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
libyuv/unit_test/compare_test.cc
Frank Barchard 64ddbbcdde [libyuv] Optimize SsimWindowRow with SIMD and use compare_win on x86 
- Route 32-bit x86 (__i386__ / _M_IX86) to compare_win.cc intrinsics
  and restrict compare_gcc.cc to 64-bit x86 (__x86_64__), avoiding
  x87 float rounding and enabling SSE2, AVX2, and AVX-512 on 32-bit
  and 64-bit MinGW, GCC, Clang, and MSan (LIBYUV_ENABLE_ROWWIN).
- Enable AVX-512BW on GCC >= 6 (GCC_HAS_AVX512) and add
  LIBYUV_TARGET_AVX512BW attributes to compare_gcc.cc and
  compare_win.cc.
- Refactor SsimWindowRow_C and Ssim8x8_C to use int32_t and float,
  and add SsimWindowRow_SSE2, SsimWindowRow_AVX2, and
  SsimWindowRow_AVX512BW in both compare_win.cc and compare_gcc.cc.
- Update TestSsimAnyWidth, Ssim, and BenchmarkSsim_Opt in
  compare_test.cc to use EXPECT_NEAR and deterministic test patterns
  (identical, different, similar).
- Update util/compare.cc and linux.mk to compute I420Ssim.

Performance (BenchmarkSsim_Opt, 1280x720):
uArch                        Before      After  Improvement
Intel Lion Cove (x64 AVX2) 120.2 us    68.1 us        1.76x
Intel Skymont (x64 SSE2)   249.6 us   146.8 us        1.70x
AMD Zen 3 (x64 AVX2)       314.9 us   150.2 us        2.10x
Intel Lion Cove (MinGW686) 203.7 us    77.9 us        2.61x
Intel Skymont (MinGW686)   338.8 us   167.5 us        2.02x
AMD Zen 3 (i686 AVX2)      738.4 us   206.5 us        3.58x

Quality (util/compare on 640x360 I420 test images):
Image Comparison                MSE       PSNR         SSIM
Identical (orig vs orig)       0.00  128.00 dB     1.000000
Similar (light +/-3 noise)     4.57   41.53 dB     0.956556
Degraded (16x16 blocky)       98.18   28.21 dB     0.858893
Very different (inverted)  11160.29    7.65 dB     0.000181

Co-authored-by: Martin Storsjö <martin@martin.st>
Bug: b/493910265
Test: out/Release/libyuv_unittest --gtest_filter=*Ssim*
Change-Id: I8603dab3f752af32d07ae9039502ca5da5a05425
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/8618641
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@google.com>
2026-10-20 15:00:52 -07:00

831 lines
26 KiB
C++
Raw Blame History

/*
* Copyright 2011 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 <string.h>
#include <time.h>
#include "../unit_test/unit_test.h"
#include "libyuv/compare.h"
#include "libyuv/cpu_id.h"
#include "libyuv/video_common.h"
#ifdef ENABLE_ROW_TESTS
#include "libyuv/compare_row.h" /* For HammingDistance_C */
#endif
namespace libyuv {
// hash seed of 5381 recommended.
static uint32_t ReferenceHashDjb2(const uint8_t* src,
uint64_t count,
uint32_t seed) {
uint32_t hash = seed;
if (count > 0) {
do {
hash = hash * 33 + *src++;
} while (--count);
}
return hash;
}
TEST_F(LibYUVCompareTest, Djb2_Test) {
const int kMaxTest = benchmark_width_ * benchmark_height_;
align_buffer_page_end(src_a, kMaxTest);
align_buffer_page_end(src_b, kMaxTest);
const char* fox =
"The quick brown fox jumps over the lazy dog"
" and feels as if he were in the seventh heaven of typography"
" together with Hermann Zapf";
uint32_t foxhash = HashDjb2(reinterpret_cast<const uint8_t*>(fox), 131, 5381);
const uint32_t kExpectedFoxHash = 2611006483u;
ASSERT_EQ(kExpectedFoxHash, foxhash);
for (int i = 0; i < kMaxTest; ++i) {
src_a[i] = (fastrand() & 0xff);
src_b[i] = (fastrand() & 0xff);
}
// Compare different buffers. Expect hash is different.
uint32_t h1 = HashDjb2(src_a, kMaxTest, 5381);
uint32_t h2 = HashDjb2(src_b, kMaxTest, 5381);
ASSERT_NE(h1, h2);
// Make last half same. Expect hash is different.
memcpy(src_a + kMaxTest / 2, src_b + kMaxTest / 2, kMaxTest / 2);
h1 = HashDjb2(src_a, kMaxTest, 5381);
h2 = HashDjb2(src_b, kMaxTest, 5381);
ASSERT_NE(h1, h2);
// Make first half same. Expect hash is different.
memcpy(src_a + kMaxTest / 2, src_a, kMaxTest / 2);
memcpy(src_b + kMaxTest / 2, src_b, kMaxTest / 2);
memcpy(src_a, src_b, kMaxTest / 2);
h1 = HashDjb2(src_a, kMaxTest, 5381);
h2 = HashDjb2(src_b, kMaxTest, 5381);
ASSERT_NE(h1, h2);
// Make same. Expect hash is same.
memcpy(src_a, src_b, kMaxTest);
h1 = HashDjb2(src_a, kMaxTest, 5381);
h2 = HashDjb2(src_b, kMaxTest, 5381);
ASSERT_EQ(h1, h2);
// Mask seed different. Expect hash is different.
memcpy(src_a, src_b, kMaxTest);
h1 = HashDjb2(src_a, kMaxTest, 5381);
h2 = HashDjb2(src_b, kMaxTest, 1234);
ASSERT_NE(h1, h2);
// Make one byte different in middle. Expect hash is different.
memcpy(src_a, src_b, kMaxTest);
++src_b[kMaxTest / 2];
h1 = HashDjb2(src_a, kMaxTest, 5381);
h2 = HashDjb2(src_b, kMaxTest, 5381);
ASSERT_NE(h1, h2);
// Make first byte different. Expect hash is different.
memcpy(src_a, src_b, kMaxTest);
++src_b[0];
h1 = HashDjb2(src_a, kMaxTest, 5381);
h2 = HashDjb2(src_b, kMaxTest, 5381);
ASSERT_NE(h1, h2);
// Make last byte different. Expect hash is different.
memcpy(src_a, src_b, kMaxTest);
++src_b[kMaxTest - 1];
h1 = HashDjb2(src_a, kMaxTest, 5381);
h2 = HashDjb2(src_b, kMaxTest, 5381);
ASSERT_NE(h1, h2);
// Make a zeros. Test different lengths. Expect hash is different.
memset(src_a, 0, kMaxTest);
h1 = HashDjb2(src_a, kMaxTest, 5381);
h2 = HashDjb2(src_a, kMaxTest / 2, 5381);
ASSERT_NE(h1, h2);
// Make a zeros and seed of zero. Test different lengths. Expect hash is same.
memset(src_a, 0, kMaxTest);
h1 = HashDjb2(src_a, kMaxTest, 0);
h2 = HashDjb2(src_a, kMaxTest / 2, 0);
ASSERT_EQ(h1, h2);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, BenchmarkDjb2_Opt) {
const int kMaxTest = benchmark_width_ * benchmark_height_;
align_buffer_page_end(src_a, kMaxTest);
for (int i = 0; i < kMaxTest; ++i) {
src_a[i] = i;
}
uint32_t h2 = ReferenceHashDjb2(src_a, kMaxTest, 5381);
uint32_t h1;
for (int i = 0; i < benchmark_iterations_; ++i) {
h1 = HashDjb2(src_a, kMaxTest, 5381);
}
ASSERT_EQ(h1, h2);
free_aligned_buffer_page_end(src_a);
}
TEST_F(LibYUVCompareTest, BenchmarkDjb2_Unaligned) {
const int kMaxTest = benchmark_width_ * benchmark_height_;
align_buffer_page_end(src_a, kMaxTest + 1);
for (int i = 0; i < kMaxTest; ++i) {
src_a[i + 1] = i;
}
uint32_t h2 = ReferenceHashDjb2(src_a + 1, kMaxTest, 5381);
uint32_t h1;
for (int i = 0; i < benchmark_iterations_; ++i) {
h1 = HashDjb2(src_a + 1, kMaxTest, 5381);
}
ASSERT_EQ(h1, h2);
free_aligned_buffer_page_end(src_a);
}
TEST_F(LibYUVCompareTest, BenchmarkARGBDetect_Opt) {
uint32_t fourcc;
const int kMaxTest = benchmark_width_ * benchmark_height_ * 4;
align_buffer_page_end(src_a, kMaxTest);
for (int i = 0; i < kMaxTest; ++i) {
src_a[i] = 255;
}
src_a[0] = 0;
fourcc = ARGBDetect(src_a, benchmark_width_ * 4, benchmark_width_,
benchmark_height_);
ASSERT_EQ(static_cast<uint32_t>(libyuv::FOURCC_BGRA), fourcc);
src_a[0] = 255;
src_a[3] = 0;
fourcc = ARGBDetect(src_a, benchmark_width_ * 4, benchmark_width_,
benchmark_height_);
ASSERT_EQ(static_cast<uint32_t>(libyuv::FOURCC_ARGB), fourcc);
src_a[3] = 255;
for (int i = 0; i < benchmark_iterations_; ++i) {
fourcc = ARGBDetect(src_a, benchmark_width_ * 4, benchmark_width_,
benchmark_height_);
}
ASSERT_EQ(0u, fourcc);
free_aligned_buffer_page_end(src_a);
}
TEST_F(LibYUVCompareTest, BenchmarkARGBDetect_Unaligned) {
uint32_t fourcc;
const int kMaxTest = benchmark_width_ * benchmark_height_ * 4 + 1;
align_buffer_page_end(src_a, kMaxTest);
for (int i = 1; i < kMaxTest; ++i) {
src_a[i] = 255;
}
src_a[0 + 1] = 0;
fourcc = ARGBDetect(src_a + 1, benchmark_width_ * 4, benchmark_width_,
benchmark_height_);
ASSERT_EQ(static_cast<uint32_t>(libyuv::FOURCC_BGRA), fourcc);
src_a[0 + 1] = 255;
src_a[3 + 1] = 0;
fourcc = ARGBDetect(src_a + 1, benchmark_width_ * 4, benchmark_width_,
benchmark_height_);
ASSERT_EQ(static_cast<uint32_t>(libyuv::FOURCC_ARGB), fourcc);
src_a[3 + 1] = 255;
for (int i = 0; i < benchmark_iterations_; ++i) {
fourcc = ARGBDetect(src_a + 1, benchmark_width_ * 4, benchmark_width_,
benchmark_height_);
}
ASSERT_EQ(0u, fourcc);
free_aligned_buffer_page_end(src_a);
}
#ifdef ENABLE_ROW_TESTS
TEST_F(LibYUVCompareTest, BenchmarkHammingDistance_Opt) {
const int kMaxWidth = 4096 * 3;
align_buffer_page_end(src_a, kMaxWidth);
align_buffer_page_end(src_b, kMaxWidth);
memset(src_a, 0, kMaxWidth);
memset(src_b, 0, kMaxWidth);
// Test known value
memcpy(src_a, "test0123test4567", 16);
memcpy(src_b, "tick0123tock4567", 16);
uint32_t h1 = HammingDistance_C(src_a, src_b, 16);
ASSERT_EQ(16u, h1);
// Test C vs OPT on random buffer
MemRandomize(src_a, kMaxWidth);
MemRandomize(src_b, kMaxWidth);
uint32_t h0 = HammingDistance_C(src_a, src_b, kMaxWidth);
int count =
benchmark_iterations_ *
((benchmark_width_ * benchmark_height_ + kMaxWidth - 1) / kMaxWidth);
for (int i = 0; i < count; ++i) {
#if defined(HAS_HAMMINGDISTANCE_NEON)
h1 = HammingDistance_NEON(src_a, src_b, kMaxWidth);
#elif defined(HAS_HAMMINGDISTANCE_AVX2)
int has_avx2 = TestCpuFlag(kCpuHasAVX2);
if (has_avx2) {
h1 = HammingDistance_AVX2(src_a, src_b, kMaxWidth);
} else {
int has_sse42 = TestCpuFlag(kCpuHasSSE42);
if (has_sse42) {
h1 = HammingDistance_SSE42(src_a, src_b, kMaxWidth);
} else {
int has_ssse3 = TestCpuFlag(kCpuHasSSSE3);
if (has_ssse3) {
h1 = HammingDistance_SSSE3(src_a, src_b, kMaxWidth);
} else {
h1 = HammingDistance_C(src_a, src_b, kMaxWidth);
}
}
}
#elif defined(HAS_HAMMINGDISTANCE_SSE42)
int has_sse42 = TestCpuFlag(kCpuHasSSE42);
if (has_sse42) {
h1 = HammingDistance_SSE42(src_a, src_b, kMaxWidth);
} else {
h1 = HammingDistance_C(src_a, src_b, kMaxWidth);
}
#else
h1 = HammingDistance_C(src_a, src_b, kMaxWidth);
#endif
}
ASSERT_EQ(h0, h1);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, BenchmarkHammingDistance_C) {
const int kMaxWidth = 4096 * 3;
align_buffer_page_end(src_a, kMaxWidth);
align_buffer_page_end(src_b, kMaxWidth);
memset(src_a, 0, kMaxWidth);
memset(src_b, 0, kMaxWidth);
// Test known value
memcpy(src_a, "test0123test4567", 16);
memcpy(src_b, "tick0123tock4567", 16);
uint32_t h1 = HammingDistance_C(src_a, src_b, 16);
ASSERT_EQ(16u, h1);
// Test C vs OPT on random buffer
MemRandomize(src_a, kMaxWidth);
MemRandomize(src_b, kMaxWidth);
uint32_t h0 = HammingDistance_C(src_a, src_b, kMaxWidth);
int count =
benchmark_iterations_ *
((benchmark_width_ * benchmark_height_ + kMaxWidth - 1) / kMaxWidth);
for (int i = 0; i < count; ++i) {
h1 = HammingDistance_C(src_a, src_b, kMaxWidth);
}
ASSERT_EQ(h0, h1);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, BenchmarkHammingDistance) {
const int kMaxWidth = 4096 * 3;
align_buffer_page_end(src_a, kMaxWidth);
align_buffer_page_end(src_b, kMaxWidth);
memset(src_a, 0, kMaxWidth);
memset(src_b, 0, kMaxWidth);
memcpy(src_a, "test0123test4567", 16);
memcpy(src_b, "tick0123tock4567", 16);
uint64_t h1 = ComputeHammingDistance(src_a, src_b, 16);
ASSERT_EQ(16u, h1);
// Test C vs OPT on random buffer
MemRandomize(src_a, kMaxWidth);
MemRandomize(src_b, kMaxWidth);
uint32_t h0 = HammingDistance_C(src_a, src_b, kMaxWidth);
int count =
benchmark_iterations_ *
((benchmark_width_ * benchmark_height_ + kMaxWidth - 1) / kMaxWidth);
for (int i = 0; i < count; ++i) {
h1 = ComputeHammingDistance(src_a, src_b, kMaxWidth);
}
ASSERT_EQ(h0, h1);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
// Tests low levels match reference C for specified size.
// The opt implementations have size limitations
// For NEON the counters are 16 bit so the shorts overflow after 65536 bytes.
// So doing one less iteration of the loop is the maximum.
#if defined(HAS_HAMMINGDISTANCE_NEON)
static const int kMaxOptCount = 65536 - 32; // 65504
#else
static const int kMaxOptCount = (1 << (32 - 3)) - 64; // 536870848
#endif
TEST_F(LibYUVCompareTest, TestHammingDistance_Opt) {
uint32_t h1 = 0;
const int kMaxWidth = (benchmark_width_ * benchmark_height_ + 63) & ~63;
align_buffer_page_end(src_a, kMaxWidth);
align_buffer_page_end(src_b, kMaxWidth);
memset(src_a, 255u, kMaxWidth);
memset(src_b, 0u, kMaxWidth);
uint64_t h0 = ComputeHammingDistance(src_a, src_b, kMaxWidth);
ASSERT_EQ(kMaxWidth * 8ULL, h0);
for (int i = 0; i < benchmark_iterations_; ++i) {
#if defined(HAS_HAMMINGDISTANCE_NEON)
h1 = HammingDistance_NEON(src_a, src_b, kMaxWidth);
#elif defined(HAS_HAMMINGDISTANCE_AVX2)
int has_avx2 = TestCpuFlag(kCpuHasAVX2);
if (has_avx2) {
h1 = HammingDistance_AVX2(src_a, src_b, kMaxWidth);
} else {
int has_sse42 = TestCpuFlag(kCpuHasSSE42);
if (has_sse42) {
h1 = HammingDistance_SSE42(src_a, src_b, kMaxWidth);
} else {
int has_ssse3 = TestCpuFlag(kCpuHasSSSE3);
if (has_ssse3) {
h1 = HammingDistance_SSSE3(src_a, src_b, kMaxWidth);
} else {
h1 = HammingDistance_C(src_a, src_b, kMaxWidth);
}
}
}
#elif defined(HAS_HAMMINGDISTANCE_SSE42)
int has_sse42 = TestCpuFlag(kCpuHasSSE42);
if (has_sse42) {
h1 = HammingDistance_SSE42(src_a, src_b, kMaxWidth);
} else {
h1 = HammingDistance_C(src_a, src_b, kMaxWidth);
}
#else
h1 = HammingDistance_C(src_a, src_b, kMaxWidth);
#endif
}
// A large count will cause the low level to potentially overflow so the
// result can not be expected to be correct.
// TODO(fbarchard): Consider expecting the low 16 bits to match.
if (kMaxWidth <= kMaxOptCount) {
ASSERT_EQ(kMaxWidth * 8U, h1);
} else {
if (kMaxWidth * 8ULL != static_cast<uint64_t>(h1)) {
printf(
"warning - HammingDistance_Opt %u does not match %llu "
"but length of %u is longer than guaranteed.\n",
h1, kMaxWidth * 8ULL, kMaxWidth);
} else {
printf(
"warning - HammingDistance_Opt %u matches but length of %u "
"is longer than guaranteed.\n",
h1, kMaxWidth);
}
}
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
#endif // ENABLE_ROW_TESTS
TEST_F(LibYUVCompareTest, TestHammingDistance) {
align_buffer_page_end(src_a, benchmark_width_ * benchmark_height_);
align_buffer_page_end(src_b, benchmark_width_ * benchmark_height_);
memset(src_a, 255u, benchmark_width_ * benchmark_height_);
memset(src_b, 0, benchmark_width_ * benchmark_height_);
uint64_t h1 = 0;
for (int i = 0; i < benchmark_iterations_; ++i) {
h1 = ComputeHammingDistance(src_a, src_b,
benchmark_width_ * benchmark_height_);
}
ASSERT_EQ(benchmark_width_ * benchmark_height_ * 8ULL, h1);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, BenchmarkSumSquareError_Opt) {
const int kMaxWidth = 4096 * 3;
align_buffer_page_end(src_a, kMaxWidth);
align_buffer_page_end(src_b, kMaxWidth);
memset(src_a, 0, kMaxWidth);
memset(src_b, 0, kMaxWidth);
memcpy(src_a, "test0123test4567", 16);
memcpy(src_b, "tick0123tock4567", 16);
uint64_t h1 = ComputeSumSquareError(src_a, src_b, 16);
ASSERT_EQ(790u, h1);
for (int i = 0; i < kMaxWidth; ++i) {
src_a[i] = i;
src_b[i] = i;
}
memset(src_a, 0, kMaxWidth);
memset(src_b, 0, kMaxWidth);
int count =
benchmark_iterations_ *
((benchmark_width_ * benchmark_height_ + kMaxWidth - 1) / kMaxWidth);
for (int i = 0; i < count; ++i) {
h1 = ComputeSumSquareError(src_a, src_b, kMaxWidth);
}
ASSERT_EQ(0u, h1);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, SumSquareError) {
const int kMaxWidth = 4096 * 3;
align_buffer_page_end(src_a, kMaxWidth);
align_buffer_page_end(src_b, kMaxWidth);
memset(src_a, 0, kMaxWidth);
memset(src_b, 0, kMaxWidth);
uint64_t err;
err = ComputeSumSquareError(src_a, src_b, kMaxWidth);
ASSERT_EQ(0u, err);
memset(src_a, 1, kMaxWidth);
err = ComputeSumSquareError(src_a, src_b, kMaxWidth);
ASSERT_EQ(static_cast<int>(err), kMaxWidth);
memset(src_a, 190, kMaxWidth);
memset(src_b, 193, kMaxWidth);
err = ComputeSumSquareError(src_a, src_b, kMaxWidth);
ASSERT_EQ(static_cast<int>(err), kMaxWidth * 3 * 3);
for (int i = 0; i < kMaxWidth; ++i) {
src_a[i] = (fastrand() & 0xff);
src_b[i] = (fastrand() & 0xff);
}
MaskCpuFlags(disable_cpu_flags_);
uint64_t c_err = ComputeSumSquareError(src_a, src_b, kMaxWidth);
MaskCpuFlags(benchmark_cpu_info_);
uint64_t opt_err = ComputeSumSquareError(src_a, src_b, kMaxWidth);
ASSERT_EQ(c_err, opt_err);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
// Test every count, aligned and unaligned, ending at the end of a page.
TEST_F(LibYUVCompareTest, TestCompareAnyCount) {
const int kMaxCount = 64 * 5 + 1;
align_buffer_page_end(src_a, kMaxCount + 1);
align_buffer_page_end(src_b, kMaxCount + 1);
MemRandomize(src_a, kMaxCount + 1);
MemRandomize(src_b, kMaxCount + 1);
for (int count = 0; count <= kMaxCount; ++count) {
for (int offset = 0; offset < 2; ++offset) {
const uint8_t* a = src_a + kMaxCount + 1 - count - offset;
const uint8_t* b = src_b + kMaxCount + 1 - count - offset;
MaskCpuFlags(disable_cpu_flags_);
uint32_t c_hash = HashDjb2(a, count, 5381);
uint64_t c_diff = ComputeHammingDistance(a, b, count);
uint64_t c_err = ComputeSumSquareError(a, b, count);
MaskCpuFlags(benchmark_cpu_info_);
uint32_t opt_hash = HashDjb2(a, count, 5381);
uint64_t opt_diff = ComputeHammingDistance(a, b, count);
uint64_t opt_err = ComputeSumSquareError(a, b, count);
EXPECT_EQ(c_hash, opt_hash) << "count " << count << " offset " << offset;
EXPECT_EQ(c_diff, opt_diff) << "count " << count << " offset " << offset;
EXPECT_EQ(c_err, opt_err) << "count " << count << " offset " << offset;
}
}
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, BenchmarkPsnr_Opt) {
align_buffer_page_end(src_a, benchmark_width_ * benchmark_height_);
align_buffer_page_end(src_b, benchmark_width_ * benchmark_height_);
for (int i = 0; i < benchmark_width_ * benchmark_height_; ++i) {
src_a[i] = i;
src_b[i] = i;
}
MaskCpuFlags(benchmark_cpu_info_);
double opt_time = get_time();
for (int i = 0; i < benchmark_iterations_; ++i) {
CalcFramePsnr(src_a, benchmark_width_, src_b, benchmark_width_,
benchmark_width_, benchmark_height_);
}
opt_time = (get_time() - opt_time) / benchmark_iterations_;
printf("BenchmarkPsnr_Opt - %8.2f us opt\n", opt_time * 1e6);
ASSERT_EQ(0, 0);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, BenchmarkPsnr_Unaligned) {
align_buffer_page_end(src_a, benchmark_width_ * benchmark_height_ + 1);
align_buffer_page_end(src_b, benchmark_width_ * benchmark_height_);
for (int i = 0; i < benchmark_width_ * benchmark_height_; ++i) {
src_a[i + 1] = i;
src_b[i] = i;
}
MaskCpuFlags(benchmark_cpu_info_);
double opt_time = get_time();
for (int i = 0; i < benchmark_iterations_; ++i) {
CalcFramePsnr(src_a + 1, benchmark_width_, src_b, benchmark_width_,
benchmark_width_, benchmark_height_);
}
opt_time = (get_time() - opt_time) / benchmark_iterations_;
printf("BenchmarkPsnr_Opt - %8.2f us opt\n", opt_time * 1e6);
ASSERT_EQ(0, 0);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, Psnr) {
const int kSrcWidth = benchmark_width_;
const int kSrcHeight = benchmark_height_;
const int b = 128;
const int kSrcPlaneSize = (kSrcWidth + b * 2) * (kSrcHeight + b * 2);
const int kSrcStride = 2 * b + kSrcWidth;
align_buffer_page_end(src_a, kSrcPlaneSize);
align_buffer_page_end(src_b, kSrcPlaneSize);
memset(src_a, 0, kSrcPlaneSize);
memset(src_b, 0, kSrcPlaneSize);
double err;
err = CalcFramePsnr(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
ASSERT_EQ(err, kMaxPsnr);
memset(src_a, 255, kSrcPlaneSize);
err = CalcFramePsnr(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
ASSERT_EQ(err, 0.0);
memset(src_a, 1, kSrcPlaneSize);
err = CalcFramePsnr(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
ASSERT_GT(err, 48.0);
ASSERT_LT(err, 49.0);
for (int i = 0; i < kSrcPlaneSize; ++i) {
src_a[i] = i;
}
err = CalcFramePsnr(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
ASSERT_GT(err, 2.0);
if (kSrcWidth * kSrcHeight >= 256) {
ASSERT_LT(err, 6.0);
}
memset(src_a, 0, kSrcPlaneSize);
memset(src_b, 0, kSrcPlaneSize);
for (int i = b; i < (kSrcHeight + b); ++i) {
for (int j = b; j < (kSrcWidth + b); ++j) {
src_a[(i * kSrcStride) + j] = (fastrand() & 0xff);
src_b[(i * kSrcStride) + j] = (fastrand() & 0xff);
}
}
MaskCpuFlags(disable_cpu_flags_);
double c_err, opt_err;
c_err = CalcFramePsnr(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
MaskCpuFlags(benchmark_cpu_info_);
opt_err = CalcFramePsnr(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
ASSERT_EQ(opt_err, c_err);
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, BenchmarkSsim_Opt) {
align_buffer_page_end(src_a, benchmark_width_ * benchmark_height_);
align_buffer_page_end(src_b, benchmark_width_ * benchmark_height_);
for (int i = 0; i < benchmark_width_ * benchmark_height_; ++i) {
src_a[i] = i;
src_b[i] = i;
}
MaskCpuFlags(benchmark_cpu_info_);
double ssim = 0.0;
for (int i = 0; i < benchmark_iterations_; ++i) {
ssim = CalcFrameSsim(src_a, benchmark_width_, src_b, benchmark_width_,
benchmark_width_, benchmark_height_);
}
if (benchmark_width_ > 8 && benchmark_height_ > 8) {
EXPECT_EQ(1.0, ssim);
} else {
EXPECT_EQ(0.0, ssim);
}
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, BenchmarkSsim_Unaligned) {
align_buffer_page_end(src_a, benchmark_width_ * benchmark_height_ + 1);
align_buffer_page_end(src_b, benchmark_width_ * benchmark_height_);
for (int i = 0; i < benchmark_width_ * benchmark_height_; ++i) {
src_a[i + 1] = i;
src_b[i] = i;
}
MaskCpuFlags(benchmark_cpu_info_);
double ssim = 0.0;
for (int i = 0; i < benchmark_iterations_; ++i) {
ssim = CalcFrameSsim(src_a + 1, benchmark_width_, src_b, benchmark_width_,
benchmark_width_, benchmark_height_);
}
if (benchmark_width_ > 8 && benchmark_height_ > 8) {
EXPECT_EQ(1.0, ssim);
} else {
EXPECT_EQ(0.0, ssim);
}
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, TestSsimAnyWidth) {
const int kMaxWidth = 140;
const int kHeight = 16;
const int kPlaneSize = kMaxWidth * kHeight + 1;
align_buffer_page_end(src_a, kPlaneSize);
align_buffer_page_end(src_b, kPlaneSize);
for (int i = 0; i < kPlaneSize; ++i) {
src_a[i] = (i * 73 + 19) & 0xff;
src_b[i] = (i * 151 + 43) & 0xff;
}
for (int width = 9; width <= kMaxWidth; ++width) {
for (int offset = 0; offset < 2; ++offset) {
const int size = width * kHeight;
const uint8_t* a = src_a + kPlaneSize - size - offset;
const uint8_t* b = src_b + kPlaneSize - size - offset;
MaskCpuFlags(disable_cpu_flags_);
double c_ssim = CalcFrameSsim(a, width, b, width, width, kHeight);
MaskCpuFlags(benchmark_cpu_info_);
double opt_ssim = CalcFrameSsim(a, width, b, width, width, kHeight);
EXPECT_NEAR(c_ssim, opt_ssim, 1e-5)
<< "width " << width << " offset " << offset;
}
}
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
TEST_F(LibYUVCompareTest, Ssim) {
const int kSrcWidth = benchmark_width_;
const int kSrcHeight = benchmark_height_;
const int b = 128;
const int kSrcPlaneSize = (kSrcWidth + b * 2) * (kSrcHeight + b * 2);
const int kSrcStride = 2 * b + kSrcWidth;
align_buffer_page_end(src_a, kSrcPlaneSize);
align_buffer_page_end(src_b, kSrcPlaneSize);
memset(src_a, 0, kSrcPlaneSize);
memset(src_b, 0, kSrcPlaneSize);
if (kSrcWidth <= 8 || kSrcHeight <= 8) {
printf("warning - Ssim size too small. Testing function executes.\n");
}
// Identical images: SSIM == 1.0.
double err;
err = CalcFrameSsim(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
if (kSrcWidth > 8 && kSrcHeight > 8) {
EXPECT_EQ(err, 1.0);
}
// Completely different constant images (255 vs 0): SSIM near 0.
memset(src_a, 255, kSrcPlaneSize);
err = CalcFrameSsim(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
if (kSrcWidth > 8 && kSrcHeight > 8) {
EXPECT_LT(err, 0.0001);
}
// Slightly different constant images (1 vs 0): SSIM ~ 0.84.
memset(src_a, 1, kSrcPlaneSize);
err = CalcFrameSsim(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
if (kSrcWidth > 8 && kSrcHeight > 8) {
EXPECT_GT(err, 0.8);
EXPECT_LT(err, 0.9);
}
// Gradient vs constant 0: SSIM near 0.
for (int i = 0; i < kSrcPlaneSize; ++i) {
src_a[i] = i;
}
err = CalcFrameSsim(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride, kSrcWidth,
kSrcHeight);
if (kSrcWidth > 8 && kSrcHeight > 8) {
EXPECT_GT(err, 0.0);
EXPECT_LT(err, 0.01);
}
// Similar images (deterministic gradient + small perturbation): SSIM ~ 0.99.
for (int i = b; i < (kSrcHeight + b); ++i) {
for (int j = b; j < (kSrcWidth + b); ++j) {
uint8_t val = static_cast<uint8_t>((i * 3 + j * 5) & 0xff);
src_a[(i * kSrcStride) + j] = val;
src_b[(i * kSrcStride) + j] =
static_cast<uint8_t>((val + ((i ^ j) & 3)) & 0xff);
}
}
MaskCpuFlags(disable_cpu_flags_);
double c_err = CalcFrameSsim(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride,
kSrcWidth, kSrcHeight);
MaskCpuFlags(benchmark_cpu_info_);
double opt_err = CalcFrameSsim(src_a + kSrcStride * b + b, kSrcStride,
src_b + kSrcStride * b + b, kSrcStride,
kSrcWidth, kSrcHeight);
if (kSrcWidth > 8 && kSrcHeight > 8) {
EXPECT_GT(opt_err, 0.9);
EXPECT_LT(opt_err, 1.0);
EXPECT_NEAR(c_err, opt_err, 1e-5);
}
free_aligned_buffer_page_end(src_a);
free_aligned_buffer_page_end(src_b);
}
} // namespace libyuv
Reference in New Issue View Git Blame Copy Permalink