]> granicus.if.org Git - libvpx/log
libvpx
8 years agoMerge "Replace arbitrary number with defs" into nextgenv2
Jingning Han [Wed, 10 Feb 2016 16:50:04 +0000 (16:50 +0000)]
Merge "Replace arbitrary number with defs" into nextgenv2

8 years agoMerge "Entropy coding for dynamic ref mv modes" into nextgenv2
Jingning Han [Wed, 10 Feb 2016 16:49:45 +0000 (16:49 +0000)]
Merge "Entropy coding for dynamic ref mv modes" into nextgenv2

8 years agoReplace arbitrary number with defs
Jingning Han [Tue, 9 Feb 2016 23:31:27 +0000 (15:31 -0800)]
Replace arbitrary number with defs

Change-Id: Ia5a68f26c67d13d3f2dd3b3f8afabb781e2c8f73

8 years agoRemove C99 struct init syle in fwd/inv config
Angie Chiang [Tue, 9 Feb 2016 23:38:27 +0000 (15:38 -0800)]
Remove C99 struct init syle in fwd/inv config

Change-Id: Ieeb458353af6c903445518eef60328c62ca5c741

8 years agoMerge "Fix partition type costing." into nextgenv2
Jingning Han [Tue, 9 Feb 2016 22:31:14 +0000 (22:31 +0000)]
Merge "Fix partition type costing." into nextgenv2

8 years agoMerge "A variety of fixes for supertx/var-tx rd costing" into nextgenv2
Debargha Mukherjee [Tue, 9 Feb 2016 17:16:23 +0000 (17:16 +0000)]
Merge "A variety of fixes for supertx/var-tx rd costing" into nextgenv2

8 years agoFix partition type costing.
Geza Lore [Tue, 9 Feb 2016 10:17:22 +0000 (10:17 +0000)]
Fix partition type costing.

This patch makes rd optimization use the same context for computing
the rate cost of coding the partitioning as the packer actually uses
when emitting it in write_modes_sb.

Change-Id: Idb1427bb2f9c37ab80c6aa182f7ff754ef0595cb

8 years agoA variety of fixes for supertx/var-tx rd costing
Debargha Mukherjee [Mon, 8 Feb 2016 17:34:43 +0000 (09:34 -0800)]
A variety of fixes for supertx/var-tx rd costing

Change-Id: I8a3d59378abb1dfa4e614b2975c2db05d4224bd5

8 years agoMerge "Fix a bug in HBD buffer size computation" into nextgenv2
Yaowu Xu [Tue, 9 Feb 2016 04:40:46 +0000 (04:40 +0000)]
Merge "Fix a bug in HBD buffer size computation" into nextgenv2

8 years agoFix a bug in HBD buffer size computation
Yaowu Xu [Tue, 9 Feb 2016 02:31:30 +0000 (18:31 -0800)]
Fix a bug in HBD buffer size computation

The value of use_highbitdepth flag is used for compute the size for
high bit depth buffer allocation, which should take value 0 or 1
depending on if the buffer is used for high bit depth or not.
Previously, the values is set to 8 or 0, this commit fixes the issue
and properly set the value for this flag to 1 or 0.

This cuts the size of highbitdepth buffer memory allocation to 2/9 of
the size prior to the fix.

Change-Id: I401518b5a6147e5d8a973e54f7ca6bc1892065e0

8 years agoEntropy coding for dynamic ref mv modes
Jingning Han [Fri, 22 Jan 2016 02:07:31 +0000 (18:07 -0800)]
Entropy coding for dynamic ref mv modes

This commit enables entropy coding for dynamic reference motion
vector modes. The probability model is contexted on the ranking
categories of the reference motion vector candidates.

Change-Id: I09b58d98a409d63ec1a407331e29f8945b7ef17d

8 years agoExplicitly set tx_type for sub8x8 blocks
Debargha Mukherjee [Mon, 8 Feb 2016 23:59:17 +0000 (15:59 -0800)]
Explicitly set tx_type for sub8x8 blocks

Fixes an issue where the tx_type was not set correctly for
sub8x8 inter and intra blocks. In the current syntax, for
sub8x8 blocks, there is still a single tx_type that is
transmitted. Ideally, this should be searched for the best
rd performance, albeit at the expense of encode speed.
For now, we just set it to DCT_DCT. Previously it was left
incorrectly as what was used for the previous non sub8x8
block.

derflr: BDRATE -0.277%

Change-Id: If76ba903bfbfd4d374cf1ac7d1daee50e92f0edd

8 years agoMerge "Enable dynamic ref motion vector mode for compound inter block" into nextgenv2
Jingning Han [Tue, 9 Feb 2016 00:04:36 +0000 (00:04 +0000)]
Merge "Enable dynamic ref motion vector mode for compound inter block" into nextgenv2

8 years agoMerge "Change to use local variables consistently" into nextgenv2
Yaowu Xu [Mon, 8 Feb 2016 23:58:23 +0000 (23:58 +0000)]
Merge "Change to use local variables consistently" into nextgenv2

8 years agoMerge "Remove a flavor of SSIM that is never really used." into nextgenv2
Yaowu Xu [Mon, 8 Feb 2016 23:39:04 +0000 (23:39 +0000)]
Merge "Remove a flavor of SSIM that is never really used." into nextgenv2

8 years agoMerge "Set a max dB value for PSNR_HVS and FAST_SSIM" into nextgenv2
Yaowu Xu [Mon, 8 Feb 2016 23:38:57 +0000 (23:38 +0000)]
Merge "Set a max dB value for PSNR_HVS and FAST_SSIM" into nextgenv2

8 years agoMerge "Fix msvc compiler warnings" into nextgenv2
Yaowu Xu [Mon, 8 Feb 2016 23:38:48 +0000 (23:38 +0000)]
Merge "Fix msvc compiler warnings" into nextgenv2

8 years agoMerge "Normalize fdct8x8 in psnrhvs computation" into nextgenv2
Yaowu Xu [Mon, 8 Feb 2016 23:38:41 +0000 (23:38 +0000)]
Merge "Normalize fdct8x8 in psnrhvs computation" into nextgenv2

8 years agoChange to use local variables consistently
Yaowu Xu [Fri, 5 Feb 2016 16:17:01 +0000 (08:17 -0800)]
Change to use local variables consistently

This commit does not change the computation, nor results.

Change-Id: I1a7bb47050220d970f075458b507c5e55d93b22e

8 years agoEnable dynamic ref motion vector mode for compound inter block
Jingning Han [Thu, 21 Jan 2016 01:40:47 +0000 (17:40 -0800)]
Enable dynamic ref motion vector mode for compound inter block

This commit enables the dynamic reference motion vector coding mode
for the compound inter blocks.

Change-Id: Ibe78fd8de6989db392cd67a9d81a69d680345ba1

8 years agoRemove a flavor of SSIM that is never really used.
Yaowu Xu [Thu, 4 Feb 2016 22:16:56 +0000 (14:16 -0800)]
Remove a flavor of SSIM that is never really used.

Change-Id: I61ea7f63acbcfeecd3f7dba5a5a38b980efc802b

8 years agoSet a max dB value for PSNR_HVS and FAST_SSIM
Yaowu Xu [Thu, 4 Feb 2016 22:03:48 +0000 (14:03 -0800)]
Set a max dB value for PSNR_HVS and FAST_SSIM

Now set at 100.0 instead of infinite

Change-Id: I41bae0c4bd95a26f9819584e7311b7945df1271a

8 years agoMerge "Experiment: use 12 taps for sharp filter" into nextgenv2
Angie Chiang [Mon, 8 Feb 2016 18:33:51 +0000 (18:33 +0000)]
Merge "Experiment: use 12 taps for sharp filter" into nextgenv2

8 years agoMerge "add convolution function with adjustable length" into nextgenv2
Angie Chiang [Mon, 8 Feb 2016 18:32:38 +0000 (18:32 +0000)]
Merge "add convolution function with adjustable length" into nextgenv2

8 years agoMerge "Fixed warnings of unused functions" into nextgenv2
Yaowu Xu [Mon, 8 Feb 2016 18:18:47 +0000 (18:18 +0000)]
Merge "Fixed warnings of unused functions" into nextgenv2

8 years agoFix msvc compiler warnings
Yaowu Xu [Mon, 8 Feb 2016 17:41:43 +0000 (09:41 -0800)]
Fix msvc compiler warnings

There were a number of compiler warnings:
1. int16_t to uint8_t in recon_intra.c;
2. double to float conversions in psnrhvs.c
3. intptr_t to int in quantize.c
4. size_t to int32_t in decoder.c

Change-Id: Id95423b17779dcfa6cf39d9a90fe8cb8b910f5df

8 years agoMerge "Fix bad merge artifacts" into nextgenv2
Yaowu Xu [Mon, 8 Feb 2016 17:25:54 +0000 (17:25 +0000)]
Merge "Fix bad merge artifacts" into nextgenv2

8 years agoNormalize fdct8x8 in psnrhvs computation
Yaowu Xu [Fri, 5 Feb 2016 00:15:42 +0000 (16:15 -0800)]
Normalize fdct8x8 in psnrhvs computation

This is to match the scale to the fdct8x8 used in original daala
psnrhvs computation.

Change-Id: Ic30b50747ba9c340bcb679f7439640046c69f90a

8 years agoExperiment: use 12 taps for sharp filter
Angie Chiang [Fri, 5 Feb 2016 19:05:56 +0000 (11:05 -0800)]
Experiment: use 12 taps for sharp filter

Set USE_12_SHARP_FILTER to 1 to turn on the experiment
The psnr percentages increase
       derf   stdhd
lowbd  +0.332 +0.318
highbd +0.476 +0.507

Change-Id: I783c0fc764ee8541645e100453c9b2073924e209

8 years agoadd convolution function with adjustable length
Angie Chiang [Fri, 5 Feb 2016 00:13:18 +0000 (16:13 -0800)]
add convolution function with adjustable length

Change-Id: I1a5b1e15a188ef11594d0c6ac0dbd42aac59cfca

8 years agoFixed warnings of unused functions
Yaowu Xu [Fri, 5 Feb 2016 17:35:34 +0000 (09:35 -0800)]
Fixed warnings of unused functions

And enabled the warning flag in configure for vp10.

Change-Id: If556d6fac65755af3d6ed7fe71b8eca0ef1b1965

8 years agoMerge "Add a speed feature for intra filter search" into nextgenv2
Hui Su [Fri, 5 Feb 2016 18:49:24 +0000 (18:49 +0000)]
Merge "Add a speed feature for intra filter search" into nextgenv2

8 years agoFix bad merge artifacts
Yaowu Xu [Fri, 5 Feb 2016 15:50:57 +0000 (07:50 -0800)]
Fix bad merge artifacts

Temporaly disable warning for unused function for vp10, needs clean
out the warnings before re-enable the flag for vp10.

Change-Id: I5636f8cd607423f6ea6963db9c2cbd688e30b495

8 years agoMerge branch 'master' into nextgenv2
Yaowu Xu [Fri, 5 Feb 2016 13:00:06 +0000 (05:00 -0800)]
Merge branch 'master' into nextgenv2

8 years agoMerge "configure: enable -Wunused-function by default"
James Zern [Fri, 5 Feb 2016 02:45:56 +0000 (02:45 +0000)]
Merge "configure: enable -Wunused-function by default"

8 years agoMerge "Vidyo patch: Optimization for 1-to-2 downsampling and upsampling."
James Zern [Fri, 5 Feb 2016 02:36:03 +0000 (02:36 +0000)]
Merge "Vidyo patch: Optimization for 1-to-2 downsampling and upsampling."

8 years agoMerge "Add 8-tap interpolation filter options for intra prediction" into nextgenv2
Hui Su [Thu, 4 Feb 2016 23:06:28 +0000 (23:06 +0000)]
Merge "Add 8-tap interpolation filter options for intra prediction" into nextgenv2

8 years agoVidyo patch: Optimization for 1-to-2 downsampling and upsampling.
Scott LaVarnway [Thu, 4 Feb 2016 22:50:26 +0000 (14:50 -0800)]
Vidyo patch: Optimization for 1-to-2 downsampling and upsampling.

Change-Id: I9cc9780f506e025aea57485a9e21f0835faf173c

8 years agoMerge "Generalize the dynamic reference motion vector coding mode" into nextgenv2
Jingning Han [Thu, 4 Feb 2016 22:28:56 +0000 (22:28 +0000)]
Merge "Generalize the dynamic reference motion vector coding mode" into nextgenv2

8 years agomerge per-segment lossless feature in misc_fixes
Yaowu Xu [Thu, 4 Feb 2016 16:31:16 +0000 (08:31 -0800)]
merge per-segment lossless feature in misc_fixes

Change-Id: I56d56781d371c99aa5cdd2db1cbc0a17437723e9

8 years agoDefine a macro to replace hardcoded values
Zoe Liu [Thu, 4 Feb 2016 13:33:22 +0000 (05:33 -0800)]
Define a macro to replace hardcoded values

The definition is for the number of frames to check to determine the
recent decay rate, further to determine the next key frame in the
first pass of the encoder.

Change-Id: Ic696d6eb518a86fa296842273cf8767ef0b0e27a

8 years agoconfigure: enable -Wunused-function by default
James Zern [Tue, 2 Feb 2016 04:52:16 +0000 (20:52 -0800)]
configure: enable -Wunused-function by default

when INLINE is defined and mips is not being targeted. otherwise keep
the old --enable-extra-warnings behavior

Change-Id: Iba576edbe5fca03efa56ce99eee11f9cafc573ad

8 years agofdct8x8_test:: quiet unused-function warning
James Zern [Thu, 4 Feb 2016 07:04:42 +0000 (23:04 -0800)]
fdct8x8_test:: quiet unused-function warning

in non-x86 builds

Change-Id: I9f6fa40d27f95a4afe92444f71c23dc9a15db466

8 years agoMerge changes from topic 'dead-code'
James Zern [Thu, 4 Feb 2016 06:25:44 +0000 (06:25 +0000)]
Merge changes from topic 'dead-code'

* changes:
  yv12config.c: remove dead CONFIG_ALPHA code
  vp9_encoder.c: remove unused macro
  vpx_mem.c: remove unused macro
  vp9_svc_layercontext.c: relocate a macro definition
  vp9_encoder.c: protect SNPRINT* macros w/CONFIG check
  vp9_aq_360.c: remove some unused macros
  vp9_diamond_search_sad_avx.c: rename (un)likely macros
  vp9_resize.c: add missing include
  vp9_aq_complexity.c: remove unused macros
  vp9_detokenize.c: remove unused macros

8 years agoMerge "Fix for supertx: ignore skip_txfm optimization" into nextgenv2
Debargha Mukherjee [Thu, 4 Feb 2016 04:28:29 +0000 (04:28 +0000)]
Merge "Fix for supertx: ignore skip_txfm optimization" into nextgenv2

8 years agoMerge changes I1171e9c7,I77292e11,I1678ab48,Id7a2c345
James Zern [Thu, 4 Feb 2016 03:43:49 +0000 (03:43 +0000)]
Merge changes I1171e9c7,I77292e11,I1678ab48,Id7a2c345

* changes:
  dct16x16_test: quiet unused-function warning
  intrapred: protect functions w/CONFIG check x2
  dct32x32_test: remove unused function
  tests: quiet some warnings w/--disable-webm-io

8 years agoMerge "vp9-skinmap. Some adjustments for model=1."
Marco Paniconi [Wed, 3 Feb 2016 23:34:07 +0000 (23:34 +0000)]
Merge "vp9-skinmap. Some adjustments for model=1."

8 years agoAdd a speed feature for intra filter search
hui su [Wed, 3 Feb 2016 19:32:25 +0000 (11:32 -0800)]
Add a speed feature for intra filter search

Seperate the prediction angle search and fitler search.
It can reduce the computation overhead of filter search by as much
as 85%, while keeping more than 50% of the coding gain.

Change-Id: Id152f71e20ebcaca8b429bdd4ca1fbeb646fc6bf

8 years agoMerge "Merge branch 'master' into nextgenv2" into nextgenv2
Yaowu Xu [Wed, 3 Feb 2016 22:54:25 +0000 (22:54 +0000)]
Merge "Merge branch 'master' into nextgenv2" into nextgenv2

8 years agovp9-skinmap. Some adjustments for model=1.
Marco [Wed, 3 Feb 2016 22:21:59 +0000 (14:21 -0800)]
vp9-skinmap. Some adjustments for model=1.

-use larger threshold on y (as in vp8).
-add distance threshold for each cluster
-use larger skin distance threshold for first cluster
-add some early exist checks.

Keep default setting to model=0.

Change-Id: I1044b99ade4bb1f215a860a019a4d84cee2f7715

8 years agoAdd 8-tap interpolation filter options for intra prediction
hui su [Wed, 13 Jan 2016 00:38:58 +0000 (16:38 -0800)]
Add 8-tap interpolation filter options for intra prediction

BD-rate performance improvement (on top of ext-intra):
derflr  0.22%
hevclr  0.36%
hevcmr  0.48%
hevchr  0.37%
stdhd   0.19%

Average speed impact on some derf clips is about 40% slower (on
top of ext-intra). Speed improvment is a to-do.

Change-Id: I8fe3fe8c5e4f60d0462778adbcc15c84dfbe7a25

8 years agoMerge "Use precise rate cost estimate for skip block mode"
Jingning Han [Wed, 3 Feb 2016 20:40:10 +0000 (20:40 +0000)]
Merge "Use precise rate cost estimate for skip block mode"

8 years agoGeneralize the dynamic reference motion vector coding mode
Jingning Han [Wed, 20 Jan 2016 00:36:25 +0000 (16:36 -0800)]
Generalize the dynamic reference motion vector coding mode

This commit generalizes dynamic reference motion vector coding mode
to support multiple candidate modes in the rate-distortion
optimization scheme and to support the selection in the bit-stream
syntax. The maximum number of modes allowed is currently limited to
4. The syntax elements for the dynamic reference motion vector
modes are using binary codes. The scheme supports single reference
frame.

It improves the compression performance
derf   0.135%
hevcmr 0.098%

Change-Id: Id053d6ce76e8365e52727bd0d12d28ce3de2e0e8

8 years agoUse precise rate cost estimate for skip block mode
Jingning Han [Wed, 3 Feb 2016 19:07:43 +0000 (11:07 -0800)]
Use precise rate cost estimate for skip block mode

It improves the compression performance of VP9 by 0.1% across all
test sets. No speed change is observed.

Change-Id: I59338c5c9e67bae22188f35fc3afbfe2a6bba6b0

8 years agoMerge "Account for zero-forcing operation in selective ref mv mode" into nextgenv2
Jingning Han [Wed, 3 Feb 2016 19:07:28 +0000 (19:07 +0000)]
Merge "Account for zero-forcing operation in selective ref mv mode" into nextgenv2

8 years agoMerge "Fix some interger overflow errors"
Hui Su [Wed, 3 Feb 2016 18:12:26 +0000 (18:12 +0000)]
Merge "Fix some interger overflow errors"

8 years agoAccount for zero-forcing operation in selective ref mv mode
Jingning Han [Fri, 15 Jan 2016 19:33:40 +0000 (11:33 -0800)]
Account for zero-forcing operation in selective ref mv mode

It makes the encoder accounts for the block zero-forcing operation
when optimizing the mode decisions.

Change-Id: I2c8e243756080b446b8a53a9679f75c4c47148cf

8 years agoMerge branch 'master' into nextgenv2
Yaowu Xu [Wed, 3 Feb 2016 16:38:48 +0000 (08:38 -0800)]
Merge branch 'master' into nextgenv2

Change-Id: Ib050607fa5c0288360ff224fd048445d16568520

8 years agoMerge "Supertx fix for 422 colour subsampling" into nextgenv2
Debargha Mukherjee [Wed, 3 Feb 2016 17:19:09 +0000 (17:19 +0000)]
Merge "Supertx fix for 422 colour subsampling" into nextgenv2

8 years agoFix for supertx: ignore skip_txfm optimization
Geza Lore [Wed, 3 Feb 2016 12:40:34 +0000 (12:40 +0000)]
Fix for supertx: ignore skip_txfm optimization

Change-Id: I80eedb548c449ec43c6b5b88c5493b665606906e

8 years agoSupertx fix for 422 colour subsampling
Julia Robson [Mon, 1 Feb 2016 15:44:50 +0000 (15:44 +0000)]
Supertx fix for 422 colour subsampling

Fixes assertion for football_422_4sif.y4m when supertx, var_tx and
ext_tx are all enabled. Problem was after subsampling, the u and v
blocks being encoded were no longer square.

Change-Id: Ie626f30a2e64538d33343a26d5124a79a6f2b985

8 years agoMerge "Add high bit depth args to arg list"
Hui Su [Wed, 3 Feb 2016 05:49:35 +0000 (05:49 +0000)]
Merge "Add high bit depth args to arg list"

8 years agodct16x16_test: quiet unused-function warning
James Zern [Wed, 3 Feb 2016 03:42:19 +0000 (19:42 -0800)]
dct16x16_test: quiet unused-function warning

in non-x86 builds

Change-Id: I1171e9c7e7ea5bcdea70bf5e3d28b07d6bb5578f

8 years agointrapred: protect functions w/CONFIG check x2
James Zern [Wed, 3 Feb 2016 03:38:37 +0000 (19:38 -0800)]
intrapred: protect functions w/CONFIG check x2

high-bitdepth version
d207e, d63e, d45e are only used with CONFIG_MISC_FIXES

Change-Id: I77292e11f51fd76d4127fd0027f876866bcf8675

8 years agodct32x32_test: remove unused function
James Zern [Wed, 3 Feb 2016 03:37:22 +0000 (19:37 -0800)]
dct32x32_test: remove unused function

idct32x32_8

Change-Id: I1678ab48639cc54be4c9294479de01beced9e1e9

8 years agotests: quiet some warnings w/--disable-webm-io
James Zern [Wed, 3 Feb 2016 03:32:25 +0000 (19:32 -0800)]
tests: quiet some warnings w/--disable-webm-io

unused-functions/variables

Change-Id: Id7a2c345442129d84ec1482e38fcdc4bc9eaad7d

8 years agoAdd high bit depth args to arg list
hui su [Wed, 3 Feb 2016 02:03:39 +0000 (18:03 -0800)]
Add high bit depth args to arg list

So that their usage info. will show up with --help.

Change-Id: I8542240dcc98e8be29ac63d081f5abb932627cbf

8 years agoyv12config.c: remove dead CONFIG_ALPHA code
James Zern [Wed, 3 Feb 2016 02:06:05 +0000 (18:06 -0800)]
yv12config.c: remove dead CONFIG_ALPHA code

the rest was removed with:
5926e7c Remove unfinished VP9 alpha channel.

Change-Id: I0c1d92765e5334e1380ad027b069b401e4ae9b8f

8 years agovp9_encoder.c: remove unused macro
James Zern [Wed, 3 Feb 2016 01:56:50 +0000 (17:56 -0800)]
vp9_encoder.c: remove unused macro

SHARP_FILTER_QTHRESH

unused since:
5373119 Merging in the Switchable interp experiment

Change-Id: I9ccc1883828babffd1886810b59aa16aca4766bf

8 years agovpx_mem.c: remove unused macro
James Zern [Wed, 3 Feb 2016 01:56:04 +0000 (17:56 -0800)]
vpx_mem.c: remove unused macro

__VPX_MEM_C__

unused since:
f58011a vpx_mem: remove vpx_memset

Change-Id: I820c972c4165275ddf21c1687748409818a7f2af

8 years agovp9_svc_layercontext.c: relocate a macro definition
James Zern [Wed, 3 Feb 2016 01:51:25 +0000 (17:51 -0800)]
vp9_svc_layercontext.c: relocate a macro definition

move SMALL_FRAME_FB_IDX closer to its first use. this is only used with
CONFIG_SPATIAL_SVC.

Change-Id: Ibf2b3e59a3aab2be2c3b3f89b4380ebe272e0ed6

8 years agovp9_encoder.c: protect SNPRINT* macros w/CONFIG check
James Zern [Wed, 3 Feb 2016 01:50:42 +0000 (17:50 -0800)]
vp9_encoder.c: protect SNPRINT* macros w/CONFIG check

these are only used with CONFIG_INTERNAL_STATS

Change-Id: Ibbcefbdc1db79d93f75e5865289f9960a8c2358a

8 years agovp9_aq_360.c: remove some unused macros
James Zern [Wed, 3 Feb 2016 01:46:44 +0000 (17:46 -0800)]
vp9_aq_360.c: remove some unused macros

ENERGY_MIN
ENERGY_MAX
ENERGY_SPAN
ENERGY_IN_BOUNDS

unused since first commit

Change-Id: I5507d36487aaee3e64f7a5fba582791ef9a533b0

8 years agovp9_diamond_search_sad_avx.c: rename (un)likely macros
James Zern [Wed, 3 Feb 2016 01:44:02 +0000 (17:44 -0800)]
vp9_diamond_search_sad_avx.c: rename (un)likely macros

avoid using '__' this is a reserved namespace for the compiler

Change-Id: I7d2be4dba2bdddc6f1010a16ad9e59a2e211b064

8 years agovp9_resize.c: add missing include
James Zern [Wed, 3 Feb 2016 01:43:33 +0000 (17:43 -0800)]
vp9_resize.c: add missing include

Change-Id: I429e3c80216102ed0a85ce348ecac0bf1e1e28eb

8 years agovp9_aq_complexity.c: remove unused macros
James Zern [Wed, 3 Feb 2016 01:42:07 +0000 (17:42 -0800)]
vp9_aq_complexity.c: remove unused macros

DEFAULT_COMPLEXITY
VAR_STRENGTH_STEP

unused since first commit

Change-Id: I4a47544098caf0d72e571633b7776146be88237c

8 years agovp9_detokenize.c: remove unused macros
James Zern [Wed, 3 Feb 2016 01:41:17 +0000 (17:41 -0800)]
vp9_detokenize.c: remove unused macros

LOW_VAL_CONTEXT_NODE
TWO_CONTEXT_NODE
THREE_CONTEXT_NODE
HIGH_LOW_CONTEXT_NODE
CAT_ONE_CONTEXT_NODE
CAT_THREEFOUR_CONTEXT_NODE
CAT_THREE_CONTEXT_NODE
CAT_FIVE_CONTEXT_NODE

unused since:
adaec4d Speeding up decode_coeffs().

Change-Id: Id07d2b0b76ff0ee6ef40d1b7a8e085e5e9e4aa8d

8 years agoFix some interger overflow errors
hui su [Tue, 2 Feb 2016 21:50:26 +0000 (13:50 -0800)]
Fix some interger overflow errors

Change-Id: I7e44bd952f28ce9925e8bdf6ee8ca2bb13de1b49

8 years agoMerge changes I4a365757,I14a3b328
James Zern [Wed, 3 Feb 2016 00:51:42 +0000 (00:51 +0000)]
Merge changes I4a365757,I14a3b328

* changes:
  vp10: remove unused (read|write)_uniform
  vp9_denoiser: mark total_adj_strong_thresh inline

8 years agoEnable adaptive motion vector referencing mode
Jingning Han [Fri, 15 Jan 2016 17:05:51 +0000 (09:05 -0800)]
Enable adaptive motion vector referencing mode

This commit allows an adaptive motion vector referencing mode
approach. It checks the available reference motion vector candidate
list and decides the amount of motion vector referencing modes. The
current implementation assumes simple binary coding for the syntax.

The compression performance is improved by
derf   0.11%
hevcmr 0.38%
stdhd  0.09%
hevchr 0.23%

The coding gains due to the new reference motion vector system are
derf   1.0%
hevcmr 1.7%
stdhd  1.4%
hevchr 1.3%

Change-Id: Idf932fc373546fe59c8741f1b933ff656e8dbc3f

8 years agoMerge "Fix a signed overflow in vp9 motion cost."
Alex Converse [Tue, 2 Feb 2016 21:12:49 +0000 (21:12 +0000)]
Merge "Fix a signed overflow in vp9 motion cost."

8 years agoMerge "Fixing a issue of calculating tx cost for SUPERTX+VAR_TX" into nextgenv2
Debargha Mukherjee [Tue, 2 Feb 2016 15:14:39 +0000 (15:14 +0000)]
Merge "Fixing a issue of calculating tx cost for SUPERTX+VAR_TX" into nextgenv2

8 years agoMerge "Loop filter search resets on overlay frame."
Paul Wilkins [Tue, 2 Feb 2016 14:44:47 +0000 (14:44 +0000)]
Merge "Loop filter search resets on overlay frame."

8 years agoMerge branch 'master' into nextgenv2
Yaowu Xu [Tue, 2 Feb 2016 13:00:05 +0000 (05:00 -0800)]
Merge branch 'master' into nextgenv2

8 years agovp10: remove unused (read|write)_uniform
James Zern [Tue, 2 Feb 2016 07:47:35 +0000 (23:47 -0800)]
vp10: remove unused (read|write)_uniform

dead code since:
5d3327e Remove palette from VP10

Change-Id: I4a36575706ea6fffefe5bc778595112ef6ff37d8

8 years agovp9_denoiser: mark total_adj_strong_thresh inline
James Zern [Tue, 2 Feb 2016 07:47:02 +0000 (23:47 -0800)]
vp9_denoiser: mark total_adj_strong_thresh inline

avoids -Wunused-function warnings when INLINE is set

Change-Id: I14a3b32837d358516b7702a2fb804bec010bb5c6

8 years agoMerge "Pass filter type instead of filter array" into nextgenv2
Angie Chiang [Tue, 2 Feb 2016 02:32:16 +0000 (02:32 +0000)]
Merge "Pass filter type instead of filter array" into nextgenv2

8 years agoPass filter type instead of filter array
Angie Chiang [Mon, 1 Feb 2016 20:49:15 +0000 (12:49 -0800)]
Pass filter type instead of filter array

Change-Id: I25f2149ddaa332722f7ab82e8f832a253c4b6ab3

8 years agoMerge "Refactor to separate restoration from loop filter" into nextgenv2
Debargha Mukherjee [Tue, 2 Feb 2016 00:17:14 +0000 (00:17 +0000)]
Merge "Refactor to separate restoration from loop filter" into nextgenv2

8 years agoFix a signed overflow in vp9 motion cost.
Alex Converse [Mon, 1 Feb 2016 17:47:39 +0000 (09:47 -0800)]
Fix a signed overflow in vp9 motion cost.

Change-Id: I5975e3aede62202d8ee6ced33889350c0a56554a

8 years agoFix automerge errors
Yaowu Xu [Mon, 1 Feb 2016 20:29:36 +0000 (12:29 -0800)]
Fix automerge errors

Change-Id: I24d415bafe617eac894427088d7b2fbe0b7e04d7

8 years agovp9: Add flag to control usage of skin detection.
Marco [Mon, 1 Feb 2016 19:52:49 +0000 (11:52 -0800)]
vp9: Add flag to control usage of skin detection.

Set off as default; on for 1 pass cbr mode, speed >=5, non-screen-content.

Change-Id: I03f2497e4028b354fd83b8a7d0e072c2a6bec878

8 years agoMerge "vp9 non-rd partition: Update condition on force_split for skin area."
Marco Paniconi [Mon, 1 Feb 2016 18:48:46 +0000 (18:48 +0000)]
Merge "vp9 non-rd partition: Update condition on force_split for skin area."

8 years agovp9 non-rd partition: Update condition on force_split for skin area.
Marco [Mon, 1 Feb 2016 17:45:35 +0000 (09:45 -0800)]
vp9 non-rd partition: Update condition on force_split for skin area.

Avoid low resolutions for now.

Change-Id: Ib3a7bb8716a033449b2afb216edd70897a280df9

8 years agovp9-denoiser bugfix: Disable postproc-denoiser under temporal denoising.
Marco [Mon, 1 Feb 2016 17:02:26 +0000 (09:02 -0800)]
vp9-denoiser bugfix: Disable postproc-denoiser under temporal denoising.

The postproc vp9_denoise() is a spatial denoise/blur function.
It was not intended to be used if temporal denoising is enabled.

Change-Id: I97d2dcb941e7cc49bbafce99d9286beb2693249d

8 years agoMerge "vp9-noise estimate: Put check to avoid possible out of bounds."
Marco Paniconi [Mon, 1 Feb 2016 15:18:16 +0000 (15:18 +0000)]
Merge "vp9-noise estimate: Put check to avoid possible out of bounds."

8 years agoMerge branch 'master' into nextgenv2
Yaowu Xu [Mon, 1 Feb 2016 13:00:05 +0000 (05:00 -0800)]
Merge branch 'master' into nextgenv2

8 years agoMerge "VP9: Refactor read_inter_block_mode_info()"
Scott LaVarnway [Mon, 1 Feb 2016 12:52:15 +0000 (12:52 +0000)]
Merge "VP9: Refactor read_inter_block_mode_info()"

8 years agovp9-noise estimate: Put check to avoid possible out of bounds.
Marco [Mon, 1 Feb 2016 06:24:55 +0000 (22:24 -0800)]
vp9-noise estimate: Put check to avoid possible out of bounds.

Put check to avoid possible out of bounds when looping
over the blocks to estimate noise level.
No change in behavior.

Change-Id: I4b7b19b7edee0ae1c35b9dc0700b1bf9b304d7f5

8 years agoMerge branch 'master' into nextgenv2
Yaowu Xu [Sun, 31 Jan 2016 13:00:05 +0000 (05:00 -0800)]
Merge branch 'master' into nextgenv2