]> granicus.if.org Git - libvpx/log
libvpx
8 years agoaom_mem: Refactor code
Urvang Joshi [Tue, 26 Jul 2016 00:12:14 +0000 (17:12 -0700)]
aom_mem: Refactor code

Change-Id: I2da9cd5da48ae97e770bccfd1233bcc70b484688

8 years agoMerge "Move CHECK_MEM_ERROR implementation to aom/internal." into nextgenv2
Yaowu Xu [Tue, 6 Sep 2016 21:26:51 +0000 (21:26 +0000)]
Merge "Move CHECK_MEM_ERROR implementation to aom/internal." into nextgenv2

8 years agoMove CHECK_MEM_ERROR implementation to aom/internal.
Alex Converse [Wed, 8 Jun 2016 23:14:57 +0000 (16:14 -0700)]
Move CHECK_MEM_ERROR implementation to aom/internal.

Allow using it in aom_dsp.

Change-Id: Ide7d58b6d11f8a45d473fc13bf730ba5bccb5516

8 years agoMake RECT_TX(>=8x8) work with VAR_TX
Yue Chen [Tue, 30 Aug 2016 00:29:33 +0000 (17:29 -0700)]
Make RECT_TX(>=8x8) work with VAR_TX

Bitstream syntax:
For a rectangular inter block, 'rect_tx' flag is sent to indicate if
the biggest rect tx is used. If no, continue to decode regular
recursive tx partition.

Change-Id: I127e35cc619b65acb5e9a0717f399cdcdb73fbf0

8 years agoMerge "Add global motion experiment to rdopt" into nextgenv2
Sarah Parker [Tue, 6 Sep 2016 18:07:31 +0000 (18:07 +0000)]
Merge "Add global motion experiment to rdopt" into nextgenv2

8 years agoMerge "Enable rectangular transforms for UV" into nextgenv2
Debargha Mukherjee [Tue, 6 Sep 2016 15:46:21 +0000 (15:46 +0000)]
Merge "Enable rectangular transforms for UV" into nextgenv2

8 years agoFix a compiler warning of unused variable
Yaowu Xu [Fri, 2 Sep 2016 22:38:59 +0000 (15:38 -0700)]
Fix a compiler warning of unused variable

Change-Id: I4a2faa32cc0847fe14dd8f40156163f4713055ca

8 years agoAvoid re-use same temp variables
Yaowu Xu [Fri, 2 Sep 2016 22:30:21 +0000 (15:30 -0700)]
Avoid re-use same temp variables

In highbd_quantize_intrin_sse2.c.

Change-Id: Iaf6360e456f1fb2f8ff06461afbfecfc0103dda3

8 years agoquantize.c: int->uint32_t for absolute values
Yaowu Xu [Fri, 2 Sep 2016 22:28:30 +0000 (15:28 -0700)]
quantize.c: int->uint32_t for absolute values

Change-Id: I784f32e0e86d873655e46cf68c5c124a698af361

8 years agoaom_dsp: AV1_IADST8x16_1D to AOM_IADST8x16_1D
Yaowu Xu [Fri, 2 Sep 2016 22:27:10 +0000 (15:27 -0700)]
aom_dsp: AV1_IADST8x16_1D to AOM_IADST8x16_1D

Change-Id: Iba415ab2d4adb3350b4747a58f69db7d02bbab68

8 years agoEnable rectangular transforms for UV
Debargha Mukherjee [Wed, 31 Aug 2016 00:43:38 +0000 (17:43 -0700)]
Enable rectangular transforms for UV

Uses an array to map block sizes, y tx sizes, and subsampling
factors to various transform sizes for UV.

Results improve by 0.1-0.2%

Change-Id: Icb58fd96bc7c01a72cbf1332fe2be4d55a0feedc

8 years agoRemoved tests and data not in use
Yaowu Xu [Fri, 2 Sep 2016 22:26:06 +0000 (15:26 -0700)]
Removed tests and data not in use

Change-Id: If688da3089ad33f18751fa2f8c46b6f5dc708bd2

8 years agoMerge "test_intra_pred_speed.cc : Fix visual studio build." into nextgenv2
Urvang Joshi [Fri, 2 Sep 2016 23:12:33 +0000 (23:12 +0000)]
Merge "test_intra_pred_speed.cc : Fix visual studio build." into nextgenv2

8 years agoMerge "Change to use AOM copyright notice" into nextgenv2
Yaowu Xu [Fri, 2 Sep 2016 22:13:23 +0000 (22:13 +0000)]
Merge "Change to use AOM copyright notice" into nextgenv2

8 years agotest_intra_pred_speed.cc : Fix visual studio build.
Urvang Joshi [Fri, 2 Sep 2016 22:05:28 +0000 (15:05 -0700)]
test_intra_pred_speed.cc : Fix visual studio build.

Visual studio doesn't like nested macros, apparently. This patch should
fix it.

Change-Id: Ifa56fae5be0b3dfd3fecd88a8a443e39135f96ab

8 years agoChange to use AOM copyright notice
Yaowu Xu [Fri, 2 Sep 2016 19:04:54 +0000 (12:04 -0700)]
Change to use AOM copyright notice

Change-Id: I2b2b70e756b7eb9611b7b33b7d5f19b3b30e0a50

8 years agoMerge "Remove unused buffer allocation functions" into nextgenv2
Yunqing Wang [Fri, 2 Sep 2016 17:52:48 +0000 (17:52 +0000)]
Merge "Remove unused buffer allocation functions" into nextgenv2

8 years agoMerge "aomcx_set_ref -> aom_cx_set_ref" into nextgenv2
Yaowu Xu [Fri, 2 Sep 2016 17:41:44 +0000 (17:41 +0000)]
Merge "aomcx_set_ref -> aom_cx_set_ref" into nextgenv2

8 years agoMerge "Change to use aom copyright notice" into nextgenv2
Yaowu Xu [Fri, 2 Sep 2016 17:41:21 +0000 (17:41 +0000)]
Merge "Change to use aom copyright notice" into nextgenv2

8 years agoMerge "Add frame info in bitstream debug tool" into nextgenv2
Angie Chiang [Fri, 2 Sep 2016 17:05:48 +0000 (17:05 +0000)]
Merge "Add frame info in bitstream debug tool" into nextgenv2

8 years agoRemove unused buffer allocation functions
Yunqing Wang [Fri, 2 Sep 2016 16:23:53 +0000 (09:23 -0700)]
Remove unused buffer allocation functions

Removed unused buffer allocation functions.

Change-Id: Ib779dde9ad6a511d88b7f7cba2604902eff7aa05

8 years agoaomcx_set_ref -> aom_cx_set_ref
Yaowu Xu [Thu, 1 Sep 2016 22:32:28 +0000 (15:32 -0700)]
aomcx_set_ref -> aom_cx_set_ref

Change-Id: I60dd645451d6d65465f099a16ac855fb0b5a57a9

8 years agoChange to use aom copyright notice
Yaowu Xu [Thu, 1 Sep 2016 21:32:49 +0000 (14:32 -0700)]
Change to use aom copyright notice

This minimize code differences between AOM master and nextgenv2

Change-Id: If144865bdf3ef0818e7aac11018b9e786444c550

8 years agoMerge "Some cleanups for unnecessary macros" into nextgenv2
Debargha Mukherjee [Fri, 2 Sep 2016 08:37:36 +0000 (08:37 +0000)]
Merge "Some cleanups for unnecessary macros" into nextgenv2

8 years agoAdd global motion experiment to rdopt
Sarah Parker [Tue, 16 Aug 2016 21:57:37 +0000 (14:57 -0700)]
Add global motion experiment to rdopt

This patch completes the global motion experiment
implementation. It modifies the format of the motion
parameters to use the mv union to facilitate faster
copying and checks for parameters equal to 0 that occur
frequently in rdopt. The rd decisions for the global motion experiment
have also been added to rdopt.
Change-Id: Idfb9f0c6d23e538221763881099c5a2a3891f5a9

8 years agoPort two daala_dering changes from AOMedia
Yaowu Xu [Thu, 1 Sep 2016 18:35:16 +0000 (11:35 -0700)]
Port two daala_dering changes from AOMedia

03394bd Remove dead code from av1_dering_search.
337b23a Changing the weights of the first CRF filter in deringing

Change-Id: I1216c146dc3f72f24ceec3d3c65c4dd6cd73623e

8 years agorename two mk files to make naming consistent
Yaowu Xu [Thu, 1 Sep 2016 18:05:32 +0000 (11:05 -0700)]
rename two mk files to make naming consistent

av1cx.mk -> av1_cx.mk
av1dx.mk -> av1_dx.mk

Change-Id: I698bd65b933c433066d5dfeb94cee680095508e4

8 years agoMerge "Add explict conversion from int64_t to int" into nextgenv2
Yaowu Xu [Fri, 2 Sep 2016 00:39:19 +0000 (00:39 +0000)]
Merge "Add explict conversion from int64_t to int" into nextgenv2

8 years agoAdd frame info in bitstream debug tool
Angie Chiang [Thu, 1 Sep 2016 23:10:50 +0000 (16:10 -0700)]
Add frame info in bitstream debug tool

Change-Id: Iead3edd8563d7900481eb199e8b003d2d3df075b

8 years agoAdd explict conversion from int64_t to int
Yaowu Xu [Thu, 1 Sep 2016 16:21:23 +0000 (09:21 -0700)]
Add explict conversion from int64_t to int

The values after right shifts should fit into 32bit int. The commit
fixes MSVC build warning when new-quant is enabled.

Change-Id: Ic89dd86fb981a1206653943658af2b6b2925a676

8 years agoMerge ".gitignore: corrent entries from vpx to aom" into nextgenv2
Yaowu Xu [Thu, 1 Sep 2016 22:33:18 +0000 (22:33 +0000)]
Merge ".gitignore: corrent entries from vpx to aom" into nextgenv2

8 years agoMerge "Add ALT_INTRA experiment." into nextgenv2
Urvang Joshi [Thu, 1 Sep 2016 21:45:31 +0000 (21:45 +0000)]
Merge "Add ALT_INTRA experiment." into nextgenv2

8 years ago.gitignore: corrent entries from vpx to aom
Yaowu Xu [Thu, 1 Sep 2016 16:10:59 +0000 (09:10 -0700)]
.gitignore: corrent entries from vpx to aom

Change-Id: I8af6a9723c31c0f868e9bd75dcc079413a3700c4

8 years agoAdd ALT_INTRA experiment.
Urvang Joshi [Thu, 1 Sep 2016 19:03:20 +0000 (12:03 -0700)]
Add ALT_INTRA experiment.

When the experiment is ON, we use Paeth predictor instead of TM
predictor.

For derf set, this gives about 0.09% improvement overall, and 0.55%
improvement if all frames are forced to be intra-only.

Also, if the EXT_INTRA experiment is also on, the improvement overall
is 0.056%, and improvement if all frames are forced to be intra-only is
0.465%.

Change-Id: Id74e107ede70a8d2107fa14fcb3f44b23a437274

8 years agoadd an explicit conversion from size_t to int
Yaowu Xu [Thu, 1 Sep 2016 15:59:46 +0000 (08:59 -0700)]
add an explicit conversion from size_t to int

Function ans_read_int() takes int as parameter, this commit uses an
explicit conversion to avoid MSVC building warning.

Change-Id: Ia405e1d5a86c0f42932fa1da29417ccbf2dd58e7

8 years agoReplace inline with INLINE
Yaowu Xu [Thu, 1 Sep 2016 15:45:22 +0000 (08:45 -0700)]
Replace inline with INLINE

This fixes msvc build errors.

Change-Id: I1344685e891db61ba569d818e0f2167b2978c299

8 years agoSome cleanups for unnecessary macros
Debargha Mukherjee [Tue, 30 Aug 2016 22:55:23 +0000 (15:55 -0700)]
Some cleanups for unnecessary macros

Remove some macros that are no longer necessary for experimentation.

Change-Id: I959bf441c8333607df4aa1ee18841f189ade8112

8 years agoPort renaming changes from AOMedia
Yaowu Xu [Tue, 30 Aug 2016 21:01:10 +0000 (14:01 -0700)]
Port renaming changes from AOMedia

Cherry-Picked the following commits:
0defd8f Changed "WebM" to "AOMedia" & "webm" to "aomedia"
54e6676 Replace "VPx" by "AVx"
5082a36 Change "Vpx" to "Avx"
7df44f1 Replace "Vp9" w/ "Av1"
967f722 Remove kVp9CodecId
828f30c Change "Vp8" to "AOM"
030b5ff AUTHORS regenerated
2524cae Add ref-mv experimental flag
016762b Change copyright notice to AOMedia form
81e5526 Replace vp9 w/ av1
9b94565 Add missing files
fa8ca9f Change "vp9" to "av1"
ec838b7  Convert "vp8" to "aom"
80edfa0 Change "VP9" to "AV1"
d1a11fb Change "vp8" to "aom"
7b58251 Point to WebM test data
dd1a5c8 Replace "VP8" with "AOM"
ff00fc0 Change "VPX" to "AOM"
01dee0b Change "vp10" to "av1" in source code
cebe6f0 Convert "vpx" to "aom"
17b0567 rename vp10*.mk to av1_*.mk
fe5f8a8 rename files vp10_* to av1_*

Change-Id: I6fc3d18eb11fc171e46140c836ad5339cf6c9419

8 years agoPort folder renaming changes from AOM
Yaowu Xu [Mon, 22 Aug 2016 23:08:15 +0000 (16:08 -0700)]
Port folder renaming changes from AOM

Manually cherry-picked commits:
ceef058 libvpx->libaom part2
3d26d91 libvpx -> libaom
cfea7dd vp10/ -> av1/
3a8eff7 Fix a build issue for a test
bf4202e Rename vpx to aom

Change-Id: I1b0eb5a40796e3aaf41c58984b4229a439a597dc

8 years agoMerge "Change buffer_alloc_sz and frame_size type to size_t" into nextgenv2
Yunqing Wang [Wed, 31 Aug 2016 23:56:02 +0000 (23:56 +0000)]
Merge "Change buffer_alloc_sz and frame_size type to size_t" into nextgenv2

8 years agoChange buffer_alloc_sz and frame_size type to size_t
Yunqing Wang [Tue, 30 Aug 2016 18:48:44 +0000 (11:48 -0700)]
Change buffer_alloc_sz and frame_size type to size_t

1. Changed buffer_alloc_sz and frame_size type to size_t.
2. Added a TODO for video resolution limits. On 32 bit systems, the maximum
resolution supported in the encoder is 4k(3840x2160). The malloc() would
fail if encoding >4k video on a 32 bit system.

Change-Id: Ibd91b28fd63d1b04e8ac9a5270a17629f239188a

8 years agoMerge "Fix Visual Studio build warnings" into nextgenv2
Yunqing Wang [Wed, 31 Aug 2016 18:50:06 +0000 (18:50 +0000)]
Merge "Fix Visual Studio build warnings" into nextgenv2

8 years agoMerge "Remove unused buffer allocation functions" into nextgenv2
Yunqing Wang [Wed, 31 Aug 2016 18:49:29 +0000 (18:49 +0000)]
Merge "Remove unused buffer allocation functions" into nextgenv2

8 years agoFix a bug in calculating the compound ref frame cost
Zoe Liu [Wed, 24 Aug 2016 23:14:00 +0000 (16:14 -0700)]
Fix a bug in calculating the compound ref frame cost

The previous ext-refs experiment did not consider the cost of the 2nd
reference frame on mode decision in the compound mode. With the fix,
using Overall PSNR, compared to the previous ext-refs RD performance
before the bug fix, all against the baseline, the improvements are:

"ext-refs" before fix: lowres -5.665%  midres: -4.833%
"ext-refs" after fix:  lowres -5.776%  midres: -5.000%
Improvement by the fix: lowres -0.111%  midres: -0.167%

Change-Id: I2eceedf2d4046b169514e049fd01baaf0bbb50c6

8 years agoMerge "Fix a bug in deciding ref frame context in ext-refs" into nextgenv2
Zoe Liu [Wed, 31 Aug 2016 16:42:42 +0000 (16:42 +0000)]
Merge "Fix a bug in deciding ref frame context in ext-refs" into nextgenv2

8 years agoMake an unmeaningful change to be added into the list
Wei-Ting [Tue, 30 Aug 2016 21:58:24 +0000 (14:58 -0700)]
Make an unmeaningful change to be added into the list

Change-Id: I578589a955bd3f3d7ff61723b574361661453f67

8 years agoFix a bug in deciding ref frame context in ext-refs
Zoe Liu [Tue, 23 Aug 2016 00:24:57 +0000 (17:24 -0700)]
Fix a bug in deciding ref frame context in ext-refs

Change-Id: Ie58b98baa870c5d2a5b7193f8fe4f84fd7ec6c16

8 years agoFix Visual Studio build warnings
Yunqing Wang [Tue, 30 Aug 2016 16:13:16 +0000 (09:13 -0700)]
Fix Visual Studio build warnings

Fixed a list of VS warnings. Warning message:
..\test\vp10_convolve_test.cc(34): warning C4244: 'initializing' : conversion
from 'ptrdiff_t' to 'int', possible loss of data

Change-Id: I9a1d3978a79fbb7b1ac028c5713ac72b6ff99172

8 years agoMerge "clpf experiment build fix" into nextgenv2
Debargha Mukherjee [Tue, 30 Aug 2016 05:45:32 +0000 (05:45 +0000)]
Merge "clpf experiment build fix" into nextgenv2

8 years agoImplement global motion parameter computation
Sarah Parker [Wed, 10 Aug 2016 00:40:53 +0000 (17:40 -0700)]
Implement global motion parameter computation

This computes global motion parameters between 2 frames by
matching corresponding points using FAST feature and then
fitting a model using RANSAC.

Change-Id: Ib6664df44090e8cfa4db9f2f9e0556931ccfe5c8

8 years agoRemove unused buffer allocation functions
Yunqing Wang [Mon, 29 Aug 2016 22:02:09 +0000 (15:02 -0700)]
Remove unused buffer allocation functions

Removed unused buffer allocation functions.

Change-Id: I5aa265a7698a2d2df736ddb89c6d93a5ee34895b

8 years agoMerge "Tiling in loop restoration + cosmetics" into nextgenv2
Debargha Mukherjee [Mon, 29 Aug 2016 20:46:13 +0000 (20:46 +0000)]
Merge "Tiling in loop restoration + cosmetics" into nextgenv2

8 years agoclpf experiment build fix
Debargha Mukherjee [Mon, 29 Aug 2016 19:22:58 +0000 (12:22 -0700)]
clpf experiment build fix

Change-Id: I729e14916ecb58b5a75756078ab96a2d340bc0d6

8 years agoTiling in loop restoration + cosmetics
Aamir Anis [Thu, 21 Jul 2016 04:29:45 +0000 (21:29 -0700)]
Tiling in loop restoration + cosmetics

Frame can be split into rectangular tiles for application of separate
bilateral or Wiener filters per tile. Some variable names changed for
better readability.

Change-Id: I13ebc4d0b0baf368e524db5ce276f03ed76af9c8

8 years agoFix for supertx with rect-tx
Debargha Mukherjee [Mon, 29 Aug 2016 18:04:17 +0000 (11:04 -0700)]
Fix for supertx with rect-tx

Change-Id: I0cc3523a8992f889f8dd203449ceb55f2a422324

8 years agofix uint32_t <-> size_t mismatch in tests
Pascal Massimino [Mon, 29 Aug 2016 07:09:09 +0000 (09:09 +0200)]
fix uint32_t <-> size_t mismatch in tests

Change-Id: Ifde4d57957219560e01ebd1657f1c0721f041054

8 years agoMerge "vp10_alloc_context_buffers: clear cm->mi* on failure" into nextgenv2
James Zern [Sat, 27 Aug 2016 16:43:52 +0000 (16:43 +0000)]
Merge "vp10_alloc_context_buffers: clear cm->mi* on failure" into nextgenv2

8 years agoMerge "Fix VS build warnings in blend_a64_mask_test.cc" into nextgenv2
Jingning Han [Fri, 26 Aug 2016 21:47:39 +0000 (21:47 +0000)]
Merge "Fix VS build warnings in blend_a64_mask_test.cc" into nextgenv2

8 years agoMerge "Remove unnecessary buffer pointers in PICK_MODE_CONTEXT" into nextgenv2
Hui Su [Fri, 26 Aug 2016 21:38:14 +0000 (21:38 +0000)]
Merge "Remove unnecessary buffer pointers in PICK_MODE_CONTEXT" into nextgenv2

8 years agoFix VS build warnings in blend_a64_mask_test.cc
Jingning Han [Fri, 26 Aug 2016 18:24:36 +0000 (11:24 -0700)]
Fix VS build warnings in blend_a64_mask_test.cc

Change-Id: Id4c764198549a60d98e5c4a74083972b97da5b81

8 years agoFix compile error in dering
Debargha Mukherjee [Fri, 26 Aug 2016 17:44:09 +0000 (10:44 -0700)]
Fix compile error in dering

Change-Id: I56890c813de1b366e4ef482d9fc6da81034636ab

8 years agoUse rectangular transforms for >= 8x8 blocks
Geza Lore [Tue, 26 Jul 2016 16:12:43 +0000 (17:12 +0100)]
Use rectangular transforms for >= 8x8 blocks

For rectangular blocks between 8x8 and 32x32, we can now code the
transform size as one bigger than the largest square that fits in
the block (eg, for 16x8, we can code a transform size of 16x16
rather than the previous maximum of 8x8), when this oversized
transform is coded in the bitstream, the codec will use the full
size rectangular transform for that block (eg 16x8 transform in
the above example).

Also fixes a scaling bug in 16x8/8x16 transforms.

Change-Id: I62ce75f1b01c46fe2fbc727ce4abef695f4fcd43

8 years agoMerge "Update VP9_PROB_COST_SHIFT to VP10_PROB_COST_SHIFT" into nextgenv2
Sarah Parker [Thu, 25 Aug 2016 18:31:22 +0000 (18:31 +0000)]
Merge "Update VP9_PROB_COST_SHIFT to VP10_PROB_COST_SHIFT" into nextgenv2

8 years agoSeparate EXT_ARFs' frame context index
Wei-ting Lin [Mon, 22 Aug 2016 22:52:32 +0000 (15:52 -0700)]
Separate EXT_ARFs' frame context index

This commit separate the frame index of EXT_ARFs' from other frame
types in the ext-refs setting.

It improves the average RD performance by

0.206% in the lowres, and
0.173% in the midres.

The overall gains for the ext-refs compared to the baseline are

5.665% in the lowres, and
4.883% in the midres.

Change-Id: I6591ad29120880c1aef0bd0b7cf15238c3f3b8f3

8 years agoMerge "Fix motion vector out of range bugs" into nextgenv2
Yunqing Wang [Thu, 25 Aug 2016 15:29:20 +0000 (15:29 +0000)]
Merge "Fix motion vector out of range bugs" into nextgenv2

8 years agoUpdate VP9_PROB_COST_SHIFT to VP10_PROB_COST_SHIFT
Sarah Parker [Thu, 25 Aug 2016 00:15:50 +0000 (17:15 -0700)]
Update VP9_PROB_COST_SHIFT to VP10_PROB_COST_SHIFT

Change-Id: Ie1416569e73e66518cdb2765d79a2fb3dd570489

8 years agoFix motion vector out of range bugs
Yunqing Wang [Thu, 25 Aug 2016 00:05:17 +0000 (17:05 -0700)]
Fix motion vector out of range bugs

2 bugs were fixed in VP9.
https://chromium-review.googlesource.com/#/c/366873/
https://chromium-review.googlesource.com/#/c/368440/
Fixed them in VP10 as well.

Change-Id: I2e53fabc6131ff80ba6dcfd4c73eb76c59b4c474

8 years agoMerge "gitignore: add some entries" into nextgenv2
Urvang Joshi [Wed, 24 Aug 2016 22:12:02 +0000 (22:12 +0000)]
Merge "gitignore: add some entries" into nextgenv2

8 years agoMerge "test_intra_pred_speed fix: use dspr2 version when HAVE_DSPR2" into nextgenv2
Urvang Joshi [Wed, 24 Aug 2016 22:10:01 +0000 (22:10 +0000)]
Merge "test_intra_pred_speed fix: use dspr2 version when HAVE_DSPR2" into nextgenv2

8 years agoRemove unnecessary buffer pointers in PICK_MODE_CONTEXT
hui su [Wed, 24 Aug 2016 19:16:51 +0000 (12:16 -0700)]
Remove unnecessary buffer pointers in PICK_MODE_CONTEXT

Change-Id: I600af6a66dc0e1310c8bfc7c16efa8a82e90856d

8 years agoMerge "Make rectangular txfm in EXT_TX work with VAR_TX" into nextgenv2
Yue Chen [Tue, 23 Aug 2016 23:54:40 +0000 (23:54 +0000)]
Merge "Make rectangular txfm in EXT_TX work with VAR_TX" into nextgenv2

8 years agoPalette code: remove the use of same if condition twice.
Urvang Joshi [Tue, 23 Aug 2016 22:01:41 +0000 (15:01 -0700)]
Palette code: remove the use of same if condition twice.

rd_pick_palette_intra_sby() method is called only when,
cpi->common.allow_screen_content_tools is on. So, no need to check that
again. We just use an assert() instead to still be safe.

Change-Id: I19785c2aac016798c8d331bbe91971b3806b73a8

8 years agoMerge "Rename CONFIG_VPX_HIGHBITDEPTH -> CONFIG_VP9_HIGHBITDEPTH" into nextgenv2
Urvang Joshi [Tue, 23 Aug 2016 19:33:52 +0000 (19:33 +0000)]
Merge "Rename CONFIG_VPX_HIGHBITDEPTH -> CONFIG_VP9_HIGHBITDEPTH" into nextgenv2

8 years agoMake rectangular txfm in EXT_TX work with VAR_TX
Yue Chen [Thu, 11 Aug 2016 23:42:08 +0000 (16:42 -0700)]
Make rectangular txfm in EXT_TX work with VAR_TX

Adapt rectangular txfm experiment to syntax/tokenization/loopfilter
framework of VAR_TX

Change-Id: Idcb005ecf5b3712de3e1cccb0d811ca16d87af24

8 years agoRename CONFIG_VPX_HIGHBITDEPTH -> CONFIG_VP9_HIGHBITDEPTH
Urvang Joshi [Tue, 23 Aug 2016 19:04:18 +0000 (12:04 -0700)]
Rename CONFIG_VPX_HIGHBITDEPTH -> CONFIG_VP9_HIGHBITDEPTH

"vpx-highbitdepth" config doesn't exist.

Change-Id: Ib6d3691454299bb381ecc75b80657fbebf9f59b2

8 years agotest_intra_pred_speed fix: use dspr2 version when HAVE_DSPR2
Urvang Joshi [Tue, 23 Aug 2016 18:29:44 +0000 (11:29 -0700)]
test_intra_pred_speed fix: use dspr2 version when HAVE_DSPR2

Change-Id: Ie7c78e19e077516615c71669022f505f8b3c80ca

8 years agogitignore: add some entries
Urvang Joshi [Tue, 23 Aug 2016 18:19:17 +0000 (11:19 -0700)]
gitignore: add some entries

Change-Id: I65507c3d132b2b3ba90cf0a7b1c729da7d3de15f

8 years agoMerge "Allow LF_UPDATE type of frames to use BWDREF" into nextgenv2
Wei-ting Lin [Tue, 23 Aug 2016 18:06:56 +0000 (18:06 +0000)]
Merge "Allow LF_UPDATE type of frames to use BWDREF" into nextgenv2

8 years agoMissing fixes for rect-tx
Debargha Mukherjee [Tue, 23 Aug 2016 07:17:00 +0000 (00:17 -0700)]
Missing fixes for rect-tx

Reintroducing some fixes that were dropped inadvertently in
course of rebasing.

Change-Id: I5f51160c586010590d4bfd5cf225fb21347b0a40

8 years agoMerge "Make type conversion explicit" into nextgenv2
Yaowu Xu [Tue, 23 Aug 2016 01:28:04 +0000 (01:28 +0000)]
Merge "Make type conversion explicit" into nextgenv2

8 years agoMerge "Various rect-tx fixes" into nextgenv2
Debargha Mukherjee [Tue, 23 Aug 2016 01:00:39 +0000 (01:00 +0000)]
Merge "Various rect-tx fixes" into nextgenv2

8 years agoMake type conversion explicit
Yaowu Xu [Mon, 22 Aug 2016 16:55:31 +0000 (09:55 -0700)]
Make type conversion explicit

This fixes two MSVC compiler warnings.

Change-Id: I55ad8833676e20c2c4a55885b99a7a9293d9623f

8 years agoMerge "Apply clang-format" into nextgenv2
Yaowu Xu [Tue, 23 Aug 2016 00:00:48 +0000 (00:00 +0000)]
Merge "Apply clang-format" into nextgenv2

8 years agoAllow LF_UPDATE type of frames to use BWDREF
Wei-ting Lin [Mon, 22 Aug 2016 18:19:30 +0000 (11:19 -0700)]
Allow LF_UPDATE type of frames to use BWDREF

Originally, only bi-pred type of frames can use BWDREF. When
extra alt-refs are inserted in a gf group, the closest alt-ref
serves as ALTREF for the frames within the corresponding
subgroup. Therefore, the original alt-ref can be used as BWDREF
for the LF_UPDATE type of frames.

This patch further swaps the virtual indices of BWDREF and ALTREF
for those frames whose BWDREF is farther than ALTREF. As a result,
the BWDREF is always the closet backward reference frame, and the
ALTREF is the farther one.

It improves the average RD performance by

0.132% in lowres, and
0.030% in midres.

The overall gains for the ext-refs compared to the baseline are

5.486% in lowres, and
4.666% in midres.

Change-Id: I22e4e5f378f19c4c89196a0a5e9214adb46c3428

8 years agoApply clang-format
Yaowu Xu [Mon, 22 Aug 2016 17:05:19 +0000 (10:05 -0700)]
Apply clang-format

Change-Id: Ie283af5f30324f54b4f749becdb48f937584707d

8 years agoVarious rect-tx fixes
Debargha Mukherjee [Thu, 11 Aug 2016 19:10:55 +0000 (12:10 -0700)]
Various rect-tx fixes

Change-Id: I02f44713b99284092ecfc50ce7ab268e91d2c6f8

8 years agoAdd integerize function back in warped_motion.c
Sarah Parker [Thu, 11 Aug 2016 22:57:09 +0000 (15:57 -0700)]
Add integerize function back in warped_motion.c

This function was previously unused and removed in
I6bc740e778658d6f81ca54888fc6fa822d3b5ee0. I am adding it back in
with previously suggested fixes.

Change-Id: Iee0afb39170d25895b11d07e71843eae6913efd1

8 years agoMerge "Palette: count Y colors only for screen content." into nextgenv2
Urvang Joshi [Fri, 19 Aug 2016 23:18:39 +0000 (23:18 +0000)]
Merge "Palette: count Y colors only for screen content." into nextgenv2

8 years agoMerge "Handle centroid rounding inside palette.c itself." into nextgenv2
Urvang Joshi [Fri, 19 Aug 2016 22:22:52 +0000 (22:22 +0000)]
Merge "Handle centroid rounding inside palette.c itself." into nextgenv2

8 years agoMerge "Insert extra ARFs' in a gf group" into nextgenv2
Wei-ting Lin [Fri, 19 Aug 2016 22:10:41 +0000 (22:10 +0000)]
Merge "Insert extra ARFs' in a gf group" into nextgenv2

8 years agoMerge "Fix compiler warnings in rdopt when warped motion is enabled" into nextgenv2
James Zern [Fri, 19 Aug 2016 21:39:59 +0000 (21:39 +0000)]
Merge "Fix compiler warnings in rdopt when warped motion is enabled" into nextgenv2

8 years agoPalette: count Y colors only for screen content.
Urvang Joshi [Fri, 19 Aug 2016 19:53:58 +0000 (12:53 -0700)]
Palette: count Y colors only for screen content.

Change-Id: Id4e12708598100df54bdfcf8cdb248161ab6ef88

8 years agoHandle centroid rounding inside palette.c itself.
Urvang Joshi [Wed, 10 Aug 2016 22:56:47 +0000 (15:56 -0700)]
Handle centroid rounding inside palette.c itself.

Mostly refactoring, but a very tiny functional change:
Do all rounding in calc_centroids() itself, instead of rounding in two
places inside palette.c

This gives a slight performance improvement for screen content:
0.078% on average.

Change-Id: I7a0e007d30ebf4e59839483a167123f31a222dd4

8 years agoFix compiler warnings in rdopt when warped motion is enabled
Sarah Parker [Fri, 19 Aug 2016 01:49:56 +0000 (18:49 -0700)]
Fix compiler warnings in rdopt when warped motion is enabled

The previous code was giving:
 unused variable ‘tmp_rate’ [-Wunused-variable]
 unused variable ‘tmp_dist’ [-Wunused-variable]
 ‘rate2_nocoeff’ may be used uninitialized in this function [-Wmaybe-uninitialized]

Change-Id: I26326d0e5ffc141ad548654356a877cd3627cea6

8 years agovp10_alloc_context_buffers: clear cm->mi* on failure
James Zern [Fri, 19 Aug 2016 17:39:46 +0000 (10:39 -0700)]
vp10_alloc_context_buffers: clear cm->mi* on failure

based on:
8b4c315 vp9_alloc_context_buffers: clear cm->mi* on failure

Change-Id: I3438a052721b960ff178cb647780f11bc33571fe

8 years agoMerge "Don't send segment probability updates when the map isn't updated." into nextgenv2
Alex Converse [Fri, 19 Aug 2016 16:52:10 +0000 (16:52 +0000)]
Merge "Don't send segment probability updates when the map isn't updated." into nextgenv2

8 years agoMerge "apply clang-format" into nextgenv2
James Zern [Fri, 19 Aug 2016 07:31:49 +0000 (07:31 +0000)]
Merge "apply clang-format" into nextgenv2

8 years agoInsert extra ARFs' in a gf group
Wei-ting Lin [Tue, 26 Jul 2016 21:41:15 +0000 (14:41 -0700)]
Insert extra ARFs' in a gf group

Insert multiple arfs in a gf group to emulate multi-layer backward
reference frames structure. At maximum, two extra ARF's are inserted
in a gf group.

It improves the RD performance by 0.317% in Avg in lowres dataset.

Change-Id: I62c32e1b0f25b978484dd113b319bebcd959bf60

8 years agoDisable global motion experiment when incompatible experiments are enabled
Sarah Parker [Thu, 18 Aug 2016 20:03:35 +0000 (13:03 -0700)]
Disable global motion experiment when incompatible experiments are enabled

This is temporary until the global motion experiment is made to work
with ext_inter and dual_filter.

Change-Id: I73624ca6f536fd98218d7e07bcd7a2c1e6f5aebd

8 years agoapply clang-format
clang-format [Thu, 18 Aug 2016 22:10:22 +0000 (15:10 -0700)]
apply clang-format

after:
253c001 Port dering experiment from aom
7208145 Adding 8x16/16x8/32x16/16x32 transforms

Change-Id: Id93e0d7b72a128701d8dec35fc2fac473944d0c1