- MCA says old version was 4 cycles and new version is 2.5 cycles/loop
- lunarlake is the only known cpu
mca -mcpu=lunarlake 100 iterations
Was vpmulhu
Iterations: 100
Instructions: 1200
Total Cycles: 426
Total uOps: 1200
Dispatch Width: 8
uOps Per Cycle: 2.82
IPC: 2.82
Block RThroughput: 4.0
Now vpsrlw
Iterations: 100
Instructions: 1200
Total Cycles: 279
Total uOps: 1400
Dispatch Width: 8
uOps Per Cycle: 5.02
IPC: 4.30
Block RThroughput: 2.5
Bug: None
Change-Id: I5a49e1cf1ed3dfb59fe9861a871df9862417c6a6
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/6697745
Reviewed-by: richard winterton <rrwinterton@gmail.com>
- Was using avgb twice for non-exact and C for exact.
On Skylake Xeon:
Now SSE3
ARGBToJ420_Opt (326 ms)
Was
Exact C
ARGBToJ420_Opt (871 ms)
Not exact AVX2
ARGBToJ420_Opt (237 ms)
Not exact SSSE3
ARGBToJ420_Opt (312 ms)
Bug: 381138208
Change-Id: I6d1081bb52e36f06736c0c6575fa82bb2268629b
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/6629605
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Ben Weiss <bweiss@google.com>
- Add +i8mm build option for sve ARGBToUV which uses usdot
- util/cpuid Get cpu count (windows, macos, linux)
- For each x86 cpu, detect hybrid (e-core)
- Includes a comment fix for ubsan unittest
- Bump version
- Apply clang format to util/*.c as well as all *.cc/*.h
Bug: 424637372
Change-Id: I08310e18051fff62c9e4e4a10d1e4361871119ac
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/6635640
Reviewed-by: Wan-Teh Chang <wtc@google.com>
- make width loop count on stack
- set YMM constants in its own asm block
- make struct for shuffle and add constants
- disable clang format on row_neon.cc function
Bug: 413781394
Change-Id: I263f6862cb7589dc31ac65d118f7ebeb65dbb24a
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/6495259
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
- use negative coefficients for UV to allow -128
- change shift to truncate instead of round for UV
- adapt all row_gcc RGB to UV into matrix functions
- add -DLIBYUV_ENABLE_ROWWIN to allow clang on Windows to use row_win.cc
Bug: 381138208
Change-Id: I6016062c859faf147a8a2cdea6c09976cbf2963c
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/6277710
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Reviewed-by: James Zern <jzern@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
- avx2 is pack/perm is mutating order
- cvt method maintains channel order on avx512
Sapphire Rapids
Benchmark of 640x360 on Sapphire Rapids
AVX512BW
[ OK ] LibYUVConvertTest.I010ToNV12_Opt (3547 ms)
[ OK ] LibYUVConvertTest.P010ToNV12_Opt (3186 ms)
AVX2
[ OK ] LibYUVConvertTest.I010ToNV12_Opt (4000 ms)
[ OK ] LibYUVConvertTest.P010ToNV12_Opt (3190 ms)
SSE2
[ OK ] LibYUVConvertTest.I010ToNV12_Opt (5433 ms)
[ OK ] LibYUVConvertTest.P010ToNV12_Opt (4840 ms)
Skylake Xeon
Now vpmovuswb
[ OK ] LibYUVConvertTest.I010ToNV12_Opt (7946 ms)
[ OK ] LibYUVConvertTest.P010ToNV12_Opt (7071 ms)
Was vpackuswb
[ OK ] LibYUVConvertTest.I010ToNV12_Opt (7684 ms)
[ OK ] LibYUVConvertTest.P010ToNV12_Opt (7059 ms)
Switch from vpunpcklwd to vpbroadcastw for scale value parameter
Was
vpunpcklwd %%xmm2,%%xmm2,%%xmm2
vbroadcastss %%xmm2,%%ymm2
Now
vpbroadcastw %%xmm2,%%ymm2
Bug: 357439226, 357721018
Change-Id: Ifc9c82ab70dba58af6efa0f57f5f7a344014652e
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/5787040
Reviewed-by: Wan-Teh Chang <wtc@google.com>
A semicolon is treated as the start of a comment by some assemblers
causing the vector length to be reported incorrectly, so use a newline
instead.
- Add volatile asm in row_gcc and row_neon64
Bug: b/5631539
Change-Id: I6b0836fcdd9247ef7b9e8ceda01df3150519ecf8
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/5666060
Reviewed-by: Justin Green <greenjustin@google.com>
- ARM Planar test use regular asm volatile syntax
- x86 row functions remove volatile from asm
Bug: 347111119, 347112532
Change-Id: I535b3dfa1a7a19824503bd95584a63b047b0e9a1
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/5637058
Reviewed-by: Justin Green <greenjustin@google.com>
- Makes ARM and Intel match and fixes some off by 1 cases
- Add ARGBToUV444MatrixRow_NEON
- Add ConvertFP16ToFP32Column_NEON
- scale_rvv fix intinsic build error
- disable row_win version of ARGBAttenuate/Unattenuate
Bug: libyuv:936, libyuv:956
Change-Id: Ied99aaad3a11a8eb69212b628c58f86ec0723c38
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4617013
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
- Convert MergeUVRow_AVX512BW to assembly
- Enable MergeUVRow_AVX512BW for Windows with clangcl
- MergeUVRow_AVX2 use vpmovzxbw and vpsllw
- MergeUVRow_16_AVX2 use vpmovzxbw and vpsllw with different shift for U and V
AMD Zen 4 640x360 100000 iterations
Was
AVX512 MergeUVPlane_Opt (884 ms)
AVX2 MergeUVPlane_Opt (945 ms)
AVX2 MergeUVPlane_16_Opt (2167 ms)
Now
AVX512 MergeUVPlane_Opt (865 ms)
AVX2 MergeUVPlane_Opt (943 ms)
SSE2 MergeUVPlane_Opt (973 ms)
AVX2 MergeUVPlane_16_Opt (2102 ms)
Bug: None
Change-Id: I658ada2a75d44c3f93be8bd3ed96f83d5fa2ab8d
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4271230
Reviewed-by: Fritz Koenig <frkoenig@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
- fix ifdefs for DetilePlane_16 to use 16 bit versions, not 8 bit. (no functional change)
Bug: b/258474032
Change-Id: Ic07e02d9801e21126ebee0ceb5779aa712a493ce
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4034812
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
- Optimized YUY2ToNV12 that reduces it from 3 steps to 2 steps
- Was SplitUV, memcpy Y, InterpolateUV
- Now YUY2ToY, YUY2ToNVUV
- rollback LIBYUV_UNLIMITED_DATA
3840x2160 1000 iterations:
Pixel 2 Cortex A73
Was YUY2ToNV12_Opt (6515 ms)
Now YUY2ToNV12_Opt (3350 ms)
AB7 Mediatek P35 Cortex A53
Was YUY2ToNV12_Opt (6435 ms)
Now YUY2ToNV12_Opt (3301 ms)
Skylake AVX2 x64
Was YUY2ToNV12_Opt (1872 ms)
Now YUY2ToNV12_Opt (1657 ms)
SSE2 x64
Was YUY2ToNV12_Opt (2008 ms)
Now YUY2ToNV12_Opt (1691 ms)
Windows Skylake AVX2 32 bit x86
Was YUY2ToNV12_Opt (2161 ms)
Now YUY2ToNV12_Opt (1628 ms)
Bug: libyuv:943
Change-Id: I6c2ba2ae765413426baf770b837de114f808f6d0
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3929843
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
- YUV to RGB use linear for first and last row.
- add assert(yuvconstants)
- rename pointers to match row functions.
- use macros that match row functions.
- use 12 bit upsampler for conversions of 10 and 12 bits
Cortex A53 AArch32
I420ToRGB24_Opt (3627 ms)
I422ToRGB24_Opt (4099 ms)
I444ToRGB24_Opt (4186 ms)
I420ToRGB24Filter_Opt (5451 ms)
I422ToRGB24Filter_Opt (5430 ms)
AVX2
Was I420ToRGB24Filter_Opt (583 ms)
Now I420ToRGB24Filter_Opt (560 ms)
Neon Cortex A7
Was I420ToRGB24Filter_Opt (5447 ms)
Now I420ToRGB24Filter_Opt (5439 ms)
Bug: libyuv:938
Change-Id: I1731f2dd591073ae11a756f06574103ba0f803c7
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3906082
Reviewed-by: Justin Green <greenjustin@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
MergeRGB and SplitRGB use a register to point to 9 shuffle tables.
- fixes an out of registers error with -mcmodel=large
InterpolateRow_16To8_NEON improves performance for I210ToI420:
On Pixel 4 for 720p x1000 images
Was I210ToI420_Opt (608 ms)
Now I210ToI420_Opt (336 ms)
On Skylake Xeon
Was I210ToI420_Opt (259 ms)
Now I210ToI420_Opt (209 ms)
Bug: libyuv:931, libyuv:930
Change-Id: I20f8244803f06da511299bf1a2ffc7945eb35221
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3717054
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Justin Green <greenjustin@google.com>
Add support for MM21 to NV12 and I420 conversion, and add SIMD
optimizations for arm, aarch64, SSE2, and SSSE3 machines.
Bug: libyuv:915, b/215425056
Change-Id: Iecb0c33287f35766a6169d4adf3b7397f1ba8b5d
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3433269
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Commit-Queue: Justin Green <greenjustin@google.com>
- reenable Intel SIMD unaffected by BIT_EXACT
- add bit exact version of ARGBAttenuate, which uses ARM version of formula.
- add bit exact version of ARGBUnatenuate, which mimics the AVX code.
Apply clang format to cleanup code.
Bug: libyuv:908, b/202888439
Change-Id: Ie842b1b3956b48f4190858e61c02998caedc2897
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3224702
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Add tests of all macros used by libyuv public headers
When a 1 step conversion is added, a 2 step test can compare
the old 2 step method to the 1 step. A 1 step unittest is
also added which compares C to SIMD. Making the 2 step
conversions measure performance of the 2 steps allows the
old 2 step performance to be compared to 1 step.
All macros used in public headers are added to an ifdef test.
Showing them in a unittest allows some diagnostics when
a test is failing.
Bug: libyuv:901
Change-Id: I7ffa6ed0cb3b506fa1b7fd4b7b1b729658c3c266
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2857916
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Mirko Bonadei <mbonadei@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Use unsigned coefficient and signed UV value in YUVTORGB.
R=fbarchard@chromium.org
Bug: libyuv:862, libyuv:863
Change-Id: I32e58b2cee383fb98104c055beb0867a7ad05bfe
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2850016
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Use unsigned coefficients on Intel.
Make C, NEON and AVX2 match under LIBYUV_UNLIMITED_DATA.
Bug: libyuv:862, libyuv:863
Change-Id: I6c02147ea3c1875c4fc23863435aea86dcf5880a
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2830180
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
These functions merge high bit depth planar RGB pixels into packed format.
Change-Id: I506935a164b069e6b2fed8bf152cb874310c0916
Bug: libyuv:886, libyuv:889
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2780468
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Planar functions pass depth instead of scale factor.
Row functions pass shift instead of depth. Add assert to C.
AVX shift instruction expects a single shift value in XMM.
Neon pass shift as input (not output).
Split Neon reimplemented as left shift on shorts by negative to achieve right shift.
Add planar unitests
Bug: libyuv:888
Change-Id: I8fe62d3d777effc5321c361cd595c58b7f93807e
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2782086
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Reviewed-by: Mirko Bonadei <mbonadei@chromium.org>
Add J420 output from RAW.
Optimize RGB24 and RAW To J420 on ARM by using NEON for the 2 step conversion.
Also fix sign-compare warning that was breaking Windows build
Bug: libyuv:887, b/183534734
Change-Id: I8c39334552dc0b28414e638708db413d6adf8d6e
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2783382
Reviewed-by: Wan-Teh Chang <wtc@google.com>
The following functions are added:
planar YUV:
I410ToAR30, I410ToARGB
planar YUVA:
I010AlphaToARGB, I210AlphaToARGB, I410AlphaToARGB
biplanar YUV:
P010ToARGB, P210ToARGB
P010ToAR30, P210ToAR30
biplanar functions can also handle 12 bit and 16 bit samples.
libyuv_unittest --gtest_filter=LibYUVConvertTest.*10*ToA*:LibYUVConvertTest.*P?1?ToA*
R=fbarchard@chromium.org
Bug: libyuv:751, libyuv:844
Change-Id: I2be02244dfa23335e1e7bc241fb0613990208de5
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2707003
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
These are 16 bit bi-planar convert functions to scale UV plane to
Y plane's size using (bi)linear filter.
libyuv_unittest --gtest_filter=*ToP41*
R=fbarchard@chromium.org
Bug: libyuv:872
Change-Id: I3cb4fafe2b2c9eedd0d91cf4c619abb9ee107bc1
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2690102
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
subq is only available for x64
sub works for both 32 bit x86 and 64 bit x64
Fox in row_gcc.cc for 32 bit x86 running out of registers.
Fix in row_neon.cc for split function argb paramter name.
Bug: libyuv:877, b/178283356, b/178713286
Change-Id: If2b12a2d6168eab08005a2cdf2c17a470a924dd1
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2656771
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
These functions convert between planar and interleaved ARGB,
optionally fill 255 to alpha / discard alpha.
This can help handle YUV(A) with Identity matrix, which is
basically planar ARGB.
libyuv_unittest --gtest_filter=LibYUVPlanarTest.*ARGBPlane*:LibYUVPlanarTest.*XRGBPlane*
R=fbarchard@google.com
Change-Id: I522a189b434f490ba1723ce51317727e7c5eb112
Bug: libyuv:877
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2649887
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
MAKEYUVCONSTANTS macro to generate struct for YUV to RGB
Fix I444AlphaToARGB unit test for ARM by adjusting C version to match Neon implementation.
Bug: libyuv:879, libyuv:878, libyuv:877, libyuv:862, b/178283356
Change-Id: Iedb171fbf668316e7d45ab9e3481de6205ed31e2
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2646472
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Duplicate I420ToARGB prototype from convert_argb.h into convert_from.h for webrtc
Apply clang format for white spacing consistency.
Bug: libyuv:838, b/151375918
Change-Id: I0f667ca5350192710dbb135e92e73e18b46135e5
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2446613
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
M420 is a row biplanar variation of NV12 supported on Microsoft webcams.
The code was hardcoded to bt.601 and should be jpeg, but the format is
very old and rare. Is a variation on NV12, so if someone needs it, it
can be re-implemented easily.
Bug: libyuv:858
Change-Id: I246167dba3c190cc76af741b8e91e58e68fde28f
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2212608
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Replace ARM64 only row function with high level function
that implements SSSE3, 32 bit Neon and C.
Compared to 2 step RAWToARGB + ARGBToRGBA on row level:
3.1x faster on ARM
6.2% faster on Intel
BUG=b/140748379
Change-Id: Ia8636d9e4fcdbe10b8c2e81610a54728e29845cd
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/1860914
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Neon and GCC Intel optimized, but win32 and mips not optimized.
BUG=libyuv:842, b/141482243
Change-Id: Ia56fa85c8cc1db51f374bd0c89b56d21ec94afa7
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/1825642
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Miguel Casas <mcasas@chromium.org>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Based on ARGBShuffle but with count adjusted and new shuffle mask
BUG=libyuv:809
Change-Id: Idd936ee6bedcf285607a68c2fc54d876b4becc01
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/1711882
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Apply clang-format to fix jpeg if() for lint fix.
Change comments about 4th pixel for open source compliance.
Rename UVToVU to SwapUV for consistency with MergeUV.
BUG=b/135532289, b/136515133
Change-Id: I9ce377c57b1d4d8f8b373c4cb44cd3f836300f79
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/1685936
Reviewed-by: Chong Zhang <chz@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Alternatives to RGB24 and AYUV for working with GPU.
BUG=libyuv:832
TESTED=out/Release/libyuv_unittest --gtest_filter=*NV21To???24* --libyuv_width=1280 --libyuv_height=720 --libyuv_repeat=1000 --libyuv_flags=-1 --libyuv_cpu_info=-1
R=rrwinterton@gmail.com
Change-Id: I5559c63f4bd4c847492fcb1571f7b03c58146689
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/1501735
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
iOS simulator has the option to build with xcode instead of clang.
GN use_xcode_clang=true enables the xcode build.
As of version Xcode 9.2, the clang version used does not support
AVX512. The version reported is version 9, but for normal clang,
version 7 is sufficient to AVX512.
When a version of XCode does support AVX512, the version check can
be updated to allow AVX512 for newer versions of XCode.
with XCode 9.2 the following macro is set.
__APPLE_CC__ 6000
Bug: libyuv:789
Test: gn gen out/Release "--args=is_debug=false target_os=\"ios\" ios_enable_code_signing=false target_cpu=\"x86\" use_xcode_clang=true"
Change-Id: I5a9a0b4a2760c7d09a4bcb464b3668979113b07e
Reviewed-on: https://chromium-review.googlesource.com/991595
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Bug: libyuv:789
Test: builds locally on linux with clang
Change-Id: I3000494d4b0b18f59d7852bc1bc0c9e422d2d63a
Reviewed-on: https://chromium-review.googlesource.com/987331
Reviewed-by: richard winterton <rrwinterton@gmail.com>
row.h adds CLANG_HAS_AVX512
function ifdefs in row.h for avx512
source code ifdefed function by function for
avx512 and avx2.
Bug: libyuv:778
Test: LibYUVConvertTest.NV21ToRGB24_Opt
Change-Id: If32b51459685d0d5785c5c1e94c8f668f8e74b55
Reviewed-on: https://chromium-review.googlesource.com/982402
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Use VMBI instructions but on AVX2 registers to avoid clockrate change.
Bug: libyuv:778
Test: LibYUVConvertTest.NV21ToRGB24_Opt
Change-Id: Id4f8ad1e0e142a380c8a46c5eab90ce145a10edd
Reviewed-on: https://chromium-review.googlesource.com/956609
Reviewed-by: richard winterton <rrwinterton@gmail.com>
AVX2 port of SSSE3 conversion to output 24 bit RGB
Bug: libyuv:778
Test: LibYUVConvertTest.NV21ToRGB24_Opt
Change-Id: I14f7815522d1b790ecd2bb39d9a3441e803b694a
Reviewed-on: https://chromium-review.googlesource.com/953303
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Bug: libyuv:750
Test: builds and runs and passes more tidy tests
Change-Id: I023699a7aa61ea3f5e4a21647112691ea5739281
Reviewed-on: https://chromium-review.googlesource.com/902170
Reviewed-by: Weiyong Yao <braveyao@chromium.org>
I422ToUYVYRow_AVX2 optimized from 7 cycles per 32 pixels to 4.6 cycles.
Instead of 2 vpermq and vpunpcklbw:
vmovdqu (%1),%%xmm2
vmovdqu 0x00(%1,%2,1),%%xmm3
vpermq $0xd8,%%ymm2,%%ymm2
vpermq $0xd8,%%ymm3,%%ymm3
vpunpcklbw %%ymm3,%%ymm2,%%ymm2
..use vpmovzxbd to expand the bytes to shorts, then vpslld and vpor
vpmovzxbd (%1),%%ymm2
vpmovzxbd 0x00(%1,%2,1),%%ymm3
vpslld $0x10,%%ymm3,%%ymm3
vpor %%ymm3,%%ymm2,%%ymm2
which reduces the port 5 bottleneck by 1 cycle.
Bug: libyuv:556
Test: out/Release/libyuv_unittest --gtest_filter=*I42?To*UY*Opt
Change-Id: I53799e53cc6b090a1a695c839094c193be3eecaf
Reviewed-on: https://chromium-review.googlesource.com/899873
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Reviewed-by: Cheng Wang <wangcheng@google.com>
