diff options
author | Frank Barchard <fbarchard@google.com> | 2023-12-09 09:28:22 -0800 |
---|---|---|
committer | libyuv LUCI CQ <libyuv-scoped@luci-project-accounts.iam.gserviceaccount.com> | 2023-12-09 18:39:41 +0000 |
commit | 6dc03dacbf7f5e2bc2b1f54fb20dc3ae4c78c100 (patch) | |
tree | 033e243be972c12939f8af18cc5df5fdbb3763e2 | |
parent | 9e61d7f9c1535e324618f0e508a775f4245c5e82 (diff) | |
download | libyuv-6dc03dacbf7f5e2bc2b1f54fb20dc3ae4c78c100.tar.gz |
Split scale_test and scale_plane_test to allow building on small devices
Bug: libyuv:956
Change-Id: I1903aa616243e891440ed92836dfb0992d31d4cd
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/5107257
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Wan-Teh Chang <wtc@google.com>
-rw-r--r-- | Android.bp | 1 | ||||
-rw-r--r-- | Android.mk | 1 | ||||
-rw-r--r-- | BUILD.gn | 1 | ||||
-rw-r--r-- | README.chromium | 2 | ||||
-rw-r--r-- | include/libyuv/version.h | 2 | ||||
-rw-r--r-- | unit_test/convert_argb_test.cc | 490 | ||||
-rw-r--r-- | unit_test/convert_test.cc | 11 | ||||
-rw-r--r-- | unit_test/scale_plane_test.cc | 470 | ||||
-rw-r--r-- | unit_test/scale_test.cc | 436 |
9 files changed, 765 insertions, 649 deletions
@@ -133,6 +133,7 @@ cc_test { "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", @@ -99,6 +99,7 @@ LOCAL_SRC_FILES := \ 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 \ @@ -324,6 +324,7 @@ if (libyuv_include_tests) { "unit_test/rotate_argb_test.cc", "unit_test/rotate_test.cc", "unit_test/scale_argb_test.cc", + "unit_test/scale_plane_test.cc", "unit_test/scale_rgb_test.cc", "unit_test/scale_test.cc", "unit_test/scale_uv_test.cc", diff --git a/README.chromium b/README.chromium index 043a63e5..069dfc2c 100644 --- a/README.chromium +++ b/README.chromium @@ -1,6 +1,6 @@ Name: libyuv URL: https://chromium.googlesource.com/libyuv/libyuv/ -Version: 1881 +Version: 1882 License: BSD License File: LICENSE Shipped: yes diff --git a/include/libyuv/version.h b/include/libyuv/version.h index 7d2d95e2..b477cb75 100644 --- a/include/libyuv/version.h +++ b/include/libyuv/version.h @@ -11,6 +11,6 @@ #ifndef INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_ -#define LIBYUV_VERSION 1881 +#define LIBYUV_VERSION 1882 #endif // INCLUDE_LIBYUV_VERSION_H_ diff --git a/unit_test/convert_argb_test.cc b/unit_test/convert_argb_test.cc index dea8a2fb..90451db7 100644 --- a/unit_test/convert_argb_test.cc +++ b/unit_test/convert_argb_test.cc @@ -1,5 +1,5 @@ /* - * Copyright 2011 The LibYuv Project Authors. All rights reserved. + * Copyright 2023 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 @@ -31,6 +31,13 @@ #include "libyuv/row.h" /* For ARGBToAR30Row_AVX2 */ #endif +#if defined(__riscv) && !defined(__clang__) +#define DISABLE_SLOW_TESTS +#undef ENABLE_FULL_TESTS +#undef ENABLE_ROW_TESTS +#define LEAN_TESTS +#endif + // Some functions fail on big endian. Enable these tests on all cpus except // PowerPC, but they are not optimized so disabled by default. #if !defined(DISABLE_SLOW_TESTS) && !defined(__powerpc__) @@ -136,6 +143,7 @@ namespace libyuv { free_aligned_buffer_page_end(src_uv); \ } +#if defined(ENABLE_FULL_TESTS) #define TESTBPTOP(SRC_FMT_PLANAR, SRC_T, SRC_BPC, SRC_SUBSAMP_X, \ SRC_SUBSAMP_Y, FMT_PLANAR, DST_T, DST_BPC, DST_SUBSAMP_X, \ DST_SUBSAMP_Y, SRC_DEPTH, TILE_WIDTH, TILE_HEIGHT) \ @@ -154,6 +162,14 @@ namespace libyuv { TESTBPTOPI(SRC_FMT_PLANAR, SRC_T, SRC_BPC, SRC_SUBSAMP_X, SRC_SUBSAMP_Y, \ FMT_PLANAR, DST_T, DST_BPC, DST_SUBSAMP_X, DST_SUBSAMP_Y, \ benchmark_width_, _Opt, +, 0, SRC_DEPTH, TILE_WIDTH, TILE_HEIGHT) +#else +#define TESTBPTOP(SRC_FMT_PLANAR, SRC_T, SRC_BPC, SRC_SUBSAMP_X, \ + SRC_SUBSAMP_Y, FMT_PLANAR, DST_T, DST_BPC, DST_SUBSAMP_X, \ + DST_SUBSAMP_Y, SRC_DEPTH, TILE_WIDTH, TILE_HEIGHT) \ + TESTBPTOPI(SRC_FMT_PLANAR, SRC_T, SRC_BPC, SRC_SUBSAMP_X, SRC_SUBSAMP_Y, \ + FMT_PLANAR, DST_T, DST_BPC, DST_SUBSAMP_X, DST_SUBSAMP_Y, \ + benchmark_width_, _Opt, +, 0, SRC_DEPTH, TILE_WIDTH, TILE_HEIGHT) +#endif TESTBPTOP(NV12, uint8_t, 1, 2, 2, I420, uint8_t, 1, 2, 2, 8, 1, 1) TESTBPTOP(NV21, uint8_t, 1, 2, 2, I420, uint8_t, 1, 2, 2, 8, 1, 1) @@ -265,8 +281,6 @@ TESTBPTOP(P012, uint16_t, 2, 2, 2, I012, uint16_t, 2, 2, 2, 12, 1, 1) #define TESTPLANARTOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ YALIGN) \ TESTPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN, benchmark_width_ + 1, _Any, +, 0) \ - TESTPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ YALIGN, benchmark_width_, _Opt, +, 0) #endif @@ -341,7 +355,7 @@ TESTPLANARTOB(I420, 2, 2, ARGBFilter, 4, 4, 1) TESTPLANARTOB(I422, 2, 1, ARGBFilter, 4, 4, 1) TESTPLANARTOB(I420, 2, 2, RGB24Filter, 3, 3, 1) TESTPLANARTOB(I422, 2, 2, RGB24Filter, 3, 3, 1) -#else +#else // FULL_TESTS TESTPLANARTOB(I420, 2, 2, ABGR, 4, 4, 1) TESTPLANARTOB(I420, 2, 2, ARGB, 4, 4, 1) TESTPLANARTOB(I420, 2, 2, BGRA, 4, 4, 1) @@ -370,218 +384,6 @@ TESTPLANARTOB(I444, 1, 1, ABGR, 4, 4, 1) TESTPLANARTOB(I444, 1, 1, ARGB, 4, 4, 1) #endif -#define TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN, W1280, N, NEG, OFF, ATTEN) \ - TEST_F(LibYUVConvertTest, FMT_PLANAR##To##FMT_B##N) { \ - const int kWidth = W1280; \ - const int kHeight = ALIGNINT(benchmark_height_, YALIGN); \ - const int kStrideB = ALIGNINT(kWidth * BPP_B, ALIGN); \ - const int kStrideUV = SUBSAMPLE(kWidth, SUBSAMP_X); \ - const int kSizeUV = kStrideUV * SUBSAMPLE(kHeight, SUBSAMP_Y); \ - align_buffer_page_end(src_y, kWidth* kHeight + OFF); \ - align_buffer_page_end(src_u, kSizeUV + OFF); \ - align_buffer_page_end(src_v, kSizeUV + OFF); \ - align_buffer_page_end(src_a, kWidth* kHeight + OFF); \ - align_buffer_page_end(dst_argb_c, kStrideB* kHeight + OFF); \ - align_buffer_page_end(dst_argb_opt, kStrideB* kHeight + OFF); \ - for (int i = 0; i < kWidth * kHeight; ++i) { \ - src_y[i + OFF] = (fastrand() & 0xff); \ - src_a[i + OFF] = (fastrand() & 0xff); \ - } \ - for (int i = 0; i < kSizeUV; ++i) { \ - src_u[i + OFF] = (fastrand() & 0xff); \ - src_v[i + OFF] = (fastrand() & 0xff); \ - } \ - memset(dst_argb_c + OFF, 1, kStrideB * kHeight); \ - memset(dst_argb_opt + OFF, 101, kStrideB * kHeight); \ - MaskCpuFlags(disable_cpu_flags_); \ - FMT_PLANAR##To##FMT_B(src_y + OFF, kWidth, src_u + OFF, kStrideUV, \ - src_v + OFF, kStrideUV, src_a + OFF, kWidth, \ - dst_argb_c + OFF, kStrideB, kWidth, NEG kHeight, \ - ATTEN); \ - MaskCpuFlags(benchmark_cpu_info_); \ - for (int i = 0; i < benchmark_iterations_; ++i) { \ - FMT_PLANAR##To##FMT_B(src_y + OFF, kWidth, src_u + OFF, kStrideUV, \ - src_v + OFF, kStrideUV, src_a + OFF, kWidth, \ - dst_argb_opt + OFF, kStrideB, kWidth, NEG kHeight, \ - ATTEN); \ - } \ - for (int i = 0; i < kWidth * BPP_B * kHeight; ++i) { \ - EXPECT_EQ(dst_argb_c[i + OFF], dst_argb_opt[i + OFF]); \ - } \ - free_aligned_buffer_page_end(src_y); \ - free_aligned_buffer_page_end(src_u); \ - free_aligned_buffer_page_end(src_v); \ - free_aligned_buffer_page_end(src_a); \ - free_aligned_buffer_page_end(dst_argb_c); \ - free_aligned_buffer_page_end(dst_argb_opt); \ - } - -#if defined(ENABLE_FULL_TESTS) -#define TESTQPLANARTOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN) \ - TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN, benchmark_width_ + 1, _Any, +, 0, 0) \ - TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN, benchmark_width_, _Unaligned, +, 2, 0) \ - TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN, benchmark_width_, _Invert, -, 0, 0) \ - TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN, benchmark_width_, _Opt, +, 0, 0) \ - TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN, benchmark_width_, _Premult, +, 0, 1) -#else -#define TESTQPLANARTOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN) \ - TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ - YALIGN, benchmark_width_, _Opt, +, 0, 0) -#endif - -#define J420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ - l, m) -#define J420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ - l, m) -#define F420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ - l, m) -#define F420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ - l, m) -#define H420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ - l, m) -#define H420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ - l, m) -#define U420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ - l, m) -#define U420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ - l, m) -#define V420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ - l, m) -#define V420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ - l, m) -#define J422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ - l, m) -#define J422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ - l, m) -#define F422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ - l, m) -#define F422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ - l, m) -#define H422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ - l, m) -#define H422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ - l, m) -#define U422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ - l, m) -#define U422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ - l, m) -#define V422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ - l, m) -#define V422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ - l, m) -#define J444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ - l, m) -#define J444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ - l, m) -#define F444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ - l, m) -#define F444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ - l, m) -#define H444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ - l, m) -#define H444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ - l, m) -#define U444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ - l, m) -#define U444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ - l, m) -#define V444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ - l, m) -#define V444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ - l, m) - -#define I420AlphaToARGBFilter(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I420AlphaToARGBMatrixFilter(a, b, c, d, e, f, g, h, i, j, \ - &kYuvI601Constants, k, l, m, kFilterBilinear) -#define I422AlphaToARGBFilter(a, b, c, d, e, f, g, h, i, j, k, l, m) \ - I422AlphaToARGBMatrixFilter(a, b, c, d, e, f, g, h, i, j, \ - &kYuvI601Constants, k, l, m, kFilterBilinear) - -#if defined(ENABLE_FULL_TESTS) -TESTQPLANARTOB(I420Alpha, 2, 2, ARGB, 4, 4, 1) -TESTQPLANARTOB(I420Alpha, 2, 2, ABGR, 4, 4, 1) -TESTQPLANARTOB(J420Alpha, 2, 2, ARGB, 4, 4, 1) -TESTQPLANARTOB(J420Alpha, 2, 2, ABGR, 4, 4, 1) -TESTQPLANARTOB(H420Alpha, 2, 2, ARGB, 4, 4, 1) -TESTQPLANARTOB(H420Alpha, 2, 2, ABGR, 4, 4, 1) -TESTQPLANARTOB(F420Alpha, 2, 2, ARGB, 4, 4, 1) -TESTQPLANARTOB(F420Alpha, 2, 2, ABGR, 4, 4, 1) -TESTQPLANARTOB(U420Alpha, 2, 2, ARGB, 4, 4, 1) -TESTQPLANARTOB(U420Alpha, 2, 2, ABGR, 4, 4, 1) -TESTQPLANARTOB(V420Alpha, 2, 2, ARGB, 4, 4, 1) -TESTQPLANARTOB(V420Alpha, 2, 2, ABGR, 4, 4, 1) -TESTQPLANARTOB(I422Alpha, 2, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(I422Alpha, 2, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(J422Alpha, 2, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(J422Alpha, 2, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(H422Alpha, 2, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(H422Alpha, 2, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(F422Alpha, 2, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(F422Alpha, 2, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(U422Alpha, 2, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(U422Alpha, 2, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(V422Alpha, 2, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(V422Alpha, 2, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(I444Alpha, 1, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(I444Alpha, 1, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(J444Alpha, 1, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(J444Alpha, 1, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(H444Alpha, 1, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(H444Alpha, 1, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(F444Alpha, 1, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(F444Alpha, 1, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(U444Alpha, 1, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(U444Alpha, 1, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(V444Alpha, 1, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(V444Alpha, 1, 1, ABGR, 4, 4, 1) -TESTQPLANARTOB(I420Alpha, 2, 2, ARGBFilter, 4, 4, 1) -TESTQPLANARTOB(I422Alpha, 2, 1, ARGBFilter, 4, 4, 1) -#else -TESTQPLANARTOB(I420Alpha, 2, 2, ARGB, 4, 4, 1) -TESTQPLANARTOB(I422Alpha, 2, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(I444Alpha, 1, 1, ARGB, 4, 4, 1) -TESTQPLANARTOB(I420Alpha, 2, 2, ARGBFilter, 4, 4, 1) -TESTQPLANARTOB(I422Alpha, 2, 1, ARGBFilter, 4, 4, 1) -#endif - #define TESTBPTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, FMT_C, BPP_B, \ W1280, N, NEG, OFF) \ TEST_F(LibYUVConvertTest, FMT_PLANAR##To##FMT_B##N) { \ @@ -636,6 +438,7 @@ TESTQPLANARTOB(I422Alpha, 2, 1, ARGBFilter, 4, 4, 1) free_aligned_buffer_page_end(dst_argb32_opt); \ } +#if defined(ENABLE_FULL_TESTS) #define TESTBPTOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, FMT_C, BPP_B) \ TESTBPTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, FMT_C, BPP_B, \ benchmark_width_ + 1, _Any, +, 0) \ @@ -645,6 +448,11 @@ TESTQPLANARTOB(I422Alpha, 2, 1, ARGBFilter, 4, 4, 1) benchmark_width_, _Invert, -, 0) \ TESTBPTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, FMT_C, BPP_B, \ benchmark_width_, _Opt, +, 0) +#else +#define TESTBPTOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, FMT_C, BPP_B) \ + TESTBPTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, FMT_C, BPP_B, \ + benchmark_width_, _Opt, +, 0) +#endif #define JNV12ToARGB(a, b, c, d, e, f, g, h) \ NV12ToARGBMatrix(a, b, c, d, e, f, &kYuvJPEGConstants, g, h) @@ -975,6 +783,7 @@ TESTATOA(AB64, uint16_t, 4, 4, 1, ABGR, uint8_t, 4, 4, 1) TESTATOA(AR64, uint16_t, 4, 4, 1, AB64, uint16_t, 4, 4, 1) TESTATOA(AB64, uint16_t, 4, 4, 1, AR64, uint16_t, 4, 4, 1) + #define TESTATOBDI(FMT_A, BPP_A, STRIDE_A, HEIGHT_A, FMT_B, BPP_B, STRIDE_B, \ HEIGHT_B, W1280, N, NEG, OFF) \ TEST_F(LibYUVConvertTest, FMT_A##To##FMT_B##Dither##N) { \ @@ -1045,6 +854,7 @@ TESTATOA(AB64, uint16_t, 4, 4, 1, AR64, uint16_t, 4, 4, 1) } \ } +#if defined(ENABLE_FULL_TESTS) #define TESTATOBD(FMT_A, BPP_A, STRIDE_A, HEIGHT_A, FMT_B, BPP_B, STRIDE_B, \ HEIGHT_B) \ TESTATOBDI(FMT_A, BPP_A, STRIDE_A, HEIGHT_A, FMT_B, BPP_B, STRIDE_B, \ @@ -1057,6 +867,12 @@ TESTATOA(AB64, uint16_t, 4, 4, 1, AR64, uint16_t, 4, 4, 1) HEIGHT_B, benchmark_width_, _Opt, +, 0) \ TESTATOBDRANDOM(FMT_A, BPP_A, STRIDE_A, HEIGHT_A, FMT_B, BPP_B, STRIDE_B, \ HEIGHT_B) +#else +#define TESTATOBD(FMT_A, BPP_A, STRIDE_A, HEIGHT_A, FMT_B, BPP_B, STRIDE_B, \ + HEIGHT_B) \ + TESTATOBDRANDOM(FMT_A, BPP_A, STRIDE_A, HEIGHT_A, FMT_B, BPP_B, STRIDE_B, \ + HEIGHT_B) +#endif #ifdef LITTLE_ENDIAN_ONLY_TEST TESTATOBD(ARGB, 4, 4, 1, RGB565, 2, 2, 1) @@ -1125,6 +941,218 @@ TESTEND(BGRAToARGB, uint8_t, 4, 4, 1) TESTEND(ABGRToARGB, uint8_t, 4, 4, 1) TESTEND(AB64ToAR64, uint16_t, 4, 4, 1) +#define TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, W1280, N, NEG, OFF, ATTEN) \ + TEST_F(LibYUVConvertTest, FMT_PLANAR##To##FMT_B##N) { \ + const int kWidth = W1280; \ + const int kHeight = ALIGNINT(benchmark_height_, YALIGN); \ + const int kStrideB = ALIGNINT(kWidth * BPP_B, ALIGN); \ + const int kStrideUV = SUBSAMPLE(kWidth, SUBSAMP_X); \ + const int kSizeUV = kStrideUV * SUBSAMPLE(kHeight, SUBSAMP_Y); \ + align_buffer_page_end(src_y, kWidth* kHeight + OFF); \ + align_buffer_page_end(src_u, kSizeUV + OFF); \ + align_buffer_page_end(src_v, kSizeUV + OFF); \ + align_buffer_page_end(src_a, kWidth* kHeight + OFF); \ + align_buffer_page_end(dst_argb_c, kStrideB* kHeight + OFF); \ + align_buffer_page_end(dst_argb_opt, kStrideB* kHeight + OFF); \ + for (int i = 0; i < kWidth * kHeight; ++i) { \ + src_y[i + OFF] = (fastrand() & 0xff); \ + src_a[i + OFF] = (fastrand() & 0xff); \ + } \ + for (int i = 0; i < kSizeUV; ++i) { \ + src_u[i + OFF] = (fastrand() & 0xff); \ + src_v[i + OFF] = (fastrand() & 0xff); \ + } \ + memset(dst_argb_c + OFF, 1, kStrideB * kHeight); \ + memset(dst_argb_opt + OFF, 101, kStrideB * kHeight); \ + MaskCpuFlags(disable_cpu_flags_); \ + FMT_PLANAR##To##FMT_B(src_y + OFF, kWidth, src_u + OFF, kStrideUV, \ + src_v + OFF, kStrideUV, src_a + OFF, kWidth, \ + dst_argb_c + OFF, kStrideB, kWidth, NEG kHeight, \ + ATTEN); \ + MaskCpuFlags(benchmark_cpu_info_); \ + for (int i = 0; i < benchmark_iterations_; ++i) { \ + FMT_PLANAR##To##FMT_B(src_y + OFF, kWidth, src_u + OFF, kStrideUV, \ + src_v + OFF, kStrideUV, src_a + OFF, kWidth, \ + dst_argb_opt + OFF, kStrideB, kWidth, NEG kHeight, \ + ATTEN); \ + } \ + for (int i = 0; i < kWidth * BPP_B * kHeight; ++i) { \ + EXPECT_EQ(dst_argb_c[i + OFF], dst_argb_opt[i + OFF]); \ + } \ + free_aligned_buffer_page_end(src_y); \ + free_aligned_buffer_page_end(src_u); \ + free_aligned_buffer_page_end(src_v); \ + free_aligned_buffer_page_end(src_a); \ + free_aligned_buffer_page_end(dst_argb_c); \ + free_aligned_buffer_page_end(dst_argb_opt); \ + } + +#if defined(ENABLE_FULL_TESTS) +#define TESTQPLANARTOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN) \ + TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, benchmark_width_ + 1, _Any, +, 0, 0) \ + TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, benchmark_width_, _Unaligned, +, 2, 0) \ + TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, benchmark_width_, _Invert, -, 0, 0) \ + TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, benchmark_width_, _Opt, +, 0, 0) \ + TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, benchmark_width_, _Premult, +, 0, 1) +#else +#define TESTQPLANARTOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN) \ + TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, benchmark_width_, _Opt, +, 0, 0) +#endif + +#define J420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ + l, m) +#define J420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ + l, m) +#define F420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ + l, m) +#define F420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ + l, m) +#define H420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ + l, m) +#define H420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ + l, m) +#define U420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ + l, m) +#define U420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ + l, m) +#define V420AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ + l, m) +#define V420AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ + l, m) +#define J422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ + l, m) +#define J422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ + l, m) +#define F422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ + l, m) +#define F422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ + l, m) +#define H422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ + l, m) +#define H422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ + l, m) +#define U422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ + l, m) +#define U422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ + l, m) +#define V422AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ + l, m) +#define V422AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ + l, m) +#define J444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ + l, m) +#define J444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvJPEGConstants, k, \ + l, m) +#define F444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ + l, m) +#define F444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvF709Constants, k, \ + l, m) +#define H444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ + l, m) +#define H444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvH709Constants, k, \ + l, m) +#define U444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ + l, m) +#define U444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuv2020Constants, k, \ + l, m) +#define V444AlphaToARGB(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToARGBMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ + l, m) +#define V444AlphaToABGR(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, &kYuvV2020Constants, k, \ + l, m) + +#define I420AlphaToARGBFilter(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I420AlphaToARGBMatrixFilter(a, b, c, d, e, f, g, h, i, j, \ + &kYuvI601Constants, k, l, m, kFilterBilinear) +#define I422AlphaToARGBFilter(a, b, c, d, e, f, g, h, i, j, k, l, m) \ + I422AlphaToARGBMatrixFilter(a, b, c, d, e, f, g, h, i, j, \ + &kYuvI601Constants, k, l, m, kFilterBilinear) + +#if defined(ENABLE_FULL_TESTS) +TESTQPLANARTOB(I420Alpha, 2, 2, ARGB, 4, 4, 1) +TESTQPLANARTOB(I420Alpha, 2, 2, ABGR, 4, 4, 1) +TESTQPLANARTOB(J420Alpha, 2, 2, ARGB, 4, 4, 1) +TESTQPLANARTOB(J420Alpha, 2, 2, ABGR, 4, 4, 1) +TESTQPLANARTOB(H420Alpha, 2, 2, ARGB, 4, 4, 1) +TESTQPLANARTOB(H420Alpha, 2, 2, ABGR, 4, 4, 1) +TESTQPLANARTOB(F420Alpha, 2, 2, ARGB, 4, 4, 1) +TESTQPLANARTOB(F420Alpha, 2, 2, ABGR, 4, 4, 1) +TESTQPLANARTOB(U420Alpha, 2, 2, ARGB, 4, 4, 1) +TESTQPLANARTOB(U420Alpha, 2, 2, ABGR, 4, 4, 1) +TESTQPLANARTOB(V420Alpha, 2, 2, ARGB, 4, 4, 1) +TESTQPLANARTOB(V420Alpha, 2, 2, ABGR, 4, 4, 1) +TESTQPLANARTOB(I422Alpha, 2, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(I422Alpha, 2, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(J422Alpha, 2, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(J422Alpha, 2, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(H422Alpha, 2, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(H422Alpha, 2, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(F422Alpha, 2, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(F422Alpha, 2, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(U422Alpha, 2, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(U422Alpha, 2, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(V422Alpha, 2, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(V422Alpha, 2, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(I444Alpha, 1, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(I444Alpha, 1, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(J444Alpha, 1, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(J444Alpha, 1, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(H444Alpha, 1, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(H444Alpha, 1, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(F444Alpha, 1, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(F444Alpha, 1, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(U444Alpha, 1, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(U444Alpha, 1, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(V444Alpha, 1, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(V444Alpha, 1, 1, ABGR, 4, 4, 1) +TESTQPLANARTOB(I420Alpha, 2, 2, ARGBFilter, 4, 4, 1) +TESTQPLANARTOB(I422Alpha, 2, 1, ARGBFilter, 4, 4, 1) +#else +TESTQPLANARTOB(I420Alpha, 2, 2, ARGB, 4, 4, 1) +TESTQPLANARTOB(I422Alpha, 2, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(I444Alpha, 1, 1, ARGB, 4, 4, 1) +TESTQPLANARTOB(I420Alpha, 2, 2, ARGBFilter, 4, 4, 1) +TESTQPLANARTOB(I422Alpha, 2, 1, ARGBFilter, 4, 4, 1) +#endif + TEST_F(LibYUVConvertTest, TestYToARGB) { uint8_t y[32]; uint8_t expectedg[32]; @@ -1261,6 +1289,7 @@ TEST_F(LibYUVConvertTest, TestDither) { free_aligned_buffer_page_end(dst_argb32_opt); \ } +#if defined(ENABLE_FULL_TESTS) #define TESTPLANARTOBD(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ YALIGN, FMT_C, BPP_C) \ TESTPLANARTOBID(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ @@ -1271,6 +1300,12 @@ TEST_F(LibYUVConvertTest, TestDither) { YALIGN, benchmark_width_, _Invert, -, 0, FMT_C, BPP_C) \ TESTPLANARTOBID(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ YALIGN, benchmark_width_, _Opt, +, 0, FMT_C, BPP_C) +#else +#define TESTPLANARTOBD(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, FMT_C, BPP_C) \ + TESTPLANARTOBID(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, benchmark_width_, _Opt, +, 0, FMT_C, BPP_C) +#endif #ifdef LITTLE_ENDIAN_ONLY_TEST TESTPLANARTOBD(I420, 2, 2, RGB565, 2, 2, 1, ARGB, 4) @@ -1578,6 +1613,7 @@ TESTQPLANARTOE(I444Alpha, 1, 1, ABGR, 1, 4, ARGB, 4) free_aligned_buffer_page_end(dst_argb_bc); \ } +#if defined(ENABLE_FULL_TESTS) #define TESTPLANETOE(FMT_A, SUB_A, BPP_A, FMT_B, SUB_B, BPP_B, FMT_C, BPP_C) \ TESTPLANETOEI(FMT_A, SUB_A, BPP_A, FMT_B, SUB_B, BPP_B, \ benchmark_width_ + 1, _Any, +, 0, FMT_C, BPP_C) \ @@ -1587,6 +1623,11 @@ TESTQPLANARTOE(I444Alpha, 1, 1, ABGR, 1, 4, ARGB, 4) _Invert, -, 0, FMT_C, BPP_C) \ TESTPLANETOEI(FMT_A, SUB_A, BPP_A, FMT_B, SUB_B, BPP_B, benchmark_width_, \ _Opt, +, 0, FMT_C, BPP_C) +#else +#define TESTPLANETOE(FMT_A, SUB_A, BPP_A, FMT_B, SUB_B, BPP_B, FMT_C, BPP_C) \ + TESTPLANETOEI(FMT_A, SUB_A, BPP_A, FMT_B, SUB_B, BPP_B, benchmark_width_, \ + _Opt, +, 0, FMT_C, BPP_C) +#endif // Caveat: Destination needs to be 4 bytes #ifdef LITTLE_ENDIAN_ONLY_TEST @@ -1703,6 +1744,8 @@ TEST_F(LibYUVConvertTest, ABGRToAR30Row_Opt) { } #endif // HAS_ABGRTOAR30ROW_AVX2 +#if !defined(LEAN_TESTS) + // Provide matrix wrappers for 12 bit YUV #define I012ToARGB(a, b, c, d, e, f, g, h, i, j) \ I012ToARGBMatrix(a, b, c, d, e, f, g, h, &kYuvI601Constants, i, j) @@ -1797,6 +1840,7 @@ TEST_F(LibYUVConvertTest, ABGRToAR30Row_Opt) { free_aligned_buffer_page_end(dst_argb_opt); \ } +#if defined(ENABLE_FULL_TESTS) #define TESTPLANAR16TOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_MASK, FMT_B, \ BPP_B, ALIGN, YALIGN) \ TESTPLANAR16TOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_MASK, FMT_B, BPP_B, \ @@ -1807,6 +1851,12 @@ TEST_F(LibYUVConvertTest, ABGRToAR30Row_Opt) { ALIGN, YALIGN, benchmark_width_, _Invert, -, 0, 0) \ TESTPLANAR16TOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_MASK, FMT_B, BPP_B, \ ALIGN, YALIGN, benchmark_width_, _Opt, +, 0, 0) +#else +#define TESTPLANAR16TOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_MASK, FMT_B, \ + BPP_B, ALIGN, YALIGN) \ + TESTPLANAR16TOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_MASK, FMT_B, BPP_B, \ + ALIGN, YALIGN, benchmark_width_, _Opt, +, 0, 0) +#endif // These conversions are only optimized for x86 #if !defined(DISABLE_SLOW_TESTS) || defined(__x86_64__) || defined(__i386__) @@ -2135,6 +2185,7 @@ TESTQPLANAR16TOB(I210Alpha, 2, 1, ARGBFilter, 4, 4, 1, 10) free_aligned_buffer_page_end(dst_argb_opt); \ } +#if defined(ENABLE_FULL_TESTS) #define TESTBP16TOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ YALIGN, S_DEPTH) \ TESTBP16TOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, YALIGN, \ @@ -2145,6 +2196,12 @@ TESTQPLANAR16TOB(I210Alpha, 2, 1, ARGBFilter, 4, 4, 1, 10) benchmark_width_, _Invert, -, 0, 0, S_DEPTH) \ TESTBP16TOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, YALIGN, \ benchmark_width_, _Opt, +, 0, 0, S_DEPTH) +#else +#define TESTBP16TOB(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \ + YALIGN, S_DEPTH) \ + TESTBP16TOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, YALIGN, \ + benchmark_width_, _Opt, +, 0, 0, S_DEPTH) +#endif #define P010ToARGB(a, b, c, d, e, f, g, h) \ P010ToARGBMatrix(a, b, c, d, e, f, &kYuvH709Constants, g, h) @@ -2639,5 +2696,6 @@ TEST_F(LibYUVConvertTest, Test565) { uint32_t checksum = HashDjb2(&pixels565[0][0], sizeof(pixels565), 5381); EXPECT_EQ(610919429u, checksum); } +#endif // !defined(LEAN_TESTS) } // namespace libyuv diff --git a/unit_test/convert_test.cc b/unit_test/convert_test.cc index 3c75c2d6..12c3398c 100644 --- a/unit_test/convert_test.cc +++ b/unit_test/convert_test.cc @@ -31,6 +31,13 @@ #include "libyuv/row.h" /* For ARGBToAR30Row_AVX2 */ #endif +#if defined(__riscv) && !defined(__clang__) +#define DISABLE_SLOW_TESTS +#undef ENABLE_FULL_TESTS +#undef ENABLE_ROW_TESTS +#define LEAN_TESTS +#endif + // Some functions fail on big endian. Enable these tests on all cpus except // PowerPC, but they are not optimized so disabled by default. #if !defined(DISABLE_SLOW_TESTS) && !defined(__powerpc__) @@ -784,6 +791,8 @@ TESTATOBP(UYVY, 2, 4, NV12, 2, 2) TESTATOBP(AYUV, 1, 4, NV12, 2, 2) TESTATOBP(AYUV, 1, 4, NV21, 2, 2) +#if !defined(LEAN_TESTS) + #ifdef HAVE_JPEG TEST_F(LibYUVConvertTest, ValidateJpeg) { const int kOff = 10; @@ -2057,4 +2066,6 @@ TEST_F(LibYUVConvertTest, TestRGB24ToI420) { } #endif +#endif // !defined(LEAN_TESTS) + } // namespace libyuv diff --git a/unit_test/scale_plane_test.cc b/unit_test/scale_plane_test.cc new file mode 100644 index 00000000..9ce47a02 --- /dev/null +++ b/unit_test/scale_plane_test.cc @@ -0,0 +1,470 @@ +/* + * Copyright 2023 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.h" + +#ifdef ENABLE_ROW_TESTS +#include "libyuv/scale_row.h" // For ScaleRowDown2Box_Odd_C +#endif + +#define STRINGIZE(line) #line +#define FILELINESTR(file, line) file ":" STRINGIZE(line) + +#if defined(__riscv) && !defined(__clang__) +#define DISABLE_SLOW_TESTS +#undef ENABLE_FULL_TESTS +#undef ENABLE_ROW_TESTS +#define LEAN_TESTS +#endif + +#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 + +namespace libyuv { + +#ifdef ENABLE_ROW_TESTS +#ifdef HAS_SCALEROWDOWN2_SSSE3 +TEST_F(LibYUVScaleTest, TestScaleRowDown2Box_Odd_SSSE3) { + SIMD_ALIGNED(uint8_t orig_pixels[128 * 2]); + SIMD_ALIGNED(uint8_t dst_pixels_opt[64]); + SIMD_ALIGNED(uint8_t dst_pixels_c[64]); + memset(orig_pixels, 0, sizeof(orig_pixels)); + memset(dst_pixels_opt, 0, sizeof(dst_pixels_opt)); + memset(dst_pixels_c, 0, sizeof(dst_pixels_c)); + + int has_ssse3 = TestCpuFlag(kCpuHasSSSE3); + if (!has_ssse3) { + printf("Warning SSSE3 not detected; Skipping test.\n"); + } else { + // TL. + orig_pixels[0] = 255u; + orig_pixels[1] = 0u; + orig_pixels[128 + 0] = 0u; + orig_pixels[128 + 1] = 0u; + // TR. + orig_pixels[2] = 0u; + orig_pixels[3] = 100u; + orig_pixels[128 + 2] = 0u; + orig_pixels[128 + 3] = 0u; + // BL. + orig_pixels[4] = 0u; + orig_pixels[5] = 0u; + orig_pixels[128 + 4] = 50u; + orig_pixels[128 + 5] = 0u; + // BR. + orig_pixels[6] = 0u; + orig_pixels[7] = 0u; + orig_pixels[128 + 6] = 0u; + orig_pixels[128 + 7] = 20u; + // Odd. + orig_pixels[126] = 4u; + orig_pixels[127] = 255u; + orig_pixels[128 + 126] = 16u; + orig_pixels[128 + 127] = 255u; + + // Test regular half size. + ScaleRowDown2Box_C(orig_pixels, 128, dst_pixels_c, 64); + + EXPECT_EQ(64u, dst_pixels_c[0]); + EXPECT_EQ(25u, dst_pixels_c[1]); + EXPECT_EQ(13u, dst_pixels_c[2]); + EXPECT_EQ(5u, dst_pixels_c[3]); + EXPECT_EQ(0u, dst_pixels_c[4]); + EXPECT_EQ(133u, dst_pixels_c[63]); + + // Test Odd width version - Last pixel is just 1 horizontal pixel. + ScaleRowDown2Box_Odd_C(orig_pixels, 128, dst_pixels_c, 64); + + EXPECT_EQ(64u, dst_pixels_c[0]); + EXPECT_EQ(25u, dst_pixels_c[1]); + EXPECT_EQ(13u, dst_pixels_c[2]); + EXPECT_EQ(5u, dst_pixels_c[3]); + EXPECT_EQ(0u, dst_pixels_c[4]); + EXPECT_EQ(10u, dst_pixels_c[63]); + + // Test one pixel less, should skip the last pixel. + memset(dst_pixels_c, 0, sizeof(dst_pixels_c)); + ScaleRowDown2Box_Odd_C(orig_pixels, 128, dst_pixels_c, 63); + + EXPECT_EQ(64u, dst_pixels_c[0]); + EXPECT_EQ(25u, dst_pixels_c[1]); + EXPECT_EQ(13u, dst_pixels_c[2]); + EXPECT_EQ(5u, dst_pixels_c[3]); + EXPECT_EQ(0u, dst_pixels_c[4]); + EXPECT_EQ(0u, dst_pixels_c[63]); + + // Test regular half size SSSE3. + ScaleRowDown2Box_SSSE3(orig_pixels, 128, dst_pixels_opt, 64); + + EXPECT_EQ(64u, dst_pixels_opt[0]); + EXPECT_EQ(25u, dst_pixels_opt[1]); + EXPECT_EQ(13u, dst_pixels_opt[2]); + EXPECT_EQ(5u, dst_pixels_opt[3]); + EXPECT_EQ(0u, dst_pixels_opt[4]); + EXPECT_EQ(133u, dst_pixels_opt[63]); + + // Compare C and SSSE3 match. + ScaleRowDown2Box_Odd_C(orig_pixels, 128, dst_pixels_c, 64); + ScaleRowDown2Box_Odd_SSSE3(orig_pixels, 128, dst_pixels_opt, 64); + for (int i = 0; i < 64; ++i) { + EXPECT_EQ(dst_pixels_c[i], dst_pixels_opt[i]); + } + } +} +#endif // HAS_SCALEROWDOWN2_SSSE3 + +extern "C" void ScaleRowDown2Box_16_NEON(const uint16_t* src_ptr, + ptrdiff_t src_stride, + uint16_t* dst, + int dst_width); + +TEST_F(LibYUVScaleTest, TestScaleRowDown2Box_16) { + SIMD_ALIGNED(uint16_t orig_pixels[2560 * 2]); + SIMD_ALIGNED(uint16_t dst_pixels_c[1280]); + SIMD_ALIGNED(uint16_t dst_pixels_opt[1280]); + + memset(orig_pixels, 0, sizeof(orig_pixels)); + memset(dst_pixels_c, 1, sizeof(dst_pixels_c)); + memset(dst_pixels_opt, 2, sizeof(dst_pixels_opt)); + + for (int i = 0; i < 2560 * 2; ++i) { + orig_pixels[i] = i; + } + ScaleRowDown2Box_16_C(&orig_pixels[0], 2560, &dst_pixels_c[0], 1280); + for (int i = 0; i < benchmark_pixels_div1280_; ++i) { +#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__) + int has_neon = TestCpuFlag(kCpuHasNEON); + if (has_neon) { + ScaleRowDown2Box_16_NEON(&orig_pixels[0], 2560, &dst_pixels_opt[0], 1280); + } else { + ScaleRowDown2Box_16_C(&orig_pixels[0], 2560, &dst_pixels_opt[0], 1280); + } +#else + ScaleRowDown2Box_16_C(&orig_pixels[0], 2560, &dst_pixels_opt[0], 1280); +#endif + } + + for (int i = 0; i < 1280; ++i) { + EXPECT_EQ(dst_pixels_c[i], dst_pixels_opt[i]); + } + + EXPECT_EQ(dst_pixels_c[0], (0 + 1 + 2560 + 2561 + 2) / 4); + EXPECT_EQ(dst_pixels_c[1279], 3839); +} +#endif // ENABLE_ROW_TESTS + +// Test scaling plane with 8 bit C vs 12 bit C and return maximum pixel +// difference. +// 0 = exact. +static int TestPlaneFilter_16(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; + int64_t src_y_plane_size = (Abs(src_width)) * (Abs(src_height)); + int src_stride_y = Abs(src_width); + int dst_y_plane_size = dst_width * dst_height; + int dst_stride_y = dst_width; + + align_buffer_page_end(src_y, src_y_plane_size); + align_buffer_page_end(src_y_16, src_y_plane_size * 2); + align_buffer_page_end(dst_y_8, dst_y_plane_size); + align_buffer_page_end(dst_y_16, dst_y_plane_size * 2); + uint16_t* p_src_y_16 = reinterpret_cast<uint16_t*>(src_y_16); + uint16_t* p_dst_y_16 = reinterpret_cast<uint16_t*>(dst_y_16); + + MemRandomize(src_y, src_y_plane_size); + memset(dst_y_8, 0, dst_y_plane_size); + memset(dst_y_16, 1, dst_y_plane_size * 2); + + for (i = 0; i < src_y_plane_size; ++i) { + p_src_y_16[i] = src_y[i] & 255; + } + + MaskCpuFlags(disable_cpu_flags); // Disable all CPU optimization. + ScalePlane(src_y, src_stride_y, src_width, src_height, dst_y_8, dst_stride_y, + dst_width, dst_height, f); + MaskCpuFlags(benchmark_cpu_info); // Enable all CPU optimization. + + for (i = 0; i < benchmark_iterations; ++i) { + ScalePlane_16(p_src_y_16, src_stride_y, src_width, src_height, p_dst_y_16, + dst_stride_y, dst_width, dst_height, f); + } + + // Expect an exact match. + int max_diff = 0; + for (i = 0; i < dst_y_plane_size; ++i) { + int abs_diff = Abs(dst_y_8[i] - p_dst_y_16[i]); + if (abs_diff > max_diff) { + max_diff = abs_diff; + } + } + + free_aligned_buffer_page_end(dst_y_8); + free_aligned_buffer_page_end(dst_y_16); + free_aligned_buffer_page_end(src_y); + free_aligned_buffer_page_end(src_y_16); + + return max_diff; +} + +// The following adjustments in dimensions ensure the scale factor will be +// exactly achieved. +// 2 is chroma subsample. +#define DX(x, nom, denom) static_cast<int>(((Abs(x) / nom + 1) / 2) * nom * 2) +#define SX(x, nom, denom) static_cast<int>(((x / nom + 1) / 2) * denom * 2) + +#define TEST_FACTOR1(name, filter, nom, denom, max_diff) \ + TEST_F(LibYUVScaleTest, DISABLED_##ScalePlaneDownBy##name##_##filter##_16) { \ + int diff = TestPlaneFilter_16( \ + 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); \ + } + +// 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, boxdiff) \ + TEST_FACTOR1(name, None, nom, denom, 0) \ + TEST_FACTOR1(name, Linear, nom, denom, boxdiff) \ + TEST_FACTOR1(name, Bilinear, nom, denom, boxdiff) \ + TEST_FACTOR1(name, Box, nom, denom, boxdiff) + +TEST_FACTOR(2, 1, 2, 0) +TEST_FACTOR(4, 1, 4, 0) +// TEST_FACTOR(8, 1, 8, 0) Disable for benchmark performance. Takes 90 seconds. +TEST_FACTOR(3by4, 3, 4, 1) +TEST_FACTOR(3by8, 3, 8, 1) +TEST_FACTOR(3, 1, 3, 0) +#undef TEST_FACTOR1 +#undef TEST_FACTOR +#undef SX +#undef DX + +TEST_F(LibYUVScaleTest, PlaneTest3x) { + const int kSrcStride = 480; + const int kDstStride = 160; + const int kSize = kSrcStride * 3; + align_buffer_page_end(orig_pixels, kSize); + for (int i = 0; i < 480 * 3; ++i) { + orig_pixels[i] = 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) { + ScalePlane(orig_pixels, kSrcStride, 480, 3, dest_pixels, kDstStride, 160, 1, + kFilterBilinear); + } + + EXPECT_EQ(225, dest_pixels[0]); + + ScalePlane(orig_pixels, kSrcStride, 480, 3, dest_pixels, kDstStride, 160, 1, + kFilterNone); + + EXPECT_EQ(225, dest_pixels[0]); + + free_aligned_buffer_page_end(dest_pixels); + free_aligned_buffer_page_end(orig_pixels); +} + +TEST_F(LibYUVScaleTest, PlaneTest4x) { + const int kSrcStride = 640; + const int kDstStride = 160; + const int kSize = kSrcStride * 4; + align_buffer_page_end(orig_pixels, kSize); + for (int i = 0; i < 640 * 4; ++i) { + orig_pixels[i] = 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) { + ScalePlane(orig_pixels, kSrcStride, 640, 4, dest_pixels, kDstStride, 160, 1, + kFilterBilinear); + } + + EXPECT_EQ(66, dest_pixels[0]); + + ScalePlane(orig_pixels, kSrcStride, 640, 4, dest_pixels, kDstStride, 160, 1, + kFilterNone); + + EXPECT_EQ(2, dest_pixels[0]); // expect the 3rd pixel of the 3rd row + + free_aligned_buffer_page_end(dest_pixels); + free_aligned_buffer_page_end(orig_pixels); +} + +// Intent is to test 200x50 to 50x200 but width and height can be parameters. +TEST_F(LibYUVScaleTest, PlaneTestRotate_None) { + const int kSize = benchmark_width_ * benchmark_height_; + align_buffer_page_end(orig_pixels, kSize); + for (int i = 0; i < kSize; ++i) { + orig_pixels[i] = i; + } + align_buffer_page_end(dest_opt_pixels, kSize); + align_buffer_page_end(dest_c_pixels, kSize); + + MaskCpuFlags(disable_cpu_flags_); // Disable all CPU optimization. + ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, benchmark_height_, + dest_c_pixels, benchmark_height_, benchmark_height_, + benchmark_width_, kFilterNone); + MaskCpuFlags(benchmark_cpu_info_); // Enable all CPU optimization. + + for (int i = 0; i < benchmark_iterations_; ++i) { + ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, + benchmark_height_, dest_opt_pixels, benchmark_height_, + benchmark_height_, benchmark_width_, kFilterNone); + } + + for (int i = 0; i < kSize; ++i) { + EXPECT_EQ(dest_c_pixels[i], dest_opt_pixels[i]); + } + + free_aligned_buffer_page_end(dest_c_pixels); + free_aligned_buffer_page_end(dest_opt_pixels); + free_aligned_buffer_page_end(orig_pixels); +} + +TEST_F(LibYUVScaleTest, PlaneTestRotate_Bilinear) { + const int kSize = benchmark_width_ * benchmark_height_; + align_buffer_page_end(orig_pixels, kSize); + for (int i = 0; i < kSize; ++i) { + orig_pixels[i] = i; + } + align_buffer_page_end(dest_opt_pixels, kSize); + align_buffer_page_end(dest_c_pixels, kSize); + + MaskCpuFlags(disable_cpu_flags_); // Disable all CPU optimization. + ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, benchmark_height_, + dest_c_pixels, benchmark_height_, benchmark_height_, + benchmark_width_, kFilterBilinear); + MaskCpuFlags(benchmark_cpu_info_); // Enable all CPU optimization. + + for (int i = 0; i < benchmark_iterations_; ++i) { + ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, + benchmark_height_, dest_opt_pixels, benchmark_height_, + benchmark_height_, benchmark_width_, kFilterBilinear); + } + + for (int i = 0; i < kSize; ++i) { + EXPECT_EQ(dest_c_pixels[i], dest_opt_pixels[i]); + } + + free_aligned_buffer_page_end(dest_c_pixels); + free_aligned_buffer_page_end(dest_opt_pixels); + free_aligned_buffer_page_end(orig_pixels); +} + +// Intent is to test 200x50 to 50x200 but width and height can be parameters. +TEST_F(LibYUVScaleTest, PlaneTestRotate_Box) { + const int kSize = benchmark_width_ * benchmark_height_; + align_buffer_page_end(orig_pixels, kSize); + for (int i = 0; i < kSize; ++i) { + orig_pixels[i] = i; + } + align_buffer_page_end(dest_opt_pixels, kSize); + align_buffer_page_end(dest_c_pixels, kSize); + + MaskCpuFlags(disable_cpu_flags_); // Disable all CPU optimization. + ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, benchmark_height_, + dest_c_pixels, benchmark_height_, benchmark_height_, + benchmark_width_, kFilterBox); + MaskCpuFlags(benchmark_cpu_info_); // Enable all CPU optimization. + + for (int i = 0; i < benchmark_iterations_; ++i) { + ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, + benchmark_height_, dest_opt_pixels, benchmark_height_, + benchmark_height_, benchmark_width_, kFilterBox); + } + + for (int i = 0; i < kSize; ++i) { + EXPECT_EQ(dest_c_pixels[i], dest_opt_pixels[i]); + } + + free_aligned_buffer_page_end(dest_c_pixels); + free_aligned_buffer_page_end(dest_opt_pixels); + free_aligned_buffer_page_end(orig_pixels); +} + +TEST_F(LibYUVScaleTest, PlaneTest1_Box) { + align_buffer_page_end(orig_pixels, 3); + align_buffer_page_end(dst_pixels, 3); + + // Pad the 1x1 byte image with invalid values before and after in case libyuv + // reads outside the memory boundaries. + orig_pixels[0] = 0; + orig_pixels[1] = 1; // scale this pixel + orig_pixels[2] = 2; + dst_pixels[0] = 3; + dst_pixels[1] = 3; + dst_pixels[2] = 3; + + libyuv::ScalePlane(orig_pixels + 1, /* src_stride= */ 1, /* src_width= */ 1, + /* src_height= */ 1, dst_pixels, /* dst_stride= */ 1, + /* dst_width= */ 1, /* dst_height= */ 2, + libyuv::kFilterBox); + + EXPECT_EQ(dst_pixels[0], 1); + EXPECT_EQ(dst_pixels[1], 1); + EXPECT_EQ(dst_pixels[2], 3); + + free_aligned_buffer_page_end(dst_pixels); + free_aligned_buffer_page_end(orig_pixels); +} + +TEST_F(LibYUVScaleTest, PlaneTest1_16_Box) { + align_buffer_page_end(orig_pixels_alloc, 3 * 2); + align_buffer_page_end(dst_pixels_alloc, 3 * 2); + uint16_t* orig_pixels = (uint16_t*)orig_pixels_alloc; + uint16_t* dst_pixels = (uint16_t*)dst_pixels_alloc; + + // Pad the 1x1 byte image with invalid values before and after in case libyuv + // reads outside the memory boundaries. + orig_pixels[0] = 0; + orig_pixels[1] = 1; // scale this pixel + orig_pixels[2] = 2; + dst_pixels[0] = 3; + dst_pixels[1] = 3; + dst_pixels[2] = 3; + + libyuv::ScalePlane_16( + orig_pixels + 1, /* src_stride= */ 1, /* src_width= */ 1, + /* src_height= */ 1, dst_pixels, /* dst_stride= */ 1, + /* dst_width= */ 1, /* dst_height= */ 2, libyuv::kFilterNone); + + EXPECT_EQ(dst_pixels[0], 1); + EXPECT_EQ(dst_pixels[1], 1); + EXPECT_EQ(dst_pixels[2], 3); + + free_aligned_buffer_page_end(dst_pixels_alloc); + free_aligned_buffer_page_end(orig_pixels_alloc); +} +} // namespace libyuv diff --git a/unit_test/scale_test.cc b/unit_test/scale_test.cc index c2232e66..6e3b9271 100644 --- a/unit_test/scale_test.cc +++ b/unit_test/scale_test.cc @@ -22,6 +22,11 @@ #define STRINGIZE(line) #line #define FILELINESTR(file, line) file ":" STRINGIZE(line) +#if defined(__riscv) && !defined(__clang__) +#define DISABLE_SLOW_TESTS +#undef ENABLE_FULL_TESTS +#endif + #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. @@ -1123,437 +1128,6 @@ TEST_SCALESWAPXY1(DISABLED_, Scale, Bilinear, 3) TEST_SCALESWAPXY1(DISABLED_, Scale, Box, 3) #endif #endif - #undef TEST_SCALESWAPXY1 -#ifdef ENABLE_ROW_TESTS -#ifdef HAS_SCALEROWDOWN2_SSSE3 -TEST_F(LibYUVScaleTest, TestScaleRowDown2Box_Odd_SSSE3) { - SIMD_ALIGNED(uint8_t orig_pixels[128 * 2]); - SIMD_ALIGNED(uint8_t dst_pixels_opt[64]); - SIMD_ALIGNED(uint8_t dst_pixels_c[64]); - memset(orig_pixels, 0, sizeof(orig_pixels)); - memset(dst_pixels_opt, 0, sizeof(dst_pixels_opt)); - memset(dst_pixels_c, 0, sizeof(dst_pixels_c)); - - int has_ssse3 = TestCpuFlag(kCpuHasSSSE3); - if (!has_ssse3) { - printf("Warning SSSE3 not detected; Skipping test.\n"); - } else { - // TL. - orig_pixels[0] = 255u; - orig_pixels[1] = 0u; - orig_pixels[128 + 0] = 0u; - orig_pixels[128 + 1] = 0u; - // TR. - orig_pixels[2] = 0u; - orig_pixels[3] = 100u; - orig_pixels[128 + 2] = 0u; - orig_pixels[128 + 3] = 0u; - // BL. - orig_pixels[4] = 0u; - orig_pixels[5] = 0u; - orig_pixels[128 + 4] = 50u; - orig_pixels[128 + 5] = 0u; - // BR. - orig_pixels[6] = 0u; - orig_pixels[7] = 0u; - orig_pixels[128 + 6] = 0u; - orig_pixels[128 + 7] = 20u; - // Odd. - orig_pixels[126] = 4u; - orig_pixels[127] = 255u; - orig_pixels[128 + 126] = 16u; - orig_pixels[128 + 127] = 255u; - - // Test regular half size. - ScaleRowDown2Box_C(orig_pixels, 128, dst_pixels_c, 64); - - EXPECT_EQ(64u, dst_pixels_c[0]); - EXPECT_EQ(25u, dst_pixels_c[1]); - EXPECT_EQ(13u, dst_pixels_c[2]); - EXPECT_EQ(5u, dst_pixels_c[3]); - EXPECT_EQ(0u, dst_pixels_c[4]); - EXPECT_EQ(133u, dst_pixels_c[63]); - - // Test Odd width version - Last pixel is just 1 horizontal pixel. - ScaleRowDown2Box_Odd_C(orig_pixels, 128, dst_pixels_c, 64); - - EXPECT_EQ(64u, dst_pixels_c[0]); - EXPECT_EQ(25u, dst_pixels_c[1]); - EXPECT_EQ(13u, dst_pixels_c[2]); - EXPECT_EQ(5u, dst_pixels_c[3]); - EXPECT_EQ(0u, dst_pixels_c[4]); - EXPECT_EQ(10u, dst_pixels_c[63]); - - // Test one pixel less, should skip the last pixel. - memset(dst_pixels_c, 0, sizeof(dst_pixels_c)); - ScaleRowDown2Box_Odd_C(orig_pixels, 128, dst_pixels_c, 63); - - EXPECT_EQ(64u, dst_pixels_c[0]); - EXPECT_EQ(25u, dst_pixels_c[1]); - EXPECT_EQ(13u, dst_pixels_c[2]); - EXPECT_EQ(5u, dst_pixels_c[3]); - EXPECT_EQ(0u, dst_pixels_c[4]); - EXPECT_EQ(0u, dst_pixels_c[63]); - - // Test regular half size SSSE3. - ScaleRowDown2Box_SSSE3(orig_pixels, 128, dst_pixels_opt, 64); - - EXPECT_EQ(64u, dst_pixels_opt[0]); - EXPECT_EQ(25u, dst_pixels_opt[1]); - EXPECT_EQ(13u, dst_pixels_opt[2]); - EXPECT_EQ(5u, dst_pixels_opt[3]); - EXPECT_EQ(0u, dst_pixels_opt[4]); - EXPECT_EQ(133u, dst_pixels_opt[63]); - - // Compare C and SSSE3 match. - ScaleRowDown2Box_Odd_C(orig_pixels, 128, dst_pixels_c, 64); - ScaleRowDown2Box_Odd_SSSE3(orig_pixels, 128, dst_pixels_opt, 64); - for (int i = 0; i < 64; ++i) { - EXPECT_EQ(dst_pixels_c[i], dst_pixels_opt[i]); - } - } -} -#endif // HAS_SCALEROWDOWN2_SSSE3 - -extern "C" void ScaleRowDown2Box_16_NEON(const uint16_t* src_ptr, - ptrdiff_t src_stride, - uint16_t* dst, - int dst_width); - -TEST_F(LibYUVScaleTest, TestScaleRowDown2Box_16) { - SIMD_ALIGNED(uint16_t orig_pixels[2560 * 2]); - SIMD_ALIGNED(uint16_t dst_pixels_c[1280]); - SIMD_ALIGNED(uint16_t dst_pixels_opt[1280]); - - memset(orig_pixels, 0, sizeof(orig_pixels)); - memset(dst_pixels_c, 1, sizeof(dst_pixels_c)); - memset(dst_pixels_opt, 2, sizeof(dst_pixels_opt)); - - for (int i = 0; i < 2560 * 2; ++i) { - orig_pixels[i] = i; - } - ScaleRowDown2Box_16_C(&orig_pixels[0], 2560, &dst_pixels_c[0], 1280); - for (int i = 0; i < benchmark_pixels_div1280_; ++i) { -#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__) - int has_neon = TestCpuFlag(kCpuHasNEON); - if (has_neon) { - ScaleRowDown2Box_16_NEON(&orig_pixels[0], 2560, &dst_pixels_opt[0], 1280); - } else { - ScaleRowDown2Box_16_C(&orig_pixels[0], 2560, &dst_pixels_opt[0], 1280); - } -#else - ScaleRowDown2Box_16_C(&orig_pixels[0], 2560, &dst_pixels_opt[0], 1280); -#endif - } - - for (int i = 0; i < 1280; ++i) { - EXPECT_EQ(dst_pixels_c[i], dst_pixels_opt[i]); - } - - EXPECT_EQ(dst_pixels_c[0], (0 + 1 + 2560 + 2561 + 2) / 4); - EXPECT_EQ(dst_pixels_c[1279], 3839); -} -#endif // ENABLE_ROW_TESTS - -// Test scaling plane with 8 bit C vs 12 bit C and return maximum pixel -// difference. -// 0 = exact. -static int TestPlaneFilter_16(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; - int64_t src_y_plane_size = (Abs(src_width)) * (Abs(src_height)); - int src_stride_y = Abs(src_width); - int dst_y_plane_size = dst_width * dst_height; - int dst_stride_y = dst_width; - - align_buffer_page_end(src_y, src_y_plane_size); - align_buffer_page_end(src_y_16, src_y_plane_size * 2); - align_buffer_page_end(dst_y_8, dst_y_plane_size); - align_buffer_page_end(dst_y_16, dst_y_plane_size * 2); - uint16_t* p_src_y_16 = reinterpret_cast<uint16_t*>(src_y_16); - uint16_t* p_dst_y_16 = reinterpret_cast<uint16_t*>(dst_y_16); - - MemRandomize(src_y, src_y_plane_size); - memset(dst_y_8, 0, dst_y_plane_size); - memset(dst_y_16, 1, dst_y_plane_size * 2); - - for (i = 0; i < src_y_plane_size; ++i) { - p_src_y_16[i] = src_y[i] & 255; - } - - MaskCpuFlags(disable_cpu_flags); // Disable all CPU optimization. - ScalePlane(src_y, src_stride_y, src_width, src_height, dst_y_8, dst_stride_y, - dst_width, dst_height, f); - MaskCpuFlags(benchmark_cpu_info); // Enable all CPU optimization. - - for (i = 0; i < benchmark_iterations; ++i) { - ScalePlane_16(p_src_y_16, src_stride_y, src_width, src_height, p_dst_y_16, - dst_stride_y, dst_width, dst_height, f); - } - - // Expect an exact match. - int max_diff = 0; - for (i = 0; i < dst_y_plane_size; ++i) { - int abs_diff = Abs(dst_y_8[i] - p_dst_y_16[i]); - if (abs_diff > max_diff) { - max_diff = abs_diff; - } - } - - free_aligned_buffer_page_end(dst_y_8); - free_aligned_buffer_page_end(dst_y_16); - free_aligned_buffer_page_end(src_y); - free_aligned_buffer_page_end(src_y_16); - - return max_diff; -} - -// The following adjustments in dimensions ensure the scale factor will be -// exactly achieved. -// 2 is chroma subsample. -#define DX(x, nom, denom) static_cast<int>(((Abs(x) / nom + 1) / 2) * nom * 2) -#define SX(x, nom, denom) static_cast<int>(((x / nom + 1) / 2) * denom * 2) - -#define TEST_FACTOR1(name, filter, nom, denom, max_diff) \ - TEST_F(LibYUVScaleTest, DISABLED_##ScalePlaneDownBy##name##_##filter##_16) { \ - int diff = TestPlaneFilter_16( \ - 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); \ - } - -// 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, boxdiff) \ - TEST_FACTOR1(name, None, nom, denom, 0) \ - TEST_FACTOR1(name, Linear, nom, denom, boxdiff) \ - TEST_FACTOR1(name, Bilinear, nom, denom, boxdiff) \ - TEST_FACTOR1(name, Box, nom, denom, boxdiff) - -TEST_FACTOR(2, 1, 2, 0) -TEST_FACTOR(4, 1, 4, 0) -// TEST_FACTOR(8, 1, 8, 0) Disable for benchmark performance. Takes 90 seconds. -TEST_FACTOR(3by4, 3, 4, 1) -TEST_FACTOR(3by8, 3, 8, 1) -TEST_FACTOR(3, 1, 3, 0) -#undef TEST_FACTOR1 -#undef TEST_FACTOR -#undef SX -#undef DX - -TEST_F(LibYUVScaleTest, PlaneTest3x) { - const int kSrcStride = 480; - const int kDstStride = 160; - const int kSize = kSrcStride * 3; - align_buffer_page_end(orig_pixels, kSize); - for (int i = 0; i < 480 * 3; ++i) { - orig_pixels[i] = 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) { - ScalePlane(orig_pixels, kSrcStride, 480, 3, dest_pixels, kDstStride, 160, 1, - kFilterBilinear); - } - - EXPECT_EQ(225, dest_pixels[0]); - - ScalePlane(orig_pixels, kSrcStride, 480, 3, dest_pixels, kDstStride, 160, 1, - kFilterNone); - - EXPECT_EQ(225, dest_pixels[0]); - - free_aligned_buffer_page_end(dest_pixels); - free_aligned_buffer_page_end(orig_pixels); -} - -TEST_F(LibYUVScaleTest, PlaneTest4x) { - const int kSrcStride = 640; - const int kDstStride = 160; - const int kSize = kSrcStride * 4; - align_buffer_page_end(orig_pixels, kSize); - for (int i = 0; i < 640 * 4; ++i) { - orig_pixels[i] = 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) { - ScalePlane(orig_pixels, kSrcStride, 640, 4, dest_pixels, kDstStride, 160, 1, - kFilterBilinear); - } - - EXPECT_EQ(66, dest_pixels[0]); - - ScalePlane(orig_pixels, kSrcStride, 640, 4, dest_pixels, kDstStride, 160, 1, - kFilterNone); - - EXPECT_EQ(2, dest_pixels[0]); // expect the 3rd pixel of the 3rd row - - free_aligned_buffer_page_end(dest_pixels); - free_aligned_buffer_page_end(orig_pixels); -} - -// Intent is to test 200x50 to 50x200 but width and height can be parameters. -TEST_F(LibYUVScaleTest, PlaneTestRotate_None) { - const int kSize = benchmark_width_ * benchmark_height_; - align_buffer_page_end(orig_pixels, kSize); - for (int i = 0; i < kSize; ++i) { - orig_pixels[i] = i; - } - align_buffer_page_end(dest_opt_pixels, kSize); - align_buffer_page_end(dest_c_pixels, kSize); - - MaskCpuFlags(disable_cpu_flags_); // Disable all CPU optimization. - ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, benchmark_height_, - dest_c_pixels, benchmark_height_, benchmark_height_, - benchmark_width_, kFilterNone); - MaskCpuFlags(benchmark_cpu_info_); // Enable all CPU optimization. - - for (int i = 0; i < benchmark_iterations_; ++i) { - ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, - benchmark_height_, dest_opt_pixels, benchmark_height_, - benchmark_height_, benchmark_width_, kFilterNone); - } - - for (int i = 0; i < kSize; ++i) { - EXPECT_EQ(dest_c_pixels[i], dest_opt_pixels[i]); - } - - free_aligned_buffer_page_end(dest_c_pixels); - free_aligned_buffer_page_end(dest_opt_pixels); - free_aligned_buffer_page_end(orig_pixels); -} - -TEST_F(LibYUVScaleTest, PlaneTestRotate_Bilinear) { - const int kSize = benchmark_width_ * benchmark_height_; - align_buffer_page_end(orig_pixels, kSize); - for (int i = 0; i < kSize; ++i) { - orig_pixels[i] = i; - } - align_buffer_page_end(dest_opt_pixels, kSize); - align_buffer_page_end(dest_c_pixels, kSize); - - MaskCpuFlags(disable_cpu_flags_); // Disable all CPU optimization. - ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, benchmark_height_, - dest_c_pixels, benchmark_height_, benchmark_height_, - benchmark_width_, kFilterBilinear); - MaskCpuFlags(benchmark_cpu_info_); // Enable all CPU optimization. - - for (int i = 0; i < benchmark_iterations_; ++i) { - ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, - benchmark_height_, dest_opt_pixels, benchmark_height_, - benchmark_height_, benchmark_width_, kFilterBilinear); - } - - for (int i = 0; i < kSize; ++i) { - EXPECT_EQ(dest_c_pixels[i], dest_opt_pixels[i]); - } - - free_aligned_buffer_page_end(dest_c_pixels); - free_aligned_buffer_page_end(dest_opt_pixels); - free_aligned_buffer_page_end(orig_pixels); -} - -// Intent is to test 200x50 to 50x200 but width and height can be parameters. -TEST_F(LibYUVScaleTest, PlaneTestRotate_Box) { - const int kSize = benchmark_width_ * benchmark_height_; - align_buffer_page_end(orig_pixels, kSize); - for (int i = 0; i < kSize; ++i) { - orig_pixels[i] = i; - } - align_buffer_page_end(dest_opt_pixels, kSize); - align_buffer_page_end(dest_c_pixels, kSize); - - MaskCpuFlags(disable_cpu_flags_); // Disable all CPU optimization. - ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, benchmark_height_, - dest_c_pixels, benchmark_height_, benchmark_height_, - benchmark_width_, kFilterBox); - MaskCpuFlags(benchmark_cpu_info_); // Enable all CPU optimization. - - for (int i = 0; i < benchmark_iterations_; ++i) { - ScalePlane(orig_pixels, benchmark_width_, benchmark_width_, - benchmark_height_, dest_opt_pixels, benchmark_height_, - benchmark_height_, benchmark_width_, kFilterBox); - } - - for (int i = 0; i < kSize; ++i) { - EXPECT_EQ(dest_c_pixels[i], dest_opt_pixels[i]); - } - - free_aligned_buffer_page_end(dest_c_pixels); - free_aligned_buffer_page_end(dest_opt_pixels); - free_aligned_buffer_page_end(orig_pixels); -} - -TEST_F(LibYUVScaleTest, PlaneTest1_Box) { - align_buffer_page_end(orig_pixels, 3); - align_buffer_page_end(dst_pixels, 3); - - // Pad the 1x1 byte image with invalid values before and after in case libyuv - // reads outside the memory boundaries. - orig_pixels[0] = 0; - orig_pixels[1] = 1; // scale this pixel - orig_pixels[2] = 2; - dst_pixels[0] = 3; - dst_pixels[1] = 3; - dst_pixels[2] = 3; - - libyuv::ScalePlane(orig_pixels + 1, /* src_stride= */ 1, /* src_width= */ 1, - /* src_height= */ 1, dst_pixels, /* dst_stride= */ 1, - /* dst_width= */ 1, /* dst_height= */ 2, - libyuv::kFilterBox); - - EXPECT_EQ(dst_pixels[0], 1); - EXPECT_EQ(dst_pixels[1], 1); - EXPECT_EQ(dst_pixels[2], 3); - - free_aligned_buffer_page_end(dst_pixels); - free_aligned_buffer_page_end(orig_pixels); -} - -TEST_F(LibYUVScaleTest, PlaneTest1_16_Box) { - align_buffer_page_end(orig_pixels_alloc, 3 * 2); - align_buffer_page_end(dst_pixels_alloc, 3 * 2); - uint16_t* orig_pixels = (uint16_t*)orig_pixels_alloc; - uint16_t* dst_pixels = (uint16_t*)dst_pixels_alloc; - - // Pad the 1x1 byte image with invalid values before and after in case libyuv - // reads outside the memory boundaries. - orig_pixels[0] = 0; - orig_pixels[1] = 1; // scale this pixel - orig_pixels[2] = 2; - dst_pixels[0] = 3; - dst_pixels[1] = 3; - dst_pixels[2] = 3; - - libyuv::ScalePlane_16( - orig_pixels + 1, /* src_stride= */ 1, /* src_width= */ 1, - /* src_height= */ 1, dst_pixels, /* dst_stride= */ 1, - /* dst_width= */ 1, /* dst_height= */ 2, libyuv::kFilterNone); - - EXPECT_EQ(dst_pixels[0], 1); - EXPECT_EQ(dst_pixels[1], 1); - EXPECT_EQ(dst_pixels[2], 3); - - free_aligned_buffer_page_end(dst_pixels_alloc); - free_aligned_buffer_page_end(orig_pixels_alloc); -} } // namespace libyuv |