Commit c8b9d995 for libheif
commit c8b9d9953373ba51ab80d8a49721a6ad15eafb9c
Author: Dirk Farin <dirk.farin@gmail.com>
Date: Fri Oct 9 23:34:26 2026 +0200
tests: add rotate_benchmark, a timing harness for HeifPixelImage::rotate_ccw()
Rotates whole 4:2:0 images at 8 and 10 bit (4032x3024, 4096x3072 and
2048x1536) by 90, 180 and 270 degrees, including the allocation of the
output image, with the planes evicted from the caches before every run,
and prints the median and the minimum per configuration as a Markdown
table. It is built with the internal-symbol tests but not registered
with ctest. It needs a Release build with WITH_REDUCED_VISIBILITY=OFF;
the comment in the source has the recipe and explains how to compare
two versions of the library.
diff --git a/tests/CMakeLists.txt b/tests/CMakeLists.txt
index 1d3567ae..da4194c3 100644
--- a/tests/CMakeLists.txt
+++ b/tests/CMakeLists.txt
@@ -61,6 +61,11 @@ add_libheif_test(sdtp_sample_limit)
add_libheif_test(extract_area_plane_checks)
add_libheif_test(extend_to_size_checks)
add_libheif_test(rotate_planes)
+
+ # Benchmark for HeifPixelImage::rotate_ccw(), not a test; see the comment in the source.
+ add_executable(rotate_benchmark rotate_benchmark.cc)
+ target_link_libraries(rotate_benchmark PRIVATE heif)
+
add_libheif_test(pixelimage_overlay)
add_libheif_test(overlay_spec_parse)
add_libheif_test(add_channel_checks)
diff --git a/tests/rotate_benchmark.cc b/tests/rotate_benchmark.cc
new file mode 100644
index 00000000..8a78eca5
--- /dev/null
+++ b/tests/rotate_benchmark.cc
@@ -0,0 +1,143 @@
+/*
+ libheif unit tests
+
+ MIT License
+
+ Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+
+ Permission is hereby granted, free of charge, to any person obtaining a copy
+ of this software and associated documentation files (the "Software"), to deal
+ in the Software without restriction, including without limitation the rights
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ copies of the Software, and to permit persons to whom the Software is
+ furnished to do so, subject to the following conditions:
+
+ The above copyright notice and this permission notice shall be included in all
+ copies or substantial portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ SOFTWARE.
+*/
+
+// Benchmark for HeifPixelImage::rotate_ccw(). Rotates whole 4:2:0 images at 8 and 10 bit,
+// including the allocation of the output image, with the planes evicted from the caches before
+// every run, and prints the median and the minimum time per configuration as a Markdown table.
+//
+// This is not a test. It needs a Release build with the internal symbols visible:
+//
+// cmake -S . -B build-bench -DCMAKE_BUILD_TYPE=Release -DWITH_REDUCED_VISIBILITY=OFF -DBUILD_TESTING=ON
+// cmake --build build-bench --target rotate_benchmark
+// build-bench/tests/rotate_benchmark [rounds] [eviction buffer in MB]
+//
+// Whole-image timings drift by a few tenths of a millisecond between runs. To compare two versions
+// of the library, build both and run their benchmarks alternately a few times rather than one
+// after the other.
+
+#include "image/pixelimage.h"
+#include "libheif/heif.h"
+
+#include <algorithm>
+#include <chrono>
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <memory>
+#include <vector>
+
+
+template<typename T>
+static void fill_plane(HeifPixelImage& img, heif_channel ch)
+{
+ uint32_t w = img.get_width(ch);
+ uint32_t h = img.get_height(ch);
+ size_t stride;
+ T* p = img.get_channel_memory<T>(ch, &stride);
+ stride /= sizeof(T);
+ for (uint32_t y = 0; y < h; y++) {
+ for (uint32_t x = 0; x < w; x++) {
+ p[y * stride + x] = static_cast<T>((x * 2654435761U) ^ (y * 40503U));
+ }
+ }
+}
+
+
+static std::shared_ptr<HeifPixelImage> make_image(uint32_t w, uint32_t h, int bits,
+ const heif_security_limits* limits)
+{
+ auto img = std::make_shared<HeifPixelImage>();
+ img->create(w, h, heif_colorspace_YCbCr, heif_chroma_420);
+ for (heif_channel ch : {heif_channel_Y, heif_channel_Cb, heif_channel_Cr}) {
+ uint32_t cw = ch == heif_channel_Y ? w : (w + 1) / 2;
+ uint32_t chh = ch == heif_channel_Y ? h : (h + 1) / 2;
+ if (img->add_channel(ch, cw, chh, bits, limits, heif_component_datatype_unsigned_integer)) {
+ return nullptr;
+ }
+ if (bits <= 8) {
+ fill_plane<uint8_t>(*img, ch);
+ }
+ else {
+ fill_plane<uint16_t>(*img, ch);
+ }
+ }
+ return img;
+}
+
+
+int main(int argc, char** argv)
+{
+ int rounds = argc > 1 ? atoi(argv[1]) : 9;
+ int evict_mb = argc > 2 ? atoi(argv[2]) : 256;
+ if (rounds < 1 || evict_mb < 1) {
+ fprintf(stderr, "usage: %s [rounds] [eviction buffer in MB]\n", argv[0]);
+ return 1;
+ }
+
+ std::vector<uint8_t> evict(static_cast<size_t>(evict_mb) << 20);
+ const heif_security_limits* limits = heif_get_global_security_limits();
+
+ const uint32_t sizes[][2] = {{4032, 3024}, {4096, 3072}, {2048, 1536}};
+
+ printf("| image | luma stride | angle | median | min |\n");
+ printf("|---|---|---|---|---|\n");
+
+ for (auto& size : sizes) {
+ for (int bits : {8, 10}) {
+ auto img = make_image(size[0], size[1], bits, limits);
+ if (!img) {
+ fprintf(stderr, "cannot create a %ux%u %d-bit image\n", size[0], size[1], bits);
+ return 1;
+ }
+
+ size_t stride;
+ img->get_channel_memory(heif_channel_Y, &stride);
+
+ for (int angle : {90, 180, 270}) {
+ std::vector<double> ms;
+ for (int round = 0; round < rounds; round++) {
+ memset(evict.data(), round, evict.size());
+
+ auto t0 = std::chrono::steady_clock::now();
+ auto result = img->rotate_ccw(angle, limits);
+ auto t1 = std::chrono::steady_clock::now();
+
+ if (!result) {
+ fprintf(stderr, "rotation failed: %s\n", result.error().message.c_str());
+ return 1;
+ }
+ ms.push_back(std::chrono::duration<double, std::milli>(t1 - t0).count());
+ }
+
+ std::sort(ms.begin(), ms.end());
+ printf("| %ux%u 4:2:0 %d-bit | %zu | %d | %.2f ms | %.2f ms |\n",
+ size[0], size[1], bits, stride, angle, ms[ms.size() / 2], ms[0]);
+ }
+ }
+ }
+
+ return 0;
+}