platform/upstream/libvpx.git
7 years agoMerge "vp9_detokenize,decode_coefs: fix signed int overflow"
James Zern [Thu, 29 Sep 2016 22:36:12 +0000 (22:36 +0000)]
Merge "vp9_detokenize,decode_coefs: fix signed int overflow"

7 years agoMerge changes I11786887,Ia91823ad
James Zern [Thu, 29 Sep 2016 19:11:35 +0000 (19:11 +0000)]
Merge changes I11786887,Ia91823ad

* changes:
  vpx_dsp/get_prob: relocate den == 0 test
  vpx_dsp/get_prob: make clip_prob branchless

7 years agovp9_detokenize,decode_coefs: fix signed int overflow
James Zern [Thu, 29 Sep 2016 03:37:18 +0000 (20:37 -0700)]
vp9_detokenize,decode_coefs: fix signed int overflow

when decoding an invalid bitstream with --enable-vp9-highbitdepth

BUG=webm:1297

Change-Id: I401d87033b4293f2ca595bc51678aad9951ecf15

7 years agovpx_dsp/get_prob: relocate den == 0 test
James Zern [Thu, 29 Sep 2016 00:39:05 +0000 (17:39 -0700)]
vpx_dsp/get_prob: relocate den == 0 test

to get_binary_prob(). the only other caller mode_mv_merge_probs() does
its own test on 0.

BUG=chromium:639712

Change-Id: I1178688706baeca2883f7aadbc254abb219a44ce

7 years agoMerge "vp9: fix compilation for g++ 6.2.x"
James Zern [Wed, 28 Sep 2016 23:36:23 +0000 (23:36 +0000)]
Merge "vp9: fix compilation for g++ 6.2.x"

7 years agoMerge "Hook up vp8_diamond_search_sad_sse3"
Johann Koenig [Wed, 28 Sep 2016 20:54:24 +0000 (20:54 +0000)]
Merge "Hook up vp8_diamond_search_sad_sse3"

7 years agoMerge "vpxdec: avoid memory leaks under most conditions"
James Zern [Wed, 28 Sep 2016 19:35:16 +0000 (19:35 +0000)]
Merge "vpxdec: avoid memory leaks under most conditions"

7 years agovpx_dsp/get_prob: make clip_prob branchless
James Zern [Wed, 28 Sep 2016 02:43:03 +0000 (19:43 -0700)]
vpx_dsp/get_prob: make clip_prob branchless

+ inline the function directly as there was only one consumer
(get_prob())

this is an attempt to reduce the amount of branches to workaround an amd
bug. this change is mildly faster or neutral across x86-64, arm.

http://support.amd.com/TechDocs/44739_12h_Rev_Gd.pdf
665 Integer Divide Instruction May Cause Unpredictable Behavior

BUG=chromium:639712

Suggested-by: Pascal Massimino <pascal.massimino@gmail.com>
Change-Id: Ia91823aded79aab469dd68095d44300e8df04ed2

7 years agovp9: fix compilation for g++ 6.2.x
Tristan Matthews [Thu, 22 Sep 2016 22:49:27 +0000 (18:49 -0400)]
vp9: fix compilation for g++ 6.2.x

Inline function called from test/dct16x16_test.cc wouldn't build due to:
  invalid operands of types ‘__gnu_cxx::__enable_if<true, double>::__type
  {aka double}’ and ‘int’ to binary ‘operator>>’
  return (abs(ref->row) >> 3) < COMPANDED_MVREF_THRESH &&

this converts the test to abs() < COMPANDED_MVREF_THRESH << 3 which
hides the promotion issue.

Regression from commit de993a847f8080d3128420c8ef8495642013bdb1

BUG=webm:1291

Change-Id: I73b5943d07d5b61b709d299114216a2371a8fd62

7 years agoMerge "Clean convolve_test.cc"
Linfeng Zhang [Wed, 28 Sep 2016 00:20:27 +0000 (00:20 +0000)]
Merge "Clean convolve_test.cc"

7 years agoMerge "Fix for compile error with range checking"
Debargha Mukherjee [Wed, 28 Sep 2016 00:05:37 +0000 (00:05 +0000)]
Merge "Fix for compile error with range checking"

7 years agoCast strto[u]l down
Johann [Tue, 27 Sep 2016 22:37:10 +0000 (15:37 -0700)]
Cast strto[u]l down

Change-Id: I945b2f8754cf484a08e5ba511cfd2d4a44181b08

7 years agoHook up vp8_diamond_search_sad_sse3
Johann [Mon, 26 Sep 2016 23:20:32 +0000 (16:20 -0700)]
Hook up vp8_diamond_search_sad_sse3

The original commit never set any 'specialize' line:
61311e61039c300ae872ccba22304e9e60dc0205

It appears the sadx4 version of function uses sdx4df calls to speed up
the search. There are no sse3 versions of the sdx4df functions, but
there are sse2 and msa versions.

There is a neon version of vpx_sad16x16x4d but not any of the smaller
versions. Perhaps if they existed this function could be expanded to use
them.

Change-Id: I936d7d6b1a3ff6dcd5a4d2322272708c47cdec13

7 years agovpxdec: avoid memory leaks under most conditions
James Zern [Tue, 27 Sep 2016 21:29:18 +0000 (14:29 -0700)]
vpxdec: avoid memory leaks under most conditions

avoids false positives when fuzzing with ASan+LSan.

Change-Id: I0d23b530ae80e5692b6951fe6e3690ea44159a5a

7 years agoMerge changes from topic 'wextra'
Johann Koenig [Tue, 27 Sep 2016 21:13:50 +0000 (21:13 +0000)]
Merge changes from topic 'wextra'

* changes:
  Expand -Wextra to more of the library
  mips: clean up wextra warnings
  Add compiler flag -Wsign-compare
  Add compiler warning flag -Wextra and fix related warnings.

7 years agoClean convolve_test.cc
Linfeng Zhang [Tue, 20 Sep 2016 21:23:15 +0000 (14:23 -0700)]
Clean convolve_test.cc

Combine test MatchesReferenceSubpixelFilter and
MatchesReferenceAveragingSubpixelFilter.

Change-Id: I75f96befbbb118cdc6b8c6001b4cdda8d88fbbd3

7 years agoExpand -Wextra to more of the library
Johann [Tue, 13 Sep 2016 21:01:38 +0000 (14:01 -0700)]
Expand -Wextra to more of the library

Suppress warnings in third_party/.

vp8 -Wclobbered issue is tracked here:
BUG=webm:1246

BUG=webm:1069

Change-Id: I9b94bf546d7b690c26a59ae67967facdce8ec45b

7 years agomips: clean up wextra warnings
Johann [Tue, 27 Sep 2016 20:17:51 +0000 (13:17 -0700)]
mips: clean up wextra warnings

Remove unused zbin variable:
warning: unused parameter ‘zbin’

Use int for loop variables to avoid unsigned conversion:
warning: comparison between signed and unsigned integer expressions

Change-Id: Icea74b870c0ee68a8bf687e796a69392af25a8ad

7 years agoAdd compiler flag -Wsign-compare
Urvang Joshi [Fri, 8 Jul 2016 23:09:36 +0000 (16:09 -0700)]
Add compiler flag -Wsign-compare

Also, fix the warnings generated by this flag.

(cherry picked from commit ebeb1155d4fa6d28e2f40c92265245f8df097fcb)

From AOM. Don't actually add -Wsign-compare. It will be covered by
-Wextra.

Switch to vpx_integer.h from df9c9d6d4c43f02c58d4e776c53323788e013cbc

BUG=webm:1069

Change-Id: I1dc6e61caa5d56af4a55b6692ab620bb3144652a

7 years agoAdd compiler warning flag -Wextra and fix related warnings.
Urvang Joshi [Thu, 14 Jul 2016 19:33:48 +0000 (12:33 -0700)]
Add compiler warning flag -Wextra and fix related warnings.

Note: some of these warnings are enabled by a combination of -Wunused
(added earlier) and -Wextra.

Cherry-picked from AOM 4790a69faaec8f03d65f64ff070f6ab4307dbb16

Expands use of (void)x; on unused variables. AOM only supports one codec
in codec_factory.h

Does not include changes to HandleDecodeResult. AOM removed
invalid_file_test.cc which does use the video parameter.

Does not enable -Wextra yet. There are more issues to fix.

BUG=webm:1069

Change-Id: I322a1366bd4fd6c0dec9e758c2d5e88e003b1cbf

7 years agoMerge "Limit max arf boost and scale motion breakout for image size."
Paul Wilkins [Tue, 27 Sep 2016 14:08:28 +0000 (14:08 +0000)]
Merge "Limit max arf boost and scale motion breakout for image size."

7 years agoFix for compile error with range checking
Peter de Rivaz [Tue, 27 Sep 2016 08:10:06 +0000 (09:10 +0100)]
Fix for compile error with range checking

Current version does not build with options:
  --enable-vp9-highbitdepth --enable-coefficient-range-checking

Change-Id: Ic3285f1a3e0d6be88da7f2cd8fa5a631368dd03b

7 years agoMerge "vp9: Reduce frame loopfilter-level for 1 pass cbr."
Marco Paniconi [Mon, 26 Sep 2016 22:05:43 +0000 (22:05 +0000)]
Merge "vp9: Reduce frame loopfilter-level for 1 pass cbr."

7 years agoMerge "Un-Revert "Restore vp8_sixtap_predict4x4_neon""
Johann Koenig [Mon, 26 Sep 2016 19:11:00 +0000 (19:11 +0000)]
Merge "Un-Revert "Restore vp8_sixtap_predict4x4_neon""

7 years agoMerge "Use shifted value for sinpi8sqrt2"
Johann Koenig [Mon, 26 Sep 2016 19:10:57 +0000 (19:10 +0000)]
Merge "Use shifted value for sinpi8sqrt2"

7 years agovp9: Reduce frame loopfilter-level for 1 pass cbr.
Marco [Mon, 26 Sep 2016 16:57:49 +0000 (09:57 -0700)]
vp9: Reduce frame loopfilter-level for 1 pass cbr.

Reduce the filt_guess for 1 pass cbr on inter-frames.
This reduces visual artifact seen in rtc clip (jimred.vga),
and improves metrics on rtc set.

Metrics on rtc set for cbr mode overall positive, most clips are up:
Speed 7 rtc: avgPSNR/SSIM up by: ~2.6/3.9%
Speed 8 rtc: avgPSNR/SSIM up by: ~1.3/2.5%

Change-Id: Ia4eccea1c19d65b583516df28823cd756c49464f

7 years agoMerge "Refactor lpf (size 4 and 8) NEON intrinsics optimization"
Linfeng Zhang [Mon, 26 Sep 2016 16:11:11 +0000 (16:11 +0000)]
Merge "Refactor lpf (size 4 and 8) NEON intrinsics optimization"

7 years agoLimit max arf boost and scale motion breakout for image size.
paulwilkins [Mon, 26 Sep 2016 14:35:23 +0000 (15:35 +0100)]
Limit max arf boost and scale motion breakout for image size.

Added a cap on the maximum boost for an arf based on interval length.
Fixed bug where by the image size was not accounted for in determining
two of the motion breakout thresholds.

Overall small gains of 0.2-0.4% psnr but on large image format clips with
slow zooms the gain may be as much as 20% or more (e.g. in_to_tree
at 1080P)

Change-Id: Id0a47391203026742daa9c97afac5705fd8c4dfb

7 years agoMerge "VP9: token decoder expansion"
Scott LaVarnway [Mon, 26 Sep 2016 12:06:50 +0000 (12:06 +0000)]
Merge "VP9: token decoder expansion"

7 years agoMerge "vp9_idct: delete dead TODOs"
James Zern [Sat, 24 Sep 2016 01:47:00 +0000 (01:47 +0000)]
Merge "vp9_idct: delete dead TODOs"

7 years agoUse shifted value for sinpi8sqrt2
Johann [Tue, 2 Aug 2016 22:59:35 +0000 (15:59 -0700)]
Use shifted value for sinpi8sqrt2

The value 35468 changes sign when stored in int16_t:
implicit conversion from 'int' to 'int16_t' (aka 'short')
changes value from 35468 to -30068

This negation requires adding back the original value to compensate.
Shifting the value keeps the value positive and saves a post-vqdmulh
shift.

This technique is used in webp and idct_dequant_full_2x_neon

BUG=b/28027557

Change-Id: I0c5ce09bea170fe08061856c2af6f841a557e0c3

7 years agoUn-Revert "Restore vp8_sixtap_predict4x4_neon"
Johann [Thu, 22 Sep 2016 21:48:58 +0000 (14:48 -0700)]
Un-Revert "Restore vp8_sixtap_predict4x4_neon"

This restores d9dce2f48eed1368a44c368fa87a506bd89ffec5

Switched to using signed shift-and-narrow. Instead of saturating
negative results to 0, it was saturating them to 255.

BUG=webm:817
BUG=webm:1273

Change-Id: I571095336aa4182e3288b17924fcaaece42b0a49

7 years agoVP9: token decoder expansion
Scott LaVarnway [Fri, 23 Sep 2016 12:03:52 +0000 (05:03 -0700)]
VP9: token decoder expansion

This version is based on Change 267683, but does not
use the macros.

Change-Id: I0619fa618decf8bdeef250584d75d70318b5d9a7

7 years agoMerge "VP9: pass TileWorkerData instead of MACROBLOCKD and vpx_reader."
Scott LaVarnway [Fri, 23 Sep 2016 11:59:15 +0000 (11:59 +0000)]
Merge "VP9: pass TileWorkerData instead of MACROBLOCKD and vpx_reader."

7 years agoMerge "vpx_idct32x32_34_add_sse2: rm unneeded transposes"
James Zern [Fri, 23 Sep 2016 02:49:26 +0000 (02:49 +0000)]
Merge "vpx_idct32x32_34_add_sse2: rm unneeded transposes"

7 years agoMerge "variance_neon: sync variance*() w/c,sse2"
James Zern [Fri, 23 Sep 2016 02:18:49 +0000 (02:18 +0000)]
Merge "variance_neon: sync variance*() w/c,sse2"

7 years agoVP9: pass TileWorkerData instead of MACROBLOCKD and vpx_reader.
Scott LaVarnway [Thu, 22 Sep 2016 20:18:36 +0000 (13:18 -0700)]
VP9: pass TileWorkerData instead of MACROBLOCKD and vpx_reader.

Change-Id: I869ef0f113c022143b531c44aefa0f1bb267052d

7 years agovpx_idct32x32_34_add_sse2: rm unneeded transposes
James Zern [Wed, 21 Sep 2016 03:22:08 +0000 (20:22 -0700)]
vpx_idct32x32_34_add_sse2: rm unneeded transposes

this change is neutral to mildly positive across various x86-64
platforms

Change-Id: I28fb5ae598fc1317b7a42c9a846ac5d57d104784

7 years agoMerge "Detect invalid highbd iht input"
Angie Chiang [Thu, 22 Sep 2016 01:06:37 +0000 (01:06 +0000)]
Merge "Detect invalid highbd iht input"

7 years agovariance_neon: sync variance*() w/c,sse2
James Zern [Thu, 22 Sep 2016 01:04:45 +0000 (18:04 -0700)]
variance_neon: sync variance*() w/c,sse2

removes some unnecessary casts and adds a few explicit uint32 ones for
larger sizes to quiet -Wshorten-64-to-32 warnings

Change-Id: I63c5fce8e62c426d5cf5c10a66a113c119a43518

7 years agoMerge "vp8: remove VP8_SET_DBG* control support"
James Zern [Thu, 22 Sep 2016 00:43:35 +0000 (00:43 +0000)]
Merge "vp8: remove VP8_SET_DBG* control support"

7 years agoDetect invalid highbd iht input
Angie Chiang [Tue, 20 Sep 2016 18:59:23 +0000 (11:59 -0700)]
Detect invalid highbd iht input

Do nothing in vp9_highbd_iht#x#_##_add_c when input magnitude is beyond
20 bits. Note that, sign bit is not included here.

In the 20 bits, we use 12 bits for input signal, 7 bits for forward
transform amplification, and 1 bit for contingency in rounding and
quantizing

BUG=https://bugs.chromium.org/p/webm/issues/detail?id=1286

Change-Id: I332c6f68df4614fc2e7d2dc4c5bb0d0cff8a245c

7 years agoKeep vp8 sixtap read within bounds
Johann [Wed, 21 Sep 2016 22:55:45 +0000 (15:55 -0700)]
Keep vp8 sixtap read within bounds

When filtering it needs 6 pixels: 2 prior to the source, the source, and
3 after the source.

When filtering 16 wide, that means 21. To accomplish this the SSE2 reads
[-2] to [5], [6] to [13], and [14] to [21], a total of 24 bytes (reading
in groups of 8 is easy)

The filter then shifts this last set to the top half of the register and
uses 'or' to combine it with the previous set.

Valgrind detected an issue reading pixels [19], [20] and [21]:
Address 0x7f581c2 is 434 bytes inside a block of size 441 alloc'd

Note: we only need pixels [16], [17], and [18] as context for [15].

To fix this, it now reads 8 bytes starting at [11], which re-loads [11]
through [13], but stops at [18] and does not over-read any values.

This is shifted by 5 and 'or'd with xmm1. Although the lower bits are
not cleared, they overlap directly with [11] through [13], so 'or'
produces the correct results.

Change-Id: I0c89c03afa660fc9b0108ac055d7bd403e493320

7 years agopredict_test: align dst buffer to 16
Johann [Wed, 21 Sep 2016 20:31:59 +0000 (13:31 -0700)]
predict_test: align dst buffer to 16

On 32 bit machines 'new' does not always appear to allocate sufficiently
aligned buffers, causing intermittent test failures.

Change-Id: I0db4fc73782012e4eef71dc0fb540e74fdbfcebe

7 years agovp8: remove VP8_SET_DBG* control support
James Zern [Wed, 7 Sep 2016 23:17:50 +0000 (16:17 -0700)]
vp8: remove VP8_SET_DBG* control support

the --enable-postproc-visualizer configure option remains as a no-op as
do the control names and values for compatibility
+ remove the corresponding debug flags from vpxdec: --pp-*

Change-Id: I4a001cd9962b59560d7d6bda6272d4ff32b8d37c

7 years agovp9_idct: delete dead TODOs
James Zern [Tue, 20 Sep 2016 03:34:22 +0000 (20:34 -0700)]
vp9_idct: delete dead TODOs

Change-Id: Icdd5494f557d83026dc078bce37997a76aa288fb

7 years agoMerge changes from topic 'Wshorten'
James Zern [Tue, 20 Sep 2016 23:17:20 +0000 (23:17 +0000)]
Merge changes from topic 'Wshorten'

* changes:
  vp8: convert some uses of unsigned long to size_t
  vp8/encoder: quiet some -Wshorten-64-to-32 warnings

7 years agoMerge "variance_avx2: sync variance functions with c-code"
James Zern [Tue, 20 Sep 2016 22:33:38 +0000 (22:33 +0000)]
Merge "variance_avx2: sync variance functions with c-code"

7 years agoMerge "examples: quiet -Wshorten-64-to-32 warnings"
James Zern [Tue, 20 Sep 2016 22:32:58 +0000 (22:32 +0000)]
Merge "examples: quiet -Wshorten-64-to-32 warnings"

7 years agoMerge "vp9_rtcd: remove non-existent highbd convolve fns"
James Zern [Tue, 20 Sep 2016 22:07:09 +0000 (22:07 +0000)]
Merge "vp9_rtcd: remove non-existent highbd convolve fns"

7 years agoMerge "Enable ssse3 bilinear tests"
Johann Koenig [Tue, 20 Sep 2016 21:46:50 +0000 (21:46 +0000)]
Merge "Enable ssse3 bilinear tests"

7 years agoMerge "Add vp8_bilinear_filter test"
Johann Koenig [Tue, 20 Sep 2016 20:30:48 +0000 (20:30 +0000)]
Merge "Add vp8_bilinear_filter test"

7 years agoMerge "Code class0 using vpx_read() / vpx_write()."
Alex Converse [Tue, 20 Sep 2016 19:19:28 +0000 (19:19 +0000)]
Merge "Code class0 using vpx_read() / vpx_write()."

7 years agovp9_rtcd: remove non-existent highbd convolve fns
James Zern [Tue, 20 Sep 2016 03:01:23 +0000 (20:01 -0700)]
vp9_rtcd: remove non-existent highbd convolve fns

these were moved to vpx_dsp

Change-Id: I307b07ae05e2333277d4b7011cba36dcf8409959

7 years agoexamples: quiet -Wshorten-64-to-32 warnings
James Zern [Tue, 20 Sep 2016 02:01:21 +0000 (19:01 -0700)]
examples: quiet -Wshorten-64-to-32 warnings

all around usage of strtol/strtoul

Change-Id: If907c89f107a068987aa71ddd93cee9a7389e4cd

7 years agovp8: convert some uses of unsigned long to size_t
James Zern [Tue, 20 Sep 2016 01:33:10 +0000 (18:33 -0700)]
vp8: convert some uses of unsigned long to size_t

similar to changes that were done in vp9 for encoded frame size
reporting. has the side-effect of quieting a -Wshorten-64-to-32 warning.

Change-Id: I89f74cb617fc29334ee351dc8dfaa3b8cfd4e5af

7 years agovp8/encoder: quiet some -Wshorten-64-to-32 warnings
James Zern [Tue, 20 Sep 2016 01:32:12 +0000 (18:32 -0700)]
vp8/encoder: quiet some -Wshorten-64-to-32 warnings

this code is similar to other existing uses and/or vp9

Change-Id: I56e646931379759d9f7332ea6d746060007c75ee

7 years agoRefactor lpf (size 4 and 8) NEON intrinsics optimization
Linfeng Zhang [Mon, 19 Sep 2016 18:37:16 +0000 (11:37 -0700)]
Refactor lpf (size 4 and 8) NEON intrinsics optimization

Also check in 8x8 8-bit transpose NEON intrinsics optimization
transpose_u8_8x8()

Change-Id: I32d321cf97ea21eab158ac4896990fc9a51681c4

7 years agovariance_avx2: sync variance functions with c-code
James Zern [Mon, 19 Sep 2016 23:17:24 +0000 (16:17 -0700)]
variance_avx2: sync variance functions with c-code

add missing int64 -> uint32 cast; quiets -Wshorten-64-to-32 warnings

Change-Id: I4850b36e18dc8b399108342be4bfe0b684aefb78

7 years agoMerge "Remove -fno-strict-aliasing flag"
Johann Koenig [Mon, 19 Sep 2016 22:49:22 +0000 (22:49 +0000)]
Merge "Remove -fno-strict-aliasing flag"

7 years agoRemove -fno-strict-aliasing flag
Johann [Sat, 27 Aug 2016 01:13:07 +0000 (18:13 -0700)]
Remove -fno-strict-aliasing flag

The referenced bug was fixed by saving neon registers. That this had any
effect was coincidental.

Both chromium and Android build with clang and neither uses this flag.

Change-Id: I470247d6fd9226fc207b42a187105581a94badc3

7 years agoCode class0 using vpx_read() / vpx_write().
Nathan E. Egge [Thu, 8 Sep 2016 21:25:49 +0000 (17:25 -0400)]
Code class0 using vpx_read() / vpx_write().

The vp9_mv_class0_tree is a balanced tree with two leafs and can
simply be coded as a boolean with probability class0[0].

Change-Id: If294dac825a5f945371092c74aa8e3f84cd962b6
(cherry picked from commit be8a8ab62ebdd111c6f2e9a33b15630570671eba)

7 years agoZero the whole rd_counts struct rather than the each member
Alex Converse [Mon, 19 Sep 2016 17:04:47 +0000 (10:04 -0700)]
Zero the whole rd_counts struct rather than the each member

Change-Id: I495aa9cec2b2b8f1ae69bdab8b3feeca76358472

7 years agoloopfilter_mb_neon: remove unused load_8x8()
James Zern [Sat, 17 Sep 2016 18:00:31 +0000 (11:00 -0700)]
loopfilter_mb_neon: remove unused load_8x8()

quiets a -Wunused-function warning for arm targets

Change-Id: I293a7e3d3d7d61d6af2fbedad5e8c25126c418b6

7 years agoMerge "Refactor lpf (size 16) NEON intrinsics optimization"
Linfeng Zhang [Sat, 17 Sep 2016 00:33:30 +0000 (00:33 +0000)]
Merge "Refactor lpf (size 16) NEON intrinsics optimization"

7 years agoMerge "vpx_codec_control: return incapable for unmatched control"
James Zern [Fri, 16 Sep 2016 17:30:44 +0000 (17:30 +0000)]
Merge "vpx_codec_control: return incapable for unmatched control"

7 years agoRefactor lpf (size 16) NEON intrinsics optimization
Linfeng Zhang [Thu, 15 Sep 2016 23:49:47 +0000 (16:49 -0700)]
Refactor lpf (size 16) NEON intrinsics optimization

Extract shared code so later lpf size 4 and 8 functions can reuse.

Change-Id: Ibb43ef1fd8651bd2e32fcc4c56cf6fa7ca237401

7 years agovpx_subpixel_8t_intrin_avx2: tolerate unversioned clang
James Zern [Fri, 16 Sep 2016 03:40:21 +0000 (20:40 -0700)]
vpx_subpixel_8t_intrin_avx2: tolerate unversioned clang

assume __clang_major__==0 has the latest version of
_mm256_broadcastsi128_si256. fixes builds with custom clang toolchains.

BUG=b/30970831

Change-Id: I90becd56278e4716bd46e2ba9d910af977e8dfa6

7 years agoMerge changes from topic 'clang-format'
James Zern [Fri, 16 Sep 2016 07:11:33 +0000 (07:11 +0000)]
Merge changes from topic 'clang-format'

* changes:
  apply clang-format
  .clang-format: update to 3.8.1

7 years agoEnable ssse3 bilinear tests
Johann [Fri, 9 Sep 2016 01:46:14 +0000 (18:46 -0700)]
Enable ssse3 bilinear tests

The code only has issues when xoffset == 0 and yoffset == 0 which
represents a simple copy. Presumably this case does not need to be
handled because the issue has existed since 2010.

BUG=webm:1287

Change-Id: Ic47e2653f3b729e99b40e53d8d2d8d1501edaaa9

7 years agoAdd vp8_bilinear_filter test
Johann [Sat, 3 Sep 2016 01:42:40 +0000 (18:42 -0700)]
Add vp8_bilinear_filter test

Build out the sixtap_predict test because the filters are
interchangeable. Add verbose failures and border checking.

Change-Id: I962f50041750dca6f8d0cd35a943424cf82ddcb1

7 years agoMerge "Revert "Restore vp8_sixtap_predict4x4_neon""
James Zern [Fri, 16 Sep 2016 06:13:41 +0000 (06:13 +0000)]
Merge "Revert "Restore vp8_sixtap_predict4x4_neon""

7 years agoRevert "Restore vp8_sixtap_predict4x4_neon"
Johann Koenig [Fri, 16 Sep 2016 06:12:49 +0000 (06:12 +0000)]
Revert "Restore vp8_sixtap_predict4x4_neon"

This reverts commit d9dce2f48eed1368a44c368fa87a506bd89ffec5.

Appears to be failing the SixtapPredict tests in some configurations and possibly test vectors as well.

Change-Id: Ica6aa83ebac47d0a76e451846e7da67b1c17a7d7

7 years agoMerge "Restore vp8_bilinear_predict4x4_neon"
Johann Koenig [Fri, 16 Sep 2016 05:33:50 +0000 (05:33 +0000)]
Merge "Restore vp8_bilinear_predict4x4_neon"

7 years agoMerge "zero structures completely"
Johann Koenig [Fri, 16 Sep 2016 04:41:22 +0000 (04:41 +0000)]
Merge "zero structures completely"

7 years agoRestore vp8_bilinear_predict4x4_neon
Johann [Fri, 2 Sep 2016 22:21:35 +0000 (15:21 -0700)]
Restore vp8_bilinear_predict4x4_neon

This function was removed when clang started introducing alignment hints
which caused the 32 bit vld1_lane_u32/vst1_lane_u32 to fail:
https://llvm.org/bugs/show_bug.cgi?id=24421

The load has been rendered safe with an implementation ~indiscernible
performance-wise that uses _u8 and over-reads just a touch.

It is still ~5x faster than C in the unaligned case and doing both
filters.

BUG=webm:892
BUG=webm:1273

Change-Id: Icf7167189391b46202f47233bb585c24c42bcc36

7 years agoMerge "Restore vp8_sixtap_predict4x4_neon"
Johann Koenig [Fri, 16 Sep 2016 04:12:07 +0000 (04:12 +0000)]
Merge "Restore vp8_sixtap_predict4x4_neon"

7 years agozero structures completely
Johann [Fri, 16 Sep 2016 00:44:24 +0000 (17:44 -0700)]
zero structures completely

Use vp[89]_zero when possible.

Expand the {} set when neither is available or nearby.

Change-Id: Ifc1f46f60100916cd798bf7be3a10f09321c99bd

7 years agovp8 postproc: expand CONFIG_POSTPROC guard
Johann [Thu, 15 Sep 2016 22:59:42 +0000 (15:59 -0700)]
vp8 postproc: expand CONFIG_POSTPROC guard

postproc.c is overloaded and used for both postproc and internal stats.
If only --enable-internal-stats is specified there are issues with
non-existent struct members and unused functions.

Change-Id: I82367f1ffce659c3918c9f964dbce94a716fbb89

7 years agoaltref test: comment out 'pass'
Johann [Tue, 13 Sep 2016 21:22:26 +0000 (14:22 -0700)]
altref test: comment out 'pass'

All the other test which do not use 'pass' (which appears to be almost
all of them) do this.

Cleans -Wextra/-Wunused-parameter:
unused parameter ‘pass’

Change-Id: I1ff3acf3f3d1e831f94dcb00ea36337afe0aefe0

7 years agoMerge "vp9 frame parallel test: Initialize cfg differently"
Johann Koenig [Thu, 15 Sep 2016 23:46:56 +0000 (23:46 +0000)]
Merge "vp9 frame parallel test: Initialize cfg differently"

7 years agovp9: Small code cleanup.
Marco [Thu, 15 Sep 2016 22:14:46 +0000 (15:14 -0700)]
vp9: Small code cleanup.

Remove the experiment LIMIT_QP_ONEPASS_VBR_LAG, as its
not currently used and no plan to use in near future.

Change-Id: Ib069f8d7225195be04b765d0ab477510dfba6a3b

7 years agoapply clang-format
clang-format [Thu, 15 Sep 2016 22:07:46 +0000 (15:07 -0700)]
apply clang-format

Change-Id: I501597b7c1e0f0c7ae2aea3ee8073f0a641b3487

7 years ago.clang-format: update to 3.8.1
James Zern [Thu, 8 Sep 2016 03:56:01 +0000 (20:56 -0700)]
.clang-format: update to 3.8.1

based on --style=Google with the following differences:
3a4
> # Generated with clang-format 3.8.1
13c14
< AllowShortCaseLabelsOnASingleLine: false
---
> AllowShortCaseLabelsOnASingleLine: true
41c42
< ConstructorInitializerAllOnOneLineOrOnePerLine: true
---
> ConstructorInitializerAllOnOneLineOrOnePerLine: false
44,45c45,46
< Cpp11BracedListStyle: true
< DerivePointerAlignment: true
---
> Cpp11BracedListStyle: false
> DerivePointerAlignment: false
73c74
< PointerAlignment: Left
---
> PointerAlignment: Right
75c76
< SortIncludes:    true
---
> SortIncludes:    false

SortIncludes will like be enabled in a future commit

Change-Id: I5c404f44081b65354e7f526411c91fbbe31ac5af

7 years agoRestore vp8_sixtap_predict4x4_neon
Johann [Fri, 2 Sep 2016 18:29:20 +0000 (11:29 -0700)]
Restore vp8_sixtap_predict4x4_neon

This function was removed when clang started introducing alignment hints
which caused the 32 bit vld1_lane_u32/vst1_lane_u32 to fail:
https://llvm.org/bugs/show_bug.cgi?id=24421

The load has been rendered safe with an implementation ~indiscernible
performance-wise that uses _u8 and over-reads just a touch.

The store, when unaligned, has a version that is ~25% slower but safe
when xoffset = 0 (second pass filter only). When the first pass filter
(or both) are in play, the new version is almost identical in speed.

Worst case performance (both filters, unaligned stores) is roughly 3-4x
faster than C.

BUG=webm:817
BUG=webm:1273

Change-Id: I1e490e94453e0872151fe0dafb05557463f6247d

7 years agovp9 frame parallel test: Initialize cfg differently
Johann [Thu, 15 Sep 2016 19:19:25 +0000 (12:19 -0700)]
vp9 frame parallel test: Initialize cfg differently

Use the canonical 'vpx_codec_dec_cfg_t()' as opposed to 'vp9_zero()'
which just hammered everything to 0.

Change-Id: Id820efef700ad92a625797f8fd58e465b15eeca4

7 years agoMerge "Documentation for building unit tests for Android"
Johann Koenig [Thu, 15 Sep 2016 19:17:13 +0000 (19:17 +0000)]
Merge "Documentation for building unit tests for Android"

7 years agoDocumentation for building unit tests for Android
Johann [Wed, 7 Sep 2016 22:18:40 +0000 (15:18 -0700)]
Documentation for building unit tests for Android

BUG=webm:1258

Change-Id: Iea142f7b0df0e047720e8c5362464932de57d564

7 years agoMerge "cosmetics,vp8: join some lines, fix table format"
James Zern [Wed, 14 Sep 2016 00:41:51 +0000 (00:41 +0000)]
Merge "cosmetics,vp8: join some lines, fix table format"

7 years agovp8 decoder: cast decoding_thread_count to int
Johann [Tue, 13 Sep 2016 21:20:19 +0000 (14:20 -0700)]
vp8 decoder: cast decoding_thread_count to int

For some reason allocated_decoding_thread_count is signed, but decoding_thread_count is not.

Cleans -Wextra/-Wsign-compare:
comparison between signed and unsigned integer expressions

Change-Id: Id0ada78100acff27c1c4ed7493c563d13c55cdcd

7 years agovp9 frame parallel test: Initialize cfg to 0
Johann [Tue, 13 Sep 2016 21:12:03 +0000 (14:12 -0700)]
vp9 frame parallel test: Initialize cfg to 0

Use vp9_zero() to set every element.

Cleans -Wextra/-Wmissing-field-initializers:
missing initializer for member ‘vpx_codec_dec_cfg::w’
missing initializer for member ‘vpx_codec_dec_cfg::h’

Change-Id: I5b41ce7d55a912e29b1d4c3e840cea80e8510fbe

7 years agovp9cx_set_ref.c: remove unused 'cfg' parameter
Johann [Tue, 13 Sep 2016 21:05:03 +0000 (14:05 -0700)]
vp9cx_set_ref.c: remove unused 'cfg' parameter

Cleans -Wextra/-Wunused-parameter warning:
warning: unused parameter ‘cfg’

Change-Id: I84eae57a50306cb66c625bb648b0a330678818db

7 years agowebmenc: remove unused 'fps' parameter
Johann [Tue, 13 Sep 2016 21:03:23 +0000 (14:03 -0700)]
webmenc: remove unused 'fps' parameter

Cleans -Wextra/-Wunused-parameter warning:
warning: unused parameter ‘fps’

Change-Id: Ia5f9338f11ae8d0708a87c6d4e7d7e924fc3b19b

7 years agovpx_codec_control: return incapable for unmatched control
James Zern [Sat, 10 Sep 2016 00:37:13 +0000 (17:37 -0700)]
vpx_codec_control: return incapable for unmatched control

VPX_CODEC_INCAPABLE rather than the more generic VPX_CODEC_ERROR

Change-Id: Id1ed7fb23a2910192713c6b2389c0b7320201f52

7 years agocosmetics,vp8: join some lines, fix table format
James Zern [Fri, 9 Sep 2016 23:39:34 +0000 (16:39 -0700)]
cosmetics,vp8: join some lines, fix table format

Change-Id: Idcf3b68f0e59bd74c9d332bbd4a7c1484ddb691a

7 years agovp8: Set the skin model to mode 1.
Marco [Fri, 9 Sep 2016 16:03:22 +0000 (09:03 -0700)]
vp8: Set the skin model to mode 1.

This change was reverted before due to a hangouts encode-time
regression investigation. But since then this change has been
cleared of causing any noticeable regression.

This mode reduces some false detection, and uses the
same model as in vp9.

Change-Id: I9c82a748c5f601d0aca9f61ee218abfbd58c62bd

7 years agoMerge "vp8: Remove TSAN warning around end of encode."
James Zern [Fri, 9 Sep 2016 03:08:18 +0000 (03:08 +0000)]
Merge "vp8: Remove TSAN warning around end of encode."

7 years agovp8: Remove TSAN warning around end of encode.
Alexander Potapenko [Tue, 6 Sep 2016 16:50:38 +0000 (18:50 +0200)]
vp8: Remove TSAN warning around end of encode.

Tsan warns when run in one pass and there is a recode
loop.

Change-Id: Ice2ecb2270f09ebd49efbd49c0e4f77d32e23c0f

7 years agoMerge "vpx_dsp: added vpx_highbd_idct32x32_1_add_sse2()"
James Zern [Thu, 8 Sep 2016 01:05:17 +0000 (01:05 +0000)]
Merge "vpx_dsp: added vpx_highbd_idct32x32_1_add_sse2()"

7 years agovpx_mem.c: remove unnecessary inline
James Zern [Wed, 7 Sep 2016 17:51:33 +0000 (10:51 -0700)]
vpx_mem.c: remove unnecessary inline

these aren't overly speed critical, best to leave it to the compiler.

Change-Id: I231c14abee5b845d7b8e8454832f2feb22c6ce45