generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge branch 'experimental' into master

Browse Source 
VP9 preview bitstream 2, commit '868ecb55a1528ca3f19286e7d1551572bf89b642'

Conflicts:
	vp9/vp9_common.mk

Change-Id: I3f0f6e692c987ff24f98ceafbb86cb9cf64ad8d3
This commit is contained in:
John Koleszar 2013-04-12 15:33:04 -07:00
commit 7f7d1357a2
209 changed files with 30236 additions and 26671 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.gitignore vendored
View File

@ -3,6 +3,8 @@
*.d
*.o
*~
/*.ivf
/*.ivf.md5
/*-*.mk
/*.asm
/*.doxy

23
build/make/configure.sh
View File

@ -460,6 +460,7 @@ write_common_target_config_h() {
#ifndef VPX_CONFIG_H
#define VPX_CONFIG_H
#define RESTRICT ${RESTRICT}
#define INLINE ${INLINE}
EOF
print_config_h ARCH "${TMP_H}" ${ARCH_LIST}
print_config_h HAVE "${TMP_H}" ${HAVE_LIST}
@ -1005,12 +1006,6 @@ process_common_toolchain() {
#error "not x32"
#endif
EOF
soft_enable runtime_cpu_detect
soft_enable mmx
soft_enable sse
soft_enable sse2
soft_enable sse3
soft_enable ssse3
case ${tgt_os} in
win*)
@ -1064,9 +1059,15 @@ EOF
;;
esac
soft_enable runtime_cpu_detect
soft_enable mmx
soft_enable sse
soft_enable sse2
soft_enable sse3
soft_enable ssse3
# We can't use 'check_cflags' until the compiler is configured and CC is
# populated.
if enabled gcc && ! disabled sse4_1 && ! check_cflags -msse4.1; then
if enabled gcc && ! disabled sse4_1 && ! check_cflags -msse4; then
RTCD_OPTIONS="${RTCD_OPTIONS}--disable-sse4_1 "
else
soft_enable sse4_1
@ -1174,6 +1175,14 @@ EOF
[ -f "${TMP_O}" ] && od -A n -t x1 "${TMP_O}" | tr -d '\n' |
grep '4f *32 *42 *45' >/dev/null 2>&1 && enable big_endian
# Try to find which inline keywords are supported
check_cc <<EOF && INLINE="inline"
static inline function() {}
EOF
check_cc <<EOF && INLINE="__attribute__((always_inline))"
static __attribute__((always_inline)) function() {}
EOF
# Almost every platform uses pthreads.
if enabled multithread; then
case ${toolchain} in

12
configure vendored
View File

@ -239,17 +239,18 @@ HAVE_LIST="
"
EXPERIMENT_LIST="
csm
lossless
new_mvref
implicit_segmentation
newbintramodes
comp_interintra_pred
tx64x64
dwtdcthybrid
cnvcontext
newcoefcontext
enable_6tap
abovesprefmv
code_nonzerocount
useselectrefmv
modelcoefprob
loop_dering
implicit_compoundinter_weight
scatterscan
"
CONFIG_LIST="
external_build
@ -647,6 +648,7 @@ process_toolchain() {
enable solution
vs_version=${tgt_cc##vs}
all_targets="${all_targets} solution"
INLINE="__forceinline"
;;
esac

16
test/altref_test.cc
View File

@ -8,19 +8,20 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "test/util.h"
namespace {
// lookahead range: [kLookAheadMin, kLookAheadMax).
const int kLookAheadMin = 5;
const int kLookAheadMax = 26;
class AltRefTest : public libvpx_test::EncoderTest,
public ::testing::TestWithParam<int> {
class AltRefTest : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<int> {
protected:
AltRefTest() : altref_count_(0) {}
AltRefTest() : EncoderTest(GET_PARAM(0)), altref_count_(0) {}
virtual ~AltRefTest() {}
virtual void SetUp() {
@ -58,7 +59,7 @@ TEST_P(AltRefTest, MonotonicTimestamps) {
const vpx_rational timebase = { 33333333, 1000000000 };
cfg_.g_timebase = timebase;
cfg_.rc_target_bitrate = 1000;
cfg_.g_lag_in_frames = GetParam();
cfg_.g_lag_in_frames = GET_PARAM(1);
libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
timebase.den, timebase.num, 0, 30);
@ -66,6 +67,7 @@ TEST_P(AltRefTest, MonotonicTimestamps) {
EXPECT_GE(altref_count(), 1);
}
INSTANTIATE_TEST_CASE_P(NonZeroLag, AltRefTest,
::testing::Range(kLookAheadMin, kLookAheadMax));
VP8_INSTANTIATE_TEST_CASE(AltRefTest,
::testing::Range(kLookAheadMin, kLookAheadMax));
} // namespace

232
test/codec_factory.h Normal file
View File

@ -0,0 +1,232 @@
/*
* Copyright (c) 2013 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.
*/
#ifndef TEST_CODEC_FACTORY_H_
#define TEST_CODEC_FACTORY_H_
extern "C" {
#include "./vpx_config.h"
#include "vpx/vpx_decoder.h"
#include "vpx/vpx_encoder.h"
#if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
#include "vpx/vp8cx.h"
#endif
#if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
#include "vpx/vp8dx.h"
#endif
}
#include "test/decode_test_driver.h"
#include "test/encode_test_driver.h"
namespace libvpx_test {
class CodecFactory {
public:
CodecFactory() {}
virtual ~CodecFactory() {}
virtual Decoder* CreateDecoder(vpx_codec_dec_cfg_t cfg,
unsigned long deadline) const = 0;
virtual Encoder* CreateEncoder(vpx_codec_enc_cfg_t cfg,
unsigned long deadline,
const unsigned long init_flags,
TwopassStatsStore *stats) const = 0;
virtual vpx_codec_err_t DefaultEncoderConfig(vpx_codec_enc_cfg_t *cfg,
int usage) const = 0;
};
/* Provide CodecTestWith<n>Params classes for a variable number of parameters
* to avoid having to include a pointer to the CodecFactory in every test
* definition.
*/
template<class T1>
class CodecTestWithParam : public ::testing::TestWithParam<
std::tr1::tuple< const libvpx_test::CodecFactory*, T1 > > {
};
template<class T1, class T2>
class CodecTestWith2Params : public ::testing::TestWithParam<
std::tr1::tuple< const libvpx_test::CodecFactory*, T1, T2 > > {
};
template<class T1, class T2, class T3>
class CodecTestWith3Params : public ::testing::TestWithParam<
std::tr1::tuple< const libvpx_test::CodecFactory*, T1, T2, T3 > > {
};
/*
* VP8 Codec Definitions
*/
#if CONFIG_VP8
class VP8Decoder : public Decoder {
public:
VP8Decoder(vpx_codec_dec_cfg_t cfg, unsigned long deadline)
: Decoder(cfg, deadline) {}
protected:
virtual const vpx_codec_iface_t* CodecInterface() const {
#if CONFIG_VP8_DECODER
return &vpx_codec_vp8_dx_algo;
#else
return NULL;
#endif
}
};
class VP8Encoder : public Encoder {
public:
VP8Encoder(vpx_codec_enc_cfg_t cfg, unsigned long deadline,
const unsigned long init_flags, TwopassStatsStore *stats)
: Encoder(cfg, deadline, init_flags, stats) {}
protected:
virtual const vpx_codec_iface_t* CodecInterface() const {
#if CONFIG_VP8_ENCODER
return &vpx_codec_vp8_cx_algo;
#else
return NULL;
#endif
}
};
class VP8CodecFactory : public CodecFactory {
public:
VP8CodecFactory() : CodecFactory() {}
virtual Decoder* CreateDecoder(vpx_codec_dec_cfg_t cfg,
unsigned long deadline) const {
#if CONFIG_VP8_DECODER
return new VP8Decoder(cfg, deadline);
#else
return NULL;
#endif
}
virtual Encoder* CreateEncoder(vpx_codec_enc_cfg_t cfg,
unsigned long deadline,
const unsigned long init_flags,
TwopassStatsStore *stats) const {
#if CONFIG_VP8_ENCODER
return new VP8Encoder(cfg, deadline, init_flags, stats);
#else
return NULL;
#endif
}
virtual vpx_codec_err_t DefaultEncoderConfig(vpx_codec_enc_cfg_t *cfg,
int usage) const {
#if CONFIG_VP8_ENCODER
return vpx_codec_enc_config_default(&vpx_codec_vp8_cx_algo, cfg, usage);
#else
return VPX_CODEC_INCAPABLE;
#endif
}
};
const libvpx_test::VP8CodecFactory kVP8;
#define VP8_INSTANTIATE_TEST_CASE(test, params)\
INSTANTIATE_TEST_CASE_P(VP8, test, \
::testing::Combine( \
::testing::Values(static_cast<const libvpx_test::CodecFactory*>( \
&libvpx_test::kVP8)), \
params))
#else
#define VP8_INSTANTIATE_TEST_CASE(test, params)
#endif // CONFIG_VP8
/*
* VP9 Codec Definitions
*/
#if CONFIG_VP9
class VP9Decoder : public Decoder {
public:
VP9Decoder(vpx_codec_dec_cfg_t cfg, unsigned long deadline)
: Decoder(cfg, deadline) {}
protected:
virtual const vpx_codec_iface_t* CodecInterface() const {
#if CONFIG_VP9_DECODER
return &vpx_codec_vp9_dx_algo;
#else
return NULL;
#endif
}
};
class VP9Encoder : public Encoder {
public:
VP9Encoder(vpx_codec_enc_cfg_t cfg, unsigned long deadline,
const unsigned long init_flags, TwopassStatsStore *stats)
: Encoder(cfg, deadline, init_flags, stats) {}
protected:
virtual const vpx_codec_iface_t* CodecInterface() const {
#if CONFIG_VP9_ENCODER
return &vpx_codec_vp9_cx_algo;
#else
return NULL;
#endif
}
};
class VP9CodecFactory : public CodecFactory {
public:
VP9CodecFactory() : CodecFactory() {}
virtual Decoder* CreateDecoder(vpx_codec_dec_cfg_t cfg,
unsigned long deadline) const {
#if CONFIG_VP9_DECODER
return new VP9Decoder(cfg, deadline);
#else
return NULL;
#endif
}
virtual Encoder* CreateEncoder(vpx_codec_enc_cfg_t cfg,
unsigned long deadline,
const unsigned long init_flags,
TwopassStatsStore *stats) const {
#if CONFIG_VP9_ENCODER
return new VP9Encoder(cfg, deadline, init_flags, stats);
#else
return NULL;
#endif
}
virtual vpx_codec_err_t DefaultEncoderConfig(vpx_codec_enc_cfg_t *cfg,
int usage) const {
#if CONFIG_VP9_ENCODER
return vpx_codec_enc_config_default(&vpx_codec_vp9_cx_algo, cfg, usage);
#else
return VPX_CODEC_INCAPABLE;
#endif
}
};
const libvpx_test::VP9CodecFactory kVP9;
#define VP9_INSTANTIATE_TEST_CASE(test, params)\
INSTANTIATE_TEST_CASE_P(VP9, test, \
::testing::Combine( \
::testing::Values(static_cast<const libvpx_test::CodecFactory*>( \
&libvpx_test::kVP9)), \
params))
#else
#define VP9_INSTANTIATE_TEST_CASE(test, params)
#endif // CONFIG_VP9
} // namespace libvpx_test
#endif // TEST_CODEC_FACTORY_H_

14
test/config_test.cc
View File

@ -8,20 +8,22 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/util.h"
#include "test/video_source.h"
namespace {
class ConfigTest : public ::libvpx_test::EncoderTest,
public ::testing::TestWithParam<enum libvpx_test::TestMode> {
public:
ConfigTest() : frame_count_in_(0), frame_count_out_(0), frame_count_max_(0) {}
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
ConfigTest() : EncoderTest(GET_PARAM(0)),
frame_count_in_(0), frame_count_out_(0), frame_count_max_(0) {}
virtual void SetUp() {
InitializeConfig();
SetMode(GetParam());
SetMode(GET_PARAM(1));
}
virtual void BeginPassHook(unsigned int /*pass*/) {
@ -57,5 +59,5 @@ TEST_P(ConfigTest, LagIsDisabled) {
EXPECT_EQ(frame_count_in_, frame_count_out_);
}
INSTANTIATE_TEST_CASE_P(OnePassModes, ConfigTest, ONE_PASS_TEST_MODES);
VP8_INSTANTIATE_TEST_CASE(ConfigTest, ONE_PASS_TEST_MODES);
} // namespace

509
test/convolve_test.cc Normal file
View File

@ -0,0 +1,509 @@
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
extern "C" {
#include "./vpx_config.h"
#include "./vp9_rtcd.h"
#include "vp9/common/vp9_filter.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"
}
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/acm_random.h"
#include "test/register_state_check.h"
#include "test/util.h"
namespace {
typedef void (*convolve_fn_t)(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h);
struct ConvolveFunctions {
ConvolveFunctions(convolve_fn_t h8, convolve_fn_t h8_avg,
convolve_fn_t v8, convolve_fn_t v8_avg,
convolve_fn_t hv8, convolve_fn_t hv8_avg)
: h8_(h8), v8_(v8), hv8_(hv8), h8_avg_(h8_avg), v8_avg_(v8_avg),
hv8_avg_(hv8_avg) {}
convolve_fn_t h8_;
convolve_fn_t v8_;
convolve_fn_t hv8_;
convolve_fn_t h8_avg_;
convolve_fn_t v8_avg_;
convolve_fn_t hv8_avg_;
};
// Reference 8-tap subpixel filter, slightly modified to fit into this test.
#define VP9_FILTER_WEIGHT 128
#define VP9_FILTER_SHIFT 7
static uint8_t clip_pixel(int x) {
return x < 0 ? 0 :
x > 255 ? 255 :
x;
}
static void filter_block2d_8_c(const uint8_t *src_ptr,
const unsigned int src_stride,
const int16_t *HFilter,
const int16_t *VFilter,
uint8_t *dst_ptr,
unsigned int dst_stride,
unsigned int output_width,
unsigned int output_height) {
// Between passes, we use an intermediate buffer whose height is extended to
// have enough horizontally filtered values as input for the vertical pass.
// This buffer is allocated to be big enough for the largest block type we
// support.
const int kInterp_Extend = 4;
const unsigned int intermediate_height =
(kInterp_Extend - 1) + output_height + kInterp_Extend;
/* Size of intermediate_buffer is max_intermediate_height * filter_max_width,
* where max_intermediate_height = (kInterp_Extend - 1) + filter_max_height
* + kInterp_Extend
* = 3 + 16 + 4
* = 23
* and filter_max_width = 16
*/
uint8_t intermediate_buffer[23 * 16];
const int intermediate_next_stride = 1 - intermediate_height * output_width;
// Horizontal pass (src -> transposed intermediate).
{
uint8_t *output_ptr = intermediate_buffer;
const int src_next_row_stride = src_stride - output_width;
unsigned int i, j;
src_ptr -= (kInterp_Extend - 1) * src_stride + (kInterp_Extend - 1);
for (i = 0; i < intermediate_height; ++i) {
for (j = 0; j < output_width; ++j) {
// Apply filter...
int temp = ((int)src_ptr[0] * HFilter[0]) +
((int)src_ptr[1] * HFilter[1]) +
((int)src_ptr[2] * HFilter[2]) +
((int)src_ptr[3] * HFilter[3]) +
((int)src_ptr[4] * HFilter[4]) +
((int)src_ptr[5] * HFilter[5]) +
((int)src_ptr[6] * HFilter[6]) +
((int)src_ptr[7] * HFilter[7]) +
(VP9_FILTER_WEIGHT >> 1); // Rounding
// Normalize back to 0-255...
*output_ptr = clip_pixel(temp >> VP9_FILTER_SHIFT);
++src_ptr;
output_ptr += intermediate_height;
}
src_ptr += src_next_row_stride;
output_ptr += intermediate_next_stride;
}
}
// Vertical pass (transposed intermediate -> dst).
{
uint8_t *src_ptr = intermediate_buffer;
const int dst_next_row_stride = dst_stride - output_width;
unsigned int i, j;
for (i = 0; i < output_height; ++i) {
for (j = 0; j < output_width; ++j) {
// Apply filter...
int temp = ((int)src_ptr[0] * VFilter[0]) +
((int)src_ptr[1] * VFilter[1]) +
((int)src_ptr[2] * VFilter[2]) +
((int)src_ptr[3] * VFilter[3]) +
((int)src_ptr[4] * VFilter[4]) +
((int)src_ptr[5] * VFilter[5]) +
((int)src_ptr[6] * VFilter[6]) +
((int)src_ptr[7] * VFilter[7]) +
(VP9_FILTER_WEIGHT >> 1); // Rounding
// Normalize back to 0-255...
*dst_ptr++ = clip_pixel(temp >> VP9_FILTER_SHIFT);
src_ptr += intermediate_height;
}
src_ptr += intermediate_next_stride;
dst_ptr += dst_next_row_stride;
}
}
}
static void block2d_average_c(uint8_t *src,
unsigned int src_stride,
uint8_t *output_ptr,
unsigned int output_stride,
unsigned int output_width,
unsigned int output_height) {
unsigned int i, j;
for (i = 0; i < output_height; ++i) {
for (j = 0; j < output_width; ++j) {
output_ptr[j] = (output_ptr[j] + src[i * src_stride + j] + 1) >> 1;
}
output_ptr += output_stride;
}
}
static void filter_average_block2d_8_c(const uint8_t *src_ptr,
const unsigned int src_stride,
const int16_t *HFilter,
const int16_t *VFilter,
uint8_t *dst_ptr,
unsigned int dst_stride,
unsigned int output_width,
unsigned int output_height) {
uint8_t tmp[16*16];
assert(output_width <= 16);
assert(output_height <= 16);
filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter, tmp, 16,
output_width, output_height);
block2d_average_c(tmp, 16, dst_ptr, dst_stride,
output_width, output_height);
}
class ConvolveTest : public PARAMS(int, int, const ConvolveFunctions*) {
public:
static void SetUpTestCase() {
// Force input_ to be unaligned, output to be 16 byte aligned.
input_ = reinterpret_cast<uint8_t*>(
vpx_memalign(kDataAlignment, kOuterBlockSize * kOuterBlockSize + 1))
+ 1;
output_ = reinterpret_cast<uint8_t*>(
vpx_memalign(kDataAlignment, kOuterBlockSize * kOuterBlockSize));
}
static void TearDownTestCase() {
vpx_free(input_ - 1);
input_ = NULL;
vpx_free(output_);
output_ = NULL;
}
protected:
static const int kDataAlignment = 16;
static const int kOuterBlockSize = 32;
static const int kInputStride = kOuterBlockSize;
static const int kOutputStride = kOuterBlockSize;
static const int kMaxDimension = 16;
int Width() const { return GET_PARAM(0); }
int Height() const { return GET_PARAM(1); }
int BorderLeft() const {
const int center = (kOuterBlockSize - Width()) / 2;
return (center + (kDataAlignment - 1)) & ~(kDataAlignment - 1);
}
int BorderTop() const { return (kOuterBlockSize - Height()) / 2; }
bool IsIndexInBorder(int i) {
return (i < BorderTop() * kOuterBlockSize ||
i >= (BorderTop() + Height()) * kOuterBlockSize ||
i % kOuterBlockSize < BorderLeft() ||
i % kOuterBlockSize >= (BorderLeft() + Width()));
}
virtual void SetUp() {
UUT_ = GET_PARAM(2);
memset(input_, 0, sizeof(input_));
/* Set up guard blocks for an inner block cetered in the outer block */
for (int i = 0; i < kOuterBlockSize * kOuterBlockSize; ++i) {
if (IsIndexInBorder(i))
output_[i] = 255;
else
output_[i] = 0;
}
::libvpx_test::ACMRandom prng;
for (int i = 0; i < kOuterBlockSize * kOuterBlockSize; ++i)
input_[i] = prng.Rand8();
}
void CheckGuardBlocks() {
for (int i = 0; i < kOuterBlockSize * kOuterBlockSize; ++i) {
if (IsIndexInBorder(i))
EXPECT_EQ(255, output_[i]);
}
}
uint8_t* input() {
return input_ + BorderTop() * kOuterBlockSize + BorderLeft();
}
uint8_t* output() {
return output_ + BorderTop() * kOuterBlockSize + BorderLeft();
}
const ConvolveFunctions* UUT_;
static uint8_t* input_;
static uint8_t* output_;
};
uint8_t* ConvolveTest::input_ = NULL;
uint8_t* ConvolveTest::output_ = NULL;
TEST_P(ConvolveTest, GuardBlocks) {
CheckGuardBlocks();
}
TEST_P(ConvolveTest, CopyHoriz) {
uint8_t* const in = input();
uint8_t* const out = output();
const int16_t filter8[8] = {0, 0, 0, 128, 0, 0, 0, 0};
REGISTER_STATE_CHECK(
UUT_->h8_(in, kInputStride, out, kOutputStride, filter8, 16, filter8, 16,
Width(), Height()));
CheckGuardBlocks();
for (int y = 0; y < Height(); ++y)
for (int x = 0; x < Width(); ++x)
ASSERT_EQ(out[y * kOutputStride + x], in[y * kInputStride + x])
<< "(" << x << "," << y << ")";
}
TEST_P(ConvolveTest, CopyVert) {
uint8_t* const in = input();
uint8_t* const out = output();
const int16_t filter8[8] = {0, 0, 0, 128, 0, 0, 0, 0};
REGISTER_STATE_CHECK(
UUT_->v8_(in, kInputStride, out, kOutputStride, filter8, 16, filter8, 16,
Width(), Height()));
CheckGuardBlocks();
for (int y = 0; y < Height(); ++y)
for (int x = 0; x < Width(); ++x)
ASSERT_EQ(out[y * kOutputStride + x], in[y * kInputStride + x])
<< "(" << x << "," << y << ")";
}
TEST_P(ConvolveTest, Copy2D) {
uint8_t* const in = input();
uint8_t* const out = output();
const int16_t filter8[8] = {0, 0, 0, 128, 0, 0, 0, 0};
REGISTER_STATE_CHECK(
UUT_->hv8_(in, kInputStride, out, kOutputStride, filter8, 16, filter8, 16,
Width(), Height()));
CheckGuardBlocks();
for (int y = 0; y < Height(); ++y)
for (int x = 0; x < Width(); ++x)
ASSERT_EQ(out[y * kOutputStride + x], in[y * kInputStride + x])
<< "(" << x << "," << y << ")";
}
const int16_t (*kTestFilterList[])[8] = {
vp9_bilinear_filters,
vp9_sub_pel_filters_6,
vp9_sub_pel_filters_8,
vp9_sub_pel_filters_8s,
vp9_sub_pel_filters_8lp
};
const int16_t kInvalidFilter[8] = { 0 };
TEST_P(ConvolveTest, MatchesReferenceSubpixelFilter) {
uint8_t* const in = input();
uint8_t* const out = output();
uint8_t ref[kOutputStride * kMaxDimension];
const int kNumFilterBanks = sizeof(kTestFilterList) /
sizeof(kTestFilterList[0]);
for (int filter_bank = 0; filter_bank < kNumFilterBanks; ++filter_bank) {
const int16_t (*filters)[8] = kTestFilterList[filter_bank];
const int kNumFilters = 16;
for (int filter_x = 0; filter_x < kNumFilters; ++filter_x) {
for (int filter_y = 0; filter_y < kNumFilters; ++filter_y) {
filter_block2d_8_c(in, kInputStride,
filters[filter_x], filters[filter_y],
ref, kOutputStride,
Width(), Height());
if (filters == vp9_sub_pel_filters_8lp || (filter_x && filter_y))
REGISTER_STATE_CHECK(
UUT_->hv8_(in, kInputStride, out, kOutputStride,
filters[filter_x], 16, filters[filter_y], 16,
Width(), Height()));
else if (filter_y)
REGISTER_STATE_CHECK(
UUT_->v8_(in, kInputStride, out, kOutputStride,
kInvalidFilter, 16, filters[filter_y], 16,
Width(), Height()));
else
REGISTER_STATE_CHECK(
UUT_->h8_(in, kInputStride, out, kOutputStride,
filters[filter_x], 16, kInvalidFilter, 16,
Width(), Height()));
CheckGuardBlocks();
for (int y = 0; y < Height(); ++y)
for (int x = 0; x < Width(); ++x)
ASSERT_EQ(ref[y * kOutputStride + x], out[y * kOutputStride + x])
<< "mismatch at (" << x << "," << y << "), "
<< "filters (" << filter_bank << ","
<< filter_x << "," << filter_y << ")";
}
}
}
}
TEST_P(ConvolveTest, MatchesReferenceAveragingSubpixelFilter) {
uint8_t* const in = input();
uint8_t* const out = output();
uint8_t ref[kOutputStride * kMaxDimension];
// Populate ref and out with some random data
::libvpx_test::ACMRandom prng;
for (int y = 0; y < Height(); ++y) {
for (int x = 0; x < Width(); ++x) {
const uint8_t r = prng.Rand8();
out[y * kOutputStride + x] = r;
ref[y * kOutputStride + x] = r;
}
}
const int kNumFilterBanks = sizeof(kTestFilterList) /
sizeof(kTestFilterList[0]);
for (int filter_bank = 0; filter_bank < kNumFilterBanks; ++filter_bank) {
const int16_t (*filters)[8] = kTestFilterList[filter_bank];
const int kNumFilters = 16;
for (int filter_x = 0; filter_x < kNumFilters; ++filter_x) {
for (int filter_y = 0; filter_y < kNumFilters; ++filter_y) {
filter_average_block2d_8_c(in, kInputStride,
filters[filter_x], filters[filter_y],
ref, kOutputStride,
Width(), Height());
if (filters == vp9_sub_pel_filters_8lp || (filter_x && filter_y))
REGISTER_STATE_CHECK(
UUT_->hv8_avg_(in, kInputStride, out, kOutputStride,
filters[filter_x], 16, filters[filter_y], 16,
Width(), Height()));
else if (filter_y)
REGISTER_STATE_CHECK(
UUT_->v8_avg_(in, kInputStride, out, kOutputStride,
filters[filter_x], 16, filters[filter_y], 16,
Width(), Height()));
else
REGISTER_STATE_CHECK(
UUT_->h8_avg_(in, kInputStride, out, kOutputStride,
filters[filter_x], 16, filters[filter_y], 16,
Width(), Height()));
CheckGuardBlocks();
for (int y = 0; y < Height(); ++y)
for (int x = 0; x < Width(); ++x)
ASSERT_EQ(ref[y * kOutputStride + x], out[y * kOutputStride + x])
<< "mismatch at (" << x << "," << y << "), "
<< "filters (" << filter_bank << ","
<< filter_x << "," << filter_y << ")";
}
}
}
}
DECLARE_ALIGNED(256, const int16_t, kChangeFilters[16][8]) = {
{ 0, 0, 0, 0, 0, 0, 0, 128},
{ 0, 0, 0, 0, 0, 0, 128},
{ 0, 0, 0, 0, 0, 128},
{ 0, 0, 0, 0, 128},
{ 0, 0, 0, 128},
{ 0, 0, 128},
{ 0, 128},
{ 128},
{ 0, 0, 0, 0, 0, 0, 0, 128},
{ 0, 0, 0, 0, 0, 0, 128},
{ 0, 0, 0, 0, 0, 128},
{ 0, 0, 0, 0, 128},
{ 0, 0, 0, 128},
{ 0, 0, 128},
{ 0, 128},
{ 128}
};
TEST_P(ConvolveTest, ChangeFilterWorks) {
uint8_t* const in = input();
uint8_t* const out = output();
REGISTER_STATE_CHECK(UUT_->h8_(in, kInputStride, out, kOutputStride,
kChangeFilters[8], 17, kChangeFilters[4], 16,
Width(), Height()));
for (int x = 0; x < Width(); ++x) {
if (x < 8)
ASSERT_EQ(in[4], out[x]) << "x == " << x;
else
ASSERT_EQ(in[12], out[x]) << "x == " << x;
}
REGISTER_STATE_CHECK(UUT_->v8_(in, kInputStride, out, kOutputStride,
kChangeFilters[4], 16, kChangeFilters[8], 17,
Width(), Height()));
for (int y = 0; y < Height(); ++y) {
if (y < 8)
ASSERT_EQ(in[4 * kInputStride], out[y * kOutputStride]) << "y == " << y;
else
ASSERT_EQ(in[12 * kInputStride], out[y * kOutputStride]) << "y == " << y;
}
REGISTER_STATE_CHECK(UUT_->hv8_(in, kInputStride, out, kOutputStride,
kChangeFilters[8], 17, kChangeFilters[8], 17,
Width(), Height()));
for (int y = 0; y < Height(); ++y) {
for (int x = 0; x < Width(); ++x) {
const int ref_x = x < 8 ? 4 : 12;
const int ref_y = y < 8 ? 4 : 12;
ASSERT_EQ(in[ref_y * kInputStride + ref_x], out[y * kOutputStride + x])
<< "x == " << x << ", y == " << y;
}
}
}
using std::tr1::make_tuple;
const ConvolveFunctions convolve8_c(
vp9_convolve8_horiz_c, vp9_convolve8_avg_horiz_c,
vp9_convolve8_vert_c, vp9_convolve8_avg_vert_c,
vp9_convolve8_c, vp9_convolve8_avg_c);
INSTANTIATE_TEST_CASE_P(C, ConvolveTest, ::testing::Values(
make_tuple(4, 4, &convolve8_c),
make_tuple(8, 4, &convolve8_c),
make_tuple(8, 8, &convolve8_c),
make_tuple(16, 8, &convolve8_c),
make_tuple(16, 16, &convolve8_c)));
}
#if HAVE_SSSE3
const ConvolveFunctions convolve8_ssse3(
vp9_convolve8_horiz_ssse3, vp9_convolve8_avg_horiz_c,
vp9_convolve8_vert_ssse3, vp9_convolve8_avg_vert_c,
vp9_convolve8_ssse3, vp9_convolve8_avg_c);
INSTANTIATE_TEST_CASE_P(SSSE3, ConvolveTest, ::testing::Values(
make_tuple(4, 4, &convolve8_ssse3),
make_tuple(8, 4, &convolve8_ssse3),
make_tuple(8, 8, &convolve8_ssse3),
make_tuple(16, 8, &convolve8_ssse3),
make_tuple(16, 16, &convolve8_ssse3)));
#endif

21
test/cq_test.cc
View File

@ -9,8 +9,12 @@
*/
#include <cmath>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "test/util.h"
namespace {
// CQ level range: [kCQLevelMin, kCQLevelMax).
const int kCQLevelMin = 4;
@ -18,12 +22,13 @@ const int kCQLevelMax = 63;
const int kCQLevelStep = 8;
const int kCQTargetBitrate = 2000;
namespace {
class CQTest : public libvpx_test::EncoderTest,
public ::testing::TestWithParam<int> {
class CQTest : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<int> {
protected:
CQTest() : cq_level_(GetParam()) { init_flags_ = VPX_CODEC_USE_PSNR; }
CQTest() : EncoderTest(GET_PARAM(0)), cq_level_(GET_PARAM(1)) {
init_flags_ = VPX_CODEC_USE_PSNR;
}
virtual ~CQTest() {}
virtual void SetUp() {
@ -100,7 +105,7 @@ TEST_P(CQTest, LinearPSNRIsHigherForCQLevel) {
EXPECT_GE(cq_psnr_lin, vbr_psnr_lin);
}
INSTANTIATE_TEST_CASE_P(CQLevelRange, CQTest,
::testing::Range(kCQLevelMin, kCQLevelMax,
kCQLevelStep));
VP8_INSTANTIATE_TEST_CASE(CQTest,
::testing::Range(kCQLevelMin, kCQLevelMax,
kCQLevelStep));
} // namespace

15
test/datarate_test.cc
View File

@ -7,17 +7,23 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/util.h"
namespace {
class DatarateTest : public ::libvpx_test::EncoderTest,
public ::testing::TestWithParam<enum libvpx_test::TestMode> {
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
public:
DatarateTest() : EncoderTest(GET_PARAM(0)) {}
protected:
virtual void SetUp() {
InitializeConfig();
SetMode(GetParam());
SetMode(GET_PARAM(1));
ResetModel();
}
@ -174,5 +180,6 @@ TEST_P(DatarateTest, ChangingDropFrameThresh) {
}
}
INSTANTIATE_TEST_CASE_P(AllModes, DatarateTest, ALL_TEST_MODES);
VP8_INSTANTIATE_TEST_CASE(DatarateTest, ALL_TEST_MODES);
} // namespace

31
test/dct16x16_test.cc
View File

@ -15,7 +15,7 @@
#include "third_party/googletest/src/include/gtest/gtest.h"
extern "C" {
#include "vp9/common/entropy.h"
#include "vp9/common/vp9_entropy.h"
#include "vp9_rtcd.h"
}
@ -26,6 +26,15 @@ using libvpx_test::ACMRandom;
namespace {
#ifdef _MSC_VER
static int round(double x) {
if (x < 0)
return (int)ceil(x - 0.5);
else
return (int)floor(x + 0.5);
}
#endif
const double PI = 3.1415926535898;
void reference2_16x16_idct_2d(double *input, double *output) {
double x;
@ -278,18 +287,10 @@ TEST(VP9Idct16x16Test, AccuracyCheck) {
<< "Error: 16x16 IDCT has error " << error
<< " at index " << j;
}
vp9_short_fdct16x16_c(in, out_c, 32);
for (int j = 0; j < 256; ++j) {
const double diff = coeff[j] - out_c[j];
const double error = diff * diff;
EXPECT_GE(1.0, error)
<< "Error: 16x16 FDCT has error " << error
<< " at index " << j;
}
}
}
#if 1
// we need enable fdct test once we re-do the 16 point fdct.
TEST(VP9Fdct16x16Test, AccuracyCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
int max_error = 0;
@ -318,10 +319,10 @@ TEST(VP9Fdct16x16Test, AccuracyCheck) {
}
EXPECT_GE(1, max_error)
<< "Error: 16x16 FDCT/IDCT has an individual roundtrip error > 1";
<< "Error: 16x16 FDCT/IDCT has an individual round trip error > 1";
EXPECT_GE(count_test_block/10, total_error)
<< "Error: 16x16 FDCT/IDCT has average roundtrip error > 1/10 per block";
EXPECT_GE(count_test_block , total_error)
<< "Error: 16x16 FDCT/IDCT has average round trip error > 1 per block";
}
TEST(VP9Fdct16x16Test, CoeffSizeCheck) {
@ -353,4 +354,6 @@ TEST(VP9Fdct16x16Test, CoeffSizeCheck) {
}
}
}
#endif
} // namespace

18
test/dct32x32_test.cc
View File

@ -36,7 +36,6 @@ static int round(double x) {
}
#endif
#if !CONFIG_DWTDCTHYBRID
static const double kPi = 3.141592653589793238462643383279502884;
static void reference2_32x32_idct_2d(double *input, double *output) {
double x;
@ -116,20 +115,9 @@ TEST(VP9Idct32x32Test, AccuracyCheck) {
<< "Error: 3x32 IDCT has error " << error
<< " at index " << j;
}
vp9_short_fdct32x32_c(in, out_c, 64);
for (int j = 0; j < 1024; ++j) {
const double diff = coeff[j] - out_c[j];
const double error = diff * diff;
EXPECT_GE(1.0, error)
<< "Error: 32x32 FDCT has error " << error
<< " at index " << j;
}
}
}
#else // CONFIG_DWTDCTHYBRID
// TODO(rbultje/debargha): add DWT-specific tests
#endif // CONFIG_DWTDCTHYBRID
TEST(VP9Fdct32x32Test, AccuracyCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
unsigned int max_error = 0;
@ -160,8 +148,8 @@ TEST(VP9Fdct32x32Test, AccuracyCheck) {
EXPECT_GE(1u, max_error)
<< "Error: 32x32 FDCT/IDCT has an individual roundtrip error > 1";
EXPECT_GE(count_test_block/10, total_error)
<< "Error: 32x32 FDCT/IDCT has average roundtrip error > 1/10 per block";
EXPECT_GE(count_test_block, total_error)
<< "Error: 32x32 FDCT/IDCT has average roundtrip error > 1 per block";
}
TEST(VP9Fdct32x32Test, CoeffSizeCheck) {

17
test/decode_test_driver.cc
View File

@ -7,16 +7,17 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "test/codec_factory.h"
#include "test/decode_test_driver.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/register_state_check.h"
#include "test/video_source.h"
namespace libvpx_test {
#if CONFIG_VP8_DECODER
vpx_codec_err_t Decoder::DecodeFrame(const uint8_t *cxdata, int size) {
vpx_codec_err_t res_dec;
InitOnce();
REGISTER_STATE_CHECK(res_dec = vpx_codec_decode(&decoder_,
cxdata, size, NULL, 0));
return res_dec;
@ -24,21 +25,23 @@ vpx_codec_err_t Decoder::DecodeFrame(const uint8_t *cxdata, int size) {
void DecoderTest::RunLoop(CompressedVideoSource *video) {
vpx_codec_dec_cfg_t dec_cfg = {0};
Decoder decoder(dec_cfg, 0);
Decoder* const decoder = codec_->CreateDecoder(dec_cfg, 0);
ASSERT_TRUE(decoder != NULL);
// Decode frames.
for (video->Begin(); video->cxdata(); video->Next()) {
vpx_codec_err_t res_dec = decoder.DecodeFrame(video->cxdata(),
video->frame_size());
ASSERT_EQ(VPX_CODEC_OK, res_dec) << decoder.DecodeError();
vpx_codec_err_t res_dec = decoder->DecodeFrame(video->cxdata(),
video->frame_size());
ASSERT_EQ(VPX_CODEC_OK, res_dec) << decoder->DecodeError();
DxDataIterator dec_iter = decoder.GetDxData();
DxDataIterator dec_iter = decoder->GetDxData();
const vpx_image_t *img = NULL;
// Get decompressed data
while ((img = dec_iter.Next()))
DecompressedFrameHook(*img, video->frame_number());
}
delete decoder;
}
#endif
} // namespace libvpx_test

33
test/decode_test_driver.h
View File

@ -14,10 +14,10 @@
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "vpx_config.h"
#include "vpx/vpx_decoder.h"
#include "vpx/vp8dx.h"
namespace libvpx_test {
class CodecFactory;
class CompressedVideoSource;
// Provides an object to handle decoding output
@ -42,12 +42,11 @@ class DxDataIterator {
class Decoder {
public:
Decoder(vpx_codec_dec_cfg_t cfg, unsigned long deadline)
: cfg_(cfg), deadline_(deadline) {
: cfg_(cfg), deadline_(deadline), init_done_(false) {
memset(&decoder_, 0, sizeof(decoder_));
Init();
}
~Decoder() {
virtual ~Decoder() {
vpx_codec_destroy(&decoder_);
}
@ -62,37 +61,45 @@ class Decoder {
}
void Control(int ctrl_id, int arg) {
InitOnce();
const vpx_codec_err_t res = vpx_codec_control_(&decoder_, ctrl_id, arg);
ASSERT_EQ(VPX_CODEC_OK, res) << DecodeError();
}
void Control(int ctrl_id, const void *arg) {
InitOnce();
const vpx_codec_err_t res = vpx_codec_control_(&decoder_, ctrl_id, arg);
ASSERT_EQ(VPX_CODEC_OK, res) << DecodeError();
}
const char *DecodeError() {
const char* DecodeError() {
const char *detail = vpx_codec_error_detail(&decoder_);
return detail ? detail : vpx_codec_error(&decoder_);
}
protected:
void Init() {
const vpx_codec_err_t res = vpx_codec_dec_init(&decoder_,
&vpx_codec_vp8_dx_algo,
&cfg_, 0);
ASSERT_EQ(VPX_CODEC_OK, res) << DecodeError();
virtual const vpx_codec_iface_t* CodecInterface() const = 0;
void InitOnce() {
if (!init_done_) {
const vpx_codec_err_t res = vpx_codec_dec_init(&decoder_,
CodecInterface(),
&cfg_, 0);
ASSERT_EQ(VPX_CODEC_OK, res) << DecodeError();
init_done_ = true;
}
}
vpx_codec_ctx_t decoder_;
vpx_codec_dec_cfg_t cfg_;
unsigned int deadline_;
bool init_done_;
};
// Common test functionality for all Decoder tests.
class DecoderTest {
public:
// Main loop.
// Main decoding loop
virtual void RunLoop(CompressedVideoSource *video);
// Hook to be called on every decompressed frame.
@ -100,9 +107,11 @@ class DecoderTest {
const unsigned int frame_number) {}
protected:
DecoderTest() {}
explicit DecoderTest(const CodecFactory *codec) : codec_(codec) {}
virtual ~DecoderTest() {}
const CodecFactory *codec_;
};
} // namespace libvpx_test

72
test/encode_test_driver.cc
View File

@ -7,11 +7,11 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vpx_config.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#if CONFIG_VP8_DECODER
#include "test/decode_test_driver.h"
#endif
#include "test/register_state_check.h"
#include "test/video_source.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
@ -45,7 +45,7 @@ void Encoder::EncodeFrameInternal(const VideoSource &video,
cfg_.g_h = img->d_h;
cfg_.g_timebase = video.timebase();
cfg_.rc_twopass_stats_in = stats_->buf();
res = vpx_codec_enc_init(&encoder_, &vpx_codec_vp8_cx_algo, &cfg_,
res = vpx_codec_enc_init(&encoder_, CodecInterface(), &cfg_,
init_flags_);
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
@ -72,6 +72,11 @@ void Encoder::Flush() {
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
void EncoderTest::InitializeConfig() {
const vpx_codec_err_t res = codec_->DefaultEncoderConfig(&cfg_, 0);
ASSERT_EQ(VPX_CODEC_OK, res);
}
void EncoderTest::SetMode(TestMode mode) {
switch (mode) {
case kRealTime:
@ -125,13 +130,17 @@ static bool compare_img(const vpx_image_t *img1,
return match;
}
void EncoderTest::MismatchHook(const vpx_image_t *img1,
const vpx_image_t *img2) {
ASSERT_TRUE(0) << "Encode/Decode mismatch found";
}
void EncoderTest::RunLoop(VideoSource *video) {
#if CONFIG_VP8_DECODER
vpx_codec_dec_cfg_t dec_cfg = {0};
#endif
stats_.Reset();
ASSERT_TRUE(passes_ == 1 || passes_ == 2);
for (unsigned int pass = 0; pass < passes_; pass++) {
last_pts_ = 0;
@ -143,34 +152,34 @@ void EncoderTest::RunLoop(VideoSource *video) {
cfg_.g_pass = VPX_RC_LAST_PASS;
BeginPassHook(pass);
Encoder encoder(cfg_, deadline_, init_flags_, &stats_);
#if CONFIG_VP8_DECODER
Decoder decoder(dec_cfg, 0);
bool has_cxdata = false;
#endif
Encoder* const encoder = codec_->CreateEncoder(cfg_, deadline_, init_flags_,
&stats_);
ASSERT_TRUE(encoder != NULL);
Decoder* const decoder = codec_->CreateDecoder(dec_cfg, 0);
bool again;
for (again = true, video->Begin(); again; video->Next()) {
again = video->img() != NULL;
PreEncodeFrameHook(video);
PreEncodeFrameHook(video, &encoder);
encoder.EncodeFrame(video, frame_flags_);
PreEncodeFrameHook(video, encoder);
encoder->EncodeFrame(video, frame_flags_);
CxDataIterator iter = encoder.GetCxData();
CxDataIterator iter = encoder->GetCxData();
bool has_cxdata = false;
bool has_dxdata = false;
while (const vpx_codec_cx_pkt_t *pkt = iter.Next()) {
pkt = MutateEncoderOutputHook(pkt);
again = true;
#if CONFIG_VP8_DECODER
vpx_codec_err_t res_dec;
#endif
switch (pkt->kind) {
case VPX_CODEC_CX_FRAME_PKT:
#if CONFIG_VP8_DECODER
has_cxdata = true;
res_dec = decoder.DecodeFrame((const uint8_t*)pkt->data.frame.buf,
pkt->data.frame.sz);
ASSERT_EQ(VPX_CODEC_OK, res_dec) << decoder.DecodeError();
#endif
if (decoder && DoDecode()) {
vpx_codec_err_t res_dec = decoder->DecodeFrame(
(const uint8_t*)pkt->data.frame.buf, pkt->data.frame.sz);
ASSERT_EQ(VPX_CODEC_OK, res_dec) << decoder->DecodeError();
has_dxdata = true;
}
ASSERT_GE(pkt->data.frame.pts, last_pts_);
last_pts_ = pkt->data.frame.pts;
FramePktHook(pkt);
@ -185,25 +194,32 @@ void EncoderTest::RunLoop(VideoSource *video) {
}
}
#if CONFIG_VP8_DECODER
if (has_cxdata) {
const vpx_image_t *img_enc = encoder.GetPreviewFrame();
DxDataIterator dec_iter = decoder.GetDxData();
if (has_dxdata && has_cxdata) {
const vpx_image_t *img_enc = encoder->GetPreviewFrame();
DxDataIterator dec_iter = decoder->GetDxData();
const vpx_image_t *img_dec = dec_iter.Next();
if(img_enc && img_dec) {
if (img_enc && img_dec) {
const bool res = compare_img(img_enc, img_dec);
ASSERT_TRUE(res)<< "Encoder/Decoder mismatch found.";
if (!res) { // Mismatch
MismatchHook(img_enc, img_dec);
}
}
if (img_dec)
DecompressedFrameHook(*img_dec, video->pts());
}
#endif
if (!Continue())
break;
}
EndPassHook();
if (decoder)
delete decoder;
delete encoder;
if (!Continue())
break;
}
}
} // namespace libvpx_test

47
test/encode_test_driver.h
View File

@ -9,14 +9,16 @@
*/
#ifndef TEST_ENCODE_TEST_DRIVER_H_
#define TEST_ENCODE_TEST_DRIVER_H_
#include "./vpx_config.h"
#include <string>
#include <vector>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "vpx/vpx_encoder.h"
#include "vpx/vp8cx.h"
namespace libvpx_test {
class CodecFactory;
class VideoSource;
enum TestMode {
@ -36,6 +38,9 @@ enum TestMode {
::libvpx_test::kOnePassGood, \
::libvpx_test::kOnePassBest)
#define TWO_PASS_TEST_MODES ::testing::Values(::libvpx_test::kTwoPassGood, \
::libvpx_test::kTwoPassBest)
// Provides an object to handle the libvpx get_cx_data() iteration pattern
class CxDataIterator {
@ -83,7 +88,7 @@ class Encoder {
public:
Encoder(vpx_codec_enc_cfg_t cfg, unsigned long deadline,
const unsigned long init_flags, TwopassStatsStore *stats)
: cfg_(cfg), deadline_(deadline), init_flags_(init_flags), stats_(stats) {
: cfg_(cfg), deadline_(deadline), init_flags_(init_flags), stats_(stats) {
memset(&encoder_, 0, sizeof(encoder_));
}
@ -112,11 +117,18 @@ class Encoder {
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
void Control(int ctrl_id, struct vpx_scaling_mode *arg) {
const vpx_codec_err_t res = vpx_codec_control_(&encoder_, ctrl_id, arg);
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
void set_deadline(unsigned long deadline) {
deadline_ = deadline;
}
protected:
virtual const vpx_codec_iface_t* CodecInterface() const = 0;
const char *EncoderError() {
const char *detail = vpx_codec_error_detail(&encoder_);
return detail ? detail : vpx_codec_error(&encoder_);
@ -145,22 +157,19 @@ class Encoder {
// classes directly, so that tests can be parameterized differently.
class EncoderTest {
protected:
EncoderTest() : abort_(false), init_flags_(0), frame_flags_(0),
last_pts_(0) {}
explicit EncoderTest(const CodecFactory *codec)
: codec_(codec), abort_(false), init_flags_(0), frame_flags_(0),
last_pts_(0) {}
virtual ~EncoderTest() {}
// Initialize the cfg_ member with the default configuration.
void InitializeConfig() {
const vpx_codec_err_t res = vpx_codec_enc_config_default(
&vpx_codec_vp8_cx_algo, &cfg_, 0);
ASSERT_EQ(VPX_CODEC_OK, res);
}
void InitializeConfig();
// Map the TestMode enum to the deadline_ and passes_ variables.
void SetMode(TestMode mode);
// Main loop.
// Main loop
virtual void RunLoop(VideoSource *video);
// Hook to be called at the beginning of a pass.
@ -182,6 +191,24 @@ class EncoderTest {
// Hook to determine whether the encode loop should continue.
virtual bool Continue() const { return !abort_; }
const CodecFactory *codec_;
// Hook to determine whether to decode frame after encoding
virtual bool DoDecode() const { return 1; }
// Hook to handle encode/decode mismatch
virtual void MismatchHook(const vpx_image_t *img1,
const vpx_image_t *img2);
// Hook to be called on every decompressed frame.
virtual void DecompressedFrameHook(const vpx_image_t& img,
vpx_codec_pts_t pts) {}
// Hook that can modify the encoder's output data
virtual const vpx_codec_cx_pkt_t * MutateEncoderOutputHook(
const vpx_codec_cx_pkt_t *pkt) {
return pkt;
}
bool abort_;
vpx_codec_enc_cfg_t cfg_;
unsigned int passes_;

153
test/error_resilience_test.cc
View File

@ -7,22 +7,37 @@
in the file PATENTS. All contributing project authors may
be found in the AUTHORS file in the root of the source tree.
*/
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "test/util.h"
namespace {
class ErrorResilienceTest : public libvpx_test::EncoderTest,
public ::testing::TestWithParam<int> {
const int kMaxErrorFrames = 8;
const int kMaxDroppableFrames = 8;
class ErrorResilienceTest : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
ErrorResilienceTest() {
psnr_ = 0.0;
nframes_ = 0;
encoding_mode_ = static_cast<libvpx_test::TestMode>(GetParam());
ErrorResilienceTest() : EncoderTest(GET_PARAM(0)),
psnr_(0.0),
nframes_(0),
mismatch_psnr_(0.0),
mismatch_nframes_(0),
encoding_mode_(GET_PARAM(1)) {
Reset();
}
virtual ~ErrorResilienceTest() {}
void Reset() {
error_nframes_ = 0;
droppable_nframes_ = 0;
}
virtual void SetUp() {
InitializeConfig();
SetMode(encoding_mode_);
@ -31,6 +46,8 @@ class ErrorResilienceTest : public libvpx_test::EncoderTest,
virtual void BeginPassHook(unsigned int /*pass*/) {
psnr_ = 0.0;
nframes_ = 0;
mismatch_psnr_ = 0.0;
mismatch_nframes_ = 0;
}
virtual bool Continue() const {
@ -42,15 +59,92 @@ class ErrorResilienceTest : public libvpx_test::EncoderTest,
nframes_++;
}
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video) {
frame_flags_ &= ~(VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF);
if (droppable_nframes_ > 0 &&
(cfg_.g_pass == VPX_RC_LAST_PASS || cfg_.g_pass == VPX_RC_ONE_PASS)) {
for (unsigned int i = 0; i < droppable_nframes_; ++i) {
if (droppable_frames_[i] == nframes_) {
std::cout << " Encoding droppable frame: "
<< droppable_frames_[i] << "\n";
frame_flags_ |= (VP8_EFLAG_NO_UPD_LAST |
VP8_EFLAG_NO_UPD_GF |
VP8_EFLAG_NO_UPD_ARF);
return;
}
}
}
}
double GetAveragePsnr() const {
if (nframes_)
return psnr_ / nframes_;
return 0.0;
}
double GetAverageMismatchPsnr() const {
if (mismatch_nframes_)
return mismatch_psnr_ / mismatch_nframes_;
return 0.0;
}
virtual bool DoDecode() const {
if (error_nframes_ > 0 &&
(cfg_.g_pass == VPX_RC_LAST_PASS || cfg_.g_pass == VPX_RC_ONE_PASS)) {
for (unsigned int i = 0; i < error_nframes_; ++i) {
if (error_frames_[i] == nframes_ - 1) {
std::cout << " Skipping decoding frame: "
<< error_frames_[i] << "\n";
return 0;
}
}
}
return 1;
}
virtual void MismatchHook(const vpx_image_t *img1,
const vpx_image_t *img2) {
double mismatch_psnr = compute_psnr(img1, img2);
mismatch_psnr_ += mismatch_psnr;
++mismatch_nframes_;
// std::cout << "Mismatch frame psnr: " << mismatch_psnr << "\n";
}
void SetErrorFrames(int num, unsigned int *list) {
if (num > kMaxErrorFrames)
num = kMaxErrorFrames;
else if (num < 0)
num = 0;
error_nframes_ = num;
for (unsigned int i = 0; i < error_nframes_; ++i)
error_frames_[i] = list[i];
}
void SetDroppableFrames(int num, unsigned int *list) {
if (num > kMaxDroppableFrames)
num = kMaxDroppableFrames;
else if (num < 0)
num = 0;
droppable_nframes_ = num;
for (unsigned int i = 0; i < droppable_nframes_; ++i)
droppable_frames_[i] = list[i];
}
unsigned int GetMismatchFrames() {
return mismatch_nframes_;
}
private:
double psnr_;
unsigned int nframes_;
unsigned int error_nframes_;
unsigned int droppable_nframes_;
double mismatch_psnr_;
unsigned int mismatch_nframes_;
unsigned int error_frames_[kMaxErrorFrames];
unsigned int droppable_frames_[kMaxDroppableFrames];
libvpx_test::TestMode encoding_mode_;
};
@ -85,6 +179,49 @@ TEST_P(ErrorResilienceTest, OnVersusOff) {
}
}
INSTANTIATE_TEST_CASE_P(OnOffTest, ErrorResilienceTest,
ONE_PASS_TEST_MODES);
TEST_P(ErrorResilienceTest, DropFramesWithoutRecovery) {
const vpx_rational timebase = { 33333333, 1000000000 };
cfg_.g_timebase = timebase;
cfg_.rc_target_bitrate = 500;
init_flags_ = VPX_CODEC_USE_PSNR;
libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
timebase.den, timebase.num, 0, 30);
// Error resilient mode ON.
cfg_.g_error_resilient = 1;
// Set an arbitrary set of error frames same as droppable frames
unsigned int num_droppable_frames = 2;
unsigned int droppable_frame_list[] = {5, 16};
SetDroppableFrames(num_droppable_frames, droppable_frame_list);
SetErrorFrames(num_droppable_frames, droppable_frame_list);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
// Test that no mismatches have been found
std::cout << " Mismatch frames: "
<< GetMismatchFrames() << "\n";
EXPECT_EQ(GetMismatchFrames(), (unsigned int) 0);
// reset previously set error/droppable frames
Reset();
// Now set an arbitrary set of error frames that are non-droppable
unsigned int num_error_frames = 3;
unsigned int error_frame_list[] = {3, 10, 20};
SetErrorFrames(num_error_frames, error_frame_list);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
// Test that dropping an arbitrary set of inter frames does not hurt too much
// Note the Average Mismatch PSNR is the average of the PSNR between
// decoded frame and encoder's version of the same frame for all frames
// with mismatch.
const double psnr_resilience_mismatch = GetAverageMismatchPsnr();
std::cout << " Mismatch PSNR: "
<< psnr_resilience_mismatch << "\n";
EXPECT_GT(psnr_resilience_mismatch, 20.0);
}
VP8_INSTANTIATE_TEST_CASE(ErrorResilienceTest, ONE_PASS_TEST_MODES);
VP9_INSTANTIATE_TEST_CASE(ErrorResilienceTest, ONE_PASS_TEST_MODES);
} // namespace

6
test/fdct4x4_test.cc
View File

@ -25,7 +25,7 @@ using libvpx_test::ACMRandom;
namespace {
TEST(Vp9FdctTest, SignBiasCheck) {
TEST(Vp9Fdct4x4Test, SignBiasCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
int16_t test_input_block[16];
int16_t test_output_block[16];
@ -88,7 +88,7 @@ TEST(Vp9FdctTest, SignBiasCheck) {
}
};
TEST(Vp9FdctTest, RoundTripErrorCheck) {
TEST(Vp9Fdct4x4Test, RoundTripErrorCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
int max_error = 0;
double total_error = 0;
@ -120,7 +120,7 @@ TEST(Vp9FdctTest, RoundTripErrorCheck) {
}
// Because the bitstream is not frozen yet, use the idct in the codebase.
vp9_short_idct4x4llm_c(test_temp_block, test_output_block, pitch);
vp9_short_idct4x4_c(test_temp_block, test_output_block, pitch);
for (int j = 0; j < 16; ++j) {
const int diff = test_input_block[j] - test_output_block[j];

2
test/fdct8x8_test.cc
View File

@ -149,7 +149,7 @@ TEST(VP9Fdct8x8Test, ExtremalCheck) {
// Initialize a test block with input range {-255, 255}.
for (int j = 0; j < 64; ++j)
test_input_block[j] = rnd.Rand8() % 2 ? 255 : -255;
test_input_block[j] = rnd.Rand8() % 2 ? 255 : -256;
const int pitch = 16;
vp9_short_fdct8x8_c(test_input_block, test_temp_block, pitch);

25
test/idct8x8_test.cc
View File

@ -120,31 +120,6 @@ TEST(VP9Idct8x8Test, AccuracyCheck) {
input[j] = rnd.Rand8() - rnd.Rand8();
const int pitch = 16;
vp9_short_fdct8x8_c(input, output_c, pitch);
reference_dct_2d(input, output_r);
for (int j = 0; j < 64; ++j) {
const double diff = output_c[j] - output_r[j];
const double error = diff * diff;
// An error in a DCT coefficient isn't that bad.
// We care more about the reconstructed pixels.
EXPECT_GE(2.0, error)
<< "Error: 8x8 FDCT/IDCT has error " << error
<< " at index " << j;
}
#if 0
// Tests that the reference iDCT and fDCT match.
reference_dct_2d(input, output_r);
reference_idct_2d(output_r, output_c);
for (int j = 0; j < 64; ++j) {
const int diff = output_c[j] -input[j];
const int error = diff * diff;
EXPECT_EQ(0, error)
<< "Error: 8x8 FDCT/IDCT has error " << error
<< " at index " << j;
}
#endif
reference_dct_2d(input, output_r);
for (int j = 0; j < 64; ++j)
coeff[j] = round(output_r[j]);

60
test/idctllm_test.cc → test/idct_test.cc
View File

@ -10,8 +10,8 @@
extern "C" {
#include "vpx_config.h"
#include "vp8_rtcd.h"
#include "./vpx_config.h"
#include "./vp8_rtcd.h"
}
#include "test/register_state_check.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
@ -20,18 +20,16 @@ typedef void (*idct_fn_t)(short *input, unsigned char *pred_ptr,
int pred_stride, unsigned char *dst_ptr,
int dst_stride);
namespace {
class IDCTTest : public ::testing::TestWithParam<idct_fn_t>
{
class IDCTTest : public ::testing::TestWithParam<idct_fn_t> {
protected:
virtual void SetUp()
{
virtual void SetUp() {
int i;
UUT = GetParam();
memset(input, 0, sizeof(input));
/* Set up guard blocks */
for(i=0; i<256; i++)
output[i] = ((i&0xF)<4&&(i<64))?0:-1;
for (i = 0; i < 256; i++)
output[i] = ((i & 0xF) < 4 && (i < 64)) ? 0 : -1;
}
idct_fn_t UUT;
@ -40,78 +38,72 @@ class IDCTTest : public ::testing::TestWithParam<idct_fn_t>
unsigned char predict[256];
};
TEST_P(IDCTTest, TestGuardBlocks)
{
TEST_P(IDCTTest, TestGuardBlocks) {
int i;
for(i=0; i<256; i++)
if((i&0xF) < 4 && i<64)
for (i = 0; i < 256; i++)
if ((i & 0xF) < 4 && i < 64)
EXPECT_EQ(0, output[i]) << i;
else
EXPECT_EQ(255, output[i]);
}
TEST_P(IDCTTest, TestAllZeros)
{
TEST_P(IDCTTest, TestAllZeros) {
int i;
REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
for(i=0; i<256; i++)
if((i&0xF) < 4 && i<64)
for (i = 0; i < 256; i++)
if ((i & 0xF) < 4 && i < 64)
EXPECT_EQ(0, output[i]) << "i==" << i;
else
EXPECT_EQ(255, output[i]) << "i==" << i;
}
TEST_P(IDCTTest, TestAllOnes)
{
TEST_P(IDCTTest, TestAllOnes) {
int i;
input[0] = 4;
REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
for(i=0; i<256; i++)
if((i&0xF) < 4 && i<64)
for (i = 0; i < 256; i++)
if ((i & 0xF) < 4 && i < 64)
EXPECT_EQ(1, output[i]) << "i==" << i;
else
EXPECT_EQ(255, output[i]) << "i==" << i;
}
TEST_P(IDCTTest, TestAddOne)
{
TEST_P(IDCTTest, TestAddOne) {
int i;
for(i=0; i<256; i++)
for (i = 0; i < 256; i++)
predict[i] = i;
input[0] = 4;
REGISTER_STATE_CHECK(UUT(input, predict, 16, output, 16));
for(i=0; i<256; i++)
if((i&0xF) < 4 && i<64)
for (i = 0; i < 256; i++)
if ((i & 0xF) < 4 && i < 64)
EXPECT_EQ(i+1, output[i]) << "i==" << i;
else
EXPECT_EQ(255, output[i]) << "i==" << i;
}
TEST_P(IDCTTest, TestWithData)
{
TEST_P(IDCTTest, TestWithData) {
int i;
for(i=0; i<16; i++)
for (i = 0; i < 16; i++)
input[i] = i;
REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
for(i=0; i<256; i++)
if((i&0xF) > 3 || i>63)
for (i = 0; i < 256; i++)
if ((i & 0xF) > 3 || i > 63)
EXPECT_EQ(255, output[i]) << "i==" << i;
else if(i == 0)
else if (i == 0)
EXPECT_EQ(11, output[i]) << "i==" << i;
else if(i == 34)
else if (i == 34)
EXPECT_EQ(1, output[i]) << "i==" << i;
else if(i == 2 || i == 17 || i == 32)
else if (i == 2 || i == 17 || i == 32)
EXPECT_EQ(3, output[i]) << "i==" << i;
else
EXPECT_EQ(0, output[i]) << "i==" << i;

16
test/keyframe_test.cc
View File

@ -9,18 +9,22 @@
*/
#include <climits>
#include <vector>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/util.h"
namespace {
class KeyframeTest : public ::libvpx_test::EncoderTest,
public ::testing::TestWithParam<enum libvpx_test::TestMode> {
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
KeyframeTest() : EncoderTest(GET_PARAM(0)) {}
virtual void SetUp() {
InitializeConfig();
SetMode(GetParam());
SetMode(GET_PARAM(1));
kf_count_ = 0;
kf_count_max_ = INT_MAX;
kf_do_force_kf_ = false;
@ -64,7 +68,7 @@ TEST_P(KeyframeTest, TestRandomVideoSource) {
// In realtime mode - auto placed keyframes are exceedingly rare, don't
// bother with this check if(GetParam() > 0)
if(GetParam() > 0)
if (GET_PARAM(1) > 0)
EXPECT_GT(kf_count_, 1);
}
@ -126,7 +130,7 @@ TEST_P(KeyframeTest, TestAutoKeyframe) {
// In realtime mode - auto placed keyframes are exceedingly rare, don't
// bother with this check
if(GetParam() > 0)
if (GET_PARAM(1) > 0)
EXPECT_EQ(2u, kf_pts_list_.size()) << " Not the right number of keyframes ";
// Verify that keyframes match the file keyframes in the file.
@ -141,5 +145,5 @@ TEST_P(KeyframeTest, TestAutoKeyframe) {
}
}
INSTANTIATE_TEST_CASE_P(AllModes, KeyframeTest, ALL_TEST_MODES);
VP8_INSTANTIATE_TEST_CASE(KeyframeTest, ALL_TEST_MODES);
} // namespace

64
test/md5_helper.h Normal file
View File

@ -0,0 +1,64 @@
/*
* Copyright (c) 2012 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef LIBVPX_TEST_MD5_HELPER_H_
#define LIBVPX_TEST_MD5_HELPER_H_
extern "C" {
#include "./md5_utils.h"
#include "vpx/vpx_decoder.h"
}
namespace libvpx_test {
class MD5 {
public:
MD5() {
MD5Init(&md5_);
}
void Add(const vpx_image_t *img) {
for (int plane = 0; plane < 3; ++plane) {
uint8_t *buf = img->planes[plane];
const int h = plane ? (img->d_h + 1) >> 1 : img->d_h;
const int w = plane ? (img->d_w + 1) >> 1 : img->d_w;
for (int y = 0; y < h; ++y) {
MD5Update(&md5_, buf, w);
buf += img->stride[plane];
}
}
}
const char *Get(void) {
static const char hex[16] = {
'0', '1', '2', '3', '4', '5', '6', '7',
'8', '9', 'a', 'b', 'c', 'd', 'e', 'f',
};
uint8_t tmp[16];
MD5Context ctx_tmp = md5_;
MD5Final(tmp, &ctx_tmp);
for (int i = 0; i < 16; i++) {
res_[i * 2 + 0] = hex[tmp[i] >> 4];
res_[i * 2 + 1] = hex[tmp[i] & 0xf];
}
res_[32] = 0;
return res_;
}
protected:
char res_[33];
MD5Context md5_;
};
} // namespace libvpx_test
#endif // LIBVPX_TEST_MD5_HELPER_H_

75
test/resize_test.cc
View File

@ -9,9 +9,12 @@
*/
#include <climits>
#include <vector>
#include "test/encode_test_driver.h"
#include "test/video_source.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "test/video_source.h"
#include "test/util.h"
namespace {
@ -49,8 +52,10 @@ class ResizingVideoSource : public ::libvpx_test::DummyVideoSource {
};
class ResizeTest : public ::libvpx_test::EncoderTest,
public ::testing::TestWithParam<enum libvpx_test::TestMode> {
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
ResizeTest() : EncoderTest(GET_PARAM(0)) {}
struct FrameInfo {
FrameInfo(vpx_codec_pts_t _pts, unsigned int _w, unsigned int _h)
: pts(_pts), w(_w), h(_h) {}
@ -62,22 +67,16 @@ class ResizeTest : public ::libvpx_test::EncoderTest,
virtual void SetUp() {
InitializeConfig();
SetMode(GetParam());
SetMode(GET_PARAM(1));
}
virtual bool Continue() const {
return !HasFatalFailure() && !abort_;
}
virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
if (pkt->data.frame.flags & VPX_FRAME_IS_KEY) {
const unsigned char *buf =
reinterpret_cast<const unsigned char *>(pkt->data.frame.buf);
const unsigned int w = (buf[6] | (buf[7] << 8)) & 0x3fff;
const unsigned int h = (buf[8] | (buf[9] << 8)) & 0x3fff;
frame_info_list_.push_back(FrameInfo(pkt->data.frame.pts, w, h));
}
virtual void DecompressedFrameHook(const vpx_image_t &img,
vpx_codec_pts_t pts) {
frame_info_list_.push_back(FrameInfo(pts, img.d_w, img.d_h));
}
std::vector< FrameInfo > frame_info_list_;
@ -100,5 +99,53 @@ TEST_P(ResizeTest, TestExternalResizeWorks) {
}
}
INSTANTIATE_TEST_CASE_P(OnePass, ResizeTest, ONE_PASS_TEST_MODES);
class ResizeInternalTest : public ResizeTest {
protected:
ResizeInternalTest() : ResizeTest(), frame0_psnr_(0.0) {}
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
libvpx_test::Encoder *encoder) {
if (video->frame() == 3) {
struct vpx_scaling_mode mode = {VP8E_FOURFIVE, VP8E_THREEFIVE};
encoder->Control(VP8E_SET_SCALEMODE, &mode);
}
if (video->frame() == 6) {
struct vpx_scaling_mode mode = {VP8E_NORMAL, VP8E_NORMAL};
encoder->Control(VP8E_SET_SCALEMODE, &mode);
}
}
virtual void PSNRPktHook(const vpx_codec_cx_pkt_t *pkt) {
if (!frame0_psnr_)
frame0_psnr_ = pkt->data.psnr.psnr[0];
EXPECT_NEAR(pkt->data.psnr.psnr[0], frame0_psnr_, 1.0);
}
double frame0_psnr_;
};
TEST_P(ResizeInternalTest, TestInternalResizeWorks) {
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 10);
init_flags_ = VPX_CODEC_USE_PSNR;
// q picked such that initial keyframe on this clip is ~30dB PSNR
cfg_.rc_min_quantizer = cfg_.rc_max_quantizer = 48;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
for (std::vector<FrameInfo>::iterator info = frame_info_list_.begin();
info != frame_info_list_.end(); ++info) {
const vpx_codec_pts_t pts = info->pts;
if (pts >= 3 && pts < 6) {
ASSERT_EQ(282U, info->w) << "Frame " << pts << " had unexpected width";
ASSERT_EQ(173U, info->h) << "Frame " << pts << " had unexpected height";
} else {
EXPECT_EQ(352U, info->w) << "Frame " << pts << " had unexpected width";
EXPECT_EQ(288U, info->h) << "Frame " << pts << " had unexpected height";
}
}
}
VP8_INSTANTIATE_TEST_CASE(ResizeTest, ONE_PASS_TEST_MODES);
VP9_INSTANTIATE_TEST_CASE(ResizeInternalTest,
::testing::Values(::libvpx_test::kOnePassBest));
} // namespace

304
test/sad_test.cc
View File

@ -15,8 +15,13 @@
extern "C" {
#include "./vpx_config.h"
#if CONFIG_VP8_ENCODER
#include "./vp8_rtcd.h"
#include "vp8/common/blockd.h"
//#include "vp8/common/blockd.h"
#endif
#if CONFIG_VP9_ENCODER
#include "./vp9_rtcd.h"
#endif
#include "vpx_mem/vpx_mem.h"
}
@ -32,14 +37,22 @@ typedef unsigned int (*sad_m_by_n_fn_t)(const unsigned char *source_ptr,
int reference_stride,
unsigned int max_sad);
typedef void (*sad_n_by_n_by_4_fn_t)(const uint8_t *src_ptr,
int src_stride,
const unsigned char * const ref_ptr[],
int ref_stride,
unsigned int *sad_array);
using libvpx_test::ACMRandom;
namespace {
class SADTest : public PARAMS(int, int, sad_m_by_n_fn_t) {
class SADTestBase : public ::testing::Test {
public:
SADTestBase(int width, int height) : width_(width), height_(height) {}
static void SetUpTestCase() {
source_data_ = reinterpret_cast<uint8_t*>(
vpx_memalign(kDataAlignment, kDataBufferSize));
vpx_memalign(kDataAlignment, kDataBlockSize));
reference_data_ = reinterpret_cast<uint8_t*>(
vpx_memalign(kDataAlignment, kDataBufferSize));
}
@ -52,36 +65,31 @@ class SADTest : public PARAMS(int, int, sad_m_by_n_fn_t) {
}
protected:
// Handle blocks up to 4 blocks 64x64 with stride up to 128
static const int kDataAlignment = 16;
static const int kDataBufferSize = 16 * 32;
static const int kDataBlockSize = 64 * 128;
static const int kDataBufferSize = 4 * kDataBlockSize;
virtual void SetUp() {
sad_fn_ = GET_PARAM(2);
height_ = GET_PARAM(1);
width_ = GET_PARAM(0);
source_stride_ = width_ * 2;
source_stride_ = (width_ + 31) & ~31;
reference_stride_ = width_ * 2;
rnd_.Reset(ACMRandom::DeterministicSeed());
}
sad_m_by_n_fn_t sad_fn_;
virtual unsigned int SAD(unsigned int max_sad) {
unsigned int ret;
REGISTER_STATE_CHECK(ret = sad_fn_(source_data_, source_stride_,
reference_data_, reference_stride_,
max_sad));
return ret;
virtual uint8_t* GetReference(int block_idx) {
return reference_data_ + block_idx * kDataBlockSize;
}
// Sum of Absolute Differences. Given two blocks, calculate the absolute
// difference between two pixels in the same relative location; accumulate.
unsigned int ReferenceSAD(unsigned int max_sad) {
unsigned int ReferenceSAD(unsigned int max_sad, int block_idx = 0) {
unsigned int sad = 0;
const uint8_t* const reference = GetReference(block_idx);
for (int h = 0; h < height_; ++h) {
for (int w = 0; w < width_; ++w) {
sad += abs(source_data_[h * source_stride_ + w]
- reference_data_[h * reference_stride_ + w]);
- reference[h * reference_stride_ + w]);
}
if (sad > max_sad) {
break;
@ -106,6 +114,32 @@ class SADTest : public PARAMS(int, int, sad_m_by_n_fn_t) {
}
}
int width_, height_;
static uint8_t* source_data_;
int source_stride_;
static uint8_t* reference_data_;
int reference_stride_;
ACMRandom rnd_;
};
class SADTest : public SADTestBase,
public ::testing::WithParamInterface<
std::tr1::tuple<int, int, sad_m_by_n_fn_t> > {
public:
SADTest() : SADTestBase(GET_PARAM(0), GET_PARAM(1)) {}
protected:
unsigned int SAD(unsigned int max_sad, int block_idx = 0) {
unsigned int ret;
const uint8_t* const reference = GetReference(block_idx);
REGISTER_STATE_CHECK(ret = GET_PARAM(2)(source_data_, source_stride_,
reference, reference_stride_,
max_sad));
return ret;
}
void CheckSad(unsigned int max_sad) {
unsigned int reference_sad, exp_sad;
@ -119,19 +153,38 @@ class SADTest : public PARAMS(int, int, sad_m_by_n_fn_t) {
ASSERT_GE(exp_sad, reference_sad);
}
}
// Handle blocks up to 16x16 with stride up to 32
int height_, width_;
static uint8_t* source_data_;
int source_stride_;
static uint8_t* reference_data_;
int reference_stride_;
ACMRandom rnd_;
};
uint8_t* SADTest::source_data_ = NULL;
uint8_t* SADTest::reference_data_ = NULL;
class SADx4Test : public SADTestBase,
public ::testing::WithParamInterface<
std::tr1::tuple<int, int, sad_n_by_n_by_4_fn_t> > {
public:
SADx4Test() : SADTestBase(GET_PARAM(0), GET_PARAM(1)) {}
protected:
void SADs(unsigned int *results) {
const uint8_t* refs[] = {GetReference(0), GetReference(1),
GetReference(2), GetReference(3)};
REGISTER_STATE_CHECK(GET_PARAM(2)(source_data_, source_stride_,
refs, reference_stride_,
results));
}
void CheckSADs() {
unsigned int reference_sad, exp_sad[4];
SADs(exp_sad);
for (int block = 0; block < 4; block++) {
reference_sad = ReferenceSAD(UINT_MAX, block);
EXPECT_EQ(exp_sad[block], reference_sad) << "block " << block;
}
}
};
uint8_t* SADTestBase::source_data_ = NULL;
uint8_t* SADTestBase::reference_data_ = NULL;
TEST_P(SADTest, MaxRef) {
FillConstant(source_data_, source_stride_, 0);
@ -139,12 +192,30 @@ TEST_P(SADTest, MaxRef) {
CheckSad(UINT_MAX);
}
TEST_P(SADx4Test, MaxRef) {
FillConstant(source_data_, source_stride_, 0);
FillConstant(GetReference(0), reference_stride_, 255);
FillConstant(GetReference(1), reference_stride_, 255);
FillConstant(GetReference(2), reference_stride_, 255);
FillConstant(GetReference(3), reference_stride_, 255);
CheckSADs();
}
TEST_P(SADTest, MaxSrc) {
FillConstant(source_data_, source_stride_, 255);
FillConstant(reference_data_, reference_stride_, 0);
CheckSad(UINT_MAX);
}
TEST_P(SADx4Test, MaxSrc) {
FillConstant(source_data_, source_stride_, 255);
FillConstant(GetReference(0), reference_stride_, 0);
FillConstant(GetReference(1), reference_stride_, 0);
FillConstant(GetReference(2), reference_stride_, 0);
FillConstant(GetReference(3), reference_stride_, 0);
CheckSADs();
}
TEST_P(SADTest, ShortRef) {
int tmp_stride = reference_stride_;
reference_stride_ >>= 1;
@ -154,6 +225,18 @@ TEST_P(SADTest, ShortRef) {
reference_stride_ = tmp_stride;
}
TEST_P(SADx4Test, ShortRef) {
int tmp_stride = reference_stride_;
reference_stride_ >>= 1;
FillRandom(source_data_, source_stride_);
FillRandom(GetReference(0), reference_stride_);
FillRandom(GetReference(1), reference_stride_);
FillRandom(GetReference(2), reference_stride_);
FillRandom(GetReference(3), reference_stride_);
CheckSADs();
reference_stride_ = tmp_stride;
}
TEST_P(SADTest, UnalignedRef) {
// The reference frame, but not the source frame, may be unaligned for
// certain types of searches.
@ -165,6 +248,20 @@ TEST_P(SADTest, UnalignedRef) {
reference_stride_ = tmp_stride;
}
TEST_P(SADx4Test, UnalignedRef) {
// The reference frame, but not the source frame, may be unaligned for
// certain types of searches.
int tmp_stride = reference_stride_;
reference_stride_ -= 1;
FillRandom(source_data_, source_stride_);
FillRandom(GetReference(0), reference_stride_);
FillRandom(GetReference(1), reference_stride_);
FillRandom(GetReference(2), reference_stride_);
FillRandom(GetReference(3), reference_stride_);
CheckSADs();
reference_stride_ = tmp_stride;
}
TEST_P(SADTest, ShortSrc) {
int tmp_stride = source_stride_;
source_stride_ >>= 1;
@ -174,6 +271,18 @@ TEST_P(SADTest, ShortSrc) {
source_stride_ = tmp_stride;
}
TEST_P(SADx4Test, ShortSrc) {
int tmp_stride = source_stride_;
source_stride_ >>= 1;
FillRandom(source_data_, source_stride_);
FillRandom(GetReference(0), reference_stride_);
FillRandom(GetReference(1), reference_stride_);
FillRandom(GetReference(2), reference_stride_);
FillRandom(GetReference(3), reference_stride_);
CheckSADs();
source_stride_ = tmp_stride;
}
TEST_P(SADTest, MaxSAD) {
// Verify that, when max_sad is set, the implementation does not return a
// value lower than the reference.
@ -184,17 +293,61 @@ TEST_P(SADTest, MaxSAD) {
using std::tr1::make_tuple;
#if CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
#define VP8_VP9_SEPARATOR ,
#else
#define VP8_VP9_SEPARATOR
#endif
#if CONFIG_VP8_ENCODER
const sad_m_by_n_fn_t sad_16x16_c = vp8_sad16x16_c;
const sad_m_by_n_fn_t sad_8x16_c = vp8_sad8x16_c;
const sad_m_by_n_fn_t sad_16x8_c = vp8_sad16x8_c;
const sad_m_by_n_fn_t sad_8x8_c = vp8_sad8x8_c;
const sad_m_by_n_fn_t sad_4x4_c = vp8_sad4x4_c;
#endif
#if CONFIG_VP9_ENCODER
const sad_m_by_n_fn_t sad_64x64_c_vp9 = vp9_sad64x64_c;
const sad_m_by_n_fn_t sad_32x32_c_vp9 = vp9_sad32x32_c;
const sad_m_by_n_fn_t sad_16x16_c_vp9 = vp9_sad16x16_c;
const sad_m_by_n_fn_t sad_8x16_c_vp9 = vp9_sad8x16_c;
const sad_m_by_n_fn_t sad_16x8_c_vp9 = vp9_sad16x8_c;
const sad_m_by_n_fn_t sad_8x8_c_vp9 = vp9_sad8x8_c;
const sad_m_by_n_fn_t sad_4x4_c_vp9 = vp9_sad4x4_c;
#endif
INSTANTIATE_TEST_CASE_P(C, SADTest, ::testing::Values(
#if CONFIG_VP8_ENCODER
make_tuple(16, 16, sad_16x16_c),
make_tuple(8, 16, sad_8x16_c),
make_tuple(16, 8, sad_16x8_c),
make_tuple(8, 8, sad_8x8_c),
make_tuple(4, 4, sad_4x4_c)));
make_tuple(4, 4, sad_4x4_c)
#endif
VP8_VP9_SEPARATOR
#if CONFIG_VP9_ENCODER
make_tuple(64, 64, sad_64x64_c_vp9),
make_tuple(32, 32, sad_32x32_c_vp9),
make_tuple(16, 16, sad_16x16_c_vp9),
make_tuple(8, 16, sad_8x16_c_vp9),
make_tuple(16, 8, sad_16x8_c_vp9),
make_tuple(8, 8, sad_8x8_c_vp9),
make_tuple(4, 4, sad_4x4_c_vp9)
#endif
));
#if CONFIG_VP9_ENCODER
const sad_n_by_n_by_4_fn_t sad_64x64x4d_c = vp9_sad64x64x4d_c;
const sad_n_by_n_by_4_fn_t sad_32x32x4d_c = vp9_sad32x32x4d_c;
const sad_n_by_n_by_4_fn_t sad_16x16x4d_c = vp9_sad16x16x4d_c;
const sad_n_by_n_by_4_fn_t sad_8x8x4d_c = vp9_sad8x8x4d_c;
const sad_n_by_n_by_4_fn_t sad_4x4x4d_c = vp9_sad4x4x4d_c;
INSTANTIATE_TEST_CASE_P(C, SADx4Test, ::testing::Values(
make_tuple(64, 64, sad_64x64x4d_c),
make_tuple(32, 32, sad_32x32x4d_c),
make_tuple(16, 16, sad_16x16x4d_c),
make_tuple(8, 8, sad_8x8x4d_c),
make_tuple(4, 4, sad_4x4x4d_c)));
#endif
// ARM tests
#if HAVE_MEDIA
@ -219,31 +372,120 @@ INSTANTIATE_TEST_CASE_P(NEON, SADTest, ::testing::Values(
// X86 tests
#if HAVE_MMX
#if CONFIG_VP8_ENCODER
const sad_m_by_n_fn_t sad_16x16_mmx = vp8_sad16x16_mmx;
const sad_m_by_n_fn_t sad_8x16_mmx = vp8_sad8x16_mmx;
const sad_m_by_n_fn_t sad_16x8_mmx = vp8_sad16x8_mmx;
const sad_m_by_n_fn_t sad_8x8_mmx = vp8_sad8x8_mmx;
const sad_m_by_n_fn_t sad_4x4_mmx = vp8_sad4x4_mmx;
#endif
#if CONFIG_VP9_ENCODER
const sad_m_by_n_fn_t sad_16x16_mmx_vp9 = vp9_sad16x16_mmx;
const sad_m_by_n_fn_t sad_8x16_mmx_vp9 = vp9_sad8x16_mmx;
const sad_m_by_n_fn_t sad_16x8_mmx_vp9 = vp9_sad16x8_mmx;
const sad_m_by_n_fn_t sad_8x8_mmx_vp9 = vp9_sad8x8_mmx;
const sad_m_by_n_fn_t sad_4x4_mmx_vp9 = vp9_sad4x4_mmx;
#endif
INSTANTIATE_TEST_CASE_P(MMX, SADTest, ::testing::Values(
#if CONFIG_VP8_ENCODER
make_tuple(16, 16, sad_16x16_mmx),
make_tuple(8, 16, sad_8x16_mmx),
make_tuple(16, 8, sad_16x8_mmx),
make_tuple(8, 8, sad_8x8_mmx),
make_tuple(4, 4, sad_4x4_mmx)));
make_tuple(4, 4, sad_4x4_mmx)
#endif
VP8_VP9_SEPARATOR
#if CONFIG_VP9_ENCODER
make_tuple(16, 16, sad_16x16_mmx_vp9),
make_tuple(8, 16, sad_8x16_mmx_vp9),
make_tuple(16, 8, sad_16x8_mmx_vp9),
make_tuple(8, 8, sad_8x8_mmx_vp9),
make_tuple(4, 4, sad_4x4_mmx_vp9)
#endif
));
#endif
#if HAVE_SSE
#if CONFIG_VP9_ENCODER
const sad_m_by_n_fn_t sad_4x4_sse_vp9 = vp9_sad4x4_sse;
INSTANTIATE_TEST_CASE_P(SSE, SADTest, ::testing::Values(
make_tuple(4, 4, sad_4x4_sse_vp9)));
const sad_n_by_n_by_4_fn_t sad_4x4x4d_sse = vp9_sad4x4x4d_sse;
INSTANTIATE_TEST_CASE_P(SSE, SADx4Test, ::testing::Values(
make_tuple(4, 4, sad_4x4x4d_sse)));
#endif
#endif
#if HAVE_SSE2
#if CONFIG_VP8_ENCODER
const sad_m_by_n_fn_t sad_16x16_wmt = vp8_sad16x16_wmt;
const sad_m_by_n_fn_t sad_8x16_wmt = vp8_sad8x16_wmt;
const sad_m_by_n_fn_t sad_16x8_wmt = vp8_sad16x8_wmt;
const sad_m_by_n_fn_t sad_8x8_wmt = vp8_sad8x8_wmt;
const sad_m_by_n_fn_t sad_4x4_wmt = vp8_sad4x4_wmt;
#endif
#if CONFIG_VP9_ENCODER
const sad_m_by_n_fn_t sad_64x64_sse2_vp9 = vp9_sad64x64_sse2;
const sad_m_by_n_fn_t sad_32x32_sse2_vp9 = vp9_sad32x32_sse2;
const sad_m_by_n_fn_t sad_16x16_sse2_vp9 = vp9_sad16x16_sse2;
const sad_m_by_n_fn_t sad_8x16_sse2_vp9 = vp9_sad8x16_sse2;
const sad_m_by_n_fn_t sad_16x8_sse2_vp9 = vp9_sad16x8_sse2;
const sad_m_by_n_fn_t sad_8x8_sse2_vp9 = vp9_sad8x8_sse2;
#endif
INSTANTIATE_TEST_CASE_P(SSE2, SADTest, ::testing::Values(
#if CONFIG_VP8_ENCODER
make_tuple(16, 16, sad_16x16_wmt),
make_tuple(8, 16, sad_8x16_wmt),
make_tuple(16, 8, sad_16x8_wmt),
make_tuple(8, 8, sad_8x8_wmt),
make_tuple(4, 4, sad_4x4_wmt)));
make_tuple(4, 4, sad_4x4_wmt)
#endif
VP8_VP9_SEPARATOR
#if CONFIG_VP9_ENCODER
make_tuple(64, 64, sad_64x64_sse2_vp9),
make_tuple(32, 32, sad_32x32_sse2_vp9),
make_tuple(16, 16, sad_16x16_sse2_vp9),
make_tuple(8, 16, sad_8x16_sse2_vp9),
make_tuple(16, 8, sad_16x8_sse2_vp9),
make_tuple(8, 8, sad_8x8_sse2_vp9)
#endif
));
#if CONFIG_VP9_ENCODER
const sad_n_by_n_by_4_fn_t sad_64x64x4d_sse2 = vp9_sad64x64x4d_sse2;
const sad_n_by_n_by_4_fn_t sad_32x32x4d_sse2 = vp9_sad32x32x4d_sse2;
const sad_n_by_n_by_4_fn_t sad_16x16x4d_sse2 = vp9_sad16x16x4d_sse2;
const sad_n_by_n_by_4_fn_t sad_16x8x4d_sse2 = vp9_sad16x8x4d_sse2;
const sad_n_by_n_by_4_fn_t sad_8x16x4d_sse2 = vp9_sad8x16x4d_sse2;
const sad_n_by_n_by_4_fn_t sad_8x8x4d_sse2 = vp9_sad8x8x4d_sse2;
INSTANTIATE_TEST_CASE_P(SSE2, SADx4Test, ::testing::Values(
make_tuple(64, 64, sad_64x64x4d_sse2),
make_tuple(32, 32, sad_32x32x4d_sse2),
make_tuple(16, 16, sad_16x16x4d_sse2),
make_tuple(16, 8, sad_16x8x4d_sse2),
make_tuple(8, 16, sad_8x16x4d_sse2),
make_tuple(8, 8, sad_8x8x4d_sse2)));
#endif
#endif
#if HAVE_SSE3
#if CONFIG_VP8_ENCODER
const sad_n_by_n_by_4_fn_t sad_16x16x4d_sse3 = vp8_sad16x16x4d_sse3;
const sad_n_by_n_by_4_fn_t sad_16x8x4d_sse3 = vp8_sad16x8x4d_sse3;
const sad_n_by_n_by_4_fn_t sad_8x16x4d_sse3 = vp8_sad8x16x4d_sse3;
const sad_n_by_n_by_4_fn_t sad_8x8x4d_sse3 = vp8_sad8x8x4d_sse3;
const sad_n_by_n_by_4_fn_t sad_4x4x4d_sse3 = vp8_sad4x4x4d_sse3;
INSTANTIATE_TEST_CASE_P(SSE3, SADx4Test, ::testing::Values(
make_tuple(16, 16, sad_16x16x4d_sse3),
make_tuple(16, 8, sad_16x8x4d_sse3),
make_tuple(8, 16, sad_8x16x4d_sse3),
make_tuple(8, 8, sad_8x8x4d_sse3),
make_tuple(4, 4, sad_4x4x4d_sse3)));
#endif
#endif
#if HAVE_SSSE3
const sad_m_by_n_fn_t sad_16x16_sse3 = vp8_sad16x16_sse3;
INSTANTIATE_TEST_CASE_P(SSE3, SADTest, ::testing::Values(

100
test/superframe_test.cc Normal file
View File

@ -0,0 +1,100 @@
/*
* Copyright (c) 2012 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <climits>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "test/util.h"
namespace {
class SuperframeTest : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<libvpx_test::TestMode> {
protected:
SuperframeTest() : EncoderTest(GET_PARAM(0)), modified_buf_(NULL),
last_sf_pts_(0) {}
virtual void SetUp() {
InitializeConfig();
SetMode(GET_PARAM(1));
sf_count_ = 0;
sf_count_max_ = INT_MAX;
}
virtual void TearDown() {
delete modified_buf_;
}
virtual bool Continue() const {
return !HasFatalFailure() && !abort_;
}
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
libvpx_test::Encoder *encoder) {
if (video->frame() == 1) {
encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 1);
}
}
virtual const vpx_codec_cx_pkt_t * MutateEncoderOutputHook(
const vpx_codec_cx_pkt_t *pkt) {
if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
return pkt;
const uint8_t *buffer = reinterpret_cast<uint8_t*>(pkt->data.frame.buf);
const uint8_t marker = buffer[pkt->data.frame.sz - 1];
const int frames = (marker & 0x7) + 1;
const int mag = ((marker >> 3) & 3) + 1;
const unsigned int index_sz = 2 + mag * frames;
if ((marker & 0xe0) == 0xc0 &&
pkt->data.frame.sz >= index_sz &&
buffer[pkt->data.frame.sz - index_sz] == marker) {
// frame is a superframe. strip off the index.
if (modified_buf_)
delete modified_buf_;
modified_buf_ = new uint8_t[pkt->data.frame.sz - index_sz];
memcpy(modified_buf_, pkt->data.frame.buf,
pkt->data.frame.sz - index_sz);
modified_pkt_ = *pkt;
modified_pkt_.data.frame.buf = modified_buf_;
modified_pkt_.data.frame.sz -= index_sz;
sf_count_++;
last_sf_pts_ = pkt->data.frame.pts;
return &modified_pkt_;
}
// Make sure we do a few frames after the last SF
abort_ |= sf_count_ > sf_count_max_ &&
pkt->data.frame.pts - last_sf_pts_ >= 5;
return pkt;
}
int sf_count_;
int sf_count_max_;
vpx_codec_cx_pkt_t modified_pkt_;
uint8_t *modified_buf_;
vpx_codec_pts_t last_sf_pts_;
};
TEST_P(SuperframeTest, TestSuperframeIndexIsOptional) {
sf_count_max_ = 0; // early exit on successful test.
cfg_.g_lag_in_frames = 25;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 40);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
EXPECT_EQ(sf_count_, 1);
}
VP9_INSTANTIATE_TEST_CASE(SuperframeTest, ::testing::Values(
::libvpx_test::kTwoPassGood));
} // namespace

36
test/test.mk
View File

@ -1,7 +1,8 @@
LIBVPX_TEST_SRCS-yes += register_state_check.h
LIBVPX_TEST_SRCS-yes += test.mk
LIBVPX_TEST_SRCS-yes += acm_random.h
LIBVPX_TEST_SRCS-yes += md5_helper.h
LIBVPX_TEST_SRCS-yes += codec_factory.h
LIBVPX_TEST_SRCS-yes += test_libvpx.cc
LIBVPX_TEST_SRCS-yes += util.h
LIBVPX_TEST_SRCS-yes += video_source.h
@ -15,17 +16,20 @@ LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += altref_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += config_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += cq_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += datarate_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += encode_test_driver.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += encode_test_driver.h
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += error_resilience_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += i420_video_source.h
LIBVPX_TEST_SRCS-yes += encode_test_driver.cc
LIBVPX_TEST_SRCS-yes += encode_test_driver.h
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += error_resilience_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += i420_video_source.h
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += keyframe_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += resize_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += ../md5_utils.h ../md5_utils.c
LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += decode_test_driver.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += decode_test_driver.h
LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += ivf_video_source.h
LIBVPX_TEST_SRCS-$(CONFIG_DECODERS) += ../md5_utils.h ../md5_utils.c
LIBVPX_TEST_SRCS-yes += decode_test_driver.cc
LIBVPX_TEST_SRCS-yes += decode_test_driver.h
LIBVPX_TEST_SRCS-$(CONFIG_DECODERS) += ivf_video_source.h
LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += test_vector_test.cc
##
@ -44,10 +48,10 @@ ifeq ($(CONFIG_VP8_ENCODER)$(CONFIG_VP8_DECODER),yesyes)
LIBVPX_TEST_SRCS-yes += vp8_boolcoder_test.cc
endif
LIBVPX_TEST_SRCS-yes += idctllm_test.cc
LIBVPX_TEST_SRCS-yes += idct_test.cc
LIBVPX_TEST_SRCS-yes += intrapred_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_POSTPROC) += pp_filter_test.cc
LIBVPX_TEST_SRCS-yes += sad_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += sad_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += set_roi.cc
LIBVPX_TEST_SRCS-yes += sixtap_predict_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += subtract_test.cc
@ -66,13 +70,18 @@ LIBVPX_TEST_SRCS-yes += vp9_boolcoder_test.cc
# IDCT test currently depends on FDCT function
LIBVPX_TEST_SRCS-yes += idct8x8_test.cc
LIBVPX_TEST_SRCS-yes += superframe_test.cc
LIBVPX_TEST_SRCS-yes += tile_independence_test.cc
endif
LIBVPX_TEST_SRCS-$(CONFIG_VP9) += convolve_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += fdct4x4_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += fdct8x8_test.cc
#LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct16x16_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct16x16_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += variance_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct32x32_test.cc
endif # VP9
@ -82,7 +91,8 @@ endif
##
## TEST DATA
##
LIBVPX_TEST_DATA-$(CONFIG_VP8_ENCODER) += hantro_collage_w352h288.yuv
LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += hantro_collage_w352h288.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-001.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-002.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-003.ivf

47
test/test_vector_test.cc
View File

@ -12,17 +12,15 @@
#include <cstdlib>
#include <string>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/decode_test_driver.h"
#include "test/ivf_video_source.h"
#include "test/util.h"
#include "test/md5_helper.h"
extern "C" {
#include "./md5_utils.h"
#include "vpx_mem/vpx_mem.h"
}
#if defined(_MSC_VER)
#define snprintf sprintf_s
#endif
namespace {
// There are 61 test vectors in total.
const char *kTestVectors[] = {
@ -59,10 +57,10 @@ const char *kTestVectors[] = {
"vp80-05-sharpness-1440.ivf", "vp80-05-sharpness-1443.ivf"
};
class TestVectorTest : public libvpx_test::DecoderTest,
public ::testing::TestWithParam<const char*> {
class TestVectorTest : public ::libvpx_test::DecoderTest,
public ::libvpx_test::CodecTestWithParam<const char*> {
protected:
TestVectorTest() : md5_file_(NULL) {}
TestVectorTest() : DecoderTest(GET_PARAM(0)), md5_file_(NULL) {}
virtual ~TestVectorTest() {
if (md5_file_)
@ -85,30 +83,9 @@ class TestVectorTest : public libvpx_test::DecoderTest,
ASSERT_NE(res, EOF) << "Read md5 data failed";
expected_md5[32] = '\0';
MD5Context md5;
MD5Init(&md5);
// Compute and update md5 for each raw in decompressed data.
for (int plane = 0; plane < 3; ++plane) {
uint8_t *buf = img.planes[plane];
for (unsigned int y = 0; y < (plane ? (img.d_h + 1) >> 1 : img.d_h);
++y) {
MD5Update(&md5, buf, (plane ? (img.d_w + 1) >> 1 : img.d_w));
buf += img.stride[plane];
}
}
uint8_t md5_sum[16];
MD5Final(md5_sum, &md5);
char actual_md5[33];
// Convert to get the actual md5.
for (int i = 0; i < 16; i++) {
snprintf(&actual_md5[i * 2], sizeof(actual_md5) - i * 2, "%02x",
md5_sum[i]);
}
actual_md5[32] = '\0';
::libvpx_test::MD5 md5_res;
md5_res.Add(&img);
const char *actual_md5 = md5_res.Get();
// Check md5 match.
ASSERT_STREQ(expected_md5, actual_md5)
@ -124,7 +101,7 @@ class TestVectorTest : public libvpx_test::DecoderTest,
// checksums match the correct md5 data, then the test is passed. Otherwise,
// the test failed.
TEST_P(TestVectorTest, MD5Match) {
const std::string filename = GetParam();
const std::string filename = GET_PARAM(1);
// Open compressed video file.
libvpx_test::IVFVideoSource video(filename);
@ -138,7 +115,7 @@ TEST_P(TestVectorTest, MD5Match) {
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
}
INSTANTIATE_TEST_CASE_P(TestVectorSequence, TestVectorTest,
::testing::ValuesIn(kTestVectors));
VP8_INSTANTIATE_TEST_CASE(TestVectorTest,
::testing::ValuesIn(kTestVectors));
} // namespace

102
test/tile_independence_test.cc Normal file
View File

@ -0,0 +1,102 @@
/*
Copyright (c) 2012 The WebM project authors. All Rights Reserved.
Use of this source code is governed by a BSD-style license
that can be found in the LICENSE file in the root of the source
tree. An additional intellectual property rights grant can be found
in the file PATENTS. All contributing project authors may
be found in the AUTHORS file in the root of the source tree.
*/
#include <cstdio>
#include <cstdlib>
#include <string>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/i420_video_source.h"
#include "test/util.h"
#include "test/md5_helper.h"
extern "C" {
#include "vpx_mem/vpx_mem.h"
}
namespace {
class TileIndependenceTest : public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWithParam<int> {
protected:
TileIndependenceTest() : EncoderTest(GET_PARAM(0)), n_tiles_(GET_PARAM(1)),
md5_fw_order_(), md5_inv_order_() {
init_flags_ = VPX_CODEC_USE_PSNR;
vpx_codec_dec_cfg_t cfg;
cfg.w = 704;
cfg.h = 144;
cfg.threads = 1;
fw_dec_ = codec_->CreateDecoder(cfg, 0);
inv_dec_ = codec_->CreateDecoder(cfg, 0);
inv_dec_->Control(VP9_INVERT_TILE_DECODE_ORDER, 1);
}
virtual ~TileIndependenceTest() {
delete fw_dec_;
delete inv_dec_;
}
virtual void SetUp() {
InitializeConfig();
SetMode(libvpx_test::kTwoPassGood);
}
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
libvpx_test::Encoder *encoder) {
if (video->frame() == 1) {
encoder->Control(VP9E_SET_TILE_COLUMNS, n_tiles_);
}
}
void UpdateMD5(::libvpx_test::Decoder *dec, const vpx_codec_cx_pkt_t *pkt,
::libvpx_test::MD5 *md5) {
dec->DecodeFrame((uint8_t *) pkt->data.frame.buf, pkt->data.frame.sz);
const vpx_image_t *img = dec->GetDxData().Next();
md5->Add(img);
}
virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
UpdateMD5(fw_dec_, pkt, &md5_fw_order_);
UpdateMD5(inv_dec_, pkt, &md5_inv_order_);
}
private:
int n_tiles_;
protected:
::libvpx_test::MD5 md5_fw_order_, md5_inv_order_;
::libvpx_test::Decoder *fw_dec_, *inv_dec_;
};
// run an encode with 2 or 4 tiles, and do the decode both in normal and
// inverted tile ordering. Ensure that the MD5 of the output in both cases
// is identical. If so, tiles are considered independent and the test passes.
TEST_P(TileIndependenceTest, MD5Match) {
const vpx_rational timebase = { 33333333, 1000000000 };
cfg_.g_timebase = timebase;
cfg_.rc_target_bitrate = 500;
cfg_.g_lag_in_frames = 25;
cfg_.rc_end_usage = VPX_VBR;
libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 704, 144,
timebase.den, timebase.num, 0, 30);
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
const char *md5_fw_str = md5_fw_order_.Get();
const char *md5_inv_str = md5_inv_order_.Get();
// could use ASSERT_EQ(!memcmp(.., .., 16) here, but this gives nicer
// output if it fails. Not sure if it's helpful since it's really just
// a MD5...
ASSERT_STREQ(md5_fw_str, md5_inv_str);
}
VP9_INSTANTIATE_TEST_CASE(TileIndependenceTest,
::testing::Range(0, 2, 1));
} // namespace

30
test/util.h
View File

@ -11,8 +11,38 @@
#ifndef TEST_UTIL_H_
#define TEST_UTIL_H_
#include <stdio.h>
#include <math.h>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "vpx/vpx_image.h"
// Macros
#define PARAMS(...) ::testing::TestWithParam< std::tr1::tuple< __VA_ARGS__ > >
#define GET_PARAM(k) std::tr1::get< k >(GetParam())
static double compute_psnr(const vpx_image_t *img1,
const vpx_image_t *img2) {
assert((img1->fmt == img2->fmt) &&
(img1->d_w == img2->d_w) &&
(img1->d_h == img2->d_h));
const unsigned int width_y = img1->d_w;
const unsigned int height_y = img1->d_h;
unsigned int i, j;
int64_t sqrerr = 0;
for (i = 0; i < height_y; ++i)
for (j = 0; j < width_y; ++j) {
int64_t d = img1->planes[VPX_PLANE_Y][i * img1->stride[VPX_PLANE_Y] + j] -
img2->planes[VPX_PLANE_Y][i * img2->stride[VPX_PLANE_Y] + j];
sqrerr += d * d;
}
double mse = sqrerr / (width_y * height_y);
double psnr = 100.0;
if (mse > 0.0) {
psnr = 10 * log10(255.0 * 255.0 / mse);
}
return psnr;
}
#endif // TEST_UTIL_H_

2
vp8/decoder/onyxd_if.c
View File

@ -302,7 +302,7 @@ int check_fragments_for_errors(VP8D_COMP *pbi)
return 1;
}
int vp8dx_receive_compressed_data(VP8D_COMP *pbi, size_t size,
const uint8_t *source,
int64_t time_stamp)

32
vp8/encoder/bitstream.c
View File

@ -50,7 +50,7 @@ const int vp8cx_base_skip_false_prob[128] =
unsigned __int64 Sectionbits[500];
#endif
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
int intra_mode_stats[10][10][10];
static unsigned int tree_update_hist [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES] [2];
extern unsigned int active_section;
@ -531,7 +531,7 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi)
vp8_convert_rfct_to_prob(cpi);
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 1;
#endif
@ -580,7 +580,7 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi)
xd->mb_to_top_edge = -((mb_row * 16)) << 3;
xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3;
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 9;
#endif
@ -593,7 +593,7 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi)
if (rf == INTRA_FRAME)
{
vp8_write(w, 0, cpi->prob_intra_coded);
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 6;
#endif
write_ymode(w, mode, pc->fc.ymode_prob);
@ -633,13 +633,13 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi)
vp8_mv_ref_probs(mv_ref_p, ct);
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
accum_mv_refs(mode, ct);
#endif
}
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 3;
#endif
@ -649,7 +649,7 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi)
{
case NEWMV:
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 5;
#endif
@ -692,7 +692,7 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi)
if (blockmode == NEW4X4)
{
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 11;
#endif
write_mv(w, &blockmv.as_mv, &best_mv, (const MV_CONTEXT *) mvc);
@ -769,7 +769,7 @@ static void write_kfmodes(VP8_COMP *cpi)
const B_PREDICTION_MODE L = left_block_mode(m, i);
const int bm = m->bmi[i].as_mode;
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
++intra_mode_stats [A] [L] [bm];
#endif
@ -1160,7 +1160,7 @@ void vp8_update_coef_probs(VP8_COMP *cpi)
#endif
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
++ tree_update_hist [i][j][k][t] [u];
#endif
@ -1181,7 +1181,7 @@ void vp8_update_coef_probs(VP8_COMP *cpi)
while (++t < ENTROPY_NODES);
/* Accum token counts for generation of default statistics */
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
t = 0;
do
@ -1527,7 +1527,7 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned char * dest
if (pc->frame_type != KEY_FRAME)
vp8_write_bit(bc, pc->refresh_last_frame);
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
if (pc->frame_type == INTER_FRAME)
active_section = 0;
@ -1550,7 +1550,7 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned char * dest
vp8_update_coef_probs(cpi);
#endif
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 2;
#endif
@ -1561,7 +1561,7 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned char * dest
{
write_kfmodes(cpi);
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 8;
#endif
}
@ -1569,7 +1569,7 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned char * dest
{
pack_inter_mode_mvs(cpi);
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 1;
#endif
}
@ -1687,7 +1687,7 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned char * dest
#endif
}
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
void print_tree_update_probs()
{
int i, j, k, l;

2
vp8/encoder/boolhuff.c
View File

@ -16,7 +16,7 @@ unsigned __int64 Sectionbits[500];
#endif
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
unsigned int active_section = 0;
#endif

2
vp8/encoder/boolhuff.h
View File

@ -67,7 +67,7 @@ static void vp8_encode_bool(BOOL_CODER *br, int bit, int probability)
unsigned int lowvalue = br->lowvalue;
register unsigned int shift;
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
#if defined(SECTIONBITS_OUTPUT)
if (bit)

6
vp8/encoder/encodemv.c
View File

@ -16,7 +16,7 @@
#include <math.h>
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
extern unsigned int active_section;
#endif
@ -359,7 +359,7 @@ void vp8_write_mvprobs(VP8_COMP *cpi)
vp8_writer *const w = cpi->bc;
MV_CONTEXT *mvc = cpi->common.fc.mvc;
int flags[2] = {0, 0};
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 4;
#endif
write_component_probs(
@ -374,7 +374,7 @@ void vp8_write_mvprobs(VP8_COMP *cpi)
if (flags[0] || flags[1])
vp8_build_component_cost_table(cpi->mb.mvcost, (const MV_CONTEXT *) cpi->common.fc.mvc, flags);
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
active_section = 5;
#endif
}

10
vp8/encoder/mcomp.c
View File

@ -18,7 +18,7 @@
#include <math.h>
#include "vp8/common/findnearmv.h"
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
static int mv_ref_ct [31] [4] [2];
static int mv_mode_cts [4] [2];
#endif
@ -1912,7 +1912,7 @@ int vp8_refining_search_sadx4(MACROBLOCK *x, BLOCK *b, BLOCKD *d,
+ mv_err_cost(&this_mv, center_mv, mvcost, x->errorperbit);
}
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
void print_mode_context(void)
{
FILE *f = fopen("modecont.c", "w");
@ -1965,8 +1965,8 @@ void print_mode_context(void)
fclose(f);
}
/* MV ref count ENTROPY_STATS stats code */
#ifdef ENTROPY_STATS
/* MV ref count VP8_ENTROPY_STATS stats code */
#ifdef VP8_ENTROPY_STATS
void init_mv_ref_counts()
{
vpx_memset(mv_ref_ct, 0, sizeof(mv_ref_ct));
@ -2020,6 +2020,6 @@ void accum_mv_refs(MB_PREDICTION_MODE m, const int ct[4])
}
}
#endif/* END MV ref count ENTROPY_STATS stats code */
#endif/* END MV ref count VP8_ENTROPY_STATS stats code */
#endif

2
vp8/encoder/mcomp.h
View File

@ -15,7 +15,7 @@
#include "block.h"
#include "vp8/common/variance.h"
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
extern void init_mv_ref_counts();
extern void accum_mv_refs(MB_PREDICTION_MODE, const int near_mv_ref_cts[4]);
#endif

10
vp8/encoder/onyx_if.c
View File

@ -111,7 +111,7 @@ extern int skip_false_count;
#endif
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
extern int intra_mode_stats[10][10][10];
#endif
@ -1805,7 +1805,7 @@ struct VP8_COMP* vp8_create_compressor(VP8_CONFIG *oxcf)
else
cpi->cyclic_refresh_map = (signed char *) NULL;
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
init_context_counters();
#endif
@ -1923,7 +1923,7 @@ struct VP8_COMP* vp8_create_compressor(VP8_CONFIG *oxcf)
cpi->mb.rd_thresh_mult[i] = 128;
}
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
init_mv_ref_counts();
#endif
@ -2060,7 +2060,7 @@ void vp8_remove_compressor(VP8_COMP **ptr)
#endif
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
print_context_counters();
print_tree_update_probs();
print_mode_context();
@ -2242,7 +2242,7 @@ void vp8_remove_compressor(VP8_COMP **ptr)
}
#endif
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
{
int i, j, k;
FILE *fmode = fopen("modecontext.c", "w");

4
vp8/encoder/tokenize.c
View File

@ -20,7 +20,7 @@
/* Global event counters used for accumulating statistics across several
compressions, then generating context.c = initial stats. */
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
_int64 context_counters[BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
#endif
void vp8_stuff_mb(VP8_COMP *cpi, MACROBLOCK *x, TOKENEXTRA **t) ;
@ -413,7 +413,7 @@ void vp8_tokenize_mb(VP8_COMP *cpi, MACROBLOCK *x, TOKENEXTRA **t)
}
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
void init_context_counters(void)
{

2
vp8/encoder/tokenize.h
View File

@ -33,7 +33,7 @@ typedef struct
int rd_cost_mby(MACROBLOCKD *);
#ifdef ENTROPY_STATS
#ifdef VP8_ENTROPY_STATS
void init_context_counters();
void print_context_counters();

0
vp8/encoder/x86/quantize_sse2.c → vp8/encoder/x86/quantize_sse2_intrinsics.c
View File

2
vp8/vp8_cx_iface.c
View File

@ -684,6 +684,8 @@ static vpx_codec_err_t image2yuvconfig(const vpx_image_t *img,
yv12->u_buffer = img->planes[VPX_PLANE_U];
yv12->v_buffer = img->planes[VPX_PLANE_V];
yv12->y_crop_width = img->d_w;
yv12->y_crop_height = img->d_h;
yv12->y_width = img->d_w;
yv12->y_height = img->d_h;
yv12->uv_width = (1 + yv12->y_width) / 2;

2
vp8/vp8_dx_iface.c
View File

@ -790,6 +790,8 @@ static vpx_codec_err_t image2yuvconfig(const vpx_image_t *img,
yv12->u_buffer = img->planes[VPX_PLANE_U];
yv12->v_buffer = img->planes[VPX_PLANE_V];
yv12->y_crop_width = img->d_w;
yv12->y_crop_height = img->d_h;
yv12->y_width = img->d_w;
yv12->y_height = img->d_h;
yv12->uv_width = yv12->y_width / 2;

6
vp8/vp8cx.mk
View File

@ -89,12 +89,12 @@ VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/subtract_mmx.asm
VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/vp8_enc_stubs_mmx.c
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/dct_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/fwalsh_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/quantize_sse2.c
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/quantize_sse2_intrinsics.c
# TODO(johann) make this generic
ifeq ($(HAVE_SSE2),yes)
vp8/encoder/x86/quantize_sse2.c.o: CFLAGS += -msse2
vp8/encoder/x86/quantize_sse2.c.d: CFLAGS += -msse2
vp8/encoder/x86/quantize_sse2_intrinsics.c.o: CFLAGS += -msse2
vp8/encoder/x86/quantize_sse2_intrinsics.c.d: CFLAGS += -msse2
endif
ifeq ($(CONFIG_TEMPORAL_DENOISING),yes)

2
vp9/common/generic/vp9_systemdependent.c
View File

@ -11,8 +11,6 @@
#include "./vpx_config.h"
#include "vp9_rtcd.h"
#include "vp9/common/vp9_subpixel.h"
#include "vp9/common/vp9_loopfilter.h"
#include "vp9/common/vp9_onyxc_int.h"
void vp9_machine_specific_config(VP9_COMMON *ctx) {

4
vp9/common/ppc/vp9_idctllm_altivec.asm → vp9/common/ppc/vp9_idct_altivec.asm
View File

@ -9,7 +9,7 @@
;
.globl short_idct4x4llm_ppc
.globl short_idct4x4_ppc
.macro load_c V, LABEL, OFF, R0, R1
lis \R0, \LABEL@ha
@ -21,7 +21,7 @@
;# r4 short *output
;# r5 int pitch
.align 2
short_idct4x4llm_ppc:
short_idct4x4_ppc:
mfspr r11, 256 ;# get old VRSAVE
oris r12, r11, 0xfff8
mtspr 256, r12 ;# set VRSAVE

75
vp9/common/ppc/vp9_systemdependent.c
View File

@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vp9/common/vp9_subpixel.h"
#include "vp9/common/vp9_loopfilter.h"
#include "recon.h"
#include "vp9/common/vp9_onyxc_int.h"
@ -17,32 +16,28 @@ void (*vp8_short_idct4x4)(short *input, short *output, int pitch);
void (*vp8_short_idct4x4_1)(short *input, short *output, int pitch);
void (*vp8_dc_only_idct)(short input_dc, short *output, int pitch);
extern void (*vp9_post_proc_down_and_across)(
unsigned char *src_ptr,
unsigned char *dst_ptr,
int src_pixels_per_line,
int dst_pixels_per_line,
int rows,
int cols,
int flimit
);
extern void (*vp9_post_proc_down_and_across)(unsigned char *src_ptr,
unsigned char *dst_ptr,
int src_pixels_per_line,
int dst_pixels_per_line,
int rows, int cols, int flimit);
extern void (*vp9_mbpost_proc_down)(unsigned char *dst, int pitch, int rows, int cols, int flimit);
extern void vp9_mbpost_proc_down_c(unsigned char *dst, int pitch, int rows, int cols, int flimit);
extern void (*vp9_mbpost_proc_across_ip)(unsigned char *src, int pitch, int rows, int cols, int flimit);
extern void vp9_mbpost_proc_across_ip_c(unsigned char *src, int pitch, int rows, int cols, int flimit);
extern void vp9_post_proc_down_and_across_c
(
unsigned char *src_ptr,
unsigned char *dst_ptr,
int src_pixels_per_line,
int dst_pixels_per_line,
int rows,
int cols,
int flimit
);
void vp9_plane_add_noise_c(unsigned char *Start, unsigned int Width, unsigned int Height, int Pitch, int q, int a);
extern void (*vp9_mbpost_proc_down)(unsigned char *dst, int pitch,
int rows, int cols, int flimit);
extern void vp9_mbpost_proc_down_c(unsigned char *dst, int pitch,
int rows, int cols, int flimit);
extern void (*vp9_mbpost_proc_across_ip)(unsigned char *src, int pitch,
int rows, int cols, int flimit);
extern void vp9_mbpost_proc_across_ip_c(unsigned char *src, int pitch,
int rows, int cols, int flimit);
extern void vp9_post_proc_down_and_across_c(unsigned char *src_ptr,
unsigned char *dst_ptr,
int src_pixels_per_line,
int dst_pixels_per_line,
int rows, int cols, int flimit);
void vp9_plane_add_noise_c(unsigned char *start,
unsigned int width, unsigned int height,
int pitch, int q, int a);
extern copy_mem_block_function *vp9_copy_mem16x16;
extern copy_mem_block_function *vp9_copy_mem8x8;
@ -60,11 +55,14 @@ extern subpixel_predict_function bilinear_predict16x16_ppc;
extern copy_mem_block_function copy_mem16x16_ppc;
void recon_b_ppc(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
void recon2b_ppc(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
void recon4b_ppc(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
void recon_b_ppc(short *diff_ptr, unsigned char *pred_ptr,
unsigned char *dst_ptr, int stride);
void recon2b_ppc(short *diff_ptr, unsigned char *pred_ptr,
unsigned char *dst_ptr, int stride);
void recon4b_ppc(short *diff_ptr, unsigned char *pred_ptr,
unsigned char *dst_ptr, int stride);
extern void short_idct4x4llm_ppc(short *input, short *output, int pitch);
extern void short_idct4x4_ppc(short *input, short *output, int pitch);
// Generic C
extern subpixel_predict_function vp9_sixtap_predict_c;
@ -80,12 +78,15 @@ extern copy_mem_block_function vp9_copy_mem16x16_c;
extern copy_mem_block_function vp9_copy_mem8x8_c;
extern copy_mem_block_function vp9_copy_mem8x4_c;
void vp9_recon_b_c(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
void vp9_recon2b_c(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
void vp9_recon4b_c(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
void vp9_recon_b_c(short *diff_ptr, unsigned char *pred_ptr,
unsigned char *dst_ptr, int stride);
void vp9_recon2b_c(short *diff_ptr, unsigned char *pred_ptr,
unsigned char *dst_ptr, int stride);
void vp9_recon4b_c(short *diff_ptr, unsigned char *pred_ptr,
unsigned char *dst_ptr, int stride);
extern void vp9_short_idct4x4llm_1_c(short *input, short *output, int pitch);
extern void vp9_short_idct4x4llm_c(short *input, short *output, int pitch);
extern void vp9_short_idct4x4_1_c(short *input, short *output, int pitch);
extern void vp9_short_idct4x4_c(short *input, short *output, int pitch);
extern void vp8_dc_only_idct_c(short input_dc, short *output, int pitch);
// PPC
@ -140,8 +141,8 @@ void vp9_machine_specific_config(void) {
vp9_sixtap_predict8x4 = sixtap_predict8x4_ppc;
vp9_sixtap_predict = sixtap_predict_ppc;
vp8_short_idct4x4_1 = vp9_short_idct4x4llm_1_c;
vp8_short_idct4x4 = short_idct4x4llm_ppc;
vp8_short_idct4x4_1 = vp9_short_idct4x4_1_c;
vp8_short_idct4x4 = short_idct4x4_ppc;
vp8_dc_only_idct = vp8_dc_only_idct_c;
vp8_lf_mbvfull = loop_filter_mbv_ppc;

42
vp9/common/vp9_alloccommon.c
View File

@ -67,20 +67,16 @@ void vp9_de_alloc_frame_buffers(VP9_COMMON *oci) {
int vp9_alloc_frame_buffers(VP9_COMMON *oci, int width, int height) {
int i;
int aligned_width, aligned_height;
vp9_de_alloc_frame_buffers(oci);
/* our internal buffers are always multiples of 16 */
if ((width & 0xf) != 0)
width += 16 - (width & 0xf);
if ((height & 0xf) != 0)
height += 16 - (height & 0xf);
aligned_width = (width + 15) & ~15;
aligned_height = (height + 15) & ~15;
for (i = 0; i < NUM_YV12_BUFFERS; i++) {
oci->fb_idx_ref_cnt[i] = 0;
oci->yv12_fb[i].flags = 0;
if (vp8_yv12_alloc_frame_buffer(&oci->yv12_fb[i], width, height,
VP9BORDERINPIXELS) < 0) {
vp9_de_alloc_frame_buffers(oci);
@ -88,15 +84,16 @@ int vp9_alloc_frame_buffers(VP9_COMMON *oci, int width, int height) {
}
}
oci->new_fb_idx = 0;
oci->lst_fb_idx = 1;
oci->gld_fb_idx = 2;
oci->alt_fb_idx = 3;
oci->new_fb_idx = NUM_YV12_BUFFERS - 1;
oci->fb_idx_ref_cnt[oci->new_fb_idx] = 1;
oci->fb_idx_ref_cnt[0] = 1;
oci->fb_idx_ref_cnt[1] = 1;
oci->fb_idx_ref_cnt[2] = 1;
oci->fb_idx_ref_cnt[3] = 1;
for (i = 0; i < 3; i++)
oci->active_ref_idx[i] = i;
for (i = 0; i < NUM_REF_FRAMES; i++) {
oci->ref_frame_map[i] = i;
oci->fb_idx_ref_cnt[i] = 1;
}
if (vp8_yv12_alloc_frame_buffer(&oci->temp_scale_frame, width, 16,
VP9BORDERINPIXELS) < 0) {
@ -110,8 +107,8 @@ int vp9_alloc_frame_buffers(VP9_COMMON *oci, int width, int height) {
return 1;
}
oci->mb_rows = height >> 4;
oci->mb_cols = width >> 4;
oci->mb_rows = aligned_height >> 4;
oci->mb_cols = aligned_width >> 4;
oci->MBs = oci->mb_rows * oci->mb_cols;
oci->mode_info_stride = oci->mb_cols + 1;
oci->mip = vpx_calloc((oci->mb_cols + 1) * (oci->mb_rows + 1), sizeof(MODE_INFO));
@ -134,7 +131,8 @@ int vp9_alloc_frame_buffers(VP9_COMMON *oci, int width, int height) {
oci->prev_mi = oci->prev_mip + oci->mode_info_stride + 1;
oci->above_context = vpx_calloc(sizeof(ENTROPY_CONTEXT_PLANES) * oci->mb_cols, 1);
oci->above_context =
vpx_calloc(sizeof(ENTROPY_CONTEXT_PLANES) * (3 + oci->mb_cols), 1);
if (!oci->above_context) {
vp9_de_alloc_frame_buffers(oci);
@ -146,6 +144,7 @@ int vp9_alloc_frame_buffers(VP9_COMMON *oci, int width, int height) {
return 0;
}
void vp9_setup_version(VP9_COMMON *cm) {
if (cm->version & 0x4) {
if (!CONFIG_EXPERIMENTAL)
@ -204,9 +203,6 @@ void vp9_create_common(VP9_COMMON *oci) {
/* Initialise reference frame sign bias structure to defaults */
vpx_memset(oci->ref_frame_sign_bias, 0, sizeof(oci->ref_frame_sign_bias));
/* Default disable buffer to buffer copying */
oci->copy_buffer_to_gf = 0;
oci->copy_buffer_to_arf = 0;
oci->kf_ymode_probs_update = 0;
}
@ -220,8 +216,4 @@ void vp9_initialize_common() {
vp9_entropy_mode_init();
vp9_entropy_mv_init();
#if CONFIG_NEWCOEFCONTEXT
vp9_init_neighbors();
#endif
}

436
vp9/common/vp9_blockd.c
View File

@ -12,15 +12,431 @@
#include "vp9/common/vp9_blockd.h"
#include "vpx_mem/vpx_mem.h"
const uint8_t vp9_block2left[TX_SIZE_MAX_SB][25] = {
{0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8},
{0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 4, 4, 4, 4, 6, 6, 6, 6, 8},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6, 8},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6, 8}
const uint8_t vp9_block2left[TX_SIZE_MAX_MB][24] = {
{ 0, 0, 0, 0,
1, 1, 1, 1,
2, 2, 2, 2,
3, 3, 3, 3,
4, 4,
5, 5,
6, 6,
7, 7 },
{ 0, 0, 0, 0,
0, 0, 0, 0,
2, 2, 2, 2,
2, 2, 2, 2,
4, 4,
4, 4,
6, 6,
6, 6 },
{ 0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0 },
};
const uint8_t vp9_block2above[TX_SIZE_MAX_SB][25] = {
{0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 4, 5, 4, 5, 6, 7, 6, 7, 8},
{0, 0, 0, 0, 2, 2, 2, 2, 0, 0, 0, 0, 2, 2, 2, 2, 4, 4, 4, 4, 6, 6, 6, 6, 8},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6, 8},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 6, 6, 6, 6, 8}
const uint8_t vp9_block2above[TX_SIZE_MAX_MB][24] = {
{ 0, 1, 2, 3,
0, 1, 2, 3,
0, 1, 2, 3,
0, 1, 2, 3,
4, 5,
4, 5,
6, 7,
6, 7 },
{ 0, 0, 0, 0,
2, 2, 2, 2,
0, 0, 0, 0,
2, 2, 2, 2,
4, 4,
4, 4,
6, 6,
6, 6 },
{ 0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0 },
};
#define S(x) x + sizeof(ENTROPY_CONTEXT_PLANES) / sizeof(ENTROPY_CONTEXT)
const uint8_t vp9_block2left_sb[TX_SIZE_MAX_SB][96] = {
{ 0, 0, 0, 0, 0, 0, 0, 0,
1, 1, 1, 1, 1, 1, 1, 1,
2, 2, 2, 2, 2, 2, 2, 2,
3, 3, 3, 3, 3, 3, 3, 3,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1),
S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2),
S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3),
4, 4, 4, 4,
5, 5, 5, 5,
S(4), S(4), S(4), S(4),
S(5), S(5), S(5), S(5),
6, 6, 6, 6,
7, 7, 7, 7,
S(6), S(6), S(6), S(6),
S(7), S(7), S(7), S(7) },
{ 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2),
S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2),
4, 4, 4, 4,
4, 4, 4, 4,
S(4), S(4), S(4), S(4),
S(4), S(4), S(4), S(4),
6, 6, 6, 6,
6, 6, 6, 6,
S(6), S(6), S(6), S(6),
S(6), S(6), S(6), S(6) },
{ 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
4, 4, 4, 4,
4, 4, 4, 4,
4, 4, 4, 4,
4, 4, 4, 4,
6, 6, 6, 6,
6, 6, 6, 6,
6, 6, 6, 6,
6, 6, 6, 6 },
{ 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0 },
};
const uint8_t vp9_block2above_sb[TX_SIZE_MAX_SB][96] = {
{ 0, 1, 2, 3, S(0), S(1), S(2), S(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3),
4, 5, S(4), S(5),
4, 5, S(4), S(5),
4, 5, S(4), S(5),
4, 5, S(4), S(5),
6, 7, S(6), S(7),
6, 7, S(6), S(7),
6, 7, S(6), S(7),
6, 7, S(6), S(7) },
{ 0, 0, 0, 0, 2, 2, 2, 2,
S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
0, 0, 0, 0, 2, 2, 2, 2,
S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
0, 0, 0, 0, 2, 2, 2, 2,
S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
0, 0, 0, 0, 2, 2, 2, 2,
S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
4, 4, 4, 4,
S(4), S(4), S(4), S(4),
4, 4, 4, 4,
S(4), S(4), S(4), S(4),
6, 6, 6, 6,
S(6), S(6), S(6), S(6),
6, 6, 6, 6,
S(6), S(6), S(6), S(6) },
{ 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
4, 4, 4, 4,
4, 4, 4, 4,
4, 4, 4, 4,
4, 4, 4, 4,
6, 6, 6, 6,
6, 6, 6, 6,
6, 6, 6, 6,
6, 6, 6, 6 },
{ 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0 },
};
#define T(x) x + 2 * (sizeof(ENTROPY_CONTEXT_PLANES) / sizeof(ENTROPY_CONTEXT))
#define U(x) x + 3 * (sizeof(ENTROPY_CONTEXT_PLANES) / sizeof(ENTROPY_CONTEXT))
const uint8_t vp9_block2left_sb64[TX_SIZE_MAX_SB][384] = {
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1), S(1),
S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2),
S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3), S(3),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1), T(1),
T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2),
T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3), T(3),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1), U(1),
U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2),
U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3), U(3),
4, 4, 4, 4, 4, 4, 4, 4,
5, 5, 5, 5, 5, 5, 5, 5,
S(4), S(4), S(4), S(4), S(4), S(4), S(4), S(4),
S(5), S(5), S(5), S(5), S(5), S(5), S(5), S(5),
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
T(5), T(5), T(5), T(5), T(5), T(5), T(5), T(5),
U(4), U(4), U(4), U(4), U(4), U(4), U(4), U(4),
U(5), U(5), U(5), U(5), U(5), U(5), U(5), U(5),
6, 6, 6, 6, 6, 6, 6, 6,
7, 7, 7, 7, 7, 7, 7, 7,
S(6), S(6), S(6), S(6), S(6), S(6), S(6), S(6),
S(7), S(7), S(7), S(7), S(7), S(7), S(7), S(7),
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
T(7), T(7), T(7), T(7), T(7), T(7), T(7), T(7),
U(6), U(6), U(6), U(6), U(6), U(6), U(6), U(6),
U(7), U(7), U(7), U(7), U(7), U(7), U(7), U(7) },
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2),
S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2),
T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2), T(2),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2),
U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2), U(2),
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
S(4), S(4), S(4), S(4), S(4), S(4), S(4), S(4),
S(4), S(4), S(4), S(4), S(4), S(4), S(4), S(4),
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
U(4), U(4), U(4), U(4), U(4), U(4), U(4), U(4),
U(4), U(4), U(4), U(4), U(4), U(4), U(4), U(4),
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
S(6), S(6), S(6), S(6), S(6), S(6), S(6), S(6),
S(6), S(6), S(6), S(6), S(6), S(6), S(6), S(6),
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
U(6), U(6), U(6), U(6), U(6), U(6), U(6), U(6),
U(6), U(6), U(6), U(6), U(6), U(6), U(6), U(6) },
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6) },
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6 },
};
const uint8_t vp9_block2above_sb64[TX_SIZE_MAX_SB][384] = {
{ 0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
0, 1, 2, 3, S(0), S(1), S(2), S(3), T(0), T(1), T(2), T(3), U(0), U(1), U(2), U(3),
4, 5, S(4), S(5), T(4), T(5), U(4), U(5),
4, 5, S(4), S(5), T(4), T(5), U(4), U(5),
4, 5, S(4), S(5), T(4), T(5), U(4), U(5),
4, 5, S(4), S(5), T(4), T(5), U(4), U(5),
4, 5, S(4), S(5), T(4), T(5), U(4), U(5),
4, 5, S(4), S(5), T(4), T(5), U(4), U(5),
4, 5, S(4), S(5), T(4), T(5), U(4), U(5),
4, 5, S(4), S(5), T(4), T(5), U(4), U(5),
6, 7, S(6), S(7), T(6), T(7), U(6), U(7),
6, 7, S(6), S(7), T(6), T(7), U(6), U(7),
6, 7, S(6), S(7), T(6), T(7), U(6), U(7),
6, 7, S(6), S(7), T(6), T(7), U(6), U(7),
6, 7, S(6), S(7), T(6), T(7), U(6), U(7),
6, 7, S(6), S(7), T(6), T(7), U(6), U(7),
6, 7, S(6), S(7), T(6), T(7), U(6), U(7),
6, 7, S(6), S(7), T(6), T(7), U(6), U(7) },
{ 0, 0, 0, 0, 2, 2, 2, 2, S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(2), T(2), T(2), T(2), U(0), U(0), U(0), U(0), U(2), U(2), U(2), U(2),
0, 0, 0, 0, 2, 2, 2, 2, S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(2), T(2), T(2), T(2), U(0), U(0), U(0), U(0), U(2), U(2), U(2), U(2),
0, 0, 0, 0, 2, 2, 2, 2, S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(2), T(2), T(2), T(2), U(0), U(0), U(0), U(0), U(2), U(2), U(2), U(2),
0, 0, 0, 0, 2, 2, 2, 2, S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(2), T(2), T(2), T(2), U(0), U(0), U(0), U(0), U(2), U(2), U(2), U(2),
0, 0, 0, 0, 2, 2, 2, 2, S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(2), T(2), T(2), T(2), U(0), U(0), U(0), U(0), U(2), U(2), U(2), U(2),
0, 0, 0, 0, 2, 2, 2, 2, S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(2), T(2), T(2), T(2), U(0), U(0), U(0), U(0), U(2), U(2), U(2), U(2),
0, 0, 0, 0, 2, 2, 2, 2, S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(2), T(2), T(2), T(2), U(0), U(0), U(0), U(0), U(2), U(2), U(2), U(2),
0, 0, 0, 0, 2, 2, 2, 2, S(0), S(0), S(0), S(0), S(2), S(2), S(2), S(2),
T(0), T(0), T(0), T(0), T(2), T(2), T(2), T(2), U(0), U(0), U(0), U(0), U(2), U(2), U(2), U(2),
4, 4, 4, 4, S(4), S(4), S(4), S(4),
T(4), T(4), T(4), T(4), U(4), U(4), U(4), U(4),
4, 4, 4, 4, S(4), S(4), S(4), S(4),
T(4), T(4), T(4), T(4), U(4), U(4), U(4), U(4),
4, 4, 4, 4, S(4), S(4), S(4), S(4),
T(4), T(4), T(4), T(4), U(4), U(4), U(4), U(4),
4, 4, 4, 4, S(4), S(4), S(4), S(4),
T(4), T(4), T(4), T(4), U(4), U(4), U(4), U(4),
6, 6, 6, 6, S(6), S(6), S(6), S(6),
T(6), T(6), T(6), T(6), U(6), U(6), U(6), U(6),
6, 6, 6, 6, S(6), S(6), S(6), S(6),
T(6), T(6), T(6), T(6), U(6), U(6), U(6), U(6),
6, 6, 6, 6, S(6), S(6), S(6), S(6),
T(6), T(6), T(6), T(6), U(6), U(6), U(6), U(6),
6, 6, 6, 6, S(6), S(6), S(6), S(6),
T(6), T(6), T(6), T(6), U(6), U(6), U(6), U(6) },
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0), S(0),
T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0), T(0),
U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0), U(0),
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
T(4), T(4), T(4), T(4), T(4), T(4), T(4), T(4),
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6),
T(6), T(6), T(6), T(6), T(6), T(6), T(6), T(6) },
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6,
6, 6, 6, 6, 6, 6, 6, 6 },
};
#undef U
#undef T
#undef S

393
vp9/common/vp9_blockd.h
View File

@ -16,9 +16,9 @@ void vpx_log(const char *format, ...);
#include "./vpx_config.h"
#include "vpx_scale/yv12config.h"
#include "vp9/common/vp9_convolve.h"
#include "vp9/common/vp9_mv.h"
#include "vp9/common/vp9_treecoder.h"
#include "vp9/common/vp9_subpixel.h"
#include "vpx_ports/mem.h"
#include "vp9/common/vp9_common.h"
@ -47,27 +47,13 @@ void vpx_log(const char *format, ...);
#define MAX_MV_REFS 9
#define MAX_MV_REF_CANDIDATES 4
#if CONFIG_DWTDCTHYBRID
#define DWT_MAX_LENGTH 64
#define DWT_TYPE 26 // 26/53/97
#define DWT_PRECISION_BITS 2
#define DWT_PRECISION_RND ((1 << DWT_PRECISION_BITS) / 2)
#define DWTDCT16X16 0
#define DWTDCT16X16_LEAN 1
#define DWTDCT8X8 2
#define DWTDCT_TYPE DWTDCT16X16_LEAN
#endif
typedef struct {
int r, c;
} POS;
typedef enum PlaneType {
PLANE_TYPE_Y_NO_DC = 0,
PLANE_TYPE_Y2,
PLANE_TYPE_UV,
typedef enum {
PLANE_TYPE_Y_WITH_DC,
PLANE_TYPE_UV,
} PLANE_TYPE;
typedef char ENTROPY_CONTEXT;
@ -75,10 +61,9 @@ typedef struct {
ENTROPY_CONTEXT y1[4];
ENTROPY_CONTEXT u[2];
ENTROPY_CONTEXT v[2];
ENTROPY_CONTEXT y2;
} ENTROPY_CONTEXT_PLANES;
#define VP9_COMBINEENTROPYCONTEXTS( Dest, A, B) \
#define VP9_COMBINEENTROPYCONTEXTS(Dest, A, B) \
Dest = ((A)!=0) + ((B)!=0);
typedef enum {
@ -86,8 +71,7 @@ typedef enum {
INTER_FRAME = 1
} FRAME_TYPE;
typedef enum
{
typedef enum {
#if CONFIG_ENABLE_6TAP
SIXTAP,
#endif
@ -98,8 +82,7 @@ typedef enum
SWITCHABLE /* should be the last one */
} INTERPOLATIONFILTERTYPE;
typedef enum
{
typedef enum {
DC_PRED, /* average of above and left pixels */
V_PRED, /* vertical prediction */
H_PRED, /* horizontal prediction */
@ -125,10 +108,8 @@ typedef enum {
SEG_LVL_ALT_Q = 0, // Use alternate Quantizer ....
SEG_LVL_ALT_LF = 1, // Use alternate loop filter value...
SEG_LVL_REF_FRAME = 2, // Optional Segment reference frame
SEG_LVL_MODE = 3, // Optional Segment mode
SEG_LVL_EOB = 4, // EOB end stop marker.
SEG_LVL_TRANSFORM = 5, // Block transform size.
SEG_LVL_MAX = 6 // Number of MB level features supported
SEG_LVL_SKIP = 3, // Optional Segment (0,0) + skip mode
SEG_LVL_MAX = 4 // Number of MB level features supported
} SEG_LVL_FEATURES;
// Segment level features.
@ -155,10 +136,7 @@ typedef enum {
#define VP9_MVREFS (1 + SPLITMV - NEARESTMV)
#if CONFIG_LOSSLESS
#define WHT_UPSCALE_FACTOR 3
#define Y2_WHT_UPSCALE_FACTOR 2
#endif
#define WHT_UPSCALE_FACTOR 2
typedef enum {
B_DC_PRED, /* average of above and left pixels */
@ -219,10 +197,7 @@ union b_mode_info {
B_PREDICTION_MODE context;
#endif
} as_mode;
struct {
int_mv first;
int_mv second;
} as_mv;
int_mv as_mv[2]; // first, second inter predictor motion vectors
};
typedef enum {
@ -274,6 +249,9 @@ typedef struct {
INTERPOLATIONFILTERTYPE interp_filter;
BLOCK_SIZE_TYPE sb_type;
#if CONFIG_CODE_NONZEROCOUNT
uint16_t nzcs[256+64*2];
#endif
} MB_MODE_INFO;
typedef struct {
@ -298,36 +276,44 @@ typedef struct blockd {
int dst;
int dst_stride;
int eob;
union b_mode_info bmi;
} BLOCKD;
typedef struct superblockd {
/* 32x32 Y and 16x16 U/V. No 2nd order transform yet. */
DECLARE_ALIGNED(16, int16_t, diff[32*32+16*16*2]);
DECLARE_ALIGNED(16, int16_t, qcoeff[32*32+16*16*2]);
DECLARE_ALIGNED(16, int16_t, dqcoeff[32*32+16*16*2]);
} SUPERBLOCKD;
struct scale_factors {
int x_num;
int x_den;
int x_offset_q4;
int x_step_q4;
int y_num;
int y_den;
int y_offset_q4;
int y_step_q4;
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
convolve_fn_t predict[2][2][8]; // horiz, vert, weight (0 - 7)
#else
convolve_fn_t predict[2][2][2]; // horiz, vert, avg
#endif
};
typedef struct macroblockd {
DECLARE_ALIGNED(16, int16_t, diff[400]); /* from idct diff */
DECLARE_ALIGNED(16, uint8_t, predictor[384]);
DECLARE_ALIGNED(16, int16_t, qcoeff[400]);
DECLARE_ALIGNED(16, int16_t, dqcoeff[400]);
DECLARE_ALIGNED(16, uint16_t, eobs[25]);
DECLARE_ALIGNED(16, int16_t, diff[64*64+32*32*2]); /* from idct diff */
DECLARE_ALIGNED(16, uint8_t, predictor[384]); // unused for superblocks
DECLARE_ALIGNED(16, int16_t, qcoeff[64*64+32*32*2]);
DECLARE_ALIGNED(16, int16_t, dqcoeff[64*64+32*32*2]);
DECLARE_ALIGNED(16, uint16_t, eobs[256+64*2]);
#if CONFIG_CODE_NONZEROCOUNT
DECLARE_ALIGNED(16, uint16_t, nzcs[256+64*2]);
#endif
SUPERBLOCKD sb_coeff_data;
/* 16 Y blocks, 4 U, 4 V, 1 DC 2nd order block, each with 16 entries. */
BLOCKD block[25];
/* 16 Y blocks, 4 U, 4 V, each with 16 entries. */
BLOCKD block[24];
int fullpixel_mask;
YV12_BUFFER_CONFIG pre; /* Filtered copy of previous frame reconstruction */
struct {
uint8_t *y_buffer, *u_buffer, *v_buffer;
} second_pre;
YV12_BUFFER_CONFIG second_pre;
YV12_BUFFER_CONFIG dst;
struct scale_factors scale_factor[2];
struct scale_factors scale_factor_uv[2];
MODE_INFO *prev_mode_info_context;
MODE_INFO *mode_info_context;
@ -337,8 +323,9 @@ typedef struct macroblockd {
int up_available;
int left_available;
int right_available;
/* Y,U,V,Y2 */
/* Y,U,V */
ENTROPY_CONTEXT_PLANES *above_context;
ENTROPY_CONTEXT_PLANES *left_context;
@ -359,6 +346,7 @@ typedef struct macroblockd {
// Probability Tree used to code Segment number
vp9_prob mb_segment_tree_probs[MB_FEATURE_TREE_PROBS];
vp9_prob mb_segment_mispred_tree_probs[MAX_MB_SEGMENTS];
#if CONFIG_NEW_MVREF
vp9_prob mb_mv_ref_probs[MAX_REF_FRAMES][MAX_MV_REF_CANDIDATES-1];
@ -387,21 +375,20 @@ typedef struct macroblockd {
unsigned int frames_since_golden;
unsigned int frames_till_alt_ref_frame;
int lossless;
/* Inverse transform function pointers. */
void (*inv_xform4x4_1_x8)(int16_t *input, int16_t *output, int pitch);
void (*inv_xform4x4_x8)(int16_t *input, int16_t *output, int pitch);
void (*inv_walsh4x4_1)(int16_t *in, int16_t *out);
void (*inv_walsh4x4_lossless)(int16_t *in, int16_t *out);
void (*inv_txm4x4_1)(int16_t *input, int16_t *output, int pitch);
void (*inv_txm4x4)(int16_t *input, int16_t *output, int pitch);
void (*itxm_add)(int16_t *input, const int16_t *dq,
uint8_t *pred, uint8_t *output, int pitch, int stride, int eob);
void (*itxm_add_y_block)(int16_t *q, const int16_t *dq,
uint8_t *pre, uint8_t *dst, int stride, struct macroblockd *xd);
void (*itxm_add_uv_block)(int16_t *q, const int16_t *dq,
uint8_t *pre, uint8_t *dst_u, uint8_t *dst_v, int stride,
struct macroblockd *xd);
struct subpix_fn_table subpix;
vp9_subpix_fn_t subpixel_predict4x4;
vp9_subpix_fn_t subpixel_predict8x4;
vp9_subpix_fn_t subpixel_predict8x8;
vp9_subpix_fn_t subpixel_predict16x16;
vp9_subpix_fn_t subpixel_predict_avg4x4;
vp9_subpix_fn_t subpixel_predict_avg8x4;
vp9_subpix_fn_t subpixel_predict_avg8x8;
vp9_subpix_fn_t subpixel_predict_avg16x16;
int allow_high_precision_mv;
int corrupted;
@ -412,74 +399,46 @@ typedef struct macroblockd {
} MACROBLOCKD;
#define ACTIVE_HT 110 // quantization stepsize threshold
#define ACTIVE_HT 110 // quantization stepsize threshold
#define ACTIVE_HT8 300
#define ACTIVE_HT8 300
#define ACTIVE_HT16 300
// convert MB_PREDICTION_MODE to B_PREDICTION_MODE
static B_PREDICTION_MODE pred_mode_conv(MB_PREDICTION_MODE mode) {
B_PREDICTION_MODE b_mode;
switch (mode) {
case DC_PRED:
b_mode = B_DC_PRED;
break;
case V_PRED:
b_mode = B_VE_PRED;
break;
case H_PRED:
b_mode = B_HE_PRED;
break;
case TM_PRED:
b_mode = B_TM_PRED;
break;
case D45_PRED:
b_mode = B_LD_PRED;
break;
case D135_PRED:
b_mode = B_RD_PRED;
break;
case D117_PRED:
b_mode = B_VR_PRED;
break;
case D153_PRED:
b_mode = B_HD_PRED;
break;
case D27_PRED:
b_mode = B_HU_PRED;
break;
case D63_PRED:
b_mode = B_VL_PRED;
break;
default :
// for debug purpose, to be removed after full testing
assert(0);
break;
case DC_PRED: return B_DC_PRED;
case V_PRED: return B_VE_PRED;
case H_PRED: return B_HE_PRED;
case TM_PRED: return B_TM_PRED;
case D45_PRED: return B_LD_PRED;
case D135_PRED: return B_RD_PRED;
case D117_PRED: return B_VR_PRED;
case D153_PRED: return B_HD_PRED;
case D27_PRED: return B_HU_PRED;
case D63_PRED: return B_VL_PRED;
default:
assert(0);
return B_MODE_COUNT; // Dummy value
}
return b_mode;
}
// transform mapping
static TX_TYPE txfm_map(B_PREDICTION_MODE bmode) {
// map transform type
TX_TYPE tx_type;
switch (bmode) {
case B_TM_PRED :
case B_RD_PRED :
tx_type = ADST_ADST;
break;
return ADST_ADST;
case B_VE_PRED :
case B_VR_PRED :
tx_type = ADST_DCT;
break;
return ADST_DCT;
case B_HE_PRED :
case B_HD_PRED :
case B_HU_PRED :
tx_type = DCT_ADST;
break;
return DCT_ADST;
#if CONFIG_NEWBINTRAMODES
case B_CONTEXT_PRED:
@ -487,33 +446,41 @@ static TX_TYPE txfm_map(B_PREDICTION_MODE bmode) {
break;
#endif
default :
tx_type = DCT_DCT;
break;
default:
return DCT_DCT;
}
return tx_type;
}
extern const uint8_t vp9_block2left[TX_SIZE_MAX_SB][25];
extern const uint8_t vp9_block2above[TX_SIZE_MAX_SB][25];
extern const uint8_t vp9_block2left[TX_SIZE_MAX_MB][24];
extern const uint8_t vp9_block2above[TX_SIZE_MAX_MB][24];
extern const uint8_t vp9_block2left_sb[TX_SIZE_MAX_SB][96];
extern const uint8_t vp9_block2above_sb[TX_SIZE_MAX_SB][96];
extern const uint8_t vp9_block2left_sb64[TX_SIZE_MAX_SB][384];
extern const uint8_t vp9_block2above_sb64[TX_SIZE_MAX_SB][384];
#define USE_ADST_FOR_I16X16_8X8 0
#define USE_ADST_FOR_I16X16_4X4 0
#define USE_ADST_FOR_I16X16_8X8 1
#define USE_ADST_FOR_I16X16_4X4 1
#define USE_ADST_FOR_I8X8_4X4 1
#define USE_ADST_PERIPHERY_ONLY 1
#define USE_ADST_FOR_SB 1
#define USE_ADST_FOR_REMOTE_EDGE 0
static TX_TYPE get_tx_type_4x4(const MACROBLOCKD *xd, const BLOCKD *b) {
static TX_TYPE get_tx_type_4x4(const MACROBLOCKD *xd, int ib) {
// TODO(debargha): explore different patterns for ADST usage when blocksize
// is smaller than the prediction size
TX_TYPE tx_type = DCT_DCT;
int ib = (int)(b - xd->block);
if (ib >= 16)
const BLOCK_SIZE_TYPE sb_type = xd->mode_info_context->mbmi.sb_type;
#if !USE_ADST_FOR_SB
if (sb_type)
return tx_type;
// TODO(rbultje, debargha): Explore ADST usage for superblocks
if (xd->mode_info_context->mbmi.sb_type)
#endif
if (ib >= (16 << (2 * sb_type))) // no chroma adst
return tx_type;
if (xd->lossless)
return DCT_DCT;
if (xd->mode_info_context->mbmi.mode == B_PRED &&
xd->q_index < ACTIVE_HT) {
const BLOCKD *b = &xd->block[ib];
tx_type = txfm_map(
#if CONFIG_NEWBINTRAMODES
b->bmi.as_mode.first == B_CONTEXT_PRED ? b->bmi.as_mode.context :
@ -521,16 +488,32 @@ static TX_TYPE get_tx_type_4x4(const MACROBLOCKD *xd, const BLOCKD *b) {
b->bmi.as_mode.first);
} else if (xd->mode_info_context->mbmi.mode == I8X8_PRED &&
xd->q_index < ACTIVE_HT) {
const BLOCKD *b = &xd->block[ib];
const int ic = (ib & 10);
#if USE_ADST_FOR_I8X8_4X4
#if USE_ADST_PERIPHERY_ONLY
// Use ADST for periphery blocks only
int ic = (ib & 10);
const int inner = ib & 5;
b += ic - ib;
tx_type = (ic != 10) ?
txfm_map(pred_mode_conv((MB_PREDICTION_MODE)b->bmi.as_mode.first)) :
DCT_DCT;
tx_type = txfm_map(pred_mode_conv(
(MB_PREDICTION_MODE)b->bmi.as_mode.first));
#if USE_ADST_FOR_REMOTE_EDGE
if (inner == 5)
tx_type = DCT_DCT;
#else
if (inner == 1) {
if (tx_type == ADST_ADST) tx_type = ADST_DCT;
else if (tx_type == DCT_ADST) tx_type = DCT_DCT;
} else if (inner == 4) {
if (tx_type == ADST_ADST) tx_type = DCT_ADST;
else if (tx_type == ADST_DCT) tx_type = DCT_DCT;
} else if (inner == 5) {
tx_type = DCT_DCT;
}
#endif
#else
// Use ADST
b += ic - ib;
tx_type = txfm_map(pred_mode_conv(
(MB_PREDICTION_MODE)b->bmi.as_mode.first));
#endif
@ -542,9 +525,22 @@ static TX_TYPE get_tx_type_4x4(const MACROBLOCKD *xd, const BLOCKD *b) {
xd->q_index < ACTIVE_HT) {
#if USE_ADST_FOR_I16X16_4X4
#if USE_ADST_PERIPHERY_ONLY
// Use ADST for periphery blocks only
tx_type = (ib < 4 || ((ib & 3) == 0)) ?
txfm_map(pred_mode_conv(xd->mode_info_context->mbmi.mode)) : DCT_DCT;
const int hmax = 4 << sb_type;
tx_type = txfm_map(pred_mode_conv(xd->mode_info_context->mbmi.mode));
#if USE_ADST_FOR_REMOTE_EDGE
if ((ib & (hmax - 1)) != 0 && ib >= hmax)
tx_type = DCT_DCT;
#else
if (ib >= 1 && ib < hmax) {
if (tx_type == ADST_ADST) tx_type = ADST_DCT;
else if (tx_type == DCT_ADST) tx_type = DCT_DCT;
} else if (ib >= 1 && (ib & (hmax - 1)) == 0) {
if (tx_type == ADST_ADST) tx_type = DCT_ADST;
else if (tx_type == ADST_DCT) tx_type = DCT_DCT;
} else if (ib != 0) {
tx_type = DCT_DCT;
}
#endif
#else
// Use ADST
tx_type = txfm_map(pred_mode_conv(xd->mode_info_context->mbmi.mode));
@ -557,29 +553,44 @@ static TX_TYPE get_tx_type_4x4(const MACROBLOCKD *xd, const BLOCKD *b) {
return tx_type;
}
static TX_TYPE get_tx_type_8x8(const MACROBLOCKD *xd, const BLOCKD *b) {
static TX_TYPE get_tx_type_8x8(const MACROBLOCKD *xd, int ib) {
// TODO(debargha): explore different patterns for ADST usage when blocksize
// is smaller than the prediction size
TX_TYPE tx_type = DCT_DCT;
int ib = (int)(b - xd->block);
if (ib >= 16)
const BLOCK_SIZE_TYPE sb_type = xd->mode_info_context->mbmi.sb_type;
#if !USE_ADST_FOR_SB
if (sb_type)
return tx_type;
// TODO(rbultje, debargha): Explore ADST usage for superblocks
if (xd->mode_info_context->mbmi.sb_type)
#endif
if (ib >= (16 << (2 * sb_type))) // no chroma adst
return tx_type;
if (xd->mode_info_context->mbmi.mode == I8X8_PRED &&
xd->q_index < ACTIVE_HT8) {
const BLOCKD *b = &xd->block[ib];
// TODO(rbultje): MB_PREDICTION_MODE / B_PREDICTION_MODE should be merged
// or the relationship otherwise modified to address this type conversion.
tx_type = txfm_map(pred_mode_conv(
(MB_PREDICTION_MODE)b->bmi.as_mode.first));
} else if (xd->mode_info_context->mbmi.mode < I8X8_PRED &&
xd->q_index < ACTIVE_HT8) {
#if USE_ADST_FOR_I8X8_4X4
#if USE_ADST_FOR_I16X16_8X8
#if USE_ADST_PERIPHERY_ONLY
// Use ADST for periphery blocks only
tx_type = (ib != 10) ?
txfm_map(pred_mode_conv(xd->mode_info_context->mbmi.mode)) : DCT_DCT;
const int hmax = 4 << sb_type;
tx_type = txfm_map(pred_mode_conv(xd->mode_info_context->mbmi.mode));
#if USE_ADST_FOR_REMOTE_EDGE
if ((ib & (hmax - 1)) != 0 && ib >= hmax)
tx_type = DCT_DCT;
#else
if (ib >= 1 && ib < hmax) {
if (tx_type == ADST_ADST) tx_type = ADST_DCT;
else if (tx_type == DCT_ADST) tx_type = DCT_DCT;
} else if (ib >= 1 && (ib & (hmax - 1)) == 0) {
if (tx_type == ADST_ADST) tx_type = DCT_ADST;
else if (tx_type == ADST_DCT) tx_type = DCT_DCT;
} else if (ib != 0) {
tx_type = DCT_DCT;
}
#endif
#else
// Use ADST
tx_type = txfm_map(pred_mode_conv(xd->mode_info_context->mbmi.mode));
@ -592,63 +603,73 @@ static TX_TYPE get_tx_type_8x8(const MACROBLOCKD *xd, const BLOCKD *b) {
return tx_type;
}
static TX_TYPE get_tx_type_16x16(const MACROBLOCKD *xd, const BLOCKD *b) {
static TX_TYPE get_tx_type_16x16(const MACROBLOCKD *xd, int ib) {
TX_TYPE tx_type = DCT_DCT;
int ib = (int)(b - xd->block);
if (ib >= 16)
const BLOCK_SIZE_TYPE sb_type = xd->mode_info_context->mbmi.sb_type;
#if !USE_ADST_FOR_SB
if (sb_type)
return tx_type;
// TODO(rbultje, debargha): Explore ADST usage for superblocks
if (xd->mode_info_context->mbmi.sb_type)
#endif
if (ib >= (16 << (2 * sb_type)))
return tx_type;
if (xd->mode_info_context->mbmi.mode < I8X8_PRED &&
xd->q_index < ACTIVE_HT16) {
tx_type = txfm_map(pred_mode_conv(xd->mode_info_context->mbmi.mode));
#if USE_ADST_PERIPHERY_ONLY
if (sb_type) {
const int hmax = 4 << sb_type;
#if USE_ADST_FOR_REMOTE_EDGE
if ((ib & (hmax - 1)) != 0 && ib >= hmax)
tx_type = DCT_DCT;
#else
if (ib >= 1 && ib < hmax) {
if (tx_type == ADST_ADST) tx_type = ADST_DCT;
else if (tx_type == DCT_ADST) tx_type = DCT_DCT;
} else if (ib >= 1 && (ib & (hmax - 1)) == 0) {
if (tx_type == ADST_ADST) tx_type = DCT_ADST;
else if (tx_type == ADST_DCT) tx_type = DCT_DCT;
} else if (ib != 0) {
tx_type = DCT_DCT;
}
#endif
}
#endif
}
return tx_type;
}
static TX_TYPE get_tx_type(const MACROBLOCKD *xd, const BLOCKD *b) {
TX_TYPE tx_type = DCT_DCT;
int ib = (int)(b - xd->block);
if (ib >= 16)
return tx_type;
if (xd->mode_info_context->mbmi.txfm_size == TX_16X16) {
tx_type = get_tx_type_16x16(xd, b);
}
if (xd->mode_info_context->mbmi.txfm_size == TX_8X8) {
ib = (ib & 8) + ((ib & 4) >> 1);
tx_type = get_tx_type_8x8(xd, &xd->block[ib]);
}
if (xd->mode_info_context->mbmi.txfm_size == TX_4X4) {
tx_type = get_tx_type_4x4(xd, b);
}
return tx_type;
}
static int get_2nd_order_usage(const MACROBLOCKD *xd) {
int has_2nd_order = (xd->mode_info_context->mbmi.mode != SPLITMV &&
xd->mode_info_context->mbmi.mode != I8X8_PRED &&
xd->mode_info_context->mbmi.mode != B_PRED &&
xd->mode_info_context->mbmi.txfm_size != TX_16X16);
if (has_2nd_order)
has_2nd_order = (get_tx_type(xd, xd->block) == DCT_DCT);
return has_2nd_order;
}
extern void vp9_build_block_doffsets(MACROBLOCKD *xd);
extern void vp9_setup_block_dptrs(MACROBLOCKD *xd);
void vp9_build_block_doffsets(MACROBLOCKD *xd);
void vp9_setup_block_dptrs(MACROBLOCKD *xd);
static void update_blockd_bmi(MACROBLOCKD *xd) {
int i;
int is_4x4;
is_4x4 = (xd->mode_info_context->mbmi.mode == SPLITMV) ||
(xd->mode_info_context->mbmi.mode == I8X8_PRED) ||
(xd->mode_info_context->mbmi.mode == B_PRED);
const MB_PREDICTION_MODE mode = xd->mode_info_context->mbmi.mode;
if (is_4x4) {
for (i = 0; i < 16; i++) {
if (mode == SPLITMV || mode == I8X8_PRED || mode == B_PRED) {
int i;
for (i = 0; i < 16; i++)
xd->block[i].bmi = xd->mode_info_context->bmi[i];
}
}
}
static TX_SIZE get_uv_tx_size(const MACROBLOCKD *xd) {
TX_SIZE tx_size_uv;
if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB64X64) {
tx_size_uv = xd->mode_info_context->mbmi.txfm_size;
} else if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB32X32) {
if (xd->mode_info_context->mbmi.txfm_size == TX_32X32)
tx_size_uv = TX_16X16;
else
tx_size_uv = xd->mode_info_context->mbmi.txfm_size;
} else {
if (xd->mode_info_context->mbmi.txfm_size == TX_16X16)
tx_size_uv = TX_8X8;
else if (xd->mode_info_context->mbmi.txfm_size == TX_8X8 &&
(xd->mode_info_context->mbmi.mode == I8X8_PRED ||
xd->mode_info_context->mbmi.mode == SPLITMV))
tx_size_uv = TX_4X4;
else
tx_size_uv = xd->mode_info_context->mbmi.txfm_size;
}
return tx_size_uv;
}
#endif // VP9_COMMON_VP9_BLOCKD_H_

21
vp9/common/vp9_coefupdateprobs.h
View File

@ -9,12 +9,25 @@
*/
#ifndef VP9_COMMON_VP9_COEFUPDATEPROBS_H_
#define VP9_COMMON_VP9_COEFUPDATEPROBS_H__
#define VP9_COMMON_VP9_COEFUPDATEPROBS_H_
/* Update probabilities for the nodes in the token entropy tree.
Generated file included by vp9_entropy.c */
#define COEF_UPDATE_PROB 252
#define COEF_UPDATE_PROB_8X8 252
#define COEF_UPDATE_PROB_16X16 252
static const vp9_prob vp9_coef_update_prob[ENTROPY_NODES] = {
252, 252, 252, 252, 252, 252, 252, 252, 252, 252, 252
};
#if CONFIG_CODE_NONZEROCOUNT
#define NZC_UPDATE_PROB_4X4 252
#define NZC_UPDATE_PROB_8X8 252
#define NZC_UPDATE_PROB_16X16 252
#define NZC_UPDATE_PROB_32X32 252
#define NZC_UPDATE_PROB_PCAT 252
#endif
#if CONFIG_MODELCOEFPROB
#define COEF_MODEL_UPDATE_PROB 16
#endif
#endif // VP9_COMMON_VP9_COEFUPDATEPROBS_H__

39
vp9/common/vp9_common.h
View File

@ -11,10 +11,11 @@
#ifndef VP9_COMMON_VP9_COMMON_H_
#define VP9_COMMON_VP9_COMMON_H_
#include <assert.h>
#include "vpx_config.h"
/* Interface header for common constant data structures and lookup tables */
#include <assert.h>
#include "./vpx_config.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx/vpx_integer.h"
@ -24,26 +25,34 @@
#define MIN(x, y) (((x) < (y)) ? (x) : (y))
#define MAX(x, y) (((x) > (y)) ? (x) : (y))
/* Only need this for fixed-size arrays, for structs just assign. */
#define ROUND_POWER_OF_TWO(value, n) (((value) + (1 << ((n) - 1))) >> (n))
#define vp9_copy(Dest, Src) { \
assert(sizeof(Dest) == sizeof(Src)); \
vpx_memcpy(Dest, Src, sizeof(Src)); \
/* If we don't want to use ROUND_POWER_OF_TWO macro
static INLINE int16_t round_power_of_two(int16_t value, int n) {
return (value + (1 << (n - 1))) >> n;
}*/
// Only need this for fixed-size arrays, for structs just assign.
#define vp9_copy(dest, src) { \
assert(sizeof(dest) == sizeof(src)); \
vpx_memcpy(dest, src, sizeof(src)); \
}
/* Use this for variably-sized arrays. */
#define vp9_copy_array(Dest, Src, N) { \
assert(sizeof(*Dest) == sizeof(*Src)); \
vpx_memcpy(Dest, Src, N * sizeof(*Src)); \
// Use this for variably-sized arrays.
#define vp9_copy_array(dest, src, n) { \
assert(sizeof(*dest) == sizeof(*src)); \
vpx_memcpy(dest, src, n * sizeof(*src)); \
}
#define vp9_zero(Dest) vpx_memset(&Dest, 0, sizeof(Dest));
#define vp9_zero(dest) vpx_memset(&dest, 0, sizeof(dest));
#define vp9_zero_array(dest, n) vpx_memset(dest, 0, n * sizeof(*dest));
#define vp9_zero_array(Dest, N) vpx_memset(Dest, 0, N * sizeof(*Dest));
static __inline uint8_t clip_pixel(int val) {
static INLINE uint8_t clip_pixel(int val) {
return (val > 255) ? 255u : (val < 0) ? 0u : val;
}
static INLINE int clamp(int value, int low, int high) {
return value < low ? low : (value > high ? high : value);
}
#endif // VP9_COMMON_VP9_COMMON_H_

850
vp9/common/vp9_convolve.c Normal file
View File

@ -0,0 +1,850 @@
/*
* Copyright (c) 2013 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 "vp9/common/vp9_convolve.h"
#include <assert.h>
#include "./vpx_config.h"
#include "./vp9_rtcd.h"
#include "vp9/common/vp9_common.h"
#include "vpx/vpx_integer.h"
#include "vpx_ports/mem.h"
#define VP9_FILTER_WEIGHT 128
#define VP9_FILTER_SHIFT 7
/* Assume a bank of 16 filters to choose from. There are two implementations
* for filter wrapping behavior, since we want to be able to pick which filter
* to start with. We could either:
*
* 1) make filter_ a pointer to the base of the filter array, and then add an
* additional offset parameter, to choose the starting filter.
* 2) use a pointer to 2 periods worth of filters, so that even if the original
* phase offset is at 15/16, we'll have valid data to read. The filter
* tables become [32][8], and the second half is duplicated.
* 3) fix the alignment of the filter tables, so that we know the 0/16 is
* always 256 byte aligned.
*
* Implementations 2 and 3 are likely preferable, as they avoid an extra 2
* parameters, and switching between them is trivial, with the
* ALIGN_FILTERS_256 macro, below.
*/
#define ALIGN_FILTERS_256 1
static void convolve_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x0, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h, int taps) {
int x, y, k, sum;
const int16_t *filter_x_base = filter_x0;
#if ALIGN_FILTERS_256
filter_x_base = (const int16_t *)(((intptr_t)filter_x0) & ~(intptr_t)0xff);
#endif
/* Adjust base pointer address for this source line */
src -= taps / 2 - 1;
for (y = 0; y < h; ++y) {
/* Pointer to filter to use */
const int16_t *filter_x = filter_x0;
/* Initial phase offset */
int x0_q4 = (filter_x - filter_x_base) / taps;
int x_q4 = x0_q4;
for (x = 0; x < w; ++x) {
/* Per-pixel src offset */
int src_x = (x_q4 - x0_q4) >> 4;
for (sum = 0, k = 0; k < taps; ++k) {
sum += src[src_x + k] * filter_x[k];
}
sum += (VP9_FILTER_WEIGHT >> 1);
dst[x] = clip_pixel(sum >> VP9_FILTER_SHIFT);
/* Adjust source and filter to use for the next pixel */
x_q4 += x_step_q4;
filter_x = filter_x_base + (x_q4 & 0xf) * taps;
}
src += src_stride;
dst += dst_stride;
}
}
static void convolve_avg_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x0, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h, int taps) {
int x, y, k, sum;
const int16_t *filter_x_base = filter_x0;
#if ALIGN_FILTERS_256
filter_x_base = (const int16_t *)(((intptr_t)filter_x0) & ~(intptr_t)0xff);
#endif
/* Adjust base pointer address for this source line */
src -= taps / 2 - 1;
for (y = 0; y < h; ++y) {
/* Pointer to filter to use */
const int16_t *filter_x = filter_x0;
/* Initial phase offset */
int x0_q4 = (filter_x - filter_x_base) / taps;
int x_q4 = x0_q4;
for (x = 0; x < w; ++x) {
/* Per-pixel src offset */
int src_x = (x_q4 - x0_q4) >> 4;
for (sum = 0, k = 0; k < taps; ++k) {
sum += src[src_x + k] * filter_x[k];
}
sum += (VP9_FILTER_WEIGHT >> 1);
dst[x] = (dst[x] + clip_pixel(sum >> VP9_FILTER_SHIFT) + 1) >> 1;
/* Adjust source and filter to use for the next pixel */
x_q4 += x_step_q4;
filter_x = filter_x_base + (x_q4 & 0xf) * taps;
}
src += src_stride;
dst += dst_stride;
}
}
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
static inline uint8_t combine_qtr(uint8_t a, uint8_t b) {
return (((a) + (b) * 3 + 2) >> 2);
}
static inline uint8_t combine_3qtr(uint8_t a, uint8_t b) {
return (((a) * 3 + (b) + 2) >> 2);
}
static inline uint8_t combine_1by8(uint8_t a, uint8_t b) {
return (((a) * 1 + (b) * 7 + 4) >> 3);
}
static inline uint8_t combine_3by8(uint8_t a, uint8_t b) {
return (((a) * 3 + (b) * 5 + 4) >> 3);
}
static inline uint8_t combine_5by8(uint8_t a, uint8_t b) {
return (((a) * 5 + (b) * 3 + 4) >> 3);
}
static inline uint8_t combine_7by8(uint8_t a, uint8_t b) {
return (((a) * 7 + (b) * 1 + 4) >> 3);
}
// TODO(debargha): Implment with a separate weight parameter
static void convolve_wtd_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x0, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h, int taps,
uint8_t (*combine)(uint8_t a, uint8_t b)) {
int x, y, k, sum;
const int16_t *filter_x_base = filter_x0;
#if ALIGN_FILTERS_256
filter_x_base = (const int16_t *)(((intptr_t)filter_x0) & ~(intptr_t)0xff);
#endif
/* Adjust base pointer address for this source line */
src -= taps / 2 - 1;
for (y = 0; y < h; ++y) {
/* Pointer to filter to use */
const int16_t *filter_x = filter_x0;
/* Initial phase offset */
int x0_q4 = (filter_x - filter_x_base) / taps;
int x_q4 = x0_q4;
for (x = 0; x < w; ++x) {
/* Per-pixel src offset */
int src_x = (x_q4 - x0_q4) >> 4;
for (sum = 0, k = 0; k < taps; ++k) {
sum += src[src_x + k] * filter_x[k];
}
sum += (VP9_FILTER_WEIGHT >> 1);
dst[x] = combine(dst[x], clip_pixel(sum >> VP9_FILTER_SHIFT));
/* Adjust source and filter to use for the next pixel */
x_q4 += x_step_q4;
filter_x = filter_x_base + (x_q4 & 0xf) * taps;
}
src += src_stride;
dst += dst_stride;
}
}
#endif
static void convolve_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y0, int y_step_q4,
int w, int h, int taps) {
int x, y, k, sum;
const int16_t *filter_y_base = filter_y0;
#if ALIGN_FILTERS_256
filter_y_base = (const int16_t *)(((intptr_t)filter_y0) & ~(intptr_t)0xff);
#endif
/* Adjust base pointer address for this source column */
src -= src_stride * (taps / 2 - 1);
for (x = 0; x < w; ++x) {
/* Pointer to filter to use */
const int16_t *filter_y = filter_y0;
/* Initial phase offset */
int y0_q4 = (filter_y - filter_y_base) / taps;
int y_q4 = y0_q4;
for (y = 0; y < h; ++y) {
/* Per-pixel src offset */
int src_y = (y_q4 - y0_q4) >> 4;
for (sum = 0, k = 0; k < taps; ++k) {
sum += src[(src_y + k) * src_stride] * filter_y[k];
}
sum += (VP9_FILTER_WEIGHT >> 1);
dst[y * dst_stride] = clip_pixel(sum >> VP9_FILTER_SHIFT);
/* Adjust source and filter to use for the next pixel */
y_q4 += y_step_q4;
filter_y = filter_y_base + (y_q4 & 0xf) * taps;
}
++src;
++dst;
}
}
static void convolve_avg_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y0, int y_step_q4,
int w, int h, int taps) {
int x, y, k, sum;
const int16_t *filter_y_base = filter_y0;
#if ALIGN_FILTERS_256
filter_y_base = (const int16_t *)(((intptr_t)filter_y0) & ~(intptr_t)0xff);
#endif
/* Adjust base pointer address for this source column */
src -= src_stride * (taps / 2 - 1);
for (x = 0; x < w; ++x) {
/* Pointer to filter to use */
const int16_t *filter_y = filter_y0;
/* Initial phase offset */
int y0_q4 = (filter_y - filter_y_base) / taps;
int y_q4 = y0_q4;
for (y = 0; y < h; ++y) {
/* Per-pixel src offset */
int src_y = (y_q4 - y0_q4) >> 4;
for (sum = 0, k = 0; k < taps; ++k) {
sum += src[(src_y + k) * src_stride] * filter_y[k];
}
sum += (VP9_FILTER_WEIGHT >> 1);
dst[y * dst_stride] =
(dst[y * dst_stride] + clip_pixel(sum >> VP9_FILTER_SHIFT) + 1) >> 1;
/* Adjust source and filter to use for the next pixel */
y_q4 += y_step_q4;
filter_y = filter_y_base + (y_q4 & 0xf) * taps;
}
++src;
++dst;
}
}
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
static void convolve_wtd_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y0, int y_step_q4,
int w, int h, int taps,
uint8_t (*combine)(uint8_t a, uint8_t b)) {
int x, y, k, sum;
const int16_t *filter_y_base = filter_y0;
#if ALIGN_FILTERS_256
filter_y_base = (const int16_t *)(((intptr_t)filter_y0) & ~(intptr_t)0xff);
#endif
/* Adjust base pointer address for this source column */
src -= src_stride * (taps / 2 - 1);
for (x = 0; x < w; ++x) {
/* Pointer to filter to use */
const int16_t *filter_y = filter_y0;
/* Initial phase offset */
int y0_q4 = (filter_y - filter_y_base) / taps;
int y_q4 = y0_q4;
for (y = 0; y < h; ++y) {
/* Per-pixel src offset */
int src_y = (y_q4 - y0_q4) >> 4;
for (sum = 0, k = 0; k < taps; ++k) {
sum += src[(src_y + k) * src_stride] * filter_y[k];
}
sum += (VP9_FILTER_WEIGHT >> 1);
dst[y * dst_stride] = combine(dst[y * dst_stride],
clip_pixel(sum >> VP9_FILTER_SHIFT));
/* Adjust source and filter to use for the next pixel */
y_q4 += y_step_q4;
filter_y = filter_y_base + (y_q4 & 0xf) * taps;
}
++src;
++dst;
}
}
#endif
static void convolve_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h, int taps) {
/* Fixed size intermediate buffer places limits on parameters.
* Maximum intermediate_height is 39, for y_step_q4 == 32,
* h == 16, taps == 8.
*/
uint8_t temp[16 * 39];
int intermediate_height = ((h * y_step_q4) >> 4) + taps - 1;
assert(w <= 16);
assert(h <= 16);
assert(taps <= 8);
assert(y_step_q4 <= 32);
if (intermediate_height < h)
intermediate_height = h;
convolve_horiz_c(src - src_stride * (taps / 2 - 1), src_stride,
temp, 16,
filter_x, x_step_q4, filter_y, y_step_q4,
w, intermediate_height, taps);
convolve_vert_c(temp + 16 * (taps / 2 - 1), 16, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, taps);
}
static void convolve_avg_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h, int taps) {
/* Fixed size intermediate buffer places limits on parameters.
* Maximum intermediate_height is 39, for y_step_q4 == 32,
* h == 16, taps == 8.
*/
uint8_t temp[16 * 39];
int intermediate_height = ((h * y_step_q4) >> 4) + taps - 1;
assert(w <= 16);
assert(h <= 16);
assert(taps <= 8);
assert(y_step_q4 <= 32);
if (intermediate_height < h)
intermediate_height = h;
convolve_horiz_c(src - src_stride * (taps / 2 - 1), src_stride,
temp, 16,
filter_x, x_step_q4, filter_y, y_step_q4,
w, intermediate_height, taps);
convolve_avg_vert_c(temp + 16 * (taps / 2 - 1), 16, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, taps);
}
void vp9_convolve8_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_horiz_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8);
}
void vp9_convolve8_avg_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_avg_horiz_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8);
}
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
void vp9_convolve8_1by8_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_horiz_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_1by8);
}
void vp9_convolve8_qtr_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_horiz_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_qtr);
}
void vp9_convolve8_3by8_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_horiz_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_3by8);
}
void vp9_convolve8_5by8_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_horiz_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_5by8);
}
void vp9_convolve8_3qtr_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_horiz_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_3qtr);
}
void vp9_convolve8_7by8_horiz_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_horiz_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_7by8);
}
#endif
void vp9_convolve8_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_vert_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8);
}
void vp9_convolve8_avg_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_avg_vert_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8);
}
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
void vp9_convolve8_1by8_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_vert_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_1by8);
}
void vp9_convolve8_qtr_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_vert_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_qtr);
}
void vp9_convolve8_3by8_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_vert_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_3by8);
}
void vp9_convolve8_5by8_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_vert_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_5by8);
}
void vp9_convolve8_3qtr_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_vert_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_3qtr);
}
void vp9_convolve8_7by8_vert_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_wtd_vert_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8, combine_7by8);
}
#endif
void vp9_convolve8_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
convolve_c(src, src_stride, dst, dst_stride,
filter_x, x_step_q4, filter_y, y_step_q4,
w, h, 8);
}
void vp9_convolve8_avg_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
/* Fixed size intermediate buffer places limits on parameters. */
DECLARE_ALIGNED_ARRAY(16, uint8_t, temp, 16 * 16);
assert(w <= 16);
assert(h <= 16);
vp9_convolve8(src, src_stride,
temp, 16,
filter_x, x_step_q4,
filter_y, y_step_q4,
w, h);
vp9_convolve_avg(temp, 16,
dst, dst_stride,
NULL, 0, /* These unused parameter should be removed! */
NULL, 0, /* These unused parameter should be removed! */
w, h);
}
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
void vp9_convolve8_1by8_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
/* Fixed size intermediate buffer places limits on parameters. */
DECLARE_ALIGNED_ARRAY(16, uint8_t, temp, 16 * 16);
assert(w <= 16);
assert(h <= 16);
vp9_convolve8(src, src_stride,
temp, 16,
filter_x, x_step_q4,
filter_y, y_step_q4,
w, h);
vp9_convolve_1by8(temp, 16,
dst, dst_stride,
NULL, 0, /* These unused parameter should be removed! */
NULL, 0, /* These unused parameter should be removed! */
w, h);
}
void vp9_convolve8_qtr_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
/* Fixed size intermediate buffer places limits on parameters. */
DECLARE_ALIGNED_ARRAY(16, uint8_t, temp, 16 * 16);
assert(w <= 16);
assert(h <= 16);
vp9_convolve8(src, src_stride,
temp, 16,
filter_x, x_step_q4,
filter_y, y_step_q4,
w, h);
vp9_convolve_qtr(temp, 16,
dst, dst_stride,
NULL, 0, /* These unused parameter should be removed! */
NULL, 0, /* These unused parameter should be removed! */
w, h);
}
void vp9_convolve8_3by8_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
/* Fixed size intermediate buffer places limits on parameters. */
DECLARE_ALIGNED_ARRAY(16, uint8_t, temp, 16 * 16);
assert(w <= 16);
assert(h <= 16);
vp9_convolve8(src, src_stride,
temp, 16,
filter_x, x_step_q4,
filter_y, y_step_q4,
w, h);
vp9_convolve_3by8(temp, 16,
dst, dst_stride,
NULL, 0, /* These unused parameter should be removed! */
NULL, 0, /* These unused parameter should be removed! */
w, h);
}
void vp9_convolve8_5by8_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
/* Fixed size intermediate buffer places limits on parameters. */
DECLARE_ALIGNED_ARRAY(16, uint8_t, temp, 16 * 16);
assert(w <= 16);
assert(h <= 16);
vp9_convolve8(src, src_stride,
temp, 16,
filter_x, x_step_q4,
filter_y, y_step_q4,
w, h);
vp9_convolve_5by8(temp, 16,
dst, dst_stride,
NULL, 0, /* These unused parameter should be removed! */
NULL, 0, /* These unused parameter should be removed! */
w, h);
}
void vp9_convolve8_3qtr_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
/* Fixed size intermediate buffer places limits on parameters. */
DECLARE_ALIGNED_ARRAY(16, uint8_t, temp, 16 * 16);
assert(w <= 16);
assert(h <= 16);
vp9_convolve8(src, src_stride,
temp, 16,
filter_x, x_step_q4,
filter_y, y_step_q4,
w, h);
vp9_convolve_3qtr(temp, 16,
dst, dst_stride,
NULL, 0, /* These unused parameter should be removed! */
NULL, 0, /* These unused parameter should be removed! */
w, h);
}
void vp9_convolve8_7by8_c(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h) {
/* Fixed size intermediate buffer places limits on parameters. */
DECLARE_ALIGNED_ARRAY(16, uint8_t, temp, 16 * 16);
assert(w <= 16);
assert(h <= 16);
vp9_convolve8(src, src_stride,
temp, 16,
filter_x, x_step_q4,
filter_y, y_step_q4,
w, h);
vp9_convolve_7by8(temp, 16,
dst, dst_stride,
NULL, 0, /* These unused parameter should be removed! */
NULL, 0, /* These unused parameter should be removed! */
w, h);
}
#endif
void vp9_convolve_copy(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h) {
if (w == 16 && h == 16) {
vp9_copy_mem16x16(src, src_stride, dst, dst_stride);
} else if (w == 8 && h == 8) {
vp9_copy_mem8x8(src, src_stride, dst, dst_stride);
} else if (w == 8 && h == 4) {
vp9_copy_mem8x4(src, src_stride, dst, dst_stride);
} else {
int r;
for (r = h; r > 0; --r) {
memcpy(dst, src, w);
src += src_stride;
dst += dst_stride;
}
}
}
void vp9_convolve_avg(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h) {
int x, y;
for (y = 0; y < h; ++y) {
for (x = 0; x < w; ++x) {
dst[x] = (dst[x] + src[x] + 1) >> 1;
}
src += src_stride;
dst += dst_stride;
}
}
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
void vp9_convolve_1by8(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h) {
int x, y;
for (y = 0; y < h; ++y) {
for (x = 0; x < w; ++x) {
dst[x] = combine_1by8(dst[x], src[x]);
}
src += src_stride;
dst += dst_stride;
}
}
void vp9_convolve_qtr(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h) {
int x, y;
for (y = 0; y < h; ++y) {
for (x = 0; x < w; ++x) {
dst[x] = combine_qtr(dst[x], src[x]);
}
src += src_stride;
dst += dst_stride;
}
}
void vp9_convolve_3by8(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h) {
int x, y;
for (y = 0; y < h; ++y) {
for (x = 0; x < w; ++x) {
dst[x] = combine_3by8(dst[x], src[x]);
}
src += src_stride;
dst += dst_stride;
}
}
void vp9_convolve_5by8(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h) {
int x, y;
for (y = 0; y < h; ++y) {
for (x = 0; x < w; ++x) {
dst[x] = combine_5by8(dst[x], src[x]);
}
src += src_stride;
dst += dst_stride;
}
}
void vp9_convolve_3qtr(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h) {
int x, y;
for (y = 0; y < h; ++y) {
for (x = 0; x < w; ++x) {
dst[x] = combine_3qtr(dst[x], src[x]);
}
src += src_stride;
dst += dst_stride;
}
}
void vp9_convolve_7by8(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int filter_x_stride,
const int16_t *filter_y, int filter_y_stride,
int w, int h) {
int x, y;
for (y = 0; y < h; ++y) {
for (x = 0; x < w; ++x) {
dst[x] = combine_7by8(dst[x], src[x]);
}
src += src_stride;
dst += dst_stride;
}
}
#endif

85
vp9/common/vp9_convolve.h Normal file
View File

@ -0,0 +1,85 @@
/*
* Copyright (c) 2013 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.
*/
#ifndef VP9_COMMON_CONVOLVE_H_
#define VP9_COMMON_CONVOLVE_H_
#include "./vpx_config.h"
#include "vpx/vpx_integer.h"
typedef void (*convolve_fn_t)(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
// Not a convolution, a block copy conforming to the convolution prototype
void vp9_convolve_copy(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
// Not a convolution, a block average conforming to the convolution prototype
void vp9_convolve_avg(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
// Not a convolution, a block wtd (1/8, 7/8) average for (dst, src)
void vp9_convolve_1by8(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
// Not a convolution, a block wtd (1/4, 3/4) average for (dst, src)
void vp9_convolve_qtr(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
// Not a convolution, a block wtd (3/8, 5/8) average for (dst, src)
void vp9_convolve_3by8(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
// Not a convolution, a block wtd (5/8, 3/8) average for (dst, src)
void vp9_convolve_5by8(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
// Not a convolution, a block wtd (3/4, 1/4) average for (dst, src)
void vp9_convolve_3qtr(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
// Not a convolution, a block wtd (7/8, 1/8) average for (dst, src)
void vp9_convolve_7by8(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int16_t *filter_x, int x_step_q4,
const int16_t *filter_y, int y_step_q4,
int w, int h);
#endif
struct subpix_fn_table {
const int16_t (*filter_x)[8];
const int16_t (*filter_y)[8];
};
#endif // VP9_COMMON_CONVOLVE_H_

8
vp9/common/vp9_debugmodes.c
View File

@ -9,6 +9,7 @@
*/
#include <stdio.h>
#include "vp9/common/vp9_blockd.h"
void vp9_print_modes_and_motion_vectors(MODE_INFO *mi, int rows, int cols,
@ -18,8 +19,7 @@ void vp9_print_modes_and_motion_vectors(MODE_INFO *mi, int rows, int cols,
int mb_index = 0;
FILE *mvs = fopen("mvs.stt", "a");
/* print out the macroblock Y modes */
mb_index = 0;
// Print out the macroblock Y modes
fprintf(mvs, "Mb Modes for Frame %d\n", frame);
for (mb_row = 0; mb_row < rows; mb_row++) {
@ -129,8 +129,8 @@ void vp9_print_modes_and_motion_vectors(MODE_INFO *mi, int rows, int cols,
mb_index = (b_row >> 2) * (cols + 1) + (b_col >> 2);
bindex = (b_row & 3) * 4 + (b_col & 3);
fprintf(mvs, "%3d:%-3d ",
mi[mb_index].bmi[bindex].as_mv.first.as_mv.row,
mi[mb_index].bmi[bindex].as_mv.first.as_mv.col);
mi[mb_index].bmi[bindex].as_mv[0].as_mv.row,
mi[mb_index].bmi[bindex].as_mv[0].as_mv.col);
}

2136
vp9/common/vp9_default_coef_probs.h
View File

File diff suppressed because it is too large Load Diff

4343
vp9/common/vp9_entropy.c
View File

File diff suppressed because it is too large Load Diff

225
vp9/common/vp9_entropy.h
View File

@ -15,7 +15,6 @@
#include "vp9/common/vp9_treecoder.h"
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_coefupdateprobs.h"
extern const int vp9_i8x8_block[4];
@ -31,10 +30,10 @@ extern const int vp9_i8x8_block[4];
#define DCT_VAL_CATEGORY3 7 /* 11-18 Extra Bits 3+1 */
#define DCT_VAL_CATEGORY4 8 /* 19-34 Extra Bits 4+1 */
#define DCT_VAL_CATEGORY5 9 /* 35-66 Extra Bits 5+1 */
#define DCT_VAL_CATEGORY6 10 /* 67+ Extra Bits 13+1 */
#define DCT_VAL_CATEGORY6 10 /* 67+ Extra Bits 14+1 */
#define DCT_EOB_TOKEN 11 /* EOB Extra Bits 0+0 */
#define MAX_ENTROPY_TOKENS 12
#define ENTROPY_NODES 11
#define MAX_ENTROPY_TOKENS 12
#define ENTROPY_NODES 11
#define EOSB_TOKEN 127 /* Not signalled, encoder only */
#define INTER_MODE_CONTEXTS 7
@ -59,31 +58,20 @@ extern vp9_extra_bit_struct vp9_extra_bits[12]; /* indexed by token value */
/* Coefficients are predicted via a 3-dimensional probability table. */
/* Outside dimension. 0 = Y no DC, 1 = Y2, 2 = UV, 3 = Y with DC */
#define BLOCK_TYPES_4X4 4
/* Outside dimension. 0 = Y with DC, 1 = UV */
#define BLOCK_TYPES 2
#define REF_TYPES 2 // intra=0, inter=1
#define BLOCK_TYPES_8X8 4
/* Middle dimension reflects the coefficient position within the transform. */
#define COEF_BANDS 6
#define BLOCK_TYPES_16X16 4
#define BLOCK_TYPES_32X32 4
/* Middle dimension is a coarsening of the coefficient's
position within the 4x4 DCT. */
#define COEF_BANDS 8
extern DECLARE_ALIGNED(16, const int, vp9_coef_bands_4x4[16]);
extern DECLARE_ALIGNED(64, const int, vp9_coef_bands_8x8[64]);
extern DECLARE_ALIGNED(16, const int, vp9_coef_bands_16x16[256]);
extern DECLARE_ALIGNED(16, const int, vp9_coef_bands_32x32[1024]);
/* Inside dimension is 3-valued measure of nearby complexity, that is,
the extent to which nearby coefficients are nonzero. For the first
coefficient (DC, unless block type is 0), we look at the (already encoded)
blocks above and to the left of the current block. The context index is
then the number (0,1,or 2) of these blocks having nonzero coefficients.
After decoding a coefficient, the measure is roughly the size of the
most recently decoded coefficient (0 for 0, 1 for 1, 2 for >1).
/* Inside dimension is measure of nearby complexity, that reflects the energy
of nearby coefficients are nonzero. For the first coefficient (DC, unless
block type is 0), we look at the (already encoded) blocks above and to the
left of the current block. The context index is then the number (0,1,or 2)
of these blocks having nonzero coefficients.
After decoding a coefficient, the measure is determined by the size of the
most recently decoded coefficient.
Note that the intuitive meaning of this measure changes as coefficients
are decoded, e.g., prior to the first token, a zero means that my neighbors
are empty while, after the first token, because of the use of end-of-block,
@ -94,21 +82,18 @@ extern DECLARE_ALIGNED(16, const int, vp9_coef_bands_32x32[1024]);
distinct bands). */
/*# define DC_TOKEN_CONTEXTS 3*/ /* 00, 0!0, !0!0 */
#define PREV_COEF_CONTEXTS 4
#define PREV_COEF_CONTEXTS 6
typedef unsigned int vp9_coeff_count[COEF_BANDS][PREV_COEF_CONTEXTS]
typedef unsigned int vp9_coeff_count[REF_TYPES][COEF_BANDS][PREV_COEF_CONTEXTS]
[MAX_ENTROPY_TOKENS];
typedef unsigned int vp9_coeff_stats[COEF_BANDS][PREV_COEF_CONTEXTS]
typedef unsigned int vp9_coeff_stats[REF_TYPES][COEF_BANDS][PREV_COEF_CONTEXTS]
[ENTROPY_NODES][2];
typedef vp9_prob vp9_coeff_probs[COEF_BANDS][PREV_COEF_CONTEXTS]
typedef vp9_prob vp9_coeff_probs[REF_TYPES][COEF_BANDS][PREV_COEF_CONTEXTS]
[ENTROPY_NODES];
#define SUBEXP_PARAM 4 /* Subexponential code parameter */
#define MODULUS_PARAM 13 /* Modulus parameter */
extern DECLARE_ALIGNED(16, const uint8_t,
vp9_prev_token_class[MAX_ENTROPY_TOKENS]);
struct VP9Common;
void vp9_default_coef_probs(struct VP9Common *);
extern DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_4x4[16]);
@ -117,38 +102,168 @@ extern DECLARE_ALIGNED(16, const int, vp9_col_scan_4x4[16]);
extern DECLARE_ALIGNED(16, const int, vp9_row_scan_4x4[16]);
extern DECLARE_ALIGNED(64, const int, vp9_default_zig_zag1d_8x8[64]);
extern DECLARE_ALIGNED(16, const int, vp9_col_scan_8x8[64]);
extern DECLARE_ALIGNED(16, const int, vp9_row_scan_8x8[64]);
extern DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_16x16[256]);
extern DECLARE_ALIGNED(16, const int, vp9_col_scan_16x16[256]);
extern DECLARE_ALIGNED(16, const int, vp9_row_scan_16x16[256]);
extern DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_32x32[1024]);
void vp9_coef_tree_initialize(void);
void vp9_adapt_coef_probs(struct VP9Common *);
static void vp9_reset_mb_tokens_context(MACROBLOCKD* const xd) {
static INLINE void vp9_reset_mb_tokens_context(MACROBLOCKD* const xd) {
/* Clear entropy contexts */
vpx_memset(xd->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
vpx_memset(xd->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
}
#if CONFIG_NEWCOEFCONTEXT
static INLINE void vp9_reset_sb_tokens_context(MACROBLOCKD* const xd) {
/* Clear entropy contexts */
vpx_memset(xd->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * 2);
vpx_memset(xd->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * 2);
}
#define MAX_NEIGHBORS 5
#define NEWCOEFCONTEXT_BAND_COND(b) ((b) >= 1)
void vp9_init_neighbors(void);
static INLINE void vp9_reset_sb64_tokens_context(MACROBLOCKD* const xd) {
/* Clear entropy contexts */
vpx_memset(xd->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * 4);
vpx_memset(xd->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * 4);
}
extern const int vp9_coef_bands8x8[64];
extern const int vp9_coef_bands4x4[16];
static int get_coef_band(const int *scan, TX_SIZE tx_size, int coef_index) {
if (tx_size == TX_4X4) {
return vp9_coef_bands4x4[scan[coef_index]];
} else {
const int pos = scan[coef_index];
const int sz = 1 << (2 + tx_size);
const int x = pos & (sz - 1), y = pos >> (2 + tx_size);
if (x >= 8 || y >= 8)
return 5;
else
return vp9_coef_bands8x8[y * 8 + x];
}
}
extern int vp9_get_coef_context(const int *scan, const int *neighbors,
int nb_pad, uint8_t *token_cache, int c, int l);
const int *vp9_get_coef_neighbors_handle(const int *scan, int *pad);
#if CONFIG_MODELCOEFPROB
#define COEFPROB_BITS 8
#define COEFPROB_MODELS (1 << COEFPROB_BITS)
// 2 => EOB and Zero nodes are unconstrained, rest are modeled
// 3 => EOB, Zero and One nodes are unconstrained, rest are modeled
#define UNCONSTRAINED_NODES 3 // Choose one of 2 or 3
// whether forward updates are model-based
#define MODEL_BASED_UPDATE 0
// if model-based how many nodes are unconstrained
#define UNCONSTRAINED_UPDATE_NODES 3
// whether backward updates are model-based
#define MODEL_BASED_ADAPT 0
#define UNCONSTRAINED_ADAPT_NODES 3
// whether to adjust the coef probs for key frames based on qindex
#define ADJUST_KF_COEF_PROBS 0
typedef vp9_prob vp9_coeff_probs_model[REF_TYPES][COEF_BANDS]
[PREV_COEF_CONTEXTS][2];
extern const vp9_prob vp9_modelcoefprobs[COEFPROB_MODELS][ENTROPY_NODES - 1];
void vp9_get_model_distribution(vp9_prob model, vp9_prob *tree_probs,
int b, int r);
void vp9_adjust_default_coef_probs(struct VP9Common *cm);
#endif // CONFIG_MODELCOEFPROB
#if CONFIG_CODE_NONZEROCOUNT
/* Alphabet for number of non-zero symbols in block */
#define NZC_0 0 /* Used for all blocks */
#define NZC_1 1 /* Used for all blocks */
#define NZC_2 2 /* Used for all blocks */
#define NZC_3TO4 3 /* Used for all blocks */
#define NZC_5TO8 4 /* Used for all blocks */
#define NZC_9TO16 5 /* Used for all blocks */
#define NZC_17TO32 6 /* Used for 8x8 and larger blocks */
#define NZC_33TO64 7 /* Used for 8x8 and larger blocks */
#define NZC_65TO128 8 /* Used for 16x16 and larger blocks */
#define NZC_129TO256 9 /* Used for 16x16 and larger blocks */
#define NZC_257TO512 10 /* Used for 32x32 and larger blocks */
#define NZC_513TO1024 11 /* Used for 32x32 and larger blocks */
/* Number of tokens for each block size */
#define NZC4X4_TOKENS 6
#define NZC8X8_TOKENS 8
#define NZC16X16_TOKENS 10
#define NZC32X32_TOKENS 12
/* Number of nodes for each block size */
#define NZC4X4_NODES 5
#define NZC8X8_NODES 7
#define NZC16X16_NODES 9
#define NZC32X32_NODES 11
/* Max number of tokens with extra bits */
#define NZC_TOKENS_EXTRA 9
/* Max number of extra bits */
#define NZC_BITS_EXTRA 9
/* Tokens without extra bits */
#define NZC_TOKENS_NOEXTRA (NZC32X32_TOKENS - NZC_TOKENS_EXTRA)
#define MAX_NZC_CONTEXTS 3
/* whether to update extra bit probabilities */
#define NZC_PCAT_UPDATE
/* nzc trees */
extern const vp9_tree_index vp9_nzc4x4_tree[];
extern const vp9_tree_index vp9_nzc8x8_tree[];
extern const vp9_tree_index vp9_nzc16x16_tree[];
extern const vp9_tree_index vp9_nzc32x32_tree[];
/* nzc encodings */
extern struct vp9_token_struct vp9_nzc4x4_encodings[NZC4X4_TOKENS];
extern struct vp9_token_struct vp9_nzc8x8_encodings[NZC8X8_TOKENS];
extern struct vp9_token_struct vp9_nzc16x16_encodings[NZC16X16_TOKENS];
extern struct vp9_token_struct vp9_nzc32x32_encodings[NZC32X32_TOKENS];
#define codenzc(x) (\
(x) <= 3 ? (x) : (x) <= 4 ? 3 : (x) <= 8 ? 4 : \
(x) <= 16 ? 5 : (x) <= 32 ? 6 : (x) <= 64 ? 7 :\
(x) <= 128 ? 8 : (x) <= 256 ? 9 : (x) <= 512 ? 10 : 11)
int vp9_get_nzc_context_y_sb64(struct VP9Common *cm, MODE_INFO *cur,
int mb_row, int mb_col, int block);
int vp9_get_nzc_context_y_sb32(struct VP9Common *cm, MODE_INFO *cur,
int mb_row, int mb_col, int block);
int vp9_get_nzc_context_y_mb16(struct VP9Common *cm, MODE_INFO *cur,
int mb_row, int mb_col, int block);
int vp9_get_nzc_context_uv_sb64(struct VP9Common *cm, MODE_INFO *cur,
int mb_row, int mb_col, int block);
int vp9_get_nzc_context_uv_sb32(struct VP9Common *cm, MODE_INFO *cur,
int mb_row, int mb_col, int block);
int vp9_get_nzc_context_uv_mb16(struct VP9Common *cm, MODE_INFO *cur,
int mb_row, int mb_col, int block);
int vp9_get_nzc_context(struct VP9Common *cm, MACROBLOCKD *xd, int block);
void vp9_update_nzc_counts(struct VP9Common *cm, MACROBLOCKD *xd,
int mb_row, int mb_col);
void vp9_adapt_nzc_probs(struct VP9Common *cm);
/* Extra bits array */
extern const int vp9_extranzcbits[NZC32X32_TOKENS];
/* Base nzc values */
extern const int vp9_basenzcvalue[NZC32X32_TOKENS];
#endif // CONFIG_CODE_NONZEROCOUNT
#include "vp9/common/vp9_coefupdateprobs.h"
const int *vp9_get_coef_neighbors_handle(const int *scan);
int vp9_get_coef_neighbor_context(const short int *qcoeff_ptr, int nodc,
const int *neigbor_handle, int rc);
extern DECLARE_ALIGNED(16, int, vp9_default_zig_zag1d_4x4_neighbors[
16 * MAX_NEIGHBORS]);
extern DECLARE_ALIGNED(16, int, vp9_row_scan_4x4_neighbors[
16 * MAX_NEIGHBORS]);
extern DECLARE_ALIGNED(16, int, vp9_col_scan_4x4_neighbors[
16 * MAX_NEIGHBORS]);
extern DECLARE_ALIGNED(16, int, vp9_default_zig_zag1d_8x8_neighbors[
64 * MAX_NEIGHBORS]);
extern DECLARE_ALIGNED(16, int, vp9_default_zig_zag1d_16x16_neighbors[
256 * MAX_NEIGHBORS]);
extern DECLARE_ALIGNED(16, int, vp9_default_zig_zag1d_32x32_neighbors[
1024 * MAX_NEIGHBORS]);
#endif // CONFIG_NEWCOEFCONTEXT
#endif // VP9_COMMON_VP9_ENTROPY_H_

159
vp9/common/vp9_entropymode.c
View File

@ -11,9 +11,10 @@
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/common/vp9_modecont.h"
#include "vp9/common/vp9_seg_common.h"
#include "vp9/common/vp9_alloccommon.h"
#include "vpx_mem/vpx_mem.h"
static const unsigned int kf_y_mode_cts[8][VP9_YMODES] = {
/* DC V H D45 135 117 153 D27 D63 TM i8x8 BPRED */
{12, 6, 5, 5, 5, 5, 5, 5, 5, 2, 22, 200},
@ -114,8 +115,6 @@ int vp9_mv_cont(const int_mv *l, const int_mv *a) {
return SUBMVREF_NORMAL;
}
const vp9_prob vp9_sub_mv_ref_prob [VP9_SUBMVREFS - 1] = { 180, 162, 25};
const vp9_prob vp9_sub_mv_ref_prob2 [SUBMVREF_COUNT][VP9_SUBMVREFS - 1] = {
{ 147, 136, 18 },
{ 106, 145, 1 },
@ -301,40 +300,32 @@ struct vp9_token_struct vp9_sub_mv_ref_encoding_array[VP9_SUBMVREFS];
void vp9_init_mbmode_probs(VP9_COMMON *x) {
unsigned int bct [VP9_YMODES] [2]; /* num Ymodes > num UV modes */
vp9_tree_probs_from_distribution(VP9_YMODES, vp9_ymode_encodings,
vp9_ymode_tree, x->fc.ymode_prob,
bct, y_mode_cts);
vp9_tree_probs_from_distribution(VP9_I32X32_MODES, vp9_sb_ymode_encodings,
vp9_sb_ymode_tree, x->fc.sb_ymode_prob,
bct, y_mode_cts);
vp9_tree_probs_from_distribution(vp9_ymode_tree, x->fc.ymode_prob,
bct, y_mode_cts, 0);
vp9_tree_probs_from_distribution(vp9_sb_ymode_tree, x->fc.sb_ymode_prob,
bct, y_mode_cts, 0);
{
int i;
for (i = 0; i < 8; i++) {
vp9_tree_probs_from_distribution(VP9_YMODES, vp9_kf_ymode_encodings,
vp9_kf_ymode_tree, x->kf_ymode_prob[i],
bct, kf_y_mode_cts[i]);
vp9_tree_probs_from_distribution(VP9_I32X32_MODES,
vp9_sb_kf_ymode_encodings,
vp9_sb_kf_ymode_tree,
vp9_tree_probs_from_distribution(vp9_kf_ymode_tree, x->kf_ymode_prob[i],
bct, kf_y_mode_cts[i], 0);
vp9_tree_probs_from_distribution(vp9_sb_kf_ymode_tree,
x->sb_kf_ymode_prob[i], bct,
kf_y_mode_cts[i]);
kf_y_mode_cts[i], 0);
}
}
{
int i;
for (i = 0; i < VP9_YMODES; i++) {
vp9_tree_probs_from_distribution(VP9_UV_MODES, vp9_uv_mode_encodings,
vp9_uv_mode_tree, x->kf_uv_mode_prob[i],
bct, kf_uv_mode_cts[i]);
vp9_tree_probs_from_distribution(VP9_UV_MODES, vp9_uv_mode_encodings,
vp9_uv_mode_tree, x->fc.uv_mode_prob[i],
bct, uv_mode_cts[i]);
vp9_tree_probs_from_distribution(vp9_uv_mode_tree, x->kf_uv_mode_prob[i],
bct, kf_uv_mode_cts[i], 0);
vp9_tree_probs_from_distribution(vp9_uv_mode_tree, x->fc.uv_mode_prob[i],
bct, uv_mode_cts[i], 0);
}
}
vp9_tree_probs_from_distribution(VP9_I8X8_MODES, vp9_i8x8_mode_encodings,
vp9_i8x8_mode_tree, x->fc.i8x8_mode_prob,
bct, i8x8_mode_cts);
vp9_tree_probs_from_distribution(vp9_i8x8_mode_tree, x->fc.i8x8_mode_prob,
bct, i8x8_mode_cts, 0);
vpx_memcpy(x->fc.sub_mv_ref_prob, vp9_sub_mv_ref_prob2,
sizeof(vp9_sub_mv_ref_prob2));
@ -344,6 +335,9 @@ void vp9_init_mbmode_probs(VP9_COMMON *x) {
#if CONFIG_COMP_INTERINTRA_PRED
x->fc.interintra_prob = VP9_DEF_INTERINTRA_PROB;
#endif
x->ref_pred_probs[0] = 120;
x->ref_pred_probs[1] = 80;
x->ref_pred_probs[2] = 40;
}
@ -351,8 +345,7 @@ static void intra_bmode_probs_from_distribution(
vp9_prob p[VP9_NKF_BINTRAMODES - 1],
unsigned int branch_ct[VP9_NKF_BINTRAMODES - 1][2],
const unsigned int events[VP9_NKF_BINTRAMODES]) {
vp9_tree_probs_from_distribution(VP9_NKF_BINTRAMODES, vp9_bmode_encodings,
vp9_bmode_tree, p, branch_ct, events);
vp9_tree_probs_from_distribution(vp9_bmode_tree, p, branch_ct, events, 0);
}
void vp9_default_bmode_probs(vp9_prob p[VP9_NKF_BINTRAMODES - 1]) {
@ -364,8 +357,7 @@ static void intra_kf_bmode_probs_from_distribution(
vp9_prob p[VP9_KF_BINTRAMODES - 1],
unsigned int branch_ct[VP9_KF_BINTRAMODES - 1][2],
const unsigned int events[VP9_KF_BINTRAMODES]) {
vp9_tree_probs_from_distribution(VP9_KF_BINTRAMODES, vp9_kf_bmode_encodings,
vp9_kf_bmode_tree, p, branch_ct, events);
vp9_tree_probs_from_distribution(vp9_kf_bmode_tree, p, branch_ct, events, 0);
}
void vp9_kf_default_bmode_probs(vp9_prob p[VP9_KF_BINTRAMODES]
@ -419,6 +411,14 @@ const int vp9_switchable_interp_map[SWITCHABLE+1] = {-1, -1, 0, 1, -1, -1};
#else
const int vp9_switchable_interp_map[SWITCHABLE+1] = {-1, 0, 1, -1, -1};
#endif
#endif // VP9_SWITCHABLE_FILTERS
// Indicates if the filter is interpolating or non-interpolating
// Note currently only the EIGHTTAP_SMOOTH is non-interpolating
#if CONFIG_ENABLE_6TAP
const int vp9_is_interpolating_filter[SWITCHABLE + 1] = {1, 0, 1, 1, 1, -1};
#else
const int vp9_is_interpolating_filter[SWITCHABLE + 1] = {0, 1, 1, 1, -1};
#endif
void vp9_entropy_mode_init() {
@ -480,7 +480,7 @@ void vp9_accum_mv_refs(VP9_COMMON *pc,
#define MVREF_COUNT_SAT 20
#define MVREF_MAX_UPDATE_FACTOR 128
void vp9_update_mode_context(VP9_COMMON *pc) {
void vp9_adapt_mode_context(VP9_COMMON *pc) {
int i, j;
unsigned int (*mv_ref_ct)[4][2];
int (*mode_context)[4];
@ -526,17 +526,17 @@ void print_mode_contexts(VP9_COMMON *pc) {
#define MODE_COUNT_SAT 20
#define MODE_MAX_UPDATE_FACTOR 144
static void update_mode_probs(int n_modes, struct vp9_token_struct *encoding,
static void update_mode_probs(int n_modes,
const vp9_tree_index *tree, unsigned int *cnt,
vp9_prob *pre_probs, vp9_prob *dst_probs) {
vp9_prob *pre_probs, vp9_prob *dst_probs,
unsigned int tok0_offset) {
#define MAX_PROBS 32
vp9_prob probs[MAX_PROBS];
unsigned int branch_ct[MAX_PROBS][2];
int t, count, factor;
assert(n_modes - 1 < MAX_PROBS);
vp9_tree_probs_from_distribution(n_modes, encoding, tree, probs,
branch_ct, cnt);
vp9_tree_probs_from_distribution(tree, probs, branch_ct, cnt, tok0_offset);
for (t = 0; t < n_modes - 1; ++t) {
count = branch_ct[t][0] + branch_ct[t][1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
@ -592,31 +592,32 @@ void vp9_adapt_mode_probs(VP9_COMMON *cm) {
#endif
#endif
update_mode_probs(VP9_YMODES, vp9_ymode_encodings, vp9_ymode_tree,
update_mode_probs(VP9_YMODES, vp9_ymode_tree,
cm->fc.ymode_counts, cm->fc.pre_ymode_prob,
cm->fc.ymode_prob);
update_mode_probs(VP9_I32X32_MODES, vp9_sb_ymode_encodings, vp9_sb_ymode_tree,
cm->fc.ymode_prob, 0);
update_mode_probs(VP9_I32X32_MODES, vp9_sb_ymode_tree,
cm->fc.sb_ymode_counts, cm->fc.pre_sb_ymode_prob,
cm->fc.sb_ymode_prob);
cm->fc.sb_ymode_prob, 0);
for (i = 0; i < VP9_YMODES; ++i) {
update_mode_probs(VP9_UV_MODES, vp9_uv_mode_encodings, vp9_uv_mode_tree,
update_mode_probs(VP9_UV_MODES, vp9_uv_mode_tree,
cm->fc.uv_mode_counts[i], cm->fc.pre_uv_mode_prob[i],
cm->fc.uv_mode_prob[i]);
cm->fc.uv_mode_prob[i], 0);
}
update_mode_probs(VP9_NKF_BINTRAMODES, vp9_bmode_encodings, vp9_bmode_tree,
update_mode_probs(VP9_NKF_BINTRAMODES, vp9_bmode_tree,
cm->fc.bmode_counts, cm->fc.pre_bmode_prob,
cm->fc.bmode_prob);
update_mode_probs(VP9_I8X8_MODES, vp9_i8x8_mode_encodings,
cm->fc.bmode_prob, 0);
update_mode_probs(VP9_I8X8_MODES,
vp9_i8x8_mode_tree, cm->fc.i8x8_mode_counts,
cm->fc.pre_i8x8_mode_prob, cm->fc.i8x8_mode_prob);
cm->fc.pre_i8x8_mode_prob, cm->fc.i8x8_mode_prob, 0);
for (i = 0; i < SUBMVREF_COUNT; ++i) {
update_mode_probs(VP9_SUBMVREFS, vp9_sub_mv_ref_encoding_array,
update_mode_probs(VP9_SUBMVREFS,
vp9_sub_mv_ref_tree, cm->fc.sub_mv_ref_counts[i],
cm->fc.pre_sub_mv_ref_prob[i], cm->fc.sub_mv_ref_prob[i]);
cm->fc.pre_sub_mv_ref_prob[i], cm->fc.sub_mv_ref_prob[i],
LEFT4X4);
}
update_mode_probs(VP9_NUMMBSPLITS, vp9_mbsplit_encodings, vp9_mbsplit_tree,
update_mode_probs(VP9_NUMMBSPLITS, vp9_mbsplit_tree,
cm->fc.mbsplit_counts, cm->fc.pre_mbsplit_prob,
cm->fc.mbsplit_prob);
cm->fc.mbsplit_prob, 0);
#if CONFIG_COMP_INTERINTRA_PRED
if (cm->use_interintra) {
int factor, interintra_prob, count;
@ -631,3 +632,65 @@ void vp9_adapt_mode_probs(VP9_COMMON *cm) {
}
#endif
}
static void set_default_lf_deltas(MACROBLOCKD *xd) {
xd->mode_ref_lf_delta_enabled = 1;
xd->mode_ref_lf_delta_update = 1;
xd->ref_lf_deltas[INTRA_FRAME] = 2;
xd->ref_lf_deltas[LAST_FRAME] = 0;
xd->ref_lf_deltas[GOLDEN_FRAME] = -2;
xd->ref_lf_deltas[ALTREF_FRAME] = -2;
xd->mode_lf_deltas[0] = 4; // BPRED
xd->mode_lf_deltas[1] = -2; // Zero
xd->mode_lf_deltas[2] = 2; // New mv
xd->mode_lf_deltas[3] = 4; // Split mv
}
void vp9_setup_past_independence(VP9_COMMON *cm, MACROBLOCKD *xd) {
// Reset the segment feature data to the default stats:
// Features disabled, 0, with delta coding (Default state).
int i;
vp9_clearall_segfeatures(xd);
xd->mb_segment_abs_delta = SEGMENT_DELTADATA;
if (cm->last_frame_seg_map)
vpx_memset(cm->last_frame_seg_map, 0, (cm->mb_rows * cm->mb_cols));
/* reset the mode ref deltas for loop filter */
vpx_memset(xd->last_ref_lf_deltas, 0, sizeof(xd->last_ref_lf_deltas));
vpx_memset(xd->last_mode_lf_deltas, 0, sizeof(xd->last_mode_lf_deltas));
set_default_lf_deltas(xd);
vp9_default_coef_probs(cm);
vp9_init_mbmode_probs(cm);
vp9_default_bmode_probs(cm->fc.bmode_prob);
vp9_kf_default_bmode_probs(cm->kf_bmode_prob);
vp9_init_mv_probs(cm);
// To force update of the sharpness
cm->last_sharpness_level = -1;
vp9_init_mode_contexts(cm);
for (i = 0; i < NUM_FRAME_CONTEXTS; i++) {
vpx_memcpy(&cm->frame_contexts[i], &cm->fc, sizeof(cm->fc));
}
vpx_memset(cm->prev_mip, 0,
(cm->mb_cols + 1) * (cm->mb_rows + 1)* sizeof(MODE_INFO));
vpx_memset(cm->mip, 0,
(cm->mb_cols + 1) * (cm->mb_rows + 1)* sizeof(MODE_INFO));
vp9_update_mode_info_border(cm, cm->mip);
vp9_update_mode_info_in_image(cm, cm->mi);
#if CONFIG_NEW_MVREF
// Defaults probabilities for encoding the MV ref id signal
vpx_memset(xd->mb_mv_ref_probs, VP9_DEFAULT_MV_REF_PROB,
sizeof(xd->mb_mv_ref_probs));
#endif
cm->ref_frame_sign_bias[GOLDEN_FRAME] = 0;
cm->ref_frame_sign_bias[ALTREF_FRAME] = 0;
cm->frame_context_idx = 0;
}

9
vp9/common/vp9_entropymode.h
View File

@ -34,8 +34,6 @@ extern const vp9_prob vp9_mbsplit_probs[VP9_NUMMBSPLITS - 1];
extern int vp9_mv_cont(const int_mv *l, const int_mv *a);
extern const vp9_prob vp9_sub_mv_ref_prob[VP9_SUBMVREFS - 1];
extern const vp9_prob vp9_sub_mv_ref_prob2[SUBMVREF_COUNT][VP9_SUBMVREFS - 1];
extern const unsigned int vp9_kf_default_bmode_counts[VP9_KF_BINTRAMODES]
@ -76,11 +74,14 @@ void vp9_entropy_mode_init(void);
struct VP9Common;
/* sets up common features to forget past dependence */
void vp9_setup_past_independence(struct VP9Common *cm, MACROBLOCKD *xd);
void vp9_init_mbmode_probs(struct VP9Common *x);
extern void vp9_init_mode_contexts(struct VP9Common *pc);
extern void vp9_update_mode_context(struct VP9Common *pc);
extern void vp9_adapt_mode_context(struct VP9Common *pc);
extern void vp9_accum_mv_refs(struct VP9Common *pc,
MB_PREDICTION_MODE m,
@ -101,6 +102,8 @@ extern const INTERPOLATIONFILTERTYPE vp9_switchable_interp
extern const int vp9_switchable_interp_map[SWITCHABLE + 1];
extern const int vp9_is_interpolating_filter[SWITCHABLE + 1];
extern const vp9_tree_index vp9_switchable_interp_tree
[2 * (VP9_SWITCHABLE_FILTERS - 1)];

240
vp9/common/vp9_entropymv.c
View File

@ -42,7 +42,10 @@ const vp9_tree_index vp9_mv_class_tree[2 * MV_CLASSES - 2] = {
-MV_CLASS_2, -MV_CLASS_3,
10, 12,
-MV_CLASS_4, -MV_CLASS_5,
-MV_CLASS_6, -MV_CLASS_7,
-MV_CLASS_6, 14,
16, 18,
-MV_CLASS_7, -MV_CLASS_8,
-MV_CLASS_9, -MV_CLASS_10,
};
struct vp9_token_struct vp9_mv_class_encodings[MV_CLASSES];
@ -62,24 +65,24 @@ const nmv_context vp9_default_nmv_context = {
{32, 64, 96},
{
{ /* vert component */
128, /* sign */
{224, 144, 192, 168, 192, 176, 192}, /* class */
{216}, /* class0 */
{136, 140, 148, 160, 176, 192, 224}, /* bits */
{{128, 128, 64}, {96, 112, 64}}, /* class0_fp */
{64, 96, 64}, /* fp */
160, /* class0_hp bit */
128, /* hp */
128, /* sign */
{224, 144, 192, 168, 192, 176, 192, 198, 198, 245}, /* class */
{216}, /* class0 */
{136, 140, 148, 160, 176, 192, 224, 234, 234, 240}, /* bits */
{{128, 128, 64}, {96, 112, 64}}, /* class0_fp */
{64, 96, 64}, /* fp */
160, /* class0_hp bit */
128, /* hp */
},
{ /* hor component */
128, /* sign */
{216, 128, 176, 160, 176, 176, 192}, /* class */
{208}, /* class0 */
{136, 140, 148, 160, 176, 192, 224}, /* bits */
{{128, 128, 64}, {96, 112, 64}}, /* class0_fp */
{64, 96, 64}, /* fp */
160, /* class0_hp bit */
128, /* hp */
128, /* sign */
{216, 128, 176, 160, 176, 176, 192, 198, 198, 208}, /* class */
{208}, /* class0 */
{136, 140, 148, 160, 176, 192, 224, 234, 234, 240}, /* bits */
{{128, 128, 64}, {96, 112, 64}}, /* class0_fp */
{64, 96, 64}, /* fp */
160, /* class0_hp bit */
128, /* hp */
}
},
};
@ -103,6 +106,9 @@ MV_CLASS_TYPE vp9_get_mv_class(int z, int *offset) {
else if (z < CLASS0_SIZE * 256) c = MV_CLASS_5;
else if (z < CLASS0_SIZE * 512) c = MV_CLASS_6;
else if (z < CLASS0_SIZE * 1024) c = MV_CLASS_7;
else if (z < CLASS0_SIZE * 2048) c = MV_CLASS_8;
else if (z < CLASS0_SIZE * 4096) c = MV_CLASS_9;
else if (z < CLASS0_SIZE * 8192) c = MV_CLASS_10;
else assert(0);
if (offset)
*offset = z - mv_class_base(c);
@ -110,11 +116,8 @@ MV_CLASS_TYPE vp9_get_mv_class(int z, int *offset) {
}
int vp9_use_nmv_hp(const MV *ref) {
if ((abs(ref->row) >> 3) < COMPANDED_MVREF_THRESH &&
(abs(ref->col) >> 3) < COMPANDED_MVREF_THRESH)
return 1;
else
return 0;
return (abs(ref->row) >> 3) < COMPANDED_MVREF_THRESH &&
(abs(ref->col) >> 3) < COMPANDED_MVREF_THRESH;
}
int vp9_get_mv_mag(MV_CLASS_TYPE c, int offset) {
@ -134,6 +137,7 @@ static void increment_nmv_component(int v,
int incr,
int usehp) {
int s, z, c, o, d, e, f;
if (!incr) return;
assert (v != 0); /* should not be zero */
s = v < 0;
mvcomp->sign[s] += incr;
@ -211,24 +215,26 @@ void vp9_increment_nmv(const MV *mv, const MV *ref, nmv_context_counts *mvctx,
}
}
static void adapt_prob(vp9_prob *dest, vp9_prob prep, vp9_prob newp,
static void adapt_prob(vp9_prob *dest, vp9_prob prep,
unsigned int ct[2]) {
int count = ct[0] + ct[1];
if (count) {
vp9_prob newp = get_binary_prob(ct[0], ct[1]);
count = count > MV_COUNT_SAT ? MV_COUNT_SAT : count;
*dest = weighted_prob(prep, newp,
MV_MAX_UPDATE_FACTOR * count / MV_COUNT_SAT);
} else {
*dest = prep;
}
}
void vp9_counts_process(nmv_context_counts *NMVcount, int usehp) {
counts_to_context(&NMVcount->comps[0], usehp);
counts_to_context(&NMVcount->comps[1], usehp);
void vp9_counts_process(nmv_context_counts *nmv_count, int usehp) {
counts_to_context(&nmv_count->comps[0], usehp);
counts_to_context(&nmv_count->comps[1], usehp);
}
void vp9_counts_to_nmv_context(
nmv_context_counts *NMVcount,
nmv_context_counts *nmv_count,
nmv_context *prob,
int usehp,
unsigned int (*branch_ct_joint)[2],
@ -241,81 +247,90 @@ void vp9_counts_to_nmv_context(
unsigned int (*branch_ct_class0_hp)[2],
unsigned int (*branch_ct_hp)[2]) {
int i, j, k;
vp9_counts_process(NMVcount, usehp);
vp9_tree_probs_from_distribution(MV_JOINTS,
vp9_mv_joint_encodings,
vp9_mv_joint_tree,
vp9_counts_process(nmv_count, usehp);
vp9_tree_probs_from_distribution(vp9_mv_joint_tree,
prob->joints,
branch_ct_joint,
NMVcount->joints);
nmv_count->joints, 0);
for (i = 0; i < 2; ++i) {
prob->comps[i].sign = get_binary_prob(NMVcount->comps[i].sign[0],
NMVcount->comps[i].sign[1]);
branch_ct_sign[i][0] = NMVcount->comps[i].sign[0];
branch_ct_sign[i][1] = NMVcount->comps[i].sign[1];
vp9_tree_probs_from_distribution(MV_CLASSES,
vp9_mv_class_encodings,
vp9_mv_class_tree,
prob->comps[i].sign = get_binary_prob(nmv_count->comps[i].sign[0],
nmv_count->comps[i].sign[1]);
branch_ct_sign[i][0] = nmv_count->comps[i].sign[0];
branch_ct_sign[i][1] = nmv_count->comps[i].sign[1];
vp9_tree_probs_from_distribution(vp9_mv_class_tree,
prob->comps[i].classes,
branch_ct_classes[i],
NMVcount->comps[i].classes);
vp9_tree_probs_from_distribution(CLASS0_SIZE,
vp9_mv_class0_encodings,
vp9_mv_class0_tree,
nmv_count->comps[i].classes, 0);
vp9_tree_probs_from_distribution(vp9_mv_class0_tree,
prob->comps[i].class0,
branch_ct_class0[i],
NMVcount->comps[i].class0);
nmv_count->comps[i].class0, 0);
for (j = 0; j < MV_OFFSET_BITS; ++j) {
prob->comps[i].bits[j] = get_binary_prob(NMVcount->comps[i].bits[j][0],
NMVcount->comps[i].bits[j][1]);
branch_ct_bits[i][j][0] = NMVcount->comps[i].bits[j][0];
branch_ct_bits[i][j][1] = NMVcount->comps[i].bits[j][1];
prob->comps[i].bits[j] = get_binary_prob(nmv_count->comps[i].bits[j][0],
nmv_count->comps[i].bits[j][1]);
branch_ct_bits[i][j][0] = nmv_count->comps[i].bits[j][0];
branch_ct_bits[i][j][1] = nmv_count->comps[i].bits[j][1];
}
}
for (i = 0; i < 2; ++i) {
for (k = 0; k < CLASS0_SIZE; ++k) {
vp9_tree_probs_from_distribution(4,
vp9_mv_fp_encodings,
vp9_mv_fp_tree,
vp9_tree_probs_from_distribution(vp9_mv_fp_tree,
prob->comps[i].class0_fp[k],
branch_ct_class0_fp[i][k],
NMVcount->comps[i].class0_fp[k]);
nmv_count->comps[i].class0_fp[k], 0);
}
vp9_tree_probs_from_distribution(4,
vp9_mv_fp_encodings,
vp9_mv_fp_tree,
vp9_tree_probs_from_distribution(vp9_mv_fp_tree,
prob->comps[i].fp,
branch_ct_fp[i],
NMVcount->comps[i].fp);
nmv_count->comps[i].fp, 0);
}
if (usehp) {
for (i = 0; i < 2; ++i) {
prob->comps[i].class0_hp =
get_binary_prob(NMVcount->comps[i].class0_hp[0],
NMVcount->comps[i].class0_hp[1]);
branch_ct_class0_hp[i][0] = NMVcount->comps[i].class0_hp[0];
branch_ct_class0_hp[i][1] = NMVcount->comps[i].class0_hp[1];
get_binary_prob(nmv_count->comps[i].class0_hp[0],
nmv_count->comps[i].class0_hp[1]);
branch_ct_class0_hp[i][0] = nmv_count->comps[i].class0_hp[0];
branch_ct_class0_hp[i][1] = nmv_count->comps[i].class0_hp[1];
prob->comps[i].hp = get_binary_prob(NMVcount->comps[i].hp[0],
NMVcount->comps[i].hp[1]);
branch_ct_hp[i][0] = NMVcount->comps[i].hp[0];
branch_ct_hp[i][1] = NMVcount->comps[i].hp[1];
prob->comps[i].hp = get_binary_prob(nmv_count->comps[i].hp[0],
nmv_count->comps[i].hp[1]);
branch_ct_hp[i][0] = nmv_count->comps[i].hp[0];
branch_ct_hp[i][1] = nmv_count->comps[i].hp[1];
}
}
}
static unsigned int adapt_probs(unsigned int i,
vp9_tree tree,
vp9_prob this_probs[],
const vp9_prob last_probs[],
const unsigned int num_events[]) {
vp9_prob this_prob;
const uint32_t left = tree[i] <= 0
? num_events[-tree[i]]
: adapt_probs(tree[i], tree, this_probs, last_probs, num_events);
const uint32_t right = tree[i + 1] <= 0
? num_events[-tree[i + 1]]
: adapt_probs(tree[i + 1], tree, this_probs, last_probs, num_events);
uint32_t weight = left + right;
if (weight) {
this_prob = get_binary_prob(left, right);
weight = weight > MV_COUNT_SAT ? MV_COUNT_SAT : weight;
this_prob = weighted_prob(last_probs[i >> 1], this_prob,
MV_MAX_UPDATE_FACTOR * weight / MV_COUNT_SAT);
} else {
this_prob = last_probs[i >> 1];
}
this_probs[i >> 1] = this_prob;
return left + right;
}
void vp9_adapt_nmv_probs(VP9_COMMON *cm, int usehp) {
int i, j, k;
nmv_context prob;
unsigned int branch_ct_joint[MV_JOINTS - 1][2];
unsigned int branch_ct_sign[2][2];
unsigned int branch_ct_classes[2][MV_CLASSES - 1][2];
unsigned int branch_ct_class0[2][CLASS0_SIZE - 1][2];
unsigned int branch_ct_bits[2][MV_OFFSET_BITS][2];
unsigned int branch_ct_class0_fp[2][CLASS0_SIZE][4 - 1][2];
unsigned int branch_ct_fp[2][4 - 1][2];
unsigned int branch_ct_class0_hp[2][2];
unsigned int branch_ct_hp[2][2];
int i, j;
#ifdef MV_COUNT_TESTING
printf("joints count: ");
for (j = 0; j < MV_JOINTS; ++j) printf("%d ", cm->fc.NMVcount.joints[j]);
@ -376,75 +391,48 @@ void vp9_adapt_nmv_probs(VP9_COMMON *cm, int usehp) {
smooth_counts(&cm->fc.NMVcount.comps[0]);
smooth_counts(&cm->fc.NMVcount.comps[1]);
#endif
vp9_counts_to_nmv_context(&cm->fc.NMVcount,
&prob,
usehp,
branch_ct_joint,
branch_ct_sign,
branch_ct_classes,
branch_ct_class0,
branch_ct_bits,
branch_ct_class0_fp,
branch_ct_fp,
branch_ct_class0_hp,
branch_ct_hp);
vp9_counts_process(&cm->fc.NMVcount, usehp);
adapt_probs(0, vp9_mv_joint_tree,
cm->fc.nmvc.joints, cm->fc.pre_nmvc.joints,
cm->fc.NMVcount.joints);
for (j = 0; j < MV_JOINTS - 1; ++j) {
adapt_prob(&cm->fc.nmvc.joints[j],
cm->fc.pre_nmvc.joints[j],
prob.joints[j],
branch_ct_joint[j]);
}
for (i = 0; i < 2; ++i) {
adapt_prob(&cm->fc.nmvc.comps[i].sign,
cm->fc.pre_nmvc.comps[i].sign,
prob.comps[i].sign,
branch_ct_sign[i]);
for (j = 0; j < MV_CLASSES - 1; ++j) {
adapt_prob(&cm->fc.nmvc.comps[i].classes[j],
cm->fc.pre_nmvc.comps[i].classes[j],
prob.comps[i].classes[j],
branch_ct_classes[i][j]);
}
for (j = 0; j < CLASS0_SIZE - 1; ++j) {
adapt_prob(&cm->fc.nmvc.comps[i].class0[j],
cm->fc.pre_nmvc.comps[i].class0[j],
prob.comps[i].class0[j],
branch_ct_class0[i][j]);
}
cm->fc.NMVcount.comps[i].sign);
adapt_probs(0, vp9_mv_class_tree,
cm->fc.nmvc.comps[i].classes, cm->fc.pre_nmvc.comps[i].classes,
cm->fc.NMVcount.comps[i].classes);
adapt_probs(0, vp9_mv_class0_tree,
cm->fc.nmvc.comps[i].class0, cm->fc.pre_nmvc.comps[i].class0,
cm->fc.NMVcount.comps[i].class0);
for (j = 0; j < MV_OFFSET_BITS; ++j) {
adapt_prob(&cm->fc.nmvc.comps[i].bits[j],
cm->fc.pre_nmvc.comps[i].bits[j],
prob.comps[i].bits[j],
branch_ct_bits[i][j]);
cm->fc.NMVcount.comps[i].bits[j]);
}
}
for (i = 0; i < 2; ++i) {
for (j = 0; j < CLASS0_SIZE; ++j) {
for (k = 0; k < 3; ++k) {
adapt_prob(&cm->fc.nmvc.comps[i].class0_fp[j][k],
cm->fc.pre_nmvc.comps[i].class0_fp[j][k],
prob.comps[i].class0_fp[j][k],
branch_ct_class0_fp[i][j][k]);
}
}
for (j = 0; j < 3; ++j) {
adapt_prob(&cm->fc.nmvc.comps[i].fp[j],
cm->fc.pre_nmvc.comps[i].fp[j],
prob.comps[i].fp[j],
branch_ct_fp[i][j]);
adapt_probs(0, vp9_mv_fp_tree,
cm->fc.nmvc.comps[i].class0_fp[j],
cm->fc.pre_nmvc.comps[i].class0_fp[j],
cm->fc.NMVcount.comps[i].class0_fp[j]);
}
adapt_probs(0, vp9_mv_fp_tree,
cm->fc.nmvc.comps[i].fp,
cm->fc.pre_nmvc.comps[i].fp,
cm->fc.NMVcount.comps[i].fp);
}
if (usehp) {
for (i = 0; i < 2; ++i) {
adapt_prob(&cm->fc.nmvc.comps[i].class0_hp,
cm->fc.pre_nmvc.comps[i].class0_hp,
prob.comps[i].class0_hp,
branch_ct_class0_hp[i]);
cm->fc.NMVcount.comps[i].class0_hp);
adapt_prob(&cm->fc.nmvc.comps[i].hp,
cm->fc.pre_nmvc.comps[i].hp,
prob.comps[i].hp,
branch_ct_hp[i]);
cm->fc.NMVcount.comps[i].hp);
}
}
}

5
vp9/common/vp9_entropymv.h
View File

@ -49,7 +49,7 @@ extern const vp9_tree_index vp9_mv_joint_tree[2 * MV_JOINTS - 2];
extern struct vp9_token_struct vp9_mv_joint_encodings [MV_JOINTS];
/* Symbols for coding magnitude class of nonzero components */
#define MV_CLASSES 8
#define MV_CLASSES 11
typedef enum {
MV_CLASS_0 = 0, /* (0, 2] integer pel */
MV_CLASS_1 = 1, /* (2, 4] integer pel */
@ -59,6 +59,9 @@ typedef enum {
MV_CLASS_5 = 5, /* (32, 64] integer pel */
MV_CLASS_6 = 6, /* (64, 128] integer pel */
MV_CLASS_7 = 7, /* (128, 256] integer pel */
MV_CLASS_8 = 8, /* (256, 512] integer pel */
MV_CLASS_9 = 9, /* (512, 1024] integer pel */
MV_CLASS_10 = 10, /* (1024,2048] integer pel */
} MV_CLASS_TYPE;
extern const vp9_tree_index vp9_mv_class_tree[2 * MV_CLASSES - 2];

204
vp9/common/vp9_extend.c
View File

@ -11,159 +11,137 @@
#include "vp9/common/vp9_extend.h"
#include "vpx_mem/vpx_mem.h"
static void copy_and_extend_plane(uint8_t *s, /* source */
int sp, /* source pitch */
uint8_t *d, /* destination */
int dp, /* destination pitch */
int h, /* height */
int w, /* width */
int et, /* extend top border */
int el, /* extend left border */
int eb, /* extend bottom border */
int er) { /* extend right border */
int i;
uint8_t *src_ptr1, *src_ptr2;
uint8_t *dest_ptr1, *dest_ptr2;
int linesize;
static void copy_and_extend_plane(const uint8_t *src, int src_pitch,
uint8_t *dst, int dst_pitch,
int w, int h,
int extend_top, int extend_left,
int extend_bottom, int extend_right) {
int i, linesize;
/* copy the left and right most columns out */
src_ptr1 = s;
src_ptr2 = s + w - 1;
dest_ptr1 = d - el;
dest_ptr2 = d + w;
// copy the left and right most columns out
const uint8_t *src_ptr1 = src;
const uint8_t *src_ptr2 = src + w - 1;
uint8_t *dst_ptr1 = dst - extend_left;
uint8_t *dst_ptr2 = dst + w;
for (i = 0; i < h; i++) {
vpx_memset(dest_ptr1, src_ptr1[0], el);
vpx_memcpy(dest_ptr1 + el, src_ptr1, w);
vpx_memset(dest_ptr2, src_ptr2[0], er);
src_ptr1 += sp;
src_ptr2 += sp;
dest_ptr1 += dp;
dest_ptr2 += dp;
vpx_memset(dst_ptr1, src_ptr1[0], extend_left);
vpx_memcpy(dst_ptr1 + extend_left, src_ptr1, w);
vpx_memset(dst_ptr2, src_ptr2[0], extend_right);
src_ptr1 += src_pitch;
src_ptr2 += src_pitch;
dst_ptr1 += dst_pitch;
dst_ptr2 += dst_pitch;
}
/* Now copy the top and bottom lines into each line of the respective
* borders
*/
src_ptr1 = d - el;
src_ptr2 = d + dp * (h - 1) - el;
dest_ptr1 = d + dp * (-et) - el;
dest_ptr2 = d + dp * (h) - el;
linesize = el + er + w;
// Now copy the top and bottom lines into each line of the respective
// borders
src_ptr1 = dst - extend_left;
src_ptr2 = dst + dst_pitch * (h - 1) - extend_left;
dst_ptr1 = dst + dst_pitch * (-extend_top) - extend_left;
dst_ptr2 = dst + dst_pitch * (h) - extend_left;
linesize = extend_left + extend_right + w;
for (i = 0; i < et; i++) {
vpx_memcpy(dest_ptr1, src_ptr1, linesize);
dest_ptr1 += dp;
for (i = 0; i < extend_top; i++) {
vpx_memcpy(dst_ptr1, src_ptr1, linesize);
dst_ptr1 += dst_pitch;
}
for (i = 0; i < eb; i++) {
vpx_memcpy(dest_ptr2, src_ptr2, linesize);
dest_ptr2 += dp;
for (i = 0; i < extend_bottom; i++) {
vpx_memcpy(dst_ptr2, src_ptr2, linesize);
dst_ptr2 += dst_pitch;
}
}
void vp9_copy_and_extend_frame(YV12_BUFFER_CONFIG *src,
void vp9_copy_and_extend_frame(const YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst) {
int et = dst->border;
int el = dst->border;
int eb = dst->border + dst->y_height - src->y_height;
int er = dst->border + dst->y_width - src->y_width;
const int et_y = dst->border;
const int el_y = dst->border;
const int eb_y = dst->border + dst->y_height - src->y_height;
const int er_y = dst->border + dst->y_width - src->y_width;
const int et_uv = dst->border >> 1;
const int el_uv = dst->border >> 1;
const int eb_uv = (dst->border >> 1) + dst->uv_height - src->uv_height;
const int er_uv = (dst->border >> 1) + dst->uv_width - src->uv_width;
copy_and_extend_plane(src->y_buffer, src->y_stride,
dst->y_buffer, dst->y_stride,
src->y_height, src->y_width,
et, el, eb, er);
et = dst->border >> 1;
el = dst->border >> 1;
eb = (dst->border >> 1) + dst->uv_height - src->uv_height;
er = (dst->border >> 1) + dst->uv_width - src->uv_width;
src->y_width, src->y_height,
et_y, el_y, eb_y, er_y);
copy_and_extend_plane(src->u_buffer, src->uv_stride,
dst->u_buffer, dst->uv_stride,
src->uv_height, src->uv_width,
et, el, eb, er);
src->uv_width, src->uv_height,
et_uv, el_uv, eb_uv, er_uv);
copy_and_extend_plane(src->v_buffer, src->uv_stride,
dst->v_buffer, dst->uv_stride,
src->uv_height, src->uv_width,
et, el, eb, er);
src->uv_width, src->uv_height,
et_y, el_y, eb_uv, er_uv);
}
void vp9_copy_and_extend_frame_with_rect(YV12_BUFFER_CONFIG *src,
void vp9_copy_and_extend_frame_with_rect(const YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst,
int srcy, int srcx,
int srch, int srcw) {
int et = dst->border;
int el = dst->border;
int eb = dst->border + dst->y_height - src->y_height;
int er = dst->border + dst->y_width - src->y_width;
int src_y_offset = srcy * src->y_stride + srcx;
int dst_y_offset = srcy * dst->y_stride + srcx;
int src_uv_offset = ((srcy * src->uv_stride) >> 1) + (srcx >> 1);
int dst_uv_offset = ((srcy * dst->uv_stride) >> 1) + (srcx >> 1);
// If the side is not touching the bounder then don't extend.
if (srcy)
et = 0;
if (srcx)
el = 0;
if (srcy + srch != src->y_height)
eb = 0;
if (srcx + srcw != src->y_width)
er = 0;
const int et_y = srcy ? 0 : dst->border;
const int el_y = srcx ? 0 : dst->border;
const int eb_y = srcy + srch != src->y_height ? 0 :
dst->border + dst->y_height - src->y_height;
const int er_y = srcx + srcw != src->y_width ? 0 :
dst->border + dst->y_width - src->y_width;
const int src_y_offset = srcy * src->y_stride + srcx;
const int dst_y_offset = srcy * dst->y_stride + srcx;
copy_and_extend_plane(src->y_buffer + src_y_offset,
src->y_stride,
dst->y_buffer + dst_y_offset,
dst->y_stride,
srch, srcw,
et, el, eb, er);
const int et_uv = (et_y + 1) >> 1;
const int el_uv = (el_y + 1) >> 1;
const int eb_uv = (eb_y + 1) >> 1;
const int er_uv = (er_y + 1) >> 1;
const int src_uv_offset = ((srcy * src->uv_stride) >> 1) + (srcx >> 1);
const int dst_uv_offset = ((srcy * dst->uv_stride) >> 1) + (srcx >> 1);
const int srch_uv = (srch + 1) >> 1;
const int srcw_uv = (srcw + 1) >> 1;
et = (et + 1) >> 1;
el = (el + 1) >> 1;
eb = (eb + 1) >> 1;
er = (er + 1) >> 1;
srch = (srch + 1) >> 1;
srcw = (srcw + 1) >> 1;
copy_and_extend_plane(src->y_buffer + src_y_offset, src->y_stride,
dst->y_buffer + dst_y_offset, dst->y_stride,
srcw, srch,
et_y, el_y, eb_y, er_y);
copy_and_extend_plane(src->u_buffer + src_uv_offset,
src->uv_stride,
dst->u_buffer + dst_uv_offset,
dst->uv_stride,
srch, srcw,
et, el, eb, er);
copy_and_extend_plane(src->u_buffer + src_uv_offset, src->uv_stride,
dst->u_buffer + dst_uv_offset, dst->uv_stride,
srcw_uv, srch_uv,
et_uv, el_uv, eb_uv, er_uv);
copy_and_extend_plane(src->v_buffer + src_uv_offset,
src->uv_stride,
dst->v_buffer + dst_uv_offset,
dst->uv_stride,
srch, srcw,
et, el, eb, er);
copy_and_extend_plane(src->v_buffer + src_uv_offset, src->uv_stride,
dst->v_buffer + dst_uv_offset, dst->uv_stride,
srcw_uv, srch_uv,
et_uv, el_uv, eb_uv, er_uv);
}
/* note the extension is only for the last row, for intra prediction purpose */
void vp9_extend_mb_row(YV12_BUFFER_CONFIG *ybf, uint8_t *YPtr,
uint8_t *UPtr, uint8_t *VPtr) {
// note the extension is only for the last row, for intra prediction purpose
void vp9_extend_mb_row(YV12_BUFFER_CONFIG *buf,
uint8_t *y, uint8_t *u, uint8_t *v) {
int i;
YPtr += ybf->y_stride * 14;
UPtr += ybf->uv_stride * 6;
VPtr += ybf->uv_stride * 6;
y += buf->y_stride * 14;
u += buf->uv_stride * 6;
v += buf->uv_stride * 6;
for (i = 0; i < 4; i++) {
YPtr[i] = YPtr[-1];
UPtr[i] = UPtr[-1];
VPtr[i] = VPtr[-1];
y[i] = y[-1];
u[i] = u[-1];
v[i] = v[-1];
}
YPtr += ybf->y_stride;
UPtr += ybf->uv_stride;
VPtr += ybf->uv_stride;
y += buf->y_stride;
u += buf->uv_stride;
v += buf->uv_stride;
for (i = 0; i < 4; i++) {
YPtr[i] = YPtr[-1];
UPtr[i] = UPtr[-1];
VPtr[i] = VPtr[-1];
y[i] = y[-1];
u[i] = u[-1];
v[i] = v[-1];
}
}

10
vp9/common/vp9_extend.h
View File

@ -14,15 +14,17 @@
#include "vpx_scale/yv12config.h"
#include "vpx/vpx_integer.h"
void vp9_extend_mb_row(YV12_BUFFER_CONFIG *ybf, uint8_t *YPtr,
uint8_t *UPtr, uint8_t *VPtr);
void vp9_copy_and_extend_frame(YV12_BUFFER_CONFIG *src,
void vp9_copy_and_extend_frame(const YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst);
void vp9_copy_and_extend_frame_with_rect(YV12_BUFFER_CONFIG *src,
void vp9_copy_and_extend_frame_with_rect(const YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst,
int srcy, int srcx,
int srch, int srcw);
void vp9_extend_mb_row(YV12_BUFFER_CONFIG *buf,
uint8_t *y, uint8_t *u, uint8_t *v);
#endif // VP9_COMMON_VP9_EXTEND_H_

1196
vp9/common/vp9_filter.c
View File

File diff suppressed because it is too large Load Diff

11
vp9/common/vp9_filter.h
View File

@ -21,10 +21,17 @@
#define SUBPEL_SHIFTS 16
extern const int16_t vp9_bilinear_filters[SUBPEL_SHIFTS][2];
extern const int16_t vp9_sub_pel_filters_6[SUBPEL_SHIFTS][6];
extern const int16_t vp9_bilinear_filters[SUBPEL_SHIFTS][8];
extern const int16_t vp9_sub_pel_filters_6[SUBPEL_SHIFTS][8];
extern const int16_t vp9_sub_pel_filters_8[SUBPEL_SHIFTS][8];
extern const int16_t vp9_sub_pel_filters_8s[SUBPEL_SHIFTS][8];
extern const int16_t vp9_sub_pel_filters_8lp[SUBPEL_SHIFTS][8];
// The VP9_BILINEAR_FILTERS_2TAP macro returns a pointer to the bilinear
// filter kernel as a 2 tap filter.
#define BF_LENGTH (sizeof(vp9_bilinear_filters[0]) / \
sizeof(vp9_bilinear_filters[0][0]))
#define BF_OFFSET (BF_LENGTH / 2 - 1)
#define VP9_BILINEAR_FILTERS_2TAP(x) (vp9_bilinear_filters[x] + BF_OFFSET)
#endif // VP9_COMMON_VP9_FILTER_H_

296
vp9/common/vp9_findnearmv.c
View File

@ -9,10 +9,11 @@
*/
#include <limits.h>
#include "vp9/common/vp9_findnearmv.h"
#include "vp9/common/vp9_sadmxn.h"
#include "vp9/common/vp9_subpelvar.h"
#include <limits.h>
const uint8_t vp9_mbsplit_offset[4][16] = {
{ 0, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
@ -32,8 +33,7 @@ static void lower_mv_precision(int_mv *mv, int usehp)
}
vp9_prob *vp9_mv_ref_probs(VP9_COMMON *pc,
vp9_prob p[4], const int context
) {
vp9_prob p[4], const int context) {
p[0] = pc->fc.vp9_mode_contexts[context][0];
p[1] = pc->fc.vp9_mode_contexts[context][1];
p[2] = pc->fc.vp9_mode_contexts[context][2];
@ -87,8 +87,8 @@ unsigned int vp9_sub_pixel_variance16x2_c(const uint8_t *src_ptr,
uint8_t temp2[2 * 16];
const int16_t *HFilter, *VFilter;
HFilter = vp9_bilinear_filters[xoffset];
VFilter = vp9_bilinear_filters[yoffset];
HFilter = VP9_BILINEAR_FILTERS_2TAP(xoffset);
VFilter = VP9_BILINEAR_FILTERS_2TAP(yoffset);
var_filter_block2d_bil_first_pass(src_ptr, FData3,
src_pixels_per_line, 1, 3, 16, HFilter);
@ -108,8 +108,8 @@ unsigned int vp9_sub_pixel_variance2x16_c(const uint8_t *src_ptr,
uint8_t temp2[2 * 16];
const int16_t *HFilter, *VFilter;
HFilter = vp9_bilinear_filters[xoffset];
VFilter = vp9_bilinear_filters[yoffset];
HFilter = VP9_BILINEAR_FILTERS_2TAP(xoffset);
VFilter = VP9_BILINEAR_FILTERS_2TAP(yoffset);
var_filter_block2d_bil_first_pass(src_ptr, FData3,
src_pixels_per_line, 1, 17, 2, HFilter);
@ -118,10 +118,12 @@ unsigned int vp9_sub_pixel_variance2x16_c(const uint8_t *src_ptr,
return vp9_variance2x16_c(temp2, 2, dst_ptr, dst_pixels_per_line, sse);
}
#if CONFIG_USESELECTREFMV
/* check a list of motion vectors by sad score using a number rows of pixels
* above and a number cols of pixels in the left to select the one with best
* score to use as ref motion vector
*/
void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
uint8_t *ref_y_buffer,
int ref_y_stride,
@ -141,130 +143,140 @@ void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
int_mv sorted_mvs[MAX_MV_REF_CANDIDATES];
int zero_seen = FALSE;
// Default all to 0,0 if nothing else available
nearest->as_int = near->as_int = 0;
vpx_memset(sorted_mvs, 0, sizeof(sorted_mvs));
if (ref_y_buffer) {
above_src = xd->dst.y_buffer - xd->dst.y_stride * 2;
above_ref = ref_y_buffer - ref_y_stride * 2;
// Default all to 0,0 if nothing else available
nearest->as_int = near->as_int = 0;
vpx_memset(sorted_mvs, 0, sizeof(sorted_mvs));
above_src = xd->dst.y_buffer - xd->dst.y_stride * 2;
above_ref = ref_y_buffer - ref_y_stride * 2;
#if CONFIG_ABOVESPREFMV
above_src -= 4;
above_ref -= 4;
above_src -= 4;
above_ref -= 4;
#else
left_src = xd->dst.y_buffer - 2;
left_ref = ref_y_buffer - 2;
left_src = xd->dst.y_buffer - 2;
left_ref = ref_y_buffer - 2;
#endif
// Limit search to the predicted best few candidates
for(i = 0; i < MAX_MV_REF_CANDIDATES; ++i) {
int_mv this_mv;
int offset = 0;
int row_offset, col_offset;
// Limit search to the predicted best few candidates
for (i = 0; i < MAX_MV_REF_CANDIDATES; ++i) {
int_mv this_mv;
int offset = 0;
int row_offset, col_offset;
this_mv.as_int = mvlist[i].as_int;
this_mv.as_int = mvlist[i].as_int;
// If we see a 0,0 vector for a second time we have reached the end of
// the list of valid candidate vectors.
if (!this_mv.as_int && zero_seen)
break;
zero_seen = zero_seen || !this_mv.as_int;
#if !CONFIG_ABOVESPREFMV
clamp_mv(&this_mv,
xd->mb_to_left_edge - LEFT_TOP_MARGIN + 24,
xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN,
xd->mb_to_top_edge - LEFT_TOP_MARGIN + 24,
xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN);
#else
clamp_mv(&this_mv,
xd->mb_to_left_edge - LEFT_TOP_MARGIN + 32,
xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN,
xd->mb_to_top_edge - LEFT_TOP_MARGIN + 24,
xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN);
#endif
row_offset = this_mv.as_mv.row >> 3;
col_offset = this_mv.as_mv.col >> 3;
offset = ref_y_stride * row_offset + col_offset;
score = 0;
if (xd->up_available) {
vp9_sub_pixel_variance16x2(above_ref + offset, ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
above_src, xd->dst.y_stride, &sse);
score += sse;
if (xd->mode_info_context->mbmi.sb_type >= BLOCK_SIZE_SB32X32) {
vp9_sub_pixel_variance16x2(above_ref + offset + 16,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
above_src + 16, xd->dst.y_stride, &sse);
score += sse;
}
if (xd->mode_info_context->mbmi.sb_type >= BLOCK_SIZE_SB64X64) {
vp9_sub_pixel_variance16x2(above_ref + offset + 32,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
above_src + 32, xd->dst.y_stride, &sse);
score += sse;
vp9_sub_pixel_variance16x2(above_ref + offset + 48,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
above_src + 48, xd->dst.y_stride, &sse);
score += sse;
}
}
#if !CONFIG_ABOVESPREFMV
if (xd->left_available) {
vp9_sub_pixel_variance2x16_c(left_ref + offset, ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
left_src, xd->dst.y_stride, &sse);
score += sse;
if (xd->mode_info_context->mbmi.sb_type >= BLOCK_SIZE_SB32X32) {
vp9_sub_pixel_variance2x16_c(left_ref + offset + ref_y_stride * 16,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
left_src + xd->dst.y_stride * 16,
xd->dst.y_stride, &sse);
score += sse;
}
if (xd->mode_info_context->mbmi.sb_type >= BLOCK_SIZE_SB64X64) {
vp9_sub_pixel_variance2x16_c(left_ref + offset + ref_y_stride * 32,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
left_src + xd->dst.y_stride * 32,
xd->dst.y_stride, &sse);
score += sse;
vp9_sub_pixel_variance2x16_c(left_ref + offset + ref_y_stride * 48,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
left_src + xd->dst.y_stride * 48,
xd->dst.y_stride, &sse);
score += sse;
}
}
#endif
// Add the entry to our list and then resort the list on score.
ref_scores[i] = score;
sorted_mvs[i].as_int = this_mv.as_int;
j = i;
while (j > 0) {
if (ref_scores[j] < ref_scores[j-1]) {
ref_scores[j] = ref_scores[j-1];
sorted_mvs[j].as_int = sorted_mvs[j-1].as_int;
ref_scores[j-1] = score;
sorted_mvs[j-1].as_int = this_mv.as_int;
j--;
} else
// If we see a 0,0 vector for a second time we have reached the end of
// the list of valid candidate vectors.
if (!this_mv.as_int && zero_seen)
break;
zero_seen = zero_seen || !this_mv.as_int;
#if !CONFIG_ABOVESPREFMV
clamp_mv(&this_mv,
xd->mb_to_left_edge - LEFT_TOP_MARGIN + 24,
xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN,
xd->mb_to_top_edge - LEFT_TOP_MARGIN + 24,
xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN);
#else
clamp_mv(&this_mv,
xd->mb_to_left_edge - LEFT_TOP_MARGIN + 32,
xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN,
xd->mb_to_top_edge - LEFT_TOP_MARGIN + 24,
xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN);
#endif
row_offset = this_mv.as_mv.row >> 3;
col_offset = this_mv.as_mv.col >> 3;
offset = ref_y_stride * row_offset + col_offset;
score = 0;
#if !CONFIG_ABOVESPREFMV
if (xd->up_available) {
#else
if (xd->up_available && xd->left_available) {
#endif
vp9_sub_pixel_variance16x2(above_ref + offset, ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
above_src, xd->dst.y_stride, &sse);
score += sse;
if (xd->mode_info_context->mbmi.sb_type >= BLOCK_SIZE_SB32X32) {
vp9_sub_pixel_variance16x2(above_ref + offset + 16,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
above_src + 16, xd->dst.y_stride, &sse);
score += sse;
}
if (xd->mode_info_context->mbmi.sb_type >= BLOCK_SIZE_SB64X64) {
vp9_sub_pixel_variance16x2(above_ref + offset + 32,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
above_src + 32, xd->dst.y_stride, &sse);
score += sse;
vp9_sub_pixel_variance16x2(above_ref + offset + 48,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
above_src + 48, xd->dst.y_stride, &sse);
score += sse;
}
}
#if !CONFIG_ABOVESPREFMV
if (xd->left_available) {
vp9_sub_pixel_variance2x16_c(left_ref + offset, ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
left_src, xd->dst.y_stride, &sse);
score += sse;
if (xd->mode_info_context->mbmi.sb_type >= BLOCK_SIZE_SB32X32) {
vp9_sub_pixel_variance2x16_c(left_ref + offset + ref_y_stride * 16,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
left_src + xd->dst.y_stride * 16,
xd->dst.y_stride, &sse);
score += sse;
}
if (xd->mode_info_context->mbmi.sb_type >= BLOCK_SIZE_SB64X64) {
vp9_sub_pixel_variance2x16_c(left_ref + offset + ref_y_stride * 32,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
left_src + xd->dst.y_stride * 32,
xd->dst.y_stride, &sse);
score += sse;
vp9_sub_pixel_variance2x16_c(left_ref + offset + ref_y_stride * 48,
ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
left_src + xd->dst.y_stride * 48,
xd->dst.y_stride, &sse);
score += sse;
}
}
#endif
// Add the entry to our list and then resort the list on score.
ref_scores[i] = score;
sorted_mvs[i].as_int = this_mv.as_int;
j = i;
while (j > 0) {
if (ref_scores[j] < ref_scores[j-1]) {
ref_scores[j] = ref_scores[j-1];
sorted_mvs[j].as_int = sorted_mvs[j-1].as_int;
ref_scores[j-1] = score;
sorted_mvs[j-1].as_int = this_mv.as_int;
j--;
} else {
break;
}
}
}
} else {
vpx_memcpy(sorted_mvs, mvlist, sizeof(sorted_mvs));
}
// Make sure all the candidates are properly clamped etc
@ -273,23 +285,35 @@ void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
clamp_mv2(&sorted_mvs[i], xd);
}
// Provided that there are non zero vectors available there will not
// be more than one 0,0 entry in the sorted list.
// The best ref mv is always set to the first entry (which gave the best
// results. The nearest is set to the first non zero vector if available and
// near to the second non zero vector if available.
// We do not use 0,0 as a nearest or near as 0,0 has its own mode.
if ( sorted_mvs[0].as_int ) {
nearest->as_int = sorted_mvs[0].as_int;
if ( sorted_mvs[1].as_int )
near->as_int = sorted_mvs[1].as_int;
else
near->as_int = sorted_mvs[2].as_int;
// Nearest may be a 0,0 or non zero vector and now matches the chosen
// "best reference". This has advantages when it is used as part of a
// compound predictor as it means a non zero vector can be paired using
// this mode with a 0 vector. The Near vector is still forced to be a
// non zero candidate if one is avaialble.
nearest->as_int = sorted_mvs[0].as_int;
if ( sorted_mvs[1].as_int ) {
near->as_int = sorted_mvs[1].as_int;
} else {
nearest->as_int = sorted_mvs[1].as_int;
near->as_int = sorted_mvs[2].as_int;
near->as_int = sorted_mvs[2].as_int;
}
// Copy back the re-ordered mv list
vpx_memcpy(mvlist, sorted_mvs, sizeof(sorted_mvs));
}
#else
void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
uint8_t *ref_y_buffer,
int ref_y_stride,
int_mv *mvlist,
int_mv *nearest,
int_mv *near) {
int i;
// Make sure all the candidates are properly clamped etc
for (i = 0; i < MAX_MV_REF_CANDIDATES; ++i) {
lower_mv_precision(&mvlist[i], xd->allow_high_precision_mv);
clamp_mv2(&mvlist[i], xd);
}
*nearest = mvlist[0];
*near = mvlist[1];
}
#endif

46
vp9/common/vp9_findnearmv.h
View File

@ -17,6 +17,9 @@
#include "vp9/common/vp9_treecoder.h"
#include "vp9/common/vp9_onyxc_int.h"
#define LEFT_TOP_MARGIN (16 << 3)
#define RIGHT_BOTTOM_MARGIN (16 << 3)
/* check a list of motion vectors by sad score using a number rows of pixels
* above and a number cols of pixels in the left to select the one with best
* score to use as ref motion vector
@ -28,9 +31,9 @@ void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
int_mv *nearest,
int_mv *near);
static void mv_bias(int refmb_ref_frame_sign_bias, int refframe, int_mv *mvp, const int *ref_frame_sign_bias) {
MV xmv;
xmv = mvp->as_mv;
static void mv_bias(int refmb_ref_frame_sign_bias, int refframe,
int_mv *mvp, const int *ref_frame_sign_bias) {
MV xmv = mvp->as_mv;
if (refmb_ref_frame_sign_bias != ref_frame_sign_bias[refframe]) {
xmv.row *= -1;
@ -40,8 +43,6 @@ static void mv_bias(int refmb_ref_frame_sign_bias, int refframe, int_mv *mvp, co
mvp->as_mv = xmv;
}
#define LEFT_TOP_MARGIN (16 << 3)
#define RIGHT_BOTTOM_MARGIN (16 << 3)
static void clamp_mv(int_mv *mv,
int mb_to_left_edge,
@ -71,10 +72,10 @@ static unsigned int check_mv_bounds(int_mv *mv,
int mb_to_right_edge,
int mb_to_top_edge,
int mb_to_bottom_edge) {
return (mv->as_mv.col < mb_to_left_edge) ||
(mv->as_mv.col > mb_to_right_edge) ||
(mv->as_mv.row < mb_to_top_edge) ||
(mv->as_mv.row > mb_to_bottom_edge);
return mv->as_mv.col < mb_to_left_edge ||
mv->as_mv.col > mb_to_right_edge ||
mv->as_mv.row < mb_to_top_edge ||
mv->as_mv.row > mb_to_bottom_edge;
}
vp9_prob *vp9_mv_ref_probs(VP9_COMMON *pc,
@ -83,21 +84,30 @@ vp9_prob *vp9_mv_ref_probs(VP9_COMMON *pc,
extern const uint8_t vp9_mbsplit_offset[4][16];
static int left_block_mv(const MODE_INFO *cur_mb, int b) {
static int left_block_mv(const MACROBLOCKD *xd,
const MODE_INFO *cur_mb, int b) {
if (!(b & 3)) {
/* On L edge, get from MB to left of us */
if (!xd->left_available)
return 0;
// On L edge, get from MB to left of us
--cur_mb;
if (cur_mb->mbmi.mode != SPLITMV)
return cur_mb->mbmi.mv[0].as_int;
b += 4;
}
return (cur_mb->bmi + b - 1)->as_mv.first.as_int;
return (cur_mb->bmi + b - 1)->as_mv[0].as_int;
}
static int left_block_second_mv(const MODE_INFO *cur_mb, int b) {
static int left_block_second_mv(const MACROBLOCKD *xd,
const MODE_INFO *cur_mb, int b) {
if (!(b & 3)) {
if (!xd->left_available)
return 0;
/* On L edge, get from MB to left of us */
--cur_mb;
@ -108,8 +118,8 @@ static int left_block_second_mv(const MODE_INFO *cur_mb, int b) {
}
return cur_mb->mbmi.second_ref_frame > 0 ?
(cur_mb->bmi + b - 1)->as_mv.second.as_int :
(cur_mb->bmi + b - 1)->as_mv.first.as_int;
(cur_mb->bmi + b - 1)->as_mv[1].as_int :
(cur_mb->bmi + b - 1)->as_mv[0].as_int;
}
static int above_block_mv(const MODE_INFO *cur_mb, int b, int mi_stride) {
@ -122,7 +132,7 @@ static int above_block_mv(const MODE_INFO *cur_mb, int b, int mi_stride) {
b += 16;
}
return (cur_mb->bmi + b - 4)->as_mv.first.as_int;
return (cur_mb->bmi + b - 4)->as_mv[0].as_int;
}
static int above_block_second_mv(const MODE_INFO *cur_mb, int b, int mi_stride) {
@ -137,8 +147,8 @@ static int above_block_second_mv(const MODE_INFO *cur_mb, int b, int mi_stride)
}
return cur_mb->mbmi.second_ref_frame > 0 ?
(cur_mb->bmi + b - 4)->as_mv.second.as_int :
(cur_mb->bmi + b - 4)->as_mv.first.as_int;
(cur_mb->bmi + b - 4)->as_mv[1].as_int :
(cur_mb->bmi + b - 4)->as_mv[0].as_int;
}
static B_PREDICTION_MODE left_block_mode(const MODE_INFO *cur_mb, int b) {

1307
vp9/common/vp9_idct.c Normal file
View File

File diff suppressed because it is too large Load Diff

85
vp9/common/vp9_idct.h Normal file
View File

@ -0,0 +1,85 @@
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_IDCT_H_
#define VP9_COMMON_VP9_IDCT_H_
#include <assert.h>
#include "./vpx_config.h"
#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_common.h"
// Constants and Macros used by all idct/dct functions
#define DCT_CONST_BITS 14
#define DCT_CONST_ROUNDING (1 << (DCT_CONST_BITS - 1))
#define pair_set_epi16(a, b) \
_mm_set1_epi32(((uint16_t)(a)) + (((uint16_t)(b)) << 16))
// Constants are round(16384 * cos(k*Pi/64)) where k = 1 to 31.
// Note: sin(k*Pi/64) = cos((32-k)*Pi/64)
static const int cospi_1_64 = 16364;
static const int cospi_2_64 = 16305;
static const int cospi_3_64 = 16207;
static const int cospi_4_64 = 16069;
static const int cospi_5_64 = 15893;
static const int cospi_6_64 = 15679;
static const int cospi_7_64 = 15426;
static const int cospi_8_64 = 15137;
static const int cospi_9_64 = 14811;
static const int cospi_10_64 = 14449;
static const int cospi_11_64 = 14053;
static const int cospi_12_64 = 13623;
static const int cospi_13_64 = 13160;
static const int cospi_14_64 = 12665;
static const int cospi_15_64 = 12140;
static const int cospi_16_64 = 11585;
static const int cospi_17_64 = 11003;
static const int cospi_18_64 = 10394;
static const int cospi_19_64 = 9760;
static const int cospi_20_64 = 9102;
static const int cospi_21_64 = 8423;
static const int cospi_22_64 = 7723;
static const int cospi_23_64 = 7005;
static const int cospi_24_64 = 6270;
static const int cospi_25_64 = 5520;
static const int cospi_26_64 = 4756;
static const int cospi_27_64 = 3981;
static const int cospi_28_64 = 3196;
static const int cospi_29_64 = 2404;
static const int cospi_30_64 = 1606;
static const int cospi_31_64 = 804;
// 16384 * sqrt(2) * sin(kPi/9) * 2 / 3
static const int sinpi_1_9 = 5283;
static const int sinpi_2_9 = 9929;
static const int sinpi_3_9 = 13377;
static const int sinpi_4_9 = 15212;
static INLINE int dct_const_round_shift(int input) {
int rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS);
assert(INT16_MIN <= rv && rv <= INT16_MAX);
return rv;
}
static INLINE int dct_32_round(int input) {
int rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS);
assert(-131072 <= rv && rv <= 131071);
return rv;
}
typedef void (*transform_1d)(int16_t*, int16_t*);
typedef struct {
transform_1d cols, rows; // vertical and horizontal
} transform_2d;
#endif // VP9_COMMON_VP9_IDCT_H_

2670
vp9/common/vp9_idctllm.c
View File

File diff suppressed because it is too large Load Diff

279
vp9/common/vp9_invtrans.c
View File

@ -11,50 +11,25 @@
#include "vp9/common/vp9_invtrans.h"
#include "./vp9_rtcd.h"
static void recon_dcblock(MACROBLOCKD *xd) {
BLOCKD *b = &xd->block[24];
int i;
for (i = 0; i < 16; i++) {
xd->block[i].dqcoeff[0] = b->diff[i];
}
}
static void recon_dcblock_8x8(MACROBLOCKD *xd) {
BLOCKD *b = &xd->block[24]; // for coeff 0, 2, 8, 10
xd->block[0].dqcoeff[0] = b->diff[0];
xd->block[4].dqcoeff[0] = b->diff[1];
xd->block[8].dqcoeff[0] = b->diff[4];
xd->block[12].dqcoeff[0] = b->diff[8];
}
void vp9_inverse_transform_b_4x4(MACROBLOCKD *xd, int block, int pitch) {
BLOCKD *b = &xd->block[block];
if (b->eob <= 1)
xd->inv_xform4x4_1_x8(b->dqcoeff, b->diff, pitch);
void vp9_inverse_transform_b_4x4(MACROBLOCKD *xd, int eob,
int16_t *dqcoeff, int16_t *diff,
int pitch) {
if (eob <= 1)
xd->inv_txm4x4_1(dqcoeff, diff, pitch);
else
xd->inv_xform4x4_x8(b->dqcoeff, b->diff, pitch);
xd->inv_txm4x4(dqcoeff, diff, pitch);
}
void vp9_inverse_transform_mby_4x4(MACROBLOCKD *xd) {
int i;
BLOCKD *blockd = xd->block;
int has_2nd_order = get_2nd_order_usage(xd);
if (has_2nd_order) {
/* do 2nd order transform on the dc block */
vp9_short_inv_walsh4x4(blockd[24].dqcoeff, blockd[24].diff);
recon_dcblock(xd);
}
for (i = 0; i < 16; i++) {
TX_TYPE tx_type = get_tx_type_4x4(xd, &xd->block[i]);
TX_TYPE tx_type = get_tx_type_4x4(xd, i);
if (tx_type != DCT_DCT) {
vp9_ihtllm(xd->block[i].dqcoeff, xd->block[i].diff, 32,
tx_type, 4, xd->block[i].eob);
vp9_short_iht4x4(xd->block[i].dqcoeff, xd->block[i].diff, 16, tx_type);
} else {
vp9_inverse_transform_b_4x4(xd, i, 32);
vp9_inverse_transform_b_4x4(xd, xd->eobs[i], xd->block[i].dqcoeff,
xd->block[i].diff, 32);
}
}
}
@ -63,7 +38,8 @@ void vp9_inverse_transform_mbuv_4x4(MACROBLOCKD *xd) {
int i;
for (i = 16; i < 24; i++) {
vp9_inverse_transform_b_4x4(xd, i, 16);
vp9_inverse_transform_b_4x4(xd, xd->eobs[i], xd->block[i].dqcoeff,
xd->block[i].diff, 16);
}
}
@ -80,29 +56,21 @@ void vp9_inverse_transform_b_8x8(int16_t *input_dqcoeff, int16_t *output_coeff,
void vp9_inverse_transform_mby_8x8(MACROBLOCKD *xd) {
int i;
BLOCKD *blockd = xd->block;
int has_2nd_order = get_2nd_order_usage(xd);
if (has_2nd_order) {
// do 2nd order transform on the dc block
vp9_short_ihaar2x2(blockd[24].dqcoeff, blockd[24].diff, 8);
recon_dcblock_8x8(xd); // need to change for 8x8
}
for (i = 0; i < 9; i += 8) {
TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[i]);
TX_TYPE tx_type = get_tx_type_8x8(xd, i);
if (tx_type != DCT_DCT) {
vp9_ihtllm(xd->block[i].dqcoeff, xd->block[i].diff, 32, tx_type, 8,
xd->block[i].eob);
vp9_short_iht8x8(xd->block[i].dqcoeff, xd->block[i].diff, 16, tx_type);
} else {
vp9_inverse_transform_b_8x8(&blockd[i].dqcoeff[0],
&blockd[i].diff[0], 32);
}
}
for (i = 2; i < 11; i += 8) {
TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[i]);
TX_TYPE tx_type = get_tx_type_8x8(xd, i);
if (tx_type != DCT_DCT) {
vp9_ihtllm(xd->block[i + 2].dqcoeff, xd->block[i].diff, 32, tx_type, 8,
xd->block[i + 2].eob);
vp9_short_iht8x8(xd->block[i + 2].dqcoeff, xd->block[i].diff,
16, tx_type);
} else {
vp9_inverse_transform_b_8x8(&blockd[i + 2].dqcoeff[0],
&blockd[i].diff[0], 32);
@ -132,9 +100,9 @@ void vp9_inverse_transform_b_16x16(int16_t *input_dqcoeff,
void vp9_inverse_transform_mby_16x16(MACROBLOCKD *xd) {
BLOCKD *bd = &xd->block[0];
TX_TYPE tx_type = get_tx_type_16x16(xd, bd);
TX_TYPE tx_type = get_tx_type_16x16(xd, 0);
if (tx_type != DCT_DCT) {
vp9_ihtllm(bd->dqcoeff, bd->diff, 32, tx_type, 16, bd->eob);
vp9_short_iht16x16(bd->dqcoeff, bd->diff, 16, tx_type);
} else {
vp9_inverse_transform_b_16x16(&xd->block[0].dqcoeff[0],
&xd->block[0].diff[0], 32);
@ -146,13 +114,208 @@ void vp9_inverse_transform_mb_16x16(MACROBLOCKD *xd) {
vp9_inverse_transform_mbuv_8x8(xd);
}
void vp9_inverse_transform_sby_32x32(SUPERBLOCKD *xd_sb) {
vp9_short_idct32x32(xd_sb->dqcoeff, xd_sb->diff, 64);
void vp9_inverse_transform_sby_32x32(MACROBLOCKD *xd) {
vp9_short_idct32x32(xd->dqcoeff, xd->diff, 64);
}
void vp9_inverse_transform_sbuv_16x16(SUPERBLOCKD *xd_sb) {
vp9_inverse_transform_b_16x16(xd_sb->dqcoeff + 1024,
xd_sb->diff + 1024, 32);
vp9_inverse_transform_b_16x16(xd_sb->dqcoeff + 1280,
xd_sb->diff + 1280, 32);
void vp9_inverse_transform_sby_16x16(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 4; n++) {
const int x_idx = n & 1, y_idx = n >> 1;
const TX_TYPE tx_type = get_tx_type_16x16(xd, (y_idx * 8 + x_idx) * 4);
if (tx_type == DCT_DCT) {
vp9_inverse_transform_b_16x16(xd->dqcoeff + n * 256,
xd->diff + x_idx * 16 + y_idx * 32 * 16,
64);
} else {
vp9_short_iht16x16(xd->dqcoeff + n * 256,
xd->diff + x_idx * 16 + y_idx * 32 * 16, 32, tx_type);
}
}
}
void vp9_inverse_transform_sby_8x8(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 16; n++) {
const int x_idx = n & 3, y_idx = n >> 2;
const TX_TYPE tx_type = get_tx_type_8x8(xd, (y_idx * 8 + x_idx) * 2);
if (tx_type == DCT_DCT) {
vp9_inverse_transform_b_8x8(xd->dqcoeff + n * 64,
xd->diff + x_idx * 8 + y_idx * 32 * 8, 64);
} else {
vp9_short_iht8x8(xd->dqcoeff + n * 64,
xd->diff + x_idx * 8 + y_idx * 32 * 8, 32, tx_type);
}
}
}
void vp9_inverse_transform_sby_4x4(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 64; n++) {
const int x_idx = n & 7, y_idx = n >> 3;
const TX_TYPE tx_type = get_tx_type_4x4(xd, y_idx * 8 + x_idx);
if (tx_type == DCT_DCT) {
vp9_inverse_transform_b_4x4(xd, xd->eobs[n], xd->dqcoeff + n * 16,
xd->diff + x_idx * 4 + y_idx * 4 * 32, 64);
} else {
vp9_short_iht4x4(xd->dqcoeff + n * 16,
xd->diff + x_idx * 4 + y_idx * 4 * 32, 32, tx_type);
}
}
}
void vp9_inverse_transform_sbuv_16x16(MACROBLOCKD *xd) {
vp9_inverse_transform_b_16x16(xd->dqcoeff + 1024,
xd->diff + 1024, 32);
vp9_inverse_transform_b_16x16(xd->dqcoeff + 1280,
xd->diff + 1280, 32);
}
void vp9_inverse_transform_sbuv_8x8(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 4; n++) {
const int x_idx = n & 1, y_idx = n >> 1;
vp9_inverse_transform_b_8x8(xd->dqcoeff + 1024 + n * 64,
xd->diff + 1024 + x_idx * 8 + y_idx * 16 * 8,
32);
vp9_inverse_transform_b_8x8(xd->dqcoeff + 1280 + n * 64,
xd->diff + 1280 + x_idx * 8 + y_idx * 16 * 8,
32);
}
}
void vp9_inverse_transform_sbuv_4x4(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 16; n++) {
const int x_idx = n & 3, y_idx = n >> 2;
vp9_inverse_transform_b_4x4(xd, xd->eobs[64 + n],
xd->dqcoeff + 1024 + n * 16,
xd->diff + 1024 + x_idx * 4 + y_idx * 16 * 4,
32);
vp9_inverse_transform_b_4x4(xd, xd->eobs[64 + 16 + n],
xd->dqcoeff + 1280 + n * 16,
xd->diff + 1280 + x_idx * 4 + y_idx * 16 * 4,
32);
}
}
void vp9_inverse_transform_sb64y_32x32(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 4; n++) {
const int x_idx = n & 1, y_idx = n >> 1;
vp9_short_idct32x32(xd->dqcoeff + n * 1024,
xd->diff + x_idx * 32 + y_idx * 32 * 64, 128);
}
}
void vp9_inverse_transform_sb64y_16x16(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 16; n++) {
const int x_idx = n & 3, y_idx = n >> 2;
const TX_TYPE tx_type = get_tx_type_16x16(xd, (y_idx * 16 + x_idx) * 4);
if (tx_type == DCT_DCT) {
vp9_inverse_transform_b_16x16(xd->dqcoeff + n * 256,
xd->diff + x_idx * 16 + y_idx * 64 * 16,
128);
} else {
vp9_short_iht16x16(xd->dqcoeff + n * 256,
xd->diff + x_idx * 16 + y_idx * 64 * 16, 64, tx_type);
}
}
}
void vp9_inverse_transform_sb64y_8x8(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 64; n++) {
const int x_idx = n & 7, y_idx = n >> 3;
const TX_TYPE tx_type = get_tx_type_8x8(xd, (y_idx * 16 + x_idx) * 2);
if (tx_type == DCT_DCT) {
vp9_inverse_transform_b_8x8(xd->dqcoeff + n * 64,
xd->diff + x_idx * 8 + y_idx * 64 * 8, 128);
} else {
vp9_short_iht8x8(xd->dqcoeff + n * 64,
xd->diff + x_idx * 8 + y_idx * 64 * 8, 64, tx_type);
}
}
}
void vp9_inverse_transform_sb64y_4x4(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 256; n++) {
const int x_idx = n & 15, y_idx = n >> 4;
const TX_TYPE tx_type = get_tx_type_4x4(xd, y_idx * 16 + x_idx);
if (tx_type == DCT_DCT) {
vp9_inverse_transform_b_4x4(xd, xd->eobs[n], xd->dqcoeff + n * 16,
xd->diff + x_idx * 4 + y_idx * 4 * 64, 128);
} else {
vp9_short_iht4x4(xd->dqcoeff + n * 16,
xd->diff + x_idx * 4 + y_idx * 4 * 64, 64, tx_type);
}
}
}
void vp9_inverse_transform_sb64uv_32x32(MACROBLOCKD *xd) {
vp9_short_idct32x32(xd->dqcoeff + 4096,
xd->diff + 4096, 64);
vp9_short_idct32x32(xd->dqcoeff + 4096 + 1024,
xd->diff + 4096 + 1024, 64);
}
void vp9_inverse_transform_sb64uv_16x16(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 4; n++) {
const int x_idx = n & 1, y_idx = n >> 1, off = x_idx * 16 + y_idx * 32 * 16;
vp9_inverse_transform_b_16x16(xd->dqcoeff + 4096 + n * 256,
xd->diff + 4096 + off, 64);
vp9_inverse_transform_b_16x16(xd->dqcoeff + 4096 + 1024 + n * 256,
xd->diff + 4096 + 1024 + off, 64);
}
}
void vp9_inverse_transform_sb64uv_8x8(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 16; n++) {
const int x_idx = n & 3, y_idx = n >> 2, off = x_idx * 8 + y_idx * 32 * 8;
vp9_inverse_transform_b_8x8(xd->dqcoeff + 4096 + n * 64,
xd->diff + 4096 + off, 64);
vp9_inverse_transform_b_8x8(xd->dqcoeff + 4096 + 1024 + n * 64,
xd->diff + 4096 + 1024 + off, 64);
}
}
void vp9_inverse_transform_sb64uv_4x4(MACROBLOCKD *xd) {
int n;
for (n = 0; n < 64; n++) {
const int x_idx = n & 7, y_idx = n >> 3, off = x_idx * 4 + y_idx * 32 * 4;
vp9_inverse_transform_b_4x4(xd, xd->eobs[256 + n],
xd->dqcoeff + 4096 + n * 16,
xd->diff + 4096 + off, 64);
vp9_inverse_transform_b_4x4(xd, xd->eobs[256 + 64 + n],
xd->dqcoeff + 4096 + 1024 + n * 16,
xd->diff + 4096 + 1024 + off, 64);
}
}

42
vp9/common/vp9_invtrans.h
View File

@ -15,31 +15,47 @@
#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_blockd.h"
extern void vp9_inverse_transform_b_4x4(MACROBLOCKD *xd, int block, int pitch);
void vp9_inverse_transform_b_4x4(MACROBLOCKD *xd, int eob,
int16_t *dqcoeff, int16_t *diff,
int pitch);
extern void vp9_inverse_transform_mb_4x4(MACROBLOCKD *xd);
void vp9_inverse_transform_mb_4x4(MACROBLOCKD *xd);
extern void vp9_inverse_transform_mby_4x4(MACROBLOCKD *xd);
void vp9_inverse_transform_mby_4x4(MACROBLOCKD *xd);
extern void vp9_inverse_transform_mbuv_4x4(MACROBLOCKD *xd);
void vp9_inverse_transform_mbuv_4x4(MACROBLOCKD *xd);
extern void vp9_inverse_transform_b_8x8(int16_t *input_dqcoeff,
void vp9_inverse_transform_b_8x8(int16_t *input_dqcoeff,
int16_t *output_coeff, int pitch);
extern void vp9_inverse_transform_mb_8x8(MACROBLOCKD *xd);
void vp9_inverse_transform_mb_8x8(MACROBLOCKD *xd);
extern void vp9_inverse_transform_mby_8x8(MACROBLOCKD *xd);
void vp9_inverse_transform_mby_8x8(MACROBLOCKD *xd);
extern void vp9_inverse_transform_mbuv_8x8(MACROBLOCKD *xd);
void vp9_inverse_transform_mbuv_8x8(MACROBLOCKD *xd);
extern void vp9_inverse_transform_b_16x16(int16_t *input_dqcoeff,
void vp9_inverse_transform_b_16x16(int16_t *input_dqcoeff,
int16_t *output_coeff, int pitch);
extern void vp9_inverse_transform_mb_16x16(MACROBLOCKD *xd);
void vp9_inverse_transform_mb_16x16(MACROBLOCKD *xd);
extern void vp9_inverse_transform_mby_16x16(MACROBLOCKD *xd);
void vp9_inverse_transform_mby_16x16(MACROBLOCKD *xd);
extern void vp9_inverse_transform_sby_32x32(SUPERBLOCKD *xd_sb);
extern void vp9_inverse_transform_sbuv_16x16(SUPERBLOCKD *xd_sb);
void vp9_inverse_transform_sby_32x32(MACROBLOCKD *xd);
void vp9_inverse_transform_sby_16x16(MACROBLOCKD *xd);
void vp9_inverse_transform_sby_8x8(MACROBLOCKD *xd);
void vp9_inverse_transform_sby_4x4(MACROBLOCKD *xd);
void vp9_inverse_transform_sbuv_16x16(MACROBLOCKD *xd);
void vp9_inverse_transform_sbuv_8x8(MACROBLOCKD *xd);
void vp9_inverse_transform_sbuv_4x4(MACROBLOCKD *xd);
void vp9_inverse_transform_sb64y_32x32(MACROBLOCKD *xd);
void vp9_inverse_transform_sb64y_16x16(MACROBLOCKD *xd);
void vp9_inverse_transform_sb64y_8x8(MACROBLOCKD *xd);
void vp9_inverse_transform_sb64y_4x4(MACROBLOCKD *xd);
void vp9_inverse_transform_sb64uv_32x32(MACROBLOCKD *xd);
void vp9_inverse_transform_sb64uv_16x16(MACROBLOCKD *xd);
void vp9_inverse_transform_sb64uv_8x8(MACROBLOCKD *xd);
void vp9_inverse_transform_sb64uv_4x4(MACROBLOCKD *xd);
#endif // VP9_COMMON_VP9_INVTRANS_H_

88
vp9/common/vp9_loopfilter.c
View File

@ -109,6 +109,9 @@ void vp9_loop_filter_frame_init(VP9_COMMON *cm,
loop_filter_info_n *lfi = &cm->lf_info;
/* update limits if sharpness has changed */
// printf("vp9_loop_filter_frame_init %d\n", default_filt_lvl);
// printf("sharpness level: %d [%d]\n",
// cm->sharpness_level, cm->last_sharpness_level);
if (cm->last_sharpness_level != cm->sharpness_level) {
vp9_loop_filter_update_sharpness(lfi, cm->sharpness_level);
cm->last_sharpness_level = cm->sharpness_level;
@ -126,7 +129,7 @@ void vp9_loop_filter_frame_init(VP9_COMMON *cm,
lvl_seg = vp9_get_segdata(xd, seg, SEG_LVL_ALT_LF);
} else { /* Delta Value */
lvl_seg += vp9_get_segdata(xd, seg, SEG_LVL_ALT_LF);
lvl_seg = (lvl_seg > 0) ? ((lvl_seg > 63) ? 63 : lvl_seg) : 0;
lvl_seg = clamp(lvl_seg, 0, 63);
}
}
@ -149,13 +152,12 @@ void vp9_loop_filter_frame_init(VP9_COMMON *cm,
/* Apply delta for Intra modes */
mode = 0; /* B_PRED */
/* Only the split mode BPRED has a further special case */
lvl_mode = lvl_ref + xd->mode_lf_deltas[mode];
lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0; /* clamp */
lvl_mode = clamp(lvl_ref + xd->mode_lf_deltas[mode], 0, 63);
lfi->lvl[seg][ref][mode] = lvl_mode;
mode = 1; /* all the rest of Intra modes */
lvl_mode = (lvl_ref > 0) ? (lvl_ref > 63 ? 63 : lvl_ref) : 0; /* clamp */
lvl_mode = clamp(lvl_ref, 0, 63);
lfi->lvl[seg][ref][mode] = lvl_mode;
/* LAST, GOLDEN, ALT */
@ -167,9 +169,7 @@ void vp9_loop_filter_frame_init(VP9_COMMON *cm,
/* Apply delta for Inter modes */
for (mode = 1; mode < 4; mode++) {
lvl_mode = lvl_ref + xd->mode_lf_deltas[mode];
lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0; /* clamp */
lvl_mode = clamp(lvl_ref + xd->mode_lf_deltas[mode], 0, 63);
lfi->lvl[seg][ref][mode] = lvl_mode;
}
}
@ -202,10 +202,12 @@ static int sb_mb_lf_skip(const MODE_INFO *const mip0,
mbmi1->mv[mbmi1->ref_frame].as_int) &&
mbmi0->ref_frame != INTRA_FRAME;
}
void vp9_loop_filter_frame(VP9_COMMON *cm,
MACROBLOCKD *xd,
int frame_filter_level,
int y_only) {
int y_only,
int dering) {
YV12_BUFFER_CONFIG *post = cm->frame_to_show;
loop_filter_info_n *lfi_n = &cm->lf_info;
struct loop_filter_info lfi;
@ -271,7 +273,6 @@ void vp9_loop_filter_frame(VP9_COMMON *cm,
vp9_loop_filter_bv(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
}
}
/* don't apply across umv border */
if (mb_row > 0 &&
@ -299,6 +300,62 @@ void vp9_loop_filter_frame(VP9_COMMON *cm,
post->uv_stride, &lfi);
}
}
#if CONFIG_LOOP_DERING
if (dering) {
if (mb_row && mb_row < cm->mb_rows - 1 &&
mb_col && mb_col < cm->mb_cols - 1) {
vp9_post_proc_down_and_across(y_ptr, y_ptr,
post->y_stride, post->y_stride,
16, 16, dering);
if (!y_only) {
vp9_post_proc_down_and_across(u_ptr, u_ptr,
post->uv_stride, post->uv_stride,
8, 8, dering);
vp9_post_proc_down_and_across(v_ptr, v_ptr,
post->uv_stride, post->uv_stride,
8, 8, dering);
}
} else {
// Adjust the filter so that no out-of-frame data is used.
uint8_t *dr_y = y_ptr, *dr_u = u_ptr, *dr_v = v_ptr;
int w_adjust = 0;
int h_adjust = 0;
if (mb_col == 0) {
dr_y += 2;
dr_u += 2;
dr_v += 2;
w_adjust += 2;
}
if (mb_col == cm->mb_cols - 1)
w_adjust += 2;
if (mb_row == 0) {
dr_y += 2 * post->y_stride;
dr_u += 2 * post->uv_stride;
dr_v += 2 * post->uv_stride;
h_adjust += 2;
}
if (mb_row == cm->mb_rows - 1)
h_adjust += 2;
vp9_post_proc_down_and_across_c(dr_y, dr_y,
post->y_stride, post->y_stride,
16 - w_adjust, 16 - h_adjust,
dering);
if (!y_only) {
vp9_post_proc_down_and_across_c(dr_u, dr_u,
post->uv_stride,
post->uv_stride,
8 - w_adjust, 8 - h_adjust,
dering);
vp9_post_proc_down_and_across_c(dr_v, dr_v,
post->uv_stride,
post->uv_stride,
8 - w_adjust, 8 - h_adjust,
dering);
}
}
}
#endif
} else {
// FIXME: Not 8x8 aware
if (mb_col > 0 &&
@ -376,16 +433,13 @@ void vp9_loop_filter_partial_frame(VP9_COMMON *cm, MACROBLOCKD *xd,
*/
if (alt_flt_enabled) {
for (i = 0; i < MAX_MB_SEGMENTS; i++) {
/* Abs value */
if (xd->mb_segment_abs_delta == SEGMENT_ABSDATA) {
// Abs value
lvl_seg[i] = vp9_get_segdata(xd, i, SEG_LVL_ALT_LF);
}
/* Delta Value */
else {
lvl_seg[i] = default_filt_lvl +
vp9_get_segdata(xd, i, SEG_LVL_ALT_LF);
lvl_seg[i] = (lvl_seg[i] > 0) ?
((lvl_seg[i] > 63) ? 63 : lvl_seg[i]) : 0;
} else {
// Delta Value
lvl_seg[i] = default_filt_lvl + vp9_get_segdata(xd, i, SEG_LVL_ALT_LF);
lvl_seg[i] = clamp(lvl_seg[i], 0, 63);
}
}
}

3
vp9/common/vp9_loopfilter.h
View File

@ -83,7 +83,8 @@ void vp9_loop_filter_frame_init(struct VP9Common *cm,
void vp9_loop_filter_frame(struct VP9Common *cm,
struct macroblockd *mbd,
int filter_level,
int y_only);
int y_only,
int dering);
void vp9_loop_filter_partial_frame(struct VP9Common *cm,
struct macroblockd *mbd,

428
vp9/common/vp9_loopfilter_filters.c
View File

@ -13,7 +13,7 @@
#include "vp9/common/vp9_loopfilter.h"
#include "vp9/common/vp9_onyxc_int.h"
static __inline int8_t signed_char_clamp(int t) {
static INLINE int8_t signed_char_clamp(int t) {
t = (t < -128 ? -128 : t);
t = (t > 127 ? 127 : t);
return (int8_t) t;
@ -21,11 +21,11 @@ static __inline int8_t signed_char_clamp(int t) {
/* should we apply any filter at all ( 11111111 yes, 00000000 no) */
static __inline int8_t filter_mask(uint8_t limit, uint8_t blimit,
uint8_t p3, uint8_t p2,
uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1,
uint8_t q2, uint8_t q3) {
static INLINE int8_t filter_mask(uint8_t limit, uint8_t blimit,
uint8_t p3, uint8_t p2,
uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1,
uint8_t q2, uint8_t q3) {
int8_t mask = 0;
mask |= (abs(p3 - p2) > limit) * -1;
mask |= (abs(p2 - p1) > limit) * -1;
@ -39,57 +39,46 @@ static __inline int8_t filter_mask(uint8_t limit, uint8_t blimit,
}
/* is there high variance internal edge ( 11111111 yes, 00000000 no) */
static __inline int8_t hevmask(uint8_t thresh, uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1) {
static INLINE int8_t hevmask(uint8_t thresh, uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1) {
int8_t hev = 0;
hev |= (abs(p1 - p0) > thresh) * -1;
hev |= (abs(q1 - q0) > thresh) * -1;
return hev;
}
static __inline void filter(int8_t mask, uint8_t hev, uint8_t *op1,
uint8_t *op0, uint8_t *oq0, uint8_t *oq1) {
int8_t ps0, qs0;
int8_t ps1, qs1;
int8_t filter, Filter1, Filter2;
int8_t u;
static INLINE void filter(int8_t mask, uint8_t hev, uint8_t *op1,
uint8_t *op0, uint8_t *oq0, uint8_t *oq1) {
int8_t filter1, filter2;
ps1 = (int8_t) *op1 ^ 0x80;
ps0 = (int8_t) *op0 ^ 0x80;
qs0 = (int8_t) *oq0 ^ 0x80;
qs1 = (int8_t) *oq1 ^ 0x80;
const int8_t ps1 = (int8_t) *op1 ^ 0x80;
const int8_t ps0 = (int8_t) *op0 ^ 0x80;
const int8_t qs0 = (int8_t) *oq0 ^ 0x80;
const int8_t qs1 = (int8_t) *oq1 ^ 0x80;
/* add outer taps if we have high edge variance */
filter = signed_char_clamp(ps1 - qs1);
filter &= hev;
// add outer taps if we have high edge variance
int8_t filter = signed_char_clamp(ps1 - qs1) & hev;
/* inner taps */
filter = signed_char_clamp(filter + 3 * (qs0 - ps0));
filter &= mask;
// inner taps
filter = signed_char_clamp(filter + 3 * (qs0 - ps0)) & mask;
/* save bottom 3 bits so that we round one side +4 and the other +3
* if it equals 4 we'll set to adjust by -1 to account for the fact
* we'd round 3 the other way
*/
Filter1 = signed_char_clamp(filter + 4);
Filter2 = signed_char_clamp(filter + 3);
Filter1 >>= 3;
Filter2 >>= 3;
u = signed_char_clamp(qs0 - Filter1);
*oq0 = u ^ 0x80;
u = signed_char_clamp(ps0 + Filter2);
*op0 = u ^ 0x80;
filter = Filter1;
// save bottom 3 bits so that we round one side +4 and the other +3
// if it equals 4 we'll set to adjust by -1 to account for the fact
// we'd round 3 the other way
filter1 = signed_char_clamp(filter + 4) >> 3;
filter2 = signed_char_clamp(filter + 3) >> 3;
/* outer tap adjustments */
*oq0 = signed_char_clamp(qs0 - filter1) ^ 0x80;
*op0 = signed_char_clamp(ps0 + filter2) ^ 0x80;
filter = filter1;
// outer tap adjustments
filter += 1;
filter >>= 1;
filter &= ~hev;
u = signed_char_clamp(qs1 - filter);
*oq1 = u ^ 0x80;
u = signed_char_clamp(ps1 + filter);
*op1 = u ^ 0x80;
*oq1 = signed_char_clamp(qs1 - filter) ^ 0x80;
*op1 = signed_char_clamp(ps1 + filter) ^ 0x80;
}
void vp9_loop_filter_horizontal_edge_c(uint8_t *s,
@ -143,11 +132,11 @@ void vp9_loop_filter_vertical_edge_c(uint8_t *s,
s += p;
} while (++i < count * 8);
}
static __inline signed char flatmask(uint8_t thresh,
uint8_t p4, uint8_t p3, uint8_t p2,
uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1, uint8_t q2,
uint8_t q3, uint8_t q4) {
static INLINE signed char flatmask4(uint8_t thresh,
uint8_t p3, uint8_t p2,
uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1,
uint8_t q2, uint8_t q3) {
int8_t flat = 0;
flat |= (abs(p1 - p0) > thresh) * -1;
flat |= (abs(q1 - q0) > thresh) * -1;
@ -155,81 +144,72 @@ static __inline signed char flatmask(uint8_t thresh,
flat |= (abs(q0 - q2) > thresh) * -1;
flat |= (abs(p3 - p0) > thresh) * -1;
flat |= (abs(q3 - q0) > thresh) * -1;
flat |= (abs(p4 - p0) > thresh) * -1;
flat |= (abs(q4 - q0) > thresh) * -1;
flat = ~flat;
return flat;
}
static INLINE signed char flatmask5(uint8_t thresh,
uint8_t p4, uint8_t p3, uint8_t p2,
uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1, uint8_t q2,
uint8_t q3, uint8_t q4) {
int8_t flat = 0;
flat |= (abs(p4 - p0) > thresh) * -1;
flat |= (abs(q4 - q0) > thresh) * -1;
flat = ~flat;
return flat & flatmask4(thresh, p3, p2, p1, p0, q0, q1, q2, q3);
}
static __inline void mbfilter(int8_t mask, uint8_t hev, uint8_t flat,
uint8_t *op4, uint8_t *op3, uint8_t *op2,
uint8_t *op1, uint8_t *op0,
uint8_t *oq0, uint8_t *oq1, uint8_t *oq2,
uint8_t *oq3, uint8_t *oq4) {
/* use a 7 tap filter [1, 1, 1, 2, 1, 1, 1] for flat line */
static INLINE void mbfilter(int8_t mask, uint8_t hev, uint8_t flat,
uint8_t *op3, uint8_t *op2,
uint8_t *op1, uint8_t *op0,
uint8_t *oq0, uint8_t *oq1,
uint8_t *oq2, uint8_t *oq3) {
// use a 7 tap filter [1, 1, 1, 2, 1, 1, 1] for flat line
if (flat && mask) {
uint8_t p0, q0;
uint8_t p1, q1;
uint8_t p2, q2;
uint8_t p3, q3;
uint8_t p4, q4;
const uint8_t p3 = *op3;
const uint8_t p2 = *op2;
const uint8_t p1 = *op1;
const uint8_t p0 = *op0;
const uint8_t q0 = *oq0;
const uint8_t q1 = *oq1;
const uint8_t q2 = *oq2;
const uint8_t q3 = *oq3;
p4 = *op4;
p3 = *op3;
p2 = *op2;
p1 = *op1;
p0 = *op0;
q0 = *oq0;
q1 = *oq1;
q2 = *oq2;
q3 = *oq3;
q4 = *oq4;
*op2 = (p4 + p4 + p3 + p2 + p2 + p1 + p0 + q0 + 4) >> 3;
*op1 = (p4 + p3 + p2 + p1 + p1 + p0 + q0 + q1 + 4) >> 3;
*op2 = (p3 + p3 + p3 + p2 + p2 + p1 + p0 + q0 + 4) >> 3;
*op1 = (p3 + p3 + p2 + p1 + p1 + p0 + q0 + q1 + 4) >> 3;
*op0 = (p3 + p2 + p1 + p0 + p0 + q0 + q1 + q2 + 4) >> 3;
*oq0 = (p2 + p1 + p0 + q0 + q0 + q1 + q2 + q3 + 4) >> 3;
*oq1 = (p1 + p0 + q0 + q1 + q1 + q2 + q3 + q4 + 4) >> 3;
*oq2 = (p0 + q0 + q1 + q2 + q2 + q3 + q4 + q4 + 4) >> 3;
*oq1 = (p1 + p0 + q0 + q1 + q1 + q2 + q3 + q3 + 4) >> 3;
*oq2 = (p0 + q0 + q1 + q2 + q2 + q3 + q3 + q3 + 4) >> 3;
} else {
int8_t ps0, qs0;
int8_t ps1, qs1;
int8_t filter, Filter1, Filter2;
int8_t u;
int8_t filter1, filter2;
ps1 = (int8_t) *op1 ^ 0x80;
ps0 = (int8_t) *op0 ^ 0x80;
qs0 = (int8_t) *oq0 ^ 0x80;
qs1 = (int8_t) *oq1 ^ 0x80;
const int8_t ps1 = (int8_t) *op1 ^ 0x80;
const int8_t ps0 = (int8_t) *op0 ^ 0x80;
const int8_t qs0 = (int8_t) *oq0 ^ 0x80;
const int8_t qs1 = (int8_t) *oq1 ^ 0x80;
/* add outer taps if we have high edge variance */
filter = signed_char_clamp(ps1 - qs1);
filter &= hev;
// add outer taps if we have high edge variance
int8_t filter = signed_char_clamp(ps1 - qs1) & hev;
/* inner taps */
filter = signed_char_clamp(filter + 3 * (qs0 - ps0));
filter &= mask;
// inner taps
filter = signed_char_clamp(filter + 3 * (qs0 - ps0)) & mask;
Filter1 = signed_char_clamp(filter + 4);
Filter2 = signed_char_clamp(filter + 3);
Filter1 >>= 3;
Filter2 >>= 3;
filter1 = signed_char_clamp(filter + 4) >> 3;
filter2 = signed_char_clamp(filter + 3) >> 3;
u = signed_char_clamp(qs0 - Filter1);
*oq0 = u ^ 0x80;
u = signed_char_clamp(ps0 + Filter2);
*op0 = u ^ 0x80;
filter = Filter1;
*oq0 = signed_char_clamp(qs0 - filter1) ^ 0x80;
*op0 = signed_char_clamp(ps0 + filter2) ^ 0x80;
filter = filter1;
/* outer tap adjustments */
// outer tap adjustments
filter += 1;
filter >>= 1;
filter &= ~hev;
u = signed_char_clamp(qs1 - filter);
*oq1 = u ^ 0x80;
u = signed_char_clamp(ps1 + filter);
*op1 = u ^ 0x80;
*oq1 = signed_char_clamp(qs1 - filter) ^ 0x80;
*op1 = signed_char_clamp(ps1 + filter) ^ 0x80;
}
}
@ -254,12 +234,11 @@ void vp9_mbloop_filter_horizontal_edge_c(uint8_t *s,
hev = hevmask(thresh[0], s[-2 * p], s[-1 * p], s[0 * p], s[1 * p]);
flat = flatmask(1,
s[-5 * p], s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p],
s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p], s[ 4 * p]);
flat = flatmask4(1, s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p],
s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p]);
mbfilter(mask, hev, flat,
s - 5 * p, s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p,
s, s + 1 * p, s + 2 * p, s + 3 * p, s + 4 * p);
s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p,
s, s + 1 * p, s + 2 * p, s + 3 * p);
++s;
} while (++i < count * 8);
@ -283,53 +262,43 @@ void vp9_mbloop_filter_vertical_edge_c(uint8_t *s,
s[0], s[1], s[2], s[3]);
hev = hevmask(thresh[0], s[-2], s[-1], s[0], s[1]);
flat = flatmask(1,
s[-5], s[-4], s[-3], s[-2], s[-1],
s[ 0], s[ 1], s[ 2], s[ 3], s[ 4]);
flat = flatmask4(1,
s[-4], s[-3], s[-2], s[-1],
s[ 0], s[ 1], s[ 2], s[ 3]);
mbfilter(mask, hev, flat,
s - 5, s - 4, s - 3, s - 2, s - 1,
s, s + 1, s + 2, s + 3, s + 4);
s - 4, s - 3, s - 2, s - 1,
s, s + 1, s + 2, s + 3);
s += p;
} while (++i < count * 8);
}
/* should we apply any filter at all ( 11111111 yes, 00000000 no) */
static __inline int8_t simple_filter_mask(uint8_t blimit,
uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1) {
/* Why does this cause problems for win32?
* error C2143: syntax error : missing ';' before 'type'
* (void) limit;
*/
int8_t mask = (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 <= blimit) * -1;
return mask;
static INLINE int8_t simple_filter_mask(uint8_t blimit,
uint8_t p1, uint8_t p0,
uint8_t q0, uint8_t q1) {
return (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 <= blimit) * -1;
}
static __inline void simple_filter(int8_t mask,
uint8_t *op1, uint8_t *op0,
uint8_t *oq0, uint8_t *oq1) {
int8_t filter, Filter1, Filter2;
int8_t p1 = (int8_t) *op1 ^ 0x80;
int8_t p0 = (int8_t) *op0 ^ 0x80;
int8_t q0 = (int8_t) *oq0 ^ 0x80;
int8_t q1 = (int8_t) *oq1 ^ 0x80;
int8_t u;
static INLINE void simple_filter(int8_t mask,
uint8_t *op1, uint8_t *op0,
uint8_t *oq0, uint8_t *oq1) {
int8_t filter1, filter2;
const int8_t p1 = (int8_t) *op1 ^ 0x80;
const int8_t p0 = (int8_t) *op0 ^ 0x80;
const int8_t q0 = (int8_t) *oq0 ^ 0x80;
const int8_t q1 = (int8_t) *oq1 ^ 0x80;
filter = signed_char_clamp(p1 - q1);
int8_t filter = signed_char_clamp(p1 - q1);
filter = signed_char_clamp(filter + 3 * (q0 - p0));
filter &= mask;
/* save bottom 3 bits so that we round one side +4 and the other +3 */
Filter1 = signed_char_clamp(filter + 4);
Filter1 >>= 3;
u = signed_char_clamp(q0 - Filter1);
*oq0 = u ^ 0x80;
// save bottom 3 bits so that we round one side +4 and the other +3
filter1 = signed_char_clamp(filter + 4) >> 3;
*oq0 = signed_char_clamp(q0 - filter1) ^ 0x80;
Filter2 = signed_char_clamp(filter + 3);
Filter2 >>= 3;
u = signed_char_clamp(p0 + Filter2);
*op0 = u ^ 0x80;
filter2 = signed_char_clamp(filter + 3) >> 3;
*op0 = signed_char_clamp(p0 + filter2) ^ 0x80;
}
void vp9_loop_filter_simple_horizontal_edge_c(uint8_t *s,
@ -481,41 +450,32 @@ void vp9_loop_filter_bvs_c(uint8_t *y_ptr, int y_stride,
vp9_loop_filter_simple_vertical_edge_c(y_ptr + 12, y_stride, blimit);
}
static __inline void wide_mbfilter(int8_t mask, uint8_t hev,
uint8_t flat, uint8_t flat2,
uint8_t *op7, uint8_t *op6, uint8_t *op5,
uint8_t *op4, uint8_t *op3, uint8_t *op2,
uint8_t *op1, uint8_t *op0, uint8_t *oq0,
uint8_t *oq1, uint8_t *oq2, uint8_t *oq3,
uint8_t *oq4, uint8_t *oq5, uint8_t *oq6,
uint8_t *oq7) {
/* use a 15 tap filter [1,1,1,1,1,1,1,2,1,1,1,1,1,1,1] for flat line */
static INLINE void wide_mbfilter(int8_t mask, uint8_t hev,
uint8_t flat, uint8_t flat2,
uint8_t *op7, uint8_t *op6, uint8_t *op5,
uint8_t *op4, uint8_t *op3, uint8_t *op2,
uint8_t *op1, uint8_t *op0, uint8_t *oq0,
uint8_t *oq1, uint8_t *oq2, uint8_t *oq3,
uint8_t *oq4, uint8_t *oq5, uint8_t *oq6,
uint8_t *oq7) {
// use a 15 tap filter [1,1,1,1,1,1,1,2,1,1,1,1,1,1,1] for flat line
if (flat2 && flat && mask) {
uint8_t p0, q0;
uint8_t p1, q1;
uint8_t p2, q2;
uint8_t p3, q3;
uint8_t p4, q4;
uint8_t p5, q5;
uint8_t p6, q6;
uint8_t p7, q7;
p7 = *op7;
p6 = *op6;
p5 = *op5;
p4 = *op4;
p3 = *op3;
p2 = *op2;
p1 = *op1;
p0 = *op0;
q0 = *oq0;
q1 = *oq1;
q2 = *oq2;
q3 = *oq3;
q4 = *oq4;
q5 = *oq5;
q6 = *oq6;
q7 = *oq7;
const uint8_t p7 = *op7;
const uint8_t p6 = *op6;
const uint8_t p5 = *op5;
const uint8_t p4 = *op4;
const uint8_t p3 = *op3;
const uint8_t p2 = *op2;
const uint8_t p1 = *op1;
const uint8_t p0 = *op0;
const uint8_t q0 = *oq0;
const uint8_t q1 = *oq1;
const uint8_t q2 = *oq2;
const uint8_t q3 = *oq3;
const uint8_t q4 = *oq4;
const uint8_t q5 = *oq5;
const uint8_t q6 = *oq6;
const uint8_t q7 = *oq7;
*op6 = (p7 * 7 + p6 * 2 +
p5 + p4 + p3 + p2 + p1 + p0 + q0 + 8) >> 4;
@ -546,68 +506,48 @@ static __inline void wide_mbfilter(int8_t mask, uint8_t hev,
*oq6 = (p0 + q0 + q1 + q2 + q3 + q4 + q5 + q6 * 2 +
q7 * 7 + 8) >> 4;
} else if (flat && mask) {
unsigned char p0, q0;
unsigned char p1, q1;
unsigned char p2, q2;
unsigned char p3, q3;
unsigned char p4, q4;
const uint8_t p3 = *op3;
const uint8_t p2 = *op2;
const uint8_t p1 = *op1;
const uint8_t p0 = *op0;
const uint8_t q0 = *oq0;
const uint8_t q1 = *oq1;
const uint8_t q2 = *oq2;
const uint8_t q3 = *oq3;
p4 = *op4;
p3 = *op3;
p2 = *op2;
p1 = *op1;
p0 = *op0;
q0 = *oq0;
q1 = *oq1;
q2 = *oq2;
q3 = *oq3;
q4 = *oq4;
*op2 = (p4 + p4 + p3 + p2 + p2 + p1 + p0 + q0 + 4) >> 3;
*op1 = (p4 + p3 + p2 + p1 + p1 + p0 + q0 + q1 + 4) >> 3;
*op2 = (p3 + p3 + p3 + p2 + p2 + p1 + p0 + q0 + 4) >> 3;
*op1 = (p3 + p3 + p2 + p1 + p1 + p0 + q0 + q1 + 4) >> 3;
*op0 = (p3 + p2 + p1 + p0 + p0 + q0 + q1 + q2 + 4) >> 3;
*oq0 = (p2 + p1 + p0 + q0 + q0 + q1 + q2 + q3 + 4) >> 3;
*oq1 = (p1 + p0 + q0 + q1 + q1 + q2 + q3 + q4 + 4) >> 3;
*oq2 = (p0 + q0 + q1 + q2 + q2 + q3 + q4 + q4 + 4) >> 3;
*oq1 = (p1 + p0 + q0 + q1 + q1 + q2 + q3 + q3 + 4) >> 3;
*oq2 = (p0 + q0 + q1 + q2 + q2 + q3 + q3 + q3 + 4) >> 3;
} else {
signed char ps0, qs0;
signed char ps1, qs1;
signed char filter, Filter1, Filter2;
signed char u;
int8_t filter1, filter2;
ps1 = (signed char) * op1 ^ 0x80;
ps0 = (signed char) * op0 ^ 0x80;
qs0 = (signed char) * oq0 ^ 0x80;
qs1 = (signed char) * oq1 ^ 0x80;
const int8_t ps1 = (int8_t) * op1 ^ 0x80;
const int8_t ps0 = (int8_t) * op0 ^ 0x80;
const int8_t qs0 = (int8_t) * oq0 ^ 0x80;
const int8_t qs1 = (int8_t) * oq1 ^ 0x80;
/* add outer taps if we have high edge variance */
filter = signed_char_clamp(ps1 - qs1);
filter &= hev;
// add outer taps if we have high edge variance
int8_t filter = signed_char_clamp(ps1 - qs1) & hev;
/* inner taps */
filter = signed_char_clamp(filter + 3 * (qs0 - ps0));
filter &= mask;
// inner taps
filter = signed_char_clamp(filter + 3 * (qs0 - ps0)) & mask;
filter1 = signed_char_clamp(filter + 4) >> 3;
filter2 = signed_char_clamp(filter + 3) >> 3;
Filter1 = signed_char_clamp(filter + 4);
Filter2 = signed_char_clamp(filter + 3);
Filter1 >>= 3;
Filter2 >>= 3;
*oq0 = signed_char_clamp(qs0 - filter1) ^ 0x80;
*op0 = signed_char_clamp(ps0 + filter2) ^ 0x80;
filter = filter1;
u = signed_char_clamp(qs0 - Filter1);
*oq0 = u ^ 0x80;
u = signed_char_clamp(ps0 + Filter2);
*op0 = u ^ 0x80;
filter = Filter1;
/* outer tap adjustments */
// outer tap adjustments
filter += 1;
filter >>= 1;
filter &= ~hev;
u = signed_char_clamp(qs1 - filter);
*oq1 = u ^ 0x80;
u = signed_char_clamp(ps1 + filter);
*op1 = u ^ 0x80;
*oq1 = signed_char_clamp(qs1 - filter) ^ 0x80;
*op1 = signed_char_clamp(ps1 + filter) ^ 0x80;
}
}
@ -636,19 +576,19 @@ void vp9_mb_lpf_horizontal_edge_w
hev = hevmask(thresh[0], s[-2 * p], s[-1 * p], s[0 * p], s[1 * p]);
flat = flatmask(1,
s[-5 * p], s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p],
s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p], s[ 4 * p]);
flat = flatmask4(1,
s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p],
s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p]);
flat2 = flatmask(1,
s[-8 * p], s[-7 * p], s[-6 * p], s[-5 * p], s[-1 * p],
s[ 0 * p], s[ 4 * p], s[ 5 * p], s[ 6 * p], s[ 7 * p]);
flat2 = flatmask5(1,
s[-8 * p], s[-7 * p], s[-6 * p], s[-5 * p], s[-1 * p],
s[ 0 * p], s[ 4 * p], s[ 5 * p], s[ 6 * p], s[ 7 * p]);
wide_mbfilter(mask, hev, flat, flat2,
s - 8 * p, s - 7 * p, s - 6 * p, s - 5 * p,
s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p,
s, s + 1 * p, s + 2 * p, s + 3 * p,
s + 4 * p, s + 5 * p, s + 6 * p, s + 7 * p);
s - 8 * p, s - 7 * p, s - 6 * p, s - 5 * p,
s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p,
s, s + 1 * p, s + 2 * p, s + 3 * p,
s + 4 * p, s + 5 * p, s + 6 * p, s + 7 * p);
++s;
} while (++i < count * 8);
@ -674,18 +614,18 @@ void vp9_mb_lpf_vertical_edge_w
s[0], s[1], s[2], s[3]);
hev = hevmask(thresh[0], s[-2], s[-1], s[0], s[1]);
flat = flatmask(1,
s[-5], s[-4], s[-3], s[-2], s[-1],
s[ 0], s[ 1], s[ 2], s[ 3], s[ 4]);
flat2 = flatmask(1,
s[-8], s[-7], s[-6], s[-5], s[-1],
s[ 0], s[ 4], s[ 5], s[ 6], s[ 7]);
flat = flatmask4(1,
s[-4], s[-3], s[-2], s[-1],
s[ 0], s[ 1], s[ 2], s[ 3]);
flat2 = flatmask5(1,
s[-8], s[-7], s[-6], s[-5], s[-1],
s[ 0], s[ 4], s[ 5], s[ 6], s[ 7]);
wide_mbfilter(mask, hev, flat, flat2,
s - 8, s - 7, s - 6, s - 5,
s - 4, s - 3, s - 2, s - 1,
s, s + 1, s + 2, s + 3,
s + 4, s + 5, s + 6, s + 7);
s - 8, s - 7, s - 6, s - 5,
s - 4, s - 3, s - 2, s - 1,
s, s + 1, s + 2, s + 3,
s + 4, s + 5, s + 6, s + 7);
s += p;
} while (++i < count * 8);
}

19
vp9/common/vp9_maskingmv.c
View File

@ -11,25 +11,19 @@
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
extern unsigned int vp9_sad16x16_sse3(
unsigned int vp9_sad16x16_sse3(
unsigned char *src_ptr,
int src_stride,
unsigned char *ref_ptr,
int ref_stride,
int max_err);
extern void vp9_sad16x16x3_sse3(
unsigned char *src_ptr,
int src_stride,
unsigned char *ref_ptr,
int ref_stride,
int *results);
extern int vp8_growmaskmb_sse3(
int vp8_growmaskmb_sse3(
unsigned char *om,
unsigned char *nm);
extern void vp8_makemask_sse3(
void vp8_makemask_sse3(
unsigned char *y,
unsigned char *u,
unsigned char *v,
@ -238,6 +232,7 @@ void grow_ymask(unsigned char *ym) {
for (i = 0; i < 256; i++)
ym[i] = nym[i];
}
void make_mb_mask(unsigned char *y, unsigned char *u, unsigned char *v,
unsigned char *ym, unsigned char *uvm,
int yp, int uvp,
@ -283,6 +278,7 @@ int compare_masks(unsigned char *sym, unsigned char *ym) {
return sad;
}
int unmasked_sad(unsigned char *src, int p, unsigned char *dst, int dp,
unsigned char *ym) {
int i, j;
@ -294,6 +290,7 @@ int unmasked_sad(unsigned char *src, int p, unsigned char *dst, int dp,
return sad;
}
int masked_motion_search(unsigned char *y, unsigned char *u, unsigned char *v,
int yp, int uvp,
unsigned char *dy, unsigned char *du, unsigned char *dv,
@ -802,5 +799,5 @@ int mainz(int argc, char *argv[]) {
}
fclose(f);
fclose(g);
return;
return 0;
}

10
vp9/common/vp9_mbpitch.c
View File

@ -20,15 +20,15 @@ static void setup_block(BLOCKD *b,
int mv_stride,
uint8_t **base,
uint8_t **base2,
int Stride,
int stride,
int offset,
BLOCKSET bs) {
if (bs == DEST) {
b->dst_stride = Stride;
b->dst_stride = stride;
b->dst = offset;
b->base_dst = base;
} else {
b->pre_stride = Stride;
b->pre_stride = stride;
b->pre = offset;
b->base_pre = base;
b->base_second_pre = base2;
@ -102,9 +102,7 @@ void vp9_setup_block_dptrs(MACROBLOCKD *xd) {
}
}
blockd[24].diff = &xd->diff[384];
for (r = 0; r < 25; r++) {
for (r = 0; r < 24; r++) {
blockd[r].qcoeff = xd->qcoeff + r * 16;
blockd[r].dqcoeff = xd->dqcoeff + r * 16;
}

2
vp9/common/vp9_modecont.c
View File

@ -12,7 +12,7 @@
#include "vp9/common/vp9_entropy.h"
const int vp9_default_mode_contexts[INTER_MODE_CONTEXTS][4] = {
{223, 1, 1, 237}, // 0,0 best: Only candidate
{1, 223, 1, 237}, // 0,0 best: Only candidate
{87, 166, 26, 219}, // 0,0 best: non zero candidates
{89, 67, 18, 125}, // 0,0 best: non zero candidates, split
{16, 141, 69, 226}, // strong nz candidate(s), no split

10
vp9/common/vp9_mv.h
View File

@ -23,4 +23,14 @@ typedef union int_mv {
MV as_mv;
} int_mv; /* facilitates faster equality tests and copies */
struct mv32 {
int32_t row;
int32_t col;
};
typedef union int_mv32 {
uint64_t as_int;
struct mv32 as_mv;
} int_mv32; /* facilitates faster equality tests and copies */
#endif // VP9_COMMON_VP9_MV_H_

313
vp9/common/vp9_mvref_common.c
View File

@ -11,64 +11,69 @@
#include "vp9/common/vp9_mvref_common.h"
#define MVREF_NEIGHBOURS 8
static int mb_mv_ref_search[MVREF_NEIGHBOURS][2] = {
{0, -1}, {-1, 0}, {-1, -1}, {0, -2},
{-2, 0}, {-1, -2}, {-2, -1}, {-2, -2}
};
static int mb_ref_distance_weight[MVREF_NEIGHBOURS] =
{ 3, 3, 2, 1, 1, 1, 1, 1 };
static int sb_mv_ref_search[MVREF_NEIGHBOURS][2] = {
{0, -1}, {-1, 0}, {1, -1}, {-1, 1},
{-1, -1}, {0, -2}, {-2, 0}, {-1, -2}
};
static int sb_ref_distance_weight[MVREF_NEIGHBOURS] =
{ 3, 3, 2, 2, 2, 1, 1, 1 };
// clamp_mv
static int sb64_mv_ref_search[MVREF_NEIGHBOURS][2] = {
{0, -1}, {-1, 0}, {1, -1}, {-1, 1},
{2, -1}, {-1, 2}, {3, -1}, {-1,-1}
};
static int sb64_ref_distance_weight[MVREF_NEIGHBOURS] =
{ 1, 1, 1, 1, 1, 1, 1, 1 };
// clamp_mv_ref
#define MV_BORDER (16 << 3) // Allow 16 pels in 1/8th pel units
static void clamp_mv(const MACROBLOCKD *xd, int_mv *mv) {
if (mv->as_mv.col < (xd->mb_to_left_edge - MV_BORDER))
mv->as_mv.col = xd->mb_to_left_edge - MV_BORDER;
else if (mv->as_mv.col > xd->mb_to_right_edge + MV_BORDER)
mv->as_mv.col = xd->mb_to_right_edge + MV_BORDER;
if (mv->as_mv.row < (xd->mb_to_top_edge - MV_BORDER))
mv->as_mv.row = xd->mb_to_top_edge - MV_BORDER;
else if (mv->as_mv.row > xd->mb_to_bottom_edge + MV_BORDER)
mv->as_mv.row = xd->mb_to_bottom_edge + MV_BORDER;
static void clamp_mv_ref(const MACROBLOCKD *xd, int_mv *mv) {
mv->as_mv.col = clamp(mv->as_mv.col, xd->mb_to_left_edge - MV_BORDER,
xd->mb_to_right_edge + MV_BORDER);
mv->as_mv.row = clamp(mv->as_mv.row, xd->mb_to_top_edge - MV_BORDER,
xd->mb_to_bottom_edge + MV_BORDER);
}
// Gets a candidate refenence motion vector from the given mode info
// structure if one exists that matches the given reference frame.
static int get_matching_candidate(
const MODE_INFO *candidate_mi,
MV_REFERENCE_FRAME ref_frame,
int_mv *c_mv
) {
int ret_val = TRUE;
static int get_matching_candidate(const MODE_INFO *candidate_mi,
MV_REFERENCE_FRAME ref_frame,
int_mv *c_mv) {
if (ref_frame == candidate_mi->mbmi.ref_frame) {
c_mv->as_int = candidate_mi->mbmi.mv[0].as_int;
} else if (ref_frame == candidate_mi->mbmi.second_ref_frame) {
c_mv->as_int = candidate_mi->mbmi.mv[1].as_int;
} else {
ret_val = FALSE;
return 0;
}
return ret_val;
return 1;
}
// Gets candidate refenence motion vector(s) from the given mode info
// structure if they exists and do NOT match the given reference frame.
static void get_non_matching_candidates(
const MODE_INFO *candidate_mi,
MV_REFERENCE_FRAME ref_frame,
MV_REFERENCE_FRAME *c_ref_frame,
int_mv *c_mv,
MV_REFERENCE_FRAME *c2_ref_frame,
int_mv *c2_mv
) {
static void get_non_matching_candidates(const MODE_INFO *candidate_mi,
MV_REFERENCE_FRAME ref_frame,
MV_REFERENCE_FRAME *c_ref_frame,
int_mv *c_mv,
MV_REFERENCE_FRAME *c2_ref_frame,
int_mv *c2_mv) {
c_mv->as_int = 0;
c2_mv->as_int = 0;
@ -85,73 +90,68 @@ static void get_non_matching_candidates(
// Second candidate
if ((candidate_mi->mbmi.second_ref_frame > INTRA_FRAME) &&
(candidate_mi->mbmi.second_ref_frame != ref_frame)) { // &&
// (candidate_mi->mbmi.mv[1].as_int != 0) &&
// (candidate_mi->mbmi.mv[1].as_int !=
// candidate_mi->mbmi.mv[0].as_int)) {
(candidate_mi->mbmi.second_ref_frame != ref_frame) &&
(candidate_mi->mbmi.mv[1].as_int != candidate_mi->mbmi.mv[0].as_int)) {
*c2_ref_frame = candidate_mi->mbmi.second_ref_frame;
c2_mv->as_int = candidate_mi->mbmi.mv[1].as_int;
}
}
}
// Performs mv adjustment based on reference frame and clamps the MV
// if it goes off the edge of the buffer.
static void scale_mv(
MACROBLOCKD *xd,
MV_REFERENCE_FRAME this_ref_frame,
MV_REFERENCE_FRAME candidate_ref_frame,
int_mv *candidate_mv,
int *ref_sign_bias
) {
if (candidate_ref_frame != this_ref_frame) {
// Performs mv sign inversion if indicated by the reference frame combination.
static void scale_mv(MACROBLOCKD *xd, MV_REFERENCE_FRAME this_ref_frame,
MV_REFERENCE_FRAME candidate_ref_frame,
int_mv *candidate_mv, int *ref_sign_bias) {
// int frame_distances[MAX_REF_FRAMES];
// int last_distance = 1;
// int gf_distance = xd->frames_since_golden;
// int arf_distance = xd->frames_till_alt_ref_frame;
//int frame_distances[MAX_REF_FRAMES];
//int last_distance = 1;
//int gf_distance = xd->frames_since_golden;
//int arf_distance = xd->frames_till_alt_ref_frame;
// Sign inversion where appropriate.
if (ref_sign_bias[candidate_ref_frame] != ref_sign_bias[this_ref_frame]) {
candidate_mv->as_mv.row = -candidate_mv->as_mv.row;
candidate_mv->as_mv.col = -candidate_mv->as_mv.col;
}
// Scale based on frame distance if the reference frames not the same.
/*frame_distances[INTRA_FRAME] = 1; // should never be used
frame_distances[LAST_FRAME] = 1;
frame_distances[GOLDEN_FRAME] =
(xd->frames_since_golden) ? xd->frames_since_golden : 1;
frame_distances[ALTREF_FRAME] =
(xd->frames_till_alt_ref_frame) ? xd->frames_till_alt_ref_frame : 1;
if (frame_distances[this_ref_frame] &&
frame_distances[candidate_ref_frame]) {
candidate_mv->as_mv.row =
(short)(((int)(candidate_mv->as_mv.row) *
frame_distances[this_ref_frame]) /
frame_distances[candidate_ref_frame]);
candidate_mv->as_mv.col =
(short)(((int)(candidate_mv->as_mv.col) *
frame_distances[this_ref_frame]) /
frame_distances[candidate_ref_frame]);
}
*/
// Sign inversion where appropriate.
if (ref_sign_bias[candidate_ref_frame] != ref_sign_bias[this_ref_frame]) {
candidate_mv->as_mv.row = -candidate_mv->as_mv.row;
candidate_mv->as_mv.col = -candidate_mv->as_mv.col;
}
// Clamp the MV so it does not point out of the frame buffer
clamp_mv(xd, candidate_mv);
/*
// Scale based on frame distance if the reference frames not the same.
frame_distances[INTRA_FRAME] = 1; // should never be used
frame_distances[LAST_FRAME] = 1;
frame_distances[GOLDEN_FRAME] =
(xd->frames_since_golden) ? xd->frames_si nce_golden : 1;
frame_distances[ALTREF_FRAME] =
(xd->frames_till_alt_ref_frame) ? xd->frames_till_alt_ref_frame : 1;
if (frame_distances[this_ref_frame] &&
frame_distances[candidate_ref_frame]) {
candidate_mv->as_mv.row =
(short)(((int)(candidate_mv->as_mv.row) *
frame_distances[this_ref_frame]) /
frame_distances[candidate_ref_frame]);
candidate_mv->as_mv.col =
(short)(((int)(candidate_mv->as_mv.col) *
frame_distances[this_ref_frame]) /
frame_distances[candidate_ref_frame]);
}
*/
}
// Adds a new candidate reference vector to the list if indeed it is new.
// If it is not new then the score of the existing candidate that it matches
// is increased and the list is resorted.
/*
// Adds a new candidate reference vector to the sorted list.
// If it is a repeat the weight of the existing entry is increased
// and the order of the list is resorted.
// This method of add plus sort has been deprecated for now as there is a
// further sort of the best candidates in vp9_find_best_ref_mvs() and the
// incremental benefit of both is small. If the decision is made to remove
// the sort in vp9_find_best_ref_mvs() for performance reasons then it may be
// worth re-instating some sort of list reordering by weight here.
//
static void addmv_and_shuffle(
int_mv *mv_list,
int *mv_scores,
int *index,
int *refmv_count,
int_mv candidate_mv,
int weight
) {
@ -162,11 +162,11 @@ static void addmv_and_shuffle(
// Check for duplicates. If there is one increase its score.
// We only compare vs the current top candidates.
insert_point = (*index < (MAX_MV_REF_CANDIDATES - 1))
? *index : (MAX_MV_REF_CANDIDATES - 1);
insert_point = (*refmv_count < (MAX_MV_REF_CANDIDATES - 1))
? *refmv_count : (MAX_MV_REF_CANDIDATES - 1);
i = insert_point;
if (*index > i)
if (*refmv_count > i)
i++;
while (i > 0) {
i--;
@ -184,7 +184,7 @@ static void addmv_and_shuffle(
mv_scores[insert_point] = weight;
i = insert_point;
}
(*index)++;
(*refmv_count)++;
}
// Reshuffle the list so that highest scoring mvs at the top.
@ -202,19 +202,42 @@ static void addmv_and_shuffle(
break;
}
}
*/
// Adds a new candidate reference vector to the list.
// The mv is thrown out if it is already in the list.
// Unlike the addmv_and_shuffle() this does not reorder the list
// but assumes that candidates are added in the order most likely to
// match distance and reference frame bias.
static void add_candidate_mv(int_mv *mv_list, int *mv_scores,
int *candidate_count, int_mv candidate_mv,
int weight) {
int i;
// Make sure we dont insert off the end of the list
const int insert_point = MIN(*candidate_count, MAX_MV_REF_CANDIDATES - 1);
// Look for duplicates
for (i = 0; i <= insert_point; ++i) {
if (candidate_mv.as_int == mv_list[i].as_int)
break;
}
// Add the candidate. If the list is already full it is only desirable that
// it should overwrite if it has a higher weight than the last entry.
if (i >= insert_point && weight > mv_scores[insert_point]) {
mv_list[insert_point].as_int = candidate_mv.as_int;
mv_scores[insert_point] = weight;
*candidate_count += (*candidate_count < MAX_MV_REF_CANDIDATES);
}
}
// This function searches the neighbourhood of a given MB/SB and populates a
// list of candidate reference vectors.
//
void vp9_find_mv_refs(
MACROBLOCKD *xd,
MODE_INFO *here,
MODE_INFO *lf_here,
MV_REFERENCE_FRAME ref_frame,
int_mv *mv_ref_list,
int *ref_sign_bias
) {
void vp9_find_mv_refs(VP9_COMMON *cm, MACROBLOCKD *xd, MODE_INFO *here,
MODE_INFO *lf_here, MV_REFERENCE_FRAME ref_frame,
int_mv *mv_ref_list, int *ref_sign_bias) {
int i;
MODE_INFO *candidate_mi;
MB_MODE_INFO * mbmi = &xd->mode_info_context->mbmi;
@ -224,17 +247,22 @@ void vp9_find_mv_refs(
MV_REFERENCE_FRAME c_ref_frame;
MV_REFERENCE_FRAME c2_ref_frame;
int candidate_scores[MAX_MV_REF_CANDIDATES];
int index = 0;
int refmv_count = 0;
int split_count = 0;
int (*mv_ref_search)[2];
int *ref_distance_weight;
int zero_seen = FALSE;
const int mb_col = (-xd->mb_to_left_edge) >> 7;
// Blank the reference vector lists and other local structures.
vpx_memset(mv_ref_list, 0, sizeof(int_mv) * MAX_MV_REF_CANDIDATES);
vpx_memset(candidate_mvs, 0, sizeof(int_mv) * MAX_MV_REF_CANDIDATES);
vpx_memset(candidate_scores, 0, sizeof(candidate_scores));
if (mbmi->sb_type) {
if (mbmi->sb_type == BLOCK_SIZE_SB64X64) {
mv_ref_search = sb64_mv_ref_search;
ref_distance_weight = sb64_ref_distance_weight;
} else if (mbmi->sb_type == BLOCK_SIZE_SB32X32) {
mv_ref_search = sb_mv_ref_search;
ref_distance_weight = sb_ref_distance_weight;
} else {
@ -245,39 +273,44 @@ void vp9_find_mv_refs(
// We first scan for candidate vectors that match the current reference frame
// Look at nearest neigbours
for (i = 0; i < 2; ++i) {
if (((mv_ref_search[i][0] << 7) >= xd->mb_to_left_edge) &&
const int mb_search_col = mb_col + mv_ref_search[i][0];
if ((mb_search_col >= cm->cur_tile_mb_col_start) &&
(mb_search_col < cm->cur_tile_mb_col_end) &&
((mv_ref_search[i][1] << 7) >= xd->mb_to_top_edge)) {
candidate_mi = here + mv_ref_search[i][0] +
(mv_ref_search[i][1] * xd->mode_info_stride);
if (get_matching_candidate(candidate_mi, ref_frame, &c_refmv)) {
clamp_mv(xd, &c_refmv);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c_refmv, ref_distance_weight[i] + 16);
add_candidate_mv(candidate_mvs, candidate_scores,
&refmv_count, c_refmv, ref_distance_weight[i] + 16);
}
split_count += (candidate_mi->mbmi.mode == SPLITMV);
}
}
// Look in the last frame
candidate_mi = lf_here;
if (get_matching_candidate(candidate_mi, ref_frame, &c_refmv)) {
clamp_mv(xd, &c_refmv);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c_refmv, 18);
// Look in the last frame if it exists
if (lf_here) {
candidate_mi = lf_here;
if (get_matching_candidate(candidate_mi, ref_frame, &c_refmv)) {
add_candidate_mv(candidate_mvs, candidate_scores,
&refmv_count, c_refmv, 18);
}
}
// More distant neigbours
for (i = 2; (i < MVREF_NEIGHBOURS) &&
(index < (MAX_MV_REF_CANDIDATES - 1)); ++i) {
if (((mv_ref_search[i][0] << 7) >= xd->mb_to_left_edge) &&
(refmv_count < (MAX_MV_REF_CANDIDATES - 1)); ++i) {
const int mb_search_col = mb_col + mv_ref_search[i][0];
if ((mb_search_col >= cm->cur_tile_mb_col_start) &&
(mb_search_col < cm->cur_tile_mb_col_end) &&
((mv_ref_search[i][1] << 7) >= xd->mb_to_top_edge)) {
candidate_mi = here + mv_ref_search[i][0] +
(mv_ref_search[i][1] * xd->mode_info_stride);
if (get_matching_candidate(candidate_mi, ref_frame, &c_refmv)) {
clamp_mv(xd, &c_refmv);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c_refmv, ref_distance_weight[i] + 16);
add_candidate_mv(candidate_mvs, candidate_scores,
&refmv_count, c_refmv, ref_distance_weight[i] + 16);
}
}
}
@ -286,9 +319,12 @@ void vp9_find_mv_refs(
// reference frame does not match. Break out when we have
// MAX_MV_REF_CANDIDATES candidates.
// Look first at spatial neighbours
if (index < (MAX_MV_REF_CANDIDATES - 1)) {
if (refmv_count < (MAX_MV_REF_CANDIDATES - 1)) {
for (i = 0; i < MVREF_NEIGHBOURS; ++i) {
if (((mv_ref_search[i][0] << 7) >= xd->mb_to_left_edge) &&
const int mb_search_col = mb_col + mv_ref_search[i][0];
if ((mb_search_col >= cm->cur_tile_mb_col_start) &&
(mb_search_col < cm->cur_tile_mb_col_end) &&
((mv_ref_search[i][1] << 7) >= xd->mb_to_top_edge)) {
candidate_mi = here + mv_ref_search[i][0] +
@ -300,24 +336,24 @@ void vp9_find_mv_refs(
if (c_ref_frame != INTRA_FRAME) {
scale_mv(xd, ref_frame, c_ref_frame, &c_refmv, ref_sign_bias);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c_refmv, ref_distance_weight[i]);
add_candidate_mv(candidate_mvs, candidate_scores,
&refmv_count, c_refmv, ref_distance_weight[i]);
}
if (c2_ref_frame != INTRA_FRAME) {
scale_mv(xd, ref_frame, c2_ref_frame, &c2_refmv, ref_sign_bias);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c2_refmv, ref_distance_weight[i]);
add_candidate_mv(candidate_mvs, candidate_scores,
&refmv_count, c2_refmv, ref_distance_weight[i]);
}
}
if (index >= (MAX_MV_REF_CANDIDATES - 1)) {
if (refmv_count >= (MAX_MV_REF_CANDIDATES - 1)) {
break;
}
}
}
// Look at the last frame
if (index < (MAX_MV_REF_CANDIDATES - 1)) {
// Look at the last frame if it exists
if (refmv_count < (MAX_MV_REF_CANDIDATES - 1) && lf_here) {
candidate_mi = lf_here;
get_non_matching_candidates(candidate_mi, ref_frame,
&c_ref_frame, &c_refmv,
@ -325,14 +361,14 @@ void vp9_find_mv_refs(
if (c_ref_frame != INTRA_FRAME) {
scale_mv(xd, ref_frame, c_ref_frame, &c_refmv, ref_sign_bias);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c_refmv, 2);
add_candidate_mv(candidate_mvs, candidate_scores,
&refmv_count, c_refmv, 2);
}
if (c2_ref_frame != INTRA_FRAME) {
scale_mv(xd, ref_frame, c2_ref_frame, &c2_refmv, ref_sign_bias);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c2_refmv, 2);
add_candidate_mv(candidate_mvs, candidate_scores,
&refmv_count, c2_refmv, 2);
}
}
@ -340,7 +376,7 @@ void vp9_find_mv_refs(
// 0,0 was best
if (candidate_mvs[0].as_int == 0) {
// 0,0 is only candidate
if (index <= 1) {
if (refmv_count <= 1) {
mbmi->mb_mode_context[ref_frame] = 0;
// non zero candidates candidates available
} else if (split_count == 0) {
@ -348,30 +384,25 @@ void vp9_find_mv_refs(
} else {
mbmi->mb_mode_context[ref_frame] = 2;
}
// Non zero best, No Split MV cases
} else if (split_count == 0) {
if (candidate_scores[0] >= 32) {
mbmi->mb_mode_context[ref_frame] = 3;
} else {
mbmi->mb_mode_context[ref_frame] = 4;
}
// Non zero best, some split mv
// Non zero best, No Split MV cases
mbmi->mb_mode_context[ref_frame] = candidate_scores[0] >= 16 ? 3 : 4;
} else {
if (candidate_scores[0] >= 32) {
mbmi->mb_mode_context[ref_frame] = 5;
} else {
mbmi->mb_mode_context[ref_frame] = 6;
}
// Non zero best, some split mv
mbmi->mb_mode_context[ref_frame] = candidate_scores[0] >= 16 ? 5 : 6;
}
// 0,0 is always a valid reference.
// Scan for 0,0 case and clamp non zero choices
for (i = 0; i < MAX_MV_REF_CANDIDATES; ++i) {
if (candidate_mvs[i].as_int == 0)
break;
if (candidate_mvs[i].as_int == 0) {
zero_seen = TRUE;
} else {
clamp_mv_ref(xd, &candidate_mvs[i]);
}
}
if (i == MAX_MV_REF_CANDIDATES) {
// 0,0 is always a valid reference. Add it if not already seen.
if (!zero_seen)
candidate_mvs[MAX_MV_REF_CANDIDATES-1].as_int = 0;
}
// Copy over the candidate list.
vpx_memcpy(mv_ref_list, candidate_mvs, sizeof(candidate_mvs));

3
vp9/common/vp9_mvref_common.h
View File

@ -14,7 +14,8 @@
#ifndef VP9_COMMON_VP9_MVREF_COMMON_H_
#define VP9_COMMON_VP9_MVREF_COMMON_H_
void vp9_find_mv_refs(MACROBLOCKD *xd,
void vp9_find_mv_refs(VP9_COMMON *cm,
MACROBLOCKD *xd,
MODE_INFO *here,
MODE_INFO *lf_here,
MV_REFERENCE_FRAME ref_frame,

40
vp9/common/vp9_onyx.h
View File

@ -16,6 +16,7 @@ extern "C"
{
#endif
#include "./vpx_config.h"
#include "vpx/internal/vpx_codec_internal.h"
#include "vpx/vp8cx.h"
#include "vpx_scale/yv12config.h"
@ -62,7 +63,7 @@ extern "C"
#include <assert.h>
static __inline void Scale2Ratio(int mode, int *hr, int *hs) {
static INLINE void Scale2Ratio(int mode, int *hr, int *hs) {
switch (mode) {
case NORMAL:
*hr = 1;
@ -89,11 +90,13 @@ extern "C"
}
typedef struct {
int Version; // 4 versions of bitstream defined 0 best quality/slowest decode, 3 lowest quality/fastest decode
int Width; // width of data passed to the compressor
int Height; // height of data passed to the compressor
int version; // 4 versions of bitstream defined:
// 0 - best quality/slowest decode,
// 3 - lowest quality/fastest decode
int width; // width of data passed to the compressor
int height; // height of data passed to the compressor
double frame_rate; // set to passed in framerate
int target_bandwidth; // bandwidth to be used in kilobits per second
int64_t target_bandwidth; // bandwidth to be used in kilobits per second
int noise_sensitivity; // parameter used for applying pre processing blur: recommendation 0
int Sharpness; // parameter used for sharpening output: recommendation 0:
@ -134,9 +137,9 @@ extern "C"
int over_shoot_pct;
// buffering parameters
int starting_buffer_level; // in seconds
int optimal_buffer_level;
int maximum_buffer_size;
int64_t starting_buffer_level; // in seconds
int64_t optimal_buffer_level;
int64_t maximum_buffer_size;
// controlling quality
int fixed_q;
@ -159,10 +162,25 @@ extern "C"
int encode_breakout; // early breakout encode threshold : for video conf recommend 800
/* Bitfield defining the error resiliency features to enable.
* Can provide decodable frames after losses in previous
* frames and decodable partitions after losses in the same frame.
*/
unsigned int error_resilient_mode;
/* Bitfield defining the parallel decoding mode where the
* decoding in successive frames may be conducted in parallel
* just by decoding the frame headers.
*/
unsigned int frame_parallel_decoding_mode;
int arnr_max_frames;
int arnr_strength;
int arnr_type;
int tile_columns;
int tile_rows;
struct vpx_fixed_buf two_pass_stats_in;
struct vpx_codec_pkt_list *output_pkt_list;
@ -195,8 +213,10 @@ extern "C"
int vp9_update_reference(VP9_PTR comp, int ref_frame_flags);
int vp9_get_reference_enc(VP9_PTR comp, VP9_REFFRAME ref_frame_flag,
YV12_BUFFER_CONFIG *sd);
int vp9_copy_reference_enc(VP9_PTR comp, VP9_REFFRAME ref_frame_flag,
YV12_BUFFER_CONFIG *sd);
int vp9_get_reference_enc(VP9_PTR ptr, int index, YV12_BUFFER_CONFIG **fb);
int vp9_set_reference_enc(VP9_PTR comp, VP9_REFFRAME ref_frame_flag,
YV12_BUFFER_CONFIG *sd);

176
vp9/common/vp9_onyxc_int.h
View File

@ -37,7 +37,16 @@ void vp9_initialize_common(void);
#define QINDEX_RANGE (MAXQ + 1)
#define NUM_YV12_BUFFERS 4
#define NUM_REF_FRAMES 3
#define NUM_REF_FRAMES_LG2 2
// 1 scratch frame for the new frame, 3 for scaled references on the encoder
// TODO(jkoleszar): These 3 extra references could probably come from the
// normal reference pool.
#define NUM_YV12_BUFFERS (NUM_REF_FRAMES + 4)
#define NUM_FRAME_CONTEXTS_LG2 2
#define NUM_FRAME_CONTEXTS (1 << NUM_FRAME_CONTEXTS_LG2)
#define COMP_PRED_CONTEXTS 2
@ -49,13 +58,23 @@ typedef struct frame_contexts {
vp9_prob i8x8_mode_prob[VP9_I8X8_MODES - 1];
vp9_prob sub_mv_ref_prob[SUBMVREF_COUNT][VP9_SUBMVREFS - 1];
vp9_prob mbsplit_prob[VP9_NUMMBSPLITS - 1];
vp9_coeff_probs coef_probs_4x4[BLOCK_TYPES_4X4];
vp9_coeff_probs hybrid_coef_probs_4x4[BLOCK_TYPES_4X4];
vp9_coeff_probs coef_probs_8x8[BLOCK_TYPES_8X8];
vp9_coeff_probs hybrid_coef_probs_8x8[BLOCK_TYPES_8X8];
vp9_coeff_probs coef_probs_16x16[BLOCK_TYPES_16X16];
vp9_coeff_probs hybrid_coef_probs_16x16[BLOCK_TYPES_16X16];
vp9_coeff_probs coef_probs_32x32[BLOCK_TYPES_32X32];
vp9_coeff_probs coef_probs_4x4[BLOCK_TYPES];
vp9_coeff_probs coef_probs_8x8[BLOCK_TYPES];
vp9_coeff_probs coef_probs_16x16[BLOCK_TYPES];
vp9_coeff_probs coef_probs_32x32[BLOCK_TYPES];
#if CONFIG_CODE_NONZEROCOUNT
vp9_prob nzc_probs_4x4[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC4X4_NODES];
vp9_prob nzc_probs_8x8[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC8X8_NODES];
vp9_prob nzc_probs_16x16[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC16X16_NODES];
vp9_prob nzc_probs_32x32[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC32X32_NODES];
vp9_prob nzc_pcat_probs[MAX_NZC_CONTEXTS]
[NZC_TOKENS_EXTRA][NZC_BITS_EXTRA];
#endif
nmv_context nmvc;
nmv_context pre_nmvc;
@ -74,21 +93,42 @@ typedef struct frame_contexts {
unsigned int sub_mv_ref_counts[SUBMVREF_COUNT][VP9_SUBMVREFS];
unsigned int mbsplit_counts[VP9_NUMMBSPLITS];
vp9_coeff_probs pre_coef_probs_4x4[BLOCK_TYPES_4X4];
vp9_coeff_probs pre_hybrid_coef_probs_4x4[BLOCK_TYPES_4X4];
vp9_coeff_probs pre_coef_probs_8x8[BLOCK_TYPES_8X8];
vp9_coeff_probs pre_hybrid_coef_probs_8x8[BLOCK_TYPES_8X8];
vp9_coeff_probs pre_coef_probs_16x16[BLOCK_TYPES_16X16];
vp9_coeff_probs pre_hybrid_coef_probs_16x16[BLOCK_TYPES_16X16];
vp9_coeff_probs pre_coef_probs_32x32[BLOCK_TYPES_32X32];
vp9_coeff_probs pre_coef_probs_4x4[BLOCK_TYPES];
vp9_coeff_probs pre_coef_probs_8x8[BLOCK_TYPES];
vp9_coeff_probs pre_coef_probs_16x16[BLOCK_TYPES];
vp9_coeff_probs pre_coef_probs_32x32[BLOCK_TYPES];
#if CONFIG_CODE_NONZEROCOUNT
vp9_prob pre_nzc_probs_4x4[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC4X4_NODES];
vp9_prob pre_nzc_probs_8x8[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC8X8_NODES];
vp9_prob pre_nzc_probs_16x16[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC16X16_NODES];
vp9_prob pre_nzc_probs_32x32[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC32X32_NODES];
vp9_prob pre_nzc_pcat_probs[MAX_NZC_CONTEXTS]
[NZC_TOKENS_EXTRA][NZC_BITS_EXTRA];
#endif
vp9_coeff_count coef_counts_4x4[BLOCK_TYPES_4X4];
vp9_coeff_count hybrid_coef_counts_4x4[BLOCK_TYPES_4X4];
vp9_coeff_count coef_counts_8x8[BLOCK_TYPES_8X8];
vp9_coeff_count hybrid_coef_counts_8x8[BLOCK_TYPES_8X8];
vp9_coeff_count coef_counts_16x16[BLOCK_TYPES_16X16];
vp9_coeff_count hybrid_coef_counts_16x16[BLOCK_TYPES_16X16];
vp9_coeff_count coef_counts_32x32[BLOCK_TYPES_32X32];
vp9_coeff_count coef_counts_4x4[BLOCK_TYPES];
vp9_coeff_count coef_counts_8x8[BLOCK_TYPES];
vp9_coeff_count coef_counts_16x16[BLOCK_TYPES];
vp9_coeff_count coef_counts_32x32[BLOCK_TYPES];
unsigned int eob_branch_counts[TX_SIZE_MAX_SB][BLOCK_TYPES][REF_TYPES]
[COEF_BANDS][PREV_COEF_CONTEXTS];
#if CONFIG_CODE_NONZEROCOUNT
unsigned int nzc_counts_4x4[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC4X4_TOKENS];
unsigned int nzc_counts_8x8[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC8X8_TOKENS];
unsigned int nzc_counts_16x16[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC16X16_TOKENS];
unsigned int nzc_counts_32x32[MAX_NZC_CONTEXTS][REF_TYPES][BLOCK_TYPES]
[NZC32X32_TOKENS];
unsigned int nzc_pcat_counts[MAX_NZC_CONTEXTS]
[NZC_TOKENS_EXTRA][NZC_BITS_EXTRA][2];
#endif
nmv_context_counts NMVcount;
vp9_prob switchable_interp_prob[VP9_SWITCHABLE_FILTERS + 1]
@ -128,13 +168,14 @@ typedef struct VP9Common {
struct vpx_internal_error_info error;
DECLARE_ALIGNED(16, int16_t, Y1dequant[QINDEX_RANGE][16]);
DECLARE_ALIGNED(16, int16_t, Y2dequant[QINDEX_RANGE][16]);
DECLARE_ALIGNED(16, int16_t, UVdequant[QINDEX_RANGE][16]);
int Width;
int Height;
int horiz_scale;
int vert_scale;
int width;
int height;
int display_width;
int display_height;
int last_width;
int last_height;
YUV_TYPE clr_type;
CLAMP_TYPE clamp_type;
@ -142,8 +183,15 @@ typedef struct VP9Common {
YV12_BUFFER_CONFIG *frame_to_show;
YV12_BUFFER_CONFIG yv12_fb[NUM_YV12_BUFFERS];
int fb_idx_ref_cnt[NUM_YV12_BUFFERS];
int new_fb_idx, lst_fb_idx, gld_fb_idx, alt_fb_idx;
int fb_idx_ref_cnt[NUM_YV12_BUFFERS]; /* reference counts */
int ref_frame_map[NUM_REF_FRAMES]; /* maps fb_idx to reference slot */
/* TODO(jkoleszar): could expand active_ref_idx to 4, with 0 as intra, and
* roll new_fb_idx into it.
*/
int active_ref_idx[3]; /* each frame can reference 3 buffers */
int new_fb_idx;
struct scale_factors active_ref_scale[3];
YV12_BUFFER_CONFIG post_proc_buffer;
YV12_BUFFER_CONFIG temp_scale_frame;
@ -173,8 +221,6 @@ typedef struct VP9Common {
int last_kf_gf_q; /* Q used on the last GF or KF */
int y1dc_delta_q;
int y2dc_delta_q;
int y2ac_delta_q;
int uvdc_delta_q;
int uvac_delta_q;
@ -201,19 +247,13 @@ typedef struct VP9Common {
int filter_level;
int last_sharpness_level;
int sharpness_level;
int refresh_last_frame; /* Two state 0 = NO, 1 = YES */
int refresh_golden_frame; /* Two state 0 = NO, 1 = YES */
int refresh_alt_ref_frame; /* Two state 0 = NO, 1 = YES */
int copy_buffer_to_gf; /* 0 none, 1 Last to GF, 2 ARF to GF */
int copy_buffer_to_arf; /* 0 none, 1 Last to ARF, 2 GF to ARF */
int dering_enabled;
int refresh_entropy_probs; /* Two state 0 = NO, 1 = YES */
int ref_frame_sign_bias[MAX_REF_FRAMES]; /* Two state 0, 1 */
/* Y,U,V,Y2 */
/* Y,U,V */
ENTROPY_CONTEXT_PLANES *above_context; /* row of context for each plane */
ENTROPY_CONTEXT_PLANES left_context[4]; /* (up to) 4 contexts "" */
@ -250,9 +290,9 @@ typedef struct VP9Common {
vp9_prob mbskip_pred_probs[MBSKIP_CONTEXTS];
FRAME_CONTEXT lfc_a; /* last alt ref entropy */
FRAME_CONTEXT lfc; /* last frame entropy */
FRAME_CONTEXT fc; /* this frame entropy */
FRAME_CONTEXT frame_contexts[NUM_FRAME_CONTEXTS];
unsigned int frame_context_idx; /* Context to use/update */
unsigned int current_video_frame;
int near_boffset[3];
@ -272,6 +312,60 @@ typedef struct VP9Common {
int use_interintra;
#endif
int error_resilient_mode;
int frame_parallel_decoding_mode;
int tile_columns, log2_tile_columns;
int cur_tile_mb_col_start, cur_tile_mb_col_end, cur_tile_col_idx;
int tile_rows, log2_tile_rows;
int cur_tile_mb_row_start, cur_tile_mb_row_end, cur_tile_row_idx;
} VP9_COMMON;
static int get_free_fb(VP9_COMMON *cm) {
int i;
for (i = 0; i < NUM_YV12_BUFFERS; i++)
if (cm->fb_idx_ref_cnt[i] == 0)
break;
assert(i < NUM_YV12_BUFFERS);
cm->fb_idx_ref_cnt[i] = 1;
return i;
}
static void ref_cnt_fb(int *buf, int *idx, int new_idx) {
if (buf[*idx] > 0)
buf[*idx]--;
*idx = new_idx;
buf[new_idx]++;
}
// TODO(debargha): merge the two functions
static void set_mb_row(VP9_COMMON *cm, MACROBLOCKD *xd,
int mb_row, int block_size) {
xd->mb_to_top_edge = -((mb_row * 16) << 3);
xd->mb_to_bottom_edge = ((cm->mb_rows - block_size - mb_row) * 16) << 3;
// Are edges available for intra prediction?
xd->up_available = (mb_row != 0);
}
static void set_mb_col(VP9_COMMON *cm, MACROBLOCKD *xd,
int mb_col, int block_size) {
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_right_edge = ((cm->mb_cols - block_size - mb_col) * 16) << 3;
// Are edges available for intra prediction?
xd->left_available = (mb_col > cm->cur_tile_mb_col_start);
xd->right_available = (mb_col + block_size < cm->cur_tile_mb_col_end);
}
static int get_mb_row(const MACROBLOCKD *xd) {
return ((-xd->mb_to_top_edge) >> 7);
}
static int get_mb_col(const MACROBLOCKD *xd) {
return ((-xd->mb_to_left_edge) >> 7);
}
#endif // VP9_COMMON_VP9_ONYXC_INT_H_

43
vp9/common/vp9_postproc.c
View File

@ -336,11 +336,8 @@ void vp9_deblock(YV12_BUFFER_CONFIG *source,
source->uv_height, source->uv_width, ppl);
}
void vp9_de_noise(YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *post,
int q,
int low_var_thresh,
int flag) {
void vp9_denoise(YV12_BUFFER_CONFIG *src, YV12_BUFFER_CONFIG *post,
int q, int low_var_thresh, int flag) {
double level = 6.0e-05 * q * q * q - .0067 * q * q + .306 * q + .0065;
int ppl = (int)(level + .5);
(void) post;
@ -424,9 +421,9 @@ static void fillrd(struct postproc_state *state, int q, int a) {
*
* INPUTS : unsigned char *Start starting address of buffer to
* add gaussian noise to
* unsigned int Width width of plane
* unsigned int Height height of plane
* int Pitch distance between subsequent lines of frame
* unsigned int width width of plane
* unsigned int height height of plane
* int pitch distance between subsequent lines of frame
* int q quantizer used to determine amount of noise
* to add
*
@ -439,25 +436,25 @@ static void fillrd(struct postproc_state *state, int q, int a) {
* SPECIAL NOTES : None.
*
****************************************************************************/
void vp9_plane_add_noise_c(uint8_t *Start, char *noise,
void vp9_plane_add_noise_c(uint8_t *start, char *noise,
char blackclamp[16],
char whiteclamp[16],
char bothclamp[16],
unsigned int Width, unsigned int Height, int Pitch) {
unsigned int width, unsigned int height, int pitch) {
unsigned int i, j;
for (i = 0; i < Height; i++) {
uint8_t *Pos = Start + i * Pitch;
char *Ref = (char *)(noise + (rand() & 0xff));
for (i = 0; i < height; i++) {
uint8_t *pos = start + i * pitch;
char *ref = (char *)(noise + (rand() & 0xff)); // NOLINT
for (j = 0; j < Width; j++) {
if (Pos[j] < blackclamp[0])
Pos[j] = blackclamp[0];
for (j = 0; j < width; j++) {
if (pos[j] < blackclamp[0])
pos[j] = blackclamp[0];
if (Pos[j] > 255 + whiteclamp[0])
Pos[j] = 255 + whiteclamp[0];
if (pos[j] > 255 + whiteclamp[0])
pos[j] = 255 + whiteclamp[0];
Pos[j] += Ref[j];
pos[j] += ref[j];
}
}
}
@ -636,8 +633,8 @@ int vp9_post_proc_frame(VP9_COMMON *oci, YV12_BUFFER_CONFIG *dest,
*dest = *oci->frame_to_show;
/* handle problem with extending borders */
dest->y_width = oci->Width;
dest->y_height = oci->Height;
dest->y_width = oci->width;
dest->y_height = oci->height;
dest->uv_height = dest->y_height / 2;
return 0;
@ -1004,8 +1001,8 @@ int vp9_post_proc_frame(VP9_COMMON *oci, YV12_BUFFER_CONFIG *dest,
*dest = oci->post_proc_buffer;
/* handle problem with extending borders */
dest->y_width = oci->Width;
dest->y_height = oci->Height;
dest->y_width = oci->width;
dest->y_height = oci->height;
dest->uv_height = dest->y_height / 2;
return 0;

24
vp9/common/vp9_postproc.h
View File

@ -13,30 +13,26 @@
#define VP9_COMMON_VP9_POSTPROC_H_
#include "vpx_ports/mem.h"
struct postproc_state {
int last_q;
int last_noise;
char noise[3072];
int last_q;
int last_noise;
char noise[3072];
DECLARE_ALIGNED(16, char, blackclamp[16]);
DECLARE_ALIGNED(16, char, whiteclamp[16]);
DECLARE_ALIGNED(16, char, bothclamp[16]);
};
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/common/vp9_ppflags.h"
int vp9_post_proc_frame(struct VP9Common *oci, YV12_BUFFER_CONFIG *dest,
vp9_ppflags_t *flags);
void vp9_denoise(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *post,
int q, int low_var_thresh, int flag);
void vp9_de_noise(YV12_BUFFER_CONFIG *source,
YV12_BUFFER_CONFIG *post,
int q,
int low_var_thresh,
int flag);
void vp9_deblock(YV12_BUFFER_CONFIG *source,
YV12_BUFFER_CONFIG *post,
int q,
int low_var_thresh,
int flag);
void vp9_deblock(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *post,
int q, int low_var_thresh, int flag);
#endif // VP9_COMMON_VP9_POSTPROC_H_

1
vp9/common/vp9_pragmas.h
View File

@ -14,6 +14,7 @@
#ifdef __INTEL_COMPILER
#pragma warning(disable:997 1011 170)
#endif
#ifdef _MSC_VER
#pragma warning(disable:4799)
#endif

134
vp9/common/vp9_pred_common.c
View File

@ -29,14 +29,15 @@ unsigned char vp9_get_pred_context(const VP9_COMMON *const cm,
// The prediction flags in these dummy entries are initialised to 0.
switch (pred_id) {
case PRED_SEG_ID:
pred_context = (m - 1)->mbmi.seg_id_predicted +
(m - cm->mode_info_stride)->mbmi.seg_id_predicted;
pred_context = (m - cm->mode_info_stride)->mbmi.seg_id_predicted;
if (xd->left_available)
pred_context += (m - 1)->mbmi.seg_id_predicted;
break;
case PRED_REF:
pred_context = (m - 1)->mbmi.ref_predicted +
(m - cm->mode_info_stride)->mbmi.ref_predicted;
pred_context = (m - cm->mode_info_stride)->mbmi.ref_predicted;
if (xd->left_available)
pred_context += (m - 1)->mbmi.ref_predicted;
break;
case PRED_COMP:
@ -61,13 +62,14 @@ unsigned char vp9_get_pred_context(const VP9_COMMON *const cm,
break;
case PRED_MBSKIP:
pred_context = (m - 1)->mbmi.mb_skip_coeff +
(m - cm->mode_info_stride)->mbmi.mb_skip_coeff;
pred_context = (m - cm->mode_info_stride)->mbmi.mb_skip_coeff;
if (xd->left_available)
pred_context += (m - 1)->mbmi.mb_skip_coeff;
break;
case PRED_SWITCHABLE_INTERP:
{
int left_in_image = (m - 1)->mbmi.mb_in_image;
int left_in_image = xd->left_available && (m - 1)->mbmi.mb_in_image;
int above_in_image = (m - cm->mode_info_stride)->mbmi.mb_in_image;
int left_mode = (m - 1)->mbmi.mode;
int above_mode = (m - cm->mode_info_stride)->mbmi.mode;
@ -98,8 +100,7 @@ unsigned char vp9_get_pred_context(const VP9_COMMON *const cm,
break;
default:
// TODO *** add error trap code.
pred_context = 0;
pred_context = 0; // *** add error trap code.
break;
}
@ -111,39 +112,23 @@ unsigned char vp9_get_pred_context(const VP9_COMMON *const cm,
vp9_prob vp9_get_pred_prob(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
PRED_ID pred_id) {
vp9_prob pred_probability;
int pred_context;
// Get the appropriate prediction context
pred_context = vp9_get_pred_context(cm, xd, pred_id);
const int pred_context = vp9_get_pred_context(cm, xd, pred_id);
switch (pred_id) {
case PRED_SEG_ID:
pred_probability = cm->segment_pred_probs[pred_context];
break;
return cm->segment_pred_probs[pred_context];
case PRED_REF:
pred_probability = cm->ref_pred_probs[pred_context];
break;
return cm->ref_pred_probs[pred_context];
case PRED_COMP:
// In keeping with convention elsewhre the probability returned is
// the probability of a "0" outcome which in this case means the
// probability of comp pred off.
pred_probability = cm->prob_comppred[pred_context];
break;
return cm->prob_comppred[pred_context];
case PRED_MBSKIP:
pred_probability = cm->mbskip_pred_probs[pred_context];
break;
return cm->mbskip_pred_probs[pred_context];
default:
// TODO *** add error trap code.
pred_probability = 128;
break;
return 128; // *** add error trap code.
}
return pred_probability;
}
// This function returns a context probability ptr for coding a given
@ -151,71 +136,41 @@ vp9_prob vp9_get_pred_prob(const VP9_COMMON *const cm,
const vp9_prob *vp9_get_pred_probs(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
PRED_ID pred_id) {
const vp9_prob *pred_probability;
int pred_context;
// Get the appropriate prediction context
pred_context = vp9_get_pred_context(cm, xd, pred_id);
const int pred_context = vp9_get_pred_context(cm, xd, pred_id);
switch (pred_id) {
case PRED_SEG_ID:
pred_probability = &cm->segment_pred_probs[pred_context];
break;
return &cm->segment_pred_probs[pred_context];
case PRED_REF:
pred_probability = &cm->ref_pred_probs[pred_context];
break;
return &cm->ref_pred_probs[pred_context];
case PRED_COMP:
// In keeping with convention elsewhre the probability returned is
// the probability of a "0" outcome which in this case means the
// probability of comp pred off.
pred_probability = &cm->prob_comppred[pred_context];
break;
return &cm->prob_comppred[pred_context];
case PRED_MBSKIP:
pred_probability = &cm->mbskip_pred_probs[pred_context];
break;
return &cm->mbskip_pred_probs[pred_context];
case PRED_SWITCHABLE_INTERP:
pred_probability = &cm->fc.switchable_interp_prob[pred_context][0];
break;
return &cm->fc.switchable_interp_prob[pred_context][0];
default:
// TODO *** add error trap code.
pred_probability = NULL;
break;
return NULL; // *** add error trap code.
}
return pred_probability;
}
// This function returns the status of the given prediction signal.
// I.e. is the predicted value for the given signal correct.
unsigned char vp9_get_pred_flag(const MACROBLOCKD *const xd,
PRED_ID pred_id) {
unsigned char pred_flag = 0;
switch (pred_id) {
case PRED_SEG_ID:
pred_flag = xd->mode_info_context->mbmi.seg_id_predicted;
break;
return xd->mode_info_context->mbmi.seg_id_predicted;
case PRED_REF:
pred_flag = xd->mode_info_context->mbmi.ref_predicted;
break;
return xd->mode_info_context->mbmi.ref_predicted;
case PRED_MBSKIP:
pred_flag = xd->mode_info_context->mbmi.mb_skip_coeff;
break;
return xd->mode_info_context->mbmi.mb_skip_coeff;
default:
// TODO *** add error trap code.
pred_flag = 0;
break;
return 0; // *** add error trap code.
}
return pred_flag;
}
// This function sets the status of the given prediction signal.
@ -277,7 +232,7 @@ void vp9_set_pred_flag(MACROBLOCKD *const xd,
break;
default:
// TODO *** add error trap code.
// *** add error trap code.
break;
}
}
@ -322,7 +277,6 @@ MV_REFERENCE_FRAME vp9_get_pred_ref(const VP9_COMMON *const cm,
MV_REFERENCE_FRAME pred_ref = LAST_FRAME;
int segment_id = xd->mode_info_context->mbmi.segment_id;
int seg_ref_active;
int i;
unsigned char frame_allowed[MAX_REF_FRAMES] = {1, 1, 1, 1};
@ -333,7 +287,7 @@ MV_REFERENCE_FRAME vp9_get_pred_ref(const VP9_COMMON *const cm,
unsigned char above_left_in_image;
// Is segment coding ennabled
seg_ref_active = vp9_segfeature_active(xd, segment_id, SEG_LVL_REF_FRAME);
int seg_ref_active = vp9_segfeature_active(xd, segment_id, SEG_LVL_REF_FRAME);
// Special case treatment if segment coding is enabled.
// Dont allow prediction of a reference frame that the segment
@ -355,9 +309,10 @@ MV_REFERENCE_FRAME vp9_get_pred_ref(const VP9_COMMON *const cm,
above_left = (m - 1 - cm->mode_info_stride)->mbmi.ref_frame;
// Are neighbours in image
left_in_image = (m - 1)->mbmi.mb_in_image;
left_in_image = (m - 1)->mbmi.mb_in_image && xd->left_available;
above_in_image = (m - cm->mode_info_stride)->mbmi.mb_in_image;
above_left_in_image = (m - 1 - cm->mode_info_stride)->mbmi.mb_in_image;
above_left_in_image = (m - 1 - cm->mode_info_stride)->mbmi.mb_in_image &&
xd->left_available;
// Adjust scores for candidate reference frames based on neigbours
if (frame_allowed[left] && left_in_image) {
@ -385,9 +340,7 @@ MV_REFERENCE_FRAME vp9_get_pred_ref(const VP9_COMMON *const cm,
// Functions to computes a set of modified reference frame probabilities
// to use when the prediction of the reference frame value fails
void vp9_calc_ref_probs(int *count, vp9_prob *probs) {
int tot_count;
tot_count = count[0] + count[1] + count[2] + count[3];
int tot_count = count[0] + count[1] + count[2] + count[3];
probs[0] = get_prob(count[0], tot_count);
tot_count -= count[0];
@ -403,19 +356,12 @@ void vp9_calc_ref_probs(int *count, vp9_prob *probs) {
// they are not allowed for a given segment.
void vp9_compute_mod_refprobs(VP9_COMMON *const cm) {
int norm_cnt[MAX_REF_FRAMES];
int intra_count;
int inter_count;
int last_count;
int gfarf_count;
int gf_count;
int arf_count;
intra_count = cm->prob_intra_coded;
inter_count = (255 - intra_count);
last_count = (inter_count * cm->prob_last_coded) / 255;
gfarf_count = inter_count - last_count;
gf_count = (gfarf_count * cm->prob_gf_coded) / 255;
arf_count = gfarf_count - gf_count;
const int intra_count = cm->prob_intra_coded;
const int inter_count = (255 - intra_count);
const int last_count = (inter_count * cm->prob_last_coded) / 255;
const int gfarf_count = inter_count - last_count;
const int gf_count = (gfarf_count * cm->prob_gf_coded) / 255;
const int arf_count = gfarf_count - gf_count;
// Work out modified reference frame probabilities to use where prediction
// of the reference frame fails

48
vp9/common/vp9_pred_common.h
View File

@ -8,48 +8,48 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/common/vp9_blockd.h"
#ifndef VP9_COMMON_VP9_PRED_COMMON_H_
#define VP9_COMMON_VP9_PRED_COMMON_H_
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_onyxc_int.h"
// Predicted items
typedef enum {
PRED_SEG_ID = 0, // Segment identifier
PRED_SEG_ID = 0, // Segment identifier
PRED_REF = 1,
PRED_COMP = 2,
PRED_MBSKIP = 3,
PRED_SWITCHABLE_INTERP = 4
} PRED_ID;
extern unsigned char vp9_get_pred_context(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
PRED_ID pred_id);
unsigned char vp9_get_pred_context(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
PRED_ID pred_id);
extern vp9_prob vp9_get_pred_prob(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
PRED_ID pred_id);
vp9_prob vp9_get_pred_prob(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
PRED_ID pred_id);
extern const vp9_prob *vp9_get_pred_probs(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
PRED_ID pred_id);
const vp9_prob *vp9_get_pred_probs(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
PRED_ID pred_id);
extern unsigned char vp9_get_pred_flag(const MACROBLOCKD *const xd,
PRED_ID pred_id);
unsigned char vp9_get_pred_flag(const MACROBLOCKD *const xd,
PRED_ID pred_id);
extern void vp9_set_pred_flag(MACROBLOCKD *const xd,
PRED_ID pred_id,
unsigned char pred_flag);
void vp9_set_pred_flag(MACROBLOCKD *const xd,
PRED_ID pred_id,
unsigned char pred_flag);
extern unsigned char vp9_get_pred_mb_segid(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
int MbIndex);
unsigned char vp9_get_pred_mb_segid(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd,
int MbIndex);
extern MV_REFERENCE_FRAME vp9_get_pred_ref(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd);
extern void vp9_compute_mod_refprobs(VP9_COMMON *const cm);
MV_REFERENCE_FRAME vp9_get_pred_ref(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd);
void vp9_compute_mod_refprobs(VP9_COMMON *const cm);
#endif // VP9_COMMON_VP9_PRED_COMMON_H_

90
vp9/common/vp9_quant_common.c
View File

@ -8,7 +8,7 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_quant_common.h"
static int dc_qlookup[QINDEX_RANGE];
@ -24,7 +24,7 @@ void vp9_init_quant_tables() {
for (i = 0; i < QINDEX_RANGE; i++) {
ac_qlookup[i] = current_val;
current_val = (int)((double)current_val * 1.02);
current_val = (int)(current_val * 1.02);
if (current_val == last_val)
current_val++;
last_val = current_val;
@ -38,88 +38,18 @@ void vp9_init_quant_tables() {
}
}
int vp9_dc_quant(int QIndex, int Delta) {
int retval;
QIndex = QIndex + Delta;
if (QIndex > MAXQ)
QIndex = MAXQ;
else if (QIndex < 0)
QIndex = 0;
retval = dc_qlookup[ QIndex ];
return retval;
int vp9_dc_quant(int qindex, int delta) {
return dc_qlookup[clamp(qindex + delta, 0, MAXQ)];
}
int vp9_dc2quant(int QIndex, int Delta) {
int retval;
QIndex = QIndex + Delta;
if (QIndex > MAXQ)
QIndex = MAXQ;
else if (QIndex < 0)
QIndex = 0;
retval = dc_qlookup[ QIndex ];
return retval;
}
int vp9_dc_uv_quant(int QIndex, int Delta) {
int retval;
QIndex = QIndex + Delta;
if (QIndex > MAXQ)
QIndex = MAXQ;
else if (QIndex < 0)
QIndex = 0;
retval = dc_qlookup[ QIndex ];
return retval;
int vp9_dc_uv_quant(int qindex, int delta) {
return dc_qlookup[clamp(qindex + delta, 0, MAXQ)];
}
int vp9_ac_yquant(int QIndex) {
int retval;
if (QIndex > MAXQ)
QIndex = MAXQ;
else if (QIndex < 0)
QIndex = 0;
retval = ac_qlookup[ QIndex ];
return retval;
int vp9_ac_yquant(int qindex) {
return ac_qlookup[clamp(qindex, 0, MAXQ)];
}
int vp9_ac2quant(int QIndex, int Delta) {
int retval;
QIndex = QIndex + Delta;
if (QIndex > MAXQ)
QIndex = MAXQ;
else if (QIndex < 0)
QIndex = 0;
retval = (ac_qlookup[ QIndex ] * 775) / 1000;
if (retval < 4)
retval = 4;
return retval;
}
int vp9_ac_uv_quant(int QIndex, int Delta) {
int retval;
QIndex = QIndex + Delta;
if (QIndex > MAXQ)
QIndex = MAXQ;
else if (QIndex < 0)
QIndex = 0;
retval = ac_qlookup[ QIndex ];
return retval;
int vp9_ac_uv_quant(int qindex, int delta) {
return ac_qlookup[clamp(qindex + delta, 0, MAXQ)];
}

15
vp9/common/vp9_quant_common.h
View File

@ -11,16 +11,15 @@
#ifndef VP9_COMMON_VP9_QUANT_COMMON_H_
#define VP9_COMMON_VP9_QUANT_COMMON_H_
#include "string.h"
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_onyxc_int.h"
extern void vp9_init_quant_tables(void);
extern int vp9_ac_yquant(int QIndex);
extern int vp9_dc_quant(int QIndex, int Delta);
extern int vp9_dc2quant(int QIndex, int Delta);
extern int vp9_ac2quant(int QIndex, int Delta);
extern int vp9_dc_uv_quant(int QIndex, int Delta);
extern int vp9_ac_uv_quant(int QIndex, int Delta);
void vp9_init_quant_tables();
int vp9_ac_yquant(int qindex);
int vp9_dc_quant(int qindex, int delta);
int vp9_dc2quant(int qindex, int delta);
int vp9_ac2quant(int qindex, int delta);
int vp9_dc_uv_quant(int qindex, int delta);
int vp9_ac_uv_quant(int qindex, int delta);
#endif // VP9_COMMON_VP9_QUANT_COMMON_H_

36
vp9/common/vp9_recon.c
View File

@ -117,7 +117,7 @@ void vp9_recon_mbuv_s_c(MACROBLOCKD *xd, uint8_t *udst, uint8_t *vdst) {
void vp9_recon_sby_s_c(MACROBLOCKD *xd, uint8_t *dst) {
int x, y, stride = xd->block[0].dst_stride;
int16_t *diff = xd->sb_coeff_data.diff;
int16_t *diff = xd->diff;
for (y = 0; y < 32; y++) {
for (x = 0; x < 32; x++) {
@ -130,8 +130,8 @@ void vp9_recon_sby_s_c(MACROBLOCKD *xd, uint8_t *dst) {
void vp9_recon_sbuv_s_c(MACROBLOCKD *xd, uint8_t *udst, uint8_t *vdst) {
int x, y, stride = xd->block[16].dst_stride;
int16_t *udiff = xd->sb_coeff_data.diff + 1024;
int16_t *vdiff = xd->sb_coeff_data.diff + 1280;
int16_t *udiff = xd->diff + 1024;
int16_t *vdiff = xd->diff + 1280;
for (y = 0; y < 16; y++) {
for (x = 0; x < 16; x++) {
@ -145,6 +145,36 @@ void vp9_recon_sbuv_s_c(MACROBLOCKD *xd, uint8_t *udst, uint8_t *vdst) {
}
}
void vp9_recon_sb64y_s_c(MACROBLOCKD *xd, uint8_t *dst) {
int x, y, stride = xd->block[0].dst_stride;
int16_t *diff = xd->diff;
for (y = 0; y < 64; y++) {
for (x = 0; x < 64; x++) {
dst[x] = clip_pixel(dst[x] + diff[x]);
}
dst += stride;
diff += 64;
}
}
void vp9_recon_sb64uv_s_c(MACROBLOCKD *xd, uint8_t *udst, uint8_t *vdst) {
int x, y, stride = xd->block[16].dst_stride;
int16_t *udiff = xd->diff + 4096;
int16_t *vdiff = xd->diff + 4096 + 1024;
for (y = 0; y < 32; y++) {
for (x = 0; x < 32; x++) {
udst[x] = clip_pixel(udst[x] + udiff[x]);
vdst[x] = clip_pixel(vdst[x] + vdiff[x]);
}
udst += stride;
vdst += stride;
udiff += 32;
vdiff += 32;
}
}
void vp9_recon_mby_c(MACROBLOCKD *xd) {
int i;

2157
vp9/common/vp9_reconinter.c
View File

File diff suppressed because it is too large Load Diff

161
vp9/common/vp9_reconinter.h
View File

@ -14,71 +14,128 @@
#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_onyxc_int.h"
extern void vp9_build_1st_inter16x16_predictors_mby(MACROBLOCKD *xd,
uint8_t *dst_y,
int dst_ystride,
int clamp_mvs);
struct subpix_fn_table;
extern void vp9_build_1st_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_uvstride);
void vp9_build_inter16x16_predictors_mby(MACROBLOCKD *xd,
uint8_t *dst_y,
int dst_ystride,
int mb_row,
int mb_col);
extern void vp9_build_1st_inter16x16_predictors_mb(MACROBLOCKD *xd,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride);
void vp9_build_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_uvstride,
int mb_row,
int mb_col);
extern void vp9_build_2nd_inter16x16_predictors_mby(MACROBLOCKD *xd,
uint8_t *dst_y,
int dst_ystride);
void vp9_build_inter16x16_predictors_mb(MACROBLOCKD *xd,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride,
int mb_row,
int mb_col);
extern void vp9_build_2nd_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_uvstride);
void vp9_build_inter32x32_predictors_sb(MACROBLOCKD *x,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride,
int mb_row,
int mb_col);
extern void vp9_build_2nd_inter16x16_predictors_mb(MACROBLOCKD *xd,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride);
void vp9_build_inter64x64_predictors_sb(MACROBLOCKD *x,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride,
int mb_row,
int mb_col);
extern void vp9_build_inter32x32_predictors_sb(MACROBLOCKD *x,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride);
void vp9_build_inter_predictors_mb(MACROBLOCKD *xd,
int mb_row,
int mb_col);
extern void vp9_build_inter64x64_predictors_sb(MACROBLOCKD *x,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride);
void vp9_build_inter4x4_predictors_mbuv(MACROBLOCKD *xd,
int mb_row,
int mb_col);
extern void vp9_build_inter_predictors_mb(MACROBLOCKD *xd);
void vp9_setup_interp_filters(MACROBLOCKD *xd,
INTERPOLATIONFILTERTYPE filter,
VP9_COMMON *cm);
extern void vp9_build_inter_predictors_b(BLOCKD *d, int pitch,
vp9_subpix_fn_t sppf);
void vp9_setup_scale_factors_for_frame(struct scale_factors *scale,
YV12_BUFFER_CONFIG *other,
int this_w, int this_h);
extern void vp9_build_2nd_inter_predictors_b(BLOCKD *d, int pitch,
vp9_subpix_fn_t sppf);
void vp9_build_inter_predictor(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int_mv *mv_q3,
const struct scale_factors *scale,
int w, int h, int do_avg,
const struct subpix_fn_table *subpix);
extern void vp9_build_inter_predictors4b(MACROBLOCKD *xd, BLOCKD *d,
int pitch);
void vp9_build_inter_predictor_q4(const uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const int_mv *fullpel_mv_q3,
const int_mv *frac_mv_q4,
const struct scale_factors *scale,
int w, int h, int do_avg,
const struct subpix_fn_table *subpix);
extern void vp9_build_2nd_inter_predictors4b(MACROBLOCKD *xd,
BLOCKD *d, int pitch);
static int scale_value_x(int val, const struct scale_factors *scale) {
return val * scale->x_num / scale->x_den;
}
extern void vp9_build_inter4x4_predictors_mbuv(MACROBLOCKD *xd);
static int scale_value_y(int val, const struct scale_factors *scale) {
return val * scale->y_num / scale->y_den;
}
extern void vp9_setup_interp_filters(MACROBLOCKD *xd,
INTERPOLATIONFILTERTYPE filter,
VP9_COMMON *cm);
static int scaled_buffer_offset(int x_offset,
int y_offset,
int stride,
const struct scale_factors *scale) {
return scale_value_y(y_offset, scale) * stride +
scale_value_x(x_offset, scale);
}
static void setup_pred_block(YV12_BUFFER_CONFIG *dst,
const YV12_BUFFER_CONFIG *src,
int mb_row, int mb_col,
const struct scale_factors *scale,
const struct scale_factors *scale_uv) {
const int recon_y_stride = src->y_stride;
const int recon_uv_stride = src->uv_stride;
int recon_yoffset;
int recon_uvoffset;
if (scale) {
recon_yoffset = scaled_buffer_offset(16 * mb_col, 16 * mb_row,
recon_y_stride, scale);
recon_uvoffset = scaled_buffer_offset(8 * mb_col, 8 * mb_row,
recon_uv_stride, scale_uv);
} else {
recon_yoffset = 16 * mb_row * recon_y_stride + 16 * mb_col;
recon_uvoffset = 8 * mb_row * recon_uv_stride + 8 * mb_col;
}
*dst = *src;
dst->y_buffer += recon_yoffset;
dst->u_buffer += recon_uvoffset;
dst->v_buffer += recon_uvoffset;
}
static void set_scale_factors(MACROBLOCKD *xd,
int ref0, int ref1,
struct scale_factors scale_factor[MAX_REF_FRAMES]) {
xd->scale_factor[0] = scale_factor[ref0 >= 0 ? ref0 : 0];
xd->scale_factor[1] = scale_factor[ref1 >= 0 ? ref1 : 0];
xd->scale_factor_uv[0] = xd->scale_factor[0];
xd->scale_factor_uv[1] = xd->scale_factor[1];
}
#endif // VP9_COMMON_VP9_RECONINTER_H_

520
vp9/common/vp9_reconintra.c
View File

@ -9,106 +9,131 @@
*/
#include <stdio.h>
#include "./vpx_config.h"
#include "vp9_rtcd.h"
#include "vp9/common/vp9_reconintra.h"
#include "vpx_mem/vpx_mem.h"
/* For skip_recon_mb(), add vp9_build_intra_predictors_mby_s(MACROBLOCKD *xd)
* and vp9_build_intra_predictors_mbuv_s(MACROBLOCKD *xd).
*/
// For skip_recon_mb(), add vp9_build_intra_predictors_mby_s(MACROBLOCKD *xd)
// and vp9_build_intra_predictors_mbuv_s(MACROBLOCKD *xd).
// Using multiplication and shifting instead of division in diagonal prediction.
// iscale table is calculated from ((1 << 16) + (i + 2) / 2) / (i+2) and used as
// ((A + B) * iscale[i] + (1 << 15)) >> 16;
// where A and B are weighted pixel values.
static const unsigned int iscale[64] = {
32768, 21845, 16384, 13107, 10923, 9362, 8192, 7282,
6554, 5958, 5461, 5041, 4681, 4369, 4096, 3855,
3641, 3449, 3277, 3121, 2979, 2849, 2731, 2621,
2521, 2427, 2341, 2260, 2185, 2114, 2048, 1986,
1928, 1872, 1820, 1771, 1725, 1680, 1638, 1598,
1560, 1524, 1489, 1456, 1425, 1394, 1365, 1337,
1311, 1285, 1260, 1237, 1214, 1192, 1170, 1150,
1130, 1111, 1092, 1074, 1057, 1040, 1024, 1008,
};
static INLINE int iscale_round(int value, int i) {
return ROUND_POWER_OF_TWO(value * iscale[i], 16);
}
static void d27_predictor(uint8_t *ypred_ptr, int y_stride, int n,
uint8_t *yabove_row, uint8_t *yleft_col) {
int r, c, h, w, v;
int a, b;
int r, c;
r = 0;
for (c = 0; c < n - 2; c++) {
if (c & 1)
a = yleft_col[r + 1];
else
a = (yleft_col[r] + yleft_col[r + 1] + 1) >> 1;
b = yabove_row[c + 2];
ypred_ptr[c] = (2 * a + (c + 1) * b + (c + 3) / 2) / (c + 3);
int a = c & 1 ? yleft_col[r + 1]
: ROUND_POWER_OF_TWO(yleft_col[r] + yleft_col[r + 1], 1);
int b = yabove_row[c + 2];
ypred_ptr[c] = iscale_round(2 * a + (c + 1) * b, 1 + c);
}
for (r = 1; r < n / 2 - 1; r++) {
for (c = 0; c < n - 2 - 2 * r; c++) {
if (c & 1)
a = yleft_col[r + 1];
else
a = (yleft_col[r] + yleft_col[r + 1] + 1) >> 1;
b = ypred_ptr[(r - 1) * y_stride + c + 2];
ypred_ptr[r * y_stride + c] = (2 * a + (c + 1) * b + (c + 3) / 2) / (c + 3);
int a = c & 1 ? yleft_col[r + 1]
: ROUND_POWER_OF_TWO(yleft_col[r] + yleft_col[r + 1], 1);
int b = ypred_ptr[(r - 1) * y_stride + c + 2];
ypred_ptr[r * y_stride + c] = iscale_round(2 * a + (c + 1) * b, 1 + c);
}
}
for (; r < n - 1; ++r) {
for (; r < n - 1; r++) {
for (c = 0; c < n; c++) {
v = (c & 1 ? yleft_col[r + 1] : (yleft_col[r] + yleft_col[r + 1] + 1) >> 1);
h = r - c / 2;
int v = c & 1 ? yleft_col[r + 1]
: ROUND_POWER_OF_TWO(yleft_col[r] + yleft_col[r + 1], 1);
int h = r - c / 2;
ypred_ptr[h * y_stride + c] = v;
}
}
c = 0;
r = n - 1;
ypred_ptr[r * y_stride] = (ypred_ptr[(r - 1) * y_stride] +
yleft_col[r] + 1) >> 1;
ypred_ptr[r * y_stride] = ROUND_POWER_OF_TWO(ypred_ptr[(r - 1) * y_stride] +
yleft_col[r], 1);
for (r = n - 2; r >= n / 2; --r) {
w = c + (n - 1 - r) * 2;
ypred_ptr[r * y_stride + w] = (ypred_ptr[(r - 1) * y_stride + w] +
ypred_ptr[r * y_stride + w - 1] + 1) >> 1;
int w = c + (n - 1 - r) * 2;
ypred_ptr[r * y_stride + w] =
ROUND_POWER_OF_TWO(ypred_ptr[(r - 1) * y_stride + w] +
ypred_ptr[r * y_stride + w - 1], 1);
}
for (c = 1; c < n; c++) {
for (r = n - 1; r >= n / 2 + c / 2; --r) {
w = c + (n - 1 - r) * 2;
ypred_ptr[r * y_stride + w] = (ypred_ptr[(r - 1) * y_stride + w] +
ypred_ptr[r * y_stride + w - 1] + 1) >> 1;
int w = c + (n - 1 - r) * 2;
ypred_ptr[r * y_stride + w] =
ROUND_POWER_OF_TWO(ypred_ptr[(r - 1) * y_stride + w] +
ypred_ptr[r * y_stride + w - 1], 1);
}
}
}
static void d63_predictor(uint8_t *ypred_ptr, int y_stride, int n,
uint8_t *yabove_row, uint8_t *yleft_col) {
int r, c, h, w, v;
int a, b;
int r, c;
c = 0;
for (r = 0; r < n - 2; r++) {
if (r & 1)
a = yabove_row[c + 1];
else
a = (yabove_row[c] + yabove_row[c + 1] + 1) >> 1;
b = yleft_col[r + 2];
ypred_ptr[r * y_stride] = (2 * a + (r + 1) * b + (r + 3) / 2) / (r + 3);
int a = r & 1 ? yabove_row[c + 1]
: ROUND_POWER_OF_TWO(yabove_row[c] + yabove_row[c + 1], 1);
int b = yleft_col[r + 2];
ypred_ptr[r * y_stride] = iscale_round(2 * a + (r + 1) * b, 1 + r);
}
for (c = 1; c < n / 2 - 1; c++) {
for (r = 0; r < n - 2 - 2 * c; r++) {
if (r & 1)
a = yabove_row[c + 1];
else
a = (yabove_row[c] + yabove_row[c + 1] + 1) >> 1;
b = ypred_ptr[(r + 2) * y_stride + c - 1];
ypred_ptr[r * y_stride + c] = (2 * a + (c + 1) * b + (c + 3) / 2) / (c + 3);
int a = r & 1 ? yabove_row[c + 1]
: ROUND_POWER_OF_TWO(yabove_row[c] + yabove_row[c + 1], 1);
int b = ypred_ptr[(r + 2) * y_stride + c - 1];
ypred_ptr[r * y_stride + c] = iscale_round(2 * a + (c + 1) * b, 1 + c);
}
}
for (; c < n - 1; ++c) {
for (r = 0; r < n; r++) {
v = (r & 1 ? yabove_row[c + 1] : (yabove_row[c] + yabove_row[c + 1] + 1) >> 1);
w = c - r / 2;
int v = r & 1 ? yabove_row[c + 1]
: ROUND_POWER_OF_TWO(yabove_row[c] + yabove_row[c + 1], 1);
int w = c - r / 2;
ypred_ptr[r * y_stride + w] = v;
}
}
r = 0;
c = n - 1;
ypred_ptr[c] = (ypred_ptr[(c - 1)] + yabove_row[c] + 1) >> 1;
ypred_ptr[c] = ROUND_POWER_OF_TWO(ypred_ptr[(c - 1)] + yabove_row[c], 1);
for (c = n - 2; c >= n / 2; --c) {
h = r + (n - 1 - c) * 2;
ypred_ptr[h * y_stride + c] = (ypred_ptr[h * y_stride + c - 1] +
ypred_ptr[(h - 1) * y_stride + c] + 1) >> 1;
int h = r + (n - 1 - c) * 2;
ypred_ptr[h * y_stride + c] =
ROUND_POWER_OF_TWO(ypred_ptr[h * y_stride + c - 1] +
ypred_ptr[(h - 1) * y_stride + c], 1);
}
for (r = 1; r < n; r++) {
for (c = n - 1; c >= n / 2 + r / 2; --c) {
h = r + (n - 1 - c) * 2;
ypred_ptr[h * y_stride + c] = (ypred_ptr[h * y_stride + c - 1] +
ypred_ptr[(h - 1) * y_stride + c] + 1) >> 1;
int h = r + (n - 1 - c) * 2;
ypred_ptr[h * y_stride + c] =
ROUND_POWER_OF_TWO(ypred_ptr[h * y_stride + c - 1] +
ypred_ptr[(h - 1) * y_stride + c], 1);
}
}
}
@ -116,27 +141,28 @@ static void d63_predictor(uint8_t *ypred_ptr, int y_stride, int n,
static void d45_predictor(uint8_t *ypred_ptr, int y_stride, int n,
uint8_t *yabove_row, uint8_t *yleft_col) {
int r, c;
for (r = 0; r < n - 1; ++r) {
for (c = 0; c <= r; ++c) {
ypred_ptr[(r - c) * y_stride + c] =
(yabove_row[r + 1] * (c + 1) +
yleft_col[r + 1] * (r - c + 1) + r / 2 + 1) / (r + 2);
ypred_ptr[(r - c) * y_stride + c] = iscale_round(
yabove_row[r + 1] * (c + 1) + yleft_col[r + 1] * (r - c + 1), r);
}
}
for (c = 0; c <= r; ++c) {
int yabove_ext = yabove_row[r]; // clip_pixel(2 * yabove_row[r] -
// yabove_row[r - 1]);
int yleft_ext = yleft_col[r]; // clip_pixel(2 * yleft_col[r] -
// yleft_col[r-1]);
ypred_ptr[(r - c) * y_stride + c] =
(yabove_ext * (c + 1) +
yleft_ext * (r - c + 1) + r / 2 + 1) / (r + 2);
iscale_round(yabove_ext * (c + 1) + yleft_ext * (r - c + 1), r);
}
for (r = 1; r < n; ++r) {
for (c = n - r; c < n; ++c) {
const int yabove_ext = ypred_ptr[(r - 1) * y_stride + c];
const int yleft_ext = ypred_ptr[r * y_stride + c - 1];
ypred_ptr[r * y_stride + c] = (yabove_ext + yleft_ext + 1) >> 1;
ypred_ptr[r * y_stride + c] =
ROUND_POWER_OF_TWO(yabove_ext + yleft_ext, 1);
}
}
}
@ -145,7 +171,7 @@ static void d117_predictor(uint8_t *ypred_ptr, int y_stride, int n,
uint8_t *yabove_row, uint8_t *yleft_col) {
int r, c;
for (c = 0; c < n; c++)
ypred_ptr[c] = (yabove_row[c - 1] + yabove_row[c] + 1) >> 1;
ypred_ptr[c] = ROUND_POWER_OF_TWO(yabove_row[c - 1] + yabove_row[c], 1);
ypred_ptr += y_stride;
for (c = 0; c < n; c++)
ypred_ptr[c] = yabove_row[c - 1];
@ -179,9 +205,10 @@ static void d135_predictor(uint8_t *ypred_ptr, int y_stride, int n,
static void d153_predictor(uint8_t *ypred_ptr, int y_stride, int n,
uint8_t *yabove_row, uint8_t *yleft_col) {
int r, c;
ypred_ptr[0] = (yabove_row[-1] + yleft_col[0] + 1) >> 1;
ypred_ptr[0] = ROUND_POWER_OF_TWO(yabove_row[-1] + yleft_col[0], 1);
for (r = 1; r < n; r++)
ypred_ptr[r * y_stride] = (yleft_col[r - 1] + yleft_col[r] + 1) >> 1;
ypred_ptr[r * y_stride] =
ROUND_POWER_OF_TWO(yleft_col[r - 1] + yleft_col[r], 1);
ypred_ptr++;
ypred_ptr[0] = yabove_row[-1];
for (r = 1; r < n; r++)
@ -248,20 +275,58 @@ void vp9_recon_intra_mbuv(MACROBLOCKD *xd) {
}
}
static INLINE int log2_minus_1(int n) {
switch (n) {
case 4: return 1;
case 8: return 2;
case 16: return 3;
case 32: return 4;
case 64: return 5;
default:
assert(0);
return 0;
}
}
void vp9_build_intra_predictors_internal(uint8_t *src, int src_stride,
uint8_t *ypred_ptr,
int y_stride, int mode, int bsize,
int up_available, int left_available) {
uint8_t *yabove_row = src - src_stride;
uint8_t yleft_col[64];
uint8_t ytop_left = yabove_row[-1];
int up_available, int left_available,
int right_available) {
int r, c, i;
uint8_t yleft_col[64], yabove_data[65], ytop_left;
uint8_t *yabove_row = yabove_data + 1;
/*
* 127 127 127 .. 127 127 127 127 127 127
* 129 A B .. Y Z
* 129 C D .. W X
* 129 E F .. U V
* 129 G H .. S T T T T T
* ..
*/
for (i = 0; i < bsize; i++) {
yleft_col[i] = src[i * src_stride - 1];
if (left_available) {
for (i = 0; i < bsize; i++)
yleft_col[i] = src[i * src_stride - 1];
} else {
vpx_memset(yleft_col, 129, bsize);
}
if (up_available) {
uint8_t *yabove_ptr = src - src_stride;
vpx_memcpy(yabove_row, yabove_ptr, bsize);
if (left_available) {
ytop_left = yabove_ptr[-1];
} else {
ytop_left = 127;
}
} else {
vpx_memset(yabove_row, 127, bsize);
ytop_left = 127;
}
yabove_row[-1] = ytop_left;
/* for Y */
switch (mode) {
case DC_PRED: {
@ -269,22 +334,7 @@ void vp9_build_intra_predictors_internal(uint8_t *src, int src_stride,
int i;
int shift;
int average = 0;
int log2_bsize_minus_1;
assert(bsize == 4 || bsize == 8 || bsize == 16 || bsize == 32 ||
bsize == 64);
if (bsize == 4) {
log2_bsize_minus_1 = 1;
} else if (bsize == 8) {
log2_bsize_minus_1 = 2;
} else if (bsize == 16) {
log2_bsize_minus_1 = 3;
} else if (bsize == 32) {
log2_bsize_minus_1 = 4;
} else {
assert(bsize == 64);
log2_bsize_minus_1 = 5;
}
int log2_bsize_minus_1 = log2_minus_1(bsize);
if (up_available || left_available) {
if (up_available) {
@ -299,7 +349,7 @@ void vp9_build_intra_predictors_internal(uint8_t *src, int src_stride,
}
}
shift = log2_bsize_minus_1 + up_available + left_available;
expected_dc = (average + (1 << (shift - 1))) >> shift;
expected_dc = ROUND_POWER_OF_TWO(average, shift);
} else {
expected_dc = 128;
}
@ -310,21 +360,19 @@ void vp9_build_intra_predictors_internal(uint8_t *src, int src_stride,
}
}
break;
case V_PRED: {
case V_PRED:
for (r = 0; r < bsize; r++) {
memcpy(ypred_ptr, yabove_row, bsize);
ypred_ptr += y_stride;
}
}
break;
case H_PRED: {
break;
case H_PRED:
for (r = 0; r < bsize; r++) {
vpx_memset(ypred_ptr, yleft_col[r], bsize);
ypred_ptr += y_stride;
}
}
break;
case TM_PRED: {
break;
case TM_PRED:
for (r = 0; r < bsize; r++) {
for (c = 0; c < bsize; c++) {
ypred_ptr[c] = clip_pixel(yleft_col[r] + yabove_row[c] - ytop_left);
@ -332,32 +380,25 @@ void vp9_build_intra_predictors_internal(uint8_t *src, int src_stride,
ypred_ptr += y_stride;
}
}
break;
case D45_PRED: {
break;
case D45_PRED:
d45_predictor(ypred_ptr, y_stride, bsize, yabove_row, yleft_col);
}
break;
case D135_PRED: {
break;
case D135_PRED:
d135_predictor(ypred_ptr, y_stride, bsize, yabove_row, yleft_col);
}
break;
case D117_PRED: {
break;
case D117_PRED:
d117_predictor(ypred_ptr, y_stride, bsize, yabove_row, yleft_col);
}
break;
case D153_PRED: {
break;
case D153_PRED:
d153_predictor(ypred_ptr, y_stride, bsize, yabove_row, yleft_col);
}
break;
case D27_PRED: {
break;
case D27_PRED:
d27_predictor(ypred_ptr, y_stride, bsize, yabove_row, yleft_col);
}
break;
case D63_PRED: {
break;
case D63_PRED:
d63_predictor(ypred_ptr, y_stride, bsize, yabove_row, yleft_col);
}
break;
break;
case I8X8_PRED:
case B_PRED:
case NEARESTMV:
@ -383,155 +424,28 @@ static void combine_interintra(MB_PREDICTION_MODE mode,
static const int scale_max = 256; // 1 << scale_bits;
static const int scale_round = 127; // (1 << (scale_bits - 1));
// This table is a function A + B*exp(-kx), where x is hor. index
static const int weights1d[32] = {
128, 122, 116, 111, 107, 103, 99, 96,
93, 90, 88, 85, 83, 81, 80, 78,
77, 76, 75, 74, 73, 72, 71, 70,
70, 69, 69, 68, 68, 68, 67, 67,
static const int weights1d[64] = {
128, 125, 122, 119, 116, 114, 111, 109,
107, 105, 103, 101, 99, 97, 96, 94,
93, 91, 90, 89, 88, 86, 85, 84,
83, 82, 81, 81, 80, 79, 78, 78,
77, 76, 76, 75, 75, 74, 74, 73,
73, 72, 72, 71, 71, 71, 70, 70,
70, 70, 69, 69, 69, 69, 68, 68,
68, 68, 68, 67, 67, 67, 67, 67,
};
// This table is a function A + B*exp(-k.sqrt(xy)), where x, y are
// hor. and vert. indices
static const int weights2d[1024] = {
128, 128, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128,
128, 122, 120, 118, 116, 115, 114, 113,
112, 111, 111, 110, 109, 109, 108, 107,
107, 106, 106, 105, 105, 104, 104, 104,
103, 103, 102, 102, 102, 101, 101, 101,
128, 120, 116, 114, 112, 111, 109, 108,
107, 106, 105, 104, 103, 102, 102, 101,
100, 100, 99, 99, 98, 97, 97, 96,
96, 96, 95, 95, 94, 94, 93, 93,
128, 118, 114, 111, 109, 107, 106, 104,
103, 102, 101, 100, 99, 98, 97, 97,
96, 95, 95, 94, 93, 93, 92, 92,
91, 91, 90, 90, 90, 89, 89, 88,
128, 116, 112, 109, 107, 105, 103, 102,
100, 99, 98, 97, 96, 95, 94, 93,
93, 92, 91, 91, 90, 90, 89, 89,
88, 88, 87, 87, 86, 86, 85, 85,
128, 115, 111, 107, 105, 103, 101, 99,
98, 97, 96, 94, 93, 93, 92, 91,
90, 89, 89, 88, 88, 87, 86, 86,
85, 85, 84, 84, 84, 83, 83, 82,
128, 114, 109, 106, 103, 101, 99, 97,
96, 95, 93, 92, 91, 90, 90, 89,
88, 87, 87, 86, 85, 85, 84, 84,
83, 83, 82, 82, 82, 81, 81, 80,
128, 113, 108, 104, 102, 99, 97, 96,
94, 93, 92, 91, 90, 89, 88, 87,
86, 85, 85, 84, 84, 83, 83, 82,
82, 81, 81, 80, 80, 79, 79, 79,
128, 112, 107, 103, 100, 98, 96, 94,
93, 91, 90, 89, 88, 87, 86, 85,
85, 84, 83, 83, 82, 82, 81, 80,
80, 80, 79, 79, 78, 78, 78, 77,
128, 111, 106, 102, 99, 97, 95, 93,
91, 90, 89, 88, 87, 86, 85, 84,
83, 83, 82, 81, 81, 80, 80, 79,
79, 78, 78, 77, 77, 77, 76, 76,
128, 111, 105, 101, 98, 96, 93, 92,
90, 89, 88, 86, 85, 84, 84, 83,
82, 81, 81, 80, 80, 79, 79, 78,
78, 77, 77, 76, 76, 76, 75, 75,
128, 110, 104, 100, 97, 94, 92, 91,
89, 88, 86, 85, 84, 83, 83, 82,
81, 80, 80, 79, 79, 78, 78, 77,
77, 76, 76, 75, 75, 75, 74, 74,
128, 109, 103, 99, 96, 93, 91, 90,
88, 87, 85, 84, 83, 82, 82, 81,
80, 79, 79, 78, 78, 77, 77, 76,
76, 75, 75, 75, 74, 74, 74, 73,
128, 109, 102, 98, 95, 93, 90, 89,
87, 86, 84, 83, 82, 81, 81, 80,
79, 78, 78, 77, 77, 76, 76, 75,
75, 75, 74, 74, 73, 73, 73, 73,
128, 108, 102, 97, 94, 92, 90, 88,
86, 85, 84, 83, 82, 81, 80, 79,
78, 78, 77, 77, 76, 76, 75, 75,
74, 74, 73, 73, 73, 73, 72, 72,
128, 107, 101, 97, 93, 91, 89, 87,
85, 84, 83, 82, 81, 80, 79, 78,
78, 77, 76, 76, 75, 75, 74, 74,
74, 73, 73, 73, 72, 72, 72, 71,
128, 107, 100, 96, 93, 90, 88, 86,
85, 83, 82, 81, 80, 79, 78, 78,
77, 76, 76, 75, 75, 74, 74, 73,
73, 73, 72, 72, 72, 71, 71, 71,
128, 106, 100, 95, 92, 89, 87, 85,
84, 83, 81, 80, 79, 78, 78, 77,
76, 76, 75, 75, 74, 74, 73, 73,
72, 72, 72, 72, 71, 71, 71, 70,
128, 106, 99, 95, 91, 89, 87, 85,
83, 82, 81, 80, 79, 78, 77, 76,
76, 75, 75, 74, 74, 73, 73, 72,
72, 72, 71, 71, 71, 71, 70, 70,
128, 105, 99, 94, 91, 88, 86, 84,
83, 81, 80, 79, 78, 77, 77, 76,
75, 75, 74, 74, 73, 73, 72, 72,
72, 71, 71, 71, 70, 70, 70, 70,
128, 105, 98, 93, 90, 88, 85, 84,
82, 81, 80, 79, 78, 77, 76, 75,
75, 74, 74, 73, 73, 72, 72, 71,
71, 71, 71, 70, 70, 70, 70, 69,
128, 104, 97, 93, 90, 87, 85, 83,
82, 80, 79, 78, 77, 76, 76, 75,
74, 74, 73, 73, 72, 72, 71, 71,
71, 70, 70, 70, 70, 69, 69, 69,
128, 104, 97, 92, 89, 86, 84, 83,
81, 80, 79, 78, 77, 76, 75, 74,
74, 73, 73, 72, 72, 71, 71, 71,
70, 70, 70, 70, 69, 69, 69, 69,
128, 104, 96, 92, 89, 86, 84, 82,
80, 79, 78, 77, 76, 75, 75, 74,
73, 73, 72, 72, 71, 71, 71, 70,
70, 70, 70, 69, 69, 69, 69, 68,
128, 103, 96, 91, 88, 85, 83, 82,
80, 79, 78, 77, 76, 75, 74, 74,
73, 72, 72, 72, 71, 71, 70, 70,
70, 70, 69, 69, 69, 69, 68, 68,
128, 103, 96, 91, 88, 85, 83, 81,
80, 78, 77, 76, 75, 75, 74, 73,
73, 72, 72, 71, 71, 70, 70, 70,
70, 69, 69, 69, 69, 68, 68, 68,
128, 102, 95, 90, 87, 84, 82, 81,
79, 78, 77, 76, 75, 74, 73, 73,
72, 72, 71, 71, 71, 70, 70, 70,
69, 69, 69, 69, 68, 68, 68, 68,
128, 102, 95, 90, 87, 84, 82, 80,
79, 77, 76, 75, 75, 74, 73, 73,
72, 72, 71, 71, 70, 70, 70, 69,
69, 69, 69, 68, 68, 68, 68, 68,
128, 102, 94, 90, 86, 84, 82, 80,
78, 77, 76, 75, 74, 73, 73, 72,
72, 71, 71, 70, 70, 70, 69, 69,
69, 69, 68, 68, 68, 68, 68, 67,
128, 101, 94, 89, 86, 83, 81, 79,
78, 77, 76, 75, 74, 73, 73, 72,
71, 71, 71, 70, 70, 69, 69, 69,
69, 68, 68, 68, 68, 68, 67, 67,
128, 101, 93, 89, 85, 83, 81, 79,
78, 76, 75, 74, 74, 73, 72, 72,
71, 71, 70, 70, 70, 69, 69, 69,
68, 68, 68, 68, 68, 67, 67, 67,
128, 101, 93, 88, 85, 82, 80, 79,
77, 76, 75, 74, 73, 73, 72, 71,
71, 70, 70, 70, 69, 69, 69, 68,
68, 68, 68, 68, 67, 67, 67, 67,
};
int size_scale = (size >= 32 ? 1 :
size == 16 ? 2 :
size == 8 ? 4 : 8);
int size_shift = size == 64 ? 1 : 0;
int size_scale = (size >= 64 ? 1:
size == 32 ? 2 :
size == 16 ? 4 :
size == 8 ? 8 : 16);
int i, j;
switch (mode) {
case V_PRED:
for (i = 0; i < size; ++i) {
for (j = 0; j < size; ++j) {
int k = i * interstride + j;
int scale = weights1d[i * size_scale >> size_shift];
int scale = weights1d[i * size_scale];
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -544,7 +458,7 @@ static void combine_interintra(MB_PREDICTION_MODE mode,
for (i = 0; i < size; ++i) {
for (j = 0; j < size; ++j) {
int k = i * interstride + j;
int scale = weights1d[j * size_scale >> size_shift];
int scale = weights1d[j * size_scale];
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -558,9 +472,8 @@ static void combine_interintra(MB_PREDICTION_MODE mode,
for (i = 0; i < size; ++i) {
for (j = 0; j < size; ++j) {
int k = i * interstride + j;
int scale = (weights2d[(i * size_scale * 32 +
j * size_scale) >> size_shift] +
weights1d[i * size_scale >> size_shift]) >> 1;
int scale = (weights1d[i * size_scale] * 3 +
weights1d[j * size_scale]) >> 2;
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -574,9 +487,8 @@ static void combine_interintra(MB_PREDICTION_MODE mode,
for (i = 0; i < size; ++i) {
for (j = 0; j < size; ++j) {
int k = i * interstride + j;
int scale = (weights2d[(i * size_scale * 32 +
j * size_scale) >> size_shift] +
weights1d[j * size_scale >> size_shift]) >> 1;
int scale = (weights1d[j * size_scale] * 3 +
weights1d[i * size_scale]) >> 2;
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -589,8 +501,7 @@ static void combine_interintra(MB_PREDICTION_MODE mode,
for (i = 0; i < size; ++i) {
for (j = 0; j < size; ++j) {
int k = i * interstride + j;
int scale = weights2d[(i * size_scale * 32 +
j * size_scale) >> size_shift];
int scale = weights1d[(i < j ? i : j) * size_scale];
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -600,8 +511,21 @@ static void combine_interintra(MB_PREDICTION_MODE mode,
break;
case D45_PRED:
case DC_PRED:
for (i = 0; i < size; ++i) {
for (j = 0; j < size; ++j) {
int k = i * interstride + j;
int scale = (weights1d[i * size_scale] +
weights1d[j * size_scale]) >> 1;
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
>> scale_bits;
}
}
break;
case TM_PRED:
case DC_PRED:
default:
// simple average
for (i = 0; i < size; ++i) {
@ -631,7 +555,7 @@ void vp9_build_interintra_16x16_predictors_mby(MACROBLOCKD *xd,
xd->dst.y_buffer, xd->dst.y_stride,
intrapredictor, 16,
xd->mode_info_context->mbmi.interintra_mode, 16,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available, xd->right_available);
combine_interintra(xd->mode_info_context->mbmi.interintra_mode,
ypred, ystride, intrapredictor, 16, 16);
}
@ -646,12 +570,12 @@ void vp9_build_interintra_16x16_predictors_mbuv(MACROBLOCKD *xd,
xd->dst.u_buffer, xd->dst.uv_stride,
uintrapredictor, 8,
xd->mode_info_context->mbmi.interintra_uv_mode, 8,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available, xd->right_available);
vp9_build_intra_predictors_internal(
xd->dst.v_buffer, xd->dst.uv_stride,
vintrapredictor, 8,
xd->mode_info_context->mbmi.interintra_uv_mode, 8,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available, xd->right_available);
combine_interintra(xd->mode_info_context->mbmi.interintra_uv_mode,
upred, uvstride, uintrapredictor, 8, 8);
combine_interintra(xd->mode_info_context->mbmi.interintra_uv_mode,
@ -666,7 +590,7 @@ void vp9_build_interintra_32x32_predictors_sby(MACROBLOCKD *xd,
xd->dst.y_buffer, xd->dst.y_stride,
intrapredictor, 32,
xd->mode_info_context->mbmi.interintra_mode, 32,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available, xd->right_available);
combine_interintra(xd->mode_info_context->mbmi.interintra_mode,
ypred, ystride, intrapredictor, 32, 32);
}
@ -681,12 +605,12 @@ void vp9_build_interintra_32x32_predictors_sbuv(MACROBLOCKD *xd,
xd->dst.u_buffer, xd->dst.uv_stride,
uintrapredictor, 16,
xd->mode_info_context->mbmi.interintra_uv_mode, 16,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available, xd->right_available);
vp9_build_intra_predictors_internal(
xd->dst.v_buffer, xd->dst.uv_stride,
vintrapredictor, 16,
xd->mode_info_context->mbmi.interintra_uv_mode, 16,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available, xd->right_available);
combine_interintra(xd->mode_info_context->mbmi.interintra_uv_mode,
upred, uvstride, uintrapredictor, 16, 16);
combine_interintra(xd->mode_info_context->mbmi.interintra_uv_mode,
@ -710,7 +634,8 @@ void vp9_build_interintra_64x64_predictors_sby(MACROBLOCKD *xd,
const int mode = xd->mode_info_context->mbmi.interintra_mode;
vp9_build_intra_predictors_internal(xd->dst.y_buffer, xd->dst.y_stride,
intrapredictor, 64, mode, 64,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
combine_interintra(xd->mode_info_context->mbmi.interintra_mode,
ypred, ystride, intrapredictor, 64, 64);
}
@ -724,10 +649,12 @@ void vp9_build_interintra_64x64_predictors_sbuv(MACROBLOCKD *xd,
const int mode = xd->mode_info_context->mbmi.interintra_uv_mode;
vp9_build_intra_predictors_internal(xd->dst.u_buffer, xd->dst.uv_stride,
uintrapredictor, 32, mode, 32,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
vp9_build_intra_predictors_internal(xd->dst.v_buffer, xd->dst.uv_stride,
vintrapredictor, 32, mode, 32,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
combine_interintra(xd->mode_info_context->mbmi.interintra_uv_mode,
upred, uvstride, uintrapredictor, 32, 32);
combine_interintra(xd->mode_info_context->mbmi.interintra_uv_mode,
@ -749,28 +676,32 @@ void vp9_build_intra_predictors_mby(MACROBLOCKD *xd) {
vp9_build_intra_predictors_internal(xd->dst.y_buffer, xd->dst.y_stride,
xd->predictor, 16,
xd->mode_info_context->mbmi.mode, 16,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
}
void vp9_build_intra_predictors_mby_s(MACROBLOCKD *xd) {
vp9_build_intra_predictors_internal(xd->dst.y_buffer, xd->dst.y_stride,
xd->dst.y_buffer, xd->dst.y_stride,
xd->mode_info_context->mbmi.mode, 16,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
}
void vp9_build_intra_predictors_sby_s(MACROBLOCKD *xd) {
vp9_build_intra_predictors_internal(xd->dst.y_buffer, xd->dst.y_stride,
xd->dst.y_buffer, xd->dst.y_stride,
xd->mode_info_context->mbmi.mode, 32,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
}
void vp9_build_intra_predictors_sb64y_s(MACROBLOCKD *xd) {
vp9_build_intra_predictors_internal(xd->dst.y_buffer, xd->dst.y_stride,
xd->dst.y_buffer, xd->dst.y_stride,
xd->mode_info_context->mbmi.mode, 64,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
}
void vp9_build_intra_predictors_mbuv_internal(MACROBLOCKD *xd,
@ -780,10 +711,12 @@ void vp9_build_intra_predictors_mbuv_internal(MACROBLOCKD *xd,
int mode, int bsize) {
vp9_build_intra_predictors_internal(xd->dst.u_buffer, xd->dst.uv_stride,
upred_ptr, uv_stride, mode, bsize,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
vp9_build_intra_predictors_internal(xd->dst.v_buffer, xd->dst.uv_stride,
vpred_ptr, uv_stride, mode, bsize,
xd->up_available, xd->left_available);
xd->up_available, xd->left_available,
xd->right_available);
}
void vp9_build_intra_predictors_mbuv(MACROBLOCKD *xd) {
@ -815,20 +748,35 @@ void vp9_build_intra_predictors_sb64uv_s(MACROBLOCKD *xd) {
32);
}
void vp9_intra8x8_predict(BLOCKD *xd,
void vp9_intra8x8_predict(MACROBLOCKD *xd,
BLOCKD *b,
int mode,
uint8_t *predictor) {
vp9_build_intra_predictors_internal(*(xd->base_dst) + xd->dst,
xd->dst_stride, predictor, 16,
mode, 8, 1, 1);
const int block4x4_idx = (b - xd->block);
const int block_idx = (block4x4_idx >> 2) | !!(block4x4_idx & 2);
const int have_top = (block_idx >> 1) || xd->up_available;
const int have_left = (block_idx & 1) || xd->left_available;
const int have_right = !(block_idx & 1) || xd->right_available;
vp9_build_intra_predictors_internal(*(b->base_dst) + b->dst,
b->dst_stride, predictor, 16,
mode, 8, have_top, have_left,
have_right);
}
void vp9_intra_uv4x4_predict(BLOCKD *xd,
void vp9_intra_uv4x4_predict(MACROBLOCKD *xd,
BLOCKD *b,
int mode,
uint8_t *predictor) {
vp9_build_intra_predictors_internal(*(xd->base_dst) + xd->dst,
xd->dst_stride, predictor, 8,
mode, 4, 1, 1);
const int block_idx = (b - xd->block) & 3;
const int have_top = (block_idx >> 1) || xd->up_available;
const int have_left = (block_idx & 1) || xd->left_available;
const int have_right = !(block_idx & 1) || xd->right_available;
vp9_build_intra_predictors_internal(*(b->base_dst) + b->dst,
b->dst_stride, predictor, 8,
mode, 4, have_top, have_left,
have_right);
}
/* TODO: try different ways of use Y-UV mode correlation

65
vp9/common/vp9_reconintra.h
View File

@ -14,37 +14,44 @@
#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_blockd.h"
extern void vp9_recon_intra_mbuv(MACROBLOCKD *xd);
extern B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
int stride, int n);
extern B_PREDICTION_MODE vp9_find_bpred_context(BLOCKD *x);
void vp9_recon_intra_mbuv(MACROBLOCKD *xd);
B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
int stride, int n,
int tx, int ty);
B_PREDICTION_MODE vp9_find_bpred_context(MACROBLOCKD *xd, BLOCKD *x);
#if CONFIG_COMP_INTERINTRA_PRED
extern void vp9_build_interintra_16x16_predictors_mb(MACROBLOCKD *xd,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride);
extern void vp9_build_interintra_16x16_predictors_mby(MACROBLOCKD *xd,
uint8_t *ypred,
int ystride);
extern void vp9_build_interintra_16x16_predictors_mbuv(MACROBLOCKD *xd,
uint8_t *upred,
uint8_t *vpred,
int uvstride);
void vp9_build_interintra_16x16_predictors_mb(MACROBLOCKD *xd,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride);
void vp9_build_interintra_16x16_predictors_mby(MACROBLOCKD *xd,
uint8_t *ypred,
int ystride);
void vp9_build_interintra_16x16_predictors_mbuv(MACROBLOCKD *xd,
uint8_t *upred,
uint8_t *vpred,
int uvstride);
#endif // CONFIG_COMP_INTERINTRA_PRED
extern void vp9_build_interintra_32x32_predictors_sb(MACROBLOCKD *xd,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride);
extern void vp9_build_interintra_64x64_predictors_sb(MACROBLOCKD *xd,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride);
void vp9_build_interintra_32x32_predictors_sb(MACROBLOCKD *xd,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride);
void vp9_build_interintra_64x64_predictors_sb(MACROBLOCKD *xd,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride);
#endif // VP9_COMMON_VP9_RECONINTRA_H_

196
vp9/common/vp9_reconintra4x4.c
View File

@ -15,17 +15,17 @@
#include "vp9_rtcd.h"
#if CONFIG_NEWBINTRAMODES
static int find_grad_measure(uint8_t *x, int stride, int n, int t,
static int find_grad_measure(uint8_t *x, int stride, int n, int tx, int ty,
int dx, int dy) {
int i, j;
int count = 0, gsum = 0, gdiv;
/* TODO: Make this code more efficient by breaking up into two loops */
for (i = -t; i < n; ++i)
for (j = -t; j < n; ++j) {
for (i = -ty; i < n; ++i)
for (j = -tx; j < n; ++j) {
int g;
if (i >= 0 && j >= 0) continue;
if (i + dy >= 0 && j + dx >= 0) continue;
if (i + dy < -t || i + dy >= n || j + dx < -t || j + dx >= n) continue;
if (i + dy < -ty || i + dy >= n || j + dx < -tx || j + dx >= n) continue;
g = abs(x[(i + dy) * stride + j + dx] - x[i * stride + j]);
gsum += g * g;
count++;
@ -36,14 +36,15 @@ static int find_grad_measure(uint8_t *x, int stride, int n, int t,
#if CONTEXT_PRED_REPLACEMENTS == 6
B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
int stride, int n) {
int stride, int n,
int tx, int ty) {
int g[8], i, imin, imax;
g[1] = find_grad_measure(ptr, stride, n, 4, 2, 1);
g[2] = find_grad_measure(ptr, stride, n, 4, 1, 1);
g[3] = find_grad_measure(ptr, stride, n, 4, 1, 2);
g[5] = find_grad_measure(ptr, stride, n, 4, -1, 2);
g[6] = find_grad_measure(ptr, stride, n, 4, -1, 1);
g[7] = find_grad_measure(ptr, stride, n, 4, -2, 1);
g[1] = find_grad_measure(ptr, stride, n, tx, ty, 2, 1);
g[2] = find_grad_measure(ptr, stride, n, tx, ty, 1, 1);
g[3] = find_grad_measure(ptr, stride, n, tx, ty, 1, 2);
g[5] = find_grad_measure(ptr, stride, n, tx, ty, -1, 2);
g[6] = find_grad_measure(ptr, stride, n, tx, ty, -1, 1);
g[7] = find_grad_measure(ptr, stride, n, tx, ty, -2, 1);
imin = 1;
for (i = 2; i < 8; i += 1 + (i == 3))
imin = (g[i] < g[imin] ? i : imin);
@ -73,12 +74,13 @@ B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
}
#elif CONTEXT_PRED_REPLACEMENTS == 4
B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
int stride, int n) {
int stride, int n,
int tx, int ty) {
int g[8], i, imin, imax;
g[1] = find_grad_measure(ptr, stride, n, 4, 2, 1);
g[3] = find_grad_measure(ptr, stride, n, 4, 1, 2);
g[5] = find_grad_measure(ptr, stride, n, 4, -1, 2);
g[7] = find_grad_measure(ptr, stride, n, 4, -2, 1);
g[1] = find_grad_measure(ptr, stride, n, tx, ty, 2, 1);
g[3] = find_grad_measure(ptr, stride, n, tx, ty, 1, 2);
g[5] = find_grad_measure(ptr, stride, n, tx, ty, -1, 2);
g[7] = find_grad_measure(ptr, stride, n, tx, ty, -2, 1);
imin = 1;
for (i = 3; i < 8; i+=2)
imin = (g[i] < g[imin] ? i : imin);
@ -104,16 +106,17 @@ B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
}
#elif CONTEXT_PRED_REPLACEMENTS == 0
B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
int stride, int n) {
int stride, int n,
int tx, int ty) {
int g[8], i, imin, imax;
g[0] = find_grad_measure(ptr, stride, n, 4, 1, 0);
g[1] = find_grad_measure(ptr, stride, n, 4, 2, 1);
g[2] = find_grad_measure(ptr, stride, n, 4, 1, 1);
g[3] = find_grad_measure(ptr, stride, n, 4, 1, 2);
g[4] = find_grad_measure(ptr, stride, n, 4, 0, 1);
g[5] = find_grad_measure(ptr, stride, n, 4, -1, 2);
g[6] = find_grad_measure(ptr, stride, n, 4, -1, 1);
g[7] = find_grad_measure(ptr, stride, n, 4, -2, 1);
g[0] = find_grad_measure(ptr, stride, n, tx, ty, 1, 0);
g[1] = find_grad_measure(ptr, stride, n, tx, ty, 2, 1);
g[2] = find_grad_measure(ptr, stride, n, tx, ty, 1, 1);
g[3] = find_grad_measure(ptr, stride, n, tx, ty, 1, 2);
g[4] = find_grad_measure(ptr, stride, n, tx, ty, 0, 1);
g[5] = find_grad_measure(ptr, stride, n, tx, ty, -1, 2);
g[6] = find_grad_measure(ptr, stride, n, tx, ty, -1, 1);
g[7] = find_grad_measure(ptr, stride, n, tx, ty, -2, 1);
imax = 0;
for (i = 1; i < 8; i++)
imax = (g[i] > g[imax] ? i : imax);
@ -144,26 +147,113 @@ B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
}
#endif
B_PREDICTION_MODE vp9_find_bpred_context(BLOCKD *x) {
B_PREDICTION_MODE vp9_find_bpred_context(MACROBLOCKD *xd, BLOCKD *x) {
const int block_idx = x - xd->block;
const int have_top = (block_idx >> 2) || xd->up_available;
const int have_left = (block_idx & 3) || xd->left_available;
uint8_t *ptr = *(x->base_dst) + x->dst;
int stride = x->dst_stride;
return vp9_find_dominant_direction(ptr, stride, 4);
int tx = have_left ? 4 : 0;
int ty = have_top ? 4 : 0;
if (!have_left && !have_top)
return B_DC_PRED;
return vp9_find_dominant_direction(ptr, stride, 4, tx, ty);
}
#endif
void vp9_intra4x4_predict(BLOCKD *x,
void vp9_intra4x4_predict(MACROBLOCKD *xd,
BLOCKD *x,
int b_mode,
uint8_t *predictor) {
int i, r, c;
const int block_idx = x - xd->block;
const int have_top = (block_idx >> 2) || xd->up_available;
const int have_left = (block_idx & 3) || xd->left_available;
const int have_right = (block_idx & 3) != 3 || xd->right_available;
uint8_t left[4], above[8], top_left;
/*
* 127 127 127 .. 127 127 127 127 127 127
* 129 A B .. Y Z
* 129 C D .. W X
* 129 E F .. U V
* 129 G H .. S T T T T T
* ..
*/
uint8_t *above = *(x->base_dst) + x->dst - x->dst_stride;
uint8_t left[4];
uint8_t top_left = above[-1];
if (have_left) {
uint8_t *left_ptr = *(x->base_dst) + x->dst - 1;
const int stride = x->dst_stride;
left[0] = (*(x->base_dst))[x->dst - 1];
left[1] = (*(x->base_dst))[x->dst - 1 + x->dst_stride];
left[2] = (*(x->base_dst))[x->dst - 1 + 2 * x->dst_stride];
left[3] = (*(x->base_dst))[x->dst - 1 + 3 * x->dst_stride];
left[0] = left_ptr[0 * stride];
left[1] = left_ptr[1 * stride];
left[2] = left_ptr[2 * stride];
left[3] = left_ptr[3 * stride];
} else {
left[0] = left[1] = left[2] = left[3] = 129;
}
if (have_top) {
uint8_t *above_ptr = *(x->base_dst) + x->dst - x->dst_stride;
if (have_left) {
top_left = above_ptr[-1];
} else {
top_left = 127;
}
above[0] = above_ptr[0];
above[1] = above_ptr[1];
above[2] = above_ptr[2];
above[3] = above_ptr[3];
if (((block_idx & 3) != 3) ||
(have_right && block_idx == 3 &&
((xd->mb_index != 3 && xd->sb_index != 3) ||
((xd->mb_index & 1) == 0 && xd->sb_index == 3)))) {
above[4] = above_ptr[4];
above[5] = above_ptr[5];
above[6] = above_ptr[6];
above[7] = above_ptr[7];
} else if (have_right) {
uint8_t *above_right = above_ptr + 4;
if (xd->sb_index == 3 && (xd->mb_index & 1))
above_right -= 32 * x->dst_stride;
if (xd->mb_index == 3)
above_right -= 16 * x->dst_stride;
above_right -= (block_idx & ~3) * x->dst_stride;
/* use a more distant above-right (from closest available top-right
* corner), but with a "localized DC" (similar'ish to TM-pred):
*
* A B C D E F G H
* I J K L
* M N O P
* Q R S T
* U V W X x1 x2 x3 x4
*
* Where:
* x1 = clip_pixel(E + X - D)
* x2 = clip_pixel(F + X - D)
* x3 = clip_pixel(G + X - D)
* x4 = clip_pixel(H + X - D)
*
* This is applied anytime when we use a "distant" above-right edge
* that is not immediately top-right to the block that we're going
* to do intra prediction for.
*/
above[4] = clip_pixel(above_right[0] + above_ptr[3] - above_right[-1]);
above[5] = clip_pixel(above_right[1] + above_ptr[3] - above_right[-1]);
above[6] = clip_pixel(above_right[2] + above_ptr[3] - above_right[-1]);
above[7] = clip_pixel(above_right[3] + above_ptr[3] - above_right[-1]);
} else {
// extend edge
above[4] = above[5] = above[6] = above[7] = above[3];
}
} else {
above[0] = above[1] = above[2] = above[3] = 127;
above[4] = above[5] = above[6] = above[7] = 127;
top_left = 127;
}
#if CONFIG_NEWBINTRAMODES
if (b_mode == B_CONTEXT_PRED)
@ -411,39 +501,3 @@ void vp9_intra4x4_predict(BLOCKD *x,
#endif
}
}
/* copy 4 bytes from the above right down so that the 4x4 prediction modes using pixels above and
* to the right prediction have filled in pixels to use.
*/
void vp9_intra_prediction_down_copy(MACROBLOCKD *xd) {
int extend_edge = xd->mb_to_right_edge == 0 && xd->mb_index < 2;
uint8_t *above_right = *(xd->block[0].base_dst) + xd->block[0].dst -
xd->block[0].dst_stride + 16;
uint32_t *dst_ptr0 = (uint32_t *)above_right;
uint32_t *dst_ptr1 =
(uint32_t *)(above_right + 4 * xd->block[0].dst_stride);
uint32_t *dst_ptr2 =
(uint32_t *)(above_right + 8 * xd->block[0].dst_stride);
uint32_t *dst_ptr3 =
(uint32_t *)(above_right + 12 * xd->block[0].dst_stride);
uint32_t *src_ptr = (uint32_t *) above_right;
if ((xd->sb_index >= 2 && xd->mb_to_right_edge == 0) ||
(xd->sb_index == 3 && xd->mb_index & 1))
src_ptr = (uint32_t *) (((uint8_t *) src_ptr) - 32 *
xd->block[0].dst_stride);
if (xd->mb_index == 3 ||
(xd->mb_to_right_edge == 0 && xd->mb_index == 2))
src_ptr = (uint32_t *) (((uint8_t *) src_ptr) - 16 *
xd->block[0].dst_stride);
if (extend_edge) {
*src_ptr = ((uint8_t *) src_ptr)[-1] * 0x01010101U;
}
*dst_ptr0 = *src_ptr;
*dst_ptr1 = *src_ptr;
*dst_ptr2 = *src_ptr;
*dst_ptr3 = *src_ptr;
}

17
vp9/common/vp9_reconintra4x4.h
View File

@ -1,17 +0,0 @@
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_RECONINTRA4X4_H_
#define VP9_COMMON_VP9_RECONINTRA4X4_H_
extern void vp9_intra_prediction_down_copy(MACROBLOCKD *xd);
#endif // VP9_COMMON_VP9_RECONINTRA4X4_H_

5
vp9/common/vp9_rtcd.c
View File

@ -12,10 +12,9 @@
#include "vp9_rtcd.h"
#include "vpx_ports/vpx_once.h"
extern void vpx_scale_rtcd(void);
void vpx_scale_rtcd(void);
void vp9_rtcd()
{
void vp9_rtcd() {
vpx_scale_rtcd();
once(setup_rtcd_internal);
}

415
vp9/common/vp9_rtcd_defs.sh
View File

@ -23,90 +23,50 @@ EOF
}
forward_decls vp9_common_forward_decls
prototype void vp9_filter_block2d_4x4_8 "const uint8_t *src_ptr, const unsigned int src_stride, const int16_t *HFilter_aligned16, const int16_t *VFilter_aligned16, uint8_t *dst_ptr, unsigned int dst_stride"
prototype void vp9_filter_block2d_8x4_8 "const uint8_t *src_ptr, const unsigned int src_stride, const int16_t *HFilter_aligned16, const int16_t *VFilter_aligned16, uint8_t *dst_ptr, unsigned int dst_stride"
prototype void vp9_filter_block2d_8x8_8 "const uint8_t *src_ptr, const unsigned int src_stride, const int16_t *HFilter_aligned16, const int16_t *VFilter_aligned16, uint8_t *dst_ptr, unsigned int dst_stride"
prototype void vp9_filter_block2d_16x16_8 "const uint8_t *src_ptr, const unsigned int src_stride, const int16_t *HFilter_aligned16, const int16_t *VFilter_aligned16, uint8_t *dst_ptr, unsigned int dst_stride"
# At the very least, MSVC 2008 has compiler bug exhibited by this code; code
# compiles warning free but a dissassembly of generated code show bugs. To be
# on the safe side, only enabled when compiled with 'gcc'.
if [ "$CONFIG_GCC" = "yes" ]; then
specialize vp9_filter_block2d_4x4_8 sse4_1 sse2
fi
specialize vp9_filter_block2d_8x4_8 ssse3 #sse4_1 sse2
specialize vp9_filter_block2d_8x8_8 ssse3 #sse4_1 sse2
specialize vp9_filter_block2d_16x16_8 ssse3 #sse4_1 sse2
#
# Dequant
#
prototype void vp9_dequantize_b "struct blockd *x"
specialize vp9_dequantize_b
prototype void vp9_dequantize_b_2x2 "struct blockd *x"
specialize vp9_dequantize_b_2x2
prototype void vp9_dequant_dc_idct_add_y_block_8x8 "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs, const int16_t *dc, struct macroblockd *xd"
specialize vp9_dequant_dc_idct_add_y_block_8x8
prototype void vp9_dequant_idct_add_y_block_8x8 "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs, struct macroblockd *xd"
prototype void vp9_dequant_idct_add_y_block_8x8 "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, struct macroblockd *xd"
specialize vp9_dequant_idct_add_y_block_8x8
prototype void vp9_dequant_idct_add_uv_block_8x8 "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dstu, uint8_t *dstv, int stride, uint16_t *eobs, struct macroblockd *xd"
prototype void vp9_dequant_idct_add_uv_block_8x8 "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dstu, uint8_t *dstv, int stride, struct macroblockd *xd"
specialize vp9_dequant_idct_add_uv_block_8x8
prototype void vp9_dequant_idct_add_16x16 "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int eob"
specialize vp9_dequant_idct_add_16x16
prototype void vp9_dequant_idct_add_8x8 "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int dc, int eob"
prototype void vp9_dequant_idct_add_8x8 "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int eob"
specialize vp9_dequant_idct_add_8x8
prototype void vp9_dequant_idct_add "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride"
prototype void vp9_dequant_idct_add "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int eob"
specialize vp9_dequant_idct_add
prototype void vp9_dequant_dc_idct_add "int16_t *input, const int16_t *dq, uint8_t *pred, uint8_t *dest, int pitch, int stride, int dc"
specialize vp9_dequant_dc_idct_add
prototype void vp9_dequant_dc_idct_add_y_block "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs, const int16_t *dcs"
specialize vp9_dequant_dc_idct_add_y_block
prototype void vp9_dequant_idct_add_y_block "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, uint16_t *eobs"
prototype void vp9_dequant_idct_add_y_block "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int stride, struct macroblockd *xd"
specialize vp9_dequant_idct_add_y_block
prototype void vp9_dequant_idct_add_uv_block "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dstu, uint8_t *dstv, int stride, uint16_t *eobs"
prototype void vp9_dequant_idct_add_uv_block "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dstu, uint8_t *dstv, int stride, struct macroblockd *xd"
specialize vp9_dequant_idct_add_uv_block
prototype void vp9_dequant_idct_add_32x32 "int16_t *q, const int16_t *dq, uint8_t *pre, uint8_t *dst, int pitch, int stride, int eob"
specialize vp9_dequant_idct_add_32x32
prototype void vp9_dequant_idct_add_uv_block_16x16 "int16_t *q, const int16_t *dq, uint8_t *dstu, uint8_t *dstv, int stride, uint16_t *eobs"
prototype void vp9_dequant_idct_add_uv_block_16x16 "int16_t *q, const int16_t *dq, uint8_t *dstu, uint8_t *dstv, int stride, struct macroblockd *xd"
specialize vp9_dequant_idct_add_uv_block_16x16
#
# RECON
#
prototype void vp9_copy_mem16x16 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
prototype void vp9_copy_mem16x16 "const uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
specialize vp9_copy_mem16x16 mmx sse2 dspr2
vp9_copy_mem16x16_dspr2=vp9_copy_mem16x16_dspr2
prototype void vp9_copy_mem8x8 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
prototype void vp9_copy_mem8x8 "const uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
specialize vp9_copy_mem8x8 mmx dspr2
vp9_copy_mem8x8_dspr2=vp9_copy_mem8x8_dspr2
prototype void vp9_copy_mem8x4 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
prototype void vp9_copy_mem8x4 "const uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
specialize vp9_copy_mem8x4 mmx
prototype void vp9_avg_mem16x16 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
specialize vp9_avg_mem16x16
prototype void vp9_avg_mem8x8 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
specialize vp9_avg_mem8x8
prototype void vp9_copy_mem8x4 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
specialize vp9_copy_mem8x4 mmx dspr2
vp9_copy_mem8x4_dspr2=vp9_copy_mem8x4_dspr2
prototype void vp9_recon_b "uint8_t *pred_ptr, int16_t *diff_ptr, uint8_t *dst_ptr, int stride"
specialize vp9_recon_b
@ -137,6 +97,12 @@ specialize vp9_recon_sby_s
prototype void vp9_recon_sbuv_s "struct macroblockd *x, uint8_t *udst, uint8_t *vdst"
specialize void vp9_recon_sbuv_s
prototype void vp9_recon_sb64y_s "struct macroblockd *x, uint8_t *dst"
specialize vp9_recon_sb64y_s
prototype void vp9_recon_sb64uv_s "struct macroblockd *x, uint8_t *udst, uint8_t *vdst"
specialize void vp9_recon_sb64uv_s
prototype void vp9_build_intra_predictors_mby_s "struct macroblockd *x"
specialize vp9_build_intra_predictors_mby_s
@ -164,15 +130,38 @@ specialize vp9_build_intra_predictors_sb64y_s;
prototype void vp9_build_intra_predictors_sb64uv_s "struct macroblockd *x"
specialize vp9_build_intra_predictors_sb64uv_s;
prototype void vp9_intra4x4_predict "struct blockd *x, int b_mode, uint8_t *predictor"
prototype void vp9_intra4x4_predict "struct macroblockd *xd, struct blockd *x, int b_mode, uint8_t *predictor"
specialize vp9_intra4x4_predict;
prototype void vp9_intra8x8_predict "struct blockd *x, int b_mode, uint8_t *predictor"
prototype void vp9_intra8x8_predict "struct macroblockd *xd, struct blockd *x, int b_mode, uint8_t *predictor"
specialize vp9_intra8x8_predict;
prototype void vp9_intra_uv4x4_predict "struct blockd *x, int b_mode, uint8_t *predictor"
prototype void vp9_intra_uv4x4_predict "struct macroblockd *xd, struct blockd *x, int b_mode, uint8_t *predictor"
specialize vp9_intra_uv4x4_predict;
if [ "$CONFIG_VP9_DECODER" = "yes" ]; then
prototype void vp9_add_residual_4x4 "const int16_t *diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride"
specialize vp9_add_residual_4x4 sse2
prototype void vp9_add_residual_8x8 "const int16_t *diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride"
specialize vp9_add_residual_8x8 sse2
prototype void vp9_add_residual_16x16 "const int16_t *diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride"
specialize vp9_add_residual_16x16 sse2
prototype void vp9_add_residual_32x32 "const int16_t *diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride"
specialize vp9_add_residual_32x32 sse2
prototype void vp9_add_constant_residual_8x8 "const int16_t diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride"
specialize vp9_add_constant_residual_8x8 sse2
prototype void vp9_add_constant_residual_16x16 "const int16_t diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride"
specialize vp9_add_constant_residual_16x16 sse2
prototype void vp9_add_constant_residual_32x32 "const int16_t diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride"
specialize vp9_add_constant_residual_32x32 sse2
fi
#
# Loopfilter
#
@ -263,171 +252,146 @@ specialize vp9_sad16x3 sse2
prototype unsigned int vp9_sad3x16 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride"
specialize vp9_sad3x16 sse2
prototype unsigned int vp9_sub_pixel_variance16x2 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
prototype unsigned int vp9_sub_pixel_variance16x2 "const uint8_t *src_ptr, const int source_stride, const int xoffset, const int yoffset, const uint8_t *ref_ptr, const int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_variance16x2 sse2
#
# Sub Pixel Filters
#
prototype void vp9_eighttap_predict16x16 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict16x16
prototype void vp9_convolve8 "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8 ssse3
prototype void vp9_eighttap_predict8x8 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict8x8
prototype void vp9_convolve8_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_horiz ssse3
prototype void vp9_eighttap_predict_avg16x16 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg16x16
prototype void vp9_convolve8_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_vert ssse3
prototype void vp9_eighttap_predict_avg8x8 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg8x8
prototype void vp9_convolve8_avg "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_avg ssse3
prototype void vp9_eighttap_predict_avg4x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg4x4
prototype void vp9_convolve8_avg_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_avg_horiz ssse3
prototype void vp9_eighttap_predict8x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict8x4
prototype void vp9_convolve8_avg_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_avg_vert ssse3
prototype void vp9_eighttap_predict4x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict4x4
#if CONFIG_IMPLICIT_COMPOUNDINTER_WEIGHT
prototype void vp9_convolve8_1by8 "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_1by8
prototype void vp9_eighttap_predict16x16_sharp "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict16x16_sharp
prototype void vp9_convolve8_qtr "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_qtr
prototype void vp9_eighttap_predict8x8_sharp "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict8x8_sharp
prototype void vp9_convolve8_3by8 "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_3by8
prototype void vp9_eighttap_predict_avg16x16_sharp "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg16x16_sharp
prototype void vp9_convolve8_5by8 "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_5by8
prototype void vp9_eighttap_predict_avg8x8_sharp "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg8x8_sharp
prototype void vp9_convolve8_3qtr "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_3qtr
prototype void vp9_eighttap_predict_avg4x4_sharp "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg4x4_sharp
prototype void vp9_convolve8_7by8 "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_7by8
prototype void vp9_eighttap_predict8x4_sharp "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict8x4_sharp
prototype void vp9_convolve8_1by8_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_1by8_horiz
prototype void vp9_eighttap_predict4x4_sharp "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict4x4_sharp
prototype void vp9_convolve8_qtr_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_qtr_horiz
prototype void vp9_eighttap_predict16x16_smooth "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict16x16_smooth
prototype void vp9_convolve8_3by8_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_3by8_horiz
prototype void vp9_eighttap_predict8x8_smooth "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict8x8_smooth
prototype void vp9_convolve8_5by8_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_5by8_horiz
prototype void vp9_eighttap_predict_avg16x16_smooth "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg16x16_smooth
prototype void vp9_convolve8_3qtr_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_3qtr_horiz
prototype void vp9_eighttap_predict_avg8x8_smooth "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg8x8_smooth
prototype void vp9_convolve8_7by8_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_7by8_horiz
prototype void vp9_eighttap_predict_avg4x4_smooth "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_avg4x4_smooth
prototype void vp9_convolve8_1by8_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_1by8_vert
prototype void vp9_eighttap_predict8x4_smooth "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict8x4_smooth
prototype void vp9_convolve8_qtr_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_qtr_vert
prototype void vp9_eighttap_predict4x4_smooth "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict4x4_smooth
prototype void vp9_convolve8_3by8_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_3by8_vert
prototype void vp9_sixtap_predict16x16 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict16x16
prototype void vp9_convolve8_5by8_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_5by8_vert
prototype void vp9_sixtap_predict8x8 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict8x8
prototype void vp9_convolve8_3qtr_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_3qtr_vert
prototype void vp9_sixtap_predict_avg16x16 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict_avg16x16
prototype void vp9_sixtap_predict_avg8x8 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict_avg8x8
prototype void vp9_sixtap_predict8x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict8x4
prototype void vp9_sixtap_predict4x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict4x4
prototype void vp9_sixtap_predict_avg4x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict_avg4x4
prototype void vp9_bilinear_predict16x16 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict16x16 sse2
prototype void vp9_bilinear_predict8x8 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict8x8 sse2
prototype void vp9_bilinear_predict_avg16x16 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict_avg16x16
prototype void vp9_bilinear_predict_avg8x8 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict_avg8x8
prototype void vp9_bilinear_predict8x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict8x4
prototype void vp9_bilinear_predict4x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict4x4
prototype void vp9_bilinear_predict_avg4x4 "uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, uint8_t *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict_avg4x4
prototype void vp9_convolve8_7by8_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
specialize vp9_convolve8_7by8_vert
#endif
#
# dct
#
prototype void vp9_short_idct4x4llm_1 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct4x4llm_1
prototype void vp9_short_idct4x4_1 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct4x4_1
prototype void vp9_short_idct4x4llm "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct4x4llm
prototype void vp9_short_idct4x4 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct4x4 sse2
prototype void vp9_short_idct8x8 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct8x8
specialize vp9_short_idct8x8 sse2
prototype void vp9_short_idct10_8x8 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct10_8x8
specialize vp9_short_idct10_8x8 sse2
prototype void vp9_short_ihaar2x2 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_ihaar2x2
prototype void vp9_short_idct1_8x8 "int16_t *input, int16_t *output"
specialize vp9_short_idct1_8x8
prototype void vp9_short_idct16x16 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct16x16
specialize vp9_short_idct16x16 sse2
prototype void vp9_short_idct10_16x16 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct10_16x16
specialize vp9_short_idct10_16x16 sse2
prototype void vp9_short_idct1_16x16 "int16_t *input, int16_t *output"
specialize vp9_short_idct1_16x16
prototype void vp9_short_idct32x32 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct32x32
specialize vp9_short_idct32x32 sse2
prototype void vp9_ihtllm "const int16_t *input, int16_t *output, int pitch, int tx_type, int tx_dim, int16_t eobs"
specialize vp9_ihtllm
prototype void vp9_short_idct1_32x32 "int16_t *input, int16_t *output"
specialize vp9_short_idct1_32x32
#
# 2nd order
#
prototype void vp9_short_inv_walsh4x4_1 "int16_t *in, int16_t *out"
specialize vp9_short_inv_walsh4x4_1
prototype void vp9_short_idct10_32x32 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct10_32x32
prototype void vp9_short_inv_walsh4x4 "int16_t *in, int16_t *out"
specialize vp9_short_inv_walsh4x4_
prototype void vp9_short_iht8x8 "int16_t *input, int16_t *output, int pitch, int tx_type"
specialize vp9_short_iht8x8
prototype void vp9_short_iht4x4 "int16_t *input, int16_t *output, int pitch, int tx_type"
specialize vp9_short_iht4x4
prototype void vp9_short_iht16x16 "int16_t *input, int16_t *output, int pitch, int tx_type"
specialize vp9_short_iht16x16
prototype void vp9_idct4_1d "int16_t *input, int16_t *output"
specialize vp9_idct4_1d sse2
# dct and add
prototype void vp9_dc_only_idct_add_8x8 "int input_dc, uint8_t *pred_ptr, uint8_t *dst_ptr, int pitch, int stride"
specialize vp9_dc_only_idct_add_8x8
prototype void vp9_dc_only_idct_add "int input_dc, uint8_t *pred_ptr, uint8_t *dst_ptr, int pitch, int stride"
specialize vp9_dc_only_idct_add
specialize vp9_dc_only_idct_add sse2
if [ "$CONFIG_LOSSLESS" = "yes" ]; then
prototype void vp9_short_inv_walsh4x4_1_x8 "int16_t *input, int16_t *output, int pitch"
prototype void vp9_short_inv_walsh4x4_x8 "int16_t *input, int16_t *output, int pitch"
prototype void vp9_short_iwalsh4x4_1 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_iwalsh4x4_1
prototype void vp9_short_iwalsh4x4 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_iwalsh4x4
prototype void vp9_dc_only_inv_walsh_add "int input_dc, uint8_t *pred_ptr, uint8_t *dst_ptr, int pitch, int stride"
prototype void vp9_short_inv_walsh4x4_1_lossless "int16_t *in, int16_t *out"
prototype void vp9_short_inv_walsh4x4_lossless "int16_t *in, int16_t *out"
fi
specialize vp9_dc_only_inv_walsh_add
prototype unsigned int vp9_sad32x3 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, int max_sad"
specialize vp9_sad32x3
@ -475,58 +439,52 @@ specialize vp9_variance4x4 mmx sse2
vp9_variance4x4_sse2=vp9_variance4x4_wmt
vp9_variance4x4_mmx=vp9_variance4x4_mmx
prototype unsigned int vp9_sub_pixel_variance64x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
specialize vp9_sub_pixel_variance64x64
prototype unsigned int vp9_sub_pixel_variance64x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_variance64x64 sse2
prototype unsigned int vp9_sub_pixel_variance32x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
specialize vp9_sub_pixel_variance32x32
prototype unsigned int vp9_sub_pixel_variance32x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_variance32x32 sse2
prototype unsigned int vp9_sub_pixel_variance16x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
prototype unsigned int vp9_sub_pixel_variance16x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_variance16x16 sse2 mmx ssse3
vp9_sub_pixel_variance16x16_sse2=vp9_sub_pixel_variance16x16_wmt
prototype unsigned int vp9_sub_pixel_variance8x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
prototype unsigned int vp9_sub_pixel_variance8x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_variance8x16 sse2 mmx
vp9_sub_pixel_variance8x16_sse2=vp9_sub_pixel_variance8x16_wmt
prototype unsigned int vp9_sub_pixel_variance16x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
prototype unsigned int vp9_sub_pixel_variance16x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_variance16x8 sse2 mmx ssse3
vp9_sub_pixel_variance16x8_sse2=vp9_sub_pixel_variance16x8_ssse3;
vp9_sub_pixel_variance16x8_sse2=vp9_sub_pixel_variance16x8_wmt
prototype unsigned int vp9_sub_pixel_variance8x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
prototype unsigned int vp9_sub_pixel_variance8x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_variance8x8 sse2 mmx
vp9_sub_pixel_variance8x8_sse2=vp9_sub_pixel_variance8x8_wmt
prototype unsigned int vp9_sub_pixel_variance4x4 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
prototype unsigned int vp9_sub_pixel_variance4x4 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_variance4x4 sse2 mmx
vp9_sub_pixel_variance4x4_sse2=vp9_sub_pixel_variance4x4_wmt
prototype unsigned int vp9_sad64x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
specialize vp9_sad64x64
specialize vp9_sad64x64 sse2
prototype unsigned int vp9_sad32x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
specialize vp9_sad32x32
specialize vp9_sad32x32 sse2
prototype unsigned int vp9_sad16x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
specialize vp9_sad16x16 mmx sse2 sse3
vp9_sad16x16_sse2=vp9_sad16x16_wmt
specialize vp9_sad16x16 mmx sse2
prototype unsigned int vp9_sad16x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
specialize vp9_sad16x8 mmx sse2
vp9_sad16x8_sse2=vp9_sad16x8_wmt
prototype unsigned int vp9_sad8x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
specialize vp9_sad8x16 mmx sse2
vp9_sad8x16_sse2=vp9_sad8x16_wmt
prototype unsigned int vp9_sad8x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
specialize vp9_sad8x8 mmx sse2
vp9_sad8x8_sse2=vp9_sad8x8_wmt
prototype unsigned int vp9_sad4x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
specialize vp9_sad4x4 mmx sse2
vp9_sad4x4_sse2=vp9_sad4x4_wmt
specialize vp9_sad4x4 mmx sse
prototype unsigned int vp9_variance_halfpixvar16x16_h "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar16x16_h mmx sse2
@ -579,76 +537,64 @@ specialize vp9_sad8x8x3 sse3
prototype void vp9_sad4x4x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad4x4x3 sse3
prototype void vp9_sad64x64x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
prototype void vp9_sad64x64x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
specialize vp9_sad64x64x8
prototype void vp9_sad32x32x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
prototype void vp9_sad32x32x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
specialize vp9_sad32x32x8
prototype void vp9_sad16x16x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
prototype void vp9_sad16x16x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
specialize vp9_sad16x16x8 sse4
prototype void vp9_sad16x8x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
prototype void vp9_sad16x8x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
specialize vp9_sad16x8x8 sse4
prototype void vp9_sad8x16x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
prototype void vp9_sad8x16x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
specialize vp9_sad8x16x8 sse4
prototype void vp9_sad8x8x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
prototype void vp9_sad8x8x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
specialize vp9_sad8x8x8 sse4
prototype void vp9_sad4x4x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
prototype void vp9_sad4x4x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
specialize vp9_sad4x4x8 sse4
prototype void vp9_sad64x64x4d "const uint8_t *src_ptr, int src_stride, const uint8_t **ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad64x64x4d
prototype void vp9_sad64x64x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
specialize vp9_sad64x64x4d sse2
prototype void vp9_sad32x32x4d "const uint8_t *src_ptr, int src_stride, const uint8_t **ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad32x32x4d
prototype void vp9_sad32x32x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
specialize vp9_sad32x32x4d sse2
prototype void vp9_sad16x16x4d "const uint8_t *src_ptr, int src_stride, const uint8_t **ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad16x16x4d sse3
prototype void vp9_sad16x16x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
specialize vp9_sad16x16x4d sse2
prototype void vp9_sad16x8x4d "const uint8_t *src_ptr, int src_stride, const uint8_t **ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad16x8x4d sse3
prototype void vp9_sad16x8x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
specialize vp9_sad16x8x4d sse2
prototype void vp9_sad8x16x4d "const uint8_t *src_ptr, int src_stride, const uint8_t **ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad8x16x4d sse3
prototype void vp9_sad8x16x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
specialize vp9_sad8x16x4d sse2
prototype void vp9_sad8x8x4d "const uint8_t *src_ptr, int src_stride, const uint8_t **ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad8x8x4d sse3
prototype void vp9_sad4x4x4d "const uint8_t *src_ptr, int src_stride, const uint8_t **ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad4x4x4d sse3
#
# Block copy
#
case $arch in
x86*)
prototype void vp9_copy32xn "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, int n"
specialize vp9_copy32xn sse2 sse3
;;
esac
prototype void vp9_sad8x8x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
specialize vp9_sad8x8x4d sse2
prototype void vp9_sad4x4x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
specialize vp9_sad4x4x4d sse
prototype unsigned int vp9_sub_pixel_mse16x16 "const uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, const uint8_t *dst_ptr, int dst_pixels_per_line, unsigned int *sse"
specialize vp9_sub_pixel_mse16x16 sse2 mmx
vp9_sub_pixel_mse16x16_sse2=vp9_sub_pixel_mse16x16_wmt
prototype unsigned int vp9_mse16x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse"
specialize vp9_mse16x16 mmx sse2
vp9_mse16x16_sse2=vp9_mse16x16_wmt
prototype unsigned int vp9_sub_pixel_mse64x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
prototype unsigned int vp9_sub_pixel_mse64x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_mse64x64
prototype unsigned int vp9_sub_pixel_mse32x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int Refstride, unsigned int *sse"
prototype unsigned int vp9_sub_pixel_mse32x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_sub_pixel_mse32x32
prototype unsigned int vp9_get_mb_ss "const int16_t *"
specialize vp9_get_mb_ss mmx sse2
# ENCODEMB INVOKE
prototype int vp9_mbblock_error "struct macroblock *mb, int dc"
prototype int vp9_mbblock_error "struct macroblock *mb"
specialize vp9_mbblock_error mmx sse2
vp9_mbblock_error_sse2=vp9_mbblock_error_xmm
@ -686,14 +632,17 @@ if [ "$CONFIG_INTERNAL_STATS" = "yes" ]; then
fi
# fdct functions
prototype void vp9_fht "const int16_t *input, int pitch, int16_t *output, int tx_type, int tx_dim"
specialize vp9_fht
prototype void vp9_short_fht4x4 "int16_t *InputData, int16_t *OutputData, int pitch, int tx_type"
specialize vp9_short_fht4x4
prototype void vp9_short_fht8x8 "int16_t *InputData, int16_t *OutputData, int pitch, int tx_type"
specialize vp9_short_fht8x8
prototype void vp9_short_fht16x16 "int16_t *InputData, int16_t *OutputData, int pitch, int tx_type"
specialize vp9_short_fht16x16
prototype void vp9_short_fdct8x8 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct8x8
prototype void vp9_short_fhaar2x2 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fhaar2x2
specialize vp9_short_fdct8x8 sse2
prototype void vp9_short_fdct4x4 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct4x4
@ -701,23 +650,17 @@ specialize vp9_short_fdct4x4
prototype void vp9_short_fdct8x4 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct8x4
prototype void vp9_short_walsh4x4 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh4x4
prototype void vp9_short_fdct32x32 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct32x32
prototype void vp9_short_fdct16x16 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct16x16
specialize vp9_short_fdct16x16 sse2
prototype void vp9_short_walsh4x4_lossless "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh4x4_lossless
prototype void vp9_short_walsh4x4 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh4x4
prototype void vp9_short_walsh4x4_x8 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh4x4_x8
prototype void vp9_short_walsh8x4_x8 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh8x4_x8
prototype void vp9_short_walsh8x4 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh8x4
#
# Motion search

13
vp9/common/vp9_sadmxn.h
View File

@ -11,14 +11,15 @@
#ifndef VP9_COMMON_VP9_SADMXN_H_
#define VP9_COMMON_VP9_SADMXN_H_
#include "./vpx_config.h"
#include "vpx/vpx_integer.h"
static __inline unsigned int sad_mx_n_c(const uint8_t *src_ptr,
int src_stride,
const uint8_t *ref_ptr,
int ref_stride,
int m,
int n) {
static INLINE unsigned int sad_mx_n_c(const uint8_t *src_ptr,
int src_stride,
const uint8_t *ref_ptr,
int ref_stride,
int m,
int n) {
int r, c;
unsigned int sad = 0;

Some files were not shown because too many files have changed in this diff Show More