_cmplt_epi32(mm_coeff2, min));
cmp2 = _mm_xor_si128(_mm_cmpgt_epi32(mm_dqcoeff, max),
- _mm_cmplt_epi32(mm_dqcoeff, min));
+ _mm_cmplt_epi32(mm_dqcoeff, min));
cmp3 = _mm_xor_si128(_mm_cmpgt_epi32(mm_dqcoeff2, max),
- _mm_cmplt_epi32(mm_dqcoeff2, min));
- test = _mm_movemask_epi8(_mm_or_si128(_mm_or_si128(cmp0, cmp1),
- _mm_or_si128(cmp2, cmp3)));
+ _mm_cmplt_epi32(mm_dqcoeff2, min));
+ test = _mm_movemask_epi8(
+ _mm_or_si128(_mm_or_si128(cmp0, cmp1), _mm_or_si128(cmp2, cmp3)));
if (!test) {
- __m128i mm_diff, error_sse2, sqcoeff_sse2;;
+ __m128i mm_diff, error_sse2, sqcoeff_sse2;
mm_coeff = _mm_packs_epi32(mm_coeff, mm_coeff2);
mm_dqcoeff = _mm_packs_epi32(mm_dqcoeff, mm_dqcoeff2);
mm_diff = _mm_sub_epi16(mm_coeff, mm_dqcoeff);
error_sse2 = _mm_madd_epi16(mm_diff, mm_diff);
sqcoeff_sse2 = _mm_madd_epi16(mm_coeff, mm_coeff);
- _mm_storeu_si128((__m128i*)temp, error_sse2);
+ _mm_storeu_si128((__m128i *)temp, error_sse2);
error = error + temp[0] + temp[1] + temp[2] + temp[3];
- _mm_storeu_si128((__m128i*)temp, sqcoeff_sse2);
+ _mm_storeu_si128((__m128i *)temp, sqcoeff_sse2);
sqcoeff += temp[0] + temp[1] + temp[2] + temp[3];
} else {
for (j = 0; j < 8; j++) {
const int64_t diff = coeff[i + j] - dqcoeff[i + j];
- error += diff * diff;
+ error += diff * diff;
sqcoeff += (int64_t)coeff[i + j] * (int64_t)coeff[i + j];
}
}
diff --git a/vp10/encoder/x86/quantize_sse2.c b/vp10/encoder/x86/quantize_sse2.c
index dabd3bd..78d5952 100644
--- a/vp10/encoder/x86/quantize_sse2.c
+++ b/vp10/encoder/x86/quantize_sse2.c
@@ -15,13 +15,12 @@
#include "vpx/vpx_integer.h"
void vp10_quantize_fp_sse2(const int16_t* coeff_ptr, intptr_t n_coeffs,
- int skip_block, const int16_t* zbin_ptr,
- const int16_t* round_ptr, const int16_t* quant_ptr,
- const int16_t* quant_shift_ptr, int16_t* qcoeff_ptr,
- int16_t* dqcoeff_ptr, const int16_t* dequant_ptr,
- uint16_t* eob_ptr,
- const int16_t* scan_ptr,
- const int16_t* iscan_ptr) {
+ int skip_block, const int16_t* zbin_ptr,
+ const int16_t* round_ptr, const int16_t* quant_ptr,
+ const int16_t* quant_shift_ptr, int16_t* qcoeff_ptr,
+ int16_t* dqcoeff_ptr, const int16_t* dequant_ptr,
+ uint16_t* eob_ptr, const int16_t* scan_ptr,
+ const int16_t* iscan_ptr) {
__m128i zero;
__m128i thr;
int16_t nzflag;
@@ -133,7 +132,7 @@
qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
nzflag = _mm_movemask_epi8(_mm_cmpgt_epi16(qcoeff0, thr)) |
- _mm_movemask_epi8(_mm_cmpgt_epi16(qcoeff1, thr));
+ _mm_movemask_epi8(_mm_cmpgt_epi16(qcoeff1, thr));
if (nzflag) {
qcoeff0 = _mm_adds_epi16(qcoeff0, round);
diff --git a/vp10/vp10_cx_iface.c b/vp10/vp10_cx_iface.c
index 8dfe1f9..0e73788 100644
--- a/vp10/vp10_cx_iface.c
+++ b/vp10/vp10_cx_iface.c
@@ -22,103 +22,100 @@
#include "vp10/vp10_iface_common.h"
struct vp10_extracfg {
- int cpu_used; // available cpu percentage in 1/16
- unsigned int enable_auto_alt_ref;
- unsigned int noise_sensitivity;
- unsigned int sharpness;
- unsigned int static_thresh;
- unsigned int tile_columns;
- unsigned int tile_rows;
- unsigned int arnr_max_frames;
- unsigned int arnr_strength;
- unsigned int min_gf_interval;
- unsigned int max_gf_interval;
- vpx_tune_metric tuning;
- unsigned int cq_level; // constrained quality level
- unsigned int rc_max_intra_bitrate_pct;
- unsigned int rc_max_inter_bitrate_pct;
- unsigned int gf_cbr_boost_pct;
- unsigned int lossless;
- unsigned int frame_parallel_decoding_mode;
- AQ_MODE aq_mode;
- unsigned int frame_periodic_boost;
- vpx_bit_depth_t bit_depth;
- vpx_tune_content content;
- vpx_color_space_t color_space;
- int color_range;
- int render_width;
- int render_height;
+ int cpu_used; // available cpu percentage in 1/16
+ unsigned int enable_auto_alt_ref;
+ unsigned int noise_sensitivity;
+ unsigned int sharpness;
+ unsigned int static_thresh;
+ unsigned int tile_columns;
+ unsigned int tile_rows;
+ unsigned int arnr_max_frames;
+ unsigned int arnr_strength;
+ unsigned int min_gf_interval;
+ unsigned int max_gf_interval;
+ vpx_tune_metric tuning;
+ unsigned int cq_level; // constrained quality level
+ unsigned int rc_max_intra_bitrate_pct;
+ unsigned int rc_max_inter_bitrate_pct;
+ unsigned int gf_cbr_boost_pct;
+ unsigned int lossless;
+ unsigned int frame_parallel_decoding_mode;
+ AQ_MODE aq_mode;
+ unsigned int frame_periodic_boost;
+ vpx_bit_depth_t bit_depth;
+ vpx_tune_content content;
+ vpx_color_space_t color_space;
+ int color_range;
+ int render_width;
+ int render_height;
};
static struct vp10_extracfg default_extra_cfg = {
- 0, // cpu_used
- 1, // enable_auto_alt_ref
- 0, // noise_sensitivity
- 0, // sharpness
- 0, // static_thresh
- 6, // tile_columns
- 0, // tile_rows
- 7, // arnr_max_frames
- 5, // arnr_strength
- 0, // min_gf_interval; 0 -> default decision
- 0, // max_gf_interval; 0 -> default decision
- VPX_TUNE_PSNR, // tuning
- 10, // cq_level
- 0, // rc_max_intra_bitrate_pct
- 0, // rc_max_inter_bitrate_pct
- 0, // gf_cbr_boost_pct
- 0, // lossless
- 1, // frame_parallel_decoding_mode
- NO_AQ, // aq_mode
- 0, // frame_periodic_delta_q
- VPX_BITS_8, // Bit depth
- VPX_CONTENT_DEFAULT, // content
- VPX_CS_UNKNOWN, // color space
- 0, // color range
- 0, // render width
- 0, // render height
+ 0, // cpu_used
+ 1, // enable_auto_alt_ref
+ 0, // noise_sensitivity
+ 0, // sharpness
+ 0, // static_thresh
+ 6, // tile_columns
+ 0, // tile_rows
+ 7, // arnr_max_frames
+ 5, // arnr_strength
+ 0, // min_gf_interval; 0 -> default decision
+ 0, // max_gf_interval; 0 -> default decision
+ VPX_TUNE_PSNR, // tuning
+ 10, // cq_level
+ 0, // rc_max_intra_bitrate_pct
+ 0, // rc_max_inter_bitrate_pct
+ 0, // gf_cbr_boost_pct
+ 0, // lossless
+ 1, // frame_parallel_decoding_mode
+ NO_AQ, // aq_mode
+ 0, // frame_periodic_delta_q
+ VPX_BITS_8, // Bit depth
+ VPX_CONTENT_DEFAULT, // content
+ VPX_CS_UNKNOWN, // color space
+ 0, // color range
+ 0, // render width
+ 0, // render height
};
struct vpx_codec_alg_priv {
- vpx_codec_priv_t base;
- vpx_codec_enc_cfg_t cfg;
- struct vp10_extracfg extra_cfg;
- VP10EncoderConfig oxcf;
- VP10_COMP *cpi;
- unsigned char *cx_data;
- size_t cx_data_sz;
- unsigned char *pending_cx_data;
- size_t pending_cx_data_sz;
- int pending_frame_count;
- size_t pending_frame_sizes[8];
+ vpx_codec_priv_t base;
+ vpx_codec_enc_cfg_t cfg;
+ struct vp10_extracfg extra_cfg;
+ VP10EncoderConfig oxcf;
+ VP10_COMP *cpi;
+ unsigned char *cx_data;
+ size_t cx_data_sz;
+ unsigned char *pending_cx_data;
+ size_t pending_cx_data_sz;
+ int pending_frame_count;
+ size_t pending_frame_sizes[8];
#if !CONFIG_MISC_FIXES
- size_t pending_frame_magnitude;
+ size_t pending_frame_magnitude;
#endif
- vpx_image_t preview_img;
- vpx_enc_frame_flags_t next_frame_flags;
- vp8_postproc_cfg_t preview_ppcfg;
+ vpx_image_t preview_img;
+ vpx_enc_frame_flags_t next_frame_flags;
+ vp8_postproc_cfg_t preview_ppcfg;
vpx_codec_pkt_list_decl(256) pkt_list;
- unsigned int fixed_kf_cntr;
+ unsigned int fixed_kf_cntr;
vpx_codec_priv_output_cx_pkt_cb_pair_t output_cx_pkt_cb;
// BufferPool that holds all reference frames.
- BufferPool *buffer_pool;
+ BufferPool *buffer_pool;
};
static VPX_REFFRAME ref_frame_to_vp10_reframe(vpx_ref_frame_type_t frame) {
switch (frame) {
- case VP8_LAST_FRAME:
- return VPX_LAST_FLAG;
- case VP8_GOLD_FRAME:
- return VPX_GOLD_FLAG;
- case VP8_ALTR_FRAME:
- return VPX_ALT_FLAG;
+ case VP8_LAST_FRAME: return VPX_LAST_FLAG;
+ case VP8_GOLD_FRAME: return VPX_GOLD_FLAG;
+ case VP8_ALTR_FRAME: return VPX_ALT_FLAG;
}
assert(0 && "Invalid Reference Frame");
return VPX_LAST_FLAG;
}
-static vpx_codec_err_t update_error_state(vpx_codec_alg_priv_t *ctx,
- const struct vpx_internal_error_info *error) {
+static vpx_codec_err_t update_error_state(
+ vpx_codec_alg_priv_t *ctx, const struct vpx_internal_error_info *error) {
const vpx_codec_err_t res = error->error_code;
if (res != VPX_CODEC_OK)
@@ -127,58 +124,60 @@
return res;
}
-
#undef ERROR
-#define ERROR(str) do {\
- ctx->base.err_detail = str;\
- return VPX_CODEC_INVALID_PARAM;\
+#define ERROR(str) \
+ do { \
+ ctx->base.err_detail = str; \
+ return VPX_CODEC_INVALID_PARAM; \
} while (0)
-#define RANGE_CHECK(p, memb, lo, hi) do {\
+#define RANGE_CHECK(p, memb, lo, hi) \
+ do { \
if (!(((p)->memb == lo || (p)->memb > (lo)) && (p)->memb <= hi)) \
- ERROR(#memb " out of range ["#lo".."#hi"]");\
+ ERROR(#memb " out of range [" #lo ".." #hi "]"); \
} while (0)
-#define RANGE_CHECK_HI(p, memb, hi) do {\
- if (!((p)->memb <= (hi))) \
- ERROR(#memb " out of range [.."#hi"]");\
+#define RANGE_CHECK_HI(p, memb, hi) \
+ do { \
+ if (!((p)->memb <= (hi))) ERROR(#memb " out of range [.." #hi "]"); \
} while (0)
-#define RANGE_CHECK_LO(p, memb, lo) do {\
- if (!((p)->memb >= (lo))) \
- ERROR(#memb " out of range ["#lo"..]");\
+#define RANGE_CHECK_LO(p, memb, lo) \
+ do { \
+ if (!((p)->memb >= (lo))) ERROR(#memb " out of range [" #lo "..]"); \
} while (0)
-#define RANGE_CHECK_BOOL(p, memb) do {\
- if (!!((p)->memb) != (p)->memb) ERROR(#memb " expected boolean");\
+#define RANGE_CHECK_BOOL(p, memb) \
+ do { \
+ if (!!((p)->memb) != (p)->memb) ERROR(#memb " expected boolean"); \
} while (0)
static vpx_codec_err_t validate_config(vpx_codec_alg_priv_t *ctx,
const vpx_codec_enc_cfg_t *cfg,
const struct vp10_extracfg *extra_cfg) {
- RANGE_CHECK(cfg, g_w, 1, 65535); // 16 bits available
- RANGE_CHECK(cfg, g_h, 1, 65535); // 16 bits available
- RANGE_CHECK(cfg, g_timebase.den, 1, 1000000000);
- RANGE_CHECK(cfg, g_timebase.num, 1, cfg->g_timebase.den);
- RANGE_CHECK_HI(cfg, g_profile, 3);
+ RANGE_CHECK(cfg, g_w, 1, 65535); // 16 bits available
+ RANGE_CHECK(cfg, g_h, 1, 65535); // 16 bits available
+ RANGE_CHECK(cfg, g_timebase.den, 1, 1000000000);
+ RANGE_CHECK(cfg, g_timebase.num, 1, cfg->g_timebase.den);
+ RANGE_CHECK_HI(cfg, g_profile, 3);
- RANGE_CHECK_HI(cfg, rc_max_quantizer, 63);
- RANGE_CHECK_HI(cfg, rc_min_quantizer, cfg->rc_max_quantizer);
+ RANGE_CHECK_HI(cfg, rc_max_quantizer, 63);
+ RANGE_CHECK_HI(cfg, rc_min_quantizer, cfg->rc_max_quantizer);
RANGE_CHECK_BOOL(extra_cfg, lossless);
- RANGE_CHECK(extra_cfg, aq_mode, 0, AQ_MODE_COUNT - 1);
+ RANGE_CHECK(extra_cfg, aq_mode, 0, AQ_MODE_COUNT - 1);
RANGE_CHECK(extra_cfg, frame_periodic_boost, 0, 1);
- RANGE_CHECK_HI(cfg, g_threads, 64);
- RANGE_CHECK_HI(cfg, g_lag_in_frames, MAX_LAG_BUFFERS);
- RANGE_CHECK(cfg, rc_end_usage, VPX_VBR, VPX_Q);
- RANGE_CHECK_HI(cfg, rc_undershoot_pct, 100);
- RANGE_CHECK_HI(cfg, rc_overshoot_pct, 100);
+ RANGE_CHECK_HI(cfg, g_threads, 64);
+ RANGE_CHECK_HI(cfg, g_lag_in_frames, MAX_LAG_BUFFERS);
+ RANGE_CHECK(cfg, rc_end_usage, VPX_VBR, VPX_Q);
+ RANGE_CHECK_HI(cfg, rc_undershoot_pct, 100);
+ RANGE_CHECK_HI(cfg, rc_overshoot_pct, 100);
RANGE_CHECK_HI(cfg, rc_2pass_vbr_bias_pct, 100);
- RANGE_CHECK(cfg, kf_mode, VPX_KF_DISABLED, VPX_KF_AUTO);
- RANGE_CHECK_BOOL(cfg, rc_resize_allowed);
- RANGE_CHECK_HI(cfg, rc_dropframe_thresh, 100);
- RANGE_CHECK_HI(cfg, rc_resize_up_thresh, 100);
+ RANGE_CHECK(cfg, kf_mode, VPX_KF_DISABLED, VPX_KF_AUTO);
+ RANGE_CHECK_BOOL(cfg, rc_resize_allowed);
+ RANGE_CHECK_HI(cfg, rc_dropframe_thresh, 100);
+ RANGE_CHECK_HI(cfg, rc_resize_up_thresh, 100);
RANGE_CHECK_HI(cfg, rc_resize_down_thresh, 100);
- RANGE_CHECK(cfg, g_pass, VPX_RC_ONE_PASS, VPX_RC_LAST_PASS);
+ RANGE_CHECK(cfg, g_pass, VPX_RC_ONE_PASS, VPX_RC_LAST_PASS);
RANGE_CHECK(extra_cfg, min_gf_interval, 0, (MAX_LAG_BUFFERS - 1));
RANGE_CHECK(extra_cfg, max_gf_interval, 0, (MAX_LAG_BUFFERS - 1));
if (extra_cfg->max_gf_interval > 0) {
@@ -186,7 +185,7 @@
}
if (extra_cfg->min_gf_interval > 0 && extra_cfg->max_gf_interval > 0) {
RANGE_CHECK(extra_cfg, max_gf_interval, extra_cfg->min_gf_interval,
- (MAX_LAG_BUFFERS - 1));
+ (MAX_LAG_BUFFERS - 1));
}
if (cfg->rc_resize_allowed == 1) {
@@ -200,11 +199,11 @@
RANGE_CHECK(cfg, ts_number_layers, 1, 1);
// VP10 does not support a lower bound on the keyframe interval in
// automatic keyframe placement mode.
- if (cfg->kf_mode != VPX_KF_DISABLED &&
- cfg->kf_min_dist != cfg->kf_max_dist &&
+ if (cfg->kf_mode != VPX_KF_DISABLED && cfg->kf_min_dist != cfg->kf_max_dist &&
cfg->kf_min_dist > 0)
- ERROR("kf_min_dist not supported in auto mode, use 0 "
- "or kf_max_dist instead.");
+ ERROR(
+ "kf_min_dist not supported in auto mode, use 0 "
+ "or kf_max_dist instead.");
RANGE_CHECK(extra_cfg, enable_auto_alt_ref, 0, 2);
RANGE_CHECK(extra_cfg, cpu_used, -8, 8);
@@ -217,12 +216,11 @@
RANGE_CHECK(extra_cfg, cq_level, 0, 63);
RANGE_CHECK(cfg, g_bit_depth, VPX_BITS_8, VPX_BITS_12);
RANGE_CHECK(cfg, g_input_bit_depth, 8, 12);
- RANGE_CHECK(extra_cfg, content,
- VPX_CONTENT_DEFAULT, VPX_CONTENT_INVALID - 1);
+ RANGE_CHECK(extra_cfg, content, VPX_CONTENT_DEFAULT, VPX_CONTENT_INVALID - 1);
// TODO(yaowu): remove this when ssim tuning is implemented for vp10
if (extra_cfg->tuning == VPX_TUNE_SSIM)
- ERROR("Option --tune=ssim is not currently supported in VP10.");
+ ERROR("Option --tune=ssim is not currently supported in VP10.");
if (cfg->g_pass == VPX_RC_LAST_PASS) {
const size_t packet_sz = sizeof(FIRSTPASS_STATS);
@@ -254,8 +252,7 @@
cfg->g_bit_depth > VPX_BITS_8) {
ERROR("Codec high bit-depth not supported in profile < 2");
}
- if (cfg->g_profile <= (unsigned int)PROFILE_1 &&
- cfg->g_input_bit_depth > 8) {
+ if (cfg->g_profile <= (unsigned int)PROFILE_1 && cfg->g_input_bit_depth > 8) {
ERROR("Source high bit-depth not supported in profile < 2");
}
if (cfg->g_profile > (unsigned int)PROFILE_1 &&
@@ -272,14 +269,14 @@
switch (img->fmt) {
case VPX_IMG_FMT_YV12:
case VPX_IMG_FMT_I420:
- case VPX_IMG_FMT_I42016:
- break;
+ case VPX_IMG_FMT_I42016: break;
case VPX_IMG_FMT_I422:
case VPX_IMG_FMT_I444:
case VPX_IMG_FMT_I440:
if (ctx->cfg.g_profile != (unsigned int)PROFILE_1) {
- ERROR("Invalid image format. I422, I444, I440 images are "
- "not supported in profile.");
+ ERROR(
+ "Invalid image format. I422, I444, I440 images are "
+ "not supported in profile.");
}
break;
case VPX_IMG_FMT_I42216:
@@ -287,13 +284,15 @@
case VPX_IMG_FMT_I44016:
if (ctx->cfg.g_profile != (unsigned int)PROFILE_1 &&
ctx->cfg.g_profile != (unsigned int)PROFILE_3) {
- ERROR("Invalid image format. 16-bit I422, I444, I440 images are "
- "not supported in profile.");
+ ERROR(
+ "Invalid image format. 16-bit I422, I444, I440 images are "
+ "not supported in profile.");
}
break;
default:
- ERROR("Invalid image format. Only YV12, I420, I422, I444 images are "
- "supported.");
+ ERROR(
+ "Invalid image format. Only YV12, I420, I422, I444 images are "
+ "supported.");
break;
}
@@ -320,37 +319,29 @@
}
static vpx_codec_err_t set_encoder_config(
- VP10EncoderConfig *oxcf,
- const vpx_codec_enc_cfg_t *cfg,
- const struct vp10_extracfg *extra_cfg) {
+ VP10EncoderConfig *oxcf, const vpx_codec_enc_cfg_t *cfg,
+ const struct vp10_extracfg *extra_cfg) {
const int is_vbr = cfg->rc_end_usage == VPX_VBR;
oxcf->profile = cfg->g_profile;
oxcf->max_threads = (int)cfg->g_threads;
- oxcf->width = cfg->g_w;
- oxcf->height = cfg->g_h;
+ oxcf->width = cfg->g_w;
+ oxcf->height = cfg->g_h;
oxcf->bit_depth = cfg->g_bit_depth;
oxcf->input_bit_depth = cfg->g_input_bit_depth;
// guess a frame rate if out of whack, use 30
oxcf->init_framerate = (double)cfg->g_timebase.den / cfg->g_timebase.num;
- if (oxcf->init_framerate > 180)
- oxcf->init_framerate = 30;
+ if (oxcf->init_framerate > 180) oxcf->init_framerate = 30;
oxcf->mode = GOOD;
switch (cfg->g_pass) {
- case VPX_RC_ONE_PASS:
- oxcf->pass = 0;
- break;
- case VPX_RC_FIRST_PASS:
- oxcf->pass = 1;
- break;
- case VPX_RC_LAST_PASS:
- oxcf->pass = 2;
- break;
+ case VPX_RC_ONE_PASS: oxcf->pass = 0; break;
+ case VPX_RC_FIRST_PASS: oxcf->pass = 1; break;
+ case VPX_RC_LAST_PASS: oxcf->pass = 2; break;
}
- oxcf->lag_in_frames = cfg->g_pass == VPX_RC_FIRST_PASS ? 0
- : cfg->g_lag_in_frames;
+ oxcf->lag_in_frames =
+ cfg->g_pass == VPX_RC_FIRST_PASS ? 0 : cfg->g_lag_in_frames;
oxcf->rc_mode = cfg->rc_end_usage;
// Convert target bandwidth from Kbit/s to Bit/s
@@ -363,55 +354,56 @@
extra_cfg->lossless ? 0 : vp10_quantizer_to_qindex(cfg->rc_min_quantizer);
oxcf->worst_allowed_q =
extra_cfg->lossless ? 0 : vp10_quantizer_to_qindex(cfg->rc_max_quantizer);
- oxcf->cq_level = vp10_quantizer_to_qindex(extra_cfg->cq_level);
+ oxcf->cq_level = vp10_quantizer_to_qindex(extra_cfg->cq_level);
oxcf->fixed_q = -1;
- oxcf->under_shoot_pct = cfg->rc_undershoot_pct;
- oxcf->over_shoot_pct = cfg->rc_overshoot_pct;
+ oxcf->under_shoot_pct = cfg->rc_undershoot_pct;
+ oxcf->over_shoot_pct = cfg->rc_overshoot_pct;
- oxcf->scaled_frame_width = cfg->rc_scaled_width;
+ oxcf->scaled_frame_width = cfg->rc_scaled_width;
oxcf->scaled_frame_height = cfg->rc_scaled_height;
if (cfg->rc_resize_allowed == 1) {
oxcf->resize_mode =
- (oxcf->scaled_frame_width == 0 || oxcf->scaled_frame_height == 0) ?
- RESIZE_DYNAMIC : RESIZE_FIXED;
+ (oxcf->scaled_frame_width == 0 || oxcf->scaled_frame_height == 0)
+ ? RESIZE_DYNAMIC
+ : RESIZE_FIXED;
} else {
oxcf->resize_mode = RESIZE_NONE;
}
- oxcf->maximum_buffer_size_ms = is_vbr ? 240000 : cfg->rc_buf_sz;
+ oxcf->maximum_buffer_size_ms = is_vbr ? 240000 : cfg->rc_buf_sz;
oxcf->starting_buffer_level_ms = is_vbr ? 60000 : cfg->rc_buf_initial_sz;
- oxcf->optimal_buffer_level_ms = is_vbr ? 60000 : cfg->rc_buf_optimal_sz;
+ oxcf->optimal_buffer_level_ms = is_vbr ? 60000 : cfg->rc_buf_optimal_sz;
- oxcf->drop_frames_water_mark = cfg->rc_dropframe_thresh;
+ oxcf->drop_frames_water_mark = cfg->rc_dropframe_thresh;
- oxcf->two_pass_vbrbias = cfg->rc_2pass_vbr_bias_pct;
- oxcf->two_pass_vbrmin_section = cfg->rc_2pass_vbr_minsection_pct;
- oxcf->two_pass_vbrmax_section = cfg->rc_2pass_vbr_maxsection_pct;
+ oxcf->two_pass_vbrbias = cfg->rc_2pass_vbr_bias_pct;
+ oxcf->two_pass_vbrmin_section = cfg->rc_2pass_vbr_minsection_pct;
+ oxcf->two_pass_vbrmax_section = cfg->rc_2pass_vbr_maxsection_pct;
- oxcf->auto_key = cfg->kf_mode == VPX_KF_AUTO &&
- cfg->kf_min_dist != cfg->kf_max_dist;
+ oxcf->auto_key =
+ cfg->kf_mode == VPX_KF_AUTO && cfg->kf_min_dist != cfg->kf_max_dist;
- oxcf->key_freq = cfg->kf_max_dist;
+ oxcf->key_freq = cfg->kf_max_dist;
- oxcf->speed = abs(extra_cfg->cpu_used);
- oxcf->encode_breakout = extra_cfg->static_thresh;
- oxcf->enable_auto_arf = extra_cfg->enable_auto_alt_ref;
- oxcf->noise_sensitivity = extra_cfg->noise_sensitivity;
- oxcf->sharpness = extra_cfg->sharpness;
+ oxcf->speed = abs(extra_cfg->cpu_used);
+ oxcf->encode_breakout = extra_cfg->static_thresh;
+ oxcf->enable_auto_arf = extra_cfg->enable_auto_alt_ref;
+ oxcf->noise_sensitivity = extra_cfg->noise_sensitivity;
+ oxcf->sharpness = extra_cfg->sharpness;
- oxcf->two_pass_stats_in = cfg->rc_twopass_stats_in;
+ oxcf->two_pass_stats_in = cfg->rc_twopass_stats_in;
#if CONFIG_FP_MB_STATS
- oxcf->firstpass_mb_stats_in = cfg->rc_firstpass_mb_stats_in;
+ oxcf->firstpass_mb_stats_in = cfg->rc_firstpass_mb_stats_in;
#endif
oxcf->color_space = extra_cfg->color_space;
oxcf->color_range = extra_cfg->color_range;
- oxcf->render_width = extra_cfg->render_width;
+ oxcf->render_width = extra_cfg->render_width;
oxcf->render_height = extra_cfg->render_height;
oxcf->arnr_max_frames = extra_cfg->arnr_max_frames;
- oxcf->arnr_strength = extra_cfg->arnr_strength;
+ oxcf->arnr_strength = extra_cfg->arnr_strength;
oxcf->min_gf_interval = extra_cfg->min_gf_interval;
oxcf->max_gf_interval = extra_cfg->max_gf_interval;
@@ -419,14 +411,14 @@
oxcf->content = extra_cfg->content;
oxcf->tile_columns = extra_cfg->tile_columns;
- oxcf->tile_rows = extra_cfg->tile_rows;
+ oxcf->tile_rows = extra_cfg->tile_rows;
- oxcf->error_resilient_mode = cfg->g_error_resilient;
+ oxcf->error_resilient_mode = cfg->g_error_resilient;
oxcf->frame_parallel_decoding_mode = extra_cfg->frame_parallel_decoding_mode;
oxcf->aq_mode = extra_cfg->aq_mode;
- oxcf->frame_periodic_boost = extra_cfg->frame_periodic_boost;
+ oxcf->frame_periodic_boost = extra_cfg->frame_periodic_boost;
/*
printf("Current VP10 Settings: \n");
@@ -464,7 +456,7 @@
}
static vpx_codec_err_t encoder_set_config(vpx_codec_alg_priv_t *ctx,
- const vpx_codec_enc_cfg_t *cfg) {
+ const vpx_codec_enc_cfg_t *cfg) {
vpx_codec_err_t res;
int force_key = 0;
@@ -494,8 +486,7 @@
vp10_change_config(ctx->cpi, &ctx->oxcf);
}
- if (force_key)
- ctx->next_frame_flags |= VPX_EFLAG_FORCE_KF;
+ if (force_key) ctx->next_frame_flags |= VPX_EFLAG_FORCE_KF;
return res;
}
@@ -503,8 +494,7 @@
static vpx_codec_err_t ctrl_get_quantizer(vpx_codec_alg_priv_t *ctx,
va_list args) {
int *const arg = va_arg(args, int *);
- if (arg == NULL)
- return VPX_CODEC_INVALID_PARAM;
+ if (arg == NULL) return VPX_CODEC_INVALID_PARAM;
*arg = vp10_get_quantizer(ctx->cpi);
return VPX_CODEC_OK;
}
@@ -512,8 +502,7 @@
static vpx_codec_err_t ctrl_get_quantizer64(vpx_codec_alg_priv_t *ctx,
va_list args) {
int *const arg = va_arg(args, int *);
- if (arg == NULL)
- return VPX_CODEC_INVALID_PARAM;
+ if (arg == NULL) return VPX_CODEC_INVALID_PARAM;
*arg = vp10_qindex_to_quantizer(vp10_get_quantizer(ctx->cpi));
return VPX_CODEC_OK;
}
@@ -629,11 +618,10 @@
return update_extra_cfg(ctx, &extra_cfg);
}
-static vpx_codec_err_t ctrl_set_rc_gf_cbr_boost_pct(
- vpx_codec_alg_priv_t *ctx, va_list args) {
+static vpx_codec_err_t ctrl_set_rc_gf_cbr_boost_pct(vpx_codec_alg_priv_t *ctx,
+ va_list args) {
struct vp10_extracfg extra_cfg = ctx->extra_cfg;
- extra_cfg.gf_cbr_boost_pct =
- CAST(VP9E_SET_GF_CBR_BOOST_PCT, args);
+ extra_cfg.gf_cbr_boost_pct = CAST(VP9E_SET_GF_CBR_BOOST_PCT, args);
return update_extra_cfg(ctx, &extra_cfg);
}
@@ -687,16 +675,13 @@
if (ctx->priv == NULL) {
vpx_codec_alg_priv_t *const priv = vpx_calloc(1, sizeof(*priv));
- if (priv == NULL)
- return VPX_CODEC_MEM_ERROR;
+ if (priv == NULL) return VPX_CODEC_MEM_ERROR;
ctx->priv = (vpx_codec_priv_t *)priv;
ctx->priv->init_flags = ctx->init_flags;
ctx->priv->enc.total_encoders = 1;
- priv->buffer_pool =
- (BufferPool *)vpx_calloc(1, sizeof(BufferPool));
- if (priv->buffer_pool == NULL)
- return VPX_CODEC_MEM_ERROR;
+ priv->buffer_pool = (BufferPool *)vpx_calloc(1, sizeof(BufferPool));
+ if (priv->buffer_pool == NULL) return VPX_CODEC_MEM_ERROR;
#if CONFIG_MULTITHREAD
if (pthread_mutex_init(&priv->buffer_pool->pool_mutex, NULL)) {
@@ -755,7 +740,8 @@
// Convert duration parameter from stream timebase to microseconds.
const uint64_t duration_us = (uint64_t)duration * 1000000 *
- (uint64_t)cfg->g_timebase.num /(uint64_t)cfg->g_timebase.den;
+ (uint64_t)cfg->g_timebase.num /
+ (uint64_t)cfg->g_timebase.den;
// If the deadline is more that the duration this frame is to be shown,
// use good quality mode. Otherwise use realtime mode.
@@ -764,11 +750,8 @@
new_mode = BEST;
}
break;
- case VPX_RC_FIRST_PASS:
- break;
- case VPX_RC_LAST_PASS:
- new_mode = deadline > 0 ? GOOD : BEST;
- break;
+ case VPX_RC_FIRST_PASS: break;
+ case VPX_RC_LAST_PASS: new_mode = deadline > 0 ? GOOD : BEST; break;
}
if (ctx->oxcf.mode != new_mode) {
@@ -795,7 +778,7 @@
marker |= ctx->pending_frame_count - 1;
#if CONFIG_MISC_FIXES
for (i = 0; i < ctx->pending_frame_count - 1; i++) {
- const size_t frame_sz = (unsigned int) ctx->pending_frame_sizes[i] - 1;
+ const size_t frame_sz = (unsigned int)ctx->pending_frame_sizes[i] - 1;
max_frame_sz = frame_sz > max_frame_sz ? frame_sz : max_frame_sz;
}
#endif
@@ -803,11 +786,9 @@
// Choose the magnitude
for (mag = 0, mask = 0xff; mag < 4; mag++) {
#if CONFIG_MISC_FIXES
- if (max_frame_sz <= mask)
- break;
+ if (max_frame_sz <= mask) break;
#else
- if (ctx->pending_frame_magnitude < mask)
- break;
+ if (ctx->pending_frame_magnitude < mask) break;
#endif
mask <<= 8;
mask |= 0xff;
@@ -872,16 +853,14 @@
unsigned int lib_flags) {
vpx_codec_frame_flags_t flags = lib_flags << 16;
- if (lib_flags & FRAMEFLAGS_KEY)
- flags |= VPX_FRAME_IS_KEY;
+ if (lib_flags & FRAMEFLAGS_KEY) flags |= VPX_FRAME_IS_KEY;
- if (cpi->droppable)
- flags |= VPX_FRAME_IS_DROPPABLE;
+ if (cpi->droppable) flags |= VPX_FRAME_IS_DROPPABLE;
return flags;
}
-static vpx_codec_err_t encoder_encode(vpx_codec_alg_priv_t *ctx,
+static vpx_codec_err_t encoder_encode(vpx_codec_alg_priv_t *ctx,
const vpx_image_t *img,
vpx_codec_pts_t pts,
unsigned long duration,
@@ -901,12 +880,11 @@
// instance for its status to determine the compressed data size.
data_sz = ctx->cfg.g_w * ctx->cfg.g_h * get_image_bps(img) / 8 *
(cpi->multi_arf_allowed ? 8 : 2);
- if (data_sz < 4096)
- data_sz = 4096;
+ if (data_sz < 4096) data_sz = 4096;
if (ctx->cx_data == NULL || ctx->cx_data_sz < data_sz) {
ctx->cx_data_sz = data_sz;
free(ctx->cx_data);
- ctx->cx_data = (unsigned char*)malloc(ctx->cx_data_sz);
+ ctx->cx_data = (unsigned char *)malloc(ctx->cx_data_sz);
if (ctx->cx_data == NULL) {
return VPX_CODEC_MEM_ERROR;
}
@@ -919,7 +897,7 @@
// Handle Flags
if (((flags & VP8_EFLAG_NO_UPD_GF) && (flags & VP8_EFLAG_FORCE_GF)) ||
- ((flags & VP8_EFLAG_NO_UPD_ARF) && (flags & VP8_EFLAG_FORCE_ARF))) {
+ ((flags & VP8_EFLAG_NO_UPD_ARF) && (flags & VP8_EFLAG_FORCE_ARF))) {
ctx->base.err_detail = "Conflicting flags.";
return VPX_CODEC_INVALID_PARAM;
}
@@ -946,16 +924,15 @@
unsigned char *cx_data;
// Set up internal flags
- if (ctx->base.init_flags & VPX_CODEC_USE_PSNR)
- cpi->b_calculate_psnr = 1;
+ if (ctx->base.init_flags & VPX_CODEC_USE_PSNR) cpi->b_calculate_psnr = 1;
if (img != NULL) {
res = image2yuvconfig(img, &sd);
// Store the original flags in to the frame buffer. Will extract the
// key frame flag when we actually encode this frame.
- if (vp10_receive_raw_frame(cpi, flags | ctx->next_frame_flags,
- &sd, dst_time_stamp, dst_end_time_stamp)) {
+ if (vp10_receive_raw_frame(cpi, flags | ctx->next_frame_flags, &sd,
+ dst_time_stamp, dst_end_time_stamp)) {
res = update_error_state(ctx, &cpi->common.error);
}
ctx->next_frame_flags = 0;
@@ -981,16 +958,15 @@
}
while (cx_data_sz >= ctx->cx_data_sz / 2 &&
- -1 != vp10_get_compressed_data(cpi, &lib_flags, &size,
- cx_data, &dst_time_stamp,
- &dst_end_time_stamp, !img)) {
+ -1 != vp10_get_compressed_data(cpi, &lib_flags, &size, cx_data,
+ &dst_time_stamp, &dst_end_time_stamp,
+ !img)) {
if (size) {
vpx_codec_cx_pkt_t pkt;
// Pack invisible frames with the next visible frame
if (!cpi->common.show_frame) {
- if (ctx->pending_cx_data == 0)
- ctx->pending_cx_data = cx_data;
+ if (ctx->pending_cx_data == 0) ctx->pending_cx_data = cx_data;
ctx->pending_cx_data_sz += size;
ctx->pending_frame_sizes[ctx->pending_frame_count++] = size;
#if !CONFIG_MISC_FIXES
@@ -1001,14 +977,13 @@
if (ctx->output_cx_pkt_cb.output_cx_pkt) {
pkt.kind = VPX_CODEC_CX_FRAME_PKT;
- pkt.data.frame.pts = ticks_to_timebase_units(timebase,
- dst_time_stamp);
- pkt.data.frame.duration =
- (unsigned long)ticks_to_timebase_units(timebase,
- dst_end_time_stamp - dst_time_stamp);
+ pkt.data.frame.pts =
+ ticks_to_timebase_units(timebase, dst_time_stamp);
+ pkt.data.frame.duration = (unsigned long)ticks_to_timebase_units(
+ timebase, dst_end_time_stamp - dst_time_stamp);
pkt.data.frame.flags = get_frame_pkt_flags(cpi, lib_flags);
pkt.data.frame.buf = ctx->pending_cx_data;
- pkt.data.frame.sz = size;
+ pkt.data.frame.sz = size;
ctx->pending_cx_data = NULL;
ctx->pending_cx_data_sz = 0;
ctx->pending_frame_count = 0;
@@ -1024,9 +999,8 @@
// Add the frame packet to the list of returned packets.
pkt.kind = VPX_CODEC_CX_FRAME_PKT;
pkt.data.frame.pts = ticks_to_timebase_units(timebase, dst_time_stamp);
- pkt.data.frame.duration =
- (unsigned long)ticks_to_timebase_units(timebase,
- dst_end_time_stamp - dst_time_stamp);
+ pkt.data.frame.duration = (unsigned long)ticks_to_timebase_units(
+ timebase, dst_end_time_stamp - dst_time_stamp);
pkt.data.frame.flags = get_frame_pkt_flags(cpi, lib_flags);
if (ctx->pending_cx_data) {
@@ -1039,7 +1013,7 @@
if (!ctx->output_cx_pkt_cb.output_cx_pkt)
size += write_superframe_index(ctx);
pkt.data.frame.buf = ctx->pending_cx_data;
- pkt.data.frame.sz = ctx->pending_cx_data_sz;
+ pkt.data.frame.sz = ctx->pending_cx_data_sz;
ctx->pending_cx_data = NULL;
ctx->pending_cx_data_sz = 0;
ctx->pending_frame_count = 0;
@@ -1048,11 +1022,11 @@
#endif
} else {
pkt.data.frame.buf = cx_data;
- pkt.data.frame.sz = size;
+ pkt.data.frame.sz = size;
}
pkt.data.frame.partition_id = -1;
- if(ctx->output_cx_pkt_cb.output_cx_pkt)
+ if (ctx->output_cx_pkt_cb.output_cx_pkt)
ctx->output_cx_pkt_cb.output_cx_pkt(&pkt,
ctx->output_cx_pkt_cb.user_priv);
else
@@ -1080,8 +1054,8 @@
YV12_BUFFER_CONFIG sd;
image2yuvconfig(&frame->img, &sd);
- vp10_set_reference_enc(ctx->cpi, ref_frame_to_vp10_reframe(frame->frame_type),
- &sd);
+ vp10_set_reference_enc(ctx->cpi,
+ ref_frame_to_vp10_reframe(frame->frame_type), &sd);
return VPX_CODEC_OK;
} else {
return VPX_CODEC_INVALID_PARAM;
@@ -1097,7 +1071,7 @@
image2yuvconfig(&frame->img, &sd);
vp10_copy_reference_enc(ctx->cpi,
- ref_frame_to_vp10_reframe(frame->frame_type), &sd);
+ ref_frame_to_vp10_reframe(frame->frame_type), &sd);
return VPX_CODEC_OK;
} else {
return VPX_CODEC_INVALID_PARAM;
@@ -1126,7 +1100,6 @@
return VPX_CODEC_INCAPABLE;
}
-
static vpx_image_t *encoder_get_preview(vpx_codec_alg_priv_t *ctx) {
YV12_BUFFER_CONFIG sd;
@@ -1147,14 +1120,13 @@
return VPX_CODEC_INVALID_PARAM;
}
-
static vpx_codec_err_t ctrl_set_active_map(vpx_codec_alg_priv_t *ctx,
va_list args) {
vpx_active_map_t *const map = va_arg(args, vpx_active_map_t *);
if (map) {
- if (!vp10_set_active_map(ctx->cpi, map->active_map,
- (int)map->rows, (int)map->cols))
+ if (!vp10_set_active_map(ctx->cpi, map->active_map, (int)map->rows,
+ (int)map->cols))
return VPX_CODEC_OK;
else
return VPX_CODEC_INVALID_PARAM;
@@ -1168,8 +1140,8 @@
vpx_active_map_t *const map = va_arg(args, vpx_active_map_t *);
if (map) {
- if (!vp10_get_active_map(ctx->cpi, map->active_map,
- (int)map->rows, (int)map->cols))
+ if (!vp10_get_active_map(ctx->cpi, map->active_map, (int)map->rows,
+ (int)map->cols))
return VPX_CODEC_OK;
else
return VPX_CODEC_INVALID_PARAM;
@@ -1183,9 +1155,9 @@
vpx_scaling_mode_t *const mode = va_arg(args, vpx_scaling_mode_t *);
if (mode) {
- const int res = vp10_set_internal_size(ctx->cpi,
- (VPX_SCALING)mode->h_scaling_mode,
- (VPX_SCALING)mode->v_scaling_mode);
+ const int res =
+ vp10_set_internal_size(ctx->cpi, (VPX_SCALING)mode->h_scaling_mode,
+ (VPX_SCALING)mode->v_scaling_mode);
return (res == 0) ? VPX_CODEC_OK : VPX_CODEC_INVALID_PARAM;
} else {
return VPX_CODEC_INVALID_PARAM;
@@ -1227,121 +1199,120 @@
va_list args) {
struct vp10_extracfg extra_cfg = ctx->extra_cfg;
int *const render_size = va_arg(args, int *);
- extra_cfg.render_width = render_size[0];
+ extra_cfg.render_width = render_size[0];
extra_cfg.render_height = render_size[1];
return update_extra_cfg(ctx, &extra_cfg);
}
static vpx_codec_ctrl_fn_map_t encoder_ctrl_maps[] = {
- {VP8_COPY_REFERENCE, ctrl_copy_reference},
+ { VP8_COPY_REFERENCE, ctrl_copy_reference },
// Setters
- {VP8_SET_REFERENCE, ctrl_set_reference},
- {VP8_SET_POSTPROC, ctrl_set_previewpp},
- {VP8E_SET_ROI_MAP, ctrl_set_roi_map},
- {VP8E_SET_ACTIVEMAP, ctrl_set_active_map},
- {VP8E_SET_SCALEMODE, ctrl_set_scale_mode},
- {VP8E_SET_CPUUSED, ctrl_set_cpuused},
- {VP8E_SET_ENABLEAUTOALTREF, ctrl_set_enable_auto_alt_ref},
- {VP8E_SET_SHARPNESS, ctrl_set_sharpness},
- {VP8E_SET_STATIC_THRESHOLD, ctrl_set_static_thresh},
- {VP9E_SET_TILE_COLUMNS, ctrl_set_tile_columns},
- {VP9E_SET_TILE_ROWS, ctrl_set_tile_rows},
- {VP8E_SET_ARNR_MAXFRAMES, ctrl_set_arnr_max_frames},
- {VP8E_SET_ARNR_STRENGTH, ctrl_set_arnr_strength},
- {VP8E_SET_ARNR_TYPE, ctrl_set_arnr_type},
- {VP8E_SET_TUNING, ctrl_set_tuning},
- {VP8E_SET_CQ_LEVEL, ctrl_set_cq_level},
- {VP8E_SET_MAX_INTRA_BITRATE_PCT, ctrl_set_rc_max_intra_bitrate_pct},
- {VP9E_SET_MAX_INTER_BITRATE_PCT, ctrl_set_rc_max_inter_bitrate_pct},
- {VP9E_SET_GF_CBR_BOOST_PCT, ctrl_set_rc_gf_cbr_boost_pct},
- {VP9E_SET_LOSSLESS, ctrl_set_lossless},
- {VP9E_SET_FRAME_PARALLEL_DECODING, ctrl_set_frame_parallel_decoding_mode},
- {VP9E_SET_AQ_MODE, ctrl_set_aq_mode},
- {VP9E_SET_FRAME_PERIODIC_BOOST, ctrl_set_frame_periodic_boost},
- {VP9E_REGISTER_CX_CALLBACK, ctrl_register_cx_callback},
- {VP9E_SET_TUNE_CONTENT, ctrl_set_tune_content},
- {VP9E_SET_COLOR_SPACE, ctrl_set_color_space},
- {VP9E_SET_COLOR_RANGE, ctrl_set_color_range},
- {VP9E_SET_NOISE_SENSITIVITY, ctrl_set_noise_sensitivity},
- {VP9E_SET_MIN_GF_INTERVAL, ctrl_set_min_gf_interval},
- {VP9E_SET_MAX_GF_INTERVAL, ctrl_set_max_gf_interval},
- {VP9E_SET_RENDER_SIZE, ctrl_set_render_size},
+ { VP8_SET_REFERENCE, ctrl_set_reference },
+ { VP8_SET_POSTPROC, ctrl_set_previewpp },
+ { VP8E_SET_ROI_MAP, ctrl_set_roi_map },
+ { VP8E_SET_ACTIVEMAP, ctrl_set_active_map },
+ { VP8E_SET_SCALEMODE, ctrl_set_scale_mode },
+ { VP8E_SET_CPUUSED, ctrl_set_cpuused },
+ { VP8E_SET_ENABLEAUTOALTREF, ctrl_set_enable_auto_alt_ref },
+ { VP8E_SET_SHARPNESS, ctrl_set_sharpness },
+ { VP8E_SET_STATIC_THRESHOLD, ctrl_set_static_thresh },
+ { VP9E_SET_TILE_COLUMNS, ctrl_set_tile_columns },
+ { VP9E_SET_TILE_ROWS, ctrl_set_tile_rows },
+ { VP8E_SET_ARNR_MAXFRAMES, ctrl_set_arnr_max_frames },
+ { VP8E_SET_ARNR_STRENGTH, ctrl_set_arnr_strength },
+ { VP8E_SET_ARNR_TYPE, ctrl_set_arnr_type },
+ { VP8E_SET_TUNING, ctrl_set_tuning },
+ { VP8E_SET_CQ_LEVEL, ctrl_set_cq_level },
+ { VP8E_SET_MAX_INTRA_BITRATE_PCT, ctrl_set_rc_max_intra_bitrate_pct },
+ { VP9E_SET_MAX_INTER_BITRATE_PCT, ctrl_set_rc_max_inter_bitrate_pct },
+ { VP9E_SET_GF_CBR_BOOST_PCT, ctrl_set_rc_gf_cbr_boost_pct },
+ { VP9E_SET_LOSSLESS, ctrl_set_lossless },
+ { VP9E_SET_FRAME_PARALLEL_DECODING, ctrl_set_frame_parallel_decoding_mode },
+ { VP9E_SET_AQ_MODE, ctrl_set_aq_mode },
+ { VP9E_SET_FRAME_PERIODIC_BOOST, ctrl_set_frame_periodic_boost },
+ { VP9E_REGISTER_CX_CALLBACK, ctrl_register_cx_callback },
+ { VP9E_SET_TUNE_CONTENT, ctrl_set_tune_content },
+ { VP9E_SET_COLOR_SPACE, ctrl_set_color_space },
+ { VP9E_SET_COLOR_RANGE, ctrl_set_color_range },
+ { VP9E_SET_NOISE_SENSITIVITY, ctrl_set_noise_sensitivity },
+ { VP9E_SET_MIN_GF_INTERVAL, ctrl_set_min_gf_interval },
+ { VP9E_SET_MAX_GF_INTERVAL, ctrl_set_max_gf_interval },
+ { VP9E_SET_RENDER_SIZE, ctrl_set_render_size },
// Getters
- {VP8E_GET_LAST_QUANTIZER, ctrl_get_quantizer},
- {VP8E_GET_LAST_QUANTIZER_64, ctrl_get_quantizer64},
- {VP9_GET_REFERENCE, ctrl_get_reference},
- {VP9E_GET_ACTIVEMAP, ctrl_get_active_map},
+ { VP8E_GET_LAST_QUANTIZER, ctrl_get_quantizer },
+ { VP8E_GET_LAST_QUANTIZER_64, ctrl_get_quantizer64 },
+ { VP9_GET_REFERENCE, ctrl_get_reference },
+ { VP9E_GET_ACTIVEMAP, ctrl_get_active_map },
- { -1, NULL},
+ { -1, NULL },
};
static vpx_codec_enc_cfg_map_t encoder_usage_cfg_map[] = {
- {
- 0,
- { // NOLINT
- 0, // g_usage
- 8, // g_threads
- 0, // g_profile
+ { 0,
+ {
+ // NOLINT
+ 0, // g_usage
+ 8, // g_threads
+ 0, // g_profile
- 320, // g_width
- 240, // g_height
- VPX_BITS_8, // g_bit_depth
- 8, // g_input_bit_depth
+ 320, // g_width
+ 240, // g_height
+ VPX_BITS_8, // g_bit_depth
+ 8, // g_input_bit_depth
- {1, 30}, // g_timebase
+ { 1, 30 }, // g_timebase
- 0, // g_error_resilient
+ 0, // g_error_resilient
- VPX_RC_ONE_PASS, // g_pass
+ VPX_RC_ONE_PASS, // g_pass
- 25, // g_lag_in_frames
+ 25, // g_lag_in_frames
- 0, // rc_dropframe_thresh
- 0, // rc_resize_allowed
- 0, // rc_scaled_width
- 0, // rc_scaled_height
- 60, // rc_resize_down_thresold
- 30, // rc_resize_up_thresold
+ 0, // rc_dropframe_thresh
+ 0, // rc_resize_allowed
+ 0, // rc_scaled_width
+ 0, // rc_scaled_height
+ 60, // rc_resize_down_thresold
+ 30, // rc_resize_up_thresold
- VPX_VBR, // rc_end_usage
- {NULL, 0}, // rc_twopass_stats_in
- {NULL, 0}, // rc_firstpass_mb_stats_in
- 256, // rc_target_bandwidth
- 0, // rc_min_quantizer
- 63, // rc_max_quantizer
- 25, // rc_undershoot_pct
- 25, // rc_overshoot_pct
+ VPX_VBR, // rc_end_usage
+ { NULL, 0 }, // rc_twopass_stats_in
+ { NULL, 0 }, // rc_firstpass_mb_stats_in
+ 256, // rc_target_bandwidth
+ 0, // rc_min_quantizer
+ 63, // rc_max_quantizer
+ 25, // rc_undershoot_pct
+ 25, // rc_overshoot_pct
- 6000, // rc_max_buffer_size
- 4000, // rc_buffer_initial_size
- 5000, // rc_buffer_optimal_size
+ 6000, // rc_max_buffer_size
+ 4000, // rc_buffer_initial_size
+ 5000, // rc_buffer_optimal_size
- 50, // rc_two_pass_vbrbias
- 0, // rc_two_pass_vbrmin_section
- 2000, // rc_two_pass_vbrmax_section
+ 50, // rc_two_pass_vbrbias
+ 0, // rc_two_pass_vbrmin_section
+ 2000, // rc_two_pass_vbrmax_section
- // keyframing settings (kf)
- VPX_KF_AUTO, // g_kfmode
- 0, // kf_min_dist
- 9999, // kf_max_dist
+ // keyframing settings (kf)
+ VPX_KF_AUTO, // g_kfmode
+ 0, // kf_min_dist
+ 9999, // kf_max_dist
- // TODO(yunqingwang): Spatial/temporal scalability are not supported
- // in VP10. The following 10 parameters are not used, which should
- // be removed later.
- 1, // ss_number_layers
- {0},
- {0}, // ss_target_bitrate
- 1, // ts_number_layers
- {0}, // ts_target_bitrate
- {0}, // ts_rate_decimator
- 0, // ts_periodicity
- {0}, // ts_layer_id
- {0}, // layer_taget_bitrate
- 0 // temporal_layering_mode
- }
- },
+ // TODO(yunqingwang): Spatial/temporal scalability are not supported
+ // in VP10. The following 10 parameters are not used, which should
+ // be removed later.
+ 1, // ss_number_layers
+ { 0 },
+ { 0 }, // ss_target_bitrate
+ 1, // ts_number_layers
+ { 0 }, // ts_target_bitrate
+ { 0 }, // ts_rate_decimator
+ 0, // ts_periodicity
+ { 0 }, // ts_layer_id
+ { 0 }, // layer_taget_bitrate
+ 0 // temporal_layering_mode
+ } },
};
#ifndef VERSION_STRING
@@ -1353,25 +1324,27 @@
#if CONFIG_VPX_HIGHBITDEPTH
VPX_CODEC_CAP_HIGHBITDEPTH |
#endif
- VPX_CODEC_CAP_ENCODER | VPX_CODEC_CAP_PSNR, // vpx_codec_caps_t
- encoder_init, // vpx_codec_init_fn_t
- encoder_destroy, // vpx_codec_destroy_fn_t
- encoder_ctrl_maps, // vpx_codec_ctrl_fn_map_t
- { // NOLINT
- NULL, // vpx_codec_peek_si_fn_t
- NULL, // vpx_codec_get_si_fn_t
- NULL, // vpx_codec_decode_fn_t
- NULL, // vpx_codec_frame_get_fn_t
- NULL // vpx_codec_set_fb_fn_t
+ VPX_CODEC_CAP_ENCODER | VPX_CODEC_CAP_PSNR, // vpx_codec_caps_t
+ encoder_init, // vpx_codec_init_fn_t
+ encoder_destroy, // vpx_codec_destroy_fn_t
+ encoder_ctrl_maps, // vpx_codec_ctrl_fn_map_t
+ {
+ // NOLINT
+ NULL, // vpx_codec_peek_si_fn_t
+ NULL, // vpx_codec_get_si_fn_t
+ NULL, // vpx_codec_decode_fn_t
+ NULL, // vpx_codec_frame_get_fn_t
+ NULL // vpx_codec_set_fb_fn_t
},
- { // NOLINT
- 1, // 1 cfg map
- encoder_usage_cfg_map, // vpx_codec_enc_cfg_map_t
- encoder_encode, // vpx_codec_encode_fn_t
- encoder_get_cxdata, // vpx_codec_get_cx_data_fn_t
- encoder_set_config, // vpx_codec_enc_config_set_fn_t
- NULL, // vpx_codec_get_global_headers_fn_t
- encoder_get_preview, // vpx_codec_get_preview_frame_fn_t
- NULL // vpx_codec_enc_mr_get_mem_loc_fn_t
+ {
+ // NOLINT
+ 1, // 1 cfg map
+ encoder_usage_cfg_map, // vpx_codec_enc_cfg_map_t
+ encoder_encode, // vpx_codec_encode_fn_t
+ encoder_get_cxdata, // vpx_codec_get_cx_data_fn_t
+ encoder_set_config, // vpx_codec_enc_config_set_fn_t
+ NULL, // vpx_codec_get_global_headers_fn_t
+ encoder_get_preview, // vpx_codec_get_preview_frame_fn_t
+ NULL // vpx_codec_enc_mr_get_mem_loc_fn_t
}
};
diff --git a/vp10/vp10_dx_iface.c b/vp10/vp10_dx_iface.c
index 444c282..23ba863 100644
--- a/vp10/vp10_dx_iface.c
+++ b/vp10/vp10_dx_iface.c
@@ -33,7 +33,7 @@
// This limit is due to framebuffer numbers.
// TODO(hkuang): Remove this limit after implementing ondemand framebuffers.
-#define FRAME_CACHE_SIZE 6 // Cache maximum 6 decoded frames.
+#define FRAME_CACHE_SIZE 6 // Cache maximum 6 decoded frames.
typedef struct cache_frame {
int fb_idx;
@@ -41,36 +41,36 @@
} cache_frame;
struct vpx_codec_alg_priv {
- vpx_codec_priv_t base;
- vpx_codec_dec_cfg_t cfg;
- vp10_stream_info_t si;
- int postproc_cfg_set;
- vp8_postproc_cfg_t postproc_cfg;
- vpx_decrypt_cb decrypt_cb;
- void *decrypt_state;
- vpx_image_t img;
- int img_avail;
- int flushed;
- int invert_tile_order;
- int last_show_frame; // Index of last output frame.
- int byte_alignment;
- int skip_loop_filter;
+ vpx_codec_priv_t base;
+ vpx_codec_dec_cfg_t cfg;
+ vp10_stream_info_t si;
+ int postproc_cfg_set;
+ vp8_postproc_cfg_t postproc_cfg;
+ vpx_decrypt_cb decrypt_cb;
+ void *decrypt_state;
+ vpx_image_t img;
+ int img_avail;
+ int flushed;
+ int invert_tile_order;
+ int last_show_frame; // Index of last output frame.
+ int byte_alignment;
+ int skip_loop_filter;
// Frame parallel related.
- int frame_parallel_decode; // frame-based threading.
- VPxWorker *frame_workers;
- int num_frame_workers;
- int next_submit_worker_id;
- int last_submit_worker_id;
- int next_output_worker_id;
- int available_threads;
- cache_frame frame_cache[FRAME_CACHE_SIZE];
- int frame_cache_write;
- int frame_cache_read;
- int num_cache_frames;
- int need_resync; // wait for key/intra-only frame
+ int frame_parallel_decode; // frame-based threading.
+ VPxWorker *frame_workers;
+ int num_frame_workers;
+ int next_submit_worker_id;
+ int last_submit_worker_id;
+ int next_output_worker_id;
+ int available_threads;
+ cache_frame frame_cache[FRAME_CACHE_SIZE];
+ int frame_cache_write;
+ int frame_cache_read;
+ int num_cache_frames;
+ int need_resync; // wait for key/intra-only frame
// BufferPool that holds all reference frames. Shared by all the FrameWorkers.
- BufferPool *buffer_pool;
+ BufferPool *buffer_pool;
// External frame buffer info to save for VP10 common.
void *ext_priv; // Private data associated with the external frame buffers.
@@ -88,8 +88,7 @@
if (!ctx->priv) {
vpx_codec_alg_priv_t *const priv =
(vpx_codec_alg_priv_t *)vpx_calloc(1, sizeof(*priv));
- if (priv == NULL)
- return VPX_CODEC_MEM_ERROR;
+ if (priv == NULL) return VPX_CODEC_MEM_ERROR;
ctx->priv = (vpx_codec_priv_t *)priv;
ctx->priv->init_flags = ctx->init_flags;
@@ -98,7 +97,9 @@
// Only do frame parallel decode when threads > 1.
priv->frame_parallel_decode =
(ctx->config.dec && (ctx->config.dec->threads > 1) &&
- (ctx->init_flags & VPX_CODEC_USE_FRAME_THREADING)) ? 1 : 0;
+ (ctx->init_flags & VPX_CODEC_USE_FRAME_THREADING))
+ ? 1
+ : 0;
if (ctx->config.dec) {
priv->cfg = *ctx->config.dec;
ctx->config.dec = &priv->cfg;
@@ -141,11 +142,10 @@
return VPX_CODEC_OK;
}
-static int parse_bitdepth_colorspace_sampling(
- BITSTREAM_PROFILE profile, struct vpx_read_bit_buffer *rb) {
+static int parse_bitdepth_colorspace_sampling(BITSTREAM_PROFILE profile,
+ struct vpx_read_bit_buffer *rb) {
vpx_color_space_t color_space;
- if (profile >= PROFILE_2)
- rb->bit_offset += 1; // Bit-depth 10 or 12.
+ if (profile >= PROFILE_2) rb->bit_offset += 1; // Bit-depth 10 or 12.
color_space = (vpx_color_space_t)vpx_rb_read_literal(rb, 3);
if (color_space != VPX_CS_SRGB) {
rb->bit_offset += 1; // [16,235] (including xvycc) vs [0,255] range.
@@ -164,17 +164,13 @@
return 1;
}
-static vpx_codec_err_t decoder_peek_si_internal(const uint8_t *data,
- unsigned int data_sz,
- vpx_codec_stream_info_t *si,
- int *is_intra_only,
- vpx_decrypt_cb decrypt_cb,
- void *decrypt_state) {
+static vpx_codec_err_t decoder_peek_si_internal(
+ const uint8_t *data, unsigned int data_sz, vpx_codec_stream_info_t *si,
+ int *is_intra_only, vpx_decrypt_cb decrypt_cb, void *decrypt_state) {
int intra_only_flag = 0;
uint8_t clear_buffer[9];
- if (data + data_sz <= data)
- return VPX_CODEC_INVALID_PARAM;
+ if (data + data_sz <= data) return VPX_CODEC_INVALID_PARAM;
si->is_kf = 0;
si->w = si->h = 0;
@@ -192,30 +188,26 @@
const int frame_marker = vpx_rb_read_literal(&rb, 2);
const BITSTREAM_PROFILE profile = vp10_read_profile(&rb);
- if (frame_marker != VPX_FRAME_MARKER)
- return VPX_CODEC_UNSUP_BITSTREAM;
+ if (frame_marker != VPX_FRAME_MARKER) return VPX_CODEC_UNSUP_BITSTREAM;
- if (profile >= MAX_PROFILES)
- return VPX_CODEC_UNSUP_BITSTREAM;
+ if (profile >= MAX_PROFILES) return VPX_CODEC_UNSUP_BITSTREAM;
if ((profile >= 2 && data_sz <= 1) || data_sz < 1)
return VPX_CODEC_UNSUP_BITSTREAM;
- if (vpx_rb_read_bit(&rb)) { // show an existing frame
+ if (vpx_rb_read_bit(&rb)) { // show an existing frame
vpx_rb_read_literal(&rb, 3); // Frame buffer to show.
return VPX_CODEC_OK;
}
- if (data_sz <= 8)
- return VPX_CODEC_UNSUP_BITSTREAM;
+ if (data_sz <= 8) return VPX_CODEC_UNSUP_BITSTREAM;
si->is_kf = !vpx_rb_read_bit(&rb);
show_frame = vpx_rb_read_bit(&rb);
error_resilient = vpx_rb_read_bit(&rb);
if (si->is_kf) {
- if (!vp10_read_sync_code(&rb))
- return VPX_CODEC_UNSUP_BITSTREAM;
+ if (!vp10_read_sync_code(&rb)) return VPX_CODEC_UNSUP_BITSTREAM;
if (!parse_bitdepth_colorspace_sampling(profile, &rb))
return VPX_CODEC_UNSUP_BITSTREAM;
@@ -226,8 +218,7 @@
rb.bit_offset += error_resilient ? 0 : 2; // reset_frame_context
if (intra_only_flag) {
- if (!vp10_read_sync_code(&rb))
- return VPX_CODEC_UNSUP_BITSTREAM;
+ if (!vp10_read_sync_code(&rb)) return VPX_CODEC_UNSUP_BITSTREAM;
if (profile > PROFILE_0) {
if (!parse_bitdepth_colorspace_sampling(profile, &rb))
return VPX_CODEC_UNSUP_BITSTREAM;
@@ -237,8 +228,7 @@
}
}
}
- if (is_intra_only != NULL)
- *is_intra_only = intra_only_flag;
+ if (is_intra_only != NULL) *is_intra_only = intra_only_flag;
return VPX_CODEC_OK;
}
@@ -251,8 +241,8 @@
static vpx_codec_err_t decoder_get_si(vpx_codec_alg_priv_t *ctx,
vpx_codec_stream_info_t *si) {
const size_t sz = (si->sz >= sizeof(vp10_stream_info_t))
- ? sizeof(vp10_stream_info_t)
- : sizeof(vpx_codec_stream_info_t);
+ ? sizeof(vp10_stream_info_t)
+ : sizeof(vpx_codec_stream_info_t);
memcpy(si, &ctx->si, sz);
si->sz = (unsigned int)sz;
@@ -264,8 +254,8 @@
ctx->base.err_detail = error;
}
-static vpx_codec_err_t update_error_state(vpx_codec_alg_priv_t *ctx,
- const struct vpx_internal_error_info *error) {
+static vpx_codec_err_t update_error_state(
+ vpx_codec_alg_priv_t *ctx, const struct vpx_internal_error_info *error) {
if (error->error_code)
set_error_detail(ctx, error->has_detail ? error->detail : NULL);
@@ -313,10 +303,8 @@
const uint8_t *data = frame_worker_data->data;
(void)arg2;
- frame_worker_data->result =
- vp10_receive_compressed_data(frame_worker_data->pbi,
- frame_worker_data->data_size,
- &data);
+ frame_worker_data->result = vp10_receive_compressed_data(
+ frame_worker_data->pbi, frame_worker_data->data_size, &data);
frame_worker_data->data_end = data;
if (frame_worker_data->pbi->common.frame_parallel_decode) {
@@ -358,25 +346,24 @@
ctx->num_cache_frames = 0;
ctx->need_resync = 1;
ctx->num_frame_workers =
- (ctx->frame_parallel_decode == 1) ? ctx->cfg.threads: 1;
+ (ctx->frame_parallel_decode == 1) ? ctx->cfg.threads : 1;
if (ctx->num_frame_workers > MAX_DECODE_THREADS)
ctx->num_frame_workers = MAX_DECODE_THREADS;
ctx->available_threads = ctx->num_frame_workers;
ctx->flushed = 0;
ctx->buffer_pool = (BufferPool *)vpx_calloc(1, sizeof(BufferPool));
- if (ctx->buffer_pool == NULL)
- return VPX_CODEC_MEM_ERROR;
+ if (ctx->buffer_pool == NULL) return VPX_CODEC_MEM_ERROR;
#if CONFIG_MULTITHREAD
- if (pthread_mutex_init(&ctx->buffer_pool->pool_mutex, NULL)) {
- set_error_detail(ctx, "Failed to allocate buffer pool mutex");
- return VPX_CODEC_MEM_ERROR;
- }
+ if (pthread_mutex_init(&ctx->buffer_pool->pool_mutex, NULL)) {
+ set_error_detail(ctx, "Failed to allocate buffer pool mutex");
+ return VPX_CODEC_MEM_ERROR;
+ }
#endif
- ctx->frame_workers = (VPxWorker *)
- vpx_malloc(ctx->num_frame_workers * sizeof(*ctx->frame_workers));
+ ctx->frame_workers = (VPxWorker *)vpx_malloc(ctx->num_frame_workers *
+ sizeof(*ctx->frame_workers));
if (ctx->frame_workers == NULL) {
set_error_detail(ctx, "Failed to allocate frame_workers");
return VPX_CODEC_MEM_ERROR;
@@ -431,8 +418,7 @@
// If postprocessing was enabled by the application and a
// configuration has not been provided, default it.
- if (!ctx->postproc_cfg_set &&
- (ctx->base.init_flags & VPX_CODEC_USE_POSTPROC))
+ if (!ctx->postproc_cfg_set && (ctx->base.init_flags & VPX_CODEC_USE_POSTPROC))
set_default_ppflags(&ctx->postproc_cfg);
init_buffer_callbacks(ctx);
@@ -462,11 +448,9 @@
const vpx_codec_err_t res =
decoder_peek_si_internal(*data, data_sz, &ctx->si, &is_intra_only,
ctx->decrypt_cb, ctx->decrypt_state);
- if (res != VPX_CODEC_OK)
- return res;
+ if (res != VPX_CODEC_OK) return res;
- if (!ctx->si.is_kf && !is_intra_only)
- return VPX_CODEC_ERROR;
+ if (!ctx->si.is_kf && !is_intra_only) return VPX_CODEC_ERROR;
}
if (!ctx->frame_parallel_decode) {
@@ -560,8 +544,7 @@
frame_worker_data->user_priv);
ctx->frame_cache[ctx->frame_cache_write].img.fb_priv =
frame_bufs[cm->new_fb_idx].raw_frame_buffer.priv;
- ctx->frame_cache_write =
- (ctx->frame_cache_write + 1) % FRAME_CACHE_SIZE;
+ ctx->frame_cache_write = (ctx->frame_cache_write + 1) % FRAME_CACHE_SIZE;
++ctx->num_cache_frames;
}
}
@@ -570,7 +553,7 @@
const uint8_t *data, unsigned int data_sz,
void *user_priv, long deadline) {
const uint8_t *data_start = data;
- const uint8_t * const data_end = data + data_sz;
+ const uint8_t *const data_end = data + data_sz;
vpx_codec_err_t res;
uint32_t frame_sizes[8];
int frame_count;
@@ -586,14 +569,12 @@
// Initialize the decoder workers on the first frame.
if (ctx->frame_workers == NULL) {
const vpx_codec_err_t res = init_decoder(ctx);
- if (res != VPX_CODEC_OK)
- return res;
+ if (res != VPX_CODEC_OK) return res;
}
res = vp10_parse_superframe_index(data, data_sz, frame_sizes, &frame_count,
- ctx->decrypt_cb, ctx->decrypt_state);
- if (res != VPX_CODEC_OK)
- return res;
+ ctx->decrypt_cb, ctx->decrypt_state);
+ if (res != VPX_CODEC_OK) return res;
if (ctx->frame_parallel_decode) {
// Decode in frame parallel mode. When decoding in this mode, the frame
@@ -606,8 +587,8 @@
for (i = 0; i < frame_count; ++i) {
const uint8_t *data_start_copy = data_start;
const uint32_t frame_size = frame_sizes[i];
- if (data_start < data
- || frame_size > (uint32_t) (data_end - data_start)) {
+ if (data_start < data ||
+ frame_size > (uint32_t)(data_end - data_start)) {
set_error_detail(ctx, "Invalid frame size in index");
return VPX_CODEC_CORRUPT_FRAME;
}
@@ -624,10 +605,9 @@
}
}
- res = decode_one(ctx, &data_start_copy, frame_size, user_priv,
- deadline);
- if (res != VPX_CODEC_OK)
- return res;
+ res =
+ decode_one(ctx, &data_start_copy, frame_size, user_priv, deadline);
+ if (res != VPX_CODEC_OK) return res;
data_start += frame_size;
}
} else {
@@ -644,8 +624,7 @@
}
res = decode_one(ctx, &data, data_sz, user_priv, deadline);
- if (res != VPX_CODEC_OK)
- return res;
+ if (res != VPX_CODEC_OK) return res;
}
} else {
// Decode in serial mode.
@@ -656,33 +635,30 @@
const uint8_t *data_start_copy = data_start;
const uint32_t frame_size = frame_sizes[i];
vpx_codec_err_t res;
- if (data_start < data
- || frame_size > (uint32_t) (data_end - data_start)) {
+ if (data_start < data ||
+ frame_size > (uint32_t)(data_end - data_start)) {
set_error_detail(ctx, "Invalid frame size in index");
return VPX_CODEC_CORRUPT_FRAME;
}
- res = decode_one(ctx, &data_start_copy, frame_size, user_priv,
- deadline);
- if (res != VPX_CODEC_OK)
- return res;
+ res =
+ decode_one(ctx, &data_start_copy, frame_size, user_priv, deadline);
+ if (res != VPX_CODEC_OK) return res;
data_start += frame_size;
}
} else {
while (data_start < data_end) {
- const uint32_t frame_size = (uint32_t) (data_end - data_start);
- const vpx_codec_err_t res = decode_one(ctx, &data_start, frame_size,
- user_priv, deadline);
- if (res != VPX_CODEC_OK)
- return res;
+ const uint32_t frame_size = (uint32_t)(data_end - data_start);
+ const vpx_codec_err_t res =
+ decode_one(ctx, &data_start, frame_size, user_priv, deadline);
+ if (res != VPX_CODEC_OK) return res;
// Account for suboptimal termination by the encoder.
while (data_start < data_end) {
- const uint8_t marker = read_marker(ctx->decrypt_cb,
- ctx->decrypt_state, data_start);
- if (marker)
- break;
+ const uint8_t marker =
+ read_marker(ctx->decrypt_cb, ctx->decrypt_state, data_start);
+ if (marker) break;
++data_start;
}
}
@@ -717,9 +693,8 @@
// Output the frames in the cache first.
if (ctx->num_cache_frames > 0) {
release_last_output_frame(ctx);
- ctx->last_show_frame = ctx->frame_cache[ctx->frame_cache_read].fb_idx;
- if (ctx->need_resync)
- return NULL;
+ ctx->last_show_frame = ctx->frame_cache[ctx->frame_cache_read].fb_idx;
+ if (ctx->need_resync) return NULL;
img = &ctx->frame_cache[ctx->frame_cache_read].img;
ctx->frame_cache_read = (ctx->frame_cache_read + 1) % FRAME_CACHE_SIZE;
--ctx->num_cache_frames;
@@ -732,8 +707,7 @@
do {
YV12_BUFFER_CONFIG sd;
const VPxWorkerInterface *const winterface = vpx_get_worker_interface();
- VPxWorker *const worker =
- &ctx->frame_workers[ctx->next_output_worker_id];
+ VPxWorker *const worker = &ctx->frame_workers[ctx->next_output_worker_id];
FrameWorkerData *const frame_worker_data =
(FrameWorkerData *)worker->data1;
ctx->next_output_worker_id =
@@ -751,8 +725,7 @@
RefCntBuffer *const frame_bufs = cm->buffer_pool->frame_bufs;
release_last_output_frame(ctx);
ctx->last_show_frame = frame_worker_data->pbi->common.new_fb_idx;
- if (ctx->need_resync)
- return NULL;
+ if (ctx->need_resync) return NULL;
yuvconfig2image(&ctx->img, &sd, frame_worker_data->user_priv);
ctx->img.fb_priv = frame_bufs[cm->new_fb_idx].raw_frame_buffer.priv;
img = &ctx->img;
@@ -763,8 +736,7 @@
frame_worker_data->received_frame = 0;
++ctx->available_threads;
ctx->need_resync = 1;
- if (ctx->flushed != 1)
- return NULL;
+ if (ctx->flushed != 1) return NULL;
}
} while (ctx->next_output_worker_id != ctx->next_submit_worker_id);
}
@@ -772,8 +744,7 @@
}
static vpx_codec_err_t decoder_set_fb_fn(
- vpx_codec_alg_priv_t *ctx,
- vpx_get_frame_buffer_cb_fn_t cb_get,
+ vpx_codec_alg_priv_t *ctx, vpx_get_frame_buffer_cb_fn_t cb_get,
vpx_release_frame_buffer_cb_fn_t cb_release, void *cb_priv) {
if (cb_get == NULL || cb_release == NULL) {
return VPX_CODEC_INVALID_PARAM;
@@ -806,7 +777,7 @@
FrameWorkerData *const frame_worker_data = (FrameWorkerData *)worker->data1;
image2yuvconfig(&frame->img, &sd);
return vp10_set_reference_dec(&frame_worker_data->pbi->common,
- (VPX_REFFRAME)frame->frame_type, &sd);
+ (VPX_REFFRAME)frame->frame_type, &sd);
} else {
return VPX_CODEC_INVALID_PARAM;
}
@@ -823,13 +794,13 @@
}
if (data) {
- vpx_ref_frame_t *frame = (vpx_ref_frame_t *) data;
+ vpx_ref_frame_t *frame = (vpx_ref_frame_t *)data;
YV12_BUFFER_CONFIG sd;
VPxWorker *const worker = ctx->frame_workers;
FrameWorkerData *const frame_worker_data = (FrameWorkerData *)worker->data1;
image2yuvconfig(&frame->img, &sd);
return vp10_copy_reference_dec(frame_worker_data->pbi,
- (VPX_REFFRAME)frame->frame_type, &sd);
+ (VPX_REFFRAME)frame->frame_type, &sd);
} else {
return VPX_CODEC_INVALID_PARAM;
}
@@ -846,7 +817,7 @@
}
if (data) {
- YV12_BUFFER_CONFIG* fb;
+ YV12_BUFFER_CONFIG *fb;
VPxWorker *const worker = ctx->frame_workers;
FrameWorkerData *const frame_worker_data = (FrameWorkerData *)worker->data1;
fb = get_ref_frame(&frame_worker_data->pbi->common, data->idx);
@@ -1025,8 +996,7 @@
ctx->byte_alignment = byte_alignment;
if (ctx->frame_workers) {
VPxWorker *const worker = ctx->frame_workers;
- FrameWorkerData *const frame_worker_data =
- (FrameWorkerData *)worker->data1;
+ FrameWorkerData *const frame_worker_data = (FrameWorkerData *)worker->data1;
frame_worker_data->pbi->common.byte_alignment = byte_alignment;
}
return VPX_CODEC_OK;
@@ -1046,29 +1016,29 @@
}
static vpx_codec_ctrl_fn_map_t decoder_ctrl_maps[] = {
- {VP8_COPY_REFERENCE, ctrl_copy_reference},
+ { VP8_COPY_REFERENCE, ctrl_copy_reference },
// Setters
- {VP8_SET_REFERENCE, ctrl_set_reference},
- {VP8_SET_POSTPROC, ctrl_set_postproc},
- {VP8_SET_DBG_COLOR_REF_FRAME, ctrl_set_dbg_options},
- {VP8_SET_DBG_COLOR_MB_MODES, ctrl_set_dbg_options},
- {VP8_SET_DBG_COLOR_B_MODES, ctrl_set_dbg_options},
- {VP8_SET_DBG_DISPLAY_MV, ctrl_set_dbg_options},
- {VP9_INVERT_TILE_DECODE_ORDER, ctrl_set_invert_tile_order},
- {VPXD_SET_DECRYPTOR, ctrl_set_decryptor},
- {VP9_SET_BYTE_ALIGNMENT, ctrl_set_byte_alignment},
- {VP9_SET_SKIP_LOOP_FILTER, ctrl_set_skip_loop_filter},
+ { VP8_SET_REFERENCE, ctrl_set_reference },
+ { VP8_SET_POSTPROC, ctrl_set_postproc },
+ { VP8_SET_DBG_COLOR_REF_FRAME, ctrl_set_dbg_options },
+ { VP8_SET_DBG_COLOR_MB_MODES, ctrl_set_dbg_options },
+ { VP8_SET_DBG_COLOR_B_MODES, ctrl_set_dbg_options },
+ { VP8_SET_DBG_DISPLAY_MV, ctrl_set_dbg_options },
+ { VP9_INVERT_TILE_DECODE_ORDER, ctrl_set_invert_tile_order },
+ { VPXD_SET_DECRYPTOR, ctrl_set_decryptor },
+ { VP9_SET_BYTE_ALIGNMENT, ctrl_set_byte_alignment },
+ { VP9_SET_SKIP_LOOP_FILTER, ctrl_set_skip_loop_filter },
// Getters
- {VP8D_GET_LAST_REF_UPDATES, ctrl_get_last_ref_updates},
- {VP8D_GET_FRAME_CORRUPTED, ctrl_get_frame_corrupted},
- {VP9_GET_REFERENCE, ctrl_get_reference},
- {VP9D_GET_DISPLAY_SIZE, ctrl_get_render_size},
- {VP9D_GET_BIT_DEPTH, ctrl_get_bit_depth},
- {VP9D_GET_FRAME_SIZE, ctrl_get_frame_size},
+ { VP8D_GET_LAST_REF_UPDATES, ctrl_get_last_ref_updates },
+ { VP8D_GET_FRAME_CORRUPTED, ctrl_get_frame_corrupted },
+ { VP9_GET_REFERENCE, ctrl_get_reference },
+ { VP9D_GET_DISPLAY_SIZE, ctrl_get_render_size },
+ { VP9D_GET_BIT_DEPTH, ctrl_get_bit_depth },
+ { VP9D_GET_FRAME_SIZE, ctrl_get_frame_size },
- { -1, NULL},
+ { -1, NULL },
};
#ifndef VERSION_STRING
@@ -1077,26 +1047,28 @@
CODEC_INTERFACE(vpx_codec_vp10_dx) = {
"WebM Project VP10 Decoder" VERSION_STRING,
VPX_CODEC_INTERNAL_ABI_VERSION,
- VPX_CODEC_CAP_DECODER|
+ VPX_CODEC_CAP_DECODER |
VPX_CODEC_CAP_EXTERNAL_FRAME_BUFFER, // vpx_codec_caps_t
- decoder_init, // vpx_codec_init_fn_t
- decoder_destroy, // vpx_codec_destroy_fn_t
- decoder_ctrl_maps, // vpx_codec_ctrl_fn_map_t
- { // NOLINT
- decoder_peek_si, // vpx_codec_peek_si_fn_t
- decoder_get_si, // vpx_codec_get_si_fn_t
- decoder_decode, // vpx_codec_decode_fn_t
- decoder_get_frame, // vpx_codec_frame_get_fn_t
- decoder_set_fb_fn, // vpx_codec_set_fb_fn_t
+ decoder_init, // vpx_codec_init_fn_t
+ decoder_destroy, // vpx_codec_destroy_fn_t
+ decoder_ctrl_maps, // vpx_codec_ctrl_fn_map_t
+ {
+ // NOLINT
+ decoder_peek_si, // vpx_codec_peek_si_fn_t
+ decoder_get_si, // vpx_codec_get_si_fn_t
+ decoder_decode, // vpx_codec_decode_fn_t
+ decoder_get_frame, // vpx_codec_frame_get_fn_t
+ decoder_set_fb_fn, // vpx_codec_set_fb_fn_t
},
- { // NOLINT
- 0,
- NULL, // vpx_codec_enc_cfg_map_t
- NULL, // vpx_codec_encode_fn_t
- NULL, // vpx_codec_get_cx_data_fn_t
- NULL, // vpx_codec_enc_config_set_fn_t
- NULL, // vpx_codec_get_global_headers_fn_t
- NULL, // vpx_codec_get_preview_frame_fn_t
- NULL // vpx_codec_enc_mr_get_mem_loc_fn_t
+ {
+ // NOLINT
+ 0,
+ NULL, // vpx_codec_enc_cfg_map_t
+ NULL, // vpx_codec_encode_fn_t
+ NULL, // vpx_codec_get_cx_data_fn_t
+ NULL, // vpx_codec_enc_config_set_fn_t
+ NULL, // vpx_codec_get_global_headers_fn_t
+ NULL, // vpx_codec_get_preview_frame_fn_t
+ NULL // vpx_codec_enc_mr_get_mem_loc_fn_t
}
};
diff --git a/vp10/vp10_iface_common.h b/vp10/vp10_iface_common.h
index b875275..b930a2e 100644
--- a/vp10/vp10_iface_common.h
+++ b/vp10/vp10_iface_common.h
@@ -12,7 +12,7 @@
#include "vpx_ports/mem.h"
-static void yuvconfig2image(vpx_image_t *img, const YV12_BUFFER_CONFIG *yv12,
+static void yuvconfig2image(vpx_image_t *img, const YV12_BUFFER_CONFIG *yv12,
void *user_priv) {
/** vpx_img_wrap() doesn't allow specifying independent strides for
* the Y, U, and V planes, nor other alignment adjustments that
@@ -61,9 +61,9 @@
// of the image.
img->fmt = (vpx_img_fmt_t)(img->fmt | VPX_IMG_FMT_HIGHBITDEPTH);
img->bit_depth = yv12->bit_depth;
- img->planes[VPX_PLANE_Y] = (uint8_t*)CONVERT_TO_SHORTPTR(yv12->y_buffer);
- img->planes[VPX_PLANE_U] = (uint8_t*)CONVERT_TO_SHORTPTR(yv12->u_buffer);
- img->planes[VPX_PLANE_V] = (uint8_t*)CONVERT_TO_SHORTPTR(yv12->v_buffer);
+ img->planes[VPX_PLANE_Y] = (uint8_t *)CONVERT_TO_SHORTPTR(yv12->y_buffer);
+ img->planes[VPX_PLANE_U] = (uint8_t *)CONVERT_TO_SHORTPTR(yv12->u_buffer);
+ img->planes[VPX_PLANE_V] = (uint8_t *)CONVERT_TO_SHORTPTR(yv12->v_buffer);
img->planes[VPX_PLANE_ALPHA] = NULL;
img->stride[VPX_PLANE_Y] = 2 * yv12->y_stride;
img->stride[VPX_PLANE_U] = 2 * yv12->uv_stride;
@@ -84,17 +84,17 @@
yv12->u_buffer = img->planes[VPX_PLANE_U];
yv12->v_buffer = img->planes[VPX_PLANE_V];
- yv12->y_crop_width = img->d_w;
+ yv12->y_crop_width = img->d_w;
yv12->y_crop_height = img->d_h;
- yv12->render_width = img->r_w;
+ yv12->render_width = img->r_w;
yv12->render_height = img->r_h;
- yv12->y_width = img->d_w;
+ yv12->y_width = img->d_w;
yv12->y_height = img->d_h;
- yv12->uv_width = img->x_chroma_shift == 1 ? (1 + yv12->y_width) / 2
- : yv12->y_width;
- yv12->uv_height = img->y_chroma_shift == 1 ? (1 + yv12->y_height) / 2
- : yv12->y_height;
+ yv12->uv_width =
+ img->x_chroma_shift == 1 ? (1 + yv12->y_width) / 2 : yv12->y_width;
+ yv12->uv_height =
+ img->y_chroma_shift == 1 ? (1 + yv12->y_height) / 2 : yv12->y_height;
yv12->uv_crop_width = yv12->uv_width;
yv12->uv_crop_height = yv12->uv_height;
@@ -124,9 +124,9 @@
} else {
yv12->flags = 0;
}
- yv12->border = (yv12->y_stride - img->w) / 2;
+ yv12->border = (yv12->y_stride - img->w) / 2;
#else
- yv12->border = (img->stride[VPX_PLANE_Y] - img->w) / 2;
+ yv12->border = (img->stride[VPX_PLANE_Y] - img->w) / 2;
#endif // CONFIG_VPX_HIGHBITDEPTH
yv12->subsampling_x = img->x_chroma_shift;
yv12->subsampling_y = img->y_chroma_shift;
