coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2026-10-28 18:07:12 +08:00
Code Issues Packages Projects Releases Wiki Activity
3,115 Commits 4 Branches 0 Tags BSD-3-Clause 65 MiB
Commit Graph

30 Commits

Author SHA1 Message Date
Frank Barchard
16af162cf9 [libyuv] Optimize RotatePlane_16 SIMD and compact 32-bit transposes 
Implement SIMD 16-bit plane rotation (TransposePlane_16,
RotatePlane90_16, RotatePlane180_16, RotatePlane270_16) used by
I010Rotate, I210Rotate, and I410Rotate, keeping tile sizes compact
with full-register writes and zero cross-lane shuffles:
- AVX2: 8x16 tile in ymm0..ymm15 with _Any tail (low-end x86 SIMD)
- AVX-512BW: 8x32 tile in zmm0..zmm15 with k1/k2 right/bottom masks
- NEON (ARM64/ARM32): 8x16 tile in v0..v23 / q0..q11 with _Any tail
- SME: 8xVL tile using ZA16 horizontal/vertical array slices
- RVV: 8xVL tile using widening unpacks with hardware vl tail
- WASM SIMD: 8x8 tile in v128 with _Any tail

Also update 32-bit ARGBRotate x86 transposes to compact vertical
in-lane tiles with full-register writes and zero cross-lane shuffles:
- TransposeWxH_32_AVX2: 4x8 tile in ymm0..ymm7 with vpmaskmovd tails
- TransposeWxH_32_AVX512BW: 4x16 tile in zmm0..zmm7 with k1/k2 tails

Add RotatePlane16_SplitMergeUV_Opt and RotatePlane16_SplitMergeUV_Odd
unit tests comparing 16-bit rotation against SplitUVPlane +
RotatePlane + MergeUVPlane across 0/90/180/270 degrees.

Performance on 1275x1650 (1000 iterations, before -> after):
CPU                     90 degree               180 degree
Coral RVV (Verilator)   7306 -> 1415c (5.16x)   2274 -> 719c (3.16x)
AMD Rome (Zen 2)        821 -> 215ms (3.82x)    259 -> 181ms (1.44x)
AMD Genoa (Zen 4)       483 -> 129ms (3.73x)    149 -> 111ms (1.34x)
AMD Turin (Zen 5)       321 -> 101ms (3.18x)    109 -> 78ms (1.40x)
AMD Milan (Zen 3)       505 -> 189ms (2.67x)    199 -> 118ms (1.69x)
Intel Skylake-X         1076 -> 447ms (2.41x)   353 -> 317ms (1.11x)
WASM SIMD (d8)          415 -> 203ms (2.04x)    370 -> 106ms (3.49x)
Intel Emerald Rapids    735 -> 422ms (1.74x)    238 -> 233ms (1.02x)
Intel Haswell           1120 -> 648ms (1.73x)   438 -> 393ms (1.11x)
Intel Ice Lake          774 -> 474ms (1.63x)    279 -> 290ms (0.96x)
ARM Neoverse-N1         864 -> 544ms (1.59x)    404 -> 266ms (1.52x)
Ampere Siryn            862 -> 576ms (1.50x)    568 -> 329ms (1.72x)
Google Axion (Maple)    554 -> 379ms (1.46x)    199 -> 142ms (1.41x)
Intel Sapphire Rapids   1126 -> 772ms (1.46x)   301 -> 288ms (1.04x)
Intel Granite Rapids    730 -> 565ms (1.29x)    280 -> 280ms (1.00x)
NVIDIA Grace            462 -> 365ms (1.26x)    192 -> 126ms (1.52x)
ARM Cortex-A53          28290 -> 24386ms (1.16x) 7178 -> 5075ms (1.41x)
Google Axion 2 (Tamar)  523 -> 453ms (1.15x)    217 -> 112ms (1.93x)

Test: libyuv_unittest --gtest_filter=*Rotate*
Bug: b/42280942
Change-Id: I91ba6cac22185b5d294fcb98344fdb8a23cc94ff
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/8545294
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@google.com>
 2026-10-12 10:57:34 -07:00
Frank Barchard
eeb06a6451 [libyuv] Optimize ARGBRotate for Intel SSE2, AVX2, and AVX-512 
Optimize 32-bit ARGBRotate (0, 90, 180, 270) across Intel x86/x64
architectures (SSE2, AVX2, AVX512BW), with matching kernels for ARM
(NEON, SVE2, SME), WASM, LoongArch (LSX, LASX), and RISC-V (RVV):

- Replace column-by-column ScaleARGBRowDownEven gather loops in
  ARGBTranspose (90/270) with full-plane horizontal tile transposes:
  - AVX512BW (TransposeWxH_32_AVX512BW / TransposeNx4_32_AVX512BW):
    Transpose 16x16 blocks across zmm0..zmm31 using 4-stage
    vpunpckldq/hdq, vpunpcklqdq/hqdq, and vshufi32x4 butterflies with
    k-masked right and bottom tails, eliminating Any wrappers and scalar
    C edge fallbacks.
  - AVX2 (TransposeWxH_32_AVX2 / TransposeNx4_32_AVX2):
    Transpose 8x8 blocks across ymm0..ymm15 using vpunpck* and
    vperm2i128 with vpmaskmovd masked tails.
  - SSE2 (Transpose4x4_32_SSE2 / Transpose4x4_32_Any_SSE2):
    Transpose 4x4 blocks in xmm registers.
- Implement ARGBRotate180 directly via ARGBMirror with negative height,
  avoiding intermediate row buffer allocations and extra copies.
- Implement ARGBRotate0 via ARGBCopy.

Speedup vs Before (1275x1650, US Letter 150 DPI, 1000 iterations):
CPU                  | 0 deg   | 90 deg  | 180 deg | 270 deg
---------------------|---------|---------|---------|--------
Intel Ice Lake       | 1.02x   | 1.34x   | 1.18x   | 1.33x
Intel Skylake-X      | 1.02x   | 1.20x   | 1.07x   | 1.13x
Intel Emerald Rapids | 1.01x   | 1.07x   | 1.00x   | 1.07x
Intel AVX-512 vs C   | 1.00x   | 2.67x   | 2.41x   | 2.63x

R=rrwinterton@gmail.com

Bug: b/42280915
Test: LIBYUV_WIDTH=1275 LIBYUV_HEIGHT=1650 LIBYUV_REPEAT=1000 ./out/Release/libyuv_unittest --gtest_filter=*ARGBRotate*
Change-Id: Ib308a225c17e46eab083a9da4033598fcf0c303c
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/8538986
Reviewed-by: richard winterton <rrwinterton@gmail.com>
 2026-10-08 16:20:46 -07:00
Frank Barchard
cb061d0378 Unittests use ASSERT instead of EXPECT 
Bug: libyuv:508639302
Change-Id: I22c35e08f3b6db1a656192877c1fb1bf4e96d6f5
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/7838659
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@google.com>
 2026-05-11 19:10:47 -07:00
Wan-Teh Chang
3cf54e90d3 Fix -Wmissing-prototypes warnings 
Declare functions as static. Declare functions in a header. Include the
header that declares the functions. Delete undeclared and unused
functions ScaleFilterRows_NEON() and ScaleRowUp2_16_NEON(). Delete
unused function ScaleY() in psnr_main.cc.

Change-Id: I182ec30611df83c61ffd01bbab595cd61fb5f1e5
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/5778601
Commit-Queue: Wan-Teh Chang <wtc@google.com>
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
 2024-08-12 19:08:24 +00:00
Sergio Garcia Murillo
f8626a7224 Add 10 bit rotate methods. 
This initial implementation is based on current unoptimized code in webrtc using just plain for loops.

Bug: libyuv:949
Change-Id: Ic87ee49c3a0b62edbaaa4255c263c1f7be4ea02b
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4110782
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
 2023-01-04 21:10:01 +00:00
Frank Barchard
287158925b use width + 1 for odd width tests 
Bug: libyuv:894, libyuv:898, libyuv:899
Change-Id: Ieba8eaeb8b06f0323824967776673e339b263220
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2809701
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Mirko Bonadei <mbonadei@chromium.org>
 2021-04-09 20:17:55 +00:00
Frank Barchard
b5e223ac4c Upstream all libyuv changes to version 1746 Prefetch for all arm functions - helps performance at higher resolutions Make MirrorPlane function public. 
Bug: libyuv:855
Change-Id: I4020face6b52767ee78d81870314285d63e98b95
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2113650
Reviewed-by: Hsiu Wang <hsiu@google.com>
 2020-03-21 20:19:44 +00:00
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
ade85fb55c remove row.h from unittests 
add SIMD_ALIGNED to unittest header.

BUG=libyuv:594
TESTED=local build passes with row.h removed from tests.
R=harryjin@google.com

Review URL: https://codereview.chromium.org/2001373002 .
 2016-05-27 10:57:49 -07:00
Frank Barchard
8030a711aa Rename rotate tests to include _Opt and disable _Odd tests 
TBR=harryjin@google.com
BUG=libyuv:543

Review URL: https://codereview.chromium.org/1577723003 .
 2016-01-11 17:30:27 -08:00
Frank Barchard
c629cb3afe add command line cpu info to allow android neon test 
in order to compare C and Neon code, a new command line flag is added.
historically environment variables controlled cpu features, but on
android apk it is easier to pass a command line option to disable cpu
optimizations.

R=harryjin@google.com
BUG=libyuv:516

Review URL: https://codereview.chromium.org/1407193009 .
 2015-11-03 17:01:48 -08:00
Frank Barchard
26db4de2ae break up unittests into categories 
R=harryjin@google.com
BUG=none

Review URL: https://codereview.chromium.org/1399523004 .
 2015-10-13 16:01:07 -07:00
Frank Barchard
16f12b58cc Replace random with fastrand 
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 .
 2015-09-22 15:47:36 -07:00
fbarchard@google.com
01ca0712c7 Test rotation with negative height 
BUG=446
TESTED=libyuvTest.NV12Rotate180_Inverted
R=bcornell@google.com

Review URL: https://webrtc-codereview.appspot.com/51169004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1420 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-06-02 22:42:45 +00:00
fbarchard@google.com
f16f33d4ce All cpu flags to be set so that instead of comparing C code, compare assembler to assembler, for benchmarking purposes. 
BUG=none
TESTED=libyuv_unittest.exe
R=bcornell@google.com

Review URL: https://webrtc-codereview.appspot.com/50499004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1346 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2015-03-26 18:22:29 +00:00
fbarchard@google.com
d10f80500f Improve cmake build. Add unittests to cmake build and automatically detect jpeg support. This change was originally generated to support the build of libyuv in naclports: https://chromium.googlesource.com/external/naclports/+/master/ports/libyuv/. Also add cmake artifacts to .gitignore file. 
BUG=366
TESTED=build and run unittests with cmake
R=harryjin@google.com

Review URL: https://webrtc-codereview.appspot.com/27009004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1146 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2014-10-28 23:37:11 +00:00
fbarchard@google.com
ae9a1388a7 Use malloc for row buffers in rotate 
BUG=296
TESTED=rotate_test
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/6329004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@922 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2013-12-26 21:41:11 +00:00
fbarchard@google.com
a14b5cdfb3 Simplify rotate unittests 
BUG=233
TEST=*Rotate*
R=ryanpetrie@google.com

Review URL: https://webrtc-codereview.appspot.com/1581004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@705 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2013-05-28 22:40:51 +00:00
fbarchard@google.com
cde587092f Replace two spaces with one after . 
BUG=none
TEST=lint
Review URL: https://webrtc-codereview.appspot.com/1063010

git-svn-id: http://libyuv.googlecode.com/svn/trunk@553 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2013-01-28 00:02:35 +00:00
fbarchard@google.com
2580ccf64d Scale benchmark just the OPT and run C version once. 
BUG=none
TEST=scale_test
Review URL: https://webrtc-codereview.appspot.com/929028

git-svn-id: http://libyuv.googlecode.com/svn/trunk@503 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-11-26 22:39:30 +00:00
fbarchard@google.com
95d845bd5b benchmark only the opt version of rotate. 
BUG=155
TEST=sudo LIBYUV_REPEAT=1000 nice --5 ./libyuv_unittest --gtest_filter=*Rotate*
Review URL: https://webrtc-codereview.appspot.com/939029

git-svn-id: http://libyuv.googlecode.com/svn/trunk@502 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-11-26 19:10:35 +00:00
fbarchard@google.com
3e46444727 Mirror source to continuous destination with Neon 
BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/937020

git-svn-id: http://libyuv.googlecode.com/svn/trunk@488 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-11-14 02:03:49 +00:00
fbarchard@google.com
3bb60b37fd Use width variable for all tests 
BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/934004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@441 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-10-24 18:31:55 +00:00
fbarchard@google.com
042acf0458 Reduce allowed error on simple conversions 
BUG=none
TEST=unittests
Review URL: https://webrtc-codereview.appspot.com/917006

git-svn-id: http://libyuv.googlecode.com/svn/trunk@431 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-10-20 06:37:21 +00:00
fbarchard@google.com
64ce0ab544 Move Neon source to its own files. 
BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/860009

git-svn-id: http://libyuv.googlecode.com/svn/trunk@396 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-10-09 00:05:29 +00:00
fbarchard@google.com
7781d943b0 Disable V210 unittest that fails valgrind, and some warning fixes for android. 
BUG=91
TEST=try
Review URL: https://webrtc-codereview.appspot.com/813004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@362 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-09-18 22:58:34 +00:00
fbarchard@google.com
b8f820af71 make const int into int 
BUG=none
TEST=lint
Review URL: https://webrtc-codereview.appspot.com/766007

git-svn-id: http://libyuv.googlecode.com/svn/trunk@338 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-08-30 22:57:02 +00:00
fbarchard@google.com
de6eca3104 Change unittest to default to 1 iteration of each test. Can be change manually to 1000 for benchmarking purposes. 
BUG=77
TEST=build\release\libyuv_unittest.exe
Review URL: https://webrtc-codereview.appspot.com/768006

git-svn-id: http://libyuv.googlecode.com/svn/trunk@336 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-08-30 21:18:10 +00:00
fbarchard@google.com
b0c9797589 Update Copyright notice to follow new chromium conventions. 
BUG=63
TEST=none
Review URL: https://webrtc-codereview.appspot.com/730004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@314 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-08-08 19:04:24 +00:00
fbarchard@google.com
d51c342521 Rotate ARGB using a scale function that reads with step, combining to form 16 bytes at a time. 
BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/674007

git-svn-id: http://libyuv.googlecode.com/svn/trunk@294 16f28f9a-4ce2-e073-06de-1de4eb20be90
 2012-06-26 23:46:25 +00:00