coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-06 16:56:55 +08:00
Code Issues Packages Projects Releases Wiki Activity
1,857 Commits 3 Branches 0 Tags BSD-3-Clause 60 MiB
Commit Graph

24 Commits

Author SHA1 Message Date
Frank Barchard
6c94ad13b5 Remove ARM NaCL macros from source 
NaCL has been disabled for awhile, so the code
will still build, but only with C versions.
This change removes the MEMACCESS() macros from
Neon and Neon64 source.

BUG=libyuv:702
TEST=try bots build for arm.
R=kjellander@chromium.org

Change-Id: Id581a5c8ff71e18cc69595e7fee9337f97c44a19
Reviewed-on: https://chromium-review.googlesource.com/528332
Reviewed-by: Cheng Wang <wangcheng@google.com>
Commit-Queue: Frank Barchard <fbarchard@google.com>
 2017-06-09 22:22:07 +00:00
Frank Barchard
44abf70187 ScaleDown odd functions adjust math so last pixel is half width source. 
existing test passes
out/Release/libyuv_unittest --gtest_filter=*Blend* --libyuv_width=33 --libyuv_height=16

new test added
BUG=libyuv:705
TEST=LibYUVScaleTest.TestScaleOdd

Change-Id: Ica91812aee2e4ed9bcc18df4962b089c2e4ae704
Reviewed-on: https://chromium-review.googlesource.com/524932
Reviewed-by: Cheng Wang <wangcheng@google.com>
Commit-Queue: Frank Barchard <fbarchard@google.com>
 2017-06-06 01:37:26 +00:00
Frank Barchard
000d2fa91a Libyuv MIPS DSPR2 optimizations. 
Optimized functions:

I444ToARGBRow_DSPR2
I422ToARGB4444Row_DSPR2
I422ToARGB1555Row_DSPR2
NV12ToARGBRow_DSPR2
BGRAToUVRow_DSPR2
BGRAToYRow_DSPR2
ABGRToUVRow_DSPR2
ARGBToYRow_DSPR2
ABGRToYRow_DSPR2
RGBAToUVRow_DSPR2
RGBAToYRow_DSPR2
ARGBToUVRow_DSPR2
RGB24ToARGBRow_DSPR2
RAWToARGBRow_DSPR2
RGB565ToARGBRow_DSPR2
ARGB1555ToARGBRow_DSPR2
ARGB4444ToARGBRow_DSPR2
ScaleAddRow_DSPR2

Bug-fixes in functions:

ScaleRowDown2_DSPR2
ScaleRowDown4_DSPR2

BUG=

Review-Url: https://codereview.chromium.org/2626123003 .
 2017-01-11 12:19:13 -08:00
Manojkumar Bhosale
288bfbefb5 Add MSA optimized remaining scale row functions 
R=fbarchard@google.com
BUG=libyuv:634

Performance Gain (vs C vectorized)
ScaleRowDown2_MSA            - ~22.3x
ScaleRowDown2_Any_MSA        - ~19.9x
ScaleRowDown2Linear_MSA      - ~31.2x
ScaleRowDown2Linear_Any_MSA  - ~29.4x
ScaleRowDown2Box_MSA         - ~20.1x
ScaleRowDown2Box_Any_MSA     - ~19.6x
ScaleRowDown4_MSA            - ~11.7x
ScaleRowDown4_Any_MSA        - ~11.2x
ScaleRowDown4Box_MSA         - ~15.1x
ScaleRowDown4Box_Any_MSA     - ~15.1x
ScaleRowDown38_MSA           - ~1x
ScaleRowDown38_Any_MSA       - ~1x
ScaleRowDown38_2_Box_MSA     - ~1.7x
ScaleRowDown38_2_Box_Any_MSA - ~1.7x
ScaleRowDown38_3_Box_MSA     - ~1.7x
ScaleRowDown38_3_Box_Any_MSA - ~1.7x
ScaleAddRow_MSA              - ~1.2x
ScaleAddRow_Any_MSA          - ~1.15x

Performance Gain (vs C non-vectorized)
ScaleRowDown2_MSA            - ~22.4x
ScaleRowDown2_Any_MSA        - ~19.8x
ScaleRowDown2Linear_MSA      - ~31.6x
ScaleRowDown2Linear_Any_MSA  - ~29.4x
ScaleRowDown2Box_MSA         - ~20.1x
ScaleRowDown2Box_Any_MSA     - ~19.6x
ScaleRowDown4_MSA            - ~11.7x
ScaleRowDown4_Any_MSA        - ~11.2x
ScaleRowDown4Box_MSA         - ~15.1x
ScaleRowDown4Box_Any_MSA     - ~15.1x
ScaleRowDown38_MSA           - ~3.2x
ScaleRowDown38_Any_MSA       - ~3.2x
ScaleRowDown38_2_Box_MSA     - ~2.4x
ScaleRowDown38_2_Box_Any_MSA - ~2.3x
ScaleRowDown38_3_Box_MSA     - ~2.9x
ScaleRowDown38_3_Box_Any_MSA - ~2.8x
ScaleAddRow_MSA              - ~8x
ScaleAddRow_Any_MSA          - ~7.46x

Review-Url: https://codereview.chromium.org/2559683002 .
 2016-12-21 13:39:44 +05:30
Manojkumar Bhosale
56b5bbb0be Add MSA optimized ARGB scaling functions 
R=fbarchard@google.com
BUG=libyuv:634

Performance Gain (vs C vectorized)
ScaleARGBRowDown2_MSA           - ~2.6x
ScaleARGBRowDown2Linear_MSA     - ~7.9x
ScaleARGBRowDown2Box_MSA        - ~3.7x
ScaleARGBRowDownEven_MSA        - ~1.2x
ScaleARGBRowDownEvenBox_MSA     - ~3.5x

ScaleARGBRowDown2_Any_MSA       - ~2.6x
ScaleARGBRowDown2Linear_Any_MSA - ~7.9x
ScaleARGBRowDown2Box_Any_MSA    - ~3.6x
ScaleARGBRowDownEven_Any_MSA    - ~1.2x
ScaleARGBRowDownEvenBox_Any_MSA - ~3.5x

Performance Gain (vs C non-vectorized)
ScaleARGBRowDown2_MSA           - 2.6x
ScaleARGBRowDown2Linear_MSA     - 13.5x
ScaleARGBRowDown2Box_MSA        - 5.8x
ScaleARGBRowDownEven_MSA        - 1.2x
ScaleARGBRowDownEvenBox_MSA     - 3.7x

ScaleARGBRowDown2_Any_MSA       - 2.6x
ScaleARGBRowDown2Linear_Any_MSA - 13.5x
ScaleARGBRowDown2Box_Any_MSA    - 5.3x
ScaleARGBRowDownEven_Any_MSA    - 1.2x
ScaleARGBRowDownEvenBox_Any_MSA - 3.7x

Review URL: https://codereview.chromium.org/2527983002 .
 2016-12-07 11:47:15 +05:30
Frank Barchard
e62309f259 clang-format libyuv 
BUG=libyuv:654
R=kjellander@chromium.org

Review URL: https://codereview.chromium.org/2469353005 .
 2016-11-07 17:37:23 -08:00
Frank Barchard
fc52d8ded2 Odd width variation of scale down by 2 for subsampling 
R=dhrosa@google.com, harryjin@google.com
BUG=libyuv:538

Review URL: https://codereview.chromium.org/1558093003 .
 2016-01-06 15:12:17 -08:00
Frank Barchard
80ca4514ef change scale down by 4 to use rounding. 
TBR=harryjin@google.com
BUG=libyuv:447

Review URL: https://codereview.chromium.org/1525033005 .
 2015-12-15 21:25:18 -08:00
Frank Barchard
ae55e41851 use rounding in scaledown by 2 
When scaling down by 2 the formula should round consistently.
(a+b+c+d+2)/4
The C version did but the SSE2 version was doing 2 averages.
avg(avg(a,b),avg(c,d))
This change uses a sum, then rounds.

R=dhrosa@google.com, harryjin@google.com
BUG=libyuv:447,libyuv:527

Review URL: https://codereview.chromium.org/1513183004 .
 2015-12-14 17:25:36 -08:00
fbarchard@google.com
05416e2d9a Box filter for YUV use rows with accumulation buffer for better memory behavior. The old code would do columns accumulated into registers, and then store the result once. This was slow from a memory point of view. The new code does a row of source at a time, updating an accumulation buffer every row. The accumulation buffer is small, and should fit cache. Before each accumulation of N rows, the buffer needs to be reset to zero. If the memset is a bottleneck, it would be faster to do the first row without an add, storing to the accumulation buffer, and then add for the remaining rows. 
BUG=425
TESTED=out\release\libyuv_unittest --gtest_filter=*ScaleTo1x1*
R=harryjin@google.com

Review URL: https://webrtc-codereview.appspot.com/52659004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1428 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-06-09 01:05:18 +00:00
fbarchard@google.com
7c09264ffc odd width support for scale by even scale factor and box scale down by 4. scale down by 4 uses scale down by 2 internally. 
BUG=431
TESTED=libyuvTest.ARGBScaleDownBy4_Bilinear

Review URL: https://webrtc-codereview.appspot.com/57399004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1412 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-05-26 17:56:51 +00:00
fbarchard@google.com
c38aeec322 scale down by 2 on argb images support odd widths using _any function. 
BUG=431
TESTED=libyuvTest.ARGBScaleDownBy2_Bilinear

Review URL: https://webrtc-codereview.appspot.com/52569004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1410 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-05-22 21:39:21 +00:00
fbarchard@google.com
31806d76d9 scale to 3/4 bug fix for odd widths. multiply to index into source by scale factor should be 4 / 3 not 3 / 4. 
BUG=433
TESTED=set LIBYUV_WIDTH=1276 out\release\libyuv_unittest.exe --gtest_catch_exceptions=0 --gtest_filter=*.Scale*
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/49219004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1391 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-04-30 17:18:13 +00:00
fbarchard@google.com
9f4636e298 AVX2 port of ScaleDownBy4. 
BUG=314
TESTED=out\release\libyuv_unittest --gtest_filter=*.ScaleDownBy4*

Review URL: https://webrtc-codereview.appspot.com/46159004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1390 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-04-30 01:58:32 +00:00
fbarchard@google.com
4e78b8dc2e scale to 3/4 or 3/8 with odd width destinations efficiently. previously if width was not multiple of what the simd loop would do (24), scaling would fall back on slower C code. This change allows SIMD to be used for most of the scaling and C for the remainder, improving efficiency. 
BUG=314
TESTED=set LIBYUV_WIDTH=1896 & ScaleDownBy3by4_*
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/48249004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1380 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-04-27 21:56:08 +00:00
fbarchard@google.com
1ffb04b43e Allow ScaleRowDown any functions to accept non-power of 2 for destination SIMD multiple. 
BUG=none
TESTED=local unittests pass
R=bcornell@google.com

Review URL: https://webrtc-codereview.appspot.com/45129004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1379 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-04-24 22:32:12 +00:00
fbarchard@google.com
2b7f6b7dee ScaleAddRows_Any_SSE2 functions for handling odd widths. 
BUG=425
TESTED=out\release\libyuv_unittest_old --gtest_filter=*.ScaleDownBy3_*
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/45219004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1377 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-04-22 00:51:56 +00:00
yang.zhang@arm.com
5f609856de Add ScaleARGBFilterCols_NEON for ARM32/64 
ARM32/64 NEON versions of ScaleARGBFilterCols_NEON are implemented.

BUG=319
TESTED=libyuvTest.* on ARM32/64 with Android
R=fbarchard@google.com

Change-Id: Ifea62bc25d846bf16cb51d13b408de7bf58dccd4

Review URL: https://webrtc-codereview.appspot.com/46699004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1361 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-04-07 03:45:29 +00:00
fbarchard@google.com
44b6ba91e4 Scale down by 4 for odd number of destination pixels using 'any' that handles SIMD for multiple of 8 pixels, and C for the remainder. 
BUG=314
TESTED=local test with width odd

Review URL: https://webrtc-codereview.appspot.com/49599004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1355 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-04-03 22:12:53 +00:00
yang.zhang@arm.com
f23d6222ac Add ScaleARGBCols_NEON for ARM32/64 
ARM32/64 NEON versions of ScaleARGBCols_NEON are implemented.

BUG=319
TESTED=libyuvTest.* on ARM32/64 with Android
R=fbarchard@google.com

Change-Id: Id9ad97f7aa5d8a34cd55ace9e648cb6ff028efd9

Review URL: https://webrtc-codereview.appspot.com/47689004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1351 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-03-31 03:03:05 +00:00
fbarchard@google.com
72673ac873 linear and point sample scale to half size for AVX2. 
BUG=314
TESTED=out\release\libyuv_unittest.exe --gtest_catch_exceptions=0 --gtest_filter=*.ScaleDownBy2*
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/44959004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1349 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-03-30 21:46:08 +00:00
fbarchard@google.com
e6ca9cc2a2 Scale down by 2 AVX2 port. Processes twice as many pixels as SSE2 and takes advantage of 3 argument instructions to reduce register usage and number of instructions. 
BUG=314
TESTED=libyuvTest.ScaleDownBy2_Box
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/42959004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1347 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-03-26 23:21:08 +00:00
fbarchard@google.com
d41fbf40dd Handle scale down by factor of 2 efficiently by calling SIMD for multiple of 16 destination pixels, and C for remainder. 
BUG=314
TESTED=out\release\libyuv_unittest.exe --gtest_catch_exceptions=0 --gtest_filter=*.ScaleDownBy2*
R=bcornell@google.com

Review URL: https://webrtc-codereview.appspot.com/48689004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1344 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-03-24 23:25:30 +00:00
yang.zhang@arm.com
d6d7de5742 Add ScaleFilterCols_NEON for ARM32/64 
ARM32/64 NEON versions of ScaleFilterCols_NEON are implemented.

BUG=319
TESTED=libyuvTest.* on ARM32/64 with Android
R=fbarchard@google.com

Change-Id: I5b0838769ffb0182155d7cd6bcc520eb81eb5c4e

Review URL: https://webrtc-codereview.appspot.com/41349004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1340 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-03-19 03:55:05 +00:00