Merge "VP8 for ARMv8 by using NEON intrinsics 14"
diff --git a/vp8/common/arm/neon/loopfilter_neon.c b/vp8/common/arm/neon/loopfilter_neon.c
index 47f522b..be77e67 100644
--- a/vp8/common/arm/neon/loopfilter_neon.c
+++ b/vp8/common/arm/neon/loopfilter_neon.c
@@ -9,8 +9,9 @@
*/
#include <arm_neon.h>
+#include "./vpx_config.h"
-static inline void vp8_loop_filter_neon(
+static INLINE void vp8_loop_filter_neon(
uint8x16_t qblimit, // flimit
uint8x16_t qlimit, // limit
uint8x16_t qthresh, // thresh
diff --git a/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.c b/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.c
index 3d024e0..b25686f 100644
--- a/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.c
+++ b/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.c
@@ -9,8 +9,9 @@
*/
#include <arm_neon.h>
+#include "./vpx_config.h"
-static inline void vp8_loop_filter_simple_horizontal_edge_neon(
+static INLINE void vp8_loop_filter_simple_horizontal_edge_neon(
unsigned char *s,
int p,
const unsigned char *blimit) {
diff --git a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c
index 68bcde4..b0952b5 100644
--- a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c
+++ b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c
@@ -9,8 +9,9 @@
*/
#include <arm_neon.h>
+#include "./vpx_config.h"
-static inline void vp8_loop_filter_simple_vertical_edge_neon(
+static INLINE void vp8_loop_filter_simple_vertical_edge_neon(
unsigned char *s,
int p,
const unsigned char *blimit) {
diff --git a/vp8/common/arm/neon/mbloopfilter_neon.c b/vp8/common/arm/neon/mbloopfilter_neon.c
index 385ce36..5351f4b 100644
--- a/vp8/common/arm/neon/mbloopfilter_neon.c
+++ b/vp8/common/arm/neon/mbloopfilter_neon.c
@@ -9,8 +9,9 @@
*/
#include <arm_neon.h>
+#include "./vpx_config.h"
-static inline void vp8_mbloop_filter_neon(
+static INLINE void vp8_mbloop_filter_neon(
uint8x16_t qblimit, // mblimit
uint8x16_t qlimit, // limit
uint8x16_t qthresh, // thresh
@@ -352,20 +353,28 @@
q9 = vcombine_u8(d18, d19);
q10 = vcombine_u8(d20, d21);
- q2tmp0 = vtrnq_u32((uint32x4_t)q3, (uint32x4_t)q7);
- q2tmp1 = vtrnq_u32((uint32x4_t)q4, (uint32x4_t)q8);
- q2tmp2 = vtrnq_u32((uint32x4_t)q5, (uint32x4_t)q9);
- q2tmp3 = vtrnq_u32((uint32x4_t)q6, (uint32x4_t)q10);
+ q2tmp0 = vtrnq_u32(vreinterpretq_u32_u8(q3), vreinterpretq_u32_u8(q7));
+ q2tmp1 = vtrnq_u32(vreinterpretq_u32_u8(q4), vreinterpretq_u32_u8(q8));
+ q2tmp2 = vtrnq_u32(vreinterpretq_u32_u8(q5), vreinterpretq_u32_u8(q9));
+ q2tmp3 = vtrnq_u32(vreinterpretq_u32_u8(q6), vreinterpretq_u32_u8(q10));
- q2tmp4 = vtrnq_u16((uint16x8_t)q2tmp0.val[0], (uint16x8_t)q2tmp2.val[0]);
- q2tmp5 = vtrnq_u16((uint16x8_t)q2tmp1.val[0], (uint16x8_t)q2tmp3.val[0]);
- q2tmp6 = vtrnq_u16((uint16x8_t)q2tmp0.val[1], (uint16x8_t)q2tmp2.val[1]);
- q2tmp7 = vtrnq_u16((uint16x8_t)q2tmp1.val[1], (uint16x8_t)q2tmp3.val[1]);
+ q2tmp4 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[0]),
+ vreinterpretq_u16_u32(q2tmp2.val[0]));
+ q2tmp5 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[0]),
+ vreinterpretq_u16_u32(q2tmp3.val[0]));
+ q2tmp6 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[1]),
+ vreinterpretq_u16_u32(q2tmp2.val[1]));
+ q2tmp7 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[1]),
+ vreinterpretq_u16_u32(q2tmp3.val[1]));
- q2tmp8 = vtrnq_u8((uint8x16_t)q2tmp4.val[0], (uint8x16_t)q2tmp5.val[0]);
- q2tmp9 = vtrnq_u8((uint8x16_t)q2tmp4.val[1], (uint8x16_t)q2tmp5.val[1]);
- q2tmp10 = vtrnq_u8((uint8x16_t)q2tmp6.val[0], (uint8x16_t)q2tmp7.val[0]);
- q2tmp11 = vtrnq_u8((uint8x16_t)q2tmp6.val[1], (uint8x16_t)q2tmp7.val[1]);
+ q2tmp8 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[0]),
+ vreinterpretq_u8_u16(q2tmp5.val[0]));
+ q2tmp9 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[1]),
+ vreinterpretq_u8_u16(q2tmp5.val[1]));
+ q2tmp10 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[0]),
+ vreinterpretq_u8_u16(q2tmp7.val[0]));
+ q2tmp11 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[1]),
+ vreinterpretq_u8_u16(q2tmp7.val[1]));
q3 = q2tmp8.val[0];
q4 = q2tmp8.val[1];
@@ -380,20 +389,28 @@
q5, q6, q7, q8, q9, q10,
&q4, &q5, &q6, &q7, &q8, &q9);
- q2tmp0 = vtrnq_u32((uint32x4_t)q3, (uint32x4_t)q7);
- q2tmp1 = vtrnq_u32((uint32x4_t)q4, (uint32x4_t)q8);
- q2tmp2 = vtrnq_u32((uint32x4_t)q5, (uint32x4_t)q9);
- q2tmp3 = vtrnq_u32((uint32x4_t)q6, (uint32x4_t)q10);
+ q2tmp0 = vtrnq_u32(vreinterpretq_u32_u8(q3), vreinterpretq_u32_u8(q7));
+ q2tmp1 = vtrnq_u32(vreinterpretq_u32_u8(q4), vreinterpretq_u32_u8(q8));
+ q2tmp2 = vtrnq_u32(vreinterpretq_u32_u8(q5), vreinterpretq_u32_u8(q9));
+ q2tmp3 = vtrnq_u32(vreinterpretq_u32_u8(q6), vreinterpretq_u32_u8(q10));
- q2tmp4 = vtrnq_u16((uint16x8_t)q2tmp0.val[0], (uint16x8_t)q2tmp2.val[0]);
- q2tmp5 = vtrnq_u16((uint16x8_t)q2tmp1.val[0], (uint16x8_t)q2tmp3.val[0]);
- q2tmp6 = vtrnq_u16((uint16x8_t)q2tmp0.val[1], (uint16x8_t)q2tmp2.val[1]);
- q2tmp7 = vtrnq_u16((uint16x8_t)q2tmp1.val[1], (uint16x8_t)q2tmp3.val[1]);
+ q2tmp4 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[0]),
+ vreinterpretq_u16_u32(q2tmp2.val[0]));
+ q2tmp5 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[0]),
+ vreinterpretq_u16_u32(q2tmp3.val[0]));
+ q2tmp6 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[1]),
+ vreinterpretq_u16_u32(q2tmp2.val[1]));
+ q2tmp7 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[1]),
+ vreinterpretq_u16_u32(q2tmp3.val[1]));
- q2tmp8 = vtrnq_u8((uint8x16_t)q2tmp4.val[0], (uint8x16_t)q2tmp5.val[0]);
- q2tmp9 = vtrnq_u8((uint8x16_t)q2tmp4.val[1], (uint8x16_t)q2tmp5.val[1]);
- q2tmp10 = vtrnq_u8((uint8x16_t)q2tmp6.val[0], (uint8x16_t)q2tmp7.val[0]);
- q2tmp11 = vtrnq_u8((uint8x16_t)q2tmp6.val[1], (uint8x16_t)q2tmp7.val[1]);
+ q2tmp8 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[0]),
+ vreinterpretq_u8_u16(q2tmp5.val[0]));
+ q2tmp9 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[1]),
+ vreinterpretq_u8_u16(q2tmp5.val[1]));
+ q2tmp10 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[0]),
+ vreinterpretq_u8_u16(q2tmp7.val[0]));
+ q2tmp11 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[1]),
+ vreinterpretq_u8_u16(q2tmp7.val[1]));
q3 = q2tmp8.val[0];
q4 = q2tmp8.val[1];
@@ -503,20 +520,28 @@
q9 = vcombine_u8(d18, d19);
q10 = vcombine_u8(d20, d21);
- q2tmp0 = vtrnq_u32((uint32x4_t)q3, (uint32x4_t)q7);
- q2tmp1 = vtrnq_u32((uint32x4_t)q4, (uint32x4_t)q8);
- q2tmp2 = vtrnq_u32((uint32x4_t)q5, (uint32x4_t)q9);
- q2tmp3 = vtrnq_u32((uint32x4_t)q6, (uint32x4_t)q10);
+ q2tmp0 = vtrnq_u32(vreinterpretq_u32_u8(q3), vreinterpretq_u32_u8(q7));
+ q2tmp1 = vtrnq_u32(vreinterpretq_u32_u8(q4), vreinterpretq_u32_u8(q8));
+ q2tmp2 = vtrnq_u32(vreinterpretq_u32_u8(q5), vreinterpretq_u32_u8(q9));
+ q2tmp3 = vtrnq_u32(vreinterpretq_u32_u8(q6), vreinterpretq_u32_u8(q10));
- q2tmp4 = vtrnq_u16((uint16x8_t)q2tmp0.val[0], (uint16x8_t)q2tmp2.val[0]);
- q2tmp5 = vtrnq_u16((uint16x8_t)q2tmp1.val[0], (uint16x8_t)q2tmp3.val[0]);
- q2tmp6 = vtrnq_u16((uint16x8_t)q2tmp0.val[1], (uint16x8_t)q2tmp2.val[1]);
- q2tmp7 = vtrnq_u16((uint16x8_t)q2tmp1.val[1], (uint16x8_t)q2tmp3.val[1]);
+ q2tmp4 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[0]),
+ vreinterpretq_u16_u32(q2tmp2.val[0]));
+ q2tmp5 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[0]),
+ vreinterpretq_u16_u32(q2tmp3.val[0]));
+ q2tmp6 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[1]),
+ vreinterpretq_u16_u32(q2tmp2.val[1]));
+ q2tmp7 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[1]),
+ vreinterpretq_u16_u32(q2tmp3.val[1]));
- q2tmp8 = vtrnq_u8((uint8x16_t)q2tmp4.val[0], (uint8x16_t)q2tmp5.val[0]);
- q2tmp9 = vtrnq_u8((uint8x16_t)q2tmp4.val[1], (uint8x16_t)q2tmp5.val[1]);
- q2tmp10 = vtrnq_u8((uint8x16_t)q2tmp6.val[0], (uint8x16_t)q2tmp7.val[0]);
- q2tmp11 = vtrnq_u8((uint8x16_t)q2tmp6.val[1], (uint8x16_t)q2tmp7.val[1]);
+ q2tmp8 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[0]),
+ vreinterpretq_u8_u16(q2tmp5.val[0]));
+ q2tmp9 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[1]),
+ vreinterpretq_u8_u16(q2tmp5.val[1]));
+ q2tmp10 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[0]),
+ vreinterpretq_u8_u16(q2tmp7.val[0]));
+ q2tmp11 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[1]),
+ vreinterpretq_u8_u16(q2tmp7.val[1]));
q3 = q2tmp8.val[0];
q4 = q2tmp8.val[1];
@@ -531,20 +556,28 @@
q5, q6, q7, q8, q9, q10,
&q4, &q5, &q6, &q7, &q8, &q9);
- q2tmp0 = vtrnq_u32((uint32x4_t)q3, (uint32x4_t)q7);
- q2tmp1 = vtrnq_u32((uint32x4_t)q4, (uint32x4_t)q8);
- q2tmp2 = vtrnq_u32((uint32x4_t)q5, (uint32x4_t)q9);
- q2tmp3 = vtrnq_u32((uint32x4_t)q6, (uint32x4_t)q10);
+ q2tmp0 = vtrnq_u32(vreinterpretq_u32_u8(q3), vreinterpretq_u32_u8(q7));
+ q2tmp1 = vtrnq_u32(vreinterpretq_u32_u8(q4), vreinterpretq_u32_u8(q8));
+ q2tmp2 = vtrnq_u32(vreinterpretq_u32_u8(q5), vreinterpretq_u32_u8(q9));
+ q2tmp3 = vtrnq_u32(vreinterpretq_u32_u8(q6), vreinterpretq_u32_u8(q10));
- q2tmp4 = vtrnq_u16((uint16x8_t)q2tmp0.val[0], (uint16x8_t)q2tmp2.val[0]);
- q2tmp5 = vtrnq_u16((uint16x8_t)q2tmp1.val[0], (uint16x8_t)q2tmp3.val[0]);
- q2tmp6 = vtrnq_u16((uint16x8_t)q2tmp0.val[1], (uint16x8_t)q2tmp2.val[1]);
- q2tmp7 = vtrnq_u16((uint16x8_t)q2tmp1.val[1], (uint16x8_t)q2tmp3.val[1]);
+ q2tmp4 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[0]),
+ vreinterpretq_u16_u32(q2tmp2.val[0]));
+ q2tmp5 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[0]),
+ vreinterpretq_u16_u32(q2tmp3.val[0]));
+ q2tmp6 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[1]),
+ vreinterpretq_u16_u32(q2tmp2.val[1]));
+ q2tmp7 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[1]),
+ vreinterpretq_u16_u32(q2tmp3.val[1]));
- q2tmp8 = vtrnq_u8((uint8x16_t)q2tmp4.val[0], (uint8x16_t)q2tmp5.val[0]);
- q2tmp9 = vtrnq_u8((uint8x16_t)q2tmp4.val[1], (uint8x16_t)q2tmp5.val[1]);
- q2tmp10 = vtrnq_u8((uint8x16_t)q2tmp6.val[0], (uint8x16_t)q2tmp7.val[0]);
- q2tmp11 = vtrnq_u8((uint8x16_t)q2tmp6.val[1], (uint8x16_t)q2tmp7.val[1]);
+ q2tmp8 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[0]),
+ vreinterpretq_u8_u16(q2tmp5.val[0]));
+ q2tmp9 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[1]),
+ vreinterpretq_u8_u16(q2tmp5.val[1]));
+ q2tmp10 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[0]),
+ vreinterpretq_u8_u16(q2tmp7.val[0]));
+ q2tmp11 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[1]),
+ vreinterpretq_u8_u16(q2tmp7.val[1]));
q3 = q2tmp8.val[0];
q4 = q2tmp8.val[1];
diff --git a/vp8/common/rtcd_defs.pl b/vp8/common/rtcd_defs.pl
index 130d965..789b0de5 100644
--- a/vp8/common/rtcd_defs.pl
+++ b/vp8/common/rtcd_defs.pl
@@ -532,7 +532,7 @@
# Denoiser filter
#
if (vpx_config("CONFIG_TEMPORAL_DENOISING") eq "yes") {
- add_proto qw/int vp8_denoiser_filter/, "struct yv12_buffer_config* mc_running_avg, struct yv12_buffer_config* running_avg, struct macroblock* signal, unsigned int motion_magnitude2, int y_offset, int uv_offset";
+ add_proto qw/int vp8_denoiser_filter/, "unsigned char *mc_running_avg_y, int mc_avg_y_stride, unsigned char *running_avg_y, int avg_y_stride, unsigned char *sig, int sig_stride, unsigned int motion_magnitude";
specialize qw/vp8_denoiser_filter sse2 neon/;
}
diff --git a/vp8/encoder/arm/neon/denoising_neon.c b/vp8/encoder/arm/neon/denoising_neon.c
index 23dc0a9..1bebe8f 100644
--- a/vp8/encoder/arm/neon/denoising_neon.c
+++ b/vp8/encoder/arm/neon/denoising_neon.c
@@ -45,10 +45,12 @@
* [16, 255] 3 6 7
*/
-int vp8_denoiser_filter_neon(YV12_BUFFER_CONFIG *mc_running_avg,
- YV12_BUFFER_CONFIG *running_avg,
- MACROBLOCK *signal, unsigned int motion_magnitude,
- int y_offset, int uv_offset) {
+int vp8_denoiser_filter_neon(unsigned char *mc_running_avg_y,
+ int mc_running_avg_y_stride,
+ unsigned char *running_avg_y,
+ int running_avg_y_stride,
+ unsigned char *sig, int sig_stride,
+ unsigned int motion_magnitude) {
/* If motion_magnitude is small, making the denoiser more aggressive by
* increasing the adjustment for each level, level1 adjustment is
* increased, the deltas stay the same.
@@ -60,14 +62,6 @@
const uint8x16_t v_level1_threshold = vdupq_n_u8(4);
const uint8x16_t v_level2_threshold = vdupq_n_u8(8);
const uint8x16_t v_level3_threshold = vdupq_n_u8(16);
-
- /* Local variables for array pointers and strides. */
- unsigned char *sig = signal->thismb;
- int sig_stride = 16;
- unsigned char *mc_running_avg_y = mc_running_avg->y_buffer + y_offset;
- int mc_running_avg_y_stride = mc_running_avg->y_stride;
- unsigned char *running_avg_y = running_avg->y_buffer + y_offset;
- int running_avg_y_stride = running_avg->y_stride;
int64x2_t v_sum_diff_total = vdupq_n_s64(0);
/* Go over lines. */
diff --git a/vp8/encoder/denoising.c b/vp8/encoder/denoising.c
index 7819265..bfce280 100644
--- a/vp8/encoder/denoising.c
+++ b/vp8/encoder/denoising.c
@@ -51,17 +51,13 @@
* [16, 255] 6 7
*/
-int vp8_denoiser_filter_c(YV12_BUFFER_CONFIG *mc_running_avg,
- YV12_BUFFER_CONFIG *running_avg, MACROBLOCK *signal,
- unsigned int motion_magnitude, int y_offset,
- int uv_offset)
+int vp8_denoiser_filter_c(unsigned char *mc_running_avg_y, int mc_avg_y_stride,
+ unsigned char *running_avg_y, int avg_y_stride,
+ unsigned char *sig, int sig_stride,
+ unsigned int motion_magnitude)
{
- unsigned char *sig = signal->thismb;
- int sig_stride = 16;
- unsigned char *mc_running_avg_y = mc_running_avg->y_buffer + y_offset;
- int mc_avg_y_stride = mc_running_avg->y_stride;
- unsigned char *running_avg_y = running_avg->y_buffer + y_offset;
- int avg_y_stride = running_avg->y_stride;
+ unsigned char *running_avg_y_start = running_avg_y;
+ unsigned char *sig_start = sig;
int r, c, i;
int sum_diff = 0;
int adj_val[3] = {3, 4, 6};
@@ -130,8 +126,7 @@
if (abs(sum_diff) > SUM_DIFF_THRESHOLD)
return COPY_BLOCK;
- vp8_copy_mem16x16(running_avg->y_buffer + y_offset, avg_y_stride,
- signal->thismb, sig_stride);
+ vp8_copy_mem16x16(running_avg_y_start, avg_y_stride, sig_start, sig_stride);
return FILTER_BLOCK;
}
@@ -285,12 +280,17 @@
if (decision == FILTER_BLOCK)
{
+ unsigned char *mc_running_avg_y =
+ denoiser->yv12_mc_running_avg.y_buffer + recon_yoffset;
+ int mc_avg_y_stride = denoiser->yv12_mc_running_avg.y_stride;
+ unsigned char *running_avg_y =
+ denoiser->yv12_running_avg[INTRA_FRAME].y_buffer + recon_yoffset;
+ int avg_y_stride = denoiser->yv12_running_avg[INTRA_FRAME].y_stride;
+
/* Filter. */
- decision = vp8_denoiser_filter(&denoiser->yv12_mc_running_avg,
- &denoiser->yv12_running_avg[INTRA_FRAME],
- x,
- motion_magnitude2,
- recon_yoffset, recon_uvoffset);
+ decision = vp8_denoiser_filter(mc_running_avg_y, mc_avg_y_stride,
+ running_avg_y, avg_y_stride,
+ x->thismb, 16, motion_magnitude2);
}
if (decision == COPY_BLOCK)
{
diff --git a/vp8/encoder/x86/denoising_sse2.c b/vp8/encoder/x86/denoising_sse2.c
index cceb826..d1f76b2 100644
--- a/vp8/encoder/x86/denoising_sse2.c
+++ b/vp8/encoder/x86/denoising_sse2.c
@@ -22,17 +22,14 @@
signed char e[16];
};
-int vp8_denoiser_filter_sse2(YV12_BUFFER_CONFIG *mc_running_avg,
- YV12_BUFFER_CONFIG *running_avg,
- MACROBLOCK *signal, unsigned int motion_magnitude,
- int y_offset, int uv_offset)
+int vp8_denoiser_filter_sse2(unsigned char *mc_running_avg_y,
+ int mc_avg_y_stride,
+ unsigned char *running_avg_y, int avg_y_stride,
+ unsigned char *sig, int sig_stride,
+ unsigned int motion_magnitude)
{
- unsigned char *sig = signal->thismb;
- int sig_stride = 16;
- unsigned char *mc_running_avg_y = mc_running_avg->y_buffer + y_offset;
- int mc_avg_y_stride = mc_running_avg->y_stride;
- unsigned char *running_avg_y = running_avg->y_buffer + y_offset;
- int avg_y_stride = running_avg->y_stride;
+ unsigned char *running_avg_y_start = running_avg_y;
+ unsigned char *sig_start = sig;
int r;
__m128i acc_diff = _mm_setzero_si128();
const __m128i k_0 = _mm_setzero_si128();
@@ -114,7 +111,6 @@
}
}
- vp8_copy_mem16x16(running_avg->y_buffer + y_offset, avg_y_stride,
- signal->thismb, sig_stride);
+ vp8_copy_mem16x16(running_avg_y_start, avg_y_stride, sig_start, sig_stride);
return FILTER_BLOCK;
}
