Adds masked variance and sad functions for wedge
Adds masked variance and sad functions needed for wedge prediction modes to come. Change-Id: I25b231bbc345e6a494316abb0a7d5cd5586a3a54
This commit is contained in:
Debargha Mukherjee
209
test/masked_sad_test.cc
Normal file
209
test/masked_sad_test.cc
Normal file
@@ -0,0 +1,209 @@
|
||||
/*
|
||||
* Copyright (c) 2015 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 <math.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
|
||||
#include "third_party/googletest/src/include/gtest/gtest.h"
|
||||
#include "test/acm_random.h"
|
||||
#include "test/clear_system_state.h"
|
||||
#include "test/register_state_check.h"
|
||||
#include "test/util.h"
|
||||
|
||||
#include "./vpx_config.h"
|
||||
#include "./vpx_dsp_rtcd.h"
|
||||
#include "vpx/vpx_integer.h"
|
||||
|
||||
using libvpx_test::ACMRandom;
|
||||
|
||||
namespace {
|
||||
const int number_of_iterations = 500;
|
||||
|
||||
typedef unsigned int (*MaskedSADFunc)(const uint8_t *a, int a_stride,
|
||||
const uint8_t *b, int b_stride,
|
||||
const uint8_t *m, int m_stride);
|
||||
typedef std::tr1::tuple<MaskedSADFunc, MaskedSADFunc> MaskedSADParam;
|
||||
|
||||
class MaskedSADTest : public ::testing::TestWithParam<MaskedSADParam> {
|
||||
public:
|
||||
virtual ~MaskedSADTest() {}
|
||||
virtual void SetUp() {
|
||||
maskedSAD_op_ = GET_PARAM(0);
|
||||
ref_maskedSAD_op_ = GET_PARAM(1);
|
||||
}
|
||||
|
||||
virtual void TearDown() { libvpx_test::ClearSystemState(); }
|
||||
|
||||
protected:
|
||||
MaskedSADFunc maskedSAD_op_;
|
||||
MaskedSADFunc ref_maskedSAD_op_;
|
||||
};
|
||||
|
||||
TEST_P(MaskedSADTest, OperationCheck) {
|
||||
unsigned int ref_ret, ret;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint8_t, src_ptr[4096]);
|
||||
DECLARE_ALIGNED(16, uint8_t, ref_ptr[4096]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[4096]);
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = 64;
|
||||
int ref_stride = 64;
|
||||
int msk_stride = 64;
|
||||
for (int i = 0; i < number_of_iterations; ++i) {
|
||||
for (int j = 0; j < 4096; j++) {
|
||||
src_ptr[j] = rnd.Rand8();
|
||||
ref_ptr[j] = rnd.Rand8();
|
||||
msk_ptr[j] = ((rnd.Rand8()&0x7f) > 64) ? rnd.Rand8()&0x3f : 64;
|
||||
assert(msk_ptr[j] <= 64);
|
||||
}
|
||||
|
||||
ref_ret = ref_maskedSAD_op_(src_ptr, src_stride, ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride);
|
||||
ASM_REGISTER_STATE_CHECK(ret = maskedSAD_op_(src_ptr, src_stride,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride));
|
||||
if (ret != ref_ret) {
|
||||
err_count++;
|
||||
if (first_failure == -1)
|
||||
first_failure = i;
|
||||
}
|
||||
}
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked SAD Test, C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure;
|
||||
}
|
||||
|
||||
#if CONFIG_VP9_HIGHBITDEPTH
|
||||
typedef unsigned int (*HighbdMaskedSADFunc)(const uint8_t *a, int a_stride,
|
||||
const uint8_t *b, int b_stride,
|
||||
const uint8_t *m, int m_stride);
|
||||
typedef std::tr1::tuple<HighbdMaskedSADFunc, HighbdMaskedSADFunc>
|
||||
HighbdMaskedSADParam;
|
||||
|
||||
class HighbdMaskedSADTest : public ::testing::
|
||||
TestWithParam<HighbdMaskedSADParam> {
|
||||
public:
|
||||
virtual ~HighbdMaskedSADTest() {}
|
||||
virtual void SetUp() {
|
||||
maskedSAD_op_ = GET_PARAM(0);
|
||||
ref_maskedSAD_op_ = GET_PARAM(1);
|
||||
}
|
||||
|
||||
virtual void TearDown() { libvpx_test::ClearSystemState(); }
|
||||
|
||||
protected:
|
||||
HighbdMaskedSADFunc maskedSAD_op_;
|
||||
HighbdMaskedSADFunc ref_maskedSAD_op_;
|
||||
};
|
||||
|
||||
TEST_P(HighbdMaskedSADTest, OperationCheck) {
|
||||
unsigned int ref_ret, ret;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint16_t, src_ptr[4096]);
|
||||
DECLARE_ALIGNED(16, uint16_t, ref_ptr[4096]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[4096]);
|
||||
uint8_t* src8_ptr = CONVERT_TO_BYTEPTR(src_ptr);
|
||||
uint8_t* ref8_ptr = CONVERT_TO_BYTEPTR(ref_ptr);
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = 64;
|
||||
int ref_stride = 64;
|
||||
int msk_stride = 64;
|
||||
for (int i = 0; i < number_of_iterations; ++i) {
|
||||
for (int j = 0; j < 4096; j++) {
|
||||
src_ptr[j] = rnd.Rand16()&0xfff;
|
||||
ref_ptr[j] = rnd.Rand16()&0xfff;
|
||||
msk_ptr[j] = ((rnd.Rand8()&0x7f) > 64) ? rnd.Rand8()&0x3f : 64;
|
||||
}
|
||||
|
||||
ref_ret = ref_maskedSAD_op_(src8_ptr, src_stride, ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride);
|
||||
ASM_REGISTER_STATE_CHECK(ret = maskedSAD_op_(src8_ptr, src_stride,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride));
|
||||
if (ret != ref_ret) {
|
||||
err_count++;
|
||||
if (first_failure == -1)
|
||||
first_failure = i;
|
||||
}
|
||||
}
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: High BD Masked SAD Test, C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure;
|
||||
}
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
|
||||
using std::tr1::make_tuple;
|
||||
|
||||
#if HAVE_SSSE3
|
||||
INSTANTIATE_TEST_CASE_P(
|
||||
SSSE3_C_COMPARE, MaskedSADTest,
|
||||
::testing::Values(
|
||||
make_tuple(&vpx_masked_sad64x64_ssse3,
|
||||
&vpx_masked_sad64x64_c),
|
||||
make_tuple(&vpx_masked_sad64x32_ssse3,
|
||||
&vpx_masked_sad64x32_c),
|
||||
make_tuple(&vpx_masked_sad32x64_ssse3,
|
||||
&vpx_masked_sad32x64_c),
|
||||
make_tuple(&vpx_masked_sad32x32_ssse3,
|
||||
&vpx_masked_sad32x32_c),
|
||||
make_tuple(&vpx_masked_sad32x16_ssse3,
|
||||
&vpx_masked_sad32x16_c),
|
||||
make_tuple(&vpx_masked_sad16x32_ssse3,
|
||||
&vpx_masked_sad16x32_c),
|
||||
make_tuple(&vpx_masked_sad16x16_ssse3,
|
||||
&vpx_masked_sad16x16_c),
|
||||
make_tuple(&vpx_masked_sad16x8_ssse3,
|
||||
&vpx_masked_sad16x8_c),
|
||||
make_tuple(&vpx_masked_sad8x16_ssse3,
|
||||
&vpx_masked_sad8x16_c),
|
||||
make_tuple(&vpx_masked_sad8x8_ssse3,
|
||||
&vpx_masked_sad8x8_c),
|
||||
make_tuple(&vpx_masked_sad8x4_ssse3,
|
||||
&vpx_masked_sad8x4_c),
|
||||
make_tuple(&vpx_masked_sad4x8_ssse3,
|
||||
&vpx_masked_sad4x8_c),
|
||||
make_tuple(&vpx_masked_sad4x4_ssse3,
|
||||
&vpx_masked_sad4x4_c)));
|
||||
#if CONFIG_VP9_HIGHBITDEPTH
|
||||
INSTANTIATE_TEST_CASE_P(
|
||||
SSSE3_C_COMPARE, HighbdMaskedSADTest,
|
||||
::testing::Values(
|
||||
make_tuple(&vp9_highbd_masked_sad64x64_ssse3,
|
||||
&vp9_highbd_masked_sad64x64_c),
|
||||
make_tuple(&vp9_highbd_masked_sad64x32_ssse3,
|
||||
&vp9_highbd_masked_sad64x32_c),
|
||||
make_tuple(&vp9_highbd_masked_sad32x64_ssse3,
|
||||
&vp9_highbd_masked_sad32x64_c),
|
||||
make_tuple(&vp9_highbd_masked_sad32x32_ssse3,
|
||||
&vp9_highbd_masked_sad32x32_c),
|
||||
make_tuple(&vp9_highbd_masked_sad32x16_ssse3,
|
||||
&vp9_highbd_masked_sad32x16_c),
|
||||
make_tuple(&vp9_highbd_masked_sad16x32_ssse3,
|
||||
&vp9_highbd_masked_sad16x32_c),
|
||||
make_tuple(&vp9_highbd_masked_sad16x16_ssse3,
|
||||
&vp9_highbd_masked_sad16x16_c),
|
||||
make_tuple(&vp9_highbd_masked_sad16x8_ssse3,
|
||||
&vp9_highbd_masked_sad16x8_c),
|
||||
make_tuple(&vp9_highbd_masked_sad8x16_ssse3,
|
||||
&vp9_highbd_masked_sad8x16_c),
|
||||
make_tuple(&vp9_highbd_masked_sad8x8_ssse3,
|
||||
&vp9_highbd_masked_sad8x8_c),
|
||||
make_tuple(&vp9_highbd_masked_sad8x4_ssse3,
|
||||
&vp9_highbd_masked_sad8x4_c),
|
||||
make_tuple(&vp9_highbd_masked_sad4x8_ssse3,
|
||||
&vp9_highbd_masked_sad4x8_c),
|
||||
make_tuple(&vp9_highbd_masked_sad4x4_ssse3,
|
||||
&vp9_highbd_masked_sad4x4_c)));
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
#endif // HAVE_SSSE3
|
||||
} // namespace
|
||||
752
test/masked_variance_test.cc
Normal file
752
test/masked_variance_test.cc
Normal file
@@ -0,0 +1,752 @@
|
||||
/*
|
||||
* Copyright (c) 2015 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 <math.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
|
||||
#include "third_party/googletest/src/include/gtest/gtest.h"
|
||||
#include "test/acm_random.h"
|
||||
#include "test/clear_system_state.h"
|
||||
#include "test/register_state_check.h"
|
||||
#include "test/util.h"
|
||||
|
||||
#include "./vpx_config.h"
|
||||
#include "./vpx_dsp_rtcd.h"
|
||||
#include "vpx/vpx_integer.h"
|
||||
#include "vpx_dsp/vpx_filter.h"
|
||||
|
||||
#define MAX_SIZE 64
|
||||
|
||||
using libvpx_test::ACMRandom;
|
||||
|
||||
namespace {
|
||||
const int number_of_iterations = 500;
|
||||
|
||||
typedef unsigned int (*MaskedVarianceFunc)(const uint8_t *a, int a_stride,
|
||||
const uint8_t *b, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
unsigned int *sse);
|
||||
|
||||
typedef std::tr1::tuple<MaskedVarianceFunc,
|
||||
MaskedVarianceFunc> MaskedVarianceParam;
|
||||
|
||||
class MaskedVarianceTest :
|
||||
public ::testing::TestWithParam<MaskedVarianceParam> {
|
||||
public:
|
||||
virtual ~MaskedVarianceTest() {}
|
||||
virtual void SetUp() {
|
||||
opt_func_ = GET_PARAM(0);
|
||||
ref_func_ = GET_PARAM(1);
|
||||
}
|
||||
|
||||
virtual void TearDown() { libvpx_test::ClearSystemState(); }
|
||||
|
||||
protected:
|
||||
MaskedVarianceFunc opt_func_;
|
||||
MaskedVarianceFunc ref_func_;
|
||||
};
|
||||
|
||||
TEST_P(MaskedVarianceTest, OperationCheck) {
|
||||
unsigned int ref_ret, opt_ret;
|
||||
unsigned int ref_sse, opt_sse;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint8_t, src_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
DECLARE_ALIGNED(16, uint8_t, ref_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = MAX_SIZE;
|
||||
int ref_stride = MAX_SIZE;
|
||||
int msk_stride = MAX_SIZE;
|
||||
|
||||
for (int i = 0; i < number_of_iterations; ++i) {
|
||||
for (int j = 0; j < MAX_SIZE*MAX_SIZE; j++) {
|
||||
src_ptr[j] = rnd.Rand8();
|
||||
ref_ptr[j] = rnd.Rand8();
|
||||
msk_ptr[j] = rnd(65);
|
||||
}
|
||||
|
||||
ref_ret = ref_func_(src_ptr, src_stride,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&ref_sse);
|
||||
ASM_REGISTER_STATE_CHECK(opt_ret = opt_func_(src_ptr, src_stride,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&opt_sse));
|
||||
|
||||
if (opt_ret != ref_ret || opt_sse != ref_sse) {
|
||||
err_count++;
|
||||
if (first_failure == -1)
|
||||
first_failure = i;
|
||||
}
|
||||
}
|
||||
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked Variance Test OperationCheck,"
|
||||
<< "C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure;
|
||||
}
|
||||
|
||||
TEST_P(MaskedVarianceTest, ExtremeValues) {
|
||||
unsigned int ref_ret, opt_ret;
|
||||
unsigned int ref_sse, opt_sse;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint8_t, src_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
DECLARE_ALIGNED(16, uint8_t, ref_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = MAX_SIZE;
|
||||
int ref_stride = MAX_SIZE;
|
||||
int msk_stride = MAX_SIZE;
|
||||
|
||||
for (int i = 0; i < 8; ++i) {
|
||||
memset(src_ptr, (i & 0x1) ? 255 : 0, MAX_SIZE*MAX_SIZE);
|
||||
memset(ref_ptr, (i & 0x2) ? 255 : 0, MAX_SIZE*MAX_SIZE);
|
||||
memset(msk_ptr, (i & 0x4) ? 64 : 0, MAX_SIZE*MAX_SIZE);
|
||||
|
||||
ref_ret = ref_func_(src_ptr, src_stride,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&ref_sse);
|
||||
ASM_REGISTER_STATE_CHECK(opt_ret = opt_func_(src_ptr, src_stride,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&opt_sse));
|
||||
|
||||
if (opt_ret != ref_ret || opt_sse != ref_sse) {
|
||||
err_count++;
|
||||
if (first_failure == -1)
|
||||
first_failure = i;
|
||||
}
|
||||
}
|
||||
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked Variance Test ExtremeValues,"
|
||||
<< "C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure;
|
||||
}
|
||||
|
||||
typedef unsigned int (*MaskedSubPixelVarianceFunc)(
|
||||
const uint8_t *a, int a_stride,
|
||||
int xoffset, int yoffset,
|
||||
const uint8_t *b, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
unsigned int *sse);
|
||||
|
||||
typedef std::tr1::tuple<MaskedSubPixelVarianceFunc,
|
||||
MaskedSubPixelVarianceFunc> MaskedSubPixelVarianceParam;
|
||||
|
||||
class MaskedSubPixelVarianceTest :
|
||||
public ::testing::TestWithParam<MaskedSubPixelVarianceParam> {
|
||||
public:
|
||||
virtual ~MaskedSubPixelVarianceTest() {}
|
||||
virtual void SetUp() {
|
||||
opt_func_ = GET_PARAM(0);
|
||||
ref_func_ = GET_PARAM(1);
|
||||
}
|
||||
|
||||
virtual void TearDown() { libvpx_test::ClearSystemState(); }
|
||||
|
||||
protected:
|
||||
MaskedSubPixelVarianceFunc opt_func_;
|
||||
MaskedSubPixelVarianceFunc ref_func_;
|
||||
};
|
||||
|
||||
TEST_P(MaskedSubPixelVarianceTest, OperationCheck) {
|
||||
unsigned int ref_ret, opt_ret;
|
||||
unsigned int ref_sse, opt_sse;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint8_t, src_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
DECLARE_ALIGNED(16, uint8_t, ref_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = (MAX_SIZE+1);
|
||||
int ref_stride = (MAX_SIZE+1);
|
||||
int msk_stride = (MAX_SIZE+1);
|
||||
int xoffset;
|
||||
int yoffset;
|
||||
|
||||
for (int i = 0; i < number_of_iterations; ++i) {
|
||||
int xoffsets[] = {0, 4, rnd(BIL_SUBPEL_SHIFTS)};
|
||||
int yoffsets[] = {0, 4, rnd(BIL_SUBPEL_SHIFTS)};
|
||||
for (int j = 0; j < (MAX_SIZE+1)*(MAX_SIZE+1); j++) {
|
||||
src_ptr[j] = rnd.Rand8();
|
||||
ref_ptr[j] = rnd.Rand8();
|
||||
msk_ptr[j] = rnd(65);
|
||||
}
|
||||
for (int k = 0; k < 3; k++) {
|
||||
xoffset = xoffsets[k];
|
||||
for (int l = 0; l < 3; l++) {
|
||||
xoffset = xoffsets[k];
|
||||
yoffset = yoffsets[l];
|
||||
|
||||
ref_ret = ref_func_(src_ptr, src_stride,
|
||||
xoffset, yoffset,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&ref_sse);
|
||||
ASM_REGISTER_STATE_CHECK(opt_ret = opt_func_(src_ptr, src_stride,
|
||||
xoffset, yoffset,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&opt_sse));
|
||||
|
||||
if (opt_ret != ref_ret || opt_sse != ref_sse) {
|
||||
err_count++;
|
||||
if (first_failure == -1)
|
||||
first_failure = i;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked Sub Pixel Variance Test OperationCheck,"
|
||||
<< "C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure;
|
||||
}
|
||||
|
||||
TEST_P(MaskedSubPixelVarianceTest, ExtremeValues) {
|
||||
unsigned int ref_ret, opt_ret;
|
||||
unsigned int ref_sse, opt_sse;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint8_t, src_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
DECLARE_ALIGNED(16, uint8_t, ref_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
int first_failure_x = -1;
|
||||
int first_failure_y = -1;
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = (MAX_SIZE+1);
|
||||
int ref_stride = (MAX_SIZE+1);
|
||||
int msk_stride = (MAX_SIZE+1);
|
||||
|
||||
for (int xoffset = 0 ; xoffset < BIL_SUBPEL_SHIFTS ; xoffset++) {
|
||||
for (int yoffset = 0 ; yoffset < BIL_SUBPEL_SHIFTS ; yoffset++) {
|
||||
for (int i = 0; i < 8; ++i) {
|
||||
memset(src_ptr, (i & 0x1) ? 255 : 0, (MAX_SIZE+1)*(MAX_SIZE+1));
|
||||
memset(ref_ptr, (i & 0x2) ? 255 : 0, (MAX_SIZE+1)*(MAX_SIZE+1));
|
||||
memset(msk_ptr, (i & 0x4) ? 64 : 0, (MAX_SIZE+1)*(MAX_SIZE+1));
|
||||
|
||||
ref_ret = ref_func_(src_ptr, src_stride,
|
||||
xoffset, yoffset,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&ref_sse);
|
||||
ASM_REGISTER_STATE_CHECK(opt_ret = opt_func_(src_ptr, src_stride,
|
||||
xoffset, yoffset,
|
||||
ref_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&opt_sse));
|
||||
|
||||
if (opt_ret != ref_ret || opt_sse != ref_sse) {
|
||||
err_count++;
|
||||
if (first_failure == -1) {
|
||||
first_failure = i;
|
||||
first_failure_x = xoffset;
|
||||
first_failure_y = yoffset;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked Variance Test ExtremeValues,"
|
||||
<< "C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure
|
||||
<< " x_offset = " << first_failure_x
|
||||
<< " y_offset = " << first_failure_y;
|
||||
}
|
||||
|
||||
#if CONFIG_VP9_HIGHBITDEPTH
|
||||
typedef std::tr1::tuple<MaskedVarianceFunc,
|
||||
MaskedVarianceFunc,
|
||||
vpx_bit_depth_t> HighbdMaskedVarianceParam;
|
||||
|
||||
class HighbdMaskedVarianceTest :
|
||||
public ::testing::TestWithParam<HighbdMaskedVarianceParam> {
|
||||
public:
|
||||
virtual ~HighbdMaskedVarianceTest() {}
|
||||
virtual void SetUp() {
|
||||
opt_func_ = GET_PARAM(0);
|
||||
ref_func_ = GET_PARAM(1);
|
||||
bit_depth_ = GET_PARAM(2);
|
||||
}
|
||||
|
||||
virtual void TearDown() { libvpx_test::ClearSystemState(); }
|
||||
|
||||
protected:
|
||||
MaskedVarianceFunc opt_func_;
|
||||
MaskedVarianceFunc ref_func_;
|
||||
vpx_bit_depth_t bit_depth_;
|
||||
};
|
||||
|
||||
TEST_P(HighbdMaskedVarianceTest, OperationCheck) {
|
||||
unsigned int ref_ret, opt_ret;
|
||||
unsigned int ref_sse, opt_sse;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint16_t, src_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
DECLARE_ALIGNED(16, uint16_t, ref_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
uint8_t* src8_ptr = CONVERT_TO_BYTEPTR(src_ptr);
|
||||
uint8_t* ref8_ptr = CONVERT_TO_BYTEPTR(ref_ptr);
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = MAX_SIZE;
|
||||
int ref_stride = MAX_SIZE;
|
||||
int msk_stride = MAX_SIZE;
|
||||
|
||||
for (int i = 0; i < number_of_iterations; ++i) {
|
||||
for (int j = 0; j < MAX_SIZE*MAX_SIZE; j++) {
|
||||
src_ptr[j] = rnd.Rand16() & ((1 << bit_depth_) - 1);
|
||||
ref_ptr[j] = rnd.Rand16() & ((1 << bit_depth_) - 1);
|
||||
msk_ptr[j] = rnd(65);
|
||||
}
|
||||
|
||||
ref_ret = ref_func_(src8_ptr, src_stride,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&ref_sse);
|
||||
ASM_REGISTER_STATE_CHECK(opt_ret = opt_func_(src8_ptr, src_stride,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&opt_sse));
|
||||
|
||||
if (opt_ret != ref_ret || opt_sse != ref_sse) {
|
||||
err_count++;
|
||||
if (first_failure == -1)
|
||||
first_failure = i;
|
||||
}
|
||||
}
|
||||
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked Variance Test OperationCheck,"
|
||||
<< "C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure;
|
||||
}
|
||||
|
||||
TEST_P(HighbdMaskedVarianceTest, ExtremeValues) {
|
||||
unsigned int ref_ret, opt_ret;
|
||||
unsigned int ref_sse, opt_sse;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint16_t, src_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
DECLARE_ALIGNED(16, uint16_t, ref_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[MAX_SIZE*MAX_SIZE]);
|
||||
uint8_t* src8_ptr = CONVERT_TO_BYTEPTR(src_ptr);
|
||||
uint8_t* ref8_ptr = CONVERT_TO_BYTEPTR(ref_ptr);
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = MAX_SIZE;
|
||||
int ref_stride = MAX_SIZE;
|
||||
int msk_stride = MAX_SIZE;
|
||||
|
||||
for (int i = 0; i < 8; ++i) {
|
||||
vpx_memset16(src_ptr, (i & 0x1) ? ((1 << bit_depth_) - 1) : 0,
|
||||
MAX_SIZE*MAX_SIZE);
|
||||
vpx_memset16(ref_ptr, (i & 0x2) ? ((1 << bit_depth_) - 1) : 0,
|
||||
MAX_SIZE*MAX_SIZE);
|
||||
memset(msk_ptr, (i & 0x4) ? 64 : 0, MAX_SIZE*MAX_SIZE);
|
||||
|
||||
ref_ret = ref_func_(src8_ptr, src_stride,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&ref_sse);
|
||||
ASM_REGISTER_STATE_CHECK(opt_ret = opt_func_(src8_ptr, src_stride,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&opt_sse));
|
||||
|
||||
if (opt_ret != ref_ret || opt_sse != ref_sse) {
|
||||
err_count++;
|
||||
if (first_failure == -1)
|
||||
first_failure = i;
|
||||
}
|
||||
}
|
||||
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked Variance Test ExtremeValues,"
|
||||
<< "C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure;
|
||||
}
|
||||
|
||||
typedef std::tr1::tuple<MaskedSubPixelVarianceFunc,
|
||||
MaskedSubPixelVarianceFunc,
|
||||
vpx_bit_depth_t> HighbdMaskedSubPixelVarianceParam;
|
||||
|
||||
class HighbdMaskedSubPixelVarianceTest :
|
||||
public ::testing::TestWithParam<HighbdMaskedSubPixelVarianceParam> {
|
||||
public:
|
||||
virtual ~HighbdMaskedSubPixelVarianceTest() {}
|
||||
virtual void SetUp() {
|
||||
opt_func_ = GET_PARAM(0);
|
||||
ref_func_ = GET_PARAM(1);
|
||||
bit_depth_ = GET_PARAM(2);
|
||||
}
|
||||
|
||||
virtual void TearDown() { libvpx_test::ClearSystemState(); }
|
||||
|
||||
protected:
|
||||
MaskedSubPixelVarianceFunc opt_func_;
|
||||
MaskedSubPixelVarianceFunc ref_func_;
|
||||
vpx_bit_depth_t bit_depth_;
|
||||
};
|
||||
|
||||
TEST_P(HighbdMaskedSubPixelVarianceTest, OperationCheck) {
|
||||
unsigned int ref_ret, opt_ret;
|
||||
unsigned int ref_sse, opt_sse;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint16_t, src_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
DECLARE_ALIGNED(16, uint16_t, ref_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
uint8_t* src8_ptr = CONVERT_TO_BYTEPTR(src_ptr);
|
||||
uint8_t* ref8_ptr = CONVERT_TO_BYTEPTR(ref_ptr);
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int first_failure_x = -1;
|
||||
int first_failure_y = -1;
|
||||
int src_stride = (MAX_SIZE+1);
|
||||
int ref_stride = (MAX_SIZE+1);
|
||||
int msk_stride = (MAX_SIZE+1);
|
||||
int xoffset, yoffset;
|
||||
|
||||
for (int i = 0; i < number_of_iterations; ++i) {
|
||||
for (xoffset = 0; xoffset < BIL_SUBPEL_SHIFTS; xoffset++) {
|
||||
for (yoffset = 0; yoffset < BIL_SUBPEL_SHIFTS; yoffset++) {
|
||||
for (int j = 0; j < (MAX_SIZE+1)*(MAX_SIZE+1); j++) {
|
||||
src_ptr[j] = rnd.Rand16() & ((1 << bit_depth_) - 1);
|
||||
ref_ptr[j] = rnd.Rand16() & ((1 << bit_depth_) - 1);
|
||||
msk_ptr[j] = rnd(65);
|
||||
}
|
||||
|
||||
ref_ret = ref_func_(src8_ptr, src_stride,
|
||||
xoffset, yoffset,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&ref_sse);
|
||||
ASM_REGISTER_STATE_CHECK(opt_ret = opt_func_(src8_ptr, src_stride,
|
||||
xoffset, yoffset,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&opt_sse));
|
||||
|
||||
if (opt_ret != ref_ret || opt_sse != ref_sse) {
|
||||
err_count++;
|
||||
if (first_failure == -1) {
|
||||
first_failure = i;
|
||||
first_failure_x = xoffset;
|
||||
first_failure_y = yoffset;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked Sub Pixel Variance Test OperationCheck,"
|
||||
<< "C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure
|
||||
<< " x_offset = " << first_failure_x
|
||||
<< " y_offset = " << first_failure_y;
|
||||
}
|
||||
|
||||
TEST_P(HighbdMaskedSubPixelVarianceTest, ExtremeValues) {
|
||||
unsigned int ref_ret, opt_ret;
|
||||
unsigned int ref_sse, opt_sse;
|
||||
ACMRandom rnd(ACMRandom::DeterministicSeed());
|
||||
DECLARE_ALIGNED(16, uint16_t, src_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
DECLARE_ALIGNED(16, uint16_t, ref_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
DECLARE_ALIGNED(16, uint8_t, msk_ptr[(MAX_SIZE+1)*(MAX_SIZE+1)]);
|
||||
uint8_t* src8_ptr = CONVERT_TO_BYTEPTR(src_ptr);
|
||||
uint8_t* ref8_ptr = CONVERT_TO_BYTEPTR(ref_ptr);
|
||||
int first_failure_x = -1;
|
||||
int first_failure_y = -1;
|
||||
int err_count = 0;
|
||||
int first_failure = -1;
|
||||
int src_stride = (MAX_SIZE+1);
|
||||
int ref_stride = (MAX_SIZE+1);
|
||||
int msk_stride = (MAX_SIZE+1);
|
||||
|
||||
for (int xoffset = 0 ; xoffset < BIL_SUBPEL_SHIFTS ; xoffset++) {
|
||||
for (int yoffset = 0 ; yoffset < BIL_SUBPEL_SHIFTS ; yoffset++) {
|
||||
for (int i = 0; i < 8; ++i) {
|
||||
vpx_memset16(src_ptr, (i & 0x1) ? ((1 << bit_depth_) - 1) : 0,
|
||||
(MAX_SIZE+1)*(MAX_SIZE+1));
|
||||
vpx_memset16(ref_ptr, (i & 0x2) ? ((1 << bit_depth_) - 1) : 0,
|
||||
(MAX_SIZE+1)*(MAX_SIZE+1));
|
||||
memset(msk_ptr, (i & 0x4) ? 64 : 0, (MAX_SIZE+1)*(MAX_SIZE+1));
|
||||
|
||||
ref_ret = ref_func_(src8_ptr, src_stride,
|
||||
xoffset, yoffset,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&ref_sse);
|
||||
ASM_REGISTER_STATE_CHECK(opt_ret = opt_func_(src8_ptr, src_stride,
|
||||
xoffset, yoffset,
|
||||
ref8_ptr, ref_stride,
|
||||
msk_ptr, msk_stride,
|
||||
&opt_sse));
|
||||
|
||||
if (opt_ret != ref_ret || opt_sse != ref_sse) {
|
||||
err_count++;
|
||||
if (first_failure == -1) {
|
||||
first_failure = i;
|
||||
first_failure_x = xoffset;
|
||||
first_failure_y = yoffset;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
EXPECT_EQ(0, err_count)
|
||||
<< "Error: Masked Variance Test ExtremeValues,"
|
||||
<< "C output doesn't match SSSE3 output. "
|
||||
<< "First failed at test case " << first_failure
|
||||
<< " x_offset = " << first_failure_x
|
||||
<< " y_offset = " << first_failure_y;
|
||||
}
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
|
||||
using std::tr1::make_tuple;
|
||||
|
||||
#if HAVE_SSSE3
|
||||
INSTANTIATE_TEST_CASE_P(
|
||||
SSSE3_C_COMPARE, MaskedVarianceTest,
|
||||
::testing::Values(
|
||||
make_tuple(&vpx_masked_variance64x64_ssse3,
|
||||
&vpx_masked_variance64x64_c),
|
||||
make_tuple(&vpx_masked_variance64x32_ssse3,
|
||||
&vpx_masked_variance64x32_c),
|
||||
make_tuple(&vpx_masked_variance32x64_ssse3,
|
||||
&vpx_masked_variance32x64_c),
|
||||
make_tuple(&vpx_masked_variance32x32_ssse3,
|
||||
&vpx_masked_variance32x32_c),
|
||||
make_tuple(&vpx_masked_variance32x16_ssse3,
|
||||
&vpx_masked_variance32x16_c),
|
||||
make_tuple(&vpx_masked_variance16x32_ssse3,
|
||||
&vpx_masked_variance16x32_c),
|
||||
make_tuple(&vpx_masked_variance16x16_ssse3,
|
||||
&vpx_masked_variance16x16_c),
|
||||
make_tuple(&vpx_masked_variance16x8_ssse3,
|
||||
&vpx_masked_variance16x8_c),
|
||||
make_tuple(&vpx_masked_variance8x16_ssse3,
|
||||
&vpx_masked_variance8x16_c),
|
||||
make_tuple(&vpx_masked_variance8x8_ssse3,
|
||||
&vpx_masked_variance8x8_c),
|
||||
make_tuple(&vpx_masked_variance8x4_ssse3,
|
||||
&vpx_masked_variance8x4_c),
|
||||
make_tuple(&vpx_masked_variance4x8_ssse3,
|
||||
&vpx_masked_variance4x8_c),
|
||||
make_tuple(&vpx_masked_variance4x4_ssse3,
|
||||
&vpx_masked_variance4x4_c)));
|
||||
|
||||
INSTANTIATE_TEST_CASE_P(
|
||||
SSSE3_C_COMPARE, MaskedSubPixelVarianceTest,
|
||||
::testing::Values(
|
||||
make_tuple(&vpx_masked_sub_pixel_variance64x64_ssse3,
|
||||
&vpx_masked_sub_pixel_variance64x64_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance64x32_ssse3,
|
||||
&vpx_masked_sub_pixel_variance64x32_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance32x64_ssse3,
|
||||
&vpx_masked_sub_pixel_variance32x64_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance32x32_ssse3,
|
||||
&vpx_masked_sub_pixel_variance32x32_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance32x16_ssse3,
|
||||
&vpx_masked_sub_pixel_variance32x16_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance16x32_ssse3,
|
||||
&vpx_masked_sub_pixel_variance16x32_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance16x16_ssse3,
|
||||
&vpx_masked_sub_pixel_variance16x16_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance16x8_ssse3,
|
||||
&vpx_masked_sub_pixel_variance16x8_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance8x16_ssse3,
|
||||
&vpx_masked_sub_pixel_variance8x16_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance8x8_ssse3,
|
||||
&vpx_masked_sub_pixel_variance8x8_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance8x4_ssse3,
|
||||
&vpx_masked_sub_pixel_variance8x4_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance4x8_ssse3,
|
||||
&vpx_masked_sub_pixel_variance4x8_c),
|
||||
make_tuple(&vpx_masked_sub_pixel_variance4x4_ssse3,
|
||||
&vpx_masked_sub_pixel_variance4x4_c)));
|
||||
|
||||
#if CONFIG_VP9_HIGHBITDEPTH
|
||||
INSTANTIATE_TEST_CASE_P(
|
||||
SSSE3_C_COMPARE, HighbdMaskedVarianceTest,
|
||||
::testing::Values(
|
||||
make_tuple(&vp9_highbd_masked_variance64x64_ssse3,
|
||||
&vp9_highbd_masked_variance64x64_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance64x32_ssse3,
|
||||
&vp9_highbd_masked_variance64x32_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance32x64_ssse3,
|
||||
&vp9_highbd_masked_variance32x64_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance32x32_ssse3,
|
||||
&vp9_highbd_masked_variance32x32_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance32x16_ssse3,
|
||||
&vp9_highbd_masked_variance32x16_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance16x32_ssse3,
|
||||
&vp9_highbd_masked_variance16x32_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance16x16_ssse3,
|
||||
&vp9_highbd_masked_variance16x16_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance16x8_ssse3,
|
||||
&vp9_highbd_masked_variance16x8_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance8x16_ssse3,
|
||||
&vp9_highbd_masked_variance8x16_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance8x8_ssse3,
|
||||
&vp9_highbd_masked_variance8x8_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance8x4_ssse3,
|
||||
&vp9_highbd_masked_variance8x4_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance4x8_ssse3,
|
||||
&vp9_highbd_masked_variance4x8_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_variance4x4_ssse3,
|
||||
&vp9_highbd_masked_variance4x4_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_10_masked_variance64x64_ssse3,
|
||||
&vp9_highbd_10_masked_variance64x64_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance64x32_ssse3,
|
||||
&vp9_highbd_10_masked_variance64x32_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance32x64_ssse3,
|
||||
&vp9_highbd_10_masked_variance32x64_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance32x32_ssse3,
|
||||
&vp9_highbd_10_masked_variance32x32_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance32x16_ssse3,
|
||||
&vp9_highbd_10_masked_variance32x16_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance16x32_ssse3,
|
||||
&vp9_highbd_10_masked_variance16x32_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance16x16_ssse3,
|
||||
&vp9_highbd_10_masked_variance16x16_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance16x8_ssse3,
|
||||
&vp9_highbd_10_masked_variance16x8_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance8x16_ssse3,
|
||||
&vp9_highbd_10_masked_variance8x16_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance8x8_ssse3,
|
||||
&vp9_highbd_10_masked_variance8x8_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance8x4_ssse3,
|
||||
&vp9_highbd_10_masked_variance8x4_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance4x8_ssse3,
|
||||
&vp9_highbd_10_masked_variance4x8_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_variance4x4_ssse3,
|
||||
&vp9_highbd_10_masked_variance4x4_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_12_masked_variance64x64_ssse3,
|
||||
&vp9_highbd_12_masked_variance64x64_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance64x32_ssse3,
|
||||
&vp9_highbd_12_masked_variance64x32_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance32x64_ssse3,
|
||||
&vp9_highbd_12_masked_variance32x64_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance32x32_ssse3,
|
||||
&vp9_highbd_12_masked_variance32x32_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance32x16_ssse3,
|
||||
&vp9_highbd_12_masked_variance32x16_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance16x32_ssse3,
|
||||
&vp9_highbd_12_masked_variance16x32_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance16x16_ssse3,
|
||||
&vp9_highbd_12_masked_variance16x16_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance16x8_ssse3,
|
||||
&vp9_highbd_12_masked_variance16x8_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance8x16_ssse3,
|
||||
&vp9_highbd_12_masked_variance8x16_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance8x8_ssse3,
|
||||
&vp9_highbd_12_masked_variance8x8_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance8x4_ssse3,
|
||||
&vp9_highbd_12_masked_variance8x4_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance4x8_ssse3,
|
||||
&vp9_highbd_12_masked_variance4x8_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_variance4x4_ssse3,
|
||||
&vp9_highbd_12_masked_variance4x4_c, VPX_BITS_12)));
|
||||
|
||||
INSTANTIATE_TEST_CASE_P(
|
||||
SSSE3_C_COMPARE, HighbdMaskedSubPixelVarianceTest,
|
||||
::testing::Values(
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance64x64_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance64x64_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance64x32_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance64x32_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance32x64_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance32x64_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance32x32_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance32x32_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance32x16_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance32x16_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance16x32_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance16x32_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance16x16_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance16x16_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance16x8_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance16x8_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance8x16_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance8x16_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance8x8_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance8x8_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance8x4_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance8x4_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance4x8_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance4x8_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_masked_sub_pixel_variance4x4_ssse3,
|
||||
&vp9_highbd_masked_sub_pixel_variance4x4_c, VPX_BITS_8),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance64x64_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance64x64_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance64x32_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance64x32_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance32x64_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance32x64_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance32x32_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance32x32_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance32x16_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance32x16_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance16x32_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance16x32_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance16x16_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance16x16_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance16x8_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance16x8_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance8x16_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance8x16_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance8x8_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance8x8_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance8x4_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance8x4_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance4x8_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance4x8_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_10_masked_sub_pixel_variance4x4_ssse3,
|
||||
&vp9_highbd_10_masked_sub_pixel_variance4x4_c, VPX_BITS_10),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance64x64_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance64x64_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance64x32_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance64x32_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance32x64_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance32x64_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance32x32_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance32x32_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance32x16_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance32x16_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance16x32_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance16x32_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance16x16_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance16x16_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance16x8_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance16x8_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance8x16_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance8x16_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance8x8_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance8x8_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance8x4_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance8x4_c, VPX_BITS_12) ,
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance4x8_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance4x8_c, VPX_BITS_12),
|
||||
make_tuple(&vp9_highbd_12_masked_sub_pixel_variance4x4_ssse3,
|
||||
&vp9_highbd_12_masked_sub_pixel_variance4x4_c, VPX_BITS_12)));
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
|
||||
#endif // HAVE_SSSE3
|
||||
} // namespace
|
||||
@@ -168,6 +168,11 @@ LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += vp10_dct_test.cc
|
||||
LIBVPX_TEST_SRCS-$(CONFIG_ANS) += vp10_ans_test.cc
|
||||
|
||||
LIBVPX_TEST_SRCS-$(CONFIG_VP10_ENCODER) += sum_squares_test.cc
|
||||
|
||||
ifeq ($(CONFIG_EXT_INTER),yes)
|
||||
LIBVPX_TEST_SRCS-$(HAVE_SSSE3) += masked_variance_test.cc
|
||||
LIBVPX_TEST_SRCS-$(HAVE_SSSE3) += masked_sad_test.cc
|
||||
endif
|
||||
endif # VP10
|
||||
|
||||
## Multi-codec / unconditional whitebox tests.
|
||||
|
||||
102
vpx_dsp/sad.c
102
vpx_dsp/sad.c
@@ -316,3 +316,105 @@ highbd_sadMxNxK(4, 4, 8)
|
||||
highbd_sadMxNx4D(4, 4)
|
||||
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
|
||||
#if CONFIG_VP10 && CONFIG_EXT_INTER
|
||||
static INLINE unsigned int masked_sad(const uint8_t *a, int a_stride,
|
||||
const uint8_t *b, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
int width, int height) {
|
||||
int y, x;
|
||||
unsigned int sad = 0;
|
||||
|
||||
for (y = 0; y < height; y++) {
|
||||
for (x = 0; x < width; x++)
|
||||
sad += m[x] * abs(a[x] - b[x]);
|
||||
|
||||
a += a_stride;
|
||||
b += b_stride;
|
||||
m += m_stride;
|
||||
}
|
||||
sad = (sad + 31) >> 6;
|
||||
|
||||
return sad;
|
||||
}
|
||||
|
||||
#define MASKSADMxN(m, n) \
|
||||
unsigned int vpx_masked_sad##m##x##n##_c(const uint8_t *src, int src_stride, \
|
||||
const uint8_t *ref, int ref_stride, \
|
||||
const uint8_t *msk, int msk_stride) { \
|
||||
return masked_sad(src, src_stride, ref, ref_stride, msk, msk_stride, m, n); \
|
||||
}
|
||||
|
||||
#if CONFIG_EXT_PARTITION
|
||||
MASKSADMxN(128, 128)
|
||||
MASKSADMxN(128, 64)
|
||||
MASKSADMxN(64, 128)
|
||||
#endif // CONFIG_EXT_PARTITION
|
||||
MASKSADMxN(64, 64)
|
||||
MASKSADMxN(64, 32)
|
||||
MASKSADMxN(32, 64)
|
||||
MASKSADMxN(32, 32)
|
||||
MASKSADMxN(32, 16)
|
||||
MASKSADMxN(16, 32)
|
||||
MASKSADMxN(16, 16)
|
||||
MASKSADMxN(16, 8)
|
||||
MASKSADMxN(8, 16)
|
||||
MASKSADMxN(8, 8)
|
||||
MASKSADMxN(8, 4)
|
||||
MASKSADMxN(4, 8)
|
||||
MASKSADMxN(4, 4)
|
||||
|
||||
#if CONFIG_VP9_HIGHBITDEPTH
|
||||
static INLINE unsigned int highbd_masked_sad(const uint8_t *a8, int a_stride,
|
||||
const uint8_t *b8, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
int width, int height) {
|
||||
int y, x;
|
||||
unsigned int sad = 0;
|
||||
const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
|
||||
const uint16_t *b = CONVERT_TO_SHORTPTR(b8);
|
||||
|
||||
for (y = 0; y < height; y++) {
|
||||
for (x = 0; x < width; x++)
|
||||
sad += m[x] * abs(a[x] - b[x]);
|
||||
|
||||
a += a_stride;
|
||||
b += b_stride;
|
||||
m += m_stride;
|
||||
}
|
||||
sad = (sad + 31) >> 6;
|
||||
|
||||
return sad;
|
||||
}
|
||||
|
||||
#define HIGHBD_MASKSADMXN(m, n) \
|
||||
unsigned int vpx_highbd_masked_sad##m##x##n##_c(const uint8_t *src, \
|
||||
int src_stride, \
|
||||
const uint8_t *ref, \
|
||||
int ref_stride, \
|
||||
const uint8_t *msk, \
|
||||
int msk_stride) { \
|
||||
return highbd_masked_sad(src, src_stride, ref, ref_stride, \
|
||||
msk, msk_stride, m, n); \
|
||||
}
|
||||
|
||||
#if CONFIG_EXT_PARTITION
|
||||
HIGHBD_MASKSADMXN(128, 128)
|
||||
HIGHBD_MASKSADMXN(128, 64)
|
||||
HIGHBD_MASKSADMXN(64, 128)
|
||||
#endif // CONFIG_EXT_PARTITION
|
||||
HIGHBD_MASKSADMXN(64, 64)
|
||||
HIGHBD_MASKSADMXN(64, 32)
|
||||
HIGHBD_MASKSADMXN(32, 64)
|
||||
HIGHBD_MASKSADMXN(32, 32)
|
||||
HIGHBD_MASKSADMXN(32, 16)
|
||||
HIGHBD_MASKSADMXN(16, 32)
|
||||
HIGHBD_MASKSADMXN(16, 16)
|
||||
HIGHBD_MASKSADMXN(16, 8)
|
||||
HIGHBD_MASKSADMXN(8, 16)
|
||||
HIGHBD_MASKSADMXN(8, 8)
|
||||
HIGHBD_MASKSADMXN(8, 4)
|
||||
HIGHBD_MASKSADMXN(4, 8)
|
||||
HIGHBD_MASKSADMXN(4, 4)
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
#endif // CONFIG_VP10 && CONFIG_EXT_INTER
|
||||
|
||||
@@ -15,8 +15,9 @@
|
||||
#include "vpx/vpx_integer.h"
|
||||
|
||||
#include "vpx_dsp/variance.h"
|
||||
#include "vpx_dsp/vpx_filter.h"
|
||||
|
||||
static const uint8_t bilinear_filters[8][2] = {
|
||||
const uint8_t vpx_bilinear_filters[BIL_SUBPEL_SHIFTS][2] = {
|
||||
{ 128, 0 },
|
||||
{ 112, 16 },
|
||||
{ 96, 32 },
|
||||
@@ -175,9 +176,9 @@ uint32_t vpx_sub_pixel_variance##W##x##H##_c(const uint8_t *a, int a_stride, \
|
||||
uint8_t temp2[H * W]; \
|
||||
\
|
||||
var_filter_block2d_bil_first_pass(a, fdata3, a_stride, 1, H + 1, W, \
|
||||
bilinear_filters[xoffset]); \
|
||||
vpx_bilinear_filters[xoffset]); \
|
||||
var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
bilinear_filters[yoffset]); \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
return vpx_variance##W##x##H##_c(temp2, W, b, b_stride, sse); \
|
||||
}
|
||||
@@ -195,9 +196,9 @@ uint32_t vpx_sub_pixel_avg_variance##W##x##H##_c(const uint8_t *a, \
|
||||
DECLARE_ALIGNED(16, uint8_t, temp3[H * W]); \
|
||||
\
|
||||
var_filter_block2d_bil_first_pass(a, fdata3, a_stride, 1, H + 1, W, \
|
||||
bilinear_filters[xoffset]); \
|
||||
vpx_bilinear_filters[xoffset]); \
|
||||
var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
bilinear_filters[yoffset]); \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
vpx_comp_avg_pred(temp3, second_pred, W, H, temp2, W); \
|
||||
\
|
||||
@@ -500,9 +501,9 @@ uint32_t vpx_highbd_8_sub_pixel_variance##W##x##H##_c( \
|
||||
uint16_t temp2[H * W]; \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
|
||||
W, bilinear_filters[xoffset]); \
|
||||
W, vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
bilinear_filters[yoffset]); \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
return vpx_highbd_8_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), W, dst, \
|
||||
dst_stride, sse); \
|
||||
@@ -517,9 +518,9 @@ uint32_t vpx_highbd_10_sub_pixel_variance##W##x##H##_c( \
|
||||
uint16_t temp2[H * W]; \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
|
||||
W, bilinear_filters[xoffset]); \
|
||||
W, vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
bilinear_filters[yoffset]); \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
return vpx_highbd_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), \
|
||||
W, dst, dst_stride, sse); \
|
||||
@@ -534,9 +535,9 @@ uint32_t vpx_highbd_12_sub_pixel_variance##W##x##H##_c( \
|
||||
uint16_t temp2[H * W]; \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
|
||||
W, bilinear_filters[xoffset]); \
|
||||
W, vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
bilinear_filters[yoffset]); \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
return vpx_highbd_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), \
|
||||
W, dst, dst_stride, sse); \
|
||||
@@ -554,9 +555,9 @@ uint32_t vpx_highbd_8_sub_pixel_avg_variance##W##x##H##_c( \
|
||||
DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
|
||||
W, bilinear_filters[xoffset]); \
|
||||
W, vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
bilinear_filters[yoffset]); \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
vpx_highbd_comp_avg_pred(temp3, second_pred, W, H, \
|
||||
CONVERT_TO_BYTEPTR(temp2), W); \
|
||||
@@ -576,9 +577,9 @@ uint32_t vpx_highbd_10_sub_pixel_avg_variance##W##x##H##_c( \
|
||||
DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
|
||||
W, bilinear_filters[xoffset]); \
|
||||
W, vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
bilinear_filters[yoffset]); \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
vpx_highbd_comp_avg_pred(temp3, second_pred, W, H, \
|
||||
CONVERT_TO_BYTEPTR(temp2), W); \
|
||||
@@ -598,9 +599,9 @@ uint32_t vpx_highbd_12_sub_pixel_avg_variance##W##x##H##_c( \
|
||||
DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \
|
||||
W, bilinear_filters[xoffset]); \
|
||||
W, vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
bilinear_filters[yoffset]); \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
vpx_highbd_comp_avg_pred(temp3, second_pred, W, H, \
|
||||
CONVERT_TO_BYTEPTR(temp2), W); \
|
||||
@@ -654,3 +655,323 @@ void vpx_highbd_comp_avg_pred(uint16_t *comp_pred, const uint8_t *pred8,
|
||||
}
|
||||
}
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
|
||||
#if CONFIG_VP10 && CONFIG_EXT_INTER
|
||||
void masked_variance(const uint8_t *a, int a_stride,
|
||||
const uint8_t *b, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
int w, int h, unsigned int *sse, int *sum) {
|
||||
int i, j;
|
||||
|
||||
int64_t sum64 = 0;
|
||||
uint64_t sse64 = 0;
|
||||
|
||||
for (i = 0; i < h; i++) {
|
||||
for (j = 0; j < w; j++) {
|
||||
const int diff = (a[j] - b[j]) * (m[j]);
|
||||
sum64 += diff;
|
||||
sse64 += diff * diff;
|
||||
}
|
||||
|
||||
a += a_stride;
|
||||
b += b_stride;
|
||||
m += m_stride;
|
||||
}
|
||||
*sum = (sum64 >= 0) ? ((sum64 + 31) >> 6) : -((-sum64 + 31) >> 6);
|
||||
*sse = (sse64 + 2047) >> 12;
|
||||
}
|
||||
|
||||
#define MASK_VAR(W, H) \
|
||||
unsigned int vpx_masked_variance##W##x##H##_c(const uint8_t *a, int a_stride, \
|
||||
const uint8_t *b, int b_stride, \
|
||||
const uint8_t *m, int m_stride, \
|
||||
unsigned int *sse) { \
|
||||
int sum; \
|
||||
masked_variance(a, a_stride, b, b_stride, m, m_stride, W, H, sse, &sum); \
|
||||
return *sse - (((int64_t)sum * sum) / (W * H)); \
|
||||
}
|
||||
|
||||
#define MASK_SUBPIX_VAR(W, H) \
|
||||
unsigned int vpx_masked_sub_pixel_variance##W##x##H##_c( \
|
||||
const uint8_t *src, int src_stride, \
|
||||
int xoffset, int yoffset, \
|
||||
const uint8_t *dst, int dst_stride, \
|
||||
const uint8_t *msk, int msk_stride, \
|
||||
unsigned int *sse) { \
|
||||
uint16_t fdata3[(H + 1) * W]; \
|
||||
uint8_t temp2[H * W]; \
|
||||
\
|
||||
var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, W, \
|
||||
vpx_bilinear_filters[xoffset]); \
|
||||
var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
return vpx_masked_variance##W##x##H##_c(temp2, W, dst, dst_stride, \
|
||||
msk, msk_stride, sse); \
|
||||
}
|
||||
|
||||
MASK_VAR(4, 4)
|
||||
MASK_SUBPIX_VAR(4, 4)
|
||||
|
||||
MASK_VAR(4, 8)
|
||||
MASK_SUBPIX_VAR(4, 8)
|
||||
|
||||
MASK_VAR(8, 4)
|
||||
MASK_SUBPIX_VAR(8, 4)
|
||||
|
||||
MASK_VAR(8, 8)
|
||||
MASK_SUBPIX_VAR(8, 8)
|
||||
|
||||
MASK_VAR(8, 16)
|
||||
MASK_SUBPIX_VAR(8, 16)
|
||||
|
||||
MASK_VAR(16, 8)
|
||||
MASK_SUBPIX_VAR(16, 8)
|
||||
|
||||
MASK_VAR(16, 16)
|
||||
MASK_SUBPIX_VAR(16, 16)
|
||||
|
||||
MASK_VAR(16, 32)
|
||||
MASK_SUBPIX_VAR(16, 32)
|
||||
|
||||
MASK_VAR(32, 16)
|
||||
MASK_SUBPIX_VAR(32, 16)
|
||||
|
||||
MASK_VAR(32, 32)
|
||||
MASK_SUBPIX_VAR(32, 32)
|
||||
|
||||
MASK_VAR(32, 64)
|
||||
MASK_SUBPIX_VAR(32, 64)
|
||||
|
||||
MASK_VAR(64, 32)
|
||||
MASK_SUBPIX_VAR(64, 32)
|
||||
|
||||
MASK_VAR(64, 64)
|
||||
MASK_SUBPIX_VAR(64, 64)
|
||||
|
||||
#if CONFIG_EXT_PARTITION
|
||||
MASK_VAR(64, 128)
|
||||
MASK_SUBPIX_VAR(64, 128)
|
||||
|
||||
MASK_VAR(128, 64)
|
||||
MASK_SUBPIX_VAR(128, 64)
|
||||
|
||||
MASK_VAR(128, 128)
|
||||
MASK_SUBPIX_VAR(128, 128)
|
||||
#endif // CONFIG_EXT_PARTITION
|
||||
|
||||
#if CONFIG_VP9_HIGHBITDEPTH
|
||||
void highbd_masked_variance64(const uint8_t *a8, int a_stride,
|
||||
const uint8_t *b8, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
int w, int h,
|
||||
uint64_t *sse64, int *sum) {
|
||||
int i, j;
|
||||
uint16_t *a = CONVERT_TO_SHORTPTR(a8);
|
||||
uint16_t *b = CONVERT_TO_SHORTPTR(b8);
|
||||
|
||||
int64_t sum64 = 0;
|
||||
*sse64 = 0;
|
||||
|
||||
for (i = 0; i < h; i++) {
|
||||
for (j = 0; j < w; j++) {
|
||||
const int diff = (a[j] - b[j]) * (m[j]);
|
||||
sum64 += diff;
|
||||
*sse64 += (int64_t)diff * diff;
|
||||
}
|
||||
|
||||
a += a_stride;
|
||||
b += b_stride;
|
||||
m += m_stride;
|
||||
}
|
||||
*sum = (sum64 >= 0) ? ((sum64 + 31) >> 6) : -((-sum64 + 31) >> 6);
|
||||
*sse64 = (*sse64 + 2047) >> 12;
|
||||
}
|
||||
|
||||
void highbd_masked_variance(const uint8_t *a8, int a_stride,
|
||||
const uint8_t *b8, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
int w, int h,
|
||||
unsigned int *sse, int *sum) {
|
||||
uint64_t sse64;
|
||||
highbd_masked_variance64(a8, a_stride, b8, b_stride, m, m_stride,
|
||||
w, h, &sse64, sum);
|
||||
*sse = (unsigned int)sse64;
|
||||
}
|
||||
|
||||
void highbd_10_masked_variance(const uint8_t *a8, int a_stride,
|
||||
const uint8_t *b8, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
int w, int h,
|
||||
unsigned int *sse, int *sum) {
|
||||
uint64_t sse64;
|
||||
highbd_masked_variance64(a8, a_stride, b8, b_stride, m, m_stride,
|
||||
w, h, &sse64, sum);
|
||||
*sum = ROUND_POWER_OF_TWO(*sum, 2);
|
||||
*sse = (unsigned int)ROUND_POWER_OF_TWO(sse64, 4);
|
||||
}
|
||||
|
||||
void highbd_12_masked_variance(const uint8_t *a8, int a_stride,
|
||||
const uint8_t *b8, int b_stride,
|
||||
const uint8_t *m, int m_stride,
|
||||
int w, int h,
|
||||
unsigned int *sse, int *sum) {
|
||||
uint64_t sse64;
|
||||
highbd_masked_variance64(a8, a_stride, b8, b_stride, m, m_stride,
|
||||
w, h, &sse64, sum);
|
||||
*sum = ROUND_POWER_OF_TWO(*sum, 4);
|
||||
*sse = (unsigned int)ROUND_POWER_OF_TWO(sse64, 8);
|
||||
}
|
||||
|
||||
#define HIGHBD_MASK_VAR(W, H) \
|
||||
unsigned int vpx_highbd_masked_variance##W##x##H##_c(const uint8_t *a, \
|
||||
int a_stride, \
|
||||
const uint8_t *b, \
|
||||
int b_stride, \
|
||||
const uint8_t *m, \
|
||||
int m_stride, \
|
||||
unsigned int *sse) { \
|
||||
int sum; \
|
||||
highbd_masked_variance(a, a_stride, b, b_stride, m, m_stride, \
|
||||
W, H, sse, &sum); \
|
||||
return *sse - (((int64_t)sum * sum) / (W * H)); \
|
||||
} \
|
||||
\
|
||||
unsigned int vpx_highbd_10_masked_variance##W##x##H##_c(const uint8_t *a, \
|
||||
int a_stride, \
|
||||
const uint8_t *b, \
|
||||
int b_stride, \
|
||||
const uint8_t *m, \
|
||||
int m_stride, \
|
||||
unsigned int *sse) { \
|
||||
int sum; \
|
||||
highbd_10_masked_variance(a, a_stride, b, b_stride, m, m_stride, \
|
||||
W, H, sse, &sum); \
|
||||
return *sse - (((int64_t)sum * sum) / (W * H)); \
|
||||
} \
|
||||
\
|
||||
unsigned int vpx_highbd_12_masked_variance##W##x##H##_c(const uint8_t *a, \
|
||||
int a_stride, \
|
||||
const uint8_t *b, \
|
||||
int b_stride, \
|
||||
const uint8_t *m, \
|
||||
int m_stride, \
|
||||
unsigned int *sse) { \
|
||||
int sum; \
|
||||
highbd_12_masked_variance(a, a_stride, b, b_stride, m, m_stride, \
|
||||
W, H, sse, &sum); \
|
||||
return *sse - (((int64_t)sum * sum) / (W * H)); \
|
||||
}
|
||||
|
||||
#define HIGHBD_MASK_SUBPIX_VAR(W, H) \
|
||||
unsigned int vpx_highbd_masked_sub_pixel_variance##W##x##H##_c( \
|
||||
const uint8_t *src, int src_stride, \
|
||||
int xoffset, int yoffset, \
|
||||
const uint8_t *dst, int dst_stride, \
|
||||
const uint8_t *msk, int msk_stride, \
|
||||
unsigned int *sse) { \
|
||||
uint16_t fdata3[(H + 1) * W]; \
|
||||
uint16_t temp2[H * W]; \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, \
|
||||
H + 1, W, \
|
||||
vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
return vpx_highbd_masked_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), \
|
||||
W, dst, dst_stride, \
|
||||
msk, msk_stride, sse); \
|
||||
} \
|
||||
\
|
||||
unsigned int vpx_highbd_10_masked_sub_pixel_variance##W##x##H##_c( \
|
||||
const uint8_t *src, int src_stride, \
|
||||
int xoffset, int yoffset, \
|
||||
const uint8_t *dst, int dst_stride, \
|
||||
const uint8_t *msk, int msk_stride, \
|
||||
unsigned int *sse) { \
|
||||
uint16_t fdata3[(H + 1) * W]; \
|
||||
uint16_t temp2[H * W]; \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, \
|
||||
H + 1, W, \
|
||||
vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
return vpx_highbd_10_masked_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), \
|
||||
W, dst, dst_stride, \
|
||||
msk, msk_stride, sse); \
|
||||
} \
|
||||
\
|
||||
unsigned int vpx_highbd_12_masked_sub_pixel_variance##W##x##H##_c( \
|
||||
const uint8_t *src, int src_stride, \
|
||||
int xoffset, int yoffset, \
|
||||
const uint8_t *dst, int dst_stride, \
|
||||
const uint8_t *msk, int msk_stride, \
|
||||
unsigned int *sse) { \
|
||||
uint16_t fdata3[(H + 1) * W]; \
|
||||
uint16_t temp2[H * W]; \
|
||||
\
|
||||
highbd_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, \
|
||||
H + 1, W, \
|
||||
vpx_bilinear_filters[xoffset]); \
|
||||
highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
|
||||
vpx_bilinear_filters[yoffset]); \
|
||||
\
|
||||
return vpx_highbd_12_masked_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), \
|
||||
W, dst, dst_stride, \
|
||||
msk, msk_stride, sse); \
|
||||
}
|
||||
|
||||
HIGHBD_MASK_VAR(4, 4)
|
||||
HIGHBD_MASK_SUBPIX_VAR(4, 4)
|
||||
|
||||
HIGHBD_MASK_VAR(4, 8)
|
||||
HIGHBD_MASK_SUBPIX_VAR(4, 8)
|
||||
|
||||
HIGHBD_MASK_VAR(8, 4)
|
||||
HIGHBD_MASK_SUBPIX_VAR(8, 4)
|
||||
|
||||
HIGHBD_MASK_VAR(8, 8)
|
||||
HIGHBD_MASK_SUBPIX_VAR(8, 8)
|
||||
|
||||
HIGHBD_MASK_VAR(8, 16)
|
||||
HIGHBD_MASK_SUBPIX_VAR(8, 16)
|
||||
|
||||
HIGHBD_MASK_VAR(16, 8)
|
||||
HIGHBD_MASK_SUBPIX_VAR(16, 8)
|
||||
|
||||
HIGHBD_MASK_VAR(16, 16)
|
||||
HIGHBD_MASK_SUBPIX_VAR(16, 16)
|
||||
|
||||
HIGHBD_MASK_VAR(16, 32)
|
||||
HIGHBD_MASK_SUBPIX_VAR(16, 32)
|
||||
|
||||
HIGHBD_MASK_VAR(32, 16)
|
||||
HIGHBD_MASK_SUBPIX_VAR(32, 16)
|
||||
|
||||
HIGHBD_MASK_VAR(32, 32)
|
||||
HIGHBD_MASK_SUBPIX_VAR(32, 32)
|
||||
|
||||
HIGHBD_MASK_VAR(32, 64)
|
||||
HIGHBD_MASK_SUBPIX_VAR(32, 64)
|
||||
|
||||
HIGHBD_MASK_VAR(64, 32)
|
||||
HIGHBD_MASK_SUBPIX_VAR(64, 32)
|
||||
|
||||
HIGHBD_MASK_VAR(64, 64)
|
||||
HIGHBD_MASK_SUBPIX_VAR(64, 64)
|
||||
|
||||
#if CONFIG_EXT_PARTITION
|
||||
HIGHBD_MASK_VAR(64, 128)
|
||||
HIGHBD_MASK_SUBPIX_VAR(64, 128)
|
||||
|
||||
HIGHBD_MASK_VAR(128, 64)
|
||||
HIGHBD_MASK_SUBPIX_VAR(128, 64)
|
||||
|
||||
HIGHBD_MASK_VAR(128, 128)
|
||||
HIGHBD_MASK_SUBPIX_VAR(128, 128)
|
||||
#endif // CONFIG_EXT_PARTITION
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
#endif // CONFIG_VP10 && CONFIG_EXT_INTER
|
||||
|
||||
@@ -293,6 +293,13 @@ DSP_SRCS-$(HAVE_SSE4_1) += x86/sad_sse4.asm
|
||||
DSP_SRCS-$(HAVE_AVX2) += x86/sad4d_avx2.c
|
||||
DSP_SRCS-$(HAVE_AVX2) += x86/sad_avx2.c
|
||||
|
||||
ifeq ($(CONFIG_VP10_ENCODER),yes)
|
||||
ifeq ($(CONFIG_EXT_INTER),yes)
|
||||
DSP_SRCS-$(HAVE_SSSE3) += x86/masked_sad_intrin_ssse3.c
|
||||
DSP_SRCS-$(HAVE_SSSE3) += x86/masked_variance_intrin_ssse3.c
|
||||
endif #CONFIG_EXT_INTER
|
||||
endif #CONFIG_VP10_ENCODER
|
||||
|
||||
ifeq ($(CONFIG_USE_X86INC),yes)
|
||||
DSP_SRCS-$(HAVE_SSE) += x86/sad4d_sse2.asm
|
||||
DSP_SRCS-$(HAVE_SSE) += x86/sad_sse2.asm
|
||||
|
||||
@@ -1464,6 +1464,154 @@ add_proto qw/unsigned int vpx_get4x4sse_cs/, "const unsigned char *src_ptr, int
|
||||
|
||||
add_proto qw/void vpx_comp_avg_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride";
|
||||
|
||||
if (vpx_config("CONFIG_EXT_INTER") eq "yes") {
|
||||
add_proto qw/unsigned int vpx_masked_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance4x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance4x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad4x4 ssse3/;
|
||||
|
||||
if (vpx_config("CONFIG_EXT_PARTITION") eq "yes") {
|
||||
add_proto qw/unsigned int vpx_masked_variance128x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance128x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masdctked_variance128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_variance64x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_variance64x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance128x128/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance128x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sub_pixel_variance64x128/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_masked_sub_pixel_variance64x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad128x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad128x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_masked_sad64x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_masked_sad64x128/;
|
||||
}
|
||||
}
|
||||
|
||||
if (vpx_config("CONFIG_AFFINE_MOTION") eq "yes") {
|
||||
add_proto qw/void vpx_upsampled_pred/, "uint8_t *comp_pred, int width, int height, const uint8_t *ref, int ref_stride";
|
||||
specialize qw/vpx_upsampled_pred sse2/;
|
||||
@@ -1700,6 +1848,346 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
|
||||
add_proto qw/unsigned int vpx_highbd_12_mse8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_mse8x8 sse2/;
|
||||
|
||||
if (vpx_config("CONFIG_EXT_INTER") eq "yes") {
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance4x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance4x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_variance4x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance4x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance4x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance4x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad64x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad32x64 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad64x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad32x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad16x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad32x32 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad16x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad16x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad8x16 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad8x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad8x4 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad4x8 ssse3/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad4x4 ssse3/;
|
||||
|
||||
if (vpx_config("CONFIG_EXT_PARTITION") eq "yes") {
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance128x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance128x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_variance64x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_variance64x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance128x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance128x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_variance64x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance64x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance128x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance128x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_variance64x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_variance64x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance128x128/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance128x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sub_pixel_variance64x128/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_masked_sub_pixel_variance64x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance128x128/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance128x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_10_masked_sub_pixel_variance64x128/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_10_masked_sub_pixel_variance64x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance128x128/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance128x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_12_masked_sub_pixel_variance64x128/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *m, int m_stride, unsigned int *sse";
|
||||
specialize qw/vpx_highbd_12_masked_sub_pixel_variance64x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad128x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad128x128/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad128x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad128x64/;
|
||||
|
||||
add_proto qw/unsigned int vpx_highbd_masked_sad64x128/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *mask, int mask_stride";
|
||||
specialize qw/vpx_highbd_masked_sad64x128/;
|
||||
}
|
||||
}
|
||||
|
||||
add_proto qw/void vpx_highbd_comp_avg_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, const uint8_t *ref8, int ref_stride";
|
||||
|
||||
#
|
||||
|
||||
@@ -27,6 +27,10 @@ extern "C" {
|
||||
|
||||
typedef int16_t InterpKernel[SUBPEL_TAPS];
|
||||
|
||||
#define BIL_SUBPEL_BITS 3
|
||||
#define BIL_SUBPEL_SHIFTS (1 << BIL_SUBPEL_BITS)
|
||||
extern const uint8_t vpx_bilinear_filters[BIL_SUBPEL_SHIFTS][2];
|
||||
|
||||
#ifdef __cplusplus
|
||||
} // extern "C"
|
||||
#endif
|
||||
|
||||
367
vpx_dsp/x86/masked_sad_intrin_ssse3.c
Normal file
367
vpx_dsp/x86/masked_sad_intrin_ssse3.c
Normal file
@@ -0,0 +1,367 @@
|
||||
/*
|
||||
* Copyright (c) 2015 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 <stdlib.h>
|
||||
#include <emmintrin.h>
|
||||
#include <tmmintrin.h>
|
||||
|
||||
#include "vpx_ports/mem.h"
|
||||
#include "./vpx_config.h"
|
||||
#include "vpx/vpx_integer.h"
|
||||
|
||||
static INLINE __m128i width8_load_2rows(const uint8_t *ptr, int stride) {
|
||||
__m128i temp1 = _mm_loadl_epi64((const __m128i*)ptr);
|
||||
__m128i temp2 = _mm_loadl_epi64((const __m128i*)(ptr + stride));
|
||||
return _mm_unpacklo_epi64(temp1, temp2);
|
||||
}
|
||||
|
||||
static INLINE __m128i width4_load_4rows(const uint8_t *ptr, int stride) {
|
||||
__m128i temp1 = _mm_cvtsi32_si128(*(const uint32_t*)ptr);
|
||||
__m128i temp2 = _mm_cvtsi32_si128(*(const uint32_t*)(ptr + stride));
|
||||
__m128i temp3 = _mm_unpacklo_epi32(temp1, temp2);
|
||||
temp1 = _mm_cvtsi32_si128(*(const uint32_t*)(ptr + stride * 2));
|
||||
temp2 = _mm_cvtsi32_si128(*(const uint32_t*)(ptr + stride * 3));
|
||||
temp1 = _mm_unpacklo_epi32(temp1, temp2);
|
||||
return _mm_unpacklo_epi64(temp3, temp1);
|
||||
}
|
||||
|
||||
static INLINE unsigned int masked_sad_ssse3(const uint8_t *a_ptr, int a_stride,
|
||||
const uint8_t *b_ptr, int b_stride,
|
||||
const uint8_t *m_ptr, int m_stride,
|
||||
int width, int height);
|
||||
|
||||
static INLINE unsigned int masked_sad8xh_ssse3(const uint8_t *a_ptr,
|
||||
int a_stride,
|
||||
const uint8_t *b_ptr,
|
||||
int b_stride,
|
||||
const uint8_t *m_ptr,
|
||||
int m_stride,
|
||||
int height);
|
||||
|
||||
static INLINE unsigned int masked_sad4xh_ssse3(const uint8_t *a_ptr,
|
||||
int a_stride,
|
||||
const uint8_t *b_ptr,
|
||||
int b_stride,
|
||||
const uint8_t *m_ptr,
|
||||
int m_stride,
|
||||
int height);
|
||||
|
||||
#define MASKSADMXN_SSSE3(m, n) \
|
||||
unsigned int vpx_masked_sad##m##x##n##_ssse3(const uint8_t *src, \
|
||||
int src_stride, \
|
||||
const uint8_t *ref, \
|
||||
int ref_stride, \
|
||||
const uint8_t *msk, \
|
||||
int msk_stride) { \
|
||||
return masked_sad_ssse3(src, src_stride, ref, ref_stride, msk, msk_stride, \
|
||||
m, n); \
|
||||
}
|
||||
|
||||
MASKSADMXN_SSSE3(64, 64)
|
||||
MASKSADMXN_SSSE3(64, 32)
|
||||
MASKSADMXN_SSSE3(32, 64)
|
||||
MASKSADMXN_SSSE3(32, 32)
|
||||
MASKSADMXN_SSSE3(32, 16)
|
||||
MASKSADMXN_SSSE3(16, 32)
|
||||
MASKSADMXN_SSSE3(16, 16)
|
||||
MASKSADMXN_SSSE3(16, 8)
|
||||
|
||||
#define MASKSAD8XN_SSSE3(n) \
|
||||
unsigned int vpx_masked_sad8x##n##_ssse3(const uint8_t *src, \
|
||||
int src_stride, \
|
||||
const uint8_t *ref, \
|
||||
int ref_stride, \
|
||||
const uint8_t *msk, \
|
||||
int msk_stride) { \
|
||||
return masked_sad8xh_ssse3(src, src_stride, ref, ref_stride, msk, \
|
||||
msk_stride, n); \
|
||||
}
|
||||
|
||||
MASKSAD8XN_SSSE3(16)
|
||||
MASKSAD8XN_SSSE3(8)
|
||||
MASKSAD8XN_SSSE3(4)
|
||||
|
||||
#define MASKSAD4XN_SSSE3(n) \
|
||||
unsigned int vpx_masked_sad4x##n##_ssse3(const uint8_t *src, int src_stride, \
|
||||
const uint8_t *ref, int ref_stride, \
|
||||
const uint8_t *msk, int msk_stride) { \
|
||||
return masked_sad4xh_ssse3(src, src_stride, ref, ref_stride, msk, \
|
||||
msk_stride, n); \
|
||||
}
|
||||
|
||||
MASKSAD4XN_SSSE3(8)
|
||||
MASKSAD4XN_SSSE3(4)
|
||||
|
||||
// For width a multiple of 16
|
||||
// Assumes values in m are <=64 and w = 16, 32, or 64
|
||||
static INLINE unsigned int masked_sad_ssse3(const uint8_t *a_ptr, int a_stride,
|
||||
const uint8_t *b_ptr, int b_stride,
|
||||
const uint8_t *m_ptr, int m_stride,
|
||||
int width, int height) {
|
||||
int y, x;
|
||||
__m128i a, b, m, temp1, temp2;
|
||||
__m128i res = _mm_setzero_si128();
|
||||
__m128i one = _mm_set1_epi16(1);
|
||||
// For each row
|
||||
for (y = 0; y < height; y++) {
|
||||
// Covering the full width
|
||||
for (x = 0; x < width; x += 16) {
|
||||
// Load a, b, m in xmm registers
|
||||
a = _mm_loadu_si128((const __m128i*)(a_ptr + x));
|
||||
b = _mm_loadu_si128((const __m128i*)(b_ptr + x));
|
||||
m = _mm_loadu_si128((const __m128i*)(m_ptr + x));
|
||||
|
||||
// Calculate the difference between a & b
|
||||
temp1 = _mm_subs_epu8(a, b);
|
||||
temp2 = _mm_subs_epu8(b, a);
|
||||
temp1 = _mm_or_si128(temp1, temp2);
|
||||
|
||||
// Multiply by m and add together
|
||||
temp2 = _mm_maddubs_epi16(temp1, m);
|
||||
// Pad out row result to 32 bit integers & add to running total
|
||||
res = _mm_add_epi32(res, _mm_madd_epi16(temp2, one));
|
||||
}
|
||||
// Move onto the next row
|
||||
a_ptr += a_stride;
|
||||
b_ptr += b_stride;
|
||||
m_ptr += m_stride;
|
||||
}
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
// sad = (sad + 31) >> 6;
|
||||
return (_mm_cvtsi128_si32(res) + 31) >> 6;
|
||||
}
|
||||
|
||||
static INLINE unsigned int masked_sad8xh_ssse3(const uint8_t *a_ptr,
|
||||
int a_stride,
|
||||
const uint8_t *b_ptr,
|
||||
int b_stride,
|
||||
const uint8_t *m_ptr,
|
||||
int m_stride,
|
||||
int height) {
|
||||
int y;
|
||||
__m128i a, b, m, temp1, temp2, row_res;
|
||||
__m128i res = _mm_setzero_si128();
|
||||
__m128i one = _mm_set1_epi16(1);
|
||||
// Add the masked SAD for 2 rows at a time
|
||||
for (y = 0; y < height; y += 2) {
|
||||
// Load a, b, m in xmm registers
|
||||
a = width8_load_2rows(a_ptr, a_stride);
|
||||
b = width8_load_2rows(b_ptr, b_stride);
|
||||
m = width8_load_2rows(m_ptr, m_stride);
|
||||
|
||||
// Calculate the difference between a & b
|
||||
temp1 = _mm_subs_epu8(a, b);
|
||||
temp2 = _mm_subs_epu8(b, a);
|
||||
temp1 = _mm_or_si128(temp1, temp2);
|
||||
|
||||
// Multiply by m and add together
|
||||
row_res = _mm_maddubs_epi16(temp1, m);
|
||||
|
||||
// Pad out row result to 32 bit integers & add to running total
|
||||
res = _mm_add_epi32(res, _mm_madd_epi16(row_res, one));
|
||||
|
||||
// Move onto the next rows
|
||||
a_ptr += a_stride * 2;
|
||||
b_ptr += b_stride * 2;
|
||||
m_ptr += m_stride * 2;
|
||||
}
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
// sad = (sad + 31) >> 6;
|
||||
return (_mm_cvtsi128_si32(res) + 31) >> 6;
|
||||
}
|
||||
|
||||
static INLINE unsigned int masked_sad4xh_ssse3(const uint8_t *a_ptr,
|
||||
int a_stride,
|
||||
const uint8_t *b_ptr,
|
||||
int b_stride,
|
||||
const uint8_t *m_ptr,
|
||||
int m_stride,
|
||||
int height) {
|
||||
int y;
|
||||
__m128i a, b, m, temp1, temp2, row_res;
|
||||
__m128i res = _mm_setzero_si128();
|
||||
__m128i one = _mm_set1_epi16(1);
|
||||
// Add the masked SAD for 4 rows at a time
|
||||
for (y = 0; y < height; y += 4) {
|
||||
// Load a, b, m in xmm registers
|
||||
a = width4_load_4rows(a_ptr, a_stride);
|
||||
b = width4_load_4rows(b_ptr, b_stride);
|
||||
m = width4_load_4rows(m_ptr, m_stride);
|
||||
|
||||
// Calculate the difference between a & b
|
||||
temp1 = _mm_subs_epu8(a, b);
|
||||
temp2 = _mm_subs_epu8(b, a);
|
||||
temp1 = _mm_or_si128(temp1, temp2);
|
||||
|
||||
// Multiply by m and add together
|
||||
row_res = _mm_maddubs_epi16(temp1, m);
|
||||
|
||||
// Pad out row result to 32 bit integers & add to running total
|
||||
res = _mm_add_epi32(res, _mm_madd_epi16(row_res, one));
|
||||
|
||||
// Move onto the next rows
|
||||
a_ptr += a_stride * 4;
|
||||
b_ptr += b_stride * 4;
|
||||
m_ptr += m_stride * 4;
|
||||
}
|
||||
// Pad out row result to 32 bit integers & add to running total
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
// sad = (sad + 31) >> 6;
|
||||
return (_mm_cvtsi128_si32(res) + 31) >> 6;
|
||||
}
|
||||
|
||||
#if CONFIG_VP9_HIGHBITDEPTH
|
||||
static INLINE __m128i highbd_width4_load_2rows(const uint16_t *ptr,
|
||||
int stride) {
|
||||
__m128i temp1 = _mm_loadl_epi64((const __m128i*)ptr);
|
||||
__m128i temp2 = _mm_loadl_epi64((const __m128i*)(ptr + stride));
|
||||
return _mm_unpacklo_epi64(temp1, temp2);
|
||||
}
|
||||
|
||||
static INLINE unsigned int highbd_masked_sad_ssse3(const uint8_t *a8_ptr,
|
||||
int a_stride,
|
||||
const uint8_t *b8_ptr,
|
||||
int b_stride,
|
||||
const uint8_t *m_ptr,
|
||||
int m_stride,
|
||||
int width, int height);
|
||||
|
||||
static INLINE unsigned int highbd_masked_sad4xh_ssse3(const uint8_t *a8_ptr,
|
||||
int a_stride,
|
||||
const uint8_t *b8_ptr,
|
||||
int b_stride,
|
||||
const uint8_t *m_ptr,
|
||||
int m_stride,
|
||||
int height);
|
||||
|
||||
#define HIGHBD_MASKSADMXN_SSSE3(m, n) \
|
||||
unsigned int vpx_highbd_masked_sad##m##x##n##_ssse3(const uint8_t *src, \
|
||||
int src_stride, \
|
||||
const uint8_t *ref, \
|
||||
int ref_stride, \
|
||||
const uint8_t *msk, \
|
||||
int msk_stride) { \
|
||||
return highbd_masked_sad_ssse3(src, src_stride, ref, ref_stride, msk, \
|
||||
msk_stride, m, n); \
|
||||
}
|
||||
|
||||
HIGHBD_MASKSADMXN_SSSE3(64, 64)
|
||||
HIGHBD_MASKSADMXN_SSSE3(64, 32)
|
||||
HIGHBD_MASKSADMXN_SSSE3(32, 64)
|
||||
HIGHBD_MASKSADMXN_SSSE3(32, 32)
|
||||
HIGHBD_MASKSADMXN_SSSE3(32, 16)
|
||||
HIGHBD_MASKSADMXN_SSSE3(16, 32)
|
||||
HIGHBD_MASKSADMXN_SSSE3(16, 16)
|
||||
HIGHBD_MASKSADMXN_SSSE3(16, 8)
|
||||
HIGHBD_MASKSADMXN_SSSE3(8, 16)
|
||||
HIGHBD_MASKSADMXN_SSSE3(8, 8)
|
||||
HIGHBD_MASKSADMXN_SSSE3(8, 4)
|
||||
|
||||
#define HIGHBD_MASKSAD4XN_SSSE3(n) \
|
||||
unsigned int vpx_highbd_masked_sad4x##n##_ssse3(const uint8_t *src, \
|
||||
int src_stride, \
|
||||
const uint8_t *ref, \
|
||||
int ref_stride, \
|
||||
const uint8_t *msk, \
|
||||
int msk_stride) { \
|
||||
return highbd_masked_sad4xh_ssse3(src, src_stride, ref, ref_stride, msk, \
|
||||
msk_stride, n); \
|
||||
}
|
||||
|
||||
HIGHBD_MASKSAD4XN_SSSE3(8)
|
||||
HIGHBD_MASKSAD4XN_SSSE3(4)
|
||||
|
||||
// For width a multiple of 8
|
||||
// Assumes values in m are <=64
|
||||
static INLINE unsigned int highbd_masked_sad_ssse3(const uint8_t *a8_ptr,
|
||||
int a_stride,
|
||||
const uint8_t *b8_ptr,
|
||||
int b_stride,
|
||||
const uint8_t *m_ptr,
|
||||
int m_stride,
|
||||
int width, int height) {
|
||||
int y, x;
|
||||
__m128i a, b, m, temp1, temp2;
|
||||
const uint16_t *a_ptr = CONVERT_TO_SHORTPTR(a8_ptr);
|
||||
const uint16_t *b_ptr = CONVERT_TO_SHORTPTR(b8_ptr);
|
||||
__m128i res = _mm_setzero_si128();
|
||||
// For each row
|
||||
for (y = 0; y < height; y++) {
|
||||
// Covering the full width
|
||||
for (x = 0; x < width; x += 8) {
|
||||
// Load a, b, m in xmm registers
|
||||
a = _mm_loadu_si128((const __m128i*)(a_ptr + x));
|
||||
b = _mm_loadu_si128((const __m128i*)(b_ptr + x));
|
||||
m = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i*)(m_ptr + x)),
|
||||
_mm_setzero_si128());
|
||||
|
||||
// Calculate the difference between a & b
|
||||
temp1 = _mm_subs_epu16(a, b);
|
||||
temp2 = _mm_subs_epu16(b, a);
|
||||
temp1 = _mm_or_si128(temp1, temp2);
|
||||
|
||||
// Add result of multiplying by m and add pairs together to running total
|
||||
res = _mm_add_epi32(res, _mm_madd_epi16(temp1, m));
|
||||
}
|
||||
// Move onto the next row
|
||||
a_ptr += a_stride;
|
||||
b_ptr += b_stride;
|
||||
m_ptr += m_stride;
|
||||
}
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
// sad = (sad + 31) >> 6;
|
||||
return (_mm_cvtsi128_si32(res) + 31) >> 6;
|
||||
}
|
||||
|
||||
static INLINE unsigned int highbd_masked_sad4xh_ssse3(const uint8_t *a8_ptr,
|
||||
int a_stride,
|
||||
const uint8_t *b8_ptr,
|
||||
int b_stride,
|
||||
const uint8_t *m_ptr,
|
||||
int m_stride,
|
||||
int height) {
|
||||
int y;
|
||||
__m128i a, b, m, temp1, temp2;
|
||||
const uint16_t *a_ptr = CONVERT_TO_SHORTPTR(a8_ptr);
|
||||
const uint16_t *b_ptr = CONVERT_TO_SHORTPTR(b8_ptr);
|
||||
__m128i res = _mm_setzero_si128();
|
||||
// Add the masked SAD for 2 rows at a time
|
||||
for (y = 0; y < height; y += 2) {
|
||||
// Load a, b, m in xmm registers
|
||||
a = highbd_width4_load_2rows(a_ptr, a_stride);
|
||||
b = highbd_width4_load_2rows(b_ptr, b_stride);
|
||||
temp1 = _mm_loadl_epi64((const __m128i*)m_ptr);
|
||||
temp2 = _mm_loadl_epi64((const __m128i*)(m_ptr + m_stride));
|
||||
m = _mm_unpacklo_epi8(_mm_unpacklo_epi32(temp1, temp2),
|
||||
_mm_setzero_si128());
|
||||
|
||||
// Calculate the difference between a & b
|
||||
temp1 = _mm_subs_epu16(a, b);
|
||||
temp2 = _mm_subs_epu16(b, a);
|
||||
temp1 = _mm_or_si128(temp1, temp2);
|
||||
|
||||
// Multiply by m and add together
|
||||
res = _mm_add_epi32(res, _mm_madd_epi16(temp1, m));
|
||||
|
||||
// Move onto the next rows
|
||||
a_ptr += a_stride * 2;
|
||||
b_ptr += b_stride * 2;
|
||||
m_ptr += m_stride * 2;
|
||||
}
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
res = _mm_hadd_epi32(res, _mm_setzero_si128());
|
||||
// sad = (sad + 31) >> 6;
|
||||
return (_mm_cvtsi128_si32(res) + 31) >> 6;
|
||||
}
|
||||
#endif // CONFIG_VP9_HIGHBITDEPTH
|
||||
2096
vpx_dsp/x86/masked_variance_intrin_ssse3.c
Normal file
2096
vpx_dsp/x86/masked_variance_intrin_ssse3.c
Normal file
File diff suppressed because it is too large
Load Diff
Reference in New Issue
Block a user