Commits · 2d601aaf3482720de65cc6d60ec25eb6a4afa46e · submodule / libyuv

08 Oct, 2015 5 commits

merge neon source files back into single libyuv library · 2d601aaf

Frank Barchard authored Oct 08, 2015

previously the neon source code was broken into a separate
library built with -mfpu=neon for the neon assembly, while
the C code was built without neon.

In this change, the neon code is added to the main library
and all code built with neon.

TBR=harryjin@google.com
BUG=libyuv:371

Review URL: https://codereview.chromium.org/1392043003 .

2d601aaf

fix jpeg and bt.709 yuvconstants for neon64. · 76a599ec

Frank Barchard authored Oct 08, 2015

yuv constants for bt.601 were previously ported to neon64, as well
as the code to respect other color spaces.  But the jpeg and bt.709
colour conversion constants were still in armv7 form.  This changes
the constants for aarch64 builds to be compatible with the code.

yuv constants are now passed as const *

Remove Yvu constants which were used for older version on nv21 but not new code.

TBR=harryjin@google.com
BUG=none

Review URL: https://codereview.chromium.org/1398623002 .

76a599ec

Fix for AVX2 dither function. · fae8e66d

Frank Barchard authored Oct 08, 2015

Fix for 64 bit gcc parameter in dither function which requires m not r,
when ABI uses register.

BUG=none

Review URL: https://codereview.chromium.org/1399463002 .

fae8e66d

port ARGB to 565 dithering AVX2 code to GCC. · 8f0caded

Frank Barchard authored Oct 08, 2015

Previously the assembly code was only available to Windows.
This CL ports the AVX2 code to GCC syntax.

TBR=harryjin@google.com
BUG=libyuv:492

Review URL: https://codereview.chromium.org/1391273003 .

8f0caded

port ARGB to 565 dithering SSE2 code to GCC. · cc89e3a7

Frank Barchard authored Oct 08, 2015

Previously the assembly code was only available to Windows.
This CL ports the SSE2 code to GCC syntax.

When running a profiler on all the unittests, this function
was the slowest of all functions that still ran in C code.
   3.71%  libyuv_unittest  libyuv_unittest      [.] ARGBToRGB565DitherRow_C

Was
ARGBToRGB565Dither_Opt (2894 ms)
Now
ARGBToRGB565Dither_Opt (432 ms)

TBR=harryjin@google.com
BUG=libyuv:492

Review URL: https://codereview.chromium.org/1397673002 .

cc89e3a7

07 Oct, 2015 5 commits

fix avx2 box filter bug for yuv down sampling. · 3e38762d

Frank Barchard authored Oct 07, 2015

offset to second group of pixels was off by 16.
should have been 32, not 16.
requires avx2 hardware and wide image for test.

R=harryjin@google.com
TBR=harryjin@google.com
BUG=libyuv:492,libyuv:501

Review URL: https://codereview.chromium.org/1395603002 .

3e38762d

Pass yuvconstants to YUV conversions for neon 64 bit · 013080f2

Frank Barchard authored Oct 07, 2015

SETUP provided by zhongwei.yao@linaro.org

Previously the 64 bit Neon code had hard coded constants in the setup macro
for YUV conversion, while 32 bit Neon code supported the yuvconstants
parameter.

This change accepts the constants passed to the YUV conversion row function,
allowing different color spaces to be respected - naming JPEG and BT.709.
As well as the existing BT.601.

TBR=harryjin@google.com
BUG=libyuv:472

Review URL: https://codereview.chromium.org/1384323002 .

013080f2

Reimplement NV21ToARGB to allow different color matrix. · 914a9856

Frank Barchard authored Oct 07, 2015

Low level for NV21ToARGB written to accept yuv matrix used by
other YUV to ARGB functions.
Previously NV21 was implemented for Windows using NV12 with a different
matrix that swapped U and V.  But the Arm version of the low level does
not allow the matrix U and V contributions to be swapped.
Using a new low level function that reads NV21 and uses the same
yuvconstants as other YUV conversion functions allows an Arm port of
this function.

TBR=harryjin@google.com
BUG=libyuv:500

Review URL: https://codereview.chromium.org/1388273002 .

914a9856

add box scaling avx2 optimization for gcc · 68fa59c8

Frank Barchard authored Oct 07, 2015

TBR=harryjin@google.com
BUG=libyuv:492

Review URL: https://codereview.chromium.org/1392803002 .

68fa59c8

Add J444ToARGB conversion function. · f00bc9ef

Frank Barchard authored Oct 07, 2015

J444 is JPeg YUV color space with 444 subsampling.
This implementation uses the existing I444ToARGB conversion, which is
BT.601 color space with 444 subsampling, but passing in the jpeg
color matrix constants.

TBR=harryjin@google.com
BUG=449

Review URL: https://codereview.chromium.org/1387313002 .

f00bc9ef

06 Oct, 2015 1 commit

port scale box filter sse2 to gcc · d7029399

Frank Barchard authored Oct 06, 2015

TBR=harryjin@google.com
BUG=libyuv:492

Review URL: https://codereview.chromium.org/1393653002 .

d7029399

02 Oct, 2015 1 commit

Speed up rounding to byte test · f4c1ac10

Frank Barchard authored Oct 02, 2015

R=harryjin@google.com
BUG=libyuv:492

Review URL: https://codereview.chromium.org/1367403007 .

f4c1ac10

01 Oct, 2015 1 commit

test xsave before calling xgetbv. · 3eefeaeb

Frank Barchard authored Oct 01, 2015

R=agl@chromium.org, harryjin@google.com
BUG=libyuv:497

Review URL: https://codereview.chromium.org/1382803002 .

3eefeaeb

30 Sep, 2015 1 commit

Remove sse2 functions that also have ssse3 · 2cc1a2b2

Frank Barchard authored Sep 30, 2015

ARGBBlendRow_SSE2, ARGBAttenuateRow_SSE2, and MirrorRow_SSE2
Since vast majority of CPUs have SSSE3 now, removing the SSE2
improves the performance of CPU dispatching.

R=harryjin@google.com
BUG=none

Review URL: https://codereview.chromium.org/1377053003 .

2cc1a2b2

25 Sep, 2015 5 commits

Width use memory instead of register for 32 bit fpic. · d039ad6e

Frank Barchard authored Sep 25, 2015

Code runs out of registers on 32 bit fpic builts.

TBR=harryjin@google.com
BUG=libyuv:496

Review URL: https://codereview.chromium.org/1369053002 .

d039ad6e

win64 version of I422AlphaToARGB. · febc26a2

Frank Barchard authored Sep 25, 2015

Was
I420AlphaToARGB_Premult (8861 ms)
I420AlphaToARGB_Opt (7119 ms)
Now
I420AlphaToABGR_Premult (2840 ms)
I420AlphaToARGB_Opt (484 ms)

C function switched to 1 step.
Was
I420AlphaToARGB_Premult (8862 ms)
I420AlphaToABGR_Opt (6718 ms)

Now
I420AlphaToARGB_Premult (8706 ms)
I420AlphaToARGB_Opt (6541 ms)

R=harryjin@google.com
BUG=libyuv:496, libyuv:473

Review URL: https://codereview.chromium.org/1359183003 .

febc26a2

AVX2 1 step I422AlphaToARGB for gcc and win. · 9a0e12f5

Frank Barchard authored Sep 25, 2015

C     I420AlphaToARGB_Opt (5169 ms)
SSSE3 I420AlphaToARGB_Opt (432 ms)
AVX2  I420AlphaToARGB_Opt (358 ms)

and with premultiplication as 2 step process:
I420AlphaToARGB_Premult (7029 ms)
I420AlphaToARGB_Premult (757 ms)
I420AlphaToARGB_Premult (508 ms)

R=harryjin@google.com
BUG=libyuv:496,libyuv:473

Review URL: https://codereview.chromium.org/1372653003 .

9a0e12f5

I420Alpha row function in 1 pass. · e365cdde

Frank Barchard authored Sep 25, 2015

API change - I420AlphaToARGB takes flag indicating if RGB should be
premultiplied by alpha.

This version implements an efficient SSSE3 version for Windows.
C version done in 2 steps.

Was
libyuvTest.I420AlphaToARGB_Any (1136 ms)
libyuvTest.I420AlphaToARGB_Unaligned (1210 ms)
libyuvTest.I420AlphaToARGB_Invert (966 ms)
libyuvTest.I420AlphaToARGB_Opt (1031 ms)
libyuvTest.I420AlphaToABGR_Any (1020 ms)
libyuvTest.I420AlphaToABGR_Unaligned (1359 ms)
libyuvTest.I420AlphaToABGR_Invert (1082 ms)
libyuvTest.I420AlphaToABGR_Opt (986 ms)

R=harryjin@google.com
BUG=libyuv:496

Review URL: https://codereview.chromium.org/1367093002 .

e365cdde

switch from ebp to ebx. · d4594bee

Frank Barchard authored Sep 25, 2015

ebx encodes more efficiently (1 byte less) for most address modes, than ebp.
previously it was used for 411 format, but the reader uses pinsrw now avoiding
gpr register.

BUG=libyuv:488
R=harryjin@google.com

Review URL: https://codereview.chromium.org/1365003003 .

d4594bee

24 Sep, 2015 1 commit

Fix nv12 64 bit gcc increment. · 8fb2048e

Frank Barchard authored Sep 24, 2015

Should be 16 bytes, but was 0x16 causing memory corruption.

TBR=harryjin@google.com
BUG=libyuv:492

Review URL: https://codereview.chromium.org/1368693002 .

8fb2048e

23 Sep, 2015 2 commits

NV12ToARGB_AVX2 ported to gcc · accc04e6

Frank Barchard authored Sep 23, 2015

TBR=harryjin@google.com
BUG=none

Review URL: https://codereview.chromium.org/1364913002 .

accc04e6

YUY2ToARGB avx2 in 1 step conversion. · 000cf89c

Frank Barchard authored Sep 23, 2015

Includes UYVYToARGB ssse3 fix.

Was
YUY2ToARGB_Opt (433 ms)
69.79%  libyuv_unittest  libyuv_unittest      [.] I422ToARGBRow_AVX2
20.73%  libyuv_unittest  libyuv_unittest      [.] YUY2ToUV422Row_AVX2
 6.04%  libyuv_unittest  libyuv_unittest      [.] YUY2ToYRow_AVX2
 0.77%  libyuv_unittest  libyuv_unittest      [.] YUY2ToARGBRow_AVX2

Now
YUY2ToARGB_Opt (280 ms)
95.66%  libyuv_unittest  libyuv_unittest      [.] YUY2ToARGBRow_AVX2

BUG=libyuv:494
R=harryjin@google.com

Review URL: https://codereview.chromium.org/1364813002 .

000cf89c

22 Sep, 2015 5 commits

Replace random with fastrand · 16f12b58

Frank Barchard authored Sep 22, 2015

random / rand is slow and impacts performance testing.
Although its only called to clear a frame once, a typical profile shows
it high in the overall profile, when doing 1000 frames for a benchmark.

95.10%  libyuv_unittest  libyuv_unittest      [.] YUY2ToARGBRow_SSSE3
 2.01%  libyuv_unittest  libc-2.19.so         [.] __random_r
 1.13%  libyuv_unittest  libc-2.19.so         [.] __random

Replace random is a faster version for unittests.

set LIBYUV_WIDTH=1280
set LIBYUV_HEIGHT=720
set LIBYUV_REPEAT=999
set LIBYUV_FLAGS=-1
out\release\libyuv_unittest --gtest_filter=*YUY2ToARGB*  | findms

Was
libyuvTest.YUY2ToARGB_Opt (497 ms)

Now
libyuvTest.YUY2ToARGB_Opt (454 ms)

R=harryjin@google.com
BUG=none

Review URL: https://codereview.chromium.org/1361813002 .

16f12b58

Fix git markers introduced on landing previous CL · 2b92ec8d
Frank Barchard authored Sep 22, 2015
```
BUG=none

Review URL: https://codereview.chromium.org/1359023003 .
```
2b92ec8d

yuy2 to rgb gcc versions · 5f3d4270

Frank Barchard authored Sep 22, 2015

read in read function for yuv conversion

R=harryjin@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1355393002 .

5f3d4270

Read Y channel in read function for yuv conversion. · 03cd8584

Frank Barchard authored Sep 22, 2015

Allows reader to support YUY2 format.
Also contains fix for win64 build for yuv conversion.

TBR=harryjin@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1355333002 .

03cd8584

yuvconstants for all YUV to RGB conversion functions. · f96890a0
Frank Barchard authored Sep 22, 2015
```
R=harryjin@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1363503002 .
```
f96890a0

18 Sep, 2015 3 commits

move constants into common · 62c49dc8

Frank Barchard authored Sep 18, 2015

R=harryjin@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1359443005 .

62c49dc8

port I444 to ARGB to matrix. Add I444 to ABGR. · 0381673d

Frank Barchard authored Sep 18, 2015

R=harryjin@google.com
BUG=libyuv:488,libyuv:490

Review URL: https://codereview.chromium.org/1348763005 .

0381673d

I444ToABGR for android · 28427a53

Frank Barchard authored Sep 18, 2015

Reimplements I444ToARGB as a matrix function.
new I444ToABGR as matrix functions with wrappers and any functions.
Allows for future J444 and H444 versions.
I444ToABGR user level function added.

BUG=libyuv:490, libyuv:449
R=harryjin@google.com

Review URL: https://codereview.chromium.org/1355733002 .

28427a53

17 Sep, 2015 6 commits

NEON J422ToABGR and H422ToABGR missing prototypes · 158d4079
Frank Barchard authored Sep 17, 2015
```
TBR=harryjin@google.com
BUG=none

Review URL: https://codereview.chromium.org/1351993003 .
```
158d4079

NEON constants · bdfd59a7

Frank Barchard authored Sep 17, 2015

TBR=harryjin@google.com
BUG=none

Review URL: https://codereview.chromium.org/1351553005 .

bdfd59a7

j422toabgr neon port using i422toabgr matrix function. · 28ce7d94
Frank Barchard authored Sep 17, 2015
```
R=harryjin@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1353923003 .
```
28ce7d94

J422ToARGB Neon but not aarch64 · 6fcbae14

Frank Barchard authored Sep 17, 2015

TBR=harryjin@google.com
BUG=libyuv:493

Review URL: https://codereview.chromium.org/1348203004 .

6fcbae14

Add H422ToARGB armv7 neon version. · 6a6b67e7

Frank Barchard authored Sep 17, 2015

Patch provided by zhongwei.yao@linaro.org

R=fbarchard@chromium.org, fbarchard@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1344393002 .

6a6b67e7

Remove redundent rounding test from scale. · ef09d53f

Frank Barchard authored Sep 17, 2015

There is still a rounding test in color test.

TBR=harryjin@google.com
BUG=libyuv:492

Review URL: https://codereview.chromium.org/1347893003 .

ef09d53f

15 Sep, 2015 2 commits

j422 not available on aarch64 · bb0a521c

Frank Barchard authored Sep 15, 2015

The aarch64 version does not have I422ToARGBMatrix yet,
so adding this to the ifdef section of row.h

R=harryjin@google.com
TBR=harryjin@google.com, zhongwei.yao@linaro.org
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1347853002 .

bb0a521c

Add J422ToARGB armv7 neon version. · 509c6442

Frank Barchard authored Sep 15, 2015

R=fbarchard@chromium.org, fbarchard@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1334173005 .

509c6442

14 Sep, 2015 2 commits

neon64 use yuvconstants like 32 bit code. · 73c32d92

Frank Barchard authored Sep 14, 2015

TBR=harryjin@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1345643002 .

73c32d92

use struct instead of vectors · a67927c9

Frank Barchard authored Sep 14, 2015

TBR=harryjin@google.com
BUG=libyuv:488

Review URL: https://codereview.chromium.org/1345623003 .

a67927c9