RGBScale function using 3 steps: RGB24ToARGB, ARGBScale, ARGBToRGB24

1920x1080 to/from 1280x720 to ARGB on Intel Skylake Xeon
RGBScaleTo1920x1080_Bilinear (2625 ms)
RGBScaleFrom1920x1080_Bilinear (2115 ms)
ARGBScaleTo1920x1080_Bilinear (1668 ms)
ARGBScaleFrom1920x1080_Bilinear (1164 ms)

Bug: b/224814071
Change-Id: Ifc7611b597409771728b13c9c39e5a7e06131021
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3537341
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
This commit is contained in:
Frank Barchard 2022-03-18 18:25:49 -07:00 committed by libyuv LUCI CQ
parent 95b14b2446
commit 124bf08fee
10 changed files with 417 additions and 22 deletions

View File

@ -71,6 +71,7 @@ cc_library {
"source/scale_msa.cc",
"source/scale_neon.cc",
"source/scale_neon64.cc",
"source/scale_rgb.cc",
"source/scale_uv.cc",
"source/video_common.cc",
],
@ -130,6 +131,7 @@ cc_test {
"unit_test/rotate_argb_test.cc",
"unit_test/rotate_test.cc",
"unit_test/scale_argb_test.cc",
"unit_test/scale_rgb_test.cc",
"unit_test/scale_test.cc",
"unit_test/scale_uv_test.cc",
"unit_test/unit_test.cc",

View File

@ -45,6 +45,7 @@ LOCAL_SRC_FILES := \
source/scale_msa.cc \
source/scale_neon.cc \
source/scale_neon64.cc \
source/scale_rgb.cc \
source/scale_uv.cc \
source/scale_win.cc \
source/video_common.cc
@ -97,6 +98,7 @@ LOCAL_SRC_FILES := \
unit_test/rotate_argb_test.cc \
unit_test/rotate_test.cc \
unit_test/scale_argb_test.cc \
unit_test/scale_rgb_test.cc \
unit_test/scale_test.cc \
unit_test/scale_uv_test.cc \
unit_test/unit_test.cc \

View File

@ -94,6 +94,7 @@ static_library("libyuv_internal") {
"include/libyuv/row.h",
"include/libyuv/scale.h",
"include/libyuv/scale_argb.h",
"include/libyuv/scale_rgb.h",
"include/libyuv/scale_row.h",
"include/libyuv/scale_uv.h",
"include/libyuv/version.h",
@ -130,6 +131,7 @@ static_library("libyuv_internal") {
"source/scale_argb.cc",
"source/scale_common.cc",
"source/scale_gcc.cc",
"source/scale_rgb.cc",
"source/scale_uv.cc",
"source/scale_win.cc",
"source/video_common.cc",
@ -258,6 +260,7 @@ if (libyuv_include_tests) {
"unit_test/rotate_argb_test.cc",
"unit_test/rotate_test.cc",
"unit_test/scale_argb_test.cc",
"unit_test/scale_rgb_test.cc",
"unit_test/scale_test.cc",
"unit_test/scale_uv_test.cc",
"unit_test/unit_test.cc",

View File

@ -1,8 +1,9 @@
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 1814
Version: 1815
License: BSD
License File: LICENSE
Description:
libyuv is an open source project that includes YUV conversion and scaling functionality.

View File

@ -0,0 +1,42 @@
/*
* Copyright 2022 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef INCLUDE_LIBYUV_SCALE_RGB_H_
#define INCLUDE_LIBYUV_SCALE_RGB_H_
#include "libyuv/basic_types.h"
#include "libyuv/scale.h" // For FilterMode
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// RGB can be RAW, RGB24 or YUV24
// RGB scales 24 bit images by converting a row at a time to ARGB
// and using ARGB row functions to scale, then convert to RGB.
// TODO(fbarchard): Allow input/output formats to be specified.
LIBYUV_API
int RGBScale(const uint8_t* src_rgb,
int src_stride_rgb,
int src_width,
int src_height,
uint8_t* dst_rgb,
int dst_stride_rgb,
int dst_width,
int dst_height,
enum FilterMode filtering);
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // INCLUDE_LIBYUV_SCALE_UV_H_

View File

@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 1814
#define LIBYUV_VERSION 1815
#endif // INCLUDE_LIBYUV_VERSION_H_

View File

@ -52,6 +52,7 @@ LOCAL_OBJ_FILES := \
source/scale_msa.o \
source/scale_neon.o \
source/scale_neon64.o \
source/scale_rgb.o \
source/scale_uv.o \
source/scale_win.o \
source/video_common.o

View File

@ -2091,7 +2091,7 @@ int RGB24ToJ420(const uint8_t* src_rgb24,
#endif
#endif // HAS_RGB24TOYJROW
{
{
#if !defined(HAS_RGB24TOYJROW)
// Allocate 2 rows of ARGB.
const int kRowSize = (width * 4 + 31) & ~31;
@ -2128,8 +2128,8 @@ int RGB24ToJ420(const uint8_t* src_rgb24,
#if !defined(HAS_RGB24TOYJROW)
free_aligned_buffer_64(row);
#endif
}
return 0;
}
return 0;
}
#undef HAS_RGB24TOYJROW

66
source/scale_rgb.cc Normal file
View File

@ -0,0 +1,66 @@
/*
* Copyright 2022 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/scale.h" /* For FilterMode */
#include <assert.h>
#include <string.h>
#include "libyuv/convert_argb.h"
#include "libyuv/convert_from_argb.h"
#include "libyuv/row.h"
#include "libyuv/scale_argb.h"
#include "libyuv/scale_rgb.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// Scale a 24 bit image.
// Converts to ARGB as intermediate step
LIBYUV_API
int RGBScale(const uint8_t* src_rgb,
int src_stride_rgb,
int src_width,
int src_height,
uint8_t* dst_rgb,
int dst_stride_rgb,
int dst_width,
int dst_height,
enum FilterMode filtering) {
int r;
uint8_t* src_argb =
(uint8_t*)malloc(src_width * src_height * 4 + dst_width * dst_height * 4);
uint8_t* dst_argb = src_argb + src_width * src_height * 4;
if (!src_argb) {
return 1;
}
r = RGB24ToARGB(src_rgb, src_stride_rgb, src_argb, src_width * 4, src_width,
src_height);
if (!r) {
r = ARGBScale(src_argb, src_width * 4, src_width, src_height, dst_argb,
dst_width * 4, dst_width, dst_height, filtering);
if (!r) {
r = ARGBToRGB24(dst_argb, dst_width * 4, dst_rgb, dst_stride_rgb,
dst_width, dst_height);
}
}
free(src_argb);
return r;
}
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif

278
unit_test/scale_rgb_test.cc Normal file
View File

@ -0,0 +1,278 @@
/*
* Copyright 2022 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <stdlib.h>
#include <time.h>
#include "../unit_test/unit_test.h"
#include "libyuv/cpu_id.h"
#include "libyuv/scale_rgb.h"
namespace libyuv {
#define STRINGIZE(line) #line
#define FILELINESTR(file, line) file ":" STRINGIZE(line)
#if !defined(DISABLE_SLOW_TESTS) || defined(__x86_64__) || defined(__i386__)
// SLOW TESTS are those that are unoptimized C code.
// FULL TESTS are optimized but test many variations of the same code.
#define ENABLE_FULL_TESTS
#endif
// Test scaling with C vs Opt and return maximum pixel difference. 0 = exact.
static int RGBTestFilter(int src_width,
int src_height,
int dst_width,
int dst_height,
FilterMode f,
int benchmark_iterations,
int disable_cpu_flags,
int benchmark_cpu_info) {
if (!SizeValid(src_width, src_height, dst_width, dst_height)) {
return 0;
}
int i, j;
const int b = 0; // 128 to test for padding/stride.
int64_t src_rgb_plane_size =
(Abs(src_width) + b * 3) * (Abs(src_height) + b * 3) * 3LL;
int src_stride_rgb = (b * 3 + Abs(src_width)) * 3;
align_buffer_page_end(src_rgb, src_rgb_plane_size);
if (!src_rgb) {
printf("Skipped. Alloc failed " FILELINESTR(__FILE__, __LINE__) "\n");
return 0;
}
MemRandomize(src_rgb, src_rgb_plane_size);
int64_t dst_rgb_plane_size = (dst_width + b * 3) * (dst_height + b * 3) * 3LL;
int dst_stride_rgb = (b * 3 + dst_width) * 3;
align_buffer_page_end(dst_rgb_c, dst_rgb_plane_size);
align_buffer_page_end(dst_rgb_opt, dst_rgb_plane_size);
if (!dst_rgb_c || !dst_rgb_opt) {
printf("Skipped. Alloc failed " FILELINESTR(__FILE__, __LINE__) "\n");
return 0;
}
memset(dst_rgb_c, 2, dst_rgb_plane_size);
memset(dst_rgb_opt, 3, dst_rgb_plane_size);
// Warm up both versions for consistent benchmarks.
MaskCpuFlags(disable_cpu_flags); // Disable all CPU optimization.
RGBScale(src_rgb + (src_stride_rgb * b) + b * 3, src_stride_rgb, src_width,
src_height, dst_rgb_c + (dst_stride_rgb * b) + b * 3, dst_stride_rgb,
dst_width, dst_height, f);
MaskCpuFlags(benchmark_cpu_info); // Enable all CPU optimization.
RGBScale(src_rgb + (src_stride_rgb * b) + b * 3, src_stride_rgb, src_width,
src_height, dst_rgb_opt + (dst_stride_rgb * b) + b * 3, dst_stride_rgb,
dst_width, dst_height, f);
MaskCpuFlags(disable_cpu_flags); // Disable all CPU optimization.
double c_time = get_time();
RGBScale(src_rgb + (src_stride_rgb * b) + b * 3, src_stride_rgb, src_width,
src_height, dst_rgb_c + (dst_stride_rgb * b) + b * 3, dst_stride_rgb,
dst_width, dst_height, f);
c_time = (get_time() - c_time);
MaskCpuFlags(benchmark_cpu_info); // Enable all CPU optimization.
double opt_time = get_time();
for (i = 0; i < benchmark_iterations; ++i) {
RGBScale(src_rgb + (src_stride_rgb * b) + b * 3, src_stride_rgb, src_width,
src_height, dst_rgb_opt + (dst_stride_rgb * b) + b * 3, dst_stride_rgb,
dst_width, dst_height, f);
}
opt_time = (get_time() - opt_time) / benchmark_iterations;
// Report performance of C vs OPT
printf("filter %d - %8d us C - %8d us OPT\n", f,
static_cast<int>(c_time * 1e6), static_cast<int>(opt_time * 1e6));
// C version may be a little off from the optimized. Order of
// operations may introduce rounding somewhere. So do a difference
// of the buffers and look to see that the max difference isn't
// over 2.
int max_diff = 0;
for (i = b; i < (dst_height + b); ++i) {
for (j = b * 3; j < (dst_width + b) * 3; ++j) {
int abs_diff = Abs(dst_rgb_c[(i * dst_stride_rgb) + j] -
dst_rgb_opt[(i * dst_stride_rgb) + j]);
if (abs_diff > max_diff) {
max_diff = abs_diff;
}
}
}
free_aligned_buffer_page_end(dst_rgb_c);
free_aligned_buffer_page_end(dst_rgb_opt);
free_aligned_buffer_page_end(src_rgb);
return max_diff;
}
// The following adjustments in dimensions ensure the scale factor will be
// exactly achieved.
#define DX(x, nom, denom) static_cast<int>((Abs(x) / nom) * nom)
#define SX(x, nom, denom) static_cast<int>((x / nom) * denom)
#define TEST_FACTOR1(name, filter, nom, denom, max_diff) \
TEST_F(LibYUVScaleTest, RGBScaleDownBy##name##_##filter) { \
int diff = RGBTestFilter( \
SX(benchmark_width_, nom, denom), SX(benchmark_height_, nom, denom), \
DX(benchmark_width_, nom, denom), DX(benchmark_height_, nom, denom), \
kFilter##filter, benchmark_iterations_, disable_cpu_flags_, \
benchmark_cpu_info_); \
EXPECT_LE(diff, max_diff); \
}
#if defined(ENABLE_FULL_TESTS)
// Test a scale factor with all 4 filters. Expect unfiltered to be exact, but
// filtering is different fixed point implementations for SSSE3, Neon and C.
#define TEST_FACTOR(name, nom, denom) \
TEST_FACTOR1(name, None, nom, denom, 0) \
TEST_FACTOR1(name, Linear, nom, denom, 3) \
TEST_FACTOR1(name, Bilinear, nom, denom, 3) \
TEST_FACTOR1(name, Box, nom, denom, 3)
#else
// Test a scale factor with Bilinear.
#define TEST_FACTOR(name, nom, denom) \
TEST_FACTOR1(name, Bilinear, nom, denom, 3)
#endif
TEST_FACTOR(2, 1, 2)
TEST_FACTOR(4, 1, 4)
// TEST_FACTOR(8, 1, 8) Disable for benchmark performance.
TEST_FACTOR(3by4, 3, 4)
TEST_FACTOR(3by8, 3, 8)
TEST_FACTOR(3, 1, 3)
#undef TEST_FACTOR1
#undef TEST_FACTOR
#undef SX
#undef DX
#define TEST_SCALETO1(name, width, height, filter, max_diff) \
TEST_F(LibYUVScaleTest, name##To##width##x##height##_##filter) { \
int diff = RGBTestFilter(benchmark_width_, benchmark_height_, width, \
height, kFilter##filter, benchmark_iterations_, \
disable_cpu_flags_, benchmark_cpu_info_); \
EXPECT_LE(diff, max_diff); \
} \
TEST_F(LibYUVScaleTest, name##From##width##x##height##_##filter) { \
int diff = RGBTestFilter(width, height, Abs(benchmark_width_), \
Abs(benchmark_height_), kFilter##filter, \
benchmark_iterations_, disable_cpu_flags_, \
benchmark_cpu_info_); \
EXPECT_LE(diff, max_diff); \
}
#if defined(ENABLE_FULL_TESTS)
/// Test scale to a specified size with all 4 filters.
#define TEST_SCALETO(name, width, height) \
TEST_SCALETO1(name, width, height, None, 0) \
TEST_SCALETO1(name, width, height, Linear, 3) \
TEST_SCALETO1(name, width, height, Bilinear, 3)
#else
#define TEST_SCALETO(name, width, height) \
TEST_SCALETO1(name, width, height, Bilinear, 3)
#endif
TEST_SCALETO(RGBScale, 1, 1)
// TEST_SCALETO(RGBScale, 256, 144) /* 128x72 * 3 */
TEST_SCALETO(RGBScale, 320, 240)
TEST_SCALETO(RGBScale, 569, 480)
TEST_SCALETO(RGBScale, 640, 360)
#ifndef DISABLE_SLOW_TESTS
TEST_SCALETO(RGBScale, 1280, 720)
TEST_SCALETO(RGBScale, 1920, 1080)
#endif // DISABLE_SLOW_TESTS
#undef TEST_SCALETO1
#undef TEST_SCALETO
#define TEST_SCALESWAPXY1(name, filter, max_diff) \
TEST_F(LibYUVScaleTest, name##SwapXY_##filter) { \
int diff = \
RGBTestFilter(benchmark_width_, benchmark_height_, benchmark_height_, \
benchmark_width_, kFilter##filter, benchmark_iterations_, \
disable_cpu_flags_, benchmark_cpu_info_); \
EXPECT_LE(diff, max_diff); \
}
#if defined(ENABLE_FULL_TESTS)
// Test scale with swapped width and height with all 3 filters.
TEST_SCALESWAPXY1(RGBScale, None, 0)
TEST_SCALESWAPXY1(RGBScale, Linear, 0)
TEST_SCALESWAPXY1(RGBScale, Bilinear, 0)
#else
TEST_SCALESWAPXY1(RGBScale, Bilinear, 0)
#endif
#undef TEST_SCALESWAPXY1
TEST_F(LibYUVScaleTest, RGBTest3x) {
const int kSrcStride = 480 * 3;
const int kDstStride = 160 * 3;
const int kSize = kSrcStride * 3;
align_buffer_page_end(orig_pixels, kSize);
for (int i = 0; i < 480 * 3; ++i) {
orig_pixels[i * 3 + 0] = i;
orig_pixels[i * 3 + 1] = 255 - i;
}
align_buffer_page_end(dest_pixels, kDstStride);
int iterations160 = (benchmark_width_ * benchmark_height_ + (160 - 1)) / 160 *
benchmark_iterations_;
for (int i = 0; i < iterations160; ++i) {
RGBScale(orig_pixels, kSrcStride, 480, 3, dest_pixels, kDstStride, 160, 1,
kFilterBilinear);
}
EXPECT_EQ(225, dest_pixels[0]);
EXPECT_EQ(255 - 225, dest_pixels[1]);
RGBScale(orig_pixels, kSrcStride, 480, 3, dest_pixels, kDstStride, 160, 1,
kFilterNone);
EXPECT_EQ(225, dest_pixels[0]);
EXPECT_EQ(255 - 225, dest_pixels[1]);
free_aligned_buffer_page_end(dest_pixels);
free_aligned_buffer_page_end(orig_pixels);
}
TEST_F(LibYUVScaleTest, RGBTest4x) {
const int kSrcStride = 640 * 3;
const int kDstStride = 160 * 3;
const int kSize = kSrcStride * 4;
align_buffer_page_end(orig_pixels, kSize);
for (int i = 0; i < 640 * 4; ++i) {
orig_pixels[i * 3 + 0] = i;
orig_pixels[i * 3 + 1] = 255 - i;
}
align_buffer_page_end(dest_pixels, kDstStride);
int iterations160 = (benchmark_width_ * benchmark_height_ + (160 - 1)) / 160 *
benchmark_iterations_;
for (int i = 0; i < iterations160; ++i) {
RGBScale(orig_pixels, kSrcStride, 640, 4, dest_pixels, kDstStride, 160, 1,
kFilterBilinear);
}
EXPECT_EQ(66, dest_pixels[0]);
EXPECT_EQ(190, dest_pixels[1]);
RGBScale(orig_pixels, kSrcStride, 64, 4, dest_pixels, kDstStride, 16, 1,
kFilterNone);
EXPECT_EQ(2, dest_pixels[0]); // expect the 3rd pixel of the 3rd row
EXPECT_EQ(255 - 2, dest_pixels[1]);
free_aligned_buffer_page_end(dest_pixels);
free_aligned_buffer_page_end(orig_pixels);
}
} // namespace libyuv