libvpx: Pull from upstream

source/libvpx/test/convolve_test.cc

 /*
  *  Copyright (c) 2010 The WebM 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 <string.h>
 #include "test/acm_random.h"
+#include "test/clear_system_state.h"
 #include "test/register_state_check.h"
 #include "test/util.h"
 #include "third_party/googletest/src/include/gtest/gtest.h"
 
 #include "./vpx_config.h"
 #include "./vp9_rtcd.h"
+#include "vp9/common/vp9_common.h"
 #include "vp9/common/vp9_filter.h"
 #include "vpx_mem/vpx_mem.h"
 #include "vpx_ports/mem.h"
 
 namespace {
 
 static const unsigned int kMaxDimension = 64;
 
 typedef void (*ConvolveFunc)(const uint8_t *src, ptrdiff_t src_stride,
                              uint8_t *dst, ptrdiff_t dst_stride,
                              const int16_t *filter_x, int filter_x_stride,
                              const int16_t *filter_y, int filter_y_stride,
                              int w, int h);
 
 struct ConvolveFunctions {
-  ConvolveFunctions(ConvolveFunc h8, ConvolveFunc h8_avg,
+  ConvolveFunctions(ConvolveFunc copy, ConvolveFunc avg,
+                    ConvolveFunc h8, ConvolveFunc h8_avg,
                     ConvolveFunc v8, ConvolveFunc v8_avg,
                     ConvolveFunc hv8, ConvolveFunc hv8_avg,
                     int bd)
-      : h8_(h8), v8_(v8), hv8_(hv8), h8_avg_(h8_avg), v8_avg_(v8_avg),
-        hv8_avg_(hv8_avg), use_highbd_(bd) {}
+      : copy_(copy), avg_(avg), h8_(h8), v8_(v8), hv8_(hv8), h8_avg_(h8_avg),
+        v8_avg_(v8_avg), hv8_avg_(hv8_avg), use_highbd_(bd) {}
 
+  ConvolveFunc copy_;
+  ConvolveFunc avg_;
   ConvolveFunc h8_;
   ConvolveFunc v8_;
   ConvolveFunc hv8_;
   ConvolveFunc h8_avg_;
   ConvolveFunc v8_avg_;
   ConvolveFunc hv8_avg_;
   int use_highbd_;  // 0 if high bitdepth not used, else the actual bit depth.
 };
 
 typedef std::tr1::tuple<int, int, const ConvolveFunctions *> ConvolveParam;
@@ skipping 240 matching lines @@
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
 class ConvolveTest : public ::testing::TestWithParam<ConvolveParam> {
  public:
   static void SetUpTestCase() {
     // Force input_ to be unaligned, output to be 16 byte aligned.
     input_ = reinterpret_cast<uint8_t*>(
         vpx_memalign(kDataAlignment, kInputBufferSize + 1)) + 1;
     output_ = reinterpret_cast<uint8_t*>(
         vpx_memalign(kDataAlignment, kOutputBufferSize));
+    output_ref_ = reinterpret_cast<uint8_t*>(
+        vpx_memalign(kDataAlignment, kOutputBufferSize));
 #if CONFIG_VP9_HIGHBITDEPTH
     input16_ = reinterpret_cast<uint16_t*>(
         vpx_memalign(kDataAlignment,
                      (kInputBufferSize + 1) * sizeof(uint16_t))) + 1;
     output16_ = reinterpret_cast<uint16_t*>(
         vpx_memalign(kDataAlignment, (kOutputBufferSize) * sizeof(uint16_t)));
+    output16_ref_ = reinterpret_cast<uint16_t*>(
+        vpx_memalign(kDataAlignment, (kOutputBufferSize) * sizeof(uint16_t)));
 #endif
   }
 
+  virtual void TearDown() { libvpx_test::ClearSystemState(); }
+
   static void TearDownTestCase() {
     vpx_free(input_ - 1);
     input_ = NULL;
     vpx_free(output_);
     output_ = NULL;
+    vpx_free(output_ref_);
+    output_ref_ = NULL;
 #if CONFIG_VP9_HIGHBITDEPTH
     vpx_free(input16_ - 1);
     input16_ = NULL;
     vpx_free(output16_);
     output16_ = NULL;
+    vpx_free(output16_ref_);
+    output16_ref_ = NULL;
 #endif
   }
 
  protected:
   static const int kDataAlignment = 16;
   static const int kOuterBlockSize = 256;
   static const int kInputStride = kOuterBlockSize;
   static const int kOutputStride = kOuterBlockSize;
   static const int kInputBufferSize = kOuterBlockSize * kOuterBlockSize;
   static const int kOutputBufferSize = kOuterBlockSize * kOuterBlockSize;
@@ skipping 45 matching lines @@
     }
   }
 
   void SetConstantInput(int value) {
     memset(input_, value, kInputBufferSize);
 #if CONFIG_VP9_HIGHBITDEPTH
     vpx_memset16(input16_, value, kInputBufferSize);
 #endif
   }
 
+  void CopyOutputToRef() {
+    vpx_memcpy(output_ref_, output_, kOutputBufferSize);
+#if CONFIG_VP9_HIGHBITDEPTH
+    vpx_memcpy(output16_ref_, output16_, kOutputBufferSize);
+#endif
+  }
+
   void CheckGuardBlocks() {
     for (int i = 0; i < kOutputBufferSize; ++i) {
       if (IsIndexInBorder(i))
         EXPECT_EQ(255, output_[i]);
     }
   }
 
   uint8_t *input() const {
 #if CONFIG_VP9_HIGHBITDEPTH
     if (UUT_->use_highbd_ == 0) {
@@ skipping 13 matching lines @@
       return output_ + BorderTop() * kOuterBlockSize + BorderLeft();
     } else {
       return CONVERT_TO_BYTEPTR(output16_ + BorderTop() * kOuterBlockSize +
                                 BorderLeft());
     }
 #else
     return output_ + BorderTop() * kOuterBlockSize + BorderLeft();
 #endif
   }
 
+  uint8_t *output_ref() const {
+#if CONFIG_VP9_HIGHBITDEPTH
+    if (UUT_->use_highbd_ == 0) {
+      return output_ref_ + BorderTop() * kOuterBlockSize + BorderLeft();
+    } else {
+      return CONVERT_TO_BYTEPTR(output16_ref_ + BorderTop() * kOuterBlockSize +
+                                BorderLeft());
+    }
+#else
+    return output_ref_ + BorderTop() * kOuterBlockSize + BorderLeft();
+#endif
+  }
+
   uint16_t lookup(uint8_t *list, int index) const {
 #if CONFIG_VP9_HIGHBITDEPTH
     if (UUT_->use_highbd_ == 0) {
       return list[index];
     } else {
       return CONVERT_TO_SHORTPTR(list)[index];
     }
 #else
     return list[index];
 #endif
@@ skipping 58 matching lines @@
     }
 #else
     filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
                        dst_ptr, dst_stride, output_width, output_height);
 #endif
   }
 
   const ConvolveFunctions* UUT_;
   static uint8_t* input_;
   static uint8_t* output_;
+  static uint8_t* output_ref_;
 #if CONFIG_VP9_HIGHBITDEPTH
   static uint16_t* input16_;
   static uint16_t* output16_;
+  static uint16_t* output16_ref_;
   int mask_;
 #endif
 };
 
 uint8_t* ConvolveTest::input_ = NULL;
 uint8_t* ConvolveTest::output_ = NULL;
+uint8_t* ConvolveTest::output_ref_ = NULL;
 #if CONFIG_VP9_HIGHBITDEPTH
 uint16_t* ConvolveTest::input16_ = NULL;
 uint16_t* ConvolveTest::output16_ = NULL;
+uint16_t* ConvolveTest::output16_ref_ = NULL;
 #endif
 
 TEST_P(ConvolveTest, GuardBlocks) {
   CheckGuardBlocks();
 }
 
+TEST_P(ConvolveTest, Copy) {
+  uint8_t* const in = input();
+  uint8_t* const out = output();
+
+  ASM_REGISTER_STATE_CHECK(
+      UUT_->copy_(in, kInputStride, out, kOutputStride, NULL, 0, NULL, 0,
+                  Width(), Height()));
+
+  CheckGuardBlocks();
+
+  for (int y = 0; y < Height(); ++y)
+    for (int x = 0; x < Width(); ++x)
+      ASSERT_EQ(lookup(out, y * kOutputStride + x),
+                lookup(in, y * kInputStride + x))
+          << "(" << x << "," << y << ")";
+}
+
+TEST_P(ConvolveTest, Avg) {
+  uint8_t* const in = input();
+  uint8_t* const out = output();
+  uint8_t* const out_ref = output_ref();
+  CopyOutputToRef();
+
+  ASM_REGISTER_STATE_CHECK(
+      UUT_->avg_(in, kInputStride, out, kOutputStride, NULL, 0, NULL, 0,
+                Width(), Height()));
+
+  CheckGuardBlocks();
+
+  for (int y = 0; y < Height(); ++y)
+    for (int x = 0; x < Width(); ++x)
+      ASSERT_EQ(lookup(out, y * kOutputStride + x),
+                ROUND_POWER_OF_TWO(lookup(in, y * kInputStride + x) +
+                                   lookup(out_ref, y * kOutputStride + x), 1))
+          << "(" << x << "," << y << ")";
+}
+
 TEST_P(ConvolveTest, CopyHoriz) {
   uint8_t* const in = input();
   uint8_t* const out = output();
   DECLARE_ALIGNED(256, const int16_t, filter8[8]) = {0, 0, 0, 128, 0, 0, 0, 0};
 
   ASM_REGISTER_STATE_CHECK(
       UUT_->h8_(in, kInputStride, out, kOutputStride, filter8, 16, filter8, 16,
                 Width(), Height()));
 
   CheckGuardBlocks();
@@ skipping 657 matching lines @@
                                 int filter_x_stride,
                                 const int16_t *filter_y,
                                 int filter_y_stride,
                                 int w, int h) {
   vp9_highbd_convolve8_avg_sse2(src, src_stride, dst, dst_stride,
                                 filter_x, filter_x_stride,
                                 filter_y, filter_y_stride, w, h, 12);
 }
 #endif  // HAVE_SSE2 && ARCH_X86_64
 
+void wrap_convolve_copy_c_8(const uint8_t *src, ptrdiff_t src_stride,
+                            uint8_t *dst, ptrdiff_t dst_stride,
+                            const int16_t *filter_x,
+                            int filter_x_stride,
+                            const int16_t *filter_y,
+                            int filter_y_stride,
+                            int w, int h) {
+  vp9_highbd_convolve_copy_c(src, src_stride, dst, dst_stride,
+                             filter_x, filter_x_stride,
+                             filter_y, filter_y_stride, w, h, 8);
+}
+
+void wrap_convolve_avg_c_8(const uint8_t *src, ptrdiff_t src_stride,
+                           uint8_t *dst, ptrdiff_t dst_stride,
+                           const int16_t *filter_x,
+                           int filter_x_stride,
+                           const int16_t *filter_y,
+                           int filter_y_stride,
+                           int w, int h) {
+  vp9_highbd_convolve_avg_c(src, src_stride, dst, dst_stride,
+                            filter_x, filter_x_stride,
+                            filter_y, filter_y_stride, w, h, 8);
+}
+
 void wrap_convolve8_horiz_c_8(const uint8_t *src, ptrdiff_t src_stride,
                               uint8_t *dst, ptrdiff_t dst_stride,
                               const int16_t *filter_x,
                               int filter_x_stride,
                               const int16_t *filter_y,
                               int filter_y_stride,
                               int w, int h) {
   vp9_highbd_convolve8_horiz_c(src, src_stride, dst, dst_stride,
                                filter_x, filter_x_stride,
                                filter_y, filter_y_stride, w, h, 8);
@@ skipping 52 matching lines @@
                             const int16_t *filter_x,
                             int filter_x_stride,
                             const int16_t *filter_y,
                             int filter_y_stride,
                             int w, int h) {
   vp9_highbd_convolve8_avg_c(src, src_stride, dst, dst_stride,
                              filter_x, filter_x_stride,
                              filter_y, filter_y_stride, w, h, 8);
 }
 
+void wrap_convolve_copy_c_10(const uint8_t *src, ptrdiff_t src_stride,
+                             uint8_t *dst, ptrdiff_t dst_stride,
+                             const int16_t *filter_x,
+                             int filter_x_stride,
+                             const int16_t *filter_y,
+                             int filter_y_stride,
+                             int w, int h) {
+  vp9_highbd_convolve_copy_c(src, src_stride, dst, dst_stride,
+                             filter_x, filter_x_stride,
+                             filter_y, filter_y_stride, w, h, 10);
+}
+
+void wrap_convolve_avg_c_10(const uint8_t *src, ptrdiff_t src_stride,
+                            uint8_t *dst, ptrdiff_t dst_stride,
+                            const int16_t *filter_x,
+                            int filter_x_stride,
+                            const int16_t *filter_y,
+                            int filter_y_stride,
+                            int w, int h) {
+  vp9_highbd_convolve_avg_c(src, src_stride, dst, dst_stride,
+                            filter_x, filter_x_stride,
+                            filter_y, filter_y_stride, w, h, 10);
+}
+
 void wrap_convolve8_horiz_c_10(const uint8_t *src, ptrdiff_t src_stride,
                                uint8_t *dst, ptrdiff_t dst_stride,
                                const int16_t *filter_x,
                                int filter_x_stride,
                                const int16_t *filter_y,
                                int filter_y_stride,
                                int w, int h) {
   vp9_highbd_convolve8_horiz_c(src, src_stride, dst, dst_stride,
                                filter_x, filter_x_stride,
                                filter_y, filter_y_stride, w, h, 10);
@@ skipping 52 matching lines @@
                              const int16_t *filter_x,
                              int filter_x_stride,
                              const int16_t *filter_y,
                              int filter_y_stride,
                              int w, int h) {
   vp9_highbd_convolve8_avg_c(src, src_stride, dst, dst_stride,
                              filter_x, filter_x_stride,
                              filter_y, filter_y_stride, w, h, 10);
 }
 
+void wrap_convolve_copy_c_12(const uint8_t *src, ptrdiff_t src_stride,
+                             uint8_t *dst, ptrdiff_t dst_stride,
+                             const int16_t *filter_x,
+                             int filter_x_stride,
+                             const int16_t *filter_y,
+                             int filter_y_stride,
+                             int w, int h) {
+  vp9_highbd_convolve_copy_c(src, src_stride, dst, dst_stride,
+                             filter_x, filter_x_stride,
+                             filter_y, filter_y_stride, w, h, 12);
+}
+
+void wrap_convolve_avg_c_12(const uint8_t *src, ptrdiff_t src_stride,
+                            uint8_t *dst, ptrdiff_t dst_stride,
+                            const int16_t *filter_x,
+                            int filter_x_stride,
+                            const int16_t *filter_y,
+                            int filter_y_stride,
+                            int w, int h) {
+  vp9_highbd_convolve_avg_c(src, src_stride, dst, dst_stride,
+                            filter_x, filter_x_stride,
+                            filter_y, filter_y_stride, w, h, 12);
+}
+
 void wrap_convolve8_horiz_c_12(const uint8_t *src, ptrdiff_t src_stride,
                                uint8_t *dst, ptrdiff_t dst_stride,
                                const int16_t *filter_x,
                                int filter_x_stride,
                                const int16_t *filter_y,
                                int filter_y_stride,
                                int w, int h) {
   vp9_highbd_convolve8_horiz_c(src, src_stride, dst, dst_stride,
                                filter_x, filter_x_stride,
                                filter_y, filter_y_stride, w, h, 12);
@@ skipping 53 matching lines @@
                              int filter_x_stride,
                              const int16_t *filter_y,
                              int filter_y_stride,
                              int w, int h) {
   vp9_highbd_convolve8_avg_c(src, src_stride, dst, dst_stride,
                              filter_x, filter_x_stride,
                              filter_y, filter_y_stride, w, h, 12);
 }
 
 const ConvolveFunctions convolve8_c(
+    wrap_convolve_copy_c_8, wrap_convolve_avg_c_8,
     wrap_convolve8_horiz_c_8, wrap_convolve8_avg_horiz_c_8,
     wrap_convolve8_vert_c_8, wrap_convolve8_avg_vert_c_8,
     wrap_convolve8_c_8, wrap_convolve8_avg_c_8, 8);
 INSTANTIATE_TEST_CASE_P(C_8, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_c),
     make_tuple(8, 4, &convolve8_c),
     make_tuple(4, 8, &convolve8_c),
     make_tuple(8, 8, &convolve8_c),
     make_tuple(16, 8, &convolve8_c),
     make_tuple(8, 16, &convolve8_c),
     make_tuple(16, 16, &convolve8_c),
     make_tuple(32, 16, &convolve8_c),
     make_tuple(16, 32, &convolve8_c),
     make_tuple(32, 32, &convolve8_c),
     make_tuple(64, 32, &convolve8_c),
     make_tuple(32, 64, &convolve8_c),
     make_tuple(64, 64, &convolve8_c)));
 const ConvolveFunctions convolve10_c(
+    wrap_convolve_copy_c_10, wrap_convolve_avg_c_10,
     wrap_convolve8_horiz_c_10, wrap_convolve8_avg_horiz_c_10,
     wrap_convolve8_vert_c_10, wrap_convolve8_avg_vert_c_10,
     wrap_convolve8_c_10, wrap_convolve8_avg_c_10, 10);
 INSTANTIATE_TEST_CASE_P(C_10, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve10_c),
     make_tuple(8, 4, &convolve10_c),
     make_tuple(4, 8, &convolve10_c),
     make_tuple(8, 8, &convolve10_c),
     make_tuple(16, 8, &convolve10_c),
     make_tuple(8, 16, &convolve10_c),
     make_tuple(16, 16, &convolve10_c),
     make_tuple(32, 16, &convolve10_c),
     make_tuple(16, 32, &convolve10_c),
     make_tuple(32, 32, &convolve10_c),
     make_tuple(64, 32, &convolve10_c),
     make_tuple(32, 64, &convolve10_c),
     make_tuple(64, 64, &convolve10_c)));
 const ConvolveFunctions convolve12_c(
+    wrap_convolve_copy_c_12, wrap_convolve_avg_c_12,
     wrap_convolve8_horiz_c_12, wrap_convolve8_avg_horiz_c_12,
     wrap_convolve8_vert_c_12, wrap_convolve8_avg_vert_c_12,
     wrap_convolve8_c_12, wrap_convolve8_avg_c_12, 12);
 INSTANTIATE_TEST_CASE_P(C_12, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve12_c),
     make_tuple(8, 4, &convolve12_c),
     make_tuple(4, 8, &convolve12_c),
     make_tuple(8, 8, &convolve12_c),
     make_tuple(16, 8, &convolve12_c),
     make_tuple(8, 16, &convolve12_c),
     make_tuple(16, 16, &convolve12_c),
     make_tuple(32, 16, &convolve12_c),
     make_tuple(16, 32, &convolve12_c),
     make_tuple(32, 32, &convolve12_c),
     make_tuple(64, 32, &convolve12_c),
     make_tuple(32, 64, &convolve12_c),
     make_tuple(64, 64, &convolve12_c)));
 
 #else
 
 const ConvolveFunctions convolve8_c(
+    vp9_convolve_copy_c, vp9_convolve_avg_c,
     vp9_convolve8_horiz_c, vp9_convolve8_avg_horiz_c,
     vp9_convolve8_vert_c, vp9_convolve8_avg_vert_c,
     vp9_convolve8_c, vp9_convolve8_avg_c, 0);
 
 INSTANTIATE_TEST_CASE_P(C, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_c),
     make_tuple(8, 4, &convolve8_c),
     make_tuple(4, 8, &convolve8_c),
     make_tuple(8, 8, &convolve8_c),
     make_tuple(16, 8, &convolve8_c),
     make_tuple(8, 16, &convolve8_c),
     make_tuple(16, 16, &convolve8_c),
     make_tuple(32, 16, &convolve8_c),
     make_tuple(16, 32, &convolve8_c),
     make_tuple(32, 32, &convolve8_c),
     make_tuple(64, 32, &convolve8_c),
     make_tuple(32, 64, &convolve8_c),
     make_tuple(64, 64, &convolve8_c)));
 #endif
 
 #if HAVE_SSE2 && ARCH_X86_64
 #if CONFIG_VP9_HIGHBITDEPTH
 const ConvolveFunctions convolve8_sse2(
+    wrap_convolve_copy_c_8, wrap_convolve_avg_c_8,
     wrap_convolve8_horiz_sse2_8, wrap_convolve8_avg_horiz_sse2_8,
     wrap_convolve8_vert_sse2_8, wrap_convolve8_avg_vert_sse2_8,
     wrap_convolve8_sse2_8, wrap_convolve8_avg_sse2_8, 8);
 const ConvolveFunctions convolve10_sse2(
+    wrap_convolve_copy_c_10, wrap_convolve_avg_c_10,
     wrap_convolve8_horiz_sse2_10, wrap_convolve8_avg_horiz_sse2_10,
     wrap_convolve8_vert_sse2_10, wrap_convolve8_avg_vert_sse2_10,
     wrap_convolve8_sse2_10, wrap_convolve8_avg_sse2_10, 10);
 const ConvolveFunctions convolve12_sse2(
+    wrap_convolve_copy_c_12, wrap_convolve_avg_c_12,
     wrap_convolve8_horiz_sse2_12, wrap_convolve8_avg_horiz_sse2_12,
     wrap_convolve8_vert_sse2_12, wrap_convolve8_avg_vert_sse2_12,
     wrap_convolve8_sse2_12, wrap_convolve8_avg_sse2_12, 12);
 INSTANTIATE_TEST_CASE_P(SSE2, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_sse2),
     make_tuple(8, 4, &convolve8_sse2),
     make_tuple(4, 8, &convolve8_sse2),
     make_tuple(8, 8, &convolve8_sse2),
     make_tuple(16, 8, &convolve8_sse2),
     make_tuple(8, 16, &convolve8_sse2),
@@ skipping 25 matching lines @@
     make_tuple(8, 16, &convolve12_sse2),
     make_tuple(16, 16, &convolve12_sse2),
     make_tuple(32, 16, &convolve12_sse2),
     make_tuple(16, 32, &convolve12_sse2),
     make_tuple(32, 32, &convolve12_sse2),
     make_tuple(64, 32, &convolve12_sse2),
     make_tuple(32, 64, &convolve12_sse2),
     make_tuple(64, 64, &convolve12_sse2)));
 #else
 const ConvolveFunctions convolve8_sse2(
+    vp9_convolve_copy_sse2, vp9_convolve_avg_sse2,
     vp9_convolve8_horiz_sse2, vp9_convolve8_avg_horiz_sse2,
     vp9_convolve8_vert_sse2, vp9_convolve8_avg_vert_sse2,
     vp9_convolve8_sse2, vp9_convolve8_avg_sse2, 0);
 
 INSTANTIATE_TEST_CASE_P(SSE2, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_sse2),
     make_tuple(8, 4, &convolve8_sse2),
     make_tuple(4, 8, &convolve8_sse2),
     make_tuple(8, 8, &convolve8_sse2),
     make_tuple(16, 8, &convolve8_sse2),
     make_tuple(8, 16, &convolve8_sse2),
     make_tuple(16, 16, &convolve8_sse2),
     make_tuple(32, 16, &convolve8_sse2),
     make_tuple(16, 32, &convolve8_sse2),
     make_tuple(32, 32, &convolve8_sse2),
     make_tuple(64, 32, &convolve8_sse2),
     make_tuple(32, 64, &convolve8_sse2),
     make_tuple(64, 64, &convolve8_sse2)));
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 #endif
 
 #if HAVE_SSSE3
 const ConvolveFunctions convolve8_ssse3(
+    vp9_convolve_copy_c, vp9_convolve_avg_c,
     vp9_convolve8_horiz_ssse3, vp9_convolve8_avg_horiz_ssse3,
     vp9_convolve8_vert_ssse3, vp9_convolve8_avg_vert_ssse3,
     vp9_convolve8_ssse3, vp9_convolve8_avg_ssse3, 0);
 
 INSTANTIATE_TEST_CASE_P(SSSE3, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_ssse3),
     make_tuple(8, 4, &convolve8_ssse3),
     make_tuple(4, 8, &convolve8_ssse3),
     make_tuple(8, 8, &convolve8_ssse3),
     make_tuple(16, 8, &convolve8_ssse3),
     make_tuple(8, 16, &convolve8_ssse3),
     make_tuple(16, 16, &convolve8_ssse3),
     make_tuple(32, 16, &convolve8_ssse3),
     make_tuple(16, 32, &convolve8_ssse3),
     make_tuple(32, 32, &convolve8_ssse3),
     make_tuple(64, 32, &convolve8_ssse3),
     make_tuple(32, 64, &convolve8_ssse3),
     make_tuple(64, 64, &convolve8_ssse3)));
 #endif
 
 #if HAVE_AVX2 && HAVE_SSSE3
 const ConvolveFunctions convolve8_avx2(
+    vp9_convolve_copy_c, vp9_convolve_avg_c,
     vp9_convolve8_horiz_avx2, vp9_convolve8_avg_horiz_ssse3,
     vp9_convolve8_vert_avx2, vp9_convolve8_avg_vert_ssse3,
     vp9_convolve8_avx2, vp9_convolve8_avg_ssse3, 0);
 
 INSTANTIATE_TEST_CASE_P(AVX2, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_avx2),
     make_tuple(8, 4, &convolve8_avx2),
     make_tuple(4, 8, &convolve8_avx2),
     make_tuple(8, 8, &convolve8_avx2),
     make_tuple(8, 16, &convolve8_avx2),
     make_tuple(16, 8, &convolve8_avx2),
     make_tuple(16, 16, &convolve8_avx2),
     make_tuple(32, 16, &convolve8_avx2),
     make_tuple(16, 32, &convolve8_avx2),
     make_tuple(32, 32, &convolve8_avx2),
     make_tuple(64, 32, &convolve8_avx2),
     make_tuple(32, 64, &convolve8_avx2),
     make_tuple(64, 64, &convolve8_avx2)));
 #endif  // HAVE_AVX2 && HAVE_SSSE3
 
+#if HAVE_NEON
 #if HAVE_NEON_ASM
 const ConvolveFunctions convolve8_neon(
+    vp9_convolve_copy_neon, vp9_convolve_avg_neon,
     vp9_convolve8_horiz_neon, vp9_convolve8_avg_horiz_neon,
     vp9_convolve8_vert_neon, vp9_convolve8_avg_vert_neon,
     vp9_convolve8_neon, vp9_convolve8_avg_neon, 0);
+#else  // HAVE_NEON
+const ConvolveFunctions convolve8_neon(
+    vp9_convolve_copy_neon, vp9_convolve_avg_neon,
+    vp9_convolve8_horiz_neon, vp9_convolve8_avg_horiz_neon,
+    vp9_convolve8_vert_neon, vp9_convolve8_avg_vert_neon,
+    vp9_convolve8_neon, vp9_convolve8_avg_neon, 0);
+#endif  // HAVE_NEON_ASM
 
 INSTANTIATE_TEST_CASE_P(NEON, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_neon),
     make_tuple(8, 4, &convolve8_neon),
     make_tuple(4, 8, &convolve8_neon),
     make_tuple(8, 8, &convolve8_neon),
     make_tuple(16, 8, &convolve8_neon),
     make_tuple(8, 16, &convolve8_neon),
     make_tuple(16, 16, &convolve8_neon),
     make_tuple(32, 16, &convolve8_neon),
     make_tuple(16, 32, &convolve8_neon),
     make_tuple(32, 32, &convolve8_neon),
     make_tuple(64, 32, &convolve8_neon),
     make_tuple(32, 64, &convolve8_neon),
     make_tuple(64, 64, &convolve8_neon)));
-#endif
+#endif  // HAVE_NEON
 
 #if HAVE_DSPR2
 const ConvolveFunctions convolve8_dspr2(
+    vp9_convolve_copy_dspr2, vp9_convolve_avg_dspr2,
     vp9_convolve8_horiz_dspr2, vp9_convolve8_avg_horiz_dspr2,
     vp9_convolve8_vert_dspr2, vp9_convolve8_avg_vert_dspr2,
     vp9_convolve8_dspr2, vp9_convolve8_avg_dspr2, 0);
 
 INSTANTIATE_TEST_CASE_P(DSPR2, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_dspr2),
     make_tuple(8, 4, &convolve8_dspr2),
     make_tuple(4, 8, &convolve8_dspr2),
     make_tuple(8, 8, &convolve8_dspr2),
     make_tuple(16, 8, &convolve8_dspr2),
     make_tuple(8, 16, &convolve8_dspr2),
     make_tuple(16, 16, &convolve8_dspr2),
     make_tuple(32, 16, &convolve8_dspr2),
     make_tuple(16, 32, &convolve8_dspr2),
     make_tuple(32, 32, &convolve8_dspr2),
     make_tuple(64, 32, &convolve8_dspr2),
     make_tuple(32, 64, &convolve8_dspr2),
     make_tuple(64, 64, &convolve8_dspr2)));
 #endif
 }  // namespace