libvpx: Pull from upstream
Current HEAD: 687c56e8026978f8ccfd5b3fbfee4dbd1de466e1
git log from upstream:
7045aec SAD32xh and SAD64xh for AVX2
7c4992c Remove the dependency in token storing locations
f0c3da9 Alter adjustment of two pass GF/ARF boost with Q.
73ae6e4 Add highbitdepth function for vp9_avg_8x8
e1111fb Remove unused VAR_BASED_FIXED_PARTITION flag
5e766cc Use rate/distortion thresholds to control non-RD partition search
6f77bff Updates to aggressive denoising mode.
e3bf55d Correct the logic of ready_for_new_data.
6356d21 vp9_denoiser_sse2.c: solve windows build error.
e2612fb Add init and reset functions for RD_COST struct
94ecfa3 Reset rate cost value in rd mode search
8514d03 vp9_denoiser_sse2.c: eliminate gcc warnings
50c59cd Adds a set of end-to-end encode tests
ed100c0 Fix an ioc issue in super_block_uvrd
716ae78 Change initialization of static_scene_max_gf_interval.
68b550f [spatial svc]Another workaround to avoid using prev_mi
d5130af Revert "Move input frame scaling into the recode loop"
4680329 Revert "[spatial svc]Another workaround to avoid using prev_mi"
23fc1f7 Fix in bit-shift operation for highbitdepth decode
91657ab fix CONFIG_SPATIAL_SVC warning
09ea74f Some updates for Speed 6/VAR_BASED_PARTITION.
<...>
TBR=tomfinegan@chromium.org
Review URL: https://codereview.chromium.org/668403002
git-svn-id: http://src.chromium.org/svn/trunk/deps/third_party/libvpx@292609 4ff67af0-8c30-449e-8e8b-ad334ec8d88c
diff --git a/README.chromium b/README.chromium
index f4874b8..ef84c90 100644
--- a/README.chromium
+++ b/README.chromium
@@ -5,9 +5,9 @@
License File: source/libvpx/LICENSE
Security Critical: yes
-Date: Monday September 22 2014
+Date: Monday October 21 2014
Branch: master
-Commit: 38b6aed8fdf50e701c070e708ff7933cc3a61117
+Commit: 687c56e8026978f8ccfd5b3fbfee4dbd1de466e1
Description:
Contains the sources used to compile libvpx binaries used by Google Chrome and
diff --git a/libvpx_srcs.gni b/libvpx_srcs.gni
index 6c1a1b0..eff2cfd 100644
--- a/libvpx_srcs.gni
+++ b/libvpx_srcs.gni
@@ -206,6 +206,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
@@ -371,8 +372,10 @@
"//third_party/libvpx/source/libvpx/vp8/encoder/x86/vp8_enc_stubs_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/common/x86/vp9_idct_intrin_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/common/x86/vp9_loopfilter_intrin_sse2.c",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_dct32x32_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_dct_sse2.c",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_quantize_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_variance_sse2.c",
]
libvpx_srcs_x86_sse3 = [
@@ -593,6 +596,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
@@ -764,8 +768,10 @@
"//third_party/libvpx/source/libvpx/vp8/encoder/x86/vp8_enc_stubs_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/common/x86/vp9_idct_intrin_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/common/x86/vp9_loopfilter_intrin_sse2.c",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_dct32x32_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_dct_sse2.c",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_quantize_sse2.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/x86/vp9_variance_sse2.c",
]
libvpx_srcs_x86_64_sse3 = [
@@ -989,6 +995,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
@@ -1354,6 +1361,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
@@ -1708,6 +1716,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
@@ -2122,6 +2131,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
@@ -2423,6 +2433,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
@@ -2720,6 +2731,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
@@ -3017,6 +3029,7 @@
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_cyclicrefresh.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_aq_variance.h",
+ "//third_party/libvpx/source/libvpx/vp9/encoder/vp9_avg.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.c",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_bitstream.h",
"//third_party/libvpx/source/libvpx/vp9/encoder/vp9_block.h",
diff --git a/libvpx_srcs_arm.gypi b/libvpx_srcs_arm.gypi
index 56b0428..3292fb0 100644
--- a/libvpx_srcs_arm.gypi
+++ b/libvpx_srcs_arm.gypi
@@ -235,6 +235,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_arm64.gypi b/libvpx_srcs_arm64.gypi
index c72f537..cb4c8ee 100644
--- a/libvpx_srcs_arm64.gypi
+++ b/libvpx_srcs_arm64.gypi
@@ -236,6 +236,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_arm_neon.gypi b/libvpx_srcs_arm_neon.gypi
index 1f8c17a..a14b108 100644
--- a/libvpx_srcs_arm_neon.gypi
+++ b/libvpx_srcs_arm_neon.gypi
@@ -288,6 +288,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_arm_neon_cpu_detect.gypi b/libvpx_srcs_arm_neon_cpu_detect.gypi
index 9bd4eda..8f1d3de 100644
--- a/libvpx_srcs_arm_neon_cpu_detect.gypi
+++ b/libvpx_srcs_arm_neon_cpu_detect.gypi
@@ -256,6 +256,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_generic.gypi b/libvpx_srcs_generic.gypi
index 4f940cb..078e369 100644
--- a/libvpx_srcs_generic.gypi
+++ b/libvpx_srcs_generic.gypi
@@ -201,6 +201,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_mips.gypi b/libvpx_srcs_mips.gypi
index 228baaa..4323ed8 100644
--- a/libvpx_srcs_mips.gypi
+++ b/libvpx_srcs_mips.gypi
@@ -203,6 +203,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_nacl.gypi b/libvpx_srcs_nacl.gypi
index 4f940cb..078e369 100644
--- a/libvpx_srcs_nacl.gypi
+++ b/libvpx_srcs_nacl.gypi
@@ -201,6 +201,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_x86.gypi b/libvpx_srcs_x86.gypi
index f5b3995..f27b786 100644
--- a/libvpx_srcs_x86.gypi
+++ b/libvpx_srcs_x86.gypi
@@ -245,6 +245,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_x86_64.gypi b/libvpx_srcs_x86_64.gypi
index f29c6c8..3596ec2 100644
--- a/libvpx_srcs_x86_64.gypi
+++ b/libvpx_srcs_x86_64.gypi
@@ -248,6 +248,7 @@
'<(libvpx_source)/vp9/encoder/vp9_aq_cyclicrefresh.h',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.c',
'<(libvpx_source)/vp9/encoder/vp9_aq_variance.h',
+ '<(libvpx_source)/vp9/encoder/vp9_avg.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.c',
'<(libvpx_source)/vp9/encoder/vp9_bitstream.h',
'<(libvpx_source)/vp9/encoder/vp9_block.h',
diff --git a/libvpx_srcs_x86_64_intrinsics.gypi b/libvpx_srcs_x86_64_intrinsics.gypi
index bb1b203..28d4bc4 100644
--- a/libvpx_srcs_x86_64_intrinsics.gypi
+++ b/libvpx_srcs_x86_64_intrinsics.gypi
@@ -36,8 +36,10 @@
'<(libvpx_source)/vp8/encoder/x86/vp8_enc_stubs_sse2.c',
'<(libvpx_source)/vp9/common/x86/vp9_idct_intrin_sse2.c',
'<(libvpx_source)/vp9/common/x86/vp9_loopfilter_intrin_sse2.c',
+ '<(libvpx_source)/vp9/encoder/x86/vp9_avg_intrin_sse2.c',
'<(libvpx_source)/vp9/encoder/x86/vp9_dct32x32_sse2.c',
'<(libvpx_source)/vp9/encoder/x86/vp9_dct_sse2.c',
+ '<(libvpx_source)/vp9/encoder/x86/vp9_quantize_sse2.c',
'<(libvpx_source)/vp9/encoder/x86/vp9_variance_sse2.c',
],
'cflags': [ '-msse2', ],
diff --git a/libvpx_srcs_x86_intrinsics.gypi b/libvpx_srcs_x86_intrinsics.gypi
index bb1b203..28d4bc4 100644
--- a/libvpx_srcs_x86_intrinsics.gypi
+++ b/libvpx_srcs_x86_intrinsics.gypi
@@ -36,8 +36,10 @@
'<(libvpx_source)/vp8/encoder/x86/vp8_enc_stubs_sse2.c',
'<(libvpx_source)/vp9/common/x86/vp9_idct_intrin_sse2.c',
'<(libvpx_source)/vp9/common/x86/vp9_loopfilter_intrin_sse2.c',
+ '<(libvpx_source)/vp9/encoder/x86/vp9_avg_intrin_sse2.c',
'<(libvpx_source)/vp9/encoder/x86/vp9_dct32x32_sse2.c',
'<(libvpx_source)/vp9/encoder/x86/vp9_dct_sse2.c',
+ '<(libvpx_source)/vp9/encoder/x86/vp9_quantize_sse2.c',
'<(libvpx_source)/vp9/encoder/x86/vp9_variance_sse2.c',
],
'cflags': [ '-msse2', ],
diff --git a/source/config/linux/arm-neon-cpu-detect/vp9_rtcd.h b/source/config/linux/arm-neon-cpu-detect/vp9_rtcd.h
index 26cf5e2..13f5411 100644
--- a/source/config/linux/arm-neon-cpu-detect/vp9_rtcd.h
+++ b/source/config/linux/arm-neon-cpu-detect/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,9 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_c
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
diff --git a/source/config/linux/arm-neon-cpu-detect/vpx_config.asm b/source/config/linux/arm-neon-cpu-detect/vpx_config.asm
index 7a9b810..c3e7b00 100644
--- a/source/config/linux/arm-neon-cpu-detect/vpx_config.asm
+++ b/source/config/linux/arm-neon-cpu-detect/vpx_config.asm
@@ -88,5 +88,5 @@
.equ CONFIG_SPATIAL_SVC , 0
.equ CONFIG_VP9_TEMPORAL_DENOISING , 0
.equ CONFIG_FP_MB_STATS , 0
-.equ CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH , 0
+.equ CONFIG_EMULATE_HARDWARE , 0
.section .note.GNU-stack,"",%progbits
diff --git a/source/config/linux/arm-neon-cpu-detect/vpx_config.h b/source/config/linux/arm-neon-cpu-detect/vpx_config.h
index 04fa0d8..78ffd73 100644
--- a/source/config/linux/arm-neon-cpu-detect/vpx_config.h
+++ b/source/config/linux/arm-neon-cpu-detect/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/linux/arm-neon/vp9_rtcd.h b/source/config/linux/arm-neon/vp9_rtcd.h
index 8a31aed..376e295 100644
--- a/source/config/linux/arm-neon/vp9_rtcd.h
+++ b/source/config/linux/arm-neon/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,9 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_c
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
diff --git a/source/config/linux/arm-neon/vpx_config.asm b/source/config/linux/arm-neon/vpx_config.asm
index e47e4a7..f4abc3b 100644
--- a/source/config/linux/arm-neon/vpx_config.asm
+++ b/source/config/linux/arm-neon/vpx_config.asm
@@ -88,5 +88,5 @@
.equ CONFIG_SPATIAL_SVC , 0
.equ CONFIG_VP9_TEMPORAL_DENOISING , 0
.equ CONFIG_FP_MB_STATS , 0
-.equ CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH , 0
+.equ CONFIG_EMULATE_HARDWARE , 0
.section .note.GNU-stack,"",%progbits
diff --git a/source/config/linux/arm-neon/vpx_config.h b/source/config/linux/arm-neon/vpx_config.h
index 6c0ea64..32ffc77 100644
--- a/source/config/linux/arm-neon/vpx_config.h
+++ b/source/config/linux/arm-neon/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/linux/arm/vp9_rtcd.h b/source/config/linux/arm/vp9_rtcd.h
index ad509a7..ec0624b 100644
--- a/source/config/linux/arm/vp9_rtcd.h
+++ b/source/config/linux/arm/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,9 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_c
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
diff --git a/source/config/linux/arm/vpx_config.asm b/source/config/linux/arm/vpx_config.asm
index d6ea277..8f7487f 100644
--- a/source/config/linux/arm/vpx_config.asm
+++ b/source/config/linux/arm/vpx_config.asm
@@ -88,5 +88,5 @@
.equ CONFIG_SPATIAL_SVC , 0
.equ CONFIG_VP9_TEMPORAL_DENOISING , 0
.equ CONFIG_FP_MB_STATS , 0
-.equ CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH , 0
+.equ CONFIG_EMULATE_HARDWARE , 0
.section .note.GNU-stack,"",%progbits
diff --git a/source/config/linux/arm/vpx_config.h b/source/config/linux/arm/vpx_config.h
index e2dff1d..dbce81d 100644
--- a/source/config/linux/arm/vpx_config.h
+++ b/source/config/linux/arm/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/linux/arm64/vp9_rtcd.h b/source/config/linux/arm64/vp9_rtcd.h
index d1034cd..57cbf96 100644
--- a/source/config/linux/arm64/vp9_rtcd.h
+++ b/source/config/linux/arm64/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,9 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_c
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
diff --git a/source/config/linux/arm64/vpx_config.asm b/source/config/linux/arm64/vpx_config.asm
index 7240985..9423d65 100644
--- a/source/config/linux/arm64/vpx_config.asm
+++ b/source/config/linux/arm64/vpx_config.asm
@@ -88,5 +88,5 @@
.equ CONFIG_SPATIAL_SVC , 0
.equ CONFIG_VP9_TEMPORAL_DENOISING , 0
.equ CONFIG_FP_MB_STATS , 0
-.equ CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH , 0
+.equ CONFIG_EMULATE_HARDWARE , 0
.section .note.GNU-stack,"",%progbits
diff --git a/source/config/linux/arm64/vpx_config.h b/source/config/linux/arm64/vpx_config.h
index 4f0107f..fc49515 100644
--- a/source/config/linux/arm64/vpx_config.h
+++ b/source/config/linux/arm64/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/linux/generic/vp9_rtcd.h b/source/config/linux/generic/vp9_rtcd.h
index b60c290..0483682 100644
--- a/source/config/linux/generic/vp9_rtcd.h
+++ b/source/config/linux/generic/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,9 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_c
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
diff --git a/source/config/linux/generic/vpx_config.asm b/source/config/linux/generic/vpx_config.asm
index 776ff66..71aee39 100644
--- a/source/config/linux/generic/vpx_config.asm
+++ b/source/config/linux/generic/vpx_config.asm
@@ -88,5 +88,5 @@
.equ CONFIG_SPATIAL_SVC , 0
.equ CONFIG_VP9_TEMPORAL_DENOISING , 0
.equ CONFIG_FP_MB_STATS , 0
-.equ CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH , 0
+.equ CONFIG_EMULATE_HARDWARE , 0
.section .note.GNU-stack,"",%progbits
diff --git a/source/config/linux/generic/vpx_config.h b/source/config/linux/generic/vpx_config.h
index d02c859..a24d041 100644
--- a/source/config/linux/generic/vpx_config.h
+++ b/source/config/linux/generic/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/linux/ia32/vp9_rtcd.h b/source/config/linux/ia32/vp9_rtcd.h
index 660b652..2d6c530 100644
--- a/source/config/linux/ia32/vp9_rtcd.h
+++ b/source/config/linux/ia32/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,10 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+unsigned int vp9_avg_8x8_sse2(const uint8_t *, int p);
+RTCD_EXTERN unsigned int (*vp9_avg_8x8)(const uint8_t *, int p);
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
int64_t vp9_block_error_sse2(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
RTCD_EXTERN int64_t (*vp9_block_error)(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
@@ -426,7 +430,8 @@
RTCD_EXTERN unsigned int (*vp9_mse8x8)(const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse);
void vp9_quantize_b_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
-#define vp9_quantize_b vp9_quantize_b_c
+void vp9_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
+RTCD_EXTERN void (*vp9_quantize_b)(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
void vp9_quantize_b_32x32_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
#define vp9_quantize_b_32x32 vp9_quantize_b_32x32_c
@@ -882,6 +887,8 @@
(void)flags;
+ vp9_avg_8x8 = vp9_avg_8x8_c;
+ if (flags & HAS_SSE2) vp9_avg_8x8 = vp9_avg_8x8_sse2;
vp9_block_error = vp9_block_error_c;
if (flags & HAS_SSE2) vp9_block_error = vp9_block_error_sse2;
vp9_convolve8 = vp9_convolve8_c;
@@ -1047,6 +1054,8 @@
if (flags & HAS_SSE2) vp9_mse8x16 = vp9_mse8x16_sse2;
vp9_mse8x8 = vp9_mse8x8_c;
if (flags & HAS_SSE2) vp9_mse8x8 = vp9_mse8x8_sse2;
+ vp9_quantize_b = vp9_quantize_b_c;
+ if (flags & HAS_SSE2) vp9_quantize_b = vp9_quantize_b_sse2;
vp9_sad16x16 = vp9_sad16x16_c;
if (flags & HAS_SSE2) vp9_sad16x16 = vp9_sad16x16_sse2;
vp9_sad16x16_avg = vp9_sad16x16_avg_c;
diff --git a/source/config/linux/ia32/vpx_config.asm b/source/config/linux/ia32/vpx_config.asm
index b47dd1f..4923184 100644
--- a/source/config/linux/ia32/vpx_config.asm
+++ b/source/config/linux/ia32/vpx_config.asm
@@ -85,4 +85,4 @@
%define CONFIG_SPATIAL_SVC 0
%define CONFIG_VP9_TEMPORAL_DENOISING 0
%define CONFIG_FP_MB_STATS 0
-%define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+%define CONFIG_EMULATE_HARDWARE 0
diff --git a/source/config/linux/ia32/vpx_config.h b/source/config/linux/ia32/vpx_config.h
index 58bda73..093cfa0 100644
--- a/source/config/linux/ia32/vpx_config.h
+++ b/source/config/linux/ia32/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/linux/mips64el/vp9_rtcd.h b/source/config/linux/mips64el/vp9_rtcd.h
index b60c290..0483682 100644
--- a/source/config/linux/mips64el/vp9_rtcd.h
+++ b/source/config/linux/mips64el/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,9 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_c
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
diff --git a/source/config/linux/mips64el/vpx_config.h b/source/config/linux/mips64el/vpx_config.h
index e3a5448..c49041e 100644
--- a/source/config/linux/mips64el/vpx_config.h
+++ b/source/config/linux/mips64el/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/linux/mipsel/vp9_rtcd.h b/source/config/linux/mipsel/vp9_rtcd.h
index b60c290..0483682 100644
--- a/source/config/linux/mipsel/vp9_rtcd.h
+++ b/source/config/linux/mipsel/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,9 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_c
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
diff --git a/source/config/linux/mipsel/vpx_config.h b/source/config/linux/mipsel/vpx_config.h
index 8c0ba89..d74f9b2 100644
--- a/source/config/linux/mipsel/vpx_config.h
+++ b/source/config/linux/mipsel/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/linux/x64/vp9_rtcd.h b/source/config/linux/x64/vp9_rtcd.h
index 253f565..f1417de 100644
--- a/source/config/linux/x64/vp9_rtcd.h
+++ b/source/config/linux/x64/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,10 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+unsigned int vp9_avg_8x8_sse2(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_sse2
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
int64_t vp9_block_error_sse2(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_sse2
@@ -429,6 +433,7 @@
#define vp9_mse8x8 vp9_mse8x8_sse2
void vp9_quantize_b_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
+void vp9_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
void vp9_quantize_b_ssse3(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
RTCD_EXTERN void (*vp9_quantize_b)(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
@@ -953,7 +958,7 @@
if (flags & HAS_SSSE3) vp9_idct8x8_12_add = vp9_idct8x8_12_add_ssse3;
vp9_idct8x8_64_add = vp9_idct8x8_64_add_sse2;
if (flags & HAS_SSSE3) vp9_idct8x8_64_add = vp9_idct8x8_64_add_ssse3;
- vp9_quantize_b = vp9_quantize_b_c;
+ vp9_quantize_b = vp9_quantize_b_sse2;
if (flags & HAS_SSSE3) vp9_quantize_b = vp9_quantize_b_ssse3;
vp9_quantize_b_32x32 = vp9_quantize_b_32x32_c;
if (flags & HAS_SSSE3) vp9_quantize_b_32x32 = vp9_quantize_b_32x32_ssse3;
diff --git a/source/config/linux/x64/vpx_config.asm b/source/config/linux/x64/vpx_config.asm
index e66e8b7..159294e 100644
--- a/source/config/linux/x64/vpx_config.asm
+++ b/source/config/linux/x64/vpx_config.asm
@@ -85,4 +85,4 @@
%define CONFIG_SPATIAL_SVC 0
%define CONFIG_VP9_TEMPORAL_DENOISING 0
%define CONFIG_FP_MB_STATS 0
-%define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+%define CONFIG_EMULATE_HARDWARE 0
diff --git a/source/config/linux/x64/vpx_config.h b/source/config/linux/x64/vpx_config.h
index 61dd009..220ad67 100644
--- a/source/config/linux/x64/vpx_config.h
+++ b/source/config/linux/x64/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/mac/ia32/vp9_rtcd.h b/source/config/mac/ia32/vp9_rtcd.h
index fa60726..35449ac 100644
--- a/source/config/mac/ia32/vp9_rtcd.h
+++ b/source/config/mac/ia32/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,10 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+unsigned int vp9_avg_8x8_sse2(const uint8_t *, int p);
+RTCD_EXTERN unsigned int (*vp9_avg_8x8)(const uint8_t *, int p);
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
@@ -392,7 +396,8 @@
#define vp9_mse8x8 vp9_mse8x8_c
void vp9_quantize_b_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
-#define vp9_quantize_b vp9_quantize_b_c
+void vp9_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
+RTCD_EXTERN void (*vp9_quantize_b)(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
void vp9_quantize_b_32x32_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
#define vp9_quantize_b_32x32 vp9_quantize_b_32x32_c
@@ -749,6 +754,8 @@
(void)flags;
+ vp9_avg_8x8 = vp9_avg_8x8_c;
+ if (flags & HAS_SSE2) vp9_avg_8x8 = vp9_avg_8x8_sse2;
vp9_convolve8 = vp9_convolve8_c;
if (flags & HAS_SSE2) vp9_convolve8 = vp9_convolve8_sse2;
if (flags & HAS_SSSE3) vp9_convolve8 = vp9_convolve8_ssse3;
@@ -846,6 +853,8 @@
if (flags & HAS_SSE2) vp9_lpf_vertical_8 = vp9_lpf_vertical_8_sse2;
vp9_lpf_vertical_8_dual = vp9_lpf_vertical_8_dual_c;
if (flags & HAS_SSE2) vp9_lpf_vertical_8_dual = vp9_lpf_vertical_8_dual_sse2;
+ vp9_quantize_b = vp9_quantize_b_c;
+ if (flags & HAS_SSE2) vp9_quantize_b = vp9_quantize_b_sse2;
vp9_sad16x16x3 = vp9_sad16x16x3_c;
if (flags & HAS_SSE3) vp9_sad16x16x3 = vp9_sad16x16x3_sse3;
if (flags & HAS_SSSE3) vp9_sad16x16x3 = vp9_sad16x16x3_ssse3;
diff --git a/source/config/mac/ia32/vpx_config.asm b/source/config/mac/ia32/vpx_config.asm
index 6caaebf..e65c0d9 100644
--- a/source/config/mac/ia32/vpx_config.asm
+++ b/source/config/mac/ia32/vpx_config.asm
@@ -85,4 +85,4 @@
%define CONFIG_SPATIAL_SVC 0
%define CONFIG_VP9_TEMPORAL_DENOISING 0
%define CONFIG_FP_MB_STATS 0
-%define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+%define CONFIG_EMULATE_HARDWARE 0
diff --git a/source/config/mac/ia32/vpx_config.h b/source/config/mac/ia32/vpx_config.h
index 3e5d038..c7b5890 100644
--- a/source/config/mac/ia32/vpx_config.h
+++ b/source/config/mac/ia32/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/mac/x64/vp9_rtcd.h b/source/config/mac/x64/vp9_rtcd.h
index 253f565..f1417de 100644
--- a/source/config/mac/x64/vp9_rtcd.h
+++ b/source/config/mac/x64/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,10 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+unsigned int vp9_avg_8x8_sse2(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_sse2
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
int64_t vp9_block_error_sse2(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_sse2
@@ -429,6 +433,7 @@
#define vp9_mse8x8 vp9_mse8x8_sse2
void vp9_quantize_b_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
+void vp9_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
void vp9_quantize_b_ssse3(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
RTCD_EXTERN void (*vp9_quantize_b)(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
@@ -953,7 +958,7 @@
if (flags & HAS_SSSE3) vp9_idct8x8_12_add = vp9_idct8x8_12_add_ssse3;
vp9_idct8x8_64_add = vp9_idct8x8_64_add_sse2;
if (flags & HAS_SSSE3) vp9_idct8x8_64_add = vp9_idct8x8_64_add_ssse3;
- vp9_quantize_b = vp9_quantize_b_c;
+ vp9_quantize_b = vp9_quantize_b_sse2;
if (flags & HAS_SSSE3) vp9_quantize_b = vp9_quantize_b_ssse3;
vp9_quantize_b_32x32 = vp9_quantize_b_32x32_c;
if (flags & HAS_SSSE3) vp9_quantize_b_32x32 = vp9_quantize_b_32x32_ssse3;
diff --git a/source/config/mac/x64/vpx_config.asm b/source/config/mac/x64/vpx_config.asm
index e66e8b7..159294e 100644
--- a/source/config/mac/x64/vpx_config.asm
+++ b/source/config/mac/x64/vpx_config.asm
@@ -85,4 +85,4 @@
%define CONFIG_SPATIAL_SVC 0
%define CONFIG_VP9_TEMPORAL_DENOISING 0
%define CONFIG_FP_MB_STATS 0
-%define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+%define CONFIG_EMULATE_HARDWARE 0
diff --git a/source/config/mac/x64/vpx_config.h b/source/config/mac/x64/vpx_config.h
index 61dd009..220ad67 100644
--- a/source/config/mac/x64/vpx_config.h
+++ b/source/config/mac/x64/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/nacl/vp9_rtcd.h b/source/config/nacl/vp9_rtcd.h
index b60c290..0483682 100644
--- a/source/config/nacl/vp9_rtcd.h
+++ b/source/config/nacl/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,9 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_c
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_c
diff --git a/source/config/nacl/vpx_config.asm b/source/config/nacl/vpx_config.asm
index 776ff66..71aee39 100644
--- a/source/config/nacl/vpx_config.asm
+++ b/source/config/nacl/vpx_config.asm
@@ -88,5 +88,5 @@
.equ CONFIG_SPATIAL_SVC , 0
.equ CONFIG_VP9_TEMPORAL_DENOISING , 0
.equ CONFIG_FP_MB_STATS , 0
-.equ CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH , 0
+.equ CONFIG_EMULATE_HARDWARE , 0
.section .note.GNU-stack,"",%progbits
diff --git a/source/config/nacl/vpx_config.h b/source/config/nacl/vpx_config.h
index d02c859..a24d041 100644
--- a/source/config/nacl/vpx_config.h
+++ b/source/config/nacl/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/win/ia32/vp9_rtcd.h b/source/config/win/ia32/vp9_rtcd.h
index 660b652..2d6c530 100644
--- a/source/config/win/ia32/vp9_rtcd.h
+++ b/source/config/win/ia32/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,10 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+unsigned int vp9_avg_8x8_sse2(const uint8_t *, int p);
+RTCD_EXTERN unsigned int (*vp9_avg_8x8)(const uint8_t *, int p);
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
int64_t vp9_block_error_sse2(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
RTCD_EXTERN int64_t (*vp9_block_error)(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
@@ -426,7 +430,8 @@
RTCD_EXTERN unsigned int (*vp9_mse8x8)(const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse);
void vp9_quantize_b_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
-#define vp9_quantize_b vp9_quantize_b_c
+void vp9_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
+RTCD_EXTERN void (*vp9_quantize_b)(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
void vp9_quantize_b_32x32_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
#define vp9_quantize_b_32x32 vp9_quantize_b_32x32_c
@@ -882,6 +887,8 @@
(void)flags;
+ vp9_avg_8x8 = vp9_avg_8x8_c;
+ if (flags & HAS_SSE2) vp9_avg_8x8 = vp9_avg_8x8_sse2;
vp9_block_error = vp9_block_error_c;
if (flags & HAS_SSE2) vp9_block_error = vp9_block_error_sse2;
vp9_convolve8 = vp9_convolve8_c;
@@ -1047,6 +1054,8 @@
if (flags & HAS_SSE2) vp9_mse8x16 = vp9_mse8x16_sse2;
vp9_mse8x8 = vp9_mse8x8_c;
if (flags & HAS_SSE2) vp9_mse8x8 = vp9_mse8x8_sse2;
+ vp9_quantize_b = vp9_quantize_b_c;
+ if (flags & HAS_SSE2) vp9_quantize_b = vp9_quantize_b_sse2;
vp9_sad16x16 = vp9_sad16x16_c;
if (flags & HAS_SSE2) vp9_sad16x16 = vp9_sad16x16_sse2;
vp9_sad16x16_avg = vp9_sad16x16_avg_c;
diff --git a/source/config/win/ia32/vpx_config.asm b/source/config/win/ia32/vpx_config.asm
index a223733..da535f4 100644
--- a/source/config/win/ia32/vpx_config.asm
+++ b/source/config/win/ia32/vpx_config.asm
@@ -85,4 +85,4 @@
%define CONFIG_SPATIAL_SVC 0
%define CONFIG_VP9_TEMPORAL_DENOISING 0
%define CONFIG_FP_MB_STATS 0
-%define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+%define CONFIG_EMULATE_HARDWARE 0
diff --git a/source/config/win/ia32/vpx_config.h b/source/config/win/ia32/vpx_config.h
index d579e01..ddd1db5 100644
--- a/source/config/win/ia32/vpx_config.h
+++ b/source/config/win/ia32/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/config/win/x64/vp9_rtcd.h b/source/config/win/x64/vp9_rtcd.h
index 253f565..f1417de 100644
--- a/source/config/win/x64/vp9_rtcd.h
+++ b/source/config/win/x64/vp9_rtcd.h
@@ -12,8 +12,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -29,6 +29,10 @@
extern "C" {
#endif
+unsigned int vp9_avg_8x8_c(const uint8_t *, int p);
+unsigned int vp9_avg_8x8_sse2(const uint8_t *, int p);
+#define vp9_avg_8x8 vp9_avg_8x8_sse2
+
int64_t vp9_block_error_c(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
int64_t vp9_block_error_sse2(const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz);
#define vp9_block_error vp9_block_error_sse2
@@ -429,6 +433,7 @@
#define vp9_mse8x8 vp9_mse8x8_sse2
void vp9_quantize_b_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
+void vp9_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
void vp9_quantize_b_ssse3(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
RTCD_EXTERN void (*vp9_quantize_b)(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan);
@@ -953,7 +958,7 @@
if (flags & HAS_SSSE3) vp9_idct8x8_12_add = vp9_idct8x8_12_add_ssse3;
vp9_idct8x8_64_add = vp9_idct8x8_64_add_sse2;
if (flags & HAS_SSSE3) vp9_idct8x8_64_add = vp9_idct8x8_64_add_ssse3;
- vp9_quantize_b = vp9_quantize_b_c;
+ vp9_quantize_b = vp9_quantize_b_sse2;
if (flags & HAS_SSSE3) vp9_quantize_b = vp9_quantize_b_ssse3;
vp9_quantize_b_32x32 = vp9_quantize_b_32x32_c;
if (flags & HAS_SSSE3) vp9_quantize_b_32x32 = vp9_quantize_b_32x32_ssse3;
diff --git a/source/config/win/x64/vpx_config.asm b/source/config/win/x64/vpx_config.asm
index 49033d7..93f24d1 100644
--- a/source/config/win/x64/vpx_config.asm
+++ b/source/config/win/x64/vpx_config.asm
@@ -85,4 +85,4 @@
%define CONFIG_SPATIAL_SVC 0
%define CONFIG_VP9_TEMPORAL_DENOISING 0
%define CONFIG_FP_MB_STATS 0
-%define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+%define CONFIG_EMULATE_HARDWARE 0
diff --git a/source/config/win/x64/vpx_config.h b/source/config/win/x64/vpx_config.h
index 936eb36..2d02d26 100644
--- a/source/config/win/x64/vpx_config.h
+++ b/source/config/win/x64/vpx_config.h
@@ -97,5 +97,5 @@
#define CONFIG_SPATIAL_SVC 0
#define CONFIG_VP9_TEMPORAL_DENOISING 0
#define CONFIG_FP_MB_STATS 0
-#define CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH 0
+#define CONFIG_EMULATE_HARDWARE 0
#endif /* VPX_CONFIG_H */
diff --git a/source/libvpx/args.h b/source/libvpx/args.h
index 04e0acd..1f37151 100644
--- a/source/libvpx/args.h
+++ b/source/libvpx/args.h
@@ -51,6 +51,7 @@
unsigned int arg_parse_uint(const struct arg *arg);
int arg_parse_int(const struct arg *arg);
struct vpx_rational arg_parse_rational(const struct arg *arg);
+int arg_parse_enum(const struct arg *arg);
int arg_parse_enum_or_int(const struct arg *arg);
#ifdef __cplusplus
} // extern "C"
diff --git a/source/libvpx/build/make/configure.sh b/source/libvpx/build/make/configure.sh
index ab6687f..56e9f44 100755
--- a/source/libvpx/build/make/configure.sh
+++ b/source/libvpx/build/make/configure.sh
@@ -736,6 +736,9 @@
# PIC is probably what we want when building shared libs
enabled shared && soft_enable pic
+ # Minimum iOS version for all target platforms (darwin and iphonesimulator).
+ IOS_VERSION_MIN="6.0"
+
# Handle darwin variants. Newer SDKs allow targeting older
# platforms, so find the newest SDK available.
case ${toolchain} in
@@ -788,8 +791,8 @@
add_ldflags "-mmacosx-version-min=10.9"
;;
*-iphonesimulator-*)
- add_cflags "-miphoneos-version-min=5.0"
- add_ldflags "-miphoneos-version-min=5.0"
+ add_cflags "-miphoneos-version-min=${IOS_VERSION_MIN}"
+ add_ldflags "-miphoneos-version-min=${IOS_VERSION_MIN}"
osx_sdk_dir="$(xcrun --sdk iphonesimulator --show-sdk-path)"
add_cflags "-isysroot ${osx_sdk_dir}"
add_ldflags "-isysroot ${osx_sdk_dir}"
@@ -970,18 +973,28 @@
;;
darwin*)
-
XCRUN_FIND="xcrun --sdk iphoneos -find"
CXX="$(${XCRUN_FIND} clang++)"
CC="$(${XCRUN_FIND} clang)"
AR="$(${XCRUN_FIND} ar)"
- LD="$(${XCRUN_FIND} ld)"
AS="$(${XCRUN_FIND} as)"
STRIP="$(${XCRUN_FIND} strip)"
NM="$(${XCRUN_FIND} nm)"
RANLIB="$(${XCRUN_FIND} ranlib)"
AS_SFX=.s
+ # Special handling of ld for armv6 because libclang_rt.ios.a does
+ # not contain armv6 support in Apple's clang package:
+ # Apple LLVM version 5.1 (clang-503.0.40) (based on LLVM 3.4svn).
+ # TODO(tomfinegan): Remove this. Our minimum iOS version (6.0)
+ # renders support for armv6 unnecessary because the 3GS and up
+ # support neon.
+ if [ "${tgt_isa}" = "armv6" ]; then
+ LD="$(${XCRUN_FIND} ld)"
+ else
+ LD="${CXX:-$(${XCRUN_FIND} ld)}"
+ fi
+
# ASFLAGS is written here instead of using check_add_asflags
# because we need to overwrite all of ASFLAGS and purge the
# options that were put in above
@@ -989,7 +1002,13 @@
alt_libc="$(xcrun --sdk iphoneos --show-sdk-path)"
add_cflags -arch ${tgt_isa} -isysroot ${alt_libc}
- add_ldflags -arch ${tgt_isa} -ios_version_min 7.0
+ add_ldflags -arch ${tgt_isa}
+
+ if [ "${LD}" = "${CXX}" ]; then
+ add_ldflags -miphoneos-version-min="${IOS_VERSION_MIN}"
+ else
+ add_ldflags -ios_version_min "${IOS_VERSION_MIN}"
+ fi
for d in lib usr/lib usr/lib/system; do
try_dir="${alt_libc}/${d}"
diff --git a/source/libvpx/configure b/source/libvpx/configure
index 7b9c211..3ed976c 100755
--- a/source/libvpx/configure
+++ b/source/libvpx/configure
@@ -281,7 +281,7 @@
spatial_svc
vp9_temporal_denoising
fp_mb_stats
- emulate_hardware_highbitdepth
+ emulate_hardware
"
CONFIG_LIST="
external_build
diff --git a/source/libvpx/examples/simple_decoder.c b/source/libvpx/examples/simple_decoder.c
index 3f7d6aa..c58b014 100644
--- a/source/libvpx/examples/simple_decoder.c
+++ b/source/libvpx/examples/simple_decoder.c
@@ -33,24 +33,25 @@
//
// Initializing The Codec
// ----------------------
-// The decoder is initialized by the following code. This is an example for
-// the VP8 decoder, but the code is analogous for all algorithms. Replace
-// `vpx_codec_vp8_dx()` with a pointer to the interface exposed by the
-// algorithm you want to use. The `cfg` argument is left as NULL in this
-// example, because we want the algorithm to determine the stream
-// configuration (width/height) and allocate memory automatically. This
-// parameter is generally only used if you need to preallocate memory,
-// particularly in External Memory Allocation mode.
+// The libvpx decoder is initialized by the call to vpx_codec_dec_init().
+// Determining the codec interface to use is handled by VpxVideoReader and the
+// functions prefixed with vpx_video_reader_. Discussion of those functions is
+// beyond the scope of this example, but the main gist is to open the input file
+// and parse just enough of it to determine if it's a VPx file and which VPx
+// codec is contained within the file.
+// Note the NULL pointer passed to vpx_codec_dec_init(). We do that in this
+// example because we want the algorithm to determine the stream configuration
+// (width/height) and allocate memory automatically.
//
// Decoding A Frame
// ----------------
// Once the frame has been read into memory, it is decoded using the
// `vpx_codec_decode` function. The call takes a pointer to the data
-// (`frame`) and the length of the data (`frame_sz`). No application data
+// (`frame`) and the length of the data (`frame_size`). No application data
// is associated with the frame in this example, so the `user_priv`
// parameter is NULL. The `deadline` parameter is left at zero for this
-// example. This parameter is generally only used when doing adaptive
-// postprocessing.
+// example. This parameter is generally only used when doing adaptive post
+// processing.
//
// Codecs may produce a variable number of output frames for every call to
// `vpx_codec_decode`. These frames are retrieved by the
@@ -72,14 +73,13 @@
// --------------
// This example does not special case any error return codes. If there was
// an error, a descriptive message is printed and the program exits. With
-// few exeptions, vpx_codec functions return an enumerated error status,
+// few exceptions, vpx_codec functions return an enumerated error status,
// with the value `0` indicating success.
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
-#include "vpx/vp8dx.h"
#include "vpx/vpx_decoder.h"
#include "./tools_common.h"
diff --git a/source/libvpx/examples/vp8cx_set_ref.c b/source/libvpx/examples/vp8cx_set_ref.c
index b0961a2..5e29d80 100644
--- a/source/libvpx/examples/vp8cx_set_ref.c
+++ b/source/libvpx/examples/vp8cx_set_ref.c
@@ -178,7 +178,7 @@
}
// Flush encoder.
- while (encode_frame(&codec, NULL, -1, writer)) {};
+ while (encode_frame(&codec, NULL, -1, writer)) {}
printf("\n");
fclose(infile);
diff --git a/source/libvpx/examples/vp9_spatial_svc_encoder.c b/source/libvpx/examples/vp9_spatial_svc_encoder.c
index 9cd716b..53ede94 100644
--- a/source/libvpx/examples/vp9_spatial_svc_encoder.c
+++ b/source/libvpx/examples/vp9_spatial_svc_encoder.c
@@ -61,12 +61,30 @@
static const arg_def_t max_bitrate_arg =
ARG_DEF(NULL, "max-bitrate", 1, "Maximum bitrate");
+#if CONFIG_VP9_HIGHBITDEPTH
+static const struct arg_enum_list bitdepth_enum[] = {
+ {"8", VPX_BITS_8},
+ {"10", VPX_BITS_10},
+ {"12", VPX_BITS_12},
+ {NULL, 0}
+};
+
+static const arg_def_t bitdepth_arg =
+ ARG_DEF_ENUM("d", "bit-depth", 1, "Bit depth for codec 8, 10 or 12. ",
+ bitdepth_enum);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
+
static const arg_def_t *svc_args[] = {
&frames_arg, &width_arg, &height_arg,
&timebase_arg, &bitrate_arg, &skip_frames_arg, &spatial_layers_arg,
&kf_dist_arg, &scale_factors_arg, &passes_arg, &pass_arg,
&fpf_name_arg, &min_q_arg, &max_q_arg, &min_bitrate_arg,
- &max_bitrate_arg, &temporal_layers_arg, NULL
+ &max_bitrate_arg, &temporal_layers_arg,
+#if CONFIG_VP9_HIGHBITDEPTH
+ &bitdepth_arg,
+#endif
+ NULL
};
static const uint32_t default_frames_to_skip = 0;
@@ -165,7 +183,7 @@
enc_cfg->kf_min_dist = arg_parse_uint(&arg);
enc_cfg->kf_max_dist = enc_cfg->kf_min_dist;
} else if (arg_match(&arg, &scale_factors_arg, argi)) {
- snprintf(string_options, 1024, "%s scale-factors=%s",
+ snprintf(string_options, sizeof(string_options), "%s scale-factors=%s",
string_options, arg.val);
} else if (arg_match(&arg, &passes_arg, argi)) {
passes = arg_parse_uint(&arg);
@@ -180,15 +198,36 @@
} else if (arg_match(&arg, &fpf_name_arg, argi)) {
fpf_file_name = arg.val;
} else if (arg_match(&arg, &min_q_arg, argi)) {
- snprintf(string_options, 1024, "%s min-quantizers=%s",
+ snprintf(string_options, sizeof(string_options), "%s min-quantizers=%s",
string_options, arg.val);
} else if (arg_match(&arg, &max_q_arg, argi)) {
- snprintf(string_options, 1024, "%s max-quantizers=%s",
+ snprintf(string_options, sizeof(string_options), "%s max-quantizers=%s",
string_options, arg.val);
} else if (arg_match(&arg, &min_bitrate_arg, argi)) {
min_bitrate = arg_parse_uint(&arg);
} else if (arg_match(&arg, &max_bitrate_arg, argi)) {
max_bitrate = arg_parse_uint(&arg);
+#if CONFIG_VP9_HIGHBITDEPTH
+ } else if (arg_match(&arg, &bitdepth_arg, argi)) {
+ enc_cfg->g_bit_depth = arg_parse_enum_or_int(&arg);
+ switch (enc_cfg->g_bit_depth) {
+ case VPX_BITS_8:
+ enc_cfg->g_input_bit_depth = 8;
+ enc_cfg->g_profile = 0;
+ break;
+ case VPX_BITS_10:
+ enc_cfg->g_input_bit_depth = 10;
+ enc_cfg->g_profile = 2;
+ break;
+ case VPX_BITS_12:
+ enc_cfg->g_input_bit_depth = 12;
+ enc_cfg->g_profile = 2;
+ break;
+ default:
+ die("Error: Invalid bit depth selected (%d)\n", enc_cfg->g_bit_depth);
+ break;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
} else {
++argj;
}
@@ -291,8 +330,17 @@
parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg);
// Allocate image buffer
- if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, enc_cfg.g_w, enc_cfg.g_h, 32))
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (!vpx_img_alloc(&raw, enc_cfg.g_input_bit_depth == 8 ?
+ VPX_IMG_FMT_I420 : VPX_IMG_FMT_I42016,
+ enc_cfg.g_w, enc_cfg.g_h, 32)) {
die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
+ }
+#else
+ if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, enc_cfg.g_w, enc_cfg.g_h, 32)) {
+ die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (!(infile = fopen(app_input.input_filename, "rb")))
die("Failed to open %s for reading\n", app_input.input_filename);
diff --git a/source/libvpx/examples/vpx_temporal_svc_encoder.c b/source/libvpx/examples/vpx_temporal_svc_encoder.c
index 1674804..ecae2fe 100644
--- a/source/libvpx/examples/vpx_temporal_svc_encoder.c
+++ b/source/libvpx/examples/vpx_temporal_svc_encoder.c
@@ -461,13 +461,27 @@
FILE *infile = NULL;
struct RateControlMetrics rc;
int64_t cx_time = 0;
+ const int min_args_base = 11;
+#if CONFIG_VP9_HIGHBITDEPTH
+ vpx_bit_depth_t bit_depth = VPX_BITS_8;
+ int input_bit_depth = 8;
+ const int min_args = min_args_base + 1;
+#else
+ const int min_args = min_args_base;
+#endif // CONFIG_VP9_HIGHBITDEPTH
exec_name = argv[0];
// Check usage and arguments.
- if (argc < 11) {
+ if (argc < min_args) {
+#if CONFIG_VP9_HIGHBITDEPTH
+ die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
+ "<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
+ "<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n", argv[0]);
+#else
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<Rate_0> ... <Rate_nlayers-1> \n", argv[0]);
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
encoder = get_vpx_encoder_by_name(argv[3]);
@@ -487,13 +501,38 @@
die("Invalid layering mode (0..12) %s", argv[10]);
}
- if (argc != 11 + mode_to_num_layers[layering_mode]) {
+ if (argc != min_args + mode_to_num_layers[layering_mode]) {
die("Invalid number of arguments");
}
+#if CONFIG_VP9_HIGHBITDEPTH
+ switch (strtol(argv[argc-1], NULL, 0)) {
+ case 8:
+ bit_depth = VPX_BITS_8;
+ input_bit_depth = 8;
+ break;
+ case 10:
+ bit_depth = VPX_BITS_10;
+ input_bit_depth = 10;
+ break;
+ case 12:
+ bit_depth = VPX_BITS_12;
+ input_bit_depth = 12;
+ break;
+ default:
+ die("Invalid bit depth (8, 10, 12) %s", argv[argc-1]);
+ }
+ if (!vpx_img_alloc(&raw,
+ bit_depth == VPX_BITS_8 ? VPX_IMG_FMT_I420 :
+ VPX_IMG_FMT_I42016,
+ width, height, 32)) {
+ die("Failed to allocate image", width, height);
+ }
+#else
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {
die("Failed to allocate image", width, height);
}
+#endif // CONFIG_VP9_HIGHBITDEPTH
// Populate encoder configuration.
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
@@ -506,6 +545,14 @@
cfg.g_w = width;
cfg.g_h = height;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (bit_depth != VPX_BITS_8) {
+ cfg.g_bit_depth = bit_depth;
+ cfg.g_input_bit_depth = input_bit_depth;
+ cfg.g_profile = 2;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
// Timebase format e.g. 30fps: numerator=1, demoninator = 30.
cfg.g_timebase.num = strtol(argv[6], NULL, 0);
cfg.g_timebase.den = strtol(argv[7], NULL, 0);
@@ -515,7 +562,9 @@
die("Invalid speed setting: must be positive");
}
- for (i = 11; (int)i < 11 + mode_to_num_layers[layering_mode]; ++i) {
+ for (i = min_args_base;
+ (int)i < min_args_base + mode_to_num_layers[layering_mode];
+ ++i) {
cfg.ts_target_bitrate[i - 11] = strtol(argv[i], NULL, 0);
}
@@ -576,7 +625,13 @@
cfg.ss_number_layers = 1;
// Initialize codec.
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (vpx_codec_enc_init(
+ &codec, encoder->codec_interface(), &cfg,
+ bit_depth == VPX_BITS_8 ? 0 : VPX_CODEC_USE_HIGHBITDEPTH))
+#else
if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
+#endif // CONFIG_VP9_HIGHBITDEPTH
die_codec(&codec, "Failed to initialize encoder");
if (strncmp(encoder->name, "vp8", 3) == 0) {
diff --git a/source/libvpx/test/convolve_test.cc b/source/libvpx/test/convolve_test.cc
index de947aa..470c436 100644
--- a/source/libvpx/test/convolve_test.cc
+++ b/source/libvpx/test/convolve_test.cc
@@ -36,7 +36,7 @@
ConvolveFunc hv8, ConvolveFunc hv8_avg,
int bd)
: h8_(h8), v8_(v8), hv8_(hv8), h8_avg_(h8_avg), v8_avg_(v8_avg),
- hv8_avg_(hv8_avg), use_high_bd_(bd) {}
+ hv8_avg_(hv8_avg), use_highbd_(bd) {}
ConvolveFunc h8_;
ConvolveFunc v8_;
@@ -44,7 +44,7 @@
ConvolveFunc h8_avg_;
ConvolveFunc v8_avg_;
ConvolveFunc hv8_avg_;
- int use_high_bd_; // 0 if high bitdepth not used, else the actual bit depth.
+ int use_highbd_; // 0 if high bitdepth not used, else the actual bit depth.
};
typedef std::tr1::tuple<int, int, const ConvolveFunctions *> ConvolveParam;
@@ -170,15 +170,15 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void high_filter_block2d_8_c(const uint16_t *src_ptr,
- const unsigned int src_stride,
- const int16_t *HFilter,
- const int16_t *VFilter,
- uint16_t *dst_ptr,
- unsigned int dst_stride,
- unsigned int output_width,
- unsigned int output_height,
- int bd) {
+void highbd_filter_block2d_8_c(const uint16_t *src_ptr,
+ const unsigned int src_stride,
+ const int16_t *HFilter,
+ const int16_t *VFilter,
+ uint16_t *dst_ptr,
+ unsigned int dst_stride,
+ unsigned int output_width,
+ unsigned int output_height,
+ int bd) {
// Between passes, we use an intermediate buffer whose height is extended to
// have enough horizontally filtered values as input for the vertical pass.
// This buffer is allocated to be big enough for the largest block type we
@@ -217,7 +217,7 @@
(VP9_FILTER_WEIGHT >> 1); // Rounding
// Normalize back to 0-255...
- *output_ptr = clip_pixel_high(temp >> VP9_FILTER_SHIFT, bd);
+ *output_ptr = clip_pixel_highbd(temp >> VP9_FILTER_SHIFT, bd);
++src_ptr;
output_ptr += intermediate_height;
}
@@ -245,7 +245,7 @@
(VP9_FILTER_WEIGHT >> 1); // Rounding
// Normalize back to 0-255...
- *dst_ptr++ = clip_pixel_high(temp >> VP9_FILTER_SHIFT, bd);
+ *dst_ptr++ = clip_pixel_highbd(temp >> VP9_FILTER_SHIFT, bd);
src_ptr += intermediate_height;
}
src_ptr += intermediate_next_stride;
@@ -254,13 +254,13 @@
}
}
-void high_block2d_average_c(uint16_t *src,
- unsigned int src_stride,
- uint16_t *output_ptr,
- unsigned int output_stride,
- unsigned int output_width,
- unsigned int output_height,
- int bd) {
+void highbd_block2d_average_c(uint16_t *src,
+ unsigned int src_stride,
+ uint16_t *output_ptr,
+ unsigned int output_stride,
+ unsigned int output_width,
+ unsigned int output_height,
+ int bd) {
unsigned int i, j;
for (i = 0; i < output_height; ++i) {
for (j = 0; j < output_width; ++j) {
@@ -270,23 +270,23 @@
}
}
-void high_filter_average_block2d_8_c(const uint16_t *src_ptr,
- const unsigned int src_stride,
- const int16_t *HFilter,
- const int16_t *VFilter,
- uint16_t *dst_ptr,
- unsigned int dst_stride,
- unsigned int output_width,
- unsigned int output_height,
- int bd) {
+void highbd_filter_average_block2d_8_c(const uint16_t *src_ptr,
+ const unsigned int src_stride,
+ const int16_t *HFilter,
+ const int16_t *VFilter,
+ uint16_t *dst_ptr,
+ unsigned int dst_stride,
+ unsigned int output_width,
+ unsigned int output_height,
+ int bd) {
uint16_t tmp[kMaxDimension * kMaxDimension];
assert(output_width <= kMaxDimension);
assert(output_height <= kMaxDimension);
- high_filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter, tmp, 64,
- output_width, output_height, bd);
- high_block2d_average_c(tmp, 64, dst_ptr, dst_stride,
- output_width, output_height, bd);
+ highbd_filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter, tmp, 64,
+ output_width, output_height, bd);
+ highbd_block2d_average_c(tmp, 64, dst_ptr, dst_stride,
+ output_width, output_height, bd);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
@@ -346,8 +346,8 @@
virtual void SetUp() {
UUT_ = GET_PARAM(2);
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ != 0)
- mask_ = (1 << UUT_->use_high_bd_) - 1;
+ if (UUT_->use_highbd_ != 0)
+ mask_ = (1 << UUT_->use_highbd_) - 1;
else
mask_ = 255;
#endif
@@ -391,7 +391,7 @@
uint8_t *input() const {
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
return input_ + BorderTop() * kOuterBlockSize + BorderLeft();
} else {
return CONVERT_TO_BYTEPTR(input16_ + BorderTop() * kOuterBlockSize +
@@ -404,7 +404,7 @@
uint8_t *output() const {
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
return output_ + BorderTop() * kOuterBlockSize + BorderLeft();
} else {
return CONVERT_TO_BYTEPTR(output16_ + BorderTop() * kOuterBlockSize +
@@ -417,7 +417,7 @@
uint16_t lookup(uint8_t *list, int index) const {
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
return list[index];
} else {
return CONVERT_TO_SHORTPTR(list)[index];
@@ -429,7 +429,7 @@
void assign_val(uint8_t *list, int index, uint16_t val) const {
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
list[index] = (uint8_t) val;
} else {
CONVERT_TO_SHORTPTR(list)[index] = val;
@@ -448,16 +448,16 @@
unsigned int output_width,
unsigned int output_height) {
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
filter_average_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
dst_ptr, dst_stride, output_width,
output_height);
} else {
- high_filter_average_block2d_8_c(CONVERT_TO_SHORTPTR(src_ptr), src_stride,
- HFilter, VFilter,
- CONVERT_TO_SHORTPTR(dst_ptr), dst_stride,
- output_width, output_height,
- UUT_->use_high_bd_);
+ highbd_filter_average_block2d_8_c(CONVERT_TO_SHORTPTR(src_ptr),
+ src_stride, HFilter, VFilter,
+ CONVERT_TO_SHORTPTR(dst_ptr),
+ dst_stride, output_width, output_height,
+ UUT_->use_highbd_);
}
#else
filter_average_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
@@ -475,14 +475,14 @@
unsigned int output_width,
unsigned int output_height) {
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
dst_ptr, dst_stride, output_width, output_height);
} else {
- high_filter_block2d_8_c(CONVERT_TO_SHORTPTR(src_ptr), src_stride,
- HFilter, VFilter,
- CONVERT_TO_SHORTPTR(dst_ptr), dst_stride,
- output_width, output_height, UUT_->use_high_bd_);
+ highbd_filter_block2d_8_c(CONVERT_TO_SHORTPTR(src_ptr), src_stride,
+ HFilter, VFilter,
+ CONVERT_TO_SHORTPTR(dst_ptr), dst_stride,
+ output_width, output_height, UUT_->use_highbd_);
}
#else
filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
@@ -598,7 +598,7 @@
uint8_t ref8[kOutputStride * kMaxDimension];
uint16_t ref16[kOutputStride * kMaxDimension];
uint8_t* ref;
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
ref = ref8;
} else {
ref = CONVERT_TO_BYTEPTR(ref16);
@@ -657,7 +657,7 @@
uint8_t ref8[kOutputStride * kMaxDimension];
uint16_t ref16[kOutputStride * kMaxDimension];
uint8_t* ref;
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
ref = ref8;
} else {
ref = CONVERT_TO_BYTEPTR(ref16);
@@ -672,7 +672,7 @@
for (int x = 0; x < Width(); ++x) {
uint16_t r;
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ == 0 || UUT_->use_high_bd_ == 8) {
+ if (UUT_->use_highbd_ == 0 || UUT_->use_highbd_ == 8) {
r = prng.Rand8Extremes();
} else {
r = prng.Rand16() & mask_;
@@ -736,7 +736,7 @@
uint8_t ref8[kOutputStride * kMaxDimension];
uint16_t ref16[kOutputStride * kMaxDimension];
uint8_t *ref;
- if (UUT_->use_high_bd_ == 0) {
+ if (UUT_->use_highbd_ == 0) {
ref = ref8;
} else {
ref = CONVERT_TO_BYTEPTR(ref16);
@@ -751,7 +751,7 @@
for (int x = 0; x < Width(); ++x) {
uint16_t r;
#if CONFIG_VP9_HIGHBITDEPTH
- if (UUT_->use_high_bd_ == 0 || UUT_->use_high_bd_ == 8) {
+ if (UUT_->use_highbd_ == 0 || UUT_->use_highbd_ == 8) {
r = prng.Rand8Extremes();
} else {
r = prng.Rand16() & mask_;
@@ -978,9 +978,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_horiz_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 8);
+ vp9_highbd_convolve8_horiz_sse2(src, src_stride, dst, dst_stride, filter_x,
+ filter_x_stride, filter_y, filter_y_stride,
+ w, h, 8);
}
void wrap_convolve8_avg_horiz_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -990,8 +990,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_horiz_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_avg_horiz_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_vert_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1001,8 +1002,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_vert_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_vert_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_avg_vert_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1012,9 +1014,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_vert_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 8);
+ vp9_highbd_convolve8_avg_vert_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1024,8 +1026,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_avg_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1035,8 +1038,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_avg_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_horiz_sse2_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1046,8 +1050,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_horiz_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_horiz_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_avg_horiz_sse2_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1057,8 +1062,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_horiz_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_avg_horiz_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_vert_sse2_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1068,8 +1074,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_vert_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_vert_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_avg_vert_sse2_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1079,8 +1086,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_vert_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_avg_vert_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_sse2_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1090,8 +1098,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_avg_sse2_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1101,9 +1110,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 10);
+ vp9_highbd_convolve8_avg_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_horiz_sse2_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1113,9 +1122,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_horiz_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_horiz_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_avg_horiz_sse2_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1125,9 +1134,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_horiz_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_avg_horiz_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_vert_sse2_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1137,9 +1146,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_vert_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_vert_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_avg_vert_sse2_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1149,8 +1158,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_vert_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 12);
+ vp9_highbd_convolve8_avg_vert_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_sse2_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1160,8 +1170,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 12);
+ vp9_highbd_convolve8_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_avg_sse2_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1171,8 +1182,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_sse2(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 12);
+ vp9_highbd_convolve8_avg_sse2(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
#endif // HAVE_SSE2 && ARCH_X86_64
@@ -1183,8 +1195,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_horiz_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_horiz_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_avg_horiz_c_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1194,8 +1207,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_vert_c_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1205,8 +1219,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_vert_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_vert_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_avg_vert_c_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1216,8 +1231,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_vert_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_avg_vert_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_c_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1227,8 +1243,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 8);
+ vp9_highbd_convolve8_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_avg_c_8(const uint8_t *src, ptrdiff_t src_stride,
@@ -1238,9 +1255,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 8);
+ vp9_highbd_convolve8_avg_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 8);
}
void wrap_convolve8_horiz_c_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1250,8 +1267,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_horiz_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_horiz_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_avg_horiz_c_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1261,9 +1279,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 10);
+ vp9_highbd_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_vert_c_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1273,8 +1291,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_vert_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_vert_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_avg_vert_c_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1284,8 +1303,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_vert_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_avg_vert_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_c_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1295,8 +1315,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_avg_c_10(const uint8_t *src, ptrdiff_t src_stride,
@@ -1306,8 +1327,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride, w, h, 10);
+ vp9_highbd_convolve8_avg_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 10);
}
void wrap_convolve8_horiz_c_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1317,9 +1339,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_horiz_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_horiz_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_avg_horiz_c_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1329,9 +1351,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_vert_c_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1341,9 +1363,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_vert_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_vert_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_avg_vert_c_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1353,9 +1375,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_vert_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_avg_vert_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_c_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1365,9 +1387,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
void wrap_convolve8_avg_c_12(const uint8_t *src, ptrdiff_t src_stride,
@@ -1377,9 +1399,9 @@
const int16_t *filter_y,
int filter_y_stride,
int w, int h) {
- vp9_high_convolve8_avg_c(src, src_stride, dst, dst_stride, filter_x,
- filter_x_stride, filter_y, filter_y_stride,
- w, h, 12);
+ vp9_highbd_convolve8_avg_c(src, src_stride, dst, dst_stride,
+ filter_x, filter_x_stride,
+ filter_y, filter_y_stride, w, h, 12);
}
const ConvolveFunctions convolve8_c(
diff --git a/source/libvpx/test/dct16x16_test.cc b/source/libvpx/test/dct16x16_test.cc
index d1ce109..5222d49 100644
--- a/source/libvpx/test/dct16x16_test.cc
+++ b/source/libvpx/test/dct16x16_test.cc
@@ -287,11 +287,11 @@
#if CONFIG_VP9_HIGHBITDEPTH
void idct16x16_10(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_idct16x16_256_add_c(in, out, stride, 10);
+ vp9_highbd_idct16x16_256_add_c(in, out, stride, 10);
}
void idct16x16_12(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_idct16x16_256_add_c(in, out, stride, 12);
+ vp9_highbd_idct16x16_256_add_c(in, out, stride, 12);
}
void idct16x16_10_ref(const tran_low_t *in, uint8_t *out, int stride,
@@ -305,11 +305,11 @@
}
void iht16x16_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
- vp9_high_iht16x16_256_add_c(in, out, stride, tx_type, 10);
+ vp9_highbd_iht16x16_256_add_c(in, out, stride, tx_type, 10);
}
void iht16x16_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
- vp9_high_iht16x16_256_add_c(in, out, stride, tx_type, 12);
+ vp9_highbd_iht16x16_256_add_c(in, out, stride, tx_type, 12);
}
#endif
@@ -443,7 +443,7 @@
void RunQuantCheck(int dc_thred, int ac_thred) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
- const int count_test_block = 1000;
+ const int count_test_block = 100000;
DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, kNumCoeffs);
DECLARE_ALIGNED_ARRAY(16, int16_t, input_extreme_block, kNumCoeffs);
DECLARE_ALIGNED_ARRAY(16, tran_low_t, output_ref_block, kNumCoeffs);
@@ -700,7 +700,7 @@
TEST_P(Trans16x16HT, QuantCheck) {
// The encoder skips any non-DC intra prediction modes,
// when the quantization step size goes beyond 988.
- RunQuantCheck(549, 988);
+ RunQuantCheck(429, 729);
}
using std::tr1::make_tuple;
@@ -709,8 +709,8 @@
INSTANTIATE_TEST_CASE_P(
C, Trans16x16DCT,
::testing::Values(
- make_tuple(&vp9_high_fdct16x16_c, &idct16x16_10, 0, VPX_BITS_10),
- make_tuple(&vp9_high_fdct16x16_c, &idct16x16_12, 0, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fdct16x16_c, &idct16x16_10, 0, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fdct16x16_c, &idct16x16_12, 0, VPX_BITS_12),
make_tuple(&vp9_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
@@ -723,14 +723,14 @@
INSTANTIATE_TEST_CASE_P(
C, Trans16x16HT,
::testing::Values(
- make_tuple(&vp9_high_fht16x16_c, &iht16x16_10, 0, VPX_BITS_10),
- make_tuple(&vp9_high_fht16x16_c, &iht16x16_10, 1, VPX_BITS_10),
- make_tuple(&vp9_high_fht16x16_c, &iht16x16_10, 2, VPX_BITS_10),
- make_tuple(&vp9_high_fht16x16_c, &iht16x16_10, 3, VPX_BITS_10),
- make_tuple(&vp9_high_fht16x16_c, &iht16x16_12, 0, VPX_BITS_12),
- make_tuple(&vp9_high_fht16x16_c, &iht16x16_12, 1, VPX_BITS_12),
- make_tuple(&vp9_high_fht16x16_c, &iht16x16_12, 2, VPX_BITS_12),
- make_tuple(&vp9_high_fht16x16_c, &iht16x16_12, 3, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_10, 0, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_10, 1, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_10, 2, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_10, 3, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_12, 0, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_12, 1, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_12, 2, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_12, 3, VPX_BITS_12),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 0, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 1, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 2, VPX_BITS_8),
@@ -745,7 +745,7 @@
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 3, VPX_BITS_8)));
#endif
-#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
NEON, Trans16x16DCT,
::testing::Values(
@@ -753,7 +753,7 @@
&vp9_idct16x16_256_add_neon, 0, VPX_BITS_8)));
#endif
-#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSE2, Trans16x16DCT,
::testing::Values(
@@ -772,7 +772,7 @@
VPX_BITS_8)));
#endif
-#if HAVE_SSSE3 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSSE3 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSSE3, Trans16x16DCT,
::testing::Values(
diff --git a/source/libvpx/test/dct32x32_test.cc b/source/libvpx/test/dct32x32_test.cc
index c7a1931..4f08be5 100644
--- a/source/libvpx/test/dct32x32_test.cc
+++ b/source/libvpx/test/dct32x32_test.cc
@@ -80,11 +80,11 @@
#if CONFIG_VP9_HIGHBITDEPTH
void idct32x32_10(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_idct32x32_1024_add_c(in, out, stride, 10);
+ vp9_highbd_idct32x32_1024_add_c(in, out, stride, 10);
}
void idct32x32_12(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_idct32x32_1024_add_c(in, out, stride, 12);
+ vp9_highbd_idct32x32_1024_add_c(in, out, stride, 12);
}
#endif
@@ -311,13 +311,13 @@
INSTANTIATE_TEST_CASE_P(
C, Trans32x32Test,
::testing::Values(
- make_tuple(&vp9_high_fdct32x32_c,
+ make_tuple(&vp9_highbd_fdct32x32_c,
&idct32x32_10, 0, VPX_BITS_10),
- make_tuple(&vp9_high_fdct32x32_rd_c,
+ make_tuple(&vp9_highbd_fdct32x32_rd_c,
&idct32x32_10, 1, VPX_BITS_10),
- make_tuple(&vp9_high_fdct32x32_c,
+ make_tuple(&vp9_highbd_fdct32x32_c,
&idct32x32_12, 0, VPX_BITS_12),
- make_tuple(&vp9_high_fdct32x32_rd_c,
+ make_tuple(&vp9_highbd_fdct32x32_rd_c,
&idct32x32_12, 1, VPX_BITS_12),
make_tuple(&vp9_fdct32x32_c,
&vp9_idct32x32_1024_add_c, 0, VPX_BITS_8),
@@ -333,7 +333,7 @@
&vp9_idct32x32_1024_add_c, 1, VPX_BITS_8)));
#endif
-#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
NEON, Trans32x32Test,
::testing::Values(
@@ -343,7 +343,7 @@
&vp9_idct32x32_1024_add_neon, 1, VPX_BITS_8)));
#endif
-#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSE2, Trans32x32Test,
::testing::Values(
@@ -353,7 +353,7 @@
&vp9_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
#endif
-#if HAVE_AVX2 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_AVX2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
AVX2, Trans32x32Test,
::testing::Values(
diff --git a/source/libvpx/test/fdct4x4_test.cc b/source/libvpx/test/fdct4x4_test.cc
index f803c8e..ef35902 100644
--- a/source/libvpx/test/fdct4x4_test.cc
+++ b/source/libvpx/test/fdct4x4_test.cc
@@ -53,27 +53,27 @@
#if CONFIG_VP9_HIGHBITDEPTH
void idct4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_idct4x4_16_add_c(in, out, stride, 10);
+ vp9_highbd_idct4x4_16_add_c(in, out, stride, 10);
}
void idct4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_idct4x4_16_add_c(in, out, stride, 12);
+ vp9_highbd_idct4x4_16_add_c(in, out, stride, 12);
}
void iht4x4_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
- vp9_high_iht4x4_16_add_c(in, out, stride, tx_type, 10);
+ vp9_highbd_iht4x4_16_add_c(in, out, stride, tx_type, 10);
}
void iht4x4_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
- vp9_high_iht4x4_16_add_c(in, out, stride, tx_type, 12);
+ vp9_highbd_iht4x4_16_add_c(in, out, stride, tx_type, 12);
}
void iwht4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_iwht4x4_16_add_c(in, out, stride, 10);
+ vp9_highbd_iwht4x4_16_add_c(in, out, stride, 10);
}
void iwht4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_iwht4x4_16_add_c(in, out, stride, 12);
+ vp9_highbd_iwht4x4_16_add_c(in, out, stride, 12);
}
#endif
@@ -408,8 +408,8 @@
INSTANTIATE_TEST_CASE_P(
C, Trans4x4DCT,
::testing::Values(
- make_tuple(&vp9_high_fdct4x4_c, &idct4x4_10, 0, VPX_BITS_10),
- make_tuple(&vp9_high_fdct4x4_c, &idct4x4_12, 0, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fdct4x4_c, &idct4x4_10, 0, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fdct4x4_c, &idct4x4_12, 0, VPX_BITS_12),
make_tuple(&vp9_fdct4x4_c, &vp9_idct4x4_16_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
@@ -422,14 +422,14 @@
INSTANTIATE_TEST_CASE_P(
C, Trans4x4HT,
::testing::Values(
- make_tuple(&vp9_high_fht4x4_c, &iht4x4_10, 0, VPX_BITS_10),
- make_tuple(&vp9_high_fht4x4_c, &iht4x4_10, 1, VPX_BITS_10),
- make_tuple(&vp9_high_fht4x4_c, &iht4x4_10, 2, VPX_BITS_10),
- make_tuple(&vp9_high_fht4x4_c, &iht4x4_10, 3, VPX_BITS_10),
- make_tuple(&vp9_high_fht4x4_c, &iht4x4_12, 0, VPX_BITS_12),
- make_tuple(&vp9_high_fht4x4_c, &iht4x4_12, 1, VPX_BITS_12),
- make_tuple(&vp9_high_fht4x4_c, &iht4x4_12, 2, VPX_BITS_12),
- make_tuple(&vp9_high_fht4x4_c, &iht4x4_12, 3, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 0, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 1, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 2, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 3, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 0, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 1, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 2, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 3, VPX_BITS_12),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8),
@@ -448,8 +448,8 @@
INSTANTIATE_TEST_CASE_P(
C, Trans4x4WHT,
::testing::Values(
- make_tuple(&vp9_high_fwht4x4_c, &iwht4x4_10, 0, VPX_BITS_10),
- make_tuple(&vp9_high_fwht4x4_c, &iwht4x4_12, 0, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_10, 0, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_12, 0, VPX_BITS_12),
make_tuple(&vp9_fwht4x4_c, &vp9_iwht4x4_16_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
@@ -458,7 +458,7 @@
make_tuple(&vp9_fwht4x4_c, &vp9_iwht4x4_16_add_c, 0, VPX_BITS_8)));
#endif
-#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
NEON, Trans4x4DCT,
::testing::Values(
@@ -473,14 +473,15 @@
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 3, VPX_BITS_8)));
#endif
-#if CONFIG_USE_X86INC && HAVE_MMX && !CONFIG_VP9_HIGHBITDEPTH
+#if CONFIG_USE_X86INC && HAVE_MMX && !CONFIG_VP9_HIGHBITDEPTH && \
+ !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
MMX, Trans4x4WHT,
::testing::Values(
make_tuple(&vp9_fwht4x4_mmx, &vp9_iwht4x4_16_add_c, 0, VPX_BITS_8)));
#endif
-#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4DCT,
::testing::Values(
diff --git a/source/libvpx/test/fdct8x8_test.cc b/source/libvpx/test/fdct8x8_test.cc
index 60d0be5..103556d 100644
--- a/source/libvpx/test/fdct8x8_test.cc
+++ b/source/libvpx/test/fdct8x8_test.cc
@@ -82,19 +82,19 @@
#if CONFIG_VP9_HIGHBITDEPTH
void idct8x8_10(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_idct8x8_64_add_c(in, out, stride, 10);
+ vp9_highbd_idct8x8_64_add_c(in, out, stride, 10);
}
void idct8x8_12(const tran_low_t *in, uint8_t *out, int stride) {
- vp9_high_idct8x8_64_add_c(in, out, stride, 12);
+ vp9_highbd_idct8x8_64_add_c(in, out, stride, 12);
}
void iht8x8_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
- vp9_high_iht8x8_64_add_c(in, out, stride, tx_type, 10);
+ vp9_highbd_iht8x8_64_add_c(in, out, stride, tx_type, 10);
}
void iht8x8_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
- vp9_high_iht8x8_64_add_c(in, out, stride, tx_type, 12);
+ vp9_highbd_iht8x8_64_add_c(in, out, stride, tx_type, 12);
}
#endif
@@ -532,8 +532,8 @@
INSTANTIATE_TEST_CASE_P(
C, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_high_fdct8x8_c, &idct8x8_10, 0, VPX_BITS_10),
- make_tuple(&vp9_high_fdct8x8_c, &idct8x8_12, 0, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fdct8x8_c, &idct8x8_10, 0, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fdct8x8_c, &idct8x8_12, 0, VPX_BITS_12),
make_tuple(&vp9_fdct8x8_c, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
@@ -546,14 +546,14 @@
INSTANTIATE_TEST_CASE_P(
C, FwdTrans8x8HT,
::testing::Values(
- make_tuple(&vp9_high_fht8x8_c, &iht8x8_10, 0, VPX_BITS_10),
- make_tuple(&vp9_high_fht8x8_c, &iht8x8_10, 1, VPX_BITS_10),
- make_tuple(&vp9_high_fht8x8_c, &iht8x8_10, 2, VPX_BITS_10),
- make_tuple(&vp9_high_fht8x8_c, &iht8x8_10, 3, VPX_BITS_10),
- make_tuple(&vp9_high_fht8x8_c, &iht8x8_12, 0, VPX_BITS_12),
- make_tuple(&vp9_high_fht8x8_c, &iht8x8_12, 1, VPX_BITS_12),
- make_tuple(&vp9_high_fht8x8_c, &iht8x8_12, 2, VPX_BITS_12),
- make_tuple(&vp9_high_fht8x8_c, &iht8x8_12, 3, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_10, 0, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_10, 1, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_10, 2, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_10, 3, VPX_BITS_10),
+ make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_12, 0, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_12, 1, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_12, 2, VPX_BITS_12),
+ make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_12, 3, VPX_BITS_12),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 0, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 1, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 2, VPX_BITS_8),
@@ -568,7 +568,7 @@
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 3, VPX_BITS_8)));
#endif
-#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
NEON, FwdTrans8x8DCT,
::testing::Values(
@@ -583,7 +583,7 @@
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_neon, 3, VPX_BITS_8)));
#endif
-#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8DCT,
::testing::Values(
@@ -598,7 +598,8 @@
make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 3, VPX_BITS_8)));
#endif
-#if HAVE_SSSE3 && ARCH_X86_64 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSSE3 && ARCH_X86_64 && !CONFIG_VP9_HIGHBITDEPTH && \
+ !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSSE3, FwdTrans8x8DCT,
::testing::Values(
diff --git a/source/libvpx/test/frame_size_tests.cc b/source/libvpx/test/frame_size_tests.cc
index 1c9a522..95cc66a 100644
--- a/source/libvpx/test/frame_size_tests.cc
+++ b/source/libvpx/test/frame_size_tests.cc
@@ -84,4 +84,13 @@
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
#endif
}
+
+TEST_F(VP9FrameSizeTestsLarge, OneByOneVideo) {
+ ::libvpx_test::RandomVideoSource video;
+
+ video.SetSize(1, 1);
+ video.set_limit(2);
+ expected_res_ = VPX_CODEC_OK;
+ ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+}
} // namespace
diff --git a/source/libvpx/test/i420_video_source.h b/source/libvpx/test/i420_video_source.h
index c3315f9..0a18480 100644
--- a/source/libvpx/test/i420_video_source.h
+++ b/source/libvpx/test/i420_video_source.h
@@ -13,104 +13,22 @@
#include <cstdlib>
#include <string>
-#include "test/video_source.h"
+#include "test/yuv_video_source.h"
namespace libvpx_test {
// This class extends VideoSource to allow parsing of raw yv12
// so that we can do actual file encodes.
-class I420VideoSource : public VideoSource {
+class I420VideoSource : public YUVVideoSource {
public:
I420VideoSource(const std::string &file_name,
unsigned int width, unsigned int height,
int rate_numerator, int rate_denominator,
unsigned int start, int limit)
- : file_name_(file_name),
- input_file_(NULL),
- img_(NULL),
- start_(start),
- limit_(limit),
- frame_(0),
- width_(0),
- height_(0),
- framerate_numerator_(rate_numerator),
- framerate_denominator_(rate_denominator) {
- // This initializes raw_sz_, width_, height_ and allocates an img.
- SetSize(width, height);
- }
-
- virtual ~I420VideoSource() {
- vpx_img_free(img_);
- if (input_file_)
- fclose(input_file_);
- }
-
- virtual void Begin() {
- if (input_file_)
- fclose(input_file_);
- input_file_ = OpenTestDataFile(file_name_);
- ASSERT_TRUE(input_file_ != NULL) << "Input file open failed. Filename: "
- << file_name_;
- if (start_) {
- fseek(input_file_, static_cast<unsigned>(raw_sz_) * start_, SEEK_SET);
- }
-
- frame_ = start_;
- FillFrame();
- }
-
- virtual void Next() {
- ++frame_;
- FillFrame();
- }
-
- virtual vpx_image_t *img() const { return (frame_ < limit_) ? img_ : NULL; }
-
- // Models a stream where Timebase = 1/FPS, so pts == frame.
- virtual vpx_codec_pts_t pts() const { return frame_; }
-
- virtual unsigned long duration() const { return 1; }
-
- virtual vpx_rational_t timebase() const {
- const vpx_rational_t t = { framerate_denominator_, framerate_numerator_ };
- return t;
- }
-
- virtual unsigned int frame() const { return frame_; }
-
- virtual unsigned int limit() const { return limit_; }
-
- void SetSize(unsigned int width, unsigned int height) {
- if (width != width_ || height != height_) {
- vpx_img_free(img_);
- img_ = vpx_img_alloc(NULL, VPX_IMG_FMT_I420, width, height, 1);
- ASSERT_TRUE(img_ != NULL);
- width_ = width;
- height_ = height;
- raw_sz_ = width * height * 3 / 2;
- }
- }
-
- virtual void FillFrame() {
- ASSERT_TRUE(input_file_ != NULL);
- // Read a frame from input_file.
- if (fread(img_->img_data, raw_sz_, 1, input_file_) == 0) {
- limit_ = frame_;
- }
- }
-
- protected:
- std::string file_name_;
- FILE *input_file_;
- vpx_image_t *img_;
- size_t raw_sz_;
- unsigned int start_;
- unsigned int limit_;
- unsigned int frame_;
- unsigned int width_;
- unsigned int height_;
- int framerate_numerator_;
- int framerate_denominator_;
+ : YUVVideoSource(file_name, VPX_IMG_FMT_I420,
+ width, height,
+ rate_numerator, rate_denominator,
+ start, limit) {}
};
} // namespace libvpx_test
diff --git a/source/libvpx/test/lpf_8_test.cc b/source/libvpx/test/lpf_8_test.cc
new file mode 100644
index 0000000..abc4107
--- /dev/null
+++ b/source/libvpx/test/lpf_8_test.cc
@@ -0,0 +1,587 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 <cmath>
+#include <cstdlib>
+#include <string>
+
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+
+#include "./vpx_config.h"
+#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_entropy.h"
+#include "vpx/vpx_integer.h"
+
+using libvpx_test::ACMRandom;
+
+namespace {
+// Horizontally and Vertically need 32x32: 8 Coeffs preceeding filtered section
+// 16 Coefs within filtered section
+// 8 Coeffs following filtered section
+const int kNumCoeffs = 1024;
+
+const int number_of_iterations = 10000;
+
+#if CONFIG_VP9_HIGHBITDEPTH
+typedef void (*loop_op_t)(uint16_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd);
+typedef void (*dual_loop_op_t)(uint16_t *s, int p, const uint8_t *blimit0,
+ const uint8_t *limit0, const uint8_t *thresh0,
+ const uint8_t *blimit1, const uint8_t *limit1,
+ const uint8_t *thresh1, int bd);
+#else
+typedef void (*loop_op_t)(uint8_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count);
+typedef void (*dual_loop_op_t)(uint8_t *s, int p, const uint8_t *blimit0,
+ const uint8_t *limit0, const uint8_t *thresh0,
+ const uint8_t *blimit1, const uint8_t *limit1,
+ const uint8_t *thresh1);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
+typedef std::tr1::tuple<loop_op_t, loop_op_t, int> loop8_param_t;
+typedef std::tr1::tuple<dual_loop_op_t, dual_loop_op_t, int> dualloop8_param_t;
+
+#if HAVE_SSE2
+#if CONFIG_VP9_HIGHBITDEPTH
+void wrapper_vertical_16_sse2(uint16_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd) {
+ vp9_highbd_lpf_vertical_16_sse2(s, p, blimit, limit, thresh, bd);
+}
+
+void wrapper_vertical_16_c(uint16_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd) {
+ vp9_highbd_lpf_vertical_16_c(s, p, blimit, limit, thresh, bd);
+}
+
+void wrapper_vertical_16_dual_sse2(uint16_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd) {
+ vp9_highbd_lpf_vertical_16_dual_sse2(s, p, blimit, limit, thresh, bd);
+}
+
+void wrapper_vertical_16_dual_c(uint16_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd) {
+ vp9_highbd_lpf_vertical_16_dual_c(s, p, blimit, limit, thresh, bd);
+}
+#else
+void wrapper_vertical_16_sse2(uint8_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count) {
+ vp9_lpf_vertical_16_sse2(s, p, blimit, limit, thresh);
+}
+
+void wrapper_vertical_16_c(uint8_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count) {
+ vp9_lpf_vertical_16_c(s, p, blimit, limit, thresh);
+}
+
+void wrapper_vertical_16_dual_sse2(uint8_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count) {
+ vp9_lpf_vertical_16_dual_sse2(s, p, blimit, limit, thresh);
+}
+
+void wrapper_vertical_16_dual_c(uint8_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count) {
+ vp9_lpf_vertical_16_dual_c(s, p, blimit, limit, thresh);
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+#endif // HAVE_SSE2
+
+class Loop8Test6Param : public ::testing::TestWithParam<loop8_param_t> {
+ public:
+ virtual ~Loop8Test6Param() {}
+ virtual void SetUp() {
+ loopfilter_op_ = GET_PARAM(0);
+ ref_loopfilter_op_ = GET_PARAM(1);
+ bit_depth_ = GET_PARAM(2);
+ mask_ = (1 << bit_depth_) - 1;
+ }
+
+ virtual void TearDown() { libvpx_test::ClearSystemState(); }
+
+ protected:
+ int bit_depth_;
+ int mask_;
+ loop_op_t loopfilter_op_;
+ loop_op_t ref_loopfilter_op_;
+};
+
+class Loop8Test9Param : public ::testing::TestWithParam<dualloop8_param_t> {
+ public:
+ virtual ~Loop8Test9Param() {}
+ virtual void SetUp() {
+ loopfilter_op_ = GET_PARAM(0);
+ ref_loopfilter_op_ = GET_PARAM(1);
+ bit_depth_ = GET_PARAM(2);
+ mask_ = (1 << bit_depth_) - 1;
+ }
+
+ virtual void TearDown() { libvpx_test::ClearSystemState(); }
+
+ protected:
+ int bit_depth_;
+ int mask_;
+ dual_loop_op_t loopfilter_op_;
+ dual_loop_op_t ref_loopfilter_op_;
+};
+
+TEST_P(Loop8Test6Param, OperationCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = number_of_iterations;
+#if CONFIG_VP9_HIGHBITDEPTH
+ int32_t bd = bit_depth_;
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, s, kNumCoeffs);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, ref_s, kNumCoeffs);
+#else
+ DECLARE_ALIGNED_ARRAY(8, uint8_t, s, kNumCoeffs);
+ DECLARE_ALIGNED_ARRAY(8, uint8_t, ref_s, kNumCoeffs);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ int err_count_total = 0;
+ int first_failure = -1;
+ for (int i = 0; i < count_test_block; ++i) {
+ int err_count = 0;
+ uint8_t tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, blimit[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, limit[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, thresh[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ int32_t p = kNumCoeffs/32;
+ int count = 1;
+
+ uint16_t tmp_s[kNumCoeffs];
+ int j = 0;
+ while (j < kNumCoeffs) {
+ uint8_t val = rnd.Rand8();
+ if (val & 0x80) { // 50% chance to choose a new value.
+ tmp_s[j] = rnd.Rand16();
+ j++;
+ } else { // 50% chance to repeat previous value in row X times
+ int k = 0;
+ while (k++ < ((val & 0x1f) + 1) && j < kNumCoeffs) {
+ if (j < 1) {
+ tmp_s[j] = rnd.Rand16();
+ } else if (val & 0x20) { // Increment by an value within the limit
+ tmp_s[j] = (tmp_s[j - 1] + (*limit - 1));
+ } else { // Decrement by an value within the limit
+ tmp_s[j] = (tmp_s[j - 1] - (*limit - 1));
+ }
+ j++;
+ }
+ }
+ }
+ for (j = 0; j < kNumCoeffs; j++) {
+ if (i % 2) {
+ s[j] = tmp_s[j] & mask_;
+ } else {
+ s[j] = tmp_s[p * (j % p) + j / p] & mask_;
+ }
+ ref_s[j] = s[j];
+ }
+#if CONFIG_VP9_HIGHBITDEPTH
+ ref_loopfilter_op_(ref_s + 8 + p * 8, p, blimit, limit, thresh, count, bd);
+ ASM_REGISTER_STATE_CHECK(
+ loopfilter_op_(s + 8 + p * 8, p, blimit, limit, thresh, count, bd));
+#else
+ ref_loopfilter_op_(ref_s+8+p*8, p, blimit, limit, thresh, count);
+ ASM_REGISTER_STATE_CHECK(
+ loopfilter_op_(s + 8 + p * 8, p, blimit, limit, thresh, count));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
+ for (int j = 0; j < kNumCoeffs; ++j) {
+ err_count += ref_s[j] != s[j];
+ }
+ if (err_count && !err_count_total) {
+ first_failure = i;
+ }
+ err_count_total += err_count;
+ }
+ EXPECT_EQ(0, err_count_total)
+ << "Error: Loop8Test6Param, C output doesn't match SSE2 "
+ "loopfilter output. "
+ << "First failed at test case " << first_failure;
+}
+
+TEST_P(Loop8Test6Param, ValueCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = number_of_iterations;
+#if CONFIG_VP9_HIGHBITDEPTH
+ const int32_t bd = bit_depth_;
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, s, kNumCoeffs);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, ref_s, kNumCoeffs);
+#else
+ DECLARE_ALIGNED_ARRAY(8, uint8_t, s, kNumCoeffs);
+ DECLARE_ALIGNED_ARRAY(8, uint8_t, ref_s, kNumCoeffs);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ int err_count_total = 0;
+ int first_failure = -1;
+ for (int i = 0; i < count_test_block; ++i) {
+ int err_count = 0;
+ uint8_t tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, blimit[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, limit[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, thresh[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ int32_t p = kNumCoeffs / 32;
+ int count = 1;
+ for (int j = 0; j < kNumCoeffs; ++j) {
+ s[j] = rnd.Rand16() & mask_;
+ ref_s[j] = s[j];
+ }
+#if CONFIG_VP9_HIGHBITDEPTH
+ ref_loopfilter_op_(ref_s + 8 + p * 8, p, blimit, limit, thresh, count, bd);
+ ASM_REGISTER_STATE_CHECK(
+ loopfilter_op_(s + 8 + p * 8, p, blimit, limit, thresh, count, bd));
+#else
+ ref_loopfilter_op_(ref_s+8+p*8, p, blimit, limit, thresh, count);
+ ASM_REGISTER_STATE_CHECK(
+ loopfilter_op_(s + 8 + p * 8, p, blimit, limit, thresh, count));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ for (int j = 0; j < kNumCoeffs; ++j) {
+ err_count += ref_s[j] != s[j];
+ }
+ if (err_count && !err_count_total) {
+ first_failure = i;
+ }
+ err_count_total += err_count;
+ }
+ EXPECT_EQ(0, err_count_total)
+ << "Error: Loop8Test6Param, C output doesn't match SSE2 "
+ "loopfilter output. "
+ << "First failed at test case " << first_failure;
+}
+
+TEST_P(Loop8Test9Param, OperationCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = number_of_iterations;
+#if CONFIG_VP9_HIGHBITDEPTH
+ const int32_t bd = bit_depth_;
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, s, kNumCoeffs);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, ref_s, kNumCoeffs);
+#else
+ DECLARE_ALIGNED_ARRAY(8, uint8_t, s, kNumCoeffs);
+ DECLARE_ALIGNED_ARRAY(8, uint8_t, ref_s, kNumCoeffs);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ int err_count_total = 0;
+ int first_failure = -1;
+ for (int i = 0; i < count_test_block; ++i) {
+ int err_count = 0;
+ uint8_t tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, blimit0[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, limit0[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, thresh0[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, blimit1[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, limit1[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, thresh1[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ int32_t p = kNumCoeffs / 32;
+ uint16_t tmp_s[kNumCoeffs];
+ int j = 0;
+ const uint8_t limit = *limit0 < *limit1 ? *limit0 : *limit1;
+ while (j < kNumCoeffs) {
+ uint8_t val = rnd.Rand8();
+ if (val & 0x80) { // 50% chance to choose a new value.
+ tmp_s[j] = rnd.Rand16();
+ j++;
+ } else { // 50% chance to repeat previous value in row X times.
+ int k = 0;
+ while (k++ < ((val & 0x1f) + 1) && j < kNumCoeffs) {
+ if (j < 1) {
+ tmp_s[j] = rnd.Rand16();
+ } else if (val & 0x20) { // Increment by a value within the limit.
+ tmp_s[j] = (tmp_s[j - 1] + (limit - 1));
+ } else { // Decrement by an value within the limit.
+ tmp_s[j] = (tmp_s[j - 1] - (limit - 1));
+ }
+ j++;
+ }
+ }
+ }
+ for (j = 0; j < kNumCoeffs; j++) {
+ if (i % 2) {
+ s[j] = tmp_s[j] & mask_;
+ } else {
+ s[j] = tmp_s[p * (j % p) + j / p] & mask_;
+ }
+ ref_s[j] = s[j];
+ }
+#if CONFIG_VP9_HIGHBITDEPTH
+ ref_loopfilter_op_(ref_s + 8 + p * 8, p, blimit0, limit0, thresh0,
+ blimit1, limit1, thresh1, bd);
+ ASM_REGISTER_STATE_CHECK(
+ loopfilter_op_(s + 8 + p * 8, p, blimit0, limit0, thresh0,
+ blimit1, limit1, thresh1, bd));
+#else
+ ref_loopfilter_op_(ref_s + 8 + p * 8, p, blimit0, limit0, thresh0,
+ blimit1, limit1, thresh1);
+ ASM_REGISTER_STATE_CHECK(
+ loopfilter_op_(s + 8 + p * 8, p, blimit0, limit0, thresh0,
+ blimit1, limit1, thresh1));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ for (int j = 0; j < kNumCoeffs; ++j) {
+ err_count += ref_s[j] != s[j];
+ }
+ if (err_count && !err_count_total) {
+ first_failure = i;
+ }
+ err_count_total += err_count;
+ }
+ EXPECT_EQ(0, err_count_total)
+ << "Error: Loop8Test9Param, C output doesn't match SSE2 "
+ "loopfilter output. "
+ << "First failed at test case " << first_failure;
+}
+
+TEST_P(Loop8Test9Param, ValueCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = number_of_iterations;
+#if CONFIG_VP9_HIGHBITDEPTH
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, s, kNumCoeffs);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, ref_s, kNumCoeffs);
+#else
+ DECLARE_ALIGNED_ARRAY(8, uint8_t, s, kNumCoeffs);
+ DECLARE_ALIGNED_ARRAY(8, uint8_t, ref_s, kNumCoeffs);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ int err_count_total = 0;
+ int first_failure = -1;
+ for (int i = 0; i < count_test_block; ++i) {
+ int err_count = 0;
+ uint8_t tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, blimit0[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, limit0[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, thresh0[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, blimit1[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, limit1[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ tmp = rnd.Rand8();
+ DECLARE_ALIGNED(16, const uint8_t, thresh1[16]) = {
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
+ tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp
+ };
+ int32_t p = kNumCoeffs / 32; // TODO(pdlf) can we have non-square here?
+ for (int j = 0; j < kNumCoeffs; ++j) {
+ s[j] = rnd.Rand16() & mask_;
+ ref_s[j] = s[j];
+ }
+#if CONFIG_VP9_HIGHBITDEPTH
+ const int32_t bd = bit_depth_;
+ ref_loopfilter_op_(ref_s + 8 + p * 8, p, blimit0, limit0, thresh0,
+ blimit1, limit1, thresh1, bd);
+ ASM_REGISTER_STATE_CHECK(
+ loopfilter_op_(s + 8 + p * 8, p, blimit0, limit0,
+ thresh0, blimit1, limit1, thresh1, bd));
+#else
+ ref_loopfilter_op_(ref_s + 8 + p * 8, p, blimit0, limit0, thresh0,
+ blimit1, limit1, thresh1);
+ ASM_REGISTER_STATE_CHECK(
+ loopfilter_op_(s + 8 + p * 8, p, blimit0, limit0, thresh0,
+ blimit1, limit1, thresh1));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ for (int j = 0; j < kNumCoeffs; ++j) {
+ err_count += ref_s[j] != s[j];
+ }
+ if (err_count && !err_count_total) {
+ first_failure = i;
+ }
+ err_count_total += err_count;
+ }
+ EXPECT_EQ(0, err_count_total)
+ << "Error: Loop8Test9Param, C output doesn't match SSE2"
+ "loopfilter output. "
+ << "First failed at test case " << first_failure;
+}
+
+using std::tr1::make_tuple;
+
+#if HAVE_SSE2
+#if CONFIG_VP9_HIGHBITDEPTH
+INSTANTIATE_TEST_CASE_P(
+ SSE2_C_COMPARE_SINGLE, Loop8Test6Param,
+ ::testing::Values(
+ make_tuple(&vp9_highbd_lpf_horizontal_4_sse2,
+ &vp9_highbd_lpf_horizontal_4_c, 8),
+ make_tuple(&vp9_highbd_lpf_vertical_4_sse2,
+ &vp9_highbd_lpf_vertical_4_c, 8),
+ make_tuple(&vp9_highbd_lpf_horizontal_8_sse2,
+ &vp9_highbd_lpf_horizontal_8_c, 8),
+ make_tuple(&vp9_highbd_lpf_horizontal_16_sse2,
+ &vp9_highbd_lpf_horizontal_16_c, 8),
+ make_tuple(&vp9_highbd_lpf_vertical_8_sse2,
+ &vp9_highbd_lpf_vertical_8_c, 8),
+ make_tuple(&wrapper_vertical_16_sse2,
+ &wrapper_vertical_16_c, 8),
+ make_tuple(&vp9_highbd_lpf_horizontal_4_sse2,
+ &vp9_highbd_lpf_horizontal_4_c, 10),
+ make_tuple(&vp9_highbd_lpf_vertical_4_sse2,
+ &vp9_highbd_lpf_vertical_4_c, 10),
+ make_tuple(&vp9_highbd_lpf_horizontal_8_sse2,
+ &vp9_highbd_lpf_horizontal_8_c, 10),
+ make_tuple(&vp9_highbd_lpf_horizontal_16_sse2,
+ &vp9_highbd_lpf_horizontal_16_c, 10),
+ make_tuple(&vp9_highbd_lpf_vertical_8_sse2,
+ &vp9_highbd_lpf_vertical_8_c, 10),
+ make_tuple(&wrapper_vertical_16_sse2,
+ &wrapper_vertical_16_c, 10),
+ make_tuple(&vp9_highbd_lpf_horizontal_4_sse2,
+ &vp9_highbd_lpf_horizontal_4_c, 12),
+ make_tuple(&vp9_highbd_lpf_vertical_4_sse2,
+ &vp9_highbd_lpf_vertical_4_c, 12),
+ make_tuple(&vp9_highbd_lpf_horizontal_8_sse2,
+ &vp9_highbd_lpf_horizontal_8_c, 12),
+ make_tuple(&vp9_highbd_lpf_horizontal_16_sse2,
+ &vp9_highbd_lpf_horizontal_16_c, 12),
+ make_tuple(&vp9_highbd_lpf_vertical_8_sse2,
+ &vp9_highbd_lpf_vertical_8_c, 12),
+ make_tuple(&wrapper_vertical_16_sse2,
+ &wrapper_vertical_16_c, 12)));
+#else
+INSTANTIATE_TEST_CASE_P(
+ SSE2_C_COMPARE_SINGLE, Loop8Test6Param,
+ ::testing::Values(
+ make_tuple(&vp9_lpf_horizontal_8_sse2, &vp9_lpf_horizontal_8_c, 8),
+ make_tuple(&vp9_lpf_horizontal_16_sse2, &vp9_lpf_horizontal_16_c, 8),
+ make_tuple(&vp9_lpf_vertical_8_sse2, &vp9_lpf_vertical_8_c, 8)));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+#endif
+
+#if HAVE_SSE2
+#if CONFIG_VP9_HIGHBITDEPTH
+INSTANTIATE_TEST_CASE_P(
+ SSE2_C_COMPARE_DUAL, Loop8Test6Param,
+ ::testing::Values(
+ make_tuple(&wrapper_vertical_16_dual_sse2,
+ &wrapper_vertical_16_dual_c, 8),
+ make_tuple(&wrapper_vertical_16_dual_sse2,
+ &wrapper_vertical_16_dual_c, 10),
+ make_tuple(&wrapper_vertical_16_dual_sse2,
+ &wrapper_vertical_16_dual_c, 12)));
+#else
+INSTANTIATE_TEST_CASE_P(
+ SSE2_C_COMPARE_DUAL, Loop8Test6Param,
+ ::testing::Values(
+ make_tuple(&wrapper_vertical_16_sse2, &wrapper_vertical_16_c, 8)));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+#endif // HAVE_SSE2
+
+#if HAVE_SSE2
+#if CONFIG_VP9_HIGHBITDEPTH
+INSTANTIATE_TEST_CASE_P(
+ SSE_C_COMPARE_DUAL, Loop8Test9Param,
+ ::testing::Values(
+ make_tuple(&vp9_highbd_lpf_horizontal_4_dual_sse2,
+ &vp9_highbd_lpf_horizontal_4_dual_c, 8),
+ make_tuple(&vp9_highbd_lpf_horizontal_8_dual_sse2,
+ &vp9_highbd_lpf_horizontal_8_dual_c, 8),
+ make_tuple(&vp9_highbd_lpf_vertical_4_dual_sse2,
+ &vp9_highbd_lpf_vertical_4_dual_c, 8),
+ make_tuple(&vp9_highbd_lpf_vertical_8_dual_sse2,
+ &vp9_highbd_lpf_vertical_8_dual_c, 8),
+ make_tuple(&vp9_highbd_lpf_horizontal_4_dual_sse2,
+ &vp9_highbd_lpf_horizontal_4_dual_c, 10),
+ make_tuple(&vp9_highbd_lpf_horizontal_8_dual_sse2,
+ &vp9_highbd_lpf_horizontal_8_dual_c, 10),
+ make_tuple(&vp9_highbd_lpf_vertical_4_dual_sse2,
+ &vp9_highbd_lpf_vertical_4_dual_c, 10),
+ make_tuple(&vp9_highbd_lpf_vertical_8_dual_sse2,
+ &vp9_highbd_lpf_vertical_8_dual_c, 10),
+ make_tuple(&vp9_highbd_lpf_horizontal_4_dual_sse2,
+ &vp9_highbd_lpf_horizontal_4_dual_c, 12),
+ make_tuple(&vp9_highbd_lpf_horizontal_8_dual_sse2,
+ &vp9_highbd_lpf_horizontal_8_dual_c, 12),
+ make_tuple(&vp9_highbd_lpf_vertical_4_dual_sse2,
+ &vp9_highbd_lpf_vertical_4_dual_c, 12),
+ make_tuple(&vp9_highbd_lpf_vertical_8_dual_sse2,
+ &vp9_highbd_lpf_vertical_8_dual_c, 12)));
+#else
+INSTANTIATE_TEST_CASE_P(
+ SSE_C_COMPARE_DUAL, Loop8Test9Param,
+ ::testing::Values(
+ make_tuple(&vp9_lpf_horizontal_4_dual_sse2,
+ &vp9_lpf_horizontal_4_dual_c, 8),
+ make_tuple(&vp9_lpf_horizontal_8_dual_sse2,
+ &vp9_lpf_horizontal_8_dual_c, 8),
+ make_tuple(&vp9_lpf_vertical_4_dual_sse2,
+ &vp9_lpf_vertical_4_dual_c, 8),
+ make_tuple(&vp9_lpf_vertical_8_dual_sse2,
+ &vp9_lpf_vertical_8_dual_c, 8)));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+#endif
+} // namespace
diff --git a/source/libvpx/test/partial_idct_test.cc b/source/libvpx/test/partial_idct_test.cc
index 9c24fee..536273e 100644
--- a/source/libvpx/test/partial_idct_test.cc
+++ b/source/libvpx/test/partial_idct_test.cc
@@ -260,7 +260,7 @@
TX_4X4, 1)));
#endif
-#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSE2, PartialIDctTest,
::testing::Values(
@@ -294,7 +294,8 @@
TX_4X4, 1)));
#endif
-#if HAVE_SSSE3 && ARCH_X86_64 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSSE3 && ARCH_X86_64 && !CONFIG_VP9_HIGHBITDEPTH && \
+ !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSSE3_64, PartialIDctTest,
::testing::Values(
@@ -304,7 +305,7 @@
TX_8X8, 12)));
#endif
-#if HAVE_SSSE3 && !CONFIG_VP9_HIGHBITDEPTH
+#if HAVE_SSSE3 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSSE3, PartialIDctTest,
::testing::Values(
diff --git a/source/libvpx/test/sad_test.cc b/source/libvpx/test/sad_test.cc
index 5377c1e..c7042fe 100644
--- a/source/libvpx/test/sad_test.cc
+++ b/source/libvpx/test/sad_test.cc
@@ -625,6 +625,20 @@
#if HAVE_AVX2
#if CONFIG_VP9_ENCODER
+const SadMxNVp9Func sad_64x64_avx2_vp9 = vp9_sad64x64_avx2;
+const SadMxNVp9Func sad_64x32_avx2_vp9 = vp9_sad64x32_avx2;
+const SadMxNVp9Func sad_32x64_avx2_vp9 = vp9_sad32x64_avx2;
+const SadMxNVp9Func sad_32x32_avx2_vp9 = vp9_sad32x32_avx2;
+const SadMxNVp9Func sad_32x16_avx2_vp9 = vp9_sad32x16_avx2;
+const SadMxNVp9Param avx2_vp9_tests[] = {
+ make_tuple(64, 64, sad_64x64_avx2_vp9),
+ make_tuple(64, 32, sad_64x32_avx2_vp9),
+ make_tuple(32, 64, sad_32x64_avx2_vp9),
+ make_tuple(32, 32, sad_32x32_avx2_vp9),
+ make_tuple(32, 16, sad_32x16_avx2_vp9),
+};
+INSTANTIATE_TEST_CASE_P(AVX2, SADVP9Test, ::testing::ValuesIn(avx2_vp9_tests));
+
const SadMxNx4Func sad_64x64x4d_avx2 = vp9_sad64x64x4d_avx2;
const SadMxNx4Func sad_32x32x4d_avx2 = vp9_sad32x32x4d_avx2;
INSTANTIATE_TEST_CASE_P(AVX2, SADx4Test, ::testing::Values(
diff --git a/source/libvpx/test/svc_test.cc b/source/libvpx/test/svc_test.cc
index fdde702..67e83e3 100644
--- a/source/libvpx/test/svc_test.cc
+++ b/source/libvpx/test/svc_test.cc
@@ -225,10 +225,9 @@
EXPECT_EQ(received_frames, n);
}
- void DropLayersAndMakeItVP9Comaptible(struct vpx_fixed_buf *const inputs,
- const int num_super_frames,
- const int remained_spatial_layers,
- const bool is_multiple_frame_contexts) {
+ void DropEnhancementLayers(struct vpx_fixed_buf *const inputs,
+ const int num_super_frames,
+ const int remained_spatial_layers) {
ASSERT_TRUE(inputs != NULL);
ASSERT_GT(num_super_frames, 0);
ASSERT_GT(remained_spatial_layers, 0);
@@ -250,45 +249,6 @@
if (frame_count == 0) {
// There's no super frame but only a single frame.
ASSERT_EQ(1, remained_spatial_layers);
- if (is_multiple_frame_contexts) {
- // Make a new super frame.
- uint8_t marker = 0xc1;
- unsigned int mask;
- int mag;
-
- // Choose the magnitude.
- for (mag = 0, mask = 0xff; mag < 4; ++mag) {
- if (inputs[i].sz < mask)
- break;
- mask <<= 8;
- mask |= 0xff;
- }
- marker |= mag << 3;
- int index_sz = 2 + (mag + 1) * 2;
-
- inputs[i].buf = realloc(inputs[i].buf, inputs[i].sz + index_sz + 16);
- ASSERT_TRUE(inputs[i].buf != NULL);
- uint8_t *frame_data = static_cast<uint8_t*>(inputs[i].buf);
- frame_data[0] &= ~2; // Set the show_frame flag to 0.
- frame_data += inputs[i].sz;
- // Add an one byte frame with show_existing_frame.
- *frame_data++ = 0x88;
-
- // Write the super frame index.
- *frame_data++ = marker;
-
- frame_sizes[0] = inputs[i].sz;
- frame_sizes[1] = 1;
- for (int j = 0; j < 2; ++j) {
- unsigned int this_sz = frame_sizes[j];
- for (int k = 0; k <= mag; k++) {
- *frame_data++ = this_sz & 0xff;
- this_sz >>= 8;
- }
- }
- *frame_data++ = marker;
- inputs[i].sz += index_sz + 1;
- }
} else {
// Found a super frame.
uint8_t *frame_data = static_cast<uint8_t*>(inputs[i].buf);
@@ -304,16 +264,13 @@
}
ASSERT_LT(frame, frame_count) << "Couldn't find a visible frame. "
<< "remained_spatial_layers: " << remained_spatial_layers
- << " super_frame: " << i
- << " is_multiple_frame_context: " << is_multiple_frame_contexts;
- if (frame == frame_count - 1 && !is_multiple_frame_contexts)
+ << " super_frame: " << i;
+ if (frame == frame_count - 1)
continue;
frame_data += frame_sizes[frame];
// We need to add one more frame for multiple frame contexts.
- if (is_multiple_frame_contexts)
- ++frame;
uint8_t marker =
static_cast<const uint8_t*>(inputs[i].buf)[inputs[i].sz - 1];
const uint32_t mag = ((marker >> 3) & 0x3) + 1;
@@ -323,35 +280,14 @@
marker |= frame;
// Copy existing frame sizes.
- memmove(frame_data + (is_multiple_frame_contexts ? 2 : 1),
- frame_start + inputs[i].sz - index_sz + 1, new_index_sz - 2);
- if (is_multiple_frame_contexts) {
- // Add a one byte frame with flag show_existing_frame.
- *frame_data++ = 0x88 | (remained_spatial_layers - 1);
- }
+ memmove(frame_data + 1, frame_start + inputs[i].sz - index_sz + 1,
+ new_index_sz - 2);
// New marker.
frame_data[0] = marker;
frame_data += (mag * (frame + 1) + 1);
- if (is_multiple_frame_contexts) {
- // Write the frame size for the one byte frame.
- frame_data -= mag;
- *frame_data++ = 1;
- for (uint32_t j = 1; j < mag; ++j) {
- *frame_data++ = 0;
- }
- }
-
*frame_data++ = marker;
inputs[i].sz = frame_data - frame_start;
-
- if (is_multiple_frame_contexts) {
- // Change the show frame flag to 0 for all frames.
- for (int j = 0; j < frame; ++j) {
- frame_start[0] &= ~2;
- frame_start += frame_sizes[j];
- }
- }
}
}
}
@@ -555,7 +491,7 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, false);
+ DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -573,13 +509,13 @@
Pass2EncodeNFrames(&stats_buf, 10, 5, &outputs[0]);
DecodeNFrames(&outputs[0], 10);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 4, false);
+ DropEnhancementLayers(&outputs[0], 10, 4);
DecodeNFrames(&outputs[0], 10);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 3, false);
+ DropEnhancementLayers(&outputs[0], 10, 3);
DecodeNFrames(&outputs[0], 10);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 2, false);
+ DropEnhancementLayers(&outputs[0], 10, 2);
DecodeNFrames(&outputs[0], 10);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, false);
+ DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
@@ -616,9 +552,9 @@
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 20, 3, &outputs[0]);
DecodeNFrames(&outputs[0], 20);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 20, 2, false);
+ DropEnhancementLayers(&outputs[0], 20, 2);
DecodeNFrames(&outputs[0], 20);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 20, 1, false);
+ DropEnhancementLayers(&outputs[0], 20, 1);
DecodeNFrames(&outputs[0], 20);
FreeBitstreamBuffers(&outputs[0], 20);
@@ -649,7 +585,6 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 2, true);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -667,7 +602,7 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, true);
+ DropEnhancementLayers(&outputs[0], 10, 1);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -686,7 +621,6 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 2, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 2, true);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -707,32 +641,13 @@
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 3, &outputs[0]);
- vpx_fixed_buf outputs_new[10];
- for (int i = 0; i < 10; ++i) {
- outputs_new[i].buf = malloc(outputs[i].sz + 16);
- ASSERT_TRUE(outputs_new[i].buf != NULL);
- memcpy(outputs_new[i].buf, outputs[i].buf, outputs[i].sz);
- outputs_new[i].sz = outputs[i].sz;
- }
- DropLayersAndMakeItVP9Comaptible(&outputs_new[0], 10, 3, true);
- DecodeNFrames(&outputs_new[0], 10);
-
- for (int i = 0; i < 10; ++i) {
- memcpy(outputs_new[i].buf, outputs[i].buf, outputs[i].sz);
- outputs_new[i].sz = outputs[i].sz;
- }
- DropLayersAndMakeItVP9Comaptible(&outputs_new[0], 10, 2, true);
- DecodeNFrames(&outputs_new[0], 10);
-
- for (int i = 0; i < 10; ++i) {
- memcpy(outputs_new[i].buf, outputs[i].buf, outputs[i].sz);
- outputs_new[i].sz = outputs[i].sz;
- }
- DropLayersAndMakeItVP9Comaptible(&outputs_new[0], 10, 1, true);
- DecodeNFrames(&outputs_new[0], 10);
+ DecodeNFrames(&outputs[0], 10);
+ DropEnhancementLayers(&outputs[0], 10, 2);
+ DecodeNFrames(&outputs[0], 10);
+ DropEnhancementLayers(&outputs[0], 10, 1);
+ DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
- FreeBitstreamBuffers(&outputs_new[0], 10);
}
TEST_F(SvcTest, TwoPassEncode2TemporalLayers) {
@@ -769,7 +684,6 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, true);
DecodeNFrames(&outputs[0], 10);
FreeBitstreamBuffers(&outputs[0], 10);
}
@@ -814,7 +728,6 @@
vpx_fixed_buf outputs[10];
memset(&outputs[0], 0, sizeof(outputs));
Pass2EncodeNFrames(&stats_buf, 10, 1, &outputs[0]);
- DropLayersAndMakeItVP9Comaptible(&outputs[0], 10, 1, true);
vpx_fixed_buf base_layer[5];
for (int i = 0; i < 5; ++i)
diff --git a/source/libvpx/test/test-data.mk b/source/libvpx/test/test-data.mk
new file mode 100644
index 0000000..e4dae3a
--- /dev/null
+++ b/source/libvpx/test/test-data.mk
@@ -0,0 +1,726 @@
+LIBVPX_TEST_SRCS-yes += test-data.mk
+
+# Encoder test source
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += hantro_collage_w352h288.yuv
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += hantro_odd.yuv
+
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_420.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_422.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_444.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_440.yuv
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_420.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_422.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_444.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_440.yuv
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_420.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_422.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_444.y4m
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_440.yuv
+
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += rush_hour_444.y4m
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += screendata.y4m
+
+# Test vectors
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-001.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-001.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-002.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-002.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-003.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-003.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-004.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-004.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-005.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-005.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-006.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-006.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-007.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-007.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-008.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-008.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-009.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-009.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-010.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-010.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-011.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-011.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-012.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-012.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-013.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-013.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-014.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-014.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-015.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-015.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-016.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-016.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-017.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-017.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-018.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-018.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1400.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1400.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1411.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1411.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1416.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1416.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1417.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1417.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1402.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1402.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1412.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1412.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1418.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1418.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1424.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1424.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-01.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-01.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-02.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-02.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-03.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-03.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-04.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-04.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1401.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1401.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1403.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1403.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1407.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1407.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1408.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1408.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1409.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1409.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1410.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1410.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1413.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1413.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1414.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1414.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1415.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1415.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1425.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1425.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1426.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1426.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1427.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1427.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1432.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1432.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1435.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1435.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1436.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1436.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1437.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1437.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1441.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1441.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1442.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1442.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1404.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1404.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1405.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1405.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1406.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1406.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1428.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1428.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1429.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1429.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1430.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1430.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1431.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1431.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1433.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1433.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1434.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1434.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1438.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1438.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1439.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1439.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1440.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1440.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1443.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1443.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-06-smallsize.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-06-smallsize.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-00.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-00.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-01.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-01.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-02.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-02.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-03.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-03.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-04.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-04.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-05.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-05.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-06.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-06.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-07.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-07.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-09.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-09.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-11.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-11.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-12.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-12.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-13.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-13.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-14.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-14.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-15.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-15.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-17.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-17.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-19.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-19.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-20.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-20.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-21.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-21.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-22.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-22.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-23.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-23.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-24.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-24.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-25.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-25.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-26.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-26.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-27.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-27.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-28.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-28.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-29.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-29.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-30.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-30.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-31.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-31.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-33.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-33.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-35.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-35.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-36.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-36.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-37.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-37.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-38.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-38.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-39.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-39.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-40.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-40.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-41.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-41.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-42.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-42.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-43.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-43.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-44.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-44.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-45.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-45.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-46.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-46.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-47.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-47.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-48.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-48.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-49.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-49.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-50.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-50.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-51.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-51.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-52.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-52.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-53.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-53.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-54.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-54.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-55.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-55.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-56.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-56.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-57.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-57.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-58.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-58.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-59.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-59.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-60.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-60.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-61.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-61.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-62.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-62.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-63.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-63.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-1.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-1.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-2.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-3.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-3.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-4.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-4.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-5.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-5.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-6.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-6.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-7.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-7.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x64.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x64.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x66.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x66.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x64.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x64.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x66.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x66.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x64.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x64.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x66.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x66.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x64.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x64.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x66.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x66.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x64.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x64.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x66.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x66.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x64.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x64.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x66.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x66.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x64.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x64.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x66.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x66.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x08.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x08.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x10.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x10.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x18.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x18.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x32.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x32.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x34.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x34.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x64.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x64.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x66.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x66.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-lf-1920x1080.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-lf-1920x1080.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-deltaq.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-deltaq.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x196.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x196.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x198.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x198.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x200.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x200.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x202.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x202.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x208.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x208.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x210.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x210.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x224.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x224.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x226.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x226.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x196.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x196.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x198.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x198.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x200.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x200.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x202.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x202.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x208.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x208.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x210.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x210.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x224.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x224.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x226.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x226.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x196.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x196.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x198.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x198.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x200.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x200.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x202.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x202.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x208.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x208.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x210.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x210.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x224.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x224.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x226.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x226.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x196.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x196.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x198.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x198.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x200.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x200.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x202.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x202.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x208.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x208.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x210.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x210.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x224.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x224.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x226.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x226.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x196.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x196.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x198.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x198.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x200.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x200.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x202.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x202.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x208.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x208.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x210.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x210.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x224.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x224.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x226.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x226.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x196.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x196.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x198.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x198.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x200.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x200.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x202.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x202.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x208.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x208.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x210.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x210.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x224.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x224.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x226.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x226.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x196.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x196.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x198.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x198.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x200.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x200.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x202.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x202.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x208.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x208.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x210.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x210.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x224.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x224.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x226.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x226.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x196.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x196.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x198.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x198.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x200.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x200.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x202.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x202.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x208.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x208.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x210.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x210.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x224.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x224.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x226.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x226.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-05-resize.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-05-resize.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-06-bilinear.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-06-bilinear.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-07-frame_parallel.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-07-frame_parallel.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile-4x1.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile-4x1.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile-4x4.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile-4x4.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x2.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x2_frame_parallel.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x2_frame_parallel.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x4.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x4.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x4_frame_parallel.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x4_frame_parallel.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x8.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x8.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x8_frame_parallel.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x8_frame_parallel.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-aq2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-aq2.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-lf_deltas.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-lf_deltas.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-subpixel-00.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-subpixel-00.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-10-show-existing-frame.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-10-show-existing-frame.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-10-show-existing-frame2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-10-show-existing-frame2.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-351x287.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-351x287.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-351x288.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-351x288.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-352x287.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-352x287.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_1.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_1.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_2.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_2.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_3.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_3.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-13-largescaling.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-13-largescaling.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-2-4-8-16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-2-4-8-16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-2.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-4.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-4.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-8.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-8.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-1.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-1.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-2.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-4.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-4.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-8-4-2-1.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-8-4-2-1.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-8.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-8.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-1.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-1.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-4.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-4.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-8.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-8.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-1.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-1.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-2.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-8.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-8.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-1.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-1.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-16.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-16.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-2.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-4.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-4.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-15-segkey.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-15-segkey.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-15-segkey_adpq.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-15-segkey_adpq.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-16-intra-only.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-16-intra-only.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-17-show-existing-frame.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-17-show-existing-frame.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-18-resize.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-18-resize.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-01.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-01.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-02.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-02.webm.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp91-2-04-yuv444.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp91-2-04-yuv444.webm.md5
+
+# Invalid files for testing libvpx error checking.
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-01-v2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-01-v2.webm.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-02-v2.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-02-v2.webm.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-03-v3.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-03-v3.webm.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-00.webm.ivf.s5861_r01-05_b6-.v2.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-00.webm.ivf.s5861_r01-05_b6-.v2.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-z.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-z.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-05-resize.ivf.s59293_r01-05_b6-.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-05-resize.ivf.s59293_r01-05_b6-.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-08-tile_1x2_frame_parallel.webm.ivf.s47039_r01-05_b6-.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-08-tile_1x2_frame_parallel.webm.ivf.s47039_r01-05_b6-.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-08-tile_1x4_frame_parallel_all_key.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-08-tile_1x4_frame_parallel_all_key.webm.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-aq2.webm.ivf.s3984_r01-05_b6-.v2.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-aq2.webm.ivf.s3984_r01-05_b6-.v2.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s19552_r01-05_b6-.v2.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s19552_r01-05_b6-.v2.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s20492_r01-05_b6-.v2.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s20492_r01-05_b6-.v2.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s73804_r01-05_b6-.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s73804_r01-05_b6-.ivf.res
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp91-2-mixedrefcsp-444to420.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp91-2-mixedrefcsp-444to420.ivf.res
+
+ifeq ($(CONFIG_DECODE_PERF_TESTS),yes)
+# BBB VP9 streams
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-bbb_426x240_tile_1x1_180kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-bbb_640x360_tile_1x2_337kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-bbb_854x480_tile_1x2_651kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-bbb_1280x720_tile_1x4_1310kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-bbb_1920x1080_tile_1x1_2581kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-bbb_1920x1080_tile_1x4_2586kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-bbb_1920x1080_tile_1x4_fpm_2304kbps.webm
+# Sintel VP9 streams
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-sintel_426x182_tile_1x1_171kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-sintel_640x272_tile_1x2_318kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-sintel_854x364_tile_1x2_621kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-sintel_1280x546_tile_1x4_1257kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-sintel_1920x818_tile_1x4_fpm_2279kbps.webm
+# TOS VP9 streams
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-tos_426x178_tile_1x1_181kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-tos_640x266_tile_1x2_336kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-tos_854x356_tile_1x2_656kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-tos_854x356_tile_1x2_fpm_546kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-tos_1280x534_tile_1x4_1306kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-tos_1280x534_tile_1x4_fpm_952kbps.webm
+LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-tos_1920x800_tile_1x4_fpm_2335kbps.webm
+endif # CONFIG_DECODE_PERF_TESTS
+
+ifeq ($(CONFIG_ENCODE_PERF_TESTS),yes)
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += desktop_640_360_30.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += kirland_640_480_30.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += macmarcomoving_640_480_30.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += macmarcostationary_640_480_30.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += niklas_1280_720_30.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += niklas_640_480_30.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += tacomanarrows_640_480_30.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += tacomasmallcameramovement_640_480_30.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += thaloundeskmtg_640_480_30.yuv
+endif # CONFIG_ENCODE_PERF_TESTS
diff --git a/source/libvpx/test/test-data.sha1 b/source/libvpx/test/test-data.sha1
index e6114ab..69d1d2f 100644
--- a/source/libvpx/test/test-data.sha1
+++ b/source/libvpx/test/test-data.sha1
@@ -17,12 +17,15 @@
a432f96ff0a787268e2f94a8092ab161a18d1b06 park_joy_90p_10_420.y4m
0b194cc312c3a2e84d156a221b0a5eb615dfddc5 park_joy_90p_10_422.y4m
ff0e0a21dc2adc95b8c1b37902713700655ced17 park_joy_90p_10_444.y4m
+c934da6fb8cc54ee2a8c17c54cf6076dac37ead0 park_joy_90p_10_440.yuv
614c32ae1eca391e867c70d19974f0d62664dd99 park_joy_90p_12_420.y4m
c92825f1ea25c5c37855083a69faac6ac4641a9e park_joy_90p_12_422.y4m
b592189b885b6cc85db55cc98512a197d73d3b34 park_joy_90p_12_444.y4m
+82c1bfcca368c2f22bad7d693d690d5499ecdd11 park_joy_90p_12_440.yuv
4e0eb61e76f0684188d9bc9f3ce61f6b6b77bb2c park_joy_90p_8_420.y4m
7a193ff7dfeb96ba5f82b2afd7afa9e1fe83d947 park_joy_90p_8_422.y4m
bdb7856e6bc93599bdda05c2e773a9f22b6c6d03 park_joy_90p_8_444.y4m
+81e1f3843748438b8f2e71db484eb22daf72e939 park_joy_90p_8_440.yuv
b1f1c3ec79114b9a0651af24ce634afb44a9a419 rush_hour_444.y4m
5184c46ddca8b1fadd16742e8500115bc8f749da vp80-00-comprehensive-001.ivf
65bf1bbbced81b97bd030f376d1b7f61a224793f vp80-00-comprehensive-002.ivf
@@ -691,6 +694,8 @@
368dccdde5288c13c25695d2eacdc7402cadf613 vp90-2-19-skip.webm.md5
ffe460282df2b0e7d4603c2158653ad96f574b02 vp90-2-19-skip-01.webm
bd21bc9eda4a4a36b221d71ede3a139fc3c7bd85 vp90-2-19-skip-01.webm.md5
+178f5bd239e38cc1cc2657a7a5e1a9f52ad2d3fe vp90-2-19-skip-02.webm
+9020d5e260bd7df08e2b3d4b86f8623cee3daea2 vp90-2-19-skip-02.webm.md5
b03c408cf23158638da18dbc3323b99a1635c68a invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf
0a3884edb3fd8f9d9b500223e650f7de257b67d8 invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf.res
5e67e24e7f53fd189e565513cef8519b1bd6c712 invalid-vp90-2-05-resize.ivf.s59293_r01-05_b6-.ivf
diff --git a/source/libvpx/test/test.mk b/source/libvpx/test/test.mk
index abf815c..30c13a1 100644
--- a/source/libvpx/test/test.mk
+++ b/source/libvpx/test/test.mk
@@ -23,6 +23,7 @@
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += error_resilience_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += i420_video_source.h
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += y4m_video_source.h
+LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += yuv_video_source.h
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += altref_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += config_test.cc
@@ -30,6 +31,7 @@
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += keyframe_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += external_frame_buffer_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += invalid_file_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += user_priv_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += active_map_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += borders_test.cc
@@ -37,6 +39,7 @@
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += frame_size_tests.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += resize_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_lossless_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_end_to_end_test.cc
LIBVPX_TEST_SRCS-yes += decode_test_driver.cc
LIBVPX_TEST_SRCS-yes += decode_test_driver.h
@@ -60,7 +63,6 @@
endif
LIBVPX_TEST_SRCS-$(CONFIG_DECODERS) += decode_api_test.cc
-LIBVPX_TEST_SRCS-$(CONFIG_DECODERS) += invalid_file_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_DECODERS) += test_vector_test.cc
# Currently we only support decoder perf tests for vp9. Also they read from WebM
@@ -103,6 +105,10 @@
LIBVPX_TEST_SRCS-yes += sixtap_predict_test.cc
LIBVPX_TEST_SRCS-yes += vpx_scale_test.cc
+ifeq ($(CONFIG_VP8_ENCODER)$(CONFIG_TEMPORAL_DENOISING),yesyes)
+LIBVPX_TEST_SRCS-$(HAVE_SSE2) += vp8_denoiser_sse2_test.cc
+endif
+
endif # VP8
## VP9
@@ -128,755 +134,22 @@
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += fdct8x8_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += variance_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_subtract_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += lpf_8_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_avg_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9) += vp9_intrapred_test.cc
ifeq ($(CONFIG_VP9_ENCODER),yes)
LIBVPX_TEST_SRCS-$(CONFIG_SPATIAL_SVC) += svc_test.cc
endif
+ifeq ($(CONFIG_VP9_ENCODER)$(CONFIG_VP9_TEMPORAL_DENOISING),yesyes)
+LIBVPX_TEST_SRCS-$(HAVE_SSE2) += vp9_denoiser_sse2_test.cc
+endif
+
endif # VP9
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += sad_test.cc
endif # CONFIG_SHARED
-
-##
-## TEST DATA
-##
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += hantro_collage_w352h288.yuv
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += hantro_odd.yuv
-
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_420.y4m
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_422.y4m
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_444.y4m
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_420.y4m
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_422.y4m
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_444.y4m
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_420.y4m
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_422.y4m
-LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_444.y4m
-
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += rush_hour_444.y4m
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += screendata.y4m
-
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-001.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-001.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-002.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-002.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-003.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-003.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-004.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-004.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-005.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-005.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-006.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-006.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-007.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-007.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-008.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-008.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-009.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-009.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-010.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-010.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-011.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-011.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-012.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-012.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-013.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-013.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-014.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-014.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-015.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-015.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-016.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-016.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-017.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-017.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-018.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-018.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1400.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1400.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1411.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1411.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1416.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1416.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1417.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1417.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1402.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1402.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1412.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1412.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1418.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1418.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1424.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1424.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-01.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-01.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-02.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-02.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-03.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-03.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-04.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-04.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1401.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1401.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1403.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1403.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1407.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1407.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1408.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1408.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1409.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1409.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1410.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1410.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1413.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1413.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1414.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1414.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1415.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1415.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1425.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1425.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1426.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1426.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1427.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1427.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1432.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1432.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1435.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1435.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1436.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1436.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1437.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1437.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1441.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1441.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1442.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1442.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1404.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1404.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1405.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1405.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1406.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1406.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1428.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1428.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1429.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1429.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1430.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1430.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1431.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1431.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1433.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1433.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1434.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1434.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1438.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1438.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1439.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1439.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1440.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1440.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1443.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1443.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-06-smallsize.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-06-smallsize.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-00.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-00.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-01.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-01.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-02.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-02.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-03.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-03.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-04.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-04.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-05.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-05.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-06.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-06.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-07.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-07.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-09.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-09.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-11.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-11.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-12.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-12.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-13.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-13.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-14.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-14.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-15.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-15.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-17.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-17.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-19.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-19.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-20.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-20.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-21.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-21.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-22.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-22.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-23.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-23.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-24.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-24.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-25.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-25.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-26.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-26.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-27.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-27.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-28.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-28.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-29.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-29.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-30.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-30.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-31.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-31.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-33.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-33.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-35.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-35.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-36.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-36.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-37.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-37.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-38.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-38.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-39.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-39.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-40.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-40.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-41.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-41.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-42.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-42.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-43.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-43.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-44.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-44.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-45.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-45.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-46.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-46.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-47.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-47.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-48.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-48.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-49.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-49.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-50.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-50.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-51.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-51.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-52.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-52.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-53.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-53.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-54.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-54.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-55.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-55.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-56.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-56.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-57.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-57.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-58.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-58.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-59.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-59.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-60.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-60.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-61.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-61.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-62.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-62.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-63.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-00-quantizer-63.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-1.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-1.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-2.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-3.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-3.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-4.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-4.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-5.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-5.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-6.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-6.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-7.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-01-sharpness-7.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x64.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x64.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x66.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-08x66.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x64.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x64.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x66.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-10x66.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x64.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x64.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x66.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-16x66.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x64.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x64.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x66.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-18x66.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x64.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x64.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x66.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-32x66.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x64.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x64.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x66.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-34x66.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x64.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x64.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x66.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-64x66.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x08.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x08.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x10.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x10.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x18.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x18.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x32.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x32.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x34.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x34.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x64.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x64.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x66.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-66x66.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-lf-1920x1080.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-02-size-lf-1920x1080.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-deltaq.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-deltaq.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x196.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x196.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x198.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x198.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x200.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x200.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x202.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x202.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x208.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x208.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x210.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x210.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x224.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x224.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x226.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-196x226.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x196.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x196.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x198.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x198.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x200.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x200.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x202.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x202.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x208.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x208.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x210.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x210.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x224.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x224.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x226.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-198x226.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x196.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x196.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x198.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x198.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x200.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x200.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x202.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x202.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x208.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x208.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x210.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x210.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x224.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x224.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x226.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-200x226.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x196.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x196.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x198.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x198.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x200.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x200.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x202.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x202.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x208.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x208.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x210.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x210.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x224.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x224.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x226.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-202x226.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x196.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x196.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x198.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x198.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x200.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x200.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x202.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x202.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x208.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x208.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x210.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x210.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x224.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x224.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x226.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-208x226.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x196.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x196.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x198.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x198.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x200.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x200.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x202.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x202.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x208.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x208.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x210.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x210.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x224.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x224.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x226.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-210x226.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x196.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x196.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x198.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x198.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x200.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x200.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x202.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x202.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x208.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x208.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x210.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x210.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x224.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x224.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x226.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-224x226.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x196.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x196.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x198.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x198.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x200.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x200.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x202.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x202.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x208.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x208.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x210.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x210.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x224.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x224.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x226.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-03-size-226x226.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-05-resize.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-05-resize.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-06-bilinear.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-06-bilinear.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-07-frame_parallel.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-07-frame_parallel.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile-4x1.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile-4x1.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile-4x4.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile-4x4.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x2.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x2_frame_parallel.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x2_frame_parallel.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x4.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x4.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x4_frame_parallel.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x4_frame_parallel.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x8.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x8.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x8_frame_parallel.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-08-tile_1x8_frame_parallel.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-aq2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-aq2.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-lf_deltas.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-lf_deltas.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-subpixel-00.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-09-subpixel-00.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-10-show-existing-frame.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-10-show-existing-frame.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-10-show-existing-frame2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-10-show-existing-frame2.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-351x287.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-351x287.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-351x288.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-351x288.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-352x287.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-11-size-352x287.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_1.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_1.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_2.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_2.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_3.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-12-droppable_3.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-13-largescaling.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-13-largescaling.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-2.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-4.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-4.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-8.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-8.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-1.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-1.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-1.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-1.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-1.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-1.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-1.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-1.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-4.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-4.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-8.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-8.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-2-16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-2.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-2.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-2.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-8.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-8.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-4-16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-4.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-4.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-4.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-4.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-8-16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-8.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-8.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-2-4-8-16.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-1-2-4-8-16.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-8-4-2-1.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-14-resize-fp-tiles-16-8-4-2-1.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-15-segkey.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-15-segkey.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-15-segkey_adpq.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-15-segkey_adpq.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-16-intra-only.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-16-intra-only.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-17-show-existing-frame.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-17-show-existing-frame.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-18-resize.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-18-resize.ivf.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-01.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-19-skip-01.webm.md5
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp91-2-04-yuv444.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp91-2-04-yuv444.webm.md5
-
-# Invalid files for testing libvpx error checking.
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-01-v2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-01-v2.webm.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-02-v2.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-02-v2.webm.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-03-v3.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-03-v3.webm.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-00.webm.ivf.s5861_r01-05_b6-.v2.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-00.webm.ivf.s5861_r01-05_b6-.v2.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-z.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-z.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-05-resize.ivf.s59293_r01-05_b6-.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-05-resize.ivf.s59293_r01-05_b6-.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-08-tile_1x2_frame_parallel.webm.ivf.s47039_r01-05_b6-.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-08-tile_1x2_frame_parallel.webm.ivf.s47039_r01-05_b6-.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-08-tile_1x4_frame_parallel_all_key.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-08-tile_1x4_frame_parallel_all_key.webm.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-aq2.webm.ivf.s3984_r01-05_b6-.v2.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-aq2.webm.ivf.s3984_r01-05_b6-.v2.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s19552_r01-05_b6-.v2.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s19552_r01-05_b6-.v2.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s20492_r01-05_b6-.v2.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s20492_r01-05_b6-.v2.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s73804_r01-05_b6-.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s73804_r01-05_b6-.ivf.res
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp91-2-mixedrefcsp-444to420.ivf
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp91-2-mixedrefcsp-444to420.ivf.res
-
-ifeq ($(CONFIG_DECODE_PERF_TESTS),yes)
-# BBB VP9 streams
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-bbb_426x240_tile_1x1_180kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-bbb_640x360_tile_1x2_337kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-bbb_854x480_tile_1x2_651kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-bbb_1280x720_tile_1x4_1310kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-bbb_1920x1080_tile_1x1_2581kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-bbb_1920x1080_tile_1x4_2586kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-bbb_1920x1080_tile_1x4_fpm_2304kbps.webm
-#Sintel VP9 streams
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-sintel_426x182_tile_1x1_171kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-sintel_640x272_tile_1x2_318kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-sintel_854x364_tile_1x2_621kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-sintel_1280x546_tile_1x4_1257kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-sintel_1920x818_tile_1x4_fpm_2279kbps.webm
-# TOS VP9 streams
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-tos_426x178_tile_1x1_181kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-tos_640x266_tile_1x2_336kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-tos_854x356_tile_1x2_656kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-tos_854x356_tile_1x2_fpm_546kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-tos_1280x534_tile_1x4_1306kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-tos_1280x534_tile_1x4_fpm_952kbps.webm
-LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += \
- vp90-2-tos_1920x800_tile_1x4_fpm_2335kbps.webm
-endif # CONFIG_DECODE_PERF_TESTS
-
-ifeq ($(CONFIG_ENCODE_PERF_TESTS),yes)
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += desktop_640_360_30.yuv
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += kirland_640_480_30.yuv
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += macmarcomoving_640_480_30.yuv
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += macmarcostationary_640_480_30.yuv
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += niklas_640_480_30.yuv
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += tacomanarrows_640_480_30.yuv
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += tacomasmallcameramovement_640_480_30.yuv
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += thaloundeskmtg_640_480_30.yuv
-LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += niklas_1280_720_30.yuv
-endif # CONFIG_ENCODE_PERF_TESTS
+include $(SRC_PATH_BARE)/test/test-data.mk
diff --git a/source/libvpx/test/test_vectors.cc b/source/libvpx/test/test_vectors.cc
index cccebf8..7efa8c0 100644
--- a/source/libvpx/test/test_vectors.cc
+++ b/source/libvpx/test/test_vectors.cc
@@ -182,7 +182,8 @@
"vp90-2-15-segkey.webm", "vp90-2-15-segkey_adpq.webm",
"vp90-2-16-intra-only.webm", "vp90-2-17-show-existing-frame.webm",
"vp90-2-18-resize.ivf", "vp90-2-19-skip.webm",
- "vp90-2-19-skip-01.webm", "vp91-2-04-yuv444.webm",
+ "vp90-2-19-skip-01.webm", "vp90-2-19-skip-02.webm",
+ "vp91-2-04-yuv444.webm",
};
const int kNumVP9TestVectors = NELEMENTS(kVP9TestVectors);
#endif // CONFIG_VP9_DECODER
diff --git a/source/libvpx/test/vp8_denoiser_sse2_test.cc b/source/libvpx/test/vp8_denoiser_sse2_test.cc
new file mode 100644
index 0000000..d4abdad
--- /dev/null
+++ b/source/libvpx/test/vp8_denoiser_sse2_test.cc
@@ -0,0 +1,116 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 <math.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+
+#include "vp8/encoder/denoising.h"
+#include "vp8/common/reconinter.h"
+#include "vpx/vpx_integer.h"
+#include "vpx_mem/vpx_mem.h"
+
+using libvpx_test::ACMRandom;
+
+namespace {
+
+const int kNumPixels = 16 * 16;
+class VP8DenoiserTest : public ::testing::TestWithParam<int> {
+ public:
+ virtual ~VP8DenoiserTest() {}
+
+ virtual void SetUp() {
+ increase_denoising_ = GetParam();
+ }
+
+ virtual void TearDown() { libvpx_test::ClearSystemState(); }
+
+ protected:
+ int increase_denoising_;
+};
+
+TEST_P(VP8DenoiserTest, BitexactCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = 4000;
+ const int stride = 16;
+
+ // Allocate the space for input and output,
+ // where sig_block_c/_sse2 is the block to be denoised,
+ // mc_avg_block is the denoised reference block,
+ // avg_block_c is the denoised result from C code,
+ // avg_block_sse2 is the denoised result from SSE2 code.
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, sig_block_c, kNumPixels);
+ // Since in VP8 denoiser, the source signal will be changed,
+ // we need another copy of the source signal as the input of sse2 code.
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, sig_block_sse2, kNumPixels);
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, mc_avg_block, kNumPixels);
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, avg_block_c, kNumPixels);
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, avg_block_sse2, kNumPixels);
+
+ for (int i = 0; i < count_test_block; ++i) {
+ // Generate random motion magnitude, 20% of which exceed the threshold.
+ const int motion_magnitude_ran =
+ rnd.Rand8() % static_cast<int>(MOTION_MAGNITUDE_THRESHOLD * 1.2);
+
+ // Initialize a test block with random number in range [0, 255].
+ for (int j = 0; j < kNumPixels; ++j) {
+ int temp = 0;
+ sig_block_sse2[j] = sig_block_c[j] = rnd.Rand8();
+ // The pixels in mc_avg_block are generated by adding a random
+ // number in range [-19, 19] to corresponding pixels in sig_block.
+ temp = sig_block_c[j] + (rnd.Rand8() % 2 == 0 ? -1 : 1) *
+ (rnd.Rand8() % 20);
+ // Clip.
+ mc_avg_block[j] = (temp < 0) ? 0 : ((temp > 255) ? 255 : temp);
+ }
+
+ // Test denosiser on Y component.
+ ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_c(
+ mc_avg_block, stride, avg_block_c, stride, sig_block_c, stride,
+ motion_magnitude_ran, increase_denoising_));
+
+ ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_sse2(
+ mc_avg_block, stride, avg_block_sse2, stride, sig_block_sse2, stride,
+ motion_magnitude_ran, increase_denoising_));
+
+ // Check bitexactness.
+ for (int h = 0; h < 16; ++h) {
+ for (int w = 0; w < 16; ++w) {
+ EXPECT_EQ(avg_block_c[h * stride + w], avg_block_sse2[h * stride + w]);
+ }
+ }
+
+ // Test denoiser on UV component.
+ ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_uv_c(
+ mc_avg_block, stride, avg_block_c, stride, sig_block_c, stride,
+ motion_magnitude_ran, increase_denoising_));
+
+ ASM_REGISTER_STATE_CHECK(vp8_denoiser_filter_uv_sse2(
+ mc_avg_block, stride, avg_block_sse2, stride, sig_block_sse2, stride,
+ motion_magnitude_ran, increase_denoising_));
+
+ // Check bitexactness.
+ for (int h = 0; h < 16; ++h) {
+ for (int w = 0; w < 16; ++w) {
+ EXPECT_EQ(avg_block_c[h * stride + w], avg_block_sse2[h * stride + w]);
+ }
+ }
+ }
+}
+
+// Test for all block size.
+INSTANTIATE_TEST_CASE_P(SSE2, VP8DenoiserTest, ::testing::Values(0, 1));
+} // namespace
diff --git a/source/libvpx/test/vp9_avg_test.cc b/source/libvpx/test/vp9_avg_test.cc
new file mode 100644
index 0000000..c2e472b
--- /dev/null
+++ b/source/libvpx/test/vp9_avg_test.cc
@@ -0,0 +1,150 @@
+/*
+ * Copyright (c) 2012 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 <string.h>
+#include <limits.h>
+#include <stdio.h>
+
+#include "./vpx_config.h"
+#if CONFIG_VP9_ENCODER
+#include "./vp9_rtcd.h"
+#endif
+#include "vpx_mem/vpx_mem.h"
+
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+using libvpx_test::ACMRandom;
+
+namespace {
+class AverageTestBase : public ::testing::Test {
+ public:
+ AverageTestBase(int width, int height) : width_(width), height_(height) {}
+
+ static void SetUpTestCase() {
+ source_data_ = reinterpret_cast<uint8_t*>(
+ vpx_memalign(kDataAlignment, kDataBlockSize));
+ }
+
+ static void TearDownTestCase() {
+ vpx_free(source_data_);
+ source_data_ = NULL;
+ }
+
+ virtual void TearDown() {
+ libvpx_test::ClearSystemState();
+ }
+
+ protected:
+ // Handle blocks up to 4 blocks 64x64 with stride up to 128
+ static const int kDataAlignment = 16;
+ static const int kDataBlockSize = 64 * 128;
+
+ virtual void SetUp() {
+ source_stride_ = (width_ + 31) & ~31;
+ rnd_.Reset(ACMRandom::DeterministicSeed());
+ }
+
+ // Sum Pixels
+ unsigned int ReferenceAverage(const uint8_t* source, int pitch ) {
+ unsigned int average = 0;
+ for (int h = 0; h < 8; ++h)
+ for (int w = 0; w < 8; ++w)
+ average += source[h * source_stride_ + w];
+ return ((average + 32) >> 6);
+ }
+
+ void FillConstant(uint8_t fill_constant) {
+ for (int i = 0; i < width_ * height_; ++i) {
+ source_data_[i] = fill_constant;
+ }
+ }
+
+ void FillRandom() {
+ for (int i = 0; i < width_ * height_; ++i) {
+ source_data_[i] = rnd_.Rand8();
+ }
+ }
+
+ int width_, height_;
+ static uint8_t* source_data_;
+ int source_stride_;
+
+ ACMRandom rnd_;
+};
+typedef unsigned int (*AverageFunction)(const uint8_t* s, int pitch);
+
+typedef std::tr1::tuple<int, int, int, AverageFunction> AvgFunc;
+
+class AverageTest
+ : public AverageTestBase,
+ public ::testing::WithParamInterface<AvgFunc>{
+ public:
+ AverageTest() : AverageTestBase(GET_PARAM(0), GET_PARAM(1)) {}
+
+ protected:
+ void CheckAverages() {
+ unsigned int expected = ReferenceAverage(source_data_+ GET_PARAM(2),
+ source_stride_);
+
+ ASM_REGISTER_STATE_CHECK(GET_PARAM(3)(source_data_+ GET_PARAM(2),
+ source_stride_));
+ unsigned int actual = GET_PARAM(3)(source_data_+ GET_PARAM(2),
+ source_stride_);
+
+ EXPECT_EQ(expected, actual);
+ }
+};
+
+
+uint8_t* AverageTestBase::source_data_ = NULL;
+
+TEST_P(AverageTest, MinValue) {
+ FillConstant(0);
+ CheckAverages();
+}
+
+TEST_P(AverageTest, MaxValue) {
+ FillConstant(255);
+ CheckAverages();
+}
+
+TEST_P(AverageTest, Random) {
+ // The reference frame, but not the source frame, may be unaligned for
+ // certain types of searches.
+ for (int i = 0; i < 1000; i++) {
+ FillRandom();
+ CheckAverages();
+ }
+}
+
+using std::tr1::make_tuple;
+
+INSTANTIATE_TEST_CASE_P(
+ C, AverageTest,
+ ::testing::Values(
+ make_tuple(16, 16, 1, &vp9_avg_8x8_c)));
+
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(
+ SSE2, AverageTest,
+ ::testing::Values(
+ make_tuple(16, 16, 0, &vp9_avg_8x8_sse2),
+ make_tuple(16, 16, 5, &vp9_avg_8x8_sse2),
+ make_tuple(32, 32, 15, &vp9_avg_8x8_sse2)));
+
+#endif
+
+} // namespace
diff --git a/source/libvpx/test/vp9_denoiser_sse2_test.cc b/source/libvpx/test/vp9_denoiser_sse2_test.cc
new file mode 100644
index 0000000..0ecba07
--- /dev/null
+++ b/source/libvpx/test/vp9_denoiser_sse2_test.cc
@@ -0,0 +1,101 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 <math.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+
+#include "vpx_scale/yv12config.h"
+#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_reconinter.h"
+#include "vp9/encoder/vp9_context_tree.h"
+#include "vp9/encoder/vp9_denoiser.h"
+
+using libvpx_test::ACMRandom;
+
+namespace {
+
+const int kNumPixels = 64 * 64;
+class VP9DenoiserTest : public ::testing::TestWithParam<BLOCK_SIZE> {
+ public:
+ virtual ~VP9DenoiserTest() {}
+
+ virtual void SetUp() {
+ bs_ = GetParam();
+ }
+
+ virtual void TearDown() { libvpx_test::ClearSystemState(); }
+
+ protected:
+ BLOCK_SIZE bs_;
+};
+
+TEST_P(VP9DenoiserTest, BitexactCheck) {
+ ACMRandom rnd(ACMRandom::DeterministicSeed());
+ const int count_test_block = 4000;
+
+ // Allocate the space for input and output,
+ // where sig_block is the block to be denoised,
+ // mc_avg_block is the denoised reference block,
+ // avg_block_c is the denoised result from C code,
+ // avg_block_sse2 is the denoised result from SSE2 code.
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, sig_block, kNumPixels);
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, mc_avg_block, kNumPixels);
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, avg_block_c, kNumPixels);
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, avg_block_sse2, kNumPixels);
+
+ for (int i = 0; i < count_test_block; ++i) {
+ // Generate random motion magnitude, 20% of which exceed the threshold.
+ const int motion_magnitude_random =
+ rnd.Rand8() % static_cast<int>(MOTION_MAGNITUDE_THRESHOLD * 1.2);
+
+ // Initialize a test block with random number in range [0, 255].
+ for (int j = 0; j < kNumPixels; ++j) {
+ int temp = 0;
+ sig_block[j] = rnd.Rand8();
+ // The pixels in mc_avg_block are generated by adding a random
+ // number in range [-19, 19] to corresponding pixels in sig_block.
+ temp = sig_block[j] + ((rnd.Rand8() % 2 == 0) ? -1 : 1) *
+ (rnd.Rand8() % 20);
+ // Clip.
+ mc_avg_block[j] = (temp < 0) ? 0 : ((temp > 255) ? 255 : temp);
+ }
+
+ ASM_REGISTER_STATE_CHECK(vp9_denoiser_filter_c(
+ sig_block, 64, mc_avg_block, 64, avg_block_c,
+ 64, 0, bs_, motion_magnitude_random));
+
+ ASM_REGISTER_STATE_CHECK(vp9_denoiser_filter_sse2(
+ sig_block, 64, mc_avg_block, 64, avg_block_sse2,
+ 64, 0, bs_, motion_magnitude_random));
+
+ // Test bitexactness.
+ for (int h = 0; h < (4 << b_height_log2_lookup[bs_]); ++h) {
+ for (int w = 0; w < (4 << b_width_log2_lookup[bs_]); ++w) {
+ EXPECT_EQ(avg_block_c[h * 64 + w], avg_block_sse2[h * 64 + w]);
+ }
+ }
+ }
+}
+
+// Test for all block size.
+INSTANTIATE_TEST_CASE_P(
+ SSE2, VP9DenoiserTest,
+ ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4, BLOCK_8X8,
+ BLOCK_8X16, BLOCK_16X8, BLOCK_16X16, BLOCK_16X32,
+ BLOCK_32X16, BLOCK_32X32, BLOCK_32X64, BLOCK_64X32,
+ BLOCK_64X64));
+} // namespace
diff --git a/source/libvpx/test/vp9_end_to_end_test.cc b/source/libvpx/test/vp9_end_to_end_test.cc
new file mode 100644
index 0000000..a8f6793
--- /dev/null
+++ b/source/libvpx/test/vp9_end_to_end_test.cc
@@ -0,0 +1,155 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 "test/codec_factory.h"
+#include "test/encode_test_driver.h"
+#include "test/y4m_video_source.h"
+#include "test/yuv_video_source.h"
+#include "test/util.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+namespace {
+
+const unsigned int kWidth = 160;
+const unsigned int kHeight = 90;
+const unsigned int kFramerate = 50;
+const unsigned int kFrames = 10;
+const int kBitrate = 500;
+const int kCpuUsed = 2;
+const double psnr_threshold = 35.0;
+
+typedef struct {
+ const char *filename;
+ unsigned int input_bit_depth;
+ vpx_img_fmt fmt;
+ vpx_bit_depth_t bit_depth;
+ unsigned int profile;
+} TestVideoParam;
+
+const TestVideoParam TestVectors[] = {
+ {"park_joy_90p_8_420.y4m", 8, VPX_IMG_FMT_I420, VPX_BITS_8, 0},
+ {"park_joy_90p_8_422.y4m", 8, VPX_IMG_FMT_I422, VPX_BITS_8, 1},
+ {"park_joy_90p_8_444.y4m", 8, VPX_IMG_FMT_I444, VPX_BITS_8, 1},
+ {"park_joy_90p_8_440.yuv", 8, VPX_IMG_FMT_I440, VPX_BITS_8, 1},
+#if CONFIG_VP9_HIGHBITDEPTH
+ {"park_joy_90p_10_420.y4m", 10, VPX_IMG_FMT_I42016, VPX_BITS_10, 2},
+ {"park_joy_90p_10_422.y4m", 10, VPX_IMG_FMT_I42216, VPX_BITS_10, 3},
+ {"park_joy_90p_10_444.y4m", 10, VPX_IMG_FMT_I44416, VPX_BITS_10, 3},
+ {"park_joy_90p_10_440.yuv", 10, VPX_IMG_FMT_I44016, VPX_BITS_10, 3},
+ {"park_joy_90p_12_420.y4m", 12, VPX_IMG_FMT_I42016, VPX_BITS_12, 2},
+ {"park_joy_90p_12_422.y4m", 12, VPX_IMG_FMT_I42216, VPX_BITS_12, 3},
+ {"park_joy_90p_12_444.y4m", 12, VPX_IMG_FMT_I44416, VPX_BITS_12, 3},
+ {"park_joy_90p_12_440.yuv", 12, VPX_IMG_FMT_I44016, VPX_BITS_12, 3},
+#endif // CONFIG_VP9_HIGHBITDEPTH
+};
+
+int is_extension_y4m(const char *filename) {
+ const char *dot = strrchr(filename, '.');
+ if (!dot || dot == filename)
+ return 0;
+ else
+ return !strcmp(dot, ".y4m");
+}
+
+class EndToEndTestLarge
+ : public ::libvpx_test::EncoderTest,
+ public ::libvpx_test::CodecTestWith2Params<libvpx_test::TestMode, \
+ TestVideoParam> {
+ protected:
+ EndToEndTestLarge()
+ : EncoderTest(GET_PARAM(0)),
+ psnr_(0.0),
+ nframes_(0),
+ encoding_mode_(GET_PARAM(1)) {
+ }
+
+ virtual ~EndToEndTestLarge() {}
+
+ virtual void SetUp() {
+ InitializeConfig();
+ SetMode(encoding_mode_);
+ if (encoding_mode_ != ::libvpx_test::kRealTime) {
+ cfg_.g_lag_in_frames = 5;
+ cfg_.rc_end_usage = VPX_VBR;
+ } else {
+ cfg_.g_lag_in_frames = 0;
+ cfg_.rc_end_usage = VPX_CBR;
+ }
+ test_video_param_ = GET_PARAM(2);
+ }
+
+ virtual void BeginPassHook(unsigned int) {
+ psnr_ = 0.0;
+ nframes_ = 0;
+ }
+
+ virtual void PSNRPktHook(const vpx_codec_cx_pkt_t *pkt) {
+ psnr_ += pkt->data.psnr.psnr[0];
+ nframes_++;
+ }
+
+ virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
+ ::libvpx_test::Encoder *encoder) {
+ if (video->frame() == 1) {
+ encoder->Control(VP8E_SET_CPUUSED, kCpuUsed);
+ if (encoding_mode_ != ::libvpx_test::kRealTime) {
+ encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 1);
+ encoder->Control(VP8E_SET_ARNR_MAXFRAMES, 7);
+ encoder->Control(VP8E_SET_ARNR_STRENGTH, 5);
+ encoder->Control(VP8E_SET_ARNR_TYPE, 3);
+ }
+ }
+ }
+
+ double GetAveragePsnr() const {
+ if (nframes_)
+ return psnr_ / nframes_;
+ return 0.0;
+ }
+
+ TestVideoParam test_video_param_;
+
+ private:
+ double psnr_;
+ unsigned int nframes_;
+ libvpx_test::TestMode encoding_mode_;
+};
+
+TEST_P(EndToEndTestLarge, EndtoEndPSNRTest) {
+ cfg_.rc_target_bitrate = kBitrate;
+ cfg_.g_error_resilient = 0;
+ cfg_.g_profile = test_video_param_.profile;
+ cfg_.g_input_bit_depth = test_video_param_.input_bit_depth;
+ cfg_.g_bit_depth = test_video_param_.bit_depth;
+ init_flags_ = VPX_CODEC_USE_PSNR;
+
+ libvpx_test::VideoSource *video;
+ if (is_extension_y4m(test_video_param_.filename)) {
+ video = new libvpx_test::Y4mVideoSource(test_video_param_.filename,
+ 0, kFrames);
+ } else {
+ video = new libvpx_test::YUVVideoSource(test_video_param_.filename,
+ test_video_param_.fmt,
+ kWidth, kHeight,
+ kFramerate, 1, 0, kFrames);
+ }
+
+ ASSERT_NO_FATAL_FAILURE(RunLoop(video));
+ const double psnr = GetAveragePsnr();
+ EXPECT_GT(psnr, psnr_threshold);
+ delete(video);
+}
+
+VP9_INSTANTIATE_TEST_CASE(
+ EndToEndTestLarge,
+ ::testing::Values(::libvpx_test::kTwoPassGood, ::libvpx_test::kOnePassGood),
+ ::testing::ValuesIn(TestVectors));
+
+} // namespace
diff --git a/source/libvpx/test/vp9_intrapred_test.cc b/source/libvpx/test/vp9_intrapred_test.cc
index 7d08d9e..694db1b 100644
--- a/source/libvpx/test/vp9_intrapred_test.cc
+++ b/source/libvpx/test/vp9_intrapred_test.cc
@@ -134,150 +134,164 @@
#if ARCH_X86_64
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_8, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_high_dc_predictor_32x32_sse2,
- &vp9_high_dc_predictor_32x32_c, 32, 8),
- make_tuple(&vp9_high_tm_predictor_16x16_sse2,
- &vp9_high_tm_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_high_tm_predictor_32x32_sse2,
- &vp9_high_tm_predictor_32x32_c, 32, 8),
- make_tuple(&vp9_high_dc_predictor_4x4_sse,
- &vp9_high_dc_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_high_dc_predictor_8x8_sse2,
- &vp9_high_dc_predictor_8x8_c, 8, 8),
- make_tuple(&vp9_high_dc_predictor_16x16_sse2,
- &vp9_high_dc_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_high_v_predictor_4x4_sse,
- &vp9_high_v_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_high_v_predictor_8x8_sse2,
- &vp9_high_v_predictor_8x8_c, 8, 8),
- make_tuple(&vp9_high_v_predictor_16x16_sse2,
- &vp9_high_v_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_high_v_predictor_32x32_sse2,
- &vp9_high_v_predictor_32x32_c, 32, 8),
- make_tuple(&vp9_high_tm_predictor_4x4_sse,
- &vp9_high_tm_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_high_tm_predictor_8x8_sse2,
- &vp9_high_tm_predictor_8x8_c, 8, 8)));
+ make_tuple(&vp9_highbd_dc_predictor_32x32_sse2,
+ &vp9_highbd_dc_predictor_32x32_c, 32, 8),
+ make_tuple(&vp9_highbd_tm_predictor_16x16_sse2,
+ &vp9_highbd_tm_predictor_16x16_c, 16, 8),
+ make_tuple(&vp9_highbd_tm_predictor_32x32_sse2,
+ &vp9_highbd_tm_predictor_32x32_c, 32, 8),
+ make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
+ &vp9_highbd_dc_predictor_4x4_c, 4, 8),
+ make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
+ &vp9_highbd_dc_predictor_8x8_c, 8, 8),
+ make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
+ &vp9_highbd_dc_predictor_16x16_c, 16, 8),
+ make_tuple(&vp9_highbd_v_predictor_4x4_sse,
+ &vp9_highbd_v_predictor_4x4_c, 4, 8),
+ make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
+ &vp9_highbd_v_predictor_8x8_c, 8, 8),
+ make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
+ &vp9_highbd_v_predictor_16x16_c, 16, 8),
+ make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
+ &vp9_highbd_v_predictor_32x32_c, 32, 8),
+ make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
+ &vp9_highbd_tm_predictor_4x4_c, 4, 8),
+ make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
+ &vp9_highbd_tm_predictor_8x8_c, 8, 8)));
#else
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_8, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_high_dc_predictor_4x4_sse,
- &vp9_high_dc_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_high_dc_predictor_8x8_sse2,
- &vp9_high_dc_predictor_8x8_c, 8, 8),
- make_tuple(&vp9_high_dc_predictor_16x16_sse2,
- &vp9_high_dc_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_high_v_predictor_4x4_sse,
- &vp9_high_v_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_high_v_predictor_8x8_sse2,
- &vp9_high_v_predictor_8x8_c, 8, 8),
- make_tuple(&vp9_high_v_predictor_16x16_sse2,
- &vp9_high_v_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_high_v_predictor_32x32_sse2,
- &vp9_high_v_predictor_32x32_c, 32, 8),
- make_tuple(&vp9_high_tm_predictor_4x4_sse,
- &vp9_high_tm_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_high_tm_predictor_8x8_sse2,
- &vp9_high_tm_predictor_8x8_c, 8, 8)));
+ make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
+ &vp9_highbd_dc_predictor_4x4_c, 4, 8),
+ make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
+ &vp9_highbd_dc_predictor_8x8_c, 8, 8),
+ make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
+ &vp9_highbd_dc_predictor_16x16_c, 16, 8),
+ make_tuple(&vp9_highbd_v_predictor_4x4_sse,
+ &vp9_highbd_v_predictor_4x4_c, 4, 8),
+ make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
+ &vp9_highbd_v_predictor_8x8_c, 8, 8),
+ make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
+ &vp9_highbd_v_predictor_16x16_c, 16, 8),
+ make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
+ &vp9_highbd_v_predictor_32x32_c, 32, 8),
+ make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
+ &vp9_highbd_tm_predictor_4x4_c, 4, 8),
+ make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
+ &vp9_highbd_tm_predictor_8x8_c, 8, 8)));
#endif
#if ARCH_X86_64
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_10, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_high_dc_predictor_32x32_sse2,
- &vp9_high_dc_predictor_32x32_c, 32, 10),
- make_tuple(&vp9_high_tm_predictor_16x16_sse2,
- &vp9_high_tm_predictor_16x16_c, 16, 10),
- make_tuple(&vp9_high_tm_predictor_32x32_sse2,
- &vp9_high_tm_predictor_32x32_c, 32, 10),
- make_tuple(&vp9_high_dc_predictor_4x4_sse,
- &vp9_high_dc_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_high_dc_predictor_8x8_sse2,
- &vp9_high_dc_predictor_8x8_c, 8, 10),
- make_tuple(&vp9_high_dc_predictor_16x16_sse2,
- &vp9_high_dc_predictor_16x16_c, 16, 10),
- make_tuple(&vp9_high_v_predictor_4x4_sse,
- &vp9_high_v_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_high_v_predictor_8x8_sse2,
- &vp9_high_v_predictor_8x8_c, 8, 10),
- make_tuple(&vp9_high_v_predictor_16x16_sse2,
- &vp9_high_v_predictor_16x16_c, 16, 10),
- make_tuple(&vp9_high_v_predictor_32x32_sse2,
- &vp9_high_v_predictor_32x32_c, 32, 10),
- make_tuple(&vp9_high_tm_predictor_4x4_sse,
- &vp9_high_tm_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_high_tm_predictor_8x8_sse2,
- &vp9_high_tm_predictor_8x8_c, 8, 10)));
+ make_tuple(&vp9_highbd_dc_predictor_32x32_sse2,
+ &vp9_highbd_dc_predictor_32x32_c, 32,
+ 10),
+ make_tuple(&vp9_highbd_tm_predictor_16x16_sse2,
+ &vp9_highbd_tm_predictor_16x16_c, 16,
+ 10),
+ make_tuple(&vp9_highbd_tm_predictor_32x32_sse2,
+ &vp9_highbd_tm_predictor_32x32_c, 32,
+ 10),
+ make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
+ &vp9_highbd_dc_predictor_4x4_c, 4, 10),
+ make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
+ &vp9_highbd_dc_predictor_8x8_c, 8, 10),
+ make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
+ &vp9_highbd_dc_predictor_16x16_c, 16,
+ 10),
+ make_tuple(&vp9_highbd_v_predictor_4x4_sse,
+ &vp9_highbd_v_predictor_4x4_c, 4, 10),
+ make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
+ &vp9_highbd_v_predictor_8x8_c, 8, 10),
+ make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
+ &vp9_highbd_v_predictor_16x16_c, 16,
+ 10),
+ make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
+ &vp9_highbd_v_predictor_32x32_c, 32,
+ 10),
+ make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
+ &vp9_highbd_tm_predictor_4x4_c, 4, 10),
+ make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
+ &vp9_highbd_tm_predictor_8x8_c, 8, 10)));
#else
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_10, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_high_dc_predictor_4x4_sse,
- &vp9_high_dc_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_high_dc_predictor_8x8_sse2,
- &vp9_high_dc_predictor_8x8_c, 8, 10),
- make_tuple(&vp9_high_dc_predictor_16x16_sse2,
- &vp9_high_dc_predictor_16x16_c, 16, 10),
- make_tuple(&vp9_high_v_predictor_4x4_sse,
- &vp9_high_v_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_high_v_predictor_8x8_sse2,
- &vp9_high_v_predictor_8x8_c, 8, 10),
- make_tuple(&vp9_high_v_predictor_16x16_sse2,
- &vp9_high_v_predictor_16x16_c, 16, 10),
- make_tuple(&vp9_high_v_predictor_32x32_sse2,
- &vp9_high_v_predictor_32x32_c, 32, 10),
- make_tuple(&vp9_high_tm_predictor_4x4_sse,
- &vp9_high_tm_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_high_tm_predictor_8x8_sse2,
- &vp9_high_tm_predictor_8x8_c, 8, 10)));
+ make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
+ &vp9_highbd_dc_predictor_4x4_c, 4, 10),
+ make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
+ &vp9_highbd_dc_predictor_8x8_c, 8, 10),
+ make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
+ &vp9_highbd_dc_predictor_16x16_c, 16,
+ 10),
+ make_tuple(&vp9_highbd_v_predictor_4x4_sse,
+ &vp9_highbd_v_predictor_4x4_c, 4, 10),
+ make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
+ &vp9_highbd_v_predictor_8x8_c, 8, 10),
+ make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
+ &vp9_highbd_v_predictor_16x16_c, 16, 10),
+ make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
+ &vp9_highbd_v_predictor_32x32_c, 32, 10),
+ make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
+ &vp9_highbd_tm_predictor_4x4_c, 4, 10),
+ make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
+ &vp9_highbd_tm_predictor_8x8_c, 8, 10)));
#endif
#if ARCH_X86_64
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_12, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_high_dc_predictor_32x32_sse2,
- &vp9_high_dc_predictor_32x32_c, 32, 12),
- make_tuple(&vp9_high_tm_predictor_16x16_sse2,
- &vp9_high_tm_predictor_16x16_c, 16, 12),
- make_tuple(&vp9_high_tm_predictor_32x32_sse2,
- &vp9_high_tm_predictor_32x32_c, 32, 12),
- make_tuple(&vp9_high_dc_predictor_4x4_sse,
- &vp9_high_dc_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_high_dc_predictor_8x8_sse2,
- &vp9_high_dc_predictor_8x8_c, 8, 12),
- make_tuple(&vp9_high_dc_predictor_16x16_sse2,
- &vp9_high_dc_predictor_16x16_c, 16, 12),
- make_tuple(&vp9_high_v_predictor_4x4_sse,
- &vp9_high_v_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_high_v_predictor_8x8_sse2,
- &vp9_high_v_predictor_8x8_c, 8, 12),
- make_tuple(&vp9_high_v_predictor_16x16_sse2,
- &vp9_high_v_predictor_16x16_c, 16, 12),
- make_tuple(&vp9_high_v_predictor_32x32_sse2,
- &vp9_high_v_predictor_32x32_c, 32, 12),
- make_tuple(&vp9_high_tm_predictor_4x4_sse,
- &vp9_high_tm_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_high_tm_predictor_8x8_sse2,
- &vp9_high_tm_predictor_8x8_c, 8, 12)));
+ make_tuple(&vp9_highbd_dc_predictor_32x32_sse2,
+ &vp9_highbd_dc_predictor_32x32_c, 32,
+ 12),
+ make_tuple(&vp9_highbd_tm_predictor_16x16_sse2,
+ &vp9_highbd_tm_predictor_16x16_c, 16,
+ 12),
+ make_tuple(&vp9_highbd_tm_predictor_32x32_sse2,
+ &vp9_highbd_tm_predictor_32x32_c, 32,
+ 12),
+ make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
+ &vp9_highbd_dc_predictor_4x4_c, 4, 12),
+ make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
+ &vp9_highbd_dc_predictor_8x8_c, 8, 12),
+ make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
+ &vp9_highbd_dc_predictor_16x16_c, 16,
+ 12),
+ make_tuple(&vp9_highbd_v_predictor_4x4_sse,
+ &vp9_highbd_v_predictor_4x4_c, 4, 12),
+ make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
+ &vp9_highbd_v_predictor_8x8_c, 8, 12),
+ make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
+ &vp9_highbd_v_predictor_16x16_c, 16,
+ 12),
+ make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
+ &vp9_highbd_v_predictor_32x32_c, 32,
+ 12),
+ make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
+ &vp9_highbd_tm_predictor_4x4_c, 4, 12),
+ make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
+ &vp9_highbd_tm_predictor_8x8_c, 8, 12)));
#else
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_12, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_high_dc_predictor_4x4_sse,
- &vp9_high_dc_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_high_dc_predictor_8x8_sse2,
- &vp9_high_dc_predictor_8x8_c, 8, 12),
- make_tuple(&vp9_high_dc_predictor_16x16_sse2,
- &vp9_high_dc_predictor_16x16_c, 16, 12),
- make_tuple(&vp9_high_v_predictor_4x4_sse,
- &vp9_high_v_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_high_v_predictor_8x8_sse2,
- &vp9_high_v_predictor_8x8_c, 8, 12),
- make_tuple(&vp9_high_v_predictor_16x16_sse2,
- &vp9_high_v_predictor_16x16_c, 16, 12),
- make_tuple(&vp9_high_v_predictor_32x32_sse2,
- &vp9_high_v_predictor_32x32_c, 32, 12),
- make_tuple(&vp9_high_tm_predictor_4x4_sse,
- &vp9_high_tm_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_high_tm_predictor_8x8_sse2,
- &vp9_high_tm_predictor_8x8_c, 8, 12)));
+ make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
+ &vp9_highbd_dc_predictor_4x4_c, 4, 12),
+ make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
+ &vp9_highbd_dc_predictor_8x8_c, 8, 12),
+ make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
+ &vp9_highbd_dc_predictor_16x16_c, 16,
+ 12),
+ make_tuple(&vp9_highbd_v_predictor_4x4_sse,
+ &vp9_highbd_v_predictor_4x4_c, 4, 12),
+ make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
+ &vp9_highbd_v_predictor_8x8_c, 8, 12),
+ make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
+ &vp9_highbd_v_predictor_16x16_c, 16, 12),
+ make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
+ &vp9_highbd_v_predictor_32x32_c, 32, 12),
+ make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
+ &vp9_highbd_tm_predictor_4x4_c, 4, 12),
+ make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
+ &vp9_highbd_tm_predictor_8x8_c, 8, 12)));
#endif
#endif // CONFIG_VP9_HIGHBITDEPTH
#endif // HAVE_SSE2
diff --git a/source/libvpx/test/vp9_lossless_test.cc b/source/libvpx/test/vp9_lossless_test.cc
index b3b9c92..67215d3 100644
--- a/source/libvpx/test/vp9_lossless_test.cc
+++ b/source/libvpx/test/vp9_lossless_test.cc
@@ -19,17 +19,17 @@
const int kMaxPsnr = 100;
-class LosslessTestLarge : public ::libvpx_test::EncoderTest,
+class LosslessTest : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
- LosslessTestLarge()
+ LosslessTest()
: EncoderTest(GET_PARAM(0)),
psnr_(kMaxPsnr),
nframes_(0),
encoding_mode_(GET_PARAM(1)) {
}
- virtual ~LosslessTestLarge() {}
+ virtual ~LosslessTest() {}
virtual void SetUp() {
InitializeConfig();
@@ -67,7 +67,7 @@
libvpx_test::TestMode encoding_mode_;
};
-TEST_P(LosslessTestLarge, TestLossLessEncoding) {
+TEST_P(LosslessTest, TestLossLessEncoding) {
const vpx_rational timebase = { 33333333, 1000000000 };
cfg_.g_timebase = timebase;
cfg_.rc_target_bitrate = 2000;
@@ -85,7 +85,7 @@
EXPECT_GE(psnr_lossless, kMaxPsnr);
}
-TEST_P(LosslessTestLarge, TestLossLessEncoding444) {
+TEST_P(LosslessTest, TestLossLessEncoding444) {
libvpx_test::Y4mVideoSource video("rush_hour_444.y4m", 0, 10);
cfg_.g_profile = 1;
@@ -102,7 +102,7 @@
EXPECT_GE(psnr_lossless, kMaxPsnr);
}
-TEST_P(LosslessTestLarge, TestLossLessEncodingCtrl) {
+TEST_P(LosslessTest, TestLossLessEncodingCtrl) {
const vpx_rational timebase = { 33333333, 1000000000 };
cfg_.g_timebase = timebase;
cfg_.rc_target_bitrate = 2000;
@@ -121,5 +121,8 @@
EXPECT_GE(psnr_lossless, kMaxPsnr);
}
-VP9_INSTANTIATE_TEST_CASE(LosslessTestLarge, ALL_TEST_MODES);
+VP9_INSTANTIATE_TEST_CASE(LosslessTest,
+ ::testing::Values(::libvpx_test::kRealTime,
+ ::libvpx_test::kOnePassGood,
+ ::libvpx_test::kTwoPassGood));
} // namespace
diff --git a/source/libvpx/test/yuv_video_source.h b/source/libvpx/test/yuv_video_source.h
new file mode 100644
index 0000000..3c852b2
--- /dev/null
+++ b/source/libvpx/test/yuv_video_source.h
@@ -0,0 +1,151 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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.
+ */
+#ifndef TEST_YUV_VIDEO_SOURCE_H_
+#define TEST_YUV_VIDEO_SOURCE_H_
+
+#include <cstdio>
+#include <cstdlib>
+#include <string>
+
+#include "test/video_source.h"
+#include "vpx/vpx_image.h"
+
+namespace libvpx_test {
+
+// This class extends VideoSource to allow parsing of raw YUV
+// formats of various color sampling and bit-depths so that we can
+// do actual file encodes.
+class YUVVideoSource : public VideoSource {
+ public:
+ YUVVideoSource(const std::string &file_name, vpx_img_fmt format,
+ unsigned int width, unsigned int height,
+ int rate_numerator, int rate_denominator,
+ unsigned int start, int limit)
+ : file_name_(file_name),
+ input_file_(NULL),
+ img_(NULL),
+ start_(start),
+ limit_(limit),
+ frame_(0),
+ width_(0),
+ height_(0),
+ format_(VPX_IMG_FMT_NONE),
+ framerate_numerator_(rate_numerator),
+ framerate_denominator_(rate_denominator) {
+ // This initializes format_, raw_size_, width_, height_ and allocates img.
+ SetSize(width, height, format);
+ }
+
+ virtual ~YUVVideoSource() {
+ vpx_img_free(img_);
+ if (input_file_)
+ fclose(input_file_);
+ }
+
+ virtual void Begin() {
+ if (input_file_)
+ fclose(input_file_);
+ input_file_ = OpenTestDataFile(file_name_);
+ ASSERT_TRUE(input_file_ != NULL) << "Input file open failed. Filename: "
+ << file_name_;
+ if (start_)
+ fseek(input_file_, static_cast<unsigned>(raw_size_) * start_, SEEK_SET);
+
+ frame_ = start_;
+ FillFrame();
+ }
+
+ virtual void Next() {
+ ++frame_;
+ FillFrame();
+ }
+
+ virtual vpx_image_t *img() const { return (frame_ < limit_) ? img_ : NULL; }
+
+ // Models a stream where Timebase = 1/FPS, so pts == frame.
+ virtual vpx_codec_pts_t pts() const { return frame_; }
+
+ virtual unsigned long duration() const { return 1; }
+
+ virtual vpx_rational_t timebase() const {
+ const vpx_rational_t t = { framerate_denominator_, framerate_numerator_ };
+ return t;
+ }
+
+ virtual unsigned int frame() const { return frame_; }
+
+ virtual unsigned int limit() const { return limit_; }
+
+ virtual void SetSize(unsigned int width, unsigned int height,
+ vpx_img_fmt format) {
+ if (width != width_ || height != height_ || format != format_) {
+ vpx_img_free(img_);
+ img_ = vpx_img_alloc(NULL, format, width, height, 1);
+ ASSERT_TRUE(img_ != NULL);
+ width_ = width;
+ height_ = height;
+ format_ = format;
+ switch (format) {
+ case VPX_IMG_FMT_I420:
+ raw_size_ = width * height * 3 / 2;
+ break;
+ case VPX_IMG_FMT_I422:
+ raw_size_ = width * height * 2;
+ break;
+ case VPX_IMG_FMT_I440:
+ raw_size_ = width * height * 2;
+ break;
+ case VPX_IMG_FMT_I444:
+ raw_size_ = width * height * 3;
+ break;
+ case VPX_IMG_FMT_I42016:
+ raw_size_ = width * height * 3;
+ break;
+ case VPX_IMG_FMT_I42216:
+ raw_size_ = width * height * 4;
+ break;
+ case VPX_IMG_FMT_I44016:
+ raw_size_ = width * height * 4;
+ break;
+ case VPX_IMG_FMT_I44416:
+ raw_size_ = width * height * 6;
+ break;
+ default:
+ ASSERT_TRUE(0);
+ }
+ }
+ }
+
+ virtual void FillFrame() {
+ ASSERT_TRUE(input_file_ != NULL);
+ // Read a frame from input_file.
+ if (fread(img_->img_data, raw_size_, 1, input_file_) == 0) {
+ limit_ = frame_;
+ }
+ }
+
+ protected:
+ std::string file_name_;
+ FILE *input_file_;
+ vpx_image_t *img_;
+ size_t raw_size_;
+ unsigned int start_;
+ unsigned int limit_;
+ unsigned int frame_;
+ unsigned int width_;
+ unsigned int height_;
+ vpx_img_fmt format_;
+ int framerate_numerator_;
+ int framerate_denominator_;
+};
+
+} // namespace libvpx_test
+
+#endif // TEST_YUV_VIDEO_SOURCE_H_
diff --git a/source/libvpx/tools_common.c b/source/libvpx/tools_common.c
index 2ec1711..e243a91 100644
--- a/source/libvpx/tools_common.c
+++ b/source/libvpx/tools_common.c
@@ -224,7 +224,8 @@
for (plane = 0; plane < 3; ++plane) {
const unsigned char *buf = img->planes[plane];
const int stride = img->stride[plane];
- const int w = vpx_img_plane_width(img, plane);
+ const int w = vpx_img_plane_width(img, plane) *
+ ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) ? 2 : 1);
const int h = vpx_img_plane_height(img, plane);
int y;
@@ -267,3 +268,219 @@
return kMaxPSNR;
}
}
+
+// TODO(debargha): Consolidate the functions below into a separate file.
+#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
+static void highbd_img_upshift(vpx_image_t *dst, vpx_image_t *src,
+ int input_shift) {
+ // Note the offset is 1 less than half.
+ const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0;
+ int plane;
+ if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
+ dst->x_chroma_shift != src->x_chroma_shift ||
+ dst->y_chroma_shift != src->y_chroma_shift ||
+ dst->fmt != src->fmt || input_shift < 0) {
+ fatal("Unsupported image conversion");
+ }
+ switch (src->fmt) {
+ case VPX_IMG_FMT_I42016:
+ case VPX_IMG_FMT_I42216:
+ case VPX_IMG_FMT_I44416:
+ case VPX_IMG_FMT_I44016:
+ break;
+ default:
+ fatal("Unsupported image conversion");
+ break;
+ }
+ for (plane = 0; plane < 3; plane++) {
+ int w = src->d_w;
+ int h = src->d_h;
+ int x, y;
+ if (plane) {
+ w = (w + src->x_chroma_shift) >> src->x_chroma_shift;
+ h = (h + src->y_chroma_shift) >> src->y_chroma_shift;
+ }
+ for (y = 0; y < h; y++) {
+ uint16_t *p_src =
+ (uint16_t *)(src->planes[plane] + y * src->stride[plane]);
+ uint16_t *p_dst =
+ (uint16_t *)(dst->planes[plane] + y * dst->stride[plane]);
+ for (x = 0; x < w; x++)
+ *p_dst++ = (*p_src++ << input_shift) + offset;
+ }
+ }
+}
+
+static void lowbd_img_upshift(vpx_image_t *dst, vpx_image_t *src,
+ int input_shift) {
+ // Note the offset is 1 less than half.
+ const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0;
+ int plane;
+ if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
+ dst->x_chroma_shift != src->x_chroma_shift ||
+ dst->y_chroma_shift != src->y_chroma_shift ||
+ dst->fmt != src->fmt + VPX_IMG_FMT_HIGHBITDEPTH ||
+ input_shift < 0) {
+ fatal("Unsupported image conversion");
+ }
+ switch (src->fmt) {
+ case VPX_IMG_FMT_I420:
+ case VPX_IMG_FMT_I422:
+ case VPX_IMG_FMT_I444:
+ case VPX_IMG_FMT_I440:
+ break;
+ default:
+ fatal("Unsupported image conversion");
+ break;
+ }
+ for (plane = 0; plane < 3; plane++) {
+ int w = src->d_w;
+ int h = src->d_h;
+ int x, y;
+ if (plane) {
+ w = (w + src->x_chroma_shift) >> src->x_chroma_shift;
+ h = (h + src->y_chroma_shift) >> src->y_chroma_shift;
+ }
+ for (y = 0; y < h; y++) {
+ uint8_t *p_src = src->planes[plane] + y * src->stride[plane];
+ uint16_t *p_dst =
+ (uint16_t *)(dst->planes[plane] + y * dst->stride[plane]);
+ for (x = 0; x < w; x++) {
+ *p_dst++ = (*p_src++ << input_shift) + offset;
+ }
+ }
+ }
+}
+
+void vpx_img_upshift(vpx_image_t *dst, vpx_image_t *src,
+ int input_shift) {
+ if (src->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
+ highbd_img_upshift(dst, src, input_shift);
+ } else {
+ lowbd_img_upshift(dst, src, input_shift);
+ }
+}
+
+void vpx_img_truncate_16_to_8(vpx_image_t *dst, vpx_image_t *src) {
+ int plane;
+ if (dst->fmt + VPX_IMG_FMT_HIGHBITDEPTH != src->fmt ||
+ dst->d_w != src->d_w || dst->d_h != src->d_h ||
+ dst->x_chroma_shift != src->x_chroma_shift ||
+ dst->y_chroma_shift != src->y_chroma_shift) {
+ fatal("Unsupported image conversion");
+ }
+ switch (dst->fmt) {
+ case VPX_IMG_FMT_I420:
+ case VPX_IMG_FMT_I422:
+ case VPX_IMG_FMT_I444:
+ case VPX_IMG_FMT_I440:
+ break;
+ default:
+ fatal("Unsupported image conversion");
+ break;
+ }
+ for (plane = 0; plane < 3; plane++) {
+ int w = src->d_w;
+ int h = src->d_h;
+ int x, y;
+ if (plane) {
+ w = (w + src->x_chroma_shift) >> src->x_chroma_shift;
+ h = (h + src->y_chroma_shift) >> src->y_chroma_shift;
+ }
+ for (y = 0; y < h; y++) {
+ uint16_t *p_src =
+ (uint16_t *)(src->planes[plane] + y * src->stride[plane]);
+ uint8_t *p_dst = dst->planes[plane] + y * dst->stride[plane];
+ for (x = 0; x < w; x++) {
+ *p_dst++ = *p_src++;
+ }
+ }
+ }
+}
+
+static void highbd_img_downshift(vpx_image_t *dst, vpx_image_t *src,
+ int down_shift) {
+ int plane;
+ if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
+ dst->x_chroma_shift != src->x_chroma_shift ||
+ dst->y_chroma_shift != src->y_chroma_shift ||
+ dst->fmt != src->fmt || down_shift < 0) {
+ fatal("Unsupported image conversion");
+ }
+ switch (src->fmt) {
+ case VPX_IMG_FMT_I42016:
+ case VPX_IMG_FMT_I42216:
+ case VPX_IMG_FMT_I44416:
+ case VPX_IMG_FMT_I44016:
+ break;
+ default:
+ fatal("Unsupported image conversion");
+ break;
+ }
+ for (plane = 0; plane < 3; plane++) {
+ int w = src->d_w;
+ int h = src->d_h;
+ int x, y;
+ if (plane) {
+ w = (w + src->x_chroma_shift) >> src->x_chroma_shift;
+ h = (h + src->y_chroma_shift) >> src->y_chroma_shift;
+ }
+ for (y = 0; y < h; y++) {
+ uint16_t *p_src =
+ (uint16_t *)(src->planes[plane] + y * src->stride[plane]);
+ uint16_t *p_dst =
+ (uint16_t *)(dst->planes[plane] + y * dst->stride[plane]);
+ for (x = 0; x < w; x++)
+ *p_dst++ = *p_src++ >> down_shift;
+ }
+ }
+}
+
+static void lowbd_img_downshift(vpx_image_t *dst, vpx_image_t *src,
+ int down_shift) {
+ int plane;
+ if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
+ dst->x_chroma_shift != src->x_chroma_shift ||
+ dst->y_chroma_shift != src->y_chroma_shift ||
+ src->fmt != dst->fmt + VPX_IMG_FMT_HIGHBITDEPTH ||
+ down_shift < 0) {
+ fatal("Unsupported image conversion");
+ }
+ switch (dst->fmt) {
+ case VPX_IMG_FMT_I420:
+ case VPX_IMG_FMT_I422:
+ case VPX_IMG_FMT_I444:
+ case VPX_IMG_FMT_I440:
+ break;
+ default:
+ fatal("Unsupported image conversion");
+ break;
+ }
+ for (plane = 0; plane < 3; plane++) {
+ int w = src->d_w;
+ int h = src->d_h;
+ int x, y;
+ if (plane) {
+ w = (w + src->x_chroma_shift) >> src->x_chroma_shift;
+ h = (h + src->y_chroma_shift) >> src->y_chroma_shift;
+ }
+ for (y = 0; y < h; y++) {
+ uint16_t *p_src =
+ (uint16_t *)(src->planes[plane] + y * src->stride[plane]);
+ uint8_t *p_dst = dst->planes[plane] + y * dst->stride[plane];
+ for (x = 0; x < w; x++) {
+ *p_dst++ = *p_src++ >> down_shift;
+ }
+ }
+ }
+}
+
+void vpx_img_downshift(vpx_image_t *dst, vpx_image_t *src,
+ int down_shift) {
+ if (dst->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
+ highbd_img_downshift(dst, src, down_shift);
+ } else {
+ lowbd_img_downshift(dst, src, down_shift);
+ }
+}
+#endif // CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
diff --git a/source/libvpx/tools_common.h b/source/libvpx/tools_common.h
index c1f466b..de6c38f 100644
--- a/source/libvpx/tools_common.h
+++ b/source/libvpx/tools_common.h
@@ -149,6 +149,12 @@
double sse_to_psnr(double samples, double peak, double mse);
+#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
+void vpx_img_upshift(vpx_image_t *dst, vpx_image_t *src, int input_shift);
+void vpx_img_downshift(vpx_image_t *dst, vpx_image_t *src, int down_shift);
+void vpx_img_truncate_16_to_8(vpx_image_t *dst, vpx_image_t *src);
+#endif
+
#ifdef __cplusplus
} /* extern "C" */
#endif
diff --git a/source/libvpx/vp8/common/arm/neon/loopfilter_neon.c b/source/libvpx/vp8/common/arm/neon/loopfilter_neon.c
index e103476..9d6807a 100644
--- a/source/libvpx/vp8/common/arm/neon/loopfilter_neon.c
+++ b/source/libvpx/vp8/common/arm/neon/loopfilter_neon.c
@@ -10,6 +10,7 @@
#include <arm_neon.h>
#include "./vpx_config.h"
+#include "vpx_ports/arm.h"
static INLINE void vp8_loop_filter_neon(
uint8x16_t qblimit, // flimit
@@ -251,38 +252,56 @@
return;
}
-#if (__GNUC__ == 4 && (__GNUC_MINOR__ == 6))
-#warning Using GCC 4.6 is not recommended
-// Some versions of gcc4.6 do not correctly process vst4_lane_u8. When built
-// with any gcc4.6, use the C code.
-extern void vp8_loop_filter_vertical_edge_c(unsigned char *s, int p,
- const unsigned char *blimit,
- const unsigned char *limit,
- const unsigned char *thresh,
- int count);
-
-void vp8_loop_filter_vertical_edge_y_neon(
- unsigned char *src,
- int pitch,
- unsigned char blimit,
- unsigned char limit,
- unsigned char thresh) {
- vp8_loop_filter_vertical_edge_c(src, pitch, &blimit, &limit, &thresh, 2);
-}
-
-void vp8_loop_filter_vertical_edge_uv_neon(
- unsigned char *u,
- int pitch,
- unsigned char blimit,
- unsigned char limit,
- unsigned char thresh,
- unsigned char *v) {
- vp8_loop_filter_vertical_edge_c(u, pitch, &blimit, &limit, &thresh, 1);
- vp8_loop_filter_vertical_edge_c(v, pitch, &blimit, &limit, &thresh, 1);
-}
-#else
static INLINE void write_4x8(unsigned char *dst, int pitch,
const uint8x8x4_t result) {
+#ifdef VPX_INCOMPATIBLE_GCC
+ /*
+ * uint8x8x4_t result
+ 00 01 02 03 | 04 05 06 07
+ 10 11 12 13 | 14 15 16 17
+ 20 21 22 23 | 24 25 26 27
+ 30 31 32 33 | 34 35 36 37
+ ---
+ * after vtrn_u16
+ 00 01 20 21 | 04 05 24 25
+ 02 03 22 23 | 06 07 26 27
+ 10 11 30 31 | 14 15 34 35
+ 12 13 32 33 | 16 17 36 37
+ ---
+ * after vtrn_u8
+ 00 10 20 30 | 04 14 24 34
+ 01 11 21 31 | 05 15 25 35
+ 02 12 22 32 | 06 16 26 36
+ 03 13 23 33 | 07 17 27 37
+ */
+ const uint16x4x2_t r02_u16 = vtrn_u16(vreinterpret_u16_u8(result.val[0]),
+ vreinterpret_u16_u8(result.val[2]));
+ const uint16x4x2_t r13_u16 = vtrn_u16(vreinterpret_u16_u8(result.val[1]),
+ vreinterpret_u16_u8(result.val[3]));
+ const uint8x8x2_t r01_u8 = vtrn_u8(vreinterpret_u8_u16(r02_u16.val[0]),
+ vreinterpret_u8_u16(r13_u16.val[0]));
+ const uint8x8x2_t r23_u8 = vtrn_u8(vreinterpret_u8_u16(r02_u16.val[1]),
+ vreinterpret_u8_u16(r13_u16.val[1]));
+ const uint32x2_t x_0_4 = vreinterpret_u32_u8(r01_u8.val[0]);
+ const uint32x2_t x_1_5 = vreinterpret_u32_u8(r01_u8.val[1]);
+ const uint32x2_t x_2_6 = vreinterpret_u32_u8(r23_u8.val[0]);
+ const uint32x2_t x_3_7 = vreinterpret_u32_u8(r23_u8.val[1]);
+ vst1_lane_u32((uint32_t *)dst, x_0_4, 0);
+ dst += pitch;
+ vst1_lane_u32((uint32_t *)dst, x_1_5, 0);
+ dst += pitch;
+ vst1_lane_u32((uint32_t *)dst, x_2_6, 0);
+ dst += pitch;
+ vst1_lane_u32((uint32_t *)dst, x_3_7, 0);
+ dst += pitch;
+ vst1_lane_u32((uint32_t *)dst, x_0_4, 1);
+ dst += pitch;
+ vst1_lane_u32((uint32_t *)dst, x_1_5, 1);
+ dst += pitch;
+ vst1_lane_u32((uint32_t *)dst, x_2_6, 1);
+ dst += pitch;
+ vst1_lane_u32((uint32_t *)dst, x_3_7, 1);
+#else
vst4_lane_u8(dst, result, 0);
dst += pitch;
vst4_lane_u8(dst, result, 1);
@@ -298,6 +317,7 @@
vst4_lane_u8(dst, result, 6);
dst += pitch;
vst4_lane_u8(dst, result, 7);
+#endif // VPX_INCOMPATIBLE_GCC
}
void vp8_loop_filter_vertical_edge_y_neon(
@@ -528,4 +548,3 @@
vd = v - 2;
write_4x8(vd, pitch, q4ResultH);
}
-#endif // (__GNUC__ == 4 && (__GNUC_MINOR__ == 6))
diff --git a/source/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c b/source/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c
index d5178bb..e1c8609 100644
--- a/source/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c
+++ b/source/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c
@@ -10,45 +10,9 @@
#include <arm_neon.h>
#include "./vpx_config.h"
+#include "vpx_ports/arm.h"
-#if (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7))
-static INLINE void write_2x8(unsigned char *dst, int pitch,
- const uint8x8x2_t result,
- const uint8x8x2_t result2) {
- vst2_lane_u8(dst, result, 0);
- dst += pitch;
- vst2_lane_u8(dst, result, 1);
- dst += pitch;
- vst2_lane_u8(dst, result, 2);
- dst += pitch;
- vst2_lane_u8(dst, result, 3);
- dst += pitch;
- vst2_lane_u8(dst, result, 4);
- dst += pitch;
- vst2_lane_u8(dst, result, 5);
- dst += pitch;
- vst2_lane_u8(dst, result, 6);
- dst += pitch;
- vst2_lane_u8(dst, result, 7);
- dst += pitch;
-
- vst2_lane_u8(dst, result2, 0);
- dst += pitch;
- vst2_lane_u8(dst, result2, 1);
- dst += pitch;
- vst2_lane_u8(dst, result2, 2);
- dst += pitch;
- vst2_lane_u8(dst, result2, 3);
- dst += pitch;
- vst2_lane_u8(dst, result2, 4);
- dst += pitch;
- vst2_lane_u8(dst, result2, 5);
- dst += pitch;
- vst2_lane_u8(dst, result2, 6);
- dst += pitch;
- vst2_lane_u8(dst, result2, 7);
-}
-#else
+#ifdef VPX_INCOMPATIBLE_GCC
static INLINE void write_2x4(unsigned char *dst, int pitch,
const uint8x8x2_t result) {
/*
@@ -88,30 +52,47 @@
dst += pitch * 8;
write_2x4(dst, pitch, result2);
}
-#endif
-
-
-#if (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7))
-static INLINE
-uint8x8x4_t read_4x8(unsigned char *src, int pitch, uint8x8x4_t x) {
- x = vld4_lane_u8(src, x, 0);
- src += pitch;
- x = vld4_lane_u8(src, x, 1);
- src += pitch;
- x = vld4_lane_u8(src, x, 2);
- src += pitch;
- x = vld4_lane_u8(src, x, 3);
- src += pitch;
- x = vld4_lane_u8(src, x, 4);
- src += pitch;
- x = vld4_lane_u8(src, x, 5);
- src += pitch;
- x = vld4_lane_u8(src, x, 6);
- src += pitch;
- x = vld4_lane_u8(src, x, 7);
- return x;
-}
#else
+static INLINE void write_2x8(unsigned char *dst, int pitch,
+ const uint8x8x2_t result,
+ const uint8x8x2_t result2) {
+ vst2_lane_u8(dst, result, 0);
+ dst += pitch;
+ vst2_lane_u8(dst, result, 1);
+ dst += pitch;
+ vst2_lane_u8(dst, result, 2);
+ dst += pitch;
+ vst2_lane_u8(dst, result, 3);
+ dst += pitch;
+ vst2_lane_u8(dst, result, 4);
+ dst += pitch;
+ vst2_lane_u8(dst, result, 5);
+ dst += pitch;
+ vst2_lane_u8(dst, result, 6);
+ dst += pitch;
+ vst2_lane_u8(dst, result, 7);
+ dst += pitch;
+
+ vst2_lane_u8(dst, result2, 0);
+ dst += pitch;
+ vst2_lane_u8(dst, result2, 1);
+ dst += pitch;
+ vst2_lane_u8(dst, result2, 2);
+ dst += pitch;
+ vst2_lane_u8(dst, result2, 3);
+ dst += pitch;
+ vst2_lane_u8(dst, result2, 4);
+ dst += pitch;
+ vst2_lane_u8(dst, result2, 5);
+ dst += pitch;
+ vst2_lane_u8(dst, result2, 6);
+ dst += pitch;
+ vst2_lane_u8(dst, result2, 7);
+}
+#endif // VPX_INCOMPATIBLE_GCC
+
+
+#ifdef VPX_INCOMPATIBLE_GCC
static INLINE
uint8x8x4_t read_4x8(unsigned char *src, int pitch, uint8x8x4_t x) {
const uint8x8_t a = vld1_u8(src);
@@ -169,7 +150,27 @@
return x;
}
-#endif
+#else
+static INLINE
+uint8x8x4_t read_4x8(unsigned char *src, int pitch, uint8x8x4_t x) {
+ x = vld4_lane_u8(src, x, 0);
+ src += pitch;
+ x = vld4_lane_u8(src, x, 1);
+ src += pitch;
+ x = vld4_lane_u8(src, x, 2);
+ src += pitch;
+ x = vld4_lane_u8(src, x, 3);
+ src += pitch;
+ x = vld4_lane_u8(src, x, 4);
+ src += pitch;
+ x = vld4_lane_u8(src, x, 5);
+ src += pitch;
+ x = vld4_lane_u8(src, x, 6);
+ src += pitch;
+ x = vld4_lane_u8(src, x, 7);
+ return x;
+}
+#endif // VPX_INCOMPATIBLE_GCC
static INLINE void vp8_loop_filter_simple_vertical_edge_neon(
unsigned char *s,
diff --git a/source/libvpx/vp8/encoder/arm/neon/vp8_shortwalsh4x4_neon.c b/source/libvpx/vp8/encoder/arm/neon/vp8_shortwalsh4x4_neon.c
index ffa3d91..5ad9465 100644
--- a/source/libvpx/vp8/encoder/arm/neon/vp8_shortwalsh4x4_neon.c
+++ b/source/libvpx/vp8/encoder/arm/neon/vp8_shortwalsh4x4_neon.c
@@ -9,11 +9,9 @@
*/
#include <arm_neon.h>
+#include "vpx_ports/arm.h"
-#if (__GNUC__ == 4 && (__GNUC_MINOR__ == 6))
-#warning Using GCC 4.6 is not recommended
-// Some versions of gcc4.6 do not correctly process this function. When built
-// with any gcc4.6, use the C code.
+#ifdef VPX_INCOMPATIBLE_GCC
#include "./vp8_rtcd.h"
void vp8_short_walsh4x4_neon(
int16_t *input,
@@ -128,4 +126,4 @@
vst1q_s16(output + 8, q1s16);
return;
}
-#endif // (__GNUC__ == 4 && (__GNUC_MINOR__ == 6))
+#endif // VPX_INCOMPATIBLE_GCC
diff --git a/source/libvpx/vp8/encoder/denoising.c b/source/libvpx/vp8/encoder/denoising.c
index 12f9734..75b2a3b 100644
--- a/source/libvpx/vp8/encoder/denoising.c
+++ b/source/libvpx/vp8/encoder/denoising.c
@@ -390,9 +390,9 @@
denoiser->denoise_pars.scale_motion_thresh = 16;
denoiser->denoise_pars.scale_increase_filter = 1;
denoiser->denoise_pars.denoise_mv_bias = 60;
- denoiser->denoise_pars.pickmode_mv_bias = 60;
- denoiser->denoise_pars.qp_thresh = 100;
- denoiser->denoise_pars.consec_zerolast = 10;
+ denoiser->denoise_pars.pickmode_mv_bias = 75;
+ denoiser->denoise_pars.qp_thresh = 85;
+ denoiser->denoise_pars.consec_zerolast = 15;
denoiser->denoise_pars.spatial_blur = 20;
}
}
@@ -453,17 +453,17 @@
// Bitrate thresholds and noise metric (nmse) thresholds for switching to
// aggressive mode.
// TODO(marpan): Adjust thresholds, including effect on resolution.
- denoiser->bitrate_threshold = 200000; // (bits/sec).
+ denoiser->bitrate_threshold = 300000; // (bits/sec).
denoiser->threshold_aggressive_mode = 35;
- if (width * height > 640 * 480) {
- denoiser->bitrate_threshold = 500000;
- denoiser->threshold_aggressive_mode = 100;
+ if (width * height > 1280 * 720) {
+ denoiser->bitrate_threshold = 2000000;
+ denoiser->threshold_aggressive_mode = 1400;
} else if (width * height > 960 * 540) {
denoiser->bitrate_threshold = 800000;
denoiser->threshold_aggressive_mode = 150;
- } else if (width * height > 1280 * 720) {
- denoiser->bitrate_threshold = 2000000;
- denoiser->threshold_aggressive_mode = 1400;
+ } else if (width * height > 640 * 480) {
+ denoiser->bitrate_threshold = 500000;
+ denoiser->threshold_aggressive_mode = 100;
}
return 0;
}
diff --git a/source/libvpx/vp8/encoder/denoising.h b/source/libvpx/vp8/encoder/denoising.h
index fb7930b..6c1f9e2 100644
--- a/source/libvpx/vp8/encoder/denoising.h
+++ b/source/libvpx/vp8/encoder/denoising.h
@@ -27,6 +27,8 @@
#define SUM_DIFF_FROM_AVG_THRESH_UV (8 * 8 * 8)
#define MOTION_MAGNITUDE_THRESHOLD_UV (8*3)
+#define MAX_GF_ARF_DENOISE_RANGE (16)
+
enum vp8_denoiser_decision
{
COPY_BLOCK,
diff --git a/source/libvpx/vp8/encoder/mcomp.c b/source/libvpx/vp8/encoder/mcomp.c
index 54abe76..545f2c8 100644
--- a/source/libvpx/vp8/encoder/mcomp.c
+++ b/source/libvpx/vp8/encoder/mcomp.c
@@ -393,8 +393,8 @@
#endif
/* central mv */
- bestmv->as_mv.row <<= 3;
- bestmv->as_mv.col <<= 3;
+ bestmv->as_mv.row *= 8;
+ bestmv->as_mv.col *= 8;
startmv = *bestmv;
/* calculate central point error */
diff --git a/source/libvpx/vp8/encoder/pickinter.c b/source/libvpx/vp8/encoder/pickinter.c
index 43f8957..9d5556d 100644
--- a/source/libvpx/vp8/encoder/pickinter.c
+++ b/source/libvpx/vp8/encoder/pickinter.c
@@ -516,9 +516,8 @@
// Adjust rd for ZEROMV and LAST, if LAST is the closest reference frame.
if (this_mode == ZEROMV &&
x->e_mbd.mode_info_context->mbmi.ref_frame == LAST_FRAME &&
- (denoise_aggressive || cpi->closest_reference_frame == LAST_FRAME))
- {
- this_rd = ((int64_t)this_rd) * rd_adj / 100;
+ (denoise_aggressive || cpi->closest_reference_frame == LAST_FRAME)) {
+ this_rd = ((int64_t)this_rd) * rd_adj / 100;
}
check_for_encode_breakout(*sse, x);
@@ -1083,7 +1082,14 @@
{
/* Store for later use by denoiser. */
- if (this_mode == ZEROMV && sse < zero_mv_sse )
+ // Dont' denoise with GOLDEN OR ALTREF is they are old reference
+ // frames (greater than MAX_GF_ARF_DENOISE_RANGE frames in past).
+ int skip_old_reference = ((this_ref_frame != LAST_FRAME) &&
+ (cpi->common.current_video_frame -
+ cpi->current_ref_frames[this_ref_frame] >
+ MAX_GF_ARF_DENOISE_RANGE)) ? 1 : 0;
+ if (this_mode == ZEROMV && sse < zero_mv_sse &&
+ !skip_old_reference)
{
zero_mv_sse = sse;
x->best_zeromv_reference_frame =
@@ -1092,7 +1098,7 @@
/* Store the best NEWMV in x for later use in the denoiser. */
if (x->e_mbd.mode_info_context->mbmi.mode == NEWMV &&
- sse < best_sse)
+ sse < best_sse && !skip_old_reference)
{
best_sse = sse;
x->best_sse_inter_mode = NEWMV;
diff --git a/source/libvpx/vp8/vp8_dx_iface.c b/source/libvpx/vp8/vp8_dx_iface.c
index 3ab8ed0..5aa274d 100644
--- a/source/libvpx/vp8/vp8_dx_iface.c
+++ b/source/libvpx/vp8/vp8_dx_iface.c
@@ -112,22 +112,19 @@
* structure. More memory may be required at the time the stream
* information becomes known.
*/
- if (!ctx->priv)
- {
- vp8_init_ctx(ctx);
- priv = (vpx_codec_alg_priv_t *)ctx->priv;
+ if (!ctx->priv) {
+ vp8_init_ctx(ctx);
+ priv = (vpx_codec_alg_priv_t *)ctx->priv;
- /* initialize number of fragments to zero */
- priv->fragments.count = 0;
- /* is input fragments enabled? */
- priv->fragments.enabled =
- (priv->base.init_flags & VPX_CODEC_USE_INPUT_FRAGMENTS);
+ /* initialize number of fragments to zero */
+ priv->fragments.count = 0;
+ /* is input fragments enabled? */
+ priv->fragments.enabled =
+ (priv->base.init_flags & VPX_CODEC_USE_INPUT_FRAGMENTS);
- /*post processing level initialized to do nothing */
- }
- else
- {
- priv = (vpx_codec_alg_priv_t *)ctx->priv;
+ /*post processing level initialized to do nothing */
+ } else {
+ priv = (vpx_codec_alg_priv_t *)ctx->priv;
}
priv->yv12_frame_buffers.use_frame_threads =
@@ -138,11 +135,10 @@
if (priv->yv12_frame_buffers.use_frame_threads &&
((ctx->priv->init_flags & VPX_CODEC_USE_ERROR_CONCEALMENT) ||
- (ctx->priv->init_flags & VPX_CODEC_USE_INPUT_FRAGMENTS)))
- {
- /* row-based threading, error concealment, and input fragments will
- * not be supported when using frame-based threading */
- res = VPX_CODEC_INVALID_PARAM;
+ (ctx->priv->init_flags & VPX_CODEC_USE_INPUT_FRAGMENTS))) {
+ /* row-based threading, error concealment, and input fragments will
+ * not be supported when using frame-based threading */
+ res = VPX_CODEC_INVALID_PARAM;
}
return res;
diff --git a/source/libvpx/vp9/common/vp9_blockd.c b/source/libvpx/vp9/common/vp9_blockd.c
index e13445f..7094a01 100644
--- a/source/libvpx/vp9/common/vp9_blockd.c
+++ b/source/libvpx/vp9/common/vp9_blockd.c
@@ -92,7 +92,7 @@
void *arg) {
int plane;
- for (plane = 0; plane < MAX_MB_PLANE; plane++)
+ for (plane = 0; plane < MAX_MB_PLANE; ++plane)
vp9_foreach_transformed_block_in_plane(xd, bsize, plane, visit, arg);
}
diff --git a/source/libvpx/vp9/common/vp9_blockd.h b/source/libvpx/vp9/common/vp9_blockd.h
index 702efe0..1234d54 100644
--- a/source/libvpx/vp9/common/vp9_blockd.h
+++ b/source/libvpx/vp9/common/vp9_blockd.h
@@ -17,14 +17,10 @@
#include "vpx_ports/mem.h"
#include "vpx_scale/yv12config.h"
-#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_common_data.h"
-#include "vp9/common/vp9_enums.h"
#include "vp9/common/vp9_filter.h"
-#include "vp9/common/vp9_idct.h"
#include "vp9/common/vp9_mv.h"
#include "vp9/common/vp9_scale.h"
-#include "vp9/common/vp9_seg_common.h"
#ifdef __cplusplus
extern "C" {
@@ -47,9 +43,9 @@
PLANE_TYPES
} PLANE_TYPE;
-typedef char ENTROPY_CONTEXT;
+#define MAX_MB_PLANE 3
-typedef char PARTITION_CONTEXT;
+typedef char ENTROPY_CONTEXT;
static INLINE int combine_entropy_contexts(ENTROPY_CONTEXT a,
ENTROPY_CONTEXT b) {
@@ -111,17 +107,6 @@
MAX_REF_FRAMES = 4
} MV_REFERENCE_FRAME;
-static INLINE int b_width_log2(BLOCK_SIZE sb_type) {
- return b_width_log2_lookup[sb_type];
-}
-static INLINE int b_height_log2(BLOCK_SIZE sb_type) {
- return b_height_log2_lookup[sb_type];
-}
-
-static INLINE int mi_width_log2(BLOCK_SIZE sb_type) {
- return mi_width_log2_lookup[sb_type];
-}
-
// This structure now relates to 8x8 block regions.
typedef struct {
// Common for both INTER and INTRA blocks
@@ -173,8 +158,6 @@
MV_PRECISION_Q4
};
-enum { MAX_MB_PLANE = 3 };
-
struct buf_2d {
uint8_t *buf;
int stride;
@@ -312,7 +295,7 @@
static INLINE void txfrm_block_to_raster_xy(BLOCK_SIZE plane_bsize,
TX_SIZE tx_size, int block,
int *x, int *y) {
- const int bwl = b_width_log2(plane_bsize);
+ const int bwl = b_width_log2_lookup[plane_bsize];
const int tx_cols_log2 = bwl - tx_size;
const int tx_cols = 1 << tx_cols_log2;
const int raster_mb = block >> (tx_size << 1);
diff --git a/source/libvpx/vp9/common/vp9_common.h b/source/libvpx/vp9/common/vp9_common.h
index 8305e7f..6801dd3 100644
--- a/source/libvpx/vp9/common/vp9_common.h
+++ b/source/libvpx/vp9/common/vp9_common.h
@@ -65,7 +65,7 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE uint16_t clip_pixel_high(int val, int bd) {
+static INLINE uint16_t clip_pixel_highbd(int val, int bd) {
switch (bd) {
case 8:
default:
@@ -77,8 +77,22 @@
}
}
+// Note:
+// tran_low_t is the datatype used for final transform coefficients.
+// tran_high_t is the datatype used for intermediate transform stages.
+typedef int64_t tran_high_t;
+typedef int32_t tran_low_t;
+
#define CONVERT_TO_SHORTPTR(x) ((uint16_t*)(((uintptr_t)x) << 1))
#define CONVERT_TO_BYTEPTR(x) ((uint8_t*)(((uintptr_t)x) >> 1 ))
+
+#else
+
+// Note:
+// tran_low_t is the datatype used for final transform coefficients.
+// tran_high_t is the datatype used for intermediate transform stages.
+typedef int32_t tran_high_t;
+typedef int16_t tran_low_t;
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_DEBUG
diff --git a/source/libvpx/vp9/common/vp9_common_data.c b/source/libvpx/vp9/common/vp9_common_data.c
index d4c1b71..2aaa009 100644
--- a/source/libvpx/vp9/common/vp9_common_data.c
+++ b/source/libvpx/vp9/common/vp9_common_data.c
@@ -8,8 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
-#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_common_data.h"
// Log 2 conversion lookup tables for block width and height
@@ -36,7 +34,6 @@
const int num_pels_log2_lookup[BLOCK_SIZES] =
{4, 5, 5, 6, 7, 7, 8, 9, 9, 10, 11, 11, 12};
-
const PARTITION_TYPE partition_lookup[][BLOCK_SIZES] = {
{ // 4X4
// 4X4, 4X8,8X4,8X8,8X16,16X8,16X16,16X32,32X16,32X32,32X64,64X32,64X64
diff --git a/source/libvpx/vp9/common/vp9_convolve.c b/source/libvpx/vp9/common/vp9_convolve.c
index ad70e59..7b65651 100644
--- a/source/libvpx/vp9/common/vp9_convolve.c
+++ b/source/libvpx/vp9/common/vp9_convolve.c
@@ -130,16 +130,14 @@
// --Require an additional SUBPEL_TAPS rows for the 8-tap filter tails.
// --((64 - 1) * 32 + 15) >> 4 + 8 = 135.
uint8_t temp[135 * 64];
- int intermediate_height = (((h - 1) * y_step_q4 + 15) >> 4) + SUBPEL_TAPS;
+ int intermediate_height =
+ (((h - 1) * y_step_q4 + y0_q4) >> SUBPEL_BITS) + SUBPEL_TAPS;
assert(w <= 64);
assert(h <= 64);
assert(y_step_q4 <= 32);
assert(x_step_q4 <= 32);
- if (intermediate_height < h)
- intermediate_height = h;
-
convolve_horiz(src - src_stride * (SUBPEL_TAPS / 2 - 1), src_stride, temp, 64,
x_filters, x0_q4, x_step_q4, w, intermediate_height);
convolve_vert(temp + 64 * (SUBPEL_TAPS / 2 - 1), 64, dst, dst_stride,
@@ -284,11 +282,11 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-static void high_convolve_horiz(const uint8_t *src8, ptrdiff_t src_stride,
- uint8_t *dst8, ptrdiff_t dst_stride,
- const InterpKernel *x_filters,
- int x0_q4, int x_step_q4,
- int w, int h, int bd) {
+static void highbd_convolve_horiz(const uint8_t *src8, ptrdiff_t src_stride,
+ uint8_t *dst8, ptrdiff_t dst_stride,
+ const InterpKernel *x_filters,
+ int x0_q4, int x_step_q4,
+ int w, int h, int bd) {
int x, y;
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
@@ -301,7 +299,7 @@
int k, sum = 0;
for (k = 0; k < SUBPEL_TAPS; ++k)
sum += src_x[k] * x_filter[k];
- dst[x] = clip_pixel_high(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
+ dst[x] = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
x_q4 += x_step_q4;
}
src += src_stride;
@@ -309,11 +307,11 @@
}
}
-static void high_convolve_avg_horiz(const uint8_t *src8, ptrdiff_t src_stride,
- uint8_t *dst8, ptrdiff_t dst_stride,
- const InterpKernel *x_filters,
- int x0_q4, int x_step_q4,
- int w, int h, int bd) {
+static void highbd_convolve_avg_horiz(const uint8_t *src8, ptrdiff_t src_stride,
+ uint8_t *dst8, ptrdiff_t dst_stride,
+ const InterpKernel *x_filters,
+ int x0_q4, int x_step_q4,
+ int w, int h, int bd) {
int x, y;
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
@@ -327,7 +325,7 @@
for (k = 0; k < SUBPEL_TAPS; ++k)
sum += src_x[k] * x_filter[k];
dst[x] = ROUND_POWER_OF_TWO(dst[x] +
- clip_pixel_high(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd), 1);
+ clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd), 1);
x_q4 += x_step_q4;
}
src += src_stride;
@@ -335,11 +333,11 @@
}
}
-static void high_convolve_vert(const uint8_t *src8, ptrdiff_t src_stride,
- uint8_t *dst8, ptrdiff_t dst_stride,
- const InterpKernel *y_filters,
- int y0_q4, int y_step_q4, int w, int h,
- int bd) {
+static void highbd_convolve_vert(const uint8_t *src8, ptrdiff_t src_stride,
+ uint8_t *dst8, ptrdiff_t dst_stride,
+ const InterpKernel *y_filters,
+ int y0_q4, int y_step_q4, int w, int h,
+ int bd) {
int x, y;
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
@@ -352,7 +350,7 @@
int k, sum = 0;
for (k = 0; k < SUBPEL_TAPS; ++k)
sum += src_y[k * src_stride] * y_filter[k];
- dst[y * dst_stride] = clip_pixel_high(
+ dst[y * dst_stride] = clip_pixel_highbd(
ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
y_q4 += y_step_q4;
}
@@ -361,11 +359,11 @@
}
}
-static void high_convolve_avg_vert(const uint8_t *src8, ptrdiff_t src_stride,
- uint8_t *dst8, ptrdiff_t dst_stride,
- const InterpKernel *y_filters,
- int y0_q4, int y_step_q4, int w, int h,
- int bd) {
+static void highbd_convolve_avg_vert(const uint8_t *src8, ptrdiff_t src_stride,
+ uint8_t *dst8, ptrdiff_t dst_stride,
+ const InterpKernel *y_filters,
+ int y0_q4, int y_step_q4, int w, int h,
+ int bd) {
int x, y;
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
@@ -379,7 +377,7 @@
for (k = 0; k < SUBPEL_TAPS; ++k)
sum += src_y[k * src_stride] * y_filter[k];
dst[y * dst_stride] = ROUND_POWER_OF_TWO(dst[y * dst_stride] +
- clip_pixel_high(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd), 1);
+ clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd), 1);
y_q4 += y_step_q4;
}
++src;
@@ -387,13 +385,13 @@
}
}
-static void high_convolve(const uint8_t *src, ptrdiff_t src_stride,
- uint8_t *dst, ptrdiff_t dst_stride,
- const InterpKernel *const x_filters,
- int x0_q4, int x_step_q4,
- const InterpKernel *const y_filters,
- int y0_q4, int y_step_q4,
- int w, int h, int bd) {
+static void highbd_convolve(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const InterpKernel *const x_filters,
+ int x0_q4, int x_step_q4,
+ const InterpKernel *const y_filters,
+ int y0_q4, int y_step_q4,
+ int w, int h, int bd) {
// Note: Fixed size intermediate buffer, temp, places limits on parameters.
// 2d filtering proceeds in 2 steps:
// (1) Interpolate horizontally into an intermediate buffer, temp.
@@ -407,119 +405,117 @@
// --Require an additional SUBPEL_TAPS rows for the 8-tap filter tails.
// --((64 - 1) * 32 + 15) >> 4 + 8 = 135.
uint16_t temp[64 * 135];
- int intermediate_height = (((h - 1) * y_step_q4 + 15) >> 4) + SUBPEL_TAPS;
+ int intermediate_height =
+ (((h - 1) * y_step_q4 + y0_q4) >> SUBPEL_BITS) + SUBPEL_TAPS;
assert(w <= 64);
assert(h <= 64);
assert(y_step_q4 <= 32);
assert(x_step_q4 <= 32);
- if (intermediate_height < h)
- intermediate_height = h;
-
- high_convolve_horiz(src - src_stride * (SUBPEL_TAPS / 2 - 1),
- src_stride, CONVERT_TO_BYTEPTR(temp), 64,
- x_filters, x0_q4, x_step_q4, w,
- intermediate_height, bd);
- high_convolve_vert(CONVERT_TO_BYTEPTR(temp) + 64 * (SUBPEL_TAPS / 2 - 1),
- 64, dst, dst_stride, y_filters, y0_q4, y_step_q4,
- w, h, bd);
+ highbd_convolve_horiz(src - src_stride * (SUBPEL_TAPS / 2 - 1),
+ src_stride, CONVERT_TO_BYTEPTR(temp), 64,
+ x_filters, x0_q4, x_step_q4, w,
+ intermediate_height, bd);
+ highbd_convolve_vert(CONVERT_TO_BYTEPTR(temp) + 64 * (SUBPEL_TAPS / 2 - 1),
+ 64, dst, dst_stride, y_filters, y0_q4, y_step_q4,
+ w, h, bd);
}
-void vp9_high_convolve8_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
+void vp9_highbd_convolve8_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_q4,
+ const int16_t *filter_y, int y_step_q4,
+ int w, int h, int bd) {
+ const InterpKernel *const filters_x = get_filter_base(filter_x);
+ const int x0_q4 = get_filter_offset(filter_x, filters_x);
+ (void)filter_y;
+ (void)y_step_q4;
+
+ highbd_convolve_horiz(src, src_stride, dst, dst_stride, filters_x,
+ x0_q4, x_step_q4, w, h, bd);
+}
+
+void vp9_highbd_convolve8_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_q4,
+ const int16_t *filter_y, int y_step_q4,
+ int w, int h, int bd) {
+ const InterpKernel *const filters_x = get_filter_base(filter_x);
+ const int x0_q4 = get_filter_offset(filter_x, filters_x);
+ (void)filter_y;
+ (void)y_step_q4;
+
+ highbd_convolve_avg_horiz(src, src_stride, dst, dst_stride, filters_x,
+ x0_q4, x_step_q4, w, h, bd);
+}
+
+void vp9_highbd_convolve8_vert_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_q4,
+ const int16_t *filter_y, int y_step_q4,
+ int w, int h, int bd) {
+ const InterpKernel *const filters_y = get_filter_base(filter_y);
+ const int y0_q4 = get_filter_offset(filter_y, filters_y);
+ (void)filter_x;
+ (void)x_step_q4;
+
+ highbd_convolve_vert(src, src_stride, dst, dst_stride, filters_y,
+ y0_q4, y_step_q4, w, h, bd);
+}
+
+void vp9_highbd_convolve8_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_q4,
+ const int16_t *filter_y, int y_step_q4,
+ int w, int h, int bd) {
+ const InterpKernel *const filters_y = get_filter_base(filter_y);
+ const int y0_q4 = get_filter_offset(filter_y, filters_y);
+ (void)filter_x;
+ (void)x_step_q4;
+
+ highbd_convolve_avg_vert(src, src_stride, dst, dst_stride, filters_y,
+ y0_q4, y_step_q4, w, h, bd);
+}
+
+void vp9_highbd_convolve8_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_q4,
+ const int16_t *filter_y, int y_step_q4,
+ int w, int h, int bd) {
+ const InterpKernel *const filters_x = get_filter_base(filter_x);
+ const int x0_q4 = get_filter_offset(filter_x, filters_x);
+
+ const InterpKernel *const filters_y = get_filter_base(filter_y);
+ const int y0_q4 = get_filter_offset(filter_y, filters_y);
+
+ highbd_convolve(src, src_stride, dst, dst_stride,
+ filters_x, x0_q4, x_step_q4,
+ filters_y, y0_q4, y_step_q4, w, h, bd);
+}
+
+void vp9_highbd_convolve8_avg_c(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h, int bd) {
- const InterpKernel *const filters_x = get_filter_base(filter_x);
- const int x0_q4 = get_filter_offset(filter_x, filters_x);
- (void)filter_y;
- (void)y_step_q4;
-
- high_convolve_horiz(src, src_stride, dst, dst_stride, filters_x,
- x0_q4, x_step_q4, w, h, bd);
-}
-
-void vp9_high_convolve8_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
- uint8_t *dst, ptrdiff_t dst_stride,
- const int16_t *filter_x, int x_step_q4,
- const int16_t *filter_y, int y_step_q4,
- int w, int h, int bd) {
- const InterpKernel *const filters_x = get_filter_base(filter_x);
- const int x0_q4 = get_filter_offset(filter_x, filters_x);
- (void)filter_y;
- (void)y_step_q4;
-
- high_convolve_avg_horiz(src, src_stride, dst, dst_stride, filters_x,
- x0_q4, x_step_q4, w, h, bd);
-}
-
-void vp9_high_convolve8_vert_c(const uint8_t *src, ptrdiff_t src_stride,
- uint8_t *dst, ptrdiff_t dst_stride,
- const int16_t *filter_x, int x_step_q4,
- const int16_t *filter_y, int y_step_q4,
- int w, int h, int bd) {
- const InterpKernel *const filters_y = get_filter_base(filter_y);
- const int y0_q4 = get_filter_offset(filter_y, filters_y);
- (void)filter_x;
- (void)x_step_q4;
-
- high_convolve_vert(src, src_stride, dst, dst_stride, filters_y,
- y0_q4, y_step_q4, w, h, bd);
-}
-
-void vp9_high_convolve8_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride,
- uint8_t *dst, ptrdiff_t dst_stride,
- const int16_t *filter_x, int x_step_q4,
- const int16_t *filter_y, int y_step_q4,
- int w, int h, int bd) {
- const InterpKernel *const filters_y = get_filter_base(filter_y);
- const int y0_q4 = get_filter_offset(filter_y, filters_y);
- (void)filter_x;
- (void)x_step_q4;
-
- high_convolve_avg_vert(src, src_stride, dst, dst_stride, filters_y,
- y0_q4, y_step_q4, w, h, bd);
-}
-
-void vp9_high_convolve8_c(const uint8_t *src, ptrdiff_t src_stride,
- uint8_t *dst, ptrdiff_t dst_stride,
- const int16_t *filter_x, int x_step_q4,
- const int16_t *filter_y, int y_step_q4,
- int w, int h, int bd) {
- const InterpKernel *const filters_x = get_filter_base(filter_x);
- const int x0_q4 = get_filter_offset(filter_x, filters_x);
-
- const InterpKernel *const filters_y = get_filter_base(filter_y);
- const int y0_q4 = get_filter_offset(filter_y, filters_y);
-
- high_convolve(src, src_stride, dst, dst_stride,
- filters_x, x0_q4, x_step_q4,
- filters_y, y0_q4, y_step_q4, w, h, bd);
-}
-
-void vp9_high_convolve8_avg_c(const uint8_t *src, ptrdiff_t src_stride,
- uint8_t *dst, ptrdiff_t dst_stride,
- const int16_t *filter_x, int x_step_q4,
- const int16_t *filter_y, int y_step_q4,
- int w, int h, int bd) {
// Fixed size intermediate buffer places limits on parameters.
DECLARE_ALIGNED_ARRAY(16, uint16_t, temp, 64 * 64);
assert(w <= 64);
assert(h <= 64);
- vp9_high_convolve8_c(src, src_stride, CONVERT_TO_BYTEPTR(temp), 64,
- filter_x, x_step_q4, filter_y, y_step_q4, w, h, bd);
- vp9_high_convolve_avg_c(CONVERT_TO_BYTEPTR(temp), 64, dst, dst_stride,
- NULL, 0, NULL, 0, w, h, bd);
+ vp9_highbd_convolve8_c(src, src_stride, CONVERT_TO_BYTEPTR(temp), 64,
+ filter_x, x_step_q4, filter_y, y_step_q4, w, h, bd);
+ vp9_highbd_convolve_avg_c(CONVERT_TO_BYTEPTR(temp), 64, dst, dst_stride,
+ NULL, 0, NULL, 0, w, h, bd);
}
-void vp9_high_convolve_copy_c(const uint8_t *src8, ptrdiff_t src_stride,
- uint8_t *dst8, ptrdiff_t dst_stride,
- const int16_t *filter_x, int filter_x_stride,
- const int16_t *filter_y, int filter_y_stride,
- int w, int h, int bd) {
+void vp9_highbd_convolve_copy_c(const uint8_t *src8, ptrdiff_t src_stride,
+ uint8_t *dst8, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int filter_x_stride,
+ const int16_t *filter_y, int filter_y_stride,
+ int w, int h, int bd) {
int r;
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
@@ -536,11 +532,11 @@
}
}
-void vp9_high_convolve_avg_c(const uint8_t *src8, ptrdiff_t src_stride,
- uint8_t *dst8, ptrdiff_t dst_stride,
- const int16_t *filter_x, int filter_x_stride,
- const int16_t *filter_y, int filter_y_stride,
- int w, int h, int bd) {
+void vp9_highbd_convolve_avg_c(const uint8_t *src8, ptrdiff_t src_stride,
+ uint8_t *dst8, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int filter_x_stride,
+ const int16_t *filter_y, int filter_y_stride,
+ int w, int h, int bd) {
int x, y;
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
diff --git a/source/libvpx/vp9/common/vp9_convolve.h b/source/libvpx/vp9/common/vp9_convolve.h
index faf70b1..8b044c8 100644
--- a/source/libvpx/vp9/common/vp9_convolve.h
+++ b/source/libvpx/vp9/common/vp9_convolve.h
@@ -24,11 +24,11 @@
int w, int h);
#if CONFIG_VP9_HIGHBITDEPTH
-typedef void (*high_convolve_fn_t)(const uint8_t *src, ptrdiff_t src_stride,
- uint8_t *dst, ptrdiff_t dst_stride,
- const int16_t *filter_x, int x_step_q4,
- const int16_t *filter_y, int y_step_q4,
- int w, int h, int bd);
+typedef void (*highbd_convolve_fn_t)(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_q4,
+ const int16_t *filter_y, int y_step_q4,
+ int w, int h, int bd);
#endif
#ifdef __cplusplus
diff --git a/source/libvpx/vp9/common/vp9_debugmodes.c b/source/libvpx/vp9/common/vp9_debugmodes.c
index 3f16841..d9dace6 100644
--- a/source/libvpx/vp9/common/vp9_debugmodes.c
+++ b/source/libvpx/vp9/common/vp9_debugmodes.c
@@ -27,7 +27,7 @@
int mi_row, mi_col;
int mi_index = 0;
// TODO(hkuang): Fix this debug function.
- MODE_INFO **mi = NULL;
+ MODE_INFO **mi = &cm->mi;
int rows = cm->mi_rows;
int cols = cm->mi_cols;
char prefix = descriptor[0];
@@ -53,7 +53,7 @@
int mi_index = 0;
FILE *mvs = fopen(file, "a");
// TODO(hkuang): Fix this debug function.
- MODE_INFO **mi = NULL;
+ MODE_INFO **mi = &cm->mi;
int rows = cm->mi_rows;
int cols = cm->mi_cols;
diff --git a/source/libvpx/vp9/common/vp9_entropy.h b/source/libvpx/vp9/common/vp9_entropy.h
index 8cdfc5c..239c049 100644
--- a/source/libvpx/vp9/common/vp9_entropy.h
+++ b/source/libvpx/vp9/common/vp9_entropy.h
@@ -13,8 +13,8 @@
#include "vpx/vpx_integer.h"
-#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_common.h"
+#include "vp9/common/vp9_prob.h"
#include "vp9/common/vp9_scan.h"
#ifdef __cplusplus
diff --git a/source/libvpx/vp9/common/vp9_entropymode.h b/source/libvpx/vp9/common/vp9_entropymode.h
index 533757b..6831d3f 100644
--- a/source/libvpx/vp9/common/vp9_entropymode.h
+++ b/source/libvpx/vp9/common/vp9_entropymode.h
@@ -20,8 +20,6 @@
#endif
#define TX_SIZE_CONTEXTS 2
-#define SWITCHABLE_FILTERS 3 // number of switchable filters
-#define SWITCHABLE_FILTER_CONTEXTS (SWITCHABLE_FILTERS + 1)
struct VP9Common;
diff --git a/source/libvpx/vp9/common/vp9_entropymv.h b/source/libvpx/vp9/common/vp9_entropymv.h
index e7033e4..75e6861 100644
--- a/source/libvpx/vp9/common/vp9_entropymv.h
+++ b/source/libvpx/vp9/common/vp9_entropymv.h
@@ -13,7 +13,9 @@
#define VP9_COMMON_VP9_ENTROPYMV_H_
#include "./vpx_config.h"
-#include "vp9/common/vp9_blockd.h"
+
+#include "vp9/common/vp9_mv.h"
+#include "vp9/common/vp9_prob.h"
#ifdef __cplusplus
extern "C" {
diff --git a/source/libvpx/vp9/common/vp9_enums.h b/source/libvpx/vp9/common/vp9_enums.h
index 8817fdb..f83d21f 100644
--- a/source/libvpx/vp9/common/vp9_enums.h
+++ b/source/libvpx/vp9/common/vp9_enums.h
@@ -67,6 +67,7 @@
PARTITION_INVALID = PARTITION_TYPES
} PARTITION_TYPE;
+typedef char PARTITION_CONTEXT;
#define PARTITION_PLOFFSET 4 // number of probability models per block size
#define PARTITION_CONTEXTS (4 * PARTITION_PLOFFSET)
diff --git a/source/libvpx/vp9/common/vp9_filter.h b/source/libvpx/vp9/common/vp9_filter.h
index 8c359c7..3377d45 100644
--- a/source/libvpx/vp9/common/vp9_filter.h
+++ b/source/libvpx/vp9/common/vp9_filter.h
@@ -35,6 +35,13 @@
SWITCHABLE = 4 /* should be the last one */
} INTERP_FILTER;
+// Number of switchable filters
+#define SWITCHABLE_FILTERS 3
+
+// The codec can operate in four possible inter prediction filter mode:
+// 8-tap, 8-tap-smooth, 8-tap-sharp, and switching between the three.
+#define SWITCHABLE_FILTER_CONTEXTS (SWITCHABLE_FILTERS + 1)
+
typedef int16_t InterpKernel[SUBPEL_TAPS];
const InterpKernel *vp9_get_interp_kernel(INTERP_FILTER filter);
diff --git a/source/libvpx/vp9/common/vp9_idct.c b/source/libvpx/vp9/common/vp9_idct.c
index b196fc5..d5b6f39 100644
--- a/source/libvpx/vp9/common/vp9_idct.c
+++ b/source/libvpx/vp9/common/vp9_idct.c
@@ -8,49 +8,48 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-#include <assert.h>
#include <math.h>
-#include "./vpx_config.h"
#include "./vp9_rtcd.h"
#include "vp9/common/vp9_systemdependent.h"
#include "vp9/common/vp9_blockd.h"
-#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_idct.h"
-#if CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH
-// When CONFIG_EMULATE_HW_HIGHBITDEPTH is 1 the transform performs strict
-// overflow wrapping to match expected hardware implementations.
+#if CONFIG_EMULATE_HARDWARE
+// When CONFIG_EMULATE_HARDWARE is 1 the transform performs a
+// non-normative method to handle overflows. A stream that causes
+// overflows in the inverse transform is considered invalid in VP9,
+// and a hardware implementer is free to choose any reasonable
+// method to handle overflows. However to aid in hardware
+// verification they can use a specific implementation of the
+// WRAPLOW() macro below that is identical to their intended
+// hardware implementation (and also use configure options to trigger
+// the C-implementation of the transform).
+//
+// The particular WRAPLOW implementation below performs strict
+// overflow wrapping to match common hardware implementations.
// bd of 8 uses trans_low with 16bits, need to remove 16bits
// bd of 10 uses trans_low with 18bits, need to remove 14bits
// bd of 12 uses trans_low with 20bits, need to remove 12bits
// bd of x uses trans_low with 8+x bits, need to remove 24-x bits
-#define WRAPLOW(x) ((((int32_t)x) << (24 - bd)) >> (24 - bd))
+#define WRAPLOW(x, bd) ((((int32_t)(x)) << (24 - bd)) >> (24 - bd))
#else
-#define WRAPLOW(x) (x)
-#endif // CONFIG_EMULATE_HARDWARE_HIGHBITDEPTH
+#define WRAPLOW(x, bd) (x)
+#endif // CONFIG_EMULATE_HARDWARE
#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE tran_low_t clamp_high(tran_high_t value, tran_low_t low,
- tran_low_t high) {
- return value < low ? low : (value > high ? high : value);
-}
-
-static INLINE tran_low_t clip_pixel_bd_high(tran_high_t dest,
- tran_high_t trans, int bd) {
- trans = WRAPLOW(trans);
- switch (bd) {
- case 8:
- default:
- return clamp_high(WRAPLOW(dest + trans), 0, 255);
- case 10:
- return clamp_high(WRAPLOW(dest + trans), 0, 1023);
- case 12:
- return clamp_high(WRAPLOW(dest + trans), 0, 4095);
- }
+static INLINE uint16_t highbd_clip_pixel_add(uint16_t dest, tran_high_t trans,
+ int bd) {
+ trans = WRAPLOW(trans, bd);
+ return clip_pixel_highbd(WRAPLOW(dest + trans, bd), bd);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
+static INLINE uint8_t clip_pixel_add(uint8_t dest, tran_high_t trans) {
+ trans = WRAPLOW(trans, 8);
+ return clip_pixel(WRAPLOW(dest + trans, 8));
+}
+
void vp9_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
/* 4-point reversible, orthonormal inverse Walsh-Hadamard in 3.5 adds,
0.5 shifts per pixel. */
@@ -72,10 +71,10 @@
c1 = e1 - c1;
a1 -= b1;
d1 += c1;
- op[0] = a1;
- op[1] = b1;
- op[2] = c1;
- op[3] = d1;
+ op[0] = WRAPLOW(a1, 8);
+ op[1] = WRAPLOW(b1, 8);
+ op[2] = WRAPLOW(c1, 8);
+ op[3] = WRAPLOW(d1, 8);
ip += 4;
op += 4;
}
@@ -93,10 +92,10 @@
c1 = e1 - c1;
a1 -= b1;
d1 += c1;
- dest[stride * 0] = clip_pixel(dest[stride * 0] + a1);
- dest[stride * 1] = clip_pixel(dest[stride * 1] + b1);
- dest[stride * 2] = clip_pixel(dest[stride * 2] + c1);
- dest[stride * 3] = clip_pixel(dest[stride * 3] + d1);
+ dest[stride * 0] = clip_pixel_add(dest[stride * 0], a1);
+ dest[stride * 1] = clip_pixel_add(dest[stride * 1], b1);
+ dest[stride * 2] = clip_pixel_add(dest[stride * 2], c1);
+ dest[stride * 3] = clip_pixel_add(dest[stride * 3], d1);
ip++;
dest++;
@@ -113,17 +112,17 @@
a1 = ip[0] >> UNIT_QUANT_SHIFT;
e1 = a1 >> 1;
a1 -= e1;
- op[0] = a1;
- op[1] = op[2] = op[3] = e1;
+ op[0] = WRAPLOW(a1, 8);
+ op[1] = op[2] = op[3] = WRAPLOW(e1, 8);
ip = tmp;
for (i = 0; i < 4; i++) {
e1 = ip[0] >> 1;
a1 = ip[0] - e1;
- dest[dest_stride * 0] = clip_pixel(dest[dest_stride * 0] + a1);
- dest[dest_stride * 1] = clip_pixel(dest[dest_stride * 1] + e1);
- dest[dest_stride * 2] = clip_pixel(dest[dest_stride * 2] + e1);
- dest[dest_stride * 3] = clip_pixel(dest[dest_stride * 3] + e1);
+ dest[dest_stride * 0] = clip_pixel_add(dest[dest_stride * 0], a1);
+ dest[dest_stride * 1] = clip_pixel_add(dest[dest_stride * 1], e1);
+ dest[dest_stride * 2] = clip_pixel_add(dest[dest_stride * 2], e1);
+ dest[dest_stride * 3] = clip_pixel_add(dest[dest_stride * 3], e1);
ip++;
dest++;
}
@@ -135,18 +134,18 @@
// stage 1
temp1 = (input[0] + input[2]) * cospi_16_64;
temp2 = (input[0] - input[2]) * cospi_16_64;
- step[0] = dct_const_round_shift(temp1);
- step[1] = dct_const_round_shift(temp2);
+ step[0] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step[1] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64;
temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64;
- step[2] = dct_const_round_shift(temp1);
- step[3] = dct_const_round_shift(temp2);
+ step[2] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step[3] = WRAPLOW(dct_const_round_shift(temp2), 8);
// stage 2
- output[0] = step[0] + step[3];
- output[1] = step[1] + step[2];
- output[2] = step[1] - step[2];
- output[3] = step[0] - step[3];
+ output[0] = WRAPLOW(step[0] + step[3], 8);
+ output[1] = WRAPLOW(step[1] + step[2], 8);
+ output[2] = WRAPLOW(step[1] - step[2], 8);
+ output[3] = WRAPLOW(step[0] - step[3], 8);
}
void vp9_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
@@ -167,9 +166,10 @@
for (j = 0; j < 4; ++j)
temp_in[j] = out[j * 4 + i];
idct4(temp_in, temp_out);
- for (j = 0; j < 4; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 4)
- + dest[j * stride + i]);
+ for (j = 0; j < 4; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 4));
+ }
}
}
@@ -177,15 +177,15 @@
int dest_stride) {
int i;
tran_high_t a1;
- tran_low_t out = dct_const_round_shift(input[0] * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
+ tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), 8);
+ out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), 8);
a1 = ROUND_POWER_OF_TWO(out, 4);
for (i = 0; i < 4; i++) {
- dest[0] = clip_pixel(dest[0] + a1);
- dest[1] = clip_pixel(dest[1] + a1);
- dest[2] = clip_pixel(dest[2] + a1);
- dest[3] = clip_pixel(dest[3] + a1);
+ dest[0] = clip_pixel_add(dest[0], a1);
+ dest[1] = clip_pixel_add(dest[1], a1);
+ dest[2] = clip_pixel_add(dest[2], a1);
+ dest[3] = clip_pixel_add(dest[3], a1);
dest += dest_stride;
}
}
@@ -200,39 +200,39 @@
step1[3] = input[6];
temp1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
temp2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
- step1[4] = dct_const_round_shift(temp1);
- step1[7] = dct_const_round_shift(temp2);
+ step1[4] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[7] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
temp2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
// stage 2 & stage 3 - even half
idct4(step1, step1);
// stage 2 - odd half
- step2[4] = step1[4] + step1[5];
- step2[5] = step1[4] - step1[5];
- step2[6] = -step1[6] + step1[7];
- step2[7] = step1[6] + step1[7];
+ step2[4] = WRAPLOW(step1[4] + step1[5], 8);
+ step2[5] = WRAPLOW(step1[4] - step1[5], 8);
+ step2[6] = WRAPLOW(-step1[6] + step1[7], 8);
+ step2[7] = WRAPLOW(step1[6] + step1[7], 8);
// stage 3 -odd half
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[7] = step2[7];
// stage 4
- output[0] = step1[0] + step1[7];
- output[1] = step1[1] + step1[6];
- output[2] = step1[2] + step1[5];
- output[3] = step1[3] + step1[4];
- output[4] = step1[3] - step1[4];
- output[5] = step1[2] - step1[5];
- output[6] = step1[1] - step1[6];
- output[7] = step1[0] - step1[7];
+ output[0] = WRAPLOW(step1[0] + step1[7], 8);
+ output[1] = WRAPLOW(step1[1] + step1[6], 8);
+ output[2] = WRAPLOW(step1[2] + step1[5], 8);
+ output[3] = WRAPLOW(step1[3] + step1[4], 8);
+ output[4] = WRAPLOW(step1[3] - step1[4], 8);
+ output[5] = WRAPLOW(step1[2] - step1[5], 8);
+ output[6] = WRAPLOW(step1[1] - step1[6], 8);
+ output[7] = WRAPLOW(step1[0] - step1[7], 8);
}
void vp9_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
@@ -253,21 +253,22 @@
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
idct8(temp_in, temp_out);
- for (j = 0; j < 8; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 5)
- + dest[j * stride + i]);
+ for (j = 0; j < 8; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 5));
+ }
}
}
void vp9_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
int i, j;
tran_high_t a1;
- tran_low_t out = dct_const_round_shift(input[0] * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
+ tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), 8);
+ out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), 8);
a1 = ROUND_POWER_OF_TWO(out, 5);
for (j = 0; j < 8; ++j) {
for (i = 0; i < 8; ++i)
- dest[i] = clip_pixel(dest[i] + a1);
+ dest[i] = clip_pixel_add(dest[i], a1);
dest += stride;
}
}
@@ -308,10 +309,10 @@
// The overall dynamic range is 14b (input) + 14b (multiplication scaling)
// + 1b (addition) = 29b.
// Hence the output bit depth is 15b.
- output[0] = dct_const_round_shift(s0);
- output[1] = dct_const_round_shift(s1);
- output[2] = dct_const_round_shift(s2);
- output[3] = dct_const_round_shift(s3);
+ output[0] = WRAPLOW(dct_const_round_shift(s0), 8);
+ output[1] = WRAPLOW(dct_const_round_shift(s1), 8);
+ output[2] = WRAPLOW(dct_const_round_shift(s2), 8);
+ output[3] = WRAPLOW(dct_const_round_shift(s3), 8);
}
void vp9_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride,
@@ -340,11 +341,13 @@
for (j = 0; j < 4; ++j)
temp_in[j] = out[j * 4 + i];
IHT_4[tx_type].cols(temp_in, temp_out);
- for (j = 0; j < 4; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 4)
- + dest[j * stride + i]);
+ for (j = 0; j < 4; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 4));
+ }
}
}
+
static void iadst8(const tran_low_t *input, tran_low_t *output) {
int s0, s1, s2, s3, s4, s5, s6, s7;
@@ -373,14 +376,14 @@
s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
- x0 = dct_const_round_shift(s0 + s4);
- x1 = dct_const_round_shift(s1 + s5);
- x2 = dct_const_round_shift(s2 + s6);
- x3 = dct_const_round_shift(s3 + s7);
- x4 = dct_const_round_shift(s0 - s4);
- x5 = dct_const_round_shift(s1 - s5);
- x6 = dct_const_round_shift(s2 - s6);
- x7 = dct_const_round_shift(s3 - s7);
+ x0 = WRAPLOW(dct_const_round_shift(s0 + s4), 8);
+ x1 = WRAPLOW(dct_const_round_shift(s1 + s5), 8);
+ x2 = WRAPLOW(dct_const_round_shift(s2 + s6), 8);
+ x3 = WRAPLOW(dct_const_round_shift(s3 + s7), 8);
+ x4 = WRAPLOW(dct_const_round_shift(s0 - s4), 8);
+ x5 = WRAPLOW(dct_const_round_shift(s1 - s5), 8);
+ x6 = WRAPLOW(dct_const_round_shift(s2 - s6), 8);
+ x7 = WRAPLOW(dct_const_round_shift(s3 - s7), 8);
// stage 2
s0 = x0;
@@ -392,14 +395,14 @@
s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
- x0 = s0 + s2;
- x1 = s1 + s3;
- x2 = s0 - s2;
- x3 = s1 - s3;
- x4 = dct_const_round_shift(s4 + s6);
- x5 = dct_const_round_shift(s5 + s7);
- x6 = dct_const_round_shift(s4 - s6);
- x7 = dct_const_round_shift(s5 - s7);
+ x0 = WRAPLOW(s0 + s2, 8);
+ x1 = WRAPLOW(s1 + s3, 8);
+ x2 = WRAPLOW(s0 - s2, 8);
+ x3 = WRAPLOW(s1 - s3, 8);
+ x4 = WRAPLOW(dct_const_round_shift(s4 + s6), 8);
+ x5 = WRAPLOW(dct_const_round_shift(s5 + s7), 8);
+ x6 = WRAPLOW(dct_const_round_shift(s4 - s6), 8);
+ x7 = WRAPLOW(dct_const_round_shift(s5 - s7), 8);
// stage 3
s2 = cospi_16_64 * (x2 + x3);
@@ -407,19 +410,19 @@
s6 = cospi_16_64 * (x6 + x7);
s7 = cospi_16_64 * (x6 - x7);
- x2 = dct_const_round_shift(s2);
- x3 = dct_const_round_shift(s3);
- x6 = dct_const_round_shift(s6);
- x7 = dct_const_round_shift(s7);
+ x2 = WRAPLOW(dct_const_round_shift(s2), 8);
+ x3 = WRAPLOW(dct_const_round_shift(s3), 8);
+ x6 = WRAPLOW(dct_const_round_shift(s6), 8);
+ x7 = WRAPLOW(dct_const_round_shift(s7), 8);
- output[0] = x0;
- output[1] = -x4;
- output[2] = x6;
- output[3] = -x2;
- output[4] = x3;
- output[5] = -x7;
- output[6] = x5;
- output[7] = -x1;
+ output[0] = WRAPLOW(x0, 8);
+ output[1] = WRAPLOW(-x4, 8);
+ output[2] = WRAPLOW(x6, 8);
+ output[3] = WRAPLOW(-x2, 8);
+ output[4] = WRAPLOW(x3, 8);
+ output[5] = WRAPLOW(-x7, 8);
+ output[6] = WRAPLOW(x5, 8);
+ output[7] = WRAPLOW(-x1, 8);
}
static const transform_2d IHT_8[] = {
@@ -449,9 +452,10 @@
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
ht.cols(temp_in, temp_out);
- for (j = 0; j < 8; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 5)
- + dest[j * stride + i]);
+ for (j = 0; j < 8; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 5));
+ }
}
}
@@ -474,9 +478,10 @@
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
idct8(temp_in, temp_out);
- for (j = 0; j < 8; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 5)
- + dest[j * stride + i]);
+ for (j = 0; j < 8; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 5));
+ }
}
}
@@ -514,23 +519,23 @@
temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
- step2[8] = dct_const_round_shift(temp1);
- step2[15] = dct_const_round_shift(temp2);
+ step2[8] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[15] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
- step2[9] = dct_const_round_shift(temp1);
- step2[14] = dct_const_round_shift(temp2);
+ step2[9] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[14] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
- step2[11] = dct_const_round_shift(temp1);
- step2[12] = dct_const_round_shift(temp2);
+ step2[11] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[12] = WRAPLOW(dct_const_round_shift(temp2), 8);
// stage 3
step1[0] = step2[0];
@@ -540,109 +545,109 @@
temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
- step1[4] = dct_const_round_shift(temp1);
- step1[7] = dct_const_round_shift(temp2);
+ step1[4] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[7] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
- step1[8] = step2[8] + step2[9];
- step1[9] = step2[8] - step2[9];
- step1[10] = -step2[10] + step2[11];
- step1[11] = step2[10] + step2[11];
- step1[12] = step2[12] + step2[13];
- step1[13] = step2[12] - step2[13];
- step1[14] = -step2[14] + step2[15];
- step1[15] = step2[14] + step2[15];
+ step1[8] = WRAPLOW(step2[8] + step2[9], 8);
+ step1[9] = WRAPLOW(step2[8] - step2[9], 8);
+ step1[10] = WRAPLOW(-step2[10] + step2[11], 8);
+ step1[11] = WRAPLOW(step2[10] + step2[11], 8);
+ step1[12] = WRAPLOW(step2[12] + step2[13], 8);
+ step1[13] = WRAPLOW(step2[12] - step2[13], 8);
+ step1[14] = WRAPLOW(-step2[14] + step2[15], 8);
+ step1[15] = WRAPLOW(step2[14] + step2[15], 8);
// stage 4
temp1 = (step1[0] + step1[1]) * cospi_16_64;
temp2 = (step1[0] - step1[1]) * cospi_16_64;
- step2[0] = dct_const_round_shift(temp1);
- step2[1] = dct_const_round_shift(temp2);
+ step2[0] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[1] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
- step2[2] = dct_const_round_shift(temp1);
- step2[3] = dct_const_round_shift(temp2);
- step2[4] = step1[4] + step1[5];
- step2[5] = step1[4] - step1[5];
- step2[6] = -step1[6] + step1[7];
- step2[7] = step1[6] + step1[7];
+ step2[2] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[3] = WRAPLOW(dct_const_round_shift(temp2), 8);
+ step2[4] = WRAPLOW(step1[4] + step1[5], 8);
+ step2[5] = WRAPLOW(step1[4] - step1[5], 8);
+ step2[6] = WRAPLOW(-step1[6] + step1[7], 8);
+ step2[7] = WRAPLOW(step1[6] + step1[7], 8);
step2[8] = step1[8];
step2[15] = step1[15];
temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
- step2[9] = dct_const_round_shift(temp1);
- step2[14] = dct_const_round_shift(temp2);
+ step2[9] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[14] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
step2[11] = step1[11];
step2[12] = step1[12];
// stage 5
- step1[0] = step2[0] + step2[3];
- step1[1] = step2[1] + step2[2];
- step1[2] = step2[1] - step2[2];
- step1[3] = step2[0] - step2[3];
+ step1[0] = WRAPLOW(step2[0] + step2[3], 8);
+ step1[1] = WRAPLOW(step2[1] + step2[2], 8);
+ step1[2] = WRAPLOW(step2[1] - step2[2], 8);
+ step1[3] = WRAPLOW(step2[0] - step2[3], 8);
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[7] = step2[7];
- step1[8] = step2[8] + step2[11];
- step1[9] = step2[9] + step2[10];
- step1[10] = step2[9] - step2[10];
- step1[11] = step2[8] - step2[11];
- step1[12] = -step2[12] + step2[15];
- step1[13] = -step2[13] + step2[14];
- step1[14] = step2[13] + step2[14];
- step1[15] = step2[12] + step2[15];
+ step1[8] = WRAPLOW(step2[8] + step2[11], 8);
+ step1[9] = WRAPLOW(step2[9] + step2[10], 8);
+ step1[10] = WRAPLOW(step2[9] - step2[10], 8);
+ step1[11] = WRAPLOW(step2[8] - step2[11], 8);
+ step1[12] = WRAPLOW(-step2[12] + step2[15], 8);
+ step1[13] = WRAPLOW(-step2[13] + step2[14], 8);
+ step1[14] = WRAPLOW(step2[13] + step2[14], 8);
+ step1[15] = WRAPLOW(step2[12] + step2[15], 8);
// stage 6
- step2[0] = step1[0] + step1[7];
- step2[1] = step1[1] + step1[6];
- step2[2] = step1[2] + step1[5];
- step2[3] = step1[3] + step1[4];
- step2[4] = step1[3] - step1[4];
- step2[5] = step1[2] - step1[5];
- step2[6] = step1[1] - step1[6];
- step2[7] = step1[0] - step1[7];
+ step2[0] = WRAPLOW(step1[0] + step1[7], 8);
+ step2[1] = WRAPLOW(step1[1] + step1[6], 8);
+ step2[2] = WRAPLOW(step1[2] + step1[5], 8);
+ step2[3] = WRAPLOW(step1[3] + step1[4], 8);
+ step2[4] = WRAPLOW(step1[3] - step1[4], 8);
+ step2[5] = WRAPLOW(step1[2] - step1[5], 8);
+ step2[6] = WRAPLOW(step1[1] - step1[6], 8);
+ step2[7] = WRAPLOW(step1[0] - step1[7], 8);
step2[8] = step1[8];
step2[9] = step1[9];
temp1 = (-step1[10] + step1[13]) * cospi_16_64;
temp2 = (step1[10] + step1[13]) * cospi_16_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = (-step1[11] + step1[12]) * cospi_16_64;
temp2 = (step1[11] + step1[12]) * cospi_16_64;
- step2[11] = dct_const_round_shift(temp1);
- step2[12] = dct_const_round_shift(temp2);
+ step2[11] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[12] = WRAPLOW(dct_const_round_shift(temp2), 8);
step2[14] = step1[14];
step2[15] = step1[15];
// stage 7
- output[0] = step2[0] + step2[15];
- output[1] = step2[1] + step2[14];
- output[2] = step2[2] + step2[13];
- output[3] = step2[3] + step2[12];
- output[4] = step2[4] + step2[11];
- output[5] = step2[5] + step2[10];
- output[6] = step2[6] + step2[9];
- output[7] = step2[7] + step2[8];
- output[8] = step2[7] - step2[8];
- output[9] = step2[6] - step2[9];
- output[10] = step2[5] - step2[10];
- output[11] = step2[4] - step2[11];
- output[12] = step2[3] - step2[12];
- output[13] = step2[2] - step2[13];
- output[14] = step2[1] - step2[14];
- output[15] = step2[0] - step2[15];
+ output[0] = WRAPLOW(step2[0] + step2[15], 8);
+ output[1] = WRAPLOW(step2[1] + step2[14], 8);
+ output[2] = WRAPLOW(step2[2] + step2[13], 8);
+ output[3] = WRAPLOW(step2[3] + step2[12], 8);
+ output[4] = WRAPLOW(step2[4] + step2[11], 8);
+ output[5] = WRAPLOW(step2[5] + step2[10], 8);
+ output[6] = WRAPLOW(step2[6] + step2[9], 8);
+ output[7] = WRAPLOW(step2[7] + step2[8], 8);
+ output[8] = WRAPLOW(step2[7] - step2[8], 8);
+ output[9] = WRAPLOW(step2[6] - step2[9], 8);
+ output[10] = WRAPLOW(step2[5] - step2[10], 8);
+ output[11] = WRAPLOW(step2[4] - step2[11], 8);
+ output[12] = WRAPLOW(step2[3] - step2[12], 8);
+ output[13] = WRAPLOW(step2[2] - step2[13], 8);
+ output[14] = WRAPLOW(step2[1] - step2[14], 8);
+ output[15] = WRAPLOW(step2[0] - step2[15], 8);
}
void vp9_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest,
@@ -664,9 +669,10 @@
for (j = 0; j < 16; ++j)
temp_in[j] = out[j * 16 + i];
idct16(temp_in, temp_out);
- for (j = 0; j < 16; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
- + dest[j * stride + i]);
+ for (j = 0; j < 16; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 6));
+ }
}
}
@@ -718,22 +724,22 @@
s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
- x0 = dct_const_round_shift(s0 + s8);
- x1 = dct_const_round_shift(s1 + s9);
- x2 = dct_const_round_shift(s2 + s10);
- x3 = dct_const_round_shift(s3 + s11);
- x4 = dct_const_round_shift(s4 + s12);
- x5 = dct_const_round_shift(s5 + s13);
- x6 = dct_const_round_shift(s6 + s14);
- x7 = dct_const_round_shift(s7 + s15);
- x8 = dct_const_round_shift(s0 - s8);
- x9 = dct_const_round_shift(s1 - s9);
- x10 = dct_const_round_shift(s2 - s10);
- x11 = dct_const_round_shift(s3 - s11);
- x12 = dct_const_round_shift(s4 - s12);
- x13 = dct_const_round_shift(s5 - s13);
- x14 = dct_const_round_shift(s6 - s14);
- x15 = dct_const_round_shift(s7 - s15);
+ x0 = WRAPLOW(dct_const_round_shift(s0 + s8), 8);
+ x1 = WRAPLOW(dct_const_round_shift(s1 + s9), 8);
+ x2 = WRAPLOW(dct_const_round_shift(s2 + s10), 8);
+ x3 = WRAPLOW(dct_const_round_shift(s3 + s11), 8);
+ x4 = WRAPLOW(dct_const_round_shift(s4 + s12), 8);
+ x5 = WRAPLOW(dct_const_round_shift(s5 + s13), 8);
+ x6 = WRAPLOW(dct_const_round_shift(s6 + s14), 8);
+ x7 = WRAPLOW(dct_const_round_shift(s7 + s15), 8);
+ x8 = WRAPLOW(dct_const_round_shift(s0 - s8), 8);
+ x9 = WRAPLOW(dct_const_round_shift(s1 - s9), 8);
+ x10 = WRAPLOW(dct_const_round_shift(s2 - s10), 8);
+ x11 = WRAPLOW(dct_const_round_shift(s3 - s11), 8);
+ x12 = WRAPLOW(dct_const_round_shift(s4 - s12), 8);
+ x13 = WRAPLOW(dct_const_round_shift(s5 - s13), 8);
+ x14 = WRAPLOW(dct_const_round_shift(s6 - s14), 8);
+ x15 = WRAPLOW(dct_const_round_shift(s7 - s15), 8);
// stage 2
s0 = x0;
@@ -753,22 +759,22 @@
s14 = - x14 * cospi_12_64 + x15 * cospi_20_64;
s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
- x0 = s0 + s4;
- x1 = s1 + s5;
- x2 = s2 + s6;
- x3 = s3 + s7;
- x4 = s0 - s4;
- x5 = s1 - s5;
- x6 = s2 - s6;
- x7 = s3 - s7;
- x8 = dct_const_round_shift(s8 + s12);
- x9 = dct_const_round_shift(s9 + s13);
- x10 = dct_const_round_shift(s10 + s14);
- x11 = dct_const_round_shift(s11 + s15);
- x12 = dct_const_round_shift(s8 - s12);
- x13 = dct_const_round_shift(s9 - s13);
- x14 = dct_const_round_shift(s10 - s14);
- x15 = dct_const_round_shift(s11 - s15);
+ x0 = WRAPLOW(s0 + s4, 8);
+ x1 = WRAPLOW(s1 + s5, 8);
+ x2 = WRAPLOW(s2 + s6, 8);
+ x3 = WRAPLOW(s3 + s7, 8);
+ x4 = WRAPLOW(s0 - s4, 8);
+ x5 = WRAPLOW(s1 - s5, 8);
+ x6 = WRAPLOW(s2 - s6, 8);
+ x7 = WRAPLOW(s3 - s7, 8);
+ x8 = WRAPLOW(dct_const_round_shift(s8 + s12), 8);
+ x9 = WRAPLOW(dct_const_round_shift(s9 + s13), 8);
+ x10 = WRAPLOW(dct_const_round_shift(s10 + s14), 8);
+ x11 = WRAPLOW(dct_const_round_shift(s11 + s15), 8);
+ x12 = WRAPLOW(dct_const_round_shift(s8 - s12), 8);
+ x13 = WRAPLOW(dct_const_round_shift(s9 - s13), 8);
+ x14 = WRAPLOW(dct_const_round_shift(s10 - s14), 8);
+ x15 = WRAPLOW(dct_const_round_shift(s11 - s15), 8);
// stage 3
s0 = x0;
@@ -788,22 +794,22 @@
s14 = - x14 * cospi_24_64 + x15 * cospi_8_64;
s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
- x0 = s0 + s2;
- x1 = s1 + s3;
- x2 = s0 - s2;
- x3 = s1 - s3;
- x4 = dct_const_round_shift(s4 + s6);
- x5 = dct_const_round_shift(s5 + s7);
- x6 = dct_const_round_shift(s4 - s6);
- x7 = dct_const_round_shift(s5 - s7);
- x8 = s8 + s10;
- x9 = s9 + s11;
- x10 = s8 - s10;
- x11 = s9 - s11;
- x12 = dct_const_round_shift(s12 + s14);
- x13 = dct_const_round_shift(s13 + s15);
- x14 = dct_const_round_shift(s12 - s14);
- x15 = dct_const_round_shift(s13 - s15);
+ x0 = WRAPLOW(check_range(s0 + s2), 8);
+ x1 = WRAPLOW(check_range(s1 + s3), 8);
+ x2 = WRAPLOW(check_range(s0 - s2), 8);
+ x3 = WRAPLOW(check_range(s1 - s3), 8);
+ x4 = WRAPLOW(dct_const_round_shift(s4 + s6), 8);
+ x5 = WRAPLOW(dct_const_round_shift(s5 + s7), 8);
+ x6 = WRAPLOW(dct_const_round_shift(s4 - s6), 8);
+ x7 = WRAPLOW(dct_const_round_shift(s5 - s7), 8);
+ x8 = WRAPLOW(check_range(s8 + s10), 8);
+ x9 = WRAPLOW(check_range(s9 + s11), 8);
+ x10 = WRAPLOW(check_range(s8 - s10), 8);
+ x11 = WRAPLOW(check_range(s9 - s11), 8);
+ x12 = WRAPLOW(dct_const_round_shift(s12 + s14), 8);
+ x13 = WRAPLOW(dct_const_round_shift(s13 + s15), 8);
+ x14 = WRAPLOW(dct_const_round_shift(s12 - s14), 8);
+ x15 = WRAPLOW(dct_const_round_shift(s13 - s15), 8);
// stage 4
s2 = (- cospi_16_64) * (x2 + x3);
@@ -815,31 +821,31 @@
s14 = (- cospi_16_64) * (x14 + x15);
s15 = cospi_16_64 * (x14 - x15);
- x2 = dct_const_round_shift(s2);
- x3 = dct_const_round_shift(s3);
- x6 = dct_const_round_shift(s6);
- x7 = dct_const_round_shift(s7);
- x10 = dct_const_round_shift(s10);
- x11 = dct_const_round_shift(s11);
- x14 = dct_const_round_shift(s14);
- x15 = dct_const_round_shift(s15);
+ x2 = WRAPLOW(dct_const_round_shift(s2), 8);
+ x3 = WRAPLOW(dct_const_round_shift(s3), 8);
+ x6 = WRAPLOW(dct_const_round_shift(s6), 8);
+ x7 = WRAPLOW(dct_const_round_shift(s7), 8);
+ x10 = WRAPLOW(dct_const_round_shift(s10), 8);
+ x11 = WRAPLOW(dct_const_round_shift(s11), 8);
+ x14 = WRAPLOW(dct_const_round_shift(s14), 8);
+ x15 = WRAPLOW(dct_const_round_shift(s15), 8);
- output[0] = x0;
- output[1] = -x8;
- output[2] = x12;
- output[3] = -x4;
- output[4] = x6;
- output[5] = x14;
- output[6] = x10;
- output[7] = x2;
- output[8] = x3;
- output[9] = x11;
- output[10] = x15;
- output[11] = x7;
- output[12] = x5;
- output[13] = -x13;
- output[14] = x9;
- output[15] = -x1;
+ output[0] = WRAPLOW(x0, 8);
+ output[1] = WRAPLOW(-x8, 8);
+ output[2] = WRAPLOW(x12, 8);
+ output[3] = WRAPLOW(-x4, 8);
+ output[4] = WRAPLOW(x6, 8);
+ output[5] = WRAPLOW(x14, 8);
+ output[6] = WRAPLOW(x10, 8);
+ output[7] = WRAPLOW(x2, 8);
+ output[8] = WRAPLOW(x3, 8);
+ output[9] = WRAPLOW(x11, 8);
+ output[10] = WRAPLOW(x15, 8);
+ output[11] = WRAPLOW(x7, 8);
+ output[12] = WRAPLOW(x5, 8);
+ output[13] = WRAPLOW(-x13, 8);
+ output[14] = WRAPLOW(x9, 8);
+ output[15] = WRAPLOW(-x1, 8);
}
static const transform_2d IHT_16[] = {
@@ -869,9 +875,10 @@
for (j = 0; j < 16; ++j)
temp_in[j] = out[j * 16 + i];
ht.cols(temp_in, temp_out);
- for (j = 0; j < 16; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
- + dest[j * stride + i]);
+ for (j = 0; j < 16; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 6));
+ }
}
}
@@ -895,21 +902,22 @@
for (j = 0; j < 16; ++j)
temp_in[j] = out[j*16 + i];
idct16(temp_in, temp_out);
- for (j = 0; j < 16; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
- + dest[j * stride + i]);
+ for (j = 0; j < 16; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 6));
+ }
}
}
void vp9_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
int i, j;
tran_high_t a1;
- tran_low_t out = dct_const_round_shift(input[0] * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
+ tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), 8);
+ out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), 8);
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 16; ++j) {
for (i = 0; i < 16; ++i)
- dest[i] = clip_pixel(dest[i] + a1);
+ dest[i] = clip_pixel_add(dest[i], a1);
dest += stride;
}
}
@@ -938,43 +946,43 @@
temp1 = input[1] * cospi_31_64 - input[31] * cospi_1_64;
temp2 = input[1] * cospi_1_64 + input[31] * cospi_31_64;
- step1[16] = dct_const_round_shift(temp1);
- step1[31] = dct_const_round_shift(temp2);
+ step1[16] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[31] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[17] * cospi_15_64 - input[15] * cospi_17_64;
temp2 = input[17] * cospi_17_64 + input[15] * cospi_15_64;
- step1[17] = dct_const_round_shift(temp1);
- step1[30] = dct_const_round_shift(temp2);
+ step1[17] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[30] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[9] * cospi_23_64 - input[23] * cospi_9_64;
temp2 = input[9] * cospi_9_64 + input[23] * cospi_23_64;
- step1[18] = dct_const_round_shift(temp1);
- step1[29] = dct_const_round_shift(temp2);
+ step1[18] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[29] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[25] * cospi_7_64 - input[7] * cospi_25_64;
temp2 = input[25] * cospi_25_64 + input[7] * cospi_7_64;
- step1[19] = dct_const_round_shift(temp1);
- step1[28] = dct_const_round_shift(temp2);
+ step1[19] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[28] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[5] * cospi_27_64 - input[27] * cospi_5_64;
temp2 = input[5] * cospi_5_64 + input[27] * cospi_27_64;
- step1[20] = dct_const_round_shift(temp1);
- step1[27] = dct_const_round_shift(temp2);
+ step1[20] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[27] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[21] * cospi_11_64 - input[11] * cospi_21_64;
temp2 = input[21] * cospi_21_64 + input[11] * cospi_11_64;
- step1[21] = dct_const_round_shift(temp1);
- step1[26] = dct_const_round_shift(temp2);
+ step1[21] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[26] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[13] * cospi_19_64 - input[19] * cospi_13_64;
temp2 = input[13] * cospi_13_64 + input[19] * cospi_19_64;
- step1[22] = dct_const_round_shift(temp1);
- step1[25] = dct_const_round_shift(temp2);
+ step1[22] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[25] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[29] * cospi_3_64 - input[3] * cospi_29_64;
temp2 = input[29] * cospi_29_64 + input[3] * cospi_3_64;
- step1[23] = dct_const_round_shift(temp1);
- step1[24] = dct_const_round_shift(temp2);
+ step1[23] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[24] = WRAPLOW(dct_const_round_shift(temp2), 8);
// stage 2
step2[0] = step1[0];
@@ -988,40 +996,40 @@
temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
- step2[8] = dct_const_round_shift(temp1);
- step2[15] = dct_const_round_shift(temp2);
+ step2[8] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[15] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
- step2[9] = dct_const_round_shift(temp1);
- step2[14] = dct_const_round_shift(temp2);
+ step2[9] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[14] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
- step2[11] = dct_const_round_shift(temp1);
- step2[12] = dct_const_round_shift(temp2);
+ step2[11] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[12] = WRAPLOW(dct_const_round_shift(temp2), 8);
- step2[16] = step1[16] + step1[17];
- step2[17] = step1[16] - step1[17];
- step2[18] = -step1[18] + step1[19];
- step2[19] = step1[18] + step1[19];
- step2[20] = step1[20] + step1[21];
- step2[21] = step1[20] - step1[21];
- step2[22] = -step1[22] + step1[23];
- step2[23] = step1[22] + step1[23];
- step2[24] = step1[24] + step1[25];
- step2[25] = step1[24] - step1[25];
- step2[26] = -step1[26] + step1[27];
- step2[27] = step1[26] + step1[27];
- step2[28] = step1[28] + step1[29];
- step2[29] = step1[28] - step1[29];
- step2[30] = -step1[30] + step1[31];
- step2[31] = step1[30] + step1[31];
+ step2[16] = WRAPLOW(step1[16] + step1[17], 8);
+ step2[17] = WRAPLOW(step1[16] - step1[17], 8);
+ step2[18] = WRAPLOW(-step1[18] + step1[19], 8);
+ step2[19] = WRAPLOW(step1[18] + step1[19], 8);
+ step2[20] = WRAPLOW(step1[20] + step1[21], 8);
+ step2[21] = WRAPLOW(step1[20] - step1[21], 8);
+ step2[22] = WRAPLOW(-step1[22] + step1[23], 8);
+ step2[23] = WRAPLOW(step1[22] + step1[23], 8);
+ step2[24] = WRAPLOW(step1[24] + step1[25], 8);
+ step2[25] = WRAPLOW(step1[24] - step1[25], 8);
+ step2[26] = WRAPLOW(-step1[26] + step1[27], 8);
+ step2[27] = WRAPLOW(step1[26] + step1[27], 8);
+ step2[28] = WRAPLOW(step1[28] + step1[29], 8);
+ step2[29] = WRAPLOW(step1[28] - step1[29], 8);
+ step2[30] = WRAPLOW(-step1[30] + step1[31], 8);
+ step2[31] = WRAPLOW(step1[30] + step1[31], 8);
// stage 3
step1[0] = step2[0];
@@ -1031,42 +1039,42 @@
temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
- step1[4] = dct_const_round_shift(temp1);
- step1[7] = dct_const_round_shift(temp2);
+ step1[4] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[7] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
- step1[8] = step2[8] + step2[9];
- step1[9] = step2[8] - step2[9];
- step1[10] = -step2[10] + step2[11];
- step1[11] = step2[10] + step2[11];
- step1[12] = step2[12] + step2[13];
- step1[13] = step2[12] - step2[13];
- step1[14] = -step2[14] + step2[15];
- step1[15] = step2[14] + step2[15];
+ step1[8] = WRAPLOW(step2[8] + step2[9], 8);
+ step1[9] = WRAPLOW(step2[8] - step2[9], 8);
+ step1[10] = WRAPLOW(-step2[10] + step2[11], 8);
+ step1[11] = WRAPLOW(step2[10] + step2[11], 8);
+ step1[12] = WRAPLOW(step2[12] + step2[13], 8);
+ step1[13] = WRAPLOW(step2[12] - step2[13], 8);
+ step1[14] = WRAPLOW(-step2[14] + step2[15], 8);
+ step1[15] = WRAPLOW(step2[14] + step2[15], 8);
step1[16] = step2[16];
step1[31] = step2[31];
temp1 = -step2[17] * cospi_4_64 + step2[30] * cospi_28_64;
temp2 = step2[17] * cospi_28_64 + step2[30] * cospi_4_64;
- step1[17] = dct_const_round_shift(temp1);
- step1[30] = dct_const_round_shift(temp2);
+ step1[17] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[30] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = -step2[18] * cospi_28_64 - step2[29] * cospi_4_64;
temp2 = -step2[18] * cospi_4_64 + step2[29] * cospi_28_64;
- step1[18] = dct_const_round_shift(temp1);
- step1[29] = dct_const_round_shift(temp2);
+ step1[18] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[29] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[19] = step2[19];
step1[20] = step2[20];
temp1 = -step2[21] * cospi_20_64 + step2[26] * cospi_12_64;
temp2 = step2[21] * cospi_12_64 + step2[26] * cospi_20_64;
- step1[21] = dct_const_round_shift(temp1);
- step1[26] = dct_const_round_shift(temp2);
+ step1[21] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[26] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = -step2[22] * cospi_12_64 - step2[25] * cospi_20_64;
temp2 = -step2[22] * cospi_20_64 + step2[25] * cospi_12_64;
- step1[22] = dct_const_round_shift(temp1);
- step1[25] = dct_const_round_shift(temp2);
+ step1[22] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[25] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[23] = step2[23];
step1[24] = step2[24];
step1[27] = step2[27];
@@ -1075,87 +1083,87 @@
// stage 4
temp1 = (step1[0] + step1[1]) * cospi_16_64;
temp2 = (step1[0] - step1[1]) * cospi_16_64;
- step2[0] = dct_const_round_shift(temp1);
- step2[1] = dct_const_round_shift(temp2);
+ step2[0] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[1] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
- step2[2] = dct_const_round_shift(temp1);
- step2[3] = dct_const_round_shift(temp2);
- step2[4] = step1[4] + step1[5];
- step2[5] = step1[4] - step1[5];
- step2[6] = -step1[6] + step1[7];
- step2[7] = step1[6] + step1[7];
+ step2[2] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[3] = WRAPLOW(dct_const_round_shift(temp2), 8);
+ step2[4] = WRAPLOW(step1[4] + step1[5], 8);
+ step2[5] = WRAPLOW(step1[4] - step1[5], 8);
+ step2[6] = WRAPLOW(-step1[6] + step1[7], 8);
+ step2[7] = WRAPLOW(step1[6] + step1[7], 8);
step2[8] = step1[8];
step2[15] = step1[15];
temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
- step2[9] = dct_const_round_shift(temp1);
- step2[14] = dct_const_round_shift(temp2);
+ step2[9] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[14] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
step2[11] = step1[11];
step2[12] = step1[12];
- step2[16] = step1[16] + step1[19];
- step2[17] = step1[17] + step1[18];
- step2[18] = step1[17] - step1[18];
- step2[19] = step1[16] - step1[19];
- step2[20] = -step1[20] + step1[23];
- step2[21] = -step1[21] + step1[22];
- step2[22] = step1[21] + step1[22];
- step2[23] = step1[20] + step1[23];
+ step2[16] = WRAPLOW(step1[16] + step1[19], 8);
+ step2[17] = WRAPLOW(step1[17] + step1[18], 8);
+ step2[18] = WRAPLOW(step1[17] - step1[18], 8);
+ step2[19] = WRAPLOW(step1[16] - step1[19], 8);
+ step2[20] = WRAPLOW(-step1[20] + step1[23], 8);
+ step2[21] = WRAPLOW(-step1[21] + step1[22], 8);
+ step2[22] = WRAPLOW(step1[21] + step1[22], 8);
+ step2[23] = WRAPLOW(step1[20] + step1[23], 8);
- step2[24] = step1[24] + step1[27];
- step2[25] = step1[25] + step1[26];
- step2[26] = step1[25] - step1[26];
- step2[27] = step1[24] - step1[27];
- step2[28] = -step1[28] + step1[31];
- step2[29] = -step1[29] + step1[30];
- step2[30] = step1[29] + step1[30];
- step2[31] = step1[28] + step1[31];
+ step2[24] = WRAPLOW(step1[24] + step1[27], 8);
+ step2[25] = WRAPLOW(step1[25] + step1[26], 8);
+ step2[26] = WRAPLOW(step1[25] - step1[26], 8);
+ step2[27] = WRAPLOW(step1[24] - step1[27], 8);
+ step2[28] = WRAPLOW(-step1[28] + step1[31], 8);
+ step2[29] = WRAPLOW(-step1[29] + step1[30], 8);
+ step2[30] = WRAPLOW(step1[29] + step1[30], 8);
+ step2[31] = WRAPLOW(step1[28] + step1[31], 8);
// stage 5
- step1[0] = step2[0] + step2[3];
- step1[1] = step2[1] + step2[2];
- step1[2] = step2[1] - step2[2];
- step1[3] = step2[0] - step2[3];
+ step1[0] = WRAPLOW(step2[0] + step2[3], 8);
+ step1[1] = WRAPLOW(step2[1] + step2[2], 8);
+ step1[2] = WRAPLOW(step2[1] - step2[2], 8);
+ step1[3] = WRAPLOW(step2[0] - step2[3], 8);
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[7] = step2[7];
- step1[8] = step2[8] + step2[11];
- step1[9] = step2[9] + step2[10];
- step1[10] = step2[9] - step2[10];
- step1[11] = step2[8] - step2[11];
- step1[12] = -step2[12] + step2[15];
- step1[13] = -step2[13] + step2[14];
- step1[14] = step2[13] + step2[14];
- step1[15] = step2[12] + step2[15];
+ step1[8] = WRAPLOW(step2[8] + step2[11], 8);
+ step1[9] = WRAPLOW(step2[9] + step2[10], 8);
+ step1[10] = WRAPLOW(step2[9] - step2[10], 8);
+ step1[11] = WRAPLOW(step2[8] - step2[11], 8);
+ step1[12] = WRAPLOW(-step2[12] + step2[15], 8);
+ step1[13] = WRAPLOW(-step2[13] + step2[14], 8);
+ step1[14] = WRAPLOW(step2[13] + step2[14], 8);
+ step1[15] = WRAPLOW(step2[12] + step2[15], 8);
step1[16] = step2[16];
step1[17] = step2[17];
temp1 = -step2[18] * cospi_8_64 + step2[29] * cospi_24_64;
temp2 = step2[18] * cospi_24_64 + step2[29] * cospi_8_64;
- step1[18] = dct_const_round_shift(temp1);
- step1[29] = dct_const_round_shift(temp2);
+ step1[18] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[29] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = -step2[19] * cospi_8_64 + step2[28] * cospi_24_64;
temp2 = step2[19] * cospi_24_64 + step2[28] * cospi_8_64;
- step1[19] = dct_const_round_shift(temp1);
- step1[28] = dct_const_round_shift(temp2);
+ step1[19] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[28] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = -step2[20] * cospi_24_64 - step2[27] * cospi_8_64;
temp2 = -step2[20] * cospi_8_64 + step2[27] * cospi_24_64;
- step1[20] = dct_const_round_shift(temp1);
- step1[27] = dct_const_round_shift(temp2);
+ step1[20] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[27] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = -step2[21] * cospi_24_64 - step2[26] * cospi_8_64;
temp2 = -step2[21] * cospi_8_64 + step2[26] * cospi_24_64;
- step1[21] = dct_const_round_shift(temp1);
- step1[26] = dct_const_round_shift(temp2);
+ step1[21] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[26] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[22] = step2[22];
step1[23] = step2[23];
step1[24] = step2[24];
@@ -1164,62 +1172,62 @@
step1[31] = step2[31];
// stage 6
- step2[0] = step1[0] + step1[7];
- step2[1] = step1[1] + step1[6];
- step2[2] = step1[2] + step1[5];
- step2[3] = step1[3] + step1[4];
- step2[4] = step1[3] - step1[4];
- step2[5] = step1[2] - step1[5];
- step2[6] = step1[1] - step1[6];
- step2[7] = step1[0] - step1[7];
+ step2[0] = WRAPLOW(step1[0] + step1[7], 8);
+ step2[1] = WRAPLOW(step1[1] + step1[6], 8);
+ step2[2] = WRAPLOW(step1[2] + step1[5], 8);
+ step2[3] = WRAPLOW(step1[3] + step1[4], 8);
+ step2[4] = WRAPLOW(step1[3] - step1[4], 8);
+ step2[5] = WRAPLOW(step1[2] - step1[5], 8);
+ step2[6] = WRAPLOW(step1[1] - step1[6], 8);
+ step2[7] = WRAPLOW(step1[0] - step1[7], 8);
step2[8] = step1[8];
step2[9] = step1[9];
temp1 = (-step1[10] + step1[13]) * cospi_16_64;
temp2 = (step1[10] + step1[13]) * cospi_16_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = (-step1[11] + step1[12]) * cospi_16_64;
temp2 = (step1[11] + step1[12]) * cospi_16_64;
- step2[11] = dct_const_round_shift(temp1);
- step2[12] = dct_const_round_shift(temp2);
+ step2[11] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step2[12] = WRAPLOW(dct_const_round_shift(temp2), 8);
step2[14] = step1[14];
step2[15] = step1[15];
- step2[16] = step1[16] + step1[23];
- step2[17] = step1[17] + step1[22];
- step2[18] = step1[18] + step1[21];
- step2[19] = step1[19] + step1[20];
- step2[20] = step1[19] - step1[20];
- step2[21] = step1[18] - step1[21];
- step2[22] = step1[17] - step1[22];
- step2[23] = step1[16] - step1[23];
+ step2[16] = WRAPLOW(step1[16] + step1[23], 8);
+ step2[17] = WRAPLOW(step1[17] + step1[22], 8);
+ step2[18] = WRAPLOW(step1[18] + step1[21], 8);
+ step2[19] = WRAPLOW(step1[19] + step1[20], 8);
+ step2[20] = WRAPLOW(step1[19] - step1[20], 8);
+ step2[21] = WRAPLOW(step1[18] - step1[21], 8);
+ step2[22] = WRAPLOW(step1[17] - step1[22], 8);
+ step2[23] = WRAPLOW(step1[16] - step1[23], 8);
- step2[24] = -step1[24] + step1[31];
- step2[25] = -step1[25] + step1[30];
- step2[26] = -step1[26] + step1[29];
- step2[27] = -step1[27] + step1[28];
- step2[28] = step1[27] + step1[28];
- step2[29] = step1[26] + step1[29];
- step2[30] = step1[25] + step1[30];
- step2[31] = step1[24] + step1[31];
+ step2[24] = WRAPLOW(-step1[24] + step1[31], 8);
+ step2[25] = WRAPLOW(-step1[25] + step1[30], 8);
+ step2[26] = WRAPLOW(-step1[26] + step1[29], 8);
+ step2[27] = WRAPLOW(-step1[27] + step1[28], 8);
+ step2[28] = WRAPLOW(step1[27] + step1[28], 8);
+ step2[29] = WRAPLOW(step1[26] + step1[29], 8);
+ step2[30] = WRAPLOW(step1[25] + step1[30], 8);
+ step2[31] = WRAPLOW(step1[24] + step1[31], 8);
// stage 7
- step1[0] = step2[0] + step2[15];
- step1[1] = step2[1] + step2[14];
- step1[2] = step2[2] + step2[13];
- step1[3] = step2[3] + step2[12];
- step1[4] = step2[4] + step2[11];
- step1[5] = step2[5] + step2[10];
- step1[6] = step2[6] + step2[9];
- step1[7] = step2[7] + step2[8];
- step1[8] = step2[7] - step2[8];
- step1[9] = step2[6] - step2[9];
- step1[10] = step2[5] - step2[10];
- step1[11] = step2[4] - step2[11];
- step1[12] = step2[3] - step2[12];
- step1[13] = step2[2] - step2[13];
- step1[14] = step2[1] - step2[14];
- step1[15] = step2[0] - step2[15];
+ step1[0] = WRAPLOW(step2[0] + step2[15], 8);
+ step1[1] = WRAPLOW(step2[1] + step2[14], 8);
+ step1[2] = WRAPLOW(step2[2] + step2[13], 8);
+ step1[3] = WRAPLOW(step2[3] + step2[12], 8);
+ step1[4] = WRAPLOW(step2[4] + step2[11], 8);
+ step1[5] = WRAPLOW(step2[5] + step2[10], 8);
+ step1[6] = WRAPLOW(step2[6] + step2[9], 8);
+ step1[7] = WRAPLOW(step2[7] + step2[8], 8);
+ step1[8] = WRAPLOW(step2[7] - step2[8], 8);
+ step1[9] = WRAPLOW(step2[6] - step2[9], 8);
+ step1[10] = WRAPLOW(step2[5] - step2[10], 8);
+ step1[11] = WRAPLOW(step2[4] - step2[11], 8);
+ step1[12] = WRAPLOW(step2[3] - step2[12], 8);
+ step1[13] = WRAPLOW(step2[2] - step2[13], 8);
+ step1[14] = WRAPLOW(step2[1] - step2[14], 8);
+ step1[15] = WRAPLOW(step2[0] - step2[15], 8);
step1[16] = step2[16];
step1[17] = step2[17];
@@ -1227,58 +1235,58 @@
step1[19] = step2[19];
temp1 = (-step2[20] + step2[27]) * cospi_16_64;
temp2 = (step2[20] + step2[27]) * cospi_16_64;
- step1[20] = dct_const_round_shift(temp1);
- step1[27] = dct_const_round_shift(temp2);
+ step1[20] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[27] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = (-step2[21] + step2[26]) * cospi_16_64;
temp2 = (step2[21] + step2[26]) * cospi_16_64;
- step1[21] = dct_const_round_shift(temp1);
- step1[26] = dct_const_round_shift(temp2);
+ step1[21] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[26] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = (-step2[22] + step2[25]) * cospi_16_64;
temp2 = (step2[22] + step2[25]) * cospi_16_64;
- step1[22] = dct_const_round_shift(temp1);
- step1[25] = dct_const_round_shift(temp2);
+ step1[22] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[25] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = (-step2[23] + step2[24]) * cospi_16_64;
temp2 = (step2[23] + step2[24]) * cospi_16_64;
- step1[23] = dct_const_round_shift(temp1);
- step1[24] = dct_const_round_shift(temp2);
+ step1[23] = WRAPLOW(dct_const_round_shift(temp1), 8);
+ step1[24] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[28] = step2[28];
step1[29] = step2[29];
step1[30] = step2[30];
step1[31] = step2[31];
// final stage
- output[0] = step1[0] + step1[31];
- output[1] = step1[1] + step1[30];
- output[2] = step1[2] + step1[29];
- output[3] = step1[3] + step1[28];
- output[4] = step1[4] + step1[27];
- output[5] = step1[5] + step1[26];
- output[6] = step1[6] + step1[25];
- output[7] = step1[7] + step1[24];
- output[8] = step1[8] + step1[23];
- output[9] = step1[9] + step1[22];
- output[10] = step1[10] + step1[21];
- output[11] = step1[11] + step1[20];
- output[12] = step1[12] + step1[19];
- output[13] = step1[13] + step1[18];
- output[14] = step1[14] + step1[17];
- output[15] = step1[15] + step1[16];
- output[16] = step1[15] - step1[16];
- output[17] = step1[14] - step1[17];
- output[18] = step1[13] - step1[18];
- output[19] = step1[12] - step1[19];
- output[20] = step1[11] - step1[20];
- output[21] = step1[10] - step1[21];
- output[22] = step1[9] - step1[22];
- output[23] = step1[8] - step1[23];
- output[24] = step1[7] - step1[24];
- output[25] = step1[6] - step1[25];
- output[26] = step1[5] - step1[26];
- output[27] = step1[4] - step1[27];
- output[28] = step1[3] - step1[28];
- output[29] = step1[2] - step1[29];
- output[30] = step1[1] - step1[30];
- output[31] = step1[0] - step1[31];
+ output[0] = WRAPLOW(step1[0] + step1[31], 8);
+ output[1] = WRAPLOW(step1[1] + step1[30], 8);
+ output[2] = WRAPLOW(step1[2] + step1[29], 8);
+ output[3] = WRAPLOW(step1[3] + step1[28], 8);
+ output[4] = WRAPLOW(step1[4] + step1[27], 8);
+ output[5] = WRAPLOW(step1[5] + step1[26], 8);
+ output[6] = WRAPLOW(step1[6] + step1[25], 8);
+ output[7] = WRAPLOW(step1[7] + step1[24], 8);
+ output[8] = WRAPLOW(step1[8] + step1[23], 8);
+ output[9] = WRAPLOW(step1[9] + step1[22], 8);
+ output[10] = WRAPLOW(step1[10] + step1[21], 8);
+ output[11] = WRAPLOW(step1[11] + step1[20], 8);
+ output[12] = WRAPLOW(step1[12] + step1[19], 8);
+ output[13] = WRAPLOW(step1[13] + step1[18], 8);
+ output[14] = WRAPLOW(step1[14] + step1[17], 8);
+ output[15] = WRAPLOW(step1[15] + step1[16], 8);
+ output[16] = WRAPLOW(step1[15] - step1[16], 8);
+ output[17] = WRAPLOW(step1[14] - step1[17], 8);
+ output[18] = WRAPLOW(step1[13] - step1[18], 8);
+ output[19] = WRAPLOW(step1[12] - step1[19], 8);
+ output[20] = WRAPLOW(step1[11] - step1[20], 8);
+ output[21] = WRAPLOW(step1[10] - step1[21], 8);
+ output[22] = WRAPLOW(step1[9] - step1[22], 8);
+ output[23] = WRAPLOW(step1[8] - step1[23], 8);
+ output[24] = WRAPLOW(step1[7] - step1[24], 8);
+ output[25] = WRAPLOW(step1[6] - step1[25], 8);
+ output[26] = WRAPLOW(step1[5] - step1[26], 8);
+ output[27] = WRAPLOW(step1[4] - step1[27], 8);
+ output[28] = WRAPLOW(step1[3] - step1[28], 8);
+ output[29] = WRAPLOW(step1[2] - step1[29], 8);
+ output[30] = WRAPLOW(step1[1] - step1[30], 8);
+ output[31] = WRAPLOW(step1[0] - step1[31], 8);
}
void vp9_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest,
@@ -1313,9 +1321,10 @@
for (j = 0; j < 32; ++j)
temp_in[j] = out[j * 32 + i];
idct32(temp_in, temp_out);
- for (j = 0; j < 32; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
- + dest[j * stride + i]);
+ for (j = 0; j < 32; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 6));
+ }
}
}
@@ -1339,9 +1348,10 @@
for (j = 0; j < 32; ++j)
temp_in[j] = out[j * 32 + i];
idct32(temp_in, temp_out);
- for (j = 0; j < 32; ++j)
- dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
- + dest[j * stride + i]);
+ for (j = 0; j < 32; ++j) {
+ dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
+ ROUND_POWER_OF_TWO(temp_out[j], 6));
+ }
}
}
@@ -1349,13 +1359,13 @@
int i, j;
tran_high_t a1;
- tran_low_t out = dct_const_round_shift(input[0] * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
+ tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), 8);
+ out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), 8);
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 32; ++j) {
for (i = 0; i < 32; ++i)
- dest[i] = clip_pixel(dest[i] + a1);
+ dest[i] = clip_pixel_add(dest[i], a1);
dest += stride;
}
}
@@ -1448,8 +1458,8 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
/* 4-point reversible, orthonormal inverse Walsh-Hadamard in 3.5 adds,
0.5 shifts per pixel. */
int i;
@@ -1471,10 +1481,10 @@
c1 = e1 - c1;
a1 -= b1;
d1 += c1;
- op[0] = WRAPLOW(a1);
- op[1] = WRAPLOW(b1);
- op[2] = WRAPLOW(c1);
- op[3] = WRAPLOW(d1);
+ op[0] = WRAPLOW(a1, bd);
+ op[1] = WRAPLOW(b1, bd);
+ op[2] = WRAPLOW(c1, bd);
+ op[3] = WRAPLOW(d1, bd);
ip += 4;
op += 4;
}
@@ -1492,39 +1502,18 @@
c1 = e1 - c1;
a1 -= b1;
d1 += c1;
- dest[stride * 0] = clip_pixel_bd_high(dest[stride * 0], a1, bd);
- dest[stride * 1] = clip_pixel_bd_high(dest[stride * 1], b1, bd);
- dest[stride * 2] = clip_pixel_bd_high(dest[stride * 2], c1, bd);
- dest[stride * 3] = clip_pixel_bd_high(dest[stride * 3], d1, bd);
+ dest[stride * 0] = highbd_clip_pixel_add(dest[stride * 0], a1, bd);
+ dest[stride * 1] = highbd_clip_pixel_add(dest[stride * 1], b1, bd);
+ dest[stride * 2] = highbd_clip_pixel_add(dest[stride * 2], c1, bd);
+ dest[stride * 3] = highbd_clip_pixel_add(dest[stride * 3], d1, bd);
ip++;
dest++;
}
}
-static void high_idct4(const tran_low_t *input, tran_low_t *output, int bd) {
- tran_low_t step[4];
- tran_high_t temp1, temp2;
- (void) bd;
- // stage 1
- temp1 = (input[0] + input[2]) * cospi_16_64;
- temp2 = (input[0] - input[2]) * cospi_16_64;
- step[0] = WRAPLOW(dct_const_round_shift(temp1));
- step[1] = WRAPLOW(dct_const_round_shift(temp2));
- temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64;
- temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64;
- step[2] = WRAPLOW(dct_const_round_shift(temp1));
- step[3] = WRAPLOW(dct_const_round_shift(temp2));
-
- // stage 2
- output[0] = WRAPLOW(step[0] + step[3]);
- output[1] = WRAPLOW(step[1] + step[2]);
- output[2] = WRAPLOW(step[1] - step[2]);
- output[3] = WRAPLOW(step[0] - step[3]);
-}
-
-void vp9_high_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest8,
- int dest_stride, int bd) {
+void vp9_highbd_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest8,
+ int dest_stride, int bd) {
int i;
tran_high_t a1, e1;
tran_low_t tmp[4];
@@ -1536,24 +1525,49 @@
a1 = ip[0] >> UNIT_QUANT_SHIFT;
e1 = a1 >> 1;
a1 -= e1;
- op[0] = WRAPLOW(a1);
- op[1] = op[2] = op[3] = WRAPLOW(e1);
+ op[0] = WRAPLOW(a1, bd);
+ op[1] = op[2] = op[3] = WRAPLOW(e1, bd);
ip = tmp;
for (i = 0; i < 4; i++) {
e1 = ip[0] >> 1;
a1 = ip[0] - e1;
- dest[dest_stride * 0] = clip_pixel_bd_high(dest[dest_stride * 0], a1, bd);
- dest[dest_stride * 1] = clip_pixel_bd_high(dest[dest_stride * 1], e1, bd);
- dest[dest_stride * 2] = clip_pixel_bd_high(dest[dest_stride * 2], e1, bd);
- dest[dest_stride * 3] = clip_pixel_bd_high(dest[dest_stride * 3], e1, bd);
+ dest[dest_stride * 0] = highbd_clip_pixel_add(
+ dest[dest_stride * 0], a1, bd);
+ dest[dest_stride * 1] = highbd_clip_pixel_add(
+ dest[dest_stride * 1], e1, bd);
+ dest[dest_stride * 2] = highbd_clip_pixel_add(
+ dest[dest_stride * 2], e1, bd);
+ dest[dest_stride * 3] = highbd_clip_pixel_add(
+ dest[dest_stride * 3], e1, bd);
ip++;
dest++;
}
}
-void vp9_high_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+static void highbd_idct4(const tran_low_t *input, tran_low_t *output, int bd) {
+ tran_low_t step[4];
+ tran_high_t temp1, temp2;
+ (void) bd;
+ // stage 1
+ temp1 = (input[0] + input[2]) * cospi_16_64;
+ temp2 = (input[0] - input[2]) * cospi_16_64;
+ step[0] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step[1] = WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64;
+ temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64;
+ step[2] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step[3] = WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ // stage 2
+ output[0] = WRAPLOW(step[0] + step[3], bd);
+ output[1] = WRAPLOW(step[1] + step[2], bd);
+ output[2] = WRAPLOW(step[1] - step[2], bd);
+ output[3] = WRAPLOW(step[0] - step[3], bd);
+}
+
+void vp9_highbd_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
tran_low_t out[4 * 4];
tran_low_t *outptr = out;
int i, j;
@@ -1562,7 +1576,7 @@
// Rows
for (i = 0; i < 4; ++i) {
- high_idct4(input, outptr, bd);
+ highbd_idct4(input, outptr, bd);
input += 4;
outptr += 4;
}
@@ -1571,33 +1585,34 @@
for (i = 0; i < 4; ++i) {
for (j = 0; j < 4; ++j)
temp_in[j] = out[j * 4 + i];
- high_idct4(temp_in, temp_out, bd);
- for (j = 0; j < 4; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ highbd_idct4(temp_in, temp_out, bd);
+ for (j = 0; j < 4; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 4), bd);
+ }
}
}
-void vp9_high_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest8,
- int dest_stride, int bd) {
+void vp9_highbd_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest8,
+ int dest_stride, int bd) {
int i;
tran_high_t a1;
- tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
+ tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), bd);
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
- out = WRAPLOW(dct_const_round_shift(out * cospi_16_64));
+ out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), bd);
a1 = ROUND_POWER_OF_TWO(out, 4);
for (i = 0; i < 4; i++) {
- dest[0] = clip_pixel_bd_high(dest[0], a1, bd);
- dest[1] = clip_pixel_bd_high(dest[1], a1, bd);
- dest[2] = clip_pixel_bd_high(dest[2], a1, bd);
- dest[3] = clip_pixel_bd_high(dest[3], a1, bd);
+ dest[0] = highbd_clip_pixel_add(dest[0], a1, bd);
+ dest[1] = highbd_clip_pixel_add(dest[1], a1, bd);
+ dest[2] = highbd_clip_pixel_add(dest[2], a1, bd);
+ dest[3] = highbd_clip_pixel_add(dest[3], a1, bd);
dest += dest_stride;
}
}
-static void high_idct8(const tran_low_t *input, tran_low_t *output, int bd) {
+static void highbd_idct8(const tran_low_t *input, tran_low_t *output, int bd) {
tran_low_t step1[8], step2[8];
tran_high_t temp1, temp2;
// stage 1
@@ -1607,43 +1622,43 @@
step1[3] = input[6];
temp1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
temp2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
- step1[4] = WRAPLOW(dct_const_round_shift(temp1));
- step1[7] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[4] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[7] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
temp2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
- step1[5] = WRAPLOW(dct_const_round_shift(temp1));
- step1[6] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), bd);
// stage 2 & stage 3 - even half
- high_idct4(step1, step1, bd);
+ highbd_idct4(step1, step1, bd);
// stage 2 - odd half
- step2[4] = WRAPLOW(step1[4] + step1[5]);
- step2[5] = WRAPLOW(step1[4] - step1[5]);
- step2[6] = WRAPLOW(-step1[6] + step1[7]);
- step2[7] = WRAPLOW(step1[6] + step1[7]);
+ step2[4] = WRAPLOW(step1[4] + step1[5], bd);
+ step2[5] = WRAPLOW(step1[4] - step1[5], bd);
+ step2[6] = WRAPLOW(-step1[6] + step1[7], bd);
+ step2[7] = WRAPLOW(step1[6] + step1[7], bd);
// stage 3 - odd half
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = WRAPLOW(dct_const_round_shift(temp1));
- step1[6] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), bd);
step1[7] = step2[7];
// stage 4
- output[0] = WRAPLOW(step1[0] + step1[7]);
- output[1] = WRAPLOW(step1[1] + step1[6]);
- output[2] = WRAPLOW(step1[2] + step1[5]);
- output[3] = WRAPLOW(step1[3] + step1[4]);
- output[4] = WRAPLOW(step1[3] - step1[4]);
- output[5] = WRAPLOW(step1[2] - step1[5]);
- output[6] = WRAPLOW(step1[1] - step1[6]);
- output[7] = WRAPLOW(step1[0] - step1[7]);
+ output[0] = WRAPLOW(step1[0] + step1[7], bd);
+ output[1] = WRAPLOW(step1[1] + step1[6], bd);
+ output[2] = WRAPLOW(step1[2] + step1[5], bd);
+ output[3] = WRAPLOW(step1[3] + step1[4], bd);
+ output[4] = WRAPLOW(step1[3] - step1[4], bd);
+ output[5] = WRAPLOW(step1[2] - step1[5], bd);
+ output[6] = WRAPLOW(step1[1] - step1[6], bd);
+ output[7] = WRAPLOW(step1[0] - step1[7], bd);
}
-void vp9_high_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
int i, j;
@@ -1652,7 +1667,7 @@
// First transform rows.
for (i = 0; i < 8; ++i) {
- high_idct8(input, outptr, bd);
+ highbd_idct8(input, outptr, bd);
input += 8;
outptr += 8;
}
@@ -1661,30 +1676,30 @@
for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
- high_idct8(temp_in, temp_out, bd);
- for (j = 0; j < 8; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(dest[j * stride + i],
- ROUND_POWER_OF_TWO(temp_out[j], 5),
- bd);
+ highbd_idct8(temp_in, temp_out, bd);
+ for (j = 0; j < 8; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
+ dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 5), bd);
+ }
}
}
-void vp9_high_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
int i, j;
tran_high_t a1;
- tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
+ tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), bd);
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
- out = WRAPLOW(dct_const_round_shift(out * cospi_16_64));
+ out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), bd);
a1 = ROUND_POWER_OF_TWO(out, 5);
for (j = 0; j < 8; ++j) {
for (i = 0; i < 8; ++i)
- dest[i] = clip_pixel_bd_high(dest[i], a1, bd);
+ dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
dest += stride;
}
}
-static void high_iadst4(const tran_low_t *input, tran_low_t *output, int bd) {
+static void highbd_iadst4(const tran_low_t *input, tran_low_t *output, int bd) {
tran_high_t s0, s1, s2, s3, s4, s5, s6, s7;
tran_high_t x0 = input[0];
@@ -1721,19 +1736,19 @@
// The overall dynamic range is 14b (input) + 14b (multiplication scaling)
// + 1b (addition) = 29b.
// Hence the output bit depth is 15b.
- output[0] = WRAPLOW(dct_const_round_shift(s0));
- output[1] = WRAPLOW(dct_const_round_shift(s1));
- output[2] = WRAPLOW(dct_const_round_shift(s2));
- output[3] = WRAPLOW(dct_const_round_shift(s3));
+ output[0] = WRAPLOW(dct_const_round_shift(s0), bd);
+ output[1] = WRAPLOW(dct_const_round_shift(s1), bd);
+ output[2] = WRAPLOW(dct_const_round_shift(s2), bd);
+ output[3] = WRAPLOW(dct_const_round_shift(s3), bd);
}
-void vp9_high_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int tx_type, int bd) {
- const high_transform_2d IHT_4[] = {
- { high_idct4, high_idct4 }, // DCT_DCT = 0
- { high_iadst4, high_idct4 }, // ADST_DCT = 1
- { high_idct4, high_iadst4 }, // DCT_ADST = 2
- { high_iadst4, high_iadst4 } // ADST_ADST = 3
+void vp9_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int tx_type, int bd) {
+ const highbd_transform_2d IHT_4[] = {
+ { highbd_idct4, highbd_idct4 }, // DCT_DCT = 0
+ { highbd_iadst4, highbd_idct4 }, // ADST_DCT = 1
+ { highbd_idct4, highbd_iadst4 }, // DCT_ADST = 2
+ { highbd_iadst4, highbd_iadst4 } // ADST_ADST = 3
};
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
@@ -1754,13 +1769,14 @@
for (j = 0; j < 4; ++j)
temp_in[j] = out[j * 4 + i];
IHT_4[tx_type].cols(temp_in, temp_out, bd);
- for (j = 0; j < 4; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ for (j = 0; j < 4; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 4), bd);
+ }
}
}
-static void high_iadst8(const tran_low_t *input, tran_low_t *output, int bd) {
+static void highbd_iadst8(const tran_low_t *input, tran_low_t *output, int bd) {
tran_high_t s0, s1, s2, s3, s4, s5, s6, s7;
tran_high_t x0 = input[7];
@@ -1788,14 +1804,14 @@
s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
- x0 = WRAPLOW(dct_const_round_shift(s0 + s4));
- x1 = WRAPLOW(dct_const_round_shift(s1 + s5));
- x2 = WRAPLOW(dct_const_round_shift(s2 + s6));
- x3 = WRAPLOW(dct_const_round_shift(s3 + s7));
- x4 = WRAPLOW(dct_const_round_shift(s0 - s4));
- x5 = WRAPLOW(dct_const_round_shift(s1 - s5));
- x6 = WRAPLOW(dct_const_round_shift(s2 - s6));
- x7 = WRAPLOW(dct_const_round_shift(s3 - s7));
+ x0 = WRAPLOW(dct_const_round_shift(s0 + s4), bd);
+ x1 = WRAPLOW(dct_const_round_shift(s1 + s5), bd);
+ x2 = WRAPLOW(dct_const_round_shift(s2 + s6), bd);
+ x3 = WRAPLOW(dct_const_round_shift(s3 + s7), bd);
+ x4 = WRAPLOW(dct_const_round_shift(s0 - s4), bd);
+ x5 = WRAPLOW(dct_const_round_shift(s1 - s5), bd);
+ x6 = WRAPLOW(dct_const_round_shift(s2 - s6), bd);
+ x7 = WRAPLOW(dct_const_round_shift(s3 - s7), bd);
// stage 2
s0 = x0;
@@ -1807,14 +1823,14 @@
s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
- x0 = s0 + s2;
- x1 = s1 + s3;
- x2 = s0 - s2;
- x3 = s1 - s3;
- x4 = WRAPLOW(dct_const_round_shift(s4 + s6));
- x5 = WRAPLOW(dct_const_round_shift(s5 + s7));
- x6 = WRAPLOW(dct_const_round_shift(s4 - s6));
- x7 = WRAPLOW(dct_const_round_shift(s5 - s7));
+ x0 = WRAPLOW(s0 + s2, bd);
+ x1 = WRAPLOW(s1 + s3, bd);
+ x2 = WRAPLOW(s0 - s2, bd);
+ x3 = WRAPLOW(s1 - s3, bd);
+ x4 = WRAPLOW(dct_const_round_shift(s4 + s6), bd);
+ x5 = WRAPLOW(dct_const_round_shift(s5 + s7), bd);
+ x6 = WRAPLOW(dct_const_round_shift(s4 - s6), bd);
+ x7 = WRAPLOW(dct_const_round_shift(s5 - s7), bd);
// stage 3
s2 = cospi_16_64 * (x2 + x3);
@@ -1822,35 +1838,35 @@
s6 = cospi_16_64 * (x6 + x7);
s7 = cospi_16_64 * (x6 - x7);
- x2 = WRAPLOW(dct_const_round_shift(s2));
- x3 = WRAPLOW(dct_const_round_shift(s3));
- x6 = WRAPLOW(dct_const_round_shift(s6));
- x7 = WRAPLOW(dct_const_round_shift(s7));
+ x2 = WRAPLOW(dct_const_round_shift(s2), bd);
+ x3 = WRAPLOW(dct_const_round_shift(s3), bd);
+ x6 = WRAPLOW(dct_const_round_shift(s6), bd);
+ x7 = WRAPLOW(dct_const_round_shift(s7), bd);
- output[0] = WRAPLOW(x0);
- output[1] = WRAPLOW(-x4);
- output[2] = WRAPLOW(x6);
- output[3] = WRAPLOW(-x2);
- output[4] = WRAPLOW(x3);
- output[5] = WRAPLOW(-x7);
- output[6] = WRAPLOW(x5);
- output[7] = WRAPLOW(-x1);
+ output[0] = WRAPLOW(x0, bd);
+ output[1] = WRAPLOW(-x4, bd);
+ output[2] = WRAPLOW(x6, bd);
+ output[3] = WRAPLOW(-x2, bd);
+ output[4] = WRAPLOW(x3, bd);
+ output[5] = WRAPLOW(-x7, bd);
+ output[6] = WRAPLOW(x5, bd);
+ output[7] = WRAPLOW(-x1, bd);
}
-static const high_transform_2d HIGH_IHT_8[] = {
- { high_idct8, high_idct8 }, // DCT_DCT = 0
- { high_iadst8, high_idct8 }, // ADST_DCT = 1
- { high_idct8, high_iadst8 }, // DCT_ADST = 2
- { high_iadst8, high_iadst8 } // ADST_ADST = 3
+static const highbd_transform_2d HIGH_IHT_8[] = {
+ { highbd_idct8, highbd_idct8 }, // DCT_DCT = 0
+ { highbd_iadst8, highbd_idct8 }, // ADST_DCT = 1
+ { highbd_idct8, highbd_iadst8 }, // DCT_ADST = 2
+ { highbd_iadst8, highbd_iadst8 } // ADST_ADST = 3
};
-void vp9_high_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int tx_type, int bd) {
+void vp9_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int tx_type, int bd) {
int i, j;
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
tran_low_t temp_in[8], temp_out[8];
- const high_transform_2d ht = HIGH_IHT_8[tx_type];
+ const highbd_transform_2d ht = HIGH_IHT_8[tx_type];
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
// Inverse transform row vectors.
@@ -1865,14 +1881,15 @@
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
ht.cols(temp_in, temp_out, bd);
- for (j = 0; j < 8; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ for (j = 0; j < 8; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 5), bd);
+ }
}
}
-void vp9_high_idct8x8_10_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct8x8_10_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
tran_low_t out[8 * 8] = { 0 };
tran_low_t *outptr = out;
int i, j;
@@ -1882,7 +1899,7 @@
// First transform rows.
// Only first 4 row has non-zero coefs.
for (i = 0; i < 4; ++i) {
- high_idct8(input, outptr, bd);
+ highbd_idct8(input, outptr, bd);
input += 8;
outptr += 8;
}
@@ -1890,14 +1907,15 @@
for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
- high_idct8(temp_in, temp_out, bd);
- for (j = 0; j < 8; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ highbd_idct8(temp_in, temp_out, bd);
+ for (j = 0; j < 8; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 5), bd);
+ }
}
}
-static void high_idct16(const tran_low_t *input, tran_low_t *output, int bd) {
+static void highbd_idct16(const tran_low_t *input, tran_low_t *output, int bd) {
tran_low_t step1[16], step2[16];
tran_high_t temp1, temp2;
(void) bd;
@@ -1932,23 +1950,23 @@
temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
- step2[8] = WRAPLOW(dct_const_round_shift(temp1));
- step2[15] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[8] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[15] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
- step2[9] = WRAPLOW(dct_const_round_shift(temp1));
- step2[14] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[9] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[14] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
- step2[10] = WRAPLOW(dct_const_round_shift(temp1));
- step2[13] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
- step2[11] = WRAPLOW(dct_const_round_shift(temp1));
- step2[12] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[11] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[12] = WRAPLOW(dct_const_round_shift(temp2), bd);
// stage 3
step1[0] = step2[0];
@@ -1958,113 +1976,113 @@
temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
- step1[4] = WRAPLOW(dct_const_round_shift(temp1));
- step1[7] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[4] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[7] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
- step1[5] = WRAPLOW(dct_const_round_shift(temp1));
- step1[6] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), bd);
- step1[8] = WRAPLOW(step2[8] + step2[9]);
- step1[9] = WRAPLOW(step2[8] - step2[9]);
- step1[10] = WRAPLOW(-step2[10] + step2[11]);
- step1[11] = WRAPLOW(step2[10] + step2[11]);
- step1[12] = WRAPLOW(step2[12] + step2[13]);
- step1[13] = WRAPLOW(step2[12] - step2[13]);
- step1[14] = WRAPLOW(-step2[14] + step2[15]);
- step1[15] = WRAPLOW(step2[14] + step2[15]);
+ step1[8] = WRAPLOW(step2[8] + step2[9], bd);
+ step1[9] = WRAPLOW(step2[8] - step2[9], bd);
+ step1[10] = WRAPLOW(-step2[10] + step2[11], bd);
+ step1[11] = WRAPLOW(step2[10] + step2[11], bd);
+ step1[12] = WRAPLOW(step2[12] + step2[13], bd);
+ step1[13] = WRAPLOW(step2[12] - step2[13], bd);
+ step1[14] = WRAPLOW(-step2[14] + step2[15], bd);
+ step1[15] = WRAPLOW(step2[14] + step2[15], bd);
// stage 4
temp1 = (step1[0] + step1[1]) * cospi_16_64;
temp2 = (step1[0] - step1[1]) * cospi_16_64;
- step2[0] = WRAPLOW(dct_const_round_shift(temp1));
- step2[1] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[0] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[1] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
- step2[2] = WRAPLOW(dct_const_round_shift(temp1));
- step2[3] = WRAPLOW(dct_const_round_shift(temp2));
- step2[4] = WRAPLOW(step1[4] + step1[5]);
- step2[5] = WRAPLOW(step1[4] - step1[5]);
- step2[6] = WRAPLOW(-step1[6] + step1[7]);
- step2[7] = WRAPLOW(step1[6] + step1[7]);
+ step2[2] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[3] = WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[4] = WRAPLOW(step1[4] + step1[5], bd);
+ step2[5] = WRAPLOW(step1[4] - step1[5], bd);
+ step2[6] = WRAPLOW(-step1[6] + step1[7], bd);
+ step2[7] = WRAPLOW(step1[6] + step1[7], bd);
step2[8] = step1[8];
step2[15] = step1[15];
temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
- step2[9] = WRAPLOW(dct_const_round_shift(temp1));
- step2[14] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[9] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[14] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
- step2[10] = WRAPLOW(dct_const_round_shift(temp1));
- step2[13] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), bd);
step2[11] = step1[11];
step2[12] = step1[12];
// stage 5
- step1[0] = WRAPLOW(step2[0] + step2[3]);
- step1[1] = WRAPLOW(step2[1] + step2[2]);
- step1[2] = WRAPLOW(step2[1] - step2[2]);
- step1[3] = WRAPLOW(step2[0] - step2[3]);
+ step1[0] = WRAPLOW(step2[0] + step2[3], bd);
+ step1[1] = WRAPLOW(step2[1] + step2[2], bd);
+ step1[2] = WRAPLOW(step2[1] - step2[2], bd);
+ step1[3] = WRAPLOW(step2[0] - step2[3], bd);
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = WRAPLOW(dct_const_round_shift(temp1));
- step1[6] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), bd);
step1[7] = step2[7];
- step1[8] = WRAPLOW(step2[8] + step2[11]);
- step1[9] = WRAPLOW(step2[9] + step2[10]);
- step1[10] = WRAPLOW(step2[9] - step2[10]);
- step1[11] = WRAPLOW(step2[8] - step2[11]);
- step1[12] = WRAPLOW(-step2[12] + step2[15]);
- step1[13] = WRAPLOW(-step2[13] + step2[14]);
- step1[14] = WRAPLOW(step2[13] + step2[14]);
- step1[15] = WRAPLOW(step2[12] + step2[15]);
+ step1[8] = WRAPLOW(step2[8] + step2[11], bd);
+ step1[9] = WRAPLOW(step2[9] + step2[10], bd);
+ step1[10] = WRAPLOW(step2[9] - step2[10], bd);
+ step1[11] = WRAPLOW(step2[8] - step2[11], bd);
+ step1[12] = WRAPLOW(-step2[12] + step2[15], bd);
+ step1[13] = WRAPLOW(-step2[13] + step2[14], bd);
+ step1[14] = WRAPLOW(step2[13] + step2[14], bd);
+ step1[15] = WRAPLOW(step2[12] + step2[15], bd);
// stage 6
- step2[0] = WRAPLOW(step1[0] + step1[7]);
- step2[1] = WRAPLOW(step1[1] + step1[6]);
- step2[2] = WRAPLOW(step1[2] + step1[5]);
- step2[3] = WRAPLOW(step1[3] + step1[4]);
- step2[4] = WRAPLOW(step1[3] - step1[4]);
- step2[5] = WRAPLOW(step1[2] - step1[5]);
- step2[6] = WRAPLOW(step1[1] - step1[6]);
- step2[7] = WRAPLOW(step1[0] - step1[7]);
+ step2[0] = WRAPLOW(step1[0] + step1[7], bd);
+ step2[1] = WRAPLOW(step1[1] + step1[6], bd);
+ step2[2] = WRAPLOW(step1[2] + step1[5], bd);
+ step2[3] = WRAPLOW(step1[3] + step1[4], bd);
+ step2[4] = WRAPLOW(step1[3] - step1[4], bd);
+ step2[5] = WRAPLOW(step1[2] - step1[5], bd);
+ step2[6] = WRAPLOW(step1[1] - step1[6], bd);
+ step2[7] = WRAPLOW(step1[0] - step1[7], bd);
step2[8] = step1[8];
step2[9] = step1[9];
temp1 = (-step1[10] + step1[13]) * cospi_16_64;
temp2 = (step1[10] + step1[13]) * cospi_16_64;
- step2[10] = WRAPLOW(dct_const_round_shift(temp1));
- step2[13] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = (-step1[11] + step1[12]) * cospi_16_64;
temp2 = (step1[11] + step1[12]) * cospi_16_64;
- step2[11] = WRAPLOW(dct_const_round_shift(temp1));
- step2[12] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[11] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[12] = WRAPLOW(dct_const_round_shift(temp2), bd);
step2[14] = step1[14];
step2[15] = step1[15];
// stage 7
- output[0] = WRAPLOW(step2[0] + step2[15]);
- output[1] = WRAPLOW(step2[1] + step2[14]);
- output[2] = WRAPLOW(step2[2] + step2[13]);
- output[3] = WRAPLOW(step2[3] + step2[12]);
- output[4] = WRAPLOW(step2[4] + step2[11]);
- output[5] = WRAPLOW(step2[5] + step2[10]);
- output[6] = WRAPLOW(step2[6] + step2[9]);
- output[7] = WRAPLOW(step2[7] + step2[8]);
- output[8] = WRAPLOW(step2[7] - step2[8]);
- output[9] = WRAPLOW(step2[6] - step2[9]);
- output[10] = WRAPLOW(step2[5] - step2[10]);
- output[11] = WRAPLOW(step2[4] - step2[11]);
- output[12] = WRAPLOW(step2[3] - step2[12]);
- output[13] = WRAPLOW(step2[2] - step2[13]);
- output[14] = WRAPLOW(step2[1] - step2[14]);
- output[15] = WRAPLOW(step2[0] - step2[15]);
+ output[0] = WRAPLOW(step2[0] + step2[15], bd);
+ output[1] = WRAPLOW(step2[1] + step2[14], bd);
+ output[2] = WRAPLOW(step2[2] + step2[13], bd);
+ output[3] = WRAPLOW(step2[3] + step2[12], bd);
+ output[4] = WRAPLOW(step2[4] + step2[11], bd);
+ output[5] = WRAPLOW(step2[5] + step2[10], bd);
+ output[6] = WRAPLOW(step2[6] + step2[9], bd);
+ output[7] = WRAPLOW(step2[7] + step2[8], bd);
+ output[8] = WRAPLOW(step2[7] - step2[8], bd);
+ output[9] = WRAPLOW(step2[6] - step2[9], bd);
+ output[10] = WRAPLOW(step2[5] - step2[10], bd);
+ output[11] = WRAPLOW(step2[4] - step2[11], bd);
+ output[12] = WRAPLOW(step2[3] - step2[12], bd);
+ output[13] = WRAPLOW(step2[2] - step2[13], bd);
+ output[14] = WRAPLOW(step2[1] - step2[14], bd);
+ output[15] = WRAPLOW(step2[0] - step2[15], bd);
}
-void vp9_high_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
int i, j;
@@ -2073,7 +2091,7 @@
// First transform rows.
for (i = 0; i < 16; ++i) {
- high_idct16(input, outptr, bd);
+ highbd_idct16(input, outptr, bd);
input += 16;
outptr += 16;
}
@@ -2082,14 +2100,16 @@
for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j)
temp_in[j] = out[j * 16 + i];
- high_idct16(temp_in, temp_out, bd);
- for (j = 0; j < 16; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ highbd_idct16(temp_in, temp_out, bd);
+ for (j = 0; j < 16; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
+ }
}
}
-static void high_iadst16(const tran_low_t *input, tran_low_t *output, int bd) {
+static void highbd_iadst16(const tran_low_t *input, tran_low_t *output,
+ int bd) {
tran_high_t s0, s1, s2, s3, s4, s5, s6, s7, s8;
tran_high_t s9, s10, s11, s12, s13, s14, s15;
@@ -2135,22 +2155,22 @@
s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
- x0 = WRAPLOW(dct_const_round_shift(s0 + s8));
- x1 = WRAPLOW(dct_const_round_shift(s1 + s9));
- x2 = WRAPLOW(dct_const_round_shift(s2 + s10));
- x3 = WRAPLOW(dct_const_round_shift(s3 + s11));
- x4 = WRAPLOW(dct_const_round_shift(s4 + s12));
- x5 = WRAPLOW(dct_const_round_shift(s5 + s13));
- x6 = WRAPLOW(dct_const_round_shift(s6 + s14));
- x7 = WRAPLOW(dct_const_round_shift(s7 + s15));
- x8 = WRAPLOW(dct_const_round_shift(s0 - s8));
- x9 = WRAPLOW(dct_const_round_shift(s1 - s9));
- x10 = WRAPLOW(dct_const_round_shift(s2 - s10));
- x11 = WRAPLOW(dct_const_round_shift(s3 - s11));
- x12 = WRAPLOW(dct_const_round_shift(s4 - s12));
- x13 = WRAPLOW(dct_const_round_shift(s5 - s13));
- x14 = WRAPLOW(dct_const_round_shift(s6 - s14));
- x15 = WRAPLOW(dct_const_round_shift(s7 - s15));
+ x0 = WRAPLOW(dct_const_round_shift(s0 + s8), bd);
+ x1 = WRAPLOW(dct_const_round_shift(s1 + s9), bd);
+ x2 = WRAPLOW(dct_const_round_shift(s2 + s10), bd);
+ x3 = WRAPLOW(dct_const_round_shift(s3 + s11), bd);
+ x4 = WRAPLOW(dct_const_round_shift(s4 + s12), bd);
+ x5 = WRAPLOW(dct_const_round_shift(s5 + s13), bd);
+ x6 = WRAPLOW(dct_const_round_shift(s6 + s14), bd);
+ x7 = WRAPLOW(dct_const_round_shift(s7 + s15), bd);
+ x8 = WRAPLOW(dct_const_round_shift(s0 - s8), bd);
+ x9 = WRAPLOW(dct_const_round_shift(s1 - s9), bd);
+ x10 = WRAPLOW(dct_const_round_shift(s2 - s10), bd);
+ x11 = WRAPLOW(dct_const_round_shift(s3 - s11), bd);
+ x12 = WRAPLOW(dct_const_round_shift(s4 - s12), bd);
+ x13 = WRAPLOW(dct_const_round_shift(s5 - s13), bd);
+ x14 = WRAPLOW(dct_const_round_shift(s6 - s14), bd);
+ x15 = WRAPLOW(dct_const_round_shift(s7 - s15), bd);
// stage 2
s0 = x0;
@@ -2170,22 +2190,22 @@
s14 = -x14 * cospi_12_64 + x15 * cospi_20_64;
s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
- x0 = WRAPLOW(s0 + s4);
- x1 = WRAPLOW(s1 + s5);
- x2 = WRAPLOW(s2 + s6);
- x3 = WRAPLOW(s3 + s7);
- x4 = WRAPLOW(s0 - s4);
- x5 = WRAPLOW(s1 - s5);
- x6 = WRAPLOW(s2 - s6);
- x7 = WRAPLOW(s3 - s7);
- x8 = WRAPLOW(dct_const_round_shift(s8 + s12));
- x9 = WRAPLOW(dct_const_round_shift(s9 + s13));
- x10 = WRAPLOW(dct_const_round_shift(s10 + s14));
- x11 = WRAPLOW(dct_const_round_shift(s11 + s15));
- x12 = WRAPLOW(dct_const_round_shift(s8 - s12));
- x13 = WRAPLOW(dct_const_round_shift(s9 - s13));
- x14 = WRAPLOW(dct_const_round_shift(s10 - s14));
- x15 = WRAPLOW(dct_const_round_shift(s11 - s15));
+ x0 = WRAPLOW(s0 + s4, bd);
+ x1 = WRAPLOW(s1 + s5, bd);
+ x2 = WRAPLOW(s2 + s6, bd);
+ x3 = WRAPLOW(s3 + s7, bd);
+ x4 = WRAPLOW(s0 - s4, bd);
+ x5 = WRAPLOW(s1 - s5, bd);
+ x6 = WRAPLOW(s2 - s6, bd);
+ x7 = WRAPLOW(s3 - s7, bd);
+ x8 = WRAPLOW(dct_const_round_shift(s8 + s12), bd);
+ x9 = WRAPLOW(dct_const_round_shift(s9 + s13), bd);
+ x10 = WRAPLOW(dct_const_round_shift(s10 + s14), bd);
+ x11 = WRAPLOW(dct_const_round_shift(s11 + s15), bd);
+ x12 = WRAPLOW(dct_const_round_shift(s8 - s12), bd);
+ x13 = WRAPLOW(dct_const_round_shift(s9 - s13), bd);
+ x14 = WRAPLOW(dct_const_round_shift(s10 - s14), bd);
+ x15 = WRAPLOW(dct_const_round_shift(s11 - s15), bd);
// stage 3
s0 = x0;
@@ -2205,22 +2225,22 @@
s14 = -x14 * cospi_24_64 + x15 * cospi_8_64;
s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
- x0 = WRAPLOW(s0 + s2);
- x1 = WRAPLOW(s1 + s3);
- x2 = WRAPLOW(s0 - s2);
- x3 = WRAPLOW(s1 - s3);
- x4 = WRAPLOW(dct_const_round_shift(s4 + s6));
- x5 = WRAPLOW(dct_const_round_shift(s5 + s7));
- x6 = WRAPLOW(dct_const_round_shift(s4 - s6));
- x7 = WRAPLOW(dct_const_round_shift(s5 - s7));
- x8 = WRAPLOW(s8 + s10);
- x9 = WRAPLOW(s9 + s11);
- x10 = WRAPLOW(s8 - s10);
- x11 = WRAPLOW(s9 - s11);
- x12 = WRAPLOW(dct_const_round_shift(s12 + s14));
- x13 = WRAPLOW(dct_const_round_shift(s13 + s15));
- x14 = WRAPLOW(dct_const_round_shift(s12 - s14));
- x15 = WRAPLOW(dct_const_round_shift(s13 - s15));
+ x0 = WRAPLOW(s0 + s2, bd);
+ x1 = WRAPLOW(s1 + s3, bd);
+ x2 = WRAPLOW(s0 - s2, bd);
+ x3 = WRAPLOW(s1 - s3, bd);
+ x4 = WRAPLOW(dct_const_round_shift(s4 + s6), bd);
+ x5 = WRAPLOW(dct_const_round_shift(s5 + s7), bd);
+ x6 = WRAPLOW(dct_const_round_shift(s4 - s6), bd);
+ x7 = WRAPLOW(dct_const_round_shift(s5 - s7), bd);
+ x8 = WRAPLOW(s8 + s10, bd);
+ x9 = WRAPLOW(s9 + s11, bd);
+ x10 = WRAPLOW(s8 - s10, bd);
+ x11 = WRAPLOW(s9 - s11, bd);
+ x12 = WRAPLOW(dct_const_round_shift(s12 + s14), bd);
+ x13 = WRAPLOW(dct_const_round_shift(s13 + s15), bd);
+ x14 = WRAPLOW(dct_const_round_shift(s12 - s14), bd);
+ x15 = WRAPLOW(dct_const_round_shift(s13 - s15), bd);
// stage 4
s2 = (- cospi_16_64) * (x2 + x3);
@@ -2232,47 +2252,47 @@
s14 = (- cospi_16_64) * (x14 + x15);
s15 = cospi_16_64 * (x14 - x15);
- x2 = WRAPLOW(dct_const_round_shift(s2));
- x3 = WRAPLOW(dct_const_round_shift(s3));
- x6 = WRAPLOW(dct_const_round_shift(s6));
- x7 = WRAPLOW(dct_const_round_shift(s7));
- x10 = WRAPLOW(dct_const_round_shift(s10));
- x11 = WRAPLOW(dct_const_round_shift(s11));
- x14 = WRAPLOW(dct_const_round_shift(s14));
- x15 = WRAPLOW(dct_const_round_shift(s15));
+ x2 = WRAPLOW(dct_const_round_shift(s2), bd);
+ x3 = WRAPLOW(dct_const_round_shift(s3), bd);
+ x6 = WRAPLOW(dct_const_round_shift(s6), bd);
+ x7 = WRAPLOW(dct_const_round_shift(s7), bd);
+ x10 = WRAPLOW(dct_const_round_shift(s10), bd);
+ x11 = WRAPLOW(dct_const_round_shift(s11), bd);
+ x14 = WRAPLOW(dct_const_round_shift(s14), bd);
+ x15 = WRAPLOW(dct_const_round_shift(s15), bd);
- output[0] = WRAPLOW(x0);
- output[1] = WRAPLOW(-x8);
- output[2] = WRAPLOW(x12);
- output[3] = WRAPLOW(-x4);
- output[4] = WRAPLOW(x6);
- output[5] = WRAPLOW(x14);
- output[6] = WRAPLOW(x10);
- output[7] = WRAPLOW(x2);
- output[8] = WRAPLOW(x3);
- output[9] = WRAPLOW(x11);
- output[10] = WRAPLOW(x15);
- output[11] = WRAPLOW(x7);
- output[12] = WRAPLOW(x5);
- output[13] = WRAPLOW(-x13);
- output[14] = WRAPLOW(x9);
- output[15] = WRAPLOW(-x1);
+ output[0] = WRAPLOW(x0, bd);
+ output[1] = WRAPLOW(-x8, bd);
+ output[2] = WRAPLOW(x12, bd);
+ output[3] = WRAPLOW(-x4, bd);
+ output[4] = WRAPLOW(x6, bd);
+ output[5] = WRAPLOW(x14, bd);
+ output[6] = WRAPLOW(x10, bd);
+ output[7] = WRAPLOW(x2, bd);
+ output[8] = WRAPLOW(x3, bd);
+ output[9] = WRAPLOW(x11, bd);
+ output[10] = WRAPLOW(x15, bd);
+ output[11] = WRAPLOW(x7, bd);
+ output[12] = WRAPLOW(x5, bd);
+ output[13] = WRAPLOW(-x13, bd);
+ output[14] = WRAPLOW(x9, bd);
+ output[15] = WRAPLOW(-x1, bd);
}
-static const high_transform_2d HIGH_IHT_16[] = {
- { high_idct16, high_idct16 }, // DCT_DCT = 0
- { high_iadst16, high_idct16 }, // ADST_DCT = 1
- { high_idct16, high_iadst16 }, // DCT_ADST = 2
- { high_iadst16, high_iadst16 } // ADST_ADST = 3
+static const highbd_transform_2d HIGH_IHT_16[] = {
+ { highbd_idct16, highbd_idct16 }, // DCT_DCT = 0
+ { highbd_iadst16, highbd_idct16 }, // ADST_DCT = 1
+ { highbd_idct16, highbd_iadst16 }, // DCT_ADST = 2
+ { highbd_iadst16, highbd_iadst16 } // ADST_ADST = 3
};
-void vp9_high_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int tx_type, int bd) {
+void vp9_highbd_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int tx_type, int bd) {
int i, j;
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
tran_low_t temp_in[16], temp_out[16];
- const high_transform_2d ht = HIGH_IHT_16[tx_type];
+ const highbd_transform_2d ht = HIGH_IHT_16[tx_type];
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
// Rows
@@ -2287,14 +2307,15 @@
for (j = 0; j < 16; ++j)
temp_in[j] = out[j * 16 + i];
ht.cols(temp_in, temp_out, bd);
- for (j = 0; j < 16; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ for (j = 0; j < 16; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
+ }
}
}
-void vp9_high_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
tran_low_t out[16 * 16] = { 0 };
tran_low_t *outptr = out;
int i, j;
@@ -2304,7 +2325,7 @@
// First transform rows. Since all non-zero dct coefficients are in
// upper-left 4x4 area, we only need to calculate first 4 rows here.
for (i = 0; i < 4; ++i) {
- high_idct16(input, outptr, bd);
+ highbd_idct16(input, outptr, bd);
input += 16;
outptr += 16;
}
@@ -2313,30 +2334,31 @@
for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j)
temp_in[j] = out[j*16 + i];
- high_idct16(temp_in, temp_out, bd);
- for (j = 0; j < 16; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ highbd_idct16(temp_in, temp_out, bd);
+ for (j = 0; j < 16; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
+ }
}
}
-void vp9_high_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
int i, j;
tran_high_t a1;
- tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
+ tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), bd);
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
- out = WRAPLOW(dct_const_round_shift(out * cospi_16_64));
+ out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), bd);
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 16; ++j) {
for (i = 0; i < 16; ++i)
- dest[i] = clip_pixel_bd_high(dest[i], a1, bd);
+ dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
dest += stride;
}
}
-static void high_idct32(const tran_low_t *input, tran_low_t *output, int bd) {
+static void highbd_idct32(const tran_low_t *input, tran_low_t *output, int bd) {
tran_low_t step1[32], step2[32];
tran_high_t temp1, temp2;
(void) bd;
@@ -2361,43 +2383,43 @@
temp1 = input[1] * cospi_31_64 - input[31] * cospi_1_64;
temp2 = input[1] * cospi_1_64 + input[31] * cospi_31_64;
- step1[16] = WRAPLOW(dct_const_round_shift(temp1));
- step1[31] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[16] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[31] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[17] * cospi_15_64 - input[15] * cospi_17_64;
temp2 = input[17] * cospi_17_64 + input[15] * cospi_15_64;
- step1[17] = WRAPLOW(dct_const_round_shift(temp1));
- step1[30] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[17] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[30] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[9] * cospi_23_64 - input[23] * cospi_9_64;
temp2 = input[9] * cospi_9_64 + input[23] * cospi_23_64;
- step1[18] = WRAPLOW(dct_const_round_shift(temp1));
- step1[29] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[18] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[29] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[25] * cospi_7_64 - input[7] * cospi_25_64;
temp2 = input[25] * cospi_25_64 + input[7] * cospi_7_64;
- step1[19] = WRAPLOW(dct_const_round_shift(temp1));
- step1[28] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[19] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[28] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[5] * cospi_27_64 - input[27] * cospi_5_64;
temp2 = input[5] * cospi_5_64 + input[27] * cospi_27_64;
- step1[20] = WRAPLOW(dct_const_round_shift(temp1));
- step1[27] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[20] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[27] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[21] * cospi_11_64 - input[11] * cospi_21_64;
temp2 = input[21] * cospi_21_64 + input[11] * cospi_11_64;
- step1[21] = WRAPLOW(dct_const_round_shift(temp1));
- step1[26] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[21] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[26] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[13] * cospi_19_64 - input[19] * cospi_13_64;
temp2 = input[13] * cospi_13_64 + input[19] * cospi_19_64;
- step1[22] = WRAPLOW(dct_const_round_shift(temp1));
- step1[25] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[22] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[25] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[29] * cospi_3_64 - input[3] * cospi_29_64;
temp2 = input[29] * cospi_29_64 + input[3] * cospi_3_64;
- step1[23] = WRAPLOW(dct_const_round_shift(temp1));
- step1[24] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[23] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[24] = WRAPLOW(dct_const_round_shift(temp2), bd);
// stage 2
step2[0] = step1[0];
@@ -2411,40 +2433,40 @@
temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
- step2[8] = WRAPLOW(dct_const_round_shift(temp1));
- step2[15] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[8] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[15] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
- step2[9] = WRAPLOW(dct_const_round_shift(temp1));
- step2[14] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[9] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[14] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
- step2[10] = WRAPLOW(dct_const_round_shift(temp1));
- step2[13] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
- step2[11] = WRAPLOW(dct_const_round_shift(temp1));
- step2[12] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[11] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[12] = WRAPLOW(dct_const_round_shift(temp2), bd);
- step2[16] = WRAPLOW(step1[16] + step1[17]);
- step2[17] = WRAPLOW(step1[16] - step1[17]);
- step2[18] = WRAPLOW(-step1[18] + step1[19]);
- step2[19] = WRAPLOW(step1[18] + step1[19]);
- step2[20] = WRAPLOW(step1[20] + step1[21]);
- step2[21] = WRAPLOW(step1[20] - step1[21]);
- step2[22] = WRAPLOW(-step1[22] + step1[23]);
- step2[23] = WRAPLOW(step1[22] + step1[23]);
- step2[24] = WRAPLOW(step1[24] + step1[25]);
- step2[25] = WRAPLOW(step1[24] - step1[25]);
- step2[26] = WRAPLOW(-step1[26] + step1[27]);
- step2[27] = WRAPLOW(step1[26] + step1[27]);
- step2[28] = WRAPLOW(step1[28] + step1[29]);
- step2[29] = WRAPLOW(step1[28] - step1[29]);
- step2[30] = WRAPLOW(-step1[30] + step1[31]);
- step2[31] = WRAPLOW(step1[30] + step1[31]);
+ step2[16] = WRAPLOW(step1[16] + step1[17], bd);
+ step2[17] = WRAPLOW(step1[16] - step1[17], bd);
+ step2[18] = WRAPLOW(-step1[18] + step1[19], bd);
+ step2[19] = WRAPLOW(step1[18] + step1[19], bd);
+ step2[20] = WRAPLOW(step1[20] + step1[21], bd);
+ step2[21] = WRAPLOW(step1[20] - step1[21], bd);
+ step2[22] = WRAPLOW(-step1[22] + step1[23], bd);
+ step2[23] = WRAPLOW(step1[22] + step1[23], bd);
+ step2[24] = WRAPLOW(step1[24] + step1[25], bd);
+ step2[25] = WRAPLOW(step1[24] - step1[25], bd);
+ step2[26] = WRAPLOW(-step1[26] + step1[27], bd);
+ step2[27] = WRAPLOW(step1[26] + step1[27], bd);
+ step2[28] = WRAPLOW(step1[28] + step1[29], bd);
+ step2[29] = WRAPLOW(step1[28] - step1[29], bd);
+ step2[30] = WRAPLOW(-step1[30] + step1[31], bd);
+ step2[31] = WRAPLOW(step1[30] + step1[31], bd);
// stage 3
step1[0] = step2[0];
@@ -2454,42 +2476,42 @@
temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
- step1[4] = WRAPLOW(dct_const_round_shift(temp1));
- step1[7] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[4] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[7] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
- step1[5] = WRAPLOW(dct_const_round_shift(temp1));
- step1[6] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), bd);
- step1[8] = WRAPLOW(step2[8] + step2[9]);
- step1[9] = WRAPLOW(step2[8] - step2[9]);
- step1[10] = WRAPLOW(-step2[10] + step2[11]);
- step1[11] = WRAPLOW(step2[10] + step2[11]);
- step1[12] = WRAPLOW(step2[12] + step2[13]);
- step1[13] = WRAPLOW(step2[12] - step2[13]);
- step1[14] = WRAPLOW(-step2[14] + step2[15]);
- step1[15] = WRAPLOW(step2[14] + step2[15]);
+ step1[8] = WRAPLOW(step2[8] + step2[9], bd);
+ step1[9] = WRAPLOW(step2[8] - step2[9], bd);
+ step1[10] = WRAPLOW(-step2[10] + step2[11], bd);
+ step1[11] = WRAPLOW(step2[10] + step2[11], bd);
+ step1[12] = WRAPLOW(step2[12] + step2[13], bd);
+ step1[13] = WRAPLOW(step2[12] - step2[13], bd);
+ step1[14] = WRAPLOW(-step2[14] + step2[15], bd);
+ step1[15] = WRAPLOW(step2[14] + step2[15], bd);
step1[16] = step2[16];
step1[31] = step2[31];
temp1 = -step2[17] * cospi_4_64 + step2[30] * cospi_28_64;
temp2 = step2[17] * cospi_28_64 + step2[30] * cospi_4_64;
- step1[17] = WRAPLOW(dct_const_round_shift(temp1));
- step1[30] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[17] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[30] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = -step2[18] * cospi_28_64 - step2[29] * cospi_4_64;
temp2 = -step2[18] * cospi_4_64 + step2[29] * cospi_28_64;
- step1[18] = WRAPLOW(dct_const_round_shift(temp1));
- step1[29] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[18] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[29] = WRAPLOW(dct_const_round_shift(temp2), bd);
step1[19] = step2[19];
step1[20] = step2[20];
temp1 = -step2[21] * cospi_20_64 + step2[26] * cospi_12_64;
temp2 = step2[21] * cospi_12_64 + step2[26] * cospi_20_64;
- step1[21] = WRAPLOW(dct_const_round_shift(temp1));
- step1[26] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[21] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[26] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = -step2[22] * cospi_12_64 - step2[25] * cospi_20_64;
temp2 = -step2[22] * cospi_20_64 + step2[25] * cospi_12_64;
- step1[22] = WRAPLOW(dct_const_round_shift(temp1));
- step1[25] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[22] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[25] = WRAPLOW(dct_const_round_shift(temp2), bd);
step1[23] = step2[23];
step1[24] = step2[24];
step1[27] = step2[27];
@@ -2498,87 +2520,87 @@
// stage 4
temp1 = (step1[0] + step1[1]) * cospi_16_64;
temp2 = (step1[0] - step1[1]) * cospi_16_64;
- step2[0] = WRAPLOW(dct_const_round_shift(temp1));
- step2[1] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[0] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[1] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
- step2[2] = WRAPLOW(dct_const_round_shift(temp1));
- step2[3] = WRAPLOW(dct_const_round_shift(temp2));
- step2[4] = WRAPLOW(step1[4] + step1[5]);
- step2[5] = WRAPLOW(step1[4] - step1[5]);
- step2[6] = WRAPLOW(-step1[6] + step1[7]);
- step2[7] = WRAPLOW(step1[6] + step1[7]);
+ step2[2] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[3] = WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[4] = WRAPLOW(step1[4] + step1[5], bd);
+ step2[5] = WRAPLOW(step1[4] - step1[5], bd);
+ step2[6] = WRAPLOW(-step1[6] + step1[7], bd);
+ step2[7] = WRAPLOW(step1[6] + step1[7], bd);
step2[8] = step1[8];
step2[15] = step1[15];
temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
- step2[9] = WRAPLOW(dct_const_round_shift(temp1));
- step2[14] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[9] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[14] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
- step2[10] = WRAPLOW(dct_const_round_shift(temp1));
- step2[13] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), bd);
step2[11] = step1[11];
step2[12] = step1[12];
- step2[16] = WRAPLOW(step1[16] + step1[19]);
- step2[17] = WRAPLOW(step1[17] + step1[18]);
- step2[18] = WRAPLOW(step1[17] - step1[18]);
- step2[19] = WRAPLOW(step1[16] - step1[19]);
- step2[20] = WRAPLOW(-step1[20] + step1[23]);
- step2[21] = WRAPLOW(-step1[21] + step1[22]);
- step2[22] = WRAPLOW(step1[21] + step1[22]);
- step2[23] = WRAPLOW(step1[20] + step1[23]);
+ step2[16] = WRAPLOW(step1[16] + step1[19], bd);
+ step2[17] = WRAPLOW(step1[17] + step1[18], bd);
+ step2[18] = WRAPLOW(step1[17] - step1[18], bd);
+ step2[19] = WRAPLOW(step1[16] - step1[19], bd);
+ step2[20] = WRAPLOW(-step1[20] + step1[23], bd);
+ step2[21] = WRAPLOW(-step1[21] + step1[22], bd);
+ step2[22] = WRAPLOW(step1[21] + step1[22], bd);
+ step2[23] = WRAPLOW(step1[20] + step1[23], bd);
- step2[24] = WRAPLOW(step1[24] + step1[27]);
- step2[25] = WRAPLOW(step1[25] + step1[26]);
- step2[26] = WRAPLOW(step1[25] - step1[26]);
- step2[27] = WRAPLOW(step1[24] - step1[27]);
- step2[28] = WRAPLOW(-step1[28] + step1[31]);
- step2[29] = WRAPLOW(-step1[29] + step1[30]);
- step2[30] = WRAPLOW(step1[29] + step1[30]);
- step2[31] = WRAPLOW(step1[28] + step1[31]);
+ step2[24] = WRAPLOW(step1[24] + step1[27], bd);
+ step2[25] = WRAPLOW(step1[25] + step1[26], bd);
+ step2[26] = WRAPLOW(step1[25] - step1[26], bd);
+ step2[27] = WRAPLOW(step1[24] - step1[27], bd);
+ step2[28] = WRAPLOW(-step1[28] + step1[31], bd);
+ step2[29] = WRAPLOW(-step1[29] + step1[30], bd);
+ step2[30] = WRAPLOW(step1[29] + step1[30], bd);
+ step2[31] = WRAPLOW(step1[28] + step1[31], bd);
// stage 5
- step1[0] = WRAPLOW(step2[0] + step2[3]);
- step1[1] = WRAPLOW(step2[1] + step2[2]);
- step1[2] = WRAPLOW(step2[1] - step2[2]);
- step1[3] = WRAPLOW(step2[0] - step2[3]);
+ step1[0] = WRAPLOW(step2[0] + step2[3], bd);
+ step1[1] = WRAPLOW(step2[1] + step2[2], bd);
+ step1[2] = WRAPLOW(step2[1] - step2[2], bd);
+ step1[3] = WRAPLOW(step2[0] - step2[3], bd);
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = WRAPLOW(dct_const_round_shift(temp1));
- step1[6] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[5] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = WRAPLOW(dct_const_round_shift(temp2), bd);
step1[7] = step2[7];
- step1[8] = WRAPLOW(step2[8] + step2[11]);
- step1[9] = WRAPLOW(step2[9] + step2[10]);
- step1[10] = WRAPLOW(step2[9] - step2[10]);
- step1[11] = WRAPLOW(step2[8] - step2[11]);
- step1[12] = WRAPLOW(-step2[12] + step2[15]);
- step1[13] = WRAPLOW(-step2[13] + step2[14]);
- step1[14] = WRAPLOW(step2[13] + step2[14]);
- step1[15] = WRAPLOW(step2[12] + step2[15]);
+ step1[8] = WRAPLOW(step2[8] + step2[11], bd);
+ step1[9] = WRAPLOW(step2[9] + step2[10], bd);
+ step1[10] = WRAPLOW(step2[9] - step2[10], bd);
+ step1[11] = WRAPLOW(step2[8] - step2[11], bd);
+ step1[12] = WRAPLOW(-step2[12] + step2[15], bd);
+ step1[13] = WRAPLOW(-step2[13] + step2[14], bd);
+ step1[14] = WRAPLOW(step2[13] + step2[14], bd);
+ step1[15] = WRAPLOW(step2[12] + step2[15], bd);
step1[16] = step2[16];
step1[17] = step2[17];
temp1 = -step2[18] * cospi_8_64 + step2[29] * cospi_24_64;
temp2 = step2[18] * cospi_24_64 + step2[29] * cospi_8_64;
- step1[18] = WRAPLOW(dct_const_round_shift(temp1));
- step1[29] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[18] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[29] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = -step2[19] * cospi_8_64 + step2[28] * cospi_24_64;
temp2 = step2[19] * cospi_24_64 + step2[28] * cospi_8_64;
- step1[19] = WRAPLOW(dct_const_round_shift(temp1));
- step1[28] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[19] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[28] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = -step2[20] * cospi_24_64 - step2[27] * cospi_8_64;
temp2 = -step2[20] * cospi_8_64 + step2[27] * cospi_24_64;
- step1[20] = WRAPLOW(dct_const_round_shift(temp1));
- step1[27] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[20] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[27] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = -step2[21] * cospi_24_64 - step2[26] * cospi_8_64;
temp2 = -step2[21] * cospi_8_64 + step2[26] * cospi_24_64;
- step1[21] = WRAPLOW(dct_const_round_shift(temp1));
- step1[26] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[21] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[26] = WRAPLOW(dct_const_round_shift(temp2), bd);
step1[22] = step2[22];
step1[23] = step2[23];
step1[24] = step2[24];
@@ -2587,62 +2609,62 @@
step1[31] = step2[31];
// stage 6
- step2[0] = WRAPLOW(step1[0] + step1[7]);
- step2[1] = WRAPLOW(step1[1] + step1[6]);
- step2[2] = WRAPLOW(step1[2] + step1[5]);
- step2[3] = WRAPLOW(step1[3] + step1[4]);
- step2[4] = WRAPLOW(step1[3] - step1[4]);
- step2[5] = WRAPLOW(step1[2] - step1[5]);
- step2[6] = WRAPLOW(step1[1] - step1[6]);
- step2[7] = WRAPLOW(step1[0] - step1[7]);
+ step2[0] = WRAPLOW(step1[0] + step1[7], bd);
+ step2[1] = WRAPLOW(step1[1] + step1[6], bd);
+ step2[2] = WRAPLOW(step1[2] + step1[5], bd);
+ step2[3] = WRAPLOW(step1[3] + step1[4], bd);
+ step2[4] = WRAPLOW(step1[3] - step1[4], bd);
+ step2[5] = WRAPLOW(step1[2] - step1[5], bd);
+ step2[6] = WRAPLOW(step1[1] - step1[6], bd);
+ step2[7] = WRAPLOW(step1[0] - step1[7], bd);
step2[8] = step1[8];
step2[9] = step1[9];
temp1 = (-step1[10] + step1[13]) * cospi_16_64;
temp2 = (step1[10] + step1[13]) * cospi_16_64;
- step2[10] = WRAPLOW(dct_const_round_shift(temp1));
- step2[13] = WRAPLOW(dct_const_round_shift(temp2));
+ step2[10] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = (-step1[11] + step1[12]) * cospi_16_64;
temp2 = (step1[11] + step1[12]) * cospi_16_64;
- step2[11] = WRAPLOW(dct_const_round_shift(temp1));
- step2[12] = WRAPLOW(dct_const_round_shift(temp2));
- step2[14] = WRAPLOW(step1[14]);
- step2[15] = WRAPLOW(step1[15]);
+ step2[11] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[12] = WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[14] = step1[14];
+ step2[15] = step1[15];
- step2[16] = WRAPLOW(step1[16] + step1[23]);
- step2[17] = WRAPLOW(step1[17] + step1[22]);
- step2[18] = WRAPLOW(step1[18] + step1[21]);
- step2[19] = WRAPLOW(step1[19] + step1[20]);
- step2[20] = WRAPLOW(step1[19] - step1[20]);
- step2[21] = WRAPLOW(step1[18] - step1[21]);
- step2[22] = WRAPLOW(step1[17] - step1[22]);
- step2[23] = WRAPLOW(step1[16] - step1[23]);
+ step2[16] = WRAPLOW(step1[16] + step1[23], bd);
+ step2[17] = WRAPLOW(step1[17] + step1[22], bd);
+ step2[18] = WRAPLOW(step1[18] + step1[21], bd);
+ step2[19] = WRAPLOW(step1[19] + step1[20], bd);
+ step2[20] = WRAPLOW(step1[19] - step1[20], bd);
+ step2[21] = WRAPLOW(step1[18] - step1[21], bd);
+ step2[22] = WRAPLOW(step1[17] - step1[22], bd);
+ step2[23] = WRAPLOW(step1[16] - step1[23], bd);
- step2[24] = WRAPLOW(-step1[24] + step1[31]);
- step2[25] = WRAPLOW(-step1[25] + step1[30]);
- step2[26] = WRAPLOW(-step1[26] + step1[29]);
- step2[27] = WRAPLOW(-step1[27] + step1[28]);
- step2[28] = WRAPLOW(step1[27] + step1[28]);
- step2[29] = WRAPLOW(step1[26] + step1[29]);
- step2[30] = WRAPLOW(step1[25] + step1[30]);
- step2[31] = WRAPLOW(step1[24] + step1[31]);
+ step2[24] = WRAPLOW(-step1[24] + step1[31], bd);
+ step2[25] = WRAPLOW(-step1[25] + step1[30], bd);
+ step2[26] = WRAPLOW(-step1[26] + step1[29], bd);
+ step2[27] = WRAPLOW(-step1[27] + step1[28], bd);
+ step2[28] = WRAPLOW(step1[27] + step1[28], bd);
+ step2[29] = WRAPLOW(step1[26] + step1[29], bd);
+ step2[30] = WRAPLOW(step1[25] + step1[30], bd);
+ step2[31] = WRAPLOW(step1[24] + step1[31], bd);
// stage 7
- step1[0] = WRAPLOW(step2[0] + step2[15]);
- step1[1] = WRAPLOW(step2[1] + step2[14]);
- step1[2] = WRAPLOW(step2[2] + step2[13]);
- step1[3] = WRAPLOW(step2[3] + step2[12]);
- step1[4] = WRAPLOW(step2[4] + step2[11]);
- step1[5] = WRAPLOW(step2[5] + step2[10]);
- step1[6] = WRAPLOW(step2[6] + step2[9]);
- step1[7] = WRAPLOW(step2[7] + step2[8]);
- step1[8] = WRAPLOW(step2[7] - step2[8]);
- step1[9] = WRAPLOW(step2[6] - step2[9]);
- step1[10] = WRAPLOW(step2[5] - step2[10]);
- step1[11] = WRAPLOW(step2[4] - step2[11]);
- step1[12] = WRAPLOW(step2[3] - step2[12]);
- step1[13] = WRAPLOW(step2[2] - step2[13]);
- step1[14] = WRAPLOW(step2[1] - step2[14]);
- step1[15] = WRAPLOW(step2[0] - step2[15]);
+ step1[0] = WRAPLOW(step2[0] + step2[15], bd);
+ step1[1] = WRAPLOW(step2[1] + step2[14], bd);
+ step1[2] = WRAPLOW(step2[2] + step2[13], bd);
+ step1[3] = WRAPLOW(step2[3] + step2[12], bd);
+ step1[4] = WRAPLOW(step2[4] + step2[11], bd);
+ step1[5] = WRAPLOW(step2[5] + step2[10], bd);
+ step1[6] = WRAPLOW(step2[6] + step2[9], bd);
+ step1[7] = WRAPLOW(step2[7] + step2[8], bd);
+ step1[8] = WRAPLOW(step2[7] - step2[8], bd);
+ step1[9] = WRAPLOW(step2[6] - step2[9], bd);
+ step1[10] = WRAPLOW(step2[5] - step2[10], bd);
+ step1[11] = WRAPLOW(step2[4] - step2[11], bd);
+ step1[12] = WRAPLOW(step2[3] - step2[12], bd);
+ step1[13] = WRAPLOW(step2[2] - step2[13], bd);
+ step1[14] = WRAPLOW(step2[1] - step2[14], bd);
+ step1[15] = WRAPLOW(step2[0] - step2[15], bd);
step1[16] = step2[16];
step1[17] = step2[17];
@@ -2650,62 +2672,62 @@
step1[19] = step2[19];
temp1 = (-step2[20] + step2[27]) * cospi_16_64;
temp2 = (step2[20] + step2[27]) * cospi_16_64;
- step1[20] = WRAPLOW(dct_const_round_shift(temp1));
- step1[27] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[20] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[27] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = (-step2[21] + step2[26]) * cospi_16_64;
temp2 = (step2[21] + step2[26]) * cospi_16_64;
- step1[21] = WRAPLOW(dct_const_round_shift(temp1));
- step1[26] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[21] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[26] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = (-step2[22] + step2[25]) * cospi_16_64;
temp2 = (step2[22] + step2[25]) * cospi_16_64;
- step1[22] = WRAPLOW(dct_const_round_shift(temp1));
- step1[25] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[22] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[25] = WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = (-step2[23] + step2[24]) * cospi_16_64;
temp2 = (step2[23] + step2[24]) * cospi_16_64;
- step1[23] = WRAPLOW(dct_const_round_shift(temp1));
- step1[24] = WRAPLOW(dct_const_round_shift(temp2));
+ step1[23] = WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[24] = WRAPLOW(dct_const_round_shift(temp2), bd);
step1[28] = step2[28];
step1[29] = step2[29];
step1[30] = step2[30];
step1[31] = step2[31];
// final stage
- output[0] = WRAPLOW(step1[0] + step1[31]);
- output[1] = WRAPLOW(step1[1] + step1[30]);
- output[2] = WRAPLOW(step1[2] + step1[29]);
- output[3] = WRAPLOW(step1[3] + step1[28]);
- output[4] = WRAPLOW(step1[4] + step1[27]);
- output[5] = WRAPLOW(step1[5] + step1[26]);
- output[6] = WRAPLOW(step1[6] + step1[25]);
- output[7] = WRAPLOW(step1[7] + step1[24]);
- output[8] = WRAPLOW(step1[8] + step1[23]);
- output[9] = WRAPLOW(step1[9] + step1[22]);
- output[10] = WRAPLOW(step1[10] + step1[21]);
- output[11] = WRAPLOW(step1[11] + step1[20]);
- output[12] = WRAPLOW(step1[12] + step1[19]);
- output[13] = WRAPLOW(step1[13] + step1[18]);
- output[14] = WRAPLOW(step1[14] + step1[17]);
- output[15] = WRAPLOW(step1[15] + step1[16]);
- output[16] = WRAPLOW(step1[15] - step1[16]);
- output[17] = WRAPLOW(step1[14] - step1[17]);
- output[18] = WRAPLOW(step1[13] - step1[18]);
- output[19] = WRAPLOW(step1[12] - step1[19]);
- output[20] = WRAPLOW(step1[11] - step1[20]);
- output[21] = WRAPLOW(step1[10] - step1[21]);
- output[22] = WRAPLOW(step1[9] - step1[22]);
- output[23] = WRAPLOW(step1[8] - step1[23]);
- output[24] = WRAPLOW(step1[7] - step1[24]);
- output[25] = WRAPLOW(step1[6] - step1[25]);
- output[26] = WRAPLOW(step1[5] - step1[26]);
- output[27] = WRAPLOW(step1[4] - step1[27]);
- output[28] = WRAPLOW(step1[3] - step1[28]);
- output[29] = WRAPLOW(step1[2] - step1[29]);
- output[30] = WRAPLOW(step1[1] - step1[30]);
- output[31] = WRAPLOW(step1[0] - step1[31]);
+ output[0] = WRAPLOW(step1[0] + step1[31], bd);
+ output[1] = WRAPLOW(step1[1] + step1[30], bd);
+ output[2] = WRAPLOW(step1[2] + step1[29], bd);
+ output[3] = WRAPLOW(step1[3] + step1[28], bd);
+ output[4] = WRAPLOW(step1[4] + step1[27], bd);
+ output[5] = WRAPLOW(step1[5] + step1[26], bd);
+ output[6] = WRAPLOW(step1[6] + step1[25], bd);
+ output[7] = WRAPLOW(step1[7] + step1[24], bd);
+ output[8] = WRAPLOW(step1[8] + step1[23], bd);
+ output[9] = WRAPLOW(step1[9] + step1[22], bd);
+ output[10] = WRAPLOW(step1[10] + step1[21], bd);
+ output[11] = WRAPLOW(step1[11] + step1[20], bd);
+ output[12] = WRAPLOW(step1[12] + step1[19], bd);
+ output[13] = WRAPLOW(step1[13] + step1[18], bd);
+ output[14] = WRAPLOW(step1[14] + step1[17], bd);
+ output[15] = WRAPLOW(step1[15] + step1[16], bd);
+ output[16] = WRAPLOW(step1[15] - step1[16], bd);
+ output[17] = WRAPLOW(step1[14] - step1[17], bd);
+ output[18] = WRAPLOW(step1[13] - step1[18], bd);
+ output[19] = WRAPLOW(step1[12] - step1[19], bd);
+ output[20] = WRAPLOW(step1[11] - step1[20], bd);
+ output[21] = WRAPLOW(step1[10] - step1[21], bd);
+ output[22] = WRAPLOW(step1[9] - step1[22], bd);
+ output[23] = WRAPLOW(step1[8] - step1[23], bd);
+ output[24] = WRAPLOW(step1[7] - step1[24], bd);
+ output[25] = WRAPLOW(step1[6] - step1[25], bd);
+ output[26] = WRAPLOW(step1[5] - step1[26], bd);
+ output[27] = WRAPLOW(step1[4] - step1[27], bd);
+ output[28] = WRAPLOW(step1[3] - step1[28], bd);
+ output[29] = WRAPLOW(step1[2] - step1[29], bd);
+ output[30] = WRAPLOW(step1[1] - step1[30], bd);
+ output[31] = WRAPLOW(step1[0] - step1[31], bd);
}
-void vp9_high_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
tran_low_t out[32 * 32];
tran_low_t *outptr = out;
int i, j;
@@ -2725,7 +2747,7 @@
zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];
if (zero_coeff[0] | zero_coeff[1])
- high_idct32(input, outptr, bd);
+ highbd_idct32(input, outptr, bd);
else
vpx_memset(outptr, 0, sizeof(tran_low_t) * 32);
input += 32;
@@ -2736,15 +2758,16 @@
for (i = 0; i < 32; ++i) {
for (j = 0; j < 32; ++j)
temp_in[j] = out[j * 32 + i];
- high_idct32(temp_in, temp_out, bd);
- for (j = 0; j < 32; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ highbd_idct32(temp_in, temp_out, bd);
+ for (j = 0; j < 32; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
+ }
}
}
-void vp9_high_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
tran_low_t out[32 * 32] = {0};
tran_low_t *outptr = out;
int i, j;
@@ -2754,7 +2777,7 @@
// Rows
// Only upper-left 8x8 has non-zero coeff.
for (i = 0; i < 8; ++i) {
- high_idct32(input, outptr, bd);
+ highbd_idct32(input, outptr, bd);
input += 32;
outptr += 32;
}
@@ -2762,50 +2785,51 @@
for (i = 0; i < 32; ++i) {
for (j = 0; j < 32; ++j)
temp_in[j] = out[j * 32 + i];
- high_idct32(temp_in, temp_out, bd);
- for (j = 0; j < 32; ++j)
- dest[j * stride + i] = clip_pixel_bd_high(
+ highbd_idct32(temp_in, temp_out, bd);
+ for (j = 0; j < 32; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
+ }
}
}
-void vp9_high_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest8,
- int stride, int bd) {
+void vp9_highbd_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
int i, j;
int a1;
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
- tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
- out = WRAPLOW(dct_const_round_shift(out * cospi_16_64));
+ tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), bd);
+ out = WRAPLOW(dct_const_round_shift(out * cospi_16_64), bd);
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 32; ++j) {
for (i = 0; i < 32; ++i)
- dest[i] = clip_pixel_bd_high(dest[i], a1, bd);
+ dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
dest += stride;
}
}
// idct
-void vp9_high_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd) {
+void vp9_highbd_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, int bd) {
if (eob > 1)
- vp9_high_idct4x4_16_add(input, dest, stride, bd);
+ vp9_highbd_idct4x4_16_add(input, dest, stride, bd);
else
- vp9_high_idct4x4_1_add(input, dest, stride, bd);
+ vp9_highbd_idct4x4_1_add(input, dest, stride, bd);
}
-void vp9_high_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd) {
+void vp9_highbd_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, int bd) {
if (eob > 1)
- vp9_high_iwht4x4_16_add(input, dest, stride, bd);
+ vp9_highbd_iwht4x4_16_add(input, dest, stride, bd);
else
- vp9_high_iwht4x4_1_add(input, dest, stride, bd);
+ vp9_highbd_iwht4x4_1_add(input, dest, stride, bd);
}
-void vp9_high_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd) {
+void vp9_highbd_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, int bd) {
// If dc is 1, then input[0] is the reconstructed value, do not need
// dequantization. Also, when dc is 1, dc is counted in eobs, namely eobs >=1.
@@ -2815,64 +2839,64 @@
// Combine that with code here.
// DC only DCT coefficient
if (eob == 1) {
- vp9_high_idct8x8_1_add(input, dest, stride, bd);
+ vp9_highbd_idct8x8_1_add(input, dest, stride, bd);
} else if (eob <= 10) {
- vp9_high_idct8x8_10_add(input, dest, stride, bd);
+ vp9_highbd_idct8x8_10_add(input, dest, stride, bd);
} else {
- vp9_high_idct8x8_64_add(input, dest, stride, bd);
+ vp9_highbd_idct8x8_64_add(input, dest, stride, bd);
}
}
-void vp9_high_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd) {
+void vp9_highbd_idct16x16_add(const tran_low_t *input, uint8_t *dest,
+ int stride, int eob, int bd) {
// The calculation can be simplified if there are not many non-zero dct
// coefficients. Use eobs to separate different cases.
// DC only DCT coefficient.
if (eob == 1) {
- vp9_high_idct16x16_1_add(input, dest, stride, bd);
+ vp9_highbd_idct16x16_1_add(input, dest, stride, bd);
} else if (eob <= 10) {
- vp9_high_idct16x16_10_add(input, dest, stride, bd);
+ vp9_highbd_idct16x16_10_add(input, dest, stride, bd);
} else {
- vp9_high_idct16x16_256_add(input, dest, stride, bd);
+ vp9_highbd_idct16x16_256_add(input, dest, stride, bd);
}
}
-void vp9_high_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd) {
+void vp9_highbd_idct32x32_add(const tran_low_t *input, uint8_t *dest,
+ int stride, int eob, int bd) {
// Non-zero coeff only in upper-left 8x8
if (eob == 1) {
- vp9_high_idct32x32_1_add(input, dest, stride, bd);
+ vp9_highbd_idct32x32_1_add(input, dest, stride, bd);
} else if (eob <= 34) {
- vp9_high_idct32x32_34_add(input, dest, stride, bd);
+ vp9_highbd_idct32x32_34_add(input, dest, stride, bd);
} else {
- vp9_high_idct32x32_1024_add(input, dest, stride, bd);
+ vp9_highbd_idct32x32_1024_add(input, dest, stride, bd);
}
}
// iht
-void vp9_high_iht4x4_add(TX_TYPE tx_type, const tran_low_t *input,
- uint8_t *dest, int stride, int eob, int bd) {
+void vp9_highbd_iht4x4_add(TX_TYPE tx_type, const tran_low_t *input,
+ uint8_t *dest, int stride, int eob, int bd) {
if (tx_type == DCT_DCT)
- vp9_high_idct4x4_add(input, dest, stride, eob, bd);
+ vp9_highbd_idct4x4_add(input, dest, stride, eob, bd);
else
- vp9_high_iht4x4_16_add(input, dest, stride, tx_type, bd);
+ vp9_highbd_iht4x4_16_add(input, dest, stride, tx_type, bd);
}
-void vp9_high_iht8x8_add(TX_TYPE tx_type, const tran_low_t *input,
- uint8_t *dest, int stride, int eob, int bd) {
+void vp9_highbd_iht8x8_add(TX_TYPE tx_type, const tran_low_t *input,
+ uint8_t *dest, int stride, int eob, int bd) {
if (tx_type == DCT_DCT) {
- vp9_high_idct8x8_add(input, dest, stride, eob, bd);
+ vp9_highbd_idct8x8_add(input, dest, stride, eob, bd);
} else {
- vp9_high_iht8x8_64_add(input, dest, stride, tx_type, bd);
+ vp9_highbd_iht8x8_64_add(input, dest, stride, tx_type, bd);
}
}
-void vp9_high_iht16x16_add(TX_TYPE tx_type, const tran_low_t *input,
+void vp9_highbd_iht16x16_add(TX_TYPE tx_type, const tran_low_t *input,
uint8_t *dest, int stride, int eob, int bd) {
if (tx_type == DCT_DCT) {
- vp9_high_idct16x16_add(input, dest, stride, eob, bd);
+ vp9_highbd_idct16x16_add(input, dest, stride, eob, bd);
} else {
- vp9_high_iht16x16_256_add(input, dest, stride, tx_type, bd);
+ vp9_highbd_iht16x16_256_add(input, dest, stride, tx_type, bd);
}
}
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/source/libvpx/vp9/common/vp9_idct.h b/source/libvpx/vp9/common/vp9_idct.h
index 694be3c..12569b9 100644
--- a/source/libvpx/vp9/common/vp9_idct.h
+++ b/source/libvpx/vp9/common/vp9_idct.h
@@ -14,7 +14,6 @@
#include <assert.h>
#include "./vpx_config.h"
-#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
@@ -22,7 +21,6 @@
extern "C" {
#endif
-
// Constants and Macros used by all idct/dct functions
#define DCT_CONST_BITS 14
#define DCT_CONST_ROUNDING (1 << (DCT_CONST_BITS - 1))
@@ -36,17 +34,6 @@
#define dual_set_epi16(a, b) \
_mm_set_epi16(b, b, b, b, a, a, a, a)
-// Note:
-// tran_low_t is the datatype used for final transform coefficients.
-// tran_high_t is the datatype used for intermediate transform stages.
-#if CONFIG_VP9_HIGHBITDEPTH
-typedef int64_t tran_high_t;
-typedef int32_t tran_low_t;
-#else
-typedef int32_t tran_high_t;
-typedef int16_t tran_low_t;
-#endif
-
// Constants:
// for (int i = 1; i< 32; ++i)
// printf("static const int cospi_%d_64 = %.0f;\n", i,
@@ -90,8 +77,7 @@
static const tran_high_t sinpi_3_9 = 13377;
static const tran_high_t sinpi_4_9 = 15212;
-static INLINE tran_low_t dct_const_round_shift(tran_high_t input) {
- tran_high_t rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS);
+static INLINE tran_low_t check_range(tran_high_t input) {
#if CONFIG_VP9_HIGHBITDEPTH
// For valid highbitdepth VP9 streams, intermediate stage coefficients will
// stay within the ranges:
@@ -105,10 +91,15 @@
// this range for every intermediate coefficient can burdensome for a decoder,
// therefore the following assertion is only enabled when configured with
// --enable-coefficient-range-checking.
- assert(INT16_MIN <= rv);
- assert(rv <= INT16_MAX);
+ assert(INT16_MIN <= input);
+ assert(input <= INT16_MAX);
#endif
- return (tran_low_t)rv;
+ return (tran_low_t)input;
+}
+
+static INLINE tran_low_t dct_const_round_shift(tran_high_t input) {
+ tran_high_t rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS);
+ return check_range(rv);
}
typedef void (*transform_1d)(const tran_low_t*, tran_low_t*);
@@ -118,11 +109,11 @@
} transform_2d;
#if CONFIG_VP9_HIGHBITDEPTH
-typedef void (*high_transform_1d)(const tran_low_t*, tran_low_t*, int bd);
+typedef void (*highbd_transform_1d)(const tran_low_t*, tran_low_t*, int bd);
typedef struct {
- high_transform_1d cols, rows; // vertical and horizontal
-} high_transform_2d;
+ highbd_transform_1d cols, rows; // vertical and horizontal
+} highbd_transform_2d;
#endif // CONFIG_VP9_HIGHBITDEPTH
void vp9_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
@@ -144,22 +135,22 @@
int stride, int eob);
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd);
-void vp9_high_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd);
-void vp9_high_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd);
-void vp9_high_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride,
+void vp9_highbd_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob, int bd);
-void vp9_high_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride,
+void vp9_highbd_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob, int bd);
-void vp9_high_iht4x4_add(TX_TYPE tx_type, const tran_low_t *input,
- uint8_t *dest, int stride, int eob, int bd);
-void vp9_high_iht8x8_add(TX_TYPE tx_type, const tran_low_t *input,
- uint8_t *dest, int stride, int eob, int bd);
-void vp9_high_iht16x16_add(TX_TYPE tx_type, const tran_low_t *input,
+void vp9_highbd_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, int bd);
+void vp9_highbd_idct16x16_add(const tran_low_t *input, uint8_t *dest,
+ int stride, int eob, int bd);
+void vp9_highbd_idct32x32_add(const tran_low_t *input, uint8_t *dest,
+ int stride, int eob, int bd);
+void vp9_highbd_iht4x4_add(TX_TYPE tx_type, const tran_low_t *input,
uint8_t *dest, int stride, int eob, int bd);
+void vp9_highbd_iht8x8_add(TX_TYPE tx_type, const tran_low_t *input,
+ uint8_t *dest, int stride, int eob, int bd);
+void vp9_highbd_iht16x16_add(TX_TYPE tx_type, const tran_low_t *input,
+ uint8_t *dest, int stride, int eob, int bd);
#endif // CONFIG_VP9_HIGHBITDEPTH
#ifdef __cplusplus
} // extern "C"
diff --git a/source/libvpx/vp9/common/vp9_loopfilter.c b/source/libvpx/vp9/common/vp9_loopfilter.c
index 102eb71..aca8d7b 100644
--- a/source/libvpx/vp9/common/vp9_loopfilter.c
+++ b/source/libvpx/vp9/common/vp9_loopfilter.c
@@ -392,6 +392,107 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static void highbd_filter_selectively_vert_row2(PLANE_TYPE plane_type,
+ uint16_t *s, int pitch,
+ unsigned int mask_16x16_l,
+ unsigned int mask_8x8_l,
+ unsigned int mask_4x4_l,
+ unsigned int mask_4x4_int_l,
+ const loop_filter_info_n *lfi_n,
+ const uint8_t *lfl, int bd) {
+ const int mask_shift = plane_type ? 4 : 8;
+ const int mask_cutoff = plane_type ? 0xf : 0xff;
+ const int lfl_forward = plane_type ? 4 : 8;
+
+ unsigned int mask_16x16_0 = mask_16x16_l & mask_cutoff;
+ unsigned int mask_8x8_0 = mask_8x8_l & mask_cutoff;
+ unsigned int mask_4x4_0 = mask_4x4_l & mask_cutoff;
+ unsigned int mask_4x4_int_0 = mask_4x4_int_l & mask_cutoff;
+ unsigned int mask_16x16_1 = (mask_16x16_l >> mask_shift) & mask_cutoff;
+ unsigned int mask_8x8_1 = (mask_8x8_l >> mask_shift) & mask_cutoff;
+ unsigned int mask_4x4_1 = (mask_4x4_l >> mask_shift) & mask_cutoff;
+ unsigned int mask_4x4_int_1 = (mask_4x4_int_l >> mask_shift) & mask_cutoff;
+ unsigned int mask;
+
+ for (mask = mask_16x16_0 | mask_8x8_0 | mask_4x4_0 | mask_4x4_int_0 |
+ mask_16x16_1 | mask_8x8_1 | mask_4x4_1 | mask_4x4_int_1;
+ mask; mask >>= 1) {
+ const loop_filter_thresh *lfi0 = lfi_n->lfthr + *lfl;
+ const loop_filter_thresh *lfi1 = lfi_n->lfthr + *(lfl + lfl_forward);
+
+ // TODO(yunqingwang): count in loopfilter functions should be removed.
+ if (mask & 1) {
+ if ((mask_16x16_0 | mask_16x16_1) & 1) {
+ if ((mask_16x16_0 & mask_16x16_1) & 1) {
+ vp9_highbd_lpf_vertical_16_dual(s, pitch, lfi0->mblim, lfi0->lim,
+ lfi0->hev_thr, bd);
+ } else if (mask_16x16_0 & 1) {
+ vp9_highbd_lpf_vertical_16(s, pitch, lfi0->mblim, lfi0->lim,
+ lfi0->hev_thr, bd);
+ } else {
+ vp9_highbd_lpf_vertical_16(s + 8 *pitch, pitch, lfi1->mblim,
+ lfi1->lim, lfi1->hev_thr, bd);
+ }
+ }
+
+ if ((mask_8x8_0 | mask_8x8_1) & 1) {
+ if ((mask_8x8_0 & mask_8x8_1) & 1) {
+ vp9_highbd_lpf_vertical_8_dual(s, pitch, lfi0->mblim, lfi0->lim,
+ lfi0->hev_thr, lfi1->mblim, lfi1->lim,
+ lfi1->hev_thr, bd);
+ } else if (mask_8x8_0 & 1) {
+ vp9_highbd_lpf_vertical_8(s, pitch, lfi0->mblim, lfi0->lim,
+ lfi0->hev_thr, 1, bd);
+ } else {
+ vp9_highbd_lpf_vertical_8(s + 8 * pitch, pitch, lfi1->mblim,
+ lfi1->lim, lfi1->hev_thr, 1, bd);
+ }
+ }
+
+ if ((mask_4x4_0 | mask_4x4_1) & 1) {
+ if ((mask_4x4_0 & mask_4x4_1) & 1) {
+ vp9_highbd_lpf_vertical_4_dual(s, pitch, lfi0->mblim, lfi0->lim,
+ lfi0->hev_thr, lfi1->mblim, lfi1->lim,
+ lfi1->hev_thr, bd);
+ } else if (mask_4x4_0 & 1) {
+ vp9_highbd_lpf_vertical_4(s, pitch, lfi0->mblim, lfi0->lim,
+ lfi0->hev_thr, 1, bd);
+ } else {
+ vp9_highbd_lpf_vertical_4(s + 8 * pitch, pitch, lfi1->mblim,
+ lfi1->lim, lfi1->hev_thr, 1, bd);
+ }
+ }
+
+ if ((mask_4x4_int_0 | mask_4x4_int_1) & 1) {
+ if ((mask_4x4_int_0 & mask_4x4_int_1) & 1) {
+ vp9_highbd_lpf_vertical_4_dual(s + 4, pitch, lfi0->mblim, lfi0->lim,
+ lfi0->hev_thr, lfi1->mblim, lfi1->lim,
+ lfi1->hev_thr, bd);
+ } else if (mask_4x4_int_0 & 1) {
+ vp9_highbd_lpf_vertical_4(s + 4, pitch, lfi0->mblim, lfi0->lim,
+ lfi0->hev_thr, 1, bd);
+ } else {
+ vp9_highbd_lpf_vertical_4(s + 8 * pitch + 4, pitch, lfi1->mblim,
+ lfi1->lim, lfi1->hev_thr, 1, bd);
+ }
+ }
+ }
+
+ s += 8;
+ lfl += 1;
+ mask_16x16_0 >>= 1;
+ mask_8x8_0 >>= 1;
+ mask_4x4_0 >>= 1;
+ mask_4x4_int_0 >>= 1;
+ mask_16x16_1 >>= 1;
+ mask_8x8_1 >>= 1;
+ mask_4x4_1 >>= 1;
+ mask_4x4_int_1 >>= 1;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static void filter_selectively_horiz(uint8_t *s, int pitch,
unsigned int mask_16x16,
unsigned int mask_8x8,
@@ -419,7 +520,7 @@
}
} else if (mask_8x8 & 1) {
if ((mask_8x8 & 3) == 3) {
- // Next block's thresholds
+ // Next block's thresholds.
const loop_filter_thresh *lfin = lfi_n->lfthr + *(lfl + 1);
vp9_lpf_horizontal_8_dual(s, pitch, lfi->mblim, lfi->lim,
@@ -448,7 +549,7 @@
}
} else if (mask_4x4 & 1) {
if ((mask_4x4 & 3) == 3) {
- // Next block's thresholds
+ // Next block's thresholds.
const loop_filter_thresh *lfin = lfi_n->lfthr + *(lfl + 1);
vp9_lpf_horizontal_4_dual(s, pitch, lfi->mblim, lfi->lim,
@@ -488,6 +589,112 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static void highbd_filter_selectively_horiz(uint16_t *s, int pitch,
+ unsigned int mask_16x16,
+ unsigned int mask_8x8,
+ unsigned int mask_4x4,
+ unsigned int mask_4x4_int,
+ const loop_filter_info_n *lfi_n,
+ const uint8_t *lfl, int bd) {
+ unsigned int mask;
+ int count;
+
+ for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int;
+ mask; mask >>= count) {
+ const loop_filter_thresh *lfi = lfi_n->lfthr + *lfl;
+
+ count = 1;
+ if (mask & 1) {
+ if (mask_16x16 & 1) {
+ if ((mask_16x16 & 3) == 3) {
+ vp9_highbd_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, 2, bd);
+ count = 2;
+ } else {
+ vp9_highbd_lpf_horizontal_16(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, 1, bd);
+ }
+ } else if (mask_8x8 & 1) {
+ if ((mask_8x8 & 3) == 3) {
+ // Next block's thresholds.
+ const loop_filter_thresh *lfin = lfi_n->lfthr + *(lfl + 1);
+
+ vp9_highbd_lpf_horizontal_8_dual(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, lfin->mblim, lfin->lim,
+ lfin->hev_thr, bd);
+
+ if ((mask_4x4_int & 3) == 3) {
+ vp9_highbd_lpf_horizontal_4_dual(s + 4 * pitch, pitch, lfi->mblim,
+ lfi->lim, lfi->hev_thr,
+ lfin->mblim, lfin->lim,
+ lfin->hev_thr, bd);
+ } else {
+ if (mask_4x4_int & 1) {
+ vp9_highbd_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim,
+ lfi->lim, lfi->hev_thr, 1, bd);
+ } else if (mask_4x4_int & 2) {
+ vp9_highbd_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim,
+ lfin->lim, lfin->hev_thr, 1, bd);
+ }
+ }
+ count = 2;
+ } else {
+ vp9_highbd_lpf_horizontal_8(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, 1, bd);
+
+ if (mask_4x4_int & 1) {
+ vp9_highbd_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim,
+ lfi->lim, lfi->hev_thr, 1, bd);
+ }
+ }
+ } else if (mask_4x4 & 1) {
+ if ((mask_4x4 & 3) == 3) {
+ // Next block's thresholds.
+ const loop_filter_thresh *lfin = lfi_n->lfthr + *(lfl + 1);
+
+ vp9_highbd_lpf_horizontal_4_dual(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, lfin->mblim, lfin->lim,
+ lfin->hev_thr, bd);
+ if ((mask_4x4_int & 3) == 3) {
+ vp9_highbd_lpf_horizontal_4_dual(s + 4 * pitch, pitch, lfi->mblim,
+ lfi->lim, lfi->hev_thr,
+ lfin->mblim, lfin->lim,
+ lfin->hev_thr, bd);
+ } else {
+ if (mask_4x4_int & 1) {
+ vp9_highbd_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim,
+ lfi->lim, lfi->hev_thr, 1, bd);
+ } else if (mask_4x4_int & 2) {
+ vp9_highbd_lpf_horizontal_4(s + 8 + 4 * pitch, pitch, lfin->mblim,
+ lfin->lim, lfin->hev_thr, 1, bd);
+ }
+ }
+ count = 2;
+ } else {
+ vp9_highbd_lpf_horizontal_4(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, 1, bd);
+
+ if (mask_4x4_int & 1) {
+ vp9_highbd_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim,
+ lfi->lim, lfi->hev_thr, 1, bd);
+ }
+ }
+ } else if (mask_4x4_int & 1) {
+ vp9_highbd_lpf_horizontal_4(s + 4 * pitch, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, 1, bd);
+ }
+ }
+ s += 8 * count;
+ lfl += count;
+ mask_16x16 >>= count;
+ mask_8x8 >>= count;
+ mask_4x4 >>= count;
+ mask_4x4_int >>= count;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
// This function ors into the current lfm structure, where to do loop
// filters for the specific mi we are looking at. It uses information
// including the block_size_type (32x16, 32x32, etc.), the transform size,
@@ -903,14 +1110,53 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static void highbd_filter_selectively_vert(uint16_t *s, int pitch,
+ unsigned int mask_16x16,
+ unsigned int mask_8x8,
+ unsigned int mask_4x4,
+ unsigned int mask_4x4_int,
+ const loop_filter_info_n *lfi_n,
+ const uint8_t *lfl, int bd) {
+ unsigned int mask;
+
+ for (mask = mask_16x16 | mask_8x8 | mask_4x4 | mask_4x4_int;
+ mask; mask >>= 1) {
+ const loop_filter_thresh *lfi = lfi_n->lfthr + *lfl;
+
+ if (mask & 1) {
+ if (mask_16x16 & 1) {
+ vp9_highbd_lpf_vertical_16(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, bd);
+ } else if (mask_8x8 & 1) {
+ vp9_highbd_lpf_vertical_8(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, 1, bd);
+ } else if (mask_4x4 & 1) {
+ vp9_highbd_lpf_vertical_4(s, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, 1, bd);
+ }
+ }
+ if (mask_4x4_int & 1)
+ vp9_highbd_lpf_vertical_4(s + 4, pitch, lfi->mblim, lfi->lim,
+ lfi->hev_thr, 1, bd);
+ s += 8;
+ lfl += 1;
+ mask_16x16 >>= 1;
+ mask_8x8 >>= 1;
+ mask_4x4 >>= 1;
+ mask_4x4_int >>= 1;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static void filter_block_plane_non420(VP9_COMMON *cm,
struct macroblockd_plane *plane,
MODE_INFO *mi_8x8,
int mi_row, int mi_col) {
const int ss_x = plane->subsampling_x;
const int ss_y = plane->subsampling_y;
- const int row_step = 1 << ss_x;
- const int col_step = 1 << ss_y;
+ const int row_step = 1 << ss_y;
+ const int col_step = 1 << ss_x;
const int row_step_stride = cm->mi_stride * row_step;
struct buf_2d *const dst = &plane->dst;
uint8_t* const dst0 = dst->buf;
@@ -1001,12 +1247,32 @@
// Disable filtering on the leftmost column
border_mask = ~(mi_col == 0);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ highbd_filter_selectively_vert(CONVERT_TO_SHORTPTR(dst->buf),
+ dst->stride,
+ mask_16x16_c & border_mask,
+ mask_8x8_c & border_mask,
+ mask_4x4_c & border_mask,
+ mask_4x4_int[r],
+ &cm->lf_info, &lfl[r << 3],
+ (int)cm->bit_depth);
+ } else {
+ filter_selectively_vert(dst->buf, dst->stride,
+ mask_16x16_c & border_mask,
+ mask_8x8_c & border_mask,
+ mask_4x4_c & border_mask,
+ mask_4x4_int[r],
+ &cm->lf_info, &lfl[r << 3]);
+ }
+#else
filter_selectively_vert(dst->buf, dst->stride,
mask_16x16_c & border_mask,
mask_8x8_c & border_mask,
mask_4x4_c & border_mask,
mask_4x4_int[r],
&cm->lf_info, &lfl[r << 3]);
+#endif // CONFIG_VP9_HIGHBITDEPTH
dst->buf += 8 * dst->stride;
mi_8x8 += row_step_stride;
}
@@ -1030,13 +1296,32 @@
mask_8x8_r = mask_8x8[r];
mask_4x4_r = mask_4x4[r];
}
-
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ highbd_filter_selectively_horiz(CONVERT_TO_SHORTPTR(dst->buf),
+ dst->stride,
+ mask_16x16_r,
+ mask_8x8_r,
+ mask_4x4_r,
+ mask_4x4_int_r,
+ &cm->lf_info, &lfl[r << 3],
+ (int)cm->bit_depth);
+ } else {
+ filter_selectively_horiz(dst->buf, dst->stride,
+ mask_16x16_r,
+ mask_8x8_r,
+ mask_4x4_r,
+ mask_4x4_int_r,
+ &cm->lf_info, &lfl[r << 3]);
+ }
+#else
filter_selectively_horiz(dst->buf, dst->stride,
mask_16x16_r,
mask_8x8_r,
mask_4x4_r,
mask_4x4_int_r,
&cm->lf_info, &lfl[r << 3]);
+#endif // CONFIG_VP9_HIGHBITDEPTH
dst->buf += 8 * dst->stride;
}
}
@@ -1062,7 +1347,29 @@
unsigned int mask_4x4_l = mask_4x4 & 0xffff;
unsigned int mask_4x4_int_l = mask_4x4_int & 0xffff;
- // Disable filtering on the leftmost column
+ // Disable filtering on the leftmost column.
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ highbd_filter_selectively_vert_row2(plane->plane_type,
+ CONVERT_TO_SHORTPTR(dst->buf),
+ dst->stride,
+ mask_16x16_l,
+ mask_8x8_l,
+ mask_4x4_l,
+ mask_4x4_int_l,
+ &cm->lf_info, &lfm->lfl_y[r << 3],
+ (int)cm->bit_depth);
+ } else {
+ filter_selectively_vert_row2(plane->plane_type,
+ dst->buf, dst->stride,
+ mask_16x16_l,
+ mask_8x8_l,
+ mask_4x4_l,
+ mask_4x4_int_l,
+ &cm->lf_info,
+ &lfm->lfl_y[r << 3]);
+ }
+#else
filter_selectively_vert_row2(plane->plane_type,
dst->buf, dst->stride,
mask_16x16_l,
@@ -1070,7 +1377,7 @@
mask_4x4_l,
mask_4x4_int_l,
&cm->lf_info, &lfm->lfl_y[r << 3]);
-
+#endif // CONFIG_VP9_HIGHBITDEPTH
dst->buf += 16 * dst->stride;
mask_16x16 >>= 16;
mask_8x8 >>= 16;
@@ -1100,12 +1407,35 @@
mask_4x4_r = mask_4x4 & 0xff;
}
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ highbd_filter_selectively_horiz(CONVERT_TO_SHORTPTR(dst->buf),
+ dst->stride,
+ mask_16x16_r,
+ mask_8x8_r,
+ mask_4x4_r,
+ mask_4x4_int & 0xff,
+ &cm->lf_info,
+ &lfm->lfl_y[r << 3],
+ (int)cm->bit_depth);
+ } else {
+ filter_selectively_horiz(dst->buf, dst->stride,
+ mask_16x16_r,
+ mask_8x8_r,
+ mask_4x4_r,
+ mask_4x4_int & 0xff,
+ &cm->lf_info,
+ &lfm->lfl_y[r << 3]);
+ }
+#else
filter_selectively_horiz(dst->buf, dst->stride,
mask_16x16_r,
mask_8x8_r,
mask_4x4_r,
mask_4x4_int & 0xff,
- &cm->lf_info, &lfm->lfl_y[r << 3]);
+ &cm->lf_info,
+ &lfm->lfl_y[r << 3]);
+#endif // CONFIG_VP9_HIGHBITDEPTH
dst->buf += 8 * dst->stride;
mask_16x16 >>= 8;
@@ -1135,14 +1465,39 @@
unsigned int mask_4x4_l = mask_4x4 & 0xff;
unsigned int mask_4x4_int_l = mask_4x4_int & 0xff;
- // Disable filtering on the leftmost column
+ // Disable filtering on the leftmost column.
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ highbd_filter_selectively_vert_row2(plane->plane_type,
+ CONVERT_TO_SHORTPTR(dst->buf),
+ dst->stride,
+ mask_16x16_l,
+ mask_8x8_l,
+ mask_4x4_l,
+ mask_4x4_int_l,
+ &cm->lf_info,
+ &lfm->lfl_uv[r << 1],
+ (int)cm->bit_depth);
+ } else {
+ filter_selectively_vert_row2(plane->plane_type,
+ dst->buf, dst->stride,
+ mask_16x16_l,
+ mask_8x8_l,
+ mask_4x4_l,
+ mask_4x4_int_l,
+ &cm->lf_info,
+ &lfm->lfl_uv[r << 1]);
+ }
+#else
filter_selectively_vert_row2(plane->plane_type,
dst->buf, dst->stride,
mask_16x16_l,
mask_8x8_l,
mask_4x4_l,
mask_4x4_int_l,
- &cm->lf_info, &lfm->lfl_uv[r << 1]);
+ &cm->lf_info,
+ &lfm->lfl_uv[r << 1]);
+#endif // CONFIG_VP9_HIGHBITDEPTH
dst->buf += 16 * dst->stride;
mask_16x16 >>= 8;
@@ -1177,12 +1532,35 @@
mask_4x4_r = mask_4x4 & 0xf;
}
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ highbd_filter_selectively_horiz(CONVERT_TO_SHORTPTR(dst->buf),
+ dst->stride,
+ mask_16x16_r,
+ mask_8x8_r,
+ mask_4x4_r,
+ mask_4x4_int_r,
+ &cm->lf_info,
+ &lfm->lfl_uv[r << 1],
+ (int)cm->bit_depth);
+ } else {
+ filter_selectively_horiz(dst->buf, dst->stride,
+ mask_16x16_r,
+ mask_8x8_r,
+ mask_4x4_r,
+ mask_4x4_int_r,
+ &cm->lf_info,
+ &lfm->lfl_uv[r << 1]);
+ }
+#else
filter_selectively_horiz(dst->buf, dst->stride,
mask_16x16_r,
mask_8x8_r,
mask_4x4_r,
mask_4x4_int_r,
- &cm->lf_info, &lfm->lfl_uv[r << 1]);
+ &cm->lf_info,
+ &lfm->lfl_uv[r << 1]);
+#endif // CONFIG_VP9_HIGHBITDEPTH
dst->buf += 8 * dst->stride;
mask_16x16 >>= 4;
diff --git a/source/libvpx/vp9/common/vp9_loopfilter_filters.c b/source/libvpx/vp9/common/vp9_loopfilter_filters.c
index 25d3311..2e32c40 100644
--- a/source/libvpx/vp9/common/vp9_loopfilter_filters.c
+++ b/source/libvpx/vp9/common/vp9_loopfilter_filters.c
@@ -17,6 +17,20 @@
return (int8_t)clamp(t, -128, 127);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static INLINE int16_t signed_char_clamp_high(int t, int bd) {
+ switch (bd) {
+ case 10:
+ return (int16_t)clamp(t, -128*4, 128*4-1);
+ case 12:
+ return (int16_t)clamp(t, -128*16, 128*16-1);
+ case 8:
+ default:
+ return (int16_t)clamp(t, -128, 128-1);
+ }
+}
+#endif
+
// should we apply any filter at all: 11111111 yes, 00000000 no
static INLINE int8_t filter_mask(uint8_t limit, uint8_t blimit,
uint8_t p3, uint8_t p2,
@@ -337,3 +351,394 @@
const uint8_t *limit, const uint8_t *thresh) {
mb_lpf_vertical_edge_w(s, p, blimit, limit, thresh, 16);
}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+// Should we apply any filter at all: 11111111 yes, 00000000 no ?
+static INLINE int8_t highbd_filter_mask(uint8_t limit, uint8_t blimit,
+ uint16_t p3, uint16_t p2,
+ uint16_t p1, uint16_t p0,
+ uint16_t q0, uint16_t q1,
+ uint16_t q2, uint16_t q3, int bd) {
+ int8_t mask = 0;
+ int16_t limit16 = (uint16_t)limit << (bd - 8);
+ int16_t blimit16 = (uint16_t)blimit << (bd - 8);
+ mask |= (abs(p3 - p2) > limit16) * -1;
+ mask |= (abs(p2 - p1) > limit16) * -1;
+ mask |= (abs(p1 - p0) > limit16) * -1;
+ mask |= (abs(q1 - q0) > limit16) * -1;
+ mask |= (abs(q2 - q1) > limit16) * -1;
+ mask |= (abs(q3 - q2) > limit16) * -1;
+ mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 > blimit16) * -1;
+ return ~mask;
+}
+
+static INLINE int8_t highbd_flat_mask4(uint8_t thresh,
+ uint16_t p3, uint16_t p2,
+ uint16_t p1, uint16_t p0,
+ uint16_t q0, uint16_t q1,
+ uint16_t q2, uint16_t q3, int bd) {
+ int8_t mask = 0;
+ int16_t thresh16 = (uint16_t)thresh << (bd - 8);
+ mask |= (abs(p1 - p0) > thresh16) * -1;
+ mask |= (abs(q1 - q0) > thresh16) * -1;
+ mask |= (abs(p2 - p0) > thresh16) * -1;
+ mask |= (abs(q2 - q0) > thresh16) * -1;
+ mask |= (abs(p3 - p0) > thresh16) * -1;
+ mask |= (abs(q3 - q0) > thresh16) * -1;
+ return ~mask;
+}
+
+static INLINE int8_t highbd_flat_mask5(uint8_t thresh,
+ uint16_t p4, uint16_t p3,
+ uint16_t p2, uint16_t p1,
+ uint16_t p0, uint16_t q0,
+ uint16_t q1, uint16_t q2,
+ uint16_t q3, uint16_t q4, int bd) {
+ int8_t mask = ~highbd_flat_mask4(thresh, p3, p2, p1, p0, q0, q1, q2, q3, bd);
+ int16_t thresh16 = (uint16_t)thresh << (bd - 8);
+ mask |= (abs(p4 - p0) > thresh16) * -1;
+ mask |= (abs(q4 - q0) > thresh16) * -1;
+ return ~mask;
+}
+
+// Is there high edge variance internal edge:
+// 11111111_11111111 yes, 00000000_00000000 no ?
+static INLINE int16_t highbd_hev_mask(uint8_t thresh, uint16_t p1, uint16_t p0,
+ uint16_t q0, uint16_t q1, int bd) {
+ int16_t hev = 0;
+ int16_t thresh16 = (uint16_t)thresh << (bd - 8);
+ hev |= (abs(p1 - p0) > thresh16) * -1;
+ hev |= (abs(q1 - q0) > thresh16) * -1;
+ return hev;
+}
+
+static INLINE void highbd_filter4(int8_t mask, uint8_t thresh, uint16_t *op1,
+ uint16_t *op0, uint16_t *oq0, uint16_t *oq1,
+ int bd) {
+ int16_t filter1, filter2;
+ // ^0x80 equivalent to subtracting 0x80 from the values to turn them
+ // into -128 to +127 instead of 0 to 255.
+ int shift = bd - 8;
+ const int16_t ps1 = (int16_t)*op1 - (0x80 << shift);
+ const int16_t ps0 = (int16_t)*op0 - (0x80 << shift);
+ const int16_t qs0 = (int16_t)*oq0 - (0x80 << shift);
+ const int16_t qs1 = (int16_t)*oq1 - (0x80 << shift);
+ const uint16_t hev = highbd_hev_mask(thresh, *op1, *op0, *oq0, *oq1, bd);
+
+ // Add outer taps if we have high edge variance.
+ int16_t filter = signed_char_clamp_high(ps1 - qs1, bd) & hev;
+
+ // Inner taps.
+ filter = signed_char_clamp_high(filter + 3 * (qs0 - ps0), bd) & mask;
+
+ // Save bottom 3 bits so that we round one side +4 and the other +3
+ // if it equals 4 we'll set to adjust by -1 to account for the fact
+ // we'd round 3 the other way.
+ filter1 = signed_char_clamp_high(filter + 4, bd) >> 3;
+ filter2 = signed_char_clamp_high(filter + 3, bd) >> 3;
+
+ *oq0 = signed_char_clamp_high(qs0 - filter1, bd) + (0x80 << shift);
+ *op0 = signed_char_clamp_high(ps0 + filter2, bd) + (0x80 << shift);
+
+ // Outer tap adjustments.
+ filter = ROUND_POWER_OF_TWO(filter1, 1) & ~hev;
+
+ *oq1 = signed_char_clamp_high(qs1 - filter, bd) + (0x80 << shift);
+ *op1 = signed_char_clamp_high(ps1 + filter, bd) + (0x80 << shift);
+}
+
+void vp9_highbd_lpf_horizontal_4_c(uint16_t *s, int p /* pitch */,
+ const uint8_t *blimit, const uint8_t *limit,
+ const uint8_t *thresh, int count, int bd) {
+ int i;
+
+ // loop filter designed to work using chars so that we can make maximum use
+ // of 8 bit simd instructions.
+ for (i = 0; i < 8 * count; ++i) {
+ const uint16_t p3 = s[-4 * p];
+ const uint16_t p2 = s[-3 * p];
+ const uint16_t p1 = s[-2 * p];
+ const uint16_t p0 = s[-p];
+ const uint16_t q0 = s[0 * p];
+ const uint16_t q1 = s[1 * p];
+ const uint16_t q2 = s[2 * p];
+ const uint16_t q3 = s[3 * p];
+ const int8_t mask = highbd_filter_mask(*limit, *blimit,
+ p3, p2, p1, p0, q0, q1, q2, q3, bd);
+ highbd_filter4(mask, *thresh, s - 2 * p, s - 1 * p, s, s + 1 * p, bd);
+ ++s;
+ }
+}
+
+void vp9_highbd_lpf_horizontal_4_dual_c(uint16_t *s, int p,
+ const uint8_t *blimit0,
+ const uint8_t *limit0,
+ const uint8_t *thresh0,
+ const uint8_t *blimit1,
+ const uint8_t *limit1,
+ const uint8_t *thresh1,
+ int bd) {
+ vp9_highbd_lpf_horizontal_4_c(s, p, blimit0, limit0, thresh0, 1, bd);
+ vp9_highbd_lpf_horizontal_4_c(s + 8, p, blimit1, limit1, thresh1, 1, bd);
+}
+
+void vp9_highbd_lpf_vertical_4_c(uint16_t *s, int pitch, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd) {
+ int i;
+
+ // loop filter designed to work using chars so that we can make maximum use
+ // of 8 bit simd instructions.
+ for (i = 0; i < 8 * count; ++i) {
+ const uint16_t p3 = s[-4], p2 = s[-3], p1 = s[-2], p0 = s[-1];
+ const uint16_t q0 = s[0], q1 = s[1], q2 = s[2], q3 = s[3];
+ const int8_t mask = highbd_filter_mask(*limit, *blimit,
+ p3, p2, p1, p0, q0, q1, q2, q3, bd);
+ highbd_filter4(mask, *thresh, s - 2, s - 1, s, s + 1, bd);
+ s += pitch;
+ }
+}
+
+void vp9_highbd_lpf_vertical_4_dual_c(uint16_t *s, int pitch,
+ const uint8_t *blimit0,
+ const uint8_t *limit0,
+ const uint8_t *thresh0,
+ const uint8_t *blimit1,
+ const uint8_t *limit1,
+ const uint8_t *thresh1,
+ int bd) {
+ vp9_highbd_lpf_vertical_4_c(s, pitch, blimit0, limit0, thresh0, 1, bd);
+ vp9_highbd_lpf_vertical_4_c(s + 8 * pitch, pitch, blimit1, limit1,
+ thresh1, 1, bd);
+}
+
+static INLINE void highbd_filter8(int8_t mask, uint8_t thresh, uint8_t flat,
+ uint16_t *op3, uint16_t *op2,
+ uint16_t *op1, uint16_t *op0,
+ uint16_t *oq0, uint16_t *oq1,
+ uint16_t *oq2, uint16_t *oq3, int bd) {
+ if (flat && mask) {
+ const uint16_t p3 = *op3, p2 = *op2, p1 = *op1, p0 = *op0;
+ const uint16_t q0 = *oq0, q1 = *oq1, q2 = *oq2, q3 = *oq3;
+
+ // 7-tap filter [1, 1, 1, 2, 1, 1, 1]
+ *op2 = ROUND_POWER_OF_TWO(p3 + p3 + p3 + 2 * p2 + p1 + p0 + q0, 3);
+ *op1 = ROUND_POWER_OF_TWO(p3 + p3 + p2 + 2 * p1 + p0 + q0 + q1, 3);
+ *op0 = ROUND_POWER_OF_TWO(p3 + p2 + p1 + 2 * p0 + q0 + q1 + q2, 3);
+ *oq0 = ROUND_POWER_OF_TWO(p2 + p1 + p0 + 2 * q0 + q1 + q2 + q3, 3);
+ *oq1 = ROUND_POWER_OF_TWO(p1 + p0 + q0 + 2 * q1 + q2 + q3 + q3, 3);
+ *oq2 = ROUND_POWER_OF_TWO(p0 + q0 + q1 + 2 * q2 + q3 + q3 + q3, 3);
+ } else {
+ highbd_filter4(mask, thresh, op1, op0, oq0, oq1, bd);
+ }
+}
+
+void vp9_highbd_lpf_horizontal_8_c(uint16_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd) {
+ int i;
+
+ // loop filter designed to work using chars so that we can make maximum use
+ // of 8 bit simd instructions.
+ for (i = 0; i < 8 * count; ++i) {
+ const uint16_t p3 = s[-4 * p], p2 = s[-3 * p], p1 = s[-2 * p], p0 = s[-p];
+ const uint16_t q0 = s[0 * p], q1 = s[1 * p], q2 = s[2 * p], q3 = s[3 * p];
+
+ const int8_t mask = highbd_filter_mask(*limit, *blimit,
+ p3, p2, p1, p0, q0, q1, q2, q3, bd);
+ const int8_t flat = highbd_flat_mask4(1, p3, p2, p1, p0, q0, q1, q2, q3,
+ bd);
+ highbd_filter8(mask, *thresh, flat,
+ s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p,
+ s, s + 1 * p, s + 2 * p, s + 3 * p, bd);
+ ++s;
+ }
+}
+
+void vp9_highbd_lpf_horizontal_8_dual_c(uint16_t *s, int p,
+ const uint8_t *blimit0,
+ const uint8_t *limit0,
+ const uint8_t *thresh0,
+ const uint8_t *blimit1,
+ const uint8_t *limit1,
+ const uint8_t *thresh1,
+ int bd) {
+ vp9_highbd_lpf_horizontal_8_c(s, p, blimit0, limit0, thresh0, 1, bd);
+ vp9_highbd_lpf_horizontal_8_c(s + 8, p, blimit1, limit1, thresh1, 1, bd);
+}
+
+void vp9_highbd_lpf_vertical_8_c(uint16_t *s, int pitch, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd) {
+ int i;
+
+ for (i = 0; i < 8 * count; ++i) {
+ const uint16_t p3 = s[-4], p2 = s[-3], p1 = s[-2], p0 = s[-1];
+ const uint16_t q0 = s[0], q1 = s[1], q2 = s[2], q3 = s[3];
+ const int8_t mask = highbd_filter_mask(*limit, *blimit,
+ p3, p2, p1, p0, q0, q1, q2, q3, bd);
+ const int8_t flat = highbd_flat_mask4(1, p3, p2, p1, p0, q0, q1, q2, q3,
+ bd);
+ highbd_filter8(mask, *thresh, flat,
+ s - 4, s - 3, s - 2, s - 1,
+ s, s + 1, s + 2, s + 3,
+ bd);
+ s += pitch;
+ }
+}
+
+void vp9_highbd_lpf_vertical_8_dual_c(uint16_t *s, int pitch,
+ const uint8_t *blimit0,
+ const uint8_t *limit0,
+ const uint8_t *thresh0,
+ const uint8_t *blimit1,
+ const uint8_t *limit1,
+ const uint8_t *thresh1,
+ int bd) {
+ vp9_highbd_lpf_vertical_8_c(s, pitch, blimit0, limit0, thresh0, 1, bd);
+ vp9_highbd_lpf_vertical_8_c(s + 8 * pitch, pitch, blimit1, limit1,
+ thresh1, 1, bd);
+}
+
+static INLINE void highbd_filter16(int8_t mask, uint8_t thresh,
+ uint8_t flat, uint8_t flat2,
+ uint16_t *op7, uint16_t *op6,
+ uint16_t *op5, uint16_t *op4,
+ uint16_t *op3, uint16_t *op2,
+ uint16_t *op1, uint16_t *op0,
+ uint16_t *oq0, uint16_t *oq1,
+ uint16_t *oq2, uint16_t *oq3,
+ uint16_t *oq4, uint16_t *oq5,
+ uint16_t *oq6, uint16_t *oq7, int bd) {
+ if (flat2 && flat && mask) {
+ const uint16_t p7 = *op7;
+ const uint16_t p6 = *op6;
+ const uint16_t p5 = *op5;
+ const uint16_t p4 = *op4;
+ const uint16_t p3 = *op3;
+ const uint16_t p2 = *op2;
+ const uint16_t p1 = *op1;
+ const uint16_t p0 = *op0;
+ const uint16_t q0 = *oq0;
+ const uint16_t q1 = *oq1;
+ const uint16_t q2 = *oq2;
+ const uint16_t q3 = *oq3;
+ const uint16_t q4 = *oq4;
+ const uint16_t q5 = *oq5;
+ const uint16_t q6 = *oq6;
+ const uint16_t q7 = *oq7;
+
+ // 15-tap filter [1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1]
+ *op6 = ROUND_POWER_OF_TWO(p7 * 7 + p6 * 2 + p5 + p4 + p3 + p2 + p1 + p0 +
+ q0, 4);
+ *op5 = ROUND_POWER_OF_TWO(p7 * 6 + p6 + p5 * 2 + p4 + p3 + p2 + p1 + p0 +
+ q0 + q1, 4);
+ *op4 = ROUND_POWER_OF_TWO(p7 * 5 + p6 + p5 + p4 * 2 + p3 + p2 + p1 + p0 +
+ q0 + q1 + q2, 4);
+ *op3 = ROUND_POWER_OF_TWO(p7 * 4 + p6 + p5 + p4 + p3 * 2 + p2 + p1 + p0 +
+ q0 + q1 + q2 + q3, 4);
+ *op2 = ROUND_POWER_OF_TWO(p7 * 3 + p6 + p5 + p4 + p3 + p2 * 2 + p1 + p0 +
+ q0 + q1 + q2 + q3 + q4, 4);
+ *op1 = ROUND_POWER_OF_TWO(p7 * 2 + p6 + p5 + p4 + p3 + p2 + p1 * 2 + p0 +
+ q0 + q1 + q2 + q3 + q4 + q5, 4);
+ *op0 = ROUND_POWER_OF_TWO(p7 + p6 + p5 + p4 + p3 + p2 + p1 + p0 * 2 +
+ q0 + q1 + q2 + q3 + q4 + q5 + q6, 4);
+ *oq0 = ROUND_POWER_OF_TWO(p6 + p5 + p4 + p3 + p2 + p1 + p0 +
+ q0 * 2 + q1 + q2 + q3 + q4 + q5 + q6 + q7, 4);
+ *oq1 = ROUND_POWER_OF_TWO(p5 + p4 + p3 + p2 + p1 + p0 +
+ q0 + q1 * 2 + q2 + q3 + q4 + q5 + q6 + q7 * 2, 4);
+ *oq2 = ROUND_POWER_OF_TWO(p4 + p3 + p2 + p1 + p0 +
+ q0 + q1 + q2 * 2 + q3 + q4 + q5 + q6 + q7 * 3, 4);
+ *oq3 = ROUND_POWER_OF_TWO(p3 + p2 + p1 + p0 +
+ q0 + q1 + q2 + q3 * 2 + q4 + q5 + q6 + q7 * 4, 4);
+ *oq4 = ROUND_POWER_OF_TWO(p2 + p1 + p0 +
+ q0 + q1 + q2 + q3 + q4 * 2 + q5 + q6 + q7 * 5, 4);
+ *oq5 = ROUND_POWER_OF_TWO(p1 + p0 +
+ q0 + q1 + q2 + q3 + q4 + q5 * 2 + q6 + q7 * 6, 4);
+ *oq6 = ROUND_POWER_OF_TWO(p0 +
+ q0 + q1 + q2 + q3 + q4 + q5 + q6 * 2 + q7 * 7, 4);
+ } else {
+ highbd_filter8(mask, thresh, flat, op3, op2, op1, op0, oq0, oq1, oq2, oq3,
+ bd);
+ }
+}
+
+void vp9_highbd_lpf_horizontal_16_c(uint16_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int count, int bd) {
+ int i;
+
+ // loop filter designed to work using chars so that we can make maximum use
+ // of 8 bit simd instructions.
+ for (i = 0; i < 8 * count; ++i) {
+ const uint16_t p3 = s[-4 * p];
+ const uint16_t p2 = s[-3 * p];
+ const uint16_t p1 = s[-2 * p];
+ const uint16_t p0 = s[-p];
+ const uint16_t q0 = s[0 * p];
+ const uint16_t q1 = s[1 * p];
+ const uint16_t q2 = s[2 * p];
+ const uint16_t q3 = s[3 * p];
+ const int8_t mask = highbd_filter_mask(*limit, *blimit,
+ p3, p2, p1, p0, q0, q1, q2, q3, bd);
+ const int8_t flat = highbd_flat_mask4(1, p3, p2, p1, p0, q0, q1, q2, q3,
+ bd);
+ const int8_t flat2 = highbd_flat_mask5(
+ 1, s[-8 * p], s[-7 * p], s[-6 * p], s[-5 * p], p0,
+ q0, s[4 * p], s[5 * p], s[6 * p], s[7 * p], bd);
+
+ highbd_filter16(mask, *thresh, flat, flat2,
+ s - 8 * p, s - 7 * p, s - 6 * p, s - 5 * p,
+ s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p,
+ s, s + 1 * p, s + 2 * p, s + 3 * p,
+ s + 4 * p, s + 5 * p, s + 6 * p, s + 7 * p,
+ bd);
+ ++s;
+ }
+}
+
+static void highbd_mb_lpf_vertical_edge_w(uint16_t *s, int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh,
+ int count, int bd) {
+ int i;
+
+ for (i = 0; i < count; ++i) {
+ const uint16_t p3 = s[-4];
+ const uint16_t p2 = s[-3];
+ const uint16_t p1 = s[-2];
+ const uint16_t p0 = s[-1];
+ const uint16_t q0 = s[0];
+ const uint16_t q1 = s[1];
+ const uint16_t q2 = s[2];
+ const uint16_t q3 = s[3];
+ const int8_t mask = highbd_filter_mask(*limit, *blimit,
+ p3, p2, p1, p0, q0, q1, q2, q3, bd);
+ const int8_t flat = highbd_flat_mask4(1, p3, p2, p1, p0, q0, q1, q2, q3,
+ bd);
+ const int8_t flat2 = highbd_flat_mask5(1, s[-8], s[-7], s[-6], s[-5], p0,
+ q0, s[4], s[5], s[6], s[7], bd);
+
+ highbd_filter16(mask, *thresh, flat, flat2,
+ s - 8, s - 7, s - 6, s - 5, s - 4, s - 3, s - 2, s - 1,
+ s, s + 1, s + 2, s + 3, s + 4, s + 5, s + 6, s + 7,
+ bd);
+ s += p;
+ }
+}
+
+void vp9_highbd_lpf_vertical_16_c(uint16_t *s, int p, const uint8_t *blimit,
+ const uint8_t *limit, const uint8_t *thresh,
+ int bd) {
+ highbd_mb_lpf_vertical_edge_w(s, p, blimit, limit, thresh, 8, bd);
+}
+
+void vp9_highbd_lpf_vertical_16_dual_c(uint16_t *s, int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh,
+ int bd) {
+ highbd_mb_lpf_vertical_edge_w(s, p, blimit, limit, thresh, 16, bd);
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/source/libvpx/vp9/common/vp9_mvref_common.c b/source/libvpx/vp9/common/vp9_mvref_common.c
index a09afff..3b34050 100644
--- a/source/libvpx/vp9/common/vp9_mvref_common.c
+++ b/source/libvpx/vp9/common/vp9_mvref_common.c
@@ -24,10 +24,7 @@
? cm->prev_mi[mi_row * xd->mi_stride + mi_col].src_mi
: NULL;
const MB_MODE_INFO *const prev_mbmi = prev_mi ? &prev_mi->src_mi->mbmi : NULL;
-
-
const POSITION *const mv_ref_search = mv_ref_blocks[mi->mbmi.sb_type];
-
int different_ref_found = 0;
int context_counter = 0;
@@ -109,10 +106,10 @@
}
void vp9_find_mv_refs(const VP9_COMMON *cm, const MACROBLOCKD *xd,
- const TileInfo *const tile,
- MODE_INFO *mi, MV_REFERENCE_FRAME ref_frame,
- int_mv *mv_ref_list,
- int mi_row, int mi_col) {
+ const TileInfo *const tile,
+ MODE_INFO *mi, MV_REFERENCE_FRAME ref_frame,
+ int_mv *mv_ref_list,
+ int mi_row, int mi_col) {
find_mv_refs_idx(cm, xd, tile, mi, ref_frame, mv_ref_list, -1,
mi_row, mi_col);
}
@@ -127,7 +124,6 @@
}
}
-
void vp9_find_best_ref_mvs(MACROBLOCKD *xd, int allow_hp,
int_mv *mvlist, int_mv *nearest, int_mv *near) {
int i;
diff --git a/source/libvpx/vp9/common/vp9_onyxc_int.h b/source/libvpx/vp9/common/vp9_onyxc_int.h
index 792e9d9..f1eda91 100644
--- a/source/libvpx/vp9/common/vp9_onyxc_int.h
+++ b/source/libvpx/vp9/common/vp9_onyxc_int.h
@@ -143,7 +143,6 @@
int prev_mi_idx;
int mi_alloc_size;
MODE_INFO *mip_array[2];
- MODE_INFO **mi_grid_base_array[2];
MODE_INFO *mip; /* Base of allocated array */
MODE_INFO *mi; /* Corresponds to upper left visible macroblock */
@@ -180,6 +179,7 @@
// VPX_BITS_8 in profile 0 or 1, VPX_BITS_10 or VPX_BITS_12 in profile 2 or 3.
vpx_bit_depth_t bit_depth;
+ vpx_bit_depth_t dequant_bit_depth; // bit_depth of current dequantizer
#if CONFIG_VP9_POSTPROC
struct postproc_state postproc_state;
@@ -328,11 +328,11 @@
const PARTITION_CONTEXT *above_ctx = xd->above_seg_context + mi_col;
const PARTITION_CONTEXT *left_ctx = xd->left_seg_context + (mi_row & MI_MASK);
- const int bsl = mi_width_log2(bsize);
+ const int bsl = mi_width_log2_lookup[bsize];
const int bs = 1 << bsl;
int above = 0, left = 0, i;
- assert(b_width_log2(bsize) == b_height_log2(bsize));
+ assert(b_width_log2_lookup[bsize] == b_height_log2_lookup[bsize]);
assert(bsl >= 0);
for (i = 0; i < bs; i++) {
diff --git a/source/libvpx/vp9/common/vp9_postproc.c b/source/libvpx/vp9/common/vp9_postproc.c
index e4e6ce7..575ffbc 100644
--- a/source/libvpx/vp9/common/vp9_postproc.c
+++ b/source/libvpx/vp9/common/vp9_postproc.c
@@ -19,6 +19,9 @@
#include "vpx_scale/vpx_scale.h"
#include "vpx_scale/yv12config.h"
+#if CONFIG_VP9_HIGHBITDEPTH
+#include "vp9/common/vp9_common.h"
+#endif
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/common/vp9_postproc.h"
#include "vp9/common/vp9_systemdependent.h"
@@ -152,6 +155,84 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+void vp9_highbd_post_proc_down_and_across_c(const uint16_t *src_ptr,
+ uint16_t *dst_ptr,
+ int src_pixels_per_line,
+ int dst_pixels_per_line,
+ int rows,
+ int cols,
+ int flimit) {
+ uint16_t const *p_src;
+ uint16_t *p_dst;
+ int row;
+ int col;
+ int i;
+ int v;
+ int pitch = src_pixels_per_line;
+ uint16_t d[8];
+
+ for (row = 0; row < rows; row++) {
+ // post_proc_down for one row.
+ p_src = src_ptr;
+ p_dst = dst_ptr;
+
+ for (col = 0; col < cols; col++) {
+ int kernel = 4;
+ int v = p_src[col];
+
+ for (i = -2; i <= 2; i++) {
+ if (abs(v - p_src[col + i * pitch]) > flimit)
+ goto down_skip_convolve;
+
+ kernel += kernel5[2 + i] * p_src[col + i * pitch];
+ }
+
+ v = (kernel >> 3);
+
+ down_skip_convolve:
+ p_dst[col] = v;
+ }
+
+ /* now post_proc_across */
+ p_src = dst_ptr;
+ p_dst = dst_ptr;
+
+ for (i = 0; i < 8; i++)
+ d[i] = p_src[i];
+
+ for (col = 0; col < cols; col++) {
+ int kernel = 4;
+ v = p_src[col];
+
+ d[col & 7] = v;
+
+ for (i = -2; i <= 2; i++) {
+ if (abs(v - p_src[col + i]) > flimit)
+ goto across_skip_convolve;
+
+ kernel += kernel5[2 + i] * p_src[col + i];
+ }
+
+ d[col & 7] = (kernel >> 3);
+
+ across_skip_convolve:
+ if (col >= 2)
+ p_dst[col - 2] = d[(col - 2) & 7];
+ }
+
+ /* handle the last two pixels */
+ p_dst[col - 2] = d[(col - 2) & 7];
+ p_dst[col - 1] = d[(col - 1) & 7];
+
+
+ /* next row */
+ src_ptr += pitch;
+ dst_ptr += dst_pixels_per_line;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static int q2mbl(int x) {
if (x < 20) x = 20;
@@ -162,10 +243,46 @@
void vp9_mbpost_proc_across_ip_c(uint8_t *src, int pitch,
int rows, int cols, int flimit) {
int r, c, i;
-
uint8_t *s = src;
uint8_t d[16];
+ for (r = 0; r < rows; r++) {
+ int sumsq = 0;
+ int sum = 0;
+
+ for (i = -8; i <= 6; i++) {
+ sumsq += s[i] * s[i];
+ sum += s[i];
+ d[i + 8] = 0;
+ }
+
+ for (c = 0; c < cols + 8; c++) {
+ int x = s[c + 7] - s[c - 8];
+ int y = s[c + 7] + s[c - 8];
+
+ sum += x;
+ sumsq += x * y;
+
+ d[c & 15] = s[c];
+
+ if (sumsq * 15 - sum * sum < flimit) {
+ d[c & 15] = (8 + sum + s[c]) >> 4;
+ }
+
+ s[c - 8] = d[(c - 8) & 15];
+ }
+ s += pitch;
+ }
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+void vp9_highbd_mbpost_proc_across_ip_c(uint16_t *src, int pitch,
+ int rows, int cols, int flimit) {
+ int r, c, i;
+
+ uint16_t *s = src;
+ uint16_t d[16];
+
for (r = 0; r < rows; r++) {
int sumsq = 0;
@@ -196,6 +313,7 @@
s += pitch;
}
}
+#endif // CONFIG_VP9_HIGHBITDEPTH
void vp9_mbpost_proc_down_c(uint8_t *dst, int pitch,
int rows, int cols, int flimit) {
@@ -229,6 +347,40 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+void vp9_highbd_mbpost_proc_down_c(uint16_t *dst, int pitch,
+ int rows, int cols, int flimit) {
+ int r, c, i;
+ const int16_t *rv3 = &vp9_rv[63 & rand()]; // NOLINT
+
+ for (c = 0; c < cols; c++) {
+ uint16_t *s = &dst[c];
+ int sumsq = 0;
+ int sum = 0;
+ uint16_t d[16];
+ const int16_t *rv2 = rv3 + ((c * 17) & 127);
+
+ for (i = -8; i <= 6; i++) {
+ sumsq += s[i * pitch] * s[i * pitch];
+ sum += s[i * pitch];
+ }
+
+ for (r = 0; r < rows + 8; r++) {
+ sumsq += s[7 * pitch] * s[ 7 * pitch] - s[-8 * pitch] * s[-8 * pitch];
+ sum += s[7 * pitch] - s[-8 * pitch];
+ d[r & 15] = s[0];
+
+ if (sumsq * 15 - sum * sum < flimit) {
+ d[r & 15] = (rv2[r & 127] + sum + s[0]) >> 4;
+ }
+
+ s[-8 * pitch] = d[(r - 8) & 15];
+ s += pitch;
+ }
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static void deblock_and_de_macro_block(YV12_BUFFER_CONFIG *source,
YV12_BUFFER_CONFIG *post,
int q,
@@ -239,6 +391,51 @@
(void) low_var_thresh;
(void) flag;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (source->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_post_proc_down_and_across(CONVERT_TO_SHORTPTR(source->y_buffer),
+ CONVERT_TO_SHORTPTR(post->y_buffer),
+ source->y_stride, post->y_stride,
+ source->y_height, source->y_width,
+ ppl);
+
+ vp9_highbd_mbpost_proc_across_ip(CONVERT_TO_SHORTPTR(post->y_buffer),
+ post->y_stride, post->y_height,
+ post->y_width, q2mbl(q));
+
+ vp9_highbd_mbpost_proc_down(CONVERT_TO_SHORTPTR(post->y_buffer),
+ post->y_stride, post->y_height,
+ post->y_width, q2mbl(q));
+
+ vp9_highbd_post_proc_down_and_across(CONVERT_TO_SHORTPTR(source->u_buffer),
+ CONVERT_TO_SHORTPTR(post->u_buffer),
+ source->uv_stride, post->uv_stride,
+ source->uv_height, source->uv_width,
+ ppl);
+ vp9_highbd_post_proc_down_and_across(CONVERT_TO_SHORTPTR(source->v_buffer),
+ CONVERT_TO_SHORTPTR(post->v_buffer),
+ source->uv_stride, post->uv_stride,
+ source->uv_height, source->uv_width,
+ ppl);
+ } else {
+ vp9_post_proc_down_and_across(source->y_buffer, post->y_buffer,
+ source->y_stride, post->y_stride,
+ source->y_height, source->y_width, ppl);
+
+ vp9_mbpost_proc_across_ip(post->y_buffer, post->y_stride, post->y_height,
+ post->y_width, q2mbl(q));
+
+ vp9_mbpost_proc_down(post->y_buffer, post->y_stride, post->y_height,
+ post->y_width, q2mbl(q));
+
+ vp9_post_proc_down_and_across(source->u_buffer, post->u_buffer,
+ source->uv_stride, post->uv_stride,
+ source->uv_height, source->uv_width, ppl);
+ vp9_post_proc_down_and_across(source->v_buffer, post->v_buffer,
+ source->uv_stride, post->uv_stride,
+ source->uv_height, source->uv_width, ppl);
+ }
+#else
vp9_post_proc_down_and_across(source->y_buffer, post->y_buffer,
source->y_stride, post->y_stride,
source->y_height, source->y_width, ppl);
@@ -255,6 +452,7 @@
vp9_post_proc_down_and_across(source->v_buffer, post->v_buffer,
source->uv_stride, post->uv_stride,
source->uv_height, source->uv_width, ppl);
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
void vp9_deblock(const YV12_BUFFER_CONFIG *src, YV12_BUFFER_CONFIG *dst,
@@ -271,10 +469,26 @@
uint8_t *const dsts[3] = {dst->y_buffer, dst->u_buffer, dst->v_buffer};
const int dst_strides[3] = {dst->y_stride, dst->uv_stride, dst->uv_stride};
- for (i = 0; i < MAX_MB_PLANE; ++i)
+ for (i = 0; i < MAX_MB_PLANE; ++i) {
+#if CONFIG_VP9_HIGHBITDEPTH
+ assert((src->flags & YV12_FLAG_HIGHBITDEPTH) ==
+ (dst->flags & YV12_FLAG_HIGHBITDEPTH));
+ if (src->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_post_proc_down_and_across(CONVERT_TO_SHORTPTR(srcs[i]),
+ CONVERT_TO_SHORTPTR(dsts[i]),
+ src_strides[i], dst_strides[i],
+ src_heights[i], src_widths[i], ppl);
+ } else {
+ vp9_post_proc_down_and_across(srcs[i], dsts[i],
+ src_strides[i], dst_strides[i],
+ src_heights[i], src_widths[i], ppl);
+ }
+#else
vp9_post_proc_down_and_across(srcs[i], dsts[i],
src_strides[i], dst_strides[i],
src_heights[i], src_widths[i], ppl);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ }
}
void vp9_denoise(const YV12_BUFFER_CONFIG *src, YV12_BUFFER_CONFIG *dst,
@@ -293,15 +507,32 @@
for (i = 0; i < MAX_MB_PLANE; ++i) {
const int src_stride = src_strides[i];
- const uint8_t *const src = srcs[i] + 2 * src_stride + 2;
const int src_width = src_widths[i] - 4;
const int src_height = src_heights[i] - 4;
-
const int dst_stride = dst_strides[i];
- uint8_t *const dst = dsts[i] + 2 * dst_stride + 2;
+#if CONFIG_VP9_HIGHBITDEPTH
+ assert((src->flags & YV12_FLAG_HIGHBITDEPTH) ==
+ (dst->flags & YV12_FLAG_HIGHBITDEPTH));
+ if (src->flags & YV12_FLAG_HIGHBITDEPTH) {
+ const uint16_t *const src = CONVERT_TO_SHORTPTR(srcs[i] + 2 * src_stride
+ + 2);
+ uint16_t *const dst = CONVERT_TO_SHORTPTR(dsts[i] + 2 * dst_stride + 2);
+ vp9_highbd_post_proc_down_and_across(src, dst, src_stride, dst_stride,
+ src_height, src_width, ppl);
+ } else {
+ const uint8_t *const src = srcs[i] + 2 * src_stride + 2;
+ uint8_t *const dst = dsts[i] + 2 * dst_stride + 2;
+
+ vp9_post_proc_down_and_across(src, dst, src_stride, dst_stride,
+ src_height, src_width, ppl);
+ }
+#else
+ const uint8_t *const src = srcs[i] + 2 * src_stride + 2;
+ uint8_t *const dst = dsts[i] + 2 * dst_stride + 2;
vp9_post_proc_down_and_across(src, dst, src_stride, dst_stride,
src_height, src_width, ppl);
+#endif
}
}
@@ -405,6 +636,9 @@
#if CONFIG_VP9_POSTPROC || CONFIG_INTERNAL_STATS
if (vp9_realloc_frame_buffer(&cm->post_proc_buffer, cm->width, cm->height,
cm->subsampling_x, cm->subsampling_y,
+#if CONFIG_VP9_HIGHBITDEPTH
+ cm->use_highbitdepth,
+#endif
VP9_DEC_BORDER_IN_PIXELS, NULL, NULL, NULL) < 0)
vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate post-processing buffer");
diff --git a/source/libvpx/vp9/common/vp9_prob.h b/source/libvpx/vp9/common/vp9_prob.h
index fa0e36d..bc1511a 100644
--- a/source/libvpx/vp9/common/vp9_prob.h
+++ b/source/libvpx/vp9/common/vp9_prob.h
@@ -14,7 +14,6 @@
#include "./vpx_config.h"
#include "vpx_ports/mem.h"
-#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_common.h"
diff --git a/source/libvpx/vp9/common/vp9_quant_common.h b/source/libvpx/vp9/common/vp9_quant_common.h
index b626605..4bae4a8 100644
--- a/source/libvpx/vp9/common/vp9_quant_common.h
+++ b/source/libvpx/vp9/common/vp9_quant_common.h
@@ -12,7 +12,7 @@
#define VP9_COMMON_VP9_QUANT_COMMON_H_
#include "vpx/vpx_codec.h"
-#include "vp9/common/vp9_blockd.h"
+#include "vp9/common/vp9_seg_common.h"
#ifdef __cplusplus
extern "C" {
diff --git a/source/libvpx/vp9/common/vp9_reconinter.c b/source/libvpx/vp9/common/vp9_reconinter.c
index b49f130..3492a23 100644
--- a/source/libvpx/vp9/common/vp9_reconinter.c
+++ b/source/libvpx/vp9/common/vp9_reconinter.c
@@ -153,19 +153,19 @@
int w, int h, int ref,
const InterpKernel *kernel,
int xs, int ys, int bd) {
- sf->high_predict[subpel_x != 0][subpel_y != 0][ref](
+ sf->highbd_predict[subpel_x != 0][subpel_y != 0][ref](
src, src_stride, dst, dst_stride,
kernel[subpel_x], xs, kernel[subpel_y], ys, w, h, bd);
}
-void vp9_high_build_inter_predictor(const uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- const MV *src_mv,
- const struct scale_factors *sf,
- int w, int h, int ref,
- const InterpKernel *kernel,
- enum mv_precision precision,
- int x, int y, int bd) {
+void vp9_highbd_build_inter_predictor(const uint8_t *src, int src_stride,
+ uint8_t *dst, int dst_stride,
+ const MV *src_mv,
+ const struct scale_factors *sf,
+ int w, int h, int ref,
+ const InterpKernel *kernel,
+ enum mv_precision precision,
+ int x, int y, int bd) {
const int is_q4 = precision == MV_PRECISION_Q4;
const MV mv_q4 = { is_q4 ? src_mv->row : src_mv->row * 2,
is_q4 ? src_mv->col : src_mv->col * 2 };
@@ -288,8 +288,9 @@
uint8_t *pre;
MV32 scaled_mv;
int xs, ys, subpel_x, subpel_y;
+ const int is_scaled = vp9_is_scaled(sf);
- if (vp9_is_scaled(sf)) {
+ if (is_scaled) {
pre = pre_buf->buf + scaled_buffer_offset(x, y, pre_buf->stride, sf);
scaled_mv = vp9_scale_mv(&mv_q4, mi_x + x, mi_y + y, sf);
xs = sf->x_step_q4;
@@ -385,12 +386,6 @@
? average_split_mvs(pd, mi, ref, block)
: mi->mbmi.mv[ref].as_mv;
-
- // TODO(jkoleszar): This clamping is done in the incorrect place for the
- // scaling case. It needs to be done on the scaled MV, not the pre-scaling
- // MV. Note however that it performs the subsampling aware scaling so
- // that the result is always q4.
- // mv_precision precision is MV_PRECISION_Q4.
const MV mv_q4 = clamp_mv_to_umv_border_sb(xd, &mv, bw, bh,
pd->subsampling_x,
pd->subsampling_y);
@@ -400,6 +395,7 @@
subpel_x, subpel_y;
uint8_t *ref_frame, *buf_ptr;
const YV12_BUFFER_CONFIG *ref_buf = xd->block_refs[ref]->buf;
+ const int is_scaled = vp9_is_scaled(sf);
// Get reference frame pointer, width and height.
if (plane == 0) {
@@ -412,7 +408,7 @@
ref_frame = plane == 1 ? ref_buf->u_buffer : ref_buf->v_buffer;
}
- if (vp9_is_scaled(sf)) {
+ if (is_scaled) {
// Co-ordinate of containing block to pixel precision.
int x_start = (-xd->mb_to_left_edge >> (3 + pd->subsampling_x));
int y_start = (-xd->mb_to_top_edge >> (3 + pd->subsampling_y));
@@ -451,7 +447,8 @@
subpel_x = scaled_mv.col & SUBPEL_MASK;
subpel_y = scaled_mv.row & SUBPEL_MASK;
- // Calculate the top left corner of the best matching block in the reference frame.
+ // Calculate the top left corner of the best matching block in the
+ // reference frame.
x0 += scaled_mv.col >> SUBPEL_BITS;
y0 += scaled_mv.row >> SUBPEL_BITS;
x0_16 += scaled_mv.col;
@@ -463,20 +460,20 @@
// Do border extension if there is motion or the
// width/height is not a multiple of 8 pixels.
- if (scaled_mv.col || scaled_mv.row ||
+ if (is_scaled || scaled_mv.col || scaled_mv.row ||
(frame_width & 0x7) || (frame_height & 0x7)) {
// Get reference block bottom right coordinate.
int x1 = ((x0_16 + (w - 1) * xs) >> SUBPEL_BITS) + 1;
int y1 = ((y0_16 + (h - 1) * ys) >> SUBPEL_BITS) + 1;
int x_pad = 0, y_pad = 0;
- if (subpel_x || (sf->x_step_q4 & SUBPEL_MASK)) {
+ if (subpel_x || (sf->x_step_q4 != SUBPEL_SHIFTS)) {
x0 -= VP9_INTERP_EXTEND - 1;
x1 += VP9_INTERP_EXTEND;
x_pad = 1;
}
- if (subpel_y || (sf->y_step_q4 & SUBPEL_MASK)) {
+ if (subpel_y || (sf->y_step_q4 != SUBPEL_SHIFTS)) {
y0 -= VP9_INTERP_EXTEND - 1;
y1 += VP9_INTERP_EXTEND;
y_pad = 1;
diff --git a/source/libvpx/vp9/common/vp9_reconinter.h b/source/libvpx/vp9/common/vp9_reconinter.h
index e70cc4c..3eaf07c 100644
--- a/source/libvpx/vp9/common/vp9_reconinter.h
+++ b/source/libvpx/vp9/common/vp9_reconinter.h
@@ -40,14 +40,14 @@
int x, int y);
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_build_inter_predictor(const uint8_t *src, int src_stride,
- uint8_t *dst, int dst_stride,
- const MV *mv_q3,
- const struct scale_factors *sf,
- int w, int h, int do_avg,
- const InterpKernel *kernel,
- enum mv_precision precision,
- int x, int y, int bd);
+void vp9_highbd_build_inter_predictor(const uint8_t *src, int src_stride,
+ uint8_t *dst, int dst_stride,
+ const MV *mv_q3,
+ const struct scale_factors *sf,
+ int w, int h, int do_avg,
+ const InterpKernel *kernel,
+ enum mv_precision precision,
+ int x, int y, int bd);
#endif
static INLINE int scaled_buffer_offset(int x_offset, int y_offset, int stride,
diff --git a/source/libvpx/vp9/common/vp9_reconintra.c b/source/libvpx/vp9/common/vp9_reconintra.c
index 7ebd2ea..720bb44 100644
--- a/source/libvpx/vp9/common/vp9_reconintra.c
+++ b/source/libvpx/vp9/common/vp9_reconintra.c
@@ -41,11 +41,11 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-#define intra_pred_high_sized(type, size) \
- void vp9_high_##type##_predictor_##size##x##size##_c( \
+#define intra_pred_highbd_sized(type, size) \
+ void vp9_highbd_##type##_predictor_##size##x##size##_c( \
uint16_t *dst, ptrdiff_t stride, const uint16_t *above, \
const uint16_t *left, int bd) { \
- high_##type##_predictor(dst, stride, size, above, left, bd); \
+ highbd_##type##_predictor(dst, stride, size, above, left, bd); \
}
#define intra_pred_allsizes(type) \
@@ -53,10 +53,10 @@
intra_pred_sized(type, 8) \
intra_pred_sized(type, 16) \
intra_pred_sized(type, 32) \
- intra_pred_high_sized(type, 4) \
- intra_pred_high_sized(type, 8) \
- intra_pred_high_sized(type, 16) \
- intra_pred_high_sized(type, 32)
+ intra_pred_highbd_sized(type, 4) \
+ intra_pred_highbd_sized(type, 8) \
+ intra_pred_highbd_sized(type, 16) \
+ intra_pred_highbd_sized(type, 32)
#else
@@ -68,9 +68,9 @@
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE void high_d207_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_d207_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r, c;
(void) above;
(void) bd;
@@ -102,9 +102,9 @@
}
}
-static INLINE void high_d63_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_d63_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r, c;
(void) left;
(void) bd;
@@ -120,9 +120,9 @@
}
}
-static INLINE void high_d45_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_d45_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+ const uint16_t *above,
+ const uint16_t *left, int bd) {
int r, c;
(void) left;
(void) bd;
@@ -137,9 +137,9 @@
}
}
-static INLINE void high_d117_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_d117_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r, c;
(void) bd;
@@ -168,9 +168,9 @@
}
}
-static INLINE void high_d135_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_d135_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r, c;
(void) bd;
dst[0] = ROUND_POWER_OF_TWO(left[0] + above[-1] * 2 + above[0], 2);
@@ -190,9 +190,9 @@
}
}
-static INLINE void high_d153_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_d153_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r, c;
(void) bd;
dst[0] = ROUND_POWER_OF_TWO(above[-1] + left[0], 1);
@@ -218,9 +218,9 @@
}
}
-static INLINE void high_v_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_v_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r;
(void) left;
(void) bd;
@@ -230,9 +230,9 @@
}
}
-static INLINE void high_h_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above, const uint16_t *left,
- int bd) {
+static INLINE void highbd_h_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r;
(void) above;
(void) bd;
@@ -242,23 +242,23 @@
}
}
-static INLINE void high_tm_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_tm_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r, c;
int ytop_left = above[-1];
(void) bd;
for (r = 0; r < bs; r++) {
for (c = 0; c < bs; c++)
- dst[c] = clip_pixel_high(left[r] + above[c] - ytop_left, bd);
+ dst[c] = clip_pixel_highbd(left[r] + above[c] - ytop_left, bd);
dst += stride;
}
}
-static INLINE void high_dc_128_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_dc_128_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int r;
(void) above;
(void) left;
@@ -269,9 +269,9 @@
}
}
-static INLINE void high_dc_left_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_dc_left_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int i, r, expected_dc, sum = 0;
(void) above;
(void) bd;
@@ -286,9 +286,9 @@
}
}
-static INLINE void high_dc_top_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_dc_top_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int i, r, expected_dc, sum = 0;
(void) left;
(void) bd;
@@ -303,9 +303,9 @@
}
}
-static INLINE void high_dc_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
+static INLINE void highbd_dc_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
int i, r, expected_dc, sum = 0;
const int count = 2 * bs;
(void) bd;
@@ -602,20 +602,20 @@
INIT_ALL_SIZES(dc_pred[1][1], dc);
#if CONFIG_VP9_HIGHBITDEPTH
- INIT_ALL_SIZES(pred_high[V_PRED], high_v);
- INIT_ALL_SIZES(pred_high[H_PRED], high_h);
- INIT_ALL_SIZES(pred_high[D207_PRED], high_d207);
- INIT_ALL_SIZES(pred_high[D45_PRED], high_d45);
- INIT_ALL_SIZES(pred_high[D63_PRED], high_d63);
- INIT_ALL_SIZES(pred_high[D117_PRED], high_d117);
- INIT_ALL_SIZES(pred_high[D135_PRED], high_d135);
- INIT_ALL_SIZES(pred_high[D153_PRED], high_d153);
- INIT_ALL_SIZES(pred_high[TM_PRED], high_tm);
+ INIT_ALL_SIZES(pred_high[V_PRED], highbd_v);
+ INIT_ALL_SIZES(pred_high[H_PRED], highbd_h);
+ INIT_ALL_SIZES(pred_high[D207_PRED], highbd_d207);
+ INIT_ALL_SIZES(pred_high[D45_PRED], highbd_d45);
+ INIT_ALL_SIZES(pred_high[D63_PRED], highbd_d63);
+ INIT_ALL_SIZES(pred_high[D117_PRED], highbd_d117);
+ INIT_ALL_SIZES(pred_high[D135_PRED], highbd_d135);
+ INIT_ALL_SIZES(pred_high[D153_PRED], highbd_d153);
+ INIT_ALL_SIZES(pred_high[TM_PRED], highbd_tm);
- INIT_ALL_SIZES(dc_pred_high[0][0], high_dc_128);
- INIT_ALL_SIZES(dc_pred_high[0][1], high_dc_top);
- INIT_ALL_SIZES(dc_pred_high[1][0], high_dc_left);
- INIT_ALL_SIZES(dc_pred_high[1][1], high_dc);
+ INIT_ALL_SIZES(dc_pred_high[0][0], highbd_dc_128);
+ INIT_ALL_SIZES(dc_pred_high[0][1], highbd_dc_top);
+ INIT_ALL_SIZES(dc_pred_high[1][0], highbd_dc_left);
+ INIT_ALL_SIZES(dc_pred_high[1][1], highbd_dc);
#endif // CONFIG_VP9_HIGHBITDEPTH
#undef intra_pred_allsizes
diff --git a/source/libvpx/vp9/common/vp9_rtcd_defs.pl b/source/libvpx/vp9/common/vp9_rtcd_defs.pl
index 0f52ae1..0530f3a 100644
--- a/source/libvpx/vp9/common/vp9_rtcd_defs.pl
+++ b/source/libvpx/vp9/common/vp9_rtcd_defs.pl
@@ -5,8 +5,8 @@
*/
#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_enums.h"
-#include "vp9/common/vp9_idct.h"
struct macroblockd;
@@ -331,6 +331,8 @@
# dct
#
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
+ # Note as optimized versions of these functions are added we need to add a check to ensure
+ # that when CONFIG_EMULATE_HARDWARE is on, it defaults to the C versions only.
add_proto qw/void vp9_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct4x4_1_add/;
@@ -380,69 +382,123 @@
add_proto qw/void vp9_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_iwht4x4_16_add/;
+
} else {
- add_proto qw/void vp9_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct4x4_1_add sse2 neon_asm dspr2/;
- $vp9_idct4x4_1_add_neon_asm=vp9_idct4x4_1_add_neon;
+ # Force C versions if CONFIG_EMULATE_HARDWARE is 1
+ if (vpx_config("CONFIG_EMULATE_HARDWARE") eq "yes") {
+ add_proto qw/void vp9_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct4x4_1_add/;
- add_proto qw/void vp9_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct4x4_16_add sse2 neon_asm dspr2/;
- $vp9_idct4x4_16_add_neon_asm=vp9_idct4x4_16_add_neon;
+ add_proto qw/void vp9_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct4x4_16_add/;
- add_proto qw/void vp9_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct8x8_1_add sse2 neon_asm dspr2/;
- $vp9_idct8x8_1_add_neon_asm=vp9_idct8x8_1_add_neon;
+ add_proto qw/void vp9_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct8x8_1_add/;
- add_proto qw/void vp9_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct8x8_64_add sse2 neon_asm dspr2/, "$ssse3_x86_64";
- $vp9_idct8x8_64_add_neon_asm=vp9_idct8x8_64_add_neon;
+ add_proto qw/void vp9_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct8x8_64_add/;
- add_proto qw/void vp9_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct8x8_12_add sse2 neon_asm dspr2/, "$ssse3_x86_64";
- $vp9_idct8x8_12_add_neon_asm=vp9_idct8x8_12_add_neon;
+ add_proto qw/void vp9_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct8x8_12_add/;
- add_proto qw/void vp9_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct16x16_1_add sse2 neon_asm dspr2/;
- $vp9_idct16x16_1_add_neon_asm=vp9_idct16x16_1_add_neon;
+ add_proto qw/void vp9_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct16x16_1_add/;
- add_proto qw/void vp9_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct16x16_256_add sse2 ssse3 neon_asm dspr2/;
- $vp9_idct16x16_256_add_neon_asm=vp9_idct16x16_256_add_neon;
+ add_proto qw/void vp9_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct16x16_256_add/;
- add_proto qw/void vp9_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct16x16_10_add sse2 ssse3 neon_asm dspr2/;
- $vp9_idct16x16_10_add_neon_asm=vp9_idct16x16_10_add_neon;
+ add_proto qw/void vp9_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct16x16_10_add/;
- add_proto qw/void vp9_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct32x32_1024_add sse2 neon_asm dspr2/;
- $vp9_idct32x32_1024_add_neon_asm=vp9_idct32x32_1024_add_neon;
+ add_proto qw/void vp9_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct32x32_1024_add/;
- add_proto qw/void vp9_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct32x32_34_add sse2 neon_asm dspr2/;
- $vp9_idct32x32_34_add_neon_asm=vp9_idct32x32_1024_add_neon;
+ add_proto qw/void vp9_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct32x32_34_add/;
- add_proto qw/void vp9_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_idct32x32_1_add sse2 neon_asm dspr2/;
- $vp9_idct32x32_1_add_neon_asm=vp9_idct32x32_1_add_neon;
+ add_proto qw/void vp9_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct32x32_1_add/;
- add_proto qw/void vp9_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
- specialize qw/vp9_iht4x4_16_add sse2 neon_asm dspr2/;
- $vp9_iht4x4_16_add_neon_asm=vp9_iht4x4_16_add_neon;
+ add_proto qw/void vp9_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ specialize qw/vp9_iht4x4_16_add/;
- add_proto qw/void vp9_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
- specialize qw/vp9_iht8x8_64_add sse2 neon_asm dspr2/;
- $vp9_iht8x8_64_add_neon_asm=vp9_iht8x8_64_add_neon;
+ add_proto qw/void vp9_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ specialize qw/vp9_iht8x8_64_add/;
- add_proto qw/void vp9_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, int tx_type";
- specialize qw/vp9_iht16x16_256_add sse2 dspr2/;
+ add_proto qw/void vp9_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, int tx_type";
+ specialize qw/vp9_iht16x16_256_add/;
- # dct and add
+ # dct and add
- add_proto qw/void vp9_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_iwht4x4_1_add/;
+ add_proto qw/void vp9_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_iwht4x4_1_add/;
- add_proto qw/void vp9_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vp9_iwht4x4_16_add/;
+ add_proto qw/void vp9_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_iwht4x4_16_add/;
+ } else {
+ add_proto qw/void vp9_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct4x4_1_add sse2 neon_asm dspr2/;
+ $vp9_idct4x4_1_add_neon_asm=vp9_idct4x4_1_add_neon;
+
+ add_proto qw/void vp9_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct4x4_16_add sse2 neon_asm dspr2/;
+ $vp9_idct4x4_16_add_neon_asm=vp9_idct4x4_16_add_neon;
+
+ add_proto qw/void vp9_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct8x8_1_add sse2 neon_asm dspr2/;
+ $vp9_idct8x8_1_add_neon_asm=vp9_idct8x8_1_add_neon;
+
+ add_proto qw/void vp9_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct8x8_64_add sse2 neon_asm dspr2/, "$ssse3_x86_64";
+ $vp9_idct8x8_64_add_neon_asm=vp9_idct8x8_64_add_neon;
+
+ add_proto qw/void vp9_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct8x8_12_add sse2 neon_asm dspr2/, "$ssse3_x86_64";
+ $vp9_idct8x8_12_add_neon_asm=vp9_idct8x8_12_add_neon;
+
+ add_proto qw/void vp9_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct16x16_1_add sse2 neon_asm dspr2/;
+ $vp9_idct16x16_1_add_neon_asm=vp9_idct16x16_1_add_neon;
+
+ add_proto qw/void vp9_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct16x16_256_add sse2 ssse3 neon_asm dspr2/;
+ $vp9_idct16x16_256_add_neon_asm=vp9_idct16x16_256_add_neon;
+
+ add_proto qw/void vp9_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct16x16_10_add sse2 ssse3 neon_asm dspr2/;
+ $vp9_idct16x16_10_add_neon_asm=vp9_idct16x16_10_add_neon;
+
+ add_proto qw/void vp9_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct32x32_1024_add sse2 neon_asm dspr2/;
+ $vp9_idct32x32_1024_add_neon_asm=vp9_idct32x32_1024_add_neon;
+
+ add_proto qw/void vp9_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct32x32_34_add sse2 neon_asm dspr2/;
+ $vp9_idct32x32_34_add_neon_asm=vp9_idct32x32_1024_add_neon;
+
+ add_proto qw/void vp9_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_idct32x32_1_add sse2 neon_asm dspr2/;
+ $vp9_idct32x32_1_add_neon_asm=vp9_idct32x32_1_add_neon;
+
+ add_proto qw/void vp9_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ specialize qw/vp9_iht4x4_16_add sse2 neon_asm dspr2/;
+ $vp9_iht4x4_16_add_neon_asm=vp9_iht4x4_16_add_neon;
+
+ add_proto qw/void vp9_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ specialize qw/vp9_iht8x8_64_add sse2 neon_asm dspr2/;
+ $vp9_iht8x8_64_add_neon_asm=vp9_iht8x8_64_add_neon;
+
+ add_proto qw/void vp9_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, int tx_type";
+ specialize qw/vp9_iht16x16_256_add sse2 dspr2/;
+
+ # dct and add
+
+ add_proto qw/void vp9_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_iwht4x4_1_add/;
+
+ add_proto qw/void vp9_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
+ specialize qw/vp9_iwht4x4_16_add/;
+ }
}
# High bitdepth functions
@@ -450,241 +506,296 @@
#
# Intra prediction
#
- add_proto qw/void vp9_high_d207_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d207_predictor_4x4/;
+ add_proto qw/void vp9_highbd_d207_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d207_predictor_4x4/;
- add_proto qw/void vp9_high_d45_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d45_predictor_4x4/;
+ add_proto qw/void vp9_highbd_d45_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d45_predictor_4x4/;
- add_proto qw/void vp9_high_d63_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d63_predictor_4x4/;
+ add_proto qw/void vp9_highbd_d63_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d63_predictor_4x4/;
- add_proto qw/void vp9_high_h_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_h_predictor_4x4/;
+ add_proto qw/void vp9_highbd_h_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_h_predictor_4x4/;
- add_proto qw/void vp9_high_d117_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d117_predictor_4x4/;
+ add_proto qw/void vp9_highbd_d117_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d117_predictor_4x4/;
- add_proto qw/void vp9_high_d135_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d135_predictor_4x4/;
+ add_proto qw/void vp9_highbd_d135_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d135_predictor_4x4/;
- add_proto qw/void vp9_high_d153_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d153_predictor_4x4/;
+ add_proto qw/void vp9_highbd_d153_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d153_predictor_4x4/;
- add_proto qw/void vp9_high_v_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_v_predictor_4x4 neon/, "$sse_x86inc";
+ add_proto qw/void vp9_highbd_v_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_v_predictor_4x4 neon/, "$sse_x86inc";
- add_proto qw/void vp9_high_tm_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_tm_predictor_4x4/, "$sse_x86inc";
+ add_proto qw/void vp9_highbd_tm_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_tm_predictor_4x4/, "$sse_x86inc";
- add_proto qw/void vp9_high_dc_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_predictor_4x4/, "$sse_x86inc";
+ add_proto qw/void vp9_highbd_dc_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_predictor_4x4/, "$sse_x86inc";
- add_proto qw/void vp9_high_dc_top_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_top_predictor_4x4/;
+ add_proto qw/void vp9_highbd_dc_top_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_top_predictor_4x4/;
- add_proto qw/void vp9_high_dc_left_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_left_predictor_4x4/;
+ add_proto qw/void vp9_highbd_dc_left_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_left_predictor_4x4/;
- add_proto qw/void vp9_high_dc_128_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_128_predictor_4x4/;
+ add_proto qw/void vp9_highbd_dc_128_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_128_predictor_4x4/;
- add_proto qw/void vp9_high_d207_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d207_predictor_8x8/;
+ add_proto qw/void vp9_highbd_d207_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d207_predictor_8x8/;
- add_proto qw/void vp9_high_d45_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d45_predictor_8x8/;
+ add_proto qw/void vp9_highbd_d45_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d45_predictor_8x8/;
- add_proto qw/void vp9_high_d63_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d63_predictor_8x8/;
+ add_proto qw/void vp9_highbd_d63_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d63_predictor_8x8/;
- add_proto qw/void vp9_high_h_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_h_predictor_8x8/;
+ add_proto qw/void vp9_highbd_h_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_h_predictor_8x8/;
- add_proto qw/void vp9_high_d117_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d117_predictor_8x8/;
+ add_proto qw/void vp9_highbd_d117_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d117_predictor_8x8/;
- add_proto qw/void vp9_high_d135_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d135_predictor_8x8/;
+ add_proto qw/void vp9_highbd_d135_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d135_predictor_8x8/;
- add_proto qw/void vp9_high_d153_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d153_predictor_8x8/;
+ add_proto qw/void vp9_highbd_d153_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d153_predictor_8x8/;
- add_proto qw/void vp9_high_v_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_v_predictor_8x8/, "$sse2_x86inc";
+ add_proto qw/void vp9_highbd_v_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_v_predictor_8x8/, "$sse2_x86inc";
- add_proto qw/void vp9_high_tm_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_tm_predictor_8x8/, "$sse2_x86inc";
+ add_proto qw/void vp9_highbd_tm_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_tm_predictor_8x8/, "$sse2_x86inc";
- add_proto qw/void vp9_high_dc_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_predictor_8x8/, "$sse2_x86inc";;
+ add_proto qw/void vp9_highbd_dc_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_predictor_8x8/, "$sse2_x86inc";;
- add_proto qw/void vp9_high_dc_top_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_top_predictor_8x8/;
+ add_proto qw/void vp9_highbd_dc_top_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_top_predictor_8x8/;
- add_proto qw/void vp9_high_dc_left_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_left_predictor_8x8/;
+ add_proto qw/void vp9_highbd_dc_left_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_left_predictor_8x8/;
- add_proto qw/void vp9_high_dc_128_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_128_predictor_8x8/;
+ add_proto qw/void vp9_highbd_dc_128_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_128_predictor_8x8/;
- add_proto qw/void vp9_high_d207_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d207_predictor_16x16/;
+ add_proto qw/void vp9_highbd_d207_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d207_predictor_16x16/;
- add_proto qw/void vp9_high_d45_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d45_predictor_16x16/;
+ add_proto qw/void vp9_highbd_d45_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d45_predictor_16x16/;
- add_proto qw/void vp9_high_d63_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d63_predictor_16x16/;
+ add_proto qw/void vp9_highbd_d63_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d63_predictor_16x16/;
- add_proto qw/void vp9_high_h_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_h_predictor_16x16/;
+ add_proto qw/void vp9_highbd_h_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_h_predictor_16x16/;
- add_proto qw/void vp9_high_d117_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d117_predictor_16x16/;
+ add_proto qw/void vp9_highbd_d117_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d117_predictor_16x16/;
- add_proto qw/void vp9_high_d135_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d135_predictor_16x16/;
+ add_proto qw/void vp9_highbd_d135_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d135_predictor_16x16/;
- add_proto qw/void vp9_high_d153_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d153_predictor_16x16/;
+ add_proto qw/void vp9_highbd_d153_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d153_predictor_16x16/;
- add_proto qw/void vp9_high_v_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_v_predictor_16x16 neon/, "$sse2_x86inc";
+ add_proto qw/void vp9_highbd_v_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_v_predictor_16x16 neon/, "$sse2_x86inc";
- add_proto qw/void vp9_high_tm_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_tm_predictor_16x16/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_tm_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_tm_predictor_16x16/, "$sse2_x86_64";
- add_proto qw/void vp9_high_dc_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_predictor_16x16/, "$sse2_x86inc";
+ add_proto qw/void vp9_highbd_dc_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_predictor_16x16/, "$sse2_x86inc";
- add_proto qw/void vp9_high_dc_top_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_top_predictor_16x16/;
+ add_proto qw/void vp9_highbd_dc_top_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_top_predictor_16x16/;
- add_proto qw/void vp9_high_dc_left_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_left_predictor_16x16/;
+ add_proto qw/void vp9_highbd_dc_left_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_left_predictor_16x16/;
- add_proto qw/void vp9_high_dc_128_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_128_predictor_16x16/;
+ add_proto qw/void vp9_highbd_dc_128_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_128_predictor_16x16/;
- add_proto qw/void vp9_high_d207_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d207_predictor_32x32/;
+ add_proto qw/void vp9_highbd_d207_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d207_predictor_32x32/;
- add_proto qw/void vp9_high_d45_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d45_predictor_32x32/;
+ add_proto qw/void vp9_highbd_d45_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d45_predictor_32x32/;
- add_proto qw/void vp9_high_d63_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d63_predictor_32x32/;
+ add_proto qw/void vp9_highbd_d63_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d63_predictor_32x32/;
- add_proto qw/void vp9_high_h_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_h_predictor_32x32/;
+ add_proto qw/void vp9_highbd_h_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_h_predictor_32x32/;
- add_proto qw/void vp9_high_d117_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d117_predictor_32x32/;
+ add_proto qw/void vp9_highbd_d117_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d117_predictor_32x32/;
- add_proto qw/void vp9_high_d135_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d135_predictor_32x32/;
+ add_proto qw/void vp9_highbd_d135_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d135_predictor_32x32/;
- add_proto qw/void vp9_high_d153_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_d153_predictor_32x32/;
+ add_proto qw/void vp9_highbd_d153_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_d153_predictor_32x32/;
- add_proto qw/void vp9_high_v_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_v_predictor_32x32/, "$sse2_x86inc";
+ add_proto qw/void vp9_highbd_v_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_v_predictor_32x32/, "$sse2_x86inc";
- add_proto qw/void vp9_high_tm_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_tm_predictor_32x32/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_tm_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_tm_predictor_32x32/, "$sse2_x86_64";
- add_proto qw/void vp9_high_dc_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_predictor_32x32/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_dc_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_predictor_32x32/, "$sse2_x86_64";
- add_proto qw/void vp9_high_dc_top_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_top_predictor_32x32/;
+ add_proto qw/void vp9_highbd_dc_top_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_top_predictor_32x32/;
- add_proto qw/void vp9_high_dc_left_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_left_predictor_32x32/;
+ add_proto qw/void vp9_highbd_dc_left_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_left_predictor_32x32/;
- add_proto qw/void vp9_high_dc_128_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bps";
- specialize qw/vp9_high_dc_128_predictor_32x32/;
+ add_proto qw/void vp9_highbd_dc_128_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vp9_highbd_dc_128_predictor_32x32/;
#
# Sub Pixel Filters
#
- add_proto qw/void vp9_high_convolve_copy/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
- specialize qw/vp9_high_convolve_copy/;
+ add_proto qw/void vp9_highbd_convolve_copy/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
+ specialize qw/vp9_highbd_convolve_copy/;
- add_proto qw/void vp9_high_convolve_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
- specialize qw/vp9_high_convolve_avg/;
+ add_proto qw/void vp9_highbd_convolve_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
+ specialize qw/vp9_highbd_convolve_avg/;
- add_proto qw/void vp9_high_convolve8/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
- specialize qw/vp9_high_convolve8/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_convolve8/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
+ specialize qw/vp9_highbd_convolve8/, "$sse2_x86_64";
- add_proto qw/void vp9_high_convolve8_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
- specialize qw/vp9_high_convolve8_horiz/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_convolve8_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
+ specialize qw/vp9_highbd_convolve8_horiz/, "$sse2_x86_64";
- add_proto qw/void vp9_high_convolve8_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
- specialize qw/vp9_high_convolve8_vert/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_convolve8_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
+ specialize qw/vp9_highbd_convolve8_vert/, "$sse2_x86_64";
- add_proto qw/void vp9_high_convolve8_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
- specialize qw/vp9_high_convolve8_avg/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_convolve8_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
+ specialize qw/vp9_highbd_convolve8_avg/, "$sse2_x86_64";
- add_proto qw/void vp9_high_convolve8_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
- specialize qw/vp9_high_convolve8_avg_horiz/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_convolve8_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
+ specialize qw/vp9_highbd_convolve8_avg_horiz/, "$sse2_x86_64";
- add_proto qw/void vp9_high_convolve8_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
- specialize qw/vp9_high_convolve8_avg_vert/, "$sse2_x86_64";
+ add_proto qw/void vp9_highbd_convolve8_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
+ specialize qw/vp9_highbd_convolve8_avg_vert/, "$sse2_x86_64";
+
+ #
+ # Loopfilter
+ #
+ add_proto qw/void vp9_highbd_lpf_vertical_16/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
+ specialize qw/vp9_highbd_lpf_vertical_16 sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_vertical_16_dual/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
+ specialize qw/vp9_highbd_lpf_vertical_16_dual sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_vertical_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count, int bd";
+ specialize qw/vp9_highbd_lpf_vertical_8 sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_vertical_8_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
+ specialize qw/vp9_highbd_lpf_vertical_8_dual sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_vertical_4/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count, int bd";
+ specialize qw/vp9_highbd_lpf_vertical_4 sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_vertical_4_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
+ specialize qw/vp9_highbd_lpf_vertical_4_dual sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_horizontal_16/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count, int bd";
+ specialize qw/vp9_highbd_lpf_horizontal_16 sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_horizontal_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count, int bd";
+ specialize qw/vp9_highbd_lpf_horizontal_8 sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_horizontal_8_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
+ specialize qw/vp9_highbd_lpf_horizontal_8_dual sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_horizontal_4/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count, int bd";
+ specialize qw/vp9_highbd_lpf_horizontal_4 sse2/;
+
+ add_proto qw/void vp9_highbd_lpf_horizontal_4_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
+ specialize qw/vp9_highbd_lpf_horizontal_4_dual sse2/;
+
+ #
+ # post proc
+ #
+ if (vpx_config("CONFIG_VP9_POSTPROC") eq "yes") {
+ add_proto qw/void vp9_highbd_mbpost_proc_down/, "uint16_t *dst, int pitch, int rows, int cols, int flimit";
+ specialize qw/vp9_highbd_mbpost_proc_down/;
+
+ add_proto qw/void vp9_highbd_mbpost_proc_across_ip/, "uint16_t *src, int pitch, int rows, int cols, int flimit";
+ specialize qw/vp9_highbd_mbpost_proc_across_ip/;
+
+ add_proto qw/void vp9_highbd_post_proc_down_and_across/, "const uint16_t *src_ptr, uint16_t *dst_ptr, int src_pixels_per_line, int dst_pixels_per_line, int rows, int cols, int flimit";
+ specialize qw/vp9_highbd_post_proc_down_and_across/;
+
+ add_proto qw/void vp9_highbd_plane_add_noise/, "uint8_t *Start, char *noise, char blackclamp[16], char whiteclamp[16], char bothclamp[16], unsigned int Width, unsigned int Height, int Pitch";
+ specialize qw/vp9_highbd_plane_add_noise/;
+ }
#
# dct
#
- add_proto qw/void vp9_high_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct4x4_1_add/;
+ # Note as optimized versions of these functions are added we need to add a check to ensure
+ # that when CONFIG_EMULATE_HARDWARE is on, it defaults to the C versions only.
+ add_proto qw/void vp9_highbd_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct4x4_1_add/;
- add_proto qw/void vp9_high_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct4x4_16_add/;
+ add_proto qw/void vp9_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct4x4_16_add/;
- add_proto qw/void vp9_high_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct8x8_1_add/;
+ add_proto qw/void vp9_highbd_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct8x8_1_add/;
- add_proto qw/void vp9_high_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct8x8_64_add/;
+ add_proto qw/void vp9_highbd_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct8x8_64_add/;
- add_proto qw/void vp9_high_idct8x8_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct8x8_10_add/;
+ add_proto qw/void vp9_highbd_idct8x8_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct8x8_10_add/;
- add_proto qw/void vp9_high_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct16x16_1_add/;
+ add_proto qw/void vp9_highbd_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct16x16_1_add/;
- add_proto qw/void vp9_high_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct16x16_256_add/;
+ add_proto qw/void vp9_highbd_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct16x16_256_add/;
- add_proto qw/void vp9_high_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct16x16_10_add/;
+ add_proto qw/void vp9_highbd_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct16x16_10_add/;
- add_proto qw/void vp9_high_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct32x32_1024_add/;
+ add_proto qw/void vp9_highbd_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct32x32_1024_add/;
- add_proto qw/void vp9_high_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct32x32_34_add/;
+ add_proto qw/void vp9_highbd_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct32x32_34_add/;
- add_proto qw/void vp9_high_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_idct32x32_1_add/;
+ add_proto qw/void vp9_highbd_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_idct32x32_1_add/;
- add_proto qw/void vp9_high_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
- specialize qw/vp9_high_iht4x4_16_add/;
+ add_proto qw/void vp9_highbd_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
+ specialize qw/vp9_highbd_iht4x4_16_add/;
- add_proto qw/void vp9_high_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
- specialize qw/vp9_high_iht8x8_64_add/;
+ add_proto qw/void vp9_highbd_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
+ specialize qw/vp9_highbd_iht8x8_64_add/;
- add_proto qw/void vp9_high_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, int tx_type, int bd";
- specialize qw/vp9_high_iht16x16_256_add/;
+ add_proto qw/void vp9_highbd_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, int tx_type, int bd";
+ specialize qw/vp9_highbd_iht16x16_256_add/;
# dct and add
- add_proto qw/void vp9_high_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_iwht4x4_1_add/;
+ add_proto qw/void vp9_highbd_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_iwht4x4_1_add/;
- add_proto qw/void vp9_high_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
- specialize qw/vp9_high_iwht4x4_16_add/;
+ add_proto qw/void vp9_highbd_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
+ specialize qw/vp9_highbd_iwht4x4_16_add/;
}
#
@@ -820,22 +931,22 @@
specialize qw/vp9_sub_pixel_avg_variance4x4/, "$sse_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vp9_sad64x64 neon/, "$sse2_x86inc";
+specialize qw/vp9_sad64x64 neon avx2/, "$sse2_x86inc";
add_proto qw/unsigned int vp9_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vp9_sad32x64/, "$sse2_x86inc";
+specialize qw/vp9_sad32x64 avx2/, "$sse2_x86inc";
add_proto qw/unsigned int vp9_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vp9_sad64x32/, "$sse2_x86inc";
+specialize qw/vp9_sad64x32 avx2/, "$sse2_x86inc";
add_proto qw/unsigned int vp9_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vp9_sad32x16/, "$sse2_x86inc";
+specialize qw/vp9_sad32x16 avx2/, "$sse2_x86inc";
add_proto qw/unsigned int vp9_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
specialize qw/vp9_sad16x32/, "$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";
-specialize qw/vp9_sad32x32 neon/, "$sse2_x86inc";
+specialize qw/vp9_sad32x32 neon avx2/, "$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";
specialize qw/vp9_sad16x16 neon/, "$sse2_x86inc";
@@ -859,22 +970,22 @@
specialize qw/vp9_sad4x4/, "$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";
-specialize qw/vp9_sad64x64_avg/, "$sse2_x86inc";
+specialize qw/vp9_sad64x64_avg avx2/, "$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";
-specialize qw/vp9_sad32x64_avg/, "$sse2_x86inc";
+specialize qw/vp9_sad32x64_avg avx2/, "$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";
-specialize qw/vp9_sad64x32_avg/, "$sse2_x86inc";
+specialize qw/vp9_sad64x32_avg avx2/, "$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";
-specialize qw/vp9_sad32x16_avg/, "$sse2_x86inc";
+specialize qw/vp9_sad32x16_avg avx2/, "$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";
specialize qw/vp9_sad16x32_avg/, "$sse2_x86inc";
add_proto qw/unsigned int vp9_sad32x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vp9_sad32x32_avg/, "$sse2_x86inc";
+specialize qw/vp9_sad32x32_avg avx2/, "$sse2_x86inc";
add_proto qw/unsigned int vp9_sad16x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
specialize qw/vp9_sad16x16_avg/, "$sse2_x86inc";
@@ -999,11 +1110,28 @@
add_proto qw/unsigned int vp9_get_mb_ss/, "const int16_t *";
specialize qw/vp9_get_mb_ss/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_avg_8x8/, "const uint8_t *, int p";
+specialize qw/vp9_avg_8x8 sse2/;
+
+if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/unsigned int vp9_highbd_avg_8x8/, "const uint8_t *, int p";
+ specialize qw/vp9_highbd_avg_8x8/;
+}
+
# ENCODEMB INVOKE
add_proto qw/void vp9_subtract_block/, "int rows, int cols, int16_t *diff_ptr, ptrdiff_t diff_stride, const uint8_t *src_ptr, ptrdiff_t src_stride, const uint8_t *pred_ptr, ptrdiff_t pred_stride";
specialize qw/vp9_subtract_block neon/, "$sse2_x86inc";
+#
+# Denoiser
+#
+if (vpx_config("CONFIG_VP9_TEMPORAL_DENOISING") eq "yes") {
+ add_proto qw/int vp9_denoiser_filter/, "const uint8_t *sig, int sig_stride, const uint8_t *mc_avg, int mc_avg_stride, uint8_t *avg, int avg_stride, int increase_denoising, BLOCK_SIZE bs, int motion_magnitude";
+ specialize qw/vp9_denoiser_filter sse2/;
+}
+
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
# the transform coefficients are held in 32-bit
# values, so the assembler code for vp9_block_error can no longer be used.
@@ -1032,7 +1160,7 @@
specialize qw/vp9_quantize_fp_32x32/, "$ssse3_x86_64";
add_proto qw/void vp9_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
- specialize qw/vp9_quantize_b/, "$ssse3_x86_64";
+ specialize qw/vp9_quantize_b sse2/, "$ssse3_x86_64";
add_proto qw/void vp9_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
specialize qw/vp9_quantize_b_32x32/, "$ssse3_x86_64";
@@ -1154,647 +1282,644 @@
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
# variance
- add_proto qw/unsigned int vp9_high_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance32x16/;
- add_proto qw/unsigned int vp9_high_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance16x32/;
- add_proto qw/unsigned int vp9_high_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance64x32/;
- add_proto qw/unsigned int vp9_high_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance32x64/;
- add_proto qw/unsigned int vp9_high_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance32x32/;
- add_proto qw/unsigned int vp9_high_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance64x64/;
- add_proto qw/unsigned int vp9_high_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance16x16/;
- add_proto qw/unsigned int vp9_high_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance16x8/;
- add_proto qw/unsigned int vp9_high_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance8x16/;
- add_proto qw/unsigned int vp9_high_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance8x8/;
- add_proto qw/unsigned int vp9_high_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance8x4/;
- add_proto qw/unsigned int vp9_high_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance4x8/;
- add_proto qw/unsigned int vp9_high_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_variance4x4/;
- add_proto qw/void vp9_high_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
- specialize qw/vp9_high_get8x8var/;
+ add_proto qw/void vp9_highbd_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
+ specialize qw/vp9_highbd_get8x8var/;
- add_proto qw/void vp9_high_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
- specialize qw/vp9_high_get16x16var/;
+ add_proto qw/void vp9_highbd_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
+ specialize qw/vp9_highbd_get16x16var/;
- add_proto qw/unsigned int vp9_high_10_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_10_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance32x16/;
- add_proto qw/unsigned int vp9_high_10_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_10_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance16x32/;
- add_proto qw/unsigned int vp9_high_10_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_10_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance64x32/;
- add_proto qw/unsigned int vp9_high_10_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_10_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance32x64/;
- add_proto qw/unsigned int vp9_high_10_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_10_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance32x32/;
- add_proto qw/unsigned int vp9_high_10_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_10_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance64x64/;
- add_proto qw/unsigned int vp9_high_10_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_10_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance16x16/;
- add_proto qw/unsigned int vp9_high_10_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_10_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance16x8/;
- add_proto qw/unsigned int vp9_high_10_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_10_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance8x16/;
- add_proto qw/unsigned int vp9_high_10_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_10_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance8x8/;
- add_proto qw/unsigned int vp9_high_10_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_10_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance8x4/;
- add_proto qw/unsigned int vp9_high_10_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_10_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance4x8/;
- add_proto qw/unsigned int vp9_high_10_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_10_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_variance4x4/;
- add_proto qw/void vp9_high_10_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
- specialize qw/vp9_high_10_get8x8var/;
+ add_proto qw/void vp9_highbd_10_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
+ specialize qw/vp9_highbd_10_get8x8var/;
- add_proto qw/void vp9_high_10_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
- specialize qw/vp9_high_10_get16x16var/;
+ add_proto qw/void vp9_highbd_10_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
+ specialize qw/vp9_highbd_10_get16x16var/;
- add_proto qw/unsigned int vp9_high_12_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_12_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance32x16/;
- add_proto qw/unsigned int vp9_high_12_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_12_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance16x32/;
- add_proto qw/unsigned int vp9_high_12_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_12_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance64x32/;
- add_proto qw/unsigned int vp9_high_12_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_12_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance32x64/;
- add_proto qw/unsigned int vp9_high_12_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_12_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance32x32/;
- add_proto qw/unsigned int vp9_high_12_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_12_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance64x64/;
- add_proto qw/unsigned int vp9_high_12_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_12_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance16x16/;
- add_proto qw/unsigned int vp9_high_12_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_12_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance16x8/;
- add_proto qw/unsigned int vp9_high_12_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_12_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance8x16/;
- add_proto qw/unsigned int vp9_high_12_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_12_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance8x8/;
- add_proto qw/unsigned int vp9_high_12_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_12_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance8x4/;
- add_proto qw/unsigned int vp9_high_12_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_12_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance4x8/;
- add_proto qw/unsigned int vp9_high_12_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_12_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_variance4x4/;
- add_proto qw/void vp9_high_12_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
- specialize qw/vp9_high_12_get8x8var/;
+ add_proto qw/void vp9_highbd_12_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
+ specialize qw/vp9_highbd_12_get8x8var/;
- add_proto qw/void vp9_high_12_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
- specialize qw/vp9_high_12_get16x16var/;
+ add_proto qw/void vp9_highbd_12_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
+ specialize qw/vp9_highbd_12_get16x16var/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance64x64/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance64x64/, "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_high_sub_pixel_avg_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance64x64/, "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_highbd_sub_pixel_avg_variance64x64/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance32x64/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance32x64/, "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_high_sub_pixel_avg_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance32x64/, "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_highbd_sub_pixel_avg_variance32x64/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance64x32/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance64x32/, "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_high_sub_pixel_avg_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance64x32/, "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_highbd_sub_pixel_avg_variance64x32/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance32x16/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance32x16/, "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_high_sub_pixel_avg_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance32x16/, "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_highbd_sub_pixel_avg_variance32x16/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance16x32/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance16x32/, "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_high_sub_pixel_avg_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance16x32/, "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_highbd_sub_pixel_avg_variance16x32/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance32x32/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance32x32/, "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_high_sub_pixel_avg_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance32x32/, "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_highbd_sub_pixel_avg_variance32x32/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance16x16/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance16x16/, "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_high_sub_pixel_avg_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance16x16/, "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_highbd_sub_pixel_avg_variance16x16/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance8x16/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance8x16/, "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_high_sub_pixel_avg_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance8x16/, "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_highbd_sub_pixel_avg_variance8x16/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance16x8/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance16x8/, "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_high_sub_pixel_avg_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance16x8/, "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_highbd_sub_pixel_avg_variance16x8/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance8x8/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance8x8/, "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_high_sub_pixel_avg_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance8x8/, "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_highbd_sub_pixel_avg_variance8x8/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance8x4/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance8x4/, "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_high_sub_pixel_avg_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance8x4/, "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_highbd_sub_pixel_avg_variance8x4/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance4x8/;
- add_proto qw/unsigned int vp9_high_sub_pixel_avg_variance4x8/, "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_high_sub_pixel_avg_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_avg_variance4x8/, "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_highbd_sub_pixel_avg_variance4x8/;
- add_proto qw/unsigned int vp9_high_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_sub_pixel_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_sub_pixel_variance4x4/;
- add_proto qw/unsigned int vp9_high_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_high_sub_pixel_avg_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_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_highbd_sub_pixel_avg_variance4x4/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance64x64/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance64x64/, "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_high_10_sub_pixel_avg_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance64x64/, "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_highbd_10_sub_pixel_avg_variance64x64/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance32x64/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance32x64/, "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_high_10_sub_pixel_avg_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance32x64/, "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_highbd_10_sub_pixel_avg_variance32x64/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance64x32/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance64x32/, "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_high_10_sub_pixel_avg_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance64x32/, "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_highbd_10_sub_pixel_avg_variance64x32/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance32x16/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance32x16/, "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_high_10_sub_pixel_avg_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance32x16/, "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_highbd_10_sub_pixel_avg_variance32x16/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance16x32/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance16x32/, "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_high_10_sub_pixel_avg_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance16x32/, "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_highbd_10_sub_pixel_avg_variance16x32/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance32x32/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance32x32/, "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_high_10_sub_pixel_avg_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance32x32/, "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_highbd_10_sub_pixel_avg_variance32x32/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance16x16/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance16x16/, "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_high_10_sub_pixel_avg_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance16x16/, "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_highbd_10_sub_pixel_avg_variance16x16/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance8x16/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance8x16/, "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_high_10_sub_pixel_avg_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance8x16/, "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_highbd_10_sub_pixel_avg_variance8x16/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance16x8/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance16x8/, "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_high_10_sub_pixel_avg_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance16x8/, "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_highbd_10_sub_pixel_avg_variance16x8/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance8x8/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance8x8/, "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_high_10_sub_pixel_avg_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance8x8/, "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_highbd_10_sub_pixel_avg_variance8x8/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance8x4/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance8x4/, "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_high_10_sub_pixel_avg_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance8x4/, "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_highbd_10_sub_pixel_avg_variance8x4/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance4x8/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_avg_variance4x8/, "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_high_10_sub_pixel_avg_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_avg_variance4x8/, "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_highbd_10_sub_pixel_avg_variance4x8/;
- add_proto qw/unsigned int vp9_high_10_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_10_sub_pixel_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_10_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_sub_pixel_variance4x4/;
- add_proto qw/unsigned int vp9_high_10_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_high_10_sub_pixel_avg_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_10_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_highbd_10_sub_pixel_avg_variance4x4/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance64x64/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance64x64/, "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_high_12_sub_pixel_avg_variance64x64/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance64x64/, "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_highbd_12_sub_pixel_avg_variance64x64/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance32x64/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance32x64/, "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_high_12_sub_pixel_avg_variance32x64/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance32x64/, "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_highbd_12_sub_pixel_avg_variance32x64/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance64x32/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance64x32/, "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_high_12_sub_pixel_avg_variance64x32/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance64x32/, "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_highbd_12_sub_pixel_avg_variance64x32/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance32x16/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance32x16/, "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_high_12_sub_pixel_avg_variance32x16/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance32x16/, "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_highbd_12_sub_pixel_avg_variance32x16/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance16x32/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance16x32/, "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_high_12_sub_pixel_avg_variance16x32/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance16x32/, "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_highbd_12_sub_pixel_avg_variance16x32/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance32x32/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance32x32/, "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_high_12_sub_pixel_avg_variance32x32/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance32x32/, "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_highbd_12_sub_pixel_avg_variance32x32/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance16x16/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance16x16/, "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_high_12_sub_pixel_avg_variance16x16/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance16x16/, "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_highbd_12_sub_pixel_avg_variance16x16/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance8x16/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance8x16/, "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_high_12_sub_pixel_avg_variance8x16/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance8x16/, "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_highbd_12_sub_pixel_avg_variance8x16/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance16x8/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance16x8/, "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_high_12_sub_pixel_avg_variance16x8/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance16x8/, "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_highbd_12_sub_pixel_avg_variance16x8/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance8x8/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance8x8/, "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_high_12_sub_pixel_avg_variance8x8/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance8x8/, "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_highbd_12_sub_pixel_avg_variance8x8/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance8x4/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance8x4/, "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_high_12_sub_pixel_avg_variance8x4/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance8x4/, "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_highbd_12_sub_pixel_avg_variance8x4/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance4x8/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_avg_variance4x8/, "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_high_12_sub_pixel_avg_variance4x8/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_avg_variance4x8/, "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_highbd_12_sub_pixel_avg_variance4x8/;
- add_proto qw/unsigned int vp9_high_12_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vp9_high_12_sub_pixel_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_12_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_sub_pixel_variance4x4/;
- add_proto qw/unsigned int vp9_high_12_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_high_12_sub_pixel_avg_variance4x4/;
+ add_proto qw/unsigned int vp9_highbd_12_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_highbd_12_sub_pixel_avg_variance4x4/;
- add_proto qw/unsigned int vp9_high_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad64x64/;
+ add_proto qw/unsigned int vp9_highbd_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad64x64/;
- add_proto qw/unsigned int vp9_high_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad32x64/;
+ add_proto qw/unsigned int vp9_highbd_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad32x64/;
- add_proto qw/unsigned int vp9_high_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad64x32/;
+ add_proto qw/unsigned int vp9_highbd_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad64x32/;
- add_proto qw/unsigned int vp9_high_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad32x16/;
+ add_proto qw/unsigned int vp9_highbd_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad32x16/;
- add_proto qw/unsigned int vp9_high_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad16x32/;
+ add_proto qw/unsigned int vp9_highbd_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad16x32/;
- add_proto qw/unsigned int vp9_high_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad32x32/;
+ add_proto qw/unsigned int vp9_highbd_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad32x32/;
- add_proto qw/unsigned int vp9_high_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad16x16/;
+ add_proto qw/unsigned int vp9_highbd_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad16x16/;
- add_proto qw/unsigned int vp9_high_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad16x8/;
+ add_proto qw/unsigned int vp9_highbd_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad16x8/;
- add_proto qw/unsigned int vp9_high_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad8x16/;
+ add_proto qw/unsigned int vp9_highbd_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad8x16/;
- add_proto qw/unsigned int vp9_high_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad8x8/;
+ add_proto qw/unsigned int vp9_highbd_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad8x8/;
- add_proto qw/unsigned int vp9_high_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad8x4/;
+ add_proto qw/unsigned int vp9_highbd_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad8x4/;
- add_proto qw/unsigned int vp9_high_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad4x8/;
+ add_proto qw/unsigned int vp9_highbd_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad4x8/;
- add_proto qw/unsigned int vp9_high_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
- specialize qw/vp9_high_sad4x4/;
+ add_proto qw/unsigned int vp9_highbd_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride";
+ specialize qw/vp9_highbd_sad4x4/;
- add_proto qw/unsigned int vp9_high_sad64x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad64x64_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad64x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad64x64_avg/;
- add_proto qw/unsigned int vp9_high_sad32x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad32x64_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad32x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad32x64_avg/;
- add_proto qw/unsigned int vp9_high_sad64x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad64x32_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad64x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad64x32_avg/;
- add_proto qw/unsigned int vp9_high_sad32x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad32x16_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad32x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad32x16_avg/;
- add_proto qw/unsigned int vp9_high_sad16x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad16x32_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad16x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad16x32_avg/;
- add_proto qw/unsigned int vp9_high_sad32x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad32x32_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad32x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad32x32_avg/;
- add_proto qw/unsigned int vp9_high_sad16x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad16x16_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad16x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad16x16_avg/;
- add_proto qw/unsigned int vp9_high_sad16x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad16x8_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad16x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad16x8_avg/;
- add_proto qw/unsigned int vp9_high_sad8x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad8x16_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad8x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad8x16_avg/;
- add_proto qw/unsigned int vp9_high_sad8x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad8x8_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad8x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad8x8_avg/;
- add_proto qw/unsigned int vp9_high_sad8x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad8x4_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad8x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad8x4_avg/;
- add_proto qw/unsigned int vp9_high_sad4x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad4x8_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad4x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad4x8_avg/;
- add_proto qw/unsigned int vp9_high_sad4x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
- specialize qw/vp9_high_sad4x4_avg/;
+ add_proto qw/unsigned int vp9_highbd_sad4x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
+ specialize qw/vp9_highbd_sad4x4_avg/;
- add_proto qw/void vp9_high_sad64x64x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad64x64x3/;
+ add_proto qw/void vp9_highbd_sad64x64x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad64x64x3/;
- add_proto qw/void vp9_high_sad32x32x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad32x32x3/;
+ add_proto qw/void vp9_highbd_sad32x32x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad32x32x3/;
- add_proto qw/void vp9_high_sad16x16x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad16x16x3/;
+ add_proto qw/void vp9_highbd_sad16x16x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad16x16x3/;
- add_proto qw/void vp9_high_sad16x8x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad16x8x3/;
+ add_proto qw/void vp9_highbd_sad16x8x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad16x8x3/;
- add_proto qw/void vp9_high_sad8x16x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad8x16x3/;
+ add_proto qw/void vp9_highbd_sad8x16x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad8x16x3/;
- add_proto qw/void vp9_high_sad8x8x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad8x8x3/;
+ add_proto qw/void vp9_highbd_sad8x8x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad8x8x3/;
- add_proto qw/void vp9_high_sad4x4x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad4x4x3/;
+ add_proto qw/void vp9_highbd_sad4x4x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad4x4x3/;
- add_proto qw/void vp9_high_sad64x64x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad64x64x8/;
+ add_proto qw/void vp9_highbd_sad64x64x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad64x64x8/;
- add_proto qw/void vp9_high_sad32x32x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad32x32x8/;
+ add_proto qw/void vp9_highbd_sad32x32x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad32x32x8/;
- add_proto qw/void vp9_high_sad16x16x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad16x16x8/;
+ add_proto qw/void vp9_highbd_sad16x16x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad16x16x8/;
- add_proto qw/void vp9_high_sad16x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad16x8x8/;
+ add_proto qw/void vp9_highbd_sad16x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad16x8x8/;
- add_proto qw/void vp9_high_sad8x16x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad8x16x8/;
+ add_proto qw/void vp9_highbd_sad8x16x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad8x16x8/;
- add_proto qw/void vp9_high_sad8x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad8x8x8/;
+ add_proto qw/void vp9_highbd_sad8x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad8x8x8/;
- add_proto qw/void vp9_high_sad8x4x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad8x4x8/;
+ add_proto qw/void vp9_highbd_sad8x4x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad8x4x8/;
- add_proto qw/void vp9_high_sad4x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad4x8x8/;
+ add_proto qw/void vp9_highbd_sad4x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad4x8x8/;
- add_proto qw/void vp9_high_sad4x4x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
- specialize qw/vp9_high_sad4x4x8/;
+ add_proto qw/void vp9_highbd_sad4x4x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+ specialize qw/vp9_highbd_sad4x4x8/;
- add_proto qw/void vp9_high_sad64x64x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad64x64x4d/;
+ add_proto qw/void vp9_highbd_sad64x64x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad64x64x4d/;
- add_proto qw/void vp9_high_sad32x64x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad32x64x4d/;
+ add_proto qw/void vp9_highbd_sad32x64x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad32x64x4d/;
- add_proto qw/void vp9_high_sad64x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad64x32x4d/;
+ add_proto qw/void vp9_highbd_sad64x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad64x32x4d/;
- add_proto qw/void vp9_high_sad32x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad32x16x4d/;
+ add_proto qw/void vp9_highbd_sad32x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad32x16x4d/;
- add_proto qw/void vp9_high_sad16x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad16x32x4d/;
+ add_proto qw/void vp9_highbd_sad16x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad16x32x4d/;
- add_proto qw/void vp9_high_sad32x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad32x32x4d/;
+ add_proto qw/void vp9_highbd_sad32x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad32x32x4d/;
- add_proto qw/void vp9_high_sad16x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad16x16x4d/;
+ add_proto qw/void vp9_highbd_sad16x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad16x16x4d/;
- add_proto qw/void vp9_high_sad16x8x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad16x8x4d/;
+ add_proto qw/void vp9_highbd_sad16x8x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad16x8x4d/;
- add_proto qw/void vp9_high_sad8x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad8x16x4d/;
+ add_proto qw/void vp9_highbd_sad8x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad8x16x4d/;
- add_proto qw/void vp9_high_sad8x8x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad8x8x4d/;
+ add_proto qw/void vp9_highbd_sad8x8x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad8x8x4d/;
# TODO(jingning): need to convert these 4x8/8x4 functions into sse2 form
- add_proto qw/void vp9_high_sad8x4x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad8x4x4d/;
+ add_proto qw/void vp9_highbd_sad8x4x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad8x4x4d/;
- add_proto qw/void vp9_high_sad4x8x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad4x8x4d/;
+ add_proto qw/void vp9_highbd_sad4x8x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad4x8x4d/;
- add_proto qw/void vp9_high_sad4x4x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
- specialize qw/vp9_high_sad4x4x4d/;
+ add_proto qw/void vp9_highbd_sad4x4x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+ specialize qw/vp9_highbd_sad4x4x4d/;
- add_proto qw/unsigned int vp9_high_mse16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_mse16x16/;
+ add_proto qw/unsigned int vp9_highbd_mse16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_mse16x16/;
- add_proto qw/unsigned int vp9_high_mse8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_mse8x16/;
+ add_proto qw/unsigned int vp9_highbd_mse8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_mse8x16/;
- add_proto qw/unsigned int vp9_high_mse16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_mse16x8/;
+ add_proto qw/unsigned int vp9_highbd_mse16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_mse16x8/;
- add_proto qw/unsigned int vp9_high_mse8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_mse8x8/;
+ add_proto qw/unsigned int vp9_highbd_mse8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_mse8x8/;
- add_proto qw/unsigned int vp9_high_10_mse16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_10_mse16x16/;
+ add_proto qw/unsigned int vp9_highbd_10_mse16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_mse16x16/;
- add_proto qw/unsigned int vp9_high_10_mse8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_10_mse8x16/;
+ add_proto qw/unsigned int vp9_highbd_10_mse8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_mse8x16/;
- add_proto qw/unsigned int vp9_high_10_mse16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_10_mse16x8/;
+ add_proto qw/unsigned int vp9_highbd_10_mse16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_mse16x8/;
- add_proto qw/unsigned int vp9_high_10_mse8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_10_mse8x8/;
+ add_proto qw/unsigned int vp9_highbd_10_mse8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_10_mse8x8/;
- add_proto qw/unsigned int vp9_high_12_mse16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_12_mse16x16/;
+ add_proto qw/unsigned int vp9_highbd_12_mse16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_mse16x16/;
- add_proto qw/unsigned int vp9_high_12_mse8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_12_mse8x16/;
+ add_proto qw/unsigned int vp9_highbd_12_mse8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_mse8x16/;
- add_proto qw/unsigned int vp9_high_12_mse16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_12_mse16x8/;
+ add_proto qw/unsigned int vp9_highbd_12_mse16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_mse16x8/;
- add_proto qw/unsigned int vp9_high_12_mse8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
- specialize qw/vp9_high_12_mse8x8/;
+ add_proto qw/unsigned int vp9_highbd_12_mse8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
+ specialize qw/vp9_highbd_12_mse8x8/;
# ENCODEMB INVOKE
- add_proto qw/int64_t vp9_high_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz, int bd";
- specialize qw/vp9_high_block_error/;
+ add_proto qw/int64_t vp9_highbd_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz, int bd";
+ specialize qw/vp9_highbd_block_error/;
- add_proto qw/void vp9_high_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, int bd";
- specialize qw/vp9_high_subtract_block/;
+ add_proto qw/void vp9_highbd_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, int bd";
+ specialize qw/vp9_highbd_subtract_block/;
- add_proto qw/void vp9_high_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
- specialize qw/vp9_high_quantize_fp/;
+ add_proto qw/void vp9_highbd_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
+ specialize qw/vp9_highbd_quantize_fp/;
- add_proto qw/void vp9_high_quantize_fp_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
- specialize qw/vp9_high_quantize_fp_32x32/;
+ add_proto qw/void vp9_highbd_quantize_fp_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
+ specialize qw/vp9_highbd_quantize_fp_32x32/;
- add_proto qw/void vp9_high_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
- specialize qw/vp9_high_quantize_b/;
+ add_proto qw/void vp9_highbd_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
+ specialize qw/vp9_highbd_quantize_b/;
- add_proto qw/void vp9_high_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
- specialize qw/vp9_high_quantize_b_32x32/;
+ add_proto qw/void vp9_highbd_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
+ specialize qw/vp9_highbd_quantize_b_32x32/;
#
# Structured Similarity (SSIM)
#
if (vpx_config("CONFIG_INTERNAL_STATS") eq "yes") {
- add_proto qw/void vp9_high_ssim_parms_8x8/, "uint16_t *s, int sp, uint16_t *r, int rp, uint32_t *sum_s, uint32_t *sum_r, uint32_t *sum_sq_s, uint32_t *sum_sq_r, uint32_t *sum_sxr";
- specialize qw/vp9_high_ssim_parms_8x8/;
-
- add_proto qw/void vp9_high_ssim_parms_8x8_shift/, "uint16_t *s, int sp, uint16_t *r, int rp, uint32_t *sum_s, uint32_t *sum_r, uint32_t *sum_sq_s, uint32_t *sum_sq_r, uint32_t *sum_sxr, unsigned int bd, unsigned int shift";
- specialize qw/vp9_high_ssim_parms_8x8_shift/;
+ add_proto qw/void vp9_highbd_ssim_parms_8x8/, "uint16_t *s, int sp, uint16_t *r, int rp, uint32_t *sum_s, uint32_t *sum_r, uint32_t *sum_sq_s, uint32_t *sum_sq_r, uint32_t *sum_sxr";
+ specialize qw/vp9_highbd_ssim_parms_8x8/;
}
# fdct functions
- add_proto qw/void vp9_high_fht4x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/vp9_high_fht4x4/;
+ add_proto qw/void vp9_highbd_fht4x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+ specialize qw/vp9_highbd_fht4x4/;
- add_proto qw/void vp9_high_fht8x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/vp9_high_fht8x8/;
+ add_proto qw/void vp9_highbd_fht8x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+ specialize qw/vp9_highbd_fht8x8/;
- add_proto qw/void vp9_high_fht16x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/vp9_high_fht16x16/;
+ add_proto qw/void vp9_highbd_fht16x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+ specialize qw/vp9_highbd_fht16x16/;
- add_proto qw/void vp9_high_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fwht4x4/;
+ add_proto qw/void vp9_highbd_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fwht4x4/;
- add_proto qw/void vp9_high_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fdct4x4/;
+ add_proto qw/void vp9_highbd_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fdct4x4/;
- add_proto qw/void vp9_high_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fdct8x8_1/;
+ add_proto qw/void vp9_highbd_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fdct8x8_1/;
- add_proto qw/void vp9_high_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fdct8x8/;
+ add_proto qw/void vp9_highbd_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fdct8x8/;
- add_proto qw/void vp9_high_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fdct16x16_1/;
+ add_proto qw/void vp9_highbd_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fdct16x16_1/;
- add_proto qw/void vp9_high_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fdct16x16/;
+ add_proto qw/void vp9_highbd_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fdct16x16/;
- add_proto qw/void vp9_high_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fdct32x32_1/;
+ add_proto qw/void vp9_highbd_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fdct32x32_1/;
- add_proto qw/void vp9_high_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fdct32x32/;
+ add_proto qw/void vp9_highbd_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fdct32x32/;
- add_proto qw/void vp9_high_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_high_fdct32x32_rd/;
+ add_proto qw/void vp9_highbd_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vp9_highbd_fdct32x32_rd/;
- add_proto qw/void vp9_high_temporal_filter_apply/, "uint8_t *frame1, unsigned int stride, uint8_t *frame2, unsigned int block_width, unsigned int block_height, int strength, int filter_weight, unsigned int *accumulator, uint16_t *count";
- specialize qw/vp9_high_temporal_filter_apply/;
+ add_proto qw/void vp9_highbd_temporal_filter_apply/, "uint8_t *frame1, unsigned int stride, uint8_t *frame2, unsigned int block_width, unsigned int block_height, int strength, int filter_weight, unsigned int *accumulator, uint16_t *count";
+ specialize qw/vp9_highbd_temporal_filter_apply/;
}
# End vp9_high encoder functions
diff --git a/source/libvpx/vp9/common/vp9_scale.c b/source/libvpx/vp9/common/vp9_scale.c
index 63e2b53..6db8f9c 100644
--- a/source/libvpx/vp9/common/vp9_scale.c
+++ b/source/libvpx/vp9/common/vp9_scale.c
@@ -47,7 +47,7 @@
void vp9_setup_scale_factors_for_frame(struct scale_factors *sf,
int other_w, int other_h,
int this_w, int this_h,
- int use_high) {
+ int use_highbd) {
#else
void vp9_setup_scale_factors_for_frame(struct scale_factors *sf,
int other_w, int other_h,
@@ -119,47 +119,47 @@
sf->predict[1][1][0] = vp9_convolve8;
sf->predict[1][1][1] = vp9_convolve8_avg;
#if CONFIG_VP9_HIGHBITDEPTH
- if (use_high) {
+ if (use_highbd) {
if (sf->x_step_q4 == 16) {
if (sf->y_step_q4 == 16) {
// No scaling in either direction.
- sf->high_predict[0][0][0] = vp9_high_convolve_copy;
- sf->high_predict[0][0][1] = vp9_high_convolve_avg;
- sf->high_predict[0][1][0] = vp9_high_convolve8_vert;
- sf->high_predict[0][1][1] = vp9_high_convolve8_avg_vert;
- sf->high_predict[1][0][0] = vp9_high_convolve8_horiz;
- sf->high_predict[1][0][1] = vp9_high_convolve8_avg_horiz;
+ sf->highbd_predict[0][0][0] = vp9_highbd_convolve_copy;
+ sf->highbd_predict[0][0][1] = vp9_highbd_convolve_avg;
+ sf->highbd_predict[0][1][0] = vp9_highbd_convolve8_vert;
+ sf->highbd_predict[0][1][1] = vp9_highbd_convolve8_avg_vert;
+ sf->highbd_predict[1][0][0] = vp9_highbd_convolve8_horiz;
+ sf->highbd_predict[1][0][1] = vp9_highbd_convolve8_avg_horiz;
} else {
// No scaling in x direction. Must always scale in the y direction.
- sf->high_predict[0][0][0] = vp9_high_convolve8_vert;
- sf->high_predict[0][0][1] = vp9_high_convolve8_avg_vert;
- sf->high_predict[0][1][0] = vp9_high_convolve8_vert;
- sf->high_predict[0][1][1] = vp9_high_convolve8_avg_vert;
- sf->high_predict[1][0][0] = vp9_high_convolve8;
- sf->high_predict[1][0][1] = vp9_high_convolve8_avg;
+ sf->highbd_predict[0][0][0] = vp9_highbd_convolve8_vert;
+ sf->highbd_predict[0][0][1] = vp9_highbd_convolve8_avg_vert;
+ sf->highbd_predict[0][1][0] = vp9_highbd_convolve8_vert;
+ sf->highbd_predict[0][1][1] = vp9_highbd_convolve8_avg_vert;
+ sf->highbd_predict[1][0][0] = vp9_highbd_convolve8;
+ sf->highbd_predict[1][0][1] = vp9_highbd_convolve8_avg;
}
} else {
if (sf->y_step_q4 == 16) {
// No scaling in the y direction. Must always scale in the x direction.
- sf->high_predict[0][0][0] = vp9_high_convolve8_horiz;
- sf->high_predict[0][0][1] = vp9_high_convolve8_avg_horiz;
- sf->high_predict[0][1][0] = vp9_high_convolve8;
- sf->high_predict[0][1][1] = vp9_high_convolve8_avg;
- sf->high_predict[1][0][0] = vp9_high_convolve8_horiz;
- sf->high_predict[1][0][1] = vp9_high_convolve8_avg_horiz;
+ sf->highbd_predict[0][0][0] = vp9_highbd_convolve8_horiz;
+ sf->highbd_predict[0][0][1] = vp9_highbd_convolve8_avg_horiz;
+ sf->highbd_predict[0][1][0] = vp9_highbd_convolve8;
+ sf->highbd_predict[0][1][1] = vp9_highbd_convolve8_avg;
+ sf->highbd_predict[1][0][0] = vp9_highbd_convolve8_horiz;
+ sf->highbd_predict[1][0][1] = vp9_highbd_convolve8_avg_horiz;
} else {
// Must always scale in both directions.
- sf->high_predict[0][0][0] = vp9_high_convolve8;
- sf->high_predict[0][0][1] = vp9_high_convolve8_avg;
- sf->high_predict[0][1][0] = vp9_high_convolve8;
- sf->high_predict[0][1][1] = vp9_high_convolve8_avg;
- sf->high_predict[1][0][0] = vp9_high_convolve8;
- sf->high_predict[1][0][1] = vp9_high_convolve8_avg;
+ sf->highbd_predict[0][0][0] = vp9_highbd_convolve8;
+ sf->highbd_predict[0][0][1] = vp9_highbd_convolve8_avg;
+ sf->highbd_predict[0][1][0] = vp9_highbd_convolve8;
+ sf->highbd_predict[0][1][1] = vp9_highbd_convolve8_avg;
+ sf->highbd_predict[1][0][0] = vp9_highbd_convolve8;
+ sf->highbd_predict[1][0][1] = vp9_highbd_convolve8_avg;
}
}
// 2D subpel motion always gets filtered in both directions.
- sf->high_predict[1][1][0] = vp9_high_convolve8;
- sf->high_predict[1][1][1] = vp9_high_convolve8_avg;
+ sf->highbd_predict[1][1][0] = vp9_highbd_convolve8;
+ sf->highbd_predict[1][1][1] = vp9_highbd_convolve8_avg;
}
#endif
}
diff --git a/source/libvpx/vp9/common/vp9_scale.h b/source/libvpx/vp9/common/vp9_scale.h
index 2e923db..a1601a7 100644
--- a/source/libvpx/vp9/common/vp9_scale.h
+++ b/source/libvpx/vp9/common/vp9_scale.h
@@ -33,7 +33,7 @@
convolve_fn_t predict[2][2][2]; // horiz, vert, avg
#if CONFIG_VP9_HIGHBITDEPTH
- high_convolve_fn_t high_predict[2][2][2]; // horiz, vert, avg
+ highbd_convolve_fn_t highbd_predict[2][2][2]; // horiz, vert, avg
#endif
};
diff --git a/source/libvpx/vp9/common/vp9_scan.c b/source/libvpx/vp9/common/vp9_scan.c
index 1ec5a0c..d6fb8b2 100644
--- a/source/libvpx/vp9/common/vp9_scan.c
+++ b/source/libvpx/vp9/common/vp9_scan.c
@@ -233,37 +233,467 @@
// in {top, left, topleft, topright, bottomleft} order
// for each position in raster scan order.
// -1 indicates the neighbor does not exist.
-DECLARE_ALIGNED(16, static int16_t,
- default_scan_4x4_neighbors[17 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- col_scan_4x4_neighbors[17 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- row_scan_4x4_neighbors[17 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- col_scan_8x8_neighbors[65 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- row_scan_8x8_neighbors[65 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- default_scan_8x8_neighbors[65 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- col_scan_16x16_neighbors[257 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- row_scan_16x16_neighbors[257 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- default_scan_16x16_neighbors[257 * MAX_NEIGHBORS]);
-DECLARE_ALIGNED(16, static int16_t,
- default_scan_32x32_neighbors[1025 * MAX_NEIGHBORS]);
+DECLARE_ALIGNED(16, static const int16_t,
+ default_scan_4x4_neighbors[17 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 0, 0, 1, 4, 4, 4, 1, 1, 8, 8, 5, 8, 2, 2, 2, 5, 9, 12, 6, 9,
+ 3, 6, 10, 13, 7, 10, 11, 14, 0, 0,
+};
-DECLARE_ALIGNED(16, static int16_t, vp9_default_iscan_4x4[16]);
-DECLARE_ALIGNED(16, static int16_t, vp9_col_iscan_4x4[16]);
-DECLARE_ALIGNED(16, static int16_t, vp9_row_iscan_4x4[16]);
-DECLARE_ALIGNED(16, static int16_t, vp9_col_iscan_8x8[64]);
-DECLARE_ALIGNED(16, static int16_t, vp9_row_iscan_8x8[64]);
-DECLARE_ALIGNED(16, static int16_t, vp9_default_iscan_8x8[64]);
-DECLARE_ALIGNED(16, static int16_t, vp9_col_iscan_16x16[256]);
-DECLARE_ALIGNED(16, static int16_t, vp9_row_iscan_16x16[256]);
-DECLARE_ALIGNED(16, static int16_t, vp9_default_iscan_16x16[256]);
-DECLARE_ALIGNED(16, static int16_t, vp9_default_iscan_32x32[1024]);
+DECLARE_ALIGNED(16, static const int16_t,
+ col_scan_4x4_neighbors[17 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 4, 4, 0, 0, 8, 8, 1, 1, 5, 5, 1, 1, 9, 9, 2, 2, 6, 6, 2, 2, 3,
+ 3, 10, 10, 7, 7, 11, 11, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t,
+ row_scan_4x4_neighbors[17 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 0, 0, 1, 1, 4, 4, 2, 2, 5, 5, 4, 4, 8, 8, 6, 6, 8, 8, 9, 9, 12,
+ 12, 10, 10, 13, 13, 14, 14, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t,
+ col_scan_8x8_neighbors[65 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 8, 8, 0, 0, 16, 16, 1, 1, 24, 24, 9, 9, 1, 1, 32, 32, 17, 17, 2,
+ 2, 25, 25, 10, 10, 40, 40, 2, 2, 18, 18, 33, 33, 3, 3, 48, 48, 11, 11, 26,
+ 26, 3, 3, 41, 41, 19, 19, 34, 34, 4, 4, 27, 27, 12, 12, 49, 49, 42, 42, 20,
+ 20, 4, 4, 35, 35, 5, 5, 28, 28, 50, 50, 43, 43, 13, 13, 36, 36, 5, 5, 21, 21,
+ 51, 51, 29, 29, 6, 6, 44, 44, 14, 14, 6, 6, 37, 37, 52, 52, 22, 22, 7, 7, 30,
+ 30, 45, 45, 15, 15, 38, 38, 23, 23, 53, 53, 31, 31, 46, 46, 39, 39, 54, 54,
+ 47, 47, 55, 55, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t,
+ row_scan_8x8_neighbors[65 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 1, 1, 0, 0, 8, 8, 2, 2, 8, 8, 9, 9, 3, 3, 16, 16, 10, 10, 16, 16,
+ 4, 4, 17, 17, 24, 24, 11, 11, 18, 18, 25, 25, 24, 24, 5, 5, 12, 12, 19, 19,
+ 32, 32, 26, 26, 6, 6, 33, 33, 32, 32, 20, 20, 27, 27, 40, 40, 13, 13, 34, 34,
+ 40, 40, 41, 41, 28, 28, 35, 35, 48, 48, 21, 21, 42, 42, 14, 14, 48, 48, 36,
+ 36, 49, 49, 43, 43, 29, 29, 56, 56, 22, 22, 50, 50, 57, 57, 44, 44, 37, 37,
+ 51, 51, 30, 30, 58, 58, 52, 52, 45, 45, 59, 59, 38, 38, 60, 60, 46, 46, 53,
+ 53, 54, 54, 61, 61, 62, 62, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t,
+ default_scan_8x8_neighbors[65 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 0, 0, 8, 8, 1, 8, 1, 1, 9, 16, 16, 16, 2, 9, 2, 2, 10, 17, 17,
+ 24, 24, 24, 3, 10, 3, 3, 18, 25, 25, 32, 11, 18, 32, 32, 4, 11, 26, 33, 19,
+ 26, 4, 4, 33, 40, 12, 19, 40, 40, 5, 12, 27, 34, 34, 41, 20, 27, 13, 20, 5,
+ 5, 41, 48, 48, 48, 28, 35, 35, 42, 21, 28, 6, 6, 6, 13, 42, 49, 49, 56, 36,
+ 43, 14, 21, 29, 36, 7, 14, 43, 50, 50, 57, 22, 29, 37, 44, 15, 22, 44, 51,
+ 51, 58, 30, 37, 23, 30, 52, 59, 45, 52, 38, 45, 31, 38, 53, 60, 46, 53, 39,
+ 46, 54, 61, 47, 54, 55, 62, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t,
+ col_scan_16x16_neighbors[257 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 16, 16, 32, 32, 0, 0, 48, 48, 1, 1, 64, 64,
+ 17, 17, 80, 80, 33, 33, 1, 1, 49, 49, 96, 96, 2, 2, 65, 65,
+ 18, 18, 112, 112, 34, 34, 81, 81, 2, 2, 50, 50, 128, 128, 3, 3,
+ 97, 97, 19, 19, 66, 66, 144, 144, 82, 82, 35, 35, 113, 113, 3, 3,
+ 51, 51, 160, 160, 4, 4, 98, 98, 129, 129, 67, 67, 20, 20, 83, 83,
+ 114, 114, 36, 36, 176, 176, 4, 4, 145, 145, 52, 52, 99, 99, 5, 5,
+ 130, 130, 68, 68, 192, 192, 161, 161, 21, 21, 115, 115, 84, 84, 37, 37,
+ 146, 146, 208, 208, 53, 53, 5, 5, 100, 100, 177, 177, 131, 131, 69, 69,
+ 6, 6, 224, 224, 116, 116, 22, 22, 162, 162, 85, 85, 147, 147, 38, 38,
+ 193, 193, 101, 101, 54, 54, 6, 6, 132, 132, 178, 178, 70, 70, 163, 163,
+ 209, 209, 7, 7, 117, 117, 23, 23, 148, 148, 7, 7, 86, 86, 194, 194,
+ 225, 225, 39, 39, 179, 179, 102, 102, 133, 133, 55, 55, 164, 164, 8, 8,
+ 71, 71, 210, 210, 118, 118, 149, 149, 195, 195, 24, 24, 87, 87, 40, 40,
+ 56, 56, 134, 134, 180, 180, 226, 226, 103, 103, 8, 8, 165, 165, 211, 211,
+ 72, 72, 150, 150, 9, 9, 119, 119, 25, 25, 88, 88, 196, 196, 41, 41,
+ 135, 135, 181, 181, 104, 104, 57, 57, 227, 227, 166, 166, 120, 120, 151, 151,
+ 197, 197, 73, 73, 9, 9, 212, 212, 89, 89, 136, 136, 182, 182, 10, 10,
+ 26, 26, 105, 105, 167, 167, 228, 228, 152, 152, 42, 42, 121, 121, 213, 213,
+ 58, 58, 198, 198, 74, 74, 137, 137, 183, 183, 168, 168, 10, 10, 90, 90,
+ 229, 229, 11, 11, 106, 106, 214, 214, 153, 153, 27, 27, 199, 199, 43, 43,
+ 184, 184, 122, 122, 169, 169, 230, 230, 59, 59, 11, 11, 75, 75, 138, 138,
+ 200, 200, 215, 215, 91, 91, 12, 12, 28, 28, 185, 185, 107, 107, 154, 154,
+ 44, 44, 231, 231, 216, 216, 60, 60, 123, 123, 12, 12, 76, 76, 201, 201,
+ 170, 170, 232, 232, 139, 139, 92, 92, 13, 13, 108, 108, 29, 29, 186, 186,
+ 217, 217, 155, 155, 45, 45, 13, 13, 61, 61, 124, 124, 14, 14, 233, 233,
+ 77, 77, 14, 14, 171, 171, 140, 140, 202, 202, 30, 30, 93, 93, 109, 109,
+ 46, 46, 156, 156, 62, 62, 187, 187, 15, 15, 125, 125, 218, 218, 78, 78,
+ 31, 31, 172, 172, 47, 47, 141, 141, 94, 94, 234, 234, 203, 203, 63, 63,
+ 110, 110, 188, 188, 157, 157, 126, 126, 79, 79, 173, 173, 95, 95, 219, 219,
+ 142, 142, 204, 204, 235, 235, 111, 111, 158, 158, 127, 127, 189, 189, 220,
+ 220, 143, 143, 174, 174, 205, 205, 236, 236, 159, 159, 190, 190, 221, 221,
+ 175, 175, 237, 237, 206, 206, 222, 222, 191, 191, 238, 238, 207, 207, 223,
+ 223, 239, 239, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t,
+ row_scan_16x16_neighbors[257 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 1, 1, 0, 0, 2, 2, 16, 16, 3, 3, 17, 17,
+ 16, 16, 4, 4, 32, 32, 18, 18, 5, 5, 33, 33, 32, 32, 19, 19,
+ 48, 48, 6, 6, 34, 34, 20, 20, 49, 49, 48, 48, 7, 7, 35, 35,
+ 64, 64, 21, 21, 50, 50, 36, 36, 64, 64, 8, 8, 65, 65, 51, 51,
+ 22, 22, 37, 37, 80, 80, 66, 66, 9, 9, 52, 52, 23, 23, 81, 81,
+ 67, 67, 80, 80, 38, 38, 10, 10, 53, 53, 82, 82, 96, 96, 68, 68,
+ 24, 24, 97, 97, 83, 83, 39, 39, 96, 96, 54, 54, 11, 11, 69, 69,
+ 98, 98, 112, 112, 84, 84, 25, 25, 40, 40, 55, 55, 113, 113, 99, 99,
+ 12, 12, 70, 70, 112, 112, 85, 85, 26, 26, 114, 114, 100, 100, 128, 128,
+ 41, 41, 56, 56, 71, 71, 115, 115, 13, 13, 86, 86, 129, 129, 101, 101,
+ 128, 128, 72, 72, 130, 130, 116, 116, 27, 27, 57, 57, 14, 14, 87, 87,
+ 42, 42, 144, 144, 102, 102, 131, 131, 145, 145, 117, 117, 73, 73, 144, 144,
+ 88, 88, 132, 132, 103, 103, 28, 28, 58, 58, 146, 146, 118, 118, 43, 43,
+ 160, 160, 147, 147, 89, 89, 104, 104, 133, 133, 161, 161, 119, 119, 160, 160,
+ 74, 74, 134, 134, 148, 148, 29, 29, 59, 59, 162, 162, 176, 176, 44, 44,
+ 120, 120, 90, 90, 105, 105, 163, 163, 177, 177, 149, 149, 176, 176, 135, 135,
+ 164, 164, 178, 178, 30, 30, 150, 150, 192, 192, 75, 75, 121, 121, 60, 60,
+ 136, 136, 193, 193, 106, 106, 151, 151, 179, 179, 192, 192, 45, 45, 165, 165,
+ 166, 166, 194, 194, 91, 91, 180, 180, 137, 137, 208, 208, 122, 122, 152, 152,
+ 208, 208, 195, 195, 76, 76, 167, 167, 209, 209, 181, 181, 224, 224, 107, 107,
+ 196, 196, 61, 61, 153, 153, 224, 224, 182, 182, 168, 168, 210, 210, 46, 46,
+ 138, 138, 92, 92, 183, 183, 225, 225, 211, 211, 240, 240, 197, 197, 169, 169,
+ 123, 123, 154, 154, 198, 198, 77, 77, 212, 212, 184, 184, 108, 108, 226, 226,
+ 199, 199, 62, 62, 227, 227, 241, 241, 139, 139, 213, 213, 170, 170, 185, 185,
+ 155, 155, 228, 228, 242, 242, 124, 124, 93, 93, 200, 200, 243, 243, 214, 214,
+ 215, 215, 229, 229, 140, 140, 186, 186, 201, 201, 78, 78, 171, 171, 109, 109,
+ 156, 156, 244, 244, 216, 216, 230, 230, 94, 94, 245, 245, 231, 231, 125, 125,
+ 202, 202, 246, 246, 232, 232, 172, 172, 217, 217, 141, 141, 110, 110, 157,
+ 157, 187, 187, 247, 247, 126, 126, 233, 233, 218, 218, 248, 248, 188, 188,
+ 203, 203, 142, 142, 173, 173, 158, 158, 249, 249, 234, 234, 204, 204, 219,
+ 219, 174, 174, 189, 189, 250, 250, 220, 220, 190, 190, 205, 205, 235, 235,
+ 206, 206, 236, 236, 251, 251, 221, 221, 252, 252, 222, 222, 237, 237, 238,
+ 238, 253, 253, 254, 254, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t,
+ default_scan_16x16_neighbors[257 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 0, 0, 16, 16, 1, 16, 1, 1, 32, 32, 17, 32,
+ 2, 17, 2, 2, 48, 48, 18, 33, 33, 48, 3, 18, 49, 64, 64, 64,
+ 34, 49, 3, 3, 19, 34, 50, 65, 4, 19, 65, 80, 80, 80, 35, 50,
+ 4, 4, 20, 35, 66, 81, 81, 96, 51, 66, 96, 96, 5, 20, 36, 51,
+ 82, 97, 21, 36, 67, 82, 97, 112, 5, 5, 52, 67, 112, 112, 37, 52,
+ 6, 21, 83, 98, 98, 113, 68, 83, 6, 6, 113, 128, 22, 37, 53, 68,
+ 84, 99, 99, 114, 128, 128, 114, 129, 69, 84, 38, 53, 7, 22, 7, 7,
+ 129, 144, 23, 38, 54, 69, 100, 115, 85, 100, 115, 130, 144, 144, 130, 145,
+ 39, 54, 70, 85, 8, 23, 55, 70, 116, 131, 101, 116, 145, 160, 24, 39,
+ 8, 8, 86, 101, 131, 146, 160, 160, 146, 161, 71, 86, 40, 55, 9, 24,
+ 117, 132, 102, 117, 161, 176, 132, 147, 56, 71, 87, 102, 25, 40, 147, 162,
+ 9, 9, 176, 176, 162, 177, 72, 87, 41, 56, 118, 133, 133, 148, 103, 118,
+ 10, 25, 148, 163, 57, 72, 88, 103, 177, 192, 26, 41, 163, 178, 192, 192,
+ 10, 10, 119, 134, 73, 88, 149, 164, 104, 119, 134, 149, 42, 57, 178, 193,
+ 164, 179, 11, 26, 58, 73, 193, 208, 89, 104, 135, 150, 120, 135, 27, 42,
+ 74, 89, 208, 208, 150, 165, 179, 194, 165, 180, 105, 120, 194, 209, 43, 58,
+ 11, 11, 136, 151, 90, 105, 151, 166, 180, 195, 59, 74, 121, 136, 209, 224,
+ 195, 210, 224, 224, 166, 181, 106, 121, 75, 90, 12, 27, 181, 196, 12, 12,
+ 210, 225, 152, 167, 167, 182, 137, 152, 28, 43, 196, 211, 122, 137, 91, 106,
+ 225, 240, 44, 59, 13, 28, 107, 122, 182, 197, 168, 183, 211, 226, 153, 168,
+ 226, 241, 60, 75, 197, 212, 138, 153, 29, 44, 76, 91, 13, 13, 183, 198,
+ 123, 138, 45, 60, 212, 227, 198, 213, 154, 169, 169, 184, 227, 242, 92, 107,
+ 61, 76, 139, 154, 14, 29, 14, 14, 184, 199, 213, 228, 108, 123, 199, 214,
+ 228, 243, 77, 92, 30, 45, 170, 185, 155, 170, 185, 200, 93, 108, 124, 139,
+ 214, 229, 46, 61, 200, 215, 229, 244, 15, 30, 109, 124, 62, 77, 140, 155,
+ 215, 230, 31, 46, 171, 186, 186, 201, 201, 216, 78, 93, 230, 245, 125, 140,
+ 47, 62, 216, 231, 156, 171, 94, 109, 231, 246, 141, 156, 63, 78, 202, 217,
+ 187, 202, 110, 125, 217, 232, 172, 187, 232, 247, 79, 94, 157, 172, 126, 141,
+ 203, 218, 95, 110, 233, 248, 218, 233, 142, 157, 111, 126, 173, 188, 188, 203,
+ 234, 249, 219, 234, 127, 142, 158, 173, 204, 219, 189, 204, 143, 158, 235,
+ 250, 174, 189, 205, 220, 159, 174, 220, 235, 221, 236, 175, 190, 190, 205,
+ 236, 251, 206, 221, 237, 252, 191, 206, 222, 237, 207, 222, 238, 253, 223,
+ 238, 239, 254, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t,
+ default_scan_32x32_neighbors[1025 * MAX_NEIGHBORS]) = {
+ 0, 0, 0, 0, 0, 0, 32, 32, 1, 32, 1, 1, 64, 64, 33, 64,
+ 2, 33, 96, 96, 2, 2, 65, 96, 34, 65, 128, 128, 97, 128, 3, 34,
+ 66, 97, 3, 3, 35, 66, 98, 129, 129, 160, 160, 160, 4, 35, 67, 98,
+ 192, 192, 4, 4, 130, 161, 161, 192, 36, 67, 99, 130, 5, 36, 68, 99,
+ 193, 224, 162, 193, 224, 224, 131, 162, 37, 68, 100, 131, 5, 5, 194, 225,
+ 225, 256, 256, 256, 163, 194, 69, 100, 132, 163, 6, 37, 226, 257, 6, 6,
+ 195, 226, 257, 288, 101, 132, 288, 288, 38, 69, 164, 195, 133, 164, 258, 289,
+ 227, 258, 196, 227, 7, 38, 289, 320, 70, 101, 320, 320, 7, 7, 165, 196,
+ 39, 70, 102, 133, 290, 321, 259, 290, 228, 259, 321, 352, 352, 352, 197, 228,
+ 134, 165, 71, 102, 8, 39, 322, 353, 291, 322, 260, 291, 103, 134, 353, 384,
+ 166, 197, 229, 260, 40, 71, 8, 8, 384, 384, 135, 166, 354, 385, 323, 354,
+ 198, 229, 292, 323, 72, 103, 261, 292, 9, 40, 385, 416, 167, 198, 104, 135,
+ 230, 261, 355, 386, 416, 416, 293, 324, 324, 355, 9, 9, 41, 72, 386, 417,
+ 199, 230, 136, 167, 417, 448, 262, 293, 356, 387, 73, 104, 387, 418, 231, 262,
+ 10, 41, 168, 199, 325, 356, 418, 449, 105, 136, 448, 448, 42, 73, 294, 325,
+ 200, 231, 10, 10, 357, 388, 137, 168, 263, 294, 388, 419, 74, 105, 419, 450,
+ 449, 480, 326, 357, 232, 263, 295, 326, 169, 200, 11, 42, 106, 137, 480, 480,
+ 450, 481, 358, 389, 264, 295, 201, 232, 138, 169, 389, 420, 43, 74, 420, 451,
+ 327, 358, 11, 11, 481, 512, 233, 264, 451, 482, 296, 327, 75, 106, 170, 201,
+ 482, 513, 512, 512, 390, 421, 359, 390, 421, 452, 107, 138, 12, 43, 202, 233,
+ 452, 483, 265, 296, 328, 359, 139, 170, 44, 75, 483, 514, 513, 544, 234, 265,
+ 297, 328, 422, 453, 12, 12, 391, 422, 171, 202, 76, 107, 514, 545, 453, 484,
+ 544, 544, 266, 297, 203, 234, 108, 139, 329, 360, 298, 329, 140, 171, 515,
+ 546, 13, 44, 423, 454, 235, 266, 545, 576, 454, 485, 45, 76, 172, 203, 330,
+ 361, 576, 576, 13, 13, 267, 298, 546, 577, 77, 108, 204, 235, 455, 486, 577,
+ 608, 299, 330, 109, 140, 547, 578, 14, 45, 14, 14, 141, 172, 578, 609, 331,
+ 362, 46, 77, 173, 204, 15, 15, 78, 109, 205, 236, 579, 610, 110, 141, 15, 46,
+ 142, 173, 47, 78, 174, 205, 16, 16, 79, 110, 206, 237, 16, 47, 111, 142,
+ 48, 79, 143, 174, 80, 111, 175, 206, 17, 48, 17, 17, 207, 238, 49, 80,
+ 81, 112, 18, 18, 18, 49, 50, 81, 82, 113, 19, 50, 51, 82, 83, 114, 608, 608,
+ 484, 515, 360, 391, 236, 267, 112, 143, 19, 19, 640, 640, 609, 640, 516, 547,
+ 485, 516, 392, 423, 361, 392, 268, 299, 237, 268, 144, 175, 113, 144, 20, 51,
+ 20, 20, 672, 672, 641, 672, 610, 641, 548, 579, 517, 548, 486, 517, 424, 455,
+ 393, 424, 362, 393, 300, 331, 269, 300, 238, 269, 176, 207, 145, 176, 114,
+ 145, 52, 83, 21, 52, 21, 21, 704, 704, 673, 704, 642, 673, 611, 642, 580,
+ 611, 549, 580, 518, 549, 487, 518, 456, 487, 425, 456, 394, 425, 363, 394,
+ 332, 363, 301, 332, 270, 301, 239, 270, 208, 239, 177, 208, 146, 177, 115,
+ 146, 84, 115, 53, 84, 22, 53, 22, 22, 705, 736, 674, 705, 643, 674, 581, 612,
+ 550, 581, 519, 550, 457, 488, 426, 457, 395, 426, 333, 364, 302, 333, 271,
+ 302, 209, 240, 178, 209, 147, 178, 85, 116, 54, 85, 23, 54, 706, 737, 675,
+ 706, 582, 613, 551, 582, 458, 489, 427, 458, 334, 365, 303, 334, 210, 241,
+ 179, 210, 86, 117, 55, 86, 707, 738, 583, 614, 459, 490, 335, 366, 211, 242,
+ 87, 118, 736, 736, 612, 643, 488, 519, 364, 395, 240, 271, 116, 147, 23, 23,
+ 768, 768, 737, 768, 644, 675, 613, 644, 520, 551, 489, 520, 396, 427, 365,
+ 396, 272, 303, 241, 272, 148, 179, 117, 148, 24, 55, 24, 24, 800, 800, 769,
+ 800, 738, 769, 676, 707, 645, 676, 614, 645, 552, 583, 521, 552, 490, 521,
+ 428, 459, 397, 428, 366, 397, 304, 335, 273, 304, 242, 273, 180, 211, 149,
+ 180, 118, 149, 56, 87, 25, 56, 25, 25, 832, 832, 801, 832, 770, 801, 739,
+ 770, 708, 739, 677, 708, 646, 677, 615, 646, 584, 615, 553, 584, 522, 553,
+ 491, 522, 460, 491, 429, 460, 398, 429, 367, 398, 336, 367, 305, 336, 274,
+ 305, 243, 274, 212, 243, 181, 212, 150, 181, 119, 150, 88, 119, 57, 88, 26,
+ 57, 26, 26, 833, 864, 802, 833, 771, 802, 709, 740, 678, 709, 647, 678, 585,
+ 616, 554, 585, 523, 554, 461, 492, 430, 461, 399, 430, 337, 368, 306, 337,
+ 275, 306, 213, 244, 182, 213, 151, 182, 89, 120, 58, 89, 27, 58, 834, 865,
+ 803, 834, 710, 741, 679, 710, 586, 617, 555, 586, 462, 493, 431, 462, 338,
+ 369, 307, 338, 214, 245, 183, 214, 90, 121, 59, 90, 835, 866, 711, 742, 587,
+ 618, 463, 494, 339, 370, 215, 246, 91, 122, 864, 864, 740, 771, 616, 647,
+ 492, 523, 368, 399, 244, 275, 120, 151, 27, 27, 896, 896, 865, 896, 772, 803,
+ 741, 772, 648, 679, 617, 648, 524, 555, 493, 524, 400, 431, 369, 400, 276,
+ 307, 245, 276, 152, 183, 121, 152, 28, 59, 28, 28, 928, 928, 897, 928, 866,
+ 897, 804, 835, 773, 804, 742, 773, 680, 711, 649, 680, 618, 649, 556, 587,
+ 525, 556, 494, 525, 432, 463, 401, 432, 370, 401, 308, 339, 277, 308, 246,
+ 277, 184, 215, 153, 184, 122, 153, 60, 91, 29, 60, 29, 29, 960, 960, 929,
+ 960, 898, 929, 867, 898, 836, 867, 805, 836, 774, 805, 743, 774, 712, 743,
+ 681, 712, 650, 681, 619, 650, 588, 619, 557, 588, 526, 557, 495, 526, 464,
+ 495, 433, 464, 402, 433, 371, 402, 340, 371, 309, 340, 278, 309, 247, 278,
+ 216, 247, 185, 216, 154, 185, 123, 154, 92, 123, 61, 92, 30, 61, 30, 30,
+ 961, 992, 930, 961, 899, 930, 837, 868, 806, 837, 775, 806, 713, 744, 682,
+ 713, 651, 682, 589, 620, 558, 589, 527, 558, 465, 496, 434, 465, 403, 434,
+ 341, 372, 310, 341, 279, 310, 217, 248, 186, 217, 155, 186, 93, 124, 62, 93,
+ 31, 62, 962, 993, 931, 962, 838, 869, 807, 838, 714, 745, 683, 714, 590, 621,
+ 559, 590, 466, 497, 435, 466, 342, 373, 311, 342, 218, 249, 187, 218, 94,
+ 125, 63, 94, 963, 994, 839, 870, 715, 746, 591, 622, 467, 498, 343, 374, 219,
+ 250, 95, 126, 868, 899, 744, 775, 620, 651, 496, 527, 372, 403, 248, 279,
+ 124, 155, 900, 931, 869, 900, 776, 807, 745, 776, 652, 683, 621, 652, 528,
+ 559, 497, 528, 404, 435, 373, 404, 280, 311, 249, 280, 156, 187, 125, 156,
+ 932, 963, 901, 932, 870, 901, 808, 839, 777, 808, 746, 777, 684, 715, 653,
+ 684, 622, 653, 560, 591, 529, 560, 498, 529, 436, 467, 405, 436, 374, 405,
+ 312, 343, 281, 312, 250, 281, 188, 219, 157, 188, 126, 157, 964, 995, 933,
+ 964, 902, 933, 871, 902, 840, 871, 809, 840, 778, 809, 747, 778, 716, 747,
+ 685, 716, 654, 685, 623, 654, 592, 623, 561, 592, 530, 561, 499, 530, 468,
+ 499, 437, 468, 406, 437, 375, 406, 344, 375, 313, 344, 282, 313, 251, 282,
+ 220, 251, 189, 220, 158, 189, 127, 158, 965, 996, 934, 965, 903, 934, 841,
+ 872, 810, 841, 779, 810, 717, 748, 686, 717, 655, 686, 593, 624, 562, 593,
+ 531, 562, 469, 500, 438, 469, 407, 438, 345, 376, 314, 345, 283, 314, 221,
+ 252, 190, 221, 159, 190, 966, 997, 935, 966, 842, 873, 811, 842, 718, 749,
+ 687, 718, 594, 625, 563, 594, 470, 501, 439, 470, 346, 377, 315, 346, 222,
+ 253, 191, 222, 967, 998, 843, 874, 719, 750, 595, 626, 471, 502, 347, 378,
+ 223, 254, 872, 903, 748, 779, 624, 655, 500, 531, 376, 407, 252, 283, 904,
+ 935, 873, 904, 780, 811, 749, 780, 656, 687, 625, 656, 532, 563, 501, 532,
+ 408, 439, 377, 408, 284, 315, 253, 284, 936, 967, 905, 936, 874, 905, 812,
+ 843, 781, 812, 750, 781, 688, 719, 657, 688, 626, 657, 564, 595, 533, 564,
+ 502, 533, 440, 471, 409, 440, 378, 409, 316, 347, 285, 316, 254, 285, 968,
+ 999, 937, 968, 906, 937, 875, 906, 844, 875, 813, 844, 782, 813, 751, 782,
+ 720, 751, 689, 720, 658, 689, 627, 658, 596, 627, 565, 596, 534, 565, 503,
+ 534, 472, 503, 441, 472, 410, 441, 379, 410, 348, 379, 317, 348, 286, 317,
+ 255, 286, 969, 1000, 938, 969, 907, 938, 845, 876, 814, 845, 783, 814, 721,
+ 752, 690, 721, 659, 690, 597, 628, 566, 597, 535, 566, 473, 504, 442, 473,
+ 411, 442, 349, 380, 318, 349, 287, 318, 970, 1001, 939, 970, 846, 877, 815,
+ 846, 722, 753, 691, 722, 598, 629, 567, 598, 474, 505, 443, 474, 350, 381,
+ 319, 350, 971, 1002, 847, 878, 723, 754, 599, 630, 475, 506, 351, 382, 876,
+ 907, 752, 783, 628, 659, 504, 535, 380, 411, 908, 939, 877, 908, 784, 815,
+ 753, 784, 660, 691, 629, 660, 536, 567, 505, 536, 412, 443, 381, 412, 940,
+ 971, 909, 940, 878, 909, 816, 847, 785, 816, 754, 785, 692, 723, 661, 692,
+ 630, 661, 568, 599, 537, 568, 506, 537, 444, 475, 413, 444, 382, 413, 972,
+ 1003, 941, 972, 910, 941, 879, 910, 848, 879, 817, 848, 786, 817, 755, 786,
+ 724, 755, 693, 724, 662, 693, 631, 662, 600, 631, 569, 600, 538, 569, 507,
+ 538, 476, 507, 445, 476, 414, 445, 383, 414, 973, 1004, 942, 973, 911, 942,
+ 849, 880, 818, 849, 787, 818, 725, 756, 694, 725, 663, 694, 601, 632, 570,
+ 601, 539, 570, 477, 508, 446, 477, 415, 446, 974, 1005, 943, 974, 850, 881,
+ 819, 850, 726, 757, 695, 726, 602, 633, 571, 602, 478, 509, 447, 478, 975,
+ 1006, 851, 882, 727, 758, 603, 634, 479, 510, 880, 911, 756, 787, 632, 663,
+ 508, 539, 912, 943, 881, 912, 788, 819, 757, 788, 664, 695, 633, 664, 540,
+ 571, 509, 540, 944, 975, 913, 944, 882, 913, 820, 851, 789, 820, 758, 789,
+ 696, 727, 665, 696, 634, 665, 572, 603, 541, 572, 510, 541, 976, 1007, 945,
+ 976, 914, 945, 883, 914, 852, 883, 821, 852, 790, 821, 759, 790, 728, 759,
+ 697, 728, 666, 697, 635, 666, 604, 635, 573, 604, 542, 573, 511, 542, 977,
+ 1008, 946, 977, 915, 946, 853, 884, 822, 853, 791, 822, 729, 760, 698, 729,
+ 667, 698, 605, 636, 574, 605, 543, 574, 978, 1009, 947, 978, 854, 885, 823,
+ 854, 730, 761, 699, 730, 606, 637, 575, 606, 979, 1010, 855, 886, 731, 762,
+ 607, 638, 884, 915, 760, 791, 636, 667, 916, 947, 885, 916, 792, 823, 761,
+ 792, 668, 699, 637, 668, 948, 979, 917, 948, 886, 917, 824, 855, 793, 824,
+ 762, 793, 700, 731, 669, 700, 638, 669, 980, 1011, 949, 980, 918, 949, 887,
+ 918, 856, 887, 825, 856, 794, 825, 763, 794, 732, 763, 701, 732, 670, 701,
+ 639, 670, 981, 1012, 950, 981, 919, 950, 857, 888, 826, 857, 795, 826, 733,
+ 764, 702, 733, 671, 702, 982, 1013, 951, 982, 858, 889, 827, 858, 734, 765,
+ 703, 734, 983, 1014, 859, 890, 735, 766, 888, 919, 764, 795, 920, 951, 889,
+ 920, 796, 827, 765, 796, 952, 983, 921, 952, 890, 921, 828, 859, 797, 828,
+ 766, 797, 984, 1015, 953, 984, 922, 953, 891, 922, 860, 891, 829, 860, 798,
+ 829, 767, 798, 985, 1016, 954, 985, 923, 954, 861, 892, 830, 861, 799, 830,
+ 986, 1017, 955, 986, 862, 893, 831, 862, 987, 1018, 863, 894, 892, 923, 924,
+ 955, 893, 924, 956, 987, 925, 956, 894, 925, 988, 1019, 957, 988, 926, 957,
+ 895, 926, 989, 1020, 958, 989, 927, 958, 990, 1021, 959, 990, 991, 1022, 0, 0,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_default_iscan_4x4[16]) = {
+ 0, 2, 5, 8, 1, 3, 9, 12, 4, 7, 11, 14, 6, 10, 13, 15,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_col_iscan_4x4[16]) = {
+ 0, 3, 7, 11, 1, 5, 9, 12, 2, 6, 10, 14, 4, 8, 13, 15,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_row_iscan_4x4[16]) = {
+ 0, 1, 3, 5, 2, 4, 6, 9, 7, 8, 11, 13, 10, 12, 14, 15,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_col_iscan_8x8[64]) = {
+ 0, 3, 8, 15, 22, 32, 40, 47, 1, 5, 11, 18, 26, 34, 44, 51,
+ 2, 7, 13, 20, 28, 38, 46, 54, 4, 10, 16, 24, 31, 41, 50, 56,
+ 6, 12, 21, 27, 35, 43, 52, 58, 9, 17, 25, 33, 39, 48, 55, 60,
+ 14, 23, 30, 37, 45, 53, 59, 62, 19, 29, 36, 42, 49, 57, 61, 63,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_row_iscan_8x8[64]) = {
+ 0, 1, 2, 5, 8, 12, 19, 24, 3, 4, 7, 10, 15, 20, 30, 39,
+ 6, 9, 13, 16, 21, 27, 37, 46, 11, 14, 17, 23, 28, 34, 44, 52,
+ 18, 22, 25, 31, 35, 41, 50, 57, 26, 29, 33, 38, 43, 49, 55, 59,
+ 32, 36, 42, 47, 51, 54, 60, 61, 40, 45, 48, 53, 56, 58, 62, 63,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_default_iscan_8x8[64]) = {
+ 0, 2, 5, 9, 14, 22, 31, 37, 1, 4, 8, 13, 19, 26, 38, 44,
+ 3, 6, 10, 17, 24, 30, 42, 49, 7, 11, 15, 21, 29, 36, 47, 53,
+ 12, 16, 20, 27, 34, 43, 52, 57, 18, 23, 28, 35, 41, 48, 56, 60,
+ 25, 32, 39, 45, 50, 55, 59, 62, 33, 40, 46, 51, 54, 58, 61, 63,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_col_iscan_16x16[256]) = {
+ 0, 4, 11, 20, 31, 43, 59, 75, 85, 109, 130, 150, 165, 181, 195, 198,
+ 1, 6, 14, 23, 34, 47, 64, 81, 95, 114, 135, 153, 171, 188, 201, 212,
+ 2, 8, 16, 25, 38, 52, 67, 83, 101, 116, 136, 157, 172, 190, 205, 216,
+ 3, 10, 18, 29, 41, 55, 71, 89, 103, 119, 141, 159, 176, 194, 208, 218,
+ 5, 12, 21, 32, 45, 58, 74, 93, 104, 123, 144, 164, 179, 196, 210, 223,
+ 7, 15, 26, 37, 49, 63, 78, 96, 112, 129, 146, 166, 182, 200, 215, 228,
+ 9, 19, 28, 39, 54, 69, 86, 102, 117, 132, 151, 170, 187, 206, 220, 230,
+ 13, 24, 35, 46, 60, 73, 91, 108, 122, 137, 154, 174, 189, 207, 224, 235,
+ 17, 30, 40, 53, 66, 82, 98, 115, 126, 142, 161, 180, 197, 213, 227, 237,
+ 22, 36, 48, 62, 76, 92, 105, 120, 133, 147, 167, 186, 203, 219, 232, 240,
+ 27, 44, 56, 70, 84, 99, 113, 127, 140, 156, 175, 193, 209, 226, 236, 244,
+ 33, 51, 68, 79, 94, 110, 125, 138, 149, 162, 184, 202, 217, 229, 241, 247,
+ 42, 61, 77, 90, 106, 121, 134, 148, 160, 173, 191, 211, 225, 238, 245, 251,
+ 50, 72, 87, 100, 118, 128, 145, 158, 168, 183, 204, 222, 233, 242, 249, 253,
+ 57, 80, 97, 111, 131, 143, 155, 169, 178, 192, 214, 231, 239, 246, 250, 254,
+ 65, 88, 107, 124, 139, 152, 163, 177, 185, 199, 221, 234, 243, 248, 252, 255,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_row_iscan_16x16[256]) = {
+ 0, 1, 2, 4, 6, 9, 12, 17, 22, 29, 36, 43, 54, 64, 76, 86,
+ 3, 5, 7, 11, 15, 19, 25, 32, 38, 48, 59, 68, 84, 99, 115, 130,
+ 8, 10, 13, 18, 23, 27, 33, 42, 51, 60, 72, 88, 103, 119, 142, 167,
+ 14, 16, 20, 26, 31, 37, 44, 53, 61, 73, 85, 100, 116, 135, 161, 185,
+ 21, 24, 30, 35, 40, 47, 55, 65, 74, 81, 94, 112, 133, 154, 179, 205,
+ 28, 34, 39, 45, 50, 58, 67, 77, 87, 96, 106, 121, 146, 169, 196, 212,
+ 41, 46, 49, 56, 63, 70, 79, 90, 98, 107, 122, 138, 159, 182, 207, 222,
+ 52, 57, 62, 69, 75, 83, 93, 102, 110, 120, 134, 150, 176, 195, 215, 226,
+ 66, 71, 78, 82, 91, 97, 108, 113, 127, 136, 148, 168, 188, 202, 221, 232,
+ 80, 89, 92, 101, 105, 114, 125, 131, 139, 151, 162, 177, 192, 208, 223, 234,
+ 95, 104, 109, 117, 123, 128, 143, 144, 155, 165, 175, 190, 206, 219, 233, 239,
+ 111, 118, 124, 129, 140, 147, 157, 164, 170, 181, 191, 203, 224, 230, 240,
+ 243, 126, 132, 137, 145, 153, 160, 174, 178, 184, 197, 204, 216, 231, 237,
+ 244, 246, 141, 149, 156, 166, 172, 180, 189, 199, 200, 210, 220, 228, 238,
+ 242, 249, 251, 152, 163, 171, 183, 186, 193, 201, 211, 214, 218, 227, 236,
+ 245, 247, 252, 253, 158, 173, 187, 194, 198, 209, 213, 217, 225, 229, 235,
+ 241, 248, 250, 254, 255,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_default_iscan_16x16[256]) = {
+ 0, 2, 5, 9, 17, 24, 36, 44, 55, 72, 88, 104, 128, 143, 166, 179,
+ 1, 4, 8, 13, 20, 30, 40, 54, 66, 79, 96, 113, 141, 154, 178, 196,
+ 3, 7, 11, 18, 25, 33, 46, 57, 71, 86, 101, 119, 148, 164, 186, 201,
+ 6, 12, 16, 23, 31, 39, 53, 64, 78, 92, 110, 127, 153, 169, 193, 208,
+ 10, 14, 19, 28, 37, 47, 58, 67, 84, 98, 114, 133, 161, 176, 198, 214,
+ 15, 21, 26, 34, 43, 52, 65, 77, 91, 106, 120, 140, 165, 185, 205, 221,
+ 22, 27, 32, 41, 48, 60, 73, 85, 99, 116, 130, 151, 175, 190, 211, 225,
+ 29, 35, 42, 49, 59, 69, 81, 95, 108, 125, 139, 155, 182, 197, 217, 229,
+ 38, 45, 51, 61, 68, 80, 93, 105, 118, 134, 150, 168, 191, 207, 223, 234,
+ 50, 56, 63, 74, 83, 94, 109, 117, 129, 147, 163, 177, 199, 213, 228, 238,
+ 62, 70, 76, 87, 97, 107, 122, 131, 145, 159, 172, 188, 210, 222, 235, 242,
+ 75, 82, 90, 102, 112, 124, 138, 146, 157, 173, 187, 202, 219, 230, 240, 245,
+ 89, 100, 111, 123, 132, 142, 156, 167, 180, 189, 203, 216, 231, 237, 246, 250,
+ 103, 115, 126, 136, 149, 162, 171, 183, 194, 204, 215, 224, 236, 241, 248,
+ 252, 121, 135, 144, 158, 170, 181, 192, 200, 209, 218, 227, 233, 243, 244,
+ 251, 254, 137, 152, 160, 174, 184, 195, 206, 212, 220, 226, 232, 239, 247,
+ 249, 253, 255,
+};
+
+DECLARE_ALIGNED(16, static const int16_t, vp9_default_iscan_32x32[1024]) = {
+ 0, 2, 5, 10, 17, 25, 38, 47, 62, 83, 101, 121, 145, 170, 193, 204,
+ 210, 219, 229, 233, 245, 257, 275, 299, 342, 356, 377, 405, 455, 471, 495,
+ 527, 1, 4, 8, 15, 22, 30, 45, 58, 74, 92, 112, 133, 158, 184, 203, 215, 222,
+ 228, 234, 237, 256, 274, 298, 317, 355, 376, 404, 426, 470, 494, 526, 551,
+ 3, 7, 12, 18, 28, 36, 52, 64, 82, 102, 118, 142, 164, 189, 208, 217, 224,
+ 231, 235, 238, 273, 297, 316, 329, 375, 403, 425, 440, 493, 525, 550, 567,
+ 6, 11, 16, 23, 31, 43, 60, 73, 90, 109, 126, 150, 173, 196, 211, 220, 226,
+ 232, 236, 239, 296, 315, 328, 335, 402, 424, 439, 447, 524, 549, 566, 575,
+ 9, 14, 19, 29, 37, 50, 65, 78, 95, 116, 134, 157, 179, 201, 214, 223, 244,
+ 255, 272, 295, 341, 354, 374, 401, 454, 469, 492, 523, 582, 596, 617, 645,
+ 13, 20, 26, 35, 44, 54, 72, 85, 105, 123, 140, 163, 182, 205, 216, 225,
+ 254, 271, 294, 314, 353, 373, 400, 423, 468, 491, 522, 548, 595, 616, 644,
+ 666, 21, 27, 33, 42, 53, 63, 80, 94, 113, 132, 151, 172, 190, 209, 218, 227,
+ 270, 293, 313, 327, 372, 399, 422, 438, 490, 521, 547, 565, 615, 643, 665,
+ 680, 24, 32, 39, 48, 57, 71, 88, 104, 120, 139, 159, 178, 197, 212, 221, 230,
+ 292, 312, 326, 334, 398, 421, 437, 446, 520, 546, 564, 574, 642, 664, 679,
+ 687, 34, 40, 46, 56, 68, 81, 96, 111, 130, 147, 167, 186, 243, 253, 269, 291,
+ 340, 352, 371, 397, 453, 467, 489, 519, 581, 594, 614, 641, 693, 705, 723,
+ 747, 41, 49, 55, 67, 77, 91, 107, 124, 138, 161, 177, 194, 252, 268, 290,
+ 311, 351, 370, 396, 420, 466, 488, 518, 545, 593, 613, 640, 663, 704, 722,
+ 746, 765, 51, 59, 66, 76, 89, 99, 119, 131, 149, 168, 181, 200, 267, 289,
+ 310, 325, 369, 395, 419, 436, 487, 517, 544, 563, 612, 639, 662, 678, 721,
+ 745, 764, 777, 61, 69, 75, 87, 100, 114, 129, 144, 162, 180, 191, 207, 288,
+ 309, 324, 333, 394, 418, 435, 445, 516, 543, 562, 573, 638, 661, 677, 686,
+ 744, 763, 776, 783, 70, 79, 86, 97, 108, 122, 137, 155, 242, 251, 266, 287,
+ 339, 350, 368, 393, 452, 465, 486, 515, 580, 592, 611, 637, 692, 703, 720,
+ 743, 788, 798, 813, 833, 84, 93, 103, 110, 125, 141, 154, 171, 250, 265, 286,
+ 308, 349, 367, 392, 417, 464, 485, 514, 542, 591, 610, 636, 660, 702, 719,
+ 742, 762, 797, 812, 832, 848, 98, 106, 115, 127, 143, 156, 169, 185, 264,
+ 285, 307, 323, 366, 391, 416, 434, 484, 513, 541, 561, 609, 635, 659, 676,
+ 718, 741, 761, 775, 811, 831, 847, 858, 117, 128, 136, 148, 160, 175, 188,
+ 198, 284, 306, 322, 332, 390, 415, 433, 444, 512, 540, 560, 572, 634, 658,
+ 675, 685, 740, 760, 774, 782, 830, 846, 857, 863, 135, 146, 152, 165, 241,
+ 249, 263, 283, 338, 348, 365, 389, 451, 463, 483, 511, 579, 590, 608, 633,
+ 691, 701, 717, 739, 787, 796, 810, 829, 867, 875, 887, 903, 153, 166, 174,
+ 183, 248, 262, 282, 305, 347, 364, 388, 414, 462, 482, 510, 539, 589, 607,
+ 632, 657, 700, 716, 738, 759, 795, 809, 828, 845, 874, 886, 902, 915, 176,
+ 187, 195, 202, 261, 281, 304, 321, 363, 387, 413, 432, 481, 509, 538, 559,
+ 606, 631, 656, 674, 715, 737, 758, 773, 808, 827, 844, 856, 885, 901, 914,
+ 923, 192, 199, 206, 213, 280, 303, 320, 331, 386, 412, 431, 443, 508, 537,
+ 558, 571, 630, 655, 673, 684, 736, 757, 772, 781, 826, 843, 855, 862, 900,
+ 913, 922, 927, 240, 247, 260, 279, 337, 346, 362, 385, 450, 461, 480, 507,
+ 578, 588, 605, 629, 690, 699, 714, 735, 786, 794, 807, 825, 866, 873, 884,
+ 899, 930, 936, 945, 957, 246, 259, 278, 302, 345, 361, 384, 411, 460, 479,
+ 506, 536, 587, 604, 628, 654, 698, 713, 734, 756, 793, 806, 824, 842, 872,
+ 883, 898, 912, 935, 944, 956, 966, 258, 277, 301, 319, 360, 383, 410, 430,
+ 478, 505, 535, 557, 603, 627, 653, 672, 712, 733, 755, 771, 805, 823, 841,
+ 854, 882, 897, 911, 921, 943, 955, 965, 972, 276, 300, 318, 330, 382, 409,
+ 429, 442, 504, 534, 556, 570, 626, 652, 671, 683, 732, 754, 770, 780, 822,
+ 840, 853, 861, 896, 910, 920, 926, 954, 964, 971, 975, 336, 344, 359, 381,
+ 449, 459, 477, 503, 577, 586, 602, 625, 689, 697, 711, 731, 785, 792, 804,
+ 821, 865, 871, 881, 895, 929, 934, 942, 953, 977, 981, 987, 995, 343, 358,
+ 380, 408, 458, 476, 502, 533, 585, 601, 624, 651, 696, 710, 730, 753, 791,
+ 803, 820, 839, 870, 880, 894, 909, 933, 941, 952, 963, 980, 986, 994, 1001,
+ 357, 379, 407, 428, 475, 501, 532, 555, 600, 623, 650, 670, 709, 729, 752,
+ 769, 802, 819, 838, 852, 879, 893, 908, 919, 940, 951, 962, 970, 985, 993,
+ 1000, 1005, 378, 406, 427, 441, 500, 531, 554, 569, 622, 649, 669, 682, 728,
+ 751, 768, 779, 818, 837, 851, 860, 892, 907, 918, 925, 950, 961, 969, 974,
+ 992, 999, 1004, 1007, 448, 457, 474, 499, 576, 584, 599, 621, 688, 695, 708,
+ 727, 784, 790, 801, 817, 864, 869, 878, 891, 928, 932, 939, 949, 976, 979,
+ 984, 991, 1008, 1010, 1013, 1017, 456, 473, 498, 530, 583, 598, 620, 648,
+ 694, 707, 726, 750, 789, 800, 816, 836, 868, 877, 890, 906, 931, 938, 948,
+ 960, 978, 983, 990, 998, 1009, 1012, 1016, 1020, 472, 497, 529, 553, 597,
+ 619, 647, 668, 706, 725, 749, 767, 799, 815, 835, 850, 876, 889, 905, 917,
+ 937, 947, 959, 968, 982, 989, 997, 1003, 1011, 1015, 1019, 1022, 496, 528,
+ 552, 568, 618, 646, 667, 681, 724, 748, 766, 778, 814, 834, 849, 859, 888,
+ 904, 916, 924, 946, 958, 967, 973, 988, 996, 1002, 1006, 1014, 1018, 1021,
+ 1023,
+};
const scan_order vp9_default_scan_orders[TX_SIZES] = {
{default_scan_4x4, vp9_default_iscan_4x4, default_scan_4x4_neighbors},
@@ -295,93 +725,3 @@
{default_scan_32x32, vp9_default_iscan_32x32, default_scan_32x32_neighbors},
}
};
-
-static int find_in_scan(const int16_t *scan, int l, int idx) {
- int n, l2 = l * l;
- for (n = 0; n < l2; n++) {
- int rc = scan[n];
- if (rc == idx)
- return n;
- }
- assert(0);
- return -1;
-}
-
-static void init_scan_neighbors(const int16_t *scan, int16_t *iscan, int l,
- int16_t *neighbors) {
- int l2 = l * l;
- int n, i, j;
-
- // dc doesn't use this type of prediction
- neighbors[MAX_NEIGHBORS * 0 + 0] = 0;
- neighbors[MAX_NEIGHBORS * 0 + 1] = 0;
- iscan[0] = find_in_scan(scan, l, 0);
- for (n = 1; n < l2; n++) {
- int rc = scan[n];
- iscan[n] = find_in_scan(scan, l, n);
- i = rc / l;
- j = rc % l;
- if (i > 0 && j > 0) {
- // col/row scan is used for adst/dct, and generally means that
- // energy decreases to zero much faster in the dimension in
- // which ADST is used compared to the direction in which DCT
- // is used. Likewise, we find much higher correlation between
- // coefficients within the direction in which DCT is used.
- // Therefore, if we use ADST/DCT, prefer the DCT neighbor coeff
- // as a context. If ADST or DCT is used in both directions, we
- // use the combination of the two as a context.
- int a = (i - 1) * l + j;
- int b = i * l + j - 1;
- if (scan == col_scan_4x4 || scan == col_scan_8x8 ||
- scan == col_scan_16x16) {
- // in the col/row scan cases (as well as left/top edge cases), we set
- // both contexts to the same value, so we can branchlessly do a+b+1>>1
- // which automatically becomes a if a == b
- neighbors[MAX_NEIGHBORS * n + 0] =
- neighbors[MAX_NEIGHBORS * n + 1] = a;
- } else if (scan == row_scan_4x4 || scan == row_scan_8x8 ||
- scan == row_scan_16x16) {
- neighbors[MAX_NEIGHBORS * n + 0] =
- neighbors[MAX_NEIGHBORS * n + 1] = b;
- } else {
- neighbors[MAX_NEIGHBORS * n + 0] = a;
- neighbors[MAX_NEIGHBORS * n + 1] = b;
- }
- } else if (i > 0) {
- neighbors[MAX_NEIGHBORS * n + 0] =
- neighbors[MAX_NEIGHBORS * n + 1] = (i - 1) * l + j;
- } else {
- assert(j > 0);
- neighbors[MAX_NEIGHBORS * n + 0] =
- neighbors[MAX_NEIGHBORS * n + 1] = i * l + j - 1;
- }
- assert(iscan[neighbors[MAX_NEIGHBORS * n + 0]] < n);
- }
- // one padding item so we don't have to add branches in code to handle
- // calls to get_coef_context() for the token after the final dc token
- neighbors[MAX_NEIGHBORS * l2 + 0] = 0;
- neighbors[MAX_NEIGHBORS * l2 + 1] = 0;
-}
-
-void vp9_init_neighbors() {
- init_scan_neighbors(default_scan_4x4, vp9_default_iscan_4x4, 4,
- default_scan_4x4_neighbors);
- init_scan_neighbors(row_scan_4x4, vp9_row_iscan_4x4, 4,
- row_scan_4x4_neighbors);
- init_scan_neighbors(col_scan_4x4, vp9_col_iscan_4x4, 4,
- col_scan_4x4_neighbors);
- init_scan_neighbors(default_scan_8x8, vp9_default_iscan_8x8, 8,
- default_scan_8x8_neighbors);
- init_scan_neighbors(row_scan_8x8, vp9_row_iscan_8x8, 8,
- row_scan_8x8_neighbors);
- init_scan_neighbors(col_scan_8x8, vp9_col_iscan_8x8, 8,
- col_scan_8x8_neighbors);
- init_scan_neighbors(default_scan_16x16, vp9_default_iscan_16x16, 16,
- default_scan_16x16_neighbors);
- init_scan_neighbors(row_scan_16x16, vp9_row_iscan_16x16, 16,
- row_scan_16x16_neighbors);
- init_scan_neighbors(col_scan_16x16, vp9_col_iscan_16x16, 16,
- col_scan_16x16_neighbors);
- init_scan_neighbors(default_scan_32x32, vp9_default_iscan_32x32, 32,
- default_scan_32x32_neighbors);
-}
diff --git a/source/libvpx/vp9/common/vp9_scan.h b/source/libvpx/vp9/common/vp9_scan.h
index 9613b67..65e2aa6 100644
--- a/source/libvpx/vp9/common/vp9_scan.h
+++ b/source/libvpx/vp9/common/vp9_scan.h
@@ -23,8 +23,6 @@
#define MAX_NEIGHBORS 2
-void vp9_init_neighbors();
-
typedef struct {
const int16_t *scan;
const int16_t *iscan;
diff --git a/source/libvpx/vp9/common/x86/vp9_asm_stubs.c b/source/libvpx/vp9/common/x86/vp9_asm_stubs.c
index 407573a..a0a5996 100644
--- a/source/libvpx/vp9/common/x86/vp9_asm_stubs.c
+++ b/source/libvpx/vp9/common/x86/vp9_asm_stubs.c
@@ -142,7 +142,7 @@
#if CONFIG_VP9_HIGHBITDEPTH
-typedef void high_filter8_1dfunction (
+typedef void highbd_filter8_1dfunction (
const uint16_t *src_ptr,
const ptrdiff_t src_pitch,
uint16_t *output_ptr,
@@ -153,87 +153,88 @@
);
#define HIGH_FUN_CONV_1D(name, step_q4, filter, dir, src_start, avg, opt) \
- void vp9_high_convolve8_##name##_##opt(const uint8_t *src8, \
- ptrdiff_t src_stride, \
- uint8_t *dst8, ptrdiff_t dst_stride, \
- const int16_t *filter_x, \
- int x_step_q4, \
- const int16_t *filter_y, \
- int y_step_q4, \
- int w, int h, int bd) { \
+ void vp9_highbd_convolve8_##name##_##opt(const uint8_t *src8, \
+ ptrdiff_t src_stride, \
+ uint8_t *dst8, \
+ ptrdiff_t dst_stride, \
+ const int16_t *filter_x, \
+ int x_step_q4, \
+ const int16_t *filter_y, \
+ int y_step_q4, \
+ int w, int h, int bd) { \
if (step_q4 == 16 && filter[3] != 128) { \
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
if (filter[0] || filter[1] || filter[2]) { \
while (w >= 16) { \
- vp9_high_filter_block1d16_##dir##8_##avg##opt(src_start, \
- src_stride, \
- dst, \
- dst_stride, \
- h, \
- filter, \
- bd); \
+ vp9_highbd_filter_block1d16_##dir##8_##avg##opt(src_start, \
+ src_stride, \
+ dst, \
+ dst_stride, \
+ h, \
+ filter, \
+ bd); \
src += 16; \
dst += 16; \
w -= 16; \
} \
while (w >= 8) { \
- vp9_high_filter_block1d8_##dir##8_##avg##opt(src_start, \
- src_stride, \
- dst, \
- dst_stride, \
- h, \
- filter, \
- bd); \
+ vp9_highbd_filter_block1d8_##dir##8_##avg##opt(src_start, \
+ src_stride, \
+ dst, \
+ dst_stride, \
+ h, \
+ filter, \
+ bd); \
src += 8; \
dst += 8; \
w -= 8; \
} \
while (w >= 4) { \
- vp9_high_filter_block1d4_##dir##8_##avg##opt(src_start, \
- src_stride, \
- dst, \
- dst_stride, \
- h, \
- filter, \
- bd); \
+ vp9_highbd_filter_block1d4_##dir##8_##avg##opt(src_start, \
+ src_stride, \
+ dst, \
+ dst_stride, \
+ h, \
+ filter, \
+ bd); \
src += 4; \
dst += 4; \
w -= 4; \
} \
} else { \
while (w >= 16) { \
- vp9_high_filter_block1d16_##dir##2_##avg##opt(src, \
- src_stride, \
- dst, \
- dst_stride, \
- h, \
- filter, \
- bd); \
+ vp9_highbd_filter_block1d16_##dir##2_##avg##opt(src, \
+ src_stride, \
+ dst, \
+ dst_stride, \
+ h, \
+ filter, \
+ bd); \
src += 16; \
dst += 16; \
w -= 16; \
} \
while (w >= 8) { \
- vp9_high_filter_block1d8_##dir##2_##avg##opt(src, \
- src_stride, \
- dst, \
- dst_stride, \
- h, \
- filter, \
- bd); \
+ vp9_highbd_filter_block1d8_##dir##2_##avg##opt(src, \
+ src_stride, \
+ dst, \
+ dst_stride, \
+ h, \
+ filter, \
+ bd); \
src += 8; \
dst += 8; \
w -= 8; \
} \
while (w >= 4) { \
- vp9_high_filter_block1d4_##dir##2_##avg##opt(src, \
- src_stride, \
- dst, \
- dst_stride, \
- h, \
- filter, \
- bd); \
+ vp9_highbd_filter_block1d4_##dir##2_##avg##opt(src, \
+ src_stride, \
+ dst, \
+ dst_stride, \
+ h, \
+ filter, \
+ bd); \
src += 4; \
dst += 4; \
w -= 4; \
@@ -241,47 +242,51 @@
} \
} \
if (w) { \
- vp9_high_convolve8_##name##_c(src8, src_stride, dst8, dst_stride, \
- filter_x, x_step_q4, filter_y, y_step_q4, \
- w, h, bd); \
+ vp9_highbd_convolve8_##name##_c(src8, src_stride, dst8, dst_stride, \
+ filter_x, x_step_q4, filter_y, y_step_q4, \
+ w, h, bd); \
} \
}
#define HIGH_FUN_CONV_2D(avg, opt) \
-void vp9_high_convolve8_##avg##opt(const uint8_t *src, ptrdiff_t src_stride, \
- uint8_t *dst, ptrdiff_t dst_stride, \
- const int16_t *filter_x, int x_step_q4, \
- const int16_t *filter_y, int y_step_q4, \
- int w, int h, int bd) { \
+void vp9_highbd_convolve8_##avg##opt(const uint8_t *src, ptrdiff_t src_stride, \
+ uint8_t *dst, ptrdiff_t dst_stride, \
+ const int16_t *filter_x, int x_step_q4, \
+ const int16_t *filter_y, int y_step_q4, \
+ int w, int h, int bd) { \
assert(w <= 64); \
assert(h <= 64); \
if (x_step_q4 == 16 && y_step_q4 == 16) { \
if (filter_x[0] || filter_x[1] || filter_x[2] || filter_x[3] == 128 || \
filter_y[0] || filter_y[1] || filter_y[2] || filter_y[3] == 128) { \
DECLARE_ALIGNED_ARRAY(16, uint16_t, fdata2, 64 * 71); \
- vp9_high_convolve8_horiz_##opt(src - 3 * src_stride, src_stride, \
- CONVERT_TO_BYTEPTR(fdata2), 64, \
- filter_x, x_step_q4, filter_y, y_step_q4, \
- w, h + 7, bd); \
- vp9_high_convolve8_##avg##vert_##opt(CONVERT_TO_BYTEPTR(fdata2) + 192, \
- 64, dst, dst_stride, \
- filter_x, x_step_q4, filter_y, \
- y_step_q4, w, h, bd); \
+ vp9_highbd_convolve8_horiz_##opt(src - 3 * src_stride, src_stride, \
+ CONVERT_TO_BYTEPTR(fdata2), 64, \
+ filter_x, x_step_q4, \
+ filter_y, y_step_q4, \
+ w, h + 7, bd); \
+ vp9_highbd_convolve8_##avg##vert_##opt(CONVERT_TO_BYTEPTR(fdata2) + 192, \
+ 64, dst, dst_stride, \
+ filter_x, x_step_q4, \
+ filter_y, y_step_q4, \
+ w, h, bd); \
} else { \
DECLARE_ALIGNED_ARRAY(16, uint16_t, fdata2, 64 * 65); \
- vp9_high_convolve8_horiz_##opt(src, src_stride, \
- CONVERT_TO_BYTEPTR(fdata2), 64, \
- filter_x, x_step_q4, filter_y, y_step_q4, \
- w, h + 1, bd); \
- vp9_high_convolve8_##avg##vert_##opt(CONVERT_TO_BYTEPTR(fdata2), 64, \
- dst, dst_stride, \
- filter_x, x_step_q4, filter_y, \
- y_step_q4, w, h, bd); \
+ vp9_highbd_convolve8_horiz_##opt(src, src_stride, \
+ CONVERT_TO_BYTEPTR(fdata2), 64, \
+ filter_x, x_step_q4, \
+ filter_y, y_step_q4, \
+ w, h + 1, bd); \
+ vp9_highbd_convolve8_##avg##vert_##opt(CONVERT_TO_BYTEPTR(fdata2), 64, \
+ dst, dst_stride, \
+ filter_x, x_step_q4, \
+ filter_y, y_step_q4, \
+ w, h, bd); \
} \
} else { \
- vp9_high_convolve8_##avg##c(src, src_stride, dst, dst_stride, \
- filter_x, x_step_q4, filter_y, y_step_q4, w, \
- h, bd); \
+ vp9_highbd_convolve8_##avg##c(src, src_stride, dst, dst_stride, \
+ filter_x, x_step_q4, filter_y, y_step_q4, w, \
+ h, bd); \
} \
}
#endif // CONFIG_VP9_HIGHBITDEPTH
@@ -485,72 +490,84 @@
FUN_CONV_2D(avg_ , sse2);
#if CONFIG_VP9_HIGHBITDEPTH && ARCH_X86_64
-high_filter8_1dfunction vp9_high_filter_block1d16_v8_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d16_h8_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d8_v8_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d8_h8_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d4_v8_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d4_h8_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d16_v8_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d16_h8_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d8_v8_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d8_h8_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d4_v8_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d4_h8_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d16_v8_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d16_h8_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d8_v8_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d8_h8_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d4_v8_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d4_h8_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d16_v8_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d16_h8_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d8_v8_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d8_h8_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d4_v8_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d4_h8_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d16_v2_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d16_h2_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d8_v2_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d8_h2_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d4_v2_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d4_h2_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d16_v2_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d16_h2_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d8_v2_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d8_h2_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d4_v2_avg_sse2;
-high_filter8_1dfunction vp9_high_filter_block1d4_h2_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d16_v2_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d16_h2_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d8_v2_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d8_h2_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d4_v2_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d4_h2_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d16_v2_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d16_h2_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d8_v2_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d8_h2_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d4_v2_avg_sse2;
+highbd_filter8_1dfunction vp9_highbd_filter_block1d4_h2_avg_sse2;
-// void vp9_high_convolve8_horiz_sse2(const uint8_t *src, ptrdiff_t src_stride,
-// uint8_t *dst, ptrdiff_t dst_stride,
-// const int16_t *filter_x, int x_step_q4,
-// const int16_t *filter_y, int y_step_q4,
-// int w, int h, int bd);
-// void vp9_high_convolve8_vert_sse2(const uint8_t *src, ptrdiff_t src_stride,
-// uint8_t *dst, ptrdiff_t dst_stride,
-// const int16_t *filter_x, int x_step_q4,
-// const int16_t *filter_y, int y_step_q4,
-// int w, int h, int bd);
-// void vp9_high_convolve8_avg_horiz_sse2(const uint8_t *src,
-// ptrdiff_t src_stride,
-// uint8_t *dst, ptrdiff_t dst_stride,
-// const int16_t *filter_x,
-// int x_step_q4,
-// const int16_t *filter_y,
-// int y_step_q4,
-// int w, int h, int bd);
-// void vp9_high_convolve8_avg_vert_sse2(const uint8_t *src,
-// ptrdiff_t src_stride,
-// uint8_t *dst, ptrdiff_t dst_stride,
-// const int16_t *filter_x, int x_step_q4,
-// const int16_t *filter_y, int y_step_q4,
-// int w, int h, int bd);
+// void vp9_highbd_convolve8_horiz_sse2(const uint8_t *src,
+// ptrdiff_t src_stride,
+// uint8_t *dst,
+// ptrdiff_t dst_stride,
+// const int16_t *filter_x,
+// int x_step_q4,
+// const int16_t *filter_y,
+// int y_step_q4,
+// int w, int h, int bd);
+// void vp9_highbd_convolve8_vert_sse2(const uint8_t *src,
+// ptrdiff_t src_stride,
+// uint8_t *dst,
+// ptrdiff_t dst_stride,
+// const int16_t *filter_x,
+// int x_step_q4,
+// const int16_t *filter_y,
+// int y_step_q4,
+// int w, int h, int bd);
+// void vp9_highbd_convolve8_avg_horiz_sse2(const uint8_t *src,
+// ptrdiff_t src_stride,
+// uint8_t *dst,
+// ptrdiff_t dst_stride,
+// const int16_t *filter_x,
+// int x_step_q4,
+// const int16_t *filter_y,
+// int y_step_q4,
+// int w, int h, int bd);
+// void vp9_highbd_convolve8_avg_vert_sse2(const uint8_t *src,
+// ptrdiff_t src_stride,
+// uint8_t *dst,
+// ptrdiff_t dst_stride,
+// const int16_t *filter_x,
+// int x_step_q4,
+// const int16_t *filter_y,
+// int y_step_q4,
+// int w, int h, int bd);
HIGH_FUN_CONV_1D(horiz, x_step_q4, filter_x, h, src, , sse2);
HIGH_FUN_CONV_1D(vert, y_step_q4, filter_y, v, src - src_stride * 3, , sse2);
HIGH_FUN_CONV_1D(avg_horiz, x_step_q4, filter_x, h, src, avg_, sse2);
HIGH_FUN_CONV_1D(avg_vert, y_step_q4, filter_y, v, src - src_stride * 3, avg_,
sse2);
-// void vp9_high_convolve8_sse2(const uint8_t *src, ptrdiff_t src_stride,
-// uint8_t *dst, ptrdiff_t dst_stride,
-// const int16_t *filter_x, int x_step_q4,
-// const int16_t *filter_y, int y_step_q4,
-// int w, int h, int bd);
-// void vp9_high_convolve8_avg_sse2(const uint8_t *src, ptrdiff_t src_stride,
-// uint8_t *dst, ptrdiff_t dst_stride,
-// const int16_t *filter_x, int x_step_q4,
-// const int16_t *filter_y, int y_step_q4,
-// int w, int h, int bd);
+// void vp9_highbd_convolve8_sse2(const uint8_t *src, ptrdiff_t src_stride,
+// uint8_t *dst, ptrdiff_t dst_stride,
+// const int16_t *filter_x, int x_step_q4,
+// const int16_t *filter_y, int y_step_q4,
+// int w, int h, int bd);
+// void vp9_highbd_convolve8_avg_sse2(const uint8_t *src, ptrdiff_t src_stride,
+// uint8_t *dst, ptrdiff_t dst_stride,
+// const int16_t *filter_x, int x_step_q4,
+// const int16_t *filter_y, int y_step_q4,
+// int w, int h, int bd);
HIGH_FUN_CONV_2D(, sse2);
HIGH_FUN_CONV_2D(avg_ , sse2);
#endif // CONFIG_VP9_HIGHBITDEPTH && ARCH_X86_64
diff --git a/source/libvpx/vp9/common/x86/vp9_high_intrapred_sse2.asm b/source/libvpx/vp9/common/x86/vp9_high_intrapred_sse2.asm
index ff45071..721126c 100644
--- a/source/libvpx/vp9/common/x86/vp9_high_intrapred_sse2.asm
+++ b/source/libvpx/vp9/common/x86/vp9_high_intrapred_sse2.asm
@@ -18,7 +18,7 @@
SECTION .text
INIT_MMX sse
-cglobal high_dc_predictor_4x4, 4, 5, 4, dst, stride, above, left, goffset
+cglobal highbd_dc_predictor_4x4, 4, 5, 4, dst, stride, above, left, goffset
GET_GOT goffsetq
movq m0, [aboveq]
@@ -45,7 +45,7 @@
RET
INIT_XMM sse2
-cglobal high_dc_predictor_8x8, 4, 5, 4, dst, stride, above, left, goffset
+cglobal highbd_dc_predictor_8x8, 4, 5, 4, dst, stride, above, left, goffset
GET_GOT goffsetq
pxor m1, m1
@@ -80,7 +80,7 @@
RET
INIT_XMM sse2
-cglobal high_dc_predictor_16x16, 4, 5, 5, dst, stride, above, left, goffset
+cglobal highbd_dc_predictor_16x16, 4, 5, 5, dst, stride, above, left, goffset
GET_GOT goffsetq
pxor m1, m1
@@ -124,7 +124,7 @@
%if ARCH_X86_64
INIT_XMM sse2
-cglobal high_dc_predictor_32x32, 4, 5, 9, dst, stride, above, left, goffset
+cglobal highbd_dc_predictor_32x32, 4, 5, 9, dst, stride, above, left, goffset
GET_GOT goffsetq
pxor m1, m1
@@ -184,7 +184,7 @@
%endif
INIT_MMX sse
-cglobal high_v_predictor_4x4, 3, 3, 1, dst, stride, above
+cglobal highbd_v_predictor_4x4, 3, 3, 1, dst, stride, above
movq m0, [aboveq]
movq [dstq ], m0
movq [dstq+strideq*2], m0
@@ -194,7 +194,7 @@
RET
INIT_XMM sse2
-cglobal high_v_predictor_8x8, 3, 3, 1, dst, stride, above
+cglobal highbd_v_predictor_8x8, 3, 3, 1, dst, stride, above
mova m0, [aboveq]
DEFINE_ARGS dst, stride, stride3
lea stride3q, [strideq*3]
@@ -210,7 +210,7 @@
RET
INIT_XMM sse2
-cglobal high_v_predictor_16x16, 3, 4, 2, dst, stride, above
+cglobal highbd_v_predictor_16x16, 3, 4, 2, dst, stride, above
mova m0, [aboveq]
mova m1, [aboveq+16]
DEFINE_ARGS dst, stride, stride3, nlines4
@@ -231,7 +231,7 @@
REP_RET
INIT_XMM sse2
-cglobal high_v_predictor_32x32, 3, 4, 4, dst, stride, above
+cglobal highbd_v_predictor_32x32, 3, 4, 4, dst, stride, above
mova m0, [aboveq]
mova m1, [aboveq+16]
mova m2, [aboveq+32]
@@ -262,7 +262,7 @@
REP_RET
INIT_MMX sse
-cglobal high_tm_predictor_4x4, 5, 6, 5, dst, stride, above, left, bps, one
+cglobal highbd_tm_predictor_4x4, 5, 6, 5, dst, stride, above, left, bps, one
movd m1, [aboveq-2]
movq m0, [aboveq]
pshufw m1, m1, 0x0
@@ -300,7 +300,7 @@
REP_RET
INIT_XMM sse2
-cglobal high_tm_predictor_8x8, 5, 6, 5, dst, stride, above, left, bps, one
+cglobal highbd_tm_predictor_8x8, 5, 6, 5, dst, stride, above, left, bps, one
movd m1, [aboveq-2]
mova m0, [aboveq]
pshuflw m1, m1, 0x0
@@ -345,7 +345,7 @@
%if ARCH_X86_64
INIT_XMM sse2
-cglobal high_tm_predictor_16x16, 5, 6, 8, dst, stride, above, left, bps, one
+cglobal highbd_tm_predictor_16x16, 5, 6, 8, dst, stride, above, left, bps, one
movd m2, [aboveq-2]
mova m0, [aboveq]
mova m1, [aboveq+16]
@@ -399,7 +399,7 @@
REP_RET
INIT_XMM sse2
-cglobal high_tm_predictor_32x32, 5, 6, 12, dst, stride, above, left, bps, one
+cglobal highbd_tm_predictor_32x32, 5, 6, 12, dst, stride, above, left, bps, one
movd m0, [aboveq-2]
mova m1, [aboveq]
mova m2, [aboveq+16]
diff --git a/source/libvpx/vp9/common/x86/vp9_high_loopfilter_intrin_sse2.c b/source/libvpx/vp9/common/x86/vp9_high_loopfilter_intrin_sse2.c
new file mode 100644
index 0000000..7e63f38
--- /dev/null
+++ b/source/libvpx/vp9/common/x86/vp9_high_loopfilter_intrin_sse2.c
@@ -0,0 +1,1119 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 <emmintrin.h> // SSE2
+
+#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_loopfilter.h"
+#include "vpx_ports/emmintrin_compat.h"
+
+static INLINE __m128i signed_char_clamp_bd_sse2(__m128i value, int bd) {
+ __m128i ubounded;
+ __m128i lbounded;
+ __m128i retval;
+
+ const __m128i zero = _mm_set1_epi16(0);
+ const __m128i one = _mm_set1_epi16(1);
+ const __m128i t80 = _mm_slli_epi16(_mm_set1_epi16(0x80), bd - 8);
+ const __m128i max = _mm_subs_epi16(
+ _mm_subs_epi16(_mm_slli_epi16(one, bd), one), t80);
+ const __m128i min = _mm_subs_epi16(zero, t80);
+ ubounded = _mm_cmpgt_epi16(value, max);
+ lbounded = _mm_cmplt_epi16(value, min);
+ retval = _mm_andnot_si128(_mm_or_si128(ubounded, lbounded), value);
+ ubounded = _mm_and_si128(ubounded, max);
+ lbounded = _mm_and_si128(lbounded, min);
+ retval = _mm_or_si128(retval, ubounded);
+ retval = _mm_or_si128(retval, lbounded);
+ return retval;
+}
+
+// TODO(debargha, peter): Break up large functions into smaller ones
+// in this file.
+static void highbd_mb_lpf_horizontal_edge_w_sse2_8(uint16_t *s,
+ int p,
+ const uint8_t *_blimit,
+ const uint8_t *_limit,
+ const uint8_t *_thresh,
+ int bd) {
+ const __m128i zero = _mm_set1_epi16(0);
+ const __m128i one = _mm_set1_epi16(1);
+ const __m128i blimit = _mm_slli_epi16(
+ _mm_unpacklo_epi8(
+ _mm_load_si128((const __m128i *)_blimit), zero), bd - 8);
+ const __m128i limit = _mm_slli_epi16(
+ _mm_unpacklo_epi8(_mm_load_si128((const __m128i *)_limit), zero), bd - 8);
+ const __m128i thresh = _mm_slli_epi16(
+ _mm_unpacklo_epi8(
+ _mm_load_si128((const __m128i *)_thresh), zero), bd - 8);
+ __m128i q7, p7, q6, p6, q5, p5, q4, p4, q3, p3, q2, p2, q1, p1, q0, p0;
+ __m128i mask, hev, flat, flat2, abs_p1p0, abs_q1q0;
+ __m128i ps1, qs1, ps0, qs0;
+ __m128i abs_p0q0, abs_p1q1, ffff, work;
+ __m128i filt, work_a, filter1, filter2;
+ __m128i flat2_q6, flat2_p6, flat2_q5, flat2_p5, flat2_q4, flat2_p4;
+ __m128i flat2_q3, flat2_p3, flat2_q2, flat2_p2, flat2_q1, flat2_p1;
+ __m128i flat2_q0, flat2_p0;
+ __m128i flat_q2, flat_p2, flat_q1, flat_p1, flat_q0, flat_p0;
+ __m128i pixelFilter_p, pixelFilter_q;
+ __m128i pixetFilter_p2p1p0, pixetFilter_q2q1q0;
+ __m128i sum_p7, sum_q7, sum_p3, sum_q3;
+ __m128i t4, t3, t80, t1;
+ __m128i eight, four;
+
+ q4 = _mm_load_si128((__m128i *)(s + 4 * p));
+ p4 = _mm_load_si128((__m128i *)(s - 5 * p));
+ q3 = _mm_load_si128((__m128i *)(s + 3 * p));
+ p3 = _mm_load_si128((__m128i *)(s - 4 * p));
+ q2 = _mm_load_si128((__m128i *)(s + 2 * p));
+ p2 = _mm_load_si128((__m128i *)(s - 3 * p));
+ q1 = _mm_load_si128((__m128i *)(s + 1 * p));
+ p1 = _mm_load_si128((__m128i *)(s - 2 * p));
+ q0 = _mm_load_si128((__m128i *)(s + 0 * p));
+ p0 = _mm_load_si128((__m128i *)(s - 1 * p));
+
+ // highbd_filter_mask
+ abs_p1p0 = _mm_or_si128(_mm_subs_epu16(p1, p0), _mm_subs_epu16(p0, p1));
+ abs_q1q0 = _mm_or_si128(_mm_subs_epu16(q1, q0), _mm_subs_epu16(q0, q1));
+
+ ffff = _mm_cmpeq_epi16(abs_p1p0, abs_p1p0);
+
+ abs_p0q0 = _mm_or_si128(_mm_subs_epu16(p0, q0), _mm_subs_epu16(q0, p0));
+ abs_p1q1 = _mm_or_si128(_mm_subs_epu16(p1, q1), _mm_subs_epu16(q1, p1));
+
+ // highbd_hev_mask (in C code this is actually called from highbd_filter4)
+ flat = _mm_max_epi16(abs_p1p0, abs_q1q0);
+ hev = _mm_subs_epu16(flat, thresh);
+ hev = _mm_xor_si128(_mm_cmpeq_epi16(hev, zero), ffff);
+
+ abs_p0q0 =_mm_adds_epu16(abs_p0q0, abs_p0q0); // abs(p0 - q0) * 2
+ abs_p1q1 = _mm_srli_epi16(abs_p1q1, 1); // abs(p1 - q1) / 2
+ mask = _mm_subs_epu16(_mm_adds_epu16(abs_p0q0, abs_p1q1), blimit);
+ mask = _mm_xor_si128(_mm_cmpeq_epi16(mask, zero), ffff);
+ mask = _mm_and_si128(mask, _mm_adds_epu16(limit, one));
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p1, p0),
+ _mm_subs_epu16(p0, p1)),
+ _mm_or_si128(_mm_subs_epu16(q1, q0),
+ _mm_subs_epu16(q0, q1)));
+ mask = _mm_max_epi16(work, mask);
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p2, p1),
+ _mm_subs_epu16(p1, p2)),
+ _mm_or_si128(_mm_subs_epu16(q2, q1),
+ _mm_subs_epu16(q1, q2)));
+ mask = _mm_max_epi16(work, mask);
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p3, p2),
+ _mm_subs_epu16(p2, p3)),
+ _mm_or_si128(_mm_subs_epu16(q3, q2),
+ _mm_subs_epu16(q2, q3)));
+ mask = _mm_max_epi16(work, mask);
+
+ mask = _mm_subs_epu16(mask, limit);
+ mask = _mm_cmpeq_epi16(mask, zero); // return ~mask
+
+ // lp filter
+ // highbd_filter4
+ t4 = _mm_set1_epi16(4);
+ t3 = _mm_set1_epi16(3);
+ t80 = _mm_slli_epi16(_mm_set1_epi16(0x80), bd - 8);
+ t1 = _mm_set1_epi16(0x1);
+
+ ps1 = _mm_subs_epi16(p1, t80);
+ qs1 = _mm_subs_epi16(q1, t80);
+ ps0 = _mm_subs_epi16(p0, t80);
+ qs0 = _mm_subs_epi16(q0, t80);
+
+ filt = _mm_and_si128(
+ signed_char_clamp_bd_sse2(_mm_subs_epi16(ps1, qs1), bd), hev);
+ work_a = _mm_subs_epi16(qs0, ps0);
+ filt = _mm_adds_epi16(filt, work_a);
+ filt = _mm_adds_epi16(filt, work_a);
+ filt = signed_char_clamp_bd_sse2(_mm_adds_epi16(filt, work_a), bd);
+ filt = _mm_and_si128(filt, mask);
+
+ filter1 = signed_char_clamp_bd_sse2(_mm_adds_epi16(filt, t4), bd);
+ filter2 = signed_char_clamp_bd_sse2(_mm_adds_epi16(filt, t3), bd);
+
+ // Filter1 >> 3
+ filter1 = _mm_srai_epi16(filter1, 0x3);
+ filter2 = _mm_srai_epi16(filter2, 0x3);
+
+ qs0 = _mm_adds_epi16(
+ signed_char_clamp_bd_sse2(_mm_subs_epi16(qs0, filter1), bd),
+ t80);
+ ps0 = _mm_adds_epi16(
+ signed_char_clamp_bd_sse2(_mm_adds_epi16(ps0, filter2), bd),
+ t80);
+ filt = _mm_adds_epi16(filter1, t1);
+ filt = _mm_srai_epi16(filt, 1);
+ filt = _mm_andnot_si128(hev, filt);
+
+ qs1 = _mm_adds_epi16(
+ signed_char_clamp_bd_sse2(_mm_subs_epi16(qs1, filt), bd),
+ t80);
+ ps1 = _mm_adds_epi16(
+ signed_char_clamp_bd_sse2(_mm_adds_epi16(ps1, filt), bd),
+ t80);
+ // end highbd_filter4
+ // loopfilter done
+
+ // highbd_flat_mask4
+ flat = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p2, p0),
+ _mm_subs_epu16(p0, p2)),
+ _mm_or_si128(_mm_subs_epu16(p3, p0),
+ _mm_subs_epu16(p0, p3)));
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(q2, q0),
+ _mm_subs_epu16(q0, q2)),
+ _mm_or_si128(_mm_subs_epu16(q3, q0),
+ _mm_subs_epu16(q0, q3)));
+ flat = _mm_max_epi16(work, flat);
+ work = _mm_max_epi16(abs_p1p0, abs_q1q0);
+ flat = _mm_max_epi16(work, flat);
+ flat = _mm_subs_epu16(flat, _mm_slli_epi16(one, bd - 8));
+ flat = _mm_cmpeq_epi16(flat, zero);
+ // end flat_mask4
+
+ // flat & mask = flat && mask (as used in filter8)
+ // (because, in both vars, each block of 16 either all 1s or all 0s)
+ flat = _mm_and_si128(flat, mask);
+
+ p5 = _mm_load_si128((__m128i *)(s - 6 * p));
+ q5 = _mm_load_si128((__m128i *)(s + 5 * p));
+ p6 = _mm_load_si128((__m128i *)(s - 7 * p));
+ q6 = _mm_load_si128((__m128i *)(s + 6 * p));
+ p7 = _mm_load_si128((__m128i *)(s - 8 * p));
+ q7 = _mm_load_si128((__m128i *)(s + 7 * p));
+
+ // highbd_flat_mask5 (arguments passed in are p0, q0, p4-p7, q4-q7
+ // but referred to as p0-p4 & q0-q4 in fn)
+ flat2 = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p4, p0),
+ _mm_subs_epu16(p0, p4)),
+ _mm_or_si128(_mm_subs_epu16(q4, q0),
+ _mm_subs_epu16(q0, q4)));
+
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p5, p0),
+ _mm_subs_epu16(p0, p5)),
+ _mm_or_si128(_mm_subs_epu16(q5, q0),
+ _mm_subs_epu16(q0, q5)));
+ flat2 = _mm_max_epi16(work, flat2);
+
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p6, p0),
+ _mm_subs_epu16(p0, p6)),
+ _mm_or_si128(_mm_subs_epu16(q6, q0),
+ _mm_subs_epu16(q0, q6)));
+ flat2 = _mm_max_epi16(work, flat2);
+
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p7, p0),
+ _mm_subs_epu16(p0, p7)),
+ _mm_or_si128(_mm_subs_epu16(q7, q0),
+ _mm_subs_epu16(q0, q7)));
+ flat2 = _mm_max_epi16(work, flat2);
+
+ flat2 = _mm_subs_epu16(flat2, _mm_slli_epi16(one, bd - 8));
+ flat2 = _mm_cmpeq_epi16(flat2, zero);
+ flat2 = _mm_and_si128(flat2, flat); // flat2 & flat & mask
+ // end highbd_flat_mask5
+
+ // ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ // flat and wide flat calculations
+ eight = _mm_set1_epi16(8);
+ four = _mm_set1_epi16(4);
+
+ pixelFilter_p = _mm_add_epi16(_mm_add_epi16(p6, p5),
+ _mm_add_epi16(p4, p3));
+ pixelFilter_q = _mm_add_epi16(_mm_add_epi16(q6, q5),
+ _mm_add_epi16(q4, q3));
+
+ pixetFilter_p2p1p0 = _mm_add_epi16(p0, _mm_add_epi16(p2, p1));
+ pixelFilter_p = _mm_add_epi16(pixelFilter_p, pixetFilter_p2p1p0);
+
+ pixetFilter_q2q1q0 = _mm_add_epi16(q0, _mm_add_epi16(q2, q1));
+ pixelFilter_q = _mm_add_epi16(pixelFilter_q, pixetFilter_q2q1q0);
+ pixelFilter_p = _mm_add_epi16(eight, _mm_add_epi16(pixelFilter_p,
+ pixelFilter_q));
+ pixetFilter_p2p1p0 = _mm_add_epi16(four,
+ _mm_add_epi16(pixetFilter_p2p1p0,
+ pixetFilter_q2q1q0));
+ flat2_p0 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_p,
+ _mm_add_epi16(p7, p0)), 4);
+ flat2_q0 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_p,
+ _mm_add_epi16(q7, q0)), 4);
+ flat_p0 = _mm_srli_epi16(_mm_add_epi16(pixetFilter_p2p1p0,
+ _mm_add_epi16(p3, p0)), 3);
+ flat_q0 = _mm_srli_epi16(_mm_add_epi16(pixetFilter_p2p1p0,
+ _mm_add_epi16(q3, q0)), 3);
+
+ sum_p7 = _mm_add_epi16(p7, p7);
+ sum_q7 = _mm_add_epi16(q7, q7);
+ sum_p3 = _mm_add_epi16(p3, p3);
+ sum_q3 = _mm_add_epi16(q3, q3);
+
+ pixelFilter_q = _mm_sub_epi16(pixelFilter_p, p6);
+ pixelFilter_p = _mm_sub_epi16(pixelFilter_p, q6);
+ flat2_p1 = _mm_srli_epi16(
+ _mm_add_epi16(pixelFilter_p, _mm_add_epi16(sum_p7, p1)), 4);
+ flat2_q1 = _mm_srli_epi16(
+ _mm_add_epi16(pixelFilter_q, _mm_add_epi16(sum_q7, q1)), 4);
+
+ pixetFilter_q2q1q0 = _mm_sub_epi16(pixetFilter_p2p1p0, p2);
+ pixetFilter_p2p1p0 = _mm_sub_epi16(pixetFilter_p2p1p0, q2);
+ flat_p1 = _mm_srli_epi16(_mm_add_epi16(pixetFilter_p2p1p0,
+ _mm_add_epi16(sum_p3, p1)), 3);
+ flat_q1 = _mm_srli_epi16(_mm_add_epi16(pixetFilter_q2q1q0,
+ _mm_add_epi16(sum_q3, q1)), 3);
+
+ sum_p7 = _mm_add_epi16(sum_p7, p7);
+ sum_q7 = _mm_add_epi16(sum_q7, q7);
+ sum_p3 = _mm_add_epi16(sum_p3, p3);
+ sum_q3 = _mm_add_epi16(sum_q3, q3);
+
+ pixelFilter_p = _mm_sub_epi16(pixelFilter_p, q5);
+ pixelFilter_q = _mm_sub_epi16(pixelFilter_q, p5);
+ flat2_p2 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_p,
+ _mm_add_epi16(sum_p7, p2)), 4);
+ flat2_q2 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_q,
+ _mm_add_epi16(sum_q7, q2)), 4);
+
+ pixetFilter_p2p1p0 = _mm_sub_epi16(pixetFilter_p2p1p0, q1);
+ pixetFilter_q2q1q0 = _mm_sub_epi16(pixetFilter_q2q1q0, p1);
+ flat_p2 = _mm_srli_epi16(_mm_add_epi16(pixetFilter_p2p1p0,
+ _mm_add_epi16(sum_p3, p2)), 3);
+ flat_q2 = _mm_srli_epi16(_mm_add_epi16(pixetFilter_q2q1q0,
+ _mm_add_epi16(sum_q3, q2)), 3);
+
+ sum_p7 = _mm_add_epi16(sum_p7, p7);
+ sum_q7 = _mm_add_epi16(sum_q7, q7);
+ pixelFilter_p = _mm_sub_epi16(pixelFilter_p, q4);
+ pixelFilter_q = _mm_sub_epi16(pixelFilter_q, p4);
+ flat2_p3 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_p,
+ _mm_add_epi16(sum_p7, p3)), 4);
+ flat2_q3 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_q,
+ _mm_add_epi16(sum_q7, q3)), 4);
+
+ sum_p7 = _mm_add_epi16(sum_p7, p7);
+ sum_q7 = _mm_add_epi16(sum_q7, q7);
+ pixelFilter_p = _mm_sub_epi16(pixelFilter_p, q3);
+ pixelFilter_q = _mm_sub_epi16(pixelFilter_q, p3);
+ flat2_p4 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_p,
+ _mm_add_epi16(sum_p7, p4)), 4);
+ flat2_q4 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_q,
+ _mm_add_epi16(sum_q7, q4)), 4);
+
+ sum_p7 = _mm_add_epi16(sum_p7, p7);
+ sum_q7 = _mm_add_epi16(sum_q7, q7);
+ pixelFilter_p = _mm_sub_epi16(pixelFilter_p, q2);
+ pixelFilter_q = _mm_sub_epi16(pixelFilter_q, p2);
+ flat2_p5 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_p,
+ _mm_add_epi16(sum_p7, p5)), 4);
+ flat2_q5 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_q,
+ _mm_add_epi16(sum_q7, q5)), 4);
+
+ sum_p7 = _mm_add_epi16(sum_p7, p7);
+ sum_q7 = _mm_add_epi16(sum_q7, q7);
+ pixelFilter_p = _mm_sub_epi16(pixelFilter_p, q1);
+ pixelFilter_q = _mm_sub_epi16(pixelFilter_q, p1);
+ flat2_p6 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_p,
+ _mm_add_epi16(sum_p7, p6)), 4);
+ flat2_q6 = _mm_srli_epi16(_mm_add_epi16(pixelFilter_q,
+ _mm_add_epi16(sum_q7, q6)), 4);
+
+ // wide flat
+ // ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+ // highbd_filter8
+ p2 = _mm_andnot_si128(flat, p2);
+ // p2 remains unchanged if !(flat && mask)
+ flat_p2 = _mm_and_si128(flat, flat_p2);
+ // when (flat && mask)
+ p2 = _mm_or_si128(p2, flat_p2); // full list of p2 values
+ q2 = _mm_andnot_si128(flat, q2);
+ flat_q2 = _mm_and_si128(flat, flat_q2);
+ q2 = _mm_or_si128(q2, flat_q2); // full list of q2 values
+
+ ps1 = _mm_andnot_si128(flat, ps1);
+ // p1 takes the value assigned to in in filter4 if !(flat && mask)
+ flat_p1 = _mm_and_si128(flat, flat_p1);
+ // when (flat && mask)
+ p1 = _mm_or_si128(ps1, flat_p1); // full list of p1 values
+ qs1 = _mm_andnot_si128(flat, qs1);
+ flat_q1 = _mm_and_si128(flat, flat_q1);
+ q1 = _mm_or_si128(qs1, flat_q1); // full list of q1 values
+
+ ps0 = _mm_andnot_si128(flat, ps0);
+ // p0 takes the value assigned to in in filter4 if !(flat && mask)
+ flat_p0 = _mm_and_si128(flat, flat_p0);
+ // when (flat && mask)
+ p0 = _mm_or_si128(ps0, flat_p0); // full list of p0 values
+ qs0 = _mm_andnot_si128(flat, qs0);
+ flat_q0 = _mm_and_si128(flat, flat_q0);
+ q0 = _mm_or_si128(qs0, flat_q0); // full list of q0 values
+ // end highbd_filter8
+
+ // highbd_filter16
+ p6 = _mm_andnot_si128(flat2, p6);
+ // p6 remains unchanged if !(flat2 && flat && mask)
+ flat2_p6 = _mm_and_si128(flat2, flat2_p6);
+ // get values for when (flat2 && flat && mask)
+ p6 = _mm_or_si128(p6, flat2_p6); // full list of p6 values
+ q6 = _mm_andnot_si128(flat2, q6);
+ // q6 remains unchanged if !(flat2 && flat && mask)
+ flat2_q6 = _mm_and_si128(flat2, flat2_q6);
+ // get values for when (flat2 && flat && mask)
+ q6 = _mm_or_si128(q6, flat2_q6); // full list of q6 values
+ _mm_store_si128((__m128i *)(s - 7 * p), p6);
+ _mm_store_si128((__m128i *)(s + 6 * p), q6);
+
+ p5 = _mm_andnot_si128(flat2, p5);
+ // p5 remains unchanged if !(flat2 && flat && mask)
+ flat2_p5 = _mm_and_si128(flat2, flat2_p5);
+ // get values for when (flat2 && flat && mask)
+ p5 = _mm_or_si128(p5, flat2_p5);
+ // full list of p5 values
+ q5 = _mm_andnot_si128(flat2, q5);
+ // q5 remains unchanged if !(flat2 && flat && mask)
+ flat2_q5 = _mm_and_si128(flat2, flat2_q5);
+ // get values for when (flat2 && flat && mask)
+ q5 = _mm_or_si128(q5, flat2_q5);
+ // full list of q5 values
+ _mm_store_si128((__m128i *)(s - 6 * p), p5);
+ _mm_store_si128((__m128i *)(s + 5 * p), q5);
+
+ p4 = _mm_andnot_si128(flat2, p4);
+ // p4 remains unchanged if !(flat2 && flat && mask)
+ flat2_p4 = _mm_and_si128(flat2, flat2_p4);
+ // get values for when (flat2 && flat && mask)
+ p4 = _mm_or_si128(p4, flat2_p4); // full list of p4 values
+ q4 = _mm_andnot_si128(flat2, q4);
+ // q4 remains unchanged if !(flat2 && flat && mask)
+ flat2_q4 = _mm_and_si128(flat2, flat2_q4);
+ // get values for when (flat2 && flat && mask)
+ q4 = _mm_or_si128(q4, flat2_q4); // full list of q4 values
+ _mm_store_si128((__m128i *)(s - 5 * p), p4);
+ _mm_store_si128((__m128i *)(s + 4 * p), q4);
+
+ p3 = _mm_andnot_si128(flat2, p3);
+ // p3 takes value from highbd_filter8 if !(flat2 && flat && mask)
+ flat2_p3 = _mm_and_si128(flat2, flat2_p3);
+ // get values for when (flat2 && flat && mask)
+ p3 = _mm_or_si128(p3, flat2_p3); // full list of p3 values
+ q3 = _mm_andnot_si128(flat2, q3);
+ // q3 takes value from highbd_filter8 if !(flat2 && flat && mask)
+ flat2_q3 = _mm_and_si128(flat2, flat2_q3);
+ // get values for when (flat2 && flat && mask)
+ q3 = _mm_or_si128(q3, flat2_q3); // full list of q3 values
+ _mm_store_si128((__m128i *)(s - 4 * p), p3);
+ _mm_store_si128((__m128i *)(s + 3 * p), q3);
+
+ p2 = _mm_andnot_si128(flat2, p2);
+ // p2 takes value from highbd_filter8 if !(flat2 && flat && mask)
+ flat2_p2 = _mm_and_si128(flat2, flat2_p2);
+ // get values for when (flat2 && flat && mask)
+ p2 = _mm_or_si128(p2, flat2_p2);
+ // full list of p2 values
+ q2 = _mm_andnot_si128(flat2, q2);
+ // q2 takes value from highbd_filter8 if !(flat2 && flat && mask)
+ flat2_q2 = _mm_and_si128(flat2, flat2_q2);
+ // get values for when (flat2 && flat && mask)
+ q2 = _mm_or_si128(q2, flat2_q2); // full list of q2 values
+ _mm_store_si128((__m128i *)(s - 3 * p), p2);
+ _mm_store_si128((__m128i *)(s + 2 * p), q2);
+
+ p1 = _mm_andnot_si128(flat2, p1);
+ // p1 takes value from highbd_filter8 if !(flat2 && flat && mask)
+ flat2_p1 = _mm_and_si128(flat2, flat2_p1);
+ // get values for when (flat2 && flat && mask)
+ p1 = _mm_or_si128(p1, flat2_p1); // full list of p1 values
+ q1 = _mm_andnot_si128(flat2, q1);
+ // q1 takes value from highbd_filter8 if !(flat2 && flat && mask)
+ flat2_q1 = _mm_and_si128(flat2, flat2_q1);
+ // get values for when (flat2 && flat && mask)
+ q1 = _mm_or_si128(q1, flat2_q1); // full list of q1 values
+ _mm_store_si128((__m128i *)(s - 2 * p), p1);
+ _mm_store_si128((__m128i *)(s + 1 * p), q1);
+
+ p0 = _mm_andnot_si128(flat2, p0);
+ // p0 takes value from highbd_filter8 if !(flat2 && flat && mask)
+ flat2_p0 = _mm_and_si128(flat2, flat2_p0);
+ // get values for when (flat2 && flat && mask)
+ p0 = _mm_or_si128(p0, flat2_p0); // full list of p0 values
+ q0 = _mm_andnot_si128(flat2, q0);
+ // q0 takes value from highbd_filter8 if !(flat2 && flat && mask)
+ flat2_q0 = _mm_and_si128(flat2, flat2_q0);
+ // get values for when (flat2 && flat && mask)
+ q0 = _mm_or_si128(q0, flat2_q0); // full list of q0 values
+ _mm_store_si128((__m128i *)(s - 1 * p), p0);
+ _mm_store_si128((__m128i *)(s - 0 * p), q0);
+}
+
+static void highbd_mb_lpf_horizontal_edge_w_sse2_16(uint16_t *s,
+ int p,
+ const uint8_t *_blimit,
+ const uint8_t *_limit,
+ const uint8_t *_thresh,
+ int bd) {
+ highbd_mb_lpf_horizontal_edge_w_sse2_8(s, p, _blimit, _limit, _thresh, bd);
+ highbd_mb_lpf_horizontal_edge_w_sse2_8(s + 8, p, _blimit, _limit, _thresh,
+ bd);
+}
+
+// TODO(yunqingwang): remove count and call these 2 functions(8 or 16) directly.
+void vp9_highbd_lpf_horizontal_16_sse2(uint16_t *s, int p,
+ const uint8_t *_blimit,
+ const uint8_t *_limit,
+ const uint8_t *_thresh,
+ int count, int bd) {
+ if (count == 1)
+ highbd_mb_lpf_horizontal_edge_w_sse2_8(s, p, _blimit, _limit, _thresh, bd);
+ else
+ highbd_mb_lpf_horizontal_edge_w_sse2_16(s, p, _blimit, _limit, _thresh, bd);
+}
+
+void vp9_highbd_lpf_horizontal_8_sse2(uint16_t *s, int p,
+ const uint8_t *_blimit,
+ const uint8_t *_limit,
+ const uint8_t *_thresh,
+ int count, int bd) {
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, flat_op2, 16);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, flat_op1, 16);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, flat_op0, 16);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, flat_oq2, 16);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, flat_oq1, 16);
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, flat_oq0, 16);
+ const __m128i zero = _mm_set1_epi16(0);
+ const __m128i blimit = _mm_slli_epi16(
+ _mm_unpacklo_epi8(_mm_load_si128((const __m128i *)_blimit), zero),
+ bd - 8);
+ const __m128i limit = _mm_slli_epi16(
+ _mm_unpacklo_epi8(_mm_load_si128((const __m128i *)_limit), zero),
+ bd - 8);
+ const __m128i thresh = _mm_slli_epi16(
+ _mm_unpacklo_epi8(_mm_load_si128((const __m128i *)_thresh), zero),
+ bd - 8);
+ __m128i mask, hev, flat;
+ __m128i p3 = _mm_load_si128((__m128i *)(s - 4 * p));
+ __m128i q3 = _mm_load_si128((__m128i *)(s + 3 * p));
+ __m128i p2 = _mm_load_si128((__m128i *)(s - 3 * p));
+ __m128i q2 = _mm_load_si128((__m128i *)(s + 2 * p));
+ __m128i p1 = _mm_load_si128((__m128i *)(s - 2 * p));
+ __m128i q1 = _mm_load_si128((__m128i *)(s + 1 * p));
+ __m128i p0 = _mm_load_si128((__m128i *)(s - 1 * p));
+ __m128i q0 = _mm_load_si128((__m128i *)(s + 0 * p));
+ const __m128i one = _mm_set1_epi16(1);
+ const __m128i ffff = _mm_cmpeq_epi16(one, one);
+ __m128i abs_p1q1, abs_p0q0, abs_q1q0, abs_p1p0, work;
+ const __m128i four = _mm_set1_epi16(4);
+ __m128i workp_a, workp_b, workp_shft;
+
+ const __m128i t4 = _mm_set1_epi16(4);
+ const __m128i t3 = _mm_set1_epi16(3);
+ const __m128i t80 = _mm_slli_epi16(_mm_set1_epi16(0x80), bd - 8);
+ const __m128i t1 = _mm_set1_epi16(0x1);
+ const __m128i ps1 = _mm_subs_epi16(p1, t80);
+ const __m128i ps0 = _mm_subs_epi16(p0, t80);
+ const __m128i qs0 = _mm_subs_epi16(q0, t80);
+ const __m128i qs1 = _mm_subs_epi16(q1, t80);
+ __m128i filt;
+ __m128i work_a;
+ __m128i filter1, filter2;
+
+ (void)count;
+
+ // filter_mask and hev_mask
+ abs_p1p0 = _mm_or_si128(_mm_subs_epu16(p1, p0),
+ _mm_subs_epu16(p0, p1));
+ abs_q1q0 = _mm_or_si128(_mm_subs_epu16(q1, q0),
+ _mm_subs_epu16(q0, q1));
+
+ abs_p0q0 = _mm_or_si128(_mm_subs_epu16(p0, q0),
+ _mm_subs_epu16(q0, p0));
+ abs_p1q1 = _mm_or_si128(_mm_subs_epu16(p1, q1),
+ _mm_subs_epu16(q1, p1));
+ flat = _mm_max_epi16(abs_p1p0, abs_q1q0);
+ hev = _mm_subs_epu16(flat, thresh);
+ hev = _mm_xor_si128(_mm_cmpeq_epi16(hev, zero), ffff);
+
+ abs_p0q0 =_mm_adds_epu16(abs_p0q0, abs_p0q0);
+ abs_p1q1 = _mm_srli_epi16(abs_p1q1, 1);
+ mask = _mm_subs_epu16(_mm_adds_epu16(abs_p0q0, abs_p1q1), blimit);
+ mask = _mm_xor_si128(_mm_cmpeq_epi16(mask, zero), ffff);
+ // mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 > blimit) * -1;
+ // So taking maximums continues to work:
+ mask = _mm_and_si128(mask, _mm_adds_epu16(limit, one));
+ mask = _mm_max_epi16(abs_p1p0, mask);
+ // mask |= (abs(p1 - p0) > limit) * -1;
+ mask = _mm_max_epi16(abs_q1q0, mask);
+ // mask |= (abs(q1 - q0) > limit) * -1;
+
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p2, p1),
+ _mm_subs_epu16(p1, p2)),
+ _mm_or_si128(_mm_subs_epu16(q2, q1),
+ _mm_subs_epu16(q1, q2)));
+ mask = _mm_max_epi16(work, mask);
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p3, p2),
+ _mm_subs_epu16(p2, p3)),
+ _mm_or_si128(_mm_subs_epu16(q3, q2),
+ _mm_subs_epu16(q2, q3)));
+ mask = _mm_max_epi16(work, mask);
+ mask = _mm_subs_epu16(mask, limit);
+ mask = _mm_cmpeq_epi16(mask, zero);
+
+ // flat_mask4
+ flat = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p2, p0),
+ _mm_subs_epu16(p0, p2)),
+ _mm_or_si128(_mm_subs_epu16(q2, q0),
+ _mm_subs_epu16(q0, q2)));
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p3, p0),
+ _mm_subs_epu16(p0, p3)),
+ _mm_or_si128(_mm_subs_epu16(q3, q0),
+ _mm_subs_epu16(q0, q3)));
+ flat = _mm_max_epi16(work, flat);
+ flat = _mm_max_epi16(abs_p1p0, flat);
+ flat = _mm_max_epi16(abs_q1q0, flat);
+ flat = _mm_subs_epu16(flat, _mm_slli_epi16(one, bd - 8));
+ flat = _mm_cmpeq_epi16(flat, zero);
+ flat = _mm_and_si128(flat, mask); // flat & mask
+
+ // Added before shift for rounding part of ROUND_POWER_OF_TWO
+
+ workp_a = _mm_add_epi16(_mm_add_epi16(p3, p3), _mm_add_epi16(p2, p1));
+ workp_a = _mm_add_epi16(_mm_add_epi16(workp_a, four), p0);
+ workp_b = _mm_add_epi16(_mm_add_epi16(q0, p2), p3);
+ workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
+ _mm_store_si128((__m128i *)&flat_op2[0], workp_shft);
+
+ workp_b = _mm_add_epi16(_mm_add_epi16(q0, q1), p1);
+ workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
+ _mm_store_si128((__m128i *)&flat_op1[0], workp_shft);
+
+ workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p3), q2);
+ workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p1), p0);
+ workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
+ _mm_store_si128((__m128i *)&flat_op0[0], workp_shft);
+
+ workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p3), q3);
+ workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p0), q0);
+ workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
+ _mm_store_si128((__m128i *)&flat_oq0[0], workp_shft);
+
+ workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p2), q3);
+ workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q0), q1);
+ workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
+ _mm_store_si128((__m128i *)&flat_oq1[0], workp_shft);
+
+ workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p1), q3);
+ workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q1), q2);
+ workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
+ _mm_store_si128((__m128i *)&flat_oq2[0], workp_shft);
+
+ // lp filter
+ filt = signed_char_clamp_bd_sse2(_mm_subs_epi16(ps1, qs1), bd);
+ filt = _mm_and_si128(filt, hev);
+ work_a = _mm_subs_epi16(qs0, ps0);
+ filt = _mm_adds_epi16(filt, work_a);
+ filt = _mm_adds_epi16(filt, work_a);
+ filt = _mm_adds_epi16(filt, work_a);
+ // (vp9_filter + 3 * (qs0 - ps0)) & mask
+ filt = signed_char_clamp_bd_sse2(filt, bd);
+ filt = _mm_and_si128(filt, mask);
+
+ filter1 = _mm_adds_epi16(filt, t4);
+ filter2 = _mm_adds_epi16(filt, t3);
+
+ // Filter1 >> 3
+ filter1 = signed_char_clamp_bd_sse2(filter1, bd);
+ filter1 = _mm_srai_epi16(filter1, 3);
+
+ // Filter2 >> 3
+ filter2 = signed_char_clamp_bd_sse2(filter2, bd);
+ filter2 = _mm_srai_epi16(filter2, 3);
+
+ // filt >> 1
+ filt = _mm_adds_epi16(filter1, t1);
+ filt = _mm_srai_epi16(filt, 1);
+ // filter = ROUND_POWER_OF_TWO(filter1, 1) & ~hev;
+ filt = _mm_andnot_si128(hev, filt);
+
+ work_a = signed_char_clamp_bd_sse2(_mm_subs_epi16(qs0, filter1), bd);
+ work_a = _mm_adds_epi16(work_a, t80);
+ q0 = _mm_load_si128((__m128i *)flat_oq0);
+ work_a = _mm_andnot_si128(flat, work_a);
+ q0 = _mm_and_si128(flat, q0);
+ q0 = _mm_or_si128(work_a, q0);
+
+ work_a = signed_char_clamp_bd_sse2(_mm_subs_epi16(qs1, filt), bd);
+ work_a = _mm_adds_epi16(work_a, t80);
+ q1 = _mm_load_si128((__m128i *)flat_oq1);
+ work_a = _mm_andnot_si128(flat, work_a);
+ q1 = _mm_and_si128(flat, q1);
+ q1 = _mm_or_si128(work_a, q1);
+
+ work_a = _mm_loadu_si128((__m128i *)(s + 2 * p));
+ q2 = _mm_load_si128((__m128i *)flat_oq2);
+ work_a = _mm_andnot_si128(flat, work_a);
+ q2 = _mm_and_si128(flat, q2);
+ q2 = _mm_or_si128(work_a, q2);
+
+ work_a = signed_char_clamp_bd_sse2(_mm_adds_epi16(ps0, filter2), bd);
+ work_a = _mm_adds_epi16(work_a, t80);
+ p0 = _mm_load_si128((__m128i *)flat_op0);
+ work_a = _mm_andnot_si128(flat, work_a);
+ p0 = _mm_and_si128(flat, p0);
+ p0 = _mm_or_si128(work_a, p0);
+
+ work_a = signed_char_clamp_bd_sse2(_mm_adds_epi16(ps1, filt), bd);
+ work_a = _mm_adds_epi16(work_a, t80);
+ p1 = _mm_load_si128((__m128i *)flat_op1);
+ work_a = _mm_andnot_si128(flat, work_a);
+ p1 = _mm_and_si128(flat, p1);
+ p1 = _mm_or_si128(work_a, p1);
+
+ work_a = _mm_loadu_si128((__m128i *)(s - 3 * p));
+ p2 = _mm_load_si128((__m128i *)flat_op2);
+ work_a = _mm_andnot_si128(flat, work_a);
+ p2 = _mm_and_si128(flat, p2);
+ p2 = _mm_or_si128(work_a, p2);
+
+ _mm_store_si128((__m128i *)(s - 3 * p), p2);
+ _mm_store_si128((__m128i *)(s - 2 * p), p1);
+ _mm_store_si128((__m128i *)(s - 1 * p), p0);
+ _mm_store_si128((__m128i *)(s + 0 * p), q0);
+ _mm_store_si128((__m128i *)(s + 1 * p), q1);
+ _mm_store_si128((__m128i *)(s + 2 * p), q2);
+}
+
+void vp9_highbd_lpf_horizontal_8_dual_sse2(uint16_t *s, int p,
+ const uint8_t *_blimit0,
+ const uint8_t *_limit0,
+ const uint8_t *_thresh0,
+ const uint8_t *_blimit1,
+ const uint8_t *_limit1,
+ const uint8_t *_thresh1,
+ int bd) {
+ vp9_highbd_lpf_horizontal_8_sse2(s, p, _blimit0, _limit0, _thresh0, 1, bd);
+ vp9_highbd_lpf_horizontal_8_sse2(s + 8, p, _blimit1, _limit1, _thresh1,
+ 1, bd);
+}
+
+void vp9_highbd_lpf_horizontal_4_sse2(uint16_t *s, int p,
+ const uint8_t *_blimit,
+ const uint8_t *_limit,
+ const uint8_t *_thresh,
+ int count, int bd) {
+ const __m128i zero = _mm_set1_epi16(0);
+ const __m128i blimit = _mm_slli_epi16(
+ _mm_unpacklo_epi8(
+ _mm_load_si128((const __m128i *)_blimit), zero), bd - 8);
+ const __m128i limit = _mm_slli_epi16(
+ _mm_unpacklo_epi8(
+ _mm_load_si128((const __m128i *)_limit), zero), bd - 8);
+ const __m128i thresh = _mm_slli_epi16(
+ _mm_unpacklo_epi8(
+ _mm_load_si128((const __m128i *)_thresh), zero), bd - 8);
+ __m128i mask, hev, flat;
+ __m128i p3 = _mm_loadu_si128((__m128i *)(s - 4 * p));
+ __m128i p2 = _mm_loadu_si128((__m128i *)(s - 3 * p));
+ __m128i p1 = _mm_loadu_si128((__m128i *)(s - 2 * p));
+ __m128i p0 = _mm_loadu_si128((__m128i *)(s - 1 * p));
+ __m128i q0 = _mm_loadu_si128((__m128i *)(s - 0 * p));
+ __m128i q1 = _mm_loadu_si128((__m128i *)(s + 1 * p));
+ __m128i q2 = _mm_loadu_si128((__m128i *)(s + 2 * p));
+ __m128i q3 = _mm_loadu_si128((__m128i *)(s + 3 * p));
+ const __m128i abs_p1p0 = _mm_or_si128(_mm_subs_epu16(p1, p0),
+ _mm_subs_epu16(p0, p1));
+ const __m128i abs_q1q0 = _mm_or_si128(_mm_subs_epu16(q1, q0),
+ _mm_subs_epu16(q0, q1));
+ const __m128i ffff = _mm_cmpeq_epi16(abs_p1p0, abs_p1p0);
+ const __m128i one = _mm_set1_epi16(1);
+ __m128i abs_p0q0 = _mm_or_si128(_mm_subs_epu16(p0, q0),
+ _mm_subs_epu16(q0, p0));
+ __m128i abs_p1q1 = _mm_or_si128(_mm_subs_epu16(p1, q1),
+ _mm_subs_epu16(q1, p1));
+ __m128i work;
+ const __m128i t4 = _mm_set1_epi16(4);
+ const __m128i t3 = _mm_set1_epi16(3);
+ const __m128i t80 = _mm_slli_epi16(_mm_set1_epi16(0x80), bd - 8);
+ const __m128i tff80 = _mm_slli_epi16(_mm_set1_epi16(0xff80), bd - 8);
+ const __m128i tffe0 = _mm_slli_epi16(_mm_set1_epi16(0xffe0), bd - 8);
+ const __m128i t1f = _mm_srli_epi16(_mm_set1_epi16(0x1fff), 16 - bd);
+ // equivalent to shifting 0x1f left by bitdepth - 8
+ // and setting new bits to 1
+ const __m128i t1 = _mm_set1_epi16(0x1);
+ const __m128i t7f = _mm_srli_epi16(_mm_set1_epi16(0x7fff), 16 - bd);
+ // equivalent to shifting 0x7f left by bitdepth - 8
+ // and setting new bits to 1
+ const __m128i ps1 = _mm_subs_epi16(_mm_loadu_si128((__m128i *)(s - 2 * p)),
+ t80);
+ const __m128i ps0 = _mm_subs_epi16(_mm_loadu_si128((__m128i *)(s - 1 * p)),
+ t80);
+ const __m128i qs0 = _mm_subs_epi16(_mm_loadu_si128((__m128i *)(s + 0 * p)),
+ t80);
+ const __m128i qs1 = _mm_subs_epi16(_mm_loadu_si128((__m128i *)(s + 1 * p)),
+ t80);
+ __m128i filt;
+ __m128i work_a;
+ __m128i filter1, filter2;
+
+ (void)count;
+
+ // filter_mask and hev_mask
+ flat = _mm_max_epi16(abs_p1p0, abs_q1q0);
+ hev = _mm_subs_epu16(flat, thresh);
+ hev = _mm_xor_si128(_mm_cmpeq_epi16(hev, zero), ffff);
+
+ abs_p0q0 =_mm_adds_epu16(abs_p0q0, abs_p0q0);
+ abs_p1q1 = _mm_srli_epi16(abs_p1q1, 1);
+ mask = _mm_subs_epu16(_mm_adds_epu16(abs_p0q0, abs_p1q1), blimit);
+ mask = _mm_xor_si128(_mm_cmpeq_epi16(mask, zero), ffff);
+ // mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 > blimit) * -1;
+ // So taking maximums continues to work:
+ mask = _mm_and_si128(mask, _mm_adds_epu16(limit, one));
+ mask = _mm_max_epi16(flat, mask);
+ // mask |= (abs(p1 - p0) > limit) * -1;
+ // mask |= (abs(q1 - q0) > limit) * -1;
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(p2, p1),
+ _mm_subs_epu16(p1, p2)),
+ _mm_or_si128(_mm_subs_epu16(p3, p2),
+ _mm_subs_epu16(p2, p3)));
+ mask = _mm_max_epi16(work, mask);
+ work = _mm_max_epi16(_mm_or_si128(_mm_subs_epu16(q2, q1),
+ _mm_subs_epu16(q1, q2)),
+ _mm_or_si128(_mm_subs_epu16(q3, q2),
+ _mm_subs_epu16(q2, q3)));
+ mask = _mm_max_epi16(work, mask);
+ mask = _mm_subs_epu16(mask, limit);
+ mask = _mm_cmpeq_epi16(mask, zero);
+
+ // filter4
+ filt = signed_char_clamp_bd_sse2(_mm_subs_epi16(ps1, qs1), bd);
+ filt = _mm_and_si128(filt, hev);
+ work_a = _mm_subs_epi16(qs0, ps0);
+ filt = _mm_adds_epi16(filt, work_a);
+ filt = _mm_adds_epi16(filt, work_a);
+ filt = signed_char_clamp_bd_sse2(_mm_adds_epi16(filt, work_a), bd);
+ // (vp9_filter + 3 * (qs0 - ps0)) & mask
+ filt = _mm_and_si128(filt, mask);
+
+ filter1 = signed_char_clamp_bd_sse2(_mm_adds_epi16(filt, t4), bd);
+ filter2 = signed_char_clamp_bd_sse2(_mm_adds_epi16(filt, t3), bd);
+
+ // Filter1 >> 3
+ work_a = _mm_cmpgt_epi16(zero, filter1); // get the values that are <0
+ filter1 = _mm_srli_epi16(filter1, 3);
+ work_a = _mm_and_si128(work_a, tffe0); // sign bits for the values < 0
+ filter1 = _mm_and_si128(filter1, t1f); // clamp the range
+ filter1 = _mm_or_si128(filter1, work_a); // reinsert the sign bits
+
+ // Filter2 >> 3
+ work_a = _mm_cmpgt_epi16(zero, filter2);
+ filter2 = _mm_srli_epi16(filter2, 3);
+ work_a = _mm_and_si128(work_a, tffe0);
+ filter2 = _mm_and_si128(filter2, t1f);
+ filter2 = _mm_or_si128(filter2, work_a);
+
+ // filt >> 1
+ filt = _mm_adds_epi16(filter1, t1);
+ work_a = _mm_cmpgt_epi16(zero, filt);
+ filt = _mm_srli_epi16(filt, 1);
+ work_a = _mm_and_si128(work_a, tff80);
+ filt = _mm_and_si128(filt, t7f);
+ filt = _mm_or_si128(filt, work_a);
+
+ filt = _mm_andnot_si128(hev, filt);
+
+ q0 = _mm_adds_epi16(
+ signed_char_clamp_bd_sse2(_mm_subs_epi16(qs0, filter1), bd), t80);
+ q1 = _mm_adds_epi16(
+ signed_char_clamp_bd_sse2(_mm_subs_epi16(qs1, filt), bd), t80);
+ p0 = _mm_adds_epi16(
+ signed_char_clamp_bd_sse2(_mm_adds_epi16(ps0, filter2), bd), t80);
+ p1 = _mm_adds_epi16(
+ signed_char_clamp_bd_sse2(_mm_adds_epi16(ps1, filt), bd), t80);
+
+ _mm_storeu_si128((__m128i *)(s - 2 * p), p1);
+ _mm_storeu_si128((__m128i *)(s - 1 * p), p0);
+ _mm_storeu_si128((__m128i *)(s + 0 * p), q0);
+ _mm_storeu_si128((__m128i *)(s + 1 * p), q1);
+}
+
+void vp9_highbd_lpf_horizontal_4_dual_sse2(uint16_t *s, int p,
+ const uint8_t *_blimit0,
+ const uint8_t *_limit0,
+ const uint8_t *_thresh0,
+ const uint8_t *_blimit1,
+ const uint8_t *_limit1,
+ const uint8_t *_thresh1,
+ int bd) {
+ vp9_highbd_lpf_horizontal_4_sse2(s, p, _blimit0, _limit0, _thresh0, 1, bd);
+ vp9_highbd_lpf_horizontal_4_sse2(s + 8, p, _blimit1, _limit1, _thresh1, 1,
+ bd);
+}
+
+static INLINE void highbd_transpose(uint16_t *src[], int in_p,
+ uint16_t *dst[], int out_p,
+ int num_8x8_to_transpose) {
+ int idx8x8 = 0;
+ __m128i p0, p1, p2, p3, p4, p5, p6, p7, x0, x1, x2, x3, x4, x5, x6, x7;
+ do {
+ uint16_t *in = src[idx8x8];
+ uint16_t *out = dst[idx8x8];
+
+ p0 = _mm_loadu_si128((__m128i *)(in + 0*in_p)); // 00 01 02 03 04 05 06 07
+ p1 = _mm_loadu_si128((__m128i *)(in + 1*in_p)); // 10 11 12 13 14 15 16 17
+ p2 = _mm_loadu_si128((__m128i *)(in + 2*in_p)); // 20 21 22 23 24 25 26 27
+ p3 = _mm_loadu_si128((__m128i *)(in + 3*in_p)); // 30 31 32 33 34 35 36 37
+ p4 = _mm_loadu_si128((__m128i *)(in + 4*in_p)); // 40 41 42 43 44 45 46 47
+ p5 = _mm_loadu_si128((__m128i *)(in + 5*in_p)); // 50 51 52 53 54 55 56 57
+ p6 = _mm_loadu_si128((__m128i *)(in + 6*in_p)); // 60 61 62 63 64 65 66 67
+ p7 = _mm_loadu_si128((__m128i *)(in + 7*in_p)); // 70 71 72 73 74 75 76 77
+ // 00 10 01 11 02 12 03 13
+ x0 = _mm_unpacklo_epi16(p0, p1);
+ // 20 30 21 31 22 32 23 33
+ x1 = _mm_unpacklo_epi16(p2, p3);
+ // 40 50 41 51 42 52 43 53
+ x2 = _mm_unpacklo_epi16(p4, p5);
+ // 60 70 61 71 62 72 63 73
+ x3 = _mm_unpacklo_epi16(p6, p7);
+ // 00 10 20 30 01 11 21 31
+ x4 = _mm_unpacklo_epi32(x0, x1);
+ // 40 50 60 70 41 51 61 71
+ x5 = _mm_unpacklo_epi32(x2, x3);
+ // 00 10 20 30 40 50 60 70
+ x6 = _mm_unpacklo_epi64(x4, x5);
+ // 01 11 21 31 41 51 61 71
+ x7 = _mm_unpackhi_epi64(x4, x5);
+
+ _mm_storeu_si128((__m128i *)(out + 0*out_p), x6);
+ // 00 10 20 30 40 50 60 70
+ _mm_storeu_si128((__m128i *)(out + 1*out_p), x7);
+ // 01 11 21 31 41 51 61 71
+
+ // 02 12 22 32 03 13 23 33
+ x4 = _mm_unpackhi_epi32(x0, x1);
+ // 42 52 62 72 43 53 63 73
+ x5 = _mm_unpackhi_epi32(x2, x3);
+ // 02 12 22 32 42 52 62 72
+ x6 = _mm_unpacklo_epi64(x4, x5);
+ // 03 13 23 33 43 53 63 73
+ x7 = _mm_unpackhi_epi64(x4, x5);
+
+ _mm_storeu_si128((__m128i *)(out + 2*out_p), x6);
+ // 02 12 22 32 42 52 62 72
+ _mm_storeu_si128((__m128i *)(out + 3*out_p), x7);
+ // 03 13 23 33 43 53 63 73
+
+ // 04 14 05 15 06 16 07 17
+ x0 = _mm_unpackhi_epi16(p0, p1);
+ // 24 34 25 35 26 36 27 37
+ x1 = _mm_unpackhi_epi16(p2, p3);
+ // 44 54 45 55 46 56 47 57
+ x2 = _mm_unpackhi_epi16(p4, p5);
+ // 64 74 65 75 66 76 67 77
+ x3 = _mm_unpackhi_epi16(p6, p7);
+ // 04 14 24 34 05 15 25 35
+ x4 = _mm_unpacklo_epi32(x0, x1);
+ // 44 54 64 74 45 55 65 75
+ x5 = _mm_unpacklo_epi32(x2, x3);
+ // 04 14 24 34 44 54 64 74
+ x6 = _mm_unpacklo_epi64(x4, x5);
+ // 05 15 25 35 45 55 65 75
+ x7 = _mm_unpackhi_epi64(x4, x5);
+
+ _mm_storeu_si128((__m128i *)(out + 4*out_p), x6);
+ // 04 14 24 34 44 54 64 74
+ _mm_storeu_si128((__m128i *)(out + 5*out_p), x7);
+ // 05 15 25 35 45 55 65 75
+
+ // 06 16 26 36 07 17 27 37
+ x4 = _mm_unpackhi_epi32(x0, x1);
+ // 46 56 66 76 47 57 67 77
+ x5 = _mm_unpackhi_epi32(x2, x3);
+ // 06 16 26 36 46 56 66 76
+ x6 = _mm_unpacklo_epi64(x4, x5);
+ // 07 17 27 37 47 57 67 77
+ x7 = _mm_unpackhi_epi64(x4, x5);
+
+ _mm_storeu_si128((__m128i *)(out + 6*out_p), x6);
+ // 06 16 26 36 46 56 66 76
+ _mm_storeu_si128((__m128i *)(out + 7*out_p), x7);
+ // 07 17 27 37 47 57 67 77
+ } while (++idx8x8 < num_8x8_to_transpose);
+}
+
+static INLINE void highbd_transpose8x16(uint16_t *in0, uint16_t *in1,
+ int in_p, uint16_t *out, int out_p) {
+ uint16_t *src0[1];
+ uint16_t *src1[1];
+ uint16_t *dest0[1];
+ uint16_t *dest1[1];
+ src0[0] = in0;
+ src1[0] = in1;
+ dest0[0] = out;
+ dest1[0] = out + 8;
+ highbd_transpose(src0, in_p, dest0, out_p, 1);
+ highbd_transpose(src1, in_p, dest1, out_p, 1);
+}
+
+void vp9_highbd_lpf_vertical_4_sse2(uint16_t *s, int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh,
+ int count, int bd) {
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, t_dst, 8 * 8);
+ uint16_t *src[1];
+ uint16_t *dst[1];
+ (void)count;
+
+ // Transpose 8x8
+ src[0] = s - 4;
+ dst[0] = t_dst;
+
+ highbd_transpose(src, p, dst, 8, 1);
+
+ // Loop filtering
+ vp9_highbd_lpf_horizontal_4_sse2(t_dst + 4 * 8, 8, blimit, limit, thresh, 1,
+ bd);
+
+ src[0] = t_dst;
+ dst[0] = s - 4;
+
+ // Transpose back
+ highbd_transpose(src, 8, dst, p, 1);
+}
+
+void vp9_highbd_lpf_vertical_4_dual_sse2(uint16_t *s, int p,
+ const uint8_t *blimit0,
+ const uint8_t *limit0,
+ const uint8_t *thresh0,
+ const uint8_t *blimit1,
+ const uint8_t *limit1,
+ const uint8_t *thresh1,
+ int bd) {
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, t_dst, 16 * 8);
+ uint16_t *src[2];
+ uint16_t *dst[2];
+
+ // Transpose 8x16
+ highbd_transpose8x16(s - 4, s - 4 + p * 8, p, t_dst, 16);
+
+ // Loop filtering
+ vp9_highbd_lpf_horizontal_4_dual_sse2(t_dst + 4 * 16, 16, blimit0, limit0,
+ thresh0, blimit1, limit1, thresh1, bd);
+ src[0] = t_dst;
+ src[1] = t_dst + 8;
+ dst[0] = s - 4;
+ dst[1] = s - 4 + p * 8;
+
+ // Transpose back
+ highbd_transpose(src, 16, dst, p, 2);
+}
+
+void vp9_highbd_lpf_vertical_8_sse2(uint16_t *s, int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh,
+ int count, int bd) {
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, t_dst, 8 * 8);
+ uint16_t *src[1];
+ uint16_t *dst[1];
+ (void)count;
+
+ // Transpose 8x8
+ src[0] = s - 4;
+ dst[0] = t_dst;
+
+ highbd_transpose(src, p, dst, 8, 1);
+
+ // Loop filtering
+ vp9_highbd_lpf_horizontal_8_sse2(t_dst + 4 * 8, 8, blimit, limit, thresh, 1,
+ bd);
+
+ src[0] = t_dst;
+ dst[0] = s - 4;
+
+ // Transpose back
+ highbd_transpose(src, 8, dst, p, 1);
+}
+
+void vp9_highbd_lpf_vertical_8_dual_sse2(uint16_t *s, int p,
+ const uint8_t *blimit0,
+ const uint8_t *limit0,
+ const uint8_t *thresh0,
+ const uint8_t *blimit1,
+ const uint8_t *limit1,
+ const uint8_t *thresh1,
+ int bd) {
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, t_dst, 16 * 8);
+ uint16_t *src[2];
+ uint16_t *dst[2];
+
+ // Transpose 8x16
+ highbd_transpose8x16(s - 4, s - 4 + p * 8, p, t_dst, 16);
+
+ // Loop filtering
+ vp9_highbd_lpf_horizontal_8_dual_sse2(t_dst + 4 * 16, 16, blimit0, limit0,
+ thresh0, blimit1, limit1, thresh1, bd);
+ src[0] = t_dst;
+ src[1] = t_dst + 8;
+
+ dst[0] = s - 4;
+ dst[1] = s - 4 + p * 8;
+
+ // Transpose back
+ highbd_transpose(src, 16, dst, p, 2);
+}
+
+void vp9_highbd_lpf_vertical_16_sse2(uint16_t *s, int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh,
+ int bd) {
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, t_dst, 8 * 16);
+ uint16_t *src[2];
+ uint16_t *dst[2];
+
+ src[0] = s - 8;
+ src[1] = s;
+ dst[0] = t_dst;
+ dst[1] = t_dst + 8 * 8;
+
+ // Transpose 16x8
+ highbd_transpose(src, p, dst, 8, 2);
+
+ // Loop filtering
+ highbd_mb_lpf_horizontal_edge_w_sse2_8(t_dst + 8 * 8, 8, blimit, limit,
+ thresh, bd);
+ src[0] = t_dst;
+ src[1] = t_dst + 8 * 8;
+ dst[0] = s - 8;
+ dst[1] = s;
+
+ // Transpose back
+ highbd_transpose(src, 8, dst, p, 2);
+}
+
+void vp9_highbd_lpf_vertical_16_dual_sse2(uint16_t *s,
+ int p,
+ const uint8_t *blimit,
+ const uint8_t *limit,
+ const uint8_t *thresh,
+ int bd) {
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, t_dst, 256);
+
+ // Transpose 16x16
+ highbd_transpose8x16(s - 8, s - 8 + 8 * p, p, t_dst, 16);
+ highbd_transpose8x16(s, s + 8 * p, p, t_dst + 8 * 16, 16);
+
+ // Loop filtering
+ highbd_mb_lpf_horizontal_edge_w_sse2_16(t_dst + 8 * 16, 16, blimit, limit,
+ thresh, bd);
+
+ // Transpose back
+ highbd_transpose8x16(t_dst, t_dst + 8 * 16, 16, s - 8, p);
+ highbd_transpose8x16(t_dst + 8, t_dst + 8 + 8 * 16, 16, s - 8 + 8 * p, p);
+}
diff --git a/source/libvpx/vp9/common/x86/vp9_high_subpixel_8t_sse2.asm b/source/libvpx/vp9/common/x86/vp9_high_subpixel_8t_sse2.asm
index 4bdbb83..29ec151 100644
--- a/source/libvpx/vp9/common/x86/vp9_high_subpixel_8t_sse2.asm
+++ b/source/libvpx/vp9/common/x86/vp9_high_subpixel_8t_sse2.asm
@@ -206,8 +206,8 @@
; unsigned int output_height,
; short *filter
;)
-global sym(vp9_high_filter_block1d4_v8_sse2) PRIVATE
-sym(vp9_high_filter_block1d4_v8_sse2):
+global sym(vp9_highbd_filter_block1d4_v8_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d4_v8_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -276,8 +276,8 @@
; unsigned int output_height,
; short *filter
;)
-global sym(vp9_high_filter_block1d8_v8_sse2) PRIVATE
-sym(vp9_high_filter_block1d8_v8_sse2):
+global sym(vp9_highbd_filter_block1d8_v8_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d8_v8_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -335,8 +335,8 @@
; unsigned int output_height,
; short *filter
;)
-global sym(vp9_high_filter_block1d16_v8_sse2) PRIVATE
-sym(vp9_high_filter_block1d16_v8_sse2):
+global sym(vp9_highbd_filter_block1d16_v8_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d16_v8_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -389,8 +389,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d4_v8_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d4_v8_avg_sse2):
+global sym(vp9_highbd_filter_block1d4_v8_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d4_v8_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -450,8 +450,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d8_v8_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d8_v8_avg_sse2):
+global sym(vp9_highbd_filter_block1d8_v8_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d8_v8_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -499,8 +499,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d16_v8_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d16_v8_avg_sse2):
+global sym(vp9_highbd_filter_block1d16_v8_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d16_v8_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -561,8 +561,8 @@
; unsigned int output_height,
; short *filter
;)
-global sym(vp9_high_filter_block1d4_h8_sse2) PRIVATE
-sym(vp9_high_filter_block1d4_h8_sse2):
+global sym(vp9_highbd_filter_block1d4_h8_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d4_h8_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -636,8 +636,8 @@
; unsigned int output_height,
; short *filter
;)
-global sym(vp9_high_filter_block1d8_h8_sse2) PRIVATE
-sym(vp9_high_filter_block1d8_h8_sse2):
+global sym(vp9_highbd_filter_block1d8_h8_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d8_h8_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -702,8 +702,8 @@
; unsigned int output_height,
; short *filter
;)
-global sym(vp9_high_filter_block1d16_h8_sse2) PRIVATE
-sym(vp9_high_filter_block1d16_h8_sse2):
+global sym(vp9_highbd_filter_block1d16_h8_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d16_h8_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -770,8 +770,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d4_h8_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d4_h8_avg_sse2):
+global sym(vp9_highbd_filter_block1d4_h8_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d4_h8_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -836,8 +836,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d8_h8_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d8_h8_avg_sse2):
+global sym(vp9_highbd_filter_block1d8_h8_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d8_h8_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -893,8 +893,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d16_h8_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d16_h8_avg_sse2):
+global sym(vp9_highbd_filter_block1d16_h8_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d16_h8_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
diff --git a/source/libvpx/vp9/common/x86/vp9_high_subpixel_bilinear_sse2.asm b/source/libvpx/vp9/common/x86/vp9_high_subpixel_bilinear_sse2.asm
index b7d4a61..9378412 100644
--- a/source/libvpx/vp9/common/x86/vp9_high_subpixel_bilinear_sse2.asm
+++ b/source/libvpx/vp9/common/x86/vp9_high_subpixel_bilinear_sse2.asm
@@ -171,8 +171,8 @@
%endm
%endif
-global sym(vp9_high_filter_block1d4_v2_sse2) PRIVATE
-sym(vp9_high_filter_block1d4_v2_sse2):
+global sym(vp9_highbd_filter_block1d4_v2_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d4_v2_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -196,8 +196,8 @@
ret
%if ARCH_X86_64
-global sym(vp9_high_filter_block1d8_v2_sse2) PRIVATE
-sym(vp9_high_filter_block1d8_v2_sse2):
+global sym(vp9_highbd_filter_block1d8_v2_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d8_v2_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -222,8 +222,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d16_v2_sse2) PRIVATE
-sym(vp9_high_filter_block1d16_v2_sse2):
+global sym(vp9_highbd_filter_block1d16_v2_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d16_v2_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -251,8 +251,8 @@
ret
%endif
-global sym(vp9_high_filter_block1d4_v2_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d4_v2_avg_sse2):
+global sym(vp9_highbd_filter_block1d4_v2_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d4_v2_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -276,8 +276,8 @@
ret
%if ARCH_X86_64
-global sym(vp9_high_filter_block1d8_v2_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d8_v2_avg_sse2):
+global sym(vp9_highbd_filter_block1d8_v2_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d8_v2_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -302,8 +302,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d16_v2_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d16_v2_avg_sse2):
+global sym(vp9_highbd_filter_block1d16_v2_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d16_v2_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -331,8 +331,8 @@
ret
%endif
-global sym(vp9_high_filter_block1d4_h2_sse2) PRIVATE
-sym(vp9_high_filter_block1d4_h2_sse2):
+global sym(vp9_highbd_filter_block1d4_h2_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d4_h2_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -357,8 +357,8 @@
ret
%if ARCH_X86_64
-global sym(vp9_high_filter_block1d8_h2_sse2) PRIVATE
-sym(vp9_high_filter_block1d8_h2_sse2):
+global sym(vp9_highbd_filter_block1d8_h2_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d8_h2_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -383,8 +383,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d16_h2_sse2) PRIVATE
-sym(vp9_high_filter_block1d16_h2_sse2):
+global sym(vp9_highbd_filter_block1d16_h2_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d16_h2_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -412,8 +412,8 @@
ret
%endif
-global sym(vp9_high_filter_block1d4_h2_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d4_h2_avg_sse2):
+global sym(vp9_highbd_filter_block1d4_h2_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d4_h2_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -438,8 +438,8 @@
ret
%if ARCH_X86_64
-global sym(vp9_high_filter_block1d8_h2_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d8_h2_avg_sse2):
+global sym(vp9_highbd_filter_block1d8_h2_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d8_h2_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
@@ -464,8 +464,8 @@
pop rbp
ret
-global sym(vp9_high_filter_block1d16_h2_avg_sse2) PRIVATE
-sym(vp9_high_filter_block1d16_h2_avg_sse2):
+global sym(vp9_highbd_filter_block1d16_h2_avg_sse2) PRIVATE
+sym(vp9_highbd_filter_block1d16_h2_avg_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
diff --git a/source/libvpx/vp9/decoder/vp9_decodeframe.c b/source/libvpx/vp9/decoder/vp9_decodeframe.c
index 4e85caf..dc712f0 100644
--- a/source/libvpx/vp9/decoder/vp9_decodeframe.c
+++ b/source/libvpx/vp9/decoder/vp9_decodeframe.c
@@ -196,6 +196,64 @@
if (eob > 0) {
TX_TYPE tx_type = DCT_DCT;
tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ if (xd->lossless) {
+ tx_type = DCT_DCT;
+ vp9_highbd_iwht4x4_add(dqcoeff, dst, stride, eob, xd->bd);
+ } else {
+ const PLANE_TYPE plane_type = pd->plane_type;
+ switch (tx_size) {
+ case TX_4X4:
+ tx_type = get_tx_type_4x4(plane_type, xd, block);
+ vp9_highbd_iht4x4_add(tx_type, dqcoeff, dst, stride, eob, xd->bd);
+ break;
+ case TX_8X8:
+ tx_type = get_tx_type(plane_type, xd);
+ vp9_highbd_iht8x8_add(tx_type, dqcoeff, dst, stride, eob, xd->bd);
+ break;
+ case TX_16X16:
+ tx_type = get_tx_type(plane_type, xd);
+ vp9_highbd_iht16x16_add(tx_type, dqcoeff, dst, stride, eob, xd->bd);
+ break;
+ case TX_32X32:
+ tx_type = DCT_DCT;
+ vp9_highbd_idct32x32_add(dqcoeff, dst, stride, eob, xd->bd);
+ break;
+ default:
+ assert(0 && "Invalid transform size");
+ }
+ }
+ } else {
+ if (xd->lossless) {
+ tx_type = DCT_DCT;
+ vp9_iwht4x4_add(dqcoeff, dst, stride, eob);
+ } else {
+ const PLANE_TYPE plane_type = pd->plane_type;
+ switch (tx_size) {
+ case TX_4X4:
+ tx_type = get_tx_type_4x4(plane_type, xd, block);
+ vp9_iht4x4_add(tx_type, dqcoeff, dst, stride, eob);
+ break;
+ case TX_8X8:
+ tx_type = get_tx_type(plane_type, xd);
+ vp9_iht8x8_add(tx_type, dqcoeff, dst, stride, eob);
+ break;
+ case TX_16X16:
+ tx_type = get_tx_type(plane_type, xd);
+ vp9_iht16x16_add(tx_type, dqcoeff, dst, stride, eob);
+ break;
+ case TX_32X32:
+ tx_type = DCT_DCT;
+ vp9_idct32x32_add(dqcoeff, dst, stride, eob);
+ break;
+ default:
+ assert(0 && "Invalid transform size");
+ return;
+ }
+ }
+ }
+#else
if (xd->lossless) {
tx_type = DCT_DCT;
vp9_iwht4x4_add(dqcoeff, dst, stride, eob);
@@ -220,8 +278,10 @@
break;
default:
assert(0 && "Invalid transform size");
+ return;
}
}
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (eob == 1) {
vpx_memset(dqcoeff, 0, 2 * sizeof(dqcoeff[0]));
@@ -258,7 +318,7 @@
dst = &pd->dst.buf[4 * y * pd->dst.stride + 4 * x];
vp9_predict_intra_block(xd, block >> (tx_size << 1),
- b_width_log2(plane_bsize), tx_size, mode,
+ b_width_log2_lookup[plane_bsize], tx_size, mode,
dst, pd->dst.stride, dst, pd->dst.stride,
x, y, plane);
@@ -324,22 +384,6 @@
return &xd->mi[0].mbmi;
}
-static void set_ref(VP9_COMMON *const cm, MACROBLOCKD *const xd,
- int idx, int mi_row, int mi_col) {
- MB_MODE_INFO *const mbmi = &xd->mi[0].src_mi->mbmi;
- RefBuffer *ref_buffer = &cm->frame_refs[mbmi->ref_frame[idx] - LAST_FRAME];
- xd->block_refs[idx] = ref_buffer;
- if (!vp9_is_valid_scale(&ref_buffer->sf))
- vpx_internal_error(&cm->error, VPX_CODEC_UNSUP_BITSTREAM,
- "Invalid scale factors");
- if (ref_buffer->buf->corrupted)
- vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
- "Block reference is corrupt");
- vp9_setup_pre_planes(xd, idx, ref_buffer->buf, mi_row, mi_col,
- &ref_buffer->sf);
- xd->corrupted |= ref_buffer->buf->corrupted;
-}
-
static void decode_block(VP9_COMMON *const cm, MACROBLOCKD *const xd,
const TileInfo *const tile,
int mi_row, int mi_col,
@@ -364,11 +408,6 @@
vp9_foreach_transformed_block(xd, bsize,
predict_and_reconstruct_intra_block, &arg);
} else {
- // Setup
- set_ref(cm, xd, 0, mi_row, mi_col);
- if (has_second_ref(mbmi))
- set_ref(cm, xd, 1, mi_row, mi_col);
-
// Prediction
vp9_dec_build_inter_predictors_sb(xd, mi_row, mi_col, bsize);
@@ -592,13 +631,18 @@
update |= read_delta_q(rb, &cm->y_dc_delta_q);
update |= read_delta_q(rb, &cm->uv_dc_delta_q);
update |= read_delta_q(rb, &cm->uv_ac_delta_q);
- if (update)
+ if (update || cm->bit_depth != cm->dequant_bit_depth) {
vp9_init_dequantizer(cm);
+ cm->dequant_bit_depth = cm->bit_depth;
+ }
xd->lossless = cm->base_qindex == 0 &&
cm->y_dc_delta_q == 0 &&
cm->uv_dc_delta_q == 0 &&
cm->uv_ac_delta_q == 0;
+#if CONFIG_VP9_HIGHBITDEPTH
+ xd->bd = (int)cm->bit_depth;
+#endif
}
static INTERP_FILTER read_interp_filter(struct vp9_read_bit_buffer *rb) {
@@ -612,10 +656,8 @@
void vp9_read_frame_size(struct vp9_read_bit_buffer *rb,
int *width, int *height) {
- const int w = vp9_rb_read_literal(rb, 16) + 1;
- const int h = vp9_rb_read_literal(rb, 16) + 1;
- *width = w;
- *height = h;
+ *width = vp9_rb_read_literal(rb, 16) + 1;
+ *height = vp9_rb_read_literal(rb, 16) + 1;
}
static void setup_display_size(VP9_COMMON *cm, struct vp9_read_bit_buffer *rb) {
@@ -670,6 +712,8 @@
vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate frame buffer");
}
+ cm->frame_bufs[cm->new_fb_idx].buf.subsampling_x = cm->subsampling_x;
+ cm->frame_bufs[cm->new_fb_idx].buf.subsampling_y = cm->subsampling_y;
cm->frame_bufs[cm->new_fb_idx].buf.bit_depth = (unsigned int)cm->bit_depth;
}
@@ -703,7 +747,7 @@
if (!found)
vp9_read_frame_size(rb, &width, &height);
- if (width <=0 || height <= 0)
+ if (width <= 0 || height <= 0)
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
"Invalid frame size");
@@ -722,8 +766,8 @@
RefBuffer *const ref_frame = &cm->frame_refs[i];
if (!valid_ref_frame_img_fmt(
ref_frame->buf->bit_depth,
- ref_frame->buf->uv_crop_width < ref_frame->buf->y_crop_width,
- ref_frame->buf->uv_crop_height < ref_frame->buf->y_crop_height,
+ ref_frame->buf->subsampling_x,
+ ref_frame->buf->subsampling_y,
cm->bit_depth,
cm->subsampling_x,
cm->subsampling_y))
@@ -746,6 +790,8 @@
vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate frame buffer");
}
+ cm->frame_bufs[cm->new_fb_idx].buf.subsampling_x = cm->subsampling_x;
+ cm->frame_bufs[cm->new_fb_idx].buf.subsampling_y = cm->subsampling_y;
cm->frame_bufs[cm->new_fb_idx].buf.bit_depth = (unsigned int)cm->bit_depth;
}
@@ -1139,8 +1185,17 @@
static void read_bitdepth_colorspace_sampling(
VP9_COMMON *cm, struct vp9_read_bit_buffer *rb) {
- if (cm->profile >= PROFILE_2)
+ if (cm->profile >= PROFILE_2) {
cm->bit_depth = vp9_rb_read_bit(rb) ? VPX_BITS_12 : VPX_BITS_10;
+#if CONFIG_VP9_HIGHBITDEPTH
+ cm->use_highbitdepth = 1;
+#endif
+ } else {
+ cm->bit_depth = VPX_BITS_8;
+#if CONFIG_VP9_HIGHBITDEPTH
+ cm->use_highbitdepth = 0;
+#endif
+ }
cm->color_space = (COLOR_SPACE)vp9_rb_read_literal(rb, 3);
if (cm->color_space != SRGB) {
vp9_rb_read_bit(rb); // [16,235] (including xvycc) vs [0,255] range
@@ -1244,6 +1299,10 @@
// case (normative).
cm->color_space = BT_601;
cm->subsampling_y = cm->subsampling_x = 1;
+ cm->bit_depth = VPX_BITS_8;
+#if CONFIG_VP9_HIGHBITDEPTH
+ cm->use_highbitdepth = 0;
+#endif
}
pbi->refresh_frame_flags = vp9_rb_read_literal(rb, REF_FRAMES);
@@ -1284,6 +1343,9 @@
}
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+ get_frame_new_buffer(cm)->bit_depth = cm->bit_depth;
+#endif
if (pbi->need_resync) {
vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
diff --git a/source/libvpx/vp9/decoder/vp9_decodemv.c b/source/libvpx/vp9/decoder/vp9_decodemv.c
index ef2dc80..a01fe84 100644
--- a/source/libvpx/vp9/decoder/vp9_decodemv.c
+++ b/source/libvpx/vp9/decoder/vp9_decodemv.c
@@ -223,7 +223,6 @@
fr = vp9_read_tree(r, vp9_mv_fp_tree, class0 ? mvcomp->class0_fp[d]
: mvcomp->fp);
-
// High precision part (if hp is not used, the default value of the hp is 1)
hp = usehp ? vp9_read(r, class0 ? mvcomp->class0_hp : mvcomp->hp)
: 1;
@@ -435,11 +434,16 @@
for (ref = 0; ref < 1 + is_compound; ++ref) {
const MV_REFERENCE_FRAME frame = mbmi->ref_frame[ref];
- const int ref_idx = frame - LAST_FRAME;
- if (cm->frame_refs[ref_idx].sf.x_scale_fp == REF_INVALID_SCALE ||
- cm->frame_refs[ref_idx].sf.y_scale_fp == REF_INVALID_SCALE )
+ RefBuffer *ref_buf = &cm->frame_refs[frame - LAST_FRAME];
+ xd->block_refs[ref] = ref_buf;
+ if ((!vp9_is_valid_scale(&ref_buf->sf)))
vpx_internal_error(&cm->error, VPX_CODEC_UNSUP_BITSTREAM,
"Reference frame has invalid dimensions");
+ if (ref_buf->buf->corrupted)
+ vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME,
+ "Block reference is corrupt");
+ vp9_setup_pre_planes(xd, ref, ref_buf->buf, mi_row, mi_col,
+ &ref_buf->sf);
vp9_find_mv_refs(cm, xd, tile, mi, frame, mbmi->ref_mvs[frame],
mi_row, mi_col);
}
diff --git a/source/libvpx/vp9/decoder/vp9_decoder.c b/source/libvpx/vp9/decoder/vp9_decoder.c
index 6ee3d70..baf6ab7 100644
--- a/source/libvpx/vp9/decoder/vp9_decoder.c
+++ b/source/libvpx/vp9/decoder/vp9_decoder.c
@@ -38,7 +38,6 @@
if (!init_done) {
vp9_rtcd();
- vp9_init_neighbors();
vp9_init_intra_predictors();
init_done = 1;
}
@@ -69,6 +68,7 @@
cm->current_video_frame = 0;
pbi->ready_for_new_data = 1;
cm->bit_depth = VPX_BITS_8;
+ cm->dequant_bit_depth = VPX_BITS_8;
// vp9_init_dequantizer() is first called here. Add check in
// frame_init_dequantizer() to avoid unnecessary calling of
@@ -232,6 +232,8 @@
cm->frame_refs[0].buf->corrupted = 1;
}
+ pbi->ready_for_new_data = 0;
+
// Check if the previous frame was a frame without any references to it.
if (cm->new_fb_idx >= 0 && cm->frame_bufs[cm->new_fb_idx].ref_count == 0)
cm->release_fb_cb(cm->cb_priv,
@@ -279,8 +281,6 @@
cm->current_video_frame++;
}
- pbi->ready_for_new_data = 0;
-
cm->error.setjmp = 0;
return retcode;
}
@@ -296,12 +296,12 @@
if (pbi->ready_for_new_data == 1)
return ret;
+ pbi->ready_for_new_data = 1;
+
/* no raw frame to show!!! */
if (!cm->show_frame)
return ret;
- pbi->ready_for_new_data = 1;
-
#if CONFIG_VP9_POSTPROC
if (!cm->show_existing_frame) {
ret = vp9_post_proc_frame(cm, sd, flags);
diff --git a/source/libvpx/vp9/decoder/vp9_detokenize.c b/source/libvpx/vp9/decoder/vp9_detokenize.c
index 5778748..421229a 100644
--- a/source/libvpx/vp9/decoder/vp9_detokenize.c
+++ b/source/libvpx/vp9/decoder/vp9_detokenize.c
@@ -190,7 +190,11 @@
}
}
v = (val * dqv) >> dq_shift;
+#if CONFIG_COEFFICIENT_RANGE_CHECKING
+ dqcoeff[scan[c]] = check_range(vp9_read_bit(r) ? -v : v);
+#else
dqcoeff[scan[c]] = vp9_read_bit(r) ? -v : v;
+#endif
token_cache[scan[c]] = vp9_pt_energy_class[token];
++c;
ctx = get_coef_context(nb, token_cache, c);
diff --git a/source/libvpx/vp9/decoder/vp9_dthread.c b/source/libvpx/vp9/decoder/vp9_dthread.c
index 62ea6c1..69e4fde 100644
--- a/source/libvpx/vp9/decoder/vp9_dthread.c
+++ b/source/libvpx/vp9/decoder/vp9_dthread.c
@@ -223,14 +223,18 @@
CHECK_MEM_ERROR(cm, lf_sync->mutex_,
vpx_malloc(sizeof(*lf_sync->mutex_) * rows));
- for (i = 0; i < rows; ++i) {
- pthread_mutex_init(&lf_sync->mutex_[i], NULL);
+ if (lf_sync->mutex_) {
+ for (i = 0; i < rows; ++i) {
+ pthread_mutex_init(&lf_sync->mutex_[i], NULL);
+ }
}
CHECK_MEM_ERROR(cm, lf_sync->cond_,
vpx_malloc(sizeof(*lf_sync->cond_) * rows));
- for (i = 0; i < rows; ++i) {
- pthread_cond_init(&lf_sync->cond_[i], NULL);
+ if (lf_sync->cond_) {
+ for (i = 0; i < rows; ++i) {
+ pthread_cond_init(&lf_sync->cond_[i], NULL);
+ }
}
}
#endif // CONFIG_MULTITHREAD
diff --git a/source/libvpx/vp9/encoder/vp9_aq_variance.c b/source/libvpx/vp9/encoder/vp9_aq_variance.c
index b96f00f..7d75f09 100644
--- a/source/libvpx/vp9/encoder/vp9_aq_variance.c
+++ b/source/libvpx/vp9/encoder/vp9_aq_variance.c
@@ -34,6 +34,9 @@
#define SEGMENT_ID(i) segment_id[(i) - ENERGY_MIN]
DECLARE_ALIGNED(16, static const uint8_t, vp9_64_zeros[64]) = {0};
+#if CONFIG_VP9_HIGHBITDEPTH
+DECLARE_ALIGNED(16, static const uint16_t, vp9_highbd_64_zeros[64]) = {0};
+#endif
unsigned int vp9_vaq_segment_id(int energy) {
ENERGY_IN_BOUNDS(energy);
@@ -126,14 +129,40 @@
const int bw = 8 * num_8x8_blocks_wide_lookup[bs] - right_overflow;
const int bh = 8 * num_8x8_blocks_high_lookup[bs] - bottom_overflow;
int avg;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ highbd_variance(x->plane[0].src.buf, x->plane[0].src.stride,
+ CONVERT_TO_BYTEPTR(vp9_highbd_64_zeros), 0, bw, bh,
+ &sse, &avg);
+ sse >>= 2 * (xd->bd - 8);
+ avg >>= (xd->bd - 8);
+ } else {
+ variance(x->plane[0].src.buf, x->plane[0].src.stride,
+ vp9_64_zeros, 0, bw, bh, &sse, &avg);
+ }
+#else
variance(x->plane[0].src.buf, x->plane[0].src.stride,
vp9_64_zeros, 0, bw, bh, &sse, &avg);
+#endif // CONFIG_VP9_HIGHBITDEPTH
var = sse - (((int64_t)avg * avg) / (bw * bh));
return (256 * var) / (bw * bh);
} else {
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
+ x->plane[0].src.stride,
+ CONVERT_TO_BYTEPTR(vp9_highbd_64_zeros),
+ 0, &sse);
+ } else {
+ var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
+ x->plane[0].src.stride,
+ vp9_64_zeros, 0, &sse);
+ }
+#else
var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
x->plane[0].src.stride,
vp9_64_zeros, 0, &sse);
+#endif // CONFIG_VP9_HIGHBITDEPTH
return (256 * var) >> num_pels_log2_lookup[bs];
}
}
diff --git a/source/libvpx/vp9/encoder/vp9_avg.c b/source/libvpx/vp9/encoder/vp9_avg.c
new file mode 100644
index 0000000..e9810c8
--- /dev/null
+++ b/source/libvpx/vp9/encoder/vp9_avg.c
@@ -0,0 +1,33 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 "vp9/common/vp9_common.h"
+#include "vpx_ports/mem.h"
+
+unsigned int vp9_avg_8x8_c(const uint8_t *s, int p) {
+ int i, j;
+ int sum = 0;
+ for (i = 0; i < 8; ++i, s+=p)
+ for (j = 0; j < 8; sum += s[j], ++j) {}
+
+ return (sum + 32) >> 6;
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+unsigned int vp9_highbd_avg_8x8_c(const uint8_t *s8, int p) {
+ int i, j;
+ int sum = 0;
+ const uint16_t* s = CONVERT_TO_SHORTPTR(s8);
+ for (i = 0; i < 8; ++i, s+=p)
+ for (j = 0; j < 8; sum += s[j], ++j) {}
+
+ return (sum + 32) >> 6;
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
diff --git a/source/libvpx/vp9/encoder/vp9_bitstream.c b/source/libvpx/vp9/encoder/vp9_bitstream.c
index f658dda..421e049 100644
--- a/source/libvpx/vp9/encoder/vp9_bitstream.c
+++ b/source/libvpx/vp9/encoder/vp9_bitstream.c
@@ -120,16 +120,28 @@
}
static void pack_mb_tokens(vp9_writer *w,
- TOKENEXTRA **tp, const TOKENEXTRA *const stop) {
+ TOKENEXTRA **tp, const TOKENEXTRA *const stop,
+ vpx_bit_depth_t bit_depth) {
TOKENEXTRA *p = *tp;
while (p < stop && p->token != EOSB_TOKEN) {
const int t = p->token;
const struct vp9_token *const a = &vp9_coef_encodings[t];
- const vp9_extra_bit *const b = &vp9_extra_bits[t];
int i = 0;
int v = a->value;
int n = a->len;
+#if CONFIG_VP9_HIGHBITDEPTH
+ const vp9_extra_bit *b;
+ if (bit_depth == VPX_BITS_12)
+ b = &vp9_extra_bits_high12[t];
+ else if (bit_depth == VPX_BITS_10)
+ b = &vp9_extra_bits_high10[t];
+ else
+ b = &vp9_extra_bits[t];
+#else
+ const vp9_extra_bit *const b = &vp9_extra_bits[t];
+ (void) bit_depth;
+#endif // CONFIG_VP9_HIGHBITDEPTH
/* skip one or two nodes */
if (p->skip_eob_node) {
@@ -387,7 +399,7 @@
}
assert(*tok < tok_end);
- pack_mb_tokens(w, tok, tok_end);
+ pack_mb_tokens(w, tok, tok_end, cm->bit_depth);
}
static void write_partition(const VP9_COMMON *const cm,
@@ -419,7 +431,7 @@
const VP9_COMMON *const cm = &cpi->common;
MACROBLOCKD *const xd = &cpi->mb.e_mbd;
- const int bsl = b_width_log2(bsize);
+ const int bsl = b_width_log2_lookup[bsize];
const int bs = (1 << bsl) / 4;
PARTITION_TYPE partition;
BLOCK_SIZE subsize;
@@ -923,26 +935,27 @@
size_t total_size = 0;
const int tile_cols = 1 << cm->log2_tile_cols;
const int tile_rows = 1 << cm->log2_tile_rows;
+ TileInfo tile[4][1 << 6];
+ TOKENEXTRA *pre_tok = cpi->tok;
+ int tile_tok = 0;
vpx_memset(cm->above_seg_context, 0, sizeof(*cm->above_seg_context) *
mi_cols_aligned_to_sb(cm->mi_cols));
- tok[0][0] = cpi->tok;
- for (tile_row = 0; tile_row < tile_rows; tile_row++) {
- if (tile_row)
- tok[tile_row][0] = tok[tile_row - 1][tile_cols - 1] +
- cpi->tok_count[tile_row - 1][tile_cols - 1];
+ for (tile_row = 0; tile_row < tile_rows; ++tile_row) {
+ for (tile_col = 0; tile_col < tile_cols; ++tile_col) {
+ vp9_tile_init(&tile[tile_row][tile_col], cm, tile_row, tile_col);
- for (tile_col = 1; tile_col < tile_cols; tile_col++)
- tok[tile_row][tile_col] = tok[tile_row][tile_col - 1] +
- cpi->tok_count[tile_row][tile_col - 1];
+ tok[tile_row][tile_col] = pre_tok + tile_tok;
+ pre_tok = tok[tile_row][tile_col];
+ tile_tok = allocated_tokens(tile[tile_row][tile_col]);
+ }
}
for (tile_row = 0; tile_row < tile_rows; tile_row++) {
for (tile_col = 0; tile_col < tile_cols; tile_col++) {
- TileInfo tile;
+ const TileInfo * const ptile = &tile[tile_row][tile_col];
- vp9_tile_init(&tile, cm, tile_row, tile_col);
tok_end = tok[tile_row][tile_col] + cpi->tok_count[tile_row][tile_col];
if (tile_col < tile_cols - 1 || tile_row < tile_rows - 1)
@@ -950,7 +963,7 @@
else
vp9_start_encode(&residual_bc, data_ptr + total_size);
- write_modes(cpi, &tile, &residual_bc, &tok[tile_row][tile_col], tok_end);
+ write_modes(cpi, ptile, &residual_bc, &tok[tile_row][tile_col], tok_end);
assert(tok[tile_row][tile_col] == tok_end);
vp9_stop_encode(&residual_bc);
if (tile_col < tile_cols - 1 || tile_row < tile_rows - 1) {
@@ -1001,7 +1014,11 @@
((cpi->svc.number_temporal_layers > 1 &&
cpi->oxcf.rc_mode == VPX_CBR) ||
(cpi->svc.number_spatial_layers > 1 &&
- cpi->svc.layer_context[cpi->svc.spatial_layer_id].is_key_frame))) {
+ cpi->svc.layer_context[cpi->svc.spatial_layer_id].is_key_frame) ||
+ (is_two_pass_svc(cpi) &&
+ cpi->svc.encode_empty_frame_state == ENCODING &&
+ cpi->svc.layer_context[0].frames_from_key_frame <
+ cpi->svc.number_temporal_layers + 1))) {
found = 0;
}
vp9_wb_write_bit(wb, found);
@@ -1093,8 +1110,7 @@
// will change to show_frame flag to 0, then add an one byte frame with
// show_existing_frame flag which tells the decoder which frame we want to
// show.
- if (!cm->show_frame ||
- (is_two_pass_svc(cpi) && cm->error_resilient_mode == 0))
+ if (!cm->show_frame)
vp9_wb_write_bit(wb, cm->intra_only);
if (!cm->error_resilient_mode)
diff --git a/source/libvpx/vp9/encoder/vp9_block.h b/source/libvpx/vp9/encoder/vp9_block.h
index 767bd7f..5194c4c 100644
--- a/source/libvpx/vp9/encoder/vp9_block.h
+++ b/source/libvpx/vp9/encoder/vp9_block.h
@@ -13,8 +13,6 @@
#include "vp9/common/vp9_entropymv.h"
#include "vp9/common/vp9_entropy.h"
-#include "vpx_ports/mem.h"
-#include "vp9/common/vp9_onyxc_int.h"
#ifdef __cplusplus
extern "C" {
@@ -122,8 +120,8 @@
void (*fwd_txm4x4)(const int16_t *input, tran_low_t *output, int stride);
void (*itxm_add)(const tran_low_t *input, uint8_t *dest, int stride, int eob);
#if CONFIG_VP9_HIGHBITDEPTH
- void (*high_itxm_add)(const tran_low_t *input, uint8_t *dest, int stride,
- int eob, int bd);
+ void (*highbd_itxm_add)(const tran_low_t *input, uint8_t *dest, int stride,
+ int eob, int bd);
#endif
};
diff --git a/source/libvpx/vp9/encoder/vp9_context_tree.h b/source/libvpx/vp9/encoder/vp9_context_tree.h
index 97f0741..6b28ee5 100644
--- a/source/libvpx/vp9/encoder/vp9_context_tree.h
+++ b/source/libvpx/vp9/encoder/vp9_context_tree.h
@@ -11,9 +11,10 @@
#ifndef VP9_ENCODER_VP9_CONTEXT_TREE_H_
#define VP9_ENCODER_VP9_CONTEXT_TREE_H_
-#include "vp9/common/vp9_onyxc_int.h"
+#include "vp9/common/vp9_blockd.h"
struct VP9_COMP;
+struct VP9Common;
// Structure to hold snapshot of coding context during the mode picking process
typedef struct {
diff --git a/source/libvpx/vp9/encoder/vp9_dct.c b/source/libvpx/vp9/encoder/vp9_dct.c
index eff8996..1090d04 100644
--- a/source/libvpx/vp9/encoder/vp9_dct.c
+++ b/source/libvpx/vp9/encoder/vp9_dct.c
@@ -1440,59 +1440,62 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_fdct4x4_c(const int16_t *input, tran_low_t *output, int stride) {
+void vp9_highbd_fdct4x4_c(const int16_t *input, tran_low_t *output,
+ int stride) {
vp9_fdct4x4_c(input, output, stride);
}
-void vp9_high_fht4x4_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp9_highbd_fht4x4_c(const int16_t *input, tran_low_t *output,
+ int stride, int tx_type) {
vp9_fht4x4_c(input, output, stride, tx_type);
}
-void vp9_high_fdct8x8_1_c(const int16_t *input, tran_low_t *final_output,
- int stride) {
+void vp9_highbd_fdct8x8_1_c(const int16_t *input, tran_low_t *final_output,
+ int stride) {
vp9_fdct8x8_1_c(input, final_output, stride);
}
-void vp9_high_fdct8x8_c(const int16_t *input, tran_low_t *final_output,
- int stride) {
+void vp9_highbd_fdct8x8_c(const int16_t *input, tran_low_t *final_output,
+ int stride) {
vp9_fdct8x8_c(input, final_output, stride);
}
-void vp9_high_fdct16x16_1_c(const int16_t *input, tran_low_t *output,
- int stride) {
+void vp9_highbd_fdct16x16_1_c(const int16_t *input, tran_low_t *output,
+ int stride) {
vp9_fdct16x16_1_c(input, output, stride);
}
-void vp9_high_fdct16x16_c(const int16_t *input, tran_low_t *output,
- int stride) {
+void vp9_highbd_fdct16x16_c(const int16_t *input, tran_low_t *output,
+ int stride) {
vp9_fdct16x16_c(input, output, stride);
}
-void vp9_high_fht8x8_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp9_highbd_fht8x8_c(const int16_t *input, tran_low_t *output,
+ int stride, int tx_type) {
vp9_fht8x8_c(input, output, stride, tx_type);
}
-void vp9_high_fwht4x4_c(const int16_t *input, tran_low_t *output, int stride) {
+void vp9_highbd_fwht4x4_c(const int16_t *input, tran_low_t *output,
+ int stride) {
vp9_fwht4x4_c(input, output, stride);
}
-void vp9_high_fht16x16_c(const int16_t *input, tran_low_t *output,
- int stride, int tx_type) {
+void vp9_highbd_fht16x16_c(const int16_t *input, tran_low_t *output,
+ int stride, int tx_type) {
vp9_fht16x16_c(input, output, stride, tx_type);
}
-void vp9_high_fdct32x32_1_c(const int16_t *input, tran_low_t *out, int stride) {
+void vp9_highbd_fdct32x32_1_c(const int16_t *input, tran_low_t *out,
+ int stride) {
vp9_fdct32x32_1_c(input, out, stride);
}
-void vp9_high_fdct32x32_c(const int16_t *input, tran_low_t *out, int stride) {
+void vp9_highbd_fdct32x32_c(const int16_t *input, tran_low_t *out, int stride) {
vp9_fdct32x32_c(input, out, stride);
}
-void vp9_high_fdct32x32_rd_c(const int16_t *input, tran_low_t *out,
- int stride) {
+void vp9_highbd_fdct32x32_rd_c(const int16_t *input, tran_low_t *out,
+ int stride) {
vp9_fdct32x32_rd_c(input, out, stride);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/source/libvpx/vp9/encoder/vp9_denoiser.c b/source/libvpx/vp9/encoder/vp9_denoiser.c
index 681b2a5..4deeed2 100644
--- a/source/libvpx/vp9/encoder/vp9_denoiser.c
+++ b/source/libvpx/vp9/encoder/vp9_denoiser.c
@@ -31,9 +31,6 @@
static void make_grayscale(YV12_BUFFER_CONFIG *yuv);
#endif
-static const int widths[] = {4, 4, 8, 8, 8, 16, 16, 16, 32, 32, 32, 64, 64};
-static const int heights[] = {4, 8, 4, 8, 16, 8, 16, 32, 16, 32, 64, 32, 64};
-
static int absdiff_thresh(BLOCK_SIZE bs, int increase_denoising) {
(void)bs;
return 3 + (increase_denoising ? 1 : 0);
@@ -52,7 +49,9 @@
}
static unsigned int sse_thresh(BLOCK_SIZE bs, int increase_denoising) {
- return widths[bs] * heights[bs] * (increase_denoising ? 60 : 40);
+ return (4 << b_width_log2_lookup[bs]) *
+ (4 << b_height_log2_lookup[bs]) *
+ (increase_denoising ? 60 : 40);
}
static int sse_diff_thresh(BLOCK_SIZE bs, int increase_denoising,
@@ -61,25 +60,31 @@
noise_motion_thresh(bs, increase_denoising)) {
return 0;
} else {
- return widths[bs] * heights[bs] * 20;
+ return (4 << b_width_log2_lookup[bs]) *
+ (4 << b_height_log2_lookup[bs]) * 20;
}
}
-static int total_adj_strong_thresh(BLOCK_SIZE bs, int increase_denoising) {
- return widths[bs] * heights[bs] * (increase_denoising ? 3 : 2);
+int total_adj_strong_thresh(BLOCK_SIZE bs, int increase_denoising) {
+ return (4 << b_width_log2_lookup[bs]) *
+ (4 << b_height_log2_lookup[bs]) * (increase_denoising ? 3 : 2);
}
static int total_adj_weak_thresh(BLOCK_SIZE bs, int increase_denoising) {
- return widths[bs] * heights[bs] * (increase_denoising ? 3 : 2);
+ return (4 << b_width_log2_lookup[bs]) *
+ (4 << b_height_log2_lookup[bs]) * (increase_denoising ? 3 : 2);
}
-static VP9_DENOISER_DECISION denoiser_filter(const uint8_t *sig, int sig_stride,
- const uint8_t *mc_avg,
- int mc_avg_stride,
- uint8_t *avg, int avg_stride,
- int increase_denoising,
- BLOCK_SIZE bs,
- int motion_magnitude) {
+// TODO(jackychen): If increase_denoising is enabled in the future,
+// we might need to update the code for calculating 'total_adj' in
+// case the C code is not bit-exact with corresponding sse2 code.
+int vp9_denoiser_filter_c(const uint8_t *sig, int sig_stride,
+ const uint8_t *mc_avg,
+ int mc_avg_stride,
+ uint8_t *avg, int avg_stride,
+ int increase_denoising,
+ BLOCK_SIZE bs,
+ int motion_magnitude) {
int r, c;
const uint8_t *sig_start = sig;
const uint8_t *mc_avg_start = mc_avg;
@@ -102,8 +107,8 @@
}
// First attempt to apply a strong temporal denoising filter.
- for (r = 0; r < heights[bs]; ++r) {
- for (c = 0; c < widths[bs]; ++c) {
+ for (r = 0; r < (4 << b_height_log2_lookup[bs]); ++r) {
+ for (c = 0; c < (4 << b_width_log2_lookup[bs]); ++c) {
diff = mc_avg[c] - sig[c];
absdiff = abs(diff);
@@ -143,7 +148,7 @@
// Otherwise, we try to dampen the filter if the delta is not too high.
delta = ((abs(total_adj) - total_adj_strong_thresh(bs, increase_denoising))
- >> 8) + 1;
+ >> num_pels_log2_lookup[bs]) + 1;
if (delta >= delta_thresh(bs, increase_denoising)) {
return COPY_BLOCK;
@@ -152,8 +157,8 @@
mc_avg = mc_avg_start;
avg = avg_start;
sig = sig_start;
- for (r = 0; r < heights[bs]; ++r) {
- for (c = 0; c < widths[bs]; ++c) {
+ for (r = 0; r < (4 << b_height_log2_lookup[bs]); ++r) {
+ for (c = 0; c < (4 << b_width_log2_lookup[bs]); ++c) {
diff = mc_avg[c] - sig[c];
adj = abs(diff);
if (adj > delta) {
@@ -193,8 +198,8 @@
static void copy_block(uint8_t *dest, int dest_stride,
const uint8_t *src, int src_stride, BLOCK_SIZE bs) {
int r;
- for (r = 0; r < heights[bs]; ++r) {
- vpx_memcpy(dest, src, widths[bs]);
+ for (r = 0; r < (4 << b_height_log2_lookup[bs]); ++r) {
+ vpx_memcpy(dest, src, (4 << b_width_log2_lookup[bs]));
dest += dest_stride;
src += src_stride;
}
@@ -336,10 +341,10 @@
&motion_magnitude);
if (decision == FILTER_BLOCK) {
- decision = denoiser_filter(src.buf, src.stride,
- mc_avg_start, mc_avg.y_stride,
- avg_start, avg.y_stride,
- 0, bs, motion_magnitude);
+ decision = vp9_denoiser_filter(src.buf, src.stride,
+ mc_avg_start, mc_avg.y_stride,
+ avg_start, avg.y_stride,
+ 0, bs, motion_magnitude);
}
if (decision == FILTER_BLOCK) {
diff --git a/source/libvpx/vp9/encoder/vp9_denoiser.h b/source/libvpx/vp9/encoder/vp9_denoiser.h
index fa714b1..421dfcd 100644
--- a/source/libvpx/vp9/encoder/vp9_denoiser.h
+++ b/source/libvpx/vp9/encoder/vp9_denoiser.h
@@ -55,6 +55,10 @@
#endif
int border);
+#if CONFIG_VP9_TEMPORAL_DENOISING
+int total_adj_strong_thresh(BLOCK_SIZE bs, int increase_denoising);
+#endif
+
void vp9_denoiser_free(VP9_DENOISER *denoiser);
#ifdef __cplusplus
diff --git a/source/libvpx/vp9/encoder/vp9_encodeframe.c b/source/libvpx/vp9/encoder/vp9_encodeframe.c
index be5ee7b..6eff8c5 100644
--- a/source/libvpx/vp9/encoder/vp9_encodeframe.c
+++ b/source/libvpx/vp9/encoder/vp9_encodeframe.c
@@ -61,16 +61,51 @@
// Eventually this should be replaced by custom no-reference routines,
// which will be faster.
static const uint8_t VP9_VAR_OFFS[64] = {
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128,
- 128, 128, 128, 128, 128, 128, 128, 128
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128
};
+#if CONFIG_VP9_HIGHBITDEPTH
+static const uint16_t VP9_HIGH_VAR_OFFS_8[64] = {
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128,
+ 128, 128, 128, 128, 128, 128, 128, 128
+};
+
+static const uint16_t VP9_HIGH_VAR_OFFS_10[64] = {
+ 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
+ 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
+ 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
+ 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
+ 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
+ 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
+ 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4,
+ 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4, 128*4
+};
+
+static const uint16_t VP9_HIGH_VAR_OFFS_12[64] = {
+ 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
+ 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
+ 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
+ 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
+ 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
+ 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
+ 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16,
+ 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16, 128*16
+};
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static unsigned int get_sby_perpixel_variance(VP9_COMP *cpi,
const struct buf_2d *ref,
BLOCK_SIZE bs) {
@@ -80,6 +115,32 @@
return ROUND_POWER_OF_TWO(var, num_pels_log2_lookup[bs]);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static unsigned int high_get_sby_perpixel_variance(
+ VP9_COMP *cpi, const struct buf_2d *ref, BLOCK_SIZE bs, int bd) {
+ unsigned int var, sse;
+ switch (bd) {
+ case 10:
+ var = cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
+ CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_10),
+ 0, &sse);
+ break;
+ case 12:
+ var = cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
+ CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_12),
+ 0, &sse);
+ break;
+ case 8:
+ default:
+ var = cpi->fn_ptr[bs].vf(ref->buf, ref->stride,
+ CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_8),
+ 0, &sse);
+ break;
+ }
+ return ROUND_POWER_OF_TWO(var, num_pels_log2_lookup[bs]);
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static unsigned int get_sby_perpixel_diff_variance(VP9_COMP *cpi,
const struct buf_2d *ref,
int mi_row, int mi_col,
@@ -335,10 +396,11 @@
const int block_width = num_8x8_blocks_wide_lookup[bsize];
const int block_height = num_8x8_blocks_high_lookup[bsize];
// TODO(debargha): Choose this more intelligently.
- const int64_t threshold_multiplier = 25;
- int64_t threshold = threshold_multiplier * cpi->common.base_qindex;
+ const int threshold_multiplier = cm->frame_type == KEY_FRAME ? 64 : 4;
+ int64_t threshold =
+ (int64_t)(threshold_multiplier *
+ vp9_convert_qindex_to_q(cm->base_qindex, cm->bit_depth));
assert(block_height == block_width);
-
tree_to_node(data, bsize, &vt);
// Split none is available only if we have more than half a block size
@@ -350,29 +412,47 @@
return 1;
}
- // Vertical split is available on all but the bottom border.
- if (mi_row + block_height / 2 < cm->mi_rows &&
- vt.part_variances->vert[0].variance < threshold &&
- vt.part_variances->vert[1].variance < threshold) {
- BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_VERT);
- set_block_size(cpi, mi_row, mi_col, subsize);
- set_block_size(cpi, mi_row, mi_col + block_width / 2, subsize);
- return 1;
+ // Only allow split for blocks above 16x16.
+ if (bsize > BLOCK_16X16) {
+ // Vertical split is available on all but the bottom border.
+ if (mi_row + block_height / 2 < cm->mi_rows &&
+ vt.part_variances->vert[0].variance < threshold &&
+ vt.part_variances->vert[1].variance < threshold) {
+ BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_VERT);
+ set_block_size(cpi, mi_row, mi_col, subsize);
+ set_block_size(cpi, mi_row, mi_col + block_width / 2, subsize);
+ return 1;
+ }
+
+ // Horizontal split is available on all but the right border.
+ if (mi_col + block_width / 2 < cm->mi_cols &&
+ vt.part_variances->horz[0].variance < threshold &&
+ vt.part_variances->horz[1].variance < threshold) {
+ BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_HORZ);
+ set_block_size(cpi, mi_row, mi_col, subsize);
+ set_block_size(cpi, mi_row + block_height / 2, mi_col, subsize);
+ return 1;
+ }
}
- // Horizontal split is available on all but the right border.
- if (mi_col + block_width / 2 < cm->mi_cols &&
- vt.part_variances->horz[0].variance < threshold &&
- vt.part_variances->horz[1].variance < threshold) {
- BLOCK_SIZE subsize = get_subsize(bsize, PARTITION_HORZ);
- set_block_size(cpi, mi_row, mi_col, subsize);
- set_block_size(cpi, mi_row + block_height / 2, mi_col, subsize);
- return 1;
+ // This will only allow 8x8 if the 16x16 variance is very large.
+ if (bsize == BLOCK_16X16) {
+ if (mi_col + block_width / 2 < cm->mi_cols &&
+ mi_row + block_height / 2 < cm->mi_rows &&
+ vt.part_variances->none.variance < (threshold << 6)) {
+ set_block_size(cpi, mi_row, mi_col, bsize);
+ return 1;
+ }
}
return 0;
}
-// TODO(debargha): Fix this function and make it work as expected.
+// This function chooses partitioning based on the variance
+// between source and reconstructed last, where variance is
+// computed for 8x8 downsampled inputs. Some things to check:
+// using the last source rather than reconstructed last, and
+// allowing for small downsampling (4x4 or 2x2) for selection
+// of smaller block sizes (i.e., < 16x16).
static void choose_partitioning(VP9_COMP *cpi,
const TileInfo *const tile,
int mi_row, int mi_col) {
@@ -391,6 +471,7 @@
const YV12_BUFFER_CONFIG *yv12 = get_ref_frame_buffer(cpi, LAST_FRAME);
const struct scale_factors *const sf = &cm->frame_refs[LAST_FRAME - 1].sf;
+ vp9_clear_system_state();
vp9_zero(vt);
set_offsets(cpi, tile, mi_row, mi_col, BLOCK_64X64);
@@ -419,6 +500,22 @@
} else {
d = VP9_VAR_OFFS;
dp = 0;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ switch (xd->bd) {
+ case 10:
+ d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_10);
+ break;
+ case 12:
+ d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_12);
+ break;
+ case 8:
+ default:
+ d = CONVERT_TO_BYTEPTR(VP9_HIGH_VAR_OFFS_8);
+ break;
+ }
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
// Fill in the entire tree of 8x8 variances for splits.
@@ -434,10 +531,28 @@
int y_idx = y16_idx + ((k >> 1) << 3);
unsigned int sse = 0;
int sum = 0;
- if (x_idx < pixels_wide && y_idx < pixels_high)
- vp9_get8x8var(s + y_idx * sp + x_idx, sp,
- d + y_idx * dp + x_idx, dp, &sse, &sum);
- fill_variance(sse, sum, 64, &vst->split[k].part_variances.none);
+
+ if (x_idx < pixels_wide && y_idx < pixels_high) {
+ int s_avg, d_avg;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ s_avg = vp9_highbd_avg_8x8(s + y_idx * sp + x_idx, sp);
+ d_avg = vp9_highbd_avg_8x8(d + y_idx * dp + x_idx, dp);
+ } else {
+ s_avg = vp9_avg_8x8(s + y_idx * sp + x_idx, sp);
+ d_avg = vp9_avg_8x8(d + y_idx * dp + x_idx, dp);
+ }
+#else
+ s_avg = vp9_avg_8x8(s + y_idx * sp + x_idx, sp);
+ d_avg = vp9_avg_8x8(d + y_idx * dp + x_idx, dp);
+#endif
+ sum = s_avg - d_avg;
+ sse = sum * sum;
+ }
+ // For an 8x8 block we have just one value the average of all 64
+ // pixels, so use 1. This means of course that there is no variance
+ // in an 8x8 block.
+ fill_variance(sse, sum, 1, &vst->split[k].part_variances.none);
}
}
}
@@ -453,8 +568,8 @@
// Now go through the entire structure, splitting every block size until
// we get to one that's got a variance lower than our threshold, or we
// hit 8x8.
- if (!set_vt_partitioning(cpi, &vt, BLOCK_64X64,
- mi_row, mi_col)) {
+ if ( mi_col + 8 > cm->mi_cols || mi_row + 8 > cm->mi_rows ||
+ !set_vt_partitioning(cpi, &vt, BLOCK_64X64, mi_row, mi_col)) {
for (i = 0; i < 4; ++i) {
const int x32_idx = ((i & 1) << 2);
const int y32_idx = ((i >> 1) << 2);
@@ -463,31 +578,15 @@
for (j = 0; j < 4; ++j) {
const int x16_idx = ((j & 1) << 1);
const int y16_idx = ((j >> 1) << 1);
- // NOTE: This is a temporary hack to disable 8x8 partitions,
- // since it works really bad - possibly due to a bug
-#define DISABLE_8X8_VAR_BASED_PARTITION
-#ifdef DISABLE_8X8_VAR_BASED_PARTITION
- if (mi_row + y32_idx + y16_idx + 1 < cm->mi_rows &&
- mi_row + x32_idx + x16_idx + 1 < cm->mi_cols) {
- set_block_size(cpi,
- (mi_row + y32_idx + y16_idx),
- (mi_col + x32_idx + x16_idx),
- BLOCK_16X16);
- } else {
- for (k = 0; k < 4; ++k) {
- const int x8_idx = (k & 1);
- const int y8_idx = (k >> 1);
- set_block_size(cpi,
- (mi_row + y32_idx + y16_idx + y8_idx),
- (mi_col + x32_idx + x16_idx + x8_idx),
- BLOCK_8X8);
- }
- }
-#else
- if (!set_vt_partitioning(cpi, &vt.split[i].split[j], tile,
+ // NOTE: Since this uses 8x8 downsampling for variance calculation
+ // we cannot really select block size 8x8 (or even 8x16/16x8),
+ // since we do not sufficient samples for variance.
+ // For now, 8x8 partition is only set if the variance of the 16x16
+ // block is very high. This is controlled in set_vt_partitioning.
+ if (!set_vt_partitioning(cpi, &vt.split[i].split[j],
BLOCK_16X16,
- (mi_row + y32_idx + y16_idx),
- (mi_col + x32_idx + x16_idx), 2)) {
+ mi_row + y32_idx + y16_idx,
+ mi_col + x32_idx + x16_idx)) {
for (k = 0; k < 4; ++k) {
const int x8_idx = (k & 1);
const int y8_idx = (k >> 1);
@@ -497,7 +596,6 @@
BLOCK_8X8);
}
}
-#endif
}
}
}
@@ -684,10 +782,9 @@
}
static void rd_pick_sb_modes(VP9_COMP *cpi, const TileInfo *const tile,
- int mi_row, int mi_col,
- int *totalrate, int64_t *totaldist,
+ int mi_row, int mi_col, RD_COST *rd_cost,
BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
- int64_t best_rd, int block) {
+ int64_t best_rd) {
VP9_COMMON *const cm = &cpi->common;
MACROBLOCK *const x = &cpi->mb;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -704,19 +801,6 @@
// Use the lower precision, but faster, 32x32 fdct for mode selection.
x->use_lp32x32fdct = 1;
- // TODO(JBB): Most other places in the code instead of calling the function
- // and then checking if its not the first 8x8 we put the check in the
- // calling function. Do that here.
- if (bsize < BLOCK_8X8) {
- // When ab_index = 0 all sub-blocks are handled, so for ab_index != 0
- // there is nothing to be done.
- if (block != 0) {
- *totalrate = 0;
- *totaldist = 0;
- return;
- }
- }
-
set_offsets(cpi, tile, mi_row, mi_col, bsize);
mbmi = &xd->mi[0].src_mi->mbmi;
mbmi->sb_type = bsize;
@@ -734,7 +818,17 @@
// Set to zero to make sure we do not use the previous encoded frame stats
mbmi->skip = 0;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ x->source_variance =
+ high_get_sby_perpixel_variance(cpi, &x->plane[0].src, bsize, xd->bd);
+ } else {
+ x->source_variance =
+ get_sby_perpixel_variance(cpi, &x->plane[0].src, bsize);
+ }
+#else
x->source_variance = get_sby_perpixel_variance(cpi, &x->plane[0].src, bsize);
+#endif // CONFIG_VP9_HIGHBITDEPTH
// Save rdmult before it might be changed, so it can be restored later.
orig_rdmult = x->rdmult;
@@ -774,28 +868,33 @@
// Find best coding mode & reconstruct the MB so it is available
// as a predictor for MBs that follow in the SB
if (frame_is_intra_only(cm)) {
- vp9_rd_pick_intra_mode_sb(cpi, x, totalrate, totaldist, bsize, ctx,
- best_rd);
+ vp9_rd_pick_intra_mode_sb(cpi, x, rd_cost, bsize, ctx, best_rd);
} else {
if (bsize >= BLOCK_8X8) {
if (vp9_segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP))
- vp9_rd_pick_inter_mode_sb_seg_skip(cpi, x, totalrate, totaldist, bsize,
+ vp9_rd_pick_inter_mode_sb_seg_skip(cpi, x, rd_cost, bsize,
ctx, best_rd);
else
vp9_rd_pick_inter_mode_sb(cpi, x, tile, mi_row, mi_col,
- totalrate, totaldist, bsize, ctx, best_rd);
+ rd_cost, bsize, ctx, best_rd);
} else {
- vp9_rd_pick_inter_mode_sub8x8(cpi, x, tile, mi_row, mi_col, totalrate,
- totaldist, bsize, ctx, best_rd);
+ vp9_rd_pick_inter_mode_sub8x8(cpi, x, tile, mi_row, mi_col, rd_cost,
+ bsize, ctx, best_rd);
}
}
+ if (aq_mode == VARIANCE_AQ && rd_cost->rate != INT_MAX) {
+ vp9_clear_system_state();
+ rd_cost->rate = (int)round(rd_cost->rate * rdmult_ratio);
+ rd_cost->rdcost = RDCOST(x->rdmult, x->rddiv, rd_cost->rate, rd_cost->dist);
+ }
+
x->rdmult = orig_rdmult;
- if (aq_mode == VARIANCE_AQ && *totalrate != INT_MAX) {
- vp9_clear_system_state();
- *totalrate = (int)round(*totalrate * rdmult_ratio);
- }
+ // TODO(jingning) The rate-distortion optimization flow needs to be
+ // refactored to provide proper exit/return handle.
+ if (rd_cost->rate == INT_MAX)
+ rd_cost->rdcost = INT64_MAX;
}
static void update_stats(VP9_COMMON *cm, const MACROBLOCK *x) {
@@ -925,7 +1024,7 @@
MACROBLOCK *const x = &cpi->mb;
MACROBLOCKD *const xd = &x->e_mbd;
- const int bsl = b_width_log2(bsize), hbs = (1 << bsl) / 4;
+ const int bsl = b_width_log2_lookup[bsize], hbs = (1 << bsl) / 4;
int ctx;
PARTITION_TYPE partition;
BLOCK_SIZE subsize = bsize;
@@ -1297,12 +1396,18 @@
*(xd->mi[0].src_mi) = ctx->mic;
xd->mi[0].src_mi = &xd->mi[0];
-
- // For in frame adaptive Q, check for reseting the segment_id and updating
- // the cyclic refresh map.
- if ((cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ) && seg->enabled) {
- vp9_cyclic_refresh_update_segment(cpi, &xd->mi[0].src_mi->mbmi,
- mi_row, mi_col, bsize, 1);
+ if (seg->enabled && cpi->oxcf.aq_mode) {
+ // For in frame complexity AQ or variance AQ, copy segment_id from
+ // segmentation_map.
+ if (cpi->oxcf.aq_mode == COMPLEXITY_AQ ||
+ cpi->oxcf.aq_mode == VARIANCE_AQ ) {
+ const uint8_t *const map = seg->update_map ? cpi->segmentation_map
+ : cm->last_frame_seg_map;
+ mbmi->segment_id = vp9_get_segment_id(cm, map, bsize, mi_row, mi_col);
+ } else {
+ // Setting segmentation map for cyclic_refresh
+ vp9_cyclic_refresh_update_segment(cpi, mbmi, mi_row, mi_col, bsize, 1);
+ }
vp9_init_plane_quantizers(cpi, x);
}
@@ -1348,7 +1453,7 @@
MACROBLOCK *const x = &cpi->mb;
MACROBLOCKD *const xd = &x->e_mbd;
- const int bsl = b_width_log2(bsize), hbs = (1 << bsl) / 4;
+ const int bsl = b_width_log2_lookup[bsize], hbs = (1 << bsl) / 4;
int ctx;
PARTITION_TYPE partition;
BLOCK_SIZE subsize;
@@ -1411,17 +1516,16 @@
update_partition_context(xd, mi_row, mi_col, subsize, bsize);
}
-static void rd_use_partition(VP9_COMP *cpi,
- const TileInfo *const tile,
- MODE_INFO *mi_8x8,
- TOKENEXTRA **tp, int mi_row, int mi_col,
+static void rd_use_partition(VP9_COMP *cpi, const TileInfo *const tile,
+ MODE_INFO *mi_8x8, TOKENEXTRA **tp,
+ int mi_row, int mi_col,
BLOCK_SIZE bsize, int *rate, int64_t *dist,
int do_recon, PC_TREE *pc_tree) {
VP9_COMMON *const cm = &cpi->common;
MACROBLOCK *const x = &cpi->mb;
MACROBLOCKD *const xd = &x->e_mbd;
const int mis = cm->mi_stride;
- const int bsl = b_width_log2(bsize);
+ const int bsl = b_width_log2_lookup[bsize];
const int mi_step = num_4x4_blocks_wide_lookup[bsize] / 2;
const int bss = (1 << bsl) / 4;
int i, pl;
@@ -1429,15 +1533,7 @@
BLOCK_SIZE subsize;
ENTROPY_CONTEXT l[16 * MAX_MB_PLANE], a[16 * MAX_MB_PLANE];
PARTITION_CONTEXT sl[8], sa[8];
- int last_part_rate = INT_MAX;
- int64_t last_part_dist = INT64_MAX;
- int64_t last_part_rd = INT64_MAX;
- int none_rate = INT_MAX;
- int64_t none_dist = INT64_MAX;
- int64_t none_rd = INT64_MAX;
- int chosen_rate = INT_MAX;
- int64_t chosen_dist = INT64_MAX;
- int64_t chosen_rd = INT64_MAX;
+ RD_COST last_part_rdc, none_rdc, chosen_rdc;
BLOCK_SIZE sub_subsize = BLOCK_4X4;
int splits_below = 0;
BLOCK_SIZE bs_type = mi_8x8[0].src_mi->mbmi.sb_type;
@@ -1450,6 +1546,10 @@
assert(num_4x4_blocks_wide_lookup[bsize] ==
num_4x4_blocks_high_lookup[bsize]);
+ vp9_rd_cost_reset(&last_part_rdc);
+ vp9_rd_cost_reset(&none_rdc);
+ vp9_rd_cost_reset(&chosen_rdc);
+
partition = partition_lookup[bsl][bs_type];
subsize = get_subsize(bsize, partition);
@@ -1483,14 +1583,15 @@
mi_row + (mi_step >> 1) < cm->mi_rows &&
mi_col + (mi_step >> 1) < cm->mi_cols) {
pc_tree->partitioning = PARTITION_NONE;
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &none_rate, &none_dist, bsize,
- ctx, INT64_MAX, 0);
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &none_rdc, bsize,
+ ctx, INT64_MAX);
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- if (none_rate < INT_MAX) {
- none_rate += cpi->partition_cost[pl][PARTITION_NONE];
- none_rd = RDCOST(x->rdmult, x->rddiv, none_rate, none_dist);
+ if (none_rdc.rate < INT_MAX) {
+ none_rdc.rate += cpi->partition_cost[pl][PARTITION_NONE];
+ none_rdc.rdcost = RDCOST(x->rdmult, x->rddiv, none_rdc.rate,
+ none_rdc.dist);
}
restore_context(cpi, mi_row, mi_col, a, l, sa, sl, bsize);
@@ -1501,84 +1602,81 @@
switch (partition) {
case PARTITION_NONE:
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &last_part_rate,
- &last_part_dist, bsize, ctx, INT64_MAX, 0);
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &last_part_rdc,
+ bsize, ctx, INT64_MAX);
break;
case PARTITION_HORZ:
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &last_part_rate,
- &last_part_dist, subsize, &pc_tree->horizontal[0],
- INT64_MAX, 0);
- if (last_part_rate != INT_MAX &&
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &last_part_rdc,
+ subsize, &pc_tree->horizontal[0],
+ INT64_MAX);
+ if (last_part_rdc.rate != INT_MAX &&
bsize >= BLOCK_8X8 && mi_row + (mi_step >> 1) < cm->mi_rows) {
- int rt = 0;
- int64_t dt = 0;
+ RD_COST tmp_rdc;
PICK_MODE_CONTEXT *ctx = &pc_tree->horizontal[0];
+ vp9_rd_cost_init(&tmp_rdc);
update_state(cpi, ctx, mi_row, mi_col, subsize, 0);
encode_superblock(cpi, tp, 0, mi_row, mi_col, subsize, ctx);
- rd_pick_sb_modes(cpi, tile, mi_row + (mi_step >> 1), mi_col, &rt, &dt,
- subsize, &pc_tree->horizontal[1], INT64_MAX, 1);
- if (rt == INT_MAX || dt == INT64_MAX) {
- last_part_rate = INT_MAX;
- last_part_dist = INT64_MAX;
+ rd_pick_sb_modes(cpi, tile, mi_row + (mi_step >> 1), mi_col, &tmp_rdc,
+ subsize, &pc_tree->horizontal[1], INT64_MAX);
+ if (tmp_rdc.rate == INT_MAX || tmp_rdc.dist == INT64_MAX) {
+ vp9_rd_cost_reset(&last_part_rdc);
break;
}
-
- last_part_rate += rt;
- last_part_dist += dt;
+ last_part_rdc.rate += tmp_rdc.rate;
+ last_part_rdc.dist += tmp_rdc.dist;
+ last_part_rdc.rdcost += tmp_rdc.rdcost;
}
break;
case PARTITION_VERT:
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &last_part_rate,
- &last_part_dist, subsize, &pc_tree->vertical[0],
- INT64_MAX, 0);
- if (last_part_rate != INT_MAX &&
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &last_part_rdc,
+ subsize, &pc_tree->vertical[0], INT64_MAX);
+ if (last_part_rdc.rate != INT_MAX &&
bsize >= BLOCK_8X8 && mi_col + (mi_step >> 1) < cm->mi_cols) {
- int rt = 0;
- int64_t dt = 0;
+ RD_COST tmp_rdc;
PICK_MODE_CONTEXT *ctx = &pc_tree->vertical[0];
+ vp9_rd_cost_init(&tmp_rdc);
update_state(cpi, ctx, mi_row, mi_col, subsize, 0);
encode_superblock(cpi, tp, 0, mi_row, mi_col, subsize, ctx);
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col + (mi_step >> 1), &rt, &dt,
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col + (mi_step >> 1), &tmp_rdc,
subsize, &pc_tree->vertical[bsize > BLOCK_8X8],
- INT64_MAX, 1);
- if (rt == INT_MAX || dt == INT64_MAX) {
- last_part_rate = INT_MAX;
- last_part_dist = INT64_MAX;
+ INT64_MAX);
+ if (tmp_rdc.rate == INT_MAX || tmp_rdc.dist == INT64_MAX) {
+ vp9_rd_cost_reset(&last_part_rdc);
break;
}
- last_part_rate += rt;
- last_part_dist += dt;
+ last_part_rdc.rate += tmp_rdc.rate;
+ last_part_rdc.dist += tmp_rdc.dist;
+ last_part_rdc.rdcost += tmp_rdc.rdcost;
}
break;
case PARTITION_SPLIT:
if (bsize == BLOCK_8X8) {
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &last_part_rate,
- &last_part_dist, subsize, pc_tree->leaf_split[0],
- INT64_MAX, 0);
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &last_part_rdc,
+ subsize, pc_tree->leaf_split[0], INT64_MAX);
break;
}
- last_part_rate = 0;
- last_part_dist = 0;
+ last_part_rdc.rate = 0;
+ last_part_rdc.dist = 0;
+ last_part_rdc.rdcost = 0;
for (i = 0; i < 4; i++) {
int x_idx = (i & 1) * (mi_step >> 1);
int y_idx = (i >> 1) * (mi_step >> 1);
int jj = i >> 1, ii = i & 0x01;
- int rt;
- int64_t dt;
-
+ RD_COST tmp_rdc;
if ((mi_row + y_idx >= cm->mi_rows) || (mi_col + x_idx >= cm->mi_cols))
continue;
+ vp9_rd_cost_init(&tmp_rdc);
rd_use_partition(cpi, tile, mi_8x8 + jj * bss * mis + ii * bss, tp,
- mi_row + y_idx, mi_col + x_idx, subsize, &rt, &dt,
+ mi_row + y_idx, mi_col + x_idx, subsize,
+ &tmp_rdc.rate, &tmp_rdc.dist,
i != 3, pc_tree->split[i]);
- if (rt == INT_MAX || dt == INT64_MAX) {
- last_part_rate = INT_MAX;
- last_part_dist = INT64_MAX;
+ if (tmp_rdc.rate == INT_MAX || tmp_rdc.dist == INT64_MAX) {
+ vp9_rd_cost_reset(&last_part_rdc);
break;
}
- last_part_rate += rt;
- last_part_dist += dt;
+ last_part_rdc.rate += tmp_rdc.rate;
+ last_part_rdc.dist += tmp_rdc.dist;
}
break;
default:
@@ -1587,9 +1685,10 @@
}
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- if (last_part_rate < INT_MAX) {
- last_part_rate += cpi->partition_cost[pl][partition];
- last_part_rd = RDCOST(x->rdmult, x->rddiv, last_part_rate, last_part_dist);
+ if (last_part_rdc.rate < INT_MAX) {
+ last_part_rdc.rate += cpi->partition_cost[pl][partition];
+ last_part_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
+ last_part_rdc.rate, last_part_rdc.dist);
}
if (do_partition_search
@@ -1601,8 +1700,8 @@
&& (mi_col + mi_step < cm->mi_cols ||
mi_col + (mi_step >> 1) == cm->mi_cols)) {
BLOCK_SIZE split_subsize = get_subsize(bsize, PARTITION_SPLIT);
- chosen_rate = 0;
- chosen_dist = 0;
+ chosen_rdc.rate = 0;
+ chosen_rdc.dist = 0;
restore_context(cpi, mi_row, mi_col, a, l, sa, sl, bsize);
pc_tree->partitioning = PARTITION_SPLIT;
@@ -1610,8 +1709,7 @@
for (i = 0; i < 4; i++) {
int x_idx = (i & 1) * (mi_step >> 1);
int y_idx = (i >> 1) * (mi_step >> 1);
- int rt = 0;
- int64_t dt = 0;
+ RD_COST tmp_rdc;
ENTROPY_CONTEXT l[16 * MAX_MB_PLANE], a[16 * MAX_MB_PLANE];
PARTITION_CONTEXT sl[8], sa[8];
@@ -1620,20 +1718,18 @@
save_context(cpi, mi_row, mi_col, a, l, sa, sl, bsize);
pc_tree->split[i]->partitioning = PARTITION_NONE;
- rd_pick_sb_modes(cpi, tile, mi_row + y_idx, mi_col + x_idx, &rt, &dt,
- split_subsize, &pc_tree->split[i]->none,
- INT64_MAX, i);
+ rd_pick_sb_modes(cpi, tile, mi_row + y_idx, mi_col + x_idx, &tmp_rdc,
+ split_subsize, &pc_tree->split[i]->none, INT64_MAX);
restore_context(cpi, mi_row, mi_col, a, l, sa, sl, bsize);
- if (rt == INT_MAX || dt == INT64_MAX) {
- chosen_rate = INT_MAX;
- chosen_dist = INT64_MAX;
+ if (tmp_rdc.rate == INT_MAX || tmp_rdc.dist == INT64_MAX) {
+ vp9_rd_cost_reset(&chosen_rdc);
break;
}
- chosen_rate += rt;
- chosen_dist += dt;
+ chosen_rdc.rate += tmp_rdc.rate;
+ chosen_rdc.dist += tmp_rdc.dist;
if (i != 3)
encode_sb(cpi, tile, tp, mi_row + y_idx, mi_col + x_idx, 0,
@@ -1641,38 +1737,36 @@
pl = partition_plane_context(xd, mi_row + y_idx, mi_col + x_idx,
split_subsize);
- chosen_rate += cpi->partition_cost[pl][PARTITION_NONE];
+ chosen_rdc.rate += cpi->partition_cost[pl][PARTITION_NONE];
}
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- if (chosen_rate < INT_MAX) {
- chosen_rate += cpi->partition_cost[pl][PARTITION_SPLIT];
- chosen_rd = RDCOST(x->rdmult, x->rddiv, chosen_rate, chosen_dist);
+ if (chosen_rdc.rate < INT_MAX) {
+ chosen_rdc.rate += cpi->partition_cost[pl][PARTITION_SPLIT];
+ chosen_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
+ chosen_rdc.rate, chosen_rdc.dist);
}
}
// If last_part is better set the partitioning to that.
- if (last_part_rd < chosen_rd) {
+ if (last_part_rdc.rdcost < chosen_rdc.rdcost) {
mi_8x8[0].src_mi->mbmi.sb_type = bsize;
if (bsize >= BLOCK_8X8)
pc_tree->partitioning = partition;
- chosen_rate = last_part_rate;
- chosen_dist = last_part_dist;
- chosen_rd = last_part_rd;
+ chosen_rdc = last_part_rdc;
}
// If none was better set the partitioning to that.
- if (none_rd < chosen_rd) {
+ if (none_rdc.rdcost < chosen_rdc.rdcost) {
if (bsize >= BLOCK_8X8)
pc_tree->partitioning = PARTITION_NONE;
- chosen_rate = none_rate;
- chosen_dist = none_dist;
+ chosen_rdc = none_rdc;
}
restore_context(cpi, mi_row, mi_col, a, l, sa, sl, bsize);
// We must have chosen a partitioning and encoding or we'll fail later on.
// No other opportunities for success.
- if ( bsize == BLOCK_64X64)
- assert(chosen_rate < INT_MAX && chosen_dist < INT64_MAX);
+ if (bsize == BLOCK_64X64)
+ assert(chosen_rdc.rate < INT_MAX && chosen_rdc.dist < INT64_MAX);
if (do_recon) {
int output_enabled = (bsize == BLOCK_64X64);
@@ -1682,18 +1776,18 @@
// closer to the target.
if ((cpi->oxcf.aq_mode == COMPLEXITY_AQ) && cm->seg.update_map) {
vp9_select_in_frame_q_segment(cpi, mi_row, mi_col,
- output_enabled, chosen_rate);
+ output_enabled, chosen_rdc.rate);
}
if (cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ)
vp9_cyclic_refresh_set_rate_and_dist_sb(cpi->cyclic_refresh,
- chosen_rate, chosen_dist);
+ chosen_rdc.rate, chosen_rdc.dist);
encode_sb(cpi, tile, tp, mi_row, mi_col, output_enabled, bsize,
pc_tree);
}
- *rate = chosen_rate;
- *dist = chosen_dist;
+ *rate = chosen_rdc.rate;
+ *dist = chosen_rdc.dist;
}
static const BLOCK_SIZE min_partition_size[BLOCK_SIZES] = {
@@ -1863,7 +1957,7 @@
int bh, bw;
BLOCK_SIZE min_size = BLOCK_32X32;
BLOCK_SIZE max_size = BLOCK_8X8;
- int bsl = mi_width_log2(BLOCK_64X64);
+ int bsl = mi_width_log2_lookup[BLOCK_64X64];
const int search_range_ctrl = (((mi_row + mi_col) >> bsl) +
get_chessboard_index(cm->current_video_frame)) & 0x1;
// Trap case where we do not have a prediction.
@@ -2022,10 +2116,9 @@
// unlikely to be selected depending on previous rate-distortion optimization
// results, for encoding speed-up.
static void rd_pick_partition(VP9_COMP *cpi, const TileInfo *const tile,
- TOKENEXTRA **tp, int mi_row,
- int mi_col, BLOCK_SIZE bsize, int *rate,
- int64_t *dist, int64_t best_rd,
- PC_TREE *pc_tree) {
+ TOKENEXTRA **tp, int mi_row, int mi_col,
+ BLOCK_SIZE bsize, RD_COST *rd_cost,
+ int64_t best_rd, PC_TREE *pc_tree) {
VP9_COMMON *const cm = &cpi->common;
MACROBLOCK *const x = &cpi->mb;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -2036,9 +2129,7 @@
PICK_MODE_CONTEXT *ctx = &pc_tree->none;
int i, pl;
BLOCK_SIZE subsize;
- int this_rate, sum_rate = 0, best_rate = INT_MAX;
- int64_t this_dist, sum_dist = 0, best_dist = INT64_MAX;
- int64_t sum_rd = 0;
+ RD_COST this_rdc, sum_rdc, best_rdc;
int do_split = bsize >= BLOCK_8X8;
int do_rect = 1;
@@ -2066,6 +2157,11 @@
assert(num_8x8_blocks_wide_lookup[bsize] ==
num_8x8_blocks_high_lookup[bsize]);
+ vp9_rd_cost_init(&this_rdc);
+ vp9_rd_cost_init(&sum_rdc);
+ vp9_rd_cost_reset(&best_rdc);
+ best_rdc.rdcost = best_rd;
+
set_offsets(cpi, tile, mi_row, mi_col, bsize);
if (bsize == BLOCK_16X16 && cpi->oxcf.aq_mode)
@@ -2157,28 +2253,29 @@
// PARTITION_NONE
if (partition_none_allowed) {
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &this_rate, &this_dist, bsize,
- ctx, best_rd, 0);
- if (this_rate != INT_MAX) {
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &this_rdc, bsize, ctx,
+ best_rdc.rdcost);
+ if (this_rdc.rate != INT_MAX) {
if (bsize >= BLOCK_8X8) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- this_rate += cpi->partition_cost[pl][PARTITION_NONE];
+ this_rdc.rate += cpi->partition_cost[pl][PARTITION_NONE];
+ this_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
+ this_rdc.rate, this_rdc.dist);
}
- sum_rd = RDCOST(x->rdmult, x->rddiv, this_rate, this_dist);
- if (sum_rd < best_rd) {
+ if (this_rdc.rdcost < best_rdc.rdcost) {
int64_t dist_breakout_thr = cpi->sf.partition_search_breakout_dist_thr;
int rate_breakout_thr = cpi->sf.partition_search_breakout_rate_thr;
- best_rate = this_rate;
- best_dist = this_dist;
- best_rd = sum_rd;
+ best_rdc = this_rdc;
if (bsize >= BLOCK_8X8)
pc_tree->partitioning = PARTITION_NONE;
// Adjust dist breakout threshold according to the partition size.
- dist_breakout_thr >>= 8 - (b_width_log2(bsize) +
- b_height_log2(bsize));
+ dist_breakout_thr >>= 8 - (b_width_log2_lookup[bsize] +
+ b_height_log2_lookup[bsize]);
+
+ rate_breakout_thr *= num_pels_log2_lookup[bsize];
// If all y, u, v transform blocks in this partition are skippable, and
// the dist & rate are within the thresholds, the partition search is
@@ -2186,8 +2283,8 @@
// The dist & rate thresholds are set to 0 at speed 0 to disable the
// early termination at that speed.
if (!x->e_mbd.lossless &&
- (ctx->skippable && best_dist < dist_breakout_thr &&
- best_rate < rate_breakout_thr)) {
+ (ctx->skippable && best_rdc.dist < dist_breakout_thr &&
+ best_rdc.rate < rate_breakout_thr)) {
do_split = 0;
do_rect = 0;
}
@@ -2247,7 +2344,6 @@
store_pred_mv(x, ctx);
// PARTITION_SPLIT
- sum_rd = 0;
// TODO(jingning): use the motion vectors given by the above search as
// the starting point of motion search in the following partition type check.
if (do_split) {
@@ -2257,14 +2353,12 @@
if (cpi->sf.adaptive_pred_interp_filter && partition_none_allowed)
pc_tree->leaf_split[0]->pred_interp_filter =
ctx->mic.mbmi.interp_filter;
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &sum_rate, &sum_dist, subsize,
- pc_tree->leaf_split[0], best_rd, 0);
- if (sum_rate == INT_MAX)
- sum_rd = INT64_MAX;
- else
- sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &sum_rdc, subsize,
+ pc_tree->leaf_split[0], best_rdc.rdcost);
+ if (sum_rdc.rate == INT_MAX)
+ sum_rdc.rdcost = INT64_MAX;
} else {
- for (i = 0; i < 4 && sum_rd < best_rd; ++i) {
+ for (i = 0; i < 4 && sum_rdc.rdcost < best_rdc.rdcost; ++i) {
const int x_idx = (i & 1) * mi_step;
const int y_idx = (i >> 1) * mi_step;
@@ -2276,28 +2370,28 @@
pc_tree->split[i]->index = i;
rd_pick_partition(cpi, tile, tp, mi_row + y_idx, mi_col + x_idx,
- subsize, &this_rate, &this_dist,
- best_rd - sum_rd, pc_tree->split[i]);
+ subsize, &this_rdc,
+ best_rdc.rdcost - sum_rdc.rdcost, pc_tree->split[i]);
- if (this_rate == INT_MAX) {
- sum_rd = INT64_MAX;
+ if (this_rdc.rate == INT_MAX) {
+ sum_rdc.rdcost = INT64_MAX;
+ break;
} else {
- sum_rate += this_rate;
- sum_dist += this_dist;
- sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
+ sum_rdc.rate += this_rdc.rate;
+ sum_rdc.dist += this_rdc.dist;
+ sum_rdc.rdcost += this_rdc.rdcost;
}
}
}
- if (sum_rd < best_rd && i == 4) {
+ if (sum_rdc.rdcost < best_rdc.rdcost && i == 4) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- sum_rate += cpi->partition_cost[pl][PARTITION_SPLIT];
- sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
+ sum_rdc.rate += cpi->partition_cost[pl][PARTITION_SPLIT];
+ sum_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
+ sum_rdc.rate, sum_rdc.dist);
- if (sum_rd < best_rd) {
- best_rate = sum_rate;
- best_dist = sum_dist;
- best_rd = sum_rd;
+ if (sum_rdc.rdcost < best_rdc.rdcost) {
+ best_rdc = sum_rdc;
pc_tree->partitioning = PARTITION_SPLIT;
}
} else {
@@ -2318,11 +2412,11 @@
partition_none_allowed)
pc_tree->horizontal[0].pred_interp_filter =
ctx->mic.mbmi.interp_filter;
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &sum_rate, &sum_dist, subsize,
- &pc_tree->horizontal[0], best_rd, 0);
- sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &sum_rdc, subsize,
+ &pc_tree->horizontal[0], best_rdc.rdcost);
- if (sum_rd < best_rd && mi_row + mi_step < cm->mi_rows) {
+ if (sum_rdc.rdcost < best_rdc.rdcost && mi_row + mi_step < cm->mi_rows &&
+ bsize > BLOCK_8X8) {
PICK_MODE_CONTEXT *ctx = &pc_tree->horizontal[0];
update_state(cpi, ctx, mi_row, mi_col, subsize, 0);
encode_superblock(cpi, tp, 0, mi_row, mi_col, subsize, ctx);
@@ -2333,25 +2427,24 @@
partition_none_allowed)
pc_tree->horizontal[1].pred_interp_filter =
ctx->mic.mbmi.interp_filter;
- rd_pick_sb_modes(cpi, tile, mi_row + mi_step, mi_col, &this_rate,
- &this_dist, subsize, &pc_tree->horizontal[1],
- best_rd - sum_rd, 1);
- if (this_rate == INT_MAX) {
- sum_rd = INT64_MAX;
+ rd_pick_sb_modes(cpi, tile, mi_row + mi_step, mi_col, &this_rdc,
+ subsize, &pc_tree->horizontal[1],
+ best_rdc.rdcost - sum_rdc.rdcost);
+ if (this_rdc.rate == INT_MAX) {
+ sum_rdc.rdcost = INT64_MAX;
} else {
- sum_rate += this_rate;
- sum_dist += this_dist;
- sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
+ sum_rdc.rate += this_rdc.rate;
+ sum_rdc.dist += this_rdc.dist;
+ sum_rdc.rdcost += this_rdc.rdcost;
}
}
- if (sum_rd < best_rd) {
+
+ if (sum_rdc.rdcost < best_rdc.rdcost) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- 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;
- best_rate = sum_rate;
- best_dist = sum_dist;
+ sum_rdc.rate += cpi->partition_cost[pl][PARTITION_HORZ];
+ sum_rdc.rdcost = RDCOST(x->rdmult, x->rddiv, sum_rdc.rate, sum_rdc.dist);
+ if (sum_rdc.rdcost < best_rdc.rdcost) {
+ best_rdc = sum_rdc;
pc_tree->partitioning = PARTITION_HORZ;
}
}
@@ -2367,10 +2460,10 @@
partition_none_allowed)
pc_tree->vertical[0].pred_interp_filter =
ctx->mic.mbmi.interp_filter;
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &sum_rate, &sum_dist, subsize,
- &pc_tree->vertical[0], best_rd, 0);
- sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
- if (sum_rd < best_rd && mi_col + mi_step < cm->mi_cols) {
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col, &sum_rdc, subsize,
+ &pc_tree->vertical[0], best_rdc.rdcost);
+ if (sum_rdc.rdcost < best_rdc.rdcost && mi_col + mi_step < cm->mi_cols &&
+ bsize > BLOCK_8X8) {
update_state(cpi, &pc_tree->vertical[0], mi_row, mi_col, subsize, 0);
encode_superblock(cpi, tp, 0, mi_row, mi_col, subsize,
&pc_tree->vertical[0]);
@@ -2381,26 +2474,24 @@
partition_none_allowed)
pc_tree->vertical[1].pred_interp_filter =
ctx->mic.mbmi.interp_filter;
- rd_pick_sb_modes(cpi, tile, mi_row, mi_col + mi_step, &this_rate,
- &this_dist, subsize,
- &pc_tree->vertical[1], best_rd - sum_rd,
- 1);
- if (this_rate == INT_MAX) {
- sum_rd = INT64_MAX;
+ rd_pick_sb_modes(cpi, tile, mi_row, mi_col + mi_step, &this_rdc, subsize,
+ &pc_tree->vertical[1], best_rdc.rdcost - sum_rdc.rdcost);
+ if (this_rdc.rate == INT_MAX) {
+ sum_rdc.rdcost = INT64_MAX;
} else {
- sum_rate += this_rate;
- sum_dist += this_dist;
- sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
+ sum_rdc.rate += this_rdc.rate;
+ sum_rdc.dist += this_rdc.dist;
+ sum_rdc.rdcost += this_rdc.rdcost;
}
}
- if (sum_rd < best_rd) {
+
+ if (sum_rdc.rdcost < best_rdc.rdcost) {
pl = partition_plane_context(xd, mi_row, mi_col, bsize);
- 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;
- best_dist = sum_dist;
- best_rd = sum_rd;
+ sum_rdc.rate += cpi->partition_cost[pl][PARTITION_VERT];
+ sum_rdc.rdcost = RDCOST(x->rdmult, x->rddiv,
+ sum_rdc.rate, sum_rdc.dist);
+ if (sum_rdc.rdcost < best_rdc.rdcost) {
+ best_rdc = sum_rdc;
pc_tree->partitioning = PARTITION_VERT;
}
}
@@ -2412,10 +2503,11 @@
// point. This code should be refactored so that the duplicate
// checks occur in some sub function and thus are used...
(void) best_rd;
- *rate = best_rate;
- *dist = best_dist;
+ *rd_cost = best_rdc;
- if (best_rate < INT_MAX && best_dist < INT64_MAX && pc_tree->index != 3) {
+
+ if (best_rdc.rate < INT_MAX && best_rdc.dist < INT64_MAX &&
+ pc_tree->index != 3) {
int output_enabled = (bsize == BLOCK_64X64);
// Check the projected output rate for this SB against it's target
@@ -2423,18 +2515,18 @@
// closer to the target.
if ((cpi->oxcf.aq_mode == COMPLEXITY_AQ) && cm->seg.update_map)
vp9_select_in_frame_q_segment(cpi, mi_row, mi_col, output_enabled,
- best_rate);
+ best_rdc.rate);
if (cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ)
vp9_cyclic_refresh_set_rate_and_dist_sb(cpi->cyclic_refresh,
- best_rate, best_dist);
+ best_rdc.rate, best_rdc.dist);
encode_sb(cpi, tile, tp, mi_row, mi_col, output_enabled, bsize, pc_tree);
}
if (bsize == BLOCK_64X64) {
assert(tp_orig < *tp);
- assert(best_rate < INT_MAX);
- assert(best_dist < INT64_MAX);
+ assert(best_rdc.rate < INT_MAX);
+ assert(best_rdc.dist < INT64_MAX);
} else {
assert(tp_orig == *tp);
}
@@ -2456,9 +2548,16 @@
mi_col += MI_BLOCK_SIZE) {
int dummy_rate;
int64_t dummy_dist;
-
+ RD_COST dummy_rdc;
int i;
+ const int idx_str = cm->mi_stride * mi_row + mi_col;
+ MODE_INFO *mi = cm->mi + idx_str;
+ MODE_INFO *prev_mi = NULL;
+
+ if (cm->frame_type != KEY_FRAME)
+ prev_mi = (cm->prev_mip + cm->mi_stride + 1 + idx_str)->src_mi;
+
if (sf->adaptive_pred_interp_filter) {
for (i = 0; i < 64; ++i)
cpi->leaf_tree[i].pred_interp_filter = SWITCHABLE;
@@ -2477,71 +2576,44 @@
// TODO(yunqingwang): use_lastframe_partitioning is no longer used in good-
// quality encoding. Need to evaluate it in real-time encoding later to
// decide if it can be removed too. And then, do the code cleanup.
- if ((sf->partition_search_type == SEARCH_PARTITION &&
- sf->use_lastframe_partitioning) ||
- sf->partition_search_type == FIXED_PARTITION ||
- 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 = cm->mi + idx_str;
- MODE_INFO *prev_mi = (cm->prev_mip + cm->mi_stride + 1 + idx_str)->src_mi;
- 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, mi_row, mi_col,
- sf->always_this_block_size);
- rd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, 1, cpi->pc_root);
- } else if (cpi->skippable_frame ||
- 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, 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, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, 1, cpi->pc_root);
- } else {
- GF_GROUP * gf_grp = &cpi->twopass.gf_group;
- int last_was_mid_sequence_overlay = 0;
- if ((cpi->oxcf.pass == 2) && (gf_grp->index)) {
- if (gf_grp->update_type[gf_grp->index - 1] == OVERLAY_UPDATE)
- last_was_mid_sequence_overlay = 1;
- }
- if ((cpi->rc.frames_since_key
- % sf->last_partitioning_redo_frequency) == 0
- || last_was_mid_sequence_overlay
- || cm->prev_mi == 0
- || cm->show_frame == 0
- || cm->frame_type == KEY_FRAME
- || cpi->rc.is_src_frame_alt_ref
- || ((sf->use_lastframe_partitioning ==
- LAST_FRAME_PARTITION_LOW_MOTION) &&
- sb_has_motion(cm, prev_mi, sf->lf_motion_threshold))) {
- // 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);
- rd_auto_partition_range(cpi, tile, mi_row, mi_col,
- &sf->min_partition_size,
- &sf->max_partition_size);
- }
- rd_pick_partition(cpi, tile, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, INT64_MAX,
- cpi->pc_root);
- } else {
- if (sf->constrain_copy_partition &&
- sb_has_motion(cm, prev_mi, sf->lf_motion_threshold))
- constrain_copy_partitioning(cpi, tile, mi, prev_mi,
- mi_row, mi_col, BLOCK_16X16);
- else
- 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);
- }
- }
+ 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, mi_row, mi_col,
+ sf->always_this_block_size);
+ rd_use_partition(cpi, tile, mi, tp, mi_row, mi_col, BLOCK_64X64,
+ &dummy_rate, &dummy_dist, 1, cpi->pc_root);
+ } else if (cpi->partition_search_skippable_frame) {
+ 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, 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 &&
+ cm->frame_type != KEY_FRAME ) {
+ choose_partitioning(cpi, tile, mi_row, mi_col);
+ 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 == SEARCH_PARTITION &&
+ sf->use_lastframe_partitioning &&
+ (cpi->rc.frames_since_key %
+ sf->last_partitioning_redo_frequency) &&
+ cm->prev_mi &&
+ cm->show_frame &&
+ cm->frame_type != KEY_FRAME &&
+ !cpi->rc.is_src_frame_alt_ref &&
+ ((sf->use_lastframe_partitioning !=
+ LAST_FRAME_PARTITION_LOW_MOTION) ||
+ !sb_has_motion(cm, prev_mi, sf->lf_motion_threshold))) {
+ if (sf->constrain_copy_partition &&
+ sb_has_motion(cm, prev_mi, sf->lf_motion_threshold))
+ constrain_copy_partitioning(cpi, tile, mi, prev_mi,
+ mi_row, mi_col, BLOCK_16X16);
+ else
+ 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 {
// If required set upper and lower partition size limits
if (sf->auto_min_max_partition_size) {
@@ -2551,7 +2623,7 @@
&sf->max_partition_size);
}
rd_pick_partition(cpi, tile, tp, mi_row, mi_col, BLOCK_64X64,
- &dummy_rate, &dummy_dist, INT64_MAX, cpi->pc_root);
+ &dummy_rdc, INT64_MAX, cpi->pc_root);
}
}
}
@@ -2652,7 +2724,7 @@
BLOCK_SIZE bsize, BLOCK_SIZE subsize,
PC_TREE *pc_tree) {
MACROBLOCKD *xd = &x->e_mbd;
- int bsl = b_width_log2(bsize), hbs = (1 << bsl) / 4;
+ int bsl = b_width_log2_lookup[bsize], hbs = (1 << bsl) / 4;
PARTITION_TYPE partition = pc_tree->partitioning;
assert(bsize >= BLOCK_8X8);
@@ -2771,8 +2843,13 @@
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;
- int64_t stop_thresh_rd;
+ int dist_breakout_thr = sf->partition_search_breakout_dist_thr;
+ int64_t rate_breakout_thr = sf->partition_search_breakout_rate_thr;
+
+ dist_breakout_thr >>= 8 - (b_width_log2_lookup[bsize] +
+ b_height_log2_lookup[bsize]);
+
+ rate_breakout_thr *= num_pels_log2_lookup[bsize];
best_rate = this_rate;
best_dist = this_dist;
@@ -2780,14 +2857,9 @@
if (bsize >= BLOCK_8X8)
pc_tree->partitioning = PARTITION_NONE;
- // Adjust threshold according to partition size.
- stop_thresh >>= 8 - (b_width_log2(bsize) +
- b_height_log2(bsize));
-
- stop_thresh_rd = RDCOST(x->rdmult, x->rddiv, 0, stop_thresh);
- // If obtained distortion is very small, choose current partition
- // and stop splitting.
- if (!x->e_mbd.lossless && best_rd < stop_thresh_rd) {
+ if (!x->e_mbd.lossless &&
+ this_rate < rate_breakout_thr &&
+ this_dist < dist_breakout_thr) {
do_split = 0;
do_rect = 0;
}
@@ -2974,7 +3046,7 @@
VP9_COMMON *const cm = &cpi->common;
MACROBLOCK *const x = &cpi->mb;
MACROBLOCKD *const xd = &x->e_mbd;
- const int bsl = b_width_log2(bsize), hbs = (1 << bsl) / 4;
+ const int bsl = b_width_log2_lookup[bsize], hbs = (1 << bsl) / 4;
const int mis = cm->mi_stride;
PARTITION_TYPE partition;
BLOCK_SIZE subsize;
@@ -3095,7 +3167,6 @@
int64_t dummy_dist = 0;
const int idx_str = cm->mi_stride * mi_row + mi_col;
MODE_INFO *mi = cm->mi + idx_str;
- MODE_INFO *prev_mi = (cm->prev_mip + cm->mi_stride + 1 + idx_str)->src_mi;
BLOCK_SIZE bsize;
x->in_static_area = 0;
x->source_variance = UINT_MAX;
@@ -3113,7 +3184,6 @@
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 = sf->partition_search_type == FIXED_PARTITION ?
sf->always_this_block_size :
@@ -3133,7 +3203,7 @@
&dummy_rate, &dummy_dist, 1, INT64_MAX,
cpi->pc_root);
} else {
- copy_partitioning(cm, mi, prev_mi);
+ choose_partitioning(cpi, tile, 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);
@@ -3170,9 +3240,34 @@
for (i = 0; i < cm->mb_rows; i++) {
for (j = 0; j < cm->mb_cols; j++) {
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ switch (cm->bit_depth) {
+ case VPX_BITS_8:
+ vp9_highbd_get16x16var(src, src_stride, last_src, last_stride,
+ &var16->sse, &var16->sum);
+ break;
+ case VPX_BITS_10:
+ vp9_highbd_10_get16x16var(src, src_stride, last_src, last_stride,
+ &var16->sse, &var16->sum);
+ break;
+ case VPX_BITS_12:
+ vp9_highbd_12_get16x16var(src, src_stride, last_src, last_stride,
+ &var16->sse, &var16->sum);
+ break;
+ default:
+ assert(0 && "cm->bit_depth should be VPX_BITS_8, VPX_BITS_10"
+ " or VPX_BITS_12");
+ return -1;
+ }
+ } else {
+ vp9_get16x16var(src, src_stride, last_src, last_stride,
+ &var16->sse, &var16->sum);
+ }
+#else
vp9_get16x16var(src, src_stride, last_src, last_stride,
&var16->sse, &var16->sum);
-
+#endif // CONFIG_VP9_HIGHBITDEPTH
var16->var = var16->sse -
(((uint32_t)var16->sum * var16->sum) >> 8);
@@ -3252,25 +3347,39 @@
const VP9_COMMON *const cm = &cpi->common;
const int tile_cols = 1 << cm->log2_tile_cols;
const int tile_rows = 1 << cm->log2_tile_rows;
+
int tile_col, tile_row;
- TOKENEXTRA *tok = cpi->tok;
+ TileInfo tile[4][1 << 6];
+ TOKENEXTRA *tok[4][1 << 6];
+ TOKENEXTRA *pre_tok = cpi->tok;
+ int tile_tok = 0;
for (tile_row = 0; tile_row < tile_rows; ++tile_row) {
for (tile_col = 0; tile_col < tile_cols; ++tile_col) {
- TileInfo tile;
- TOKENEXTRA *old_tok = tok;
+ vp9_tile_init(&tile[tile_row][tile_col], cm, tile_row, tile_col);
+
+ tok[tile_row][tile_col] = pre_tok + tile_tok;
+ pre_tok = tok[tile_row][tile_col];
+ tile_tok = allocated_tokens(tile[tile_row][tile_col]);
+ }
+ }
+
+ for (tile_row = 0; tile_row < tile_rows; ++tile_row) {
+ for (tile_col = 0; tile_col < tile_cols; ++tile_col) {
+ const TileInfo * const ptile = &tile[tile_row][tile_col];
+ TOKENEXTRA * const old_tok = tok[tile_row][tile_col];
int mi_row;
- vp9_tile_init(&tile, cm, tile_row, tile_col);
- for (mi_row = tile.mi_row_start; mi_row < tile.mi_row_end;
+ for (mi_row = ptile->mi_row_start; mi_row < ptile->mi_row_end;
mi_row += MI_BLOCK_SIZE) {
if (cpi->sf.use_nonrd_pick_mode && !frame_is_intra_only(cm))
- encode_nonrd_sb_row(cpi, &tile, mi_row, &tok);
+ encode_nonrd_sb_row(cpi, ptile, mi_row, &tok[tile_row][tile_col]);
else
- encode_rd_sb_row(cpi, &tile, mi_row, &tok);
+ encode_rd_sb_row(cpi, ptile, mi_row, &tok[tile_row][tile_col]);
}
- cpi->tok_count[tile_row][tile_col] = (unsigned int)(tok - old_tok);
- assert(tok - cpi->tok <= get_token_alloc(cm->mb_rows, cm->mb_cols));
+ cpi->tok_count[tile_row][tile_col] =
+ (unsigned int)(tok[tile_row][tile_col] - old_tok);
+ assert(tok[tile_row][tile_col] - old_tok <= allocated_tokens(*ptile));
}
}
}
@@ -3314,7 +3423,16 @@
cm->tx_mode = select_tx_mode(cpi);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth)
+ x->fwd_txm4x4 = xd->lossless ? vp9_fwht4x4 : vp9_fdct4x4;
+ else
+ x->fwd_txm4x4 = xd->lossless ? vp9_highbd_fwht4x4 : vp9_highbd_fdct4x4;
+ x->highbd_itxm_add = xd->lossless ? vp9_highbd_iwht4x4_add :
+ vp9_highbd_idct4x4_add;
+#else
x->fwd_txm4x4 = xd->lossless ? vp9_fwht4x4 : vp9_fdct4x4;
+#endif // CONFIG_VP9_HIGHBITDEPTH
x->itxm_add = xd->lossless ? vp9_iwht4x4_add : vp9_idct4x4_add;
if (xd->lossless) {
diff --git a/source/libvpx/vp9/encoder/vp9_encodemb.c b/source/libvpx/vp9/encoder/vp9_encodemb.c
index 2eae149..f5faa7c 100644
--- a/source/libvpx/vp9/encoder/vp9_encodemb.c
+++ b/source/libvpx/vp9/encoder/vp9_encodemb.c
@@ -51,6 +51,29 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+void vp9_highbd_subtract_block_c(int rows, int cols,
+ int16_t *diff, ptrdiff_t diff_stride,
+ const uint8_t *src8, ptrdiff_t src_stride,
+ const uint8_t *pred8, ptrdiff_t pred_stride,
+ int bd) {
+ int r, c;
+ uint16_t *src = CONVERT_TO_SHORTPTR(src8);
+ uint16_t *pred = CONVERT_TO_SHORTPTR(pred8);
+ (void) bd;
+
+ for (r = 0; r < rows; r++) {
+ for (c = 0; c < cols; c++) {
+ diff[c] = src[c] - pred[c];
+ }
+
+ diff += diff_stride;
+ pred += pred_stride;
+ src += src_stride;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
void vp9_subtract_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
struct macroblock_plane *const p = &x->plane[plane];
const struct macroblockd_plane *const pd = &x->e_mbd.plane[plane];
@@ -58,6 +81,14 @@
const int bw = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
const int bh = 4 * num_4x4_blocks_high_lookup[plane_bsize];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (x->e_mbd.cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_subtract_block(bh, bw, p->src_diff, bw, p->src.buf,
+ p->src.stride, pd->dst.buf, pd->dst.stride,
+ x->e_mbd.bd);
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
vp9_subtract_block(bh, bw, p->src_diff, bw, p->src.buf, p->src.stride,
pd->dst.buf, pd->dst.stride);
}
@@ -124,6 +155,8 @@
int64_t rd_cost0, rd_cost1;
int rate0, rate1, error0, error1, t0, t1;
int best, band, pt, i, final_eob;
+ const TOKENVALUE *dct_value_tokens;
+ const int16_t *dct_value_cost;
assert((!type && !plane) || (type && plane));
assert(eob <= default_eob);
@@ -140,9 +173,24 @@
tokens[eob][0].qc = 0;
tokens[eob][1] = tokens[eob][0];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->bd == 12) {
+ dct_value_tokens = vp9_dct_value_tokens_high12_ptr;
+ dct_value_cost = vp9_dct_value_cost_high12_ptr;
+ } else if (xd->bd == 10) {
+ dct_value_tokens = vp9_dct_value_tokens_high10_ptr;
+ dct_value_cost = vp9_dct_value_cost_high10_ptr;
+ } else {
+ dct_value_tokens = vp9_dct_value_tokens_ptr;
+ dct_value_cost = vp9_dct_value_cost_ptr;
+ }
+#else
+ dct_value_tokens = vp9_dct_value_tokens_ptr;
+ dct_value_cost = vp9_dct_value_cost_ptr;
+#endif
for (i = 0; i < eob; i++)
token_cache[scan[i]] =
- vp9_pt_energy_class[vp9_dct_value_tokens_ptr[qcoeff[scan[i]]].token];
+ vp9_pt_energy_class[dct_value_tokens[qcoeff[scan[i]]].token];
for (i = eob; i-- > 0;) {
int base_bits, d2, dx;
@@ -156,7 +204,7 @@
/* Evaluate the first possibility for this state. */
rate0 = tokens[next][0].rate;
rate1 = tokens[next][1].rate;
- t0 = (vp9_dct_value_tokens_ptr + x)->token;
+ t0 = (dct_value_tokens + x)->token;
/* Consider both possible successor states. */
if (next < default_eob) {
band = band_translate[i + 1];
@@ -169,8 +217,13 @@
UPDATE_RD_COST();
/* And pick the best. */
best = rd_cost1 < rd_cost0;
- base_bits = vp9_dct_value_cost_ptr[x];
+ base_bits = dct_value_cost[x];
dx = mul * (dqcoeff[rc] - coeff[rc]);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ dx >>= xd->bd - 8;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
d2 = dx * dx;
tokens[i][0].rate = base_bits + (best ? rate1 : rate0);
tokens[i][0].error = d2 + (best ? error1 : error0);
@@ -203,7 +256,7 @@
t0 = tokens[next][0].token == EOB_TOKEN ? EOB_TOKEN : ZERO_TOKEN;
t1 = tokens[next][1].token == EOB_TOKEN ? EOB_TOKEN : ZERO_TOKEN;
} else {
- t0 = t1 = (vp9_dct_value_tokens_ptr + x)->token;
+ t0 = t1 = (dct_value_tokens + x)->token;
}
if (next < default_eob) {
band = band_translate[i + 1];
@@ -222,10 +275,18 @@
UPDATE_RD_COST();
/* And pick the best. */
best = rd_cost1 < rd_cost0;
- base_bits = vp9_dct_value_cost_ptr[x];
+ base_bits = dct_value_cost[x];
if (shortcut) {
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ dx -= ((dequant_ptr[rc != 0] >> (xd->bd - 8)) + sz) ^ sz;
+ } else {
+ dx -= (dequant_ptr[rc != 0] + sz) ^ sz;
+ }
+#else
dx -= (dequant_ptr[rc != 0] + sz) ^ sz;
+#endif // CONFIG_VP9_HIGHBITDEPTH
d2 = dx * dx;
}
tokens[i][1].rate = base_bits + (best ? rate1 : rate0);
@@ -303,14 +364,14 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE void high_fdct32x32(int rd_transform, const int16_t *src,
- tran_low_t *dst, int src_stride) {
+static INLINE void highbd_fdct32x32(int rd_transform, const int16_t *src,
+ tran_low_t *dst, int src_stride) {
if (rd_transform)
- vp9_high_fdct32x32_rd(src, dst, src_stride);
+ vp9_highbd_fdct32x32_rd(src, dst, src_stride);
else
- vp9_high_fdct32x32(src, dst, src_stride);
+ vp9_highbd_fdct32x32(src, dst, src_stride);
}
-#endif
+#endif // CONFIG_VP9_HIGHBITDEPTH
void vp9_xform_quant_fp(MACROBLOCK *x, int plane, int block,
BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
@@ -328,6 +389,45 @@
txfrm_block_to_raster_xy(plane_bsize, tx_size, block, &i, &j);
src_diff = &p->src_diff[4 * (j * diff_stride + i)];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ switch (tx_size) {
+ case TX_32X32:
+ highbd_fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_fp_32x32(coeff, 1024, x->skip_block, p->zbin,
+ p->round_fp, p->quant_fp, p->quant_shift,
+ qcoeff, dqcoeff, pd->dequant,
+ p->zbin_extra, eob, scan_order->scan,
+ scan_order->iscan);
+ break;
+ case TX_16X16:
+ vp9_highbd_fdct16x16(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_fp(coeff, 256, x->skip_block, p->zbin, p->round_fp,
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ break;
+ case TX_8X8:
+ vp9_highbd_fdct8x8(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_fp(coeff, 64, x->skip_block, p->zbin, p->round_fp,
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ break;
+ case TX_4X4:
+ x->fwd_txm4x4(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_fp(coeff, 16, x->skip_block, p->zbin, p->round_fp,
+ p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ break;
+ default:
+ assert(0);
+ }
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
switch (tx_size) {
case TX_32X32:
fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride);
@@ -379,6 +479,40 @@
txfrm_block_to_raster_xy(plane_bsize, tx_size, block, &i, &j);
src_diff = &p->src_diff[4 * (j * diff_stride + i)];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ switch (tx_size) {
+ case TX_32X32:
+ vp9_highbd_fdct32x32_1(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_dc_32x32(coeff, x->skip_block, p->round,
+ p->quant_fp[0], qcoeff, dqcoeff,
+ pd->dequant[0], eob);
+ break;
+ case TX_16X16:
+ vp9_highbd_fdct16x16_1(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_dc(coeff, x->skip_block, p->round,
+ p->quant_fp[0], qcoeff, dqcoeff,
+ pd->dequant[0], eob);
+ break;
+ case TX_8X8:
+ vp9_highbd_fdct8x8_1(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_dc(coeff, x->skip_block, p->round,
+ p->quant_fp[0], qcoeff, dqcoeff,
+ pd->dequant[0], eob);
+ break;
+ case TX_4X4:
+ x->fwd_txm4x4(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_dc(coeff, x->skip_block, p->round,
+ p->quant_fp[0], qcoeff, dqcoeff,
+ pd->dequant[0], eob);
+ break;
+ default:
+ assert(0);
+ }
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
switch (tx_size) {
case TX_32X32:
vp9_fdct32x32_1(src_diff, coeff, diff_stride);
@@ -426,6 +560,44 @@
txfrm_block_to_raster_xy(plane_bsize, tx_size, block, &i, &j);
src_diff = &p->src_diff[4 * (j * diff_stride + i)];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ switch (tx_size) {
+ case TX_32X32:
+ highbd_fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin,
+ p->round, p->quant, p->quant_shift, qcoeff,
+ dqcoeff, pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ break;
+ case TX_16X16:
+ vp9_highbd_fdct16x16(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
+ p->quant, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ break;
+ case TX_8X8:
+ vp9_highbd_fdct8x8(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
+ p->quant, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ break;
+ case TX_4X4:
+ x->fwd_txm4x4(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
+ p->quant, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ break;
+ default:
+ assert(0);
+ }
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
switch (tx_size) {
case TX_32X32:
fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride);
@@ -520,6 +692,34 @@
if (x->skip_encode || p->eobs[block] == 0)
return;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ switch (tx_size) {
+ case TX_32X32:
+ vp9_highbd_idct32x32_add(dqcoeff, dst, pd->dst.stride,
+ p->eobs[block], xd->bd);
+ break;
+ case TX_16X16:
+ vp9_highbd_idct16x16_add(dqcoeff, dst, pd->dst.stride,
+ p->eobs[block], xd->bd);
+ break;
+ case TX_8X8:
+ vp9_highbd_idct8x8_add(dqcoeff, dst, pd->dst.stride,
+ p->eobs[block], xd->bd);
+ break;
+ case TX_4X4:
+ // this is like vp9_short_idct4x4 but has a special case around eob<=1
+ // which is significant (not just an optimization) for the lossless
+ // case.
+ x->highbd_itxm_add(dqcoeff, dst, pd->dst.stride,
+ p->eobs[block], xd->bd);
+ break;
+ default:
+ assert(0 && "Invalid transform size");
+ }
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
switch (tx_size) {
case TX_32X32:
@@ -557,8 +757,15 @@
vp9_xform_quant(x, plane, block, plane_bsize, tx_size);
- if (p->eobs[block] > 0)
+ if (p->eobs[block] > 0) {
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ x->highbd_itxm_add(dqcoeff, dst, pd->dst.stride, p->eobs[block], xd->bd);
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
x->itxm_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+ }
}
void vp9_encode_sby_pass1(MACROBLOCK *x, BLOCK_SIZE bsize) {
@@ -609,7 +816,7 @@
const scan_order *scan_order;
TX_TYPE tx_type;
PREDICTION_MODE mode;
- const int bwl = b_width_log2(plane_bsize);
+ const int bwl = b_width_log2_lookup[plane_bsize];
const int diff_stride = 4 * (1 << bwl);
uint8_t *src, *dst;
int16_t *src_diff;
@@ -622,6 +829,117 @@
src = &p->src.buf[4 * (j * src_stride + i)];
src_diff = &p->src_diff[4 * (j * diff_stride + i)];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ switch (tx_size) {
+ case TX_32X32:
+ scan_order = &vp9_default_scan_orders[TX_32X32];
+ mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
+ vp9_predict_intra_block(xd, block >> 6, bwl, TX_32X32, mode,
+ x->skip_encode ? src : dst,
+ x->skip_encode ? src_stride : dst_stride,
+ dst, dst_stride, i, j, plane);
+ if (!x->skip_recode) {
+ vp9_highbd_subtract_block(32, 32, src_diff, diff_stride,
+ src, src_stride, dst, dst_stride, xd->bd);
+ highbd_fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin,
+ p->round, p->quant, p->quant_shift,
+ qcoeff, dqcoeff, pd->dequant,
+ p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ }
+ if (!x->skip_encode && *eob) {
+ vp9_highbd_idct32x32_add(dqcoeff, dst, dst_stride, *eob, xd->bd);
+ }
+ break;
+ case TX_16X16:
+ tx_type = get_tx_type(pd->plane_type, xd);
+ scan_order = &vp9_scan_orders[TX_16X16][tx_type];
+ mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
+ vp9_predict_intra_block(xd, block >> 4, bwl, TX_16X16, mode,
+ x->skip_encode ? src : dst,
+ x->skip_encode ? src_stride : dst_stride,
+ dst, dst_stride, i, j, plane);
+ if (!x->skip_recode) {
+ vp9_highbd_subtract_block(16, 16, src_diff, diff_stride,
+ src, src_stride, dst, dst_stride, xd->bd);
+ vp9_highbd_fht16x16(src_diff, coeff, diff_stride, tx_type);
+ vp9_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
+ p->quant, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ }
+ if (!x->skip_encode && *eob) {
+ vp9_highbd_iht16x16_add(tx_type, dqcoeff, dst, dst_stride,
+ *eob, xd->bd);
+ }
+ break;
+ case TX_8X8:
+ tx_type = get_tx_type(pd->plane_type, xd);
+ scan_order = &vp9_scan_orders[TX_8X8][tx_type];
+ mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
+ vp9_predict_intra_block(xd, block >> 2, bwl, TX_8X8, mode,
+ x->skip_encode ? src : dst,
+ x->skip_encode ? src_stride : dst_stride,
+ dst, dst_stride, i, j, plane);
+ if (!x->skip_recode) {
+ vp9_highbd_subtract_block(8, 8, src_diff, diff_stride,
+ src, src_stride, dst, dst_stride, xd->bd);
+ vp9_highbd_fht8x8(src_diff, coeff, diff_stride, tx_type);
+ vp9_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
+ p->quant, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ }
+ if (!x->skip_encode && *eob) {
+ vp9_highbd_iht8x8_add(tx_type, dqcoeff, dst, dst_stride, *eob,
+ xd->bd);
+ }
+ break;
+ case TX_4X4:
+ tx_type = get_tx_type_4x4(pd->plane_type, xd, block);
+ scan_order = &vp9_scan_orders[TX_4X4][tx_type];
+ mode = plane == 0 ? get_y_mode(xd->mi[0].src_mi, block) : mbmi->uv_mode;
+ vp9_predict_intra_block(xd, block, bwl, TX_4X4, mode,
+ x->skip_encode ? src : dst,
+ x->skip_encode ? src_stride : dst_stride,
+ dst, dst_stride, i, j, plane);
+
+ if (!x->skip_recode) {
+ vp9_highbd_subtract_block(4, 4, src_diff, diff_stride,
+ src, src_stride, dst, dst_stride, xd->bd);
+ if (tx_type != DCT_DCT)
+ vp9_highbd_fht4x4(src_diff, coeff, diff_stride, tx_type);
+ else
+ x->fwd_txm4x4(src_diff, coeff, diff_stride);
+ vp9_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
+ p->quant, p->quant_shift, qcoeff, dqcoeff,
+ pd->dequant, p->zbin_extra, eob,
+ scan_order->scan, scan_order->iscan);
+ }
+
+ if (!x->skip_encode && *eob) {
+ if (tx_type == DCT_DCT) {
+ // this is like vp9_short_idct4x4 but has a special case around
+ // eob<=1 which is significant (not just an optimization) for the
+ // lossless case.
+ x->highbd_itxm_add(dqcoeff, dst, dst_stride, *eob, xd->bd);
+ } else {
+ vp9_highbd_iht4x4_16_add(dqcoeff, dst, dst_stride, tx_type, xd->bd);
+ }
+ }
+ break;
+ default:
+ assert(0);
+ return;
+ }
+ if (*eob)
+ *(args->skip) = 0;
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
switch (tx_size) {
case TX_32X32:
scan_order = &vp9_default_scan_orders[TX_32X32];
diff --git a/source/libvpx/vp9/encoder/vp9_encodemb.h b/source/libvpx/vp9/encoder/vp9_encodemb.h
index 1999718..54d2b37 100644
--- a/source/libvpx/vp9/encoder/vp9_encodemb.h
+++ b/source/libvpx/vp9/encoder/vp9_encodemb.h
@@ -13,8 +13,6 @@
#include "./vpx_config.h"
#include "vp9/encoder/vp9_block.h"
-#include "vp9/encoder/vp9_encoder.h"
-#include "vp9/common/vp9_onyxc_int.h"
#ifdef __cplusplus
extern "C" {
diff --git a/source/libvpx/vp9/encoder/vp9_encoder.c b/source/libvpx/vp9/encoder/vp9_encoder.c
index 5f5af19..1758e3f 100644
--- a/source/libvpx/vp9/encoder/vp9_encoder.c
+++ b/source/libvpx/vp9/encoder/vp9_encoder.c
@@ -144,7 +144,6 @@
if (!init_done) {
vp9_rtcd();
- vp9_init_neighbors();
vp9_init_intra_predictors();
vp9_coef_tree_initialize();
vp9_tokenize_initialize();
@@ -226,6 +225,9 @@
}
vpx_memset(&cpi->svc.scaled_frames[0], 0,
MAX_LAG_BUFFERS * sizeof(cpi->svc.scaled_frames[0]));
+
+ vp9_free_frame_buffer(&cpi->svc.empty_frame.img);
+ vpx_memset(&cpi->svc.empty_frame, 0, sizeof(cpi->svc.empty_frame));
}
static void save_coding_context(VP9_COMP *cpi) {
@@ -586,8 +588,6 @@
cpi->ref_frame_flags = 0;
init_buffer_indices(cpi);
-
- set_tile_limits(cpi);
}
static void set_rc_buffer_sizes(RATE_CONTROL *rc,
@@ -604,6 +604,612 @@
: maximum * bandwidth / 1000;
}
+#if CONFIG_VP9_HIGHBITDEPTH
+#define HIGHBD_BFP(BT, SDF, SDAF, VF, SVF, SVAF, SDX3F, SDX8F, SDX4DF) \
+ cpi->fn_ptr[BT].sdf = SDF; \
+ cpi->fn_ptr[BT].sdaf = SDAF; \
+ cpi->fn_ptr[BT].vf = VF; \
+ cpi->fn_ptr[BT].svf = SVF; \
+ cpi->fn_ptr[BT].svaf = SVAF; \
+ cpi->fn_ptr[BT].sdx3f = SDX3F; \
+ cpi->fn_ptr[BT].sdx8f = SDX8F; \
+ cpi->fn_ptr[BT].sdx4df = SDX4DF;
+
+#define MAKE_BFP_SAD_WRAPPER(fnname) \
+static unsigned int fnname##_bits8(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride); \
+} \
+static unsigned int fnname##_bits10(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride) >> 2; \
+} \
+static unsigned int fnname##_bits12(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride) >> 4; \
+}
+
+#define MAKE_BFP_SADAVG_WRAPPER(fnname) static unsigned int \
+fnname##_bits8(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ const uint8_t *second_pred) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride, second_pred); \
+} \
+static unsigned int fnname##_bits10(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ const uint8_t *second_pred) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride, \
+ second_pred) >> 2; \
+} \
+static unsigned int fnname##_bits12(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ const uint8_t *second_pred) { \
+ return fnname(src_ptr, source_stride, ref_ptr, ref_stride, \
+ second_pred) >> 4; \
+}
+
+#define MAKE_BFP_SAD3_WRAPPER(fnname) \
+static void fnname##_bits8(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+} \
+static void fnname##_bits10(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 3; i++) \
+ sad_array[i] >>= 2; \
+} \
+static void fnname##_bits12(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 3; i++) \
+ sad_array[i] >>= 4; \
+}
+
+#define MAKE_BFP_SAD8_WRAPPER(fnname) \
+static void fnname##_bits8(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+} \
+static void fnname##_bits10(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 8; i++) \
+ sad_array[i] >>= 2; \
+} \
+static void fnname##_bits12(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 8; i++) \
+ sad_array[i] >>= 4; \
+}
+#define MAKE_BFP_SAD4D_WRAPPER(fnname) \
+static void fnname##_bits8(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t* const ref_ptr[], \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+} \
+static void fnname##_bits10(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t* const ref_ptr[], \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 4; i++) \
+ sad_array[i] >>= 2; \
+} \
+static void fnname##_bits12(const uint8_t *src_ptr, \
+ int source_stride, \
+ const uint8_t* const ref_ptr[], \
+ int ref_stride, \
+ unsigned int *sad_array) { \
+ int i; \
+ fnname(src_ptr, source_stride, ref_ptr, ref_stride, sad_array); \
+ for (i = 0; i < 4; i++) \
+ sad_array[i] >>= 4; \
+}
+
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad32x16)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad32x16_avg)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad32x16x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad16x32)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad16x32_avg)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad16x32x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad64x32)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad64x32_avg)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad64x32x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad32x64)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad32x64_avg)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad32x64x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad32x32)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad32x32_avg)
+MAKE_BFP_SAD3_WRAPPER(vp9_highbd_sad32x32x3)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad32x32x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad32x32x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad64x64)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad64x64_avg)
+MAKE_BFP_SAD3_WRAPPER(vp9_highbd_sad64x64x3)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad64x64x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad64x64x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad16x16)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad16x16_avg)
+MAKE_BFP_SAD3_WRAPPER(vp9_highbd_sad16x16x3)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad16x16x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad16x16x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad16x8)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad16x8_avg)
+MAKE_BFP_SAD3_WRAPPER(vp9_highbd_sad16x8x3)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad16x8x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad16x8x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad8x16)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad8x16_avg)
+MAKE_BFP_SAD3_WRAPPER(vp9_highbd_sad8x16x3)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad8x16x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad8x16x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad8x8)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad8x8_avg)
+MAKE_BFP_SAD3_WRAPPER(vp9_highbd_sad8x8x3)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad8x8x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad8x8x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad8x4)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad8x4_avg)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad8x4x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad8x4x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad4x8)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad4x8_avg)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad4x8x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad4x8x4d)
+MAKE_BFP_SAD_WRAPPER(vp9_highbd_sad4x4)
+MAKE_BFP_SADAVG_WRAPPER(vp9_highbd_sad4x4_avg)
+MAKE_BFP_SAD3_WRAPPER(vp9_highbd_sad4x4x3)
+MAKE_BFP_SAD8_WRAPPER(vp9_highbd_sad4x4x8)
+MAKE_BFP_SAD4D_WRAPPER(vp9_highbd_sad4x4x4d)
+
+static void highbd_set_var_fns(VP9_COMP *const cpi) {
+ VP9_COMMON *const cm = &cpi->common;
+ if (cm->use_highbitdepth) {
+ switch (cm->bit_depth) {
+ case VPX_BITS_8:
+ HIGHBD_BFP(BLOCK_32X16,
+ vp9_highbd_sad32x16_bits8,
+ vp9_highbd_sad32x16_avg_bits8,
+ vp9_highbd_variance32x16,
+ vp9_highbd_sub_pixel_variance32x16,
+ vp9_highbd_sub_pixel_avg_variance32x16,
+ NULL,
+ NULL,
+ vp9_highbd_sad32x16x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_16X32,
+ vp9_highbd_sad16x32_bits8,
+ vp9_highbd_sad16x32_avg_bits8,
+ vp9_highbd_variance16x32,
+ vp9_highbd_sub_pixel_variance16x32,
+ vp9_highbd_sub_pixel_avg_variance16x32,
+ NULL,
+ NULL,
+ vp9_highbd_sad16x32x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_64X32,
+ vp9_highbd_sad64x32_bits8,
+ vp9_highbd_sad64x32_avg_bits8,
+ vp9_highbd_variance64x32,
+ vp9_highbd_sub_pixel_variance64x32,
+ vp9_highbd_sub_pixel_avg_variance64x32,
+ NULL,
+ NULL,
+ vp9_highbd_sad64x32x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_32X64,
+ vp9_highbd_sad32x64_bits8,
+ vp9_highbd_sad32x64_avg_bits8,
+ vp9_highbd_variance32x64,
+ vp9_highbd_sub_pixel_variance32x64,
+ vp9_highbd_sub_pixel_avg_variance32x64,
+ NULL,
+ NULL,
+ vp9_highbd_sad32x64x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_32X32,
+ vp9_highbd_sad32x32_bits8,
+ vp9_highbd_sad32x32_avg_bits8,
+ vp9_highbd_variance32x32,
+ vp9_highbd_sub_pixel_variance32x32,
+ vp9_highbd_sub_pixel_avg_variance32x32,
+ vp9_highbd_sad32x32x3_bits8,
+ vp9_highbd_sad32x32x8_bits8,
+ vp9_highbd_sad32x32x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_64X64,
+ vp9_highbd_sad64x64_bits8,
+ vp9_highbd_sad64x64_avg_bits8,
+ vp9_highbd_variance64x64,
+ vp9_highbd_sub_pixel_variance64x64,
+ vp9_highbd_sub_pixel_avg_variance64x64,
+ vp9_highbd_sad64x64x3_bits8,
+ vp9_highbd_sad64x64x8_bits8,
+ vp9_highbd_sad64x64x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_16X16,
+ vp9_highbd_sad16x16_bits8,
+ vp9_highbd_sad16x16_avg_bits8,
+ vp9_highbd_variance16x16,
+ vp9_highbd_sub_pixel_variance16x16,
+ vp9_highbd_sub_pixel_avg_variance16x16,
+ vp9_highbd_sad16x16x3_bits8,
+ vp9_highbd_sad16x16x8_bits8,
+ vp9_highbd_sad16x16x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_16X8,
+ vp9_highbd_sad16x8_bits8,
+ vp9_highbd_sad16x8_avg_bits8,
+ vp9_highbd_variance16x8,
+ vp9_highbd_sub_pixel_variance16x8,
+ vp9_highbd_sub_pixel_avg_variance16x8,
+ vp9_highbd_sad16x8x3_bits8,
+ vp9_highbd_sad16x8x8_bits8,
+ vp9_highbd_sad16x8x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_8X16,
+ vp9_highbd_sad8x16_bits8,
+ vp9_highbd_sad8x16_avg_bits8,
+ vp9_highbd_variance8x16,
+ vp9_highbd_sub_pixel_variance8x16,
+ vp9_highbd_sub_pixel_avg_variance8x16,
+ vp9_highbd_sad8x16x3_bits8,
+ vp9_highbd_sad8x16x8_bits8,
+ vp9_highbd_sad8x16x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_8X8,
+ vp9_highbd_sad8x8_bits8,
+ vp9_highbd_sad8x8_avg_bits8,
+ vp9_highbd_variance8x8,
+ vp9_highbd_sub_pixel_variance8x8,
+ vp9_highbd_sub_pixel_avg_variance8x8,
+ vp9_highbd_sad8x8x3_bits8,
+ vp9_highbd_sad8x8x8_bits8,
+ vp9_highbd_sad8x8x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_8X4,
+ vp9_highbd_sad8x4_bits8,
+ vp9_highbd_sad8x4_avg_bits8,
+ vp9_highbd_variance8x4,
+ vp9_highbd_sub_pixel_variance8x4,
+ vp9_highbd_sub_pixel_avg_variance8x4,
+ NULL,
+ vp9_highbd_sad8x4x8_bits8,
+ vp9_highbd_sad8x4x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_4X8,
+ vp9_highbd_sad4x8_bits8,
+ vp9_highbd_sad4x8_avg_bits8,
+ vp9_highbd_variance4x8,
+ vp9_highbd_sub_pixel_variance4x8,
+ vp9_highbd_sub_pixel_avg_variance4x8,
+ NULL,
+ vp9_highbd_sad4x8x8_bits8,
+ vp9_highbd_sad4x8x4d_bits8)
+
+ HIGHBD_BFP(BLOCK_4X4,
+ vp9_highbd_sad4x4_bits8,
+ vp9_highbd_sad4x4_avg_bits8,
+ vp9_highbd_variance4x4,
+ vp9_highbd_sub_pixel_variance4x4,
+ vp9_highbd_sub_pixel_avg_variance4x4,
+ vp9_highbd_sad4x4x3_bits8,
+ vp9_highbd_sad4x4x8_bits8,
+ vp9_highbd_sad4x4x4d_bits8)
+ break;
+
+ case VPX_BITS_10:
+ HIGHBD_BFP(BLOCK_32X16,
+ vp9_highbd_sad32x16_bits10,
+ vp9_highbd_sad32x16_avg_bits10,
+ vp9_highbd_10_variance32x16,
+ vp9_highbd_10_sub_pixel_variance32x16,
+ vp9_highbd_10_sub_pixel_avg_variance32x16,
+ NULL,
+ NULL,
+ vp9_highbd_sad32x16x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_16X32,
+ vp9_highbd_sad16x32_bits10,
+ vp9_highbd_sad16x32_avg_bits10,
+ vp9_highbd_10_variance16x32,
+ vp9_highbd_10_sub_pixel_variance16x32,
+ vp9_highbd_10_sub_pixel_avg_variance16x32,
+ NULL,
+ NULL,
+ vp9_highbd_sad16x32x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_64X32,
+ vp9_highbd_sad64x32_bits10,
+ vp9_highbd_sad64x32_avg_bits10,
+ vp9_highbd_10_variance64x32,
+ vp9_highbd_10_sub_pixel_variance64x32,
+ vp9_highbd_10_sub_pixel_avg_variance64x32,
+ NULL,
+ NULL,
+ vp9_highbd_sad64x32x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_32X64,
+ vp9_highbd_sad32x64_bits10,
+ vp9_highbd_sad32x64_avg_bits10,
+ vp9_highbd_10_variance32x64,
+ vp9_highbd_10_sub_pixel_variance32x64,
+ vp9_highbd_10_sub_pixel_avg_variance32x64,
+ NULL,
+ NULL,
+ vp9_highbd_sad32x64x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_32X32,
+ vp9_highbd_sad32x32_bits10,
+ vp9_highbd_sad32x32_avg_bits10,
+ vp9_highbd_10_variance32x32,
+ vp9_highbd_10_sub_pixel_variance32x32,
+ vp9_highbd_10_sub_pixel_avg_variance32x32,
+ vp9_highbd_sad32x32x3_bits10,
+ vp9_highbd_sad32x32x8_bits10,
+ vp9_highbd_sad32x32x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_64X64,
+ vp9_highbd_sad64x64_bits10,
+ vp9_highbd_sad64x64_avg_bits10,
+ vp9_highbd_10_variance64x64,
+ vp9_highbd_10_sub_pixel_variance64x64,
+ vp9_highbd_10_sub_pixel_avg_variance64x64,
+ vp9_highbd_sad64x64x3_bits10,
+ vp9_highbd_sad64x64x8_bits10,
+ vp9_highbd_sad64x64x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_16X16,
+ vp9_highbd_sad16x16_bits10,
+ vp9_highbd_sad16x16_avg_bits10,
+ vp9_highbd_10_variance16x16,
+ vp9_highbd_10_sub_pixel_variance16x16,
+ vp9_highbd_10_sub_pixel_avg_variance16x16,
+ vp9_highbd_sad16x16x3_bits10,
+ vp9_highbd_sad16x16x8_bits10,
+ vp9_highbd_sad16x16x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_16X8,
+ vp9_highbd_sad16x8_bits10,
+ vp9_highbd_sad16x8_avg_bits10,
+ vp9_highbd_10_variance16x8,
+ vp9_highbd_10_sub_pixel_variance16x8,
+ vp9_highbd_10_sub_pixel_avg_variance16x8,
+ vp9_highbd_sad16x8x3_bits10,
+ vp9_highbd_sad16x8x8_bits10,
+ vp9_highbd_sad16x8x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_8X16,
+ vp9_highbd_sad8x16_bits10,
+ vp9_highbd_sad8x16_avg_bits10,
+ vp9_highbd_10_variance8x16,
+ vp9_highbd_10_sub_pixel_variance8x16,
+ vp9_highbd_10_sub_pixel_avg_variance8x16,
+ vp9_highbd_sad8x16x3_bits10,
+ vp9_highbd_sad8x16x8_bits10,
+ vp9_highbd_sad8x16x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_8X8,
+ vp9_highbd_sad8x8_bits10,
+ vp9_highbd_sad8x8_avg_bits10,
+ vp9_highbd_10_variance8x8,
+ vp9_highbd_10_sub_pixel_variance8x8,
+ vp9_highbd_10_sub_pixel_avg_variance8x8,
+ vp9_highbd_sad8x8x3_bits10,
+ vp9_highbd_sad8x8x8_bits10,
+ vp9_highbd_sad8x8x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_8X4,
+ vp9_highbd_sad8x4_bits10,
+ vp9_highbd_sad8x4_avg_bits10,
+ vp9_highbd_10_variance8x4,
+ vp9_highbd_10_sub_pixel_variance8x4,
+ vp9_highbd_10_sub_pixel_avg_variance8x4,
+ NULL,
+ vp9_highbd_sad8x4x8_bits10,
+ vp9_highbd_sad8x4x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_4X8,
+ vp9_highbd_sad4x8_bits10,
+ vp9_highbd_sad4x8_avg_bits10,
+ vp9_highbd_10_variance4x8,
+ vp9_highbd_10_sub_pixel_variance4x8,
+ vp9_highbd_10_sub_pixel_avg_variance4x8,
+ NULL,
+ vp9_highbd_sad4x8x8_bits10,
+ vp9_highbd_sad4x8x4d_bits10)
+
+ HIGHBD_BFP(BLOCK_4X4,
+ vp9_highbd_sad4x4_bits10,
+ vp9_highbd_sad4x4_avg_bits10,
+ vp9_highbd_10_variance4x4,
+ vp9_highbd_10_sub_pixel_variance4x4,
+ vp9_highbd_10_sub_pixel_avg_variance4x4,
+ vp9_highbd_sad4x4x3_bits10,
+ vp9_highbd_sad4x4x8_bits10,
+ vp9_highbd_sad4x4x4d_bits10)
+ break;
+
+ case VPX_BITS_12:
+ HIGHBD_BFP(BLOCK_32X16,
+ vp9_highbd_sad32x16_bits12,
+ vp9_highbd_sad32x16_avg_bits12,
+ vp9_highbd_12_variance32x16,
+ vp9_highbd_12_sub_pixel_variance32x16,
+ vp9_highbd_12_sub_pixel_avg_variance32x16,
+ NULL,
+ NULL,
+ vp9_highbd_sad32x16x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_16X32,
+ vp9_highbd_sad16x32_bits12,
+ vp9_highbd_sad16x32_avg_bits12,
+ vp9_highbd_12_variance16x32,
+ vp9_highbd_12_sub_pixel_variance16x32,
+ vp9_highbd_12_sub_pixel_avg_variance16x32,
+ NULL,
+ NULL,
+ vp9_highbd_sad16x32x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_64X32,
+ vp9_highbd_sad64x32_bits12,
+ vp9_highbd_sad64x32_avg_bits12,
+ vp9_highbd_12_variance64x32,
+ vp9_highbd_12_sub_pixel_variance64x32,
+ vp9_highbd_12_sub_pixel_avg_variance64x32,
+ NULL,
+ NULL,
+ vp9_highbd_sad64x32x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_32X64,
+ vp9_highbd_sad32x64_bits12,
+ vp9_highbd_sad32x64_avg_bits12,
+ vp9_highbd_12_variance32x64,
+ vp9_highbd_12_sub_pixel_variance32x64,
+ vp9_highbd_12_sub_pixel_avg_variance32x64,
+ NULL,
+ NULL,
+ vp9_highbd_sad32x64x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_32X32,
+ vp9_highbd_sad32x32_bits12,
+ vp9_highbd_sad32x32_avg_bits12,
+ vp9_highbd_12_variance32x32,
+ vp9_highbd_12_sub_pixel_variance32x32,
+ vp9_highbd_12_sub_pixel_avg_variance32x32,
+ vp9_highbd_sad32x32x3_bits12,
+ vp9_highbd_sad32x32x8_bits12,
+ vp9_highbd_sad32x32x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_64X64,
+ vp9_highbd_sad64x64_bits12,
+ vp9_highbd_sad64x64_avg_bits12,
+ vp9_highbd_12_variance64x64,
+ vp9_highbd_12_sub_pixel_variance64x64,
+ vp9_highbd_12_sub_pixel_avg_variance64x64,
+ vp9_highbd_sad64x64x3_bits12,
+ vp9_highbd_sad64x64x8_bits12,
+ vp9_highbd_sad64x64x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_16X16,
+ vp9_highbd_sad16x16_bits12,
+ vp9_highbd_sad16x16_avg_bits12,
+ vp9_highbd_12_variance16x16,
+ vp9_highbd_12_sub_pixel_variance16x16,
+ vp9_highbd_12_sub_pixel_avg_variance16x16,
+ vp9_highbd_sad16x16x3_bits12,
+ vp9_highbd_sad16x16x8_bits12,
+ vp9_highbd_sad16x16x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_16X8,
+ vp9_highbd_sad16x8_bits12,
+ vp9_highbd_sad16x8_avg_bits12,
+ vp9_highbd_12_variance16x8,
+ vp9_highbd_12_sub_pixel_variance16x8,
+ vp9_highbd_12_sub_pixel_avg_variance16x8,
+ vp9_highbd_sad16x8x3_bits12,
+ vp9_highbd_sad16x8x8_bits12,
+ vp9_highbd_sad16x8x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_8X16,
+ vp9_highbd_sad8x16_bits12,
+ vp9_highbd_sad8x16_avg_bits12,
+ vp9_highbd_12_variance8x16,
+ vp9_highbd_12_sub_pixel_variance8x16,
+ vp9_highbd_12_sub_pixel_avg_variance8x16,
+ vp9_highbd_sad8x16x3_bits12,
+ vp9_highbd_sad8x16x8_bits12,
+ vp9_highbd_sad8x16x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_8X8,
+ vp9_highbd_sad8x8_bits12,
+ vp9_highbd_sad8x8_avg_bits12,
+ vp9_highbd_12_variance8x8,
+ vp9_highbd_12_sub_pixel_variance8x8,
+ vp9_highbd_12_sub_pixel_avg_variance8x8,
+ vp9_highbd_sad8x8x3_bits12,
+ vp9_highbd_sad8x8x8_bits12,
+ vp9_highbd_sad8x8x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_8X4,
+ vp9_highbd_sad8x4_bits12,
+ vp9_highbd_sad8x4_avg_bits12,
+ vp9_highbd_12_variance8x4,
+ vp9_highbd_12_sub_pixel_variance8x4,
+ vp9_highbd_12_sub_pixel_avg_variance8x4,
+ NULL,
+ vp9_highbd_sad8x4x8_bits12,
+ vp9_highbd_sad8x4x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_4X8,
+ vp9_highbd_sad4x8_bits12,
+ vp9_highbd_sad4x8_avg_bits12,
+ vp9_highbd_12_variance4x8,
+ vp9_highbd_12_sub_pixel_variance4x8,
+ vp9_highbd_12_sub_pixel_avg_variance4x8,
+ NULL,
+ vp9_highbd_sad4x8x8_bits12,
+ vp9_highbd_sad4x8x4d_bits12)
+
+ HIGHBD_BFP(BLOCK_4X4,
+ vp9_highbd_sad4x4_bits12,
+ vp9_highbd_sad4x4_avg_bits12,
+ vp9_highbd_12_variance4x4,
+ vp9_highbd_12_sub_pixel_variance4x4,
+ vp9_highbd_12_sub_pixel_avg_variance4x4,
+ vp9_highbd_sad4x4x3_bits12,
+ vp9_highbd_sad4x4x8_bits12,
+ vp9_highbd_sad4x4x4d_bits12)
+ break;
+
+ default:
+ assert(0 && "cm->bit_depth should be VPX_BITS_8, "
+ "VPX_BITS_10 or VPX_BITS_12");
+ }
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
void vp9_change_config(struct VP9_COMP *cpi, const VP9EncoderConfig *oxcf) {
VP9_COMMON *const cm = &cpi->common;
RATE_CONTROL *const rc = &cpi->rc;
@@ -619,10 +1225,8 @@
cpi->oxcf = *oxcf;
#if CONFIG_VP9_HIGHBITDEPTH
- if (cpi->oxcf.use_highbitdepth) {
- cpi->mb.e_mbd.bd = (int)cm->bit_depth;
- }
-#endif
+ cpi->mb.e_mbd.bd = (int)cm->bit_depth;
+#endif // CONFIG_VP9_HIGHBITDEPTH
rc->baseline_gf_interval = DEFAULT_GF_INTERVAL;
@@ -693,6 +1297,10 @@
cpi->ext_refresh_frame_flags_pending = 0;
cpi->ext_refresh_frame_context_pending = 0;
+#if CONFIG_VP9_HIGHBITDEPTH
+ highbd_set_var_fns(cpi);
+#endif
+
#if CONFIG_VP9_TEMPORAL_DENOISING
if (cpi->oxcf.noise_sensitivity > 0) {
vp9_denoiser_alloc(&(cpi->denoiser), cm->width, cm->height,
@@ -772,7 +1380,7 @@
vp9_rc_init(&cpi->oxcf, oxcf->pass, &cpi->rc);
cm->current_video_frame = 0;
- cpi->skippable_frame = 0;
+ cpi->partition_search_skippable_frame = 0;
// Create the encoder segmentation map and set all entries to 0
CHECK_MEM_ERROR(cm, cpi->segmentation_map,
@@ -1066,6 +1674,10 @@
vp9_sub_pixel_avg_variance4x4,
vp9_sad4x4x3, vp9_sad4x4x8, vp9_sad4x4x4d)
+#if CONFIG_VP9_HIGHBITDEPTH
+ highbd_set_var_fns(cpi);
+#endif
+
/* vp9_init_quantizer() is first called here. Add check in
* vp9_frame_init_quantizer() so that vp9_init_quantizer is only
* called later when needed. This will avoid unnecessary calls of
@@ -1098,15 +1710,16 @@
- cpi->first_time_stamp_ever) / 10000000.000;
double total_encode_time = (cpi->time_receive_data +
cpi->time_compress_data) / 1000.000;
- double dr = (double)cpi->bytes * (double) 8 / (double)1000
- / time_encoded;
+ const double dr =
+ (double)cpi->bytes * (double) 8 / (double)1000 / time_encoded;
+ const double peak = (double)((1 << cpi->oxcf.input_bit_depth) - 1);
if (cpi->b_calculate_psnr) {
const double total_psnr =
- vpx_sse_to_psnr((double)cpi->total_samples, 255.0,
+ vpx_sse_to_psnr((double)cpi->total_samples, peak,
(double)cpi->total_sq_error);
const double totalp_psnr =
- vpx_sse_to_psnr((double)cpi->totalp_samples, 255.0,
+ vpx_sse_to_psnr((double)cpi->totalp_samples, peak,
(double)cpi->totalp_sq_error);
const double total_ssim = 100 * pow(cpi->summed_quality /
cpi->summed_weights, 8.0);
@@ -1193,6 +1806,7 @@
#endif
}
+
static int64_t get_sse(const uint8_t *a, int a_stride,
const uint8_t *b, int b_stride,
int width, int height) {
@@ -1234,6 +1848,63 @@
return total_sse;
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static int64_t highbd_get_sse_shift(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int width, int height,
+ unsigned int input_shift) {
+ const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
+ const uint16_t *b = CONVERT_TO_SHORTPTR(b8);
+ int64_t total_sse = 0;
+ int x, y;
+ for (y = 0; y < height; ++y) {
+ for (x = 0; x < width; ++x) {
+ int64_t diff;
+ diff = (a[x] >> input_shift) - (b[x] >> input_shift);
+ total_sse += diff * diff;
+ }
+ a += a_stride;
+ b += b_stride;
+ }
+ return total_sse;
+}
+
+static int64_t highbd_get_sse(const uint8_t *a, int a_stride,
+ const uint8_t *b, int b_stride,
+ int width, int height) {
+ int64_t total_sse = 0;
+ int x, y;
+ const int dw = width % 16;
+ const int dh = height % 16;
+ unsigned int sse = 0;
+ int sum = 0;
+ if (dw > 0) {
+ highbd_variance(&a[width - dw], a_stride, &b[width - dw], b_stride,
+ dw, height, &sse, &sum);
+ total_sse += sse;
+ }
+ if (dh > 0) {
+ highbd_variance(&a[(height - dh) * a_stride], a_stride,
+ &b[(height - dh) * b_stride], b_stride,
+ width - dw, dh, &sse, &sum);
+ total_sse += sse;
+ }
+ for (y = 0; y < height / 16; ++y) {
+ const uint8_t *pa = a;
+ const uint8_t *pb = b;
+ for (x = 0; x < width / 16; ++x) {
+ vp9_highbd_mse16x16(pa, a_stride, pb, b_stride, &sse);
+ total_sse += sse;
+ pa += 16;
+ pb += 16;
+ }
+ a += 16 * a_stride;
+ b += 16 * b_stride;
+ }
+ return total_sse;
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
typedef struct {
double psnr[4]; // total/y/u/v
uint64_t sse[4]; // total/y/u/v
@@ -1242,6 +1913,7 @@
static void calc_psnr(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b,
PSNR_STATS *psnr) {
+ static const double peak = 255.0;
const int widths[3] = {a->y_width, a->uv_width, a->uv_width };
const int heights[3] = {a->y_height, a->uv_height, a->uv_height};
const uint8_t *a_planes[3] = {a->y_buffer, a->u_buffer, a->v_buffer };
@@ -1261,7 +1933,7 @@
w, h);
psnr->sse[1 + i] = sse;
psnr->samples[1 + i] = samples;
- psnr->psnr[1 + i] = vpx_sse_to_psnr(samples, 255.0, (double)sse);
+ psnr->psnr[1 + i] = vpx_sse_to_psnr(samples, peak, (double)sse);
total_sse += sse;
total_samples += samples;
@@ -1269,15 +1941,73 @@
psnr->sse[0] = total_sse;
psnr->samples[0] = total_samples;
- psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, 255.0,
+ psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
(double)total_sse);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static void calc_highbd_psnr(const YV12_BUFFER_CONFIG *a,
+ const YV12_BUFFER_CONFIG *b,
+ PSNR_STATS *psnr,
+ unsigned int bit_depth,
+ unsigned int in_bit_depth) {
+ const int widths[3] = {a->y_width, a->uv_width, a->uv_width };
+ const int heights[3] = {a->y_height, a->uv_height, a->uv_height};
+ const uint8_t *a_planes[3] = {a->y_buffer, a->u_buffer, a->v_buffer };
+ const int a_strides[3] = {a->y_stride, a->uv_stride, a->uv_stride};
+ const uint8_t *b_planes[3] = {b->y_buffer, b->u_buffer, b->v_buffer };
+ const int b_strides[3] = {b->y_stride, b->uv_stride, b->uv_stride};
+ int i;
+ uint64_t total_sse = 0;
+ uint32_t total_samples = 0;
+ const double peak = (double)((1 << in_bit_depth) - 1);
+ const unsigned int input_shift = bit_depth - in_bit_depth;
+
+ for (i = 0; i < 3; ++i) {
+ const int w = widths[i];
+ const int h = heights[i];
+ const uint32_t samples = w * h;
+ uint64_t sse;
+ if (a->flags & YV12_FLAG_HIGHBITDEPTH) {
+ if (input_shift) {
+ sse = highbd_get_sse_shift(a_planes[i], a_strides[i],
+ b_planes[i], b_strides[i], w, h,
+ input_shift);
+ } else {
+ sse = highbd_get_sse(a_planes[i], a_strides[i],
+ b_planes[i], b_strides[i], w, h);
+ }
+ } else {
+ sse = get_sse(a_planes[i], a_strides[i],
+ b_planes[i], b_strides[i],
+ w, h);
+ }
+ psnr->sse[1 + i] = sse;
+ psnr->samples[1 + i] = samples;
+ psnr->psnr[1 + i] = vpx_sse_to_psnr(samples, peak, (double)sse);
+
+ total_sse += sse;
+ total_samples += samples;
+ }
+
+ psnr->sse[0] = total_sse;
+ psnr->samples[0] = total_samples;
+ psnr->psnr[0] = vpx_sse_to_psnr((double)total_samples, peak,
+ (double)total_sse);
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static void generate_psnr_packet(VP9_COMP *cpi) {
struct vpx_codec_cx_pkt pkt;
int i;
PSNR_STATS psnr;
+#if CONFIG_VP9_HIGHBITDEPTH
+ calc_highbd_psnr(cpi->Source, cpi->common.frame_to_show, &psnr,
+ cpi->mb.e_mbd.bd, cpi->oxcf.input_bit_depth);
+#else
calc_psnr(cpi->Source, cpi->common.frame_to_show, &psnr);
+#endif
+
for (i = 0; i < 4; ++i) {
pkt.data.psnr.samples[i] = psnr.samples[i];
pkt.data.psnr.sse[i] = psnr.sse[i];
@@ -1386,6 +2116,36 @@
uint8_t *src = s->y_buffer;
int h = cm->height;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (s->flags & YV12_FLAG_HIGHBITDEPTH) {
+ uint16_t *src16 = CONVERT_TO_SHORTPTR(s->y_buffer);
+
+ do {
+ fwrite(src16, s->y_width, 2, yuv_rec_file);
+ src16 += s->y_stride;
+ } while (--h);
+
+ src16 = CONVERT_TO_SHORTPTR(s->u_buffer);
+ h = s->uv_height;
+
+ do {
+ fwrite(src16, s->uv_width, 2, yuv_rec_file);
+ src16 += s->uv_stride;
+ } while (--h);
+
+ src16 = CONVERT_TO_SHORTPTR(s->v_buffer);
+ h = s->uv_height;
+
+ do {
+ fwrite(src16, s->uv_width, 2, yuv_rec_file);
+ src16 += s->uv_stride;
+ } while (--h);
+
+ fflush(yuv_rec_file);
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
do {
fwrite(src, s->y_width, 1, yuv_rec_file);
src += s->y_stride;
@@ -1411,8 +2171,14 @@
}
#endif
+#if CONFIG_VP9_HIGHBITDEPTH
+static void scale_and_extend_frame_nonnormative(const YV12_BUFFER_CONFIG *src,
+ YV12_BUFFER_CONFIG *dst,
+ int bd) {
+#else
static void scale_and_extend_frame_nonnormative(const YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst) {
+#endif // CONFIG_VP9_HIGHBITDEPTH
// TODO(dkovalev): replace YV12_BUFFER_CONFIG with vpx_image_t
int i;
const uint8_t *const srcs[3] = {src->y_buffer, src->u_buffer, src->v_buffer};
@@ -1428,15 +2194,31 @@
const int dst_heights[3] = {dst->y_crop_height, dst->uv_crop_height,
dst->uv_crop_height};
- for (i = 0; i < MAX_MB_PLANE; ++i)
+ for (i = 0; i < MAX_MB_PLANE; ++i) {
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (src->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_resize_plane(srcs[i], src_heights[i], src_widths[i],
+ src_strides[i], dsts[i], dst_heights[i],
+ dst_widths[i], dst_strides[i], bd);
+ } else {
+ vp9_resize_plane(srcs[i], src_heights[i], src_widths[i], src_strides[i],
+ dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]);
+ }
+#else
vp9_resize_plane(srcs[i], src_heights[i], src_widths[i], src_strides[i],
dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]);
-
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ }
vp9_extend_frame_borders(dst);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static void scale_and_extend_frame(const YV12_BUFFER_CONFIG *src,
+ YV12_BUFFER_CONFIG *dst, int bd) {
+#else
static void scale_and_extend_frame(const YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst) {
+#endif // CONFIG_VP9_HIGHBITDEPTH
const int src_w = src->y_crop_width;
const int src_h = src->y_crop_height;
const int dst_w = dst->y_crop_width;
@@ -1460,10 +2242,24 @@
src_stride + (x / factor) * src_w / dst_w;
uint8_t *dst_ptr = dsts[i] + (y / factor) * dst_stride + (x / factor);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (src->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_convolve8(src_ptr, src_stride, dst_ptr, dst_stride,
+ kernel[x_q4 & 0xf], 16 * src_w / dst_w,
+ kernel[y_q4 & 0xf], 16 * src_h / dst_h,
+ 16 / factor, 16 / factor, bd);
+ } else {
+ vp9_convolve8(src_ptr, src_stride, dst_ptr, dst_stride,
+ kernel[x_q4 & 0xf], 16 * src_w / dst_w,
+ kernel[y_q4 & 0xf], 16 * src_h / dst_h,
+ 16 / factor, 16 / factor);
+ }
+#else
vp9_convolve8(src_ptr, src_stride, dst_ptr, dst_stride,
kernel[x_q4 & 0xf], 16 * src_w / dst_w,
kernel[y_q4 & 0xf], 16 * src_h / dst_h,
16 / factor, 16 / factor);
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
}
}
@@ -1632,9 +2428,14 @@
cm->subsampling_x, cm->subsampling_y,
#if CONFIG_VP9_HIGHBITDEPTH
cm->use_highbitdepth,
-#endif
+#endif // CONFIG_VP9_HIGHBITDEPTH
VP9_ENC_BORDER_IN_PIXELS, NULL, NULL, NULL);
+#if CONFIG_VP9_HIGHBITDEPTH
+ scale_and_extend_frame(ref, &cm->frame_bufs[new_fb].buf,
+ (int)cm->bit_depth);
+#else
scale_and_extend_frame(ref, &cm->frame_bufs[new_fb].buf);
+#endif // CONFIG_VP9_HIGHBITDEPTH
cpi->scaled_ref_idx[ref_frame - 1] = new_fb;
} else {
cpi->scaled_ref_idx[ref_frame - 1] = idx;
@@ -1698,11 +2499,12 @@
cpi->rc.total_target_vs_actual,
(cpi->rc.starting_buffer_level - cpi->rc.bits_off_target),
cpi->rc.total_actual_bits, cm->base_qindex,
- vp9_convert_qindex_to_q(cm->base_qindex),
- (double)vp9_dc_quant(cm->base_qindex, 0) / 4.0,
- vp9_convert_qindex_to_q(cpi->twopass.active_worst_quality),
+ vp9_convert_qindex_to_q(cm->base_qindex, cm->bit_depth),
+ (double)vp9_dc_quant(cm->base_qindex, 0, cm->bit_depth) / 4.0,
+ vp9_convert_qindex_to_q(cpi->twopass.active_worst_quality,
+ cm->bit_depth),
cpi->rc.avg_q,
- vp9_convert_qindex_to_q(cpi->oxcf.cq_level),
+ vp9_convert_qindex_to_q(cpi->oxcf.cq_level, cm->bit_depth),
cpi->refresh_last_frame, cpi->refresh_golden_frame,
cpi->refresh_alt_ref_frame, cm->frame_type, cpi->rc.gfu_boost,
cpi->twopass.bits_left,
@@ -1824,11 +2626,22 @@
rc->this_key_frame_forced &&
(rc->projected_frame_size < rc->max_frame_bandwidth)) {
int last_q = q;
- int kf_err = vp9_get_y_sse(cpi->Source, get_frame_new_buffer(cm));
+ int kf_err;
int high_err_target = cpi->ambient_err;
int low_err_target = cpi->ambient_err >> 1;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ kf_err = vp9_highbd_get_y_sse(cpi->Source, get_frame_new_buffer(cm),
+ cm->bit_depth);
+ } else {
+ kf_err = vp9_get_y_sse(cpi->Source, get_frame_new_buffer(cm));
+ }
+#else
+ kf_err = vp9_get_y_sse(cpi->Source, get_frame_new_buffer(cm));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
// Prevent possible divide by zero error below for perfect KF
kf_err += !kf_err;
@@ -1999,7 +2812,11 @@
YV12_BUFFER_CONFIG *scaled) {
if (cm->mi_cols * MI_SIZE != unscaled->y_width ||
cm->mi_rows * MI_SIZE != unscaled->y_height) {
+#if CONFIG_VP9_HIGHBITDEPTH
+ scale_and_extend_frame_nonnormative(unscaled, scaled, (int)cm->bit_depth);
+#else
scale_and_extend_frame_nonnormative(unscaled, scaled);
+#endif // CONFIG_VP9_HIGHBITDEPTH
return scaled;
} else {
return unscaled;
@@ -2165,7 +2982,9 @@
}
}
if (is_two_pass_svc(cpi) && cm->error_resilient_mode == 0) {
+ // Use the last frame context for the empty frame.
cm->frame_context_idx =
+ (cpi->svc.encode_empty_frame_state == ENCODING) ? FRAME_CONTEXTS - 1 :
cpi->svc.spatial_layer_id * cpi->svc.number_temporal_layers +
cpi->svc.temporal_layer_id;
@@ -2203,9 +3022,9 @@
// Check if the current frame is skippable for the partition search in the
// second pass according to the first pass stats
- if (oxcf->pass == 2 &&
+ if (cpi->sf.allow_partition_search_skip && oxcf->pass == 2 &&
(!cpi->use_svc || is_two_pass_svc(cpi))) {
- cpi->skippable_frame = is_skippable_frame(cpi);
+ cpi->partition_search_skippable_frame = is_skippable_frame(cpi);
}
// For 1 pass CBR, check if we are dropping this frame.
@@ -2289,7 +3108,17 @@
// fixed interval. Note the reconstruction error if it is the frame before
// the force key frame
if (cpi->rc.next_key_frame_forced && cpi->rc.frames_to_key == 1) {
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ cpi->ambient_err = vp9_highbd_get_y_sse(cpi->Source,
+ get_frame_new_buffer(cm),
+ cm->bit_depth);
+ } else {
+ cpi->ambient_err = vp9_get_y_sse(cpi->Source, get_frame_new_buffer(cm));
+ }
+#else
cpi->ambient_err = vp9_get_y_sse(cpi->Source, get_frame_new_buffer(cm));
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
// If the encoder forced a KEY_FRAME decision
@@ -2336,7 +3165,9 @@
cpi->ref_frame_flags = get_ref_frame_flags(cpi);
cm->last_frame_type = cm->frame_type;
- vp9_rc_postencode_update(cpi, *size);
+
+ if (!(is_two_pass_svc(cpi) && cpi->svc.encode_empty_frame_state == ENCODING))
+ vp9_rc_postencode_update(cpi, *size);
#if 0
output_frame_level_debug_stats(cpi);
@@ -2360,12 +3191,8 @@
cm->last_height = cm->height;
// reset to normal state now that we are done.
- if (!cm->show_existing_frame) {
- if (is_two_pass_svc(cpi) && cm->error_resilient_mode == 0)
- cm->last_show_frame = 0;
- else
- cm->last_show_frame = cm->show_frame;
- }
+ if (!cm->show_existing_frame)
+ cm->last_show_frame = cm->show_frame;
if (cm->show_frame) {
vp9_swap_mi_and_prev_mi(cm);
@@ -2402,7 +3229,9 @@
uint8_t *dest, unsigned int *frame_flags) {
cpi->allow_encode_breakout = ENCODE_BREAKOUT_ENABLED;
encode_frame_to_data_rate(cpi, size, dest, frame_flags);
- vp9_twopass_postencode_update(cpi);
+
+ if (!(is_two_pass_svc(cpi) && cpi->svc.encode_empty_frame_state == ENCODING))
+ vp9_twopass_postencode_update(cpi);
}
static void init_motion_estimation(VP9_COMP *cpi) {
@@ -2415,13 +3244,19 @@
}
}
-static void check_initial_width(VP9_COMP *cpi, int subsampling_x,
- int subsampling_y) {
+static void check_initial_width(VP9_COMP *cpi,
+#if CONFIG_VP9_HIGHBITDEPTH
+ int use_highbitdepth,
+#endif
+ int subsampling_x, int subsampling_y) {
VP9_COMMON *const cm = &cpi->common;
if (!cpi->initial_width) {
cm->subsampling_x = subsampling_x;
cm->subsampling_y = subsampling_y;
+#if CONFIG_VP9_HIGHBITDEPTH
+ cm->use_highbitdepth = use_highbitdepth;
+#endif
alloc_raw_frame_buffers(cpi);
alloc_ref_frame_buffers(cpi);
@@ -2441,10 +3276,14 @@
VP9_COMMON *cm = &cpi->common;
struct vpx_usec_timer timer;
int res = 0;
- const int subsampling_x = sd->uv_width < sd->y_width;
- const int subsampling_y = sd->uv_height < sd->y_height;
-
+ const int subsampling_x = sd->subsampling_x;
+ const int subsampling_y = sd->subsampling_y;
+#if CONFIG_VP9_HIGHBITDEPTH
+ const int use_highbitdepth = sd->flags & YV12_FLAG_HIGHBITDEPTH;
+ check_initial_width(cpi, use_highbitdepth, subsampling_x, subsampling_y);
+#else
check_initial_width(cpi, subsampling_x, subsampling_y);
+#endif // CONFIG_VP9_HIGHBITDEPTH
vpx_usec_timer_start(&timer);
@@ -2580,6 +3419,9 @@
if (is_two_pass_svc(cpi)) {
#if CONFIG_SPATIAL_SVC
vp9_svc_start_frame(cpi);
+ // Use a small empty frame instead of a real frame
+ if (cpi->svc.encode_empty_frame_state == ENCODING)
+ source = &cpi->svc.empty_frame;
#endif
if (oxcf->pass == 2)
vp9_restore_layer_context(cpi);
@@ -2598,6 +3440,11 @@
// Should we encode an arf frame.
arf_src_index = get_arf_src_index(cpi);
+
+ // Skip alt frame if we encode the empty frame
+ if (is_two_pass_svc(cpi) && source != NULL)
+ arf_src_index = 0;
+
if (arf_src_index) {
assert(arf_src_index <= rc->frames_to_key);
@@ -2708,7 +3555,10 @@
// For two pass encodes analyse the first pass stats and determine
// the bit allocation and other parameters for this frame / group of frames.
- if ((oxcf->pass == 2) && (!cpi->use_svc || is_two_pass_svc(cpi))) {
+ if ((oxcf->pass == 2) &&
+ (!cpi->use_svc ||
+ (is_two_pass_svc(cpi) &&
+ cpi->svc.encode_empty_frame_state != ENCODING))) {
vp9_rc_get_second_pass_params(cpi);
}
@@ -2760,7 +3610,7 @@
vp9_setup_scale_factors_for_frame(&ref_buf->sf,
buf->y_crop_width, buf->y_crop_height,
cm->width, cm->height);
-#endif
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (vp9_is_scaled(&ref_buf->sf))
vp9_extend_frame_borders(buf);
}
@@ -2776,14 +3626,14 @@
const int lossless = is_lossless_requested(oxcf);
#if CONFIG_VP9_HIGHBITDEPTH
if (cpi->oxcf.use_highbitdepth)
- cpi->mb.fwd_txm4x4 = lossless ? vp9_high_fwht4x4 : vp9_high_fdct4x4;
+ cpi->mb.fwd_txm4x4 = lossless ? vp9_highbd_fwht4x4 : vp9_highbd_fdct4x4;
else
cpi->mb.fwd_txm4x4 = lossless ? vp9_fwht4x4 : vp9_fdct4x4;
- cpi->mb.high_itxm_add = lossless ? vp9_high_iwht4x4_add :
- vp9_high_idct4x4_add;
+ cpi->mb.highbd_itxm_add = lossless ? vp9_highbd_iwht4x4_add :
+ vp9_highbd_idct4x4_add;
#else
cpi->mb.fwd_txm4x4 = lossless ? vp9_fwht4x4 : vp9_fdct4x4;
-#endif
+#endif // CONFIG_VP9_HIGHBITDEPTH
cpi->mb.itxm_add = lossless ? vp9_iwht4x4_add : vp9_idct4x4_add;
vp9_first_pass(cpi, source);
} else if (oxcf->pass == 2 &&
@@ -2836,7 +3686,12 @@
YV12_BUFFER_CONFIG *recon = cpi->common.frame_to_show;
YV12_BUFFER_CONFIG *pp = &cm->post_proc_buffer;
PSNR_STATS psnr;
+#if CONFIG_VP9_HIGHBITDEPTH
+ calc_highbd_psnr(orig, recon, &psnr, cpi->mb.e_mbd.bd,
+ cpi->oxcf.input_bit_depth);
+#else
calc_psnr(orig, recon, &psnr);
+#endif // CONFIG_VP9_HIGHBITDEPTH
cpi->total += psnr.psnr[0];
cpi->total_y += psnr.psnr[1];
@@ -2856,7 +3711,12 @@
#endif
vp9_clear_system_state();
+#if CONFIG_VP9_HIGHBITDEPTH
+ calc_highbd_psnr(orig, pp, &psnr, cpi->mb.e_mbd.bd,
+ cpi->oxcf.input_bit_depth);
+#else
calc_psnr(orig, pp, &psnr2);
+#endif // CONFIG_VP9_HIGHBITDEPTH
cpi->totalp += psnr2.psnr[0];
cpi->totalp_y += psnr2.psnr[1];
@@ -2865,12 +3725,29 @@
cpi->totalp_sq_error += psnr2.sse[0];
cpi->totalp_samples += psnr2.samples[0];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ frame_ssim2 = vp9_highbd_calc_ssim(orig, recon, &weight, xd->bd);
+ } else {
+ frame_ssim2 = vp9_calc_ssim(orig, recon, &weight);
+ }
+#else
frame_ssim2 = vp9_calc_ssim(orig, recon, &weight);
+#endif // CONFIG_VP9_HIGHBITDEPTH
cpi->summed_quality += frame_ssim2 * weight;
cpi->summed_weights += weight;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ frame_ssim2 = vp9_highbd_calc_ssim(
+ orig, &cm->post_proc_buffer, &weight, xd->bd);
+ } else {
+ frame_ssim2 = vp9_calc_ssim(orig, &cm->post_proc_buffer, &weight);
+ }
+#else
frame_ssim2 = vp9_calc_ssim(orig, &cm->post_proc_buffer, &weight);
+#endif // CONFIG_VP9_HIGHBITDEPTH
cpi->summedp_quality += frame_ssim2 * weight;
cpi->summedp_weights += weight;
@@ -2889,7 +3766,17 @@
if (cpi->b_calculate_ssimg) {
double y, u, v, frame_all;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ frame_all = vp9_highbd_calc_ssimg(cpi->Source, cm->frame_to_show, &y,
+ &u, &v, xd->bd);
+ } else {
+ frame_all = vp9_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u,
+ &v);
+ }
+#else
frame_all = vp9_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u, &v);
+#endif // CONFIG_VP9_HIGHBITDEPTH
cpi->total_ssimg_y += y;
cpi->total_ssimg_u += u;
cpi->total_ssimg_v += v;
@@ -2900,10 +3787,18 @@
#endif
- if (is_two_pass_svc(cpi) && cm->show_frame) {
- ++cpi->svc.spatial_layer_to_encode;
- if (cpi->svc.spatial_layer_to_encode >= cpi->svc.number_spatial_layers)
- cpi->svc.spatial_layer_to_encode = 0;
+ if (is_two_pass_svc(cpi)) {
+ if (cpi->svc.encode_empty_frame_state == ENCODING)
+ cpi->svc.encode_empty_frame_state = ENCODED;
+
+ if (cm->show_frame) {
+ ++cpi->svc.spatial_layer_to_encode;
+ if (cpi->svc.spatial_layer_to_encode >= cpi->svc.number_spatial_layers)
+ cpi->svc.spatial_layer_to_encode = 0;
+
+ // May need the empty frame after an visible frame.
+ cpi->svc.encode_empty_frame_state = NEED_TO_ENCODE;
+ }
}
return 0;
}
@@ -2986,15 +3881,14 @@
int vp9_set_size_literal(VP9_COMP *cpi, unsigned int width,
unsigned int height) {
VP9_COMMON *cm = &cpi->common;
-
+#if CONFIG_VP9_HIGHBITDEPTH
+ check_initial_width(cpi, 1, 1, cm->use_highbitdepth);
+#else
check_initial_width(cpi, 1, 1);
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (width) {
cm->width = width;
- if (cm->width * 5 < cpi->initial_width) {
- cm->width = cpi->initial_width / 5 + 1;
- printf("Warning: Desired width too small, changed to %d\n", cm->width);
- }
if (cm->width > cpi->initial_width) {
cm->width = cpi->initial_width;
printf("Warning: Desired width too large, changed to %d\n", cm->width);
@@ -3003,10 +3897,6 @@
if (height) {
cm->height = height;
- if (cm->height * 5 < cpi->initial_height) {
- cm->height = cpi->initial_height / 5 + 1;
- printf("Warning: Desired height too small, changed to %d\n", cm->height);
- }
if (cm->height > cpi->initial_height) {
cm->height = cpi->initial_height;
printf("Warning: Desired height too large, changed to %d\n", cm->height);
@@ -3033,6 +3923,35 @@
a->y_crop_width, a->y_crop_height);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+int vp9_highbd_get_y_sse(const YV12_BUFFER_CONFIG *a,
+ const YV12_BUFFER_CONFIG *b,
+ vpx_bit_depth_t bit_depth) {
+ unsigned int sse;
+ int sum;
+ assert(a->y_crop_width == b->y_crop_width);
+ assert(a->y_crop_height == b->y_crop_height);
+ assert((a->flags & YV12_FLAG_HIGHBITDEPTH) != 0);
+ assert((b->flags & YV12_FLAG_HIGHBITDEPTH) != 0);
+ switch (bit_depth) {
+ case VPX_BITS_8:
+ highbd_variance(a->y_buffer, a->y_stride, b->y_buffer, b->y_stride,
+ a->y_crop_width, a->y_crop_height, &sse, &sum);
+ return (int) sse;
+ case VPX_BITS_10:
+ highbd_10_variance(a->y_buffer, a->y_stride, b->y_buffer, b->y_stride,
+ a->y_crop_width, a->y_crop_height, &sse, &sum);
+ return (int) sse;
+ case VPX_BITS_12:
+ highbd_12_variance(a->y_buffer, a->y_stride, b->y_buffer, b->y_stride,
+ a->y_crop_width, a->y_crop_height, &sse, &sum);
+ return (int) sse;
+ default:
+ assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
+ return -1;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
int vp9_get_quantizer(VP9_COMP *cpi) {
return cpi->common.base_qindex;
diff --git a/source/libvpx/vp9/encoder/vp9_encoder.h b/source/libvpx/vp9/encoder/vp9_encoder.h
index 80774de..2c56b81 100644
--- a/source/libvpx/vp9/encoder/vp9_encoder.h
+++ b/source/libvpx/vp9/encoder/vp9_encoder.h
@@ -14,12 +14,10 @@
#include <stdio.h>
#include "./vpx_config.h"
-#include "vpx_ports/mem.h"
#include "vpx/internal/vpx_codec_internal.h"
#include "vpx/vp8cx.h"
#include "vp9/common/vp9_ppflags.h"
-#include "vp9/common/vp9_entropy.h"
#include "vp9/common/vp9_entropymode.h"
#include "vp9/common/vp9_onyxc_int.h"
@@ -241,7 +239,8 @@
YV12_BUFFER_CONFIG *unscaled_last_source;
YV12_BUFFER_CONFIG scaled_last_source;
- int skippable_frame;
+ // For a still frame, this flag is set to 1 to skip partition search.
+ int partition_search_skippable_frame;
int scaled_ref_idx[3];
int lst_fb_idx;
@@ -482,7 +481,21 @@
return mb_rows * mb_cols * (16 * 16 * 3 + 4);
}
+// Get the allocated token size for a tile. It does the same calculation as in
+// the frame token allocation.
+static INLINE int allocated_tokens(TileInfo tile) {
+ int tile_mb_rows = (tile.mi_row_end - tile.mi_row_start + 1) >> 1;
+ int tile_mb_cols = (tile.mi_col_end - tile.mi_col_start + 1) >> 1;
+
+ return get_token_alloc(tile_mb_rows, tile_mb_cols);
+}
+
int vp9_get_y_sse(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b);
+#if CONFIG_VP9_HIGHBITDEPTH
+int vp9_highbd_get_y_sse(const YV12_BUFFER_CONFIG *a,
+ const YV12_BUFFER_CONFIG *b,
+ vpx_bit_depth_t bit_depth);
+#endif // CONFIG_VP9_HIGHBITDEPTH
void vp9_alloc_compressor_data(VP9_COMP *cpi);
@@ -525,8 +538,8 @@
return frame_index & 0x1;
}
-static INLINE int *cond_sad_list(const struct VP9_COMP *cpi, int *sad_list) {
- return cpi->sf.mv.subpel_search_method != SUBPEL_TREE ? sad_list : NULL;
+static INLINE int *cond_cost_list(const struct VP9_COMP *cpi, int *cost_list) {
+ return cpi->sf.mv.subpel_search_method != SUBPEL_TREE ? cost_list : NULL;
}
#ifdef __cplusplus
diff --git a/source/libvpx/vp9/encoder/vp9_extend.c b/source/libvpx/vp9/encoder/vp9_extend.c
index e8517c8..c9b2131 100644
--- a/source/libvpx/vp9/encoder/vp9_extend.c
+++ b/source/libvpx/vp9/encoder/vp9_extend.c
@@ -55,6 +55,52 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static void highbd_copy_and_extend_plane(const uint8_t *src8, int src_pitch,
+ uint8_t *dst8, int dst_pitch,
+ int w, int h,
+ int extend_top, int extend_left,
+ int extend_bottom, int extend_right) {
+ int i, linesize;
+ uint16_t *src = CONVERT_TO_SHORTPTR(src8);
+ uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
+
+ // copy the left and right most columns out
+ const uint16_t *src_ptr1 = src;
+ const uint16_t *src_ptr2 = src + w - 1;
+ uint16_t *dst_ptr1 = dst - extend_left;
+ uint16_t *dst_ptr2 = dst + w;
+
+ for (i = 0; i < h; i++) {
+ vpx_memset16(dst_ptr1, src_ptr1[0], extend_left);
+ vpx_memcpy(dst_ptr1 + extend_left, src_ptr1, w * sizeof(uint16_t));
+ vpx_memset16(dst_ptr2, src_ptr2[0], extend_right);
+ src_ptr1 += src_pitch;
+ src_ptr2 += src_pitch;
+ dst_ptr1 += dst_pitch;
+ dst_ptr2 += dst_pitch;
+ }
+
+ // Now copy the top and bottom lines into each line of the respective
+ // borders
+ src_ptr1 = dst - extend_left;
+ src_ptr2 = dst + dst_pitch * (h - 1) - extend_left;
+ dst_ptr1 = dst + dst_pitch * (-extend_top) - extend_left;
+ dst_ptr2 = dst + dst_pitch * (h) - extend_left;
+ linesize = extend_left + extend_right + w;
+
+ for (i = 0; i < extend_top; i++) {
+ vpx_memcpy(dst_ptr1, src_ptr1, linesize * sizeof(uint16_t));
+ dst_ptr1 += dst_pitch;
+ }
+
+ for (i = 0; i < extend_bottom; i++) {
+ vpx_memcpy(dst_ptr2, src_ptr2, linesize * sizeof(uint16_t));
+ dst_ptr2 += dst_pitch;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
void vp9_copy_and_extend_frame(const YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst) {
// Extend src frame in buffer
@@ -64,10 +110,10 @@
// Motion estimation may use src block variance with the block size up
// to 64x64, so the right and bottom need to be extended to 64 multiple
// or up to 16, whichever is greater.
- const int eb_y = MAX(ALIGN_POWER_OF_TWO(src->y_width, 6) - src->y_width,
- 16);
- const int er_y = MAX(ALIGN_POWER_OF_TWO(src->y_height, 6) - src->y_height,
- 16);
+ const int eb_y = MAX(src->y_width + 16, ALIGN_POWER_OF_TWO(src->y_width, 6))
+ - src->y_crop_width;
+ const int er_y = MAX(src->y_height + 16, ALIGN_POWER_OF_TWO(src->y_height, 6))
+ - src->y_crop_height;
const int uv_width_subsampling = (src->uv_width != src->y_width);
const int uv_height_subsampling = (src->uv_height != src->y_height);
const int et_uv = et_y >> uv_height_subsampling;
@@ -75,19 +121,39 @@
const int eb_uv = eb_y >> uv_height_subsampling;
const int er_uv = er_y >> uv_width_subsampling;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (src->flags & YV12_FLAG_HIGHBITDEPTH) {
+ highbd_copy_and_extend_plane(src->y_buffer, src->y_stride,
+ dst->y_buffer, dst->y_stride,
+ src->y_crop_width, src->y_crop_height,
+ et_y, el_y, eb_y, er_y);
+
+ highbd_copy_and_extend_plane(src->u_buffer, src->uv_stride,
+ dst->u_buffer, dst->uv_stride,
+ src->uv_crop_width, src->uv_crop_height,
+ et_uv, el_uv, eb_uv, er_uv);
+
+ highbd_copy_and_extend_plane(src->v_buffer, src->uv_stride,
+ dst->v_buffer, dst->uv_stride,
+ src->uv_crop_width, src->uv_crop_height,
+ et_uv, el_uv, eb_uv, er_uv);
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
copy_and_extend_plane(src->y_buffer, src->y_stride,
dst->y_buffer, dst->y_stride,
- src->y_width, src->y_height,
+ src->y_crop_width, src->y_crop_height,
et_y, el_y, eb_y, er_y);
copy_and_extend_plane(src->u_buffer, src->uv_stride,
dst->u_buffer, dst->uv_stride,
- src->uv_width, src->uv_height,
+ src->uv_crop_width, src->uv_crop_height,
et_uv, el_uv, eb_uv, er_uv);
copy_and_extend_plane(src->v_buffer, src->uv_stride,
dst->v_buffer, dst->uv_stride,
- src->uv_width, src->uv_height,
+ src->uv_crop_width, src->uv_crop_height,
et_uv, el_uv, eb_uv, er_uv);
}
diff --git a/source/libvpx/vp9/encoder/vp9_firstpass.c b/source/libvpx/vp9/encoder/vp9_firstpass.c
index 0282e9f..f1baf83 100644
--- a/source/libvpx/vp9/encoder/vp9_firstpass.c
+++ b/source/libvpx/vp9/encoder/vp9_firstpass.c
@@ -38,14 +38,16 @@
#define OUTPUT_FPF 0
#define ARF_STATS_OUTPUT 0
+#define BOOST_BREAKOUT 12.5
#define BOOST_FACTOR 12.5
-#define ERR_DIVISOR 100.0
-#define FACTOR_PT_LOW 0.5
-#define FACTOR_PT_HIGH 0.9
+#define ERR_DIVISOR 128.0
+#define FACTOR_PT_LOW 0.70
+#define FACTOR_PT_HIGH 0.90
#define FIRST_PASS_Q 10.0
#define GF_MAX_BOOST 96.0
#define INTRA_MODE_PENALTY 1024
#define KF_MAX_BOOST 128.0
+#define MIN_ARF_GF_BOOST 240
#define MIN_DECAY_FACTOR 0.01
#define MIN_GF_INTERVAL 4
#define MIN_KF_BOOST 300
@@ -64,13 +66,6 @@
*b = temp;
}
-static int gfboost_qadjust(int qindex, vpx_bit_depth_t bit_depth) {
- const double q = vp9_convert_qindex_to_q(qindex, bit_depth);
- return (int)((0.00000828 * q * q * q) +
- (-0.0055 * q * q) +
- (1.32 * q) + 79.3);
-}
-
// Resets the first pass file to the given position using a relative seek from
// the current position.
static void reset_fpf_position(TWO_PASS *p,
@@ -281,6 +276,60 @@
return sse;
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static vp9_variance_fn_t highbd_get_block_variance_fn(BLOCK_SIZE bsize,
+ int bd) {
+ switch (bd) {
+ default:
+ switch (bsize) {
+ case BLOCK_8X8:
+ return vp9_highbd_mse8x8;
+ case BLOCK_16X8:
+ return vp9_highbd_mse16x8;
+ case BLOCK_8X16:
+ return vp9_highbd_mse8x16;
+ default:
+ return vp9_highbd_mse16x16;
+ }
+ break;
+ case 10:
+ switch (bsize) {
+ case BLOCK_8X8:
+ return vp9_highbd_10_mse8x8;
+ case BLOCK_16X8:
+ return vp9_highbd_10_mse16x8;
+ case BLOCK_8X16:
+ return vp9_highbd_10_mse8x16;
+ default:
+ return vp9_highbd_10_mse16x16;
+ }
+ break;
+ case 12:
+ switch (bsize) {
+ case BLOCK_8X8:
+ return vp9_highbd_12_mse8x8;
+ case BLOCK_16X8:
+ return vp9_highbd_12_mse16x8;
+ case BLOCK_8X16:
+ return vp9_highbd_12_mse8x16;
+ default:
+ return vp9_highbd_12_mse16x16;
+ }
+ break;
+ }
+}
+
+static unsigned int highbd_get_prediction_error(BLOCK_SIZE bsize,
+ const struct buf_2d *src,
+ const struct buf_2d *ref,
+ int bd) {
+ unsigned int sse;
+ const vp9_variance_fn_t fn = highbd_get_block_variance_fn(bsize, bd);
+ fn(src->buf, src->stride, ref->buf, ref->stride, &sse);
+ return sse;
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
// Refine the motion search range according to the frame dimension
// for first pass test.
static int get_search_range(const VP9_COMMON *cm) {
@@ -311,6 +360,11 @@
// Override the default variance function to use MSE.
v_fn_ptr.vf = get_block_variance_fn(bsize);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ v_fn_ptr.vf = highbd_get_block_variance_fn(bsize, xd->bd);
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
// Center the initial step/diamond search on best mv.
tmp_err = cpi->diamond_search_sad(x, &cpi->ss_cfg, &ref_mv_full, &tmp_mv,
@@ -562,6 +616,24 @@
(bsize >= BLOCK_16X16 ? TX_16X16 : TX_8X8) : TX_4X4;
vp9_encode_intra_block_plane(x, bsize, 0);
this_error = vp9_get_mb_ss(x->plane[0].src_diff);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ switch (cm->bit_depth) {
+ case VPX_BITS_8:
+ break;
+ case VPX_BITS_10:
+ this_error >>= 4;
+ break;
+ case VPX_BITS_12:
+ this_error >>= 8;
+ break;
+ default:
+ assert(0 && "cm->bit_depth should be VPX_BITS_8, "
+ "VPX_BITS_10 or VPX_BITS_12");
+ return;
+ }
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_clear_system_state();
@@ -601,8 +673,18 @@
struct buf_2d unscaled_last_source_buf_2d;
xd->plane[0].pre[0].buf = first_ref_buf->y_buffer + recon_yoffset;
- motion_error = get_prediction_error(bsize, &x->plane[0].src,
- &xd->plane[0].pre[0]);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ motion_error = highbd_get_prediction_error(
+ bsize, &x->plane[0].src, &xd->plane[0].pre[0], xd->bd);
+ } else {
+ motion_error = get_prediction_error(
+ bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
+ }
+#else
+ motion_error = get_prediction_error(
+ bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
+#endif // CONFIG_VP9_HIGHBITDEPTH
// Compute the motion error of the 0,0 motion using the last source
// frame as the reference. Skip the further motion search on
@@ -611,8 +693,18 @@
cpi->unscaled_last_source->y_buffer + recon_yoffset;
unscaled_last_source_buf_2d.stride =
cpi->unscaled_last_source->y_stride;
- raw_motion_error = get_prediction_error(bsize, &x->plane[0].src,
- &unscaled_last_source_buf_2d);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ raw_motion_error = highbd_get_prediction_error(
+ bsize, &x->plane[0].src, &unscaled_last_source_buf_2d, xd->bd);
+ } else {
+ raw_motion_error = get_prediction_error(
+ bsize, &x->plane[0].src, &unscaled_last_source_buf_2d);
+ }
+#else
+ raw_motion_error = get_prediction_error(
+ bsize, &x->plane[0].src, &unscaled_last_source_buf_2d);
+#endif // CONFIG_VP9_HIGHBITDEPTH
// TODO(pengchong): Replace the hard-coded threshold
if (raw_motion_error > 25 || lc != NULL) {
@@ -648,8 +740,18 @@
int gf_motion_error;
xd->plane[0].pre[0].buf = gld_yv12->y_buffer + recon_yoffset;
- gf_motion_error = get_prediction_error(bsize, &x->plane[0].src,
- &xd->plane[0].pre[0]);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ gf_motion_error = highbd_get_prediction_error(
+ bsize, &x->plane[0].src, &xd->plane[0].pre[0], xd->bd);
+ } else {
+ gf_motion_error = get_prediction_error(
+ bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
+ }
+#else
+ gf_motion_error = get_prediction_error(
+ bsize, &x->plane[0].src, &xd->plane[0].pre[0]);
+#endif // CONFIG_VP9_HIGHBITDEPTH
first_pass_motion_search(cpi, x, &zero_mv, &tmp_mv,
&gf_motion_error);
@@ -949,7 +1051,7 @@
// Adjustment based on actual quantizer to power term.
const double power_term =
- MIN(vp9_convert_qindex_to_q(q, bit_depth) * 0.0125 + pt_low, pt_high);
+ MIN(vp9_convert_qindex_to_q(q, bit_depth) * 0.01 + pt_low, pt_high);
// Calculate correction factor.
if (power_term < 1.0)
@@ -958,6 +1060,11 @@
return fclamp(pow(error_term, power_term), 0.05, 5.0);
}
+// Larger image formats are expected to be a little harder to code relatively
+// given the same prediction error score. This in part at least relates to the
+// increased size and hence coding cost of motion vectors.
+#define EDIV_SIZE_FACTOR 800
+
static int get_twopass_worst_quality(const VP9_COMP *cpi,
const FIRSTPASS_STATS *stats,
int section_target_bandwidth) {
@@ -971,8 +1078,10 @@
const double section_err = stats->coded_error / stats->count;
const double err_per_mb = section_err / num_mbs;
const double speed_term = 1.0 + 0.04 * oxcf->speed;
+ const double ediv_size_correction = num_mbs / EDIV_SIZE_FACTOR;
const int target_norm_bits_per_mb = ((uint64_t)section_target_bandwidth <<
BPER_MB_NORMBITS) / num_mbs;
+
int q;
int is_svc_upper_layer = 0;
if (is_two_pass_svc(cpi) && cpi->svc.spatial_layer_id > 0)
@@ -982,7 +1091,7 @@
// content at the given rate.
for (q = rc->best_quality; q < rc->worst_quality; ++q) {
const double factor =
- calc_correction_factor(err_per_mb, ERR_DIVISOR,
+ calc_correction_factor(err_per_mb, ERR_DIVISOR - ediv_size_correction,
is_svc_upper_layer ? SVC_FACTOR_PT_LOW :
FACTOR_PT_LOW, FACTOR_PT_HIGH, q,
cpi->common.bit_depth);
@@ -1065,6 +1174,8 @@
// Reset the vbr bits off target counter
cpi->rc.vbr_bits_off_target = 0;
+ cpi->rc.rate_error_estimate = 0;
+
// Static sequence monitor variables.
twopass->kf_zeromotion_pct = 100;
twopass->last_kfgroup_zeromotion_pct = 100;
@@ -1199,11 +1310,15 @@
double this_frame_mv_in_out,
double max_boost) {
double frame_boost;
+ const double lq =
+ vp9_convert_qindex_to_q(cpi->rc.avg_frame_qindex[INTER_FRAME],
+ cpi->common.bit_depth);
+ const double boost_correction = MIN((0.5 + (lq * 0.015)), 1.5);
// Underlying boost factor is based on inter error ratio.
frame_boost = (BASELINE_ERR_PER_MB * cpi->common.MBs) /
DOUBLE_DIVIDE_CHECK(this_frame->coded_error);
- frame_boost = frame_boost * BOOST_FACTOR;
+ frame_boost = frame_boost * BOOST_FACTOR * boost_correction;
// Increase boost for frames where new data coming into frame (e.g. zoom out).
// Slightly reduce boost if there is a net balance of motion out of the frame
@@ -1214,7 +1329,7 @@
else
frame_boost += frame_boost * (this_frame_mv_in_out / 2.0);
- return MIN(frame_boost, max_boost);
+ return MIN(frame_boost, max_boost * boost_correction);
}
static int calc_arf_boost(VP9_COMP *cpi, int offset,
@@ -1303,6 +1418,7 @@
arf_boost = (*f_boost + *b_boost);
if (arf_boost < ((b_frames + f_frames) * 20))
arf_boost = ((b_frames + f_frames) * 20);
+ arf_boost = MAX(arf_boost, MIN_ARF_GF_BOOST);
return arf_boost;
}
@@ -1580,6 +1696,7 @@
int b_boost = 0;
int flash_detected;
int active_max_gf_interval;
+ int active_min_gf_interval;
int64_t gf_group_bits;
double gf_group_error_left;
int gf_arf_bits;
@@ -1608,21 +1725,30 @@
// Motion breakout threshold for loop below depends on image size.
mv_ratio_accumulator_thresh = (cpi->common.width + cpi->common.height) / 4.0;
- // Work out a maximum interval for the GF group.
+ // Set a maximum and minimum interval for the GF group.
// If the image appears almost completely static we can extend beyond this.
- if (cpi->multi_arf_allowed) {
- active_max_gf_interval = rc->max_gf_interval;
- } else {
- // The value chosen depends on the active Q range. At low Q we have
- // bits to spare and are better with a smaller interval and smaller boost.
- // At high Q when there are few bits to spare we are better with a longer
- // interval to spread the cost of the GF.
- active_max_gf_interval =
- 12 + ((int)vp9_convert_qindex_to_q(rc->last_q[INTER_FRAME],
- cpi->common.bit_depth) >> 5);
+ {
+ int int_max_q =
+ (int)(vp9_convert_qindex_to_q(twopass->active_worst_quality,
+ cpi->common.bit_depth));
+ int int_lbq =
+ (int)(vp9_convert_qindex_to_q(rc->last_boosted_qindex,
+ cpi->common.bit_depth));
+ active_min_gf_interval = MIN_GF_INTERVAL + MIN(2, int_max_q / 200);
+ if (active_min_gf_interval > rc->max_gf_interval)
+ active_min_gf_interval = rc->max_gf_interval;
- if (active_max_gf_interval > rc->max_gf_interval)
- active_max_gf_interval = rc->max_gf_interval;
+ if (cpi->multi_arf_allowed) {
+ active_max_gf_interval = rc->max_gf_interval;
+ } else {
+ // The value chosen depends on the active Q range. At low Q we have
+ // bits to spare and are better with a smaller interval and smaller boost.
+ // At high Q when there are few bits to spare we are better with a longer
+ // interval to spread the cost of the GF.
+ active_max_gf_interval = 12 + MIN(4, (int_lbq / 6));
+ if (active_max_gf_interval > rc->max_gf_interval)
+ active_max_gf_interval = rc->max_gf_interval;
+ }
}
i = 0;
@@ -1678,12 +1804,12 @@
(i >= active_max_gf_interval && (zero_motion_accumulator < 0.995)) ||
(
// Don't break out with a very short interval.
- (i > MIN_GF_INTERVAL) &&
+ (i > active_min_gf_interval) &&
(!flash_detected) &&
((mv_ratio_accumulator > mv_ratio_accumulator_thresh) ||
(abs_mv_in_out_accumulator > 3.0) ||
(mv_in_out_accumulator < -2.0) ||
- ((boost_score - old_boost_score) < BOOST_FACTOR)))) {
+ ((boost_score - old_boost_score) < BOOST_BREAKOUT)))) {
boost_score = old_boost_score;
break;
}
@@ -1731,7 +1857,7 @@
(cpi->multi_arf_allowed && (rc->baseline_gf_interval >= 6) &&
(zero_motion_accumulator < 0.995)) ? 1 : 0;
} else {
- rc->gfu_boost = MAX((int)boost_score, 125);
+ rc->gfu_boost = MAX((int)boost_score, MIN_ARF_GF_BOOST);
rc->source_alt_ref_pending = 0;
}
@@ -1742,18 +1868,8 @@
gf_group_bits = calculate_total_gf_group_bits(cpi, gf_group_err);
// Calculate the extra bits to be used for boosted frame(s)
- {
- int q = rc->last_q[INTER_FRAME];
- int boost =
- (rc->gfu_boost * gfboost_qadjust(q, cpi->common.bit_depth)) / 100;
-
- // Set max and minimum boost and hence minimum allocation.
- boost = clamp(boost, 125, (rc->baseline_gf_interval + 1) * 200);
-
- // Calculate the extra bits to be used for boosted frame(s)
- gf_arf_bits = calculate_boost_bits(rc->baseline_gf_interval,
- boost, gf_group_bits);
- }
+ gf_arf_bits = calculate_boost_bits(rc->baseline_gf_interval,
+ rc->gfu_boost, gf_group_bits);
// Adjust KF group bits and error remaining.
twopass->kf_group_error_left -= (int64_t)gf_group_err;
@@ -2101,10 +2217,23 @@
twopass->modified_error_left -= kf_group_err;
}
+#define VBR_PCT_ADJUSTMENT_LIMIT 50
// For VBR...adjustment to the frame target based on error from previous frames
-void vbr_rate_correction(int * this_frame_target,
+void vbr_rate_correction(VP9_COMP *cpi,
+ int * this_frame_target,
const int64_t vbr_bits_off_target) {
- int max_delta = (*this_frame_target * 15) / 100;
+ int max_delta;
+ double position_factor = 1.0;
+
+ // How far through the clip are we.
+ // This number is used to damp the per frame rate correction.
+ // Range 0 - 1.0
+ if (cpi->twopass.total_stats.count) {
+ position_factor = sqrt((double)cpi->common.current_video_frame /
+ cpi->twopass.total_stats.count);
+ }
+ max_delta = (int)(position_factor *
+ ((*this_frame_target * VBR_PCT_ADJUSTMENT_LIMIT) / 100));
// vbr_bits_off_target > 0 means we have extra bits to spend
if (vbr_bits_off_target > 0) {
@@ -2202,7 +2331,7 @@
// Correction to rate target based on prior over or under shoot.
if (cpi->oxcf.rc_mode == VPX_VBR)
- vbr_rate_correction(&target_rate, rc->vbr_bits_off_target);
+ vbr_rate_correction(cpi, &target_rate, rc->vbr_bits_off_target);
vp9_rc_set_frame_target(cpi, target_rate);
cm->frame_type = INTER_FRAME;
@@ -2234,7 +2363,11 @@
section_target_bandwidth);
twopass->active_worst_quality = tmp_q;
rc->ni_av_qi = tmp_q;
+ rc->last_q[INTER_FRAME] = tmp_q;
rc->avg_q = vp9_convert_qindex_to_q(tmp_q, cm->bit_depth);
+ rc->avg_frame_qindex[INTER_FRAME] = tmp_q;
+ rc->last_q[KEY_FRAME] = (tmp_q + cpi->oxcf.best_allowed_q) / 2;
+ rc->avg_frame_qindex[KEY_FRAME] = rc->last_q[KEY_FRAME];
}
vp9_zero(this_frame);
if (EOF == input_stats(twopass, &this_frame))
@@ -2259,6 +2392,9 @@
cpi->ref_frame_flags &=
(~VP9_LAST_FLAG & ~VP9_GOLD_FLAG & ~VP9_ALT_FLAG);
lc->frames_from_key_frame = 0;
+ // Reset the empty frame resolution since we have a key frame.
+ cpi->svc.empty_frame_width = cm->width;
+ cpi->svc.empty_frame_height = cm->height;
}
} else {
cm->frame_type = INTER_FRAME;
@@ -2275,16 +2411,6 @@
if (rc->frames_till_gf_update_due == 0) {
define_gf_group(cpi, &this_frame_copy);
- if (twopass->gf_zeromotion_pct > 995) {
- // As long as max_thresh for encode breakout is small enough, it is ok
- // to enable it for show frame, i.e. set allow_encode_breakout to
- // ENCODE_BREAKOUT_LIMITED.
- if (!cm->show_frame)
- cpi->allow_encode_breakout = ENCODE_BREAKOUT_DISABLED;
- else
- cpi->allow_encode_breakout = ENCODE_BREAKOUT_LIMITED;
- }
-
rc->frames_till_gf_update_due = rc->baseline_gf_interval;
if (lc != NULL)
cpi->refresh_golden_frame = 1;
@@ -2294,8 +2420,9 @@
FILE *fpfile;
fpfile = fopen("arf.stt", "a");
++arf_count;
- fprintf(fpfile, "%10d %10d %10d %10ld\n",
- cm->current_video_frame, rc->kf_boost, arf_count, rc->gfu_boost);
+ fprintf(fpfile, "%10d %10ld %10d %10d %10ld\n",
+ cm->current_video_frame, rc->frames_till_gf_update_due,
+ rc->kf_boost, arf_count, rc->gfu_boost);
fclose(fpfile);
}
@@ -2314,7 +2441,7 @@
// Correction to rate target based on prior over or under shoot.
if (cpi->oxcf.rc_mode == VPX_VBR)
- vbr_rate_correction(&target_rate, rc->vbr_bits_off_target);
+ vbr_rate_correction(cpi, &target_rate, rc->vbr_bits_off_target);
vp9_rc_set_frame_target(cpi, target_rate);
@@ -2322,20 +2449,30 @@
subtract_stats(&twopass->total_left_stats, &this_frame);
}
+#define MINQ_ADJ_LIMIT 32
+#define Q_LIMIT_STEP 1
void vp9_twopass_postencode_update(VP9_COMP *cpi) {
TWO_PASS *const twopass = &cpi->twopass;
RATE_CONTROL *const rc = &cpi->rc;
+ const int bits_used = rc->base_frame_target;
// VBR correction is done through rc->vbr_bits_off_target. Based on the
// sign of this value, a limited % adjustment is made to the target rate
// of subsequent frames, to try and push it back towards 0. This method
// is designed to prevent extreme behaviour at the end of a clip
// or group of frames.
- const int bits_used = rc->base_frame_target;
rc->vbr_bits_off_target += rc->base_frame_target - rc->projected_frame_size;
-
twopass->bits_left = MAX(twopass->bits_left - bits_used, 0);
+ // Calculate the pct rc error.
+ if (rc->total_actual_bits) {
+ rc->rate_error_estimate =
+ (int)((rc->vbr_bits_off_target * 100) / rc->total_actual_bits);
+ rc->rate_error_estimate = clamp(rc->rate_error_estimate, -100, 100);
+ } else {
+ rc->rate_error_estimate = 0;
+ }
+
if (cpi->common.frame_type != KEY_FRAME &&
!vp9_is_upper_layer_key_frame(cpi)) {
twopass->kf_group_bits -= bits_used;
@@ -2345,4 +2482,32 @@
// Increment the gf group index ready for the next frame.
++twopass->gf_group.index;
+
+ // If the rate control is drifting consider adjustment ot min or maxq.
+ // Only make adjustments on gf/arf
+ if ((cpi->oxcf.rc_mode == VPX_VBR) &&
+ (cpi->twopass.gf_zeromotion_pct < VLOW_MOTION_THRESHOLD) &&
+ !cpi->rc.is_src_frame_alt_ref) {
+ const int maxq_adj_limit =
+ rc->worst_quality - twopass->active_worst_quality;
+
+ // Undershoot.
+ if (rc->rate_error_estimate > cpi->oxcf.under_shoot_pct) {
+ --twopass->extend_maxq;
+ if (rc->rolling_target_bits >= rc->rolling_actual_bits)
+ twopass->extend_minq += Q_LIMIT_STEP;
+ // Overshoot.
+ } else if (rc->rate_error_estimate < -cpi->oxcf.over_shoot_pct) {
+ --twopass->extend_minq;
+ if (rc->rolling_target_bits < rc->rolling_actual_bits)
+ twopass->extend_maxq += Q_LIMIT_STEP;
+ } else {
+ if (rc->rolling_target_bits < rc->rolling_actual_bits)
+ --twopass->extend_minq;
+ if (rc->rolling_target_bits > rc->rolling_actual_bits)
+ --twopass->extend_maxq;
+ }
+ twopass->extend_minq = clamp(twopass->extend_minq, 0, MINQ_ADJ_LIMIT);
+ twopass->extend_maxq = clamp(twopass->extend_maxq, 0, maxq_adj_limit);
+ }
}
diff --git a/source/libvpx/vp9/encoder/vp9_firstpass.h b/source/libvpx/vp9/encoder/vp9_firstpass.h
index 0b82d32..e21d869 100644
--- a/source/libvpx/vp9/encoder/vp9_firstpass.h
+++ b/source/libvpx/vp9/encoder/vp9_firstpass.h
@@ -39,6 +39,8 @@
} FIRSTPASS_MB_STATS;
#endif
+#define VLOW_MOTION_THRESHOLD 950
+
typedef struct {
double frame;
double intra_error;
@@ -110,8 +112,9 @@
int kf_zeromotion_pct;
int last_kfgroup_zeromotion_pct;
int gf_zeromotion_pct;
-
int active_worst_quality;
+ int extend_minq;
+ int extend_maxq;
GF_GROUP gf_group;
} TWO_PASS;
diff --git a/source/libvpx/vp9/encoder/vp9_mbgraph.c b/source/libvpx/vp9/encoder/vp9_mbgraph.c
index 42981d8..bd04c56 100644
--- a/source/libvpx/vp9/encoder/vp9_mbgraph.c
+++ b/source/libvpx/vp9/encoder/vp9_mbgraph.c
@@ -34,7 +34,7 @@
const int tmp_row_min = x->mv_row_min;
const int tmp_row_max = x->mv_row_max;
MV ref_full;
- int sad_list[5];
+ int cost_list[5];
// Further step/diamond searches as necessary
int step_param = mv_sf->reduce_first_step_size;
@@ -47,7 +47,7 @@
/*cpi->sf.search_method == HEX*/
vp9_hex_search(x, &ref_full, step_param, x->errorperbit, 0,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
&v_fn_ptr, 0, ref_mv, dst_mv);
// Try sub-pixel MC
@@ -58,7 +58,7 @@
cpi->find_fractional_mv_step(
x, dst_mv, ref_mv, cpi->common.allow_high_precision_mv, x->errorperbit,
&v_fn_ptr, 0, mv_sf->subpel_iters_per_step,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
NULL, NULL,
&distortion, &sse, NULL, 0, 0);
}
diff --git a/source/libvpx/vp9/encoder/vp9_mcomp.c b/source/libvpx/vp9/encoder/vp9_mcomp.c
index a25dc61..69b4193 100644
--- a/source/libvpx/vp9/encoder/vp9_mcomp.c
+++ b/source/libvpx/vp9/encoder/vp9_mcomp.c
@@ -284,17 +284,220 @@
int tc = bc; \
\
bestmv->row *= 8; \
- bestmv->col *= 8; \
- if (second_pred != NULL) { \
- DECLARE_ALIGNED_ARRAY(16, uint8_t, comp_pred, 64 * 64); \
- vp9_comp_avg_pred(comp_pred, second_pred, w, h, y + offset, y_stride); \
- besterr = vfp->vf(comp_pred, w, z, src_stride, sse1); \
- } else { \
- besterr = vfp->vf(y + offset, y_stride, z, src_stride, sse1); \
- } \
- *distortion = besterr; \
+ bestmv->col *= 8;
+
+#if CONFIG_VP9_HIGHBITDEPTH
+#define SETUP_CENTER_ERROR \
+ if (second_pred != NULL) { \
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) { \
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, comp_pred16, 64 * 64); \
+ vp9_highbd_comp_avg_pred(comp_pred16, second_pred, w, h, y + offset, \
+ y_stride); \
+ besterr = vfp->vf(CONVERT_TO_BYTEPTR(comp_pred16), w, z, src_stride, \
+ sse1); \
+ } else { \
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, comp_pred, 64 * 64); \
+ vp9_comp_avg_pred(comp_pred, second_pred, w, h, y + offset, y_stride); \
+ besterr = vfp->vf(comp_pred, w, z, src_stride, sse1); \
+ } \
+ } else { \
+ besterr = vfp->vf(y + offset, y_stride, z, src_stride, sse1); \
+ } \
+ *distortion = besterr; \
besterr += mv_err_cost(bestmv, ref_mv, mvjcost, mvcost, error_per_bit);
+#else
+
+#define SETUP_CENTER_ERROR \
+ if (second_pred != NULL) { \
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, comp_pred, 64 * 64); \
+ vp9_comp_avg_pred(comp_pred, second_pred, w, h, y + offset, y_stride); \
+ besterr = vfp->vf(comp_pred, w, z, src_stride, sse1); \
+ } else { \
+ besterr = vfp->vf(y + offset, y_stride, z, src_stride, sse1); \
+ } \
+ *distortion = besterr; \
+ besterr += mv_err_cost(bestmv, ref_mv, mvjcost, mvcost, error_per_bit);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
+
+
+
+static INLINE int divide_and_round(const int n, const int d) {
+ return ((n < 0) ^ (d < 0)) ? ((n - d / 2) / d) : ((n + d / 2) / d);
+}
+
+static INLINE int is_cost_list_wellbehaved(int *cost_list) {
+ return cost_list[0] < cost_list[1] &&
+ cost_list[0] < cost_list[2] &&
+ cost_list[0] < cost_list[3] &&
+ cost_list[0] < cost_list[4];
+}
+
+// Returns surface minima estimate at given precision in 1/2^n bits.
+// Assume a model for the cost surface: S = A(x - x0)^2 + B(y - y0)^2 + C
+// For a given set of costs S0, S1, S2, S3, S4 at points
+// (y, x) = (0, 0), (0, -1), (1, 0), (0, 1) and (-1, 0) respectively,
+// the solution for the location of the minima (x0, y0) is given by:
+// x0 = 1/2 (S1 - S3)/(S1 + S3 - 2*S0),
+// y0 = 1/2 (S4 - S2)/(S4 + S2 - 2*S0).
+// The code below is an integerized version of that.
+static void get_cost_surf_min(int *cost_list, int *ir, int *ic,
+ int bits) {
+ *ic = divide_and_round((cost_list[1] - cost_list[3]) * (1 << (bits - 1)),
+ (cost_list[1] - 2 * cost_list[0] + cost_list[3]));
+ *ir = divide_and_round((cost_list[4] - cost_list[2]) * (1 << (bits - 1)),
+ (cost_list[4] - 2 * cost_list[0] + cost_list[2]));
+}
+
+int vp9_find_best_sub_pixel_tree_pruned_evenmore(
+ const MACROBLOCK *x,
+ MV *bestmv, const MV *ref_mv,
+ int allow_hp,
+ int error_per_bit,
+ const vp9_variance_fn_ptr_t *vfp,
+ int forced_stop,
+ int iters_per_step,
+ int *cost_list,
+ int *mvjcost, int *mvcost[2],
+ int *distortion,
+ unsigned int *sse1,
+ const uint8_t *second_pred,
+ int w, int h) {
+ SETUP_SUBPEL_SEARCH;
+ SETUP_CENTER_ERROR;
+ (void) halfiters;
+ (void) quarteriters;
+ (void) eighthiters;
+ (void) whichdir;
+ (void) allow_hp;
+ (void) forced_stop;
+ (void) hstep;
+
+ if (cost_list &&
+ cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
+ cost_list[2] != INT_MAX && cost_list[3] != INT_MAX &&
+ cost_list[4] != INT_MAX &&
+ is_cost_list_wellbehaved(cost_list)) {
+ int ir, ic;
+ unsigned int minpt;
+ get_cost_surf_min(cost_list, &ir, &ic, 2);
+ if (ir != 0 || ic != 0) {
+ CHECK_BETTER(minpt, tr + 2 * ir, tc + 2 * ic);
+ }
+ } else {
+ FIRST_LEVEL_CHECKS;
+ if (halfiters > 1) {
+ SECOND_LEVEL_CHECKS;
+ }
+
+ tr = br;
+ tc = bc;
+
+ // Each subsequent iteration checks at least one point in common with
+ // the last iteration could be 2 ( if diag selected) 1/4 pel
+ // Note forced_stop: 0 - full, 1 - qtr only, 2 - half only
+ if (forced_stop != 2) {
+ hstep >>= 1;
+ FIRST_LEVEL_CHECKS;
+ if (quarteriters > 1) {
+ SECOND_LEVEL_CHECKS;
+ }
+ }
+ }
+
+ tr = br;
+ tc = bc;
+
+ if (allow_hp && vp9_use_mv_hp(ref_mv) && forced_stop == 0) {
+ hstep >>= 1;
+ FIRST_LEVEL_CHECKS;
+ if (eighthiters > 1) {
+ SECOND_LEVEL_CHECKS;
+ }
+ }
+
+ bestmv->row = br;
+ bestmv->col = bc;
+
+ if ((abs(bestmv->col - ref_mv->col) > (MAX_FULL_PEL_VAL << 3)) ||
+ (abs(bestmv->row - ref_mv->row) > (MAX_FULL_PEL_VAL << 3)))
+ return INT_MAX;
+
+ return besterr;
+}
+
+int vp9_find_best_sub_pixel_tree_pruned_more(const MACROBLOCK *x,
+ MV *bestmv, const MV *ref_mv,
+ int allow_hp,
+ int error_per_bit,
+ const vp9_variance_fn_ptr_t *vfp,
+ int forced_stop,
+ int iters_per_step,
+ int *cost_list,
+ int *mvjcost, int *mvcost[2],
+ int *distortion,
+ unsigned int *sse1,
+ const uint8_t *second_pred,
+ int w, int h) {
+ SETUP_SUBPEL_SEARCH;
+ SETUP_CENTER_ERROR;
+ if (cost_list &&
+ cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
+ cost_list[2] != INT_MAX && cost_list[3] != INT_MAX &&
+ cost_list[4] != INT_MAX &&
+ is_cost_list_wellbehaved(cost_list)) {
+ unsigned int minpt;
+ int ir, ic;
+ get_cost_surf_min(cost_list, &ir, &ic, 1);
+ if (ir != 0 || ic != 0) {
+ CHECK_BETTER(minpt, tr + ir * hstep, tc + ic * hstep);
+ }
+ } else {
+ FIRST_LEVEL_CHECKS;
+ if (halfiters > 1) {
+ SECOND_LEVEL_CHECKS;
+ }
+ }
+
+ // Each subsequent iteration checks at least one point in common with
+ // the last iteration could be 2 ( if diag selected) 1/4 pel
+
+ // Note forced_stop: 0 - full, 1 - qtr only, 2 - half only
+ if (forced_stop != 2) {
+ tr = br;
+ tc = bc;
+ hstep >>= 1;
+ FIRST_LEVEL_CHECKS;
+ if (quarteriters > 1) {
+ SECOND_LEVEL_CHECKS;
+ }
+ }
+
+ if (allow_hp && vp9_use_mv_hp(ref_mv) && forced_stop == 0) {
+ tr = br;
+ tc = bc;
+ hstep >>= 1;
+ FIRST_LEVEL_CHECKS;
+ if (eighthiters > 1) {
+ SECOND_LEVEL_CHECKS;
+ }
+ }
+ // These lines insure static analysis doesn't warn that
+ // tr and tc aren't used after the above point.
+ (void) tr;
+ (void) tc;
+
+ bestmv->row = br;
+ bestmv->col = bc;
+
+ if ((abs(bestmv->col - ref_mv->col) > (MAX_FULL_PEL_VAL << 3)) ||
+ (abs(bestmv->row - ref_mv->row) > (MAX_FULL_PEL_VAL << 3)))
+ return INT_MAX;
+
+ return besterr;
+}
+
int vp9_find_best_sub_pixel_tree_pruned(const MACROBLOCK *x,
MV *bestmv, const MV *ref_mv,
int allow_hp,
@@ -302,21 +505,21 @@
const vp9_variance_fn_ptr_t *vfp,
int forced_stop,
int iters_per_step,
- int *sad_list,
+ int *cost_list,
int *mvjcost, int *mvcost[2],
int *distortion,
unsigned int *sse1,
const uint8_t *second_pred,
int w, int h) {
SETUP_SUBPEL_SEARCH;
-
- if (sad_list &&
- sad_list[0] != INT_MAX && sad_list[1] != INT_MAX &&
- sad_list[2] != INT_MAX && sad_list[3] != INT_MAX &&
- sad_list[4] != INT_MAX) {
+ SETUP_CENTER_ERROR;
+ if (cost_list &&
+ cost_list[0] != INT_MAX && cost_list[1] != INT_MAX &&
+ cost_list[2] != INT_MAX && cost_list[3] != INT_MAX &&
+ cost_list[4] != INT_MAX) {
unsigned int left, right, up, down, diag;
- whichdir = (sad_list[1] < sad_list[3] ? 0 : 1) +
- (sad_list[2] < sad_list[4] ? 0 : 2);
+ whichdir = (cost_list[1] < cost_list[3] ? 0 : 1) +
+ (cost_list[2] < cost_list[4] ? 0 : 2);
switch (whichdir) {
case 0:
CHECK_BETTER(left, tr, tc - hstep);
@@ -394,14 +597,15 @@
const vp9_variance_fn_ptr_t *vfp,
int forced_stop,
int iters_per_step,
- int *sad_list,
+ int *cost_list,
int *mvjcost, int *mvcost[2],
int *distortion,
unsigned int *sse1,
const uint8_t *second_pred,
int w, int h) {
SETUP_SUBPEL_SEARCH;
- (void) sad_list; // to silence compiler warning
+ SETUP_CENTER_ERROR;
+ (void) cost_list; // to silence compiler warning
// Each subsequent iteration checks at least one point in
// common with the last iteration could be 2 ( if diag selected)
@@ -484,6 +688,57 @@
#define MAX_PATTERN_CANDIDATES 8 // max number of canddiates per scale
#define PATTERN_CANDIDATES_REF 3 // number of refinement candidates
+// Calculate and return a sad+mvcost list around an integer best pel.
+static INLINE void calc_int_cost_list(const MACROBLOCK *x,
+ const MV *ref_mv,
+ int sadpb,
+ const vp9_variance_fn_ptr_t *fn_ptr,
+ const MV *best_mv,
+ int *cost_list) {
+ static const MV neighbors[4] = {{0, -1}, {1, 0}, {0, 1}, {-1, 0}};
+ const struct buf_2d *const what = &x->plane[0].src;
+ const struct buf_2d *const in_what = &x->e_mbd.plane[0].pre[0];
+ const MV fcenter_mv = {ref_mv->row >> 3, ref_mv->col >> 3};
+ int br = best_mv->row;
+ int bc = best_mv->col;
+ MV this_mv;
+ int i;
+ unsigned int sse;
+
+ this_mv.row = br;
+ this_mv.col = bc;
+ cost_list[0] = fn_ptr->vf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv),
+ in_what->stride, &sse) +
+ mvsad_err_cost(x, &this_mv, &fcenter_mv, sadpb);
+ if (check_bounds(x, br, bc, 1)) {
+ for (i = 0; i < 4; i++) {
+ const MV this_mv = {br + neighbors[i].row,
+ bc + neighbors[i].col};
+ cost_list[i + 1] = fn_ptr->vf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv),
+ in_what->stride, &sse) +
+ // mvsad_err_cost(x, &this_mv, &fcenter_mv, sadpb);
+ mv_err_cost(&this_mv, &fcenter_mv, x->nmvjointcost, x->mvcost,
+ x->errorperbit);
+ }
+ } else {
+ for (i = 0; i < 4; i++) {
+ const MV this_mv = {br + neighbors[i].row,
+ bc + neighbors[i].col};
+ if (!is_mv_in(x, &this_mv))
+ cost_list[i + 1] = INT_MAX;
+ else
+ cost_list[i + 1] = fn_ptr->vf(what->buf, what->stride,
+ get_buf_from_mv(in_what, &this_mv),
+ in_what->stride, &sse) +
+ // mvsad_err_cost(x, &this_mv, &fcenter_mv, sadpb);
+ mv_err_cost(&this_mv, &fcenter_mv, x->nmvjointcost, x->mvcost,
+ x->errorperbit);
+ }
+ }
+}
+
// Generic pattern search function that searches over multiple scales.
// Each scale can have a different number of candidates and shape of
// candidates as indicated in the num_candidates and candidates arrays
@@ -494,7 +749,7 @@
int search_param,
int sad_per_bit,
int do_init_search,
- int *sad_list,
+ int *cost_list,
const vp9_variance_fn_ptr_t *vfp,
int use_mvcost,
const MV *center_mv,
@@ -646,40 +901,14 @@
}
// Returns the one-away integer pel sad values around the best as follows:
- // sad_list[0]: sad at the best integer pel
- // sad_list[1]: sad at delta {0, -1} (left) from the best integer pel
- // sad_list[2]: sad at delta { 1, 0} (bottom) from the best integer pel
- // sad_list[3]: sad at delta { 0, 1} (right) from the best integer pel
- // sad_list[4]: sad at delta {-1, 0} (top) from the best integer pel
- if (sad_list) {
- static const MV neighbors[4] = {{0, -1}, {1, 0}, {0, 1}, {-1, 0}};
- sad_list[0] = bestsad;
- if (check_bounds(x, br, bc, 1)) {
- for (i = 0; i < 4; i++) {
- const MV this_mv = {br + neighbors[i].row,
- bc + neighbors[i].col};
- sad_list[i + 1] = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride) +
- (use_mvcost ?
- mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit) :
- 0);
- }
- } else {
- for (i = 0; i < 4; i++) {
- const MV this_mv = {br + neighbors[i].row,
- bc + neighbors[i].col};
- if (!is_mv_in(x, &this_mv))
- sad_list[i + 1] = INT_MAX;
- else
- sad_list[i + 1] = vfp->sdf(what->buf, what->stride,
- get_buf_from_mv(in_what, &this_mv),
- in_what->stride) +
- (use_mvcost ?
- mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit) :
- 0);
- }
- }
+ // cost_list[0]: cost at the best integer pel
+ // cost_list[1]: cost at delta {0, -1} (left) from the best integer pel
+ // cost_list[2]: cost at delta { 1, 0} (bottom) from the best integer pel
+ // cost_list[3]: cost at delta { 0, 1} (right) from the best integer pel
+ // cost_list[4]: cost at delta {-1, 0} (top) from the best integer pel
+ if (cost_list) {
+ const MV best_mv = { br, bc };
+ calc_int_cost_list(x, &fcenter_mv, sad_per_bit, vfp, &best_mv, cost_list);
}
best_mv->row = br;
best_mv->col = bc;
@@ -687,7 +916,7 @@
}
// A specialized function where the smallest scale search candidates
-// are 4 1-away neighbors, and sad_list is non-null
+// are 4 1-away neighbors, and cost_list is non-null
// TODO(debargha): Merge this function with the one above. Also remove
// use_mvcost option since it is always 1, to save unnecessary branches.
static int vp9_pattern_search_sad(const MACROBLOCK *x,
@@ -695,7 +924,7 @@
int search_param,
int sad_per_bit,
int do_init_search,
- int *sad_list,
+ int *cost_list,
const vp9_variance_fn_ptr_t *vfp,
int use_mvcost,
const MV *center_mv,
@@ -720,8 +949,8 @@
clamp_mv(ref_mv, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max);
br = ref_mv->row;
bc = ref_mv->col;
- if (sad_list != NULL) {
- sad_list[0] = sad_list[1] = sad_list[2] = sad_list[3] = sad_list[4] =
+ if (cost_list != NULL) {
+ cost_list[0] = cost_list[1] = cost_list[2] = cost_list[3] = cost_list[4] =
INT_MAX;
}
@@ -775,7 +1004,7 @@
// If the center point is still the best, just skip this and move to
// the refinement step.
if (best_init_s != -1) {
- int do_sad = (num_candidates[0] == 4 && sad_list != NULL);
+ int do_sad = (num_candidates[0] == 4 && cost_list != NULL);
int best_site = -1;
s = best_init_s;
@@ -849,15 +1078,15 @@
} while (best_site != -1);
}
- // Note: If we enter the if below, then sad_list must be non-NULL.
+ // Note: If we enter the if below, then cost_list must be non-NULL.
if (s == 0) {
- sad_list[0] = bestsad;
+ cost_list[0] = bestsad;
if (!do_init_search || s != best_init_s) {
if (check_bounds(x, br, bc, 1 << s)) {
for (i = 0; i < num_candidates[s]; i++) {
const MV this_mv = {br + candidates[s][i].row,
bc + candidates[s][i].col};
- sad_list[i + 1] =
+ cost_list[i + 1] =
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
in_what->stride);
@@ -869,7 +1098,7 @@
bc + candidates[s][i].col};
if (!is_mv_in(x, &this_mv))
continue;
- sad_list[i + 1] =
+ cost_list[i + 1] =
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
in_what->stride);
@@ -889,15 +1118,15 @@
next_chkpts_indices[0] = (k == 0) ? num_candidates[s] - 1 : k - 1;
next_chkpts_indices[1] = k;
next_chkpts_indices[2] = (k == num_candidates[s] - 1) ? 0 : k + 1;
- sad_list[1] = sad_list[2] = sad_list[3] = sad_list[4] = INT_MAX;
- sad_list[((k + 2) % 4) + 1] = sad_list[0];
- sad_list[0] = bestsad;
+ cost_list[1] = cost_list[2] = cost_list[3] = cost_list[4] = INT_MAX;
+ cost_list[((k + 2) % 4) + 1] = cost_list[0];
+ cost_list[0] = bestsad;
if (check_bounds(x, br, bc, 1 << s)) {
for (i = 0; i < PATTERN_CANDIDATES_REF; i++) {
const MV this_mv = {br + candidates[s][next_chkpts_indices[i]].row,
bc + candidates[s][next_chkpts_indices[i]].col};
- sad_list[next_chkpts_indices[i] + 1] =
+ cost_list[next_chkpts_indices[i] + 1] =
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
in_what->stride);
@@ -908,10 +1137,10 @@
const MV this_mv = {br + candidates[s][next_chkpts_indices[i]].row,
bc + candidates[s][next_chkpts_indices[i]].col};
if (!is_mv_in(x, &this_mv)) {
- sad_list[next_chkpts_indices[i] + 1] = INT_MAX;
+ cost_list[next_chkpts_indices[i] + 1] = INT_MAX;
continue;
}
- sad_list[next_chkpts_indices[i] + 1] =
+ cost_list[next_chkpts_indices[i] + 1] =
thissad = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
in_what->stride);
@@ -929,20 +1158,20 @@
}
// Returns the one-away integer pel sad values around the best as follows:
- // sad_list[0]: sad at the best integer pel
- // sad_list[1]: sad at delta {0, -1} (left) from the best integer pel
- // sad_list[2]: sad at delta { 1, 0} (bottom) from the best integer pel
- // sad_list[3]: sad at delta { 0, 1} (right) from the best integer pel
- // sad_list[4]: sad at delta {-1, 0} (top) from the best integer pel
- if (sad_list) {
+ // cost_list[0]: sad at the best integer pel
+ // cost_list[1]: sad at delta {0, -1} (left) from the best integer pel
+ // cost_list[2]: sad at delta { 1, 0} (bottom) from the best integer pel
+ // cost_list[3]: sad at delta { 0, 1} (right) from the best integer pel
+ // cost_list[4]: sad at delta {-1, 0} (top) from the best integer pel
+ if (cost_list) {
static const MV neighbors[4] = {{0, -1}, {1, 0}, {0, 1}, {-1, 0}};
- if (sad_list[0] == INT_MAX) {
- sad_list[0] = bestsad;
+ if (cost_list[0] == INT_MAX) {
+ cost_list[0] = bestsad;
if (check_bounds(x, br, bc, 1)) {
for (i = 0; i < 4; i++) {
- const MV this_mv = {br + neighbors[i].row,
- bc + neighbors[i].col};
- sad_list[i + 1] = vfp->sdf(what->buf, what->stride,
+ const MV this_mv = { br + neighbors[i].row,
+ bc + neighbors[i].col };
+ cost_list[i + 1] = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
in_what->stride);
}
@@ -951,9 +1180,9 @@
const MV this_mv = {br + neighbors[i].row,
bc + neighbors[i].col};
if (!is_mv_in(x, &this_mv))
- sad_list[i + 1] = INT_MAX;
+ cost_list[i + 1] = INT_MAX;
else
- sad_list[i + 1] = vfp->sdf(what->buf, what->stride,
+ cost_list[i + 1] = vfp->sdf(what->buf, what->stride,
get_buf_from_mv(in_what, &this_mv),
in_what->stride);
}
@@ -963,8 +1192,8 @@
for (i = 0; i < 4; i++) {
const MV this_mv = {br + neighbors[i].row,
bc + neighbors[i].col};
- if (sad_list[i + 1] != INT_MAX) {
- sad_list[i + 1] +=
+ if (cost_list[i + 1] != INT_MAX) {
+ cost_list[i + 1] +=
mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
}
}
@@ -1014,7 +1243,7 @@
int search_param,
int sad_per_bit,
int do_init_search,
- int *sad_list,
+ int *cost_list,
const vp9_variance_fn_ptr_t *vfp,
int use_mvcost,
const MV *center_mv, MV *best_mv) {
@@ -1039,7 +1268,7 @@
{ -1024, 0}},
};
return vp9_pattern_search(x, ref_mv, search_param, sad_per_bit,
- do_init_search, sad_list, vfp, use_mvcost,
+ do_init_search, cost_list, vfp, use_mvcost,
center_mv, best_mv,
hex_num_candidates, hex_candidates);
}
@@ -1049,7 +1278,7 @@
int search_param,
int sad_per_bit,
int do_init_search,
- int *sad_list,
+ int *cost_list,
const vp9_variance_fn_ptr_t *vfp,
int use_mvcost,
const MV *center_mv,
@@ -1081,7 +1310,7 @@
{-512, 512}, {-1024, 0}},
};
return vp9_pattern_search_sad(x, ref_mv, search_param, sad_per_bit,
- do_init_search, sad_list, vfp, use_mvcost,
+ do_init_search, cost_list, vfp, use_mvcost,
center_mv, best_mv,
bigdia_num_candidates, bigdia_candidates);
}
@@ -1091,7 +1320,7 @@
int search_param,
int sad_per_bit,
int do_init_search,
- int *sad_list,
+ int *cost_list,
const vp9_variance_fn_ptr_t *vfp,
int use_mvcost,
const MV *center_mv,
@@ -1123,7 +1352,7 @@
{0, 1024}, {-1024, 1024}, {-1024, 0}},
};
return vp9_pattern_search(x, ref_mv, search_param, sad_per_bit,
- do_init_search, sad_list, vfp, use_mvcost,
+ do_init_search, cost_list, vfp, use_mvcost,
center_mv, best_mv,
square_num_candidates, square_candidates);
}
@@ -1133,13 +1362,13 @@
int search_param,
int sad_per_bit,
int do_init_search, // must be zero for fast_hex
- int *sad_list,
+ int *cost_list,
const vp9_variance_fn_ptr_t *vfp,
int use_mvcost,
const MV *center_mv,
MV *best_mv) {
return vp9_hex_search(x, ref_mv, MAX(MAX_MVSEARCH_STEPS - 2, search_param),
- sad_per_bit, do_init_search, sad_list, vfp, use_mvcost,
+ sad_per_bit, do_init_search, cost_list, vfp, use_mvcost,
center_mv, best_mv);
}
@@ -1148,13 +1377,13 @@
int search_param,
int sad_per_bit,
int do_init_search,
- int *sad_list,
+ int *cost_list,
const vp9_variance_fn_ptr_t *vfp,
int use_mvcost,
const MV *center_mv,
MV *best_mv) {
return vp9_bigdia_search(x, ref_mv, MAX(MAX_MVSEARCH_STEPS - 2, search_param),
- sad_per_bit, do_init_search, sad_list, vfp,
+ sad_per_bit, do_init_search, cost_list, vfp,
use_mvcost, center_mv, best_mv);
}
@@ -1378,10 +1607,10 @@
/* do_refine: If last step (1-away) of n-step search doesn't pick the center
point as the best match, we will do a final 1-away diamond
refining search */
-
int vp9_full_pixel_diamond(const VP9_COMP *cpi, MACROBLOCK *x,
MV *mvp_full, int step_param,
int sadpb, int further_steps, int do_refine,
+ int *cost_list,
const vp9_variance_fn_ptr_t *fn_ptr,
const MV *ref_mv, MV *dst_mv) {
MV temp_mv;
@@ -1434,6 +1663,11 @@
*dst_mv = best_mv;
}
}
+
+ // Return cost list.
+ if (cost_list) {
+ calc_int_cost_list(x, ref_mv, sadpb, fn_ptr, dst_mv, cost_list);
+ }
return bestsme;
}
@@ -1753,46 +1987,46 @@
int vp9_full_pixel_search(VP9_COMP *cpi, MACROBLOCK *x,
BLOCK_SIZE bsize, MV *mvp_full,
int step_param, int error_per_bit,
- int *sad_list,
+ int *cost_list,
const MV *ref_mv, MV *tmp_mv,
int var_max, int rd) {
const SPEED_FEATURES *const sf = &cpi->sf;
const SEARCH_METHODS method = sf->mv.search_method;
vp9_variance_fn_ptr_t *fn_ptr = &cpi->fn_ptr[bsize];
int var = 0;
- if (sad_list) {
- sad_list[0] = INT_MAX;
- sad_list[1] = INT_MAX;
- sad_list[2] = INT_MAX;
- sad_list[3] = INT_MAX;
- sad_list[4] = INT_MAX;
+ if (cost_list) {
+ cost_list[0] = INT_MAX;
+ cost_list[1] = INT_MAX;
+ cost_list[2] = INT_MAX;
+ cost_list[3] = INT_MAX;
+ cost_list[4] = INT_MAX;
}
switch (method) {
case FAST_DIAMOND:
var = vp9_fast_dia_search(x, mvp_full, step_param, error_per_bit, 0,
- sad_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case FAST_HEX:
var = vp9_fast_hex_search(x, mvp_full, step_param, error_per_bit, 0,
- sad_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case HEX:
var = vp9_hex_search(x, mvp_full, step_param, error_per_bit, 1,
- sad_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case SQUARE:
var = vp9_square_search(x, mvp_full, step_param, error_per_bit, 1,
- sad_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case BIGDIA:
var = vp9_bigdia_search(x, mvp_full, step_param, error_per_bit, 1,
- sad_list, fn_ptr, 1, ref_mv, tmp_mv);
+ cost_list, fn_ptr, 1, ref_mv, tmp_mv);
break;
case NSTEP:
var = vp9_full_pixel_diamond(cpi, x, mvp_full, step_param, error_per_bit,
MAX_MVSEARCH_STEPS - 1 - step_param,
- 1, fn_ptr, ref_mv, tmp_mv);
+ 1, cost_list, fn_ptr, ref_mv, tmp_mv);
break;
default:
assert(!"Invalid search method.");
diff --git a/source/libvpx/vp9/encoder/vp9_mcomp.h b/source/libvpx/vp9/encoder/vp9_mcomp.h
index 9b4734a..9ddca25 100644
--- a/source/libvpx/vp9/encoder/vp9_mcomp.h
+++ b/source/libvpx/vp9/encoder/vp9_mcomp.h
@@ -70,6 +70,7 @@
int vp9_full_pixel_diamond(const struct VP9_COMP *cpi, MACROBLOCK *x,
MV *mvp_full, int step_param,
int sadpb, int further_steps, int do_refine,
+ int *cost_list,
const vp9_variance_fn_ptr_t *fn_ptr,
const MV *ref_mv, MV *dst_mv);
@@ -79,7 +80,7 @@
int search_param,
int error_per_bit,
int do_init_search,
- int *sad_list,
+ int *cost_list,
const vp9_variance_fn_ptr_t *vf,
int use_mvcost,
const MV *center_mv,
@@ -99,7 +100,7 @@
const vp9_variance_fn_ptr_t *vfp,
int forced_stop, // 0 - full, 1 - qtr only, 2 - half only
int iters_per_step,
- int *sad_list,
+ int *cost_list,
int *mvjcost, int *mvcost[2],
int *distortion, unsigned int *sse1,
const uint8_t *second_pred,
@@ -107,6 +108,8 @@
extern fractional_mv_step_fp vp9_find_best_sub_pixel_tree;
extern fractional_mv_step_fp vp9_find_best_sub_pixel_tree_pruned;
+extern fractional_mv_step_fp vp9_find_best_sub_pixel_tree_pruned_more;
+extern fractional_mv_step_fp vp9_find_best_sub_pixel_tree_pruned_evenmore;
typedef int (*vp9_full_search_fn_t)(const MACROBLOCK *x,
const MV *ref_mv, int sad_per_bit,
@@ -139,7 +142,7 @@
int vp9_full_pixel_search(struct VP9_COMP *cpi, MACROBLOCK *x,
BLOCK_SIZE bsize, MV *mvp_full,
int step_param, int error_per_bit,
- int *sad_list,
+ int *cost_list,
const MV *ref_mv, MV *tmp_mv,
int var_max, int rd);
diff --git a/source/libvpx/vp9/encoder/vp9_picklpf.c b/source/libvpx/vp9/encoder/vp9_picklpf.c
index 2fc05e7..85984fd 100644
--- a/source/libvpx/vp9/encoder/vp9_picklpf.c
+++ b/source/libvpx/vp9/encoder/vp9_picklpf.c
@@ -40,7 +40,15 @@
vp9_loop_filter_frame(cm->frame_to_show, cm, &cpi->mb.e_mbd, filt_level, 1,
partial_frame);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ filt_err = vp9_highbd_get_y_sse(sd, cm->frame_to_show, cm->bit_depth);
+ } else {
+ filt_err = vp9_get_y_sse(sd, cm->frame_to_show);
+ }
+#else
filt_err = vp9_get_y_sse(sd, cm->frame_to_show);
+#endif // CONFIG_VP9_HIGHBITDEPTH
// Re-instate the unfiltered frame
vpx_yv12_copy_y(&cpi->last_frame_uf, cm->frame_to_show);
@@ -145,7 +153,26 @@
const int q = vp9_ac_quant(cm->base_qindex, 0, cm->bit_depth);
// These values were determined by linear fitting the result of the
// searched level, filt_guess = q * 0.316206 + 3.87252
+#if CONFIG_VP9_HIGHDEPTH
+ int filt_guess;
+ switch (cm->bit_depth) {
+ case VPX_BITS_8:
+ filt_guess = ROUND_POWER_OF_TWO(q * 20723 + 1015158, 18);
+ break;
+ case VPX_BITS_10:
+ filt_guess = ROUND_POWER_OF_TWO(q * 20723 + 4060632, 20);
+ break;
+ case VPX_BITS_12:
+ filt_guess = ROUND_POWER_OF_TWO(q * 20723 + 16242526, 22);
+ break;
+ default:
+ assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 "
+ "or VPX_BITS_12");
+ return;
+ }
+#else
int filt_guess = ROUND_POWER_OF_TWO(q * 20723 + 1015158, 18);
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (cm->frame_type == KEY_FRAME)
filt_guess -= 4;
lf->filter_level = clamp(filt_guess, min_filter_level, max_filter_level);
diff --git a/source/libvpx/vp9/encoder/vp9_pickmode.c b/source/libvpx/vp9/encoder/vp9_pickmode.c
index a788c1d..b74b2dd 100644
--- a/source/libvpx/vp9/encoder/vp9_pickmode.c
+++ b/source/libvpx/vp9/encoder/vp9_pickmode.c
@@ -132,7 +132,7 @@
const int tmp_row_min = x->mv_row_min;
const int tmp_row_max = x->mv_row_max;
int rv = 0;
- int sad_list[5];
+ int cost_list[5];
const YV12_BUFFER_CONFIG *scaled_ref_frame = vp9_get_scaled_ref_frame(cpi,
ref);
if (cpi->common.show_frame &&
@@ -160,7 +160,7 @@
mvp_full.row >>= 3;
vp9_full_pixel_search(cpi, x, bsize, &mvp_full, step_param, sadpb,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
&ref_mv, &tmp_mv->as_mv, INT_MAX, 0);
x->mv_col_min = tmp_col_min;
@@ -187,7 +187,7 @@
&cpi->fn_ptr[bsize],
cpi->sf.mv.subpel_force_stop,
cpi->sf.mv.subpel_iters_per_step,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
x->nmvjointcost, x->mvcost,
&dis, &x->pred_sse[ref], NULL, 0, 0);
x->pred_mv[ref] = tmp_mv->as_mv;
@@ -235,19 +235,54 @@
tx_mode_to_biggest_tx_size[cpi->common.tx_mode]);
else
xd->mi[0].src_mi->mbmi.tx_size = TX_8X8;
+
+ if (cpi->sf.partition_search_type == VAR_BASED_PARTITION &&
+ xd->mi[0].src_mi->mbmi.tx_size > TX_16X16)
+ xd->mi[0].src_mi->mbmi.tx_size = TX_16X16;
} else {
xd->mi[0].src_mi->mbmi.tx_size =
MIN(max_txsize_lookup[bsize],
tx_mode_to_biggest_tx_size[cpi->common.tx_mode]);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_model_rd_from_var_lapndz(sse - var, 1 << num_pels_log2_lookup[bsize],
+ dc_quant >> (xd->bd - 5), &rate, &dist);
+ } else {
+ vp9_model_rd_from_var_lapndz(sse - var, 1 << num_pels_log2_lookup[bsize],
+ dc_quant >> 3, &rate, &dist);
+ }
+#else
vp9_model_rd_from_var_lapndz(sse - var, 1 << num_pels_log2_lookup[bsize],
dc_quant >> 3, &rate, &dist);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
*out_rate_sum = rate >> 1;
*out_dist_sum = dist << 3;
- vp9_model_rd_from_var_lapndz(var, 1 << num_pels_log2_lookup[bsize],
- ac_quant >> 3, &rate, &dist);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_model_rd_from_var_lapndz(var,
+ 1 << num_pels_log2_lookup[bsize],
+ ac_quant >> (xd->bd - 5),
+ &rate,
+ &dist);
+ } else {
+ vp9_model_rd_from_var_lapndz(var,
+ 1 << num_pels_log2_lookup[bsize],
+ ac_quant >> 3,
+ &rate,
+ &dist);
+ }
+#else
+ vp9_model_rd_from_var_lapndz(var,
+ 1 << num_pels_log2_lookup[bsize],
+ ac_quant >> 3,
+ &rate,
+ &dist);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
*out_rate_sum += rate;
*out_dist_sum += dist << 4;
}
@@ -293,16 +328,29 @@
// The encode_breakout input
const unsigned int min_thresh =
MIN(((unsigned int)x->encode_breakout << 4), max_thresh);
+#if CONFIG_VP9_HIGHBITDEPTH
+ const int shift = 2 * xd->bd - 16;
+#endif
// Calculate threshold according to dequant value.
thresh_ac = (xd->plane[0].dequant[1] * xd->plane[0].dequant[1]) / 9;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if ((xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) && shift > 0) {
+ thresh_ac = ROUND_POWER_OF_TWO(thresh_ac, shift);
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
thresh_ac = clamp(thresh_ac, min_thresh, max_thresh);
// Adjust ac threshold according to partition size.
thresh_ac >>=
- 8 - (b_width_log2(bsize) + b_height_log2(bsize));
+ 8 - (b_width_log2_lookup[bsize] + b_height_log2_lookup[bsize]);
thresh_dc = (xd->plane[0].dequant[0] * xd->plane[0].dequant[0] >> 6);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if ((xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) && shift > 0) {
+ thresh_dc = ROUND_POWER_OF_TWO(thresh_dc, shift);
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
} else {
thresh_ac = 0;
thresh_dc = 0;
@@ -389,7 +437,7 @@
pd->dst.buf = &dst_buf_base[4 * (j * dst_stride + i)];
// Use source buffer as an approximation for the fully reconstructed buffer.
vp9_predict_intra_block(xd, block >> (2 * tx_size),
- b_width_log2(plane_bsize),
+ b_width_log2_lookup[plane_bsize],
tx_size, args->mode,
p->src.buf, src_stride,
pd->dst.buf, dst_stride,
@@ -438,9 +486,12 @@
// var_y and sse_y are saved to be used in skipping checking
unsigned int var_y = UINT_MAX;
unsigned int sse_y = UINT_MAX;
-
- const int intra_cost_penalty =
- 20 * vp9_dc_quant(cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth);
+ // Reduce the intra cost penalty for small blocks (<=16x16).
+ const int reduction_fac =
+ (cpi->sf.partition_search_type == VAR_BASED_PARTITION &&
+ bsize <= BLOCK_16X16) ? 4 : 1;
+ const int intra_cost_penalty = vp9_get_intra_cost_penalty(
+ cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth) / reduction_fac;
const int64_t inter_mode_thresh = RDCOST(x->rdmult, x->rddiv,
intra_cost_penalty, 0);
const int intra_mode_cost = 50;
@@ -449,7 +500,7 @@
const int *const rd_threshes = cpi->rd.threshes[segment_id][bsize];
const int *const rd_thresh_freq_fact = cpi->rd.thresh_freq_fact[bsize];
INTERP_FILTER filter_ref = cm->interp_filter;
- const int bsl = mi_width_log2(bsize);
+ const int bsl = mi_width_log2_lookup[bsize];
const int pred_filter_search = cm->interp_filter == SWITCHABLE ?
(((mi_row + mi_col) >> bsl) +
get_chessboard_index(cm->current_video_frame)) & 0x1 : 0;
@@ -461,14 +512,25 @@
// tmp[3] points to dst buffer, and the other 3 point to allocated buffers.
PRED_BUFFER tmp[4];
DECLARE_ALIGNED_ARRAY(16, uint8_t, pred_buf, 3 * 64 * 64);
+#if CONFIG_VP9_HIGHBITDEPTH
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, pred_buf_16, 3 * 64 * 64);
+#endif
struct buf_2d orig_dst = pd->dst;
PRED_BUFFER *best_pred = NULL;
PRED_BUFFER *this_mode_pred = NULL;
+ const int pixels_in_block = bh * bw;
if (cpi->sf.reuse_inter_pred_sby) {
int i;
for (i = 0; i < 3; i++) {
- tmp[i].data = &pred_buf[bw * bh * i];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth)
+ tmp[i].data = CONVERT_TO_BYTEPTR(&pred_buf_16[pixels_in_block * i]);
+ else
+ tmp[i].data = &pred_buf[pixels_in_block * i];
+#else
+ tmp[i].data = &pred_buf[pixels_in_block * i];
+#endif // CONFIG_VP9_HIGHBITDEPTH
tmp[i].stride = bw;
tmp[i].in_use = 0;
}
@@ -557,7 +619,8 @@
continue;
if (this_mode == NEWMV) {
- if (this_rd < (int64_t)(1 << num_pels_log2_lookup[bsize]))
+ if (cpi->sf.partition_search_type != VAR_BASED_PARTITION &&
+ this_rd < (int64_t)(1 << num_pels_log2_lookup[bsize]))
continue;
if (!combined_motion_search(cpi, x, bsize, mi_row, mi_col,
&frame_mv[NEWMV][ref_frame],
@@ -703,8 +766,18 @@
if (best_pred != NULL && cpi->sf.reuse_inter_pred_sby &&
best_pred->data != orig_dst.buf) {
pd->dst = orig_dst;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (cm->use_highbitdepth) {
+ vp9_highbd_convolve_copy(best_pred->data, bw, pd->dst.buf, pd->dst.stride,
+ NULL, 0, NULL, 0, bw, bh, xd->bd);
+ } else {
+ vp9_convolve_copy(best_pred->data, bw, pd->dst.buf, pd->dst.stride,
+ NULL, 0, NULL, 0, bw, bh);
+ }
+#else
vp9_convolve_copy(best_pred->data, bw, pd->dst.buf, pd->dst.stride, NULL, 0,
NULL, 0, bw, bh);
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
mbmi->mode = best_mode;
diff --git a/source/libvpx/vp9/encoder/vp9_quantize.c b/source/libvpx/vp9/encoder/vp9_quantize.c
index 3d2c409..2ba1f92 100644
--- a/source/libvpx/vp9/encoder/vp9_quantize.c
+++ b/source/libvpx/vp9/encoder/vp9_quantize.c
@@ -41,10 +41,10 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_quantize_dc(const tran_low_t *coeff_ptr, int skip_block,
- const int16_t *round_ptr, const int16_t quant,
- tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
- const int16_t dequant_ptr, uint16_t *eob_ptr) {
+void vp9_highbd_quantize_dc(const tran_low_t *coeff_ptr, int skip_block,
+ const int16_t *round_ptr, const int16_t quant,
+ tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
+ const int16_t dequant_ptr, uint16_t *eob_ptr) {
int eob = -1;
if (!skip_block) {
@@ -88,10 +88,14 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_quantize_dc_32x32(const tran_low_t *coeff_ptr, int skip_block,
- const int16_t *round_ptr, const int16_t quant,
- tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
- const int16_t dequant_ptr, uint16_t *eob_ptr) {
+void vp9_highbd_quantize_dc_32x32(const tran_low_t *coeff_ptr,
+ int skip_block,
+ const int16_t *round_ptr,
+ const int16_t quant,
+ tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr,
+ const int16_t dequant_ptr,
+ uint16_t *eob_ptr) {
int eob = -1;
if (!skip_block) {
@@ -154,14 +158,20 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_quantize_fp_c(const tran_low_t *coeff_ptr, intptr_t count,
- int skip_block, const int16_t *zbin_ptr,
- const int16_t *round_ptr, const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr,
- int zbin_oq_value, uint16_t *eob_ptr,
- const int16_t *scan, const int16_t *iscan) {
+void vp9_highbd_quantize_fp_c(const tran_low_t *coeff_ptr,
+ intptr_t count,
+ int skip_block,
+ const int16_t *zbin_ptr,
+ const int16_t *round_ptr,
+ const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr,
+ tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr,
+ const int16_t *dequant_ptr,
+ int zbin_oq_value,
+ uint16_t *eob_ptr,
+ const int16_t *scan,
+ const int16_t *iscan) {
int i;
int eob = -1;
// TODO(jingning) Decide the need of these arguments after the
@@ -242,17 +252,17 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_quantize_fp_32x32_c(const tran_low_t *coeff_ptr,
- intptr_t n_coeffs, int skip_block,
- const int16_t *zbin_ptr,
- const int16_t *round_ptr,
- const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr,
- tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr,
- int zbin_oq_value, uint16_t *eob_ptr,
- const int16_t *scan, const int16_t *iscan) {
+void vp9_highbd_quantize_fp_32x32_c(const tran_low_t *coeff_ptr,
+ intptr_t n_coeffs, int skip_block,
+ const int16_t *zbin_ptr,
+ const int16_t *round_ptr,
+ const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr,
+ tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr,
+ const int16_t *dequant_ptr,
+ int zbin_oq_value, uint16_t *eob_ptr,
+ const int16_t *scan, const int16_t *iscan) {
int i, eob = -1;
(void)zbin_ptr;
(void)quant_shift_ptr;
@@ -340,14 +350,14 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_quantize_b_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs,
- int skip_block, const int16_t *zbin_ptr,
- const int16_t *round_ptr, const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr, int zbin_oq_value,
- uint16_t *eob_ptr, const int16_t *scan,
- const int16_t *iscan) {
+void vp9_highbd_quantize_b_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs,
+ int skip_block, const int16_t *zbin_ptr,
+ const int16_t *round_ptr, const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr,
+ tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
+ const int16_t *dequant_ptr, int zbin_oq_value,
+ uint16_t *eob_ptr, const int16_t *scan,
+ const int16_t *iscan) {
int i, non_zero_count = (int)n_coeffs, eob = -1;
const int zbins[2] = { zbin_ptr[0] + zbin_oq_value,
zbin_ptr[1] + zbin_oq_value };
@@ -452,17 +462,17 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_quantize_b_32x32_c(const tran_low_t *coeff_ptr,
- intptr_t n_coeffs, int skip_block,
- const int16_t *zbin_ptr,
- const int16_t *round_ptr,
- const int16_t *quant_ptr,
- const int16_t *quant_shift_ptr,
- tran_low_t *qcoeff_ptr,
- tran_low_t *dqcoeff_ptr,
- const int16_t *dequant_ptr,
- int zbin_oq_value, uint16_t *eob_ptr,
- const int16_t *scan, const int16_t *iscan) {
+void vp9_highbd_quantize_b_32x32_c(const tran_low_t *coeff_ptr,
+ intptr_t n_coeffs, int skip_block,
+ const int16_t *zbin_ptr,
+ const int16_t *round_ptr,
+ const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr,
+ tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr,
+ const int16_t *dequant_ptr,
+ int zbin_oq_value, uint16_t *eob_ptr,
+ const int16_t *scan, const int16_t *iscan) {
const int zbins[2] = { ROUND_POWER_OF_TWO(zbin_ptr[0] + zbin_oq_value, 1),
ROUND_POWER_OF_TWO(zbin_ptr[1] + zbin_oq_value, 1) };
const int nzbins[2] = { zbins[0] * -1, zbins[1] * -1 };
@@ -519,7 +529,7 @@
#if CONFIG_VP9_HIGHBITDEPTH
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
- vp9_high_quantize_b(BLOCK_OFFSET(p->coeff, block),
+ vp9_highbd_quantize_b(BLOCK_OFFSET(p->coeff, block),
16, x->skip_block,
p->zbin, p->round, p->quant, p->quant_shift,
BLOCK_OFFSET(p->qcoeff, block),
@@ -641,13 +651,14 @@
x->plane[0].quant_shift = quants->y_quant_shift[qindex];
x->plane[0].zbin = quants->y_zbin[qindex];
x->plane[0].round = quants->y_round[qindex];
- x->plane[0].quant_thred[0] = cm->y_dequant[qindex][0] *
- cm->y_dequant[qindex][0];
- x->plane[0].quant_thred[1] = cm->y_dequant[qindex][1] *
- cm->y_dequant[qindex][1];
x->plane[0].zbin_extra = (int16_t)((cm->y_dequant[qindex][1] * zbin) >> 7);
xd->plane[0].dequant = cm->y_dequant[qindex];
+ x->plane[0].quant_thred[0] = (x->plane[0].zbin[0] + x->plane[0].zbin_extra) *
+ (x->plane[0].zbin[0] + x->plane[0].zbin_extra);
+ x->plane[0].quant_thred[1] = (x->plane[0].zbin[1] + x->plane[0].zbin_extra) *
+ (x->plane[0].zbin[1] + x->plane[0].zbin_extra);
+
// UV
for (i = 1; i < 3; i++) {
x->plane[i].quant = quants->uv_quant[qindex];
@@ -656,12 +667,15 @@
x->plane[i].quant_shift = quants->uv_quant_shift[qindex];
x->plane[i].zbin = quants->uv_zbin[qindex];
x->plane[i].round = quants->uv_round[qindex];
- x->plane[i].quant_thred[0] = cm->y_dequant[qindex][0] *
- cm->y_dequant[qindex][0];
- x->plane[i].quant_thred[1] = cm->y_dequant[qindex][1] *
- cm->y_dequant[qindex][1];
x->plane[i].zbin_extra = (int16_t)((cm->uv_dequant[qindex][1] * zbin) >> 7);
xd->plane[i].dequant = cm->uv_dequant[qindex];
+
+ x->plane[i].quant_thred[0] =
+ (x->plane[i].zbin[0] + x->plane[i].zbin_extra) *
+ (x->plane[i].zbin[0] + x->plane[i].zbin_extra);
+ x->plane[i].quant_thred[1] =
+ (x->plane[i].zbin[1] + x->plane[i].zbin_extra) *
+ (x->plane[i].zbin[1] + x->plane[i].zbin_extra);
}
x->skip_block = vp9_segfeature_active(&cm->seg, segment_id, SEG_LVL_SKIP);
diff --git a/source/libvpx/vp9/encoder/vp9_quantize.h b/source/libvpx/vp9/encoder/vp9_quantize.h
index d7edb0b..cee46e7 100644
--- a/source/libvpx/vp9/encoder/vp9_quantize.h
+++ b/source/libvpx/vp9/encoder/vp9_quantize.h
@@ -49,15 +49,18 @@
const int16_t *scan, const int16_t *iscan);
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_quantize_dc(const tran_low_t *coeff_ptr, int skip_block,
- const int16_t *round_ptr, const int16_t quant_ptr,
- tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
- const int16_t dequant_ptr, uint16_t *eob_ptr);
-void vp9_high_quantize_dc_32x32(const tran_low_t *coeff_ptr, int skip_block,
- const int16_t *round_ptr,
- const int16_t quant_ptr, tran_low_t *qcoeff_ptr,
- tran_low_t *dqcoeff_ptr,
- const int16_t dequant_ptr, uint16_t *eob_ptr);
+void vp9_highbd_quantize_dc(const tran_low_t *coeff_ptr, int skip_block,
+ const int16_t *round_ptr, const int16_t quant_ptr,
+ tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr,
+ const int16_t dequant_ptr, uint16_t *eob_ptr);
+void vp9_highbd_quantize_dc_32x32(const tran_low_t *coeff_ptr,
+ int skip_block,
+ const int16_t *round_ptr,
+ const int16_t quant_ptr,
+ tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr,
+ const int16_t dequant_ptr,
+ uint16_t *eob_ptr);
#endif
struct VP9_COMP;
diff --git a/source/libvpx/vp9/encoder/vp9_ratectrl.c b/source/libvpx/vp9/encoder/vp9_ratectrl.c
index 9b6c773..65bca66 100644
--- a/source/libvpx/vp9/encoder/vp9_ratectrl.c
+++ b/source/libvpx/vp9/encoder/vp9_ratectrl.c
@@ -177,6 +177,9 @@
const double q = vp9_convert_qindex_to_q(qindex, bit_depth);
int enumerator = frame_type == KEY_FRAME ? 2700000 : 1800000;
+ assert(correction_factor <= MAX_BPB_FACTOR &&
+ correction_factor >= MIN_BPB_FACTOR);
+
// q based adjustment to baseline enumerator
enumerator += (int)(enumerator * q) >> 12;
return (int)(enumerator * correction_factor / q);
@@ -187,7 +190,8 @@
vpx_bit_depth_t bit_depth) {
const int bpm = (int)(vp9_rc_bits_per_mb(frame_type, q, correction_factor,
bit_depth));
- return ((uint64_t)bpm * mbs) >> BPER_MB_NORMBITS;
+ return MAX(FRAME_OVERHEAD_BITS,
+ (int)((uint64_t)bpm * mbs) >> BPER_MB_NORMBITS);
}
int vp9_rc_clamp_pframe_target_size(const VP9_COMP *const cpi, int target) {
@@ -276,7 +280,7 @@
}
rc->last_q[KEY_FRAME] = oxcf->best_allowed_q;
- rc->last_q[INTER_FRAME] = oxcf->best_allowed_q;
+ rc->last_q[INTER_FRAME] = oxcf->worst_allowed_q;
rc->buffer_level = rc->starting_buffer_level;
rc->bits_off_target = rc->starting_buffer_level;
@@ -298,7 +302,6 @@
rc->source_alt_ref_active = 0;
rc->frames_till_gf_update_due = 0;
-
rc->ni_av_qi = oxcf->worst_allowed_q;
rc->ni_tot_qi = 0;
rc->ni_frames = 0;
@@ -410,7 +413,7 @@
rate_correction_factor,
cm->bit_depth);
// Work out a size correction factor.
- if (projected_size_based_on_q > 0)
+ if (projected_size_based_on_q > FRAME_OVERHEAD_BITS)
correction_factor = (100 * cpi->rc.projected_frame_size) /
projected_size_based_on_q;
@@ -559,7 +562,7 @@
int adjustment = 0;
int active_worst_quality;
if (cm->frame_type == KEY_FRAME)
- return rc->worst_quality;
+ return rc->worst_quality * 4 / 5;
if (cm->current_video_frame > 1)
active_worst_quality = MIN(rc->worst_quality,
rc->avg_frame_qindex[INTER_FRAME] * 5 / 4);
@@ -988,6 +991,21 @@
}
}
+ // Extenstion to max or min Q if undershoot or overshoot is outside
+ // the permitted range.
+ if ((cpi->oxcf.rc_mode == VPX_VBR) &&
+ (cpi->twopass.gf_zeromotion_pct < VLOW_MOTION_THRESHOLD)) {
+ if (frame_is_intra_only(cm) ||
+ (!rc->is_src_frame_alt_ref &&
+ (cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame))) {
+ active_best_quality -= cpi->twopass.extend_minq;
+ active_worst_quality += (cpi->twopass.extend_maxq / 2);
+ } else {
+ active_best_quality -= cpi->twopass.extend_minq / 2;
+ active_worst_quality += cpi->twopass.extend_maxq;
+ }
+ }
+
#if LIMIT_QRANGE_FOR_ALTREF_AND_KEY
vp9_clear_system_state();
// Static forced key frames Q restrictions dealt with elsewhere.
@@ -1482,9 +1500,7 @@
rc->max_gf_interval = 16;
// Extended interval for genuinely static scenes
- rc->static_scene_max_gf_interval = oxcf->key_freq >> 1;
- if (rc->static_scene_max_gf_interval > (MAX_LAG_BUFFERS * 2))
- rc->static_scene_max_gf_interval = MAX_LAG_BUFFERS * 2;
+ rc->static_scene_max_gf_interval = MAX_LAG_BUFFERS * 2;
if (is_altref_enabled(cpi)) {
if (rc->static_scene_max_gf_interval > oxcf->lag_in_frames - 1)
diff --git a/source/libvpx/vp9/encoder/vp9_ratectrl.h b/source/libvpx/vp9/encoder/vp9_ratectrl.h
index edfb9fc..bc74129 100644
--- a/source/libvpx/vp9/encoder/vp9_ratectrl.h
+++ b/source/libvpx/vp9/encoder/vp9_ratectrl.h
@@ -87,6 +87,8 @@
int long_rolling_target_bits;
int long_rolling_actual_bits;
+ int rate_error_estimate;
+
int64_t total_actual_bits;
int64_t total_target_bits;
int64_t total_target_vs_actual;
diff --git a/source/libvpx/vp9/encoder/vp9_rd.c b/source/libvpx/vp9/encoder/vp9_rd.c
index 17369d4..7f526fc 100644
--- a/source/libvpx/vp9/encoder/vp9_rd.c
+++ b/source/libvpx/vp9/encoder/vp9_rd.c
@@ -44,6 +44,18 @@
// Factor to weigh the rate for switchable interp filters.
#define SWITCHABLE_INTERP_RATE_FACTOR 1
+void vp9_rd_cost_reset(RD_COST *rd_cost) {
+ rd_cost->rate = INT_MAX;
+ rd_cost->dist = INT64_MAX;
+ rd_cost->rdcost = INT64_MAX;
+}
+
+void vp9_rd_cost_init(RD_COST *rd_cost) {
+ rd_cost->rate = 0;
+ rd_cost->dist = 0;
+ rd_cost->rdcost = 0;
+}
+
// The baseline rd thresholds for breaking out of the rd loop for
// certain modes are assumed to be based on 8x8 blocks.
// This table is used to correct for block size.
@@ -136,9 +148,9 @@
};
int vp9_compute_rd_mult(const VP9_COMP *cpi, int qindex) {
- const int q = vp9_dc_quant(qindex, 0, cpi->common.bit_depth);
+ const int64_t q = vp9_dc_quant(qindex, 0, cpi->common.bit_depth);
#if CONFIG_VP9_HIGHBITDEPTH
- int rdmult = 0;
+ int64_t rdmult = 0;
switch (cpi->common.bit_depth) {
case VPX_BITS_8:
rdmult = 88 * q * q / 24;
@@ -154,8 +166,8 @@
return -1;
}
#else
- int rdmult = 88 * q * q / 24;
-#endif
+ int64_t rdmult = 88 * q * q / 24;
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (cpi->oxcf.pass == 2 && (cpi->common.frame_type != KEY_FRAME)) {
const GF_GROUP *const gf_group = &cpi->twopass.gf_group;
const FRAME_UPDATE_TYPE frame_type = gf_group->update_type[gf_group->index];
@@ -164,7 +176,7 @@
rdmult = (rdmult * rd_frame_type_factor[frame_type]) >> 7;
rdmult += ((rdmult * rd_boost_factor[boost_index]) >> 7);
}
- return rdmult;
+ return (int)rdmult;
}
static int compute_rd_thresh_factor(int qindex, vpx_bit_depth_t bit_depth) {
@@ -187,7 +199,7 @@
#else
(void) bit_depth;
q = vp9_dc_quant(qindex, 0, VPX_BITS_8) / 4.0;
-#endif
+#endif // CONFIG_VP9_HIGHBITDEPTH
// TODO(debargha): Adjust the function below.
return MAX((int)(pow(q, RD_THRESH_POW) * 5.12), 8);
}
@@ -213,7 +225,7 @@
#else
cpi->mb.sadperbit16 = sad_per_bit16lut_8[qindex];
cpi->mb.sadperbit4 = sad_per_bit4lut_8[qindex];
-#endif
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
static void set_block_thresholds(const VP9_COMMON *cm, RD_OPT *rd) {
@@ -598,3 +610,24 @@
if (sf->disable_split_mask & (1 << i))
rd->thresh_mult_sub8x8[i] = INT_MAX;
}
+
+int vp9_get_intra_cost_penalty(int qindex, int qdelta,
+ vpx_bit_depth_t bit_depth) {
+ const int q = vp9_dc_quant(qindex, qdelta, bit_depth);
+#if CONFIG_VP9_HIGHBITDEPTH
+ switch (bit_depth) {
+ case VPX_BITS_8:
+ return 20 * q;
+ case VPX_BITS_10:
+ return 5 * q;
+ case VPX_BITS_12:
+ return ROUND_POWER_OF_TWO(5 * q, 2);
+ default:
+ assert(0 && "bit_depth should be VPX_BITS_8, VPX_BITS_10 or VPX_BITS_12");
+ return -1;
+ }
+#else
+ return 20 * q;
+#endif // CONFIG_VP9_HIGHBITDEPTH
+}
+
diff --git a/source/libvpx/vp9/encoder/vp9_rd.h b/source/libvpx/vp9/encoder/vp9_rd.h
index 5dcb2f8..1aa5266 100644
--- a/source/libvpx/vp9/encoder/vp9_rd.h
+++ b/source/libvpx/vp9/encoder/vp9_rd.h
@@ -117,6 +117,17 @@
int RDDIV;
} RD_OPT;
+typedef struct RD_COST {
+ int rate;
+ int64_t dist;
+ int64_t rdcost;
+} RD_COST;
+
+// Reset the rate distortion cost values to maximum (invalid) value.
+void vp9_rd_cost_reset(RD_COST *rd_cost);
+// Initialize the rate distortion cost values to zero.
+void vp9_rd_cost_init(RD_COST *rd_cost);
+
struct TileInfo;
struct VP9_COMP;
struct macroblock;
@@ -162,6 +173,10 @@
int mi_row, int mi_col,
const struct scale_factors *scale,
const struct scale_factors *scale_uv);
+
+int vp9_get_intra_cost_penalty(int qindex, int qdelta,
+ vpx_bit_depth_t bit_depth);
+
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/source/libvpx/vp9/encoder/vp9_rdopt.c b/source/libvpx/vp9/encoder/vp9_rdopt.c
index f05351a..eca8e58 100644
--- a/source/libvpx/vp9/encoder/vp9_rdopt.c
+++ b/source/libvpx/vp9/encoder/vp9_rdopt.c
@@ -131,7 +131,7 @@
static int raster_block_offset(BLOCK_SIZE plane_bsize,
int raster_block, int stride) {
- const int bw = b_width_log2(plane_bsize);
+ const int bw = b_width_log2_lookup[plane_bsize];
const int y = 4 * (raster_block >> bw);
const int x = 4 * (raster_block & ((1 << bw) - 1));
return y * stride + x;
@@ -169,7 +169,8 @@
static void model_rd_for_sb(VP9_COMP *cpi, BLOCK_SIZE bsize,
MACROBLOCK *x, MACROBLOCKD *xd,
- int *out_rate_sum, int64_t *out_dist_sum) {
+ int *out_rate_sum, int64_t *out_dist_sum,
+ int *skip_txfm_sb, int64_t *skip_sse_sb) {
// Note our transform coeffs are 8 times an orthogonal transform.
// Hence quantizer step is also 8 times. To get effective quantizer
// we need to divide by 8 before sending to modeling function.
@@ -180,7 +181,9 @@
unsigned int sse;
unsigned int var = 0;
unsigned int sum_sse = 0;
- const int shift = 8;
+ int64_t total_sse = 0;
+ int skip_flag = 1;
+ const int shift = 6;
int rate;
int64_t dist;
@@ -192,6 +195,12 @@
const BLOCK_SIZE bs = get_plane_block_size(bsize, pd);
const TX_SIZE max_tx_size = max_txsize_lookup[bs];
const BLOCK_SIZE unit_size = txsize_to_bsize[max_tx_size];
+ const int64_t dc_thr = p->quant_thred[0] >> shift;
+ const int64_t ac_thr = p->quant_thred[1] >> shift;
+ // The low thresholds are used to measure if the prediction errors are
+ // low enough so that we can skip the mode search.
+ const int64_t low_dc_thr = MIN(50, dc_thr >> 2);
+ const int64_t low_ac_thr = MIN(80, ac_thr >> 2);
int bw = 1 << (b_width_log2_lookup[bs] - b_width_log2_lookup[unit_size]);
int bh = 1 << (b_height_log2_lookup[bs] - b_width_log2_lookup[unit_size]);
int idx, idy;
@@ -205,32 +214,49 @@
uint8_t *src = p->src.buf + (idy * p->src.stride << lh) + (idx << lw);
uint8_t *dst = pd->dst.buf + (idy * pd->dst.stride << lh) + (idx << lh);
int block_idx = (idy << 1) + idx;
+ int low_err_skip = 0;
var = cpi->fn_ptr[unit_size].vf(src, p->src.stride,
dst, pd->dst.stride, &sse);
x->bsse[(i << 2) + block_idx] = sse;
sum_sse += sse;
+ x->skip_txfm[(i << 2) + block_idx] = 0;
if (!x->select_tx_size) {
- if (x->bsse[(i << 2) + block_idx] < p->quant_thred[0] >> shift)
- x->skip_txfm[(i << 2) + block_idx] = 1;
- else if (var < p->quant_thred[1] >> shift)
+ // Check if all ac coefficients can be quantized to zero.
+ if (var < ac_thr || var == 0) {
x->skip_txfm[(i << 2) + block_idx] = 2;
- else
- x->skip_txfm[(i << 2) + block_idx] = 0;
+
+ // Check if dc coefficient can be quantized to zero.
+ if (sse - var < dc_thr || sse == var) {
+ x->skip_txfm[(i << 2) + block_idx] = 1;
+
+ if (!sse || (var < low_ac_thr && sse - var < low_dc_thr))
+ low_err_skip = 1;
+ }
+ }
}
+ if (skip_flag && !low_err_skip)
+ skip_flag = 0;
+
if (i == 0)
x->pred_sse[ref] += sse;
}
}
+ total_sse += sum_sse;
+
// Fast approximate the modelling function.
if (cpi->oxcf.speed > 4) {
int64_t rate;
- int64_t dist;
- int64_t square_error = sse;
+ const int64_t square_error = sum_sse;
int quantizer = (pd->dequant[1] >> 3);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ quantizer >>= (xd->bd - 8);
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (quantizer < 120)
rate = (square_error * (280 - quantizer)) >> 8;
@@ -240,13 +266,26 @@
rate_sum += rate;
dist_sum += dist;
} else {
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_model_rd_from_var_lapndz(sum_sse, 1 << num_pels_log2_lookup[bs],
+ pd->dequant[1] >> (xd->bd - 5),
+ &rate, &dist);
+ } else {
+ vp9_model_rd_from_var_lapndz(sum_sse, 1 << num_pels_log2_lookup[bs],
+ pd->dequant[1] >> 3, &rate, &dist);
+ }
+#else
vp9_model_rd_from_var_lapndz(sum_sse, 1 << num_pels_log2_lookup[bs],
pd->dequant[1] >> 3, &rate, &dist);
+#endif // CONFIG_VP9_HIGHBITDEPTH
rate_sum += rate;
dist_sum += dist;
}
}
+ *skip_txfm_sb = skip_flag;
+ *skip_sse_sb = total_sse << 4;
*out_rate_sum = (int)rate_sum;
*out_dist_sum = dist_sum << 4;
}
@@ -266,6 +305,31 @@
return error;
}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+int64_t vp9_highbd_block_error_c(const tran_low_t *coeff,
+ const tran_low_t *dqcoeff,
+ intptr_t block_size,
+ int64_t *ssz, int bd) {
+ int i;
+ int64_t error = 0, sqcoeff = 0;
+ int shift = 2 * (bd - 8);
+ int rounding = shift > 0 ? 1 << (shift - 1) : 0;
+
+ for (i = 0; i < block_size; i++) {
+ const int64_t diff = coeff[i] - dqcoeff[i];
+ error += diff * diff;
+ sqcoeff += (int64_t)coeff[i] * (int64_t)coeff[i];
+ }
+ assert(error >= 0 && sqcoeff >= 0);
+ error = (error + rounding) >> shift;
+ sqcoeff = (sqcoeff + rounding) >> shift;
+
+ *ssz = sqcoeff;
+ return error;
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
/* The trailing '0' is a terminator which is used inside cost_coeffs() to
* decide whether to include cost of a trailing EOB node or not (i.e. we
* can skip this if the last coefficient in this transform block, e.g. the
@@ -351,8 +415,14 @@
return cost;
}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+static void dist_block(int plane, int block, TX_SIZE tx_size,
+ struct rdcost_block_args* args, int bd) {
+#else
static void dist_block(int plane, int block, TX_SIZE tx_size,
struct rdcost_block_args* args) {
+#endif // CONFIG_VP9_HIGHBITDEPTH
const int ss_txfrm_size = tx_size << 1;
MACROBLOCK* const x = args->x;
MACROBLOCKD* const xd = &x->e_mbd;
@@ -362,14 +432,24 @@
int shift = tx_size == TX_32X32 ? 0 : 2;
tran_low_t *const coeff = BLOCK_OFFSET(p->coeff, block);
tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
+#if CONFIG_VP9_HIGHBITDEPTH
+ args->dist = vp9_highbd_block_error(coeff, dqcoeff, 16 << ss_txfrm_size,
+ &this_sse, bd) >> shift;
+#else
args->dist = vp9_block_error(coeff, dqcoeff, 16 << ss_txfrm_size,
&this_sse) >> shift;
+#endif // CONFIG_VP9_HIGHBITDEPTH
args->sse = this_sse >> shift;
if (x->skip_encode && !is_inter_block(&xd->mi[0].src_mi->mbmi)) {
// TODO(jingning): tune the model to better capture the distortion.
int64_t p = (pd->dequant[1] * pd->dequant[1] *
(1 << ss_txfrm_size)) >> (shift + 2);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ p >>= ((xd->bd - 8) * 2);
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
args->dist += (p >> 4);
args->sse += p;
}
@@ -399,12 +479,28 @@
if (!is_inter_block(mbmi)) {
vp9_encode_block_intra(x, plane, block, plane_bsize, tx_size, &mbmi->skip);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ dist_block(plane, block, tx_size, args, xd->bd);
+ } else {
+ dist_block(plane, block, tx_size, args, 8);
+ }
+#else
dist_block(plane, block, tx_size, args);
+#endif // CONFIG_VP9_HIGHBITDEPTH
} else if (max_txsize_lookup[plane_bsize] == tx_size) {
if (x->skip_txfm[(plane << 2) + (block >> (tx_size << 1))] == 0) {
// full forward transform and quantization
vp9_xform_quant(x, plane, block, plane_bsize, tx_size);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ dist_block(plane, block, tx_size, args, xd->bd);
+ } else {
+ dist_block(plane, block, tx_size, args, 8);
+ }
+#else
dist_block(plane, block, tx_size, args);
+#endif // CONFIG_VP9_HIGHBITDEPTH
} else if (x->skip_txfm[(plane << 2) + (block >> (tx_size << 1))] == 2) {
// compute DC coefficient
tran_low_t *const coeff = BLOCK_OFFSET(x->plane[plane].coeff, block);
@@ -412,9 +508,17 @@
vp9_xform_quant_dc(x, plane, block, plane_bsize, tx_size);
args->sse = x->bsse[(plane << 2) + (block >> (tx_size << 1))] << 4;
args->dist = args->sse;
- if (!x->plane[plane].eobs[block])
- args->dist = args->sse - ((coeff[0] * coeff[0] -
- (coeff[0] - dqcoeff[0]) * (coeff[0] - dqcoeff[0])) >> 2);
+ if (x->plane[plane].eobs[block]) {
+ int64_t dc_correct = coeff[0] * coeff[0] -
+ (coeff[0] - dqcoeff[0]) * (coeff[0] - dqcoeff[0]);
+#if CONFIG_VP9_HIGHBITDEPTH
+ dc_correct >>= ((xd->bd - 8) * 2);
+#endif
+ if (tx_size != TX_32X32)
+ dc_correct >>= 2;
+
+ args->dist = MAX(0, args->sse - dc_correct);
+ }
} else {
// skip forward transform
x->plane[plane].eobs[block] = 0;
@@ -424,7 +528,15 @@
} else {
// full forward transform and quantization
vp9_xform_quant(x, plane, block, plane_bsize, tx_size);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ dist_block(plane, block, tx_size, args, xd->bd);
+ } else {
+ dist_block(plane, block, tx_size, args, 8);
+ }
+#else
dist_block(plane, block, tx_size, args);
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
rate_block(plane, block, plane_bsize, tx_size, args);
@@ -659,6 +771,9 @@
const int num_4x4_blocks_high = num_4x4_blocks_high_lookup[bsize];
int idx, idy;
uint8_t best_dst[8 * 8];
+#if CONFIG_VP9_HIGHBITDEPTH
+ uint16_t best_dst16[8 * 8];
+#endif
assert(ib < 4);
@@ -666,6 +781,108 @@
vpx_memcpy(tl, l, sizeof(tl));
xd->mi[0].src_mi->mbmi.tx_size = TX_4X4;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ for (mode = DC_PRED; mode <= TM_PRED; ++mode) {
+ int64_t this_rd;
+ int ratey = 0;
+ int64_t distortion = 0;
+ int rate = bmode_costs[mode];
+
+ if (!(cpi->sf.intra_y_mode_mask[TX_4X4] & (1 << mode)))
+ continue;
+
+ // Only do the oblique modes if the best so far is
+ // one of the neighboring directional modes
+ if (cpi->sf.mode_search_skip_flags & FLAG_SKIP_INTRA_DIRMISMATCH) {
+ if (conditional_skipintra(mode, *best_mode))
+ continue;
+ }
+
+ vpx_memcpy(tempa, ta, sizeof(ta));
+ vpx_memcpy(templ, tl, sizeof(tl));
+
+ for (idy = 0; idy < num_4x4_blocks_high; ++idy) {
+ for (idx = 0; idx < num_4x4_blocks_wide; ++idx) {
+ const int block = ib + idy * 2 + idx;
+ const uint8_t *const src = &src_init[idx * 4 + idy * 4 * src_stride];
+ uint8_t *const dst = &dst_init[idx * 4 + idy * 4 * dst_stride];
+ int16_t *const src_diff = raster_block_offset_int16(BLOCK_8X8, block,
+ p->src_diff);
+ tran_low_t *const coeff = BLOCK_OFFSET(x->plane[0].coeff, block);
+ xd->mi[0].src_mi->bmi[block].as_mode = mode;
+ vp9_predict_intra_block(xd, block, 1,
+ TX_4X4, mode,
+ x->skip_encode ? src : dst,
+ x->skip_encode ? src_stride : dst_stride,
+ dst, dst_stride, idx, idy, 0);
+ vp9_highbd_subtract_block(4, 4, src_diff, 8, src, src_stride,
+ dst, dst_stride, xd->bd);
+ if (xd->lossless) {
+ const scan_order *so = &vp9_default_scan_orders[TX_4X4];
+ vp9_highbd_fwht4x4(src_diff, coeff, 8);
+ vp9_regular_quantize_b_4x4(x, 0, block, so->scan, so->iscan);
+ ratey += cost_coeffs(x, 0, block, tempa + idx, templ + idy, TX_4X4,
+ so->scan, so->neighbors,
+ cpi->sf.use_fast_coef_costing);
+ if (RDCOST(x->rdmult, x->rddiv, ratey, distortion) >= best_rd)
+ goto next_highbd;
+ vp9_highbd_iwht4x4_add(BLOCK_OFFSET(pd->dqcoeff, block),
+ dst, dst_stride,
+ p->eobs[block], xd->bd);
+ } else {
+ int64_t unused;
+ const TX_TYPE tx_type = get_tx_type_4x4(PLANE_TYPE_Y, xd, block);
+ const scan_order *so = &vp9_scan_orders[TX_4X4][tx_type];
+ vp9_highbd_fht4x4(src_diff, coeff, 8, tx_type);
+ vp9_regular_quantize_b_4x4(x, 0, block, so->scan, so->iscan);
+ ratey += cost_coeffs(x, 0, block, tempa + idx, templ + idy, TX_4X4,
+ so->scan, so->neighbors,
+ cpi->sf.use_fast_coef_costing);
+ distortion += vp9_highbd_block_error(
+ coeff, BLOCK_OFFSET(pd->dqcoeff, block),
+ 16, &unused, xd->bd) >> 2;
+ if (RDCOST(x->rdmult, x->rddiv, ratey, distortion) >= best_rd)
+ goto next_highbd;
+ vp9_highbd_iht4x4_add(tx_type, BLOCK_OFFSET(pd->dqcoeff, block),
+ dst, dst_stride, p->eobs[block], xd->bd);
+ }
+ }
+ }
+
+ rate += ratey;
+ this_rd = RDCOST(x->rdmult, x->rddiv, rate, distortion);
+
+ if (this_rd < best_rd) {
+ *bestrate = rate;
+ *bestratey = ratey;
+ *bestdistortion = distortion;
+ best_rd = this_rd;
+ *best_mode = mode;
+ vpx_memcpy(a, tempa, sizeof(tempa));
+ vpx_memcpy(l, templ, sizeof(templ));
+ for (idy = 0; idy < num_4x4_blocks_high * 4; ++idy) {
+ vpx_memcpy(best_dst16 + idy * 8,
+ CONVERT_TO_SHORTPTR(dst_init + idy * dst_stride),
+ num_4x4_blocks_wide * 4 * sizeof(uint16_t));
+ }
+ }
+ next_highbd:
+ {}
+ }
+ if (best_rd >= rd_thresh || x->skip_encode)
+ return best_rd;
+
+ for (idy = 0; idy < num_4x4_blocks_high * 4; ++idy) {
+ vpx_memcpy(CONVERT_TO_SHORTPTR(dst_init + idy * dst_stride),
+ best_dst16 + idy * 8,
+ num_4x4_blocks_wide * 4 * sizeof(uint16_t));
+ }
+
+ return best_rd;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
for (mode = DC_PRED; mode <= TM_PRED; ++mode) {
int64_t this_rd;
int ratey = 0;
@@ -827,6 +1044,7 @@
return RDCOST(mb->rdmult, mb->rddiv, cost, total_distortion);
}
+// This function is used only for intra_only frames
static int64_t rd_pick_intra_sby_mode(VP9_COMP *cpi, MACROBLOCK *x,
int *rate, int *rate_tokenonly,
int64_t *distortion, int *skippable,
@@ -841,24 +1059,21 @@
int64_t this_distortion, this_rd;
TX_SIZE best_tx = TX_4X4;
int i;
- int *bmode_costs = cpi->mbmode_cost;
+ int *bmode_costs;
+ const MODE_INFO *above_mi = xd->mi[-xd->mi_stride].src_mi;
+ const MODE_INFO *left_mi = xd->left_available ? xd->mi[-1].src_mi : NULL;
+ const PREDICTION_MODE A = vp9_above_block_mode(mic, above_mi, 0);
+ const PREDICTION_MODE L = vp9_left_block_mode(mic, left_mi, 0);
+ bmode_costs = cpi->y_mode_costs[A][L];
if (cpi->sf.tx_size_search_method == USE_FULL_RD)
for (i = 0; i < TX_MODES; i++)
tx_cache[i] = INT64_MAX;
+ vpx_memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
/* Y Search for intra prediction mode */
for (mode = DC_PRED; mode <= TM_PRED; mode++) {
int64_t local_tx_cache[TX_MODES];
- MODE_INFO *above_mi = xd->mi[-xd->mi_stride].src_mi;
- MODE_INFO *left_mi = xd->left_available ? xd->mi[-1].src_mi : NULL;
-
- if (cpi->common.frame_type == KEY_FRAME) {
- const PREDICTION_MODE A = vp9_above_block_mode(mic, above_mi, 0);
- const PREDICTION_MODE L = vp9_left_block_mode(mic, left_mi, 0);
-
- bmode_costs = cpi->y_mode_costs[A][L];
- }
mic->mbmi.mode = mode;
super_block_yrd(cpi, x, &this_rate_tokenonly, &this_distortion,
@@ -897,21 +1112,24 @@
return best_rd;
}
-static void super_block_uvrd(const VP9_COMP *cpi, MACROBLOCK *x,
- int *rate, int64_t *distortion, int *skippable,
- int64_t *sse, BLOCK_SIZE bsize,
- int64_t ref_best_rd) {
+// Return value 0: early termination triggered, no valid rd cost available;
+// 1: rd cost values are valid.
+static int super_block_uvrd(const VP9_COMP *cpi, MACROBLOCK *x,
+ int *rate, int64_t *distortion, int *skippable,
+ int64_t *sse, BLOCK_SIZE bsize,
+ int64_t ref_best_rd) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mi[0].src_mi->mbmi;
const TX_SIZE uv_tx_size = get_uv_tx_size(mbmi, &xd->plane[1]);
int plane;
int pnrate = 0, pnskip = 1;
int64_t pndist = 0, pnsse = 0;
+ int is_cost_valid = 1;
if (ref_best_rd < 0)
- goto term;
+ is_cost_valid = 0;
- if (is_inter_block(mbmi)) {
+ if (is_inter_block(mbmi) && is_cost_valid) {
int plane;
for (plane = 1; plane < MAX_MB_PLANE; ++plane)
vp9_subtract_plane(x, bsize, plane);
@@ -926,21 +1144,25 @@
txfm_rd_in_plane(x, &pnrate, &pndist, &pnskip, &pnsse,
ref_best_rd, plane, bsize, uv_tx_size,
cpi->sf.use_fast_coef_costing);
- if (pnrate == INT_MAX)
- goto term;
+ if (pnrate == INT_MAX) {
+ is_cost_valid = 0;
+ break;
+ }
*rate += pnrate;
*distortion += pndist;
*sse += pnsse;
*skippable &= pnskip;
}
- return;
- term:
- *rate = INT_MAX;
- *distortion = INT64_MAX;
- *sse = INT64_MAX;
- *skippable = 0;
- return;
+ if (!is_cost_valid) {
+ // reset cost value
+ *rate = INT_MAX;
+ *distortion = INT64_MAX;
+ *sse = INT64_MAX;
+ *skippable = 0;
+ }
+
+ return is_cost_valid;
}
static int64_t rd_pick_intra_sbuv_mode(VP9_COMP *cpi, MACROBLOCK *x,
@@ -955,15 +1177,15 @@
int this_rate_tokenonly, this_rate, s;
int64_t this_distortion, this_sse;
+ vpx_memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
for (mode = DC_PRED; mode <= TM_PRED; ++mode) {
if (!(cpi->sf.intra_uv_mode_mask[max_tx_size] & (1 << mode)))
continue;
xd->mi[0].src_mi->mbmi.uv_mode = mode;
- super_block_uvrd(cpi, x, &this_rate_tokenonly,
- &this_distortion, &s, &this_sse, bsize, best_rd);
- if (this_rate_tokenonly == INT_MAX)
+ if (!super_block_uvrd(cpi, x, &this_rate_tokenonly,
+ &this_distortion, &s, &this_sse, bsize, best_rd))
continue;
this_rate = this_rate_tokenonly +
cpi->intra_uv_mode_cost[cpi->common.frame_type][mode];
@@ -993,6 +1215,7 @@
int64_t unused;
x->e_mbd.mi[0].src_mi->mbmi.uv_mode = DC_PRED;
+ vpx_memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
super_block_uvrd(cpi, x, rate_tokenonly, distortion,
skippable, &unused, bsize, INT64_MAX);
*rate = *rate_tokenonly + cpi->intra_uv_mode_cost[cm->frame_type][DC_PRED];
@@ -1121,6 +1344,16 @@
for (ref = 0; ref < 1 + is_compound; ++ref) {
const uint8_t *pre = &pd->pre[ref].buf[raster_block_offset(BLOCK_8X8, i,
pd->pre[ref].stride)];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_build_inter_predictor(pre, pd->pre[ref].stride,
+ dst, pd->dst.stride,
+ &mi->bmi[i].as_mv[ref].as_mv,
+ &xd->block_refs[ref]->sf, width, height,
+ ref, kernel, MV_PRECISION_Q3,
+ mi_col * MI_SIZE + 4 * (i % 2),
+ mi_row * MI_SIZE + 4 * (i / 2), xd->bd);
+ } else {
vp9_build_inter_predictor(pre, pd->pre[ref].stride,
dst, pd->dst.stride,
&mi->bmi[i].as_mv[ref].as_mv,
@@ -1129,11 +1362,32 @@
mi_col * MI_SIZE + 4 * (i % 2),
mi_row * MI_SIZE + 4 * (i / 2));
}
+#else
+ vp9_build_inter_predictor(pre, pd->pre[ref].stride,
+ dst, pd->dst.stride,
+ &mi->bmi[i].as_mv[ref].as_mv,
+ &xd->block_refs[ref]->sf, width, height, ref,
+ kernel, MV_PRECISION_Q3,
+ mi_col * MI_SIZE + 4 * (i % 2),
+ mi_row * MI_SIZE + 4 * (i / 2));
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ }
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_subtract_block(
+ height, width, raster_block_offset_int16(BLOCK_8X8, i, p->src_diff), 8,
+ src, p->src.stride, dst, pd->dst.stride, xd->bd);
+ } else {
+ vp9_subtract_block(
+ height, width, raster_block_offset_int16(BLOCK_8X8, i, p->src_diff), 8,
+ src, p->src.stride, dst, pd->dst.stride);
+ }
+#else
vp9_subtract_block(height, width,
raster_block_offset_int16(BLOCK_8X8, i, p->src_diff), 8,
- src, p->src.stride,
- dst, pd->dst.stride);
+ src, p->src.stride, dst, pd->dst.stride);
+#endif // CONFIG_VP9_HIGHBITDEPTH
k = i;
for (idy = 0; idy < height / 4; ++idy) {
@@ -1146,8 +1400,19 @@
x->fwd_txm4x4(raster_block_offset_int16(BLOCK_8X8, k, p->src_diff),
coeff, 8);
vp9_regular_quantize_b_4x4(x, 0, k, so->scan, so->iscan);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ thisdistortion += vp9_highbd_block_error(coeff,
+ BLOCK_OFFSET(pd->dqcoeff, k),
+ 16, &ssz, xd->bd);
+ } else {
+ thisdistortion += vp9_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, k),
+ 16, &ssz);
+ }
+#else
thisdistortion += vp9_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, k),
16, &ssz);
+#endif // CONFIG_VP9_HIGHBITDEPTH
thissse += ssz;
thisrate += cost_coeffs(x, 0, k, ta + (k & 1), tl + (k >> 1), TX_4X4,
so->scan, so->neighbors,
@@ -1369,7 +1634,7 @@
int sadpb = x->sadperbit4;
MV mvp_full;
int max_mv;
- int sad_list[5];
+ int cost_list[5];
/* Is the best so far sufficiently good that we cant justify doing
* and new motion search. */
@@ -1415,7 +1680,7 @@
bestsme = vp9_full_pixel_search(
cpi, x, bsize, &mvp_full, step_param, sadpb,
- cpi->sf.mv.subpel_search_method != SUBPEL_TREE ? sad_list : NULL,
+ cpi->sf.mv.subpel_search_method != SUBPEL_TREE ? cost_list : NULL,
&bsi->ref_mv[0]->as_mv, new_mv,
INT_MAX, 1);
@@ -1429,7 +1694,7 @@
sadpb, 16, &cpi->fn_ptr[bsize],
&bsi->ref_mv[0]->as_mv,
&best_mv->as_mv);
- sad_list[1] = sad_list[2] = sad_list[3] = sad_list[4] = INT_MAX;
+ cost_list[1] = cost_list[2] = cost_list[3] = cost_list[4] = INT_MAX;
if (thissme < bestsme) {
bestsme = thissme;
*new_mv = best_mv->as_mv;
@@ -1450,7 +1715,7 @@
x->errorperbit, &cpi->fn_ptr[bsize],
cpi->sf.mv.subpel_force_stop,
cpi->sf.mv.subpel_iters_per_step,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
x->nmvjointcost, x->mvcost,
&distortion,
&x->pred_sse[mbmi->ref_frame[0]],
@@ -1784,7 +2049,7 @@
int tmp_col_max = x->mv_col_max;
int tmp_row_min = x->mv_row_min;
int tmp_row_max = x->mv_row_max;
- int sad_list[5];
+ int cost_list[5];
const YV12_BUFFER_CONFIG *scaled_ref_frame = vp9_get_scaled_ref_frame(cpi,
ref);
@@ -1820,14 +2085,14 @@
}
if (cpi->sf.adaptive_motion_search && bsize < BLOCK_64X64) {
- int boffset = 2 * (b_width_log2(BLOCK_64X64) - MIN(b_height_log2(bsize),
- b_width_log2(bsize)));
+ int boffset = 2 * (b_width_log2_lookup[BLOCK_64X64] -
+ MIN(b_height_log2_lookup[bsize], b_width_log2_lookup[bsize]));
step_param = MAX(step_param, boffset);
}
if (cpi->sf.adaptive_motion_search) {
- int bwl = b_width_log2(bsize);
- int bhl = b_height_log2(bsize);
+ int bwl = b_width_log2_lookup[bsize];
+ int bhl = b_height_log2_lookup[bsize];
int i;
int tlevel = x->pred_mv_sad[ref] >> (bwl + bhl + 4);
@@ -1856,7 +2121,7 @@
mvp_full.row >>= 3;
bestsme = vp9_full_pixel_search(cpi, x, bsize, &mvp_full, step_param, sadpb,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
&ref_mv, &tmp_mv->as_mv, INT_MAX, 1);
x->mv_col_min = tmp_col_min;
@@ -1872,7 +2137,7 @@
&cpi->fn_ptr[bsize],
cpi->sf.mv.subpel_force_stop,
cpi->sf.mv.subpel_iters_per_step,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
x->nmvjointcost, x->mvcost,
&dis, &x->pred_sse[ref], NULL, 0, 0);
}
@@ -1904,7 +2169,12 @@
int_mv ref_mv[2];
int ite, ref;
// Prediction buffer from second frame.
+#if CONFIG_VP9_HIGHBITDEPTH
+ uint8_t *second_pred;
+ uint8_t *second_pred_alloc;
+#else
uint8_t *second_pred = vpx_memalign(16, pw * ph * sizeof(uint8_t));
+#endif // CONFIG_VP9_HIGHBITDEPTH
const InterpKernel *kernel = vp9_get_interp_kernel(mbmi->interp_filter);
// Do joint motion search in compound mode to get more accurate mv.
@@ -1915,6 +2185,15 @@
vp9_get_scaled_ref_frame(cpi, mbmi->ref_frame[0]),
vp9_get_scaled_ref_frame(cpi, mbmi->ref_frame[1])
};
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ second_pred_alloc = vpx_memalign(16, pw * ph * sizeof(uint16_t));
+ second_pred = CONVERT_TO_BYTEPTR(second_pred_alloc);
+ } else {
+ second_pred_alloc = vpx_memalign(16, pw * ph * sizeof(uint8_t));
+ second_pred = second_pred_alloc;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
for (ref = 0; ref < 2; ++ref) {
ref_mv[ref] = mbmi->ref_mvs[refs[ref]][0];
@@ -1953,6 +2232,28 @@
ref_yv12[1] = xd->plane[0].pre[1];
// Get pred block from second frame.
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_build_inter_predictor(ref_yv12[!id].buf,
+ ref_yv12[!id].stride,
+ second_pred, pw,
+ &frame_mv[refs[!id]].as_mv,
+ &xd->block_refs[!id]->sf,
+ pw, ph, 0,
+ kernel, MV_PRECISION_Q3,
+ mi_col * MI_SIZE, mi_row * MI_SIZE,
+ xd->bd);
+ } else {
+ vp9_build_inter_predictor(ref_yv12[!id].buf,
+ ref_yv12[!id].stride,
+ second_pred, pw,
+ &frame_mv[refs[!id]].as_mv,
+ &xd->block_refs[!id]->sf,
+ pw, ph, 0,
+ kernel, MV_PRECISION_Q3,
+ mi_col * MI_SIZE, mi_row * MI_SIZE);
+ }
+#else
vp9_build_inter_predictor(ref_yv12[!id].buf,
ref_yv12[!id].stride,
second_pred, pw,
@@ -1961,6 +2262,7 @@
pw, ph, 0,
kernel, MV_PRECISION_Q3,
mi_col * MI_SIZE, mi_row * MI_SIZE);
+#endif // CONFIG_VP9_HIGHBITDEPTH
// Compound motion search on first ref frame.
if (id)
@@ -2029,7 +2331,11 @@
x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+ vpx_free(second_pred_alloc);
+#else
vpx_free(second_pred);
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
static INLINE void restore_dst_buf(MACROBLOCKD *xd,
@@ -2042,93 +2348,12 @@
}
}
-static void rd_encode_breakout_test(VP9_COMP *cpi, MACROBLOCK *x,
- BLOCK_SIZE bsize, int *rate2,
- int64_t *distortion, int64_t *distortion_uv,
- int *disable_skip) {
- VP9_COMMON *cm = &cpi->common;
- MACROBLOCKD *xd = &x->e_mbd;
- const BLOCK_SIZE y_size = get_plane_block_size(bsize, &xd->plane[0]);
- const BLOCK_SIZE uv_size = get_plane_block_size(bsize, &xd->plane[1]);
- unsigned int var, sse;
- // Skipping threshold for ac.
- unsigned int thresh_ac;
- // Skipping threshold for dc
- unsigned int thresh_dc;
-
- var = cpi->fn_ptr[y_size].vf(x->plane[0].src.buf, x->plane[0].src.stride,
- xd->plane[0].dst.buf,
- xd->plane[0].dst.stride, &sse);
-
- if (x->encode_breakout > 0) {
- // Set a maximum for threshold to avoid big PSNR loss in low bitrate
- // case. Use extreme low threshold for static frames to limit skipping.
- const unsigned int max_thresh = (cpi->allow_encode_breakout ==
- ENCODE_BREAKOUT_LIMITED) ? 128 : 36000;
- // The encode_breakout input
- const unsigned int min_thresh =
- MIN(((unsigned int)x->encode_breakout << 4), max_thresh);
-
- // Calculate threshold according to dequant value.
- thresh_ac = (xd->plane[0].dequant[1] * xd->plane[0].dequant[1]) / 9;
- thresh_ac = clamp(thresh_ac, min_thresh, max_thresh);
-
- // Adjust threshold according to partition size.
- thresh_ac >>= 8 - (b_width_log2(bsize) +
- b_height_log2(bsize));
- thresh_dc = (xd->plane[0].dequant[0] * xd->plane[0].dequant[0] >> 6);
- } else {
- thresh_ac = 0;
- thresh_dc = 0;
- }
-
- // Y skipping condition checking
- if (sse < thresh_ac || sse == 0) {
- // dc skipping checking
- if ((sse - var) < thresh_dc || sse == var) {
- unsigned int sse_u, sse_v;
- unsigned int var_u, var_v;
-
- var_u = cpi->fn_ptr[uv_size].vf(x->plane[1].src.buf,
- x->plane[1].src.stride,
- xd->plane[1].dst.buf,
- xd->plane[1].dst.stride, &sse_u);
-
- // U skipping condition checking
- if ((sse_u * 4 < thresh_ac || sse_u == 0) &&
- (sse_u - var_u < thresh_dc || sse_u == var_u)) {
- var_v = cpi->fn_ptr[uv_size].vf(x->plane[2].src.buf,
- x->plane[2].src.stride,
- xd->plane[2].dst.buf,
- xd->plane[2].dst.stride, &sse_v);
-
- // V skipping condition checking
- if ((sse_v * 4 < thresh_ac || sse_v == 0) &&
- (sse_v - var_v < thresh_dc || sse_v == var_v)) {
- x->skip = 1;
-
- // The cost of skip bit needs to be added.
- *rate2 += vp9_cost_bit(vp9_get_skip_prob(cm, xd), 1);
-
- // Scaling factor for SSE from spatial domain to frequency domain
- // is 16. Adjust distortion accordingly.
- *distortion_uv = (sse_u + sse_v) << 4;
- *distortion = (sse << 4) + *distortion_uv;
-
- *disable_skip = 1;
- }
- }
- }
- }
-}
-
static int64_t handle_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
BLOCK_SIZE bsize,
int64_t txfm_cache[],
int *rate2, int64_t *distortion,
int *skippable,
- int *rate_y, int64_t *distortion_y,
- int *rate_uv, int64_t *distortion_uv,
+ int *rate_y, int *rate_uv,
int *disable_skip,
int_mv (*mode_mv)[MAX_REF_FRAMES],
int mi_row, int mi_col,
@@ -2148,8 +2373,13 @@
int refs[2] = { mbmi->ref_frame[0],
(mbmi->ref_frame[1] < 0 ? 0 : mbmi->ref_frame[1]) };
int_mv cur_mv[2];
- int64_t this_rd = 0;
+#if CONFIG_VP9_HIGHBITDEPTH
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, tmp_buf16, MAX_MB_PLANE * 64 * 64);
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, tmp_buf8, MAX_MB_PLANE * 64 * 64);
+ uint8_t *tmp_buf;
+#else
DECLARE_ALIGNED_ARRAY(16, uint8_t, tmp_buf, MAX_MB_PLANE * 64 * 64);
+#endif // CONFIG_VP9_HIGHBITDEPTH
int pred_exists = 0;
int intpel_mv;
int64_t rd, tmp_rd, best_rd = INT64_MAX;
@@ -2166,6 +2396,18 @@
(((mi_row + mi_col) >> bsl) +
get_chessboard_index(cm->current_video_frame)) & 0x1 : 0;
+ int skip_txfm_sb = 0;
+ int64_t skip_sse_sb = INT64_MAX;
+ int64_t distortion_y = 0, distortion_uv = 0;
+
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ tmp_buf = CONVERT_TO_BYTEPTR(tmp_buf16);
+ } else {
+ tmp_buf = tmp_buf8;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
if (pred_filter_search) {
INTERP_FILTER af = SWITCHABLE, lf = SWITCHABLE;
if (xd->up_available)
@@ -2275,6 +2517,9 @@
for (i = 0; i < SWITCHABLE_FILTERS; ++i) {
int j;
int64_t rs_rd;
+ int tmp_skip_sb = 0;
+ int64_t tmp_skip_sse = INT64_MAX;
+
mbmi->interp_filter = i;
rs = vp9_get_switchable_rate(cpi);
rs_rd = RDCOST(x->rdmult, x->rddiv, rs, 0);
@@ -2310,7 +2555,8 @@
}
}
vp9_build_inter_predictors_sb(xd, mi_row, mi_col, bsize);
- model_rd_for_sb(cpi, bsize, x, xd, &rate_sum, &dist_sum);
+ model_rd_for_sb(cpi, bsize, x, xd, &rate_sum, &dist_sum,
+ &tmp_skip_sb, &tmp_skip_sse);
rd = RDCOST(x->rdmult, x->rddiv, rate_sum, dist_sum);
rd_opt->filter_cache[i] = rd;
@@ -2339,8 +2585,6 @@
best_filter = mbmi->interp_filter;
if (cm->interp_filter == SWITCHABLE && i && !intpel_mv)
best_needs_copy = !best_needs_copy;
- vpx_memcpy(skip_txfm, x->skip_txfm, sizeof(skip_txfm));
- vpx_memcpy(bsse, x->bsse, sizeof(bsse));
}
if ((cm->interp_filter == SWITCHABLE && newbest) ||
@@ -2348,6 +2592,11 @@
cm->interp_filter == mbmi->interp_filter)) {
pred_exists = 1;
tmp_rd = best_rd;
+
+ skip_txfm_sb = tmp_skip_sb;
+ skip_sse_sb = tmp_skip_sse;
+ vpx_memcpy(skip_txfm, x->skip_txfm, sizeof(skip_txfm));
+ vpx_memcpy(bsse, x->bsse, sizeof(bsse));
}
}
restore_dst_buf(xd, orig_dst, orig_dst_stride);
@@ -2374,7 +2623,8 @@
// switchable list (ex. bilinear) is indicated at the frame level, or
// skip condition holds.
vp9_build_inter_predictors_sb(xd, mi_row, mi_col, bsize);
- model_rd_for_sb(cpi, bsize, x, xd, &tmp_rate, &tmp_dist);
+ model_rd_for_sb(cpi, bsize, x, xd, &tmp_rate, &tmp_dist,
+ &skip_txfm_sb, &skip_sse_sb);
rd = RDCOST(x->rdmult, x->rddiv, rs + tmp_rate, tmp_dist);
vpx_memcpy(skip_txfm, x->skip_txfm, sizeof(skip_txfm));
vpx_memcpy(bsse, x->bsse, sizeof(bsse));
@@ -2401,23 +2651,17 @@
if (cm->interp_filter == SWITCHABLE)
*rate2 += rs;
- if (!is_comp_pred) {
- if (cpi->allow_encode_breakout)
- rd_encode_breakout_test(cpi, x, bsize, rate2, distortion, distortion_uv,
- disable_skip);
- }
-
vpx_memcpy(x->skip_txfm, skip_txfm, sizeof(skip_txfm));
vpx_memcpy(x->bsse, bsse, sizeof(bsse));
- if (!x->skip) {
+ if (!skip_txfm_sb) {
int skippable_y, skippable_uv;
int64_t sseuv = INT64_MAX;
int64_t rdcosty = INT64_MAX;
// Y cost and distortion
vp9_subtract_plane(x, bsize, 0);
- super_block_yrd(cpi, x, rate_y, distortion_y, &skippable_y, psse,
+ super_block_yrd(cpi, x, rate_y, &distortion_y, &skippable_y, psse,
bsize, txfm_cache, ref_best_rd);
if (*rate_y == INT_MAX) {
@@ -2428,14 +2672,13 @@
}
*rate2 += *rate_y;
- *distortion += *distortion_y;
+ *distortion += distortion_y;
rdcosty = RDCOST(x->rdmult, x->rddiv, *rate2, *distortion);
rdcosty = MIN(rdcosty, RDCOST(x->rdmult, x->rddiv, 0, *psse));
- super_block_uvrd(cpi, x, rate_uv, distortion_uv, &skippable_uv, &sseuv,
- bsize, ref_best_rd - rdcosty);
- if (*rate_uv == INT_MAX) {
+ if (!super_block_uvrd(cpi, x, rate_uv, &distortion_uv, &skippable_uv,
+ &sseuv, bsize, ref_best_rd - rdcosty)) {
*rate2 = INT_MAX;
*distortion = INT64_MAX;
restore_dst_buf(xd, orig_dst, orig_dst_stride);
@@ -2444,20 +2687,27 @@
*psse += sseuv;
*rate2 += *rate_uv;
- *distortion += *distortion_uv;
+ *distortion += distortion_uv;
*skippable = skippable_y && skippable_uv;
+ } else {
+ x->skip = 1;
+ *disable_skip = 1;
+
+ // The cost of skip bit needs to be added.
+ *rate2 += vp9_cost_bit(vp9_get_skip_prob(cm, xd), 1);
+
+ *distortion = skip_sse_sb;
}
if (!is_comp_pred)
single_skippable[this_mode][refs[0]] = *skippable;
restore_dst_buf(xd, orig_dst, orig_dst_stride);
- return this_rd; // if 0, this will be re-calculated by caller
+ return 0; // The rate-distortion cost will be re-calculated by caller.
}
void vp9_rd_pick_intra_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
- int *returnrate, int64_t *returndist,
- BLOCK_SIZE bsize,
+ RD_COST *rd_cost, BLOCK_SIZE bsize,
PICK_MODE_CONTEXT *ctx, int64_t best_rd) {
VP9_COMMON *const cm = &cpi->common;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -2474,37 +2724,34 @@
if (rd_pick_intra_sby_mode(cpi, x, &rate_y, &rate_y_tokenonly,
&dist_y, &y_skip, bsize, tx_cache,
best_rd) >= best_rd) {
- *returnrate = INT_MAX;
+ rd_cost->rate = INT_MAX;
return;
}
- max_uv_tx_size = get_uv_tx_size_impl(xd->mi[0].src_mi->mbmi.tx_size, bsize,
- pd[1].subsampling_x,
- pd[1].subsampling_y);
- rd_pick_intra_sbuv_mode(cpi, x, ctx, &rate_uv, &rate_uv_tokenonly,
- &dist_uv, &uv_skip, bsize, max_uv_tx_size);
} else {
y_skip = 0;
if (rd_pick_intra_sub_8x8_y_mode(cpi, x, &rate_y, &rate_y_tokenonly,
&dist_y, best_rd) >= best_rd) {
- *returnrate = INT_MAX;
+ rd_cost->rate = INT_MAX;
return;
}
- max_uv_tx_size = get_uv_tx_size_impl(xd->mi[0].src_mi->mbmi.tx_size, bsize,
- pd[1].subsampling_x,
- pd[1].subsampling_y);
- rd_pick_intra_sbuv_mode(cpi, x, ctx, &rate_uv, &rate_uv_tokenonly,
- &dist_uv, &uv_skip, BLOCK_8X8, max_uv_tx_size);
}
+ max_uv_tx_size = get_uv_tx_size_impl(xd->mi[0].src_mi->mbmi.tx_size, bsize,
+ pd[1].subsampling_x,
+ pd[1].subsampling_y);
+ rd_pick_intra_sbuv_mode(cpi, x, ctx, &rate_uv, &rate_uv_tokenonly,
+ &dist_uv, &uv_skip, MAX(BLOCK_8X8, bsize),
+ max_uv_tx_size);
if (y_skip && uv_skip) {
- *returnrate = rate_y + rate_uv - rate_y_tokenonly - rate_uv_tokenonly +
- vp9_cost_bit(vp9_get_skip_prob(cm, xd), 1);
- *returndist = dist_y + dist_uv;
+ rd_cost->rate = rate_y + rate_uv - rate_y_tokenonly - rate_uv_tokenonly +
+ vp9_cost_bit(vp9_get_skip_prob(cm, xd), 1);
+ rd_cost->dist = dist_y + dist_uv;
vp9_zero(ctx->tx_rd_diff);
} else {
int i;
- *returnrate = rate_y + rate_uv + vp9_cost_bit(vp9_get_skip_prob(cm, xd), 0);
- *returndist = dist_y + dist_uv;
+ rd_cost->rate = rate_y + rate_uv +
+ vp9_cost_bit(vp9_get_skip_prob(cm, xd), 0);
+ rd_cost->dist = dist_y + dist_uv;
if (cpi->sf.tx_size_search_method == USE_FULL_RD)
for (i = 0; i < TX_MODES; i++) {
if (tx_cache[i] < INT64_MAX && tx_cache[cm->tx_mode] < INT64_MAX)
@@ -2515,45 +2762,43 @@
}
ctx->mic = *xd->mi[0].src_mi;
+ rd_cost->rdcost = RDCOST(x->rdmult, x->rddiv, rd_cost->rate, rd_cost->dist);
}
-// Updating rd_thresh_freq_fact[] here means that the different
-// partition/block sizes are handled independently based on the best
-// choice for the current partition. It may well be better to keep a scaled
-// best rd so far value and update rd_thresh_freq_fact based on the mode/size
-// combination that wins out.
static void update_rd_thresh_fact(VP9_COMP *cpi, int bsize,
int best_mode_index) {
if (cpi->sf.adaptive_rd_thresh > 0) {
const int top_mode = bsize < BLOCK_8X8 ? MAX_REFS : MAX_MODES;
int mode;
for (mode = 0; mode < top_mode; ++mode) {
- int *const fact = &cpi->rd.thresh_freq_fact[bsize][mode];
-
- if (mode == best_mode_index) {
- *fact -= (*fact >> 3);
- } else {
- *fact = MIN(*fact + RD_THRESH_INC,
- cpi->sf.adaptive_rd_thresh * RD_THRESH_MAX_FACT);
+ const BLOCK_SIZE min_size = MAX(bsize - 1, BLOCK_4X4);
+ const BLOCK_SIZE max_size = MIN(bsize + 2, BLOCK_64X64);
+ BLOCK_SIZE bs;
+ for (bs = min_size; bs <= max_size; ++bs) {
+ int *const fact = &cpi->rd.thresh_freq_fact[bs][mode];
+ if (mode == best_mode_index) {
+ *fact -= (*fact >> 4);
+ } else {
+ *fact = MIN(*fact + RD_THRESH_INC,
+ cpi->sf.adaptive_rd_thresh * RD_THRESH_MAX_FACT);
+ }
}
}
}
}
-int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
- const TileInfo *const tile,
- int mi_row, int mi_col,
- int *returnrate,
- int64_t *returndistortion,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far) {
+void vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
+ const TileInfo *const tile,
+ int mi_row, int mi_col,
+ RD_COST *rd_cost, BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far) {
VP9_COMMON *const cm = &cpi->common;
RD_OPT *const rd_opt = &cpi->rd;
+ SPEED_FEATURES *const sf = &cpi->sf;
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mi[0].src_mi->mbmi;
const struct segmentation *const seg = &cm->seg;
- struct macroblockd_plane *const pd = xd->plane;
PREDICTION_MODE this_mode;
MV_REFERENCE_FRAME ref_frame, second_ref_frame;
unsigned char segment_id = mbmi->segment_id;
@@ -2584,20 +2829,20 @@
int64_t dist_uv[TX_SIZES];
int skip_uv[TX_SIZES];
PREDICTION_MODE mode_uv[TX_SIZES];
- const int intra_cost_penalty =
- 20 * vp9_dc_quant(cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth);
+ const int intra_cost_penalty = vp9_get_intra_cost_penalty(
+ cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth);
int best_skip2 = 0;
uint8_t ref_frame_skip_mask[2] = { 0 };
uint16_t mode_skip_mask[MAX_REF_FRAMES] = { 0 };
- int mode_skip_start = cpi->sf.mode_skip_start + 1;
+ int mode_skip_start = sf->mode_skip_start + 1;
const int *const rd_threshes = rd_opt->threshes[segment_id][bsize];
const int *const rd_thresh_freq_fact = rd_opt->thresh_freq_fact[bsize];
- int mode_threshold[MAX_MODES];
+ int64_t mode_threshold[MAX_MODES];
int *mode_map = rd_opt->mode_map[bsize];
- const int mode_search_skip_flags = cpi->sf.mode_search_skip_flags;
+ const int mode_search_skip_flags = sf->mode_search_skip_flags;
vp9_zero(best_mbmode);
- x->skip_encode = cpi->sf.skip_encode_frame && x->q_index < QIDX_SKIP_THRESH;
+ x->skip_encode = sf->skip_encode_frame && x->q_index < QIDX_SKIP_THRESH;
estimate_ref_frame_costs(cm, xd, segment_id, ref_costs_single, ref_costs_comp,
&comp_mode_p);
@@ -2619,7 +2864,7 @@
}
}
- *returnrate = INT_MAX;
+ rd_cost->rate = INT_MAX;
for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ++ref_frame) {
x->pred_mv_sad[ref_frame] = INT_MAX;
@@ -2638,7 +2883,7 @@
// are masked out.
ref_frame_skip_mask[0] |= (1 << ref_frame);
ref_frame_skip_mask[1] |= SECOND_REF_FRAME_MASK;
- } else if (cpi->sf.reference_masking) {
+ } else if (sf->reference_masking) {
for (i = LAST_FRAME; i <= ALTREF_FRAME; ++i) {
// Skip fixed mv modes for poor references
if ((x->pred_mv_sad[ref_frame] >> 2) > x->pred_mv_sad[i]) {
@@ -2676,25 +2921,37 @@
}
if (cpi->rc.is_src_frame_alt_ref) {
- if (cpi->sf.alt_ref_search_fp) {
+ if (sf->alt_ref_search_fp) {
mode_skip_mask[ALTREF_FRAME] = 0;
ref_frame_skip_mask[0] = ~(1 << ALTREF_FRAME);
ref_frame_skip_mask[1] = SECOND_REF_FRAME_MASK;
}
}
- if (bsize > cpi->sf.max_intra_bsize) {
+ if (sf->alt_ref_search_fp)
+ if (!cm->show_frame && x->pred_mv_sad[GOLDEN_FRAME] < INT_MAX)
+ if (x->pred_mv_sad[ALTREF_FRAME] > (x->pred_mv_sad[GOLDEN_FRAME] << 1))
+ mode_skip_mask[ALTREF_FRAME] |= INTER_ALL;
+
+ if (sf->adaptive_mode_search) {
+ if (cm->show_frame && !cpi->rc.is_src_frame_alt_ref &&
+ cpi->rc.frames_since_golden >= 3)
+ if (x->pred_mv_sad[GOLDEN_FRAME] > (x->pred_mv_sad[LAST_FRAME] << 1))
+ mode_skip_mask[GOLDEN_FRAME] |= INTER_ALL;
+ }
+
+ if (bsize > sf->max_intra_bsize) {
ref_frame_skip_mask[0] |= (1 << INTRA_FRAME);
ref_frame_skip_mask[1] |= (1 << INTRA_FRAME);
}
mode_skip_mask[INTRA_FRAME] |=
- ~(cpi->sf.intra_y_mode_mask[max_txsize_lookup[bsize]]);
+ ~(sf->intra_y_mode_mask[max_txsize_lookup[bsize]]);
for (i = 0; i < MAX_MODES; ++i)
mode_threshold[i] = ((int64_t)rd_threshes[i] * rd_thresh_freq_fact[i]) >> 5;
- midx = cpi->sf.schedule_mode_search ? mode_skip_start : 0;
+ midx = sf->schedule_mode_search ? mode_skip_start : 0;
while (midx > 4) {
uint8_t end_pos = 0;
for (i = 5; i < midx; ++i) {
@@ -2758,18 +3015,18 @@
continue;
// Test best rd so far against threshold for trying this mode.
- if (best_mode_skippable && cpi->sf.schedule_mode_search)
+ if (best_mode_skippable && sf->schedule_mode_search)
mode_threshold[mode_index] <<= 1;
if (best_rd < mode_threshold[mode_index])
continue;
- if (cpi->sf.motion_field_mode_search) {
+ if (sf->motion_field_mode_search) {
const int mi_width = MIN(num_8x8_blocks_wide_lookup[bsize],
tile->mi_col_end - mi_col);
const int mi_height = MIN(num_8x8_blocks_high_lookup[bsize],
tile->mi_row_end - mi_row);
- const int bsl = mi_width_log2(bsize);
+ const int bsl = mi_width_log2_lookup[bsize];
int cb_partition_search_ctrl = (((mi_row + mi_col) >> bsl)
+ get_chessboard_index(cm->current_video_frame)) & 0x1;
MB_MODE_INFO *ref_mbmi;
@@ -2838,7 +3095,7 @@
}
if (ref_frame == INTRA_FRAME) {
- if (cpi->sf.adaptive_mode_search)
+ if (sf->adaptive_mode_search)
if ((x->source_variance << num_pels_log2_lookup[bsize]) > best_pred_sse)
continue;
@@ -2895,14 +3152,15 @@
if (ref_frame == INTRA_FRAME) {
TX_SIZE uv_tx;
+ struct macroblockd_plane *const pd = &xd->plane[1];
+ vpx_memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
super_block_yrd(cpi, x, &rate_y, &distortion_y, &skippable,
NULL, bsize, tx_cache, best_rd);
-
if (rate_y == INT_MAX)
continue;
- uv_tx = get_uv_tx_size_impl(mbmi->tx_size, bsize, pd[1].subsampling_x,
- pd[1].subsampling_y);
+ uv_tx = get_uv_tx_size_impl(mbmi->tx_size, bsize, pd->subsampling_x,
+ pd->subsampling_y);
if (rate_uv_intra[uv_tx] == INT_MAX) {
choose_intra_uv_mode(cpi, ctx, bsize, uv_tx,
&rate_uv_intra[uv_tx], &rate_uv_tokenonly[uv_tx],
@@ -2922,8 +3180,7 @@
this_rd = handle_inter_mode(cpi, x, bsize,
tx_cache,
&rate2, &distortion2, &skippable,
- &rate_y, &distortion_y,
- &rate_uv, &distortion_uv,
+ &rate_y, &rate_uv,
&disable_skip, frame_mv,
mi_row, mi_col,
single_newmv, single_inter_filter,
@@ -3004,8 +3261,9 @@
best_pred_sse = x->pred_sse[ref_frame];
}
- *returnrate = rate2;
- *returndistortion = distortion2;
+ rd_cost->rate = rate2;
+ rd_cost->dist = distortion2;
+ rd_cost->rdcost = this_rd;
best_rd = this_rd;
best_mbmode = *mbmi;
best_skip2 = this_skip2;
@@ -3020,9 +3278,14 @@
// based on qp, activity mask and history
if ((mode_search_skip_flags & FLAG_EARLY_TERMINATE) &&
(mode_index > MIN_EARLY_TERM_INDEX)) {
- const int qstep = xd->plane[0].dequant[1];
+ int qstep = xd->plane[0].dequant[1];
// TODO(debargha): Enhance this by specializing for each mode_index
int scale = 4;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ qstep >>= (xd->bd - 8);
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (x->source_variance < UINT_MAX) {
const int var_adjust = (x->source_variance < 16);
scale -= var_adjust;
@@ -3130,11 +3393,14 @@
best_mbmode.mode = ZEROMV;
}
- if (best_mode_index < 0 || best_rd >= best_rd_so_far)
- return INT64_MAX;
+ if (best_mode_index < 0 || best_rd >= best_rd_so_far) {
+ rd_cost->rate = INT_MAX;
+ rd_cost->rdcost = INT64_MAX;
+ return;
+ }
// If we used an estimate for the uv intra rd in the loop above...
- if (cpi->sf.use_uv_intra_rd_estimate) {
+ if (sf->use_uv_intra_rd_estimate) {
// Do Intra UV best rd mode selection if best mode choice above was intra.
if (best_mbmode.ref_frame[0] == INTRA_FRAME) {
TX_SIZE uv_tx_size;
@@ -3191,18 +3457,33 @@
// updating code causes PSNR loss. Need to figure out the confliction.
x->skip |= best_mode_skippable;
+ if (!x->skip && !x->select_tx_size) {
+ int has_high_freq_coeff = 0;
+ int plane;
+ int max_plane = is_inter_block(&xd->mi[0].src_mi->mbmi)
+ ? MAX_MB_PLANE : 1;
+ for (plane = 0; plane < max_plane; ++plane) {
+ x->plane[plane].eobs = ctx->eobs_pbuf[plane][1];
+ has_high_freq_coeff |= vp9_has_high_freq_in_plane(x, bsize, plane);
+ }
+
+ for (plane = max_plane; plane < MAX_MB_PLANE; ++plane) {
+ x->plane[plane].eobs = ctx->eobs_pbuf[plane][2];
+ has_high_freq_coeff |= vp9_has_high_freq_in_plane(x, bsize, plane);
+ }
+
+ best_mode_skippable |= !has_high_freq_coeff;
+ }
+
store_coding_context(x, ctx, best_mode_index, best_pred_diff,
best_tx_diff, best_filter_diff, best_mode_skippable);
-
- return best_rd;
}
-int64_t vp9_rd_pick_inter_mode_sb_seg_skip(VP9_COMP *cpi, MACROBLOCK *x,
- int *returnrate,
- int64_t *returndistortion,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far) {
+void vp9_rd_pick_inter_mode_sb_seg_skip(VP9_COMP *cpi, MACROBLOCK *x,
+ RD_COST *rd_cost,
+ BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far) {
VP9_COMMON *const cm = &cpi->common;
RD_OPT *const rd_opt = &cpi->rd;
MACROBLOCKD *const xd = &x->e_mbd;
@@ -3230,7 +3511,7 @@
for (i = LAST_FRAME; i < MAX_REF_FRAMES; ++i)
x->pred_mv_sad[i] = INT_MAX;
- *returnrate = INT_MAX;
+ rd_cost->rate = INT_MAX;
assert(vp9_segfeature_active(&cm->seg, segment_id, SEG_LVL_SKIP));
@@ -3279,11 +3560,15 @@
rate2 += ref_costs_single[LAST_FRAME];
this_rd = RDCOST(x->rdmult, x->rddiv, rate2, distortion2);
- *returnrate = rate2;
- *returndistortion = distortion2;
+ rd_cost->rate = rate2;
+ rd_cost->dist = distortion2;
+ rd_cost->rdcost = this_rd;
- if (this_rd >= best_rd_so_far)
- return INT64_MAX;
+ if (this_rd >= best_rd_so_far) {
+ rd_cost->rate = INT_MAX;
+ rd_cost->rdcost = INT64_MAX;
+ return;
+ }
assert((cm->interp_filter == SWITCHABLE) ||
(cm->interp_filter == mbmi->interp_filter));
@@ -3298,20 +3583,18 @@
swap_block_ptr(x, ctx, 1, 0, 0, MAX_MB_PLANE);
store_coding_context(x, ctx, THR_ZEROMV,
best_pred_diff, best_tx_diff, best_filter_diff, 0);
-
- return this_rd;
}
-int64_t vp9_rd_pick_inter_mode_sub8x8(VP9_COMP *cpi, MACROBLOCK *x,
- const TileInfo *const tile,
- int mi_row, int mi_col,
- int *returnrate,
- int64_t *returndistortion,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far) {
+void vp9_rd_pick_inter_mode_sub8x8(VP9_COMP *cpi, MACROBLOCK *x,
+ const TileInfo *const tile,
+ int mi_row, int mi_col,
+ RD_COST *rd_cost,
+ BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far) {
VP9_COMMON *const cm = &cpi->common;
RD_OPT *const rd_opt = &cpi->rd;
+ SPEED_FEATURES *const sf = &cpi->sf;
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = &xd->mi[0].src_mi->mbmi;
const struct segmentation *const seg = &cm->seg;
@@ -3338,14 +3621,14 @@
int64_t dist_uv;
int skip_uv;
PREDICTION_MODE mode_uv = DC_PRED;
- const int intra_cost_penalty =
- 20 * vp9_dc_quant(cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth);
+ const int intra_cost_penalty = vp9_get_intra_cost_penalty(
+ cm->base_qindex, cm->y_dc_delta_q, cm->bit_depth);
int_mv seg_mvs[4][MAX_REF_FRAMES];
b_mode_info best_bmodes[4];
int best_skip2 = 0;
int ref_frame_skip_mask[2] = { 0 };
- x->skip_encode = cpi->sf.skip_encode_frame && x->q_index < QIDX_SKIP_THRESH;
+ x->skip_encode = sf->skip_encode_frame && x->q_index < QIDX_SKIP_THRESH;
vpx_memset(x->zcoeff_blk[TX_4X4], 0, 4);
vp9_zero(best_mbmode);
@@ -3364,7 +3647,7 @@
best_filter_rd[i] = INT64_MAX;
rate_uv_intra = INT_MAX;
- *returnrate = INT_MAX;
+ rd_cost->rate = INT_MAX;
for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ref_frame++) {
if (cpi->ref_frame_flags & flag_list[ref_frame]) {
@@ -3398,7 +3681,7 @@
// Look at the reference frame of the best mode so far and set the
// skip mask to look at a subset of the remaining modes.
- if (ref_index > 2 && cpi->sf.mode_skip_start < MAX_MODES) {
+ if (ref_index > 2 && sf->mode_skip_start < MAX_MODES) {
if (ref_index == 3) {
switch (best_mbmode.ref_frame[0]) {
case INTRA_FRAME:
@@ -3443,7 +3726,7 @@
if (vp9_segfeature_active(seg, segment_id, SEG_LVL_REF_FRAME))
continue;
- if ((cpi->sf.mode_search_skip_flags & FLAG_SKIP_COMP_BESTINTRA) &&
+ if ((sf->mode_search_skip_flags & FLAG_SKIP_COMP_BESTINTRA) &&
best_mbmode.ref_frame[0] == INTRA_FRAME)
continue;
}
@@ -3547,12 +3830,12 @@
if (cm->interp_filter != BILINEAR) {
tmp_best_filter = EIGHTTAP;
- if (x->source_variance < cpi->sf.disable_filter_search_var_thresh) {
+ if (x->source_variance < sf->disable_filter_search_var_thresh) {
tmp_best_filter = EIGHTTAP;
- } else if (cpi->sf.adaptive_pred_interp_filter == 1 &&
+ } else if (sf->adaptive_pred_interp_filter == 1 &&
ctx->pred_interp_filter < SWITCHABLE) {
tmp_best_filter = ctx->pred_interp_filter;
- } else if (cpi->sf.adaptive_pred_interp_filter == 2) {
+ } else if (sf->adaptive_pred_interp_filter == 2) {
tmp_best_filter = ctx->pred_interp_filter < SWITCHABLE ?
ctx->pred_interp_filter : 0;
} else {
@@ -3605,7 +3888,7 @@
}
pred_exists = 1;
if (switchable_filter_index == 0 &&
- cpi->sf.use_rd_breakout &&
+ sf->use_rd_breakout &&
best_rd < INT64_MAX) {
if (tmp_best_rdu / 2 > best_rd) {
// skip searching the other filters if the first is
@@ -3668,10 +3951,11 @@
// then dont bother looking at UV
vp9_build_inter_predictors_sbuv(&x->e_mbd, mi_row, mi_col,
BLOCK_8X8);
- super_block_uvrd(cpi, x, &rate_uv, &distortion_uv, &uv_skippable,
- &uv_sse, BLOCK_8X8, tmp_best_rdu);
- if (rate_uv == INT_MAX)
+ vpx_memset(x->skip_txfm, 0, sizeof(x->skip_txfm));
+ if (!super_block_uvrd(cpi, x, &rate_uv, &distortion_uv, &uv_skippable,
+ &uv_sse, BLOCK_8X8, tmp_best_rdu))
continue;
+
rate2 += rate_uv;
distortion2 += distortion_uv;
skippable = skippable && uv_skippable;
@@ -3738,8 +4022,9 @@
max_plane = 1;
}
- *returnrate = rate2;
- *returndistortion = distortion2;
+ rd_cost->rate = rate2;
+ rd_cost->dist = distortion2;
+ rd_cost->rdcost = this_rd;
best_rd = this_rd;
best_yrd = best_rd -
RDCOST(x->rdmult, x->rddiv, rate_uv, distortion_uv);
@@ -3755,11 +4040,16 @@
// TODO(debargha): enhance this test with a better distortion prediction
// based on qp, activity mask and history
- if ((cpi->sf.mode_search_skip_flags & FLAG_EARLY_TERMINATE) &&
+ if ((sf->mode_search_skip_flags & FLAG_EARLY_TERMINATE) &&
(ref_index > MIN_EARLY_TERM_INDEX)) {
- const int qstep = xd->plane[0].dequant[1];
+ int qstep = xd->plane[0].dequant[1];
// TODO(debargha): Enhance this by specializing for each mode_index
int scale = 4;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ qstep >>= (xd->bd - 8);
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
if (x->source_variance < UINT_MAX) {
const int var_adjust = (x->source_variance < 16);
scale -= var_adjust;
@@ -3826,11 +4116,14 @@
break;
}
- if (best_rd >= best_rd_so_far)
- return INT64_MAX;
+ if (best_rd >= best_rd_so_far) {
+ rd_cost->rate = INT_MAX;
+ rd_cost->rdcost = INT64_MAX;
+ return;
+ }
// If we used an estimate for the uv intra rd in the loop above...
- if (cpi->sf.use_uv_intra_rd_estimate) {
+ if (sf->use_uv_intra_rd_estimate) {
// Do Intra UV best rd mode selection if best mode choice above was intra.
if (best_mbmode.ref_frame[0] == INTRA_FRAME) {
*mbmi = best_mbmode;
@@ -3843,9 +4136,10 @@
}
if (best_rd == INT64_MAX) {
- *returnrate = INT_MAX;
- *returndistortion = INT64_MAX;
- return best_rd;
+ rd_cost->rate = INT_MAX;
+ rd_cost->dist = INT64_MAX;
+ rd_cost->rdcost = INT64_MAX;
+ return;
}
assert((cm->interp_filter == SWITCHABLE) ||
@@ -3891,7 +4185,5 @@
store_coding_context(x, ctx, best_ref_index,
best_pred_diff, best_tx_diff, best_filter_diff, 0);
-
- return best_rd;
}
diff --git a/source/libvpx/vp9/encoder/vp9_rdopt.h b/source/libvpx/vp9/encoder/vp9_rdopt.h
index 52c603f..ed38ce8 100644
--- a/source/libvpx/vp9/encoder/vp9_rdopt.h
+++ b/source/libvpx/vp9/encoder/vp9_rdopt.h
@@ -23,38 +23,33 @@
struct TileInfo;
struct VP9_COMP;
struct macroblock;
+struct RD_COST;
void vp9_rd_pick_intra_mode_sb(struct VP9_COMP *cpi, struct macroblock *x,
- int *r, int64_t *d, BLOCK_SIZE bsize,
+ struct RD_COST *rd_cost, BLOCK_SIZE bsize,
PICK_MODE_CONTEXT *ctx, int64_t best_rd);
-int64_t vp9_rd_pick_inter_mode_sb(struct VP9_COMP *cpi, struct macroblock *x,
- const struct TileInfo *const tile,
- int mi_row, int mi_col,
- int *returnrate,
- int64_t *returndistortion,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far);
+void vp9_rd_pick_inter_mode_sb(struct VP9_COMP *cpi, struct macroblock *x,
+ const struct TileInfo *const tile,
+ int mi_row, int mi_col,
+ struct RD_COST *rd_cost,
+ BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far);
-int64_t vp9_rd_pick_inter_mode_sb_seg_skip(struct VP9_COMP *cpi,
- struct macroblock *x,
- int *returnrate,
- int64_t *returndistortion,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far);
+void vp9_rd_pick_inter_mode_sb_seg_skip(struct VP9_COMP *cpi,
+ struct macroblock *x,
+ struct RD_COST *rd_cost,
+ BLOCK_SIZE bsize,
+ PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far);
-int64_t vp9_rd_pick_inter_mode_sub8x8(struct VP9_COMP *cpi,
- struct macroblock *x,
- const struct TileInfo *const tile,
- int mi_row, int mi_col,
- int *returnrate,
- int64_t *returndistortion,
- BLOCK_SIZE bsize,
- PICK_MODE_CONTEXT *ctx,
- int64_t best_rd_so_far);
-
+void vp9_rd_pick_inter_mode_sub8x8(struct VP9_COMP *cpi,
+ struct macroblock *x,
+ const struct TileInfo *const tile,
+ int mi_row, int mi_col,
+ struct RD_COST *rd_cost,
+ BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx,
+ int64_t best_rd_so_far);
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/source/libvpx/vp9/encoder/vp9_resize.c b/source/libvpx/vp9/encoder/vp9_resize.c
index 4e6efae..3d361d4 100644
--- a/source/libvpx/vp9/encoder/vp9_resize.c
+++ b/source/libvpx/vp9/encoder/vp9_resize.c
@@ -312,7 +312,7 @@
static void down2_symeven(const uint8_t *const input, int length,
uint8_t *output) {
// Actual filter len = 2 * filter_len_half.
- static const int16_t *filter = vp9_down2_symeven_half_filter;
+ const int16_t *filter = vp9_down2_symeven_half_filter;
const int filter_len_half = sizeof(vp9_down2_symeven_half_filter) / 2;
int i, j;
uint8_t *optr = output;
@@ -368,7 +368,7 @@
static void down2_symodd(const uint8_t *const input, int length,
uint8_t *output) {
// Actual filter len = 2 * filter_len_half - 1.
- static const int16_t *filter = vp9_down2_symodd_half_filter;
+ const int16_t *filter = vp9_down2_symodd_half_filter;
const int filter_len_half = sizeof(vp9_down2_symodd_half_filter) / 2;
int i, j;
uint8_t *optr = output;
@@ -529,6 +529,302 @@
free(arrbuf);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static void highbd_interpolate(const uint16_t *const input, int inlength,
+ uint16_t *output, int outlength, int bd) {
+ const int64_t delta =
+ (((uint64_t)inlength << 32) + outlength / 2) / outlength;
+ const int64_t offset = inlength > outlength ?
+ (((int64_t)(inlength - outlength) << 31) + outlength / 2) / outlength :
+ -(((int64_t)(outlength - inlength) << 31) + outlength / 2) / outlength;
+ uint16_t *optr = output;
+ int x, x1, x2, sum, k, int_pel, sub_pel;
+ int64_t y;
+
+ const interp_kernel *interp_filters =
+ choose_interp_filter(inlength, outlength);
+
+ x = 0;
+ y = offset;
+ while ((y >> INTERP_PRECISION_BITS) < (INTERP_TAPS / 2 - 1)) {
+ x++;
+ y += delta;
+ }
+ x1 = x;
+ x = outlength - 1;
+ y = delta * x + offset;
+ while ((y >> INTERP_PRECISION_BITS) +
+ (int64_t)(INTERP_TAPS / 2) >= inlength) {
+ x--;
+ y -= delta;
+ }
+ x2 = x;
+ if (x1 > x2) {
+ for (x = 0, y = offset; x < outlength; ++x, y += delta) {
+ const int16_t *filter;
+ int_pel = y >> INTERP_PRECISION_BITS;
+ sub_pel = (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS)) & SUBPEL_MASK;
+ filter = interp_filters[sub_pel];
+ sum = 0;
+ for (k = 0; k < INTERP_TAPS; ++k) {
+ const int pk = int_pel - INTERP_TAPS / 2 + 1 + k;
+ sum += filter[k] *
+ input[(pk < 0 ? 0 : (pk >= inlength ? inlength - 1 : pk))];
+ }
+ *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
+ }
+ } else {
+ // Initial part.
+ for (x = 0, y = offset; x < x1; ++x, y += delta) {
+ const int16_t *filter;
+ int_pel = y >> INTERP_PRECISION_BITS;
+ sub_pel = (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS)) & SUBPEL_MASK;
+ filter = interp_filters[sub_pel];
+ sum = 0;
+ for (k = 0; k < INTERP_TAPS; ++k)
+ sum += filter[k] *
+ input[(int_pel - INTERP_TAPS / 2 + 1 + k < 0 ?
+ 0 : int_pel - INTERP_TAPS / 2 + 1 + k)];
+ *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
+ }
+ // Middle part.
+ for (; x <= x2; ++x, y += delta) {
+ const int16_t *filter;
+ int_pel = y >> INTERP_PRECISION_BITS;
+ sub_pel = (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS)) & SUBPEL_MASK;
+ filter = interp_filters[sub_pel];
+ sum = 0;
+ for (k = 0; k < INTERP_TAPS; ++k)
+ sum += filter[k] * input[int_pel - INTERP_TAPS / 2 + 1 + k];
+ *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
+ }
+ // End part.
+ for (; x < outlength; ++x, y += delta) {
+ const int16_t *filter;
+ int_pel = y >> INTERP_PRECISION_BITS;
+ sub_pel = (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS)) & SUBPEL_MASK;
+ filter = interp_filters[sub_pel];
+ sum = 0;
+ for (k = 0; k < INTERP_TAPS; ++k)
+ sum += filter[k] * input[(int_pel - INTERP_TAPS / 2 + 1 + k >=
+ inlength ? inlength - 1 :
+ int_pel - INTERP_TAPS / 2 + 1 + k)];
+ *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
+ }
+ }
+}
+
+static void highbd_down2_symeven(const uint16_t *const input, int length,
+ uint16_t *output, int bd) {
+ // Actual filter len = 2 * filter_len_half.
+ static const int16_t *filter = vp9_down2_symeven_half_filter;
+ const int filter_len_half = sizeof(vp9_down2_symeven_half_filter) / 2;
+ int i, j;
+ uint16_t *optr = output;
+ int l1 = filter_len_half;
+ int l2 = (length - filter_len_half);
+ l1 += (l1 & 1);
+ l2 += (l2 & 1);
+ if (l1 > l2) {
+ // Short input length.
+ for (i = 0; i < length; i += 2) {
+ int sum = (1 << (FILTER_BITS - 1));
+ for (j = 0; j < filter_len_half; ++j) {
+ sum += (input[(i - j < 0 ? 0 : i - j)] +
+ input[(i + 1 + j >= length ? length - 1 : i + 1 + j)]) *
+ filter[j];
+ }
+ sum >>= FILTER_BITS;
+ *optr++ = clip_pixel_highbd(sum, bd);
+ }
+ } else {
+ // Initial part.
+ for (i = 0; i < l1; i += 2) {
+ int sum = (1 << (FILTER_BITS - 1));
+ for (j = 0; j < filter_len_half; ++j) {
+ sum += (input[(i - j < 0 ? 0 : i - j)] + input[i + 1 + j]) * filter[j];
+ }
+ sum >>= FILTER_BITS;
+ *optr++ = clip_pixel_highbd(sum, bd);
+ }
+ // Middle part.
+ for (; i < l2; i += 2) {
+ int sum = (1 << (FILTER_BITS - 1));
+ for (j = 0; j < filter_len_half; ++j) {
+ sum += (input[i - j] + input[i + 1 + j]) * filter[j];
+ }
+ sum >>= FILTER_BITS;
+ *optr++ = clip_pixel_highbd(sum, bd);
+ }
+ // End part.
+ for (; i < length; i += 2) {
+ int sum = (1 << (FILTER_BITS - 1));
+ for (j = 0; j < filter_len_half; ++j) {
+ sum += (input[i - j] +
+ input[(i + 1 + j >= length ? length - 1 : i + 1 + j)]) *
+ filter[j];
+ }
+ sum >>= FILTER_BITS;
+ *optr++ = clip_pixel_highbd(sum, bd);
+ }
+ }
+}
+
+static void highbd_down2_symodd(const uint16_t *const input, int length,
+ uint16_t *output, int bd) {
+ // Actual filter len = 2 * filter_len_half - 1.
+ static const int16_t *filter = vp9_down2_symodd_half_filter;
+ const int filter_len_half = sizeof(vp9_down2_symodd_half_filter) / 2;
+ int i, j;
+ uint16_t *optr = output;
+ int l1 = filter_len_half - 1;
+ int l2 = (length - filter_len_half + 1);
+ l1 += (l1 & 1);
+ l2 += (l2 & 1);
+ if (l1 > l2) {
+ // Short input length.
+ for (i = 0; i < length; i += 2) {
+ int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
+ for (j = 1; j < filter_len_half; ++j) {
+ sum += (input[(i - j < 0 ? 0 : i - j)] +
+ input[(i + j >= length ? length - 1 : i + j)]) *
+ filter[j];
+ }
+ sum >>= FILTER_BITS;
+ *optr++ = clip_pixel_highbd(sum, bd);
+ }
+ } else {
+ // Initial part.
+ for (i = 0; i < l1; i += 2) {
+ int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
+ for (j = 1; j < filter_len_half; ++j) {
+ sum += (input[(i - j < 0 ? 0 : i - j)] + input[i + j]) * filter[j];
+ }
+ sum >>= FILTER_BITS;
+ *optr++ = clip_pixel_highbd(sum, bd);
+ }
+ // Middle part.
+ for (; i < l2; i += 2) {
+ int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
+ for (j = 1; j < filter_len_half; ++j) {
+ sum += (input[i - j] + input[i + j]) * filter[j];
+ }
+ sum >>= FILTER_BITS;
+ *optr++ = clip_pixel_highbd(sum, bd);
+ }
+ // End part.
+ for (; i < length; i += 2) {
+ int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
+ for (j = 1; j < filter_len_half; ++j) {
+ sum += (input[i - j] + input[(i + j >= length ? length - 1 : i + j)]) *
+ filter[j];
+ }
+ sum >>= FILTER_BITS;
+ *optr++ = clip_pixel_highbd(sum, bd);
+ }
+ }
+}
+
+static void highbd_resize_multistep(const uint16_t *const input,
+ int length,
+ uint16_t *output,
+ int olength,
+ uint16_t *buf,
+ int bd) {
+ int steps;
+ if (length == olength) {
+ memcpy(output, input, sizeof(uint16_t) * length);
+ return;
+ }
+ steps = get_down2_steps(length, olength);
+
+ if (steps > 0) {
+ int s;
+ uint16_t *out = NULL;
+ uint16_t *tmpbuf = NULL;
+ uint16_t *otmp, *otmp2;
+ int filteredlength = length;
+ if (!tmpbuf) {
+ tmpbuf = (uint16_t *)malloc(sizeof(uint16_t) * length);
+ otmp = tmpbuf;
+ } else {
+ otmp = buf;
+ }
+ otmp2 = otmp + get_down2_length(length, 1);
+ for (s = 0; s < steps; ++s) {
+ const int proj_filteredlength = get_down2_length(filteredlength, 1);
+ const uint16_t *const in = (s == 0 ? input : out);
+ if (s == steps - 1 && proj_filteredlength == olength)
+ out = output;
+ else
+ out = (s & 1 ? otmp2 : otmp);
+ if (filteredlength & 1)
+ highbd_down2_symodd(in, filteredlength, out, bd);
+ else
+ highbd_down2_symeven(in, filteredlength, out, bd);
+ filteredlength = proj_filteredlength;
+ }
+ if (filteredlength != olength) {
+ highbd_interpolate(out, filteredlength, output, olength, bd);
+ }
+ if (tmpbuf)
+ free(tmpbuf);
+ } else {
+ highbd_interpolate(input, length, output, olength, bd);
+ }
+}
+
+static void highbd_fill_col_to_arr(uint16_t *img, int stride, int len,
+ uint16_t *arr) {
+ int i;
+ uint16_t *iptr = img;
+ uint16_t *aptr = arr;
+ for (i = 0; i < len; ++i, iptr += stride) {
+ *aptr++ = *iptr;
+ }
+}
+
+static void highbd_fill_arr_to_col(uint16_t *img, int stride, int len,
+ uint16_t *arr) {
+ int i;
+ uint16_t *iptr = img;
+ uint16_t *aptr = arr;
+ for (i = 0; i < len; ++i, iptr += stride) {
+ *iptr = *aptr++;
+ }
+}
+
+void vp9_highbd_resize_plane(const uint8_t *const input,
+ int height,
+ int width,
+ int in_stride,
+ uint8_t *output,
+ int height2,
+ int width2,
+ int out_stride,
+ int bd) {
+ int i;
+ uint16_t *intbuf = (uint16_t *)malloc(sizeof(uint16_t) * width2 * height);
+ uint16_t *tmpbuf = (uint16_t *)malloc(sizeof(uint16_t) *
+ (width < height ? height : width));
+ uint16_t *arrbuf = (uint16_t *)malloc(sizeof(uint16_t) * (height + height2));
+ for (i = 0; i < height; ++i) {
+ highbd_resize_multistep(CONVERT_TO_SHORTPTR(input + in_stride * i), width,
+ intbuf + width2 * i, width2, tmpbuf, bd);
+ }
+ for (i = 0; i < width2; ++i) {
+ highbd_fill_col_to_arr(intbuf + i, width2, height, arrbuf);
+ highbd_resize_multistep(arrbuf, height, arrbuf + height, height2, tmpbuf,
+ bd);
+ highbd_fill_arr_to_col(CONVERT_TO_SHORTPTR(output + i), out_stride, height2,
+ arrbuf + height);
+ }
+ free(intbuf);
+ free(tmpbuf);
+ free(arrbuf);
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
void vp9_resize_frame420(const uint8_t *const y,
int y_stride,
const uint8_t *const u, const uint8_t *const v,
@@ -574,3 +870,51 @@
vp9_resize_plane(v, height, width, uv_stride,
ov, oheight, owidth, ouv_stride);
}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+void vp9_highbd_resize_frame420(const uint8_t *const y,
+ int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride,
+ int height, int width,
+ uint8_t *oy, int oy_stride,
+ uint8_t *ou, uint8_t *ov, int ouv_stride,
+ int oheight, int owidth, int bd) {
+ vp9_highbd_resize_plane(y, height, width, y_stride,
+ oy, oheight, owidth, oy_stride, bd);
+ vp9_highbd_resize_plane(u, height / 2, width / 2, uv_stride,
+ ou, oheight / 2, owidth / 2, ouv_stride, bd);
+ vp9_highbd_resize_plane(v, height / 2, width / 2, uv_stride,
+ ov, oheight / 2, owidth / 2, ouv_stride, bd);
+}
+
+void vp9_highbd_resize_frame422(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride,
+ int height, int width,
+ uint8_t *oy, int oy_stride,
+ uint8_t *ou, uint8_t *ov, int ouv_stride,
+ int oheight, int owidth, int bd) {
+ vp9_highbd_resize_plane(y, height, width, y_stride,
+ oy, oheight, owidth, oy_stride, bd);
+ vp9_highbd_resize_plane(u, height, width / 2, uv_stride,
+ ou, oheight, owidth / 2, ouv_stride, bd);
+ vp9_highbd_resize_plane(v, height, width / 2, uv_stride,
+ ov, oheight, owidth / 2, ouv_stride, bd);
+}
+
+void vp9_highbd_resize_frame444(const uint8_t *const y, int y_stride,
+ const uint8_t *const u, const uint8_t *const v,
+ int uv_stride,
+ int height, int width,
+ uint8_t *oy, int oy_stride,
+ uint8_t *ou, uint8_t *ov, int ouv_stride,
+ int oheight, int owidth, int bd) {
+ vp9_highbd_resize_plane(y, height, width, y_stride,
+ oy, oheight, owidth, oy_stride, bd);
+ vp9_highbd_resize_plane(u, height, width, uv_stride,
+ ou, oheight, owidth, ouv_stride, bd);
+ vp9_highbd_resize_plane(v, height, width, uv_stride,
+ ov, oheight, owidth, ouv_stride, bd);
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/source/libvpx/vp9/encoder/vp9_resize.h b/source/libvpx/vp9/encoder/vp9_resize.h
index 1818cd4..067af53 100644
--- a/source/libvpx/vp9/encoder/vp9_resize.h
+++ b/source/libvpx/vp9/encoder/vp9_resize.h
@@ -65,4 +65,60 @@
int oheight,
int owidth);
+#if CONFIG_VP9_HIGHBITDEPTH
+void vp9_highbd_resize_plane(const uint8_t *const input,
+ int height,
+ int width,
+ int in_stride,
+ uint8_t *output,
+ int height2,
+ int width2,
+ int out_stride,
+ int bd);
+void vp9_highbd_resize_frame420(const uint8_t *const y,
+ int y_stride,
+ const uint8_t *const u,
+ const uint8_t *const v,
+ int uv_stride,
+ int height,
+ int width,
+ uint8_t *oy,
+ int oy_stride,
+ uint8_t *ou,
+ uint8_t *ov,
+ int ouv_stride,
+ int oheight,
+ int owidth,
+ int bd);
+void vp9_highbd_resize_frame422(const uint8_t *const y,
+ int y_stride,
+ const uint8_t *const u,
+ const uint8_t *const v,
+ int uv_stride,
+ int height,
+ int width,
+ uint8_t *oy,
+ int oy_stride,
+ uint8_t *ou,
+ uint8_t *ov,
+ int ouv_stride,
+ int oheight,
+ int owidth,
+ int bd);
+void vp9_highbd_resize_frame444(const uint8_t *const y,
+ int y_stride,
+ const uint8_t *const u,
+ const uint8_t *const v,
+ int uv_stride,
+ int height,
+ int width,
+ uint8_t *oy,
+ int oy_stride,
+ uint8_t *ou,
+ uint8_t *ov,
+ int ouv_stride,
+ int oheight,
+ int owidth,
+ int bd);
+#endif // CONFIG_VP9_HIGHBITDEPTH
#endif // VP9_ENCODER_VP9_RESIZE_H_
diff --git a/source/libvpx/vp9/encoder/vp9_sad.c b/source/libvpx/vp9/encoder/vp9_sad.c
index cee6ce1..73134f2 100644
--- a/source/libvpx/vp9/encoder/vp9_sad.c
+++ b/source/libvpx/vp9/encoder/vp9_sad.c
@@ -32,7 +32,6 @@
a += a_stride;
b += b_stride;
}
-
return sad;
}
@@ -136,9 +135,9 @@
sadMxNx4D(4, 4)
#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE unsigned int high_sad(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int width, int height) {
+static INLINE unsigned int highbd_sad(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int width, int height) {
int y, x;
unsigned int sad = 0;
const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
@@ -153,9 +152,9 @@
return sad;
}
-static INLINE unsigned int high_sadb(const uint8_t *a8, int a_stride,
- const uint16_t *b, int b_stride,
- int width, int height) {
+static INLINE unsigned int highbd_sadb(const uint8_t *a8, int a_stride,
+ const uint16_t *b, int b_stride,
+ int width, int height) {
int y, x;
unsigned int sad = 0;
const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
@@ -169,103 +168,109 @@
return sad;
}
-#define high_sadMxN(m, n) \
-unsigned int vp9_high_sad##m##x##n##_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride) { \
- return high_sad(src, src_stride, ref, ref_stride, m, n); \
+#define highbd_sadMxN(m, n) \
+unsigned int vp9_highbd_sad##m##x##n##_c(const uint8_t *src, int src_stride, \
+ const uint8_t *ref, int ref_stride) { \
+ return highbd_sad(src, src_stride, ref, ref_stride, m, n); \
} \
-unsigned int vp9_high_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) { \
+unsigned int vp9_highbd_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) { \
uint16_t comp_pred[m * n]; \
- vp9_high_comp_avg_pred(comp_pred, second_pred, m, n, ref, ref_stride); \
- return high_sadb(src, src_stride, comp_pred, m, m, n); \
+ vp9_highbd_comp_avg_pred(comp_pred, second_pred, m, n, ref, ref_stride); \
+ return highbd_sadb(src, src_stride, comp_pred, m, m, n); \
}
-#define high_sadMxNxK(m, n, k) \
-void vp9_high_sad##m##x##n##x##k##_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride, \
- unsigned int *sads) { \
+#define highbd_sadMxNxK(m, n, k) \
+void vp9_highbd_sad##m##x##n##x##k##_c(const uint8_t *src, int src_stride, \
+ const uint8_t *ref, int ref_stride, \
+ unsigned int *sads) { \
int i; \
- for (i = 0; i < k; ++i) \
- sads[i] = vp9_high_sad##m##x##n##_c(src, src_stride, &ref[i], ref_stride); \
+ for (i = 0; i < k; ++i) { \
+ sads[i] = vp9_highbd_sad##m##x##n##_c(src, src_stride, &ref[i], \
+ ref_stride); \
+ } \
}
-#define high_sadMxNx4D(m, n) \
-void vp9_high_sad##m##x##n##x4d_c(const uint8_t *src, int src_stride, \
- const uint8_t *const refs[], \
- int ref_stride, unsigned int *sads) { \
+#define highbd_sadMxNx4D(m, n) \
+void vp9_highbd_sad##m##x##n##x4d_c(const uint8_t *src, int src_stride, \
+ const uint8_t *const refs[], \
+ int ref_stride, unsigned int *sads) { \
int i; \
- for (i = 0; i < 4; ++i) \
- sads[i] = vp9_high_sad##m##x##n##_c(src, src_stride, refs[i], ref_stride); \
+ for (i = 0; i < 4; ++i) { \
+ sads[i] = vp9_highbd_sad##m##x##n##_c(src, src_stride, refs[i], \
+ ref_stride); \
+ } \
}
// 64x64
-high_sadMxN(64, 64)
-high_sadMxNxK(64, 64, 3)
-high_sadMxNxK(64, 64, 8)
-high_sadMxNx4D(64, 64)
+highbd_sadMxN(64, 64)
+highbd_sadMxNxK(64, 64, 3)
+highbd_sadMxNxK(64, 64, 8)
+highbd_sadMxNx4D(64, 64)
// 64x32
-high_sadMxN(64, 32)
-high_sadMxNx4D(64, 32)
+highbd_sadMxN(64, 32)
+highbd_sadMxNx4D(64, 32)
// 32x64
-high_sadMxN(32, 64)
-high_sadMxNx4D(32, 64)
+highbd_sadMxN(32, 64)
+highbd_sadMxNx4D(32, 64)
// 32x32
-high_sadMxN(32, 32)
-high_sadMxNxK(32, 32, 3)
-high_sadMxNxK(32, 32, 8)
-high_sadMxNx4D(32, 32)
+highbd_sadMxN(32, 32)
+highbd_sadMxNxK(32, 32, 3)
+highbd_sadMxNxK(32, 32, 8)
+highbd_sadMxNx4D(32, 32)
// 32x16
-high_sadMxN(32, 16)
-high_sadMxNx4D(32, 16)
+highbd_sadMxN(32, 16)
+highbd_sadMxNx4D(32, 16)
// 16x32
-high_sadMxN(16, 32)
-high_sadMxNx4D(16, 32)
+highbd_sadMxN(16, 32)
+highbd_sadMxNx4D(16, 32)
// 16x16
-high_sadMxN(16, 16)
-high_sadMxNxK(16, 16, 3)
-high_sadMxNxK(16, 16, 8)
-high_sadMxNx4D(16, 16)
+highbd_sadMxN(16, 16)
+highbd_sadMxNxK(16, 16, 3)
+highbd_sadMxNxK(16, 16, 8)
+highbd_sadMxNx4D(16, 16)
// 16x8
-high_sadMxN(16, 8)
-high_sadMxNxK(16, 8, 3)
-high_sadMxNxK(16, 8, 8)
-high_sadMxNx4D(16, 8)
+highbd_sadMxN(16, 8)
+highbd_sadMxNxK(16, 8, 3)
+highbd_sadMxNxK(16, 8, 8)
+highbd_sadMxNx4D(16, 8)
// 8x16
-high_sadMxN(8, 16)
-high_sadMxNxK(8, 16, 3)
-high_sadMxNxK(8, 16, 8)
-high_sadMxNx4D(8, 16)
+highbd_sadMxN(8, 16)
+highbd_sadMxNxK(8, 16, 3)
+highbd_sadMxNxK(8, 16, 8)
+highbd_sadMxNx4D(8, 16)
// 8x8
-high_sadMxN(8, 8)
-high_sadMxNxK(8, 8, 3)
-high_sadMxNxK(8, 8, 8)
-high_sadMxNx4D(8, 8)
+highbd_sadMxN(8, 8)
+highbd_sadMxNxK(8, 8, 3)
+highbd_sadMxNxK(8, 8, 8)
+highbd_sadMxNx4D(8, 8)
// 8x4
-high_sadMxN(8, 4)
-high_sadMxNxK(8, 4, 8)
-high_sadMxNx4D(8, 4)
+highbd_sadMxN(8, 4)
+highbd_sadMxNxK(8, 4, 8)
+highbd_sadMxNx4D(8, 4)
// 4x8
-high_sadMxN(4, 8)
-high_sadMxNxK(4, 8, 8)
-high_sadMxNx4D(4, 8)
+highbd_sadMxN(4, 8)
+highbd_sadMxNxK(4, 8, 8)
+highbd_sadMxNx4D(4, 8)
// 4x4
-high_sadMxN(4, 4)
-high_sadMxNxK(4, 4, 3)
-high_sadMxNxK(4, 4, 8)
-high_sadMxNx4D(4, 4)
+highbd_sadMxN(4, 4)
+highbd_sadMxNxK(4, 4, 3)
+highbd_sadMxNxK(4, 4, 8)
+highbd_sadMxNx4D(4, 4)
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/source/libvpx/vp9/encoder/vp9_speed_features.c b/source/libvpx/vp9/encoder/vp9_speed_features.c
index 52e9a8e..9e3ee2c 100644
--- a/source/libvpx/vp9/encoder/vp9_speed_features.c
+++ b/source/libvpx/vp9/encoder/vp9_speed_features.c
@@ -59,7 +59,7 @@
sf->partition_search_breakout_dist_thr = (1 << 23);
else
sf->partition_search_breakout_dist_thr = (1 << 21);
- sf->partition_search_breakout_rate_thr = 500;
+ sf->partition_search_breakout_rate_thr = 80;
}
if (speed >= 2) {
@@ -70,8 +70,12 @@
sf->disable_split_mask = cm->show_frame ? DISABLE_ALL_SPLIT
: DISABLE_ALL_INTER_SPLIT;
sf->adaptive_pred_interp_filter = 0;
+ sf->partition_search_breakout_dist_thr = (1 << 24);
+ sf->partition_search_breakout_rate_thr = 120;
} else {
sf->disable_split_mask = LAST_AND_INTRA_SPLIT_ONLY;
+ sf->partition_search_breakout_dist_thr = (1 << 22);
+ sf->partition_search_breakout_rate_thr = 100;
}
sf->reference_masking = 1;
@@ -83,11 +87,7 @@
sf->comp_inter_joint_search_thresh = BLOCK_SIZES;
sf->auto_min_max_partition_size = CONSTRAIN_NEIGHBORING_MIN_MAX;
- if (MIN(cm->width, cm->height) >= 720)
- sf->partition_search_breakout_dist_thr = (1 << 24);
- else
- sf->partition_search_breakout_dist_thr = (1 << 22);
- sf->partition_search_breakout_rate_thr = 700;
+ sf->allow_partition_search_skip = 1;
}
if (speed >= 3) {
@@ -96,29 +96,27 @@
if (MIN(cm->width, cm->height) >= 720) {
sf->disable_split_mask = DISABLE_ALL_SPLIT;
sf->schedule_mode_search = cm->base_qindex < 220 ? 1 : 0;
+ sf->partition_search_breakout_dist_thr = (1 << 25);
+ sf->partition_search_breakout_rate_thr = 200;
} else {
sf->max_intra_bsize = BLOCK_32X32;
sf->disable_split_mask = DISABLE_ALL_INTER_SPLIT;
sf->schedule_mode_search = cm->base_qindex < 175 ? 1 : 0;
+ sf->partition_search_breakout_dist_thr = (1 << 23);
+ sf->partition_search_breakout_rate_thr = 120;
}
+ sf->mv.subpel_search_method = SUBPEL_TREE_PRUNED;
sf->adaptive_pred_interp_filter = 0;
sf->adaptive_mode_search = 1;
sf->cb_partition_search = !boosted;
sf->cb_pred_filter_search = 1;
sf->alt_ref_search_fp = 1;
- sf->motion_field_mode_search = !boosted;
sf->recode_loop = ALLOW_RECODE_KFMAXBW;
sf->adaptive_rd_thresh = 3;
sf->mode_skip_start = 6;
sf->intra_y_mode_mask[TX_32X32] = INTRA_DC;
sf->intra_uv_mode_mask[TX_32X32] = INTRA_DC;
sf->adaptive_interp_filter_search = 1;
-
- if (MIN(cm->width, cm->height) >= 720)
- sf->partition_search_breakout_dist_thr = (1 << 25);
- else
- sf->partition_search_breakout_dist_thr = (1 << 23);
- sf->partition_search_breakout_rate_thr = 1000;
}
if (speed >= 4) {
@@ -126,19 +124,20 @@
sf->tx_size_search_method = USE_LARGESTALL;
sf->disable_split_mask = DISABLE_ALL_SPLIT;
sf->mv.search_method = BIGDIA;
- sf->mv.subpel_search_method = SUBPEL_TREE_PRUNED;
+ sf->mv.subpel_search_method = SUBPEL_TREE_PRUNED_MORE;
sf->adaptive_rd_thresh = 4;
sf->mode_search_skip_flags |= FLAG_EARLY_TERMINATE;
sf->disable_filter_search_var_thresh = 200;
sf->use_lp32x32fdct = 1;
sf->use_fast_coef_updates = ONE_LOOP_REDUCED;
sf->use_fast_coef_costing = 1;
+ sf->motion_field_mode_search = !boosted;
if (MIN(cm->width, cm->height) >= 720)
sf->partition_search_breakout_dist_thr = (1 << 26);
else
sf->partition_search_breakout_dist_thr = (1 << 24);
- sf->partition_search_breakout_rate_thr = 1500;
+ sf->partition_search_breakout_rate_thr = 300;
}
if (speed >= 5) {
@@ -152,7 +151,6 @@
sf->intra_y_mode_mask[i] = INTRA_DC;
sf->intra_uv_mode_mask[i] = INTRA_DC;
}
- cpi->allow_encode_breakout = ENCODE_BREAKOUT_ENABLED;
}
if (speed >= 6) {
sf->mv.reduce_first_step_size = 1;
@@ -250,6 +248,7 @@
sf->intra_y_mode_mask[TX_32X32] = INTRA_DC;
sf->frame_parameter_update = 0;
sf->mv.search_method = FAST_HEX;
+
sf->inter_mode_mask[BLOCK_32X32] = INTER_NEAREST_NEAR_NEW;
sf->inter_mode_mask[BLOCK_32X64] = INTER_NEAREST;
sf->inter_mode_mask[BLOCK_64X32] = INTER_NEAREST;
@@ -272,6 +271,16 @@
sf->partition_search_type = REFERENCE_PARTITION;
sf->use_nonrd_pick_mode = 1;
sf->allow_skip_recode = 0;
+ sf->inter_mode_mask[BLOCK_32X32] = INTER_NEAREST_NEW_ZERO;
+ sf->inter_mode_mask[BLOCK_32X64] = INTER_NEAREST_NEW_ZERO;
+ sf->inter_mode_mask[BLOCK_64X32] = INTER_NEAREST_NEW_ZERO;
+ sf->inter_mode_mask[BLOCK_64X64] = INTER_NEAREST_NEW_ZERO;
+
+ if (MIN(cm->width, cm->height) >= 720)
+ sf->partition_search_breakout_dist_thr = (1 << 25);
+ else
+ sf->partition_search_breakout_dist_thr = (1 << 23);
+ sf->partition_search_breakout_rate_thr = 200;
}
if (speed >= 6) {
@@ -279,12 +288,13 @@
int i;
// Allow fancy modes at all sizes since SOURCE_VAR_BASED_PARTITION is used
for (i = 0; i < BLOCK_SIZES; ++i)
- sf->inter_mode_mask[i] = INTER_ALL;
+ sf->inter_mode_mask[i] = INTER_NEAREST_NEAR_NEW;
}
// Adaptively switch between SOURCE_VAR_BASED_PARTITION and FIXED_PARTITION.
- sf->partition_search_type = SOURCE_VAR_BASED_PARTITION;
+ sf->partition_search_type = VAR_BASED_PARTITION;
sf->search_type_check_frequency = 50;
+ sf->mv.search_method = NSTEP;
sf->tx_size_search_method = is_keyframe ? USE_LARGESTALL : USE_TX_8X8;
@@ -292,7 +302,7 @@
sf->reuse_inter_pred_sby = 1;
// Increase mode checking threshold for NEWMV.
- sf->elevate_newmv_thresh = 2000;
+ sf->elevate_newmv_thresh = 1000;
sf->mv.reduce_first_step_size = 1;
}
@@ -365,6 +375,7 @@
sf->max_delta_qindex = 0;
sf->disable_filter_search_var_thresh = 0;
sf->adaptive_interp_filter_search = 0;
+ sf->allow_partition_search_skip = 0;
for (i = 0; i < TX_SIZES; i++) {
sf->intra_y_mode_mask[i] = INTRA_ALL;
@@ -423,6 +434,10 @@
cpi->find_fractional_mv_step = vp9_find_best_sub_pixel_tree;
} else if (sf->mv.subpel_search_method == SUBPEL_TREE_PRUNED) {
cpi->find_fractional_mv_step = vp9_find_best_sub_pixel_tree_pruned;
+ } else if (sf->mv.subpel_search_method == SUBPEL_TREE_PRUNED_MORE) {
+ cpi->find_fractional_mv_step = vp9_find_best_sub_pixel_tree_pruned_more;
+ } else if (sf->mv.subpel_search_method == SUBPEL_TREE_PRUNED_EVENMORE) {
+ cpi->find_fractional_mv_step = vp9_find_best_sub_pixel_tree_pruned_evenmore;
}
cpi->mb.optimize = sf->optimize_coefficients == 1 && oxcf->pass != 1;
diff --git a/source/libvpx/vp9/encoder/vp9_speed_features.h b/source/libvpx/vp9/encoder/vp9_speed_features.h
index ed84008..951b4af 100644
--- a/source/libvpx/vp9/encoder/vp9_speed_features.h
+++ b/source/libvpx/vp9/encoder/vp9_speed_features.h
@@ -34,6 +34,9 @@
enum {
INTER_ALL = (1 << NEARESTMV) | (1 << NEARMV) | (1 << ZEROMV) | (1 << NEWMV),
INTER_NEAREST = (1 << NEARESTMV),
+ INTER_NEAREST_NEW = (1 << NEARESTMV) | (1 << NEWMV),
+ INTER_NEAREST_ZERO = (1 << NEARESTMV) | (1 << ZEROMV),
+ INTER_NEAREST_NEW_ZERO = (1 << NEARESTMV) | (1 << ZEROMV) | (1 << NEWMV),
INTER_NEAREST_NEAR_NEW = (1 << NEARESTMV) | (1 << NEARMV) | (1 << NEWMV),
INTER_NEAREST_NEAR_ZERO = (1 << NEARESTMV) | (1 << NEARMV) | (1 << ZEROMV),
};
@@ -78,7 +81,9 @@
typedef enum {
SUBPEL_TREE = 0,
- SUBPEL_TREE_PRUNED = 1,
+ SUBPEL_TREE_PRUNED = 1, // Prunes 1/2-pel searches
+ SUBPEL_TREE_PRUNED_MORE = 2, // Prunes 1/2-pel searches more aggressively
+ SUBPEL_TREE_PRUNED_EVENMORE = 3, // Prunes 1/2- and 1/4-pel searches
// Other methods to come
} SUBPEL_SEARCH_METHODS;
@@ -144,16 +149,12 @@
typedef enum {
// Search partitions using RD/NONRD criterion
- SEARCH_PARTITION = 0,
+ SEARCH_PARTITION,
// Always use a fixed size partition
- FIXED_PARTITION = 1,
+ FIXED_PARTITION,
- // Use a fixed size partition in every 64X64 SB, where the size is
- // determined based on source variance
- VAR_BASED_FIXED_PARTITION = 2,
-
- REFERENCE_PARTITION = 3,
+ REFERENCE_PARTITION,
// Use an arbitrary partitioning scheme based on source variance within
// a 64X64 SB
@@ -435,6 +436,9 @@
// Partition search early breakout thresholds.
int64_t partition_search_breakout_dist_thr;
int partition_search_breakout_rate_thr;
+
+ // Allow skipping partition search for still image frame
+ int allow_partition_search_skip;
} SPEED_FEATURES;
struct VP9_COMP;
diff --git a/source/libvpx/vp9/encoder/vp9_ssim.c b/source/libvpx/vp9/encoder/vp9_ssim.c
index 8435640..5dbfbf5 100644
--- a/source/libvpx/vp9/encoder/vp9_ssim.c
+++ b/source/libvpx/vp9/encoder/vp9_ssim.c
@@ -43,6 +43,24 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+void vp9_highbd_ssim_parms_8x8_c(uint16_t *s, int sp, uint16_t *r, int rp,
+ uint32_t *sum_s, uint32_t *sum_r,
+ uint32_t *sum_sq_s, uint32_t *sum_sq_r,
+ uint32_t *sum_sxr) {
+ int i, j;
+ for (i = 0; i < 8; i++, s += sp, r += rp) {
+ for (j = 0; j < 8; j++) {
+ *sum_s += s[j];
+ *sum_r += r[j];
+ *sum_sq_s += s[j] * s[j];
+ *sum_sq_r += r[j] * r[j];
+ *sum_sxr += s[j] * r[j];
+ }
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static const int64_t cc1 = 26634; // (64^2*(.01*255)^2
static const int64_t cc2 = 239708; // (64^2*(.03*255)^2
@@ -73,6 +91,22 @@
return similarity(sum_s, sum_r, sum_sq_s, sum_sq_r, sum_sxr, 64);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static double highbd_ssim_8x8(uint16_t *s, int sp, uint16_t *r, int rp,
+ unsigned int bd) {
+ uint32_t sum_s = 0, sum_r = 0, sum_sq_s = 0, sum_sq_r = 0, sum_sxr = 0;
+ const int oshift = bd - 8;
+ vp9_highbd_ssim_parms_8x8(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r,
+ &sum_sxr);
+ return similarity(sum_s >> oshift,
+ sum_r >> oshift,
+ sum_sq_s >> (2 * oshift),
+ sum_sq_r >> (2 * oshift),
+ sum_sxr >> (2 * oshift),
+ 64);
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
// We are using a 8x8 moving window with starting location of each 8x8 window
// on the 4x4 pixel grid. Such arrangement allows the windows to overlap
// block boundaries to penalize blocking artifacts.
@@ -94,6 +128,31 @@
ssim_total /= samples;
return ssim_total;
}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+double vp9_highbd_ssim2(uint8_t *img1, uint8_t *img2, int stride_img1,
+ int stride_img2, int width, int height,
+ unsigned int bd) {
+ int i, j;
+ int samples = 0;
+ double ssim_total = 0;
+
+ // sample point start with each 4x4 location
+ for (i = 0; i <= height - 8;
+ i += 4, img1 += stride_img1 * 4, img2 += stride_img2 * 4) {
+ for (j = 0; j <= width - 8; j += 4) {
+ double v = highbd_ssim_8x8(CONVERT_TO_SHORTPTR(img1 + j), stride_img1,
+ CONVERT_TO_SHORTPTR(img2 + j), stride_img2,
+ bd);
+ ssim_total += v;
+ samples++;
+ }
+ }
+ ssim_total /= samples;
+ return ssim_total;
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
double vp9_calc_ssim(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest,
double *weight) {
double a, b, c;
@@ -141,3 +200,55 @@
return ssim_all;
}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+double vp9_highbd_calc_ssim(YV12_BUFFER_CONFIG *source,
+ YV12_BUFFER_CONFIG *dest,
+ double *weight, unsigned int bd) {
+ double a, b, c;
+ double ssimv;
+
+ a = vp9_highbd_ssim2(source->y_buffer, dest->y_buffer,
+ source->y_stride, dest->y_stride,
+ source->y_crop_width, source->y_crop_height, bd);
+
+ b = vp9_highbd_ssim2(source->u_buffer, dest->u_buffer,
+ source->uv_stride, dest->uv_stride,
+ source->uv_crop_width, source->uv_crop_height, bd);
+
+ c = vp9_highbd_ssim2(source->v_buffer, dest->v_buffer,
+ source->uv_stride, dest->uv_stride,
+ source->uv_crop_width, source->uv_crop_height, bd);
+
+ ssimv = a * .8 + .1 * (b + c);
+
+ *weight = 1;
+
+ return ssimv;
+}
+
+double vp9_highbd_calc_ssimg(YV12_BUFFER_CONFIG *source,
+ YV12_BUFFER_CONFIG *dest, double *ssim_y,
+ double *ssim_u, double *ssim_v, unsigned int bd) {
+ double ssim_all = 0;
+ double a, b, c;
+
+ a = vp9_highbd_ssim2(source->y_buffer, dest->y_buffer,
+ source->y_stride, dest->y_stride,
+ source->y_crop_width, source->y_crop_height, bd);
+
+ b = vp9_highbd_ssim2(source->u_buffer, dest->u_buffer,
+ source->uv_stride, dest->uv_stride,
+ source->uv_crop_width, source->uv_crop_height, bd);
+
+ c = vp9_highbd_ssim2(source->v_buffer, dest->v_buffer,
+ source->uv_stride, dest->uv_stride,
+ source->uv_crop_width, source->uv_crop_height, bd);
+ *ssim_y = a;
+ *ssim_u = b;
+ *ssim_v = c;
+ ssim_all = (a * 4 + b + c) / 6;
+
+ return ssim_all;
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/source/libvpx/vp9/encoder/vp9_ssim.h b/source/libvpx/vp9/encoder/vp9_ssim.h
index d1dd1b7..e75623b 100644
--- a/source/libvpx/vp9/encoder/vp9_ssim.h
+++ b/source/libvpx/vp9/encoder/vp9_ssim.h
@@ -23,6 +23,20 @@
double vp9_calc_ssimg(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest,
double *ssim_y, double *ssim_u, double *ssim_v);
+#if CONFIG_VP9_HIGHBITDEPTH
+double vp9_highbd_calc_ssim(YV12_BUFFER_CONFIG *source,
+ YV12_BUFFER_CONFIG *dest,
+ double *weight,
+ unsigned int bd);
+
+double vp9_highbd_calc_ssimg(YV12_BUFFER_CONFIG *source,
+ YV12_BUFFER_CONFIG *dest,
+ double *ssim_y,
+ double *ssim_u,
+ double *ssim_v,
+ unsigned int bd);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/source/libvpx/vp9/encoder/vp9_svc_layercontext.c b/source/libvpx/vp9/encoder/vp9_svc_layercontext.c
index eed681c..1573557 100644
--- a/source/libvpx/vp9/encoder/vp9_svc_layercontext.c
+++ b/source/libvpx/vp9/encoder/vp9_svc_layercontext.c
@@ -14,6 +14,8 @@
#include "vp9/encoder/vp9_svc_layercontext.h"
#include "vp9/encoder/vp9_extend.h"
+#define SMALL_FRAME_FB_IDX 7
+
void vp9_init_layer_context(VP9_COMP *const cpi) {
SVC *const svc = &cpi->svc;
const VP9EncoderConfig *const oxcf = &cpi->oxcf;
@@ -28,6 +30,25 @@
layer_end = svc->number_temporal_layers;
} else {
layer_end = svc->number_spatial_layers;
+
+ if (cpi->oxcf.error_resilient_mode == 0 && cpi->oxcf.pass == 2) {
+ if (vp9_realloc_frame_buffer(&cpi->svc.empty_frame.img,
+ cpi->common.width, cpi->common.height,
+ cpi->common.subsampling_x,
+ cpi->common.subsampling_y,
+#if CONFIG_VP9_HIGHBITDEPTH
+ cpi->common.use_highbitdepth,
+#endif
+ VP9_ENC_BORDER_IN_PIXELS, NULL, NULL, NULL))
+ vpx_internal_error(&cpi->common.error, VPX_CODEC_MEM_ERROR,
+ "Failed to allocate empty frame for multiple frame "
+ "contexts");
+
+ vpx_memset(cpi->svc.empty_frame.img.buffer_alloc, 0x80,
+ cpi->svc.empty_frame.img.buffer_alloc_sz);
+ cpi->svc.empty_frame_width = cpi->common.width;
+ cpi->svc.empty_frame_height = cpi->common.height;
+ }
}
for (layer = 0; layer < layer_end; ++layer) {
@@ -310,6 +331,47 @@
get_layer_resolution(cpi->oxcf.width, cpi->oxcf.height,
lc->scaling_factor_num, lc->scaling_factor_den,
&width, &height);
+
+ // Workaround for multiple frame contexts. In some frames we can't use prev_mi
+ // since its previous frame could be changed during decoding time. The idea is
+ // we put a empty invisible frame in front of them, then we will not use
+ // prev_mi when encoding these frames.
+ if (cpi->oxcf.error_resilient_mode == 0 && cpi->oxcf.pass == 2 &&
+ cpi->svc.encode_empty_frame_state == NEED_TO_ENCODE) {
+ if ((cpi->svc.number_temporal_layers > 1 &&
+ cpi->svc.temporal_layer_id < cpi->svc.number_temporal_layers - 1) ||
+ (cpi->svc.number_spatial_layers > 1 &&
+ cpi->svc.spatial_layer_id == 0)) {
+ struct lookahead_entry *buf = vp9_lookahead_peek(cpi->lookahead, 0);
+
+ if (buf != NULL) {
+ cpi->svc.empty_frame.ts_start = buf->ts_start;
+ cpi->svc.empty_frame.ts_end = buf->ts_end;
+ cpi->svc.encode_empty_frame_state = ENCODING;
+ cpi->common.show_frame = 0;
+ cpi->ref_frame_flags = 0;
+ cpi->common.frame_type = INTER_FRAME;
+ cpi->lst_fb_idx =
+ cpi->gld_fb_idx = cpi->alt_fb_idx = SMALL_FRAME_FB_IDX;
+
+ // Gradually make the empty frame smaller to save bits. Make it half of
+ // its previous size because of the scaling factor restriction.
+ cpi->svc.empty_frame_width >>= 1;
+ cpi->svc.empty_frame_width = (cpi->svc.empty_frame_width + 1) & ~1;
+ if (cpi->svc.empty_frame_width < 16)
+ cpi->svc.empty_frame_width = 16;
+
+ cpi->svc.empty_frame_height >>= 1;
+ cpi->svc.empty_frame_height = (cpi->svc.empty_frame_height + 1) & ~1;
+ if (cpi->svc.empty_frame_height < 16)
+ cpi->svc.empty_frame_height = 16;
+
+ width = cpi->svc.empty_frame_width;
+ height = cpi->svc.empty_frame_height;
+ }
+ }
+ }
+
if (vp9_set_size_literal(cpi, width, height) != 0)
return VPX_CODEC_INVALID_PARAM;
@@ -317,7 +379,6 @@
cpi->oxcf.best_allowed_q = vp9_quantizer_to_qindex(lc->min_q);
vp9_change_config(cpi, &cpi->oxcf);
-
vp9_set_high_precision_mv(cpi, 1);
cpi->alt_ref_source = get_layer_context(cpi)->alt_ref_source;
diff --git a/source/libvpx/vp9/encoder/vp9_svc_layercontext.h b/source/libvpx/vp9/encoder/vp9_svc_layercontext.h
index 47a5456..e9645ce 100644
--- a/source/libvpx/vp9/encoder/vp9_svc_layercontext.h
+++ b/source/libvpx/vp9/encoder/vp9_svc_layercontext.h
@@ -50,6 +50,16 @@
int spatial_layer_to_encode;
+ // Workaround for multiple frame contexts
+ enum {
+ ENCODED = 0,
+ ENCODING,
+ NEED_TO_ENCODE
+ }encode_empty_frame_state;
+ struct lookahead_entry empty_frame;
+ int empty_frame_width;
+ int empty_frame_height;
+
// Store scaled source frames to be used for temporal filter to generate
// a alt ref frame.
YV12_BUFFER_CONFIG scaled_frames[MAX_LAG_BUFFERS];
diff --git a/source/libvpx/vp9/encoder/vp9_temporal_filter.c b/source/libvpx/vp9/encoder/vp9_temporal_filter.c
index 6fd796d..5599227 100644
--- a/source/libvpx/vp9/encoder/vp9_temporal_filter.c
+++ b/source/libvpx/vp9/encoder/vp9_temporal_filter.c
@@ -56,6 +56,34 @@
mv_precision_uv = MV_PRECISION_Q3;
}
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ vp9_highbd_build_inter_predictor(y_mb_ptr, stride,
+ &pred[0], 16,
+ &mv,
+ scale,
+ 16, 16,
+ which_mv,
+ kernel, MV_PRECISION_Q3, x, y, xd->bd);
+
+ vp9_highbd_build_inter_predictor(u_mb_ptr, uv_stride,
+ &pred[256], uv_block_width,
+ &mv,
+ scale,
+ uv_block_width, uv_block_height,
+ which_mv,
+ kernel, mv_precision_uv, x, y, xd->bd);
+
+ vp9_highbd_build_inter_predictor(v_mb_ptr, uv_stride,
+ &pred[512], uv_block_width,
+ &mv,
+ scale,
+ uv_block_width, uv_block_height,
+ which_mv,
+ kernel, mv_precision_uv, x, y, xd->bd);
+ return;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
vp9_build_inter_predictor(y_mb_ptr, stride,
&pred[0], 16,
&mv,
@@ -133,6 +161,54 @@
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+void vp9_highbd_temporal_filter_apply_c(uint8_t *frame1_8,
+ unsigned int stride,
+ uint8_t *frame2_8,
+ unsigned int block_width,
+ unsigned int block_height,
+ int strength,
+ int filter_weight,
+ unsigned int *accumulator,
+ uint16_t *count) {
+ uint16_t *frame1 = CONVERT_TO_SHORTPTR(frame1_8);
+ uint16_t *frame2 = CONVERT_TO_SHORTPTR(frame2_8);
+ unsigned int i, j, k;
+ int modifier;
+ int byte = 0;
+ const int rounding = strength > 0 ? 1 << (strength - 1) : 0;
+
+ for (i = 0, k = 0; i < block_height; i++) {
+ for (j = 0; j < block_width; j++, k++) {
+ int src_byte = frame1[byte];
+ int pixel_value = *frame2++;
+
+ modifier = src_byte - pixel_value;
+ // This is an integer approximation of:
+ // float coeff = (3.0 * modifer * modifier) / pow(2, strength);
+ // modifier = (int)roundf(coeff > 16 ? 0 : 16-coeff);
+ modifier *= modifier;
+ modifier *= 3;
+ modifier += rounding;
+ modifier >>= strength;
+
+ if (modifier > 16)
+ modifier = 16;
+
+ modifier = 16 - modifier;
+ modifier *= filter_weight;
+
+ count[k] += modifier;
+ accumulator[k] += modifier * pixel_value;
+
+ byte++;
+ }
+
+ byte += stride - block_width;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
static int temporal_filter_find_matching_mb_c(VP9_COMP *cpi,
uint8_t *arf_frame_buf,
uint8_t *frame_ptr_buf,
@@ -145,7 +221,7 @@
int bestsme = INT_MAX;
int distortion;
unsigned int sse;
- int sad_list[5];
+ int cost_list[5];
MV best_ref_mv1 = {0, 0};
MV best_ref_mv1_full; /* full-pixel value of best_ref_mv1 */
@@ -169,7 +245,7 @@
// Ignore mv costing by sending NULL pointer instead of cost arrays
vp9_hex_search(x, &best_ref_mv1_full, step_param, sadpb, 1,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
&cpi->fn_ptr[BLOCK_16X16], 0, &best_ref_mv1, ref_mv);
// Ignore mv costing by sending NULL pointer instead of cost array
@@ -179,7 +255,7 @@
x->errorperbit,
&cpi->fn_ptr[BLOCK_16X16],
0, mv_sf->subpel_iters_per_step,
- cond_sad_list(cpi, sad_list),
+ cond_cost_list(cpi, cost_list),
NULL, NULL,
&distortion, &sse, NULL, 0, 0);
@@ -209,13 +285,26 @@
MACROBLOCKD *mbd = &cpi->mb.e_mbd;
YV12_BUFFER_CONFIG *f = frames[alt_ref_index];
uint8_t *dst1, *dst2;
+#if CONFIG_VP9_HIGHBITDEPTH
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, predictor16, 16 * 16 * 3);
+ DECLARE_ALIGNED_ARRAY(16, uint8_t, predictor8, 16 * 16 * 3);
+ uint8_t *predictor;
+#else
DECLARE_ALIGNED_ARRAY(16, uint8_t, predictor, 16 * 16 * 3);
+#endif
const int mb_uv_height = 16 >> mbd->plane[1].subsampling_y;
const int mb_uv_width = 16 >> mbd->plane[1].subsampling_x;
// Save input state
uint8_t* input_buffer[MAX_MB_PLANE];
int i;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (mbd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ predictor = CONVERT_TO_BYTEPTR(predictor16);
+ } else {
+ predictor = predictor8;
+ }
+#endif
for (i = 0; i < MAX_MB_PLANE; i++)
input_buffer[i] = mbd->plane[i].pre[0].buf;
@@ -286,6 +375,44 @@
predictor, scale,
mb_col * 16, mb_row * 16);
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (mbd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ int adj_strength = strength + 2 * (mbd->bd - 8);
+ // Apply the filter (YUV)
+ vp9_highbd_temporal_filter_apply(f->y_buffer + mb_y_offset,
+ f->y_stride,
+ predictor, 16, 16, adj_strength,
+ filter_weight,
+ accumulator, count);
+ vp9_highbd_temporal_filter_apply(f->u_buffer + mb_uv_offset,
+ f->uv_stride, predictor + 256,
+ mb_uv_width, mb_uv_height,
+ adj_strength,
+ filter_weight, accumulator + 256,
+ count + 256);
+ vp9_highbd_temporal_filter_apply(f->v_buffer + mb_uv_offset,
+ f->uv_stride, predictor + 512,
+ mb_uv_width, mb_uv_height,
+ adj_strength, filter_weight,
+ accumulator + 512, count + 512);
+ } else {
+ // Apply the filter (YUV)
+ vp9_temporal_filter_apply(f->y_buffer + mb_y_offset, f->y_stride,
+ predictor, 16, 16,
+ strength, filter_weight,
+ accumulator, count);
+ vp9_temporal_filter_apply(f->u_buffer + mb_uv_offset, f->uv_stride,
+ predictor + 256,
+ mb_uv_width, mb_uv_height, strength,
+ filter_weight, accumulator + 256,
+ count + 256);
+ vp9_temporal_filter_apply(f->v_buffer + mb_uv_offset, f->uv_stride,
+ predictor + 512,
+ mb_uv_width, mb_uv_height, strength,
+ filter_weight, accumulator + 512,
+ count + 512);
+ }
+#else
// Apply the filter (YUV)
vp9_temporal_filter_apply(f->y_buffer + mb_y_offset, f->y_stride,
predictor, 16, 16,
@@ -301,9 +428,108 @@
mb_uv_width, mb_uv_height, strength,
filter_weight, accumulator + 512,
count + 512);
+#endif // CONFIG_VP9_HIGHBITDEPTH
}
}
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (mbd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+ uint16_t *dst1_16;
+ uint16_t *dst2_16;
+ // Normalize filter output to produce AltRef frame
+ dst1 = cpi->alt_ref_buffer.y_buffer;
+ dst1_16 = CONVERT_TO_SHORTPTR(dst1);
+ stride = cpi->alt_ref_buffer.y_stride;
+ byte = mb_y_offset;
+ for (i = 0, k = 0; i < 16; i++) {
+ for (j = 0; j < 16; j++, k++) {
+ unsigned int pval = accumulator[k] + (count[k] >> 1);
+ pval *= fixed_divide[count[k]];
+ pval >>= 19;
+
+ dst1_16[byte] = (uint16_t)pval;
+
+ // move to next pixel
+ byte++;
+ }
+
+ byte += stride - 16;
+ }
+
+ dst1 = cpi->alt_ref_buffer.u_buffer;
+ dst2 = cpi->alt_ref_buffer.v_buffer;
+ dst1_16 = CONVERT_TO_SHORTPTR(dst1);
+ dst2_16 = CONVERT_TO_SHORTPTR(dst2);
+ stride = cpi->alt_ref_buffer.uv_stride;
+ byte = mb_uv_offset;
+ for (i = 0, k = 256; i < mb_uv_height; i++) {
+ for (j = 0; j < mb_uv_width; j++, k++) {
+ int m = k + 256;
+
+ // U
+ unsigned int pval = accumulator[k] + (count[k] >> 1);
+ pval *= fixed_divide[count[k]];
+ pval >>= 19;
+ dst1_16[byte] = (uint16_t)pval;
+
+ // V
+ pval = accumulator[m] + (count[m] >> 1);
+ pval *= fixed_divide[count[m]];
+ pval >>= 19;
+ dst2_16[byte] = (uint16_t)pval;
+
+ // move to next pixel
+ byte++;
+ }
+
+ byte += stride - mb_uv_width;
+ }
+ } else {
+ // Normalize filter output to produce AltRef frame
+ dst1 = cpi->alt_ref_buffer.y_buffer;
+ stride = cpi->alt_ref_buffer.y_stride;
+ byte = mb_y_offset;
+ for (i = 0, k = 0; i < 16; i++) {
+ for (j = 0; j < 16; j++, k++) {
+ unsigned int pval = accumulator[k] + (count[k] >> 1);
+ pval *= fixed_divide[count[k]];
+ pval >>= 19;
+
+ dst1[byte] = (uint8_t)pval;
+
+ // move to next pixel
+ byte++;
+ }
+ byte += stride - 16;
+ }
+
+ dst1 = cpi->alt_ref_buffer.u_buffer;
+ dst2 = cpi->alt_ref_buffer.v_buffer;
+ stride = cpi->alt_ref_buffer.uv_stride;
+ byte = mb_uv_offset;
+ for (i = 0, k = 256; i < mb_uv_height; i++) {
+ for (j = 0; j < mb_uv_width; j++, k++) {
+ int m = k + 256;
+
+ // U
+ unsigned int pval = accumulator[k] + (count[k] >> 1);
+ pval *= fixed_divide[count[k]];
+ pval >>= 19;
+ dst1[byte] = (uint8_t)pval;
+
+ // V
+ pval = accumulator[m] + (count[m] >> 1);
+ pval *= fixed_divide[count[m]];
+ pval >>= 19;
+ dst2[byte] = (uint8_t)pval;
+
+ // move to next pixel
+ byte++;
+ }
+ byte += stride - mb_uv_width;
+ }
+ }
+#else
// Normalize filter output to produce AltRef frame
dst1 = cpi->alt_ref_buffer.y_buffer;
stride = cpi->alt_ref_buffer.y_stride;
@@ -347,6 +573,7 @@
}
byte += stride - mb_uv_width;
}
+#endif // CONFIG_VP9_HIGHBITDEPTH
mb_y_offset += 16;
mb_uv_offset += mb_uv_width;
}
@@ -449,60 +676,69 @@
frames[frames_to_blur - 1 - frame] = &buf->img;
}
- // Setup scaling factors. Scaling on each of the arnr frames is not supported
- if (is_two_pass_svc(cpi)) {
- // In spatial svc the scaling factors might be less then 1/2. So we will use
- // non-normative scaling.
- int frame_used = 0;
+ if (frames_to_blur > 0) {
+ // Setup scaling factors. Scaling on each of the arnr frames is not
+ // supported.
+ if (is_two_pass_svc(cpi)) {
+ // In spatial svc the scaling factors might be less then 1/2.
+ // So we will use non-normative scaling.
+ int frame_used = 0;
#if CONFIG_VP9_HIGHBITDEPTH
- vp9_setup_scale_factors_for_frame(&sf,
- get_frame_new_buffer(cm)->y_crop_width,
- get_frame_new_buffer(cm)->y_crop_height,
- get_frame_new_buffer(cm)->y_crop_width,
- get_frame_new_buffer(cm)->y_crop_height,
- cm->use_highbitdepth);
+ vp9_setup_scale_factors_for_frame(
+ &sf,
+ get_frame_new_buffer(cm)->y_crop_width,
+ get_frame_new_buffer(cm)->y_crop_height,
+ get_frame_new_buffer(cm)->y_crop_width,
+ get_frame_new_buffer(cm)->y_crop_height,
+ cm->use_highbitdepth);
#else
- vp9_setup_scale_factors_for_frame(&sf,
- get_frame_new_buffer(cm)->y_crop_width,
- get_frame_new_buffer(cm)->y_crop_height,
- get_frame_new_buffer(cm)->y_crop_width,
- get_frame_new_buffer(cm)->y_crop_height);
-#endif
- for (frame = 0; frame < frames_to_blur; ++frame) {
- if (cm->mi_cols * MI_SIZE != frames[frame]->y_width ||
- cm->mi_rows * MI_SIZE != frames[frame]->y_height) {
- if (vp9_realloc_frame_buffer(&cpi->svc.scaled_frames[frame_used],
- cm->width, cm->height,
- cm->subsampling_x, cm->subsampling_y,
-#if CONFIG_VP9_HIGHBITDEPTH
- cm->use_highbitdepth,
-#endif
- VP9_ENC_BORDER_IN_PIXELS, NULL, NULL,
- NULL))
- vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR,
- "Failed to reallocate alt_ref_buffer");
+ vp9_setup_scale_factors_for_frame(
+ &sf,
+ get_frame_new_buffer(cm)->y_crop_width,
+ get_frame_new_buffer(cm)->y_crop_height,
+ get_frame_new_buffer(cm)->y_crop_width,
+ get_frame_new_buffer(cm)->y_crop_height);
+#endif // CONFIG_VP9_HIGHBITDEPTH
- frames[frame] = vp9_scale_if_required(cm, frames[frame],
- &cpi->svc.scaled_frames[frame_used]);
- ++frame_used;
- }
- }
- } else {
- // ARF is produced at the native frame size and resized when coded.
+ for (frame = 0; frame < frames_to_blur; ++frame) {
+ if (cm->mi_cols * MI_SIZE != frames[frame]->y_width ||
+ cm->mi_rows * MI_SIZE != frames[frame]->y_height) {
+ if (vp9_realloc_frame_buffer(&cpi->svc.scaled_frames[frame_used],
+ cm->width, cm->height,
+ cm->subsampling_x, cm->subsampling_y,
#if CONFIG_VP9_HIGHBITDEPTH
- vp9_setup_scale_factors_for_frame(&sf,
- frames[0]->y_crop_width,
- frames[0]->y_crop_height,
- frames[0]->y_crop_width,
- frames[0]->y_crop_height,
- cm->use_highbitdepth);
-#else
- vp9_setup_scale_factors_for_frame(&sf,
- frames[0]->y_crop_width,
- frames[0]->y_crop_height,
- frames[0]->y_crop_width,
- frames[0]->y_crop_height);
+ cm->use_highbitdepth,
#endif
+ VP9_ENC_BORDER_IN_PIXELS, NULL, NULL,
+ NULL)) {
+ vpx_internal_error(&cm->error, VPX_CODEC_MEM_ERROR,
+ "Failed to reallocate alt_ref_buffer");
+ }
+ frames[frame] = vp9_scale_if_required(
+ cm, frames[frame], &cpi->svc.scaled_frames[frame_used]);
+ ++frame_used;
+ }
+ }
+ cm->mi = cm->mip + cm->mi_stride + 1;
+ cpi->mb.e_mbd.mi = cm->mi;
+ cpi->mb.e_mbd.mi[0].src_mi = &cpi->mb.e_mbd.mi[0];
+ } else {
+ // ARF is produced at the native frame size and resized when coded.
+#if CONFIG_VP9_HIGHBITDEPTH
+ vp9_setup_scale_factors_for_frame(&sf,
+ frames[0]->y_crop_width,
+ frames[0]->y_crop_height,
+ frames[0]->y_crop_width,
+ frames[0]->y_crop_height,
+ cm->use_highbitdepth);
+#else
+ vp9_setup_scale_factors_for_frame(&sf,
+ frames[0]->y_crop_width,
+ frames[0]->y_crop_height,
+ frames[0]->y_crop_width,
+ frames[0]->y_crop_height);
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ }
}
temporal_filter_iterate_c(cpi, frames, frames_to_blur,
diff --git a/source/libvpx/vp9/encoder/vp9_tokenize.c b/source/libvpx/vp9/encoder/vp9_tokenize.c
index 8b9aa91..adf01bf 100644
--- a/source/libvpx/vp9/encoder/vp9_tokenize.c
+++ b/source/libvpx/vp9/encoder/vp9_tokenize.c
@@ -261,7 +261,7 @@
}
static INLINE void add_token(TOKENEXTRA **t, const vp9_prob *context_tree,
- int16_t extra, uint8_t token,
+ int32_t extra, uint8_t token,
uint8_t skip_eob_node,
unsigned int *counts) {
(*t)->token = token;
@@ -329,7 +329,7 @@
scan = so->scan;
nb = so->neighbors;
c = 0;
-#if CONFIG_VP9_HIGH && CONFIG_HIGH_QUANT
+#if CONFIG_VP9_HIGHBITDEPTH
if (cpi->common.profile >= PROFILE_2) {
dct_value_tokens = (cpi->common.bit_depth == VPX_BITS_10 ?
vp9_dct_value_tokens_high10_ptr :
@@ -403,6 +403,24 @@
return result;
}
+static void has_high_freq_coeff(int plane, int block,
+ BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
+ void *argv) {
+ struct is_skippable_args *args = argv;
+ int eobs = (tx_size == TX_4X4) ? 3 : 10;
+ (void) plane_bsize;
+
+ *(args->skippable) |= (args->x->plane[plane].eobs[block] > eobs);
+}
+
+int vp9_has_high_freq_in_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
+ int result = 0;
+ struct is_skippable_args args = {x, &result};
+ vp9_foreach_transformed_block_in_plane(&x->e_mbd, bsize, plane,
+ has_high_freq_coeff, &args);
+ return result;
+}
+
void vp9_tokenize_sb(VP9_COMP *cpi, TOKENEXTRA **t, int dry_run,
BLOCK_SIZE bsize) {
VP9_COMMON *const cm = &cpi->common;
diff --git a/source/libvpx/vp9/encoder/vp9_tokenize.h b/source/libvpx/vp9/encoder/vp9_tokenize.h
index 063c0ba..825252b 100644
--- a/source/libvpx/vp9/encoder/vp9_tokenize.h
+++ b/source/libvpx/vp9/encoder/vp9_tokenize.h
@@ -26,12 +26,20 @@
typedef struct {
int16_t token;
+#if CONFIG_VP9_HIGHBITDEPTH
+ int32_t extra;
+#else
int16_t extra;
+#endif
} TOKENVALUE;
typedef struct {
const vp9_prob *context_tree;
+#if CONFIG_VP9_HIGHBITDEPTH
+ int32_t extra;
+#else
int16_t extra;
+#endif
uint8_t token;
uint8_t skip_eob_node;
} TOKENEXTRA;
@@ -41,6 +49,7 @@
extern struct vp9_token vp9_coef_encodings[];
int vp9_is_skippable_in_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane);
+int vp9_has_high_freq_in_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane);
struct VP9_COMP;
@@ -53,6 +62,12 @@
* fields are not.
*/
extern const TOKENVALUE *vp9_dct_value_tokens_ptr;
+#if CONFIG_VP9_HIGHBITDEPTH
+extern const int16_t *vp9_dct_value_cost_high10_ptr;
+extern const TOKENVALUE *vp9_dct_value_tokens_high10_ptr;
+extern const int16_t *vp9_dct_value_cost_high12_ptr;
+extern const TOKENVALUE *vp9_dct_value_tokens_high12_ptr;
+#endif // CONFIG_VP9_HIGHBITDEPTH
#ifdef __cplusplus
} // extern "C"
diff --git a/source/libvpx/vp9/encoder/vp9_variance.c b/source/libvpx/vp9/encoder/vp9_variance.c
index c97f93f..4555bde 100644
--- a/source/libvpx/vp9/encoder/vp9_variance.c
+++ b/source/libvpx/vp9/encoder/vp9_variance.c
@@ -269,10 +269,10 @@
}
#if CONFIG_VP9_HIGHBITDEPTH
-void high_variance64(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h, uint64_t *sse,
- uint64_t *sum) {
+void highbd_variance64(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int w, int h, uint64_t *sse,
+ uint64_t *sum) {
int i, j;
uint16_t *a = CONVERT_TO_SHORTPTR(a8);
@@ -291,40 +291,40 @@
}
}
-void high_variance(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h, unsigned int *sse,
- int *sum) {
+void highbd_variance(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int w, int h, unsigned int *sse,
+ int *sum) {
uint64_t sse_long = 0;
uint64_t sum_long = 0;
- high_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
+ highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
*sse = sse_long;
*sum = sum_long;
}
-void high_10_variance(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h, unsigned int *sse,
- int *sum) {
+void highbd_10_variance(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int w, int h, unsigned int *sse,
+ int *sum) {
uint64_t sse_long = 0;
uint64_t sum_long = 0;
- high_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
+ highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
*sum = ROUND_POWER_OF_TWO(sum_long, 2);
*sse = ROUND_POWER_OF_TWO(sse_long, 4);
}
-void high_12_variance(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h, unsigned int *sse,
- int *sum) {
+void highbd_12_variance(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int w, int h, unsigned int *sse,
+ int *sum) {
uint64_t sse_long = 0;
uint64_t sum_long = 0;
- high_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
+ highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
*sum = ROUND_POWER_OF_TWO(sum_long, 4);
*sse = ROUND_POWER_OF_TWO(sse_long, 8);
}
-static void high_var_filter_block2d_bil_first_pass(
+static void highbd_var_filter_block2d_bil_first_pass(
const uint8_t *src_ptr8,
uint16_t *output_ptr,
unsigned int src_pixels_per_line,
@@ -350,7 +350,7 @@
}
}
-static void high_var_filter_block2d_bil_second_pass(
+static void highbd_var_filter_block2d_bil_second_pass(
const uint16_t *src_ptr,
uint16_t *output_ptr,
unsigned int src_pixels_per_line,
@@ -374,257 +374,267 @@
}
}
-#define HIGH_VAR(W, H) \
-unsigned int vp9_high_variance##W##x##H##_c(const uint8_t *a, int a_stride, \
- const uint8_t *b, int b_stride, \
+#define HIGHBD_VAR(W, H) \
+unsigned int vp9_highbd_variance##W##x##H##_c(const uint8_t *a, int a_stride, \
+ const uint8_t *b, int b_stride, \
+ unsigned int *sse) { \
+ int sum; \
+ highbd_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \
+ return *sse - (((int64_t)sum * sum) / (W * H)); \
+} \
+\
+unsigned int vp9_highbd_10_variance##W##x##H##_c(const uint8_t *a, \
+ int a_stride, \
+ const uint8_t *b, \
+ int b_stride, \
+ unsigned int *sse) { \
+ int sum; \
+ highbd_10_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \
+ return *sse - (((int64_t)sum * sum) / (W * H)); \
+} \
+\
+unsigned int vp9_highbd_12_variance##W##x##H##_c(const uint8_t *a, \
+ int a_stride, \
+ const uint8_t *b, \
+ int b_stride, \
+ unsigned int *sse) { \
+ int sum; \
+ highbd_12_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \
+ return *sse - (((int64_t)sum * sum) / (W * H)); \
+}
+
+#define HIGHBD_SUBPIX_VAR(W, H) \
+unsigned int vp9_highbd_sub_pixel_variance##W##x##H##_c( \
+ const uint8_t *src, int src_stride, \
+ int xoffset, int yoffset, \
+ const uint8_t *dst, int dst_stride, \
+ unsigned int *sse) { \
+ uint16_t fdata3[(H + 1) * W]; \
+ uint16_t temp2[H * W]; \
+\
+ highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
+ W, BILINEAR_FILTERS_2TAP(xoffset)); \
+ highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
+ BILINEAR_FILTERS_2TAP(yoffset)); \
+\
+ return vp9_highbd_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), W, dst, \
+ dst_stride, sse); \
+} \
+\
+unsigned int vp9_highbd_10_sub_pixel_variance##W##x##H##_c( \
+ const uint8_t *src, int src_stride, \
+ int xoffset, int yoffset, \
+ const uint8_t *dst, int dst_stride, \
+ unsigned int *sse) { \
+ uint16_t fdata3[(H + 1) * W]; \
+ uint16_t temp2[H * W]; \
+\
+ highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
+ W, BILINEAR_FILTERS_2TAP(xoffset)); \
+ highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
+ BILINEAR_FILTERS_2TAP(yoffset)); \
+\
+ return vp9_highbd_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), \
+ W, dst, dst_stride, sse); \
+} \
+\
+unsigned int vp9_highbd_12_sub_pixel_variance##W##x##H##_c( \
+ const uint8_t *src, int src_stride, \
+ int xoffset, int yoffset, \
+ const uint8_t *dst, int dst_stride, \
+ unsigned int *sse) { \
+ uint16_t fdata3[(H + 1) * W]; \
+ uint16_t temp2[H * W]; \
+\
+ highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
+ W, BILINEAR_FILTERS_2TAP(xoffset)); \
+ highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
+ BILINEAR_FILTERS_2TAP(yoffset)); \
+\
+ return vp9_highbd_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), \
+ W, dst, dst_stride, sse); \
+}
+
+#define HIGHBD_SUBPIX_AVG_VAR(W, H) \
+unsigned int vp9_highbd_sub_pixel_avg_variance##W##x##H##_c( \
+ const uint8_t *src, int src_stride, \
+ int xoffset, int yoffset, \
+ const uint8_t *dst, int dst_stride, \
+ unsigned int *sse, \
+ const uint8_t *second_pred) { \
+ uint16_t fdata3[(H + 1) * W]; \
+ uint16_t temp2[H * W]; \
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \
+\
+ highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
+ W, BILINEAR_FILTERS_2TAP(xoffset)); \
+ highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
+ BILINEAR_FILTERS_2TAP(yoffset)); \
+\
+ vp9_highbd_comp_avg_pred(temp3, second_pred, W, H, \
+ CONVERT_TO_BYTEPTR(temp2), W); \
+\
+ return vp9_highbd_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, dst, \
+ dst_stride, sse); \
+} \
+\
+unsigned int vp9_highbd_10_sub_pixel_avg_variance##W##x##H##_c( \
+ const uint8_t *src, int src_stride, \
+ int xoffset, int yoffset, \
+ const uint8_t *dst, int dst_stride, \
+ unsigned int *sse, \
+ const uint8_t *second_pred) { \
+ uint16_t fdata3[(H + 1) * W]; \
+ uint16_t temp2[H * W]; \
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \
+\
+ highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
+ W, BILINEAR_FILTERS_2TAP(xoffset)); \
+ highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
+ BILINEAR_FILTERS_2TAP(yoffset)); \
+\
+ vp9_highbd_comp_avg_pred(temp3, second_pred, W, H, \
+ CONVERT_TO_BYTEPTR(temp2), W); \
+\
+ return vp9_highbd_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), \
+ W, dst, dst_stride, sse); \
+} \
+\
+unsigned int vp9_highbd_12_sub_pixel_avg_variance##W##x##H##_c( \
+ const uint8_t *src, int src_stride, \
+ int xoffset, int yoffset, \
+ const uint8_t *dst, int dst_stride, \
+ unsigned int *sse, \
+ const uint8_t *second_pred) { \
+ uint16_t fdata3[(H + 1) * W]; \
+ uint16_t temp2[H * W]; \
+ DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \
+\
+ highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
+ W, BILINEAR_FILTERS_2TAP(xoffset)); \
+ highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
+ BILINEAR_FILTERS_2TAP(yoffset)); \
+\
+ vp9_highbd_comp_avg_pred(temp3, second_pred, W, H, \
+ CONVERT_TO_BYTEPTR(temp2), W); \
+\
+ return vp9_highbd_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), \
+ W, dst, dst_stride, sse); \
+}
+
+#define HIGHBD_GET_VAR(S) \
+void vp9_highbd_get##S##x##S##var_c(const uint8_t *src, int src_stride, \
+ const uint8_t *ref, int ref_stride, \
+ unsigned int *sse, int *sum) { \
+ highbd_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \
+} \
+\
+void vp9_highbd_10_get##S##x##S##var_c(const uint8_t *src, int src_stride, \
+ const uint8_t *ref, int ref_stride, \
+ unsigned int *sse, int *sum) { \
+ highbd_10_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \
+} \
+\
+void vp9_highbd_12_get##S##x##S##var_c(const uint8_t *src, int src_stride, \
+ const uint8_t *ref, int ref_stride, \
+ unsigned int *sse, int *sum) { \
+ highbd_12_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \
+}
+
+#define HIGHBD_MSE(W, H) \
+unsigned int vp9_highbd_mse##W##x##H##_c(const uint8_t *src, \
+ int src_stride, \
+ const uint8_t *ref, \
+ int ref_stride, \
+ unsigned int *sse) { \
+ int sum; \
+ highbd_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \
+ return *sse; \
+} \
+\
+unsigned int vp9_highbd_10_mse##W##x##H##_c(const uint8_t *src, \
+ int src_stride, \
+ const uint8_t *ref, \
+ int ref_stride, \
unsigned int *sse) { \
int sum; \
- high_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \
- return *sse - (((int64_t)sum * sum) / (W * H)); \
-} \
-\
-unsigned int vp9_high_10_variance##W##x##H##_c(const uint8_t *a, int a_stride, \
- const uint8_t *b, int b_stride, \
- unsigned int *sse) { \
- int sum; \
- high_10_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \
- return *sse - (((int64_t)sum * sum) / (W * H)); \
-} \
-\
-unsigned int vp9_high_12_variance##W##x##H##_c(const uint8_t *a, int a_stride, \
- const uint8_t *b, int b_stride, \
- unsigned int *sse) { \
- int sum; \
- high_12_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \
- return *sse - (((int64_t)sum * sum) / (W * H)); \
-}
-
-#define HIGH_SUBPIX_VAR(W, H) \
-unsigned int vp9_high_sub_pixel_variance##W##x##H##_c( \
- const uint8_t *src, int src_stride, \
- int xoffset, int yoffset, \
- const uint8_t *dst, int dst_stride, \
- unsigned int *sse) { \
- uint16_t fdata3[(H + 1) * W]; \
- uint16_t temp2[H * W]; \
-\
- high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
- W, BILINEAR_FILTERS_2TAP(xoffset)); \
- high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
- BILINEAR_FILTERS_2TAP(yoffset)); \
-\
- return vp9_high_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), W, dst, \
- dst_stride, sse); \
-} \
-\
-unsigned int vp9_high_10_sub_pixel_variance##W##x##H##_c( \
- const uint8_t *src, int src_stride, \
- int xoffset, int yoffset, \
- const uint8_t *dst, int dst_stride, \
- unsigned int *sse) { \
- uint16_t fdata3[(H + 1) * W]; \
- uint16_t temp2[H * W]; \
-\
- high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
- W, BILINEAR_FILTERS_2TAP(xoffset)); \
- high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
- BILINEAR_FILTERS_2TAP(yoffset)); \
-\
- return vp9_high_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), W, dst, \
- dst_stride, sse); \
-} \
-\
-unsigned int vp9_high_12_sub_pixel_variance##W##x##H##_c( \
- const uint8_t *src, int src_stride, \
- int xoffset, int yoffset, \
- const uint8_t *dst, int dst_stride, \
- unsigned int *sse) { \
- uint16_t fdata3[(H + 1) * W]; \
- uint16_t temp2[H * W]; \
-\
- high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
- W, BILINEAR_FILTERS_2TAP(xoffset)); \
- high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
- BILINEAR_FILTERS_2TAP(yoffset)); \
-\
- return vp9_high_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), W, dst, \
- dst_stride, sse); \
-}
-
-#define HIGH_SUBPIX_AVG_VAR(W, H) \
-unsigned int vp9_high_sub_pixel_avg_variance##W##x##H##_c( \
- const uint8_t *src, int src_stride, \
- int xoffset, int yoffset, \
- const uint8_t *dst, int dst_stride, \
- unsigned int *sse, \
- const uint8_t *second_pred) { \
- uint16_t fdata3[(H + 1) * W]; \
- uint16_t temp2[H * W]; \
- DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \
-\
- high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
- W, BILINEAR_FILTERS_2TAP(xoffset)); \
- high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
- BILINEAR_FILTERS_2TAP(yoffset)); \
-\
- vp9_high_comp_avg_pred(temp3, second_pred, W, H, CONVERT_TO_BYTEPTR(temp2), \
- W); \
-\
- return vp9_high_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, dst, \
- dst_stride, sse); \
-} \
-\
-unsigned int vp9_high_10_sub_pixel_avg_variance##W##x##H##_c( \
- const uint8_t *src, int src_stride, \
- int xoffset, int yoffset, \
- const uint8_t *dst, int dst_stride, \
- unsigned int *sse, \
- const uint8_t *second_pred) { \
- uint16_t fdata3[(H + 1) * W]; \
- uint16_t temp2[H * W]; \
- DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \
-\
- high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
- W, BILINEAR_FILTERS_2TAP(xoffset)); \
- high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
- BILINEAR_FILTERS_2TAP(yoffset)); \
-\
- vp9_high_comp_avg_pred(temp3, second_pred, W, H, CONVERT_TO_BYTEPTR(temp2), \
- W); \
-\
- return vp9_high_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, dst, \
- dst_stride, sse); \
-} \
-\
-unsigned int vp9_high_12_sub_pixel_avg_variance##W##x##H##_c( \
- const uint8_t *src, int src_stride, \
- int xoffset, int yoffset, \
- const uint8_t *dst, int dst_stride, \
- unsigned int *sse, \
- const uint8_t *second_pred) { \
- uint16_t fdata3[(H + 1) * W]; \
- uint16_t temp2[H * W]; \
- DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \
-\
- high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
- W, BILINEAR_FILTERS_2TAP(xoffset)); \
- high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
- BILINEAR_FILTERS_2TAP(yoffset)); \
-\
- vp9_high_comp_avg_pred(temp3, second_pred, W, H, CONVERT_TO_BYTEPTR(temp2), \
- W); \
-\
- return vp9_high_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, dst, \
- dst_stride, sse); \
-}
-
-#define HIGH_GET_VAR(S) \
-void vp9_high_get##S##x##S##var_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride, \
- unsigned int *sse, int *sum) { \
- high_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \
-} \
-\
-void vp9_high_10_get##S##x##S##var_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride, \
- unsigned int *sse, int *sum) { \
- high_10_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \
-} \
-\
-void vp9_high_12_get##S##x##S##var_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride, \
- unsigned int *sse, int *sum) { \
- high_12_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \
-}
-
-#define HIGH_MSE(W, H) \
-unsigned int vp9_high_mse##W##x##H##_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride, \
- unsigned int *sse) { \
- int sum; \
- high_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \
+ highbd_10_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \
return *sse; \
} \
\
-unsigned int vp9_high_10_mse##W##x##H##_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride, \
- unsigned int *sse) { \
+unsigned int vp9_highbd_12_mse##W##x##H##_c(const uint8_t *src, \
+ int src_stride, \
+ const uint8_t *ref, \
+ int ref_stride, \
+ unsigned int *sse) { \
int sum; \
- high_10_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \
- return *sse; \
-} \
-\
-unsigned int vp9_high_12_mse##W##x##H##_c(const uint8_t *src, int src_stride, \
- const uint8_t *ref, int ref_stride, \
- unsigned int *sse) { \
- int sum; \
- high_12_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \
+ highbd_12_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \
return *sse; \
}
-HIGH_GET_VAR(8)
-HIGH_GET_VAR(16)
+HIGHBD_GET_VAR(8)
+HIGHBD_GET_VAR(16)
-HIGH_MSE(16, 16)
-HIGH_MSE(16, 8)
-HIGH_MSE(8, 16)
-HIGH_MSE(8, 8)
+HIGHBD_MSE(16, 16)
+HIGHBD_MSE(16, 8)
+HIGHBD_MSE(8, 16)
+HIGHBD_MSE(8, 8)
-HIGH_VAR(4, 4)
-HIGH_SUBPIX_VAR(4, 4)
-HIGH_SUBPIX_AVG_VAR(4, 4)
+HIGHBD_VAR(4, 4)
+HIGHBD_SUBPIX_VAR(4, 4)
+HIGHBD_SUBPIX_AVG_VAR(4, 4)
-HIGH_VAR(4, 8)
-HIGH_SUBPIX_VAR(4, 8)
-HIGH_SUBPIX_AVG_VAR(4, 8)
+HIGHBD_VAR(4, 8)
+HIGHBD_SUBPIX_VAR(4, 8)
+HIGHBD_SUBPIX_AVG_VAR(4, 8)
-HIGH_VAR(8, 4)
-HIGH_SUBPIX_VAR(8, 4)
-HIGH_SUBPIX_AVG_VAR(8, 4)
+HIGHBD_VAR(8, 4)
+HIGHBD_SUBPIX_VAR(8, 4)
+HIGHBD_SUBPIX_AVG_VAR(8, 4)
-HIGH_VAR(8, 8)
-HIGH_SUBPIX_VAR(8, 8)
-HIGH_SUBPIX_AVG_VAR(8, 8)
+HIGHBD_VAR(8, 8)
+HIGHBD_SUBPIX_VAR(8, 8)
+HIGHBD_SUBPIX_AVG_VAR(8, 8)
-HIGH_VAR(8, 16)
-HIGH_SUBPIX_VAR(8, 16)
-HIGH_SUBPIX_AVG_VAR(8, 16)
+HIGHBD_VAR(8, 16)
+HIGHBD_SUBPIX_VAR(8, 16)
+HIGHBD_SUBPIX_AVG_VAR(8, 16)
-HIGH_VAR(16, 8)
-HIGH_SUBPIX_VAR(16, 8)
-HIGH_SUBPIX_AVG_VAR(16, 8)
+HIGHBD_VAR(16, 8)
+HIGHBD_SUBPIX_VAR(16, 8)
+HIGHBD_SUBPIX_AVG_VAR(16, 8)
-HIGH_VAR(16, 16)
-HIGH_SUBPIX_VAR(16, 16)
-HIGH_SUBPIX_AVG_VAR(16, 16)
+HIGHBD_VAR(16, 16)
+HIGHBD_SUBPIX_VAR(16, 16)
+HIGHBD_SUBPIX_AVG_VAR(16, 16)
-HIGH_VAR(16, 32)
-HIGH_SUBPIX_VAR(16, 32)
-HIGH_SUBPIX_AVG_VAR(16, 32)
+HIGHBD_VAR(16, 32)
+HIGHBD_SUBPIX_VAR(16, 32)
+HIGHBD_SUBPIX_AVG_VAR(16, 32)
-HIGH_VAR(32, 16)
-HIGH_SUBPIX_VAR(32, 16)
-HIGH_SUBPIX_AVG_VAR(32, 16)
+HIGHBD_VAR(32, 16)
+HIGHBD_SUBPIX_VAR(32, 16)
+HIGHBD_SUBPIX_AVG_VAR(32, 16)
-HIGH_VAR(32, 32)
-HIGH_SUBPIX_VAR(32, 32)
-HIGH_SUBPIX_AVG_VAR(32, 32)
+HIGHBD_VAR(32, 32)
+HIGHBD_SUBPIX_VAR(32, 32)
+HIGHBD_SUBPIX_AVG_VAR(32, 32)
-HIGH_VAR(32, 64)
-HIGH_SUBPIX_VAR(32, 64)
-HIGH_SUBPIX_AVG_VAR(32, 64)
+HIGHBD_VAR(32, 64)
+HIGHBD_SUBPIX_VAR(32, 64)
+HIGHBD_SUBPIX_AVG_VAR(32, 64)
-HIGH_VAR(64, 32)
-HIGH_SUBPIX_VAR(64, 32)
-HIGH_SUBPIX_AVG_VAR(64, 32)
+HIGHBD_VAR(64, 32)
+HIGHBD_SUBPIX_VAR(64, 32)
+HIGHBD_SUBPIX_AVG_VAR(64, 32)
-HIGH_VAR(64, 64)
-HIGH_SUBPIX_VAR(64, 64)
-HIGH_SUBPIX_AVG_VAR(64, 64)
+HIGHBD_VAR(64, 64)
+HIGHBD_SUBPIX_VAR(64, 64)
+HIGHBD_SUBPIX_AVG_VAR(64, 64)
-void vp9_high_comp_avg_pred(uint16_t *comp_pred, const uint8_t *pred8,
- int width, int height, const uint8_t *ref8,
- int ref_stride) {
+void vp9_highbd_comp_avg_pred(uint16_t *comp_pred, const uint8_t *pred8,
+ int width, int height, const uint8_t *ref8,
+ int ref_stride) {
int i, j;
uint16_t *pred = CONVERT_TO_SHORTPTR(pred8);
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
diff --git a/source/libvpx/vp9/encoder/vp9_variance.h b/source/libvpx/vp9/encoder/vp9_variance.h
index c51d08d..53148f2 100644
--- a/source/libvpx/vp9/encoder/vp9_variance.h
+++ b/source/libvpx/vp9/encoder/vp9_variance.h
@@ -23,20 +23,20 @@
unsigned int *sse, int *sum);
#if CONFIG_VP9_HIGHBITDEPTH
-void high_variance(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h,
- unsigned int *sse, int *sum);
+void highbd_variance(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int w, int h,
+ unsigned int *sse, int *sum);
-void high_10_variance(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h,
- unsigned int *sse, int *sum);
+void highbd_10_variance(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int w, int h,
+ unsigned int *sse, int *sum);
-void high_12_variance(const uint8_t *a8, int a_stride,
- const uint8_t *b8, int b_stride,
- int w, int h,
- unsigned int *sse, int *sum);
+void highbd_12_variance(const uint8_t *a8, int a_stride,
+ const uint8_t *b8, int b_stride,
+ int w, int h,
+ unsigned int *sse, int *sum);
#endif
typedef unsigned int(*vp9_sad_fn_t)(const uint8_t *src_ptr,
@@ -99,8 +99,9 @@
int height, const uint8_t *ref, int ref_stride);
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_high_comp_avg_pred(uint16_t *comp_pred, const uint8_t *pred, int width,
- int height, const uint8_t *ref, int ref_stride);
+void vp9_highbd_comp_avg_pred(uint16_t *comp_pred, const uint8_t *pred,
+ int width, int height,
+ const uint8_t *ref, int ref_stride);
#endif
#ifdef __cplusplus
diff --git a/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c b/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c
new file mode 100644
index 0000000..ca6cf1a
--- /dev/null
+++ b/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c
@@ -0,0 +1,40 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 <emmintrin.h>
+#include "vpx_ports/mem.h"
+
+
+unsigned int vp9_avg_8x8_sse2(const uint8_t *s, int p) {
+ __m128i s0, s1, u0;
+ unsigned int avg = 0;
+ u0 = _mm_setzero_si128();
+ s0 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(s)), u0);
+ s1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(s + p)), u0);
+ s0 = _mm_adds_epu16(s0, s1);
+ s1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(s + 2 * p)), u0);
+ s0 = _mm_adds_epu16(s0, s1);
+ s1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(s + 3 * p)), u0);
+ s0 = _mm_adds_epu16(s0, s1);
+ s1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(s + 4 * p)), u0);
+ s0 = _mm_adds_epu16(s0, s1);
+ s1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(s + 5 * p)), u0);
+ s0 = _mm_adds_epu16(s0, s1);
+ s1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(s + 6 * p)), u0);
+ s0 = _mm_adds_epu16(s0, s1);
+ s1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(s + 7 * p)), u0);
+ s0 = _mm_adds_epu16(s0, s1);
+
+ s0 = _mm_adds_epu16(s0, _mm_srli_si128(s0, 8));
+ s0 = _mm_adds_epu16(s0, _mm_srli_epi64(s0, 32));
+ s0 = _mm_adds_epu16(s0, _mm_srli_epi64(s0, 16));
+ avg = _mm_extract_epi16(s0, 0);
+ return (avg + 32) >> 6;
+}
diff --git a/source/libvpx/vp9/encoder/x86/vp9_denoiser_sse2.c b/source/libvpx/vp9/encoder/x86/vp9_denoiser_sse2.c
new file mode 100644
index 0000000..bf5fa88
--- /dev/null
+++ b/source/libvpx/vp9/encoder/x86/vp9_denoiser_sse2.c
@@ -0,0 +1,474 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 <emmintrin.h>
+
+#include "./vpx_config.h"
+#include "./vp9_rtcd.h"
+
+#include "vpx_ports/emmintrin_compat.h"
+#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_reconinter.h"
+#include "vp9/encoder/vp9_context_tree.h"
+#include "vp9/encoder/vp9_denoiser.h"
+#include "vpx_mem/vpx_mem.h"
+
+// Compute the sum of all pixel differences of this MB.
+static INLINE int sum_diff_16x1(__m128i acc_diff) {
+ const __m128i k_1 = _mm_set1_epi16(1);
+ const __m128i acc_diff_lo = _mm_srai_epi16(
+ _mm_unpacklo_epi8(acc_diff, acc_diff), 8);
+ const __m128i acc_diff_hi = _mm_srai_epi16(
+ _mm_unpackhi_epi8(acc_diff, acc_diff), 8);
+ const __m128i acc_diff_16 = _mm_add_epi16(acc_diff_lo, acc_diff_hi);
+ const __m128i hg_fe_dc_ba = _mm_madd_epi16(acc_diff_16, k_1);
+ const __m128i hgfe_dcba = _mm_add_epi32(hg_fe_dc_ba,
+ _mm_srli_si128(hg_fe_dc_ba, 8));
+ const __m128i hgfedcba = _mm_add_epi32(hgfe_dcba,
+ _mm_srli_si128(hgfe_dcba, 4));
+ int sum_diff = _mm_cvtsi128_si32(hgfedcba);
+ return sum_diff;
+}
+
+// Denoise a 16x1 vector.
+static INLINE __m128i vp9_denoiser_16x1_sse2(const uint8_t *sig,
+ const uint8_t *mc_running_avg_y,
+ uint8_t *running_avg_y,
+ const __m128i *k_0,
+ const __m128i *k_4,
+ const __m128i *k_8,
+ const __m128i *k_16,
+ const __m128i *l3,
+ const __m128i *l32,
+ const __m128i *l21,
+ __m128i acc_diff) {
+ // Calculate differences
+ const __m128i v_sig = _mm_loadu_si128((const __m128i *)(&sig[0]));
+ const __m128i v_mc_running_avg_y = _mm_loadu_si128(
+ (const __m128i *)(&mc_running_avg_y[0]));
+ __m128i v_running_avg_y;
+ const __m128i pdiff = _mm_subs_epu8(v_mc_running_avg_y, v_sig);
+ const __m128i ndiff = _mm_subs_epu8(v_sig, v_mc_running_avg_y);
+ // Obtain the sign. FF if diff is negative.
+ const __m128i diff_sign = _mm_cmpeq_epi8(pdiff, *k_0);
+ // Clamp absolute difference to 16 to be used to get mask. Doing this
+ // allows us to use _mm_cmpgt_epi8, which operates on signed byte.
+ const __m128i clamped_absdiff = _mm_min_epu8(
+ _mm_or_si128(pdiff, ndiff), *k_16);
+ // Get masks for l2 l1 and l0 adjustments.
+ const __m128i mask2 = _mm_cmpgt_epi8(*k_16, clamped_absdiff);
+ const __m128i mask1 = _mm_cmpgt_epi8(*k_8, clamped_absdiff);
+ const __m128i mask0 = _mm_cmpgt_epi8(*k_4, clamped_absdiff);
+ // Get adjustments for l2, l1, and l0.
+ __m128i adj2 = _mm_and_si128(mask2, *l32);
+ const __m128i adj1 = _mm_and_si128(mask1, *l21);
+ const __m128i adj0 = _mm_and_si128(mask0, clamped_absdiff);
+ __m128i adj, padj, nadj;
+
+ // Combine the adjustments and get absolute adjustments.
+ adj2 = _mm_add_epi8(adj2, adj1);
+ adj = _mm_sub_epi8(*l3, adj2);
+ adj = _mm_andnot_si128(mask0, adj);
+ adj = _mm_or_si128(adj, adj0);
+
+ // Restore the sign and get positive and negative adjustments.
+ padj = _mm_andnot_si128(diff_sign, adj);
+ nadj = _mm_and_si128(diff_sign, adj);
+
+ // Calculate filtered value.
+ v_running_avg_y = _mm_adds_epu8(v_sig, padj);
+ v_running_avg_y = _mm_subs_epu8(v_running_avg_y, nadj);
+ _mm_storeu_si128((__m128i *)running_avg_y, v_running_avg_y);
+
+ // Adjustments <=7, and each element in acc_diff can fit in signed
+ // char.
+ acc_diff = _mm_adds_epi8(acc_diff, padj);
+ acc_diff = _mm_subs_epi8(acc_diff, nadj);
+ return acc_diff;
+}
+
+// Denoise a 16x1 vector with a weaker filter.
+static INLINE __m128i vp9_denoiser_adj_16x1_sse2(const uint8_t *sig,
+ const uint8_t *mc_running_avg_y,
+ uint8_t *running_avg_y,
+ const __m128i k_0,
+ const __m128i k_delta,
+ __m128i acc_diff) {
+ __m128i v_running_avg_y = _mm_loadu_si128((__m128i *)(&running_avg_y[0]));
+ // Calculate differences.
+ const __m128i v_sig = _mm_loadu_si128((const __m128i *)(&sig[0]));
+ const __m128i v_mc_running_avg_y =
+ _mm_loadu_si128((const __m128i *)(&mc_running_avg_y[0]));
+ const __m128i pdiff = _mm_subs_epu8(v_mc_running_avg_y, v_sig);
+ const __m128i ndiff = _mm_subs_epu8(v_sig, v_mc_running_avg_y);
+ // Obtain the sign. FF if diff is negative.
+ const __m128i diff_sign = _mm_cmpeq_epi8(pdiff, k_0);
+ // Clamp absolute difference to delta to get the adjustment.
+ const __m128i adj =
+ _mm_min_epu8(_mm_or_si128(pdiff, ndiff), k_delta);
+ // Restore the sign and get positive and negative adjustments.
+ __m128i padj, nadj;
+ padj = _mm_andnot_si128(diff_sign, adj);
+ nadj = _mm_and_si128(diff_sign, adj);
+ // Calculate filtered value.
+ v_running_avg_y = _mm_subs_epu8(v_running_avg_y, padj);
+ v_running_avg_y = _mm_adds_epu8(v_running_avg_y, nadj);
+ _mm_storeu_si128((__m128i *)running_avg_y, v_running_avg_y);
+
+ // Accumulate the adjustments.
+ acc_diff = _mm_subs_epi8(acc_diff, padj);
+ acc_diff = _mm_adds_epi8(acc_diff, nadj);
+ return acc_diff;
+}
+
+static int vp9_denoiser_4xM_sse2(const uint8_t *sig, int sig_stride,
+ const uint8_t *mc_running_avg_y,
+ int mc_avg_y_stride,
+ uint8_t *running_avg_y, int avg_y_stride,
+ int increase_denoising,
+ BLOCK_SIZE bs,
+ int motion_magnitude) {
+ int sum_diff_thresh;
+ int r;
+ int shift_inc = (increase_denoising &&
+ motion_magnitude <= MOTION_MAGNITUDE_THRESHOLD) ? 1 : 0;
+ unsigned char sig_buffer[2][16], mc_running_buffer[2][16],
+ running_buffer[2][16];
+ __m128i acc_diff = _mm_setzero_si128();
+ const __m128i k_0 = _mm_setzero_si128();
+ 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 + 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.
+ const __m128i l21 = _mm_set1_epi8(1);
+ int sum_diff = 0;
+
+ for (r = 0; r < ((4 << b_height_log2_lookup[bs]) >> 2); ++r) {
+ vpx_memcpy(sig_buffer[r], sig, 4);
+ vpx_memcpy(sig_buffer[r] + 4, sig + sig_stride, 4);
+ vpx_memcpy(sig_buffer[r] + 8, sig + sig_stride * 2, 4);
+ vpx_memcpy(sig_buffer[r] + 12, sig + sig_stride * 3, 4);
+ vpx_memcpy(mc_running_buffer[r], mc_running_avg_y, 4);
+ vpx_memcpy(mc_running_buffer[r] + 4, mc_running_avg_y +
+ mc_avg_y_stride, 4);
+ vpx_memcpy(mc_running_buffer[r] + 8, mc_running_avg_y +
+ mc_avg_y_stride * 2, 4);
+ vpx_memcpy(mc_running_buffer[r] + 12, mc_running_avg_y +
+ mc_avg_y_stride * 3, 4);
+ vpx_memcpy(running_buffer[r], running_avg_y, 4);
+ vpx_memcpy(running_buffer[r] + 4, running_avg_y +
+ avg_y_stride, 4);
+ vpx_memcpy(running_buffer[r] + 8, running_avg_y +
+ avg_y_stride * 2, 4);
+ vpx_memcpy(running_buffer[r] + 12, running_avg_y +
+ avg_y_stride * 3, 4);
+ acc_diff = vp9_denoiser_16x1_sse2(sig_buffer[r],
+ mc_running_buffer[r],
+ running_buffer[r],
+ &k_0, &k_4, &k_8, &k_16,
+ &l3, &l32, &l21, acc_diff);
+ vpx_memcpy(running_avg_y, running_buffer[r], 4);
+ vpx_memcpy(running_avg_y + avg_y_stride, running_buffer[r] + 4, 4);
+ vpx_memcpy(running_avg_y + avg_y_stride * 2,
+ running_buffer[r] + 8, 4);
+ vpx_memcpy(running_avg_y + avg_y_stride * 3,
+ running_buffer[r] + 12, 4);
+ // Update pointers for next iteration.
+ sig += (sig_stride << 2);
+ mc_running_avg_y += (mc_avg_y_stride << 2);
+ running_avg_y += (avg_y_stride << 2);
+ }
+
+ {
+ sum_diff = sum_diff_16x1(acc_diff);
+ sum_diff_thresh = total_adj_strong_thresh(bs, increase_denoising);
+ if (abs(sum_diff) > sum_diff_thresh) {
+ // Before returning to copy the block (i.e., apply no denoising),
+ // checK if we can still apply some (weaker) temporal filtering to
+ // this block, that would otherwise not be denoised at all. Simplest
+ // is to apply an additional adjustment to running_avg_y to bring it
+ // closer to sig. The adjustment is capped by a maximum delta, and
+ // chosen such that in most cases the resulting sum_diff will be
+ // within the accceptable range given by sum_diff_thresh.
+
+ // The delta is set by the excess of absolute pixel diff over the
+ // threshold.
+ int delta = ((abs(sum_diff) - sum_diff_thresh)
+ >> num_pels_log2_lookup[bs]) + 1;
+ // Only apply the adjustment for max delta up to 3.
+ if (delta < 4) {
+ const __m128i k_delta = _mm_set1_epi8(delta);
+ running_avg_y -= avg_y_stride * (4 << b_height_log2_lookup[bs]);
+ sum_diff = 0;
+ for (r = 0; r < ((4 << b_height_log2_lookup[bs]) >> 2); ++r) {
+ acc_diff = vp9_denoiser_adj_16x1_sse2(
+ sig_buffer[r], mc_running_buffer[r],
+ running_buffer[r], k_0, k_delta,
+ acc_diff);
+ vpx_memcpy(running_avg_y, running_buffer[r], 4);
+ vpx_memcpy(running_avg_y + avg_y_stride, running_buffer[r] + 4, 4);
+ vpx_memcpy(running_avg_y + avg_y_stride * 2,
+ running_buffer[r] + 8, 4);
+ vpx_memcpy(running_avg_y + avg_y_stride * 3,
+ running_buffer[r] + 12, 4);
+ // Update pointers for next iteration.
+ running_avg_y += (avg_y_stride << 2);
+ }
+ sum_diff = sum_diff_16x1(acc_diff);
+ if (abs(sum_diff) > sum_diff_thresh) {
+ return COPY_BLOCK;
+ }
+ } else {
+ return COPY_BLOCK;
+ }
+ }
+ }
+ return FILTER_BLOCK;
+}
+
+static int vp9_denoiser_8xM_sse2(const uint8_t *sig, int sig_stride,
+ const uint8_t *mc_running_avg_y,
+ int mc_avg_y_stride,
+ uint8_t *running_avg_y, int avg_y_stride,
+ int increase_denoising,
+ BLOCK_SIZE bs,
+ int motion_magnitude) {
+ int sum_diff_thresh;
+ int r;
+ int shift_inc = (increase_denoising &&
+ motion_magnitude <= MOTION_MAGNITUDE_THRESHOLD) ? 1 : 0;
+ unsigned char sig_buffer[8][16], mc_running_buffer[8][16],
+ running_buffer[8][16];
+ __m128i acc_diff = _mm_setzero_si128();
+ const __m128i k_0 = _mm_setzero_si128();
+ 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 + 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.
+ const __m128i l21 = _mm_set1_epi8(1);
+ int sum_diff = 0;
+
+ for (r = 0; r < ((4 << b_height_log2_lookup[bs]) >> 1); ++r) {
+ vpx_memcpy(sig_buffer[r], sig, 8);
+ vpx_memcpy(sig_buffer[r] + 8, sig + sig_stride, 8);
+ vpx_memcpy(mc_running_buffer[r], mc_running_avg_y, 8);
+ vpx_memcpy(mc_running_buffer[r] + 8, mc_running_avg_y +
+ mc_avg_y_stride, 8);
+ vpx_memcpy(running_buffer[r], running_avg_y, 8);
+ vpx_memcpy(running_buffer[r] + 8, running_avg_y +
+ avg_y_stride, 8);
+ acc_diff = vp9_denoiser_16x1_sse2(sig_buffer[r],
+ mc_running_buffer[r],
+ running_buffer[r],
+ &k_0, &k_4, &k_8, &k_16,
+ &l3, &l32, &l21, acc_diff);
+ vpx_memcpy(running_avg_y, running_buffer[r], 8);
+ vpx_memcpy(running_avg_y + avg_y_stride, running_buffer[r] + 8, 8);
+ // Update pointers for next iteration.
+ sig += (sig_stride << 1);
+ mc_running_avg_y += (mc_avg_y_stride << 1);
+ running_avg_y += (avg_y_stride << 1);
+ }
+
+ {
+ sum_diff = sum_diff_16x1(acc_diff);
+ sum_diff_thresh = total_adj_strong_thresh(bs, increase_denoising);
+ if (abs(sum_diff) > sum_diff_thresh) {
+ // Before returning to copy the block (i.e., apply no denoising),
+ // checK if we can still apply some (weaker) temporal filtering to
+ // this block, that would otherwise not be denoised at all. Simplest
+ // is to apply an additional adjustment to running_avg_y to bring it
+ // closer to sig. The adjustment is capped by a maximum delta, and
+ // chosen such that in most cases the resulting sum_diff will be
+ // within the accceptable range given by sum_diff_thresh.
+
+ // The delta is set by the excess of absolute pixel diff over the
+ // threshold.
+ int delta = ((abs(sum_diff) - sum_diff_thresh)
+ >> num_pels_log2_lookup[bs]) + 1;
+ // Only apply the adjustment for max delta up to 3.
+ if (delta < 4) {
+ const __m128i k_delta = _mm_set1_epi8(delta);
+ running_avg_y -= avg_y_stride * (4 << b_height_log2_lookup[bs]);
+ for (r = 0; r < ((4 << b_height_log2_lookup[bs]) >> 1); ++r) {
+ acc_diff = vp9_denoiser_adj_16x1_sse2(
+ sig_buffer[r], mc_running_buffer[r],
+ running_buffer[r], k_0, k_delta,
+ acc_diff);
+ vpx_memcpy(running_avg_y, running_buffer[r], 8);
+ vpx_memcpy(running_avg_y + avg_y_stride, running_buffer[r] + 8, 8);
+ // Update pointers for next iteration.
+ running_avg_y += (avg_y_stride << 1);
+ }
+ sum_diff = sum_diff_16x1(acc_diff);
+ if (abs(sum_diff) > sum_diff_thresh) {
+ return COPY_BLOCK;
+ }
+ } else {
+ return COPY_BLOCK;
+ }
+ }
+ }
+ return FILTER_BLOCK;
+}
+
+static int vp9_denoiser_64_32_16xM_sse2(const uint8_t *sig, int sig_stride,
+ const uint8_t *mc_running_avg_y,
+ int mc_avg_y_stride,
+ uint8_t *running_avg_y,
+ int avg_y_stride,
+ int increase_denoising, BLOCK_SIZE bs,
+ int motion_magnitude) {
+ int sum_diff_thresh;
+ int r, c;
+ int shift_inc = (increase_denoising &&
+ motion_magnitude <= MOTION_MAGNITUDE_THRESHOLD) ? 1 : 0;
+ __m128i acc_diff[4][4];
+ const __m128i k_0 = _mm_setzero_si128();
+ 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 + 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.
+ const __m128i l21 = _mm_set1_epi8(1);
+ int sum_diff = 0;
+
+ for (c = 0; c < 4; ++c) {
+ for (r = 0; r < 4; ++r) {
+ acc_diff[c][r] = _mm_setzero_si128();
+ }
+ }
+
+ for (r = 0; r < (4 << b_height_log2_lookup[bs]); r++) {
+ for (c = 0; c < (4 << b_width_log2_lookup[bs]); c += 16) {
+ acc_diff[c>>4][r>>4] = vp9_denoiser_16x1_sse2(
+ sig, mc_running_avg_y,
+ running_avg_y,
+ &k_0, &k_4, &k_8, &k_16,
+ &l3, &l32, &l21, acc_diff[c>>4][r>>4]);
+ // Update pointers for next iteration.
+ sig += 16;
+ mc_running_avg_y += 16;
+ running_avg_y += 16;
+ }
+
+ if ((r + 1) % 16 == 0 || (bs == BLOCK_16X8 && r == 7)) {
+ for (c = 0; c < (4 << b_width_log2_lookup[bs]); c += 16) {
+ sum_diff += sum_diff_16x1(acc_diff[c>>4][r>>4]);
+ }
+ }
+
+ // Update pointers for next iteration.
+ sig = sig - 16 * ((4 << b_width_log2_lookup[bs]) >> 4) + sig_stride;
+ mc_running_avg_y = mc_running_avg_y -
+ 16 * ((4 << b_width_log2_lookup[bs]) >> 4) +
+ mc_avg_y_stride;
+ running_avg_y = running_avg_y -
+ 16 * ((4 << b_width_log2_lookup[bs]) >> 4) +
+ avg_y_stride;
+ }
+
+ {
+ sum_diff_thresh = total_adj_strong_thresh(bs, increase_denoising);
+ if (abs(sum_diff) > sum_diff_thresh) {
+ int delta = ((abs(sum_diff) - sum_diff_thresh)
+ >> num_pels_log2_lookup[bs]) + 1;
+
+ // Only apply the adjustment for max delta up to 3.
+ if (delta < 4) {
+ const __m128i k_delta = _mm_set1_epi8(delta);
+ sig -= sig_stride * (4 << b_height_log2_lookup[bs]);
+ mc_running_avg_y -= mc_avg_y_stride * (4 << b_height_log2_lookup[bs]);
+ running_avg_y -= avg_y_stride * (4 << b_height_log2_lookup[bs]);
+ sum_diff = 0;
+ for (r = 0; r < (4 << b_height_log2_lookup[bs]); ++r) {
+ for (c = 0; c < (4 << b_width_log2_lookup[bs]); c += 16) {
+ acc_diff[c>>4][r>>4] = vp9_denoiser_adj_16x1_sse2(
+ sig, mc_running_avg_y,
+ running_avg_y, k_0,
+ k_delta, acc_diff[c>>4][r>>4]);
+ // Update pointers for next iteration.
+ sig += 16;
+ mc_running_avg_y += 16;
+ running_avg_y += 16;
+ }
+
+ if ((r + 1) % 16 == 0 || (bs == BLOCK_16X8 && r == 7)) {
+ for (c = 0; c < (4 << b_width_log2_lookup[bs]); c += 16) {
+ sum_diff += sum_diff_16x1(acc_diff[c>>4][r>>4]);
+ }
+ }
+ sig = sig - 16 * ((4 << b_width_log2_lookup[bs]) >> 4) + sig_stride;
+ mc_running_avg_y = mc_running_avg_y -
+ 16 * ((4 << b_width_log2_lookup[bs]) >> 4) +
+ mc_avg_y_stride;
+ running_avg_y = running_avg_y -
+ 16 * ((4 << b_width_log2_lookup[bs]) >> 4) +
+ avg_y_stride;
+ }
+ if (abs(sum_diff) > sum_diff_thresh) {
+ return COPY_BLOCK;
+ }
+ } else {
+ return COPY_BLOCK;
+ }
+ }
+ }
+ return FILTER_BLOCK;
+}
+
+int vp9_denoiser_filter_sse2(const uint8_t *sig, int sig_stride,
+ const uint8_t *mc_avg,
+ int mc_avg_stride,
+ uint8_t *avg, int avg_stride,
+ int increase_denoising,
+ BLOCK_SIZE bs,
+ int motion_magnitude) {
+ if (bs == BLOCK_4X4 || bs == BLOCK_4X8) {
+ return vp9_denoiser_4xM_sse2(sig, sig_stride,
+ mc_avg, mc_avg_stride,
+ avg, avg_stride,
+ increase_denoising,
+ bs, motion_magnitude);
+ } else if (bs == BLOCK_8X4 || bs == BLOCK_8X8 || bs == BLOCK_8X16) {
+ return vp9_denoiser_8xM_sse2(sig, sig_stride,
+ mc_avg, mc_avg_stride,
+ avg, avg_stride,
+ increase_denoising,
+ bs, motion_magnitude);
+ } else if (bs == BLOCK_16X8 || bs == BLOCK_16X16 || bs == BLOCK_16X32 ||
+ bs == BLOCK_32X16|| bs == BLOCK_32X32 || bs == BLOCK_32X64 ||
+ bs == BLOCK_64X32 || bs == BLOCK_64X64) {
+ return vp9_denoiser_64_32_16xM_sse2(sig, sig_stride,
+ mc_avg, mc_avg_stride,
+ avg, avg_stride,
+ increase_denoising,
+ bs, motion_magnitude);
+ } else {
+ return COPY_BLOCK;
+ }
+}
diff --git a/source/libvpx/vp9/encoder/x86/vp9_quantize_sse2.c b/source/libvpx/vp9/encoder/x86/vp9_quantize_sse2.c
new file mode 100644
index 0000000..7c1c884
--- /dev/null
+++ b/source/libvpx/vp9/encoder/x86/vp9_quantize_sse2.c
@@ -0,0 +1,225 @@
+/*
+ * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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 <emmintrin.h>
+#include <xmmintrin.h>
+
+#include "vpx/vpx_integer.h"
+
+void vp9_quantize_b_sse2(const int16_t* coeff_ptr, intptr_t n_coeffs,
+ int skip_block, const int16_t* zbin_ptr,
+ const int16_t* round_ptr, const int16_t* quant_ptr,
+ const int16_t* quant_shift_ptr, int16_t* qcoeff_ptr,
+ int16_t* dqcoeff_ptr, const int16_t* dequant_ptr,
+ int zbin_oq_value, uint16_t* eob_ptr,
+ const int16_t* scan_ptr,
+ const int16_t* iscan_ptr) {
+ __m128i zero;
+ (void)scan_ptr;
+
+ coeff_ptr += n_coeffs;
+ iscan_ptr += n_coeffs;
+ qcoeff_ptr += n_coeffs;
+ dqcoeff_ptr += n_coeffs;
+ n_coeffs = -n_coeffs;
+ zero = _mm_setzero_si128();
+ if (!skip_block) {
+ __m128i eob;
+ __m128i zbin;
+ __m128i round, quant, dequant, shift;
+ {
+ __m128i coeff0, coeff1;
+
+ // Setup global values
+ {
+ __m128i zbin_oq;
+ __m128i pw_1;
+ zbin_oq = _mm_set1_epi16(zbin_oq_value);
+ zbin = _mm_load_si128((const __m128i*)zbin_ptr);
+ round = _mm_load_si128((const __m128i*)round_ptr);
+ quant = _mm_load_si128((const __m128i*)quant_ptr);
+ zbin = _mm_add_epi16(zbin, zbin_oq);
+ pw_1 = _mm_set1_epi16(1);
+ zbin = _mm_sub_epi16(zbin, pw_1);
+ dequant = _mm_load_si128((const __m128i*)dequant_ptr);
+ shift = _mm_load_si128((const __m128i*)quant_shift_ptr);
+ }
+
+ {
+ __m128i coeff0_sign, coeff1_sign;
+ __m128i qcoeff0, qcoeff1;
+ __m128i qtmp0, qtmp1;
+ __m128i cmp_mask0, cmp_mask1;
+ // Do DC and first 15 AC
+ coeff0 = _mm_load_si128((const __m128i*)(coeff_ptr + n_coeffs));
+ coeff1 = _mm_load_si128((const __m128i*)(coeff_ptr + n_coeffs) + 1);
+
+ // Poor man's sign extract
+ coeff0_sign = _mm_srai_epi16(coeff0, 15);
+ coeff1_sign = _mm_srai_epi16(coeff1, 15);
+ qcoeff0 = _mm_xor_si128(coeff0, coeff0_sign);
+ qcoeff1 = _mm_xor_si128(coeff1, coeff1_sign);
+ qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
+
+ cmp_mask0 = _mm_cmpgt_epi16(qcoeff0, zbin);
+ zbin = _mm_unpackhi_epi64(zbin, zbin); // Switch DC to AC
+ cmp_mask1 = _mm_cmpgt_epi16(qcoeff1, zbin);
+ qcoeff0 = _mm_adds_epi16(qcoeff0, round);
+ round = _mm_unpackhi_epi64(round, round);
+ qcoeff1 = _mm_adds_epi16(qcoeff1, round);
+ qtmp0 = _mm_mulhi_epi16(qcoeff0, quant);
+ quant = _mm_unpackhi_epi64(quant, quant);
+ qtmp1 = _mm_mulhi_epi16(qcoeff1, quant);
+ qtmp0 = _mm_add_epi16(qtmp0, qcoeff0);
+ qtmp1 = _mm_add_epi16(qtmp1, qcoeff1);
+ qcoeff0 = _mm_mulhi_epi16(qtmp0, shift);
+ shift = _mm_unpackhi_epi64(shift, shift);
+ qcoeff1 = _mm_mulhi_epi16(qtmp1, shift);
+
+ // Reinsert signs
+ qcoeff0 = _mm_xor_si128(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_xor_si128(qcoeff1, coeff1_sign);
+ qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
+
+ // Mask out zbin threshold coeffs
+ qcoeff0 = _mm_and_si128(qcoeff0, cmp_mask0);
+ qcoeff1 = _mm_and_si128(qcoeff1, cmp_mask1);
+
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), qcoeff0);
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, qcoeff1);
+
+ coeff0 = _mm_mullo_epi16(qcoeff0, dequant);
+ dequant = _mm_unpackhi_epi64(dequant, dequant);
+ coeff1 = _mm_mullo_epi16(qcoeff1, dequant);
+
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), coeff0);
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, coeff1);
+ }
+
+ {
+ // Scan for eob
+ __m128i zero_coeff0, zero_coeff1;
+ __m128i nzero_coeff0, nzero_coeff1;
+ __m128i iscan0, iscan1;
+ __m128i eob1;
+ zero_coeff0 = _mm_cmpeq_epi16(coeff0, zero);
+ zero_coeff1 = _mm_cmpeq_epi16(coeff1, zero);
+ nzero_coeff0 = _mm_cmpeq_epi16(zero_coeff0, zero);
+ nzero_coeff1 = _mm_cmpeq_epi16(zero_coeff1, zero);
+ iscan0 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs));
+ iscan1 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs) + 1);
+ // Add one to convert from indices to counts
+ iscan0 = _mm_sub_epi16(iscan0, nzero_coeff0);
+ iscan1 = _mm_sub_epi16(iscan1, nzero_coeff1);
+ eob = _mm_and_si128(iscan0, nzero_coeff0);
+ eob1 = _mm_and_si128(iscan1, nzero_coeff1);
+ eob = _mm_max_epi16(eob, eob1);
+ }
+ n_coeffs += 8 * 2;
+ }
+
+ // AC only loop
+ while (n_coeffs < 0) {
+ __m128i coeff0, coeff1;
+ {
+ __m128i coeff0_sign, coeff1_sign;
+ __m128i qcoeff0, qcoeff1;
+ __m128i qtmp0, qtmp1;
+ __m128i cmp_mask0, cmp_mask1;
+
+ coeff0 = _mm_load_si128((const __m128i*)(coeff_ptr + n_coeffs));
+ coeff1 = _mm_load_si128((const __m128i*)(coeff_ptr + n_coeffs) + 1);
+
+ // Poor man's sign extract
+ coeff0_sign = _mm_srai_epi16(coeff0, 15);
+ coeff1_sign = _mm_srai_epi16(coeff1, 15);
+ qcoeff0 = _mm_xor_si128(coeff0, coeff0_sign);
+ qcoeff1 = _mm_xor_si128(coeff1, coeff1_sign);
+ qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
+
+ cmp_mask0 = _mm_cmpgt_epi16(qcoeff0, zbin);
+ cmp_mask1 = _mm_cmpgt_epi16(qcoeff1, zbin);
+ qcoeff0 = _mm_adds_epi16(qcoeff0, round);
+ qcoeff1 = _mm_adds_epi16(qcoeff1, round);
+ qtmp0 = _mm_mulhi_epi16(qcoeff0, quant);
+ qtmp1 = _mm_mulhi_epi16(qcoeff1, quant);
+ qtmp0 = _mm_add_epi16(qtmp0, qcoeff0);
+ qtmp1 = _mm_add_epi16(qtmp1, qcoeff1);
+ qcoeff0 = _mm_mulhi_epi16(qtmp0, shift);
+ qcoeff1 = _mm_mulhi_epi16(qtmp1, shift);
+
+ // Reinsert signs
+ qcoeff0 = _mm_xor_si128(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_xor_si128(qcoeff1, coeff1_sign);
+ qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign);
+ qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign);
+
+ // Mask out zbin threshold coeffs
+ qcoeff0 = _mm_and_si128(qcoeff0, cmp_mask0);
+ qcoeff1 = _mm_and_si128(qcoeff1, cmp_mask1);
+
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), qcoeff0);
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, qcoeff1);
+
+ coeff0 = _mm_mullo_epi16(qcoeff0, dequant);
+ coeff1 = _mm_mullo_epi16(qcoeff1, dequant);
+
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), coeff0);
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, coeff1);
+ }
+
+ {
+ // Scan for eob
+ __m128i zero_coeff0, zero_coeff1;
+ __m128i nzero_coeff0, nzero_coeff1;
+ __m128i iscan0, iscan1;
+ __m128i eob0, eob1;
+ zero_coeff0 = _mm_cmpeq_epi16(coeff0, zero);
+ zero_coeff1 = _mm_cmpeq_epi16(coeff1, zero);
+ nzero_coeff0 = _mm_cmpeq_epi16(zero_coeff0, zero);
+ nzero_coeff1 = _mm_cmpeq_epi16(zero_coeff1, zero);
+ iscan0 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs));
+ iscan1 = _mm_load_si128((const __m128i*)(iscan_ptr + n_coeffs) + 1);
+ // Add one to convert from indices to counts
+ iscan0 = _mm_sub_epi16(iscan0, nzero_coeff0);
+ iscan1 = _mm_sub_epi16(iscan1, nzero_coeff1);
+ eob0 = _mm_and_si128(iscan0, nzero_coeff0);
+ eob1 = _mm_and_si128(iscan1, nzero_coeff1);
+ eob0 = _mm_max_epi16(eob0, eob1);
+ eob = _mm_max_epi16(eob, eob0);
+ }
+ n_coeffs += 8 * 2;
+ }
+
+ // Accumulate EOB
+ {
+ __m128i eob_shuffled;
+ eob_shuffled = _mm_shuffle_epi32(eob, 0xe);
+ eob = _mm_max_epi16(eob, eob_shuffled);
+ eob_shuffled = _mm_shufflelo_epi16(eob, 0xe);
+ eob = _mm_max_epi16(eob, eob_shuffled);
+ eob_shuffled = _mm_shufflelo_epi16(eob, 0x1);
+ eob = _mm_max_epi16(eob, eob_shuffled);
+ *eob_ptr = _mm_extract_epi16(eob, 1);
+ }
+ } else {
+ do {
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs), zero);
+ _mm_store_si128((__m128i*)(dqcoeff_ptr + n_coeffs) + 1, zero);
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs), zero);
+ _mm_store_si128((__m128i*)(qcoeff_ptr + n_coeffs) + 1, zero);
+ n_coeffs += 8 * 2;
+ } while (n_coeffs < 0);
+ *eob_ptr = 0;
+ }
+}
diff --git a/source/libvpx/vp9/encoder/x86/vp9_sad_intrin_avx2.c b/source/libvpx/vp9/encoder/x86/vp9_sad_intrin_avx2.c
new file mode 100644
index 0000000..1131930
--- /dev/null
+++ b/source/libvpx/vp9/encoder/x86/vp9_sad_intrin_avx2.c
@@ -0,0 +1,180 @@
+/*
+ * Copyright (c) 2012 The WebM project authors. All Rights Reserved.
+ *
+ * Use 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>
+#include "vpx_ports/mem.h"
+
+#define FSAD64_H(h) \
+unsigned int vp9_sad64x##h##_avx2(const uint8_t *src_ptr, \
+ int src_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride) { \
+ int i, res; \
+ __m256i sad1_reg, sad2_reg, ref1_reg, ref2_reg; \
+ __m256i sum_sad = _mm256_setzero_si256(); \
+ __m256i sum_sad_h; \
+ __m128i sum_sad128; \
+ for (i = 0 ; i < h ; i++) { \
+ ref1_reg = _mm256_loadu_si256((__m256i const *)ref_ptr); \
+ ref2_reg = _mm256_loadu_si256((__m256i const *)(ref_ptr + 32)); \
+ sad1_reg = _mm256_sad_epu8(ref1_reg, \
+ _mm256_loadu_si256((__m256i const *)src_ptr)); \
+ sad2_reg = _mm256_sad_epu8(ref2_reg, \
+ _mm256_loadu_si256((__m256i const *)(src_ptr + 32))); \
+ sum_sad = _mm256_add_epi32(sum_sad, _mm256_add_epi32(sad1_reg, sad2_reg)); \
+ ref_ptr+= ref_stride; \
+ src_ptr+= src_stride; \
+ } \
+ sum_sad_h = _mm256_srli_si256(sum_sad, 8); \
+ sum_sad = _mm256_add_epi32(sum_sad, sum_sad_h); \
+ sum_sad128 = _mm256_extracti128_si256(sum_sad, 1); \
+ sum_sad128 = _mm_add_epi32(_mm256_castsi256_si128(sum_sad), sum_sad128); \
+ res = _mm_cvtsi128_si32(sum_sad128); \
+ return res; \
+}
+
+#define FSAD32_H(h) \
+unsigned int vp9_sad32x##h##_avx2(const uint8_t *src_ptr, \
+ int src_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride) { \
+ int i, res; \
+ __m256i sad1_reg, sad2_reg, ref1_reg, ref2_reg; \
+ __m256i sum_sad = _mm256_setzero_si256(); \
+ __m256i sum_sad_h; \
+ __m128i sum_sad128; \
+ int ref2_stride = ref_stride << 1; \
+ int src2_stride = src_stride << 1; \
+ int max = h >> 1; \
+ for (i = 0 ; i < max ; i++) { \
+ ref1_reg = _mm256_loadu_si256((__m256i const *)ref_ptr); \
+ ref2_reg = _mm256_loadu_si256((__m256i const *)(ref_ptr + ref_stride)); \
+ sad1_reg = _mm256_sad_epu8(ref1_reg, \
+ _mm256_loadu_si256((__m256i const *)src_ptr)); \
+ sad2_reg = _mm256_sad_epu8(ref2_reg, \
+ _mm256_loadu_si256((__m256i const *)(src_ptr + src_stride))); \
+ sum_sad = _mm256_add_epi32(sum_sad, _mm256_add_epi32(sad1_reg, sad2_reg)); \
+ ref_ptr+= ref2_stride; \
+ src_ptr+= src2_stride; \
+ } \
+ sum_sad_h = _mm256_srli_si256(sum_sad, 8); \
+ sum_sad = _mm256_add_epi32(sum_sad, sum_sad_h); \
+ sum_sad128 = _mm256_extracti128_si256(sum_sad, 1); \
+ sum_sad128 = _mm_add_epi32(_mm256_castsi256_si128(sum_sad), sum_sad128); \
+ res = _mm_cvtsi128_si32(sum_sad128); \
+ return res; \
+}
+
+#define FSAD64 \
+FSAD64_H(64); \
+FSAD64_H(32);
+
+#define FSAD32 \
+FSAD32_H(64); \
+FSAD32_H(32); \
+FSAD32_H(16);
+
+FSAD64;
+FSAD32;
+
+#undef FSAD64
+#undef FSAD32
+#undef FSAD64_H
+#undef FSAD32_H
+
+#define FSADAVG64_H(h) \
+unsigned int vp9_sad64x##h##_avg_avx2(const uint8_t *src_ptr, \
+ int src_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ const uint8_t *second_pred) { \
+ int i, res; \
+ __m256i sad1_reg, sad2_reg, ref1_reg, ref2_reg; \
+ __m256i sum_sad = _mm256_setzero_si256(); \
+ __m256i sum_sad_h; \
+ __m128i sum_sad128; \
+ for (i = 0 ; i < h ; i++) { \
+ ref1_reg = _mm256_loadu_si256((__m256i const *)ref_ptr); \
+ ref2_reg = _mm256_loadu_si256((__m256i const *)(ref_ptr + 32)); \
+ ref1_reg = _mm256_avg_epu8(ref1_reg, \
+ _mm256_loadu_si256((__m256i const *)second_pred)); \
+ ref2_reg = _mm256_avg_epu8(ref2_reg, \
+ _mm256_loadu_si256((__m256i const *)(second_pred +32))); \
+ sad1_reg = _mm256_sad_epu8(ref1_reg, \
+ _mm256_loadu_si256((__m256i const *)src_ptr)); \
+ sad2_reg = _mm256_sad_epu8(ref2_reg, \
+ _mm256_loadu_si256((__m256i const *)(src_ptr + 32))); \
+ sum_sad = _mm256_add_epi32(sum_sad, _mm256_add_epi32(sad1_reg, sad2_reg)); \
+ ref_ptr+= ref_stride; \
+ src_ptr+= src_stride; \
+ second_pred+= 64; \
+ } \
+ sum_sad_h = _mm256_srli_si256(sum_sad, 8); \
+ sum_sad = _mm256_add_epi32(sum_sad, sum_sad_h); \
+ sum_sad128 = _mm256_extracti128_si256(sum_sad, 1); \
+ sum_sad128 = _mm_add_epi32(_mm256_castsi256_si128(sum_sad), sum_sad128); \
+ res = _mm_cvtsi128_si32(sum_sad128); \
+ return res; \
+}
+
+#define FSADAVG32_H(h) \
+unsigned int vp9_sad32x##h##_avg_avx2(const uint8_t *src_ptr, \
+ int src_stride, \
+ const uint8_t *ref_ptr, \
+ int ref_stride, \
+ const uint8_t *second_pred) { \
+ int i, res; \
+ __m256i sad1_reg, sad2_reg, ref1_reg, ref2_reg; \
+ __m256i sum_sad = _mm256_setzero_si256(); \
+ __m256i sum_sad_h; \
+ __m128i sum_sad128; \
+ int ref2_stride = ref_stride << 1; \
+ int src2_stride = src_stride << 1; \
+ int max = h >> 1; \
+ for (i = 0 ; i < max ; i++) { \
+ ref1_reg = _mm256_loadu_si256((__m256i const *)ref_ptr); \
+ ref2_reg = _mm256_loadu_si256((__m256i const *)(ref_ptr + ref_stride)); \
+ ref1_reg = _mm256_avg_epu8(ref1_reg, \
+ _mm256_loadu_si256((__m256i const *)second_pred)); \
+ ref2_reg = _mm256_avg_epu8(ref2_reg, \
+ _mm256_loadu_si256((__m256i const *)(second_pred +32))); \
+ sad1_reg = _mm256_sad_epu8(ref1_reg, \
+ _mm256_loadu_si256((__m256i const *)src_ptr)); \
+ sad2_reg = _mm256_sad_epu8(ref2_reg, \
+ _mm256_loadu_si256((__m256i const *)(src_ptr + src_stride))); \
+ sum_sad = _mm256_add_epi32(sum_sad, \
+ _mm256_add_epi32(sad1_reg, sad2_reg)); \
+ ref_ptr+= ref2_stride; \
+ src_ptr+= src2_stride; \
+ second_pred+= 64; \
+ } \
+ sum_sad_h = _mm256_srli_si256(sum_sad, 8); \
+ sum_sad = _mm256_add_epi32(sum_sad, sum_sad_h); \
+ sum_sad128 = _mm256_extracti128_si256(sum_sad, 1); \
+ sum_sad128 = _mm_add_epi32(_mm256_castsi256_si128(sum_sad), sum_sad128); \
+ res = _mm_cvtsi128_si32(sum_sad128); \
+ return res; \
+}
+
+#define FSADAVG64 \
+FSADAVG64_H(64); \
+FSADAVG64_H(32);
+
+#define FSADAVG32 \
+FSADAVG32_H(64); \
+FSADAVG32_H(32); \
+FSADAVG32_H(16);
+
+FSADAVG64;
+FSADAVG32;
+
+#undef FSADAVG64
+#undef FSADAVG32
+#undef FSADAVG64_H
+#undef FSADAVG32_H
diff --git a/source/libvpx/vp9/vp9_common.mk b/source/libvpx/vp9/vp9_common.mk
index 07a3be8..9414120 100644
--- a/source/libvpx/vp9/vp9_common.mk
+++ b/source/libvpx/vp9/vp9_common.mk
@@ -93,6 +93,7 @@
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_high_intrapred_sse2.asm
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_high_subpixel_8t_sse2.asm
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_high_subpixel_bilinear_sse2.asm
+VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_high_loopfilter_intrin_sse2.c
endif
# common (c)
diff --git a/source/libvpx/vp9/vp9_cx_iface.c b/source/libvpx/vp9/vp9_cx_iface.c
index fbf4aa2..d0ca524 100644
--- a/source/libvpx/vp9/vp9_cx_iface.c
+++ b/source/libvpx/vp9/vp9_cx_iface.c
@@ -188,11 +188,9 @@
}
if (alt_ref_sum > REF_FRAMES - cfg->ss_number_layers)
ERROR("Not enough ref buffers for svc alt ref frames");
- if ((cfg->ss_number_layers > 3 ||
- cfg->ss_number_layers * cfg->ts_number_layers > 4) &&
+ if (cfg->ss_number_layers * cfg->ts_number_layers > 3 &&
cfg->g_error_resilient == 0)
- ERROR("Multiple frame context are not supported for more than 3 spatial "
- "layers or more than 4 spatial x temporal layers");
+ ERROR("Multiple frame context are not supported for more than 3 layers");
}
#endif
@@ -274,27 +272,49 @@
}
#if !CONFIG_VP9_HIGHBITDEPTH
- if (cfg->g_profile > (unsigned int)PROFILE_1)
+ if (cfg->g_profile > (unsigned int)PROFILE_1) {
ERROR("Profile > 1 not supported in this build configuration");
+ }
#endif
if (cfg->g_profile <= (unsigned int)PROFILE_1 &&
- extra_cfg->bit_depth > VPX_BITS_8)
+ cfg->g_bit_depth > VPX_BITS_8) {
ERROR("Codec high bit-depth not supported in profile < 2");
+ }
+ if (cfg->g_profile <= (unsigned int)PROFILE_1 &&
+ cfg->g_input_bit_depth > 8) {
+ ERROR("Source high bit-depth not supported in profile < 2");
+ }
if (cfg->g_profile > (unsigned int)PROFILE_1 &&
- extra_cfg->bit_depth == VPX_BITS_8)
+ cfg->g_bit_depth == VPX_BITS_8) {
ERROR("Codec bit-depth 8 not supported in profile > 1");
+ }
return VPX_CODEC_OK;
}
-
static vpx_codec_err_t validate_img(vpx_codec_alg_priv_t *ctx,
const vpx_image_t *img) {
switch (img->fmt) {
case VPX_IMG_FMT_YV12:
case VPX_IMG_FMT_I420:
+ case VPX_IMG_FMT_I42016:
+ break;
case VPX_IMG_FMT_I422:
case VPX_IMG_FMT_I444:
+ case VPX_IMG_FMT_I440:
+ if (ctx->cfg.g_profile != (unsigned int)PROFILE_1) {
+ ERROR("Invalid image format. I422, I444, I440 images are "
+ "not supported in profile.");
+ }
+ break;
+ case VPX_IMG_FMT_I42216:
+ case VPX_IMG_FMT_I44416:
+ case VPX_IMG_FMT_I44016:
+ if (ctx->cfg.g_profile != (unsigned int)PROFILE_1 &&
+ ctx->cfg.g_profile != (unsigned int)PROFILE_3) {
+ ERROR("Invalid image format. 16-bit I422, I444, I440 images are "
+ "not supported in profile.");
+ }
break;
default:
ERROR("Invalid image format. Only YV12, I420, I422, I444 images are "
@@ -314,9 +334,11 @@
case VPX_IMG_FMT_I420: return 12;
case VPX_IMG_FMT_I422: return 16;
case VPX_IMG_FMT_I444: return 24;
+ case VPX_IMG_FMT_I440: return 16;
case VPX_IMG_FMT_I42016: return 24;
case VPX_IMG_FMT_I42216: return 32;
case VPX_IMG_FMT_I44416: return 48;
+ case VPX_IMG_FMT_I44016: return 32;
default: assert(0 && "Invalid image format"); break;
}
return 0;
@@ -330,7 +352,7 @@
oxcf->profile = cfg->g_profile;
oxcf->width = cfg->g_w;
oxcf->height = cfg->g_h;
- oxcf->bit_depth = extra_cfg->bit_depth;
+ oxcf->bit_depth = cfg->g_bit_depth;
oxcf->input_bit_depth = cfg->g_input_bit_depth;
// guess a frame rate if out of whack, use 30
oxcf->init_framerate = (double)cfg->g_timebase.den / cfg->g_timebase.num;
diff --git a/source/libvpx/vp9/vp9_dx_iface.c b/source/libvpx/vp9/vp9_dx_iface.c
index 393c66e..85e32d3 100644
--- a/source/libvpx/vp9/vp9_dx_iface.c
+++ b/source/libvpx/vp9/vp9_dx_iface.c
@@ -437,7 +437,6 @@
// call to get_frame.
if (!(*iter)) {
img = &ctx->img;
- img->bit_depth = (int)ctx->pbi->common.bit_depth;
*iter = img;
}
}
diff --git a/source/libvpx/vp9/vp9_iface_common.h b/source/libvpx/vp9/vp9_iface_common.h
index fc98b62..00fbfdd 100644
--- a/source/libvpx/vp9/vp9_iface_common.h
+++ b/source/libvpx/vp9/vp9_iface_common.h
@@ -16,11 +16,9 @@
* the Y, U, and V planes, nor other alignment adjustments that
* might be representable by a YV12_BUFFER_CONFIG, so we just
* initialize all the fields.*/
- const int ss_x = yv12->uv_crop_width < yv12->y_crop_width;
- const int ss_y = yv12->uv_crop_height < yv12->y_crop_height;
int bps;
- if (!ss_y) {
- if (!ss_x) {
+ if (!yv12->subsampling_y) {
+ if (!yv12->subsampling_x) {
img->fmt = VPX_IMG_FMT_I444;
bps = 24;
} else {
@@ -28,16 +26,21 @@
bps = 16;
}
} else {
- img->fmt = VPX_IMG_FMT_I420;
- bps = 12;
+ if (!yv12->subsampling_x) {
+ img->fmt = VPX_IMG_FMT_I440;
+ bps = 16;
+ } else {
+ img->fmt = VPX_IMG_FMT_I420;
+ bps = 12;
+ }
}
img->bit_depth = 8;
img->w = yv12->y_stride;
img->h = ALIGN_POWER_OF_TWO(yv12->y_height + 2 * VP9_ENC_BORDER_IN_PIXELS, 3);
img->d_w = yv12->y_crop_width;
img->d_h = yv12->y_crop_height;
- img->x_chroma_shift = ss_x;
- img->y_chroma_shift = ss_y;
+ img->x_chroma_shift = yv12->subsampling_x;
+ img->y_chroma_shift = yv12->subsampling_y;
img->planes[VPX_PLANE_Y] = yv12->y_buffer;
img->planes[VPX_PLANE_U] = yv12->u_buffer;
img->planes[VPX_PLANE_V] = yv12->v_buffer;
@@ -46,6 +49,22 @@
img->stride[VPX_PLANE_U] = yv12->uv_stride;
img->stride[VPX_PLANE_V] = yv12->uv_stride;
img->stride[VPX_PLANE_ALPHA] = yv12->y_stride;
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (yv12->flags & YV12_FLAG_HIGHBITDEPTH) {
+ // vpx_image_t uses byte strides and a pointer to the first byte
+ // of the image.
+ img->fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
+ img->bit_depth = yv12->bit_depth;
+ img->planes[VPX_PLANE_Y] = (uint8_t*)CONVERT_TO_SHORTPTR(yv12->y_buffer);
+ img->planes[VPX_PLANE_U] = (uint8_t*)CONVERT_TO_SHORTPTR(yv12->u_buffer);
+ img->planes[VPX_PLANE_V] = (uint8_t*)CONVERT_TO_SHORTPTR(yv12->v_buffer);
+ img->planes[VPX_PLANE_ALPHA] = NULL;
+ img->stride[VPX_PLANE_Y] = 2 * yv12->y_stride;
+ img->stride[VPX_PLANE_U] = 2 * yv12->uv_stride;
+ img->stride[VPX_PLANE_V] = 2 * yv12->uv_stride;
+ img->stride[VPX_PLANE_ALPHA] = 2 * yv12->y_stride;
+ }
+#endif // CONFIG_VP9_HIGHBITDEPTH
img->bps = bps;
img->user_priv = user_priv;
img->img_data = yv12->buffer_alloc;
@@ -68,11 +87,39 @@
: yv12->y_width;
yv12->uv_height = img->y_chroma_shift == 1 ? (1 + yv12->y_height) / 2
: yv12->y_height;
+ yv12->uv_crop_width = yv12->uv_width;
+ yv12->uv_crop_height = yv12->uv_height;
yv12->y_stride = img->stride[VPX_PLANE_Y];
yv12->uv_stride = img->stride[VPX_PLANE_U];
+#if CONFIG_VP9_HIGHBITDEPTH
+ if (img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
+ // In vpx_image_t
+ // planes point to uint8 address of start of data
+ // stride counts uint8s to reach next row
+ // In YV12_BUFFER_CONFIG
+ // y_buffer, u_buffer, v_buffer point to uint16 address of data
+ // stride and border counts in uint16s
+ // This means that all the address calculations in the main body of code
+ // should work correctly.
+ // However, before we do any pixel operations we need to cast the address
+ // to a uint16 ponter and double its value.
+ yv12->y_buffer = CONVERT_TO_BYTEPTR(yv12->y_buffer);
+ yv12->u_buffer = CONVERT_TO_BYTEPTR(yv12->u_buffer);
+ yv12->v_buffer = CONVERT_TO_BYTEPTR(yv12->v_buffer);
+ yv12->y_stride >>= 1;
+ yv12->uv_stride >>= 1;
+ yv12->flags = YV12_FLAG_HIGHBITDEPTH;
+ } else {
+ yv12->flags = 0;
+ }
+ yv12->border = (yv12->y_stride - img->w) / 2;
+#else
yv12->border = (img->stride[VPX_PLANE_Y] - img->w) / 2;
+#endif // CONFIG_VP9_HIGHBITDEPTH
+ yv12->subsampling_x = img->x_chroma_shift;
+ yv12->subsampling_y = img->y_chroma_shift;
return VPX_CODEC_OK;
}
diff --git a/source/libvpx/vp9/vp9cx.mk b/source/libvpx/vp9/vp9cx.mk
index e450f7b..e72cb00 100644
--- a/source/libvpx/vp9/vp9cx.mk
+++ b/source/libvpx/vp9/vp9cx.mk
@@ -17,6 +17,7 @@
VP9_CX_SRCS-yes += vp9_cx_iface.c
+VP9_CX_SRCS-yes += encoder/vp9_avg.c
VP9_CX_SRCS-yes += encoder/vp9_bitstream.c
VP9_CX_SRCS-yes += encoder/vp9_context_tree.c
VP9_CX_SRCS-yes += encoder/vp9_context_tree.h
@@ -95,10 +96,12 @@
VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_variance_impl_intrin_avx2.c
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_sad4d_sse2.asm
+VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_avg_intrin_sse2.c
VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_sad4d_intrin_avx2.c
VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_subpel_variance_impl_intrin_avx2.c
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_temporal_filter_apply_sse2.asm
VP9_CX_SRCS-$(HAVE_SSE3) += encoder/x86/vp9_sad_sse3.asm
+VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_quantize_sse2.c
ifeq ($(CONFIG_USE_X86INC),yes)
VP9_CX_SRCS-$(HAVE_MMX) += encoder/x86/vp9_dct_mmx.asm
@@ -115,11 +118,16 @@
endif
VP9_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/vp9_sad_ssse3.asm
VP9_CX_SRCS-$(HAVE_SSE4_1) += encoder/x86/vp9_sad_sse4.asm
+VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_sad_intrin_avx2.c
VP9_CX_SRCS-$(ARCH_X86_64) += encoder/x86/vp9_ssim_opt_x86_64.asm
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_dct_sse2.c
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_dct32x32_sse2.c
+ifeq ($(CONFIG_VP9_TEMPORAL_DENOISING),yes)
+VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_denoiser_sse2.c
+endif
+
VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_dct32x32_avx2.c
VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_dct_avx2.c
VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_error_intrin_avx2.c
diff --git a/source/libvpx/vpx/src/svc_encodeframe.c b/source/libvpx/vpx/src/svc_encodeframe.c
index 773087d..fa3409c 100644
--- a/source/libvpx/vpx/src/svc_encodeframe.c
+++ b/source/libvpx/vpx/src/svc_encodeframe.c
@@ -350,7 +350,7 @@
}
}
- for (i = 0; i < svc_ctx->spatial_layers; ++i) {
+ for (i = 0; i < VPX_SS_MAX_LAYERS; ++i) {
if (total > 0) {
enc_cfg->ss_target_bitrate[i] = (unsigned int)
(enc_cfg->rc_target_bitrate * alloc_ratio[i] / total);
diff --git a/source/libvpx/vpx/src/vpx_image.c b/source/libvpx/vpx/src/vpx_image.c
index e58b61e..9aae12c 100644
--- a/source/libvpx/vpx/src/vpx_image.c
+++ b/source/libvpx/vpx/src/vpx_image.c
@@ -15,16 +15,16 @@
#include "vpx/vpx_integer.h"
#include "vpx_mem/vpx_mem.h"
-static vpx_image_t *img_alloc_helper(vpx_image_t *img,
- vpx_img_fmt_t fmt,
- unsigned int d_w,
- unsigned int d_h,
- unsigned int buf_align,
- unsigned int stride_align,
+static vpx_image_t *img_alloc_helper(vpx_image_t *img,
+ vpx_img_fmt_t fmt,
+ unsigned int d_w,
+ unsigned int d_h,
+ unsigned int buf_align,
+ unsigned int stride_align,
unsigned char *img_data) {
-
- unsigned int h, w, s, xcs, ycs, bps;
- int align;
+ unsigned int h, w, s, xcs, ycs, bps;
+ unsigned int stride_in_bytes;
+ int align;
/* Treat align==0 like align==1 */
if (!buf_align)
@@ -70,6 +70,7 @@
bps = 12;
break;
case VPX_IMG_FMT_I422:
+ case VPX_IMG_FMT_I440:
bps = 16;
break;
case VPX_IMG_FMT_I444:
@@ -79,6 +80,7 @@
bps = 24;
break;
case VPX_IMG_FMT_I42216:
+ case VPX_IMG_FMT_I44016:
bps = 32;
break;
case VPX_IMG_FMT_I44416:
@@ -107,9 +109,12 @@
switch (fmt) {
case VPX_IMG_FMT_I420:
+ case VPX_IMG_FMT_I440:
case VPX_IMG_FMT_YV12:
case VPX_IMG_FMT_VPXI420:
case VPX_IMG_FMT_VPXYV12:
+ case VPX_IMG_FMT_I42016:
+ case VPX_IMG_FMT_I44016:
ycs = 1;
break;
default:
@@ -124,6 +129,7 @@
h = (d_h + align) & ~align;
s = (fmt & VPX_IMG_FMT_PLANAR) ? w : bps * w / 8;
s = (s + stride_align - 1) & ~(stride_align - 1);
+ stride_in_bytes = (fmt & VPX_IMG_FMT_HIGHBITDEPTH) ? s * 2 : s;
/* Allocate the new image */
if (!img) {
@@ -162,8 +168,8 @@
img->bps = bps;
/* Calculate strides */
- img->stride[VPX_PLANE_Y] = img->stride[VPX_PLANE_ALPHA] = s;
- img->stride[VPX_PLANE_U] = img->stride[VPX_PLANE_V] = s >> xcs;
+ img->stride[VPX_PLANE_Y] = img->stride[VPX_PLANE_ALPHA] = stride_in_bytes;
+ img->stride[VPX_PLANE_U] = img->stride[VPX_PLANE_V] = stride_in_bytes >> xcs;
/* Default viewport to entire image */
if (!vpx_img_set_rect(img, 0, 0, d_w, d_h))
@@ -209,39 +215,40 @@
img->planes[VPX_PLANE_PACKED] =
img->img_data + x * img->bps / 8 + y * img->stride[VPX_PLANE_PACKED];
} else {
+ const int bytes_per_sample =
+ (img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) ? 2 : 1;
data = img->img_data;
if (img->fmt & VPX_IMG_FMT_HAS_ALPHA) {
img->planes[VPX_PLANE_ALPHA] =
- data + x + y * img->stride[VPX_PLANE_ALPHA];
+ data + x * bytes_per_sample + y * img->stride[VPX_PLANE_ALPHA];
data += img->h * img->stride[VPX_PLANE_ALPHA];
}
- img->planes[VPX_PLANE_Y] = data + x + y * img->stride[VPX_PLANE_Y];
+ img->planes[VPX_PLANE_Y] = data + x * bytes_per_sample +
+ y * img->stride[VPX_PLANE_Y];
data += img->h * img->stride[VPX_PLANE_Y];
if (!(img->fmt & VPX_IMG_FMT_UV_FLIP)) {
- img->planes[VPX_PLANE_U] = data
- + (x >> img->x_chroma_shift)
- + (y >> img->y_chroma_shift) * img->stride[VPX_PLANE_U];
+ img->planes[VPX_PLANE_U] =
+ data + (x >> img->x_chroma_shift) * bytes_per_sample +
+ (y >> img->y_chroma_shift) * img->stride[VPX_PLANE_U];
data += (img->h >> img->y_chroma_shift) * img->stride[VPX_PLANE_U];
- img->planes[VPX_PLANE_V] = data
- + (x >> img->x_chroma_shift)
- + (y >> img->y_chroma_shift) * img->stride[VPX_PLANE_V];
+ img->planes[VPX_PLANE_V] =
+ data + (x >> img->x_chroma_shift) * bytes_per_sample +
+ (y >> img->y_chroma_shift) * img->stride[VPX_PLANE_V];
} else {
- img->planes[VPX_PLANE_V] = data
- + (x >> img->x_chroma_shift)
- + (y >> img->y_chroma_shift) * img->stride[VPX_PLANE_V];
+ img->planes[VPX_PLANE_V] =
+ data + (x >> img->x_chroma_shift) * bytes_per_sample +
+ (y >> img->y_chroma_shift) * img->stride[VPX_PLANE_V];
data += (img->h >> img->y_chroma_shift) * img->stride[VPX_PLANE_V];
- img->planes[VPX_PLANE_U] = data
- + (x >> img->x_chroma_shift)
- + (y >> img->y_chroma_shift) * img->stride[VPX_PLANE_U];
+ img->planes[VPX_PLANE_U] =
+ data + (x >> img->x_chroma_shift) * bytes_per_sample +
+ (y >> img->y_chroma_shift) * img->stride[VPX_PLANE_U];
}
}
-
return 0;
}
-
return -1;
}
diff --git a/source/libvpx/vpx/vpx_encoder.h b/source/libvpx/vpx/vpx_encoder.h
index c6c7d08..044243d 100644
--- a/source/libvpx/vpx/vpx_encoder.h
+++ b/source/libvpx/vpx/vpx_encoder.h
@@ -161,7 +161,9 @@
VPX_CODEC_STATS_PKT, /**< Two-pass statistics for this frame */
VPX_CODEC_FPMB_STATS_PKT, /**< first pass mb statistics for this frame */
VPX_CODEC_PSNR_PKT, /**< PSNR statistics for this frame */
-#if CONFIG_SPATIAL_SVC
+ // TODO(minghai): This is for testing purporses. The released library can't
+ // depend on vpx_config.h
+#if defined(CONFIG_SPATIAL_SVC) && CONFIG_SPATIAL_SVC
VPX_CODEC_SPATIAL_SVC_LAYER_SIZES, /**< Sizes for each layer in this frame*/
VPX_CODEC_SPATIAL_SVC_LAYER_PSNR, /**< PSNR for each layer in this frame*/
#endif
@@ -201,7 +203,9 @@
double psnr[4]; /**< PSNR, total/y/u/v */
} psnr; /**< data for PSNR packet */
vpx_fixed_buf_t raw; /**< data for arbitrary packets */
-#if CONFIG_SPATIAL_SVC
+ // TODO(minghai): This is for testing purporses. The released library
+ // can't depend on vpx_config.h
+#if defined(CONFIG_SPATIAL_SVC) && CONFIG_SPATIAL_SVC
size_t layer_sizes[VPX_SS_MAX_LAYERS];
struct vpx_psnr_pkt layer_psnr[VPX_SS_MAX_LAYERS];
#endif
diff --git a/source/libvpx/vpx/vpx_image.h b/source/libvpx/vpx/vpx_image.h
index 0b7bb90..337e4c4 100644
--- a/source/libvpx/vpx/vpx_image.h
+++ b/source/libvpx/vpx/vpx_image.h
@@ -58,46 +58,14 @@
VPX_IMG_FMT_VPXI420 = VPX_IMG_FMT_PLANAR | 4,
VPX_IMG_FMT_I422 = VPX_IMG_FMT_PLANAR | 5,
VPX_IMG_FMT_I444 = VPX_IMG_FMT_PLANAR | 6,
- VPX_IMG_FMT_444A = VPX_IMG_FMT_PLANAR | VPX_IMG_FMT_HAS_ALPHA | 7,
+ VPX_IMG_FMT_I440 = VPX_IMG_FMT_PLANAR | 7,
+ VPX_IMG_FMT_444A = VPX_IMG_FMT_PLANAR | VPX_IMG_FMT_HAS_ALPHA | 6,
VPX_IMG_FMT_I42016 = VPX_IMG_FMT_I420 | VPX_IMG_FMT_HIGHBITDEPTH,
VPX_IMG_FMT_I42216 = VPX_IMG_FMT_I422 | VPX_IMG_FMT_HIGHBITDEPTH,
- VPX_IMG_FMT_I44416 = VPX_IMG_FMT_I444 | VPX_IMG_FMT_HIGHBITDEPTH
+ VPX_IMG_FMT_I44416 = VPX_IMG_FMT_I444 | VPX_IMG_FMT_HIGHBITDEPTH,
+ VPX_IMG_FMT_I44016 = VPX_IMG_FMT_I440 | VPX_IMG_FMT_HIGHBITDEPTH
} vpx_img_fmt_t; /**< alias for enum vpx_img_fmt */
-#if !defined(VPX_CODEC_DISABLE_COMPAT) || !VPX_CODEC_DISABLE_COMPAT
-#define IMG_FMT_PLANAR VPX_IMG_FMT_PLANAR /**< \deprecated Use #VPX_IMG_FMT_PLANAR */
-#define IMG_FMT_UV_FLIP VPX_IMG_FMT_UV_FLIP /**< \deprecated Use #VPX_IMG_FMT_UV_FLIP */
-#define IMG_FMT_HAS_ALPHA VPX_IMG_FMT_HAS_ALPHA /**< \deprecated Use #VPX_IMG_FMT_HAS_ALPHA */
-
- /*!\brief Deprecated list of supported image formats
- * \deprecated New code should use #vpx_img_fmt
- */
-#define img_fmt vpx_img_fmt
- /*!\brief alias for enum img_fmt.
- * \deprecated New code should use #vpx_img_fmt_t
- */
-#define img_fmt_t vpx_img_fmt_t
-
-#define IMG_FMT_NONE VPX_IMG_FMT_NONE /**< \deprecated Use #VPX_IMG_FMT_NONE */
-#define IMG_FMT_RGB24 VPX_IMG_FMT_RGB24 /**< \deprecated Use #VPX_IMG_FMT_RGB24 */
-#define IMG_FMT_RGB32 VPX_IMG_FMT_RGB32 /**< \deprecated Use #VPX_IMG_FMT_RGB32 */
-#define IMG_FMT_RGB565 VPX_IMG_FMT_RGB565 /**< \deprecated Use #VPX_IMG_FMT_RGB565 */
-#define IMG_FMT_RGB555 VPX_IMG_FMT_RGB555 /**< \deprecated Use #VPX_IMG_FMT_RGB555 */
-#define IMG_FMT_UYVY VPX_IMG_FMT_UYVY /**< \deprecated Use #VPX_IMG_FMT_UYVY */
-#define IMG_FMT_YUY2 VPX_IMG_FMT_YUY2 /**< \deprecated Use #VPX_IMG_FMT_YUY2 */
-#define IMG_FMT_YVYU VPX_IMG_FMT_YVYU /**< \deprecated Use #VPX_IMG_FMT_YVYU */
-#define IMG_FMT_BGR24 VPX_IMG_FMT_BGR24 /**< \deprecated Use #VPX_IMG_FMT_BGR24 */
-#define IMG_FMT_RGB32_LE VPX_IMG_FMT_RGB32_LE /**< \deprecated Use #VPX_IMG_FMT_RGB32_LE */
-#define IMG_FMT_ARGB VPX_IMG_FMT_ARGB /**< \deprecated Use #VPX_IMG_FMT_ARGB */
-#define IMG_FMT_ARGB_LE VPX_IMG_FMT_ARGB_LE /**< \deprecated Use #VPX_IMG_FMT_ARGB_LE */
-#define IMG_FMT_RGB565_LE VPX_IMG_FMT_RGB565_LE /**< \deprecated Use #VPX_IMG_FMT_RGB565_LE */
-#define IMG_FMT_RGB555_LE VPX_IMG_FMT_RGB555_LE /**< \deprecated Use #VPX_IMG_FMT_RGB555_LE */
-#define IMG_FMT_YV12 VPX_IMG_FMT_YV12 /**< \deprecated Use #VPX_IMG_FMT_YV12 */
-#define IMG_FMT_I420 VPX_IMG_FMT_I420 /**< \deprecated Use #VPX_IMG_FMT_I420 */
-#define IMG_FMT_VPXYV12 VPX_IMG_FMT_VPXYV12 /**< \deprecated Use #VPX_IMG_FMT_VPXYV12 */
-#define IMG_FMT_VPXI420 VPX_IMG_FMT_VPXI420 /**< \deprecated Use #VPX_IMG_FMT_VPXI420 */
-#endif /* VPX_CODEC_DISABLE_COMPAT */
-
/**\brief Image Descriptor */
typedef struct vpx_image {
vpx_img_fmt_t fmt; /**< Image Format */
@@ -121,13 +89,6 @@
#define VPX_PLANE_U 1 /**< U (Chroma) plane */
#define VPX_PLANE_V 2 /**< V (Chroma) plane */
#define VPX_PLANE_ALPHA 3 /**< A (Transparency) plane */
-#if !defined(VPX_CODEC_DISABLE_COMPAT) || !VPX_CODEC_DISABLE_COMPAT
-#define PLANE_PACKED VPX_PLANE_PACKED
-#define PLANE_Y VPX_PLANE_Y
-#define PLANE_U VPX_PLANE_U
-#define PLANE_V VPX_PLANE_V
-#define PLANE_ALPHA VPX_PLANE_ALPHA
-#endif
unsigned char *planes[4]; /**< pointer to the top left pixel for each plane */
int stride[4]; /**< stride between rows for each plane */
diff --git a/source/libvpx/vpx/vpx_integer.h b/source/libvpx/vpx/vpx_integer.h
index ffeefb8..500f9b9 100644
--- a/source/libvpx/vpx/vpx_integer.h
+++ b/source/libvpx/vpx/vpx_integer.h
@@ -49,9 +49,15 @@
/* Most platforms have the C99 standard integer types. */
-#if defined(__cplusplus) && !defined(__STDC_FORMAT_MACROS)
-#define __STDC_FORMAT_MACROS
-#endif
+#if defined(__cplusplus)
+# if !defined(__STDC_FORMAT_MACROS)
+# define __STDC_FORMAT_MACROS
+# endif
+# if !defined(__STDC_LIMIT_MACROS)
+# define __STDC_LIMIT_MACROS
+# endif
+#endif // __cplusplus
+
#include <stdint.h>
#endif
diff --git a/source/libvpx/vpx_ports/arm.h b/source/libvpx/vpx_ports/arm.h
index 1e4a8e2..42c98f5 100644
--- a/source/libvpx/vpx_ports/arm.h
+++ b/source/libvpx/vpx_ports/arm.h
@@ -27,6 +27,12 @@
int arm_cpu_caps(void);
+// Earlier gcc compilers have issues with some neon intrinsics
+#if !defined(__clang__) && defined(__GNUC__) && \
+ __GNUC__ == 4 && __GNUC_MINOR__ <= 6
+#define VPX_INCOMPATIBLE_GCC
+#endif
+
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/source/libvpx/vpx_scale/generic/yv12config.c b/source/libvpx/vpx_scale/generic/yv12config.c
index 475d231..00a8c16 100644
--- a/source/libvpx/vpx_scale/generic/yv12config.c
+++ b/source/libvpx/vpx_scale/generic/yv12config.c
@@ -10,7 +10,6 @@
#include <assert.h>
-#include "./vpx_config.h"
#include "vpx_scale/yv12config.h"
#include "vpx_mem/vpx_mem.h"
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
@@ -242,6 +241,8 @@
ybf->border = border;
ybf->frame_size = (int)frame_size;
+ ybf->subsampling_x = ss_x;
+ ybf->subsampling_y = ss_y;
#if CONFIG_VP9_HIGHBITDEPTH
if (use_highbitdepth) {
diff --git a/source/libvpx/vpx_scale/yv12config.h b/source/libvpx/vpx_scale/yv12config.h
index 9ff764c..b9f13fd 100644
--- a/source/libvpx/vpx_scale/yv12config.h
+++ b/source/libvpx/vpx_scale/yv12config.h
@@ -15,6 +15,7 @@
extern "C" {
#endif
+#include "./vpx_config.h"
#include "vpx/vpx_codec.h"
#include "vpx/vpx_frame_buffer.h"
#include "vpx/vpx_integer.h"
@@ -51,6 +52,8 @@
int buffer_alloc_sz;
int border;
int frame_size;
+ int subsampling_x;
+ int subsampling_y;
unsigned int bit_depth;
int corrupted;
diff --git a/source/libvpx/vpxdec.c b/source/libvpx/vpxdec.c
index cf23c29..2afdb71 100644
--- a/source/libvpx/vpxdec.c
+++ b/source/libvpx/vpxdec.c
@@ -47,52 +47,49 @@
struct WebmInputContext *webm_ctx;
};
-static const arg_def_t looparg = ARG_DEF(NULL, "loops", 1,
- "Number of times to decode the file");
-static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
- "Codec to use");
-static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
- "Output raw YV12 frames");
-static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
- "Output raw I420 frames");
-static const arg_def_t flipuvarg = ARG_DEF(NULL, "flipuv", 0,
- "Flip the chroma planes in the output");
-static const arg_def_t rawvideo = ARG_DEF(NULL, "rawvideo", 0,
- "Output raw YUV frames");
-static const arg_def_t noblitarg = ARG_DEF(NULL, "noblit", 0,
- "Don't process the decoded frames");
-static const arg_def_t progressarg = ARG_DEF(NULL, "progress", 0,
- "Show progress after each frame decodes");
-static const arg_def_t limitarg = ARG_DEF(NULL, "limit", 1,
- "Stop decoding after n frames");
-static const arg_def_t skiparg = ARG_DEF(NULL, "skip", 1,
- "Skip the first n input frames");
-static const arg_def_t postprocarg = ARG_DEF(NULL, "postproc", 0,
- "Postprocess decoded frames");
-static const arg_def_t summaryarg = ARG_DEF(NULL, "summary", 0,
- "Show timing summary");
-static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
- "Output file name pattern (see below)");
-static const arg_def_t threadsarg = ARG_DEF("t", "threads", 1,
- "Max threads to use");
-static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
- "Show version string");
-static const arg_def_t error_concealment = ARG_DEF(NULL, "error-concealment", 0,
- "Enable decoder error-concealment");
-static const arg_def_t scalearg = ARG_DEF("S", "scale", 0,
- "Scale output frames uniformly");
-static const arg_def_t continuearg =
- ARG_DEF("k", "keep-going", 0, "(debug) Continue decoding after error");
-
-static const arg_def_t fb_arg =
- ARG_DEF(NULL, "frame-buffers", 1, "Number of frame buffers to use");
-
-static const arg_def_t md5arg = ARG_DEF(NULL, "md5", 0,
- "Compute the MD5 sum of the decoded frame");
+static const arg_def_t looparg = ARG_DEF(
+ NULL, "loops", 1, "Number of times to decode the file");
+static const arg_def_t codecarg = ARG_DEF(
+ NULL, "codec", 1, "Codec to use");
+static const arg_def_t use_yv12 = ARG_DEF(
+ NULL, "yv12", 0, "Output raw YV12 frames");
+static const arg_def_t use_i420 = ARG_DEF(
+ NULL, "i420", 0, "Output raw I420 frames");
+static const arg_def_t flipuvarg = ARG_DEF(
+ NULL, "flipuv", 0, "Flip the chroma planes in the output");
+static const arg_def_t rawvideo = ARG_DEF(
+ NULL, "rawvideo", 0, "Output raw YUV frames");
+static const arg_def_t noblitarg = ARG_DEF(
+ NULL, "noblit", 0, "Don't process the decoded frames");
+static const arg_def_t progressarg = ARG_DEF(
+ NULL, "progress", 0, "Show progress after each frame decodes");
+static const arg_def_t limitarg = ARG_DEF(
+ NULL, "limit", 1, "Stop decoding after n frames");
+static const arg_def_t skiparg = ARG_DEF(
+ NULL, "skip", 1, "Skip the first n input frames");
+static const arg_def_t postprocarg = ARG_DEF(
+ NULL, "postproc", 0, "Postprocess decoded frames");
+static const arg_def_t summaryarg = ARG_DEF(
+ NULL, "summary", 0, "Show timing summary");
+static const arg_def_t outputfile = ARG_DEF(
+ "o", "output", 1, "Output file name pattern (see below)");
+static const arg_def_t threadsarg = ARG_DEF(
+ "t", "threads", 1, "Max threads to use");
+static const arg_def_t verbosearg = ARG_DEF(
+ "v", "verbose", 0, "Show version string");
+static const arg_def_t error_concealment = ARG_DEF(
+ NULL, "error-concealment", 0, "Enable decoder error-concealment");
+static const arg_def_t scalearg = ARG_DEF(
+ "S", "scale", 0, "Scale output frames uniformly");
+static const arg_def_t continuearg = ARG_DEF(
+ "k", "keep-going", 0, "(debug) Continue decoding after error");
+static const arg_def_t fb_arg = ARG_DEF(
+ NULL, "frame-buffers", 1, "Number of frame buffers to use");
+static const arg_def_t md5arg = ARG_DEF(
+ NULL, "md5", 0, "Compute the MD5 sum of the decoded frame");
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
static const arg_def_t outbitdeptharg = ARG_DEF(
- NULL, "output-bit-depth", 1,
- "Output bit-depth for decoded frames");
+ NULL, "output-bit-depth", 1, "Output bit-depth for decoded frames");
#endif
static const arg_def_t *all_args[] = {
@@ -527,174 +524,12 @@
}
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
-static void high_img_upshift(vpx_image_t *dst, vpx_image_t *src,
- int input_shift) {
- const int offset = input_shift > 0 ? (1 << (input_shift - 1)) : 0;
- int plane;
- if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
- dst->x_chroma_shift != src->x_chroma_shift ||
- dst->y_chroma_shift != src->y_chroma_shift ||
- dst->fmt != src->fmt || input_shift < 0) {
- fatal("Unsupported image conversion");
- }
- switch (src->fmt) {
- case VPX_IMG_FMT_I42016:
- case VPX_IMG_FMT_I42216:
- case VPX_IMG_FMT_I44416:
- break;
- default:
- fatal("Unsupported image conversion");
- break;
- }
- for (plane = 0; plane < 3; plane++) {
- int w = src->d_w;
- int h = src->d_h;
- int x, y;
- if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
- }
- for (y = 0; y < h; y++) {
- uint16_t *p_src = (uint16_t *)(src->planes[plane] +
- y * src->stride[plane]);
- uint16_t *p_dst = (uint16_t *)(dst->planes[plane] +
- y * dst->stride[plane]);
- for (x = 0; x < w; x++)
- *p_dst++ = (*p_src++ << input_shift) + offset;
- }
- }
-}
-
-static void low_img_upshift(vpx_image_t *dst, vpx_image_t *src,
- int input_shift) {
- const int offset = input_shift > 0 ? (1 << (input_shift - 1)) : 0;
- int plane;
- if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
- dst->x_chroma_shift != src->x_chroma_shift ||
- dst->y_chroma_shift != src->y_chroma_shift ||
- dst->fmt != src->fmt + VPX_IMG_FMT_HIGHBITDEPTH ||
- input_shift < 0) {
- fatal("Unsupported image conversion");
- }
- switch (src->fmt) {
- case VPX_IMG_FMT_I420:
- case VPX_IMG_FMT_I422:
- case VPX_IMG_FMT_I444:
- break;
- default:
- fatal("Unsupported image conversion");
- break;
- }
- for (plane = 0; plane < 3; plane++) {
- int w = src->d_w;
- int h = src->d_h;
- int x, y;
- if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
- }
- for (y = 0; y < h; y++) {
- uint8_t *p_src = src->planes[plane] + y * src->stride[plane];
- uint16_t *p_dst = (uint16_t *)(dst->planes[plane] +
- y * dst->stride[plane]);
- for (x = 0; x < w; x++) {
- *p_dst++ = (*p_src++ << input_shift) + offset;
- }
- }
- }
-}
-
-static void img_upshift(vpx_image_t *dst, vpx_image_t *src,
- int input_shift) {
- if (src->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
- high_img_upshift(dst, src, input_shift);
- } else {
- low_img_upshift(dst, src, input_shift);
- }
-}
-
-static void high_img_downshift(vpx_image_t *dst, vpx_image_t *src,
- int down_shift) {
- int plane;
- if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
- dst->x_chroma_shift != src->x_chroma_shift ||
- dst->y_chroma_shift != src->y_chroma_shift ||
- dst->fmt != src->fmt || down_shift < 0) {
- fatal("Unsupported image conversion");
- }
- switch (src->fmt) {
- case VPX_IMG_FMT_I42016:
- case VPX_IMG_FMT_I42216:
- case VPX_IMG_FMT_I44416:
- break;
- default:
- fatal("Unsupported image conversion");
- break;
- }
- for (plane = 0; plane < 3; plane++) {
- int w = src->d_w;
- int h = src->d_h;
- int x, y;
- if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
- }
- for (y = 0; y < h; y++) {
- uint16_t *p_src = (uint16_t *)(src->planes[plane] +
- y * src->stride[plane]);
- uint16_t *p_dst = (uint16_t *)(dst->planes[plane] +
- y * dst->stride[plane]);
- for (x = 0; x < w; x++)
- *p_dst++ = *p_src++ >> down_shift;
- }
- }
-}
-
-static void low_img_downshift(vpx_image_t *dst, vpx_image_t *src,
- int down_shift) {
- int plane;
- if (dst->d_w != src->d_w || dst->d_h != src->d_h ||
- dst->x_chroma_shift != src->x_chroma_shift ||
- dst->y_chroma_shift != src->y_chroma_shift ||
- src->fmt != dst->fmt + VPX_IMG_FMT_HIGHBITDEPTH ||
- down_shift < 0) {
- fatal("Unsupported image conversion");
- }
- switch (dst->fmt) {
- case VPX_IMG_FMT_I420:
- case VPX_IMG_FMT_I422:
- case VPX_IMG_FMT_I444:
- break;
- default:
- fatal("Unsupported image conversion");
- break;
- }
- for (plane = 0; plane < 3; plane++) {
- int w = src->d_w;
- int h = src->d_h;
- int x, y;
- if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
- }
- for (y = 0; y < h; y++) {
- uint16_t *p_src = (uint16_t *)(src->planes[plane] +
- y * src->stride[plane]);
- uint8_t *p_dst = dst->planes[plane] + y * dst->stride[plane];
- for (x = 0; x < w; x++) {
- *p_dst++ = *p_src++ >> down_shift;
- }
- }
- }
-}
-
-static void img_downshift(vpx_image_t *dst, vpx_image_t *src,
- int down_shift) {
- if (dst->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
- high_img_downshift(dst, src, down_shift);
- } else {
- low_img_downshift(dst, src, down_shift);
- }
+static int img_shifted_realloc_required(const vpx_image_t *img,
+ const vpx_image_t *shifted,
+ vpx_img_fmt_t required_fmt) {
+ return img->d_w != shifted->d_w ||
+ img->d_h != shifted->d_h ||
+ required_fmt != shifted->fmt;
}
#endif
@@ -933,7 +768,7 @@
if (use_y4m && !noblit) {
if (!single_file) {
fprintf(stderr, "YUV4MPEG2 not supported with output patterns,"
- " try --i420 or --yv12.\n");
+ " try --i420 or --yv12 or --rawvideo.\n");
return EXIT_FAILURE;
}
@@ -1130,22 +965,25 @@
}
// Shift up or down if necessary
if (output_bit_depth != img->bit_depth) {
+ const vpx_img_fmt_t shifted_fmt = output_bit_depth == 8 ?
+ img->fmt ^ (img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) :
+ img->fmt | VPX_IMG_FMT_HIGHBITDEPTH;
+ if (img_shifted &&
+ img_shifted_realloc_required(img, img_shifted, shifted_fmt)) {
+ vpx_img_free(img_shifted);
+ img_shifted = NULL;
+ }
if (!img_shifted) {
- if (output_bit_depth == 8) {
- img_shifted = vpx_img_alloc(
- NULL, img->fmt - VPX_IMG_FMT_HIGHBITDEPTH,
- img->d_w, img->d_h, 16);
- } else {
- img_shifted = vpx_img_alloc(
- NULL, img->fmt | VPX_IMG_FMT_HIGHBITDEPTH,
- img->d_w, img->d_h, 16);
- }
+ img_shifted = vpx_img_alloc(NULL, shifted_fmt,
+ img->d_w, img->d_h, 16);
img_shifted->bit_depth = output_bit_depth;
}
if (output_bit_depth > img->bit_depth) {
- img_upshift(img_shifted, img, output_bit_depth - img->bit_depth);
+ vpx_img_upshift(img_shifted, img,
+ output_bit_depth - img->bit_depth);
} else {
- img_downshift(img_shifted, img, img->bit_depth - output_bit_depth);
+ vpx_img_downshift(img_shifted, img,
+ img->bit_depth - output_bit_depth);
}
img = img_shifted;
}
@@ -1155,6 +993,10 @@
if (use_y4m) {
char buf[Y4M_BUFFER_SIZE] = {0};
size_t len = 0;
+ if (img->fmt == VPX_IMG_FMT_I440 || img->fmt == VPX_IMG_FMT_I44016) {
+ fprintf(stderr, "Cannot produce y4m output for 440 sampling.\n");
+ goto fail;
+ }
if (frame_out == 1) {
// Y4M file header
len = y4m_write_file_header(buf, sizeof(buf),
diff --git a/source/libvpx/vpxenc.c b/source/libvpx/vpxenc.c
index 1b0b632..0a0c071 100644
--- a/source/libvpx/vpxenc.c
+++ b/source/libvpx/vpxenc.c
@@ -128,48 +128,50 @@
return 0;
}
-static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
- "Debug mode (makes output deterministic)");
-static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
- "Output filename");
-static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
- "Input file is YV12 ");
-static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
- "Input file is I420 (default)");
-static const arg_def_t use_i422 = ARG_DEF(NULL, "i422", 0,
- "Input file is I422");
-static const arg_def_t use_i444 = ARG_DEF(NULL, "i444", 0,
- "Input file is I444");
-static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
- "Codec to use");
-static const arg_def_t passes = ARG_DEF("p", "passes", 1,
- "Number of passes (1/2)");
-static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
- "Pass to execute (1/2)");
-static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
- "First pass statistics file name");
+static const arg_def_t debugmode = ARG_DEF(
+ "D", "debug", 0, "Debug mode (makes output deterministic)");
+static const arg_def_t outputfile = ARG_DEF(
+ "o", "output", 1, "Output filename");
+static const arg_def_t use_yv12 = ARG_DEF(
+ NULL, "yv12", 0, "Input file is YV12 ");
+static const arg_def_t use_i420 = ARG_DEF(
+ NULL, "i420", 0, "Input file is I420 (default)");
+static const arg_def_t use_i422 = ARG_DEF(
+ NULL, "i422", 0, "Input file is I422");
+static const arg_def_t use_i444 = ARG_DEF(
+ NULL, "i444", 0, "Input file is I444");
+static const arg_def_t use_i440 = ARG_DEF(
+ NULL, "i440", 0, "Input file is I440");
+static const arg_def_t codecarg = ARG_DEF(
+ NULL, "codec", 1, "Codec to use");
+static const arg_def_t passes = ARG_DEF(
+ "p", "passes", 1, "Number of passes (1/2)");
+static const arg_def_t pass_arg = ARG_DEF(
+ NULL, "pass", 1, "Pass to execute (1/2)");
+static const arg_def_t fpf_name = ARG_DEF(
+ NULL, "fpf", 1, "First pass statistics file name");
#if CONFIG_FP_MB_STATS
-static const arg_def_t fpmbf_name = ARG_DEF(NULL, "fpmbf", 1,
- "First pass block statistics file name");
+static const arg_def_t fpmbf_name = ARG_DEF(
+ NULL, "fpmbf", 1, "First pass block statistics file name");
#endif
-static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
- "Stop encoding after n input frames");
-static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
- "Skip the first n input frames");
-static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
- "Deadline per frame (usec)");
-static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
- "Use Best Quality Deadline");
-static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
- "Use Good Quality Deadline");
-static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
- "Use Realtime Quality Deadline");
-static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
- "Do not print encode progress");
-static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
- "Show encoder parameters");
-static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
- "Show PSNR in status line");
+static const arg_def_t limit = ARG_DEF(
+ NULL, "limit", 1, "Stop encoding after n input frames");
+static const arg_def_t skip = ARG_DEF(
+ NULL, "skip", 1, "Skip the first n input frames");
+static const arg_def_t deadline = ARG_DEF(
+ "d", "deadline", 1, "Deadline per frame (usec)");
+static const arg_def_t best_dl = ARG_DEF(
+ NULL, "best", 0, "Use Best Quality Deadline");
+static const arg_def_t good_dl = ARG_DEF(
+ NULL, "good", 0, "Use Good Quality Deadline");
+static const arg_def_t rt_dl = ARG_DEF(
+ NULL, "rt", 0, "Use Realtime Quality Deadline");
+static const arg_def_t quietarg = ARG_DEF(
+ "q", "quiet", 0, "Do not print encode progress");
+static const arg_def_t verbosearg = ARG_DEF(
+ "v", "verbose", 0, "Show encoder parameters");
+static const arg_def_t psnrarg = ARG_DEF(
+ NULL, "psnr", 0, "Show PSNR in status line");
static const struct arg_enum_list test_decode_enum[] = {
{"off", TEST_DECODE_OFF},
@@ -177,28 +179,25 @@
{"warn", TEST_DECODE_WARN},
{NULL, 0}
};
-static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
- "Test encode/decode mismatch",
- test_decode_enum);
-static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
- "Stream frame rate (rate/scale)");
-static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
- "Output IVF (default is WebM if WebM IO is enabled)");
-static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
- "Makes encoder output partitions. Requires IVF output!");
-static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
- "Show quantizer histogram (n-buckets)");
-static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
- "Show rate histogram (n-buckets)");
-static const arg_def_t disable_warnings =
- ARG_DEF(NULL, "disable-warnings", 0,
- "Disable warnings about potentially incorrect encode settings.");
-static const arg_def_t disable_warning_prompt =
- ARG_DEF("y", "disable-warning-prompt", 0,
- "Display warnings, but do not prompt user to continue.");
-static const arg_def_t experimental_bitstream =
- ARG_DEF(NULL, "experimental-bitstream", 0,
- "Allow experimental bitstream features.");
+static const arg_def_t recontest = ARG_DEF_ENUM(
+ NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
+static const arg_def_t framerate = ARG_DEF(
+ NULL, "fps", 1, "Stream frame rate (rate/scale)");
+static const arg_def_t use_ivf = ARG_DEF(
+ NULL, "ivf", 0, "Output IVF (default is WebM if WebM IO is enabled)");
+static const arg_def_t out_part = ARG_DEF(
+ "P", "output-partitions", 0,
+ "Makes encoder output partitions. Requires IVF output!");
+static const arg_def_t q_hist_n = ARG_DEF(
+ NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
+static const arg_def_t rate_hist_n = ARG_DEF(
+ NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
+static const arg_def_t disable_warnings = ARG_DEF(
+ NULL, "disable-warnings", 0,
+ "Disable warnings about potentially incorrect encode settings.");
+static const arg_def_t disable_warning_prompt = ARG_DEF(
+ "y", "disable-warning-prompt", 0,
+ "Display warnings, but do not prompt user to continue.");
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
static const arg_def_t test16bitinternalarg = ARG_DEF(
@@ -214,16 +213,14 @@
NULL
};
-static const arg_def_t usage = ARG_DEF("u", "usage", 1,
- "Usage profile number to use");
-static const arg_def_t threads = ARG_DEF("t", "threads", 1,
- "Max number of threads to use");
-static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
- "Bitstream profile number to use");
-static const arg_def_t width = ARG_DEF("w", "width", 1,
- "Frame width");
-static const arg_def_t height = ARG_DEF("h", "height", 1,
- "Frame height");
+static const arg_def_t usage = ARG_DEF(
+ "u", "usage", 1, "Usage profile number to use");
+static const arg_def_t threads = ARG_DEF(
+ "t", "threads", 1, "Max number of threads to use");
+static const arg_def_t profile = ARG_DEF(
+ NULL, "profile", 1, "Bitstream profile number to use");
+static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
+static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
#if CONFIG_WEBM_IO
static const struct arg_enum_list stereo_mode_enum[] = {
{"mono", STEREO_FORMAT_MONO},
@@ -233,18 +230,18 @@
{"right-left", STEREO_FORMAT_RIGHT_LEFT},
{NULL, 0}
};
-static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
- "Stereo 3D video format", stereo_mode_enum);
+static const arg_def_t stereo_mode = ARG_DEF_ENUM(
+ NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
#endif
-static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
- "Output timestamp precision (fractional seconds)");
-static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
- "Enable error resiliency features");
-static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
- "Max number of frames to lag");
+static const arg_def_t timebase = ARG_DEF(
+ NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
+static const arg_def_t error_resilient = ARG_DEF(
+ NULL, "error-resilient", 1, "Enable error resiliency features");
+static const arg_def_t lag_in_frames = ARG_DEF(
+ NULL, "lag-in-frames", 1, "Max number of frames to lag");
static const arg_def_t *global_args[] = {
- &use_yv12, &use_i420, &use_i422, &use_i444,
+ &use_yv12, &use_i420, &use_i422, &use_i444, &use_i440,
&usage, &threads, &profile,
&width, &height,
#if CONFIG_WEBM_IO
@@ -258,18 +255,18 @@
&lag_in_frames, NULL
};
-static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
- "Temporal resampling threshold (buf %)");
-static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
- "Spatial resampling enabled (bool)");
-static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
- "Width of encoded frame");
-static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
- "Height of encoded frame");
-static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
- "Upscale threshold (buf %)");
-static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
- "Downscale threshold (buf %)");
+static const arg_def_t dropframe_thresh = ARG_DEF(
+ NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
+static const arg_def_t resize_allowed = ARG_DEF(
+ NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
+static const arg_def_t resize_width = ARG_DEF(
+ NULL, "resize-width", 1, "Width of encoded frame");
+static const arg_def_t resize_height = ARG_DEF(
+ NULL, "resize-height", 1, "Height of encoded frame");
+static const arg_def_t resize_up_thresh = ARG_DEF(
+ NULL, "resize-up", 1, "Upscale threshold (buf %)");
+static const arg_def_t resize_down_thresh = ARG_DEF(
+ NULL, "resize-down", 1, "Downscale threshold (buf %)");
static const struct arg_enum_list end_usage_enum[] = {
{"vbr", VPX_VBR},
{"cbr", VPX_CBR},
@@ -277,24 +274,24 @@
{"q", VPX_Q},
{NULL, 0}
};
-static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
- "Rate control mode", end_usage_enum);
-static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
- "Bitrate (kbps)");
-static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
- "Minimum (best) quantizer");
-static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
- "Maximum (worst) quantizer");
-static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
- "Datarate undershoot (min) target (%)");
-static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
- "Datarate overshoot (max) target (%)");
-static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
- "Client buffer size (ms)");
-static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
- "Client initial buffer size (ms)");
-static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
- "Client optimal buffer size (ms)");
+static const arg_def_t end_usage = ARG_DEF_ENUM(
+ NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
+static const arg_def_t target_bitrate = ARG_DEF(
+ NULL, "target-bitrate", 1, "Bitrate (kbps)");
+static const arg_def_t min_quantizer = ARG_DEF(
+ NULL, "min-q", 1, "Minimum (best) quantizer");
+static const arg_def_t max_quantizer = ARG_DEF(
+ NULL, "max-q", 1, "Maximum (worst) quantizer");
+static const arg_def_t undershoot_pct = ARG_DEF(
+ NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
+static const arg_def_t overshoot_pct = ARG_DEF(
+ NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
+static const arg_def_t buf_sz = ARG_DEF(
+ NULL, "buf-sz", 1, "Client buffer size (ms)");
+static const arg_def_t buf_initial_sz = ARG_DEF(
+ NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
+static const arg_def_t buf_optimal_sz = ARG_DEF(
+ NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
static const arg_def_t *rc_args[] = {
&dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
&resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
@@ -303,59 +300,59 @@
};
-static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
- "CBR/VBR bias (0=CBR, 100=VBR)");
-static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
- "GOP min bitrate (% of target)");
-static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
- "GOP max bitrate (% of target)");
+static const arg_def_t bias_pct = ARG_DEF(
+ NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
+static const arg_def_t minsection_pct = ARG_DEF(
+ NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
+static const arg_def_t maxsection_pct = ARG_DEF(
+ NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
static const arg_def_t *rc_twopass_args[] = {
&bias_pct, &minsection_pct, &maxsection_pct, NULL
};
-static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
- "Minimum keyframe interval (frames)");
-static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
- "Maximum keyframe interval (frames)");
-static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
- "Disable keyframe placement");
+static const arg_def_t kf_min_dist = ARG_DEF(
+ NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
+static const arg_def_t kf_max_dist = ARG_DEF(
+ NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
+static const arg_def_t kf_disabled = ARG_DEF(
+ NULL, "disable-kf", 0, "Disable keyframe placement");
static const arg_def_t *kf_args[] = {
&kf_min_dist, &kf_max_dist, &kf_disabled, NULL
};
-static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
- "Noise sensitivity (frames to blur)");
-static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
- "Loop filter sharpness (0..7)");
-static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
- "Motion detection threshold");
-static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
- "CPU Used (-16..16)");
-static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
- "Enable automatic alt reference frames");
-static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
- "AltRef max frames (0..15)");
-static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
- "AltRef filter strength (0..6)");
-static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
- "AltRef type");
+static const arg_def_t noise_sens = ARG_DEF(
+ NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
+static const arg_def_t sharpness = ARG_DEF(
+ NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
+static const arg_def_t static_thresh = ARG_DEF(
+ NULL, "static-thresh", 1, "Motion detection threshold");
+static const arg_def_t cpu_used = ARG_DEF(
+ NULL, "cpu-used", 1, "CPU Used (-16..16)");
+static const arg_def_t auto_altref = ARG_DEF(
+ NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
+static const arg_def_t arnr_maxframes = ARG_DEF(
+ NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
+static const arg_def_t arnr_strength = ARG_DEF(
+ NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
+static const arg_def_t arnr_type = ARG_DEF(
+ NULL, "arnr-type", 1, "AltRef type");
static const struct arg_enum_list tuning_enum[] = {
{"psnr", VP8_TUNE_PSNR},
{"ssim", VP8_TUNE_SSIM},
{NULL, 0}
};
-static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
- "Material to favor", tuning_enum);
-static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
- "Constant/Constrained Quality level");
-static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
- "Max I-frame bitrate (pct)");
+static const arg_def_t tune_ssim = ARG_DEF_ENUM(
+ NULL, "tune", 1, "Material to favor", tuning_enum);
+static const arg_def_t cq_level = ARG_DEF(
+ NULL, "cq-level", 1, "Constant/Constrained Quality level");
+static const arg_def_t max_intra_rate_pct = ARG_DEF(
+ NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
#if CONFIG_VP8_ENCODER
-static const arg_def_t token_parts =
- ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
+static const arg_def_t token_parts = ARG_DEF(
+ NULL, "token-parts", 1, "Number of token partitions to use, log2");
static const arg_def_t *vp8_args[] = {
&cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
&token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
@@ -373,11 +370,12 @@
#endif
#if CONFIG_VP9_ENCODER
-static const arg_def_t tile_cols =
- ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
-static const arg_def_t tile_rows =
- ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
-static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
+static const arg_def_t tile_cols = ARG_DEF(
+ NULL, "tile-columns", 1, "Number of tile columns to use, log2");
+static const arg_def_t tile_rows = ARG_DEF(
+ NULL, "tile-rows", 1, "Number of tile rows to use, log2");
+static const arg_def_t lossless = ARG_DEF(
+ NULL, "lossless", 1, "Lossless mode");
static const arg_def_t frame_parallel_decoding = ARG_DEF(
NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
static const arg_def_t aq_mode = ARG_DEF(
@@ -396,13 +394,12 @@
{NULL, 0}
};
-static const arg_def_t bitdeptharg = ARG_DEF_ENUM("b", "bit-depth", 1,
- "Bit depth for codec "
- "(8 for version <=1, "
- "10 or 12 for version 2)",
- bitdepth_enum);
-static const arg_def_t inbitdeptharg = ARG_DEF(NULL, "input-bit-depth", 1,
- "Bit depth of input");
+static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
+ "b", "bit-depth", 1,
+ "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
+ bitdepth_enum);
+static const arg_def_t inbitdeptharg = ARG_DEF(
+ NULL, "input-bit-depth", 1, "Bit depth of input");
#endif
static const struct arg_enum_list tune_content_enum[] = {
@@ -829,6 +826,8 @@
global->color_type = I422;
else if (arg_match(&arg, &use_i444, argi))
global->color_type = I444;
+ else if (arg_match(&arg, &use_i440, argi))
+ global->color_type = I440;
else if (arg_match(&arg, &quietarg, argi))
global->quiet = 1;
else if (arg_match(&arg, &verbosearg, argi))
@@ -857,8 +856,6 @@
global->disable_warnings = 1;
else if (arg_match(&arg, &disable_warning_prompt, argi))
global->disable_warning_prompt = 1;
- else if (arg_match(&arg, &experimental_bitstream, argi))
- global->experimental_bitstream = 1;
else
argj++;
}
@@ -1162,7 +1159,6 @@
if (j == config->arg_ctrl_cnt)
config->arg_ctrl_cnt++;
}
-
}
}
if (!match)
@@ -1196,12 +1192,6 @@
fatal("Stream %d: Specify stream dimensions with --width (-w) "
" and --height (-h)", stream->index);
- if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
- fatal("Stream %d: profile %d is experimental and requires the --%s flag",
- stream->index, stream->config.cfg.g_profile,
- experimental_bitstream.long_name);
- }
-
// Check that the codec bit depth is greater than the input bit depth.
if (stream->config.cfg.g_input_bit_depth >
(unsigned int)stream->config.cfg.g_bit_depth) {
@@ -1288,7 +1278,12 @@
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_I440: return "I440";
case VPX_IMG_FMT_YV12: return "YV12";
+ case VPX_IMG_FMT_I42016: return "I42016";
+ case VPX_IMG_FMT_I42216: return "I42216";
+ case VPX_IMG_FMT_I44416: return "I44416";
+ case VPX_IMG_FMT_I44016: return "I44016";
default: return "Other";
}
}
@@ -1695,7 +1690,7 @@
}
-static void show_psnr(struct stream_state *stream) {
+static void show_psnr(struct stream_state *stream, double peak) {
int i;
double ovpsnr;
@@ -1703,7 +1698,7 @@
return;
fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
- ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
+ ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
(double)stream->psnr_sse_total);
fprintf(stderr, " %.3f", ovpsnr);
@@ -1718,132 +1713,6 @@
return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
}
-#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
-static void high_img_upshift(vpx_image_t *dst, vpx_image_t *src,
- int input_shift) {
- // Note the offset is 1 less than half
- const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0;
- int plane;
- if (dst->w != src->w || dst->h != src->h ||
- dst->x_chroma_shift != src->x_chroma_shift ||
- dst->y_chroma_shift != src->y_chroma_shift ||
- dst->fmt != src->fmt || input_shift < 0) {
- fatal("Unsupported image conversion");
- }
- switch (src->fmt) {
- case VPX_IMG_FMT_I42016:
- case VPX_IMG_FMT_I42216:
- case VPX_IMG_FMT_I44416:
- break;
- default:
- fatal("Unsupported image conversion");
- break;
- }
- for (plane = 0; plane < 3; plane++) {
- int w = src->w;
- int h = src->h;
- int x, y;
- if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
- }
- for (y = 0; y < h; y++) {
- uint16_t *p_src = (uint16_t *)(src->planes[plane] +
- y * src->stride[plane]);
- uint16_t *p_dst = (uint16_t *)(dst->planes[plane] +
- y * dst->stride[plane]);
- for (x = 0; x < w; x++)
- *p_dst++ = (*p_src++ << input_shift) + offset;
- }
- }
-}
-
-static void low_img_upshift(vpx_image_t *dst, vpx_image_t *src,
- int input_shift) {
- // Note the offset is 1 less than half
- const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0;
- int plane;
- if (dst->w != src->w || dst->h != src->h ||
- dst->x_chroma_shift != src->x_chroma_shift ||
- dst->y_chroma_shift != src->y_chroma_shift ||
- dst->fmt != src->fmt + VPX_IMG_FMT_HIGHBITDEPTH ||
- input_shift < 0) {
- fatal("Unsupported image conversion");
- }
- switch (src->fmt) {
- case VPX_IMG_FMT_I420:
- case VPX_IMG_FMT_I422:
- case VPX_IMG_FMT_I444:
- break;
- default:
- fatal("Unsupported image conversion");
- break;
- }
- for (plane = 0; plane < 3; plane++) {
- int w = src->w;
- int h = src->h;
- int x, y;
- if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
- }
- for (y = 0; y < h; y++) {
- uint8_t *p_src = src->planes[plane] + y * src->stride[plane];
- uint16_t *p_dst = (uint16_t *)(dst->planes[plane] +
- y * dst->stride[plane]);
- for (x = 0; x < w; x++) {
- *p_dst++ = (*p_src++ << input_shift) + offset;
- }
- }
- }
-}
-
-static void img_upshift(vpx_image_t *dst, vpx_image_t *src,
- int input_shift) {
- if (src->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
- high_img_upshift(dst, src, input_shift);
- } else {
- low_img_upshift(dst, src, input_shift);
- }
-}
-
-static void img_cast_16_to_8(vpx_image_t *dst, vpx_image_t *src) {
- int plane;
- if (dst->fmt + VPX_IMG_FMT_HIGHBITDEPTH != src->fmt ||
- dst->d_w != src->d_w || dst->d_h != src->d_h ||
- dst->x_chroma_shift != src->x_chroma_shift ||
- dst->y_chroma_shift != src->y_chroma_shift) {
- fatal("Unsupported image conversion");
- }
- switch (dst->fmt) {
- case VPX_IMG_FMT_I420:
- case VPX_IMG_FMT_I422:
- case VPX_IMG_FMT_I444:
- break;
- default:
- fatal("Unsupported image conversion");
- break;
- }
- for (plane = 0; plane < 3; plane++) {
- int w = src->d_w;
- int h = src->d_h;
- int x, y;
- if (plane) {
- w >>= src->x_chroma_shift;
- h >>= src->y_chroma_shift;
- }
- for (y = 0; y < h; y++) {
- uint16_t *p_src = (uint16_t *)(src->planes[plane] +
- y * src->stride[plane]);
- uint8_t *p_dst = dst->planes[plane] + y * dst->stride[plane];
- for (x = 0; x < w; x++) {
- *p_dst++ = *p_src++;
- }
- }
- }
-}
-#endif
-
static void test_decode(struct stream_state *stream,
enum TestDecodeFatality fatal,
const VpxInterface *codec) {
@@ -1883,12 +1752,12 @@
if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
enc_img.d_w, enc_img.d_h, 16);
- img_cast_16_to_8(&enc_img, &ref_enc.img);
+ vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
}
if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
dec_img.d_w, dec_img.d_h, 16);
- img_cast_16_to_8(&dec_img, &ref_dec.img);
+ vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
}
}
#endif
@@ -1993,6 +1862,9 @@
case I444:
input.fmt = VPX_IMG_FMT_I444;
break;
+ case I440:
+ input.fmt = VPX_IMG_FMT_I440;
+ break;
case YV12:
input.fmt = VPX_IMG_FMT_YV12;
break;
@@ -2042,14 +1914,15 @@
/* If the input file doesn't specify its w/h (raw files), try to get
* the data from the first stream's configuration.
*/
- if (!input.width || !input.height)
- FOREACH_STREAM( {
- if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
- input.width = stream->config.cfg.g_w;
- input.height = stream->config.cfg.g_h;
- break;
- }
- });
+ if (!input.width || !input.height) {
+ FOREACH_STREAM({
+ if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
+ input.width = stream->config.cfg.g_w;
+ input.height = stream->config.cfg.g_h;
+ break;
+ }
+ });
+ }
/* Update stream configurations from the input file's parameters */
if (!input.width || !input.height)
@@ -2196,7 +2069,7 @@
input.width, input.height, 32);
allocated_raw_shift = 1;
}
- img_upshift(&raw_shift, &raw, input_shift);
+ vpx_img_upshift(&raw_shift, &raw, input_shift);
frame_to_encode = &raw_shift;
} else {
frame_to_encode = &raw;
@@ -2272,24 +2145,29 @@
if (stream_cnt > 1)
fprintf(stderr, "\n");
- if (!global.quiet)
- FOREACH_STREAM(fprintf(
- stderr,
- "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
- " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
- global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
- seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
- seen_frames ? (int64_t)stream->nbytes * 8
- * (int64_t)global.framerate.num / global.framerate.den
- / seen_frames
- : 0,
- stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
- stream->cx_time > 9999999 ? "ms" : "us",
- usec_to_fps(stream->cx_time, seen_frames));
- );
+ if (!global.quiet) {
+ FOREACH_STREAM(fprintf(stderr,
+ "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
+ " %7"PRId64" %s (%.2f fps)\033[K\n",
+ pass + 1,
+ global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
+ seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
+ seen_frames ? (int64_t)stream->nbytes * 8 *
+ (int64_t)global.framerate.num / global.framerate.den /
+ seen_frames : 0,
+ stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
+ stream->cx_time > 9999999 ? "ms" : "us",
+ usec_to_fps(stream->cx_time, seen_frames)));
+ }
- if (global.show_psnr)
- FOREACH_STREAM(show_psnr(stream));
+ if (global.show_psnr) {
+ if (global.codec->fourcc == VP9_FOURCC) {
+ FOREACH_STREAM(
+ show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
+ } else {
+ FOREACH_STREAM(show_psnr(stream, 255.0));
+ }
+ }
FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
diff --git a/source/libvpx/vpxenc.h b/source/libvpx/vpxenc.h
index 3d6728e..d867e9d 100644
--- a/source/libvpx/vpxenc.h
+++ b/source/libvpx/vpxenc.h
@@ -26,6 +26,7 @@
I420, // 4:2:0 8+ bit-depth
I422, // 4:2:2 8+ bit-depth
I444, // 4:4:4 8+ bit-depth
+ I440, // 4:4:0 8+ bit-depth
YV12, // 4:2:0 with uv flipped, only 8-bit depth
} ColorInputType;
