]> granicus.if.org Git - libvpx/log
libvpx
8 years agoCode clean on encoder rate controller
Zoe Liu [Thu, 21 Jul 2016 21:30:24 +0000 (14:30 -0700)]
Code clean on encoder rate controller

Change-Id: Iec29c00e24ac8c4f24d43142db6ae03f1b3945ac

8 years agoPalette code: Use built-in qsort() method; create remove_dup() method.
Urvang Joshi [Thu, 4 Aug 2016 23:17:58 +0000 (16:17 -0700)]
Palette code: Use built-in qsort() method; create remove_dup() method.

Change-Id: Id816413307334336a9f473540cf9aa0e789ea9e9

8 years agoMerge "Fix for lossless with rect-tx" into nextgenv2
Debargha Mukherjee [Wed, 10 Aug 2016 19:06:47 +0000 (19:06 +0000)]
Merge "Fix for lossless with rect-tx" into nextgenv2

8 years agoMerge changes I619b365d,I579a9328 into nextgenv2
James Zern [Wed, 10 Aug 2016 19:06:16 +0000 (19:06 +0000)]
Merge changes I619b365d,I579a9328 into nextgenv2

* changes:
  lossless_test: mark tests as Large
  cpu_speed_test: mark speed 0 as Large

8 years agoMerge "Palette code cleanup:" into nextgenv2
Urvang Joshi [Wed, 10 Aug 2016 19:05:11 +0000 (19:05 +0000)]
Merge "Palette code cleanup:" into nextgenv2

8 years agoChange to use proper types
Yaowu Xu [Wed, 10 Aug 2016 01:36:23 +0000 (18:36 -0700)]
Change to use proper types

block: from int64_t to int as it is a block index.
sse: from unsigned int to int64_t to reduce type conversion.

Change-Id: Iec8104ff8a3fd3a77d4e451c12918bd869966c2f

8 years agoFix for lossless with rect-tx
Peter de Rivaz [Wed, 10 Aug 2016 09:19:09 +0000 (10:19 +0100)]
Fix for lossless with rect-tx

Change-Id: Ibb1e5d5137c7717bc6a8683ad78d842c3e5f052e

8 years agolossless_test: mark tests as Large
James Zern [Wed, 10 Aug 2016 03:39:44 +0000 (20:39 -0700)]
lossless_test: mark tests as Large

Change-Id: I619b365d636737da8b1a322bab3be973de53200d

8 years agocpu_speed_test: mark speed 0 as Large
James Zern [Wed, 10 Aug 2016 03:32:51 +0000 (20:32 -0700)]
cpu_speed_test: mark speed 0 as Large

TestTuneScreen / TestScreencastQ0 are the worst offenders

Change-Id: I579a93289aa431afbfea8a280ddcb1011ab1a8cf

8 years agoMerge "vp10_highbd_quantize_fp: use const consistently" into nextgenv2
Yaowu Xu [Wed, 10 Aug 2016 03:13:42 +0000 (03:13 +0000)]
Merge "vp10_highbd_quantize_fp: use const consistently" into nextgenv2

8 years agovp10_highbd_quantize_fp: use const consistently
Yaowu Xu [Wed, 10 Aug 2016 00:59:47 +0000 (17:59 -0700)]
vp10_highbd_quantize_fp: use const consistently

Remove a few extra ones that are consistent with the definitions, this
fixes some MSVC warnings.

Change-Id: I4b26de4cca71f0ac85667bd641c448b44315941b

8 years agoMerge "remove SVC" into nextgenv2
James Zern [Wed, 10 Aug 2016 03:07:07 +0000 (03:07 +0000)]
Merge "remove SVC" into nextgenv2

8 years agoremove SVC
James Zern [Mon, 8 Aug 2016 22:09:30 +0000 (15:09 -0700)]
remove SVC

spatial/temporal scalability are not supported in VP10 currently.
+ remove the unused vp10/encoder/skin_detection.[hc]

this also enables DatarateTestLarge for VP10 which passes with no
experiments enabled. these were removed previously when only the SVC
tests should have been:
134710a Disable tests not applicable to VP10

Change-Id: I9ee7a0dd5ad3d8cc1e8fd5f0a90260fa43da387c

8 years agoMerge "Add interface to compute gm parameters in encodeframe" into nextgenv2
Sarah Parker [Wed, 10 Aug 2016 00:37:28 +0000 (00:37 +0000)]
Merge "Add interface to compute gm parameters in encodeframe" into nextgenv2

8 years agoMerge "remove vp8cx_set_ref.c" into nextgenv2
James Zern [Wed, 10 Aug 2016 00:03:54 +0000 (00:03 +0000)]
Merge "remove vp8cx_set_ref.c" into nextgenv2

8 years agoAdd interface to compute gm parameters in encodeframe
Sarah Parker [Tue, 9 Aug 2016 01:11:05 +0000 (18:11 -0700)]
Add interface to compute gm parameters in encodeframe

This patch just creates the interface for global motion computation
and calls it from encodeframe. Currently, the function
compute_global_motion_feature_based is empty and the work to do
the actual parameter calculation will be added in a future patch.

Change-Id: Ife142742140079e1c1743b66f180aeb2ecea29ae

8 years agoMerge "Fix a bug for multi_arf_allowed" into nextgenv2
Wei-ting Lin [Tue, 9 Aug 2016 21:09:56 +0000 (21:09 +0000)]
Merge "Fix a bug for multi_arf_allowed" into nextgenv2

8 years agoPalette code cleanup:
Urvang Joshi [Thu, 4 Aug 2016 18:00:17 +0000 (11:00 -0700)]
Palette code cleanup:

- Avoid some memcpy()s
- Remove indices array
- Make pre_indices array local
- Avoid rounding twice
- Other small simplifications

Change-Id: Iac3236daaad04f21f54054cdd9504de13b942a07

8 years agoremove vp8cx_set_ref.c
James Zern [Tue, 9 Aug 2016 00:14:04 +0000 (17:14 -0700)]
remove vp8cx_set_ref.c

and the related tests. vpxcx_set_ref is the binary to use for vp10.

Change-Id: I4c4ce7b36b165e6d06b87fd6b53923a1c11e4e6c

8 years agoMerge "configure: test for -Wfloat-conversion" into nextgenv2
James Zern [Mon, 8 Aug 2016 21:48:14 +0000 (21:48 +0000)]
Merge "configure: test for -Wfloat-conversion" into nextgenv2

8 years agoMerge "Optimization EXT_INTRA's filtered intra predictor (SSE4.1)" into nextgenv2
Yi Luo [Mon, 8 Aug 2016 20:55:43 +0000 (20:55 +0000)]
Merge "Optimization EXT_INTRA's filtered intra predictor (SSE4.1)" into nextgenv2

8 years agoAdd reconstruction using gm parameters
Sarah Parker [Wed, 3 Aug 2016 19:29:27 +0000 (12:29 -0700)]
Add reconstruction using gm parameters

This patch only includes inter frame reconstruction using gm
parameters when GLOBAL_MOTION and/or VP9_HIGHBITDEPTH are enabled.
GM is not currently used when EXT_INTER or DUAL_FILTER is enabled.
This will be added in a followup patch. For now, these experiments
will take precedence over GLOBAL_MOTION when they are all enabled.

Change-Id: I930ddda529c44d7245dbb56db3c9c5524cf45473

8 years agoOptimization EXT_INTRA's filtered intra predictor (SSE4.1)
Yi Luo [Tue, 2 Aug 2016 16:57:08 +0000 (09:57 -0700)]
Optimization EXT_INTRA's filtered intra predictor (SSE4.1)

- Add unit tests to verify the bit-exact result.
- In speed test, function speed (for each mode/tx_size)
  improves about 23%~35%.
- On E5-2680, park_joy_1080p, 10 frames, --kf-max-dist=1,
  encoding time improves about 1%~2%.

Change-Id: Id89f313d44eea562c02e775a6253dc4df7e046a9

8 years agoMerge "Speed filter intra mode search in EXT_INTRA experiment" into nextgenv2
Yue Chen [Sat, 6 Aug 2016 00:17:33 +0000 (00:17 +0000)]
Merge "Speed filter intra mode search in EXT_INTRA experiment" into nextgenv2

8 years agoconfigure: test for -Wfloat-conversion
James Zern [Fri, 22 Jul 2016 20:13:56 +0000 (13:13 -0700)]
configure: test for -Wfloat-conversion

supported by clang, gcc-4.9+

Change-Id: I893766de7307fef9a8b68c0cfae137c9d3b0dbe8
(cherry picked from commit 889ed5b158fc280927f2de9172d48245c3b735a7)

8 years agoMerge "warped_motion: remove unused vp10_integerize_model" into nextgenv2
James Zern [Sat, 6 Aug 2016 00:02:14 +0000 (00:02 +0000)]
Merge "warped_motion: remove unused vp10_integerize_model" into nextgenv2

8 years agoSpeed filter intra mode search in EXT_INTRA experiment
Yue Chen [Fri, 5 Aug 2016 17:37:24 +0000 (10:37 -0700)]
Speed filter intra mode search in EXT_INTRA experiment

(1) Key frame: skip filter intra modes whose directional pred
    version is relatively bad (rd >= 1.125 * best_rd)
(2) Inter frame: do not check filter intra modes if best_intra_rd
    >= 1.25 * best_rd

Encoding time overhead is reduced by:
4.9% (9.2%->4.3%, soccer_cif)
Coding gains drop by 0.021% on lowres and by 0.076% on midres

Change-Id: I29b6f7d3d3dc4b362c6d63bc447e6a429ba5dc66

8 years agoFix a bug for multi_arf_allowed
Wei-ting Lin [Fri, 5 Aug 2016 20:32:47 +0000 (13:32 -0700)]
Fix a bug for multi_arf_allowed

The ARF Index was wrong when updating the upsampled reference
frame buffer.

Compared to the baseline in which multi_arf_allowed is disabled, the
RD performance drops 2.250% in Avg using Overall PSNR in the derf
dataset. The performance decrease is especially in the following
video sequences:

foreman_cif: drops 7.489%
husky_cif: drops 6.421%
soccer_cif: drops 4.850%

However, it has a significant gain in the following video sequences:

container_cif: increases 8.043%
harbour_cif: increases 1.332%

Change-Id: I02472909eb34bd070d7544f57383e72559fa42b3

8 years agoPalette code: simpler and faster duplicate removal
Urvang Joshi [Thu, 4 Aug 2016 00:02:03 +0000 (17:02 -0700)]
Palette code: simpler and faster duplicate removal

Change-Id: I0c1baa5ca73c1f067d69239d3e31d1050b4706d2

8 years agoMerge "Code refactoring on Macros related to ref frame numbers" into nextgenv2
Zoe Liu [Fri, 5 Aug 2016 16:59:35 +0000 (16:59 +0000)]
Merge "Code refactoring on Macros related to ref frame numbers" into nextgenv2

8 years agoMerge "Make palette code faster: replace nested for loops by a single memcpy()."...
Urvang Joshi [Fri, 5 Aug 2016 00:20:11 +0000 (00:20 +0000)]
Merge "Make palette code faster: replace nested for loops by a single memcpy()." into nextgenv2

8 years agoMerge "Replace variants of 'vp8' and 'vp9' with 'vpx'" into nextgenv2
Yaowu Xu [Thu, 4 Aug 2016 23:48:59 +0000 (23:48 +0000)]
Merge "Replace variants of 'vp8' and 'vp9' with 'vpx'" into nextgenv2

8 years agoMerge changes Ic5ddba3c,Ibe7a3248 into nextgenv2
Yaowu Xu [Thu, 4 Aug 2016 23:27:35 +0000 (23:27 +0000)]
Merge changes Ic5ddba3c,Ibe7a3248 into nextgenv2

* changes:
  Fix a number of msvc warnings
  Reduce number of frames in lossless tests

8 years agoReplace variants of 'vp8' and 'vp9' with 'vpx'
Yaowu Xu [Thu, 4 Aug 2016 20:57:22 +0000 (13:57 -0700)]
Replace variants of 'vp8' and 'vp9' with 'vpx'

Change-Id: Id6cb96b0b15efdda63348d8bfe59fc0533c85ba1

8 years agoMake palette code faster: replace nested for loops by a single memcpy().
Urvang Joshi [Wed, 3 Aug 2016 23:42:11 +0000 (16:42 -0700)]
Make palette code faster: replace nested for loops by a single memcpy().

Change-Id: Ia14df45a35c98d680822454fbb8d1763884c1852

8 years agoFix a number of msvc warnings
Yaowu Xu [Wed, 3 Aug 2016 21:36:23 +0000 (14:36 -0700)]
Fix a number of msvc warnings

Change-Id: Ic5ddba3ca0c87245617b6dbc78c0f13dc952ce8b

8 years agoReduce number of frames in lossless tests
Yaowu Xu [Thu, 4 Aug 2016 21:36:29 +0000 (14:36 -0700)]
Reduce number of frames in lossless tests

This it to reduce the time necessary for these tests.

Change-Id: Ibe7a3248a6c45baf575af85fdffcffc557dd054b

8 years agovp10_inv_txfm2d_test: normalize max_error type
James Zern [Wed, 3 Aug 2016 02:54:17 +0000 (19:54 -0700)]
vp10_inv_txfm2d_test: normalize max_error type

quiets double -> int conversion warning

Change-Id: Ic860d187bc77e18b277eef28310feee1899cdbe6

8 years agoMerge ".clang-format: disable DerivePointerAlignment" into nextgenv2
James Zern [Thu, 4 Aug 2016 19:21:38 +0000 (19:21 +0000)]
Merge ".clang-format: disable DerivePointerAlignment" into nextgenv2

8 years agoMerge "more cleanup of vp8 and vp9" into nextgenv2
Yaowu Xu [Thu, 4 Aug 2016 18:56:46 +0000 (18:56 +0000)]
Merge "more cleanup of vp8 and vp9" into nextgenv2

8 years agoCode refactoring on Macros related to ref frame numbers
Zoe Liu [Wed, 3 Aug 2016 21:46:43 +0000 (14:46 -0700)]
Code refactoring on Macros related to ref frame numbers

We have renamed following Macros to avoid name confusion:

REFS_PER_FRAME --> INTER_REFS_PER_FRAME
(= ALTREF_FRAME - LAST_FRAME + 1)
MAX_REF_FRAMES --> TOTAL_REFS_PER_FRAME
(= ALTREF_FRAME - INTRA_FRAME + 1)

INTER_REFS_PER_FRAME specifies the maximum number of reference frames
that each Inter frame may use.
TOTAL_REFS_PER_FRAME is equal to INTER_REFS_PER_FRAME + 1, which
counts the INTRA_FRAME.

Further, at the encoder side, since REF_FRAMES specifies the maximum
number of the reference frames that the encoder may store, REF_FRAMES
is usually larger than INTER_REFS_PER_FRAME. For example, in the
ext-refs experiment, REF_FRAMES == 8, which allows the encoder to
store maximum 8 reference frames in the buffer, but
INTER_REFS_PER_FRAME equals to 6, which allows each Inter frame may
use up to 6 frames out of the 8 buffered frames as its references.
Hence, in order to explore the possibility to store more reference
frames in future patches, we modified a couple of array sizes to
accomodate the case that the number of buffered reference frames is
not always equal to the number of the references that are being used
by each Inter frame.

Change-Id: I19e42ef608946cc76ebfd3e965a05f4b9b93a0b3

8 years agowarped_motion: remove unused vp10_integerize_model
James Zern [Wed, 3 Aug 2016 22:52:03 +0000 (15:52 -0700)]
warped_motion: remove unused vp10_integerize_model

this function produces implicit double -> int conversion warnings and
has additional style issues.

Change-Id: I6bc740e778658d6f81ca54888fc6fa822d3b5ee0

8 years agoMerge "Adjust gm parameter computation to avoid mismatch" into nextgenv2
Sarah Parker [Wed, 3 Aug 2016 21:42:30 +0000 (21:42 +0000)]
Merge "Adjust gm parameter computation to avoid mismatch" into nextgenv2

8 years agoMerge "Cherry pick from AOMedia" into nextgenv2
Yaowu Xu [Wed, 3 Aug 2016 19:25:22 +0000 (19:25 +0000)]
Merge "Cherry pick from AOMedia" into nextgenv2

8 years agoMerge "Cherry pick from AOM:" into nextgenv2
Yaowu Xu [Wed, 3 Aug 2016 19:25:17 +0000 (19:25 +0000)]
Merge "Cherry pick from AOM:" into nextgenv2

8 years agomore cleanup of vp8 and vp9
Yaowu Xu [Wed, 3 Aug 2016 18:33:39 +0000 (11:33 -0700)]
more cleanup of vp8 and vp9

Change-Id: Ic90ebe6136f4b75645ba699d49c0bcb3764ddccf

8 years agoAdjust gm parameter computation to avoid mismatch
Sarah Parker [Tue, 2 Aug 2016 19:28:41 +0000 (12:28 -0700)]
Adjust gm parameter computation to avoid mismatch

The gm parameters need to have WARPED_PRECISION_BITS precision
until they are written to the bitstream because functions in
reconinter use these parameters before they are written to
the bitstream. Previously, the parameters weren't being converted
to WARPED_PRECISION_BITS until they were read from the bitstream
which causes an encode/decode mismatch.

Change-Id: I31e76e9d6f7d24df21af287a72f8c01f1997304d

8 years agoMerge "Fix a bug and a function name in EXT_INTRA experiment" into nextgenv2
Yue Chen [Wed, 3 Aug 2016 17:04:14 +0000 (17:04 +0000)]
Merge "Fix a bug and a function name in EXT_INTRA experiment" into nextgenv2

8 years agoCherry pick from AOMedia
Yaowu Xu [Fri, 29 Jul 2016 23:45:53 +0000 (16:45 -0700)]
Cherry pick from AOMedia

5b5fbad VP9LfSync->VP10LfSync
b752848 vp8_yv12 -> vpx_yv12
e5068cd VP9->VPX for reference frame flags

Change-Id: Ia36860499c81a5aca8cd6190e7370ec404c0df0f

8 years agoFix a bug and a function name in EXT_INTRA experiment
Yue Chen [Tue, 2 Aug 2016 22:18:44 +0000 (15:18 -0700)]
Fix a bug and a function name in EXT_INTRA experiment

(1) Apply ALLOW_FILTER_INTRA_MODES flag to the correct place, otherwise
there are bitstream mismatchs when it is 0.
(2) Rename pick_ext_intra_iframe() to pick_ext_intra_interframe().

Change-Id: Ic88c930de1d3f819750f0892df52bde55ae32a91

8 years agoCherry pick from AOM:
Yaowu Xu [Fri, 29 Jul 2016 22:38:55 +0000 (15:38 -0700)]
Cherry pick from AOM:

68e7e4d0 Remove VP9_CAP_POSTPROC
0738390c Remove vp9_temporal denoise
b89861a4 Remove vp9-postproc

Change-Id: I4ecaa0ac83a519c8174a494378fc23df610ff2a8

8 years agoDisable tests not applicable to VP10
Yaowu Xu [Mon, 1 Aug 2016 20:32:40 +0000 (13:32 -0700)]
Disable tests not applicable to VP10

As VP10 does not support multiple layers yet, we disable the ported
tests from VP9.

Change-Id: Ib7577c27e402ede481213b7a64ebee7576a025a5

8 years agoMerge "Use all possible intra ref pixels for blocks on frame boundary" into nextgenv2
Hui Su [Mon, 1 Aug 2016 17:29:09 +0000 (17:29 +0000)]
Merge "Use all possible intra ref pixels for blocks on frame boundary" into nextgenv2

8 years agoMerge "Cherry pick renaming changes from AOMedia branch" into nextgenv2
Yaowu Xu [Mon, 1 Aug 2016 17:16:06 +0000 (17:16 +0000)]
Merge "Cherry pick renaming changes from AOMedia branch" into nextgenv2

8 years agoMerge "Rename files with vp9_ prefix" into nextgenv2
Yaowu Xu [Mon, 1 Aug 2016 17:15:49 +0000 (17:15 +0000)]
Merge "Rename files with vp9_ prefix" into nextgenv2

8 years agoCherry pick renaming changes from AOMedia branch
Yaowu Xu [Fri, 29 Jul 2016 17:14:19 +0000 (10:14 -0700)]
Cherry pick renaming changes from AOMedia branch

Manually cherry-picked the following changes:
8c8d16de vp9 -> vpx in names
75b57d39 VP9_ -> VPX_ in function names
761a7088 VP9_INTERP_EXTEND -> VPX_INTERP_EXTEND
4273a52c VP9->VPX in border pixel macros
03568c31 VP9_FRAME_MARKER -> VPX_FRAME_MARKER
2334f51d VP9->VPX in fdct function names

Change-Id: Icc18dbf4b416dd0fa21033b3e19ab8a47c893508

8 years agoUse all possible intra ref pixels for blocks on frame boundary
hui su [Thu, 28 Jul 2016 17:44:58 +0000 (10:44 -0700)]
Use all possible intra ref pixels for blocks on frame boundary

Tested on lowres and midres, performance impact is neutral.

BUG=https://bugs.chromium.org/p/webm/issues/detail?id=1269

Change-Id: Idaccaf7da7b460e6201efd436c084be66b5d4ebd

8 years agoMerge "More vp8/vp9 clean up" into nextgenv2
Yaowu Xu [Fri, 29 Jul 2016 16:50:33 +0000 (16:50 +0000)]
Merge "More vp8/vp9 clean up" into nextgenv2

8 years agoRename files with vp9_ prefix
Yaowu Xu [Thu, 28 Jul 2016 18:51:48 +0000 (11:51 -0700)]
Rename files with vp9_ prefix

Change-Id: I9c51ae3a2af698efe32288b807f881385e19822b

8 years agoMore vp8/vp9 clean up
Yaowu Xu [Wed, 27 Jul 2016 22:52:42 +0000 (15:52 -0700)]
More vp8/vp9 clean up

Change-Id: I8101de20e873c19d03c7fd2977bc22003e395807

8 years agoFix a missing space
hui su [Thu, 28 Jul 2016 16:54:50 +0000 (09:54 -0700)]
Fix a missing space

Change-Id: I6a9ab351012b731308d6b4fc78c09744c6afb61a

8 years agoMerge "Separate frame context index for different frame types" into nextgenv2
Jingning Han [Thu, 28 Jul 2016 16:39:44 +0000 (16:39 +0000)]
Merge "Separate frame context index for different frame types" into nextgenv2

8 years agoRemove vp8, vp9 folders
Yaowu Xu [Mon, 25 Jul 2016 22:39:46 +0000 (15:39 -0700)]
Remove vp8, vp9 folders

Change-Id: I09b8acd22d031ece52e1fee18b998349bf1cf06b

8 years agoSeparate frame context index for different frame types
Jingning Han [Wed, 27 Jul 2016 16:12:53 +0000 (09:12 -0700)]
Separate frame context index for different frame types

This commit makes the encoder to use different frame context index
for different frame types. In the baseline setting, it sets the
frame context index of the overlay frame to be different from other
regular inter frames. In the ext-refs setting, it further allows
the backward reference frame to use a different index.

It improves the compression performance for both settings.

Baseline
lowres  0.12%

ext-refs
lowres  0.50%
midres  0.56%

Change-Id: I7c63ddec9fc296c56a86353cf2c661a740b97a97

8 years ago.clang-format: disable DerivePointerAlignment
James Zern [Tue, 26 Jul 2016 23:34:54 +0000 (16:34 -0700)]
.clang-format: disable DerivePointerAlignment

everything outside of third_party should follow 'PointerAlignment:
right' i.e., associate the '*' with the variable

+ add a note about the clang-format that generated this file

Change-Id: I13e3f4f5fb6e22a8fa7fc3d06879c995b7c41a39
(cherry picked from commit e4290800b21478a3f9548c58b4f15c5ba5393073)

8 years agoMerge "ext-intra: squeeze the derivative table" into nextgenv2
Hui Su [Tue, 26 Jul 2016 21:53:50 +0000 (21:53 +0000)]
Merge "ext-intra: squeeze the derivative table" into nextgenv2

8 years agoext-intra: squeeze the derivative table
hui su [Tue, 26 Jul 2016 01:08:08 +0000 (18:08 -0700)]
ext-intra: squeeze the derivative table

Reduce its size form 270x2 to 90.

Change-Id: Icaacc4258e43cdc19c6f06598fee17c3dd06d7e9

8 years agoMerge "Fix compilation error under Clang 4.0." into nextgenv2
Yaowu Xu [Tue, 26 Jul 2016 17:21:57 +0000 (17:21 +0000)]
Merge "Fix compilation error under Clang 4.0." into nextgenv2

8 years agoMerge "MinArfFreqTest: Don't leak video on failure." into nextgenv2
Yaowu Xu [Tue, 26 Jul 2016 17:21:46 +0000 (17:21 +0000)]
Merge "MinArfFreqTest: Don't leak video on failure." into nextgenv2

8 years agoMerge "blockiness_test: fix implicit float conversion" into nextgenv2
Yaowu Xu [Tue, 26 Jul 2016 17:21:37 +0000 (17:21 +0000)]
Merge "blockiness_test: fix implicit float conversion" into nextgenv2

8 years agoMerge "resize_test: fix implicit float->int conversion" into nextgenv2
Yaowu Xu [Tue, 26 Jul 2016 17:21:30 +0000 (17:21 +0000)]
Merge "resize_test: fix implicit float->int conversion" into nextgenv2

8 years agoMerge "Add VPX_SWAP macro" into nextgenv2
Yaowu Xu [Tue, 26 Jul 2016 17:21:22 +0000 (17:21 +0000)]
Merge "Add VPX_SWAP macro" into nextgenv2

8 years agoMerge "Make test encoder test driver less likely to leak on failure." into nextgenv2
Yaowu Xu [Tue, 26 Jul 2016 17:21:05 +0000 (17:21 +0000)]
Merge "Make test encoder test driver less likely to leak on failure." into nextgenv2

8 years agoMerge "Combine vpxcx_set_ref example for VP9 and VP10" into nextgenv2
Yunqing Wang [Tue, 26 Jul 2016 04:16:21 +0000 (04:16 +0000)]
Merge "Combine vpxcx_set_ref example for VP9 and VP10" into nextgenv2

8 years agoFix compilation error under Clang 4.0.
Ivan Krasin [Mon, 25 Jul 2016 19:11:02 +0000 (12:11 -0700)]
Fix compilation error under Clang 4.0.

The LLVM trunk has reached 4.0 and now __clang_major__ is not enough
to distinguish between old XCode Clang and the new 'real' Clang.
Using __apple_build_version__ allows to make this distinction.

BUG=chromium:631144

Change-Id: I0b6e46fddfe4f409c7b7e558bda34872e60ee2d9

8 years agoMinArfFreqTest: Don't leak video on failure.
Alex Converse [Thu, 21 Jul 2016 18:38:27 +0000 (11:38 -0700)]
MinArfFreqTest: Don't leak video on failure.

Change-Id: I250379f0ac8d4929c9032e7343290e2980fc2e77

8 years agoblockiness_test: fix implicit float conversion
James Zern [Fri, 22 Jul 2016 22:08:59 +0000 (15:08 -0700)]
blockiness_test: fix implicit float conversion

float->int as reported by -Wfloat-conversion

Change-Id: Icb0ecb9e2d54edb95813d9f2de34cb6c27b63cbd
(cherry picked from commit 5e2791b54da066cc8543c066813373c9274ff53f)

8 years agoresize_test: fix implicit float->int conversion
James Zern [Fri, 22 Jul 2016 20:11:07 +0000 (13:11 -0700)]
resize_test: fix implicit float->int conversion

Change-Id: I1efc16fa158740a06da719a1ea90c6dd6a182bb4
(cherry picked from commit 325bdddc38ad15cf7dd2bc618461a13cfb006812)

8 years agoAdd VPX_SWAP macro
Yury Gitman [Fri, 15 Jul 2016 17:32:04 +0000 (10:32 -0700)]
Add VPX_SWAP macro

Change-Id: I60e233eddef238ad918183392794084673f27d2d

8 years agoMake test encoder test driver less likely to leak on failure.
Alex Converse [Thu, 21 Jul 2016 18:36:41 +0000 (11:36 -0700)]
Make test encoder test driver less likely to leak on failure.

Individual tests still need to be updated.

Change-Id: Ic433d0f742e13560b136f136b72b2a9973970d78

8 years agoy4minput.c: correct empty loop formatting
James Zern [Wed, 20 Jul 2016 03:46:39 +0000 (20:46 -0700)]
y4minput.c: correct empty loop formatting

prefer {}s over ';'

Change-Id: I563fc82717e1deb4f42a40e03dca318c6adaa0c1

8 years agobuild/make/Makefile: add a 'test_*' default target
James Zern [Mon, 18 Jul 2016 23:30:58 +0000 (16:30 -0700)]
build/make/Makefile: add a 'test_*' default target

allows 'make test_libvpx', etc. some reworking of the makefiles would be
needed to avoid hard coding targets here.

Change-Id: I18982dbf691e7d36ab8bcf5934bab9340687b061
(cherry picked from commit 25085a6ac21fc8e2341b92e2f1f14d5a7fef30c6)

8 years agobuild/make/Makefile: remove default suffix rules
James Zern [Mon, 18 Jul 2016 23:30:40 +0000 (16:30 -0700)]
build/make/Makefile: remove default suffix rules

Change-Id: I15c8976c6478bf75ec617398f49461b310ab7569

8 years agoMerge "take II: variance_test partial clean-up" into nextgenv2
Yaowu Xu [Tue, 26 Jul 2016 04:02:36 +0000 (04:02 +0000)]
Merge "take II: variance_test partial clean-up" into nextgenv2

8 years agotake II: variance_test partial clean-up
skal [Thu, 14 Jul 2016 00:33:22 +0000 (17:33 -0700)]
take II: variance_test partial clean-up

remove some (but not all yet!) tuple mis-use, and revamp the code a lot.
Factorize some common chores into MainTestClass.

Change-Id: Id37b7330eebe80d19b9d12a454f24ff9be6b1116

8 years agoCombine vpxcx_set_ref example for VP9 and VP10
Yunqing Wang [Sat, 23 Jul 2016 00:14:22 +0000 (17:14 -0700)]
Combine vpxcx_set_ref example for VP9 and VP10

Modified the example so that the test script worked correctly. Also
included minor fixes.

Change-Id: If97525dd9b85004a07e91c384132eadae35cdbf8

8 years agoMerge "inline->INLINE: vpx_dsp/x86/blend_a64_vmask_sse4.c" into nextgenv2
Yaowu Xu [Mon, 25 Jul 2016 20:48:52 +0000 (20:48 +0000)]
Merge "inline->INLINE: vpx_dsp/x86/blend_a64_vmask_sse4.c" into nextgenv2

8 years agoMerge "Adjust adaptive_rd_thresh speed feature" into nextgenv2
Yunqing Wang [Mon, 25 Jul 2016 18:53:27 +0000 (18:53 +0000)]
Merge "Adjust adaptive_rd_thresh speed feature" into nextgenv2

8 years agoinline->INLINE: vpx_dsp/x86/blend_a64_vmask_sse4.c
Yaowu Xu [Mon, 25 Jul 2016 17:43:12 +0000 (10:43 -0700)]
inline->INLINE: vpx_dsp/x86/blend_a64_vmask_sse4.c

This fixes the build using MSVC.

Change-Id: I7788e28fd4f0c6ff3d405c4b4a2ff59eda3ba8b6

9 years agovp10/ratectrl.c: fix implicit float conversion
James Zern [Fri, 22 Jul 2016 20:23:10 +0000 (13:23 -0700)]
vp10/ratectrl.c: fix implicit float conversion

float->int as reported by -Wfloat-conversion

Change-Id: I3c55990821d922bda7a7600c00ae8d5dcc3cee94

9 years agord_pick_palette_intra_sbuv: fix implicit float conv
James Zern [Fri, 22 Jul 2016 20:21:55 +0000 (13:21 -0700)]
rd_pick_palette_intra_sbuv: fix implicit float conv

float->int as reported by -Wfloat-conversion

Change-Id: I9e3d6ce9dbb0689f214afc8d5950f209275e883d

9 years agoFix compiler errors in warped_motion.c
Sarah Parker [Fri, 22 Jul 2016 18:51:15 +0000 (11:51 -0700)]
Fix compiler errors in warped_motion.c

A few functions and variables were named incorrectly

Change-Id: Ib32fa459c51b9e9aad8bb107e1b689a96d98b368

9 years agoMerge "Restore vp10_default_scan_orders[]" into nextgenv2
James Zern [Fri, 22 Jul 2016 22:24:58 +0000 (22:24 +0000)]
Merge "Restore vp10_default_scan_orders[]" into nextgenv2

9 years agoAdjust adaptive_rd_thresh speed feature
Yunqing Wang [Thu, 21 Jul 2016 16:00:36 +0000 (09:00 -0700)]
Adjust adaptive_rd_thresh speed feature

Set adaptive_rd_thresh to 0 at speed 0. This allows a thorough mode
search, and eliminates a blocking artifact seen in an encoder test.

Borg test:
1. lowres
Overall PSNR: -0.135%; SSIM: -0.293%;
2. hdres
Overall PSNR: -0.122%; SSIM: -0.208%;
Encoder speed tests: 2% - 6% slower.

Change-Id: Ie7601cb8824df8f6f2ae0b2942bd938600f76990

9 years agoRestore vp10_default_scan_orders[]
Sarah Parker [Fri, 22 Jul 2016 18:29:31 +0000 (11:29 -0700)]
Restore vp10_default_scan_orders[]

vp10_default_scan_orders was removed in:
e5848de Rectangular transforms 4x8 & 8x4
This fixes compiler error in vp10_quantize_test.cc
Change-Id: I1b8a637e011f9426c3b41e61e00e3babc80defba

9 years agoadd .clang-format, based on Google style
James Zern [Thu, 14 Jul 2016 22:18:30 +0000 (15:18 -0700)]
add .clang-format, based on Google style

derived from clang-format 3.7.1; same as used in libaom

Change-Id: I8ea915a41d1f2ea3b0d4e4dab9ebc808e9116f11

9 years agoAdd and remove explicit type conversions
Yaowu Xu [Fri, 22 Jul 2016 17:04:06 +0000 (10:04 -0700)]
Add and remove explicit type conversions

Change-Id: I8b791fda7c64a0363549add99dc9fcae3b29beae

9 years agoFix compiling issues
Yaowu Xu [Fri, 22 Jul 2016 16:30:52 +0000 (09:30 -0700)]
Fix compiling issues

Change-Id: I530348b12a1c039842ce4e33d21046fe63878f19

9 years agoMerge "Add affine model to global motion" into nextgenv2
Sarah Parker [Thu, 21 Jul 2016 23:17:33 +0000 (23:17 +0000)]
Merge "Add affine model to global motion" into nextgenv2

9 years agoMerge "Rectangular transforms 4x8 & 8x4" into nextgenv2
Debargha Mukherjee [Thu, 21 Jul 2016 21:54:12 +0000 (21:54 +0000)]
Merge "Rectangular transforms 4x8 & 8x4" into nextgenv2