Commit 03b412d0 authored by John Koleszar's avatar John Koleszar
Browse files

VP9 profile 0 release candidate

Merge experimental branch into master

Change-Id: Ie5f89fb977d28a4d98a8dcdf1c6eb97271a3c1db
parents a43bdcd7 d0ed677a
......@@ -103,6 +103,18 @@ test::
.PHONY: testdata
testdata::
# Add compiler flags for intrinsic files
$(BUILD_PFX)%_mmx.c.d: CFLAGS += -mmmx
$(BUILD_PFX)%_mmx.c.o: CFLAGS += -mmmx
$(BUILD_PFX)%_sse2.c.d: CFLAGS += -msse2
$(BUILD_PFX)%_sse2.c.o: CFLAGS += -msse2
$(BUILD_PFX)%_sse3.c.d: CFLAGS += -msse3
$(BUILD_PFX)%_sse3.c.o: CFLAGS += -msse3
$(BUILD_PFX)%_ssse3.c.d: CFLAGS += -mssse3
$(BUILD_PFX)%_ssse3.c.o: CFLAGS += -mssse3
$(BUILD_PFX)%_sse4.c.d: CFLAGS += -msse4.1
$(BUILD_PFX)%_sse4.c.o: CFLAGS += -msse4.1
$(BUILD_PFX)%.c.d: %.c
$(if $(quiet),@echo " [DEP] $@")
$(qexec)mkdir -p $(dir $@)
......
......@@ -266,12 +266,13 @@ else
fi
TMP_H="${TMPDIRx}/vpx-conf-$$-${RANDOM}.h"
TMP_C="${TMPDIRx}/vpx-conf-$$-${RANDOM}.c"
TMP_CC="${TMPDIRx}/vpx-conf-$$-${RANDOM}.cc"
TMP_O="${TMPDIRx}/vpx-conf-$$-${RANDOM}.o"
TMP_X="${TMPDIRx}/vpx-conf-$$-${RANDOM}.x"
TMP_ASM="${TMPDIRx}/vpx-conf-$$-${RANDOM}.asm"
clean_temp_files() {
rm -f ${TMP_C} ${TMP_H} ${TMP_O} ${TMP_X} ${TMP_ASM}
rm -f ${TMP_C} ${TMP_CC} ${TMP_H} ${TMP_O} ${TMP_X} ${TMP_ASM}
}
#
......@@ -292,9 +293,9 @@ check_cc() {
check_cxx() {
log check_cxx "$@"
cat >${TMP_C}
log_file ${TMP_C}
check_cmd ${CXX} ${CXXFLAGS} "$@" -c -o ${TMP_O} ${TMP_C}
cat >${TMP_CC}
log_file ${TMP_CC}
check_cmd ${CXX} ${CXXFLAGS} "$@" -c -o ${TMP_O} ${TMP_CC}
}
check_cpp() {
......@@ -1071,7 +1072,7 @@ EOF
tune_cflags="-march="
setup_gnu_toolchain
#for 32 bit x86 builds, -O3 did not turn on this flag
enabled optimizations && check_add_cflags -fomit-frame-pointer
enabled optimizations && disabled gprof && check_add_cflags -fomit-frame-pointer
;;
vs*)
# When building with Microsoft Visual Studio the assembler is
......
......@@ -243,19 +243,11 @@ HAVE_LIST="
unistd_h
"
EXPERIMENT_LIST="
csm
new_mvref
implicit_segmentation
newbintramodes
comp_interintra_pred
enable_6tap
abovesprefmv
code_nonzerocount
useselectrefmv
modelcoefprob
loop_dering
implicit_compoundinter_weight
scatterscan
oneshotq
multiple_arf
non420
alpha
balanced_coeftree
"
CONFIG_LIST="
external_build
......@@ -608,7 +600,10 @@ process_toolchain() {
check_add_cflags -Wimplicit-function-declaration
check_add_cflags -Wuninitialized
check_add_cflags -Wunused-variable
check_add_cflags -Wunused-but-set-variable
case ${CC} in
*clang*) ;;
*) check_add_cflags -Wunused-but-set-variable ;;
esac
enabled extra_warnings || check_add_cflags -Wno-unused-function
fi
......
......@@ -34,6 +34,13 @@ class ACMRandom {
return (value >> 24) & 0xff;
}
uint8_t Rand8Extremes(void) {
// Returns a random value near 0 or near 255, to better exercise
// saturation behavior.
const uint8_t r = Rand8();
return r < 128 ? r << 4 : r >> 4;
}
int PseudoUniform(int range) {
return random_.Generate(range);
}
......
/*
* Copyright (c) 2012 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 <climits>
#include <vector>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "test/util.h"
namespace {
class BordersTest : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
BordersTest() : EncoderTest(GET_PARAM(0)) {}
virtual void SetUp() {
InitializeConfig();
SetMode(GET_PARAM(1));
}
virtual bool Continue() const {
return !HasFatalFailure() && !abort_;
}
virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
::libvpx_test::Encoder *encoder) {
if ( video->frame() == 1) {
encoder->Control(VP8E_SET_CPUUSED, 0);
encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 1);
encoder->Control(VP8E_SET_ARNR_MAXFRAMES, 7);
encoder->Control(VP8E_SET_ARNR_STRENGTH, 5);
encoder->Control(VP8E_SET_ARNR_TYPE, 3);
}
}
virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
if (pkt->data.frame.flags & VPX_FRAME_IS_KEY) {
}
}
};
TEST_P(BordersTest, TestEncodeHighBitrate) {
// Validate that this non multiple of 64 wide clip encodes and decodes
// without a mismatch when passing in a very low max q. This pushes
// the encoder to producing lots of big partitions which will likely
// extend into the border and test the border condition.
cfg_.g_lag_in_frames = 25;
cfg_.rc_2pass_vbr_minsection_pct = 5;
cfg_.rc_2pass_vbr_minsection_pct = 2000;
cfg_.rc_target_bitrate = 2000;
cfg_.rc_max_quantizer = 10;
::libvpx_test::I420VideoSource video("hantro_odd.yuv", 208, 144, 30, 1, 0,
40);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
}
TEST_P(BordersTest, TestLowBitrate) {
// Validate that this clip encodes and decodes without a mismatch
// when passing in a very high min q. This pushes the encoder to producing
// lots of small partitions which might will test the other condition.
cfg_.g_lag_in_frames = 25;
cfg_.rc_2pass_vbr_minsection_pct = 5;
cfg_.rc_2pass_vbr_minsection_pct = 2000;
cfg_.rc_target_bitrate = 200;
cfg_.rc_min_quantizer = 40;
::libvpx_test::I420VideoSource video("hantro_odd.yuv", 208, 144, 30, 1, 0,
40);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
}
VP9_INSTANTIATE_TEST_CASE(BordersTest, ::testing::Values(
::libvpx_test::kTwoPassGood));
} // namespace
......@@ -8,6 +8,10 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "test/acm_random.h"
#include "test/register_state_check.h"
#include "test/util.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
extern "C" {
#include "./vpx_config.h"
......@@ -16,10 +20,6 @@ extern "C" {
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"
}
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/acm_random.h"
#include "test/register_state_check.h"
#include "test/util.h"
namespace {
typedef void (*convolve_fn_t)(const uint8_t *src, int src_stride,
......@@ -46,27 +46,27 @@ struct ConvolveFunctions {
// Reference 8-tap subpixel filter, slightly modified to fit into this test.
#define VP9_FILTER_WEIGHT 128
#define VP9_FILTER_SHIFT 7
static uint8_t clip_pixel(int x) {
uint8_t clip_pixel(int x) {
return x < 0 ? 0 :
x > 255 ? 255 :
x;
}
static void filter_block2d_8_c(const uint8_t *src_ptr,
const unsigned int src_stride,
const int16_t *HFilter,
const int16_t *VFilter,
uint8_t *dst_ptr,
unsigned int dst_stride,
unsigned int output_width,
unsigned int output_height) {
void filter_block2d_8_c(const uint8_t *src_ptr,
const unsigned int src_stride,
const int16_t *HFilter,
const int16_t *VFilter,
uint8_t *dst_ptr,
unsigned int dst_stride,
unsigned int output_width,
unsigned int output_height) {
// Between passes, we use an intermediate buffer whose height is extended to
// have enough horizontally filtered values as input for the vertical pass.
// This buffer is allocated to be big enough for the largest block type we
// support.
const int kInterp_Extend = 4;
const unsigned int intermediate_height =
(kInterp_Extend - 1) + output_height + kInterp_Extend;
(kInterp_Extend - 1) + output_height + kInterp_Extend;
/* Size of intermediate_buffer is max_intermediate_height * filter_max_width,
* where max_intermediate_height = (kInterp_Extend - 1) + filter_max_height
......@@ -75,7 +75,7 @@ static void filter_block2d_8_c(const uint8_t *src_ptr,
* = 23
* and filter_max_width = 16
*/
uint8_t intermediate_buffer[23 * 16];
uint8_t intermediate_buffer[71 * 64];
const int intermediate_next_stride = 1 - intermediate_height * output_width;
// Horizontal pass (src -> transposed intermediate).
......@@ -87,15 +87,15 @@ static void filter_block2d_8_c(const uint8_t *src_ptr,
for (i = 0; i < intermediate_height; ++i) {
for (j = 0; j < output_width; ++j) {
// Apply filter...
int temp = ((int)src_ptr[0] * HFilter[0]) +
((int)src_ptr[1] * HFilter[1]) +
((int)src_ptr[2] * HFilter[2]) +
((int)src_ptr[3] * HFilter[3]) +
((int)src_ptr[4] * HFilter[4]) +
((int)src_ptr[5] * HFilter[5]) +
((int)src_ptr[6] * HFilter[6]) +
((int)src_ptr[7] * HFilter[7]) +
(VP9_FILTER_WEIGHT >> 1); // Rounding
const int temp = (src_ptr[0] * HFilter[0]) +
(src_ptr[1] * HFilter[1]) +
(src_ptr[2] * HFilter[2]) +
(src_ptr[3] * HFilter[3]) +
(src_ptr[4] * HFilter[4]) +
(src_ptr[5] * HFilter[5]) +
(src_ptr[6] * HFilter[6]) +
(src_ptr[7] * HFilter[7]) +
(VP9_FILTER_WEIGHT >> 1); // Rounding
// Normalize back to 0-255...
*output_ptr = clip_pixel(temp >> VP9_FILTER_SHIFT);
......@@ -115,15 +115,15 @@ static void filter_block2d_8_c(const uint8_t *src_ptr,
for (i = 0; i < output_height; ++i) {
for (j = 0; j < output_width; ++j) {
// Apply filter...
int temp = ((int)src_ptr[0] * VFilter[0]) +
((int)src_ptr[1] * VFilter[1]) +
((int)src_ptr[2] * VFilter[2]) +
((int)src_ptr[3] * VFilter[3]) +
((int)src_ptr[4] * VFilter[4]) +
((int)src_ptr[5] * VFilter[5]) +
((int)src_ptr[6] * VFilter[6]) +
((int)src_ptr[7] * VFilter[7]) +
(VP9_FILTER_WEIGHT >> 1); // Rounding
const int temp = (src_ptr[0] * VFilter[0]) +
(src_ptr[1] * VFilter[1]) +
(src_ptr[2] * VFilter[2]) +
(src_ptr[3] * VFilter[3]) +
(src_ptr[4] * VFilter[4]) +
(src_ptr[5] * VFilter[5]) +
(src_ptr[6] * VFilter[6]) +
(src_ptr[7] * VFilter[7]) +
(VP9_FILTER_WEIGHT >> 1); // Rounding
// Normalize back to 0-255...
*dst_ptr++ = clip_pixel(temp >> VP9_FILTER_SHIFT);
......@@ -135,12 +135,12 @@ static void filter_block2d_8_c(const uint8_t *src_ptr,
}
}
static void block2d_average_c(uint8_t *src,
unsigned int src_stride,
uint8_t *output_ptr,
unsigned int output_stride,
unsigned int output_width,
unsigned int output_height) {
void block2d_average_c(uint8_t *src,
unsigned int src_stride,
uint8_t *output_ptr,
unsigned int output_stride,
unsigned int output_width,
unsigned int output_height) {
unsigned int i, j;
for (i = 0; i < output_height; ++i) {
for (j = 0; j < output_width; ++j) {
......@@ -150,21 +150,21 @@ static void block2d_average_c(uint8_t *src,
}
}
static void filter_average_block2d_8_c(const uint8_t *src_ptr,
const unsigned int src_stride,
const int16_t *HFilter,
const int16_t *VFilter,
uint8_t *dst_ptr,
unsigned int dst_stride,
unsigned int output_width,
unsigned int output_height) {
uint8_t tmp[16*16];
assert(output_width <= 16);
assert(output_height <= 16);
filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter, tmp, 16,
void filter_average_block2d_8_c(const uint8_t *src_ptr,
const unsigned int src_stride,
const int16_t *HFilter,
const int16_t *VFilter,
uint8_t *dst_ptr,
unsigned int dst_stride,
unsigned int output_width,
unsigned int output_height) {
uint8_t tmp[64 * 64];
assert(output_width <= 64);
assert(output_height <= 64);
filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter, tmp, 64,
output_width, output_height);
block2d_average_c(tmp, 16, dst_ptr, dst_stride,
block2d_average_c(tmp, 64, dst_ptr, dst_stride,
output_width, output_height);
}
......@@ -173,10 +173,9 @@ class ConvolveTest : public PARAMS(int, int, const ConvolveFunctions*) {
static void SetUpTestCase() {
// Force input_ to be unaligned, output to be 16 byte aligned.
input_ = reinterpret_cast<uint8_t*>(
vpx_memalign(kDataAlignment, kOuterBlockSize * kOuterBlockSize + 1))
+ 1;
vpx_memalign(kDataAlignment, kInputBufferSize + 1)) + 1;
output_ = reinterpret_cast<uint8_t*>(
vpx_memalign(kDataAlignment, kOuterBlockSize * kOuterBlockSize));
vpx_memalign(kDataAlignment, kOutputBufferSize));
}
static void TearDownTestCase() {
......@@ -186,62 +185,63 @@ class ConvolveTest : public PARAMS(int, int, const ConvolveFunctions*) {
output_ = NULL;
}
protected:
static const int kDataAlignment = 16;
static const int kOuterBlockSize = 32;
static const int kInputStride = kOuterBlockSize;
static const int kOutputStride = kOuterBlockSize;
static const int kMaxDimension = 16;
int Width() const { return GET_PARAM(0); }
int Height() const { return GET_PARAM(1); }
int BorderLeft() const {
const int center = (kOuterBlockSize - Width()) / 2;
return (center + (kDataAlignment - 1)) & ~(kDataAlignment - 1);
}
int BorderTop() const { return (kOuterBlockSize - Height()) / 2; }
bool IsIndexInBorder(int i) {
return (i < BorderTop() * kOuterBlockSize ||
i >= (BorderTop() + Height()) * kOuterBlockSize ||
i % kOuterBlockSize < BorderLeft() ||
i % kOuterBlockSize >= (BorderLeft() + Width()));
}
protected:
static const int kDataAlignment = 16;
static const int kOuterBlockSize = 128;
static const int kInputStride = kOuterBlockSize;
static const int kOutputStride = kOuterBlockSize;
static const int kMaxDimension = 64;
static const int kInputBufferSize = kOuterBlockSize * kOuterBlockSize;
static const int kOutputBufferSize = kOuterBlockSize * kOuterBlockSize;
int Width() const { return GET_PARAM(0); }
int Height() const { return GET_PARAM(1); }
int BorderLeft() const {
const int center = (kOuterBlockSize - Width()) / 2;
return (center + (kDataAlignment - 1)) & ~(kDataAlignment - 1);
}
int BorderTop() const { return (kOuterBlockSize - Height()) / 2; }
virtual void SetUp() {
UUT_ = GET_PARAM(2);
memset(input_, 0, sizeof(input_));
/* Set up guard blocks for an inner block cetered in the outer block */
for (int i = 0; i < kOuterBlockSize * kOuterBlockSize; ++i) {
if (IsIndexInBorder(i))
output_[i] = 255;
else
output_[i] = 0;
}
bool IsIndexInBorder(int i) {
return (i < BorderTop() * kOuterBlockSize ||
i >= (BorderTop() + Height()) * kOuterBlockSize ||
i % kOuterBlockSize < BorderLeft() ||
i % kOuterBlockSize >= (BorderLeft() + Width()));
}
::libvpx_test::ACMRandom prng;
for (int i = 0; i < kOuterBlockSize * kOuterBlockSize; ++i)
input_[i] = prng.Rand8();
virtual void SetUp() {
UUT_ = GET_PARAM(2);
/* Set up guard blocks for an inner block cetered in the outer block */
for (int i = 0; i < kOutputBufferSize; ++i) {
if (IsIndexInBorder(i))
output_[i] = 255;
else
output_[i] = 0;
}
void CheckGuardBlocks() {
for (int i = 0; i < kOuterBlockSize * kOuterBlockSize; ++i) {
if (IsIndexInBorder(i))
EXPECT_EQ(255, output_[i]);
}
}
::libvpx_test::ACMRandom prng;
for (int i = 0; i < kInputBufferSize; ++i)
input_[i] = prng.Rand8Extremes();
}
uint8_t* input() {
return input_ + BorderTop() * kOuterBlockSize + BorderLeft();
void CheckGuardBlocks() {
for (int i = 0; i < kOutputBufferSize; ++i) {
if (IsIndexInBorder(i))
EXPECT_EQ(255, output_[i]);
}
}
uint8_t* output() {
return output_ + BorderTop() * kOuterBlockSize + BorderLeft();
}
uint8_t* input() const {
return input_ + BorderTop() * kOuterBlockSize + BorderLeft();
}
const ConvolveFunctions* UUT_;
static uint8_t* input_;
static uint8_t* output_;
uint8_t* output() const {
return output_ + BorderTop() * kOuterBlockSize + BorderLeft();
}
const ConvolveFunctions* UUT_;
static uint8_t* input_;
static uint8_t* output_;
};
uint8_t* ConvolveTest::input_ = NULL;
uint8_t* ConvolveTest::output_ = NULL;
......@@ -303,11 +303,33 @@ TEST_P(ConvolveTest, Copy2D) {
const int16_t (*kTestFilterList[])[8] = {
vp9_bilinear_filters,
vp9_sub_pel_filters_6,
vp9_sub_pel_filters_8,
vp9_sub_pel_filters_8s,
vp9_sub_pel_filters_8lp
};
const int kNumFilterBanks = sizeof(kTestFilterList) /
sizeof(kTestFilterList[0]);
const int kNumFilters = 16;
TEST(ConvolveTest, FiltersWontSaturateWhenAddedPairwise) {
for (int filter_bank = 0; filter_bank < kNumFilterBanks; ++filter_bank) {
const int16_t (*filters)[8] = kTestFilterList[filter_bank];
for (int i = 0; i < kNumFilters; i++) {
const int p0 = filters[i][0] + filters[i][1];
const int p1 = filters[i][2] + filters[i][3];
const int p2 = filters[i][4] + filters[i][5];
const int p3 = filters[i][6] + filters[i][7];
EXPECT_LE(p0, 128);
EXPECT_LE(p1, 128);
EXPECT_LE(p2, 128);
EXPECT_LE(p3, 128);
EXPECT_LE(p0 + p3, 128);
EXPECT_LE(p0 + p3 + p1, 128);
EXPECT_LE(p0 + p3 + p1 + p2, 128);
EXPECT_EQ(p0 + p1 + p2 + p3, 128);
}
}
}
const int16_t kInvalidFilter[8] = { 0 };
......@@ -316,12 +338,9 @@ TEST_P(ConvolveTest, MatchesReferenceSubpixelFilter) {
uint8_t* const out = output();
uint8_t ref[kOutputStride * kMaxDimension];
const int kNumFilterBanks = sizeof(kTestFilterList) /
sizeof(kTestFilterList[0]);
for (int filter_bank = 0; filter_bank < kNumFilterBanks; ++filter_bank) {
const int16_t (*filters)[8] = kTestFilterList[filter_bank];
const int kNumFilters = 16;
for (int filter_x = 0; filter_x < kNumFilters; ++filter_x) {
for (int filter_y = 0; filter_y < kNumFilters; ++filter_y) {
......@@ -368,7 +387,7 @@ TEST_P(ConvolveTest, MatchesReferenceAveragingSubpixelFilter) {
::libvpx_test::ACMRandom prng;
for (int y = 0; y < Height(); ++y) {
for (int x = 0; x < Width(); ++x) {
const uint8_t r = prng.Rand8();
const uint8_t r = prng.Rand8Extremes();
out[y * kOutputStride + x] = r;
ref[y * kOutputStride + x] = r;
......@@ -440,16 +459,17 @@ DECLARE_ALIGNED(256, const int16_t, kChangeFilters[16][8]) = {
TEST_P(ConvolveTest, ChangeFilterWorks) {
uint8_t* const in = input();
uint8_t* const out = output();
const int kPixelSelected = 4;
REGISTER_STATE_CHECK(UUT_->h8_(in, kInputStride, out, kOutputStride,
kChangeFilters[8], 17, kChangeFilters[4], 16,
Width(), Height()));
for (int x = 0; x < Width(); ++x) {
if (x < 8)
ASSERT_EQ(in[4], out[x]) << "x == " << x;
else
ASSERT_EQ(in[12], out[x]) << "x == " << x;
const int kQ4StepAdjust = x >> 4;
const int kFilterPeriodAdjust = (x >> 3) << 3;
const int ref_x = kQ4StepAdjust + kFilterPeriodAdjust + kPixelSelected;
ASSERT_EQ(in[ref_x], out[x]) << "x == " << x;
}
REGISTER_STATE_CHECK(UUT_->v8_(in, kInputStride, out, kOutputStride,
......@@ -457,10 +477,10 @@ TEST_P(ConvolveTest, ChangeFilterWorks) {
Width(), Height()));
for (int y = 0; y < Height(); ++y) {
if (y < 8)
ASSERT_EQ(in[4 * kInputStride], out[y * kOutputStride]) << "y == " << y;
else
ASSERT_EQ(in[12 * kInputStride], out[y * kOutputStride]) << "y == " << y;
const int kQ4StepAdjust = y >> 4;
const int kFilterPeriodAdjust = (y >> 3) << 3;
const int ref_y = kQ4StepAdjust + kFilterPeriodAdjust + kPixelSelected;
ASSERT_EQ(in[ref_y * kInputStride], out[y * kInputStride]) << "y == " << y;
}
REGISTER_STATE_CHECK(UUT_->hv8_(in, kInputStride, out, kOutputStride,
......@@ -468,9 +488,13 @@ TEST_P(ConvolveTest, ChangeFilterWorks) {
Width(), Height()));
for (int y = 0; y < Height(); ++y) {
const int kQ4StepAdjustY = y >> 4;
const int kFilterPeriodAdjustY = (y >> 3) << 3;
const int ref_y = kQ4StepAdjustY + kFilterPeriodAdjustY + kPixelSelected;
for (int x = 0; x < Width(); ++x) {
const int ref_x = x < 8 ? 4 : 12;
const int ref_y = y < 8 ? 4 : 12;
const int kQ4StepAdjustX = x >> 4;
const int kFilterPeriodAdjustX = (x >> 3) << 3;
const int ref_x = kQ4StepAdjustX + kFilterPeriodAdjustX + kPixelSelected;
ASSERT_EQ(in[ref_y * kInputStride + ref_x], out[y * kOutputStride + x])
<< "x == " << x << ", y == " << y;
......@@ -489,10 +513,17 @@ const ConvolveFunctions convolve8_c(
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),