Merging upstream version 127.0.

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

6 files changed, 272 insertions, 22 deletions

diff --git a/third_party/aom/test/aom_image_test.cc b/third_party/aom/test/aom_image_test.cc
index 03f4373f35..0dfb912215 100644
--- a/third_party/aom/test/aom_image_test.cc
+++ b/third_party/aom/test/aom_image_test.cc
@@ -9,6 +9,8 @@
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
  */
 
+#include <climits>
+
 #include "aom/aom_image.h"
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
@@ -70,3 +72,66 @@ TEST(AomImageTest, AomImgAllocNv12) {
   EXPECT_EQ(img.planes[AOM_PLANE_V], nullptr);
   aom_img_free(&img);
 }
+
+TEST(AomImageTest, AomImgAllocHugeWidth) {
+  // The stride (0x80000000 * 2) would overflow unsigned int.
+  aom_image_t *image =
+      aom_img_alloc(nullptr, AOM_IMG_FMT_I42016, 0x80000000, 1, 1);
+  ASSERT_EQ(image, nullptr);
+
+  // The stride (0x80000000) would overflow int.
+  image = aom_img_alloc(nullptr, AOM_IMG_FMT_I420, 0x80000000, 1, 1);
+  ASSERT_EQ(image, nullptr);
+
+  // The aligned width (UINT_MAX + 1) would overflow unsigned int.
+  image = aom_img_alloc(nullptr, AOM_IMG_FMT_I420, UINT_MAX, 1, 1);
+  ASSERT_EQ(image, nullptr);
+
+  image = aom_img_alloc_with_border(nullptr, AOM_IMG_FMT_I422, 1, INT_MAX, 1,
+                                    0x40000000, 0);
+  if (image) {
+    uint16_t *y_plane =
+        reinterpret_cast<uint16_t *>(image->planes[AOM_PLANE_Y]);
+    y_plane[0] = 0;
+    y_plane[image->d_w - 1] = 0;
+    aom_img_free(image);
+  }
+
+  image = aom_img_alloc(nullptr, AOM_IMG_FMT_I420, 0x7ffffffe, 1, 1);
+  if (image) {
+    aom_img_free(image);
+  }
+
+  image = aom_img_alloc(nullptr, AOM_IMG_FMT_I420, 285245883, 64, 1);
+  if (image) {
+    aom_img_free(image);
+  }
+
+  image = aom_img_alloc(nullptr, AOM_IMG_FMT_NV12, 285245883, 64, 1);
+  if (image) {
+    aom_img_free(image);
+  }
+
+  image = aom_img_alloc(nullptr, AOM_IMG_FMT_YV12, 285245883, 64, 1);
+  if (image) {
+    aom_img_free(image);
+  }
+
+  image = aom_img_alloc(nullptr, AOM_IMG_FMT_I42016, 65536, 2, 1);
+  if (image) {
+    uint16_t *y_plane =
+        reinterpret_cast<uint16_t *>(image->planes[AOM_PLANE_Y]);
+    y_plane[0] = 0;
+    y_plane[image->d_w - 1] = 0;
+    aom_img_free(image);
+  }
+
+  image = aom_img_alloc(nullptr, AOM_IMG_FMT_I42016, 285245883, 2, 1);
+  if (image) {
+    uint16_t *y_plane =
+        reinterpret_cast<uint16_t *>(image->planes[AOM_PLANE_Y]);
+    y_plane[0] = 0;
+    y_plane[image->d_w - 1] = 0;
+    aom_img_free(image);
+  }
+}
diff --git a/third_party/aom/test/disflow_test.cc b/third_party/aom/test/disflow_test.cc
index 4f004480e2..bee9e1261c 100644
--- a/third_party/aom/test/disflow_test.cc
+++ b/third_party/aom/test/disflow_test.cc
@@ -124,4 +124,9 @@ INSTANTIATE_TEST_SUITE_P(NEON, ComputeFlowTest,
                          ::testing::Values(aom_compute_flow_at_point_neon));
 #endif
 
+#if HAVE_SVE
+INSTANTIATE_TEST_SUITE_P(SVE, ComputeFlowTest,
+                         ::testing::Values(aom_compute_flow_at_point_sve));
+#endif
+
 }  // namespace
diff --git a/third_party/aom/test/ethread_test.cc b/third_party/aom/test/ethread_test.cc
index ce45394eb8..415f5de269 100644
--- a/third_party/aom/test/ethread_test.cc
+++ b/third_party/aom/test/ethread_test.cc
@@ -18,6 +18,7 @@
 #include "test/util.h"
 #include "test/y4m_video_source.h"
 #include "test/yuv_video_source.h"
+#include "av1/encoder/enc_enums.h"
 #include "av1/encoder/firstpass.h"
 
 namespace {
@@ -411,9 +412,7 @@ class AVxEncoderThreadTest
                                 const std::vector<size_t> ref_size_enc,
                                 const std::vector<std::string> ref_md5_enc,
                                 const std::vector<std::string> ref_md5_dec) {
-    // This value should be kept the same as MAX_NUM_THREADS
-    // in aom_thread.h
-    cfg_.g_threads = 64;
+    cfg_.g_threads = MAX_NUM_THREADS;
     ASSERT_NO_FATAL_FAILURE(RunLoop(video));
     std::vector<size_t> multi_thr_max_row_mt_size_enc;
     std::vector<std::string> multi_thr_max_row_mt_md5_enc;
diff --git a/third_party/aom/test/frame_resize_test.cc b/third_party/aom/test/frame_resize_test.cc
new file mode 100644
index 0000000000..8891304192
--- /dev/null
+++ b/third_party/aom/test/frame_resize_test.cc
@@ -0,0 +1,157 @@
+/*
+ * Copyright (c) 2024, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include "config/av1_rtcd.h"
+#include "test/acm_random.h"
+#include "test/util.h"
+#include "aom_ports/aom_timer.h"
+#include "aom_ports/bitops.h"
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+namespace {
+
+using ::testing::Combine;
+using ::testing::Values;
+using ::testing::ValuesIn;
+
+using std::make_tuple;
+using std::tuple;
+
+const int kIters = 1000;
+
+typedef tuple<int, int> FrameDimension;
+
+// Resolutions (width x height) to be tested for resizing.
+const FrameDimension kFrameDim[] = {
+  make_tuple(3840, 2160), make_tuple(2560, 1440), make_tuple(1920, 1080),
+  make_tuple(1280, 720),  make_tuple(640, 480),   make_tuple(640, 360),
+  make_tuple(256, 256),
+};
+
+// Check that two 8-bit output buffers are identical.
+void AssertOutputBufferEq(const uint8_t *p1, const uint8_t *p2, int width,
+                          int height) {
+  ASSERT_TRUE(p1 != p2) << "Buffers must be at different memory locations";
+  for (int j = 0; j < height; ++j) {
+    if (memcmp(p1, p2, sizeof(*p1) * width) == 0) {
+      p1 += width;
+      p2 += width;
+      continue;
+    }
+    for (int i = 0; i < width; ++i) {
+      ASSERT_EQ(p1[i], p2[i])
+          << width << "x" << height << " Pixel mismatch at (" << i << ", " << j
+          << ")";
+    }
+  }
+}
+
+typedef bool (*LowBDResizeFunc)(uint8_t *intbuf, uint8_t *output,
+                                int out_stride, int height, int height2,
+                                int stride, int start_wd);
+// Test parameter list:
+//  <tst_fun, dims>
+typedef tuple<LowBDResizeFunc, FrameDimension> ResizeTestParams;
+
+class AV1ResizeYTest : public ::testing::TestWithParam<ResizeTestParams> {
+ public:
+  void SetUp() {
+    test_fun_ = GET_PARAM(0);
+    frame_dim_ = GET_PARAM(1);
+    width_ = std::get<0>(frame_dim_);
+    height_ = std::get<1>(frame_dim_);
+    const int msb = get_msb(AOMMIN(width_, height_));
+    n_levels_ = AOMMAX(msb - MIN_PYRAMID_SIZE_LOG2, 1);
+
+    src_ = (uint8_t *)aom_malloc((width_ / 2) * height_ * sizeof(*src_));
+    ref_dest_ =
+        (uint8_t *)aom_calloc((width_ * height_) / 4, sizeof(*ref_dest_));
+    test_dest_ =
+        (uint8_t *)aom_calloc((width_ * height_) / 4, sizeof(*test_dest_));
+  }
+
+  void RunTest() {
+    int width2 = width_, height2 = height_;
+
+    for (int i = 0; i < (width_ / 2) * height_; i++) src_[i] = rng_.Rand8();
+    for (int level = 1; level < n_levels_; level++) {
+      width2 = (width_ >> level);
+      height2 = (height_ >> level);
+      resize_vert_dir_c(src_, ref_dest_, width2, height2 << 1, height2, width2,
+                        0);
+      test_fun_(src_, test_dest_, width2, height2 << 1, height2, width2, 0);
+
+      AssertOutputBufferEq(ref_dest_, test_dest_, width2, height2);
+    }
+  }
+
+  void SpeedTest() {
+    int width2 = width_, height2 = height_;
+
+    for (int i = 0; i < (width_ / 2) * height_; i++) src_[i] = rng_.Rand8();
+    for (int level = 1; level < n_levels_; level++) {
+      width2 = (width_ >> level);
+      height2 = (height_ >> level);
+      aom_usec_timer ref_timer;
+      aom_usec_timer_start(&ref_timer);
+      for (int j = 0; j < kIters; j++) {
+        resize_vert_dir_c(src_, ref_dest_, width2, height2 << 1, height2,
+                          width2, 0);
+      }
+      aom_usec_timer_mark(&ref_timer);
+      const int64_t ref_time = aom_usec_timer_elapsed(&ref_timer);
+
+      aom_usec_timer tst_timer;
+      aom_usec_timer_start(&tst_timer);
+      for (int j = 0; j < kIters; j++) {
+        test_fun_(src_, test_dest_, width2, height2 << 1, height2, width2, 0);
+      }
+      aom_usec_timer_mark(&tst_timer);
+      const int64_t tst_time = aom_usec_timer_elapsed(&tst_timer);
+
+      std::cout << "level: " << level << " [" << width2 << " x " << height2
+                << "] C time = " << ref_time << " , SIMD time = " << tst_time
+                << " scaling=" << float(1.00) * ref_time / tst_time << "x \n";
+    }
+  }
+
+  void TearDown() {
+    aom_free(src_);
+    aom_free(ref_dest_);
+    aom_free(test_dest_);
+  }
+
+ private:
+  LowBDResizeFunc test_fun_;
+  FrameDimension frame_dim_;
+  int width_;
+  int height_;
+  int n_levels_;
+  uint8_t *src_;
+  uint8_t *ref_dest_;
+  uint8_t *test_dest_;
+  libaom_test::ACMRandom rng_;
+};
+
+GTEST_ALLOW_UNINSTANTIATED_PARAMETERIZED_TEST(AV1ResizeYTest);
+
+TEST_P(AV1ResizeYTest, RunTest) { RunTest(); }
+
+TEST_P(AV1ResizeYTest, DISABLED_SpeedTest) { SpeedTest(); }
+
+#if HAVE_AVX2
+INSTANTIATE_TEST_SUITE_P(
+    AVX2, AV1ResizeYTest,
+    ::testing::Combine(::testing::Values(resize_vert_dir_avx2),
+                       ::testing::ValuesIn(kFrameDim)));
+#endif
+
+}  // namespace
diff --git a/third_party/aom/test/test.cmake b/third_party/aom/test/test.cmake
index e2f5da570d..2631c9fb39 100644
--- a/third_party/aom/test/test.cmake
+++ b/third_party/aom/test/test.cmake
@@ -209,6 +209,7 @@ if(NOT BUILD_SHARED_LIBS)
               "${AOM_ROOT}/test/fdct4x4_test.cc"
               "${AOM_ROOT}/test/fft_test.cc"
               "${AOM_ROOT}/test/firstpass_test.cc"
+              "${AOM_ROOT}/test/frame_resize_test.cc"
               "${AOM_ROOT}/test/fwht4x4_test.cc"
               "${AOM_ROOT}/test/hadamard_test.cc"
               "${AOM_ROOT}/test/horver_correlation_test.cc"
diff --git a/third_party/aom/test/wiener_test.cc b/third_party/aom/test/wiener_test.cc
index b995c84d8f..c38e10e3c2 100644
--- a/third_party/aom/test/wiener_test.cc
+++ b/third_party/aom/test/wiener_test.cc
@@ -397,6 +397,12 @@ INSTANTIATE_TEST_SUITE_P(NEON, WienerTest,
                          ::testing::Values(av1_compute_stats_neon));
 #endif  // HAVE_NEON
 
+#if HAVE_SVE
+
+INSTANTIATE_TEST_SUITE_P(SVE, WienerTest,
+                         ::testing::Values(av1_compute_stats_sve));
+#endif  // HAVE_SVE
+
 }  // namespace wiener_lowbd
 
 #if CONFIG_AV1_HIGHBITDEPTH
@@ -514,25 +520,27 @@ static void compute_stats_highbd_win_opt_c(int wiener_win, const uint8_t *dgd8,
 }
 
 void compute_stats_highbd_opt_c(int wiener_win, const uint8_t *dgd,
-                                const uint8_t *src, int h_start, int h_end,
-                                int v_start, int v_end, int dgd_stride,
-                                int src_stride, int64_t *M, int64_t *H,
-                                aom_bit_depth_t bit_depth) {
+                                const uint8_t *src, int16_t *d, int16_t *s,
+                                int h_start, int h_end, int v_start, int v_end,
+                                int dgd_stride, int src_stride, int64_t *M,
+                                int64_t *H, aom_bit_depth_t bit_depth) {
   if (wiener_win == WIENER_WIN || wiener_win == WIENER_WIN_CHROMA) {
     compute_stats_highbd_win_opt_c(wiener_win, dgd, src, h_start, h_end,
                                    v_start, v_end, dgd_stride, src_stride, M, H,
                                    bit_depth);
   } else {
-    av1_compute_stats_highbd_c(wiener_win, dgd, src, h_start, h_end, v_start,
-                               v_end, dgd_stride, src_stride, M, H, bit_depth);
+    av1_compute_stats_highbd_c(wiener_win, dgd, src, d, s, h_start, h_end,
+                               v_start, v_end, dgd_stride, src_stride, M, H,
+                               bit_depth);
   }
 }
 
 static const int kIterations = 100;
 typedef void (*compute_stats_Func)(int wiener_win, const uint8_t *dgd,
-                                   const uint8_t *src, int h_start, int h_end,
-                                   int v_start, int v_end, int dgd_stride,
-                                   int src_stride, int64_t *M, int64_t *H,
+                                   const uint8_t *src, int16_t *d, int16_t *s,
+                                   int h_start, int h_end, int v_start,
+                                   int v_end, int dgd_stride, int src_stride,
+                                   int64_t *M, int64_t *H,
                                    aom_bit_depth_t bit_depth);
 
 typedef std::tuple<const compute_stats_Func> WienerTestParam;
@@ -546,11 +554,17 @@ class WienerTestHighbd : public ::testing::TestWithParam<WienerTestParam> {
     dgd_buf = (uint16_t *)aom_memalign(
         32, MAX_DATA_BLOCK * MAX_DATA_BLOCK * sizeof(*dgd_buf));
     ASSERT_NE(dgd_buf, nullptr);
+    const size_t buf_size =
+        sizeof(*buf) * 6 * RESTORATION_UNITSIZE_MAX * RESTORATION_UNITSIZE_MAX;
+    buf = (int16_t *)aom_memalign(32, buf_size);
+    ASSERT_NE(buf, nullptr);
+    memset(buf, 0, buf_size);
     target_func_ = GET_PARAM(0);
   }
   void TearDown() override {
     aom_free(src_buf);
     aom_free(dgd_buf);
+    aom_free(buf);
   }
   void RunWienerTest(const int32_t wiener_win, int32_t run_times,
                      aom_bit_depth_t bit_depth);
@@ -562,6 +576,7 @@ class WienerTestHighbd : public ::testing::TestWithParam<WienerTestParam> {
   libaom_test::ACMRandom rng_;
   uint16_t *src_buf;
   uint16_t *dgd_buf;
+  int16_t *buf;
 };
 
 void WienerTestHighbd::RunWienerTest(const int32_t wiener_win,
@@ -589,6 +604,9 @@ void WienerTestHighbd::RunWienerTest(const int32_t wiener_win,
   const int dgd_stride = h_end;
   const int src_stride = MAX_DATA_BLOCK;
   const int iters = run_times == 1 ? kIterations : 2;
+  int16_t *dgd_avg = buf;
+  int16_t *src_avg =
+      buf + (3 * RESTORATION_UNITSIZE_MAX * RESTORATION_UNITSIZE_MAX);
   for (int iter = 0; iter < iters && !HasFatalFailure(); ++iter) {
     for (int i = 0; i < MAX_DATA_BLOCK * MAX_DATA_BLOCK; ++i) {
       dgd_buf[i] = rng_.Rand16() % (1 << bit_depth);
@@ -601,16 +619,17 @@ void WienerTestHighbd::RunWienerTest(const int32_t wiener_win,
     aom_usec_timer timer;
     aom_usec_timer_start(&timer);
     for (int i = 0; i < run_times; ++i) {
-      av1_compute_stats_highbd_c(wiener_win, dgd8, src8, h_start, h_end,
-                                 v_start, v_end, dgd_stride, src_stride, M_ref,
-                                 H_ref, bit_depth);
+      av1_compute_stats_highbd_c(wiener_win, dgd8, src8, dgd_avg, src_avg,
+                                 h_start, h_end, v_start, v_end, dgd_stride,
+                                 src_stride, M_ref, H_ref, bit_depth);
     }
     aom_usec_timer_mark(&timer);
     const double time1 = static_cast<double>(aom_usec_timer_elapsed(&timer));
     aom_usec_timer_start(&timer);
     for (int i = 0; i < run_times; ++i) {
-      target_func_(wiener_win, dgd8, src8, h_start, h_end, v_start, v_end,
-                   dgd_stride, src_stride, M_test, H_test, bit_depth);
+      target_func_(wiener_win, dgd8, src8, dgd_avg, src_avg, h_start, h_end,
+                   v_start, v_end, dgd_stride, src_stride, M_test, H_test,
+                   bit_depth);
     }
     aom_usec_timer_mark(&timer);
     const double time2 = static_cast<double>(aom_usec_timer_elapsed(&timer));
@@ -657,6 +676,9 @@ void WienerTestHighbd::RunWienerTest_ExtremeValues(const int32_t wiener_win,
   const int dgd_stride = h_end;
   const int src_stride = MAX_DATA_BLOCK;
   const int iters = 1;
+  int16_t *dgd_avg = buf;
+  int16_t *src_avg =
+      buf + (3 * RESTORATION_UNITSIZE_MAX * RESTORATION_UNITSIZE_MAX);
   for (int iter = 0; iter < iters && !HasFatalFailure(); ++iter) {
     // Fill with alternating extreme values to maximize difference with
     // the average.
@@ -668,12 +690,13 @@ void WienerTestHighbd::RunWienerTest_ExtremeValues(const int32_t wiener_win,
         dgd_buf + wiener_halfwin * MAX_DATA_BLOCK + wiener_halfwin);
     const uint8_t *src8 = CONVERT_TO_BYTEPTR(src_buf);
 
-    av1_compute_stats_highbd_c(wiener_win, dgd8, src8, h_start, h_end, v_start,
-                               v_end, dgd_stride, src_stride, M_ref, H_ref,
-                               bit_depth);
+    av1_compute_stats_highbd_c(wiener_win, dgd8, src8, dgd_avg, src_avg,
+                               h_start, h_end, v_start, v_end, dgd_stride,
+                               src_stride, M_ref, H_ref, bit_depth);
 
-    target_func_(wiener_win, dgd8, src8, h_start, h_end, v_start, v_end,
-                 dgd_stride, src_stride, M_test, H_test, bit_depth);
+    target_func_(wiener_win, dgd8, src8, dgd_avg, src_avg, h_start, h_end,
+                 v_start, v_end, dgd_stride, src_stride, M_test, H_test,
+                 bit_depth);
 
     int failed = 0;
     for (int i = 0; i < wiener_win2; ++i) {