Merge "Making vp9_initialize_dec() static."
diff --git a/build/make/Android.mk b/build/make/Android.mk
index 826ff2f..816334e 100644
--- a/build/make/Android.mk
+++ b/build/make/Android.mk
@@ -38,8 +38,9 @@
# For this we import the 'cpufeatures' module from the NDK sources.
# libvpx can also be configured without this runtime detection method.
# Configuring with --disable-runtime-cpu-detect will assume presence of NEON.
-# Configuring with --disable-runtime-cpu-detect --disable-neon will remove any
-# NEON dependency.
+# Configuring with --disable-runtime-cpu-detect --disable-neon \
+# --disable-neon-asm
+# will remove any NEON dependency.
# To change to building armeabi, run ./libvpx/configure again, but with
# --target=arm5te-android-gcc and modify the Application.mk file to
diff --git a/build/make/configure.sh b/build/make/configure.sh
index a2a64c2..c07b049 100755
--- a/build/make/configure.sh
+++ b/build/make/configure.sh
@@ -518,7 +518,7 @@
--enable-?*|--disable-?*)
eval `echo "$opt" | sed 's/--/action=/;s/-/ option=/;s/-/_/g'`
if echo "${ARCH_EXT_LIST}" | grep "^ *$option\$" >/dev/null; then
- [ $action = "disable" ] && RTCD_OPTIONS="${RTCD_OPTIONS}${opt} "
+ [ $action = "disable" ] && RTCD_OPTIONS="${RTCD_OPTIONS}--disable-${option} "
elif [ $action = "disable" ] && ! disabled $option ; then
echo "${CMDLINE_SELECT}" | grep "^ *$option\$" >/dev/null ||
die_unknown $opt
@@ -835,7 +835,7 @@
check_add_cflags -march=armv7-a -mfloat-abi=${float_abi}
check_add_asflags -march=armv7-a -mfloat-abi=${float_abi}
- if enabled neon
+ if enabled neon || enabled neon_asm
then
check_add_cflags -mfpu=neon #-ftree-vectorize
check_add_asflags -mfpu=neon
@@ -882,7 +882,7 @@
tune_asflags="--cpu="
if [ -z "${tune_cpu}" ]; then
if [ ${tgt_isa} = "armv7" ]; then
- if enabled neon
+ if enabled neon || enabled neon_asm
then
check_add_cflags --fpu=softvfp+vfpv3
check_add_asflags --fpu=softvfp+vfpv3
diff --git a/examples.mk b/examples.mk
index b9dcb65..28ab33a 100644
--- a/examples.mk
+++ b/examples.mk
@@ -81,13 +81,13 @@
EXAMPLES-$(CONFIG_VP9_ENCODER) += resize_util.c
endif
-EXAMPLES-$(CONFIG_ENCODERS) += vpx_temporal_scalable_patterns.c
-vpx_temporal_scalable_patterns.SRCS += ivfenc.c ivfenc.h
-vpx_temporal_scalable_patterns.SRCS += tools_common.c tools_common.h
-vpx_temporal_scalable_patterns.SRCS += video_common.h
-vpx_temporal_scalable_patterns.SRCS += video_writer.h video_writer.c
-vpx_temporal_scalable_patterns.GUID = B18C08F2-A439-4502-A78E-849BE3D60947
-vpx_temporal_scalable_patterns.DESCRIPTION = Temporal Scalability Encoder
+EXAMPLES-$(CONFIG_ENCODERS) += vpx_temporal_svc_encoder.c
+vpx_temporal_svc_encoder.SRCS += ivfenc.c ivfenc.h
+vpx_temporal_svc_encoder.SRCS += tools_common.c tools_common.h
+vpx_temporal_svc_encoder.SRCS += video_common.h
+vpx_temporal_svc_encoder.SRCS += video_writer.h video_writer.c
+vpx_temporal_svc_encoder.GUID = B18C08F2-A439-4502-A78E-849BE3D60947
+vpx_temporal_svc_encoder.DESCRIPTION = Temporal SVC Encoder
EXAMPLES-$(CONFIG_VP8_DECODER) += simple_decoder.c
simple_decoder.GUID = D3BBF1E9-2427-450D-BBFF-B2843C1D44CC
simple_decoder.SRCS += ivfdec.h ivfdec.c
diff --git a/examples/set_maps.c b/examples/set_maps.c
index 4343832..4ba38ee 100644
--- a/examples/set_maps.c
+++ b/examples/set_maps.c
@@ -64,7 +64,8 @@
static void set_roi_map(const vpx_codec_enc_cfg_t *cfg,
vpx_codec_ctx_t *codec) {
unsigned int i;
- vpx_roi_map_t roi = {0};
+ vpx_roi_map_t roi;
+ memset(&roi, 0, sizeof(roi));
roi.rows = (cfg->g_h + 15) / 16;
roi.cols = (cfg->g_w + 15) / 16;
@@ -97,7 +98,7 @@
static void set_active_map(const vpx_codec_enc_cfg_t *cfg,
vpx_codec_ctx_t *codec) {
unsigned int i;
- vpx_active_map_t map = {0};
+ vpx_active_map_t map = {0, 0, 0};
map.rows = (cfg->g_h + 15) / 16;
map.cols = (cfg->g_w + 15) / 16;
@@ -114,7 +115,7 @@
static void unset_active_map(const vpx_codec_enc_cfg_t *cfg,
vpx_codec_ctx_t *codec) {
- vpx_active_map_t map = {0};
+ vpx_active_map_t map = {0, 0, 0};
map.rows = (cfg->g_h + 15) / 16;
map.cols = (cfg->g_w + 15) / 16;
@@ -153,22 +154,23 @@
int main(int argc, char **argv) {
FILE *infile = NULL;
- vpx_codec_ctx_t codec = {0};
- vpx_codec_enc_cfg_t cfg = {0};
+ vpx_codec_ctx_t codec;
+ vpx_codec_enc_cfg_t cfg;
int frame_count = 0;
- vpx_image_t raw = {0};
+ vpx_image_t raw;
vpx_codec_err_t res;
- VpxVideoInfo info = {0};
+ VpxVideoInfo info;
VpxVideoWriter *writer = NULL;
const VpxInterface *encoder = NULL;
const int fps = 2; // TODO(dkovalev) add command line argument
const double bits_per_pixel_per_frame = 0.067;
exec_name = argv[0];
-
if (argc != 6)
die("Invalid number of arguments");
+ memset(&info, 0, sizeof(info));
+
encoder = get_vpx_encoder_by_name(argv[1]);
if (!encoder)
die("Unsupported codec.");
diff --git a/examples/vpx_temporal_scalable_patterns.c b/examples/vpx_temporal_svc_encoder.c
similarity index 99%
rename from examples/vpx_temporal_scalable_patterns.c
rename to examples/vpx_temporal_svc_encoder.c
index 07dd318..e45b50c 100644
--- a/examples/vpx_temporal_scalable_patterns.c
+++ b/examples/vpx_temporal_svc_encoder.c
@@ -8,7 +8,7 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-// This is an example demonstrating how to implement a multi-layer VP9
+// This is an example demonstrating how to implement a multi-layer VPx
// encoding scheme based on temporal scalability for video applications
// that benefit from a scalable bitstream.
diff --git a/test/sad_test.cc b/test/sad_test.cc
index 23adbda..a692891 100644
--- a/test/sad_test.cc
+++ b/test/sad_test.cc
@@ -32,7 +32,8 @@
typedef unsigned int (*sad_m_by_n_fn_t)(const unsigned char *source_ptr,
int source_stride,
const unsigned char *reference_ptr,
- int reference_stride);
+ int reference_stride,
+ unsigned int max_sad);
typedef std::tr1::tuple<int, int, sad_m_by_n_fn_t> sad_m_by_n_test_param_t;
typedef void (*sad_n_by_n_by_4_fn_t)(const uint8_t *src_ptr,
@@ -86,7 +87,7 @@
// Sum of Absolute Differences. Given two blocks, calculate the absolute
// difference between two pixels in the same relative location; accumulate.
- unsigned int ReferenceSAD(int block_idx = 0) {
+ unsigned int ReferenceSAD(unsigned int max_sad, int block_idx = 0) {
unsigned int sad = 0;
const uint8_t* const reference = GetReference(block_idx);
@@ -95,6 +96,9 @@
sad += abs(source_data_[h * source_stride_ + w]
- reference[h * reference_stride_ + w]);
}
+ if (sad > max_sad) {
+ break;
+ }
}
return sad;
}
@@ -130,20 +134,21 @@
SADTest() : SADTestBase(GET_PARAM(0), GET_PARAM(1)) {}
protected:
- unsigned int SAD(int block_idx = 0) {
+ unsigned int SAD(unsigned int max_sad, int block_idx = 0) {
unsigned int ret;
const uint8_t* const reference = GetReference(block_idx);
REGISTER_STATE_CHECK(ret = GET_PARAM(2)(source_data_, source_stride_,
- reference, reference_stride_));
+ reference, reference_stride_,
+ max_sad));
return ret;
}
void CheckSad(unsigned int max_sad) {
unsigned int reference_sad, exp_sad;
- reference_sad = ReferenceSAD();
- exp_sad = SAD();
+ reference_sad = ReferenceSAD(max_sad);
+ exp_sad = SAD(max_sad);
if (reference_sad <= max_sad) {
ASSERT_EQ(exp_sad, reference_sad);
@@ -174,7 +179,7 @@
SADs(exp_sad);
for (int block = 0; block < 4; block++) {
- reference_sad = ReferenceSAD(block);
+ reference_sad = ReferenceSAD(UINT_MAX, block);
EXPECT_EQ(exp_sad[block], reference_sad) << "block " << block;
}
diff --git a/tools_common.h b/tools_common.h
index 2124882..e033de2 100644
--- a/tools_common.h
+++ b/tools_common.h
@@ -89,7 +89,7 @@
enum VideoFileType file_type;
uint32_t width;
uint32_t height;
- int use_i420;
+ vpx_img_fmt_t fmt;
int only_i420;
uint32_t fourcc;
struct VpxRational framerate;
diff --git a/vp8/common/arm/dequantize_arm.c b/vp8/common/arm/dequantize_arm.c
index 3e37e08..1f8157f 100644
--- a/vp8/common/arm/dequantize_arm.c
+++ b/vp8/common/arm/dequantize_arm.c
@@ -12,26 +12,9 @@
#include "vpx_config.h"
#include "vp8/common/blockd.h"
-#if HAVE_NEON_ASM
-extern void vp8_dequantize_b_loop_neon(short *Q, short *DQC, short *DQ);
-#endif
-
#if HAVE_MEDIA
extern void vp8_dequantize_b_loop_v6(short *Q, short *DQC, short *DQ);
-#endif
-#if HAVE_NEON_ASM
-
-void vp8_dequantize_b_neon(BLOCKD *d, short *DQC)
-{
- short *DQ = d->dqcoeff;
- short *Q = d->qcoeff;
-
- vp8_dequantize_b_loop_neon(Q, DQC, DQ);
-}
-#endif
-
-#if HAVE_MEDIA
void vp8_dequantize_b_v6(BLOCKD *d, short *DQC)
{
short *DQ = d->dqcoeff;
diff --git a/vp8/common/arm/loopfilter_arm.c b/vp8/common/arm/loopfilter_arm.c
index 5d693c6..f37ca63 100644
--- a/vp8/common/arm/loopfilter_arm.c
+++ b/vp8/common/arm/loopfilter_arm.c
@@ -25,20 +25,24 @@
extern prototype_loopfilter(vp8_mbloop_filter_vertical_edge_armv6);
#endif
-#if HAVE_NEON_ASM
+#if HAVE_NEON_ASM || HAVE_NEON
typedef void loopfilter_y_neon(unsigned char *src, int pitch,
unsigned char blimit, unsigned char limit, unsigned char thresh);
typedef void loopfilter_uv_neon(unsigned char *u, int pitch,
unsigned char blimit, unsigned char limit, unsigned char thresh,
unsigned char *v);
+#endif
+#if HAVE_NEON_ASM
extern loopfilter_y_neon vp8_loop_filter_horizontal_edge_y_neon;
extern loopfilter_y_neon vp8_loop_filter_vertical_edge_y_neon;
-extern loopfilter_y_neon vp8_mbloop_filter_horizontal_edge_y_neon;
-extern loopfilter_y_neon vp8_mbloop_filter_vertical_edge_y_neon;
-
extern loopfilter_uv_neon vp8_loop_filter_horizontal_edge_uv_neon;
extern loopfilter_uv_neon vp8_loop_filter_vertical_edge_uv_neon;
+#endif
+
+#if HAVE_NEON
+extern loopfilter_y_neon vp8_mbloop_filter_horizontal_edge_y_neon;
+extern loopfilter_y_neon vp8_mbloop_filter_vertical_edge_y_neon;
extern loopfilter_uv_neon vp8_mbloop_filter_horizontal_edge_uv_neon;
extern loopfilter_uv_neon vp8_mbloop_filter_vertical_edge_uv_neon;
#endif
@@ -118,7 +122,7 @@
}
#endif
-#if HAVE_NEON_ASM
+#if HAVE_NEON
/* NEON loopfilter functions */
/* Horizontal MB filtering */
void vp8_loop_filter_mbh_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -146,7 +150,9 @@
if (u_ptr)
vp8_mbloop_filter_vertical_edge_uv_neon(u_ptr, uv_stride, mblim, lim, hev_thr, v_ptr);
}
+#endif
+#if HAVE_NEON_ASM
/* Horizontal B Filtering */
void vp8_loop_filter_bh_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
int y_stride, int uv_stride, loop_filter_info *lfi)
diff --git a/vp8/common/arm/neon/dequantizeb_neon.c b/vp8/common/arm/neon/dequantizeb_neon.c
index 60f69c8..54e709d 100644
--- a/vp8/common/arm/neon/dequantizeb_neon.c
+++ b/vp8/common/arm/neon/dequantizeb_neon.c
@@ -10,18 +10,16 @@
#include <arm_neon.h>
-void vp8_dequantize_b_loop_neon(
- int16_t *Q,
- int16_t *DQC,
- int16_t *DQ) {
+#include "vp8/common/blockd.h"
+
+void vp8_dequantize_b_neon(BLOCKD *d, short *DQC) {
int16x8x2_t qQ, qDQC, qDQ;
- qQ = vld2q_s16(Q);
+ qQ = vld2q_s16(d->qcoeff);
qDQC = vld2q_s16(DQC);
qDQ.val[0] = vmulq_s16(qQ.val[0], qDQC.val[0]);
qDQ.val[1] = vmulq_s16(qQ.val[1], qDQC.val[1]);
- vst2q_s16(DQ, qDQ);
- return;
+ vst2q_s16(d->dqcoeff, qDQ);
}
diff --git a/vp8/common/arm/reconintra_arm.c b/vp8/common/arm/reconintra_arm.c
index 765fc3a..e55a33c 100644
--- a/vp8/common/arm/reconintra_arm.c
+++ b/vp8/common/arm/reconintra_arm.c
@@ -14,7 +14,7 @@
#include "vp8/common/blockd.h"
#include "vpx_mem/vpx_mem.h"
-#if HAVE_NEON_ARM
+#if HAVE_NEON_ASM
extern void vp8_build_intra_predictors_mby_neon_func(
unsigned char *y_buffer,
unsigned char *ypred_ptr,
diff --git a/vp8/common/rtcd_defs.pl b/vp8/common/rtcd_defs.pl
index 0aaec1b..cbfd76a 100644
--- a/vp8/common/rtcd_defs.pl
+++ b/vp8/common/rtcd_defs.pl
@@ -29,9 +29,8 @@
# Dequant
#
add_proto qw/void vp8_dequantize_b/, "struct blockd*, short *dqc";
-specialize qw/vp8_dequantize_b mmx media neon_asm/;
+specialize qw/vp8_dequantize_b mmx media neon/;
$vp8_dequantize_b_media=vp8_dequantize_b_v6;
-$vp8_dequantize_b_neon_asm=vp8_dequantize_b_neon;
add_proto qw/void vp8_dequant_idct_add/, "short *input, short *dq, unsigned char *output, int stride";
specialize qw/vp8_dequant_idct_add mmx media neon dspr2/;
@@ -54,9 +53,8 @@
# Loopfilter
#
add_proto qw/void vp8_loop_filter_mbv/, "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi";
-specialize qw/vp8_loop_filter_mbv mmx sse2 media neon_asm dspr2/;
+specialize qw/vp8_loop_filter_mbv mmx sse2 media neon dspr2/;
$vp8_loop_filter_mbv_media=vp8_loop_filter_mbv_armv6;
-$vp8_loop_filter_mbv_neon_asm=vp8_loop_filter_mbv_neon;
$vp8_loop_filter_mbv_dspr2=vp8_loop_filter_mbv_dspr2;
add_proto qw/void vp8_loop_filter_bv/, "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi";
@@ -66,9 +64,8 @@
$vp8_loop_filter_bv_dspr2=vp8_loop_filter_bv_dspr2;
add_proto qw/void vp8_loop_filter_mbh/, "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi";
-specialize qw/vp8_loop_filter_mbh mmx sse2 media neon_asm dspr2/;
+specialize qw/vp8_loop_filter_mbh mmx sse2 media neon dspr2/;
$vp8_loop_filter_mbh_media=vp8_loop_filter_mbh_armv6;
-$vp8_loop_filter_mbh_neon_asm=vp8_loop_filter_mbh_neon;
$vp8_loop_filter_mbh_dspr2=vp8_loop_filter_mbh_dspr2;
add_proto qw/void vp8_loop_filter_bh/, "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi";
@@ -87,12 +84,12 @@
$vp8_loop_filter_simple_mbv_neon_asm=vp8_loop_filter_mbvs_neon;
add_proto qw/void vp8_loop_filter_simple_mbh/, "unsigned char *y, int ystride, const unsigned char *blimit";
-specialize qw/vp8_loop_filter_simple_mbh mmx sse2 media neon_asm/;
+specialize qw/vp8_loop_filter_simple_mbh mmx sse2 media neon/;
$vp8_loop_filter_simple_mbh_c=vp8_loop_filter_simple_horizontal_edge_c;
$vp8_loop_filter_simple_mbh_mmx=vp8_loop_filter_simple_horizontal_edge_mmx;
$vp8_loop_filter_simple_mbh_sse2=vp8_loop_filter_simple_horizontal_edge_sse2;
$vp8_loop_filter_simple_mbh_media=vp8_loop_filter_simple_horizontal_edge_armv6;
-$vp8_loop_filter_simple_mbh_neon_asm=vp8_loop_filter_mbhs_neon;
+$vp8_loop_filter_simple_mbh_neon=vp8_loop_filter_mbhs_neon;
add_proto qw/void vp8_loop_filter_simple_bv/, "unsigned char *y, int ystride, const unsigned char *blimit";
specialize qw/vp8_loop_filter_simple_bv mmx sse2 media neon_asm/;
@@ -103,12 +100,12 @@
$vp8_loop_filter_simple_bv_neon_asm=vp8_loop_filter_bvs_neon;
add_proto qw/void vp8_loop_filter_simple_bh/, "unsigned char *y, int ystride, const unsigned char *blimit";
-specialize qw/vp8_loop_filter_simple_bh mmx sse2 media neon_asm/;
+specialize qw/vp8_loop_filter_simple_bh mmx sse2 media neon/;
$vp8_loop_filter_simple_bh_c=vp8_loop_filter_bhs_c;
$vp8_loop_filter_simple_bh_mmx=vp8_loop_filter_bhs_mmx;
$vp8_loop_filter_simple_bh_sse2=vp8_loop_filter_bhs_sse2;
$vp8_loop_filter_simple_bh_media=vp8_loop_filter_bhs_armv6;
-$vp8_loop_filter_simple_bh_neon_asm=vp8_loop_filter_bhs_neon;
+$vp8_loop_filter_simple_bh_neon=vp8_loop_filter_bhs_neon;
#
# IDCT
@@ -316,23 +313,23 @@
#
# Single block SAD
#
-add_proto qw/unsigned int vp8_sad4x4/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp8_sad4x4/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad4x4 mmx sse2 neon/;
$vp8_sad4x4_sse2=vp8_sad4x4_wmt;
-add_proto qw/unsigned int vp8_sad8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp8_sad8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad8x8 mmx sse2 neon/;
$vp8_sad8x8_sse2=vp8_sad8x8_wmt;
-add_proto qw/unsigned int vp8_sad8x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp8_sad8x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad8x16 mmx sse2 neon/;
$vp8_sad8x16_sse2=vp8_sad8x16_wmt;
-add_proto qw/unsigned int vp8_sad16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp8_sad16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad16x8 mmx sse2 neon/;
$vp8_sad16x8_sse2=vp8_sad16x8_wmt;
-add_proto qw/unsigned int vp8_sad16x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp8_sad16x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad16x16 mmx sse2 sse3 media neon/;
$vp8_sad16x16_sse2=vp8_sad16x16_wmt;
$vp8_sad16x16_media=vp8_sad16x16_armv6;
@@ -555,7 +552,7 @@
# Denoiser filter
#
if (vpx_config("CONFIG_TEMPORAL_DENOISING") eq "yes") {
- 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";
+ 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, int increase_denoising";
specialize qw/vp8_denoiser_filter sse2 neon/;
}
diff --git a/vp8/common/sad_c.c b/vp8/common/sad_c.c
index 6587202..5f36fc9 100644
--- a/vp8/common/sad_c.c
+++ b/vp8/common/sad_c.c
@@ -16,7 +16,7 @@
static unsigned int sad_mx_n_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
- int m, int n)
+ unsigned int max_sad, int m, int n)
{
int r, c;
unsigned int sad = 0;
@@ -28,6 +28,9 @@
sad += abs(src_ptr[c] - ref_ptr[c]);
}
+ if (sad > max_sad)
+ break;
+
src_ptr += src_stride;
ref_ptr += ref_stride;
}
@@ -40,199 +43,204 @@
*/
unsigned int vp8_sad16x16_c(const unsigned char *src_ptr, int src_stride,
- const unsigned char *ref_ptr, int ref_stride)
+ const unsigned char *ref_ptr, int ref_stride,
+ unsigned int max_sad)
{
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 16, 16);
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 16, 16);
}
unsigned int vp8_sad8x8_c(const unsigned char *src_ptr, int src_stride,
- const unsigned char *ref_ptr, int ref_stride)
+ const unsigned char *ref_ptr, int ref_stride,
+ unsigned int max_sad)
{
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 8, 8);
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 8, 8);
}
unsigned int vp8_sad16x8_c(const unsigned char *src_ptr, int src_stride,
- const unsigned char *ref_ptr, int ref_stride)
+ const unsigned char *ref_ptr, int ref_stride,
+ unsigned int max_sad)
{
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 16, 8);
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 16, 8);
}
unsigned int vp8_sad8x16_c(const unsigned char *src_ptr, int src_stride,
- const unsigned char *ref_ptr, int ref_stride)
+ const unsigned char *ref_ptr, int ref_stride,
+ unsigned int max_sad)
{
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 8, 16);
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 8, 16);
}
unsigned int vp8_sad4x4_c(const unsigned char *src_ptr, int src_stride,
- const unsigned char *ref_ptr, int ref_stride)
+ const unsigned char *ref_ptr, int ref_stride,
+ unsigned int max_sad)
{
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 4, 4);
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 4, 4);
}
void vp8_sad16x16x3_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
+ sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
}
void vp8_sad16x16x8_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned short *sad_array)
{
- sad_array[0] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
- sad_array[3] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 3, ref_stride);
- sad_array[4] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride);
- sad_array[5] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride);
- sad_array[6] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 6, ref_stride);
- sad_array[7] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride);
+ sad_array[0] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+ sad_array[3] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+ sad_array[4] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+ sad_array[5] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+ sad_array[6] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+ sad_array[7] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
}
void vp8_sad16x8x3_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
+ sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
}
void vp8_sad16x8x8_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned short *sad_array)
{
- sad_array[0] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
- sad_array[3] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 3, ref_stride);
- sad_array[4] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride);
- sad_array[5] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride);
- sad_array[6] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 6, ref_stride);
- sad_array[7] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride);
+ sad_array[0] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+ sad_array[3] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+ sad_array[4] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+ sad_array[5] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+ sad_array[6] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+ sad_array[7] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
}
void vp8_sad8x8x3_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
+ sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
}
void vp8_sad8x8x8_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned short *sad_array)
{
- sad_array[0] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
- sad_array[3] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 3, ref_stride);
- sad_array[4] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride);
- sad_array[5] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride);
- sad_array[6] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 6, ref_stride);
- sad_array[7] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride);
+ sad_array[0] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+ sad_array[3] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+ sad_array[4] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+ sad_array[5] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+ sad_array[6] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+ sad_array[7] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
}
void vp8_sad8x16x3_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
+ sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
}
void vp8_sad8x16x8_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned short *sad_array)
{
- sad_array[0] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
- sad_array[3] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 3, ref_stride);
- sad_array[4] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride);
- sad_array[5] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride);
- sad_array[6] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 6, ref_stride);
- sad_array[7] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride);
+ sad_array[0] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+ sad_array[3] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+ sad_array[4] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+ sad_array[5] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+ sad_array[6] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+ sad_array[7] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
}
void vp8_sad4x4x3_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
+ sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
}
void vp8_sad4x4x8_c(const unsigned char *src_ptr, int src_stride,
const unsigned char *ref_ptr, int ref_stride,
unsigned short *sad_array)
{
- sad_array[0] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 0, ref_stride);
- sad_array[1] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride);
- sad_array[2] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride);
- sad_array[3] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 3, ref_stride);
- sad_array[4] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 4, ref_stride);
- sad_array[5] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 5, ref_stride);
- sad_array[6] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 6, ref_stride);
- sad_array[7] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 7, ref_stride);
+ sad_array[0] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+ sad_array[1] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+ sad_array[2] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+ sad_array[3] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+ sad_array[4] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+ sad_array[5] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+ sad_array[6] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+ sad_array[7] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
}
void vp8_sad16x16x4d_c(const unsigned char *src_ptr, int src_stride,
const unsigned char * const ref_ptr[], int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride);
- sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride);
- sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride);
- sad_array[3] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride);
+ sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+ sad_array[3] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
}
void vp8_sad16x8x4d_c(const unsigned char *src_ptr, int src_stride,
const unsigned char * const ref_ptr[], int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride);
- sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride);
- sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride);
- sad_array[3] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride);
+ sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+ sad_array[3] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
}
void vp8_sad8x8x4d_c(const unsigned char *src_ptr, int src_stride,
const unsigned char * const ref_ptr[], int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride);
- sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride);
- sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride);
- sad_array[3] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride);
+ sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+ sad_array[3] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
}
void vp8_sad8x16x4d_c(const unsigned char *src_ptr, int src_stride,
const unsigned char * const ref_ptr[], int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride);
- sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride);
- sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride);
- sad_array[3] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride);
+ sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+ sad_array[3] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
}
void vp8_sad4x4x4d_c(const unsigned char *src_ptr, int src_stride,
const unsigned char * const ref_ptr[], int ref_stride,
unsigned int *sad_array)
{
- sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[0], ref_stride);
- sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[1], ref_stride);
- sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[2], ref_stride);
- sad_array[3] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[3], ref_stride);
+ sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+ sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+ sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+ sad_array[3] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
}
/* Copy 2 macroblocks to a buffer */
diff --git a/vp8/common/variance.h b/vp8/common/variance.h
index 765c19a..89a32a7 100644
--- a/vp8/common/variance.h
+++ b/vp8/common/variance.h
@@ -22,7 +22,8 @@
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
- int ref_stride);
+ int ref_stride,
+ unsigned int max_sad);
typedef void (*vp8_copy32xn_fn_t)(
const unsigned char *src_ptr,
diff --git a/vp8/encoder/arm/neon/denoising_neon.c b/vp8/encoder/arm/neon/denoising_neon.c
index 1bebe8f..32ce65a 100644
--- a/vp8/encoder/arm/neon/denoising_neon.c
+++ b/vp8/encoder/arm/neon/denoising_neon.c
@@ -50,7 +50,8 @@
unsigned char *running_avg_y,
int running_avg_y_stride,
unsigned char *sig, int sig_stride,
- unsigned int motion_magnitude) {
+ unsigned int motion_magnitude,
+ int increase_denoising) {
/* 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.
diff --git a/vp8/encoder/block.h b/vp8/encoder/block.h
index dd733e5..34879cf 100644
--- a/vp8/encoder/block.h
+++ b/vp8/encoder/block.h
@@ -125,6 +125,7 @@
int optimize;
int q_index;
+ int increase_denoising;
#if CONFIG_TEMPORAL_DENOISING
MB_PREDICTION_MODE best_sse_inter_mode;
diff --git a/vp8/encoder/denoising.c b/vp8/encoder/denoising.c
index bfce280..1e645fb 100644
--- a/vp8/encoder/denoising.c
+++ b/vp8/encoder/denoising.c
@@ -21,6 +21,7 @@
*/
static const unsigned int SSE_DIFF_THRESHOLD = 16 * 16 * 20;
static const unsigned int SSE_THRESHOLD = 16 * 16 * 40;
+static const unsigned int SSE_THRESHOLD_HIGH = 16 * 16 * 60;
/*
* The filter function was modified to reduce the computational complexity.
@@ -54,20 +55,29 @@
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 int motion_magnitude,
+ int increase_denoising)
{
unsigned char *running_avg_y_start = running_avg_y;
unsigned char *sig_start = sig;
- int r, c, i;
+ int sum_diff_thresh;
+ int r, c;
int sum_diff = 0;
int adj_val[3] = {3, 4, 6};
-
+ int shift_inc1 = 0;
+ int shift_inc2 = 1;
/* If motion_magnitude is small, making the denoiser more aggressive by
- * increasing the adjustment for each level. */
+ * increasing the adjustment for each level. Add another increment for
+ * blocks that are labeled for increase denoising. */
if (motion_magnitude <= MOTION_MAGNITUDE_THRESHOLD)
{
- for (i = 0; i < 3; i++)
- adj_val[i] += 1;
+ if (increase_denoising) {
+ shift_inc1 = 1;
+ shift_inc2 = 2;
+ }
+ adj_val[0] += shift_inc2;
+ adj_val[1] += shift_inc2;
+ adj_val[2] += shift_inc2;
}
for (r = 0; r < 16; ++r)
@@ -81,8 +91,9 @@
diff = mc_running_avg_y[c] - sig[c];
absdiff = abs(diff);
- /* When |diff| < 4, use pixel value from last denoised raw. */
- if (absdiff <= 3)
+ // When |diff| <= |3 + shift_inc1|, use pixel value from
+ // last denoised raw.
+ if (absdiff <= 3 + shift_inc1)
{
running_avg_y[c] = mc_running_avg_y[c];
sum_diff += diff;
@@ -123,7 +134,9 @@
running_avg_y += avg_y_stride;
}
- if (abs(sum_diff) > SUM_DIFF_THRESHOLD)
+ sum_diff_thresh= SUM_DIFF_THRESHOLD;
+ if (increase_denoising) sum_diff_thresh = SUM_DIFF_THRESHOLD_HIGH;
+ if (abs(sum_diff) > sum_diff_thresh)
return COPY_BLOCK;
vp8_copy_mem16x16(running_avg_y_start, avg_y_stride, sig_start, sig_stride);
@@ -187,7 +200,7 @@
int mv_row;
int mv_col;
unsigned int motion_magnitude2;
-
+ unsigned int sse_thresh;
MV_REFERENCE_FRAME frame = x->best_reference_frame;
MV_REFERENCE_FRAME zero_frame = x->best_zeromv_reference_frame;
@@ -272,7 +285,10 @@
mv_row = x->best_sse_mv.as_mv.row;
mv_col = x->best_sse_mv.as_mv.col;
motion_magnitude2 = mv_row * mv_row + mv_col * mv_col;
- if (best_sse > SSE_THRESHOLD || motion_magnitude2
+ sse_thresh = SSE_THRESHOLD;
+ if (x->increase_denoising) sse_thresh = SSE_THRESHOLD_HIGH;
+
+ if (best_sse > sse_thresh || motion_magnitude2
> 8 * NOISE_MOTION_THRESHOLD)
{
decision = COPY_BLOCK;
@@ -290,7 +306,8 @@
/* Filter. */
decision = vp8_denoiser_filter(mc_running_avg_y, mc_avg_y_stride,
running_avg_y, avg_y_stride,
- x->thismb, 16, motion_magnitude2);
+ x->thismb, 16, motion_magnitude2,
+ x->increase_denoising);
}
if (decision == COPY_BLOCK)
{
diff --git a/vp8/encoder/denoising.h b/vp8/encoder/denoising.h
index cc9913a..ae744d2 100644
--- a/vp8/encoder/denoising.h
+++ b/vp8/encoder/denoising.h
@@ -18,6 +18,7 @@
#endif
#define SUM_DIFF_THRESHOLD (16 * 16 * 2)
+#define SUM_DIFF_THRESHOLD_HIGH (16 * 16 * 3)
#define MOTION_MAGNITUDE_THRESHOLD (8*3)
enum vp8_denoiser_decision
diff --git a/vp8/encoder/mcomp.c b/vp8/encoder/mcomp.c
index e7a1d76..54abe76 100644
--- a/vp8/encoder/mcomp.c
+++ b/vp8/encoder/mcomp.c
@@ -898,7 +898,7 @@
this_offset = base_offset + (br * (pre_stride)) + bc;
this_mv.as_mv.row = br;
this_mv.as_mv.col = bc;
- bestsad = vfp->sdf(what, what_stride, this_offset, in_what_stride)
+ bestsad = vfp->sdf(what, what_stride, this_offset, in_what_stride, UINT_MAX)
+ mvsad_err_cost(&this_mv, &fcenter_mv, mvsadcost, sad_per_bit);
#if CONFIG_MULTI_RES_ENCODING
@@ -923,7 +923,7 @@
this_mv.as_mv.row = br + hex[i].row;
this_mv.as_mv.col = bc + hex[i].col;
this_offset = base_offset + (this_mv.as_mv.row * in_what_stride) + this_mv.as_mv.col;
- thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride);
+ thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride, bestsad);
CHECK_BETTER
}
}else
@@ -934,7 +934,7 @@
this_mv.as_mv.col = bc + hex[i].col;
CHECK_POINT
this_offset = base_offset + (this_mv.as_mv.row * in_what_stride) + this_mv.as_mv.col;
- thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride);
+ thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride, bestsad);
CHECK_BETTER
}
}
@@ -960,8 +960,7 @@
this_mv.as_mv.row = br + next_chkpts[k][i].row;
this_mv.as_mv.col = bc + next_chkpts[k][i].col;
this_offset = base_offset + (this_mv.as_mv.row * (in_what_stride)) + this_mv.as_mv.col;
- thissad = vfp->sdf(what, what_stride, this_offset,
- in_what_stride);
+ thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride, bestsad);
CHECK_BETTER
}
}else
@@ -972,8 +971,7 @@
this_mv.as_mv.col = bc + next_chkpts[k][i].col;
CHECK_POINT
this_offset = base_offset + (this_mv.as_mv.row * (in_what_stride)) + this_mv.as_mv.col;
- thissad = vfp->sdf(what, what_stride, this_offset,
- in_what_stride);
+ thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride, bestsad);
CHECK_BETTER
}
}
@@ -1004,8 +1002,7 @@
this_mv.as_mv.row = br + neighbors[i].row;
this_mv.as_mv.col = bc + neighbors[i].col;
this_offset = base_offset + (this_mv.as_mv.row * (in_what_stride)) + this_mv.as_mv.col;
- thissad = vfp->sdf(what, what_stride, this_offset,
- in_what_stride);
+ thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride, bestsad);
CHECK_BETTER
}
}else
@@ -1016,8 +1013,7 @@
this_mv.as_mv.col = bc + neighbors[i].col;
CHECK_POINT
this_offset = base_offset + (this_mv.as_mv.row * (in_what_stride)) + this_mv.as_mv.col;
- thissad = vfp->sdf(what, what_stride, this_offset,
- in_what_stride);
+ thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride, bestsad);
CHECK_BETTER
}
}
@@ -1101,7 +1097,7 @@
best_address = in_what;
/* Check the starting position */
- bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride)
+ bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride, UINT_MAX)
+ mvsad_err_cost(best_mv, &fcenter_mv, mvsadcost, sad_per_bit);
/* search_param determines the length of the initial step and hence
@@ -1126,8 +1122,7 @@
{
check_here = ss[i].offset + best_address;
- thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ thissad = fn_ptr->sdf(what, what_stride, check_here, in_what_stride, bestsad);
if (thissad < bestsad)
{
@@ -1226,7 +1221,7 @@
best_address = in_what;
/* Check the starting position */
- bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride)
+ bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride, UINT_MAX)
+ mvsad_err_cost(best_mv, &fcenter_mv, mvsadcost, sad_per_bit);
/* search_param determines the length of the initial step and hence the
@@ -1294,8 +1289,7 @@
(this_row_offset > x->mv_row_min) && (this_row_offset < x->mv_row_max))
{
check_here = ss[i].offset + best_address;
- thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ thissad = fn_ptr->sdf(what, what_stride, check_here, in_what_stride, bestsad);
if (thissad < bestsad)
{
@@ -1379,7 +1373,7 @@
/* Baseline value at the centre */
bestsad = fn_ptr->sdf(what, what_stride, bestaddress,
- in_what_stride)
+ in_what_stride, UINT_MAX)
+ mvsad_err_cost(best_mv, &fcenter_mv, mvsadcost, sad_per_bit);
/* Apply further limits to prevent us looking using vectors that
@@ -1404,8 +1398,7 @@
for (c = col_min; c < col_max; c++)
{
- thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ thissad = fn_ptr->sdf(what, what_stride, check_here, in_what_stride, bestsad);
this_mv.as_mv.col = c;
thissad += mvsad_err_cost(&this_mv, &fcenter_mv,
@@ -1478,7 +1471,7 @@
/* Baseline value at the centre */
bestsad = fn_ptr->sdf(what, what_stride, bestaddress,
- in_what_stride)
+ in_what_stride, UINT_MAX)
+ mvsad_err_cost(best_mv, &fcenter_mv, mvsadcost, sad_per_bit);
/* Apply further limits to prevent us looking using vectors that stretch
@@ -1534,8 +1527,7 @@
while (c < col_max)
{
- thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ thissad = fn_ptr->sdf(what, what_stride, check_here, in_what_stride, bestsad);
if (thissad < bestsad)
{
@@ -1614,7 +1606,7 @@
/* Baseline value at the centre */
bestsad = fn_ptr->sdf(what, what_stride,
- bestaddress, in_what_stride)
+ bestaddress, in_what_stride, UINT_MAX)
+ mvsad_err_cost(best_mv, &fcenter_mv, mvsadcost, sad_per_bit);
/* Apply further limits to prevent us looking using vectors that stretch
@@ -1700,8 +1692,7 @@
while (c < col_max)
{
- thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ thissad = fn_ptr->sdf(what, what_stride, check_here , in_what_stride, bestsad);
if (thissad < bestsad)
{
@@ -1760,7 +1751,7 @@
fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3;
bestsad = fn_ptr->sdf(what, what_stride, best_address,
- in_what_stride)
+ in_what_stride, UINT_MAX)
+ mvsad_err_cost(ref_mv, &fcenter_mv, mvsadcost, error_per_bit);
for (i=0; i<search_range; i++)
@@ -1776,8 +1767,7 @@
(this_row_offset > x->mv_row_min) && (this_row_offset < x->mv_row_max))
{
check_here = (neighbors[j].row)*in_what_stride + neighbors[j].col + best_address;
- thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ thissad = fn_ptr->sdf(what, what_stride, check_here , in_what_stride, bestsad);
if (thissad < bestsad)
{
@@ -1841,7 +1831,7 @@
fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3;
bestsad = fn_ptr->sdf(what, what_stride, best_address,
- in_what_stride)
+ in_what_stride, UINT_MAX)
+ mvsad_err_cost(ref_mv, &fcenter_mv, mvsadcost, error_per_bit);
for (i=0; i<search_range; i++)
@@ -1892,8 +1882,7 @@
(this_row_offset > x->mv_row_min) && (this_row_offset < x->mv_row_max))
{
check_here = (neighbors[j].row)*in_what_stride + neighbors[j].col + best_address;
- thissad = fn_ptr->sdf(what, what_stride, check_here,
- in_what_stride);
+ thissad = fn_ptr->sdf(what, what_stride, check_here , in_what_stride, bestsad);
if (thissad < bestsad)
{
diff --git a/vp8/encoder/pickinter.c b/vp8/encoder/pickinter.c
index 39a3baf..cf6a82f 100644
--- a/vp8/encoder/pickinter.c
+++ b/vp8/encoder/pickinter.c
@@ -1177,6 +1177,7 @@
x->best_reference_frame = best_mbmode.ref_frame;
best_sse = best_rd_sse;
}
+ x->increase_denoising = 0;
vp8_denoiser_denoise_mb(&cpi->denoiser, x, best_sse, zero_mv_sse,
recon_yoffset, recon_uvoffset);
diff --git a/vp8/encoder/rdopt.c b/vp8/encoder/rdopt.c
index b273483..387701c 100644
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -1686,25 +1686,16 @@
}else if(xd->mb_to_top_edge==0)
{ /* only has left MB for sad calculation. */
near_sad[0] = near_sad[2] = INT_MAX;
- near_sad[1] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, xd->dst.y_buffer - 16, xd->dst.y_stride);
+ near_sad[1] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - 16,xd->dst.y_stride, UINT_MAX);
}else if(xd->mb_to_left_edge ==0)
{ /* only has left MB for sad calculation. */
near_sad[1] = near_sad[2] = INT_MAX;
- near_sad[0] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16,
- xd->dst.y_stride);
+ near_sad[0] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16,xd->dst.y_stride, UINT_MAX);
}else
{
- near_sad[0] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16,
- xd->dst.y_stride);
- near_sad[1] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, xd->dst.y_buffer - 16, xd->dst.y_stride);
- near_sad[2] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr,
- b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16 -16,
- xd->dst.y_stride);
+ near_sad[0] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16,xd->dst.y_stride, UINT_MAX);
+ near_sad[1] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - 16,xd->dst.y_stride, UINT_MAX);
+ near_sad[2] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16 -16,xd->dst.y_stride, UINT_MAX);
}
if(cpi->common.last_frame_type != KEY_FRAME)
@@ -1719,21 +1710,14 @@
if(xd->mb_to_bottom_edge==0) near_sad[7] = INT_MAX;
if(near_sad[4] != INT_MAX)
- near_sad[4] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, pre_y_buffer - pre_y_stride *16,
- pre_y_stride);
+ near_sad[4] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer - pre_y_stride *16, pre_y_stride, UINT_MAX);
if(near_sad[5] != INT_MAX)
- near_sad[5] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, pre_y_buffer - 16, pre_y_stride);
- near_sad[3] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, pre_y_buffer, pre_y_stride);
+ near_sad[5] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer - 16, pre_y_stride, UINT_MAX);
+ near_sad[3] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer, pre_y_stride, UINT_MAX);
if(near_sad[6] != INT_MAX)
- near_sad[6] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, pre_y_buffer + 16, pre_y_stride);
+ near_sad[6] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer + 16, pre_y_stride, UINT_MAX);
if(near_sad[7] != INT_MAX)
- near_sad[7] = cpi->fn_ptr[BLOCK_16X16].sdf(
- src_y_ptr, b->src_stride, pre_y_buffer + pre_y_stride *16,
- pre_y_stride);
+ near_sad[7] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer + pre_y_stride *16, pre_y_stride, UINT_MAX);
}
if(cpi->common.last_frame_type != KEY_FRAME)
diff --git a/vp8/encoder/x86/denoising_sse2.c b/vp8/encoder/x86/denoising_sse2.c
index d1f76b2..5112f89 100644
--- a/vp8/encoder/x86/denoising_sse2.c
+++ b/vp8/encoder/x86/denoising_sse2.c
@@ -26,19 +26,24 @@
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 int motion_magnitude,
+ int increase_denoising)
{
unsigned char *running_avg_y_start = running_avg_y;
unsigned char *sig_start = sig;
+ int sum_diff_thresh;
int r;
+ int shift_inc = (increase_denoising &&
+ motion_magnitude <= MOTION_MAGNITUDE_THRESHOLD) ? 1 : 0;
__m128i acc_diff = _mm_setzero_si128();
const __m128i k_0 = _mm_setzero_si128();
- const __m128i k_4 = _mm_set1_epi8(4);
+ const __m128i k_4 = _mm_set1_epi8(4 + shift_inc);
const __m128i k_8 = _mm_set1_epi8(8);
const __m128i k_16 = _mm_set1_epi8(16);
/* Modify each level's adjustment according to motion_magnitude. */
const __m128i l3 = _mm_set1_epi8(
- (motion_magnitude <= MOTION_MAGNITUDE_THRESHOLD) ? 7 : 6);
+ (motion_magnitude <= MOTION_MAGNITUDE_THRESHOLD) ?
+ 7 + shift_inc : 6);
/* Difference between level 3 and level 2 is 2. */
const __m128i l32 = _mm_set1_epi8(2);
/* Difference between level 2 and level 1 is 1. */
@@ -105,7 +110,9 @@
+ s.e[6] + s.e[7] + s.e[8] + s.e[9] + s.e[10] + s.e[11]
+ s.e[12] + s.e[13] + s.e[14] + s.e[15];
- if (abs(sum_diff) > SUM_DIFF_THRESHOLD)
+ sum_diff_thresh = SUM_DIFF_THRESHOLD;
+ if (increase_denoising) sum_diff_thresh = SUM_DIFF_THRESHOLD_HIGH;
+ if (abs(sum_diff) > sum_diff_thresh)
{
return COPY_BLOCK;
}
diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index 911608c..8282547 100644
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -158,10 +158,10 @@
VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6$(ASM)
# common (neon)
-VP8_COMMON_SRCS-$(ARCH_NEON_ASM) += common/arm/reconintra_arm.c
+#VP8_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/reconintra_arm.c
VP8_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/loopfilter_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/loopfiltersimpleverticaledge_neon$(ASM)
-VP8_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/buildintrapredictorsmby_neon$(ASM)
+#VP8_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/buildintrapredictorsmby_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/idct_blk_neon.c
VP8_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/idct_dequant_0_2x_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/idct_dequant_full_2x_neon$(ASM)
diff --git a/vp9/common/vp9_common.h b/vp9/common/vp9_common.h
index 2dccb70..04db7c0 100644
--- a/vp9/common/vp9_common.h
+++ b/vp9/common/vp9_common.h
@@ -45,7 +45,7 @@
vpx_memcpy(dest, src, n * sizeof(*src)); \
}
-#define vp9_zero(dest) vpx_memset(&dest, 0, sizeof(dest))
+#define vp9_zero(dest) vpx_memset(&(dest), 0, sizeof(dest))
#define vp9_zero_array(dest, n) vpx_memset(dest, 0, n * sizeof(*dest))
static INLINE uint8_t clip_pixel(int val) {
diff --git a/vp9/common/vp9_debugmodes.c b/vp9/common/vp9_debugmodes.c
index 8f150a4..d2522bb 100644
--- a/vp9/common/vp9_debugmodes.c
+++ b/vp9/common/vp9_debugmodes.c
@@ -24,10 +24,9 @@
*/
static void print_mi_data(VP9_COMMON *cm, FILE *file, const char *descriptor,
size_t member_offset) {
- int mi_row;
- int mi_col;
+ int mi_row, mi_col;
int mi_index = 0;
- MODE_INFO **mi_8x8 = cm->mi_grid_visible;
+ MODE_INFO **mi = cm->mi_grid_visible;
int rows = cm->mi_rows;
int cols = cm->mi_cols;
char prefix = descriptor[0];
@@ -38,7 +37,7 @@
fprintf(file, "%c ", prefix);
for (mi_col = 0; mi_col < cols; mi_col++) {
fprintf(file, "%2d ",
- *((int*) ((char *) (&mi_8x8[mi_index]->mbmi) +
+ *((int*) ((char *) (&mi[mi_index]->mbmi) +
member_offset)));
mi_index++;
}
@@ -52,7 +51,7 @@
int mi_col;
int mi_index = 0;
FILE *mvs = fopen(file, "a");
- MODE_INFO **mi_8x8 = cm->mi_grid_visible;
+ MODE_INFO **mi = cm->mi_grid_visible;
int rows = cm->mi_rows;
int cols = cm->mi_cols;
@@ -67,8 +66,8 @@
for (mi_row = 0; mi_row < rows; mi_row++) {
fprintf(mvs, "V ");
for (mi_col = 0; mi_col < cols; mi_col++) {
- fprintf(mvs, "%4d:%4d ", mi_8x8[mi_index]->mbmi.mv[0].as_mv.row,
- mi_8x8[mi_index]->mbmi.mv[0].as_mv.col);
+ fprintf(mvs, "%4d:%4d ", mi[mi_index]->mbmi.mv[0].as_mv.row,
+ mi[mi_index]->mbmi.mv[0].as_mv.col);
mi_index++;
}
fprintf(mvs, "\n");
diff --git a/vp9/common/vp9_loopfilter.c b/vp9/common/vp9_loopfilter.c
index 5b43e23..efd0249 100644
--- a/vp9/common/vp9_loopfilter.c
+++ b/vp9/common/vp9_loopfilter.c
@@ -619,12 +619,12 @@
// by mi_row, mi_col.
// TODO(JBB): This function only works for yv12.
void vp9_setup_mask(VP9_COMMON *const cm, const int mi_row, const int mi_col,
- MODE_INFO **mi_8x8, const int mode_info_stride,
+ 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_8x8;
- MODE_INFO **mip2 = mi_8x8;
+ MODE_INFO **mip = mi;
+ MODE_INFO **mip2 = mi;
// These are offsets to the next mi in the 64x64 block. It is what gets
// added to the mi ptr as we go through each loop. It helps us to avoids
@@ -1192,32 +1192,33 @@
}
void vp9_loop_filter_rows(const YV12_BUFFER_CONFIG *frame_buffer,
- VP9_COMMON *cm, MACROBLOCKD *xd,
+ VP9_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;
- int mi_row, mi_col;
+ const int use_420 = y_only || (planes[1].subsampling_y == 1 &&
+ planes[1].subsampling_x == 1);
LOOP_FILTER_MASK lfm;
- int use_420 = y_only || (xd->plane[1].subsampling_y == 1 &&
- xd->plane[1].subsampling_x == 1);
+ int mi_row, mi_col;
for (mi_row = start; mi_row < stop; mi_row += MI_BLOCK_SIZE) {
- MODE_INFO **mi_8x8 = cm->mi_grid_visible + mi_row * cm->mi_stride;
+ MODE_INFO **mi = cm->mi_grid_visible + mi_row * cm->mi_stride;
for (mi_col = 0; mi_col < cm->mi_cols; mi_col += MI_BLOCK_SIZE) {
int plane;
- vp9_setup_dst_planes(xd, frame_buffer, mi_row, mi_col);
+ vp9_setup_dst_planes(planes, frame_buffer, mi_row, mi_col);
// TODO(JBB): Make setup_mask work for non 420.
if (use_420)
- vp9_setup_mask(cm, mi_row, mi_col, mi_8x8 + mi_col, cm->mi_stride,
+ vp9_setup_mask(cm, mi_row, mi_col, mi + mi_col, cm->mi_stride,
&lfm);
for (plane = 0; plane < num_planes; ++plane) {
if (use_420)
- vp9_filter_block_plane(cm, &xd->plane[plane], mi_row, &lfm);
+ vp9_filter_block_plane(cm, &planes[plane], mi_row, &lfm);
else
- filter_block_plane_non420(cm, &xd->plane[plane], mi_8x8 + mi_col,
+ filter_block_plane_non420(cm, &planes[plane], mi + mi_col,
mi_row, mi_col);
}
}
@@ -1239,7 +1240,7 @@
}
end_mi_row = start_mi_row + mi_rows_to_filter;
vp9_loop_filter_frame_init(cm, frame_filter_level);
- vp9_loop_filter_rows(frame, cm, xd,
+ vp9_loop_filter_rows(frame, cm, xd->plane,
start_mi_row, end_mi_row,
y_only);
}
@@ -1247,7 +1248,7 @@
int vp9_loop_filter_worker(void *arg1, void *arg2) {
LFWorkerData *const lf_data = (LFWorkerData*)arg1;
(void)arg2;
- vp9_loop_filter_rows(lf_data->frame_buffer, lf_data->cm, &lf_data->xd,
+ vp9_loop_filter_rows(lf_data->frame_buffer, lf_data->cm, lf_data->planes,
lf_data->start, lf_data->stop, lf_data->y_only);
return 1;
}
diff --git a/vp9/common/vp9_loopfilter.h b/vp9/common/vp9_loopfilter.h
index 83463c5..6fa2773 100644
--- a/vp9/common/vp9_loopfilter.h
+++ b/vp9/common/vp9_loopfilter.h
@@ -112,15 +112,15 @@
// Apply the loop filter to [start, stop) macro block rows in frame_buffer.
void vp9_loop_filter_rows(const YV12_BUFFER_CONFIG *frame_buffer,
- struct VP9Common *cm, struct macroblockd *xd,
+ struct VP9Common *cm,
+ struct macroblockd_plane planes[MAX_MB_PLANE],
int start, int stop, int y_only);
typedef struct LoopFilterWorkerData {
const YV12_BUFFER_CONFIG *frame_buffer;
struct VP9Common *cm;
- struct macroblockd xd; // TODO(jzern): most of this is unnecessary to the
- // loopfilter. the planes are necessary as their state
- // is changed during decode.
+ struct macroblockd_plane planes[MAX_MB_PLANE];
+
int start;
int stop;
int y_only;
diff --git a/vp9/common/vp9_postproc.c b/vp9/common/vp9_postproc.c
index 5601a93..9f32104 100644
--- a/vp9/common/vp9_postproc.c
+++ b/vp9/common/vp9_postproc.c
@@ -24,61 +24,7 @@
#include "vp9/common/vp9_systemdependent.h"
#include "vp9/common/vp9_textblit.h"
-#define RGB_TO_YUV(t) \
- ( (0.257*(float)(t >> 16)) + (0.504*(float)(t >> 8 & 0xff)) + \
- (0.098*(float)(t & 0xff)) + 16), \
- (-(0.148*(float)(t >> 16)) - (0.291*(float)(t >> 8 & 0xff)) + \
- (0.439*(float)(t & 0xff)) + 128), \
- ( (0.439*(float)(t >> 16)) - (0.368*(float)(t >> 8 & 0xff)) - \
- (0.071*(float)(t & 0xff)) + 128)
-
-/* global constants */
-#if 0 && CONFIG_POSTPROC_VISUALIZER
-static const unsigned char PREDICTION_MODE_colors[MB_MODE_COUNT][3] = {
- { RGB_TO_YUV(0x98FB98) }, /* PaleGreen */
- { RGB_TO_YUV(0x00FF00) }, /* Green */
- { RGB_TO_YUV(0xADFF2F) }, /* GreenYellow */
- { RGB_TO_YUV(0x8F0000) }, /* Dark Red */
- { RGB_TO_YUV(0x008F8F) }, /* Dark Cyan */
- { RGB_TO_YUV(0x008F8F) }, /* Dark Cyan */
- { RGB_TO_YUV(0x008F8F) }, /* Dark Cyan */
- { RGB_TO_YUV(0x8F0000) }, /* Dark Red */
- { RGB_TO_YUV(0x8F0000) }, /* Dark Red */
- { RGB_TO_YUV(0x228B22) }, /* ForestGreen */
- { RGB_TO_YUV(0x006400) }, /* DarkGreen */
- { RGB_TO_YUV(0x98F5FF) }, /* Cadet Blue */
- { RGB_TO_YUV(0x6CA6CD) }, /* Sky Blue */
- { RGB_TO_YUV(0x00008B) }, /* Dark blue */
- { RGB_TO_YUV(0x551A8B) }, /* Purple */
- { RGB_TO_YUV(0xFF0000) } /* Red */
- { RGB_TO_YUV(0xCC33FF) }, /* Magenta */
-};
-
-static const unsigned char B_PREDICTION_MODE_colors[INTRA_MODES][3] = {
- { RGB_TO_YUV(0x6633ff) }, /* Purple */
- { RGB_TO_YUV(0xcc33ff) }, /* Magenta */
- { RGB_TO_YUV(0xff33cc) }, /* Pink */
- { RGB_TO_YUV(0xff3366) }, /* Coral */
- { RGB_TO_YUV(0x3366ff) }, /* Blue */
- { RGB_TO_YUV(0xed00f5) }, /* Dark Blue */
- { RGB_TO_YUV(0x2e00b8) }, /* Dark Purple */
- { RGB_TO_YUV(0xff6633) }, /* Orange */
- { RGB_TO_YUV(0x33ccff) }, /* Light Blue */
- { RGB_TO_YUV(0x8ab800) }, /* Green */
- { RGB_TO_YUV(0xffcc33) }, /* Light Orange */
- { RGB_TO_YUV(0x33ffcc) }, /* Aqua */
- { RGB_TO_YUV(0x66ff33) }, /* Light Green */
- { RGB_TO_YUV(0xccff33) }, /* Yellow */
-};
-
-static const unsigned char MV_REFERENCE_FRAME_colors[MAX_REF_FRAMES][3] = {
- { RGB_TO_YUV(0x00ff00) }, /* Blue */
- { RGB_TO_YUV(0x0000ff) }, /* Green */
- { RGB_TO_YUV(0xffff00) }, /* Yellow */
- { RGB_TO_YUV(0xff0000) }, /* Red */
-};
-#endif
-
+#if CONFIG_VP9_POSTPROC
static const short kernel5[] = {
1, 1, 4, 1, 1
};
@@ -448,163 +394,6 @@
}
}
-/* Blend the macro block with a solid colored square. Leave the
- * edges unblended to give distinction to macro blocks in areas
- * filled with the same color block.
- */
-void vp9_blend_mb_inner_c(uint8_t *y, uint8_t *u, uint8_t *v,
- int y1, int u1, int v1, int alpha, int stride) {
- int i, j;
- int y1_const = y1 * ((1 << 16) - alpha);
- int u1_const = u1 * ((1 << 16) - alpha);
- int v1_const = v1 * ((1 << 16) - alpha);
-
- y += 2 * stride + 2;
- for (i = 0; i < 12; i++) {
- for (j = 0; j < 12; j++) {
- y[j] = (y[j] * alpha + y1_const) >> 16;
- }
- y += stride;
- }
-
- stride >>= 1;
-
- u += stride + 1;
- v += stride + 1;
-
- for (i = 0; i < 6; i++) {
- for (j = 0; j < 6; j++) {
- u[j] = (u[j] * alpha + u1_const) >> 16;
- v[j] = (v[j] * alpha + v1_const) >> 16;
- }
- u += stride;
- v += stride;
- }
-}
-
-/* Blend only the edge of the macro block. Leave center
- * unblended to allow for other visualizations to be layered.
- */
-void vp9_blend_mb_outer_c(uint8_t *y, uint8_t *u, uint8_t *v,
- int y1, int u1, int v1, int alpha, int stride) {
- int i, j;
- int y1_const = y1 * ((1 << 16) - alpha);
- int u1_const = u1 * ((1 << 16) - alpha);
- int v1_const = v1 * ((1 << 16) - alpha);
-
- for (i = 0; i < 2; i++) {
- for (j = 0; j < 16; j++) {
- y[j] = (y[j] * alpha + y1_const) >> 16;
- }
- y += stride;
- }
-
- for (i = 0; i < 12; i++) {
- y[0] = (y[0] * alpha + y1_const) >> 16;
- y[1] = (y[1] * alpha + y1_const) >> 16;
- y[14] = (y[14] * alpha + y1_const) >> 16;
- y[15] = (y[15] * alpha + y1_const) >> 16;
- y += stride;
- }
-
- for (i = 0; i < 2; i++) {
- for (j = 0; j < 16; j++) {
- y[j] = (y[j] * alpha + y1_const) >> 16;
- }
- y += stride;
- }
-
- stride >>= 1;
-
- for (j = 0; j < 8; j++) {
- u[j] = (u[j] * alpha + u1_const) >> 16;
- v[j] = (v[j] * alpha + v1_const) >> 16;
- }
- u += stride;
- v += stride;
-
- for (i = 0; i < 6; i++) {
- u[0] = (u[0] * alpha + u1_const) >> 16;
- v[0] = (v[0] * alpha + v1_const) >> 16;
-
- u[7] = (u[7] * alpha + u1_const) >> 16;
- v[7] = (v[7] * alpha + v1_const) >> 16;
-
- u += stride;
- v += stride;
- }
-
- for (j = 0; j < 8; j++) {
- u[j] = (u[j] * alpha + u1_const) >> 16;
- v[j] = (v[j] * alpha + v1_const) >> 16;
- }
-}
-
-void vp9_blend_b_c(uint8_t *y, uint8_t *u, uint8_t *v,
- int y1, int u1, int v1, int alpha, int stride) {
- int i, j;
- int y1_const = y1 * ((1 << 16) - alpha);
- int u1_const = u1 * ((1 << 16) - alpha);
- int v1_const = v1 * ((1 << 16) - alpha);
-
- for (i = 0; i < 4; i++) {
- for (j = 0; j < 4; j++) {
- y[j] = (y[j] * alpha + y1_const) >> 16;
- }
- y += stride;
- }
-
- stride >>= 1;
-
- for (i = 0; i < 2; i++) {
- for (j = 0; j < 2; j++) {
- u[j] = (u[j] * alpha + u1_const) >> 16;
- v[j] = (v[j] * alpha + v1_const) >> 16;
- }
- u += stride;
- v += stride;
- }
-}
-
-static void constrain_line(int x0, int *x1, int y0, int *y1,
- int width, int height) {
- int dx;
- int dy;
-
- if (*x1 > width) {
- dx = *x1 - x0;
- dy = *y1 - y0;
-
- *x1 = width;
- if (dx)
- *y1 = ((width - x0) * dy) / dx + y0;
- }
- if (*x1 < 0) {
- dx = *x1 - x0;
- dy = *y1 - y0;
-
- *x1 = 0;
- if (dx)
- *y1 = ((0 - x0) * dy) / dx + y0;
- }
- if (*y1 > height) {
- dx = *x1 - x0;
- dy = *y1 - y0;
-
- *y1 = height;
- if (dy)
- *x1 = ((height - y0) * dx) / dy + x0;
- }
- if (*y1 < 0) {
- dx = *x1 - x0;
- dy = *y1 - y0;
-
- *y1 = 0;
- if (dy)
- *x1 = ((0 - y0) * dx) / dy + x0;
- }
-}
-
int vp9_post_proc_frame(struct VP9Common *cm,
YV12_BUFFER_CONFIG *dest, vp9_ppflags_t *ppflags) {
const int q = MIN(63, cm->lf.filter_level * 10 / 6);
@@ -643,328 +432,6 @@
ppbuf->y_width, ppbuf->y_height, ppbuf->y_stride);
}
-#if 0 && CONFIG_POSTPROC_VISUALIZER
- if (flags & VP9D_DEBUG_TXT_FRAME_INFO) {
- char message[512];
- snprintf(message, sizeof(message) -1,
- "F%1dG%1dQ%3dF%3dP%d_s%dx%d",
- (cm->frame_type == KEY_FRAME),
- cm->refresh_golden_frame,
- cm->base_qindex,
- cm->filter_level,
- flags,
- cm->mb_cols, cm->mb_rows);
- vp9_blit_text(message, ppbuf->y_buffer, ppbuf->y_stride);
- }
-
- if (flags & VP9D_DEBUG_TXT_MBLK_MODES) {
- int i, j;
- uint8_t *y_ptr;
- int mb_rows = ppbuf->y_height >> 4;
- int mb_cols = ppbuf->y_width >> 4;
- int mb_index = 0;
- MODE_INFO *mi = cm->mi;
-
- y_ptr = post->y_buffer + 4 * post->y_stride + 4;
-
- /* vp9_filter each macro block */
- for (i = 0; i < mb_rows; i++) {
- for (j = 0; j < mb_cols; j++) {
- char zz[4];
-
- snprintf(zz, sizeof(zz) - 1, "%c", mi[mb_index].mbmi.mode + 'a');
-
- vp9_blit_text(zz, y_ptr, post->y_stride);
- mb_index++;
- y_ptr += 16;
- }
-
- mb_index++; /* border */
- y_ptr += post->y_stride * 16 - post->y_width;
- }
- }
-
- if (flags & VP9D_DEBUG_TXT_DC_DIFF) {
- int i, j;
- uint8_t *y_ptr;
- int mb_rows = ppbuf->y_height >> 4;
- int mb_cols = ppbuf->y_width >> 4;
- int mb_index = 0;
- MODE_INFO *mi = cm->mi;
-
- y_ptr = post->y_buffer + 4 * post->y_stride + 4;
-
- /* vp9_filter each macro block */
- for (i = 0; i < mb_rows; i++) {
- for (j = 0; j < mb_cols; j++) {
- char zz[4];
- int dc_diff = !(mi[mb_index].mbmi.mode != I4X4_PRED &&
- mi[mb_index].mbmi.mode != SPLITMV &&
- mi[mb_index].mbmi.skip);
-
- if (cm->frame_type == KEY_FRAME)
- snprintf(zz, sizeof(zz) - 1, "a");
- else
- snprintf(zz, sizeof(zz) - 1, "%c", dc_diff + '0');
-
- vp9_blit_text(zz, y_ptr, post->y_stride);
- mb_index++;
- y_ptr += 16;
- }
-
- mb_index++; /* border */
- y_ptr += post->y_stride * 16 - post->y_width;
- }
- }
-
- if (flags & VP9D_DEBUG_TXT_RATE_INFO) {
- char message[512];
- snprintf(message, sizeof(message),
- "Bitrate: %10.2f framerate: %10.2f ",
- cm->bitrate, cm->framerate);
- vp9_blit_text(message, ppbuf->y_buffer, ppbuf->y_stride);
- }
-
- /* Draw motion vectors */
- if ((flags & VP9D_DEBUG_DRAW_MV) && ppflags->display_mv_flag) {
- int width = ppbuf->y_width;
- int height = ppbuf->y_height;
- uint8_t *y_buffer = ppbuf->y_buffer;
- int y_stride = ppbuf->y_stride;
- MODE_INFO *mi = cm->mi;
- int x0, y0;
-
- for (y0 = 0; y0 < height; y0 += 16) {
- for (x0 = 0; x0 < width; x0 += 16) {
- int x1, y1;
-
- if (!(ppflags->display_mv_flag & (1 << mi->mbmi.mode))) {
- mi++;
- continue;
- }
-
- if (mi->mbmi.mode == SPLITMV) {
- switch (mi->mbmi.partitioning) {
- case PARTITIONING_16X8 : { /* mv_top_bottom */
- union b_mode_info *bmi = &mi->bmi[0];
- MV *mv = &bmi->mv.as_mv;
-
- x1 = x0 + 8 + (mv->col >> 3);
- y1 = y0 + 4 + (mv->row >> 3);
-
- constrain_line(x0 + 8, &x1, y0 + 4, &y1, width, height);
- vp9_blit_line(x0 + 8, x1, y0 + 4, y1, y_buffer, y_stride);
-
- bmi = &mi->bmi[8];
-
- x1 = x0 + 8 + (mv->col >> 3);
- y1 = y0 + 12 + (mv->row >> 3);
-
- constrain_line(x0 + 8, &x1, y0 + 12, &y1, width, height);
- vp9_blit_line(x0 + 8, x1, y0 + 12, y1, y_buffer, y_stride);
-
- break;
- }
- case PARTITIONING_8X16 : { /* mv_left_right */
- union b_mode_info *bmi = &mi->bmi[0];
- MV *mv = &bmi->mv.as_mv;
-
- x1 = x0 + 4 + (mv->col >> 3);
- y1 = y0 + 8 + (mv->row >> 3);
-
- constrain_line(x0 + 4, &x1, y0 + 8, &y1, width, height);
- vp9_blit_line(x0 + 4, x1, y0 + 8, y1, y_buffer, y_stride);
-
- bmi = &mi->bmi[2];
-
- x1 = x0 + 12 + (mv->col >> 3);
- y1 = y0 + 8 + (mv->row >> 3);
-
- constrain_line(x0 + 12, &x1, y0 + 8, &y1, width, height);
- vp9_blit_line(x0 + 12, x1, y0 + 8, y1, y_buffer, y_stride);
-
- break;
- }
- case PARTITIONING_8X8 : { /* mv_quarters */
- union b_mode_info *bmi = &mi->bmi[0];
- MV *mv = &bmi->mv.as_mv;
-
- x1 = x0 + 4 + (mv->col >> 3);
- y1 = y0 + 4 + (mv->row >> 3);
-
- constrain_line(x0 + 4, &x1, y0 + 4, &y1, width, height);
- vp9_blit_line(x0 + 4, x1, y0 + 4, y1, y_buffer, y_stride);
-
- bmi = &mi->bmi[2];
-
- x1 = x0 + 12 + (mv->col >> 3);
- y1 = y0 + 4 + (mv->row >> 3);
-
- constrain_line(x0 + 12, &x1, y0 + 4, &y1, width, height);
- vp9_blit_line(x0 + 12, x1, y0 + 4, y1, y_buffer, y_stride);
-
- bmi = &mi->bmi[8];
-
- x1 = x0 + 4 + (mv->col >> 3);
- y1 = y0 + 12 + (mv->row >> 3);
-
- constrain_line(x0 + 4, &x1, y0 + 12, &y1, width, height);
- vp9_blit_line(x0 + 4, x1, y0 + 12, y1, y_buffer, y_stride);
-
- bmi = &mi->bmi[10];
-
- x1 = x0 + 12 + (mv->col >> 3);
- y1 = y0 + 12 + (mv->row >> 3);
-
- constrain_line(x0 + 12, &x1, y0 + 12, &y1, width, height);
- vp9_blit_line(x0 + 12, x1, y0 + 12, y1, y_buffer, y_stride);
- break;
- }
- case PARTITIONING_4X4:
- default : {
- union b_mode_info *bmi = mi->bmi;
- int bx0, by0;
-
- for (by0 = y0; by0 < (y0 + 16); by0 += 4) {
- for (bx0 = x0; bx0 < (x0 + 16); bx0 += 4) {
- MV *mv = &bmi->mv.as_mv;
-
- x1 = bx0 + 2 + (mv->col >> 3);
- y1 = by0 + 2 + (mv->row >> 3);
-
- constrain_line(bx0 + 2, &x1, by0 + 2, &y1, width, height);
- vp9_blit_line(bx0 + 2, x1, by0 + 2, y1, y_buffer, y_stride);
-
- bmi++;
- }
- }
- }
- }
- } else if (is_inter_mode(mi->mbmi.mode)) {
- MV *mv = &mi->mbmi.mv.as_mv;
- const int lx0 = x0 + 8;
- const int ly0 = y0 + 8;
-
- x1 = lx0 + (mv->col >> 3);
- y1 = ly0 + (mv->row >> 3);
-
- if (x1 != lx0 && y1 != ly0) {
- constrain_line(lx0, &x1, ly0 - 1, &y1, width, height);
- vp9_blit_line(lx0, x1, ly0 - 1, y1, y_buffer, y_stride);
-
- constrain_line(lx0, &x1, ly0 + 1, &y1, width, height);
- vp9_blit_line(lx0, x1, ly0 + 1, y1, y_buffer, y_stride);
- } else {
- vp9_blit_line(lx0, x1, ly0, y1, y_buffer, y_stride);
- }
- }
-
- mi++;
- }
- mi++;
- }
- }
-
- /* Color in block modes */
- if ((flags & VP9D_DEBUG_CLR_BLK_MODES)
- && (ppflags->display_mb_modes_flag || ppflags->display_b_modes_flag)) {
- int y, x;
- int width = ppbuf->y_width;
- int height = ppbuf->y_height;
- uint8_t *y_ptr = ppbuf->y_buffer;
- uint8_t *u_ptr = ppbuf->u_buffer;
- uint8_t *v_ptr = ppbuf->v_buffer;
- int y_stride = ppbuf->y_stride;
- MODE_INFO *mi = cm->mi;
-
- for (y = 0; y < height; y += 16) {
- for (x = 0; x < width; x += 16) {
- int Y = 0, U = 0, V = 0;
-
- if (mi->mbmi.mode == I4X4_PRED &&
- ((ppflags->display_mb_modes_flag & I4X4_PRED) ||
- ppflags->display_b_modes_flag)) {
- int by, bx;
- uint8_t *yl, *ul, *vl;
- union b_mode_info *bmi = mi->bmi;
-
- yl = y_ptr + x;
- ul = u_ptr + (x >> 1);
- vl = v_ptr + (x >> 1);
-
- for (by = 0; by < 16; by += 4) {
- for (bx = 0; bx < 16; bx += 4) {
- if ((ppflags->display_b_modes_flag & (1 << mi->mbmi.mode))
- || (ppflags->display_mb_modes_flag & I4X4_PRED)) {
- Y = B_PREDICTION_MODE_colors[bmi->as_mode][0];
- U = B_PREDICTION_MODE_colors[bmi->as_mode][1];
- V = B_PREDICTION_MODE_colors[bmi->as_mode][2];
-
- vp9_blend_b(yl + bx, ul + (bx >> 1), vl + (bx >> 1), Y, U, V,
- 0xc000, y_stride);
- }
- bmi++;
- }
-
- yl += y_stride * 4;
- ul += y_stride * 1;
- vl += y_stride * 1;
- }
- } else if (ppflags->display_mb_modes_flag & (1 << mi->mbmi.mode)) {
- Y = PREDICTION_MODE_colors[mi->mbmi.mode][0];
- U = PREDICTION_MODE_colors[mi->mbmi.mode][1];
- V = PREDICTION_MODE_colors[mi->mbmi.mode][2];
-
- vp9_blend_mb_inner(y_ptr + x, u_ptr + (x >> 1), v_ptr + (x >> 1),
- Y, U, V, 0xc000, y_stride);
- }
-
- mi++;
- }
- y_ptr += y_stride * 16;
- u_ptr += y_stride * 4;
- v_ptr += y_stride * 4;
-
- mi++;
- }
- }
-
- /* Color in frame reference blocks */
- if ((flags & VP9D_DEBUG_CLR_FRM_REF_BLKS) &&
- ppflags->display_ref_frame_flag) {
- int y, x;
- int width = ppbuf->y_width;
- int height = ppbuf->y_height;
- uint8_t *y_ptr = ppbuf->y_buffer;
- uint8_t *u_ptr = ppbuf->u_buffer;
- uint8_t *v_ptr = ppbuf->v_buffer;
- int y_stride = ppbuf->y_stride;
- MODE_INFO *mi = cm->mi;
-
- for (y = 0; y < height; y += 16) {
- for (x = 0; x < width; x += 16) {
- int Y = 0, U = 0, V = 0;
-
- if (ppflags->display_ref_frame_flag & (1 << mi->mbmi.ref_frame)) {
- Y = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][0];
- U = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][1];
- V = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][2];
-
- vp9_blend_mb_outer(y_ptr + x, u_ptr + (x >> 1), v_ptr + (x >> 1),
- Y, U, V, 0xc000, y_stride);
- }
-
- mi++;
- }
- y_ptr += y_stride * 16;
- u_ptr += y_stride * 4;
- v_ptr += y_stride * 4;
-
- mi++;
- }
- }
-#endif
-
*dest = *ppbuf;
/* handle problem with extending borders */
@@ -975,3 +442,4 @@
return 0;
}
+#endif
diff --git a/vp9/common/vp9_ppflags.h b/vp9/common/vp9_ppflags.h
index e8b04d2..1644a1b 100644
--- a/vp9/common/vp9_ppflags.h
+++ b/vp9/common/vp9_ppflags.h
@@ -33,12 +33,6 @@
int post_proc_flag;
int deblocking_level;
int noise_level;
-#if CONFIG_POSTPROC_VISUALIZER
- int display_ref_frame_flag;
- int display_mb_modes_flag;
- int display_b_modes_flag;
- int display_mv_flag;
-#endif // CONFIG_POSTPROC_VISUALIZER
} vp9_ppflags_t;
#ifdef __cplusplus
diff --git a/vp9/common/vp9_reconinter.c b/vp9/common/vp9_reconinter.c
index e722d6a..edc36d7 100644
--- a/vp9/common/vp9_reconinter.c
+++ b/vp9/common/vp9_reconinter.c
@@ -409,7 +409,7 @@
}
}
-void vp9_setup_dst_planes(MACROBLOCKD *xd,
+void vp9_setup_dst_planes(struct macroblockd_plane planes[MAX_MB_PLANE],
const YV12_BUFFER_CONFIG *src,
int mi_row, int mi_col) {
uint8_t *const buffers[4] = {src->y_buffer, src->u_buffer, src->v_buffer,
@@ -419,7 +419,7 @@
int i;
for (i = 0; i < MAX_MB_PLANE; ++i) {
- struct macroblockd_plane *const pd = &xd->plane[i];
+ struct macroblockd_plane *const pd = &planes[i];
setup_pred_plane(&pd->dst, buffers[i], strides[i], mi_row, mi_col, NULL,
pd->subsampling_x, pd->subsampling_y);
}
diff --git a/vp9/common/vp9_reconinter.h b/vp9/common/vp9_reconinter.h
index 86f3158..58c596e 100644
--- a/vp9/common/vp9_reconinter.h
+++ b/vp9/common/vp9_reconinter.h
@@ -57,7 +57,8 @@
dst->stride = stride;
}
-void vp9_setup_dst_planes(MACROBLOCKD *xd, const YV12_BUFFER_CONFIG *src,
+void vp9_setup_dst_planes(struct macroblockd_plane planes[MAX_MB_PLANE],
+ const YV12_BUFFER_CONFIG *src,
int mi_row, int mi_col);
void vp9_setup_pre_planes(MACROBLOCKD *xd, int idx,
diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl
index 47366b6..1037bfb 100644
--- a/vp9/common/vp9_rtcd_defs.pl
+++ b/vp9/common/vp9_rtcd_defs.pl
@@ -520,82 +520,82 @@
add_proto qw/unsigned int vp9_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance4x4/, "$sse_x86inc", "$ssse3_x86inc";
-add_proto qw/unsigned int vp9_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad64x64/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad32x64/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad64x32/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad32x16/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad16x32/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad32x32/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad16x16 mmx/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad16x8 mmx/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad8x16 mmx/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad8x8 mmx/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad8x4/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad4x8/, "$sse_x86inc";
-add_proto qw/unsigned int vp9_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+add_proto qw/unsigned int vp9_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp9_sad4x4 mmx/, "$sse_x86inc";
-add_proto qw/unsigned int vp9_sad64x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad64x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad64x64_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad32x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad32x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad32x64_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad64x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad64x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad64x32_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad32x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad32x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad32x16_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad16x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad16x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad16x32_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad32x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad32x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad32x32_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad16x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad16x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad16x16_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad16x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad16x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad16x8_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad8x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad8x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad8x16_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad8x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad8x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad8x8_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad8x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad8x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad8x4_avg/, "$sse2_x86inc";
-add_proto qw/unsigned int vp9_sad4x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad4x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad4x8_avg/, "$sse_x86inc";
-add_proto qw/unsigned int vp9_sad4x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+add_proto qw/unsigned int vp9_sad4x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad4x4_avg/, "$sse_x86inc";
add_proto qw/void vp9_sad64x64x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
@@ -703,7 +703,7 @@
# ENCODEMB INVOKE
add_proto qw/int64_t vp9_block_error/, "const int16_t *coeff, const int16_t *dqcoeff, intptr_t block_size, int64_t *ssz";
-specialize qw/vp9_block_error/, "$sse2_x86inc";
+specialize qw/vp9_block_error avx2/, "$sse2_x86inc";
add_proto qw/void vp9_subtract_block/, "int rows, int cols, int16_t *diff_ptr, ptrdiff_t diff_stride, const uint8_t *src_ptr, ptrdiff_t src_stride, const uint8_t *pred_ptr, ptrdiff_t pred_stride";
specialize qw/vp9_subtract_block/, "$sse2_x86inc";
diff --git a/vp9/decoder/vp9_decodeframe.c b/vp9/decoder/vp9_decodeframe.c
index 9dc0cf1..3124158 100644
--- a/vp9/decoder/vp9_decodeframe.c
+++ b/vp9/decoder/vp9_decodeframe.c
@@ -316,7 +316,7 @@
// as they are always compared to values that are in 1/8th pel units
set_mi_row_col(xd, tile, mi_row, bh, mi_col, bw, cm->mi_rows, cm->mi_cols);
- vp9_setup_dst_planes(xd, get_frame_new_buffer(cm), mi_row, mi_col);
+ vp9_setup_dst_planes(xd->plane, get_frame_new_buffer(cm), mi_row, mi_col);
return &xd->mi[0]->mbmi;
}
@@ -686,7 +686,7 @@
LFWorkerData *const lf_data = (LFWorkerData*)pbi->lf_worker.data1;
lf_data->frame_buffer = get_frame_new_buffer(cm);
lf_data->cm = cm;
- lf_data->xd = pbi->mb;
+ vp9_copy(lf_data->planes, pbi->mb.plane);
lf_data->stop = 0;
lf_data->y_only = 0;
vp9_loop_filter_frame_init(cm, cm->lf.filter_level);
@@ -1279,6 +1279,7 @@
const uint8_t *data,
const uint8_t *data_end,
uint8_t *clear_data /* buffer size MAX_VP9_HEADER_SIZE */) {
+ vp9_zero(*rb);
rb->bit_offset = 0;
rb->error_handler = error_handler;
rb->error_handler_data = &pbi->common;
@@ -1299,7 +1300,7 @@
const uint8_t **p_data_end) {
VP9_COMMON *const cm = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
- struct vp9_read_bit_buffer rb = { 0 };
+ struct vp9_read_bit_buffer rb;
uint8_t clear_data[MAX_VP9_HEADER_SIZE];
const size_t first_partition_size = read_uncompressed_header(pbi,
init_read_bit_buffer(pbi, &rb, data, data_end, clear_data));
@@ -1325,16 +1326,6 @@
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
"Truncated packet or corrupt header length");
- if (do_loopfilter_inline && pbi->lf_worker.data1 == NULL) {
- CHECK_MEM_ERROR(cm, pbi->lf_worker.data1,
- vpx_memalign(32, sizeof(LFWorkerData)));
- pbi->lf_worker.hook = (VP9WorkerHook)vp9_loop_filter_worker;
- if (pbi->max_threads > 1 && !vp9_worker_reset(&pbi->lf_worker)) {
- vpx_internal_error(&cm->error, VPX_CODEC_ERROR,
- "Loop filter thread creation failed");
- }
- }
-
init_macroblockd(cm, &pbi->mb);
if (cm->coding_use_prev_mi)
@@ -1357,9 +1348,23 @@
if (pbi->max_threads > 1 && tile_rows == 1 && tile_cols > 1 &&
cm->frame_parallel_decoding_mode) {
*p_data_end = decode_tiles_mt(pbi, data + first_partition_size, data_end);
+ // If multiple threads are used to decode tiles, then we use those threads
+ // to do parallel loopfiltering.
+ vp9_loop_filter_frame_mt(new_fb, pbi, cm, cm->lf.filter_level, 0);
} else {
+ if (do_loopfilter_inline && pbi->lf_worker.data1 == NULL) {
+ CHECK_MEM_ERROR(cm, pbi->lf_worker.data1,
+ vpx_memalign(32, sizeof(LFWorkerData)));
+ pbi->lf_worker.hook = (VP9WorkerHook)vp9_loop_filter_worker;
+ if (pbi->max_threads > 1 && !vp9_worker_reset(&pbi->lf_worker)) {
+ vpx_internal_error(&cm->error, VPX_CODEC_ERROR,
+ "Loop filter thread creation failed");
+ }
+ }
*p_data_end = decode_tiles(pbi, data + first_partition_size, data_end,
do_loopfilter_inline);
+ if (!do_loopfilter_inline)
+ vp9_loop_filter_frame(new_fb, cm, &pbi->mb, cm->lf.filter_level, 0, 0);
}
new_fb->corrupted |= xd->corrupted;
@@ -1388,16 +1393,5 @@
if (cm->refresh_frame_context)
cm->frame_contexts[cm->frame_context_idx] = cm->fc;
- // Loopfilter
- if (!do_loopfilter_inline) {
- // If multiple threads are used to decode tiles, then we use those threads
- // to do parallel loopfiltering.
- if (pbi->num_tile_workers) {
- vp9_loop_filter_frame_mt(new_fb, pbi, cm, cm->lf.filter_level, 0, 0);
- } else {
- vp9_loop_filter_frame(new_fb, cm, &pbi->mb, cm->lf.filter_level, 0, 0);
- }
- }
-
return 0;
}
diff --git a/vp9/decoder/vp9_decoder.c b/vp9/decoder/vp9_decoder.c
index b67c7d6..98b890b 100644
--- a/vp9/decoder/vp9_decoder.c
+++ b/vp9/decoder/vp9_decoder.c
@@ -304,11 +304,14 @@
int64_t *time_stamp, int64_t *time_end_stamp,
vp9_ppflags_t *flags) {
int ret = -1;
+#if !CONFIG_VP9_POSTPROC
+ (void)*flags;
+#endif
if (pbi->ready_for_new_data == 1)
return ret;
- /* ie no raw frame to show!!! */
+ /* no raw frame to show!!! */
if (pbi->common.show_frame == 0)
return ret;
@@ -319,8 +322,8 @@
#if CONFIG_VP9_POSTPROC
ret = vp9_post_proc_frame(&pbi->common, sd, flags);
#else
- *sd = *pbi->common.frame_to_show;
- ret = 0;
+ *sd = *pbi->common.frame_to_show;
+ ret = 0;
#endif /*!CONFIG_POSTPROC*/
vp9_clear_system_state();
return ret;
diff --git a/vp9/decoder/vp9_dthread.c b/vp9/decoder/vp9_dthread.c
index 5fe5ed7..bc6c418 100644
--- a/vp9/decoder/vp9_dthread.c
+++ b/vp9/decoder/vp9_dthread.c
@@ -89,7 +89,8 @@
// Implement row loopfiltering for each thread.
static void loop_filter_rows_mt(const YV12_BUFFER_CONFIG *const frame_buffer,
- VP9_COMMON *const cm, MACROBLOCKD *const xd,
+ VP9_COMMON *const cm,
+ struct macroblockd_plane planes[MAX_MB_PLANE],
int start, int stop, int y_only,
VP9LfSync *const lf_sync, int num_lf_workers) {
const int num_planes = y_only ? 1 : MAX_MB_PLANE;
@@ -107,11 +108,11 @@
sync_read(lf_sync, r, c);
- vp9_setup_dst_planes(xd, frame_buffer, mi_row, mi_col);
+ vp9_setup_dst_planes(planes, frame_buffer, mi_row, mi_col);
vp9_setup_mask(cm, mi_row, mi_col, mi + mi_col, cm->mi_stride, &lfm);
for (plane = 0; plane < num_planes; ++plane) {
- vp9_filter_block_plane(cm, &xd->plane[plane], mi_row, &lfm);
+ vp9_filter_block_plane(cm, &planes[plane], mi_row, &lfm);
}
sync_write(lf_sync, r, c, sb_cols);
@@ -124,7 +125,7 @@
TileWorkerData *const tile_data = (TileWorkerData*)arg1;
LFWorkerData *const lf_data = &tile_data->lfdata;
- loop_filter_rows_mt(lf_data->frame_buffer, lf_data->cm, &lf_data->xd,
+ loop_filter_rows_mt(lf_data->frame_buffer, lf_data->cm, lf_data->planes,
lf_data->start, lf_data->stop, lf_data->y_only,
lf_data->lf_sync, lf_data->num_lf_workers);
return 1;
@@ -135,7 +136,7 @@
void vp9_loop_filter_frame_mt(YV12_BUFFER_CONFIG *frame,
VP9Decoder *pbi, VP9_COMMON *cm,
int frame_filter_level,
- int y_only, int partial_frame) {
+ int y_only) {
VP9LfSync *const lf_sync = &pbi->lf_row_sync;
// Number of superblock rows and cols
const int sb_rows = mi_cols_aligned_to_sb(cm->mi_rows) >> MI_BLOCK_SIZE_LOG2;
@@ -186,7 +187,7 @@
// Loopfilter data
lf_data->frame_buffer = frame;
lf_data->cm = cm;
- lf_data->xd = pbi->mb;
+ vp9_copy(lf_data->planes, pbi->mb.plane);
lf_data->start = i;
lf_data->stop = sb_rows;
lf_data->y_only = y_only; // always do all planes in decoder
diff --git a/vp9/decoder/vp9_dthread.h b/vp9/decoder/vp9_dthread.h
index c3b7a29..a727e2a 100644
--- a/vp9/decoder/vp9_dthread.h
+++ b/vp9/decoder/vp9_dthread.h
@@ -52,6 +52,6 @@
struct VP9Decoder *pbi,
struct VP9Common *cm,
int frame_filter_level,
- int y_only, int partial_frame);
+ int y_only);
#endif // VP9_DECODER_VP9_DTHREAD_H_
diff --git a/vp9/encoder/vp9_block.h b/vp9/encoder/vp9_block.h
index 2ccf4f8..0504a49 100644
--- a/vp9/encoder/vp9_block.h
+++ b/vp9/encoder/vp9_block.h
@@ -20,43 +20,6 @@
extern "C" {
#endif
-// Structure to hold snapshot of coding context during the mode picking process
-typedef struct {
- MODE_INFO mic;
- uint8_t *zcoeff_blk;
- int16_t *coeff[MAX_MB_PLANE][3];
- int16_t *qcoeff[MAX_MB_PLANE][3];
- int16_t *dqcoeff[MAX_MB_PLANE][3];
- uint16_t *eobs[MAX_MB_PLANE][3];
-
- // dual buffer pointers, 0: in use, 1: best in store
- int16_t *coeff_pbuf[MAX_MB_PLANE][3];
- int16_t *qcoeff_pbuf[MAX_MB_PLANE][3];
- int16_t *dqcoeff_pbuf[MAX_MB_PLANE][3];
- uint16_t *eobs_pbuf[MAX_MB_PLANE][3];
-
- int is_coded;
- int num_4x4_blk;
- int skip;
- int_mv best_ref_mv[2];
- int_mv ref_mvs[MAX_REF_FRAMES][MAX_MV_REF_CANDIDATES];
- int rate;
- int distortion;
- int best_mode_index;
- int rddiv;
- int rdmult;
- int hybrid_pred_diff;
- int comp_pred_diff;
- int single_pred_diff;
- int64_t tx_rd_diff[TX_MODES];
- int64_t best_filter_diff[SWITCHABLE_FILTER_CONTEXTS];
-
- // motion vector cache for adaptive motion search control in partition
- // search loop
- int_mv pred_mv[MAX_REF_FRAMES];
- INTERP_FILTER pred_interp_filter;
-} PICK_MODE_CONTEXT;
-
struct macroblock_plane {
DECLARE_ALIGNED(16, int16_t, src_diff[64 * 64]);
int16_t *qcoeff;
@@ -73,18 +36,6 @@
// Zbin Over Quant value
int16_t zbin_extra;
};
-typedef struct PC_TREE {
- int index;
- PARTITION_TYPE partitioning;
- BLOCK_SIZE block_size;
- PICK_MODE_CONTEXT none;
- PICK_MODE_CONTEXT horizontal[2];
- PICK_MODE_CONTEXT vertical[2];
- union {
- struct PC_TREE *split[4];
- PICK_MODE_CONTEXT *leaf_split[4];
- };
-} PC_TREE;
/* The [2] dimension is for whether we skip the EOB node (i.e. if previous
* coefficient in this block was zero) or not. */
@@ -157,15 +108,9 @@
// Used to store sub partition's choices.
int_mv pred_mv[MAX_REF_FRAMES];
- PICK_MODE_CONTEXT *leaf_tree;
- PC_TREE *pc_tree;
- PC_TREE *pc_root;
- int partition_cost[PARTITION_CONTEXTS][PARTITION_TYPES];
-
void (*fwd_txm4x4)(const int16_t *input, int16_t *output, int stride);
};
-
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/vp9/encoder/vp9_context_tree.c b/vp9/encoder/vp9_context_tree.c
index ac9b562..9b7a932 100644
--- a/vp9/encoder/vp9_context_tree.c
+++ b/vp9/encoder/vp9_context_tree.c
@@ -9,6 +9,7 @@
*/
#include "vp9/encoder/vp9_context_tree.h"
+#include "vp9/encoder/vp9_encoder.h"
static const BLOCK_SIZE square[] = {
BLOCK_8X8,
@@ -86,7 +87,7 @@
// partition level. There are contexts for none, horizontal, vertical, and
// split. Along with a block_size value and a selected block_size which
// represents the state of our search.
-void vp9_setup_pc_tree(VP9_COMMON *cm, MACROBLOCK *x) {
+void vp9_setup_pc_tree(VP9_COMMON *cm, VP9_COMP *cpi) {
int i, j;
const int leaf_nodes = 64;
const int tree_nodes = 64 + 16 + 4 + 1;
@@ -96,23 +97,24 @@
int square_index = 1;
int nodes;
- vpx_free(x->leaf_tree);
- CHECK_MEM_ERROR(cm, x->leaf_tree, vpx_calloc(leaf_nodes,
- sizeof(*x->leaf_tree)));
- vpx_free(x->pc_tree);
- CHECK_MEM_ERROR(cm, x->pc_tree, vpx_calloc(tree_nodes, sizeof(*x->pc_tree)));
+ vpx_free(cpi->leaf_tree);
+ CHECK_MEM_ERROR(cm, cpi->leaf_tree, vpx_calloc(leaf_nodes,
+ sizeof(*cpi->leaf_tree)));
+ vpx_free(cpi->pc_tree);
+ CHECK_MEM_ERROR(cm, cpi->pc_tree, vpx_calloc(tree_nodes,
+ sizeof(*cpi->pc_tree)));
- this_pc = &x->pc_tree[0];
- this_leaf = &x->leaf_tree[0];
+ this_pc = &cpi->pc_tree[0];
+ this_leaf = &cpi->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, &x->leaf_tree[i]);
+ alloc_mode_context(cm, 1, &cpi->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) {
- PC_TREE *const tree = &x->pc_tree[pc_tree_index];
+ PC_TREE *const tree = &cpi->pc_tree[pc_tree_index];
tree->block_size = square[0];
alloc_tree_contexts(cm, tree, 4);
tree->leaf_split[0] = this_leaf++;
@@ -124,7 +126,7 @@
// from leafs to the root.
for (nodes = 16; nodes > 0; nodes >>= 2) {
for (i = 0; i < nodes; ++i) {
- PC_TREE *const tree = &x->pc_tree[pc_tree_index];
+ PC_TREE *const tree = &cpi->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++)
@@ -133,24 +135,24 @@
}
++square_index;
}
- x->pc_root = &x->pc_tree[tree_nodes - 1];
- x->pc_root[0].none.best_mode_index = 2;
+ cpi->pc_root = &cpi->pc_tree[tree_nodes - 1];
+ cpi->pc_root[0].none.best_mode_index = 2;
}
-void vp9_free_pc_tree(MACROBLOCK *x) {
+void vp9_free_pc_tree(VP9_COMP *cpi) {
const int tree_nodes = 64 + 16 + 4 + 1;
int i;
// Set up all 4x4 mode contexts
for (i = 0; i < 64; ++i)
- free_mode_context(&x->leaf_tree[i]);
+ free_mode_context(&cpi->leaf_tree[i]);
// Sets up all the leaf nodes in the tree.
for (i = 0; i < tree_nodes; ++i)
- free_tree_contexts(&x->pc_tree[i]);
+ free_tree_contexts(&cpi->pc_tree[i]);
- vpx_free(x->pc_tree);
- x->pc_tree = NULL;
- vpx_free(x->leaf_tree);
- x->leaf_tree = NULL;
+ vpx_free(cpi->pc_tree);
+ cpi->pc_tree = NULL;
+ vpx_free(cpi->leaf_tree);
+ cpi->leaf_tree = NULL;
}
diff --git a/vp9/encoder/vp9_context_tree.h b/vp9/encoder/vp9_context_tree.h
index 66a6f00..e12aac2 100644
--- a/vp9/encoder/vp9_context_tree.h
+++ b/vp9/encoder/vp9_context_tree.h
@@ -11,9 +11,61 @@
#ifndef VP9_ENCODER_VP9_CONTEXT_TREE_H_
#define VP9_ENCODER_VP9_CONTEXT_TREE_H_
-#include "vp9/encoder/vp9_encoder.h"
+#include "vp9/common/vp9_onyxc_int.h"
-void vp9_setup_pc_tree(VP9_COMMON *cm, MACROBLOCK *x);
-void vp9_free_pc_tree(MACROBLOCK *x);
+struct VP9_COMP;
+
+// Structure to hold snapshot of coding context during the mode picking process
+typedef struct {
+ MODE_INFO mic;
+ uint8_t *zcoeff_blk;
+ int16_t *coeff[MAX_MB_PLANE][3];
+ int16_t *qcoeff[MAX_MB_PLANE][3];
+ int16_t *dqcoeff[MAX_MB_PLANE][3];
+ uint16_t *eobs[MAX_MB_PLANE][3];
+
+ // dual buffer pointers, 0: in use, 1: best in store
+ int16_t *coeff_pbuf[MAX_MB_PLANE][3];
+ int16_t *qcoeff_pbuf[MAX_MB_PLANE][3];
+ int16_t *dqcoeff_pbuf[MAX_MB_PLANE][3];
+ uint16_t *eobs_pbuf[MAX_MB_PLANE][3];
+
+ int is_coded;
+ int num_4x4_blk;
+ int skip;
+ int_mv best_ref_mv[2];
+ int_mv ref_mvs[MAX_REF_FRAMES][MAX_MV_REF_CANDIDATES];
+ int rate;
+ int distortion;
+ int best_mode_index;
+ int rddiv;
+ int rdmult;
+ int hybrid_pred_diff;
+ int comp_pred_diff;
+ int single_pred_diff;
+ int64_t tx_rd_diff[TX_MODES];
+ int64_t best_filter_diff[SWITCHABLE_FILTER_CONTEXTS];
+
+ // motion vector cache for adaptive motion search control in partition
+ // search loop
+ int_mv pred_mv[MAX_REF_FRAMES];
+ INTERP_FILTER pred_interp_filter;
+} PICK_MODE_CONTEXT;
+
+typedef struct PC_TREE {
+ int index;
+ PARTITION_TYPE partitioning;
+ BLOCK_SIZE block_size;
+ PICK_MODE_CONTEXT none;
+ PICK_MODE_CONTEXT horizontal[2];
+ PICK_MODE_CONTEXT vertical[2];
+ union {
+ struct PC_TREE *split[4];
+ PICK_MODE_CONTEXT *leaf_split[4];
+ };
+} PC_TREE;
+
+void vp9_setup_pc_tree(struct VP9Common *cm, struct VP9_COMP *cpi);
+void vp9_free_pc_tree(struct VP9_COMP *cpi);
#endif /* VP9_ENCODER_VP9_CONTEXT_TREE_H_ */
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index c11f9f5..f7cb05b 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -201,7 +201,7 @@
mbmi = &xd->mi[0]->mbmi;
// Set up destination pointers.
- vp9_setup_dst_planes(xd, get_frame_new_buffer(cm), mi_row, mi_col);
+ vp9_setup_dst_planes(xd->plane, get_frame_new_buffer(cm), mi_row, mi_col);
// Set up limit values for MV components.
// Mv beyond the range do not produce new/different prediction block.
@@ -1297,14 +1297,14 @@
if (row8x8_remaining >= MI_BLOCK_SIZE &&
col8x8_remaining >= MI_BLOCK_SIZE) {
this_sad = cpi->fn_ptr[BLOCK_64X64].sdf(src, src_stride,
- pre, pre_stride);
+ pre, pre_stride, 0x7fffffff);
threshold = (1 << 12);
} else {
int r, c;
for (r = 0; r < row8x8_remaining; r += 2)
for (c = 0; c < col8x8_remaining; c += 2)
- this_sad += cpi->fn_ptr[BLOCK_16X16].sdf(src, src_stride,
- pre, pre_stride);
+ this_sad += cpi->fn_ptr[BLOCK_16X16].sdf(src, src_stride, pre,
+ pre_stride, 0x7fffffff);
threshold = (row8x8_remaining * col8x8_remaining) << 6;
}
@@ -1538,7 +1538,7 @@
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
if (none_rate < INT_MAX) {
- none_rate += x->partition_cost[pl][PARTITION_NONE];
+ none_rate += cpi->partition_cost[pl][PARTITION_NONE];
none_rd = RDCOST(x->rdmult, x->rddiv, none_rate, none_dist);
}
@@ -1636,7 +1636,7 @@
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
if (last_part_rate < INT_MAX) {
- last_part_rate += x->partition_cost[pl][partition];
+ last_part_rate += cpi->partition_cost[pl][partition];
last_part_rd = RDCOST(x->rdmult, x->rddiv, last_part_rate, last_part_dist);
}
@@ -1689,11 +1689,11 @@
pl = partition_plane_context(xd, mi_row + y_idx, mi_col + x_idx,
split_subsize);
- chosen_rate += x->partition_cost[pl][PARTITION_NONE];
+ chosen_rate += cpi->partition_cost[pl][PARTITION_NONE];
}
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
if (chosen_rate < INT_MAX) {
- chosen_rate += x->partition_cost[pl][PARTITION_SPLIT];
+ chosen_rate += cpi->partition_cost[pl][PARTITION_SPLIT];
chosen_rd = RDCOST(x->rdmult, x->rddiv, chosen_rate, chosen_dist);
}
}
@@ -1805,15 +1805,11 @@
BLOCK_SIZE *max_block_size) {
VP9_COMMON *const cm = &cpi->common;
MACROBLOCKD *const xd = &cpi->mb.e_mbd;
- MODE_INFO **mi_8x8 = xd->mi;
- const int left_in_image = xd->left_available && mi_8x8[-1];
- const int above_in_image = xd->up_available &&
- mi_8x8[-xd->mi_stride];
- MODE_INFO **above_sb64_mi_8x8;
- MODE_INFO **left_sb64_mi_8x8;
-
- int row8x8_remaining = tile->mi_row_end - mi_row;
- int col8x8_remaining = tile->mi_col_end - mi_col;
+ MODE_INFO **mi = xd->mi;
+ const int left_in_image = xd->left_available && mi[-1];
+ const int above_in_image = xd->up_available && mi[-xd->mi_stride];
+ const int row8x8_remaining = tile->mi_row_end - mi_row;
+ const int col8x8_remaining = tile->mi_col_end - mi_col;
int bh, bw;
BLOCK_SIZE min_size = BLOCK_4X4;
BLOCK_SIZE max_size = BLOCK_64X64;
@@ -1833,15 +1829,13 @@
}
// Find the min and max partition sizes used in the left SB64
if (left_in_image) {
- left_sb64_mi_8x8 = &mi_8x8[-MI_BLOCK_SIZE];
- get_sb_partition_size_range(cpi, left_sb64_mi_8x8,
- &min_size, &max_size);
+ MODE_INFO **left_sb64_mi = &mi[-MI_BLOCK_SIZE];
+ get_sb_partition_size_range(cpi, left_sb64_mi, &min_size, &max_size);
}
// Find the min and max partition sizes used in the above SB64.
if (above_in_image) {
- above_sb64_mi_8x8 = &mi_8x8[-xd->mi_stride * MI_BLOCK_SIZE];
- get_sb_partition_size_range(cpi, above_sb64_mi_8x8,
- &min_size, &max_size);
+ MODE_INFO **above_sb64_mi = &mi[-xd->mi_stride * MI_BLOCK_SIZE];
+ get_sb_partition_size_range(cpi, above_sb64_mi, &min_size, &max_size);
}
// adjust observed min and max
if (cpi->sf.auto_min_max_partition_size == RELAXED_NEIGHBORING_MIN_MAX) {
@@ -2021,7 +2015,7 @@
if (this_rate != INT_MAX) {
if (bsize >= BLOCK_8X8) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- this_rate += x->partition_cost[pl][PARTITION_NONE];
+ this_rate += cpi->partition_cost[pl][PARTITION_NONE];
}
sum_rd = RDCOST(x->rdmult, x->rddiv, this_rate, this_dist);
if (sum_rd < best_rd) {
@@ -2109,7 +2103,7 @@
if (sum_rd < best_rd && i == 4) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- sum_rate += x->partition_cost[pl][PARTITION_SPLIT];
+ sum_rate += cpi->partition_cost[pl][PARTITION_SPLIT];
sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
if (sum_rd < best_rd) {
best_rate = sum_rate;
@@ -2163,7 +2157,7 @@
}
if (sum_rd < best_rd) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- sum_rate += x->partition_cost[pl][PARTITION_HORZ];
+ sum_rate += cpi->partition_cost[pl][PARTITION_HORZ];
sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
if (sum_rd < best_rd) {
best_rd = sum_rd;
@@ -2212,7 +2206,7 @@
}
if (sum_rd < best_rd) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- sum_rate += x->partition_cost[pl][PARTITION_VERT];
+ sum_rate += cpi->partition_cost[pl][PARTITION_VERT];
sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
if (sum_rd < best_rd) {
best_rate = sum_rate;
@@ -2274,17 +2268,16 @@
int64_t dummy_dist;
int i;
- MACROBLOCK *x = &cpi->mb;
if (sf->adaptive_pred_interp_filter) {
for (i = 0; i < 64; ++i)
- x->leaf_tree[i].pred_interp_filter = SWITCHABLE;
+ cpi->leaf_tree[i].pred_interp_filter = SWITCHABLE;
for (i = 0; i < 64; ++i) {
- x->pc_tree[i].vertical[0].pred_interp_filter = SWITCHABLE;
- x->pc_tree[i].vertical[1].pred_interp_filter = SWITCHABLE;
- x->pc_tree[i].horizontal[0].pred_interp_filter = SWITCHABLE;
- x->pc_tree[i].horizontal[1].pred_interp_filter = SWITCHABLE;
+ cpi->pc_tree[i].vertical[0].pred_interp_filter = SWITCHABLE;
+ cpi->pc_tree[i].vertical[1].pred_interp_filter = SWITCHABLE;
+ cpi->pc_tree[i].horizontal[0].pred_interp_filter = SWITCHABLE;
+ cpi->pc_tree[i].horizontal[1].pred_interp_filter = SWITCHABLE;
}
}
@@ -2296,26 +2289,26 @@
sf->partition_search_type == VAR_BASED_PARTITION ||
sf->partition_search_type == VAR_BASED_FIXED_PARTITION) {
const int idx_str = cm->mi_stride * mi_row + mi_col;
- MODE_INFO **mi_8x8 = cm->mi_grid_visible + idx_str;
- MODE_INFO **prev_mi_8x8 = cm->prev_mi_grid_visible + idx_str;
+ MODE_INFO **mi = cm->mi_grid_visible + idx_str;
+ MODE_INFO **prev_mi = cm->prev_mi_grid_visible + idx_str;
cpi->mb.source_variance = UINT_MAX;
if (sf->partition_search_type == FIXED_PARTITION) {
set_offsets(cpi, tile, mi_row, mi_col, BLOCK_64X64);
- set_fixed_partitioning(cpi, tile, mi_8x8, mi_row, mi_col,
+ set_fixed_partitioning(cpi, tile, mi, mi_row, mi_col,
sf->always_this_block_size);
- rd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, 1, x->pc_root);
+ rd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ &dummy_rate, &dummy_dist, 1, cpi->pc_root);
} else if (sf->partition_search_type == VAR_BASED_FIXED_PARTITION) {
BLOCK_SIZE bsize;
set_offsets(cpi, tile, mi_row, mi_col, BLOCK_64X64);
bsize = get_rd_var_based_fixed_partition(cpi, mi_row, mi_col);
- set_fixed_partitioning(cpi, tile, mi_8x8, mi_row, mi_col, bsize);
- rd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, 1, x->pc_root);
+ set_fixed_partitioning(cpi, tile, mi, mi_row, mi_col, bsize);
+ rd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ &dummy_rate, &dummy_dist, 1, cpi->pc_root);
} else if (sf->partition_search_type == VAR_BASED_PARTITION) {
choose_partitioning(cpi, tile, mi_row, mi_col);
- rd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, 1, x->pc_root);
+ rd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ &dummy_rate, &dummy_dist, 1, cpi->pc_root);
} else {
if ((cm->current_video_frame
% sf->last_partitioning_redo_frequency) == 0
@@ -2325,7 +2318,7 @@
|| cpi->rc.is_src_frame_alt_ref
|| ((sf->use_lastframe_partitioning ==
LAST_FRAME_PARTITION_LOW_MOTION) &&
- sb_has_motion(cm, prev_mi_8x8))) {
+ sb_has_motion(cm, prev_mi))) {
// If required set upper and lower partition size limits
if (sf->auto_min_max_partition_size) {
set_offsets(cpi, tile, mi_row, mi_col, BLOCK_64X64);
@@ -2334,16 +2327,17 @@
&sf->max_partition_size);
}
rd_pick_partition(cpi, tile, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, 1, INT64_MAX, x->pc_root);
+ &dummy_rate, &dummy_dist, 1, INT64_MAX,
+ cpi->pc_root);
} else {
if (sf->constrain_copy_partition &&
- sb_has_motion(cm, prev_mi_8x8))
- constrain_copy_partitioning(cpi, tile, mi_8x8, prev_mi_8x8,
+ sb_has_motion(cm, prev_mi))
+ constrain_copy_partitioning(cpi, tile, mi, prev_mi,
mi_row, mi_col, BLOCK_16X16);
else
- copy_partitioning(cm, mi_8x8, prev_mi_8x8);
- rd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, 1, x->pc_root);
+ copy_partitioning(cm, mi, prev_mi);
+ rd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ &dummy_rate, &dummy_dist, 1, cpi->pc_root);
}
}
} else {
@@ -2355,7 +2349,7 @@
&sf->max_partition_size);
}
rd_pick_partition(cpi, tile, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, 1, INT64_MAX, x->pc_root);
+ &dummy_rate, &dummy_dist, 1, INT64_MAX, cpi->pc_root);
}
}
}
@@ -2633,7 +2627,7 @@
if (this_rate != INT_MAX) {
int pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- this_rate += x->partition_cost[pl][PARTITION_NONE];
+ this_rate += cpi->partition_cost[pl][PARTITION_NONE];
sum_rd = RDCOST(x->rdmult, x->rddiv, this_rate, this_dist);
if (sum_rd < best_rd) {
int64_t stop_thresh = 4096;
@@ -2671,7 +2665,7 @@
sum_rd = 0;
if (do_split) {
int pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- sum_rate += x->partition_cost[pl][PARTITION_SPLIT];
+ sum_rate += cpi->partition_cost[pl][PARTITION_SPLIT];
subsize = get_subsize(bsize, PARTITION_SPLIT);
for (i = 0; i < 4 && sum_rd < best_rd; ++i) {
const int x_idx = (i & 1) * ms;
@@ -2730,7 +2724,7 @@
sum_rd = INT64_MAX;
} else {
int pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- this_rate += x->partition_cost[pl][PARTITION_HORZ];
+ this_rate += cpi->partition_cost[pl][PARTITION_HORZ];
sum_rate += this_rate;
sum_dist += this_dist;
sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
@@ -2764,7 +2758,7 @@
sum_rd = INT64_MAX;
} else {
int pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- this_rate += x->partition_cost[pl][PARTITION_VERT];
+ this_rate += cpi->partition_cost[pl][PARTITION_VERT];
sum_rate += this_rate;
sum_dist += this_dist;
sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
@@ -2822,7 +2816,7 @@
static void nonrd_use_partition(VP9_COMP *cpi,
const TileInfo *const tile,
- MODE_INFO **mi_8x8,
+ MODE_INFO **mi,
TOKENEXTRA **tp,
int mi_row, int mi_col,
BLOCK_SIZE bsize, int output_enabled,
@@ -2841,7 +2835,7 @@
if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
return;
- subsize = (bsize >= BLOCK_8X8) ? mi_8x8[0]->mbmi.sb_type : BLOCK_4X4;
+ subsize = (bsize >= BLOCK_8X8) ? mi[0]->mbmi.sb_type : BLOCK_4X4;
partition = partition_lookup[bsl][subsize];
switch (partition) {
@@ -2869,7 +2863,7 @@
if (mi_row + hbs < cm->mi_rows) {
nonrd_pick_sb_modes(cpi, tile, mi_row + hbs, mi_col,
&rate, &dist, subsize);
- pc_tree->horizontal[1].mic.mbmi = mi_8x8[0]->mbmi;
+ pc_tree->horizontal[1].mic.mbmi = mi[0]->mbmi;
if (rate != INT_MAX && dist != INT64_MAX &&
*totrate != INT_MAX && *totdist != INT64_MAX) {
*totrate += rate;
@@ -2879,10 +2873,10 @@
break;
case PARTITION_SPLIT:
subsize = get_subsize(bsize, PARTITION_SPLIT);
- nonrd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col,
+ nonrd_use_partition(cpi, tile, mi, tp, mi_row, mi_col,
subsize, output_enabled, totrate, totdist,
pc_tree->split[0]);
- nonrd_use_partition(cpi, tile, mi_8x8 + hbs, tp,
+ nonrd_use_partition(cpi, tile, mi + hbs, tp,
mi_row, mi_col + hbs, subsize, output_enabled,
&rate, &dist, pc_tree->split[1]);
if (rate != INT_MAX && dist != INT64_MAX &&
@@ -2890,7 +2884,7 @@
*totrate += rate;
*totdist += dist;
}
- nonrd_use_partition(cpi, tile, mi_8x8 + hbs * mis, tp,
+ nonrd_use_partition(cpi, tile, mi + hbs * mis, tp,
mi_row + hbs, mi_col, subsize, output_enabled,
&rate, &dist, pc_tree->split[2]);
if (rate != INT_MAX && dist != INT64_MAX &&
@@ -2898,7 +2892,7 @@
*totrate += rate;
*totdist += dist;
}
- nonrd_use_partition(cpi, tile, mi_8x8 + hbs * mis + hbs, tp,
+ nonrd_use_partition(cpi, tile, mi + hbs * mis + hbs, tp,
mi_row + hbs, mi_col + hbs, subsize, output_enabled,
&rate, &dist, pc_tree->split[3]);
if (rate != INT_MAX && dist != INT64_MAX &&
@@ -2937,8 +2931,8 @@
int dummy_rate = 0;
int64_t dummy_dist = 0;
const int idx_str = cm->mi_stride * mi_row + mi_col;
- MODE_INFO **mi_8x8 = cm->mi_grid_visible + idx_str;
- MODE_INFO **prev_mi_8x8 = cm->prev_mi_grid_visible + idx_str;
+ MODE_INFO **mi = cm->mi_grid_visible + idx_str;
+ MODE_INFO **prev_mi = cm->prev_mi_grid_visible + idx_str;
BLOCK_SIZE bsize;
x->in_static_area = 0;
@@ -2949,22 +2943,22 @@
switch (cpi->sf.partition_search_type) {
case VAR_BASED_PARTITION:
choose_partitioning(cpi, tile, mi_row, mi_col);
- nonrd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col, BLOCK_64X64,
- 1, &dummy_rate, &dummy_dist, x->pc_root);
+ nonrd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ 1, &dummy_rate, &dummy_dist, cpi->pc_root);
break;
case SOURCE_VAR_BASED_PARTITION:
- set_source_var_based_partition(cpi, tile, mi_8x8, mi_row, mi_col);
- nonrd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col, BLOCK_64X64,
- 1, &dummy_rate, &dummy_dist, x->pc_root);
+ set_source_var_based_partition(cpi, tile, mi, mi_row, mi_col);
+ nonrd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ 1, &dummy_rate, &dummy_dist, cpi->pc_root);
break;
case VAR_BASED_FIXED_PARTITION:
case FIXED_PARTITION:
bsize = cpi->sf.partition_search_type == FIXED_PARTITION ?
cpi->sf.always_this_block_size :
get_nonrd_var_based_fixed_partition(cpi, mi_row, mi_col);
- set_fixed_partitioning(cpi, tile, mi_8x8, mi_row, mi_col, bsize);
- nonrd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col, BLOCK_64X64,
- 1, &dummy_rate, &dummy_dist, x->pc_root);
+ set_fixed_partitioning(cpi, tile, mi, mi_row, mi_col, bsize);
+ nonrd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ 1, &dummy_rate, &dummy_dist, cpi->pc_root);
break;
case REFERENCE_PARTITION:
if (cpi->sf.partition_check ||
@@ -2975,12 +2969,12 @@
&cpi->sf.max_partition_size);
nonrd_pick_partition(cpi, tile, tp, mi_row, mi_col, BLOCK_64X64,
&dummy_rate, &dummy_dist, 1, INT64_MAX,
- x->pc_root);
+ cpi->pc_root);
} else {
- copy_partitioning(cm, mi_8x8, prev_mi_8x8);
- nonrd_use_partition(cpi, tile, mi_8x8, tp, mi_row, mi_col,
+ copy_partitioning(cm, mi, prev_mi);
+ nonrd_use_partition(cpi, tile, mi, tp, mi_row, mi_col,
BLOCK_64X64, 1, &dummy_rate, &dummy_dist,
- x->pc_root);
+ cpi->pc_root);
}
break;
default:
@@ -3043,7 +3037,7 @@
int i;
struct macroblock_plane *const p = x->plane;
struct macroblockd_plane *const pd = xd->plane;
- PICK_MODE_CONTEXT *ctx = &x->pc_root->none;
+ PICK_MODE_CONTEXT *ctx = &cpi->pc_root->none;
for (i = 0; i < MAX_MB_PLANE; ++i) {
p[i].coeff = ctx->coeff_pbuf[i][0];
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index 00f6526..02536e3 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -187,7 +187,7 @@
vpx_free(cpi->tok);
cpi->tok = 0;
- vp9_free_pc_tree(&cpi->mb);
+ vp9_free_pc_tree(cpi);
for (i = 0; i < cpi->svc.number_spatial_layers; ++i) {
LAYER_CONTEXT *const lc = &cpi->svc.layer_context[i];
@@ -455,7 +455,7 @@
CHECK_MEM_ERROR(cm, cpi->tok, vpx_calloc(tokens, sizeof(*cpi->tok)));
}
- vp9_setup_pc_tree(&cpi->common, &cpi->mb);
+ vp9_setup_pc_tree(&cpi->common, cpi);
}
static void update_frame_size(VP9_COMP *cpi) {
@@ -2777,6 +2777,9 @@
int vp9_get_preview_raw_frame(VP9_COMP *cpi, YV12_BUFFER_CONFIG *dest,
vp9_ppflags_t *flags) {
VP9_COMMON *cm = &cpi->common;
+#if !CONFIG_VP9_POSTPROC
+ (void)flags;
+#endif
if (!cm->show_frame) {
return -1;
@@ -2785,7 +2788,6 @@
#if CONFIG_VP9_POSTPROC
ret = vp9_post_proc_frame(cm, dest, flags);
#else
-
if (cm->frame_to_show) {
*dest = *cm->frame_to_show;
dest->y_width = cm->width;
@@ -2796,7 +2798,6 @@
} else {
ret = -1;
}
-
#endif // !CONFIG_VP9_POSTPROC
vp9_clear_system_state();
return ret;
diff --git a/vp9/encoder/vp9_encoder.h b/vp9/encoder/vp9_encoder.h
index 17c826f..6ffbd3f 100644
--- a/vp9/encoder/vp9_encoder.h
+++ b/vp9/encoder/vp9_encoder.h
@@ -24,6 +24,7 @@
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/encoder/vp9_aq_cyclicrefresh.h"
+#include "vp9/encoder/vp9_context_tree.h"
#include "vp9/encoder/vp9_encodemb.h"
#include "vp9/encoder/vp9_firstpass.h"
#include "vp9/encoder/vp9_lookahead.h"
@@ -502,6 +503,11 @@
int y_mode_costs[INTRA_MODES][INTRA_MODES][INTRA_MODES];
int switchable_interp_costs[SWITCHABLE_FILTER_CONTEXTS][SWITCHABLE_FILTERS];
+ PICK_MODE_CONTEXT *leaf_tree;
+ PC_TREE *pc_tree;
+ PC_TREE *pc_root;
+ int partition_cost[PARTITION_CONTEXTS][PARTITION_TYPES];
+
#if CONFIG_MULTIPLE_ARF
// ARF tracking variables.
int multi_arf_enabled;
diff --git a/vp9/encoder/vp9_firstpass.c b/vp9/encoder/vp9_firstpass.c
index c6b6197..116a0b1 100644
--- a/vp9/encoder/vp9_firstpass.c
+++ b/vp9/encoder/vp9_firstpass.c
@@ -61,6 +61,7 @@
#define MIN_GF_INTERVAL 4
#endif
+
// #define LONG_TERM_VBR_CORRECTION
static void swap_yv12(YV12_BUFFER_CONFIG *a, YV12_BUFFER_CONFIG *b) {
@@ -475,7 +476,7 @@
TileInfo tile;
struct macroblock_plane *const p = x->plane;
struct macroblockd_plane *const pd = xd->plane;
- const PICK_MODE_CONTEXT *ctx = &x->pc_root->none;
+ const PICK_MODE_CONTEXT *ctx = &cpi->pc_root->none;
int i;
int recon_yoffset, recon_uvoffset;
@@ -540,7 +541,7 @@
vp9_setup_src_planes(x, cpi->Source, 0, 0);
vp9_setup_pre_planes(xd, 0, first_ref_buf, 0, 0, NULL);
- vp9_setup_dst_planes(xd, new_yv12, 0, 0);
+ vp9_setup_dst_planes(xd->plane, new_yv12, 0, 0);
xd->mi = cm->mi_grid_visible;
xd->mi[0] = cm->mi;
@@ -1421,10 +1422,13 @@
static void calculate_section_intra_ratio(struct twopass_rc *twopass,
const FIRSTPASS_STATS *start_pos,
int section_length) {
- FIRSTPASS_STATS next_frame = { 0 };
- FIRSTPASS_STATS sectionstats = { 0 };
+ FIRSTPASS_STATS next_frame;
+ FIRSTPASS_STATS sectionstats;
int i;
+ vp9_zero(next_frame);
+ vp9_zero(sectionstats);
+
reset_fpf_position(twopass, start_pos);
for (i = 0; i < section_length; ++i) {
@@ -1497,7 +1501,7 @@
RATE_CONTROL *const rc = &cpi->rc;
const VP9EncoderConfig *const oxcf = &cpi->oxcf;
struct twopass_rc *const twopass = &cpi->twopass;
- FIRSTPASS_STATS next_frame = { 0 };
+ FIRSTPASS_STATS next_frame;
const FIRSTPASS_STATS *start_pos;
int i;
double boost_score = 0.0;
@@ -1524,10 +1528,10 @@
int flash_detected;
int active_max_gf_interval;
- twopass->gf_group_bits = 0;
-
vp9_clear_system_state();
+ vp9_zero(next_frame);
+ twopass->gf_group_bits = 0;
start_pos = twopass->stats_in;
// Load stats for the current frame.
@@ -2149,8 +2153,8 @@
double this_frame_coded_error;
int target;
LAYER_CONTEXT *lc = NULL;
- int is_spatial_svc = (cpi->use_svc && cpi->svc.number_temporal_layers == 1);
-
+ const int is_spatial_svc = (cpi->use_svc &&
+ cpi->svc.number_temporal_layers == 1);
if (is_spatial_svc) {
lc = &cpi->svc.layer_context[cpi->svc.spatial_layer_id];
frames_left = (int)(twopass->total_stats.count -
@@ -2210,14 +2214,14 @@
this_frame_copy = this_frame;
find_next_key_frame(cpi, &this_frame_copy);
// Don't place key frame in any enhancement layers in spatial svc
- if (cpi->use_svc && cpi->svc.number_temporal_layers == 1) {
+ if (is_spatial_svc) {
lc->is_key_frame = 1;
if (cpi->svc.spatial_layer_id > 0) {
cm->frame_type = INTER_FRAME;
}
}
} else {
- if (cpi->use_svc && cpi->svc.number_temporal_layers == 1) {
+ if (is_spatial_svc) {
lc->is_key_frame = 0;
}
cm->frame_type = INTER_FRAME;
diff --git a/vp9/encoder/vp9_mbgraph.c b/vp9/encoder/vp9_mbgraph.c
index 643c742..5e87d28 100644
--- a/vp9/encoder/vp9_mbgraph.c
+++ b/vp9/encoder/vp9_mbgraph.c
@@ -20,7 +20,6 @@
#include "vp9/common/vp9_systemdependent.h"
-
static unsigned int do_16x16_motion_iteration(VP9_COMP *cpi,
const MV *ref_mv,
MV *dst_mv,
@@ -73,7 +72,8 @@
x->mv_row_max = tmp_row_max;
return vp9_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride,
- xd->plane[0].dst.buf, xd->plane[0].dst.stride);
+ xd->plane[0].dst.buf, xd->plane[0].dst.stride,
+ INT_MAX);
}
static int do_16x16_motion_search(VP9_COMP *cpi, const MV *ref_mv,
@@ -86,7 +86,8 @@
// Try zero MV first
// FIXME should really use something like near/nearest MV and/or MV prediction
err = vp9_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride,
- xd->plane[0].pre[0].buf, xd->plane[0].pre[0].stride);
+ xd->plane[0].pre[0].buf, xd->plane[0].pre[0].stride,
+ INT_MAX);
dst_mv->as_int = 0;
// Test last reference frame using the previous best mv as the
@@ -122,7 +123,8 @@
// Try zero MV first
// FIXME should really use something like near/nearest MV and/or MV prediction
err = vp9_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride,
- xd->plane[0].pre[0].buf, xd->plane[0].pre[0].stride);
+ xd->plane[0].pre[0].buf, xd->plane[0].pre[0].stride,
+ INT_MAX);
dst_mv->as_int = 0;
@@ -145,7 +147,7 @@
xd->plane[0].dst.buf, xd->plane[0].dst.stride,
0, 0, 0);
err = vp9_sad16x16(x->plane[0].src.buf, x->plane[0].src.stride,
- xd->plane[0].dst.buf, xd->plane[0].dst.stride);
+ xd->plane[0].dst.buf, xd->plane[0].dst.stride, best_err);
// find best
if (err < best_err) {
@@ -234,8 +236,9 @@
int mb_col, mb_row, offset = 0;
int mb_y_offset = 0, arf_y_offset = 0, gld_y_offset = 0;
MV arf_top_mv = {0, 0}, gld_top_mv = {0, 0};
- MODE_INFO mi_local = { { 0 } };
+ MODE_INFO mi_local;
+ vp9_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;
diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c
index 94f5cc0..4f7d6f1 100644
--- a/vp9/encoder/vp9_mcomp.c
+++ b/vp9/encoder/vp9_mcomp.c
@@ -524,7 +524,9 @@
// 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);
+ get_buf_from_mv(in_what, ref_mv), in_what->stride,
+ 0x7fffffff) + 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
@@ -540,7 +542,7 @@
bc + candidates[t][i].col};
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ in_what->stride, bestsad);
CHECK_BETTER
}
} else {
@@ -551,7 +553,7 @@
continue;
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ in_what->stride, bestsad);
CHECK_BETTER
}
}
@@ -583,7 +585,7 @@
bc + candidates[s][i].col};
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ in_what->stride, bestsad);
CHECK_BETTER
}
} else {
@@ -594,7 +596,7 @@
continue;
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ in_what->stride, bestsad);
CHECK_BETTER
}
}
@@ -621,7 +623,7 @@
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);
+ in_what->stride, bestsad);
CHECK_BETTER
}
} else {
@@ -632,7 +634,7 @@
continue;
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ in_what->stride, bestsad);
CHECK_BETTER
}
}
@@ -659,7 +661,7 @@
bc + neighbors[i].col};
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ in_what->stride, bestsad);
CHECK_BETTER
}
} else {
@@ -670,7 +672,7 @@
continue;
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
- in_what->stride);
+ in_what->stride, bestsad);
CHECK_BETTER
}
}
@@ -884,11 +886,16 @@
int r, c, i;
int start_col, end_col, start_row, end_row;
+ // The cfg and search_param parameters are not used in this search variant
+ (void)cfg;
+ (void)search_param;
+
clamp_mv(ref_mv, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max);
*best_mv = *ref_mv;
*num00 = 11;
best_sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, ref_mv), in_what->stride) +
+ get_buf_from_mv(in_what, ref_mv), in_what->stride,
+ 0x7fffffff) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
start_row = MAX(-range, x->mv_row_min - ref_mv->row);
start_col = MAX(-range, x->mv_col_min - ref_mv->col);
@@ -922,7 +929,7 @@
for (i = 0; i < end_col - c; ++i) {
const MV mv = {ref_mv->row + r, ref_mv->col + c + i};
unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &mv), in_what->stride);
+ get_buf_from_mv(in_what, &mv), in_what->stride, best_sad);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
@@ -968,7 +975,7 @@
// Check the starting position
best_sad = fn_ptr->sdf(what->buf, what->stride,
- best_address, in_what->stride) +
+ best_address, in_what->stride, 0x7fffffff) +
mvsad_err_cost(x, best_mv, &fcenter_mv, sad_per_bit);
i = 1;
@@ -979,7 +986,8 @@
best_mv->col + ss[i].mv.col};
if (is_mv_in(x, &mv)) {
int sad = fn_ptr->sdf(what->buf, what->stride,
- best_address + ss[i].offset, in_what->stride);
+ best_address + ss[i].offset, in_what->stride,
+ best_sad);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
@@ -1004,7 +1012,7 @@
if (is_mv_in(x, &this_mv)) {
int sad = fn_ptr->sdf(what->buf, what->stride,
best_address + ss[best_site].offset,
- in_what->stride);
+ in_what->stride, best_sad);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
@@ -1069,7 +1077,7 @@
best_address = in_what;
// Check the starting position
- bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride)
+ bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride, 0x7fffffff)
+ mvsad_err_cost(x, best_mv, &fcenter_mv, sad_per_bit);
i = 1;
@@ -1121,7 +1129,7 @@
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);
+ in_what_stride, bestsad);
if (thissad < bestsad) {
thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
@@ -1146,7 +1154,7 @@
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);
+ in_what_stride, bestsad);
if (thissad < bestsad) {
thissad += mvsad_err_cost(&this_mv, &fcenter_mv,
mvjsadcost, mvsadcost, sad_per_bit);
@@ -1245,7 +1253,7 @@
const int col_max = MIN(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) +
+ get_buf_from_mv(in_what, ref_mv), in_what->stride, 0x7fffffff) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
*best_mv = *ref_mv;
@@ -1253,7 +1261,7 @@
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) +
+ get_buf_from_mv(in_what, &mv), in_what->stride, best_sad) +
mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
if (sad < best_sad) {
best_sad = sad;
@@ -1278,7 +1286,7 @@
const int col_max = MIN(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) +
+ get_buf_from_mv(in_what, ref_mv), in_what->stride, 0x7fffffff) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
*best_mv = *ref_mv;
@@ -1312,7 +1320,7 @@
while (c < col_max) {
unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
- check_here, in_what->stride);
+ check_here, in_what->stride, best_sad);
if (sad < best_sad) {
const MV mv = {r, c};
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
@@ -1343,7 +1351,7 @@
const int col_max = MIN(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) +
+ get_buf_from_mv(in_what, ref_mv), in_what->stride, 0x7fffffff) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
*best_mv = *ref_mv;
@@ -1401,7 +1409,7 @@
while (c < col_max) {
unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
- check_here, in_what->stride);
+ check_here, in_what->stride, best_sad);
if (sad < best_sad) {
const MV mv = {r, c};
sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
@@ -1430,7 +1438,7 @@
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) +
+ in_what->stride, 0x7fffffff) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, error_per_bit);
int i, j;
@@ -1442,7 +1450,7 @@
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);
+ get_buf_from_mv(in_what, &mv), in_what->stride, best_sad);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit);
if (sad < best_sad) {
@@ -1475,7 +1483,7 @@
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) +
+ in_what->stride, 0x7fffffff) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, error_per_bit);
int i, j;
@@ -1516,7 +1524,7 @@
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);
+ in_what->stride, best_sad);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit);
if (sad < best_sad) {
@@ -1555,7 +1563,8 @@
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) +
+ get_buf_from_mv(in_what, ref_mv), in_what->stride,
+ second_pred, 0x7fffffff) +
mvsad_err_cost(x, ref_mv, &fcenter_mv, error_per_bit);
int i, j;
@@ -1568,7 +1577,8 @@
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);
+ get_buf_from_mv(in_what, &mv), in_what->stride,
+ second_pred, best_sad);
if (sad < best_sad) {
sad += mvsad_err_cost(x, &mv, &fcenter_mv, error_per_bit);
if (sad < best_sad) {
diff --git a/vp9/encoder/vp9_pickmode.c b/vp9/encoder/vp9_pickmode.c
index 78fba73..1e9887c 100644
--- a/vp9/encoder/vp9_pickmode.c
+++ b/vp9/encoder/vp9_pickmode.c
@@ -31,7 +31,7 @@
int_mv *tmp_mv, int *rate_mv) {
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
- struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0}};
+ struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0, 0}};
int step_param;
int sadpb = x->sadperbit16;
MV mvp_full;
@@ -110,7 +110,7 @@
MV *tmp_mv) {
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
- struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0}};
+ struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0, 0}};
int ref = mbmi->ref_frame[0];
MV ref_mv = mbmi->ref_mvs[ref][0].as_mv;
int dis;
diff --git a/vp9/encoder/vp9_quantize.c b/vp9/encoder/vp9_quantize.c
index 5206bb6..4d3086d 100644
--- a/vp9/encoder/vp9_quantize.c
+++ b/vp9/encoder/vp9_quantize.c
@@ -32,6 +32,7 @@
zbin_ptr[1] + zbin_oq_value };
const int nzbins[2] = { zbins[0] * -1,
zbins[1] * -1 };
+ (void)iscan;
vpx_memset(qcoeff_ptr, 0, count * sizeof(int16_t));
vpx_memset(dqcoeff_ptr, 0, count * sizeof(int16_t));
@@ -87,6 +88,7 @@
int idx = 0;
int idx_arr[1024];
int i, eob = -1;
+ (void)iscan;
vpx_memset(qcoeff_ptr, 0, n_coeffs * sizeof(int16_t));
vpx_memset(dqcoeff_ptr, 0, n_coeffs * sizeof(int16_t));
diff --git a/vp9/encoder/vp9_ratectrl.c b/vp9/encoder/vp9_ratectrl.c
index 124e926..a04622c 100644
--- a/vp9/encoder/vp9_ratectrl.c
+++ b/vp9/encoder/vp9_ratectrl.c
@@ -1147,10 +1147,6 @@
cpi->rc.frames_to_key--;
}
-static int test_for_kf_one_pass(VP9_COMP *cpi) {
- // Placeholder function for auto key frame
- return 0;
-}
// Use this macro to turn on/off use of alt-refs in one-pass mode.
#define USE_ALTREF_FOR_ONE_PASS 1
@@ -1182,11 +1178,12 @@
VP9_COMMON *const cm = &cpi->common;
RATE_CONTROL *const rc = &cpi->rc;
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 && test_for_kf_one_pass(cpi)))) {
+ (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;
@@ -1313,10 +1310,11 @@
VP9_COMMON *const cm = &cpi->common;
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 && test_for_kf_one_pass(cpi)))) {
+ (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;
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 410a5ff..6b7b5d5 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -297,7 +297,7 @@
fill_token_costs(x->token_costs, cm->fc.coef_probs);
for (i = 0; i < PARTITION_CONTEXTS; i++)
- vp9_cost_tokens(x->partition_cost[i], get_partition_probs(cm, i),
+ vp9_cost_tokens(cpi->partition_cost[i], get_partition_probs(cm, i),
vp9_partition_tree);
}
@@ -745,7 +745,8 @@
int use_fast_coef_casting) {
MACROBLOCKD *const xd = &x->e_mbd;
const struct macroblockd_plane *const pd = &xd->plane[plane];
- struct rdcost_block_args args = { 0 };
+ struct rdcost_block_args args;
+ vp9_zero(args);
args.x = x;
args.best_rd = ref_best_rd;
args.use_fast_coef_costing = use_fast_coef_casting;
@@ -2124,7 +2125,8 @@
// 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);
+ ref_y_ptr, ref_y_stride,
+ 0x7fffffff);
// Note if it is the best so far.
if (this_sad < best_sad) {
@@ -2311,7 +2313,7 @@
MACROBLOCKD *xd = &x->e_mbd;
const VP9_COMMON *cm = &cpi->common;
MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
- struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0}};
+ struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0, 0}};
int bestsme = INT_MAX;
int step_param;
int sadpb = x->sadperbit16;
@@ -3067,7 +3069,7 @@
int64_t best_pred_rd[REFERENCE_MODES];
int64_t best_filter_rd[SWITCHABLE_FILTER_CONTEXTS];
int64_t best_filter_diff[SWITCHABLE_FILTER_CONTEXTS];
- MB_MODE_INFO best_mbmode = { 0 };
+ MB_MODE_INFO best_mbmode;
int mode_index, best_mode_index = -1;
unsigned int ref_costs_single[MAX_REF_FRAMES], ref_costs_comp[MAX_REF_FRAMES];
vp9_prob comp_mode_p;
@@ -3093,7 +3095,7 @@
const int intra_y_mode_mask =
cpi->sf.intra_y_mode_mask[max_txsize_lookup[bsize]];
int disable_inter_mode_mask = cpi->sf.disable_inter_mode_mask[bsize];
-
+ vp9_zero(best_mbmode);
x->skip_encode = cpi->sf.skip_encode_frame && x->q_index < QIDX_SKIP_THRESH;
estimate_ref_frame_costs(cm, xd, segment_id, ref_costs_single, ref_costs_comp,
@@ -3676,7 +3678,7 @@
int64_t best_pred_rd[REFERENCE_MODES];
int64_t best_filter_rd[SWITCHABLE_FILTER_CONTEXTS];
int64_t best_filter_diff[SWITCHABLE_FILTER_CONTEXTS];
- MB_MODE_INFO best_mbmode = { 0 };
+ MB_MODE_INFO best_mbmode;
int ref_index, best_ref_index = 0;
unsigned int ref_costs_single[MAX_REF_FRAMES], ref_costs_comp[MAX_REF_FRAMES];
vp9_prob comp_mode_p;
@@ -3696,6 +3698,7 @@
x->skip_encode = cpi->sf.skip_encode_frame && x->q_index < QIDX_SKIP_THRESH;
vpx_memset(x->zcoeff_blk[TX_4X4], 0, 4);
+ vp9_zero(best_mbmode);
for (i = 0; i < 4; i++) {
int j;
diff --git a/vp9/encoder/vp9_sad.c b/vp9/encoder/vp9_sad.c
index d062636..892e905 100644
--- a/vp9/encoder/vp9_sad.c
+++ b/vp9/encoder/vp9_sad.c
@@ -35,12 +35,14 @@
#define sadMxN(m, n) \
unsigned int vp9_sad##m##x##n##_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride) { \
+ const uint8_t *ref, int ref_stride, \
+ unsigned int max_sad) { \
return sad(src, src_stride, ref, ref_stride, m, n); \
} \
unsigned int vp9_sad##m##x##n##_avg_c(const uint8_t *src, int src_stride, \
const uint8_t *ref, int ref_stride, \
- const uint8_t *second_pred) { \
+ const uint8_t *second_pred, \
+ unsigned int max_sad) { \
uint8_t comp_pred[m * n]; \
vp9_comp_avg_pred(comp_pred, second_pred, m, n, ref, ref_stride); \
return sad(src, src_stride, comp_pred, m, m, n); \
@@ -52,7 +54,8 @@
unsigned int *sads) { \
int i; \
for (i = 0; i < k; ++i) \
- sads[i] = vp9_sad##m##x##n##_c(src, src_stride, &ref[i], ref_stride); \
+ sads[i] = vp9_sad##m##x##n##_c(src, src_stride, &ref[i], ref_stride, \
+ 0x7fffffff); \
}
#define sadMxNx4D(m, n) \
@@ -61,7 +64,8 @@
unsigned int *sads) { \
int i; \
for (i = 0; i < 4; ++i) \
- sads[i] = vp9_sad##m##x##n##_c(src, src_stride, refs[i], ref_stride); \
+ sads[i] = vp9_sad##m##x##n##_c(src, src_stride, refs[i], ref_stride, \
+ 0x7fffffff); \
}
// 64x64
diff --git a/vp9/encoder/vp9_segmentation.c b/vp9/encoder/vp9_segmentation.c
index 7537d1b..574df62 100644
--- a/vp9/encoder/vp9_segmentation.c
+++ b/vp9/encoder/vp9_segmentation.c
@@ -109,7 +109,7 @@
}
static void count_segs(VP9_COMP *cpi, const TileInfo *const tile,
- MODE_INFO **mi_8x8,
+ MODE_INFO **mi,
int *no_pred_segcounts,
int (*temporal_predictor_count)[2],
int *t_unpred_seg_counts,
@@ -121,7 +121,7 @@
if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
return;
- xd->mi = mi_8x8;
+ xd->mi = mi;
segment_id = xd->mi[0]->mbmi.segment_id;
set_mi_row_col(xd, tile, mi_row, bh, mi_col, bw, cm->mi_rows, cm->mi_cols);
@@ -131,7 +131,7 @@
// Temporal prediction not allowed on key frames
if (cm->frame_type != KEY_FRAME) {
- const BLOCK_SIZE bsize = mi_8x8[0]->mbmi.sb_type;
+ 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 = vp9_get_segment_id(cm, cm->last_frame_seg_map,
bsize, mi_row, mi_col);
@@ -143,14 +143,14 @@
xd->mi[0]->mbmi.seg_id_predicted = pred_flag;
temporal_predictor_count[pred_context][pred_flag]++;
+ // Update the "unpredicted" segment count
if (!pred_flag)
- // Update the "unpredicted" segment count
t_unpred_seg_counts[segment_id]++;
}
}
static void count_segs_sb(VP9_COMP *cpi, const TileInfo *const tile,
- MODE_INFO **mi_8x8,
+ MODE_INFO **mi,
int *no_pred_segcounts,
int (*temporal_predictor_count)[2],
int *t_unpred_seg_counts,
@@ -164,22 +164,22 @@
if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
return;
- bw = num_8x8_blocks_wide_lookup[mi_8x8[0]->mbmi.sb_type];
- bh = num_8x8_blocks_high_lookup[mi_8x8[0]->mbmi.sb_type];
+ bw = num_8x8_blocks_wide_lookup[mi[0]->mbmi.sb_type];
+ bh = num_8x8_blocks_high_lookup[mi[0]->mbmi.sb_type];
if (bw == bs && bh == bs) {
- count_segs(cpi, tile, mi_8x8, no_pred_segcounts, temporal_predictor_count,
+ count_segs(cpi, tile, mi, no_pred_segcounts, temporal_predictor_count,
t_unpred_seg_counts, bs, bs, mi_row, mi_col);
} else if (bw == bs && bh < bs) {
- count_segs(cpi, tile, mi_8x8, no_pred_segcounts, temporal_predictor_count,
+ count_segs(cpi, tile, mi, no_pred_segcounts, temporal_predictor_count,
t_unpred_seg_counts, bs, hbs, mi_row, mi_col);
- count_segs(cpi, tile, mi_8x8 + hbs * mis, no_pred_segcounts,
+ count_segs(cpi, tile, mi + hbs * mis, no_pred_segcounts,
temporal_predictor_count, t_unpred_seg_counts, bs, hbs,
mi_row + hbs, mi_col);
} else if (bw < bs && bh == bs) {
- count_segs(cpi, tile, mi_8x8, no_pred_segcounts, temporal_predictor_count,
+ count_segs(cpi, tile, mi, no_pred_segcounts, temporal_predictor_count,
t_unpred_seg_counts, hbs, bs, mi_row, mi_col);
- count_segs(cpi, tile, mi_8x8 + hbs,
+ count_segs(cpi, tile, mi + hbs,
no_pred_segcounts, temporal_predictor_count, t_unpred_seg_counts,
hbs, bs, mi_row, mi_col + hbs);
} else {
@@ -192,7 +192,7 @@
const int mi_dc = hbs * (n & 1);
const int mi_dr = hbs * (n >> 1);
- count_segs_sb(cpi, tile, &mi_8x8[mi_dr * mis + mi_dc],
+ count_segs_sb(cpi, 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);
@@ -217,9 +217,6 @@
vp9_prob t_pred_tree[SEG_TREE_PROBS];
vp9_prob t_nopred_prob[PREDICTION_PROBS];
- const int mis = cm->mi_stride;
- MODE_INFO **mi_ptr, **mi;
-
// Set default state for the segment tree probabilities and the
// temporal coding probabilities
vpx_memset(seg->tree_probs, 255, sizeof(seg->tree_probs));
@@ -229,12 +226,13 @@
// predicts this one
for (tile_col = 0; tile_col < 1 << cm->log2_tile_cols; tile_col++) {
TileInfo tile;
-
+ MODE_INFO **mi_ptr;
vp9_tile_init(&tile, cm, 0, tile_col);
+
mi_ptr = cm->mi_grid_visible + tile.mi_col_start;
for (mi_row = 0; mi_row < cm->mi_rows;
- mi_row += 8, mi_ptr += 8 * mis) {
- mi = mi_ptr;
+ mi_row += 8, mi_ptr += 8 * cm->mi_stride) {
+ MODE_INFO **mi = mi_ptr;
for (mi_col = tile.mi_col_start; mi_col < tile.mi_col_end;
mi_col += 8, mi += 8)
count_segs_sb(cpi, &tile, mi, no_pred_segcounts,
diff --git a/vp9/encoder/vp9_tokenize.c b/vp9/encoder/vp9_tokenize.c
index 8ce98d9..dcca92d 100644
--- a/vp9/encoder/vp9_tokenize.c
+++ b/vp9/encoder/vp9_tokenize.c
@@ -232,7 +232,6 @@
cpi->common.fc.coef_probs[tx_size][type][ref];
unsigned int (*const eob_branch)[COEFF_CONTEXTS] =
cpi->common.counts.eob_branch[tx_size][type][ref];
-
const uint8_t *const band = get_band_translate(tx_size);
const int seg_eob = get_tx_eob(&cpi->common.seg, segment_id, tx_size);
@@ -289,14 +288,17 @@
MACROBLOCK *x;
int *skippable;
};
-
static void is_skippable(int plane, int block,
BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
void *argv) {
struct is_skippable_args *args = argv;
+ (void)plane_bsize;
+ (void)tx_size;
args->skippable[0] &= (!args->x->plane[plane].eobs[block]);
}
+// TODO(yaowu): rewrite and optimize this function to remove the usage of
+// vp9_foreach_transform_block() and simplify is_skippable().
int vp9_is_skippable_in_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
int result = 1;
struct is_skippable_args args = {x, &result};
diff --git a/vp9/encoder/vp9_variance.h b/vp9/encoder/vp9_variance.h
index 4a194b7..c47fe13 100644
--- a/vp9/encoder/vp9_variance.h
+++ b/vp9/encoder/vp9_variance.h
@@ -25,13 +25,15 @@
typedef unsigned int(*vp9_sad_fn_t)(const uint8_t *src_ptr,
int source_stride,
const uint8_t *ref_ptr,
- int ref_stride);
+ int ref_stride,
+ unsigned int max_sad);
typedef unsigned int(*vp9_sad_avg_fn_t)(const uint8_t *src_ptr,
int source_stride,
const uint8_t *ref_ptr,
int ref_stride,
- const uint8_t *second_pred);
+ const uint8_t *second_pred,
+ unsigned int max_sad);
typedef void (*vp9_sad_multi_fn_t)(const uint8_t *src_ptr,
int source_stride,
diff --git a/vp9/encoder/x86/vp9_error_intrin_avx2.c b/vp9/encoder/x86/vp9_error_intrin_avx2.c
new file mode 100644
index 0000000..c67490f
--- /dev/null
+++ b/vp9/encoder/x86/vp9_error_intrin_avx2.c
@@ -0,0 +1,72 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Usee of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include <immintrin.h> // AVX2
+#include "vpx/vpx_integer.h"
+
+
+int64_t vp9_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;
+ __m128i sse_reg128, ssz_reg128;
+ int64_t sse;
+ int i;
+ const __m256i zero_reg = _mm256_set1_epi16(0);
+
+ // init sse and ssz registerd to zero
+ sse_reg = _mm256_set1_epi16(0);
+ ssz_reg = _mm256_set1_epi16(0);
+
+ 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));
+ // dqcoeff - coeff
+ dqcoeff_reg = _mm256_sub_epi16(dqcoeff_reg, coeff_reg);
+ // madd (dqcoeff - coeff)
+ dqcoeff_reg = _mm256_madd_epi16(dqcoeff_reg, dqcoeff_reg);
+ // madd coeff
+ coeff_reg = _mm256_madd_epi16(coeff_reg, coeff_reg);
+ // expand each double word of madd (dqcoeff - coeff) to quad word
+ exp_dqcoeff_lo = _mm256_unpacklo_epi32(dqcoeff_reg, zero_reg);
+ exp_dqcoeff_hi = _mm256_unpackhi_epi32(dqcoeff_reg, zero_reg);
+ // expand each double word of madd (coeff) to quad word
+ exp_coeff_lo = _mm256_unpacklo_epi32(coeff_reg, zero_reg);
+ exp_coeff_hi = _mm256_unpackhi_epi32(coeff_reg, zero_reg);
+ // add each quad word of madd (dqcoeff - coeff) and madd (coeff)
+ sse_reg = _mm256_add_epi64(sse_reg, exp_dqcoeff_lo);
+ ssz_reg = _mm256_add_epi64(ssz_reg, exp_coeff_lo);
+ sse_reg = _mm256_add_epi64(sse_reg, exp_dqcoeff_hi);
+ ssz_reg = _mm256_add_epi64(ssz_reg, exp_coeff_hi);
+ }
+ // save the higher 64 bit of each 128 bit lane
+ sse_reg_64hi = _mm256_srli_si256(sse_reg, 8);
+ ssz_reg_64hi = _mm256_srli_si256(ssz_reg, 8);
+ // add the higher 64 bit to the low 64 bit
+ sse_reg = _mm256_add_epi64(sse_reg, sse_reg_64hi);
+ ssz_reg = _mm256_add_epi64(ssz_reg, ssz_reg_64hi);
+
+ // add each 64 bit from each of the 128 bit lane of the 256 bit
+ sse_reg128 = _mm_add_epi64(_mm256_castsi256_si128(sse_reg),
+ _mm256_extractf128_si256(sse_reg, 1));
+
+ ssz_reg128 = _mm_add_epi64(_mm256_castsi256_si128(ssz_reg),
+ _mm256_extractf128_si256(ssz_reg, 1));
+
+ // store the results
+ _mm_storel_epi64((__m128i*)(&sse), sse_reg128);
+
+ _mm_storel_epi64((__m128i*)(ssz), ssz_reg128);
+ return sse;
+}
diff --git a/vp9/vp9_dx_iface.c b/vp9/vp9_dx_iface.c
index 5f7485c..2802fba 100644
--- a/vp9/vp9_dx_iface.c
+++ b/vp9/vp9_dx_iface.c
@@ -32,21 +32,12 @@
vpx_codec_priv_t base;
vpx_codec_dec_cfg_t cfg;
vp9_stream_info_t si;
- int decoder_init;
struct VP9Decoder *pbi;
int postproc_cfg_set;
vp8_postproc_cfg_t postproc_cfg;
-#if CONFIG_POSTPROC_VISUALIZER
- unsigned int dbg_postproc_flag;
- int dbg_color_ref_frame_flag;
- int dbg_color_mb_modes_flag;
- int dbg_color_b_modes_flag;
- int dbg_display_mv_flag;
-#endif
vpx_decrypt_cb decrypt_cb;
void *decrypt_state;
vpx_image_t img;
- int img_setup;
int img_avail;
int invert_tile_order;
@@ -226,22 +217,10 @@
static void set_ppflags(const vpx_codec_alg_priv_t *ctx,
vp9_ppflags_t *flags) {
flags->post_proc_flag =
-#if CONFIG_POSTPROC_VISUALIZER
- (ctx->dbg_color_ref_frame_flag ? VP9D_DEBUG_CLR_FRM_REF_BLKS : 0) |
- (ctx->dbg_color_mb_modes_flag ? VP9D_DEBUG_CLR_BLK_MODES : 0) |
- (ctx->dbg_color_b_modes_flag ? VP9D_DEBUG_CLR_BLK_MODES : 0) |
- (ctx->dbg_display_mv_flag ? VP9D_DEBUG_DRAW_MV : 0) |
-#endif
ctx->postproc_cfg.post_proc_flag;
flags->deblocking_level = ctx->postproc_cfg.deblocking_level;
flags->noise_level = ctx->postproc_cfg.noise_level;
-#if CONFIG_POSTPROC_VISUALIZER
- flags->display_ref_frame_flag = ctx->dbg_color_ref_frame_flag;
- flags->display_mb_modes_flag = ctx->dbg_color_mb_modes_flag;
- flags->display_b_modes_flag = ctx->dbg_color_b_modes_flag;
- flags->display_mv_flag = ctx->dbg_display_mv_flag;
-#endif
}
static void init_decoder(vpx_codec_alg_priv_t *ctx) {
@@ -283,12 +262,10 @@
}
// Initialize the decoder instance on the first frame
- if (!ctx->decoder_init) {
+ if (ctx->pbi == NULL) {
init_decoder(ctx);
if (ctx->pbi == NULL)
return VPX_CODEC_ERROR;
-
- ctx->decoder_init = 1;
}
// Set these even if already initialized. The caller may have changed the
@@ -537,22 +514,7 @@
static vpx_codec_err_t ctrl_set_dbg_options(vpx_codec_alg_priv_t *ctx,
int ctrl_id, va_list args) {
-#if CONFIG_POSTPROC_VISUALIZER && CONFIG_POSTPROC
- int data = va_arg(args, int);
-
-#define MAP(id, var) case id: var = data; break;
-
- switch (ctrl_id) {
- MAP(VP8_SET_DBG_COLOR_REF_FRAME, ctx->dbg_color_ref_frame_flag);
- MAP(VP8_SET_DBG_COLOR_MB_MODES, ctx->dbg_color_mb_modes_flag);
- MAP(VP8_SET_DBG_COLOR_B_MODES, ctx->dbg_color_b_modes_flag);
- MAP(VP8_SET_DBG_DISPLAY_MV, ctx->dbg_display_mv_flag);
- }
-
- return VPX_CODEC_OK;
-#else
return VPX_CODEC_INCAPABLE;
-#endif
}
static vpx_codec_err_t ctrl_get_last_ref_updates(vpx_codec_alg_priv_t *ctx,
diff --git a/vp9/vp9cx.mk b/vp9/vp9cx.mk
index bc9a478..6e5c521 100644
--- a/vp9/vp9cx.mk
+++ b/vp9/vp9cx.mk
@@ -103,6 +103,7 @@
ifeq ($(CONFIG_USE_X86INC),yes)
VP9_CX_SRCS-$(HAVE_MMX) += encoder/x86/vp9_dct_mmx.asm
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_error_sse2.asm
+VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_error_intrin_avx2.c
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_sad_sse2.asm
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_subtract_sse2.asm
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_variance_sse2.c
diff --git a/vpxenc.c b/vpxenc.c
index 96a7ab6..28d43f2 100644
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -755,7 +755,7 @@
input->height = input->y4m.pic_h;
input->framerate.numerator = input->y4m.fps_n;
input->framerate.denominator = input->y4m.fps_d;
- input->use_i420 = 0;
+ input->fmt = input->y4m.vpx_fmt;
} else
fatal("Unsupported Y4M stream.");
} else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
@@ -1059,6 +1059,23 @@
}
}
+static const char* file_type_to_string(enum VideoFileType t) {
+ switch (t) {
+ case FILE_TYPE_RAW: return "RAW";
+ case FILE_TYPE_Y4M: return "Y4M";
+ default: return "Other";
+ }
+}
+
+static const char* image_format_to_string(vpx_img_fmt_t f) {
+ switch (f) {
+ case VPX_IMG_FMT_I420: return "I420";
+ case VPX_IMG_FMT_I422: return "I422";
+ case VPX_IMG_FMT_I444: return "I444";
+ case VPX_IMG_FMT_YV12: return "YV12";
+ default: return "Other";
+ }
+}
static void show_stream_config(struct stream_state *stream,
struct VpxEncoderConfig *global,
@@ -1070,8 +1087,10 @@
if (stream->index == 0) {
fprintf(stderr, "Codec: %s\n",
vpx_codec_iface_name(global->codec->interface()));
- fprintf(stderr, "Source file: %s Format: %s\n", input->filename,
- input->use_i420 ? "I420" : "YV12");
+ fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
+ input->filename,
+ file_type_to_string(input->file_type),
+ image_format_to_string(input->fmt));
}
if (stream->next || stream->index)
fprintf(stderr, "\nStream Index: %d\n", stream->index);
@@ -1501,7 +1520,6 @@
/* Setup default input stream settings */
input.framerate.numerator = 30;
input.framerate.denominator = 1;
- input.use_i420 = 1;
input.only_i420 = 1;
/* First parse the global configuration values, because we want to apply
@@ -1511,6 +1529,7 @@
argv = argv_dup(argc - 1, argv_ + 1);
parse_global_config(&global, argv);
+ input.fmt = global.use_i420 ? VPX_IMG_FMT_I420 : VPX_IMG_FMT_YV12;
{
/* Now parse each stream's parameters. Using a local scope here
@@ -1611,10 +1630,7 @@
frames.*/
memset(&raw, 0, sizeof(raw));
else
- vpx_img_alloc(&raw,
- input.use_i420 ? VPX_IMG_FMT_I420
- : VPX_IMG_FMT_YV12,
- input.width, input.height, 32);
+ vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
FOREACH_STREAM(stream->rate_hist =
init_rate_histogram(&stream->config.cfg,
