mirror of
https://chromium.googlesource.com/libyuv/libyuv
synced 2026-10-29 02:17:13 +08:00
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>
205 lines
4.6 KiB
Plaintext
205 lines
4.6 KiB
Plaintext
package {
|
|
default_applicable_licenses: ["external_libyuv_license"],
|
|
}
|
|
|
|
// Added automatically by a large-scale-change
|
|
//
|
|
// large-scale-change included anything that looked like it might be a license
|
|
// text as a license_text. e.g. LICENSE, NOTICE, COPYING etc.
|
|
//
|
|
// Please consider removing redundant or irrelevant files from 'license_text:'.
|
|
// See: http://go/android-license-faq
|
|
license {
|
|
name: "external_libyuv_license",
|
|
visibility: [":__subpackages__"],
|
|
license_kinds: [
|
|
"SPDX-license-identifier-BSD",
|
|
],
|
|
license_text: [
|
|
"LICENSE",
|
|
"PATENTS",
|
|
],
|
|
}
|
|
|
|
cc_library {
|
|
name: "libyuv",
|
|
vendor_available: true,
|
|
product_available: true,
|
|
host_supported: true,
|
|
vndk: {
|
|
enabled: true,
|
|
},
|
|
|
|
srcs: [
|
|
"source/compare.cc",
|
|
"source/compare_common.cc",
|
|
"source/compare_gcc.cc",
|
|
"source/compare_neon.cc",
|
|
"source/compare_neon64.cc",
|
|
"source/convert.cc",
|
|
"source/convert_argb.cc",
|
|
"source/convert_from.cc",
|
|
"source/convert_from_argb.cc",
|
|
"source/convert_jpeg.cc",
|
|
"source/convert_to_argb.cc",
|
|
"source/convert_to_i420.cc",
|
|
"source/cpu_id.cc",
|
|
"source/mjpeg_decoder.cc",
|
|
"source/mjpeg_validate.cc",
|
|
"source/planar_functions.cc",
|
|
"source/rotate.cc",
|
|
"source/rotate_any.cc",
|
|
"source/rotate_argb.cc",
|
|
"source/rotate_common.cc",
|
|
"source/rotate_gcc.cc",
|
|
"source/rotate_neon.cc",
|
|
"source/rotate_neon64.cc",
|
|
"source/rotate_rvv.cc",
|
|
"source/row_any.cc",
|
|
"source/row_common.cc",
|
|
"source/row_gcc.cc",
|
|
"source/row_neon.cc",
|
|
"source/row_neon64.cc",
|
|
"source/row_rvv.cc",
|
|
"source/scale.cc",
|
|
"source/scale_any.cc",
|
|
"source/scale_argb.cc",
|
|
"source/scale_common.cc",
|
|
"source/scale_gcc.cc",
|
|
"source/scale_neon.cc",
|
|
"source/scale_neon64.cc",
|
|
"source/scale_rgb.cc",
|
|
"source/scale_rvv.cc",
|
|
"source/scale_uv.cc",
|
|
"source/video_common.cc",
|
|
],
|
|
|
|
cflags: [
|
|
"-Wall",
|
|
"-Werror",
|
|
"-Wno-unused-parameter",
|
|
"-fexceptions",
|
|
"-DHAVE_JPEG",
|
|
"-DLIBYUV_UNLIMITED_DATA",
|
|
],
|
|
|
|
arch: {
|
|
arm: {
|
|
cflags: ["-mfpu=neon"],
|
|
},
|
|
},
|
|
|
|
shared_libs: ["libjpeg"],
|
|
|
|
export_include_dirs: ["include"],
|
|
|
|
apex_available: [
|
|
"//apex_available:platform",
|
|
"com.android.media.swcodec",
|
|
],
|
|
min_sdk_version: "29",
|
|
sanitize: {
|
|
cfi: true,
|
|
config: {
|
|
cfi_assembly_support: true,
|
|
},
|
|
},
|
|
}
|
|
|
|
// compatibilty static library until all uses of libyuv_static are replaced
|
|
// with libyuv (b/37646797)
|
|
cc_library_static {
|
|
name: "libyuv_static",
|
|
vendor_available: true,
|
|
whole_static_libs: ["libyuv"],
|
|
apex_available: [
|
|
"//apex_available:platform",
|
|
"com.android.media.swcodec",
|
|
],
|
|
min_sdk_version: "29",
|
|
}
|
|
|
|
cc_test {
|
|
name: "libyuv_unittest",
|
|
static_libs: ["libyuv"],
|
|
shared_libs: ["libjpeg"],
|
|
cflags: ["-Wall", "-Werror"],
|
|
srcs: [
|
|
"unit_test/basictypes_test.cc",
|
|
"unit_test/color_test.cc",
|
|
"unit_test/compare_test.cc",
|
|
"unit_test/convert_argb_test.cc",
|
|
"unit_test/convert_test.cc",
|
|
"unit_test/cpu_test.cc",
|
|
"unit_test/cpu_thread_test.cc",
|
|
"unit_test/math_test.cc",
|
|
"unit_test/planar_test.cc",
|
|
"unit_test/rotate_argb_test.cc",
|
|
"unit_test/rotate_test.cc",
|
|
"unit_test/scale_argb_test.cc",
|
|
"unit_test/scale_plane_test.cc",
|
|
"unit_test/scale_rgb_test.cc",
|
|
"unit_test/scale_test.cc",
|
|
"unit_test/scale_uv_test.cc",
|
|
"unit_test/unit_test.cc",
|
|
"unit_test/video_common_test.cc",
|
|
],
|
|
}
|
|
|
|
cc_test {
|
|
name: "compare",
|
|
gtest: false,
|
|
srcs: [
|
|
"util/compare.cc",
|
|
],
|
|
static_libs: ["libyuv"],
|
|
}
|
|
|
|
cc_test {
|
|
name: "cpuid",
|
|
gtest: false,
|
|
srcs: [
|
|
"util/cpuid.c",
|
|
],
|
|
static_libs: ["libyuv"],
|
|
}
|
|
|
|
cc_test {
|
|
name: "i444tonv12_eg",
|
|
gtest: false,
|
|
srcs: [
|
|
"util/i444tonv12_eg.cc",
|
|
],
|
|
static_libs: ["libyuv"],
|
|
}
|
|
|
|
cc_test {
|
|
name: "psnr",
|
|
gtest: false,
|
|
srcs: [
|
|
"util/psnr_main.cc",
|
|
"util/psnr.cc",
|
|
"util/ssim.cc",
|
|
],
|
|
static_libs: ["libyuv"],
|
|
}
|
|
|
|
cc_test {
|
|
name: "yuvconstants",
|
|
gtest: false,
|
|
srcs: [
|
|
"util/yuvconstants.c",
|
|
],
|
|
static_libs: ["libyuv"],
|
|
}
|
|
|
|
cc_test {
|
|
name: "yuvconvert",
|
|
gtest: false,
|
|
srcs: [
|
|
"util/yuvconvert.cc",
|
|
],
|
|
static_libs: ["libyuv"],
|
|
shared_libs: ["libjpeg"],
|
|
}
|