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

Merge branch 'experimental' of review:webm/libvpx

Browse Source 
Change-Id: Ib2c2236349c2ae8ee81bd01c5067dddcbac713ca
This commit is contained in:
John Koleszar 2013-01-14 16:25:26 -08:00
commit de5546c372
156 changed files with 14735 additions and 11657 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
build/make/configure.sh
View File

@ -436,10 +436,10 @@ RTCD_OPTIONS = ${RTCD_OPTIONS}
EOF
if enabled rvct; then cat >> $1 << EOF
fmt_deps = sed -e 's;^__image.axf;\$\${@:.d=.o} \$\$@;' #hide
fmt_deps = sed -e 's;^__image.axf;\${@:.d=.o} \$@;' #hide
EOF
else cat >> $1 << EOF
fmt_deps = sed -e 's;^\([a-zA-Z0-9_]*\)\.o;\$\${@:.d=.o} \$\$@;'
fmt_deps = sed -e 's;^\([a-zA-Z0-9_]*\)\.o;\${@:.d=.o} \$@;'
EOF
fi

10
configure vendored
View File

@ -239,15 +239,17 @@ HAVE_LIST="
"
EXPERIMENT_LIST="
csm
comp_intra_pred
superblocks
pred_filter
lossless
subpelrefmv
new_mvref
implicit_segmentation
newbintramodes
comp_interintra_pred
tx64x64
dwtdcthybrid
cnvcontext
newcoefcontext
enable_6tap
abovesprefmv
"
CONFIG_LIST="
external_build

197
test/dct32x32_test.cc Normal file
View File

@ -0,0 +1,197 @@
/*
* 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 <math.h>
#include <stdlib.h>
#include <string.h>
#include "third_party/googletest/src/include/gtest/gtest.h"
extern "C" {
#include "vp9/common/vp9_entropy.h"
#include "./vp9_rtcd.h"
void vp9_short_fdct32x32_c(int16_t *input, int16_t *out, int pitch);
void vp9_short_idct32x32_c(short *input, short *output, int pitch);
}
#include "test/acm_random.h"
#include "vpx/vpx_integer.h"
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
#if !CONFIG_DWTDCTHYBRID
static const double kPi = 3.141592653589793238462643383279502884;
static void reference2_32x32_idct_2d(double *input, double *output) {
double x;
for (int l = 0; l < 32; ++l) {
for (int k = 0; k < 32; ++k) {
double s = 0;
for (int i = 0; i < 32; ++i) {
for (int j = 0; j < 32; ++j) {
x = cos(kPi * j * (l + 0.5) / 32.0) *
cos(kPi * i * (k + 0.5) / 32.0) * input[i * 32 + j] / 1024;
if (i != 0)
x *= sqrt(2.0);
if (j != 0)
x *= sqrt(2.0);
s += x;
}
}
output[k * 32 + l] = s / 4;
}
}
}
static void reference_32x32_dct_1d(double in[32], double out[32], int stride) {
const double kInvSqrt2 = 0.707106781186547524400844362104;
for (int k = 0; k < 32; k++) {
out[k] = 0.0;
for (int n = 0; n < 32; n++)
out[k] += in[n] * cos(kPi * (2 * n + 1) * k / 64.0);
if (k == 0)
out[k] = out[k] * kInvSqrt2;
}
}
static void reference_32x32_dct_2d(int16_t input[32*32], double output[32*32]) {
// First transform columns
for (int i = 0; i < 32; ++i) {
double temp_in[32], temp_out[32];
for (int j = 0; j < 32; ++j)
temp_in[j] = input[j*32 + i];
reference_32x32_dct_1d(temp_in, temp_out, 1);
for (int j = 0; j < 32; ++j)
output[j * 32 + i] = temp_out[j];
}
// Then transform rows
for (int i = 0; i < 32; ++i) {
double temp_in[32], temp_out[32];
for (int j = 0; j < 32; ++j)
temp_in[j] = output[j + i*32];
reference_32x32_dct_1d(temp_in, temp_out, 1);
// Scale by some magic number
for (int j = 0; j < 32; ++j)
output[j + i * 32] = temp_out[j] / 4;
}
}
TEST(VP9Idct32x32Test, AccuracyCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 1000;
for (int i = 0; i < count_test_block; ++i) {
int16_t in[1024], coeff[1024];
int16_t out_c[1024];
double out_r[1024];
// Initialize a test block with input range [-255, 255].
for (int j = 0; j < 1024; ++j)
in[j] = rnd.Rand8() - rnd.Rand8();
reference_32x32_dct_2d(in, out_r);
for (int j = 0; j < 1024; j++)
coeff[j] = round(out_r[j]);
vp9_short_idct32x32_c(coeff, out_c, 64);
for (int j = 0; j < 1024; ++j) {
const int diff = out_c[j] - in[j];
const int error = diff * diff;
EXPECT_GE(1, error)
<< "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;
int64_t total_error = 0;
const int count_test_block = 1000;
for (int i = 0; i < count_test_block; ++i) {
int16_t test_input_block[1024];
int16_t test_temp_block[1024];
int16_t test_output_block[1024];
// Initialize a test block with input range [-255, 255].
for (int j = 0; j < 1024; ++j)
test_input_block[j] = rnd.Rand8() - rnd.Rand8();
const int pitch = 64;
vp9_short_fdct32x32_c(test_input_block, test_temp_block, pitch);
vp9_short_idct32x32_c(test_temp_block, test_output_block, pitch);
for (int j = 0; j < 1024; ++j) {
const unsigned diff = test_input_block[j] - test_output_block[j];
const unsigned error = diff * diff;
if (max_error < error)
max_error = error;
total_error += error;
}
}
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";
}
TEST(VP9Fdct32x32Test, CoeffSizeCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 1000;
for (int i = 0; i < count_test_block; ++i) {
int16_t input_block[1024], input_extreme_block[1024];
int16_t output_block[1024], output_extreme_block[1024];
// Initialize a test block with input range [-255, 255].
for (int j = 0; j < 1024; ++j) {
input_block[j] = rnd.Rand8() - rnd.Rand8();
input_extreme_block[j] = rnd.Rand8() % 2 ? 255 : -255;
}
if (i == 0)
for (int j = 0; j < 1024; ++j)
input_extreme_block[j] = 255;
const int pitch = 64;
vp9_short_fdct32x32_c(input_block, output_block, pitch);
vp9_short_fdct32x32_c(input_extreme_block, output_extreme_block, pitch);
// The minimum quant value is 4.
for (int j = 0; j < 1024; ++j) {
EXPECT_GE(4*DCT_MAX_VALUE, abs(output_block[j]))
<< "Error: 32x32 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
EXPECT_GE(4*DCT_MAX_VALUE, abs(output_extreme_block[j]))
<< "Error: 32x32 FDCT extreme has coefficient larger than "
"4*DCT_MAX_VALUE";
}
}
}
} // namespace

1
test/test.mk
View File

@ -69,6 +69,7 @@ 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) += variance_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct32x32_test.cc
endif # VP9

2
vp8/common/postproc.c
View File

@ -15,7 +15,7 @@
#include "vpx_scale/yv12config.h"
#include "postproc.h"
#include "common.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "systemdependent.h"
#include <limits.h>

2
vp8/decoder/decodframe.c
View File

@ -21,7 +21,7 @@
#include "vp8/common/alloccommon.h"
#include "vp8/common/entropymode.h"
#include "vp8/common/quant_common.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "vp8/common/setupintrarecon.h"
#include "decodemv.h"

2
vp8/decoder/onyxd_if.c
View File

@ -26,7 +26,7 @@
#include "vp8/common/quant_common.h"
#include "./vpx_scale_rtcd.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "vp8/common/systemdependent.h"
#include "vpx_ports/vpx_timer.h"
#include "detokenize.h"

2
vp8/encoder/firstpass.c
View File

@ -21,7 +21,7 @@
#include "vp8/common/systemdependent.h"
#include "mcomp.h"
#include "firstpass.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "encodemb.h"
#include "vp8/common/extend.h"
#include "vpx_mem/vpx_mem.h"

6
vp8/encoder/onyx_if.c
View File

@ -20,7 +20,7 @@
#include "mcomp.h"
#include "firstpass.h"
#include "psnr.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "vp8/common/extend.h"
#include "ratectrl.h"
#include "vp8/common/quant_common.h"
@ -2588,7 +2588,7 @@ static void scale_and_extend_source(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi)
Scale2Ratio(cm->horiz_scale, &hr, &hs);
Scale2Ratio(cm->vert_scale, &vr, &vs);
vp8_scale_frame(sd, &cpi->scaled_source, cm->temp_scale_frame.y_buffer,
vpx_scale_frame(sd, &cpi->scaled_source, cm->temp_scale_frame.y_buffer,
tmp_height, hs, hr, vs, vr, 0);
vp8_yv12_extend_frame_borders(&cpi->scaled_source);
@ -3466,7 +3466,7 @@ static void encode_frame_to_data_rate
/* Note that we should not throw out a key frame (especially when
* spatial resampling is enabled).
*/
if ((cm->frame_type == KEY_FRAME))
if (cm->frame_type == KEY_FRAME)
{
cpi->decimation_count = cpi->decimation_factor;
}

2
vp8/encoder/picklpf.c
View File

@ -14,7 +14,7 @@
#include "onyx_int.h"
#include "quantize.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "vp8/common/alloccommon.h"
#include "vp8/common/loopfilter.h"
#if ARCH_ARM

2
vp8/encoder/psnr.c
View File

@ -13,7 +13,7 @@
#include "math.h"
#include "vp8/common/systemdependent.h" /* for vp8_clear_system_state() */
#define MAX_PSNR 60
#define MAX_PSNR 100
double vp8_mse2psnr(double Samples, double Peak, double Mse)
{

2
vp8/encoder/temporal_filter.c
View File

@ -17,7 +17,7 @@
#include "mcomp.h"
#include "firstpass.h"
#include "psnr.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "vp8/common/extend.h"
#include "ratectrl.h"
#include "vp8/common/quant_common.h"

4
vp8/vp8_cx_iface.c
View File

@ -1178,7 +1178,9 @@ static vpx_codec_err_t vp8e_set_scalemode(vpx_codec_alg_priv_t *ctx,
{
int res;
vpx_scaling_mode_t scalemode = *(vpx_scaling_mode_t *)data ;
res = vp8_set_internal_size(ctx->cpi, scalemode.h_scaling_mode, scalemode.v_scaling_mode);
res = vp8_set_internal_size(ctx->cpi,
(VPX_SCALING)scalemode.h_scaling_mode,
(VPX_SCALING)scalemode.v_scaling_mode);
if (!res)
{

4
vp9/common/vp9_alloccommon.c
View File

@ -220,4 +220,8 @@ void vp9_initialize_common() {
vp9_entropy_mode_init();
vp9_entropy_mv_init();
#if CONFIG_NEWCOEFCONTEXT
vp9_init_neighbors();
#endif
}

2
vp9/common/vp9_alloccommon.h
View File

@ -23,4 +23,4 @@ void vp9_setup_version(VP9_COMMON *oci);
void vp9_update_mode_info_border(VP9_COMMON *cpi, MODE_INFO *mi_base);
void vp9_update_mode_info_in_image(VP9_COMMON *cpi, MODE_INFO *mi);
#endif
#endif // VP9_COMMON_VP9_ALLOCCOMMON_H_

23
vp9/common/vp9_blockd.c
View File

@ -12,18 +12,15 @@
#include "vp9/common/vp9_blockd.h"
#include "vpx_mem/vpx_mem.h"
const unsigned char vp9_block2left[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
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 unsigned char vp9_block2above[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
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 unsigned char vp9_block2left_8x8[25] = {
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
};
const unsigned char vp9_block2above_8x8[25] = {
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
};

143
vp9/common/vp9_blockd.h
View File

@ -45,6 +45,19 @@ void vpx_log(const char *format, ...);
#define SEGMENT_DELTADATA 0
#define SEGMENT_ABSDATA 1
#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;
@ -65,11 +78,6 @@ typedef struct {
ENTROPY_CONTEXT y2;
} ENTROPY_CONTEXT_PLANES;
extern const unsigned char vp9_block2left[25];
extern const unsigned char vp9_block2above[25];
extern const unsigned char vp9_block2left_8x8[25];
extern const unsigned char vp9_block2above_8x8[25];
#define VP9_COMBINEENTROPYCONTEXTS( Dest, A, B) \
Dest = ((A)!=0) + ((B)!=0);
@ -80,10 +88,13 @@ typedef enum {
typedef enum
{
SIXTAP = 0,
BILINEAR = 1,
EIGHTTAP = 2,
EIGHTTAP_SHARP = 3,
#if CONFIG_ENABLE_6TAP
SIXTAP,
#endif
EIGHTTAP_SMOOTH,
EIGHTTAP,
EIGHTTAP_SHARP,
BILINEAR,
SWITCHABLE /* should be the last one */
} INTERPOLATIONFILTERTYPE;
@ -101,13 +112,11 @@ typedef enum
TM_PRED, /* Truemotion prediction */
I8X8_PRED, /* 8x8 based prediction, each 8x8 has its own prediction mode */
B_PRED, /* block based prediction, each block has its own prediction mode */
NEARESTMV,
NEARMV,
ZEROMV,
NEWMV,
SPLITMV,
MB_MODE_COUNT
} MB_PREDICTION_MODE;
@ -120,15 +129,16 @@ typedef enum {
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_FEATURES;
// Segment level features.
typedef enum {
TX_4X4, // 4x4 dct transform
TX_8X8, // 8x8 dct transform
TX_16X16, // 16x16 dct transform
TX_SIZE_MAX // Number of different transforms available
TX_4X4 = 0, // 4x4 dct transform
TX_8X8 = 1, // 8x8 dct transform
TX_16X16 = 2, // 16x16 dct transform
TX_SIZE_MAX_MB = 3, // Number of different transforms available
TX_32X32 = TX_SIZE_MAX_MB, // 32x32 dct transform
TX_SIZE_MAX_SB, // Number of transforms available to SBs
} TX_SIZE;
typedef enum {
@ -205,9 +215,6 @@ union b_mode_info {
struct {
B_PREDICTION_MODE first;
TX_TYPE tx_type;
#if CONFIG_COMP_INTRA_PRED
B_PREDICTION_MODE second;
#endif
#if CONFIG_NEWBINTRAMODES
B_PREDICTION_MODE context;
#endif
@ -227,18 +234,21 @@ typedef enum {
MAX_REF_FRAMES = 4
} MV_REFERENCE_FRAME;
typedef enum {
BLOCK_SIZE_MB16X16 = 0,
BLOCK_SIZE_SB32X32 = 1,
BLOCK_SIZE_SB64X64 = 2,
} BLOCK_SIZE_TYPE;
typedef struct {
MB_PREDICTION_MODE mode, uv_mode;
#if CONFIG_COMP_INTRA_PRED
MB_PREDICTION_MODE second_mode, second_uv_mode;
#endif
#if CONFIG_COMP_INTERINTRA_PRED
MB_PREDICTION_MODE interintra_mode, interintra_uv_mode;
#endif
MV_REFERENCE_FRAME ref_frame, second_ref_frame;
TX_SIZE txfm_size;
int_mv mv[2]; // for each reference frame used
int_mv ref_mvs[MAX_REF_FRAMES][MAX_MV_REFS];
int_mv ref_mvs[MAX_REF_FRAMES][MAX_MV_REF_CANDIDATES];
int_mv best_mv, best_second_mv;
#if CONFIG_NEW_MVREF
int best_index, best_second_index;
@ -261,17 +271,9 @@ typedef struct {
// a valid predictor
unsigned char mb_in_image;
#if CONFIG_PRED_FILTER
// Flag to turn prediction signal filter on(1)/off(0 ) at the MB level
unsigned int pred_filter_enabled;
#endif
INTERPOLATIONFILTERTYPE interp_filter;
INTERPOLATIONFILTERTYPE interp_filter;
#if CONFIG_SUPERBLOCKS
// FIXME need a SB array of 4 MB_MODE_INFOs that
// only needs one encoded_as_sb.
unsigned char encoded_as_sb;
#endif
BLOCK_SIZE_TYPE sb_type;
} MB_MODE_INFO;
typedef struct {
@ -280,19 +282,19 @@ typedef struct {
} MODE_INFO;
typedef struct blockd {
short *qcoeff;
short *dqcoeff;
unsigned char *predictor;
short *diff;
short *dequant;
int16_t *qcoeff;
int16_t *dqcoeff;
uint8_t *predictor;
int16_t *diff;
int16_t *dequant;
/* 16 Y blocks, 4 U blocks, 4 V blocks each with 16 entries */
unsigned char **base_pre;
unsigned char **base_second_pre;
uint8_t **base_pre;
uint8_t **base_second_pre;
int pre;
int pre_stride;
unsigned char **base_dst;
uint8_t **base_dst;
int dst;
int dst_stride;
@ -301,12 +303,21 @@ typedef struct blockd {
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;
typedef struct macroblockd {
DECLARE_ALIGNED(16, short, diff[400]); /* from idct diff */
DECLARE_ALIGNED(16, unsigned char, predictor[384]);
DECLARE_ALIGNED(16, short, qcoeff[400]);
DECLARE_ALIGNED(16, short, dqcoeff[400]);
DECLARE_ALIGNED(16, unsigned short, eobs[25]);
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]);
SUPERBLOCKD sb_coeff_data;
/* 16 Y blocks, 4 U, 4 V, 1 DC 2nd order block, each with 16 entries. */
BLOCKD block[25];
@ -350,7 +361,7 @@ typedef struct macroblockd {
vp9_prob mb_segment_tree_probs[MB_FEATURE_TREE_PROBS];
#if CONFIG_NEW_MVREF
vp9_prob mb_mv_ref_id_probs[MAX_REF_FRAMES][3];
vp9_prob mb_mv_ref_probs[MAX_REF_FRAMES][MAX_MV_REF_CANDIDATES-1];
#endif
// Segment features
@ -377,17 +388,17 @@ typedef struct macroblockd {
unsigned int frames_till_alt_ref_frame;
/* Inverse transform function pointers. */
void (*inv_xform4x4_1_x8)(short *input, short *output, int pitch);
void (*inv_xform4x4_x8)(short *input, short *output, int pitch);
void (*inv_walsh4x4_1)(short *in, short *out);
void (*inv_walsh4x4_lossless)(short *in, short *out);
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);
vp9_subpix_fn_t subpixel_predict;
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_avg;
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;
@ -395,14 +406,7 @@ typedef struct macroblockd {
int corrupted;
#if !CONFIG_SUPERBLOCKS && (ARCH_X86 || ARCH_X86_64)
/* This is an intermediate buffer currently used in sub-pixel motion search
* to keep a copy of the reference area. This buffer can be used for other
* purpose.
*/
DECLARE_ALIGNED(32, unsigned char, y_buf[22 * 32]);
#endif
int sb_index;
int mb_index; // Index of the MB in the SB (0..3)
int q_index;
@ -490,6 +494,9 @@ static TX_TYPE txfm_map(B_PREDICTION_MODE bmode) {
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];
#define USE_ADST_FOR_I16X16_8X8 0
#define USE_ADST_FOR_I16X16_4X4 0
#define USE_ADST_FOR_I8X8_4X4 1
@ -502,11 +509,9 @@ static TX_TYPE get_tx_type_4x4(const MACROBLOCKD *xd, const BLOCKD *b) {
int ib = (int)(b - xd->block);
if (ib >= 16)
return tx_type;
#if CONFIG_SUPERBLOCKS
// TODO(rbultje, debargha): Explore ADST usage for superblocks
if (xd->mode_info_context->mbmi.encoded_as_sb)
if (xd->mode_info_context->mbmi.sb_type)
return tx_type;
#endif
if (xd->mode_info_context->mbmi.mode == B_PRED &&
xd->q_index < ACTIVE_HT) {
tx_type = txfm_map(
@ -559,11 +564,9 @@ static TX_TYPE get_tx_type_8x8(const MACROBLOCKD *xd, const BLOCKD *b) {
int ib = (int)(b - xd->block);
if (ib >= 16)
return tx_type;
#if CONFIG_SUPERBLOCKS
// TODO(rbultje, debargha): Explore ADST usage for superblocks
if (xd->mode_info_context->mbmi.encoded_as_sb)
if (xd->mode_info_context->mbmi.sb_type)
return tx_type;
#endif
if (xd->mode_info_context->mbmi.mode == I8X8_PRED &&
xd->q_index < ACTIVE_HT8) {
// TODO(rbultje): MB_PREDICTION_MODE / B_PREDICTION_MODE should be merged
@ -594,11 +597,9 @@ static TX_TYPE get_tx_type_16x16(const MACROBLOCKD *xd, const BLOCKD *b) {
int ib = (int)(b - xd->block);
if (ib >= 16)
return tx_type;
#if CONFIG_SUPERBLOCKS
// TODO(rbultje, debargha): Explore ADST usage for superblocks
if (xd->mode_info_context->mbmi.encoded_as_sb)
if (xd->mode_info_context->mbmi.sb_type)
return tx_type;
#endif
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));
@ -650,4 +651,4 @@ static void update_blockd_bmi(MACROBLOCKD *xd) {
}
}
}
#endif /* __INC_BLOCKD_H */
#endif // VP9_COMMON_VP9_BLOCKD_H_

4
vp9/common/vp9_coefupdateprobs.h
View File

@ -8,9 +8,13 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef 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
#endif // VP9_COMMON_VP9_COEFUPDATEPROBS_H__

30
vp9/common/vp9_common.h
View File

@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_COMMON_H_
#define VP9_COMMON_VP9_COMMON_H_
@ -17,25 +16,34 @@
/* Interface header for common constant data structures and lookup tables */
#include "vpx_mem/vpx_mem.h"
#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_common_types.h"
#define TRUE 1
#define FALSE 0
#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 vp9_copy( Dest, Src) { \
assert( sizeof( Dest) == sizeof( Src)); \
vpx_memcpy( Dest, Src, sizeof( Src)); \
#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)); \
#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));
#endif /* common_h */
static __inline uint8_t clip_pixel(int val) {
return (val > 255) ? 255u : (val < 0) ? 0u : val;
}
#endif // VP9_COMMON_VP9_COMMON_H_

18
vp9/common/vp9_common_types.h
View File

@ -1,18 +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_COMMON_TYPES_H_
#define VP9_COMMON_VP9_COMMON_TYPES_H_
#define TRUE 1
#define FALSE 0
#endif

3
vp9/common/vp9_debugmodes.c
View File

@ -87,9 +87,6 @@ void vp9_print_modes_and_motion_vectors(MODE_INFO *mi, int rows, int cols,
if (mi[mb_index].mbmi.mode == B_PRED) {
fprintf(mvs, "%2d ", mi[mb_index].bmi[bindex].as_mode.first);
#if CONFIG_COMP_INTRA_PRED
fprintf(mvs, "%2d ", mi[mb_index].bmi[bindex].as_mode.second);
#endif
} else
fprintf(mvs, "xx ");

2532
vp9/common/vp9_default_coef_probs.h
View File

File diff suppressed because it is too large Load Diff

1207
vp9/common/vp9_entropy.c
View File

File diff suppressed because it is too large Load Diff

64
vp9/common/vp9_entropy.h
View File

@ -8,10 +8,10 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_ENTROPY_H_
#define VP9_COMMON_VP9_ENTROPY_H_
#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_treecoder.h"
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_common.h"
@ -55,24 +55,27 @@ extern vp9_extra_bit_struct vp9_extra_bits[12]; /* indexed by token value */
#define PROB_UPDATE_BASELINE_COST 7
#define MAX_PROB 255
#define DCT_MAX_VALUE 8192
#define DCT_MAX_VALUE 16384
/* 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 4
#define BLOCK_TYPES_4X4 4
#define BLOCK_TYPES_8X8 4
#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[16]);
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
@ -91,24 +94,61 @@ extern DECLARE_ALIGNED(16, const int, vp9_coef_bands_16x16[256]);
distinct bands). */
/*# define DC_TOKEN_CONTEXTS 3*/ /* 00, 0!0, !0!0 */
#define PREV_COEF_CONTEXTS 4
#define PREV_COEF_CONTEXTS 4
typedef unsigned int vp9_coeff_count[COEF_BANDS][PREV_COEF_CONTEXTS]
[MAX_ENTROPY_TOKENS];
typedef unsigned int vp9_coeff_stats[COEF_BANDS][PREV_COEF_CONTEXTS]
[ENTROPY_NODES][2];
typedef vp9_prob vp9_coeff_probs[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 unsigned char, vp9_prev_token_class[MAX_ENTROPY_TOKENS]);
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[16]);
extern DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_4x4[16]);
extern DECLARE_ALIGNED(16, const int, vp9_col_scan[16]);
extern DECLARE_ALIGNED(16, const int, vp9_row_scan[16]);
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]);
void vp9_coef_tree_initialize(void);
extern DECLARE_ALIGNED(16, const int, vp9_default_zig_zag1d_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 *);
#endif
static 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
#define MAX_NEIGHBORS 5
#define NEWCOEFCONTEXT_BAND_COND(b) ((b) >= 1)
void vp9_init_neighbors(void);
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_

234
vp9/common/vp9_entropymode.c
View File

@ -272,13 +272,11 @@ const vp9_tree_index vp9_mv_ref_tree[8] = {
-NEWMV, -SPLITMV
};
#if CONFIG_SUPERBLOCKS
const vp9_tree_index vp9_sb_mv_ref_tree[6] = {
-ZEROMV, 2,
-NEARESTMV, 4,
-NEARMV, -NEWMV
};
#endif
const vp9_tree_index vp9_sub_mv_ref_tree[6] = {
-LEFT4X4, 2,
@ -289,19 +287,15 @@ const vp9_tree_index vp9_sub_mv_ref_tree[6] = {
struct vp9_token_struct vp9_bmode_encodings[VP9_NKF_BINTRAMODES];
struct vp9_token_struct vp9_kf_bmode_encodings[VP9_KF_BINTRAMODES];
struct vp9_token_struct vp9_ymode_encodings[VP9_YMODES];
#if CONFIG_SUPERBLOCKS
struct vp9_token_struct vp9_sb_ymode_encodings[VP9_I32X32_MODES];
struct vp9_token_struct vp9_sb_kf_ymode_encodings[VP9_I32X32_MODES];
#endif
struct vp9_token_struct vp9_kf_ymode_encodings[VP9_YMODES];
struct vp9_token_struct vp9_uv_mode_encodings[VP9_UV_MODES];
struct vp9_token_struct vp9_i8x8_mode_encodings[VP9_I8X8_MODES];
struct vp9_token_struct vp9_mbsplit_encodings[VP9_NUMMBSPLITS];
struct vp9_token_struct vp9_mv_ref_encoding_array[VP9_MVREFS];
#if CONFIG_SUPERBLOCKS
struct vp9_token_struct vp9_sb_mv_ref_encoding_array[VP9_MVREFS];
#endif
struct vp9_token_struct vp9_sub_mv_ref_encoding_array[VP9_SUBMVREFS];
void vp9_init_mbmode_probs(VP9_COMMON *x) {
@ -309,25 +303,21 @@ void vp9_init_mbmode_probs(VP9_COMMON *x) {
vp9_tree_probs_from_distribution(VP9_YMODES, vp9_ymode_encodings,
vp9_ymode_tree, x->fc.ymode_prob,
bct, y_mode_cts, 256, 1);
#if CONFIG_SUPERBLOCKS
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, 256, 1);
#endif
bct, y_mode_cts);
{
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], 256, 1);
#if CONFIG_SUPERBLOCKS
bct, kf_y_mode_cts[i]);
vp9_tree_probs_from_distribution(VP9_I32X32_MODES,
vp9_sb_kf_ymode_encodings,
vp9_sb_kf_ymode_tree,
x->sb_kf_ymode_prob[i], bct,
kf_y_mode_cts[i], 256, 1);
#endif
kf_y_mode_cts[i]);
}
}
{
@ -335,16 +325,16 @@ void vp9_init_mbmode_probs(VP9_COMMON *x) {
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], 256, 1);
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], 256, 1);
bct, uv_mode_cts[i]);
}
}
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, 256, 1);
bct, i8x8_mode_cts);
vpx_memcpy(x->fc.sub_mv_ref_prob, vp9_sub_mv_ref_prob2,
sizeof(vp9_sub_mv_ref_prob2));
@ -362,7 +352,7 @@ static void intra_bmode_probs_from_distribution(
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, 256, 1);
vp9_bmode_tree, p, branch_ct, events);
}
void vp9_default_bmode_probs(vp9_prob p[VP9_NKF_BINTRAMODES - 1]) {
@ -375,7 +365,7 @@ static void intra_kf_bmode_probs_from_distribution(
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, 256, 1);
vp9_kf_bmode_tree, p, branch_ct, events);
}
void vp9_kf_default_bmode_probs(vp9_prob p[VP9_KF_BINTRAMODES]
@ -398,9 +388,15 @@ const vp9_tree_index vp9_switchable_interp_tree[VP9_SWITCHABLE_FILTERS*2-2] = {
-1, -2
};
struct vp9_token_struct vp9_switchable_interp_encodings[VP9_SWITCHABLE_FILTERS];
#if CONFIG_ENABLE_6TAP
const INTERPOLATIONFILTERTYPE vp9_switchable_interp[VP9_SWITCHABLE_FILTERS] = {
EIGHTTAP, SIXTAP, EIGHTTAP_SHARP};
const int vp9_switchable_interp_map[SWITCHABLE+1] = {1, -1, 0, 2, -1};
SIXTAP, EIGHTTAP, EIGHTTAP_SHARP};
const int vp9_switchable_interp_map[SWITCHABLE+1] = {0, -1, 1, 2, -1, -1};
#else
const INTERPOLATIONFILTERTYPE vp9_switchable_interp[VP9_SWITCHABLE_FILTERS] = {
EIGHTTAP, EIGHTTAP_SMOOTH, EIGHTTAP_SHARP};
const int vp9_switchable_interp_map[SWITCHABLE+1] = {1, 0, 2, -1, -1};
#endif
const vp9_prob vp9_switchable_interp_prob [VP9_SWITCHABLE_FILTERS+1]
[VP9_SWITCHABLE_FILTERS-1] = {
{248, 192}, { 32, 248}, { 32, 32}, {192, 160}
@ -418,7 +414,11 @@ const vp9_prob vp9_switchable_interp_prob [VP9_SWITCHABLE_FILTERS+1]
};
const INTERPOLATIONFILTERTYPE vp9_switchable_interp[VP9_SWITCHABLE_FILTERS] = {
EIGHTTAP, EIGHTTAP_SHARP};
const int vp9_switchable_interp_map[SWITCHABLE+1] = {-1, -1, 0, 1, -1}; //8, 8s
#if CONFIG_ENABLE_6TAP
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
void vp9_entropy_mode_init() {
@ -426,10 +426,8 @@ void vp9_entropy_mode_init() {
vp9_tokens_from_tree(vp9_bmode_encodings, vp9_bmode_tree);
vp9_tokens_from_tree(vp9_ymode_encodings, vp9_ymode_tree);
vp9_tokens_from_tree(vp9_kf_ymode_encodings, vp9_kf_ymode_tree);
#if CONFIG_SUPERBLOCKS
vp9_tokens_from_tree(vp9_sb_ymode_encodings, vp9_sb_ymode_tree);
vp9_tokens_from_tree(vp9_sb_kf_ymode_encodings, vp9_sb_kf_ymode_tree);
#endif
vp9_tokens_from_tree(vp9_uv_mode_encodings, vp9_uv_mode_tree);
vp9_tokens_from_tree(vp9_i8x8_mode_encodings, vp9_i8x8_mode_tree);
vp9_tokens_from_tree(vp9_mbsplit_encodings, vp9_mbsplit_tree);
@ -438,10 +436,8 @@ void vp9_entropy_mode_init() {
vp9_tokens_from_tree_offset(vp9_mv_ref_encoding_array,
vp9_mv_ref_tree, NEARESTMV);
#if CONFIG_SUPERBLOCKS
vp9_tokens_from_tree_offset(vp9_sb_mv_ref_encoding_array,
vp9_sb_mv_ref_tree, NEARESTMV);
#endif
vp9_tokens_from_tree_offset(vp9_sub_mv_ref_encoding_array,
vp9_sub_mv_ref_tree, LEFT4X4);
}
@ -495,17 +491,14 @@ void vp9_update_mode_context(VP9_COMMON *pc) {
for (j = 0; j < INTER_MODE_CONTEXTS; j++) {
for (i = 0; i < 4; i++) {
int this_prob;
int count = mv_ref_ct[j][i][0] + mv_ref_ct[j][i][1];
int factor;
{
this_prob = count > 0 ? 256 * mv_ref_ct[j][i][0] / count : 128;
count = count > MVREF_COUNT_SAT ? MVREF_COUNT_SAT : count;
factor = (MVREF_MAX_UPDATE_FACTOR * count / MVREF_COUNT_SAT);
this_prob = (pc->fc.vp9_mode_contexts[j][i] * (256 - factor) +
this_prob * factor + 128) >> 8;
mode_context[j][i] = clip_prob(this_prob);
}
int count = mv_ref_ct[j][i][0] + mv_ref_ct[j][i][1], factor;
count = count > MVREF_COUNT_SAT ? MVREF_COUNT_SAT : count;
factor = (MVREF_MAX_UPDATE_FACTOR * count / MVREF_COUNT_SAT);
mode_context[j][i] = weighted_prob(pc->fc.vp9_mode_contexts[j][i],
get_binary_prob(mv_ref_ct[j][i][0],
mv_ref_ct[j][i][1]),
factor);
}
}
}
@ -531,25 +524,33 @@ void print_mode_contexts(VP9_COMMON *pc) {
}
#endif
// #define MODE_COUNT_TESTING
#define MODE_COUNT_SAT 20
#define MODE_MAX_UPDATE_FACTOR 144
static void update_mode_probs(int n_modes, struct vp9_token_struct *encoding,
const vp9_tree_index *tree, unsigned int *cnt,
vp9_prob *pre_probs, vp9_prob *dst_probs) {
#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);
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;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
dst_probs[t] = weighted_prob(pre_probs[t], probs[t], factor);
}
}
// #define MODE_COUNT_TESTING
void vp9_adapt_mode_probs(VP9_COMMON *cm) {
int i, t, count, factor;
unsigned int branch_ct[32][2];
vp9_prob ymode_probs[VP9_YMODES - 1];
#if CONFIG_SUPERBLOCKS
vp9_prob sb_ymode_probs[VP9_I32X32_MODES - 1];
#endif
vp9_prob uvmode_probs[VP9_UV_MODES - 1];
vp9_prob bmode_probs[VP9_NKF_BINTRAMODES - 1];
vp9_prob i8x8_mode_probs[VP9_I8X8_MODES - 1];
vp9_prob sub_mv_ref_probs[VP9_SUBMVREFS - 1];
vp9_prob mbsplit_probs[VP9_NUMMBSPLITS - 1];
#if CONFIG_COMP_INTERINTRA_PRED
vp9_prob interintra_prob;
#endif
int i;
#ifdef MODE_COUNT_TESTING
int t;
printf("static const unsigned int\nymode_counts"
"[VP9_YMODES] = {\n");
for (t = 0; t < VP9_YMODES; ++t) printf("%d, ", cm->fc.ymode_counts[t]);
@ -590,116 +591,43 @@ void vp9_adapt_mode_probs(VP9_COMMON *cm) {
printf("};\n");
#endif
#endif
vp9_tree_probs_from_distribution(
VP9_YMODES, vp9_ymode_encodings, vp9_ymode_tree,
ymode_probs, branch_ct, cm->fc.ymode_counts,
256, 1);
for (t = 0; t < VP9_YMODES - 1; ++t) {
int prob;
count = branch_ct[t][0] + branch_ct[t][1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
prob = ((int)cm->fc.pre_ymode_prob[t] * (256 - factor) +
(int)ymode_probs[t] * factor + 128) >> 8;
cm->fc.ymode_prob[t] = clip_prob(prob);
}
#if CONFIG_SUPERBLOCKS
vp9_tree_probs_from_distribution(VP9_I32X32_MODES,
vp9_sb_ymode_encodings, vp9_sb_ymode_tree,
sb_ymode_probs, branch_ct,
cm->fc.sb_ymode_counts,
256, 1);
for (t = 0; t < VP9_I32X32_MODES - 1; ++t) {
int prob;
count = branch_ct[t][0] + branch_ct[t][1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
prob = ((int)cm->fc.pre_sb_ymode_prob[t] * (256 - factor) +
(int)sb_ymode_probs[t] * factor + 128) >> 8;
cm->fc.sb_ymode_prob[t] = clip_prob(prob);
}
#endif
update_mode_probs(VP9_YMODES, vp9_ymode_encodings, 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.sb_ymode_counts, cm->fc.pre_sb_ymode_prob,
cm->fc.sb_ymode_prob);
for (i = 0; i < VP9_YMODES; ++i) {
vp9_tree_probs_from_distribution(VP9_UV_MODES, vp9_uv_mode_encodings,
vp9_uv_mode_tree, uvmode_probs, branch_ct,
cm->fc.uv_mode_counts[i], 256, 1);
for (t = 0; t < VP9_UV_MODES - 1; ++t) {
int prob;
count = branch_ct[t][0] + branch_ct[t][1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
prob = ((int)cm->fc.pre_uv_mode_prob[i][t] * (256 - factor) +
(int)uvmode_probs[t] * factor + 128) >> 8;
cm->fc.uv_mode_prob[i][t] = clip_prob(prob);
}
}
vp9_tree_probs_from_distribution(VP9_NKF_BINTRAMODES, vp9_bmode_encodings,
vp9_bmode_tree, bmode_probs, branch_ct,
cm->fc.bmode_counts, 256, 1);
for (t = 0; t < VP9_NKF_BINTRAMODES - 1; ++t) {
int prob;
count = branch_ct[t][0] + branch_ct[t][1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
prob = ((int)cm->fc.pre_bmode_prob[t] * (256 - factor) +
(int)bmode_probs[t] * factor + 128) >> 8;
cm->fc.bmode_prob[t] = clip_prob(prob);
}
vp9_tree_probs_from_distribution(VP9_I8X8_MODES, vp9_i8x8_mode_encodings,
vp9_i8x8_mode_tree, i8x8_mode_probs,
branch_ct, cm->fc.i8x8_mode_counts, 256, 1);
for (t = 0; t < VP9_I8X8_MODES - 1; ++t) {
int prob;
count = branch_ct[t][0] + branch_ct[t][1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
prob = ((int)cm->fc.pre_i8x8_mode_prob[t] * (256 - factor) +
(int)i8x8_mode_probs[t] * factor + 128) >> 8;
cm->fc.i8x8_mode_prob[t] = clip_prob(prob);
update_mode_probs(VP9_UV_MODES, vp9_uv_mode_encodings, vp9_uv_mode_tree,
cm->fc.uv_mode_counts[i], cm->fc.pre_uv_mode_prob[i],
cm->fc.uv_mode_prob[i]);
}
update_mode_probs(VP9_NKF_BINTRAMODES, vp9_bmode_encodings, 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,
vp9_i8x8_mode_tree, cm->fc.i8x8_mode_counts,
cm->fc.pre_i8x8_mode_prob, cm->fc.i8x8_mode_prob);
for (i = 0; i < SUBMVREF_COUNT; ++i) {
vp9_tree_probs_from_distribution(VP9_SUBMVREFS,
vp9_sub_mv_ref_encoding_array,
vp9_sub_mv_ref_tree, sub_mv_ref_probs,
branch_ct, cm->fc.sub_mv_ref_counts[i],
256, 1);
for (t = 0; t < VP9_SUBMVREFS - 1; ++t) {
int prob;
count = branch_ct[t][0] + branch_ct[t][1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
prob = ((int)cm->fc.pre_sub_mv_ref_prob[i][t] * (256 - factor) +
(int)sub_mv_ref_probs[t] * factor + 128) >> 8;
cm->fc.sub_mv_ref_prob[i][t] = clip_prob(prob);
}
}
vp9_tree_probs_from_distribution(VP9_NUMMBSPLITS, vp9_mbsplit_encodings,
vp9_mbsplit_tree, mbsplit_probs, branch_ct,
cm->fc.mbsplit_counts, 256, 1);
for (t = 0; t < VP9_NUMMBSPLITS - 1; ++t) {
int prob;
count = branch_ct[t][0] + branch_ct[t][1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
prob = ((int)cm->fc.pre_mbsplit_prob[t] * (256 - factor) +
(int)mbsplit_probs[t] * factor + 128) >> 8;
cm->fc.mbsplit_prob[t] = clip_prob(prob);
update_mode_probs(VP9_SUBMVREFS, vp9_sub_mv_ref_encoding_array,
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]);
}
update_mode_probs(VP9_NUMMBSPLITS, vp9_mbsplit_encodings, vp9_mbsplit_tree,
cm->fc.mbsplit_counts, cm->fc.pre_mbsplit_prob,
cm->fc.mbsplit_prob);
#if CONFIG_COMP_INTERINTRA_PRED
if (cm->use_interintra) {
int prob;
interintra_prob = vp9_bin_prob_from_distribution(cm->fc.interintra_counts);
int factor, interintra_prob, count;
interintra_prob = get_binary_prob(cm->fc.interintra_counts[0],
cm->fc.interintra_counts[1]);
count = cm->fc.interintra_counts[0] + cm->fc.interintra_counts[1];
count = count > MODE_COUNT_SAT ? MODE_COUNT_SAT : count;
factor = (MODE_MAX_UPDATE_FACTOR * count / MODE_COUNT_SAT);
prob = ((int)cm->fc.pre_interintra_prob * (256 - factor) +
(int)interintra_prob * factor + 128) >> 8;
if (prob <= 0)
cm->fc.interintra_prob = 1;
else if (prob > 255)
cm->fc.interintra_prob = 255;
else
cm->fc.interintra_prob = prob;
cm->fc.interintra_prob = weighted_prob(cm->fc.pre_interintra_prob,
interintra_prob, factor);
}
#endif
}

8
vp9/common/vp9_entropymode.h
View File

@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_ENTROPYMODE_H_
#define VP9_COMMON_VP9_ENTROPYMODE_H_
@ -17,9 +16,6 @@
#define SUBMVREF_COUNT 5
#define VP9_NUMMBSPLITS 4
#if CONFIG_COMP_INTRA_PRED
#define DEFAULT_COMP_INTRA_PROB 32
#endif
#if CONFIG_COMP_INTERINTRA_PRED
#define VP9_DEF_INTERINTRA_PROB 248
@ -98,7 +94,7 @@ void vp9_kf_default_bmode_probs(vp9_prob dest[VP9_KF_BINTRAMODES]
void vp9_adapt_mode_probs(struct VP9Common *);
#define VP9_SWITCHABLE_FILTERS 2 /* number of switchable filters */
#define VP9_SWITCHABLE_FILTERS 3 /* number of switchable filters */
extern const INTERPOLATIONFILTERTYPE vp9_switchable_interp
[VP9_SWITCHABLE_FILTERS];
@ -114,4 +110,4 @@ extern struct vp9_token_struct vp9_switchable_interp_encodings
extern const vp9_prob vp9_switchable_interp_prob[VP9_SWITCHABLE_FILTERS + 1]
[VP9_SWITCHABLE_FILTERS - 1];
#endif
#endif // VP9_COMMON_VP9_ENTROPYMODE_H_

42
vp9/common/vp9_entropymv.c
View File

@ -213,16 +213,12 @@ 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,
unsigned int ct[2]) {
int factor;
int prob;
int count = ct[0] + ct[1];
if (count) {
count = count > MV_COUNT_SAT ? MV_COUNT_SAT : count;
factor = (MV_MAX_UPDATE_FACTOR * count / MV_COUNT_SAT);
prob = ((int)prep * (256 - factor) + (int)(newp) * factor + 128) >> 8;
prob += !prob;
prob = (prob > 255 ? 255 : prob);
*dest = prob;
*dest = weighted_prob(prep, newp,
MV_MAX_UPDATE_FACTOR * count / MV_COUNT_SAT);
}
}
@ -251,11 +247,10 @@ void vp9_counts_to_nmv_context(
vp9_mv_joint_tree,
prob->joints,
branch_ct_joint,
NMVcount->joints,
256, 1);
NMVcount->joints);
for (i = 0; i < 2; ++i) {
prob->comps[i].sign =
vp9_bin_prob_from_distribution(NMVcount->comps[i].sign);
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,
@ -263,18 +258,16 @@ void vp9_counts_to_nmv_context(
vp9_mv_class_tree,
prob->comps[i].classes,
branch_ct_classes[i],
NMVcount->comps[i].classes,
256, 1);
NMVcount->comps[i].classes);
vp9_tree_probs_from_distribution(CLASS0_SIZE,
vp9_mv_class0_encodings,
vp9_mv_class0_tree,
prob->comps[i].class0,
branch_ct_class0[i],
NMVcount->comps[i].class0,
256, 1);
NMVcount->comps[i].class0);
for (j = 0; j < MV_OFFSET_BITS; ++j) {
prob->comps[i].bits[j] = vp9_bin_prob_from_distribution(
NMVcount->comps[i].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];
}
@ -286,26 +279,25 @@ void vp9_counts_to_nmv_context(
vp9_mv_fp_tree,
prob->comps[i].class0_fp[k],
branch_ct_class0_fp[i][k],
NMVcount->comps[i].class0_fp[k],
256, 1);
NMVcount->comps[i].class0_fp[k]);
}
vp9_tree_probs_from_distribution(4,
vp9_mv_fp_encodings,
vp9_mv_fp_tree,
prob->comps[i].fp,
branch_ct_fp[i],
NMVcount->comps[i].fp,
256, 1);
NMVcount->comps[i].fp);
}
if (usehp) {
for (i = 0; i < 2; ++i) {
prob->comps[i].class0_hp = vp9_bin_prob_from_distribution(
NMVcount->comps[i].class0_hp);
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];
prob->comps[i].hp =
vp9_bin_prob_from_distribution(NMVcount->comps[i].hp);
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];
}

10
vp9/common/vp9_entropymv.h
View File

@ -25,6 +25,13 @@ void vp9_adapt_nmv_probs(struct VP9Common *cm, int usehp);
int vp9_use_nmv_hp(const MV *ref);
#define VP9_NMV_UPDATE_PROB 255
#if CONFIG_NEW_MVREF
#define VP9_MVREF_UPDATE_PROB 252
#define VP9_DEFAULT_MV_REF_PROB 192
#define VP9_MV_REF_UPDATE_COST (14 << 8)
#endif
//#define MV_GROUP_UPDATE
#define LOW_PRECISION_MV_UPDATE /* Use 7 bit forward update */
@ -126,4 +133,5 @@ void vp9_counts_to_nmv_context(
unsigned int (*branch_ct_class0_hp)[2],
unsigned int (*branch_ct_hp)[2]);
void vp9_counts_process(nmv_context_counts *NMVcount, int usehp);
#endif
#endif // VP9_COMMON_VP9_ENTROPYMV_H_

12
vp9/common/vp9_extend.c
View File

@ -11,9 +11,9 @@
#include "vp9/common/vp9_extend.h"
#include "vpx_mem/vpx_mem.h"
static void copy_and_extend_plane(unsigned char *s, /* source */
static void copy_and_extend_plane(uint8_t *s, /* source */
int sp, /* source pitch */
unsigned char *d, /* destination */
uint8_t *d, /* destination */
int dp, /* destination pitch */
int h, /* height */
int w, /* width */
@ -22,8 +22,8 @@ static void copy_and_extend_plane(unsigned char *s, /* source */
int eb, /* extend bottom border */
int er) { /* extend right border */
int i;
unsigned char *src_ptr1, *src_ptr2;
unsigned char *dest_ptr1, *dest_ptr2;
uint8_t *src_ptr1, *src_ptr2;
uint8_t *dest_ptr1, *dest_ptr2;
int linesize;
/* copy the left and right most columns out */
@ -143,8 +143,8 @@ void vp9_copy_and_extend_frame_with_rect(YV12_BUFFER_CONFIG *src,
}
/* note the extension is only for the last row, for intra prediction purpose */
void vp9_extend_mb_row(YV12_BUFFER_CONFIG *ybf, unsigned char *YPtr,
unsigned char *UPtr, unsigned char *VPtr) {
void vp9_extend_mb_row(YV12_BUFFER_CONFIG *ybf, uint8_t *YPtr,
uint8_t *UPtr, uint8_t *VPtr) {
int i;
YPtr += ybf->y_stride * 14;

7
vp9/common/vp9_extend.h
View File

@ -12,9 +12,10 @@
#define VP9_COMMON_VP9_EXTEND_H_
#include "vpx_scale/yv12config.h"
#include "vpx/vpx_integer.h"
void vp9_extend_mb_row(YV12_BUFFER_CONFIG *ybf, unsigned char *YPtr,
unsigned char *UPtr, unsigned char *VPtr);
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,
YV12_BUFFER_CONFIG *dst);
@ -24,4 +25,4 @@ void vp9_copy_and_extend_frame_with_rect(YV12_BUFFER_CONFIG *src,
int srcy, int srcx,
int srch, int srcw);
#endif // __INC_EXTEND_H
#endif // VP9_COMMON_VP9_EXTEND_H_

1044
vp9/common/vp9_filter.c
View File

File diff suppressed because it is too large Load Diff

12
vp9/common/vp9_filter.h
View File

@ -13,6 +13,7 @@
#include "vpx_config.h"
#include "vpx_scale/yv12config.h"
#include "vpx/vpx_integer.h"
#define BLOCK_HEIGHT_WIDTH 4
#define VP9_FILTER_WEIGHT 128
@ -20,9 +21,10 @@
#define SUBPEL_SHIFTS 16
extern const short vp9_bilinear_filters[SUBPEL_SHIFTS][2];
extern const short vp9_sub_pel_filters_6[SUBPEL_SHIFTS][6];
extern const short vp9_sub_pel_filters_8[SUBPEL_SHIFTS][8];
extern const short vp9_sub_pel_filters_8s[SUBPEL_SHIFTS][8];
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_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];
#endif // FILTER_H
#endif // VP9_COMMON_VP9_FILTER_H_

177
vp9/common/vp9_findnearmv.c
View File

@ -14,7 +14,7 @@
#include "vp9/common/vp9_subpelvar.h"
#include <limits.h>
const unsigned char vp9_mbsplit_offset[4][16] = {
const uint8_t vp9_mbsplit_offset[4][16] = {
{ 0, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{ 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{ 0, 2, 8, 10, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
@ -42,23 +42,23 @@ vp9_prob *vp9_mv_ref_probs(VP9_COMMON *pc,
}
#define SP(x) (((x) & 7) << 1)
unsigned int vp9_sad3x16_c(const unsigned char *src_ptr,
unsigned int vp9_sad3x16_c(const uint8_t *src_ptr,
int src_stride,
const unsigned char *ref_ptr,
const uint8_t *ref_ptr,
int ref_stride) {
return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 3, 16);
}
unsigned int vp9_sad16x3_c(const unsigned char *src_ptr,
unsigned int vp9_sad16x3_c(const uint8_t *src_ptr,
int src_stride,
const unsigned char *ref_ptr,
const uint8_t *ref_ptr,
int ref_stride) {
return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 16, 3);
}
#if CONFIG_SUBPELREFMV
unsigned int vp9_variance2x16_c(const unsigned char *src_ptr,
unsigned int vp9_variance2x16_c(const uint8_t *src_ptr,
const int source_stride,
const unsigned char *ref_ptr,
const uint8_t *ref_ptr,
const int recon_stride,
unsigned int *sse) {
int sum;
@ -66,9 +66,9 @@ unsigned int vp9_variance2x16_c(const unsigned char *src_ptr,
return (*sse - (((unsigned int)sum * sum) >> 5));
}
unsigned int vp9_variance16x2_c(const unsigned char *src_ptr,
unsigned int vp9_variance16x2_c(const uint8_t *src_ptr,
const int source_stride,
const unsigned char *ref_ptr,
const uint8_t *ref_ptr,
const int recon_stride,
unsigned int *sse) {
int sum;
@ -76,16 +76,16 @@ unsigned int vp9_variance16x2_c(const unsigned char *src_ptr,
return (*sse - (((unsigned int)sum * sum) >> 5));
}
unsigned int vp9_sub_pixel_variance16x2_c(const unsigned char *src_ptr,
unsigned int vp9_sub_pixel_variance16x2_c(const uint8_t *src_ptr,
const int src_pixels_per_line,
const int xoffset,
const int yoffset,
const unsigned char *dst_ptr,
const uint8_t *dst_ptr,
const int dst_pixels_per_line,
unsigned int *sse) {
unsigned short FData3[16 * 3]; // Temp data buffer used in filtering
unsigned char temp2[2 * 16];
const short *HFilter, *VFilter;
uint16_t FData3[16 * 3]; // Temp data buffer used in filtering
uint8_t temp2[2 * 16];
const int16_t *HFilter, *VFilter;
HFilter = vp9_bilinear_filters[xoffset];
VFilter = vp9_bilinear_filters[yoffset];
@ -97,16 +97,16 @@ unsigned int vp9_sub_pixel_variance16x2_c(const unsigned char *src_ptr,
return vp9_variance16x2_c(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
}
unsigned int vp9_sub_pixel_variance2x16_c(const unsigned char *src_ptr,
unsigned int vp9_sub_pixel_variance2x16_c(const uint8_t *src_ptr,
const int src_pixels_per_line,
const int xoffset,
const int yoffset,
const unsigned char *dst_ptr,
const uint8_t *dst_ptr,
const int dst_pixels_per_line,
unsigned int *sse) {
unsigned short FData3[2 * 17]; // Temp data buffer used in filtering
unsigned char temp2[2 * 16];
const short *HFilter, *VFilter;
uint16_t FData3[2 * 17]; // Temp data buffer used in filtering
uint8_t temp2[2 * 16];
const int16_t *HFilter, *VFilter;
HFilter = vp9_bilinear_filters[xoffset];
VFilter = vp9_bilinear_filters[yoffset];
@ -117,51 +117,46 @@ unsigned int vp9_sub_pixel_variance2x16_c(const unsigned char *src_ptr,
return vp9_variance2x16_c(temp2, 2, dst_ptr, dst_pixels_per_line, sse);
}
#endif
/* 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,
unsigned char *ref_y_buffer,
uint8_t *ref_y_buffer,
int ref_y_stride,
int_mv *mvlist,
int_mv *best_mv,
int_mv *nearest,
int_mv *near) {
int i, j;
unsigned char *above_src;
unsigned char *left_src;
unsigned char *above_ref;
unsigned char *left_ref;
unsigned int score;
#if CONFIG_SUBPELREFMV
unsigned int sse;
uint8_t *above_src;
uint8_t *above_ref;
#if !CONFIG_ABOVESPREFMV
uint8_t *left_src;
uint8_t *left_ref;
#endif
unsigned int ref_scores[MAX_MV_REFS] = {0};
int_mv sorted_mvs[MAX_MV_REFS];
unsigned int score;
unsigned int sse;
unsigned int ref_scores[MAX_MV_REF_CANDIDATES] = {0};
int_mv sorted_mvs[MAX_MV_REF_CANDIDATES];
int zero_seen = FALSE;
// Default all to 0,0 if nothing else available
best_mv->as_int = nearest->as_int = near->as_int = 0;
nearest->as_int = near->as_int = 0;
vpx_memset(sorted_mvs, 0, sizeof(sorted_mvs));
#if CONFIG_SUBPELREFMV
above_src = xd->dst.y_buffer - xd->dst.y_stride * 2;
left_src = xd->dst.y_buffer - 2;
above_ref = ref_y_buffer - ref_y_stride * 2;
left_ref = ref_y_buffer - 2;
#if CONFIG_ABOVESPREFMV
above_src -= 4;
above_ref -= 4;
#else
above_src = xd->dst.y_buffer - xd->dst.y_stride * 3;
left_src = xd->dst.y_buffer - 3;
above_ref = ref_y_buffer - ref_y_stride * 3;
left_ref = ref_y_buffer - 3;
left_src = xd->dst.y_buffer - 2;
left_ref = ref_y_buffer - 2;
#endif
//for(i = 0; i < MAX_MV_REFS; ++i) {
// Limit search to the predicted best 4
for(i = 0; i < 4; ++i) {
// 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;
@ -175,42 +170,61 @@ void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
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
#if CONFIG_SUBPELREFMV
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_c(above_ref + offset, ref_y_stride,
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, xd->dst.y_stride, &sse);
score += sse;
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
vp9_sub_pixel_variance16x2_c(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);
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;
}
#endif
}
#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 CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
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),
@ -219,36 +233,22 @@ void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
xd->dst.y_stride, &sse);
score += sse;
}
#endif
}
#else
row_offset = (this_mv.as_mv.row > 0) ?
((this_mv.as_mv.row + 3) >> 3):((this_mv.as_mv.row + 4) >> 3);
col_offset = (this_mv.as_mv.col > 0) ?
((this_mv.as_mv.col + 3) >> 3):((this_mv.as_mv.col + 4) >> 3);
offset = ref_y_stride * row_offset + col_offset;
score = 0;
if (xd->up_available) {
score += vp9_sad16x3(above_src, xd->dst.y_stride,
above_ref + offset, ref_y_stride);
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
score += vp9_sad16x3(above_src + 16, xd->dst.y_stride,
above_ref + offset + 16, ref_y_stride);
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
}
if (xd->left_available) {
score += vp9_sad3x16(left_src, xd->dst.y_stride,
left_ref + offset, ref_y_stride);
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
score += vp9_sad3x16(left_src + xd->dst.y_stride * 16,
xd->dst.y_stride,
left_ref + offset + ref_y_stride * 16,
ref_y_stride);
}
#endif
}
#endif
// Add the entry to our list and then resort the list on score.
@ -268,14 +268,11 @@ void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
}
// Make sure all the candidates are properly clamped etc
for (i = 0; i < 4; ++i) {
for (i = 0; i < MAX_MV_REF_CANDIDATES; ++i) {
lower_mv_precision(&sorted_mvs[i], xd->allow_high_precision_mv);
clamp_mv2(&sorted_mvs[i], xd);
}
// Set the best mv to the first entry in the sorted list
best_mv->as_int = sorted_mvs[0].as_int;
// 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

7
vp9/common/vp9_findnearmv.h
View File

@ -22,10 +22,9 @@
* score to use as ref motion vector
*/
void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
unsigned char *ref_y_buffer,
uint8_t *ref_y_buffer,
int ref_y_stride,
int_mv *mvlist,
int_mv *best_mv,
int_mv *nearest,
int_mv *near);
@ -82,7 +81,7 @@ vp9_prob *vp9_mv_ref_probs(VP9_COMMON *pc,
vp9_prob p[VP9_MVREFS - 1],
const int context);
extern const unsigned char vp9_mbsplit_offset[4][16];
extern const uint8_t vp9_mbsplit_offset[4][16];
static int left_block_mv(const MODE_INFO *cur_mb, int b) {
if (!(b & 3)) {
@ -182,4 +181,4 @@ static B_PREDICTION_MODE above_block_mode(const MODE_INFO *cur_mb,
return (cur_mb->bmi + b - 4)->as_mode.first;
}
#endif
#endif // VP9_COMMON_VP9_FINDNEARMV_H_

4
vp9/common/vp9_header.h
View File

@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_HEADER_H_
#define VP9_COMMON_VP9_HEADER_H_
@ -38,5 +37,4 @@ typedef struct {
#define VP9_HEADER_SIZE 3
#endif
#endif
#endif // VP9_COMMON_VP9_HEADER_H_

1670
vp9/common/vp9_idctllm.c
View File

File diff suppressed because it is too large Load Diff

4
vp9/common/vp9_implicit_segmentation.c
View File

@ -33,8 +33,8 @@ typedef struct {
int min_y;
int max_x;
int max_y;
long long sum_x;
long long sum_y;
int64_t sum_x;
int64_t sum_y;
int pixels;
int seg_value;
int label;

17
vp9/common/vp9_invtrans.c
View File

@ -72,7 +72,7 @@ void vp9_inverse_transform_mb_4x4(MACROBLOCKD *xd) {
vp9_inverse_transform_mbuv_4x4(xd);
}
void vp9_inverse_transform_b_8x8(short *input_dqcoeff, short *output_coeff,
void vp9_inverse_transform_b_8x8(int16_t *input_dqcoeff, int16_t *output_coeff,
int pitch) {
vp9_short_idct8x8(input_dqcoeff, output_coeff, pitch);
}
@ -125,8 +125,8 @@ void vp9_inverse_transform_mb_8x8(MACROBLOCKD *xd) {
vp9_inverse_transform_mbuv_8x8(xd);
}
void vp9_inverse_transform_b_16x16(short *input_dqcoeff,
short *output_coeff, int pitch) {
void vp9_inverse_transform_b_16x16(int16_t *input_dqcoeff,
int16_t *output_coeff, int pitch) {
vp9_short_idct16x16(input_dqcoeff, output_coeff, pitch);
}
@ -145,3 +145,14 @@ void vp9_inverse_transform_mb_16x16(MACROBLOCKD *xd) {
vp9_inverse_transform_mby_16x16(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_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);
}

14
vp9/common/vp9_invtrans.h
View File

@ -12,6 +12,7 @@
#define VP9_COMMON_VP9_INVTRANS_H_
#include "./vpx_config.h"
#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_blockd.h"
extern void vp9_inverse_transform_b_4x4(MACROBLOCKD *xd, int block, int pitch);
@ -22,8 +23,8 @@ extern void vp9_inverse_transform_mby_4x4(MACROBLOCKD *xd);
extern void vp9_inverse_transform_mbuv_4x4(MACROBLOCKD *xd);
extern void vp9_inverse_transform_b_8x8(short *input_dqcoeff,
short *output_coeff, int pitch);
extern 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);
@ -31,11 +32,14 @@ extern void vp9_inverse_transform_mby_8x8(MACROBLOCKD *xd);
extern void vp9_inverse_transform_mbuv_8x8(MACROBLOCKD *xd);
extern void vp9_inverse_transform_b_16x16(short *input_dqcoeff,
short *output_coeff, int pitch);
extern 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);
extern void vp9_inverse_transform_mby_16x16(MACROBLOCKD *xd);
#endif // __INC_INVTRANS_H
extern void vp9_inverse_transform_sby_32x32(SUPERBLOCKD *xd_sb);
extern void vp9_inverse_transform_sbuv_16x16(SUPERBLOCKD *xd_sb);
#endif // VP9_COMMON_VP9_INVTRANS_H_

280
vp9/common/vp9_loopfilter.c
View File

@ -176,46 +176,70 @@ void vp9_loop_filter_frame_init(VP9_COMMON *cm,
}
}
void vp9_loop_filter_frame(VP9_COMMON *cm, MACROBLOCKD *xd) {
// Determine if we should skip inner-MB loop filtering within a MB
// The current condition is that the loop filtering is skipped only
// the MB uses a prediction size of 16x16 and either 16x16 transform
// is used or there is no residue at all.
static int mb_lf_skip(const MB_MODE_INFO *const mbmi) {
const MB_PREDICTION_MODE mode = mbmi->mode;
const int skip_coef = mbmi->mb_skip_coeff;
const int tx_size = mbmi->txfm_size;
return mode != B_PRED && mode != I8X8_PRED && mode != SPLITMV &&
(tx_size >= TX_16X16 || skip_coef);
}
// Determine if we should skip MB loop filtering on a MB edge within
// a superblock, the current condition is that MB loop filtering is
// skipped only when both MBs do not use inner MB loop filtering, and
// same motion vector with same reference frame
static int sb_mb_lf_skip(const MODE_INFO *const mip0,
const MODE_INFO *const mip1) {
const MB_MODE_INFO *mbmi0 = &mip0->mbmi;
const MB_MODE_INFO *mbmi1 = &mip0->mbmi;
return mb_lf_skip(mbmi0) && mb_lf_skip(mbmi1) &&
(mbmi0->ref_frame == mbmi1->ref_frame) &&
(mbmi0->mv[mbmi0->ref_frame].as_int ==
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) {
YV12_BUFFER_CONFIG *post = cm->frame_to_show;
loop_filter_info_n *lfi_n = &cm->lf_info;
struct loop_filter_info lfi;
FRAME_TYPE frame_type = cm->frame_type;
int mb_row;
int mb_col;
int filter_level;
unsigned char *y_ptr, *u_ptr, *v_ptr;
const FRAME_TYPE frame_type = cm->frame_type;
int mb_row, mb_col;
uint8_t *y_ptr, *u_ptr, *v_ptr;
/* Point at base of Mb MODE_INFO list */
const MODE_INFO *mode_info_context = cm->mi;
const int mis = cm->mode_info_stride;
/* Initialize the loop filter for this frame. */
vp9_loop_filter_frame_init(cm, xd, cm->filter_level);
vp9_loop_filter_frame_init(cm, xd, frame_filter_level);
/* Set up the buffer pointers */
y_ptr = post->y_buffer;
u_ptr = post->u_buffer;
v_ptr = post->v_buffer;
if (y_only) {
u_ptr = 0;
v_ptr = 0;
} else {
u_ptr = post->u_buffer;
v_ptr = post->v_buffer;
}
/* vp9_filter each macro block */
for (mb_row = 0; mb_row < cm->mb_rows; mb_row++) {
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++) {
int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
mode_info_context->mbmi.mode != I8X8_PRED &&
mode_info_context->mbmi.mode != SPLITMV &&
mode_info_context->mbmi.mb_skip_coeff);
const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
const MB_PREDICTION_MODE mode = mode_info_context->mbmi.mode;
const int mode_index = lfi_n->mode_lf_lut[mode];
const int seg = mode_info_context->mbmi.segment_id;
const int ref_frame = mode_info_context->mbmi.ref_frame;
int tx_type = mode_info_context->mbmi.txfm_size;
filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
const int filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
if (filter_level) {
const int skip_lf = mb_lf_skip(&mode_info_context->mbmi);
const int tx_size = mode_info_context->mbmi.txfm_size;
if (cm->filter_type == NORMAL_LOOPFILTER) {
const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
lfi.mblim = lfi_n->mblim[filter_level];
@ -223,198 +247,102 @@ void vp9_loop_filter_frame(VP9_COMMON *cm, MACROBLOCKD *xd) {
lfi.lim = lfi_n->lim[filter_level];
lfi.hev_thr = lfi_n->hev_thr[hev_index];
if (mb_col > 0
#if CONFIG_SUPERBLOCKS
&& !((mb_col & 1) && mode_info_context->mbmi.encoded_as_sb &&
mode_info_context[0].mbmi.mb_skip_coeff &&
mode_info_context[-1].mbmi.mb_skip_coeff)
#endif
)
vp9_loop_filter_mbv(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
if (!skip_lf && tx_type != TX_16X16) {
if (tx_type == TX_8X8)
vp9_loop_filter_bv8x8(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
if (mb_col > 0 &&
!((mb_col & 1) && mode_info_context->mbmi.sb_type &&
(sb_mb_lf_skip(mode_info_context - 1, mode_info_context) ||
tx_size >= TX_32X32))
) {
if (tx_size >= TX_16X16)
vp9_lpf_mbv_w(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
else
vp9_loop_filter_mbv(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
}
if (!skip_lf) {
if (tx_size >= TX_8X8) {
if (tx_size == TX_8X8 && (mode == I8X8_PRED || mode == SPLITMV))
vp9_loop_filter_bv8x8(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
else
vp9_loop_filter_bv8x8(y_ptr, NULL, NULL, post->y_stride,
post->uv_stride, &lfi);
} else {
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
#if CONFIG_SUPERBLOCKS
&& !((mb_row & 1) && mode_info_context->mbmi.encoded_as_sb &&
mode_info_context[0].mbmi.mb_skip_coeff &&
mode_info_context[-cm->mode_info_stride].mbmi.mb_skip_coeff)
#endif
)
vp9_loop_filter_mbh(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
if (!skip_lf && tx_type != TX_16X16) {
if (tx_type == TX_8X8)
vp9_loop_filter_bh8x8(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
if (mb_row > 0 &&
!((mb_row & 1) && mode_info_context->mbmi.sb_type &&
(sb_mb_lf_skip(mode_info_context - mis, mode_info_context) ||
tx_size >= TX_32X32))
) {
if (tx_size >= TX_16X16)
vp9_lpf_mbh_w(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
else
vp9_loop_filter_mbh(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
}
if (!skip_lf) {
if (tx_size >= TX_8X8) {
if (tx_size == TX_8X8 && (mode == I8X8_PRED || mode == SPLITMV))
vp9_loop_filter_bh8x8(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
else
vp9_loop_filter_bh8x8(y_ptr, NULL, NULL, post->y_stride,
post->uv_stride, &lfi);
} else {
vp9_loop_filter_bh(y_ptr, u_ptr, v_ptr, post->y_stride,
post->uv_stride, &lfi);
}
}
} else {
// FIXME: Not 8x8 aware
if (mb_col > 0
#if CONFIG_SUPERBLOCKS
&& !((mb_col & 1) && mode_info_context->mbmi.encoded_as_sb &&
mode_info_context[0].mbmi.mb_skip_coeff &&
mode_info_context[-1].mbmi.mb_skip_coeff)
#endif
)
if (mb_col > 0 &&
!(skip_lf && mb_lf_skip(&mode_info_context[-1].mbmi)) &&
!((mb_col & 1) && mode_info_context->mbmi.sb_type))
vp9_loop_filter_simple_mbv(y_ptr, post->y_stride,
lfi_n->mblim[filter_level]);
if (!skip_lf)
vp9_loop_filter_simple_bv(y_ptr, post->y_stride,
lfi_n->blim[filter_level]);
/* don't apply across umv border */
if (mb_row > 0
#if CONFIG_SUPERBLOCKS
&& !((mb_row & 1) && mode_info_context->mbmi.encoded_as_sb &&
mode_info_context[0].mbmi.mb_skip_coeff &&
mode_info_context[-cm->mode_info_stride].mbmi.mb_skip_coeff)
#endif
)
if (mb_row > 0 &&
!(skip_lf && mb_lf_skip(&mode_info_context[-mis].mbmi)) &&
!((mb_row & 1) && mode_info_context->mbmi.sb_type))
vp9_loop_filter_simple_mbh(y_ptr, post->y_stride,
lfi_n->mblim[filter_level]);
if (!skip_lf)
vp9_loop_filter_simple_bh(y_ptr, post->y_stride,
lfi_n->blim[filter_level]);
}
}
y_ptr += 16;
u_ptr += 8;
v_ptr += 8;
if (!y_only) {
u_ptr += 8;
v_ptr += 8;
}
mode_info_context++; /* step to next MB */
}
y_ptr += post->y_stride * 16 - post->y_width;
u_ptr += post->uv_stride * 8 - post->uv_width;
v_ptr += post->uv_stride * 8 - post->uv_width;
if (!y_only) {
u_ptr += post->uv_stride * 8 - post->uv_width;
v_ptr += post->uv_stride * 8 - post->uv_width;
}
mode_info_context++; /* Skip border mb */
}
}
void vp9_loop_filter_frame_yonly(VP9_COMMON *cm, MACROBLOCKD *xd,
int default_filt_lvl) {
YV12_BUFFER_CONFIG *post = cm->frame_to_show;
unsigned char *y_ptr;
int mb_row;
int mb_col;
loop_filter_info_n *lfi_n = &cm->lf_info;
struct loop_filter_info lfi;
int filter_level;
FRAME_TYPE frame_type = cm->frame_type;
/* Point at base of Mb MODE_INFO list */
const MODE_INFO *mode_info_context = cm->mi;
#if 0
if (default_filt_lvl == 0) /* no filter applied */
return;
#endif
/* Initialize the loop filter for this frame. */
vp9_loop_filter_frame_init(cm, xd, default_filt_lvl);
/* Set up the buffer pointers */
y_ptr = post->y_buffer;
/* vp9_filter each macro block */
for (mb_row = 0; mb_row < cm->mb_rows; mb_row++) {
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++) {
int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
mode_info_context->mbmi.mode != I8X8_PRED &&
mode_info_context->mbmi.mode != SPLITMV &&
mode_info_context->mbmi.mb_skip_coeff);
const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
const int seg = mode_info_context->mbmi.segment_id;
const int ref_frame = mode_info_context->mbmi.ref_frame;
int tx_type = mode_info_context->mbmi.txfm_size;
filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
if (filter_level) {
if (cm->filter_type == NORMAL_LOOPFILTER) {
const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
lfi.mblim = lfi_n->mblim[filter_level];
lfi.blim = lfi_n->blim[filter_level];
lfi.lim = lfi_n->lim[filter_level];
lfi.hev_thr = lfi_n->hev_thr[hev_index];
if (mb_col > 0)
vp9_loop_filter_mbv(y_ptr, 0, 0, post->y_stride, 0, &lfi);
if (!skip_lf && tx_type != TX_16X16) {
if (tx_type == TX_8X8)
vp9_loop_filter_bv8x8(y_ptr, 0, 0, post->y_stride, 0, &lfi);
else
vp9_loop_filter_bv(y_ptr, 0, 0, post->y_stride, 0, &lfi);
}
/* don't apply across umv border */
if (mb_row > 0)
vp9_loop_filter_mbh(y_ptr, 0, 0, post->y_stride, 0, &lfi);
if (!skip_lf && tx_type != TX_16X16) {
if (tx_type == TX_8X8)
vp9_loop_filter_bh8x8(y_ptr, 0, 0, post->y_stride, 0, &lfi);
else
vp9_loop_filter_bh(y_ptr, 0, 0, post->y_stride, 0, &lfi);
}
} else {
// FIXME: Not 8x8 aware
if (mb_col > 0)
vp9_loop_filter_simple_mbv(y_ptr, post->y_stride,
lfi_n->mblim[filter_level]);
if (!skip_lf)
vp9_loop_filter_simple_bv(y_ptr, post->y_stride,
lfi_n->blim[filter_level]);
/* don't apply across umv border */
if (mb_row > 0)
vp9_loop_filter_simple_mbh(y_ptr, post->y_stride,
lfi_n->mblim[filter_level]);
if (!skip_lf)
vp9_loop_filter_simple_bh(y_ptr, post->y_stride,
lfi_n->blim[filter_level]);
}
}
y_ptr += 16;
mode_info_context++; /* step to next MB */
}
y_ptr += post->y_stride * 16 - post->y_width;
mode_info_context++; /* Skip border mb */
}
}
void vp9_loop_filter_partial_frame(VP9_COMMON *cm, MACROBLOCKD *xd,
int default_filt_lvl) {
YV12_BUFFER_CONFIG *post = cm->frame_to_show;
unsigned char *y_ptr;
uint8_t *y_ptr;
int mb_row;
int mb_col;
int mb_cols = post->y_width >> 4;

32
vp9/common/vp9_loopfilter.h
View File

@ -49,26 +49,26 @@ struct loop_filter_info {
};
#define prototype_loopfilter(sym) \
void sym(unsigned char *src, int pitch, const unsigned char *blimit,\
void sym(uint8_t *src, int pitch, const unsigned char *blimit, \
const unsigned char *limit, const unsigned char *thresh, int count)
#define prototype_loopfilter_block(sym) \
void sym(unsigned char *y, unsigned char *u, unsigned char *v, \
void sym(uint8_t *y, uint8_t *u, uint8_t *v, \
int ystride, int uv_stride, struct loop_filter_info *lfi)
#define prototype_simple_loopfilter(sym) \
void sym(unsigned char *y, int ystride, const unsigned char *blimit)
void sym(uint8_t *y, int ystride, const unsigned char *blimit)
#if ARCH_X86 || ARCH_X86_64
#include "x86/vp9_loopfilter_x86.h"
#endif
typedef void loop_filter_uvfunction(unsigned char *u, /* source pointer */
typedef void loop_filter_uvfunction(uint8_t *u, /* source pointer */
int p, /* pitch */
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
unsigned char *v);
uint8_t *v);
/* assorted loopfilter functions which get used elsewhere */
struct VP9Common;
@ -80,17 +80,27 @@ void vp9_loop_filter_frame_init(struct VP9Common *cm,
struct macroblockd *mbd,
int default_filt_lvl);
void vp9_loop_filter_frame(struct VP9Common *cm, struct macroblockd *mbd);
void vp9_loop_filter_frame(struct VP9Common *cm,
struct macroblockd *mbd,
int filter_level,
int y_only);
void vp9_loop_filter_partial_frame(struct VP9Common *cm,
struct macroblockd *mbd,
int default_filt_lvl);
void vp9_loop_filter_frame_yonly(struct VP9Common *cm,
struct macroblockd *mbd,
int default_filt_lvl);
void vp9_loop_filter_update_sharpness(loop_filter_info_n *lfi,
int sharpness_lvl);
#endif // loopfilter_h
void vp9_mb_lpf_horizontal_edge_w(unsigned char *s, int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
int count);
void vp9_mb_lpf_vertical_edge_w(unsigned char *s, int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
int count);
#endif // VP9_COMMON_VP9_LOOPFILTER_H_

679
vp9/common/vp9_loopfilter_filters.c
View File

@ -13,20 +13,20 @@
#include "vp9/common/vp9_loopfilter.h"
#include "vp9/common/vp9_onyxc_int.h"
typedef unsigned char uc;
static __inline signed char 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 (signed char) t;
return (int8_t) t;
}
/* should we apply any filter at all ( 11111111 yes, 00000000 no) */
static __inline signed char filter_mask(uc limit, uc blimit,
uc p3, uc p2, uc p1, uc p0,
uc q0, uc q1, uc q2, uc q3) {
signed char mask = 0;
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;
mask |= (abs(p1 - p0) > limit) * -1;
@ -39,26 +39,25 @@ static __inline signed char filter_mask(uc limit, uc blimit,
}
/* is there high variance internal edge ( 11111111 yes, 00000000 no) */
static __inline signed char hevmask(uc thresh, uc p1, uc p0, uc q0, uc q1) {
signed char hev = 0;
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(signed char mask, uc hev, uc *op1,
uc *op0, uc *oq0, uc *oq1)
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;
{
signed char ps0, qs0;
signed char ps1, qs1;
signed char filter, Filter1, Filter2;
signed char u;
ps1 = (signed char) * op1 ^ 0x80;
ps0 = (signed char) * op0 ^ 0x80;
qs0 = (signed char) * oq0 ^ 0x80;
qs1 = (signed char) * oq1 ^ 0x80;
ps1 = (int8_t) *op1 ^ 0x80;
ps0 = (int8_t) *op0 ^ 0x80;
qs0 = (int8_t) *oq0 ^ 0x80;
qs1 = (int8_t) *oq1 ^ 0x80;
/* add outer taps if we have high edge variance */
filter = signed_char_clamp(ps1 - qs1);
@ -91,20 +90,16 @@ static __inline void filter(signed char mask, uc hev, uc *op1,
*oq1 = u ^ 0x80;
u = signed_char_clamp(ps1 + filter);
*op1 = u ^ 0x80;
}
void vp9_loop_filter_horizontal_edge_c
(
unsigned char *s,
int p, /* pitch */
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
int count
) {
int hev = 0; /* high edge variance */
signed char mask = 0;
void vp9_loop_filter_horizontal_edge_c(uint8_t *s,
int p, /* pitch */
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
int count) {
int hev = 0; /* high edge variance */
int8_t mask = 0;
int i = 0;
/* loop filter designed to work using chars so that we can make maximum use
@ -123,14 +118,14 @@ void vp9_loop_filter_horizontal_edge_c
} while (++i < count * 8);
}
void vp9_loop_filter_vertical_edge_c(unsigned char *s,
void vp9_loop_filter_vertical_edge_c(uint8_t *s,
int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
int count) {
int hev = 0; /* high edge variance */
signed char mask = 0;
int8_t mask = 0;
int i = 0;
/* loop filter designed to work using chars so that we can make maximum use
@ -148,27 +143,409 @@ void vp9_loop_filter_vertical_edge_c(unsigned char *s,
s += p;
} while (++i < count * 8);
}
static __inline signed char flatmask(uc thresh,
uc p4, uc p3, uc p2, uc p1, uc p0,
uc q0, uc q1, uc q2, uc q3, uc q4) {
signed char flat = 0;
flat |= (abs(p1 - p0) > 1) * -1;
flat |= (abs(q1 - q0) > 1) * -1;
flat |= (abs(p0 - p2) > 1) * -1;
flat |= (abs(q0 - q2) > 1) * -1;
flat |= (abs(p3 - p0) > 1) * -1;
flat |= (abs(q3 - q0) > 1) * -1;
flat |= (abs(p4 - p0) > 1) * -1;
flat |= (abs(q4 - q0) > 1) * -1;
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) {
int8_t flat = 0;
flat |= (abs(p1 - p0) > thresh) * -1;
flat |= (abs(q1 - q0) > thresh) * -1;
flat |= (abs(p0 - p2) > thresh) * -1;
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 void mbfilter(signed char mask, uc hev, uc flat,
uc *op4, uc *op3, uc *op2, uc *op1, uc *op0,
uc *oq0, uc *oq1, uc *oq2, uc *oq3, uc *oq4) {
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 */
if (flat && mask) {
uint8_t p0, q0;
uint8_t p1, q1;
uint8_t p2, q2;
uint8_t p3, q3;
uint8_t p4, q4;
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;
*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;
} else {
int8_t ps0, qs0;
int8_t ps1, qs1;
int8_t filter, Filter1, Filter2;
int8_t u;
ps1 = (int8_t) *op1 ^ 0x80;
ps0 = (int8_t) *op0 ^ 0x80;
qs0 = (int8_t) *oq0 ^ 0x80;
qs1 = (int8_t) *oq1 ^ 0x80;
/* add outer taps if we have high edge variance */
filter = signed_char_clamp(ps1 - qs1);
filter &= hev;
/* inner taps */
filter = signed_char_clamp(filter + 3 * (qs0 - ps0));
filter &= mask;
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;
/* 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;
}
}
void vp9_mbloop_filter_horizontal_edge_c(uint8_t *s,
int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
int count) {
int8_t hev = 0; /* high edge variance */
int8_t mask = 0;
int8_t flat = 0;
int i = 0;
/* loop filter designed to work using chars so that we can make maximum use
* of 8 bit simd instructions.
*/
do {
mask = filter_mask(limit[0], blimit[0],
s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p],
s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p]);
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]);
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;
} while (++i < count * 8);
}
void vp9_mbloop_filter_vertical_edge_c(uint8_t *s,
int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh,
int count) {
int8_t hev = 0; /* high edge variance */
int8_t mask = 0;
int8_t flat = 0;
int i = 0;
do {
mask = filter_mask(limit[0], blimit[0],
s[-4], s[-3], s[-2], s[-1],
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]);
mbfilter(mask, hev, flat,
s - 5, s - 4, s - 3, s - 2, s - 1,
s, s + 1, s + 2, s + 3, s + 4);
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 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;
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;
Filter2 = signed_char_clamp(filter + 3);
Filter2 >>= 3;
u = signed_char_clamp(p0 + Filter2);
*op0 = u ^ 0x80;
}
void vp9_loop_filter_simple_horizontal_edge_c(uint8_t *s,
int p,
const unsigned char *blimit) {
int8_t mask = 0;
int i = 0;
do {
mask = simple_filter_mask(blimit[0],
s[-2 * p], s[-1 * p],
s[0 * p], s[1 * p]);
simple_filter(mask,
s - 2 * p, s - 1 * p,
s, s + 1 * p);
++s;
} while (++i < 16);
}
void vp9_loop_filter_simple_vertical_edge_c(uint8_t *s,
int p,
const unsigned char *blimit) {
int8_t mask = 0;
int i = 0;
do {
mask = simple_filter_mask(blimit[0], s[-2], s[-1], s[0], s[1]);
simple_filter(mask, s - 2, s - 1, s, s + 1);
s += p;
} while (++i < 16);
}
/* Vertical MB Filtering */
void vp9_loop_filter_mbv_c(uint8_t *y_ptr, uint8_t *u_ptr,
uint8_t *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_vertical_edge_c(y_ptr, y_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
vp9_mbloop_filter_vertical_edge_c(u_ptr, uv_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 1);
if (v_ptr)
vp9_mbloop_filter_vertical_edge_c(v_ptr, uv_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 1);
}
/* Vertical B Filtering */
void vp9_loop_filter_bv_c(uint8_t*y_ptr, uint8_t *u_ptr,
uint8_t *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_loop_filter_vertical_edge_c(y_ptr + 4, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
vp9_loop_filter_vertical_edge_c(y_ptr + 8, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
vp9_loop_filter_vertical_edge_c(y_ptr + 12, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
vp9_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
if (v_ptr)
vp9_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
}
/* Horizontal MB filtering */
void vp9_loop_filter_mbh_c(uint8_t *y_ptr, uint8_t *u_ptr,
uint8_t *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_horizontal_edge_c(y_ptr, y_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
vp9_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 1);
if (v_ptr)
vp9_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 1);
}
/* Horizontal B Filtering */
void vp9_loop_filter_bh_c(uint8_t *y_ptr, uint8_t *u_ptr,
uint8_t *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_loop_filter_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
vp9_loop_filter_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
vp9_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
vp9_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
if (v_ptr)
vp9_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
}
void vp9_loop_filter_bh8x8_c(uint8_t *y_ptr, uint8_t *u_ptr,
uint8_t *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_horizontal_edge_c(
y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
vp9_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
if (v_ptr)
vp9_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
}
void vp9_loop_filter_bhs_c(uint8_t *y_ptr, int y_stride,
const unsigned char *blimit) {
vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 4 * y_stride,
y_stride, blimit);
vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 8 * y_stride,
y_stride, blimit);
vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 12 * y_stride,
y_stride, blimit);
}
void vp9_loop_filter_bv8x8_c(uint8_t *y_ptr, uint8_t *u_ptr,
uint8_t *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_vertical_edge_c(
y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
vp9_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
if (v_ptr)
vp9_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
}
void vp9_loop_filter_bvs_c(uint8_t *y_ptr, int y_stride,
const unsigned char *blimit) {
vp9_loop_filter_simple_vertical_edge_c(y_ptr + 4, y_stride, blimit);
vp9_loop_filter_simple_vertical_edge_c(y_ptr + 8, y_stride, blimit);
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 */
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;
*op6 = (p7 * 7 + p6 * 2 +
p5 + p4 + p3 + p2 + p1 + p0 + q0 + 8) >> 4;
*op5 = (p7 * 6 + p6 + p5 * 2 +
p4 + p3 + p2 + p1 + p0 + q0 + q1 + 8) >> 4;
*op4 = (p7 * 5 + p6 + p5 + p4 * 2 +
p3 + p2 + p1 + p0 + q0 + q1 + q2 + 8) >> 4;
*op3 = (p7 * 4 + p6 + p5 + p4 + p3 * 2 +
p2 + p1 + p0 + q0 + q1 + q2 + q3 + 8) >> 4;
*op2 = (p7 * 3 + p6 + p5 + p4 + p3 + p2 * 2 +
p1 + p0 + q0 + q1 + q2 + q3 + q4 + 8) >> 4;
*op1 = (p7 * 2 + p6 + p5 + p4 + p3 + p2 + p1 * 2 +
p0 + q0 + q1 + q2 + q3 + q4 + q5 + 8) >> 4;
*op0 = (p7 + p6 + p5 + p4 + p3 + p2 + p1 + p0 * 2 +
q0 + q1 + q2 + q3 + q4 + q5 + q6 + 8) >> 4;
*oq0 = (p6 + p5 + p4 + p3 + p2 + p1 + p0 + q0 * 2 +
q1 + q2 + q3 + q4 + q5 + q6 + q7 + 8) >> 4;
*oq1 = (p5 + p4 + p3 + p2 + p1 + p0 + q0 + q1 * 2 +
q2 + q3 + q4 + q5 + q6 + q7 * 2 + 8) >> 4;
*oq2 = (p4 + p3 + p2 + p1 + p0 + q0 + q1 + q2 * 2 +
q3 + q4 + q5 + q6 + q7 * 3 + 8) >> 4;
*oq3 = (p3 + p2 + p1 + p0 + q0 + q1 + q2 + q3 * 2 +
q4 + q5 + q6 + q7 * 4 + 8) >> 4;
*oq4 = (p2 + p1 + p0 + q0 + q1 + q2 + q3 + q4 * 2 +
q5 + q6 + q7 * 5 + 8) >> 4;
*oq5 = (p1 + p0 + q0 + q1 + q2 + q3 + q4 + q5 * 2 +
q6 + q7 * 6 + 8) >> 4;
*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;
@ -233,7 +610,8 @@ static __inline void mbfilter(signed char mask, uc hev, uc flat,
*op1 = u ^ 0x80;
}
}
void vp9_mbloop_filter_horizontal_edge_c
void vp9_mb_lpf_horizontal_edge_w
(
unsigned char *s,
int p,
@ -245,31 +623,37 @@ void vp9_mbloop_filter_horizontal_edge_c
signed char hev = 0; /* high edge variance */
signed char mask = 0;
signed char flat = 0;
signed char flat2 = 0;
int i = 0;
/* loop filter designed to work using chars so that we can make maximum use
* of 8 bit simd instructions.
*/
do {
mask = filter_mask(limit[0], blimit[0],
s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p],
s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p]);
hev = hevmask(thresh[0], s[-2 * p], s[-1 * p], s[0 * p], s[1 * p]);
flat = flatmask(thresh[0],
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]);
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);
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]);
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;
} while (++i < count * 8);
}
void vp9_mbloop_filter_vertical_edge_c
void vp9_mb_lpf_vertical_edge_w
(
unsigned char *s,
int p,
@ -281,106 +665,35 @@ void vp9_mbloop_filter_vertical_edge_c
signed char hev = 0; /* high edge variance */
signed char mask = 0;
signed char flat = 0;
signed char flat2 = 0;
int i = 0;
do {
mask = filter_mask(limit[0], blimit[0],
s[-4], s[-3], s[-2], s[-1],
s[0], s[1], s[2], s[3]);
hev = hevmask(thresh[0], s[-2], s[-1], s[0], s[1]);
flat = flatmask(thresh[0],
flat = flatmask(1,
s[-5], s[-4], s[-3], s[-2], s[-1],
s[ 0], s[ 1], s[ 2], s[ 3], s[ 4]);
mbfilter(mask, hev, flat,
s - 5, s - 4, s - 3, s - 2, s - 1,
s, 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]);
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 += p;
} while (++i < count * 8);
}
/* should we apply any filter at all ( 11111111 yes, 00000000 no) */
static __inline signed char simple_filter_mask(uc blimit,
uc p1, uc p0,
uc q0, uc q1) {
/* Why does this cause problems for win32?
* error C2143: syntax error : missing ';' before 'type'
* (void) limit;
*/
signed char mask = (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 <= blimit) * -1;
return mask;
}
static __inline void simple_filter(signed char mask,
uc *op1, uc *op0,
uc *oq0, uc *oq1) {
signed char filter, Filter1, Filter2;
signed char p1 = (signed char) * op1 ^ 0x80;
signed char p0 = (signed char) * op0 ^ 0x80;
signed char q0 = (signed char) * oq0 ^ 0x80;
signed char q1 = (signed char) * oq1 ^ 0x80;
signed char u;
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;
Filter2 = signed_char_clamp(filter + 3);
Filter2 >>= 3;
u = signed_char_clamp(p0 + Filter2);
*op0 = u ^ 0x80;
}
void vp9_loop_filter_simple_horizontal_edge_c
(
unsigned char *s,
int p,
const unsigned char *blimit
) {
signed char mask = 0;
int i = 0;
do {
mask = simple_filter_mask(blimit[0],
s[-2 * p], s[-1 * p],
s[0 * p], s[1 * p]);
simple_filter(mask,
s - 2 * p, s - 1 * p,
s, s + 1 * p);
++s;
} while (++i < 16);
}
void vp9_loop_filter_simple_vertical_edge_c
(
unsigned char *s,
int p,
const unsigned char *blimit
) {
signed char mask = 0;
int i = 0;
do {
mask = simple_filter_mask(blimit[0], s[-2], s[-1], s[0], s[1]);
simple_filter(mask, s - 2, s - 1, s, s + 1);
s += p;
} while (++i < 16);
}
/* Vertical MB Filtering */
void vp9_loop_filter_mbv_c(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_vertical_edge_c(y_ptr, y_stride,
void vp9_lpf_mbv_w_c(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mb_lpf_vertical_edge_w(y_ptr, y_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
@ -391,32 +704,10 @@ void vp9_loop_filter_mbv_c(unsigned char *y_ptr, unsigned char *u_ptr,
vp9_mbloop_filter_vertical_edge_c(v_ptr, uv_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 1);
}
/* Vertical B Filtering */
void vp9_loop_filter_bv_c(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_loop_filter_vertical_edge_c(y_ptr + 4, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
vp9_loop_filter_vertical_edge_c(y_ptr + 8, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
vp9_loop_filter_vertical_edge_c(y_ptr + 12, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
vp9_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
if (v_ptr)
vp9_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
}
/* Horizontal MB filtering */
void vp9_loop_filter_mbh_c(unsigned char *y_ptr, unsigned char *u_ptr,
void vp9_lpf_mbh_w_c(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_horizontal_edge_c(y_ptr, y_stride,
vp9_mb_lpf_horizontal_edge_w(y_ptr, y_stride,
lfi->mblim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
@ -428,53 +719,3 @@ void vp9_loop_filter_mbh_c(unsigned char *y_ptr, unsigned char *u_ptr,
lfi->mblim, lfi->lim, lfi->hev_thr, 1);
}
/* Horizontal B Filtering */
void vp9_loop_filter_bh_c(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_loop_filter_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
vp9_loop_filter_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
vp9_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 2);
if (u_ptr)
vp9_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
if (v_ptr)
vp9_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr, 1);
}
void vp9_loop_filter_bh8x8_c(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_horizontal_edge_c(
y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
}
void vp9_loop_filter_bhs_c(unsigned char *y_ptr, int y_stride,
const unsigned char *blimit) {
vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 4 * y_stride,
y_stride, blimit);
vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 8 * y_stride,
y_stride, blimit);
vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 12 * y_stride,
y_stride, blimit);
}
void vp9_loop_filter_bv8x8_c(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_vertical_edge_c(
y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
}
void vp9_loop_filter_bvs_c(unsigned char *y_ptr, int y_stride,
const unsigned char *blimit) {
vp9_loop_filter_simple_vertical_edge_c(y_ptr + 4, y_stride, blimit);
vp9_loop_filter_simple_vertical_edge_c(y_ptr + 8, y_stride, blimit);
vp9_loop_filter_simple_vertical_edge_c(y_ptr + 12, y_stride, blimit);
}

25
vp9/common/vp9_mbpitch.c
View File

@ -16,17 +16,13 @@ typedef enum {
DEST = 1
} BLOCKSET;
static void setup_block
(
BLOCKD *b,
int mv_stride,
unsigned char **base,
unsigned char **base2,
int Stride,
int offset,
BLOCKSET bs
) {
static void setup_block(BLOCKD *b,
int mv_stride,
uint8_t **base,
uint8_t **base2,
int Stride,
int offset,
BLOCKSET bs) {
if (bs == DEST) {
b->dst_stride = Stride;
b->dst = offset;
@ -37,15 +33,13 @@ static void setup_block
b->base_pre = base;
b->base_second_pre = base2;
}
}
static void setup_macroblock(MACROBLOCKD *xd, BLOCKSET bs) {
int block;
unsigned char **y, **u, **v;
unsigned char **y2 = NULL, **u2 = NULL, **v2 = NULL;
uint8_t **y, **u, **v;
uint8_t **y2 = NULL, **u2 = NULL, **v2 = NULL;
BLOCKD *blockd = xd->block;
int stride;
@ -117,7 +111,6 @@ void vp9_setup_block_dptrs(MACROBLOCKD *xd) {
}
void vp9_build_block_doffsets(MACROBLOCKD *xd) {
/* handle the destination pitch features */
setup_macroblock(xd, DEST);
setup_macroblock(xd, PRED);

4
vp9/common/vp9_modecont.h
View File

@ -8,9 +8,9 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_MODECONT_H_
#define VP9_COMMON_VP9_MODECONT_H_
extern const int vp9_default_mode_contexts[INTER_MODE_CONTEXTS][4];
#endif
#endif // VP9_COMMON_VP9_MODECONT_H_

8
vp9/common/vp9_mv.h
View File

@ -8,14 +8,14 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_MV_H_
#define VP9_COMMON_VP9_MV_H_
#include "vpx/vpx_integer.h"
typedef struct {
short row;
short col;
int16_t row;
int16_t col;
} MV;
typedef union int_mv {
@ -23,4 +23,4 @@ typedef union int_mv {
MV as_mv;
} int_mv; /* facilitates faster equality tests and copies */
#endif
#endif // VP9_COMMON_VP9_MV_H_

284
vp9/common/vp9_mvref_common.c
View File

@ -17,14 +17,13 @@ static int mb_mv_ref_search[MVREF_NEIGHBOURS][2] = {
};
static int mb_ref_distance_weight[MVREF_NEIGHBOURS] =
{ 3, 3, 2, 1, 1, 1, 1, 1 };
#if CONFIG_SUPERBLOCKS
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 };
#endif
// clamp_mv
#define MV_BORDER (16 << 3) // Allow 16 pels in 1/8th pel units
static void clamp_mv(const MACROBLOCKD *xd, int_mv *mv) {
@ -40,10 +39,29 @@ static void clamp_mv(const MACROBLOCKD *xd, int_mv *mv) {
mv->as_mv.row = 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;
// Gets a best matching candidate refenence motion vector
// from the given mode info structure (if available)
static int get_candidate_mvref(
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 ret_val;
}
// 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,
@ -52,61 +70,29 @@ static int get_candidate_mvref(
int_mv *c2_mv
) {
int ret_val = FALSE;
c_mv->as_int = 0;
c2_mv->as_int = 0;
*c_ref_frame = INTRA_FRAME;
*c2_ref_frame = INTRA_FRAME;
// Target ref frame matches candidate first ref frame
if (ref_frame == candidate_mi->mbmi.ref_frame) {
c_mv->as_int = candidate_mi->mbmi.mv[0].as_int;
*c_ref_frame = ref_frame;
ret_val = TRUE;
// If first candidate not valid neither will be.
if (candidate_mi->mbmi.ref_frame > INTRA_FRAME) {
// First candidate
if (candidate_mi->mbmi.ref_frame != ref_frame) {
*c_ref_frame = candidate_mi->mbmi.ref_frame;
c_mv->as_int = candidate_mi->mbmi.mv[0].as_int;
}
// Is there a second non zero vector we can use.
// Second candidate
if ((candidate_mi->mbmi.second_ref_frame > INTRA_FRAME) &&
(candidate_mi->mbmi.mv[1].as_int != 0) &&
(candidate_mi->mbmi.mv[1].as_int != c_mv->as_int)) {
c2_mv->as_int = candidate_mi->mbmi.mv[1].as_int;
(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)) {
*c2_ref_frame = candidate_mi->mbmi.second_ref_frame;
}
// Target ref frame matches candidate second ref frame
} else if (ref_frame == candidate_mi->mbmi.second_ref_frame) {
c_mv->as_int = candidate_mi->mbmi.mv[1].as_int;
*c_ref_frame = ref_frame;
ret_val = TRUE;
// Is there a second non zero vector we can use.
if ((candidate_mi->mbmi.ref_frame > INTRA_FRAME) &&
(candidate_mi->mbmi.mv[0].as_int != 0) &&
(candidate_mi->mbmi.mv[0].as_int != c_mv->as_int)) {
c2_mv->as_int = candidate_mi->mbmi.mv[0].as_int;
*c2_ref_frame = candidate_mi->mbmi.ref_frame;
}
// No ref frame matches so use first ref mv as first choice
} else if (candidate_mi->mbmi.ref_frame > INTRA_FRAME) {
c_mv->as_int = candidate_mi->mbmi.mv[0].as_int;
*c_ref_frame = candidate_mi->mbmi.ref_frame;
ret_val = TRUE;
// Is there a second non zero vector we can use.
if ((candidate_mi->mbmi.second_ref_frame > INTRA_FRAME) &&
(candidate_mi->mbmi.mv[1].as_int != 0) &&
(candidate_mi->mbmi.mv[1].as_int != c_mv->as_int)) {
c2_mv->as_int = candidate_mi->mbmi.mv[1].as_int;
*c2_ref_frame = candidate_mi->mbmi.second_ref_frame;
}
// If only the second ref mv is valid:- (Should not trigger in current code
// base given current possible compound prediction options).
} else if (candidate_mi->mbmi.second_ref_frame > INTRA_FRAME) {
c_mv->as_int = candidate_mi->mbmi.mv[1].as_int;
*c_ref_frame = candidate_mi->mbmi.second_ref_frame;
ret_val = TRUE;
}
return ret_val;
}
// Performs mv adjustment based on reference frame and clamps the MV
@ -170,14 +156,20 @@ static void addmv_and_shuffle(
int weight
) {
int i = *index;
int i;
int insert_point;
int duplicate_found = FALSE;
// Check for duplicates. If there is one increment its score.
// Duplicate defined as being the same full pel vector with rounding.
// 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);
i = insert_point;
if (*index > i)
i++;
while (i > 0) {
i--;
if (candidate_mv.as_int == mv_list[i].as_int) {
duplicate_found = TRUE;
mv_scores[i] += weight;
@ -185,11 +177,13 @@ static void addmv_and_shuffle(
}
}
// If no duplicate was found add the new vector and give it a weight
if (!duplicate_found) {
mv_list[*index].as_int = candidate_mv.as_int;
mv_scores[*index] = weight;
i = *index;
// If no duplicate and the new candidate is good enough then add it.
if (!duplicate_found ) {
if (weight > mv_scores[insert_point]) {
mv_list[insert_point].as_int = candidate_mv.as_int;
mv_scores[insert_point] = weight;
i = insert_point;
}
(*index)++;
}
@ -224,38 +218,32 @@ void vp9_find_mv_refs(
int i;
MODE_INFO *candidate_mi;
MB_MODE_INFO * mbmi = &xd->mode_info_context->mbmi;
int_mv candidate_mvs[MAX_MV_REFS];
int_mv candidate_mvs[MAX_MV_REF_CANDIDATES];
int_mv c_refmv;
MV_REFERENCE_FRAME c_ref_frame;
int_mv c2_refmv;
MV_REFERENCE_FRAME c_ref_frame;
MV_REFERENCE_FRAME c2_ref_frame;
int candidate_scores[MAX_MV_REFS];
int candidate_scores[MAX_MV_REF_CANDIDATES];
int index = 0;
int split_count = 0;
int ref_weight = 0;
int valid_mv_ref;
int (*mv_ref_search)[2];
int *ref_distance_weight;
// Blank the reference vector lists and other local structures.
vpx_memset(mv_ref_list, 0, sizeof(int_mv) * MAX_MV_REFS);
vpx_memset(candidate_mvs, 0, sizeof(int_mv) * MAX_MV_REFS);
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 CONFIG_SUPERBLOCKS
if (mbmi->encoded_as_sb) {
if (mbmi->sb_type) {
mv_ref_search = sb_mv_ref_search;
ref_distance_weight = sb_ref_distance_weight;
} else {
mv_ref_search = mb_mv_ref_search;
ref_distance_weight = mb_ref_distance_weight;
}
#else
mv_ref_search = mb_mv_ref_search;
ref_distance_weight = mb_ref_distance_weight;
#endif
// Populate a list with candidate reference vectors from the
// spatial neighbours.
// 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) &&
((mv_ref_search[i][1] << 7) >= xd->mb_to_top_edge)) {
@ -263,95 +251,89 @@ void vp9_find_mv_refs(
candidate_mi = here + mv_ref_search[i][0] +
(mv_ref_search[i][1] * xd->mode_info_stride);
valid_mv_ref = get_candidate_mvref(candidate_mi, ref_frame,
&c_ref_frame, &c_refmv,
&c2_ref_frame, &c2_refmv);
// If there is a valid MV candidate then add it to the list
if (valid_mv_ref) {
scale_mv(xd, ref_frame, c_ref_frame, &c_refmv, ref_sign_bias );
ref_weight = ref_distance_weight[i] +
((c_ref_frame == ref_frame) << 4);
split_count += (candidate_mi->mbmi.mode == SPLITMV);
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_weight);
// If there is a second valid mv then add it as well.
if (c2_ref_frame > INTRA_FRAME) {
scale_mv(xd, ref_frame, c2_ref_frame, &c2_refmv, ref_sign_bias );
ref_weight = ref_distance_weight[i] +
((c2_ref_frame == ref_frame) << 4);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c2_refmv, ref_weight);
}
&index, c_refmv, ref_distance_weight[i] + 16);
}
split_count += (candidate_mi->mbmi.mode == SPLITMV);
}
}
// Look at the corresponding vector in the last frame
// Look in the last frame
candidate_mi = lf_here;
valid_mv_ref = get_candidate_mvref(candidate_mi, ref_frame,
&c_ref_frame, &c_refmv,
&c2_ref_frame, &c2_refmv);
// If there is a valid MV candidate then add it to the list
if (valid_mv_ref) {
scale_mv(xd, ref_frame, c_ref_frame, &c_refmv, ref_sign_bias );
ref_weight = 2 + ((c_ref_frame == ref_frame) << 4);
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_weight);
// If there is a second valid mv then add it as well.
if (c2_ref_frame > INTRA_FRAME) {
scale_mv(xd, ref_frame, c2_ref_frame, &c2_refmv, ref_sign_bias );
ref_weight = ref_distance_weight[i] +
((c2_ref_frame == ref_frame) << 4);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c2_refmv, ref_weight);
}
&index, c_refmv, 18);
}
// Populate a list with candidate reference vectors from the
// spatial neighbours.
for (i = 2; (i < MVREF_NEIGHBOURS) && (index < (MAX_MV_REFS - 2)); ++i) {
// 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) &&
((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);
valid_mv_ref = get_candidate_mvref(candidate_mi, ref_frame,
&c_ref_frame, &c_refmv,
&c2_ref_frame, &c2_refmv);
// If there is a valid MV candidate then add it to the list
if (valid_mv_ref) {
scale_mv(xd, ref_frame, c_ref_frame, &c_refmv, ref_sign_bias );
ref_weight = ref_distance_weight[i] +
((c_ref_frame == ref_frame) << 4);
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_weight);
// If there is a second valid mv then add it as well.
if (c2_ref_frame > INTRA_FRAME) {
scale_mv(xd, ref_frame, c2_ref_frame, &c2_refmv, ref_sign_bias );
ref_weight = ref_distance_weight[i] +
((c2_ref_frame == ref_frame) << 4);
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c2_refmv, ref_weight);
}
&index, c_refmv, ref_distance_weight[i] + 16);
}
}
}
// Make sure we are able to add 0,0
if (index > (MAX_MV_REFS - 1)) {
index = (MAX_MV_REFS - 1);
// If we have not found enough candidates consider ones where the
// 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)) {
for (i = 0; i < MVREF_NEIGHBOURS; ++i) {
if (((mv_ref_search[i][0] << 7) >= xd->mb_to_left_edge) &&
((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);
get_non_matching_candidates(candidate_mi, ref_frame,
&c_ref_frame, &c_refmv,
&c2_ref_frame, &c2_refmv);
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]);
}
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]);
}
}
if (index >= (MAX_MV_REF_CANDIDATES - 1)) {
break;
}
}
}
// Look at the last frame
if (index < (MAX_MV_REF_CANDIDATES - 1)) {
candidate_mi = lf_here;
get_non_matching_candidates(candidate_mi, ref_frame,
&c_ref_frame, &c_refmv,
&c2_ref_frame, &c2_refmv);
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);
}
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);
}
}
// Define inter mode coding context.
@ -383,14 +365,12 @@ void vp9_find_mv_refs(
}
// 0,0 is always a valid reference.
for (i = 0; i < index; ++i) {
for (i = 0; i < MAX_MV_REF_CANDIDATES; ++i) {
if (candidate_mvs[i].as_int == 0)
break;
}
if (i == index) {
c_refmv.as_int = 0;
addmv_and_shuffle(candidate_mvs, candidate_scores,
&index, c_refmv, candidate_scores[3]+1 );
if (i == MAX_MV_REF_CANDIDATES) {
candidate_mvs[MAX_MV_REF_CANDIDATES-1].as_int = 0;
}
// Copy over the candidate list.

18
vp9/common/vp9_mvref_common.h
View File

@ -11,18 +11,14 @@
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/common/vp9_blockd.h"
#ifndef VP9_COMMON_VP9_MVREF_COMMON_H_
#define VP9_COMMON_VP9_MVREF_COMMON_H_
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
);
#endif
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);
#endif // VP9_COMMON_VP9_MVREF_COMMON_H_

4
vp9/common/vp9_onyx.h
View File

@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_ONYX_H_
#define VP9_COMMON_VP9_ONYX_H_
@ -20,7 +19,6 @@ extern "C"
#include "vpx/internal/vpx_codec_internal.h"
#include "vpx/vp8cx.h"
#include "vpx_scale/yv12config.h"
#include "vp9/common/vp9_type_aliases.h"
#include "vp9/common/vp9_ppflags.h"
typedef int *VP9_PTR;
@ -222,4 +220,4 @@ extern "C"
}
#endif
#endif // __INC_ONYX_H
#endif // VP9_COMMON_VP9_ONYX_H_

93
vp9/common/vp9_onyxc_int.h
View File

@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_ONYXC_INT_H_
#define VP9_COMMON_VP9_ONYXC_INT_H_
@ -45,70 +44,51 @@ void vp9_initialize_common(void);
typedef struct frame_contexts {
vp9_prob bmode_prob[VP9_NKF_BINTRAMODES - 1];
vp9_prob ymode_prob[VP9_YMODES - 1]; /* interframe intra mode probs */
#if CONFIG_SUPERBLOCKS
vp9_prob sb_ymode_prob[VP9_I32X32_MODES - 1];
#endif
vp9_prob uv_mode_prob[VP9_YMODES][VP9_UV_MODES - 1];
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_prob coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob hybrid_coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob hybrid_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob hybrid_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
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];
nmv_context nmvc;
nmv_context pre_nmvc;
vp9_prob pre_bmode_prob[VP9_NKF_BINTRAMODES - 1];
vp9_prob pre_ymode_prob[VP9_YMODES - 1]; /* interframe intra mode probs */
#if CONFIG_SUPERBLOCKS
vp9_prob pre_sb_ymode_prob[VP9_I32X32_MODES - 1];
#endif
vp9_prob pre_uv_mode_prob[VP9_YMODES][VP9_UV_MODES - 1];
vp9_prob pre_i8x8_mode_prob[VP9_I8X8_MODES - 1];
vp9_prob pre_sub_mv_ref_prob[SUBMVREF_COUNT][VP9_SUBMVREFS - 1];
vp9_prob pre_mbsplit_prob[VP9_NUMMBSPLITS - 1];
unsigned int bmode_counts[VP9_NKF_BINTRAMODES];
unsigned int ymode_counts[VP9_YMODES]; /* interframe intra mode probs */
#if CONFIG_SUPERBLOCKS
unsigned int sb_ymode_counts[VP9_I32X32_MODES];
#endif
unsigned int uv_mode_counts[VP9_YMODES][VP9_UV_MODES];
unsigned int i8x8_mode_counts[VP9_I8X8_MODES]; /* interframe intra probs */
unsigned int sub_mv_ref_counts[SUBMVREF_COUNT][VP9_SUBMVREFS];
unsigned int mbsplit_counts[VP9_NUMMBSPLITS];
vp9_prob pre_coef_probs [BLOCK_TYPES] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob pre_hybrid_coef_probs [BLOCK_TYPES] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [ENTROPY_NODES];
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_prob pre_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob pre_hybrid_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob pre_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [ENTROPY_NODES];
vp9_prob pre_hybrid_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [ENTROPY_NODES];
unsigned int coef_counts [BLOCK_TYPES] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
unsigned int hybrid_coef_counts [BLOCK_TYPES] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
unsigned int coef_counts_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
unsigned int hybrid_coef_counts_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
unsigned int coef_counts_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
unsigned int hybrid_coef_counts_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS]
[PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS];
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];
nmv_context_counts NMVcount;
vp9_prob switchable_interp_prob[VP9_SWITCHABLE_FILTERS + 1]
@ -139,16 +119,17 @@ typedef enum {
ONLY_4X4 = 0,
ALLOW_8X8 = 1,
ALLOW_16X16 = 2,
TX_MODE_SELECT = 3,
NB_TXFM_MODES = 4,
ALLOW_32X32 = 3,
TX_MODE_SELECT = 4,
NB_TXFM_MODES = 5,
} TXFM_MODE;
typedef struct VP9Common {
struct vpx_internal_error_info error;
DECLARE_ALIGNED(16, short, Y1dequant[QINDEX_RANGE][16]);
DECLARE_ALIGNED(16, short, Y2dequant[QINDEX_RANGE][16]);
DECLARE_ALIGNED(16, short, UVdequant[QINDEX_RANGE][16]);
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;
@ -234,7 +215,7 @@ typedef struct VP9Common {
/* Y,U,V,Y2 */
ENTROPY_CONTEXT_PLANES *above_context; /* row of context for each plane */
ENTROPY_CONTEXT_PLANES left_context[2]; /* (up to) 4 contexts "" */
ENTROPY_CONTEXT_PLANES left_context[4]; /* (up to) 4 contexts "" */
/* keyframe block modes are predicted by their above, left neighbors */
@ -242,9 +223,7 @@ typedef struct VP9Common {
[VP9_KF_BINTRAMODES]
[VP9_KF_BINTRAMODES - 1];
vp9_prob kf_ymode_prob[8][VP9_YMODES - 1]; /* keyframe "" */
#if CONFIG_SUPERBLOCKS
vp9_prob sb_kf_ymode_prob[8][VP9_I32X32_MODES - 1];
#endif
int kf_ymode_probs_index;
int kf_ymode_probs_update;
vp9_prob kf_uv_mode_prob[VP9_YMODES] [VP9_UV_MODES - 1];
@ -252,9 +231,8 @@ typedef struct VP9Common {
vp9_prob prob_intra_coded;
vp9_prob prob_last_coded;
vp9_prob prob_gf_coded;
#if CONFIG_SUPERBLOCKS
vp9_prob sb_coded;
#endif
vp9_prob sb32_coded;
vp9_prob sb64_coded;
// Context probabilities when using predictive coding of segment id
vp9_prob segment_pred_probs[PREDICTION_PROBS];
@ -268,7 +246,7 @@ typedef struct VP9Common {
vp9_prob prob_comppred[COMP_PRED_CONTEXTS];
// FIXME contextualize
vp9_prob prob_tx[TX_SIZE_MAX - 1];
vp9_prob prob_tx[TX_SIZE_MAX_SB - 1];
vp9_prob mbskip_pred_probs[MBSKIP_CONTEXTS];
@ -290,17 +268,10 @@ typedef struct VP9Common {
struct postproc_state postproc_state;
#endif
#if CONFIG_PRED_FILTER
/* Prediction filter variables */
int pred_filter_mode; // 0=disabled at the frame level (no MB filtered)
// 1=enabled at the frame level (all MB filtered)
// 2=specified per MB (1=filtered, 0=non-filtered)
vp9_prob prob_pred_filter_off;
#endif
#if CONFIG_COMP_INTERINTRA_PRED
int use_interintra;
#endif
} VP9_COMMON;
#endif // __INC_ONYX_INT_H
#endif // VP9_COMMON_VP9_ONYXC_INT_H_

52
vp9/common/vp9_postproc.c
View File

@ -13,7 +13,7 @@
#include "vpx_scale/yv12config.h"
#include "vp9/common/vp9_postproc.h"
#include "vp9/common/vp9_textblit.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "vp9/common/vp9_systemdependent.h"
#include "./vp9_rtcd.h"
#include "./vpx_scale_rtcd.h"
@ -132,20 +132,20 @@ const short vp9_rv[] = {
/****************************************************************************
*/
void vp9_post_proc_down_and_across_c(unsigned char *src_ptr,
unsigned char *dst_ptr,
void vp9_post_proc_down_and_across_c(uint8_t *src_ptr,
uint8_t *dst_ptr,
int src_pixels_per_line,
int dst_pixels_per_line,
int rows,
int cols,
int flimit) {
unsigned char *p_src, *p_dst;
uint8_t *p_src, *p_dst;
int row;
int col;
int i;
int v;
int pitch = src_pixels_per_line;
unsigned char d[8];
uint8_t d[8];
(void)dst_pixels_per_line;
for (row = 0; row < rows; row++) {
@ -215,12 +215,12 @@ static int q2mbl(int x) {
return x * x / 3;
}
void vp9_mbpost_proc_across_ip_c(unsigned char *src, int pitch,
void vp9_mbpost_proc_across_ip_c(uint8_t *src, int pitch,
int rows, int cols, int flimit) {
int r, c, i;
unsigned char *s = src;
unsigned char d[16];
uint8_t *s = src;
uint8_t d[16];
for (r = 0; r < rows; r++) {
@ -253,16 +253,16 @@ void vp9_mbpost_proc_across_ip_c(unsigned char *src, int pitch,
}
}
void vp9_mbpost_proc_down_c(unsigned char *dst, int pitch,
void vp9_mbpost_proc_down_c(uint8_t *dst, int pitch,
int rows, int cols, int flimit) {
int r, c, i;
const short *rv3 = &vp9_rv[63 & rand()];
for (c = 0; c < cols; c++) {
unsigned char *s = &dst[c];
uint8_t *s = &dst[c];
int sumsq = 0;
int sum = 0;
unsigned char d[16];
uint8_t d[16];
const short *rv2 = rv3 + ((c * 17) & 127);
for (i = -8; i <= 6; i++) {
@ -439,7 +439,7 @@ static void fillrd(struct postproc_state *state, int q, int a) {
* SPECIAL NOTES : None.
*
****************************************************************************/
void vp9_plane_add_noise_c(unsigned char *Start, char *noise,
void vp9_plane_add_noise_c(uint8_t *Start, char *noise,
char blackclamp[16],
char whiteclamp[16],
char bothclamp[16],
@ -447,7 +447,7 @@ void vp9_plane_add_noise_c(unsigned char *Start, char *noise,
unsigned int i, j;
for (i = 0; i < Height; i++) {
unsigned char *Pos = Start + i * Pitch;
uint8_t *Pos = Start + i * Pitch;
char *Ref = (char *)(noise + (rand() & 0xff));
for (j = 0; j < Width; j++) {
@ -466,7 +466,7 @@ void vp9_plane_add_noise_c(unsigned char *Start, char *noise,
* edges unblended to give distinction to macro blocks in areas
* filled with the same color block.
*/
void vp9_blend_mb_inner_c(unsigned char *y, unsigned char *u, unsigned char *v,
void vp9_blend_mb_inner_c(uint8_t *y, uint8_t *u, uint8_t *v,
int y1, int u1, int v1, int alpha, int stride) {
int i, j;
int y1_const = y1 * ((1 << 16) - alpha);
@ -499,7 +499,7 @@ void vp9_blend_mb_inner_c(unsigned char *y, unsigned char *u, unsigned char *v,
/* Blend only the edge of the macro block. Leave center
* unblended to allow for other visualizations to be layered.
*/
void vp9_blend_mb_outer_c(unsigned char *y, unsigned char *u, unsigned char *v,
void vp9_blend_mb_outer_c(uint8_t *y, uint8_t *u, uint8_t *v,
int y1, int u1, int v1, int alpha, int stride) {
int i, j;
int y1_const = y1 * ((1 << 16) - alpha);
@ -554,7 +554,7 @@ void vp9_blend_mb_outer_c(unsigned char *y, unsigned char *u, unsigned char *v,
}
}
void vp9_blend_b_c(unsigned char *y, unsigned char *u, unsigned char *v,
void vp9_blend_b_c(uint8_t *y, uint8_t *u, uint8_t *v,
int y1, int u1, int v1, int alpha, int stride) {
int i, j;
int y1_const = y1 * ((1 << 16) - alpha);
@ -688,7 +688,7 @@ int vp9_post_proc_frame(VP9_COMMON *oci, YV12_BUFFER_CONFIG *dest,
if (flags & VP9D_DEBUG_TXT_MBLK_MODES) {
int i, j;
unsigned char *y_ptr;
uint8_t *y_ptr;
YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
int mb_rows = post->y_height >> 4;
int mb_cols = post->y_width >> 4;
@ -717,7 +717,7 @@ int vp9_post_proc_frame(VP9_COMMON *oci, YV12_BUFFER_CONFIG *dest,
if (flags & VP9D_DEBUG_TXT_DC_DIFF) {
int i, j;
unsigned char *y_ptr;
uint8_t *y_ptr;
YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
int mb_rows = post->y_height >> 4;
int mb_cols = post->y_width >> 4;
@ -764,7 +764,7 @@ int vp9_post_proc_frame(VP9_COMMON *oci, YV12_BUFFER_CONFIG *dest,
YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
int width = post->y_width;
int height = post->y_height;
unsigned char *y_buffer = oci->post_proc_buffer.y_buffer;
uint8_t *y_buffer = oci->post_proc_buffer.y_buffer;
int y_stride = oci->post_proc_buffer.y_stride;
MODE_INFO *mi = oci->mi;
int x0, y0;
@ -906,9 +906,9 @@ int vp9_post_proc_frame(VP9_COMMON *oci, YV12_BUFFER_CONFIG *dest,
YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
int width = post->y_width;
int height = post->y_height;
unsigned char *y_ptr = oci->post_proc_buffer.y_buffer;
unsigned char *u_ptr = oci->post_proc_buffer.u_buffer;
unsigned char *v_ptr = oci->post_proc_buffer.v_buffer;
uint8_t *y_ptr = oci->post_proc_buffer.y_buffer;
uint8_t *u_ptr = oci->post_proc_buffer.u_buffer;
uint8_t *v_ptr = oci->post_proc_buffer.v_buffer;
int y_stride = oci->post_proc_buffer.y_stride;
MODE_INFO *mi = oci->mi;
@ -920,7 +920,7 @@ int vp9_post_proc_frame(VP9_COMMON *oci, YV12_BUFFER_CONFIG *dest,
((ppflags->display_mb_modes_flag & B_PRED) ||
ppflags->display_b_modes_flag)) {
int by, bx;
unsigned char *yl, *ul, *vl;
uint8_t *yl, *ul, *vl;
union b_mode_info *bmi = mi->bmi;
yl = y_ptr + x;
@ -971,9 +971,9 @@ int vp9_post_proc_frame(VP9_COMMON *oci, YV12_BUFFER_CONFIG *dest,
YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
int width = post->y_width;
int height = post->y_height;
unsigned char *y_ptr = oci->post_proc_buffer.y_buffer;
unsigned char *u_ptr = oci->post_proc_buffer.u_buffer;
unsigned char *v_ptr = oci->post_proc_buffer.v_buffer;
uint8_t *y_ptr = oci->post_proc_buffer.y_buffer;
uint8_t *u_ptr = oci->post_proc_buffer.u_buffer;
uint8_t *v_ptr = oci->post_proc_buffer.v_buffer;
int y_stride = oci->post_proc_buffer.y_stride;
MODE_INFO *mi = oci->mi;

3
vp9/common/vp9_postproc.h
View File

@ -38,4 +38,5 @@ void vp9_deblock(YV12_BUFFER_CONFIG *source,
int q,
int low_var_thresh,
int flag);
#endif
#endif // VP9_COMMON_VP9_POSTPROC_H_

4
vp9/common/vp9_ppflags.h
View File

@ -8,9 +8,9 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_PPFLAGS_H_
#define VP9_COMMON_VP9_PPFLAGS_H_
enum {
VP9D_NOFILTERING = 0,
VP9D_DEBLOCK = 1 << 0,
@ -35,4 +35,4 @@ typedef struct {
int display_mv_flag;
} vp9_ppflags_t;
#endif
#endif // VP9_COMMON_VP9_PPFLAGS_H_

6
vp9/common/vp9_pragmas.h
View File

@ -8,8 +8,8 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_PRAGMAS_H_
#define VP9_COMMON_VP9_PRAGMAS_H_
#ifdef __INTEL_COMPILER
#pragma warning(disable:997 1011 170)
@ -17,3 +17,5 @@
#ifdef _MSC_VER
#pragma warning(disable:4799)
#endif
#endif // VP9_COMMON_VP9_PRAGMAS_H_

110
vp9/common/vp9_pred_common.c
View File

@ -9,8 +9,10 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_pred_common.h"
#include "vp9/common/vp9_seg_common.h"
#include "vp9/common/vp9_treecoder.h"
// TBD prediction functions for various bitstream signals
@ -221,54 +223,57 @@ unsigned char vp9_get_pred_flag(const MACROBLOCKD *const xd,
void vp9_set_pred_flag(MACROBLOCKD *const xd,
PRED_ID pred_id,
unsigned char pred_flag) {
#if CONFIG_SUPERBLOCKS
const int mis = xd->mode_info_stride;
#endif
switch (pred_id) {
case PRED_SEG_ID:
xd->mode_info_context->mbmi.seg_id_predicted = pred_flag;
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
if (xd->mb_to_right_edge >= 0)
xd->mode_info_context[1].mbmi.seg_id_predicted = pred_flag;
if (xd->mb_to_bottom_edge >= 0) {
xd->mode_info_context[mis].mbmi.seg_id_predicted = pred_flag;
if (xd->mb_to_right_edge >= 0)
xd->mode_info_context[mis + 1].mbmi.seg_id_predicted = pred_flag;
if (xd->mode_info_context->mbmi.sb_type) {
#define sub(a, b) (b) < 0 ? (a) + (b) : (a)
const int n_mbs = 1 << xd->mode_info_context->mbmi.sb_type;
const int x_mbs = sub(n_mbs, xd->mb_to_right_edge >> 7);
const int y_mbs = sub(n_mbs, xd->mb_to_bottom_edge >> 7);
int x, y;
for (y = 0; y < y_mbs; y++) {
for (x = !y; x < x_mbs; x++) {
xd->mode_info_context[y * mis + x].mbmi.seg_id_predicted =
pred_flag;
}
}
}
#endif
break;
case PRED_REF:
xd->mode_info_context->mbmi.ref_predicted = pred_flag;
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
if (xd->mb_to_right_edge >= 0)
xd->mode_info_context[1].mbmi.ref_predicted = pred_flag;
if (xd->mb_to_bottom_edge >= 0) {
xd->mode_info_context[mis].mbmi.ref_predicted = pred_flag;
if (xd->mb_to_right_edge >= 0)
xd->mode_info_context[mis + 1].mbmi.ref_predicted = pred_flag;
if (xd->mode_info_context->mbmi.sb_type) {
const int n_mbs = 1 << xd->mode_info_context->mbmi.sb_type;
const int x_mbs = sub(n_mbs, xd->mb_to_right_edge >> 7);
const int y_mbs = sub(n_mbs, xd->mb_to_bottom_edge >> 7);
int x, y;
for (y = 0; y < y_mbs; y++) {
for (x = !y; x < x_mbs; x++) {
xd->mode_info_context[y * mis + x].mbmi.ref_predicted = pred_flag;
}
}
}
#endif
break;
case PRED_MBSKIP:
xd->mode_info_context->mbmi.mb_skip_coeff = pred_flag;
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
if (xd->mb_to_right_edge >= 0)
xd->mode_info_context[1].mbmi.mb_skip_coeff = pred_flag;
if (xd->mb_to_bottom_edge >= 0) {
xd->mode_info_context[mis].mbmi.mb_skip_coeff = pred_flag;
if (xd->mb_to_right_edge >= 0)
xd->mode_info_context[mis + 1].mbmi.mb_skip_coeff = pred_flag;
if (xd->mode_info_context->mbmi.sb_type) {
const int n_mbs = 1 << xd->mode_info_context->mbmi.sb_type;
const int x_mbs = sub(n_mbs, xd->mb_to_right_edge >> 7);
const int y_mbs = sub(n_mbs, xd->mb_to_bottom_edge >> 7);
int x, y;
for (y = 0; y < y_mbs; y++) {
for (x = !y; x < x_mbs; x++) {
xd->mode_info_context[y * mis + x].mbmi.mb_skip_coeff = pred_flag;
}
}
}
#endif
break;
default:
@ -286,25 +291,25 @@ unsigned char vp9_get_pred_mb_segid(const VP9_COMMON *const cm,
const MACROBLOCKD *const xd, int MbIndex) {
// Currently the prediction for the macroblock segment ID is
// the value stored for this macroblock in the previous frame.
#if CONFIG_SUPERBLOCKS
if (!xd->mode_info_context->mbmi.encoded_as_sb) {
#endif
if (!xd->mode_info_context->mbmi.sb_type) {
return cm->last_frame_seg_map[MbIndex];
#if CONFIG_SUPERBLOCKS
} else {
int seg_id = cm->last_frame_seg_map[MbIndex];
int mb_col = MbIndex % cm->mb_cols;
int mb_row = MbIndex / cm->mb_cols;
if (mb_col + 1 < cm->mb_cols)
seg_id = seg_id && cm->last_frame_seg_map[MbIndex + 1];
if (mb_row + 1 < cm->mb_rows) {
seg_id = seg_id && cm->last_frame_seg_map[MbIndex + cm->mb_cols];
if (mb_col + 1 < cm->mb_cols)
seg_id = seg_id && cm->last_frame_seg_map[MbIndex + cm->mb_cols + 1];
const int n_mbs = 1 << xd->mode_info_context->mbmi.sb_type;
const int mb_col = MbIndex % cm->mb_cols;
const int mb_row = MbIndex / cm->mb_cols;
const int x_mbs = MIN(n_mbs, cm->mb_cols - mb_col);
const int y_mbs = MIN(n_mbs, cm->mb_rows - mb_row);
int x, y;
unsigned seg_id = -1;
for (y = mb_row; y < mb_row + y_mbs; y++) {
for (x = mb_col; x < mb_col + x_mbs; x++) {
seg_id = MIN(seg_id, cm->last_frame_seg_map[cm->mb_cols * y + x]);
}
}
return seg_id;
}
#endif
}
MV_REFERENCE_FRAME vp9_get_pred_ref(const VP9_COMMON *const cm,
@ -383,26 +388,13 @@ void vp9_calc_ref_probs(int *count, vp9_prob *probs) {
int tot_count;
tot_count = count[0] + count[1] + count[2] + count[3];
if (tot_count) {
probs[0] = (vp9_prob)((count[0] * 255 + (tot_count >> 1)) / tot_count);
probs[0] += !probs[0];
} else
probs[0] = 128;
probs[0] = get_prob(count[0], tot_count);
tot_count -= count[0];
if (tot_count) {
probs[1] = (vp9_prob)((count[1] * 255 + (tot_count >> 1)) / tot_count);
probs[1] += !probs[1];
} else
probs[1] = 128;
probs[1] = get_prob(count[1], tot_count);
tot_count -= count[1];
if (tot_count) {
probs[2] = (vp9_prob)((count[2] * 255 + (tot_count >> 1)) / tot_count);
probs[2] += !probs[2];
} else
probs[2] = 128;
probs[2] = get_prob(count[2], tot_count);
}
// Computes a set of modified conditional probabilities for the reference frame

3
vp9/common/vp9_pred_common.h
View File

@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vp9/common/vp9_type_aliases.h"
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/common/vp9_blockd.h"
@ -53,4 +52,4 @@ 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);
#endif /* __INC_PRED_COMMON_H__ */
#endif // VP9_COMMON_VP9_PRED_COMMON_H_

6
vp9/common/vp9_quant_common.h
View File

@ -8,15 +8,19 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#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();
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);
#endif // VP9_COMMON_VP9_QUANT_COMMON_H_

135
vp9/common/vp9_recon.c
View File

@ -13,26 +13,15 @@
#include "vp9_rtcd.h"
#include "vp9/common/vp9_blockd.h"
void vp9_recon_b_c
(
unsigned char *pred_ptr,
short *diff_ptr,
unsigned char *dst_ptr,
int stride
) {
void vp9_recon_b_c(uint8_t *pred_ptr,
int16_t *diff_ptr,
uint8_t *dst_ptr,
int stride) {
int r, c;
for (r = 0; r < 4; r++) {
for (c = 0; c < 4; c++) {
int a = diff_ptr[c] + pred_ptr[c];
if (a < 0)
a = 0;
if (a > 255)
a = 255;
dst_ptr[c] = (unsigned char) a;
dst_ptr[c] = clip_pixel(diff_ptr[c] + pred_ptr[c]);
}
dst_ptr += stride;
@ -41,26 +30,15 @@ void vp9_recon_b_c
}
}
void vp9_recon_uv_b_c
(
unsigned char *pred_ptr,
short *diff_ptr,
unsigned char *dst_ptr,
int stride
) {
void vp9_recon_uv_b_c(uint8_t *pred_ptr,
int16_t *diff_ptr,
uint8_t *dst_ptr,
int stride) {
int r, c;
for (r = 0; r < 4; r++) {
for (c = 0; c < 4; c++) {
int a = diff_ptr[c] + pred_ptr[c];
if (a < 0)
a = 0;
if (a > 255)
a = 255;
dst_ptr[c] = (unsigned char) a;
dst_ptr[c] = clip_pixel(diff_ptr[c] + pred_ptr[c]);
}
dst_ptr += stride;
@ -68,26 +46,16 @@ void vp9_recon_uv_b_c
pred_ptr += 8;
}
}
void vp9_recon4b_c
(
unsigned char *pred_ptr,
short *diff_ptr,
unsigned char *dst_ptr,
int stride
) {
void vp9_recon4b_c(uint8_t *pred_ptr,
int16_t *diff_ptr,
uint8_t *dst_ptr,
int stride) {
int r, c;
for (r = 0; r < 4; r++) {
for (c = 0; c < 16; c++) {
int a = diff_ptr[c] + pred_ptr[c];
if (a < 0)
a = 0;
if (a > 255)
a = 255;
dst_ptr[c] = (unsigned char) a;
dst_ptr[c] = clip_pixel(diff_ptr[c] + pred_ptr[c]);
}
dst_ptr += stride;
@ -96,26 +64,15 @@ void vp9_recon4b_c
}
}
void vp9_recon2b_c
(
unsigned char *pred_ptr,
short *diff_ptr,
unsigned char *dst_ptr,
int stride
) {
void vp9_recon2b_c(uint8_t *pred_ptr,
int16_t *diff_ptr,
uint8_t *dst_ptr,
int stride) {
int r, c;
for (r = 0; r < 4; r++) {
for (c = 0; c < 8; c++) {
int a = diff_ptr[c] + pred_ptr[c];
if (a < 0)
a = 0;
if (a > 255)
a = 255;
dst_ptr[c] = (unsigned char) a;
dst_ptr[c] = clip_pixel(diff_ptr[c] + pred_ptr[c]);
}
dst_ptr += stride;
@ -124,21 +81,15 @@ void vp9_recon2b_c
}
}
#if CONFIG_SUPERBLOCKS
void vp9_recon_mby_s_c(MACROBLOCKD *xd, uint8_t *dst) {
int x, y;
BLOCKD *b = &xd->block[0];
int stride = b->dst_stride;
short *diff = b->diff;
int16_t *diff = b->diff;
for (y = 0; y < 16; y++) {
for (x = 0; x < 16; x++) {
int a = dst[x] + diff[x];
if (a < 0)
a = 0;
else if (a > 255)
a = 255;
dst[x] = a;
dst[x] = clip_pixel(dst[x] + diff[x]);
}
dst += stride;
diff += 16;
@ -152,23 +103,47 @@ void vp9_recon_mbuv_s_c(MACROBLOCKD *xd, uint8_t *udst, uint8_t *vdst) {
for (i = 0; i < 2; i++, dst = vdst) {
BLOCKD *b = &xd->block[16 + 4 * i];
int stride = b->dst_stride;
short *diff = b->diff;
int16_t *diff = b->diff;
for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++) {
int a = dst[x] + diff[x];
if (a < 0)
a = 0;
else if (a > 255)
a = 255;
dst[x] = a;
dst[x] = clip_pixel(dst[x] + diff[x]);
}
dst += stride;
diff += 8;
}
}
}
#endif
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;
for (y = 0; y < 32; y++) {
for (x = 0; x < 32; x++) {
dst[x] = clip_pixel(dst[x] + diff[x]);
}
dst += stride;
diff += 32;
}
}
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;
for (y = 0; y < 16; y++) {
for (x = 0; x < 16; x++) {
udst[x] = clip_pixel(udst[x] + udiff[x]);
vdst[x] = clip_pixel(vdst[x] + vdiff[x]);
}
udst += stride;
vdst += stride;
udiff += 16;
vdiff += 16;
}
}
void vp9_recon_mby_c(MACROBLOCKD *xd) {
int i;

464
vp9/common/vp9_reconinter.c
View File

@ -18,45 +18,58 @@
void vp9_setup_interp_filters(MACROBLOCKD *xd,
INTERPOLATIONFILTERTYPE mcomp_filter_type,
VP9_COMMON *cm) {
#if CONFIG_ENABLE_6TAP
if (mcomp_filter_type == SIXTAP) {
xd->subpixel_predict = vp9_sixtap_predict;
xd->subpixel_predict4x4 = vp9_sixtap_predict4x4;
xd->subpixel_predict8x4 = vp9_sixtap_predict8x4;
xd->subpixel_predict8x8 = vp9_sixtap_predict8x8;
xd->subpixel_predict16x16 = vp9_sixtap_predict16x16;
xd->subpixel_predict_avg = vp9_sixtap_predict_avg;
xd->subpixel_predict_avg4x4 = vp9_sixtap_predict_avg4x4;
xd->subpixel_predict_avg8x8 = vp9_sixtap_predict_avg8x8;
xd->subpixel_predict_avg16x16 = vp9_sixtap_predict_avg16x16;
} else if (mcomp_filter_type == EIGHTTAP || mcomp_filter_type == SWITCHABLE) {
xd->subpixel_predict = vp9_eighttap_predict;
} else {
#endif
if (mcomp_filter_type == EIGHTTAP || mcomp_filter_type == SWITCHABLE) {
xd->subpixel_predict4x4 = vp9_eighttap_predict4x4;
xd->subpixel_predict8x4 = vp9_eighttap_predict8x4;
xd->subpixel_predict8x8 = vp9_eighttap_predict8x8;
xd->subpixel_predict16x16 = vp9_eighttap_predict16x16;
xd->subpixel_predict_avg = vp9_eighttap_predict_avg4x4;
xd->subpixel_predict_avg4x4 = vp9_eighttap_predict_avg4x4;
xd->subpixel_predict_avg8x8 = vp9_eighttap_predict_avg8x8;
xd->subpixel_predict_avg16x16 = vp9_eighttap_predict_avg16x16;
} else if (mcomp_filter_type == EIGHTTAP_SMOOTH) {
xd->subpixel_predict4x4 = vp9_eighttap_predict4x4_smooth;
xd->subpixel_predict8x4 = vp9_eighttap_predict8x4_smooth;
xd->subpixel_predict8x8 = vp9_eighttap_predict8x8_smooth;
xd->subpixel_predict16x16 = vp9_eighttap_predict16x16_smooth;
xd->subpixel_predict_avg4x4 = vp9_eighttap_predict_avg4x4_smooth;
xd->subpixel_predict_avg8x8 = vp9_eighttap_predict_avg8x8_smooth;
xd->subpixel_predict_avg16x16 = vp9_eighttap_predict_avg16x16_smooth;
} else if (mcomp_filter_type == EIGHTTAP_SHARP) {
xd->subpixel_predict = vp9_eighttap_predict_sharp;
xd->subpixel_predict4x4 = vp9_eighttap_predict4x4_sharp;
xd->subpixel_predict8x4 = vp9_eighttap_predict8x4_sharp;
xd->subpixel_predict8x8 = vp9_eighttap_predict8x8_sharp;
xd->subpixel_predict16x16 = vp9_eighttap_predict16x16_sharp;
xd->subpixel_predict_avg = vp9_eighttap_predict_avg4x4_sharp;
xd->subpixel_predict_avg4x4 = vp9_eighttap_predict_avg4x4_sharp;
xd->subpixel_predict_avg8x8 = vp9_eighttap_predict_avg8x8_sharp;
xd->subpixel_predict_avg16x16 = vp9_eighttap_predict_avg16x16_sharp_c;
}
else {
xd->subpixel_predict = vp9_bilinear_predict4x4;
} else {
xd->subpixel_predict4x4 = vp9_bilinear_predict4x4;
xd->subpixel_predict8x4 = vp9_bilinear_predict8x4;
xd->subpixel_predict8x8 = vp9_bilinear_predict8x8;
xd->subpixel_predict16x16 = vp9_bilinear_predict16x16;
xd->subpixel_predict_avg = vp9_bilinear_predict_avg4x4;
xd->subpixel_predict_avg4x4 = vp9_bilinear_predict_avg4x4;
xd->subpixel_predict_avg8x8 = vp9_bilinear_predict_avg8x8;
xd->subpixel_predict_avg16x16 = vp9_bilinear_predict_avg16x16;
}
#if CONFIG_ENABLE_6TAP
}
#endif
}
void vp9_copy_mem16x16_c(unsigned char *src,
void vp9_copy_mem16x16_c(uint8_t *src,
int src_stride,
unsigned char *dst,
uint8_t *dst,
int dst_stride) {
int r;
@ -91,9 +104,9 @@ void vp9_copy_mem16x16_c(unsigned char *src,
}
}
void vp9_avg_mem16x16_c(unsigned char *src,
void vp9_avg_mem16x16_c(uint8_t *src,
int src_stride,
unsigned char *dst,
uint8_t *dst,
int dst_stride) {
int r;
@ -109,9 +122,9 @@ void vp9_avg_mem16x16_c(unsigned char *src,
}
}
void vp9_copy_mem8x8_c(unsigned char *src,
void vp9_copy_mem8x8_c(uint8_t *src,
int src_stride,
unsigned char *dst,
uint8_t *dst,
int dst_stride) {
int r;
@ -134,9 +147,9 @@ void vp9_copy_mem8x8_c(unsigned char *src,
}
}
void vp9_avg_mem8x8_c(unsigned char *src,
void vp9_avg_mem8x8_c(uint8_t *src,
int src_stride,
unsigned char *dst,
uint8_t *dst,
int dst_stride) {
int r;
@ -152,9 +165,9 @@ void vp9_avg_mem8x8_c(unsigned char *src,
}
}
void vp9_copy_mem8x4_c(unsigned char *src,
void vp9_copy_mem8x4_c(uint8_t *src,
int src_stride,
unsigned char *dst,
uint8_t *dst,
int dst_stride) {
int r;
@ -179,9 +192,9 @@ void vp9_copy_mem8x4_c(unsigned char *src,
void vp9_build_inter_predictors_b(BLOCKD *d, int pitch, vp9_subpix_fn_t sppf) {
int r;
unsigned char *ptr_base;
unsigned char *ptr;
unsigned char *pred_ptr = d->predictor;
uint8_t *ptr_base;
uint8_t *ptr;
uint8_t *pred_ptr = d->predictor;
int_mv mv;
ptr_base = *(d->base_pre);
@ -221,9 +234,9 @@ void vp9_build_inter_predictors_b(BLOCKD *d, int pitch, vp9_subpix_fn_t sppf) {
void vp9_build_2nd_inter_predictors_b(BLOCKD *d, int pitch,
vp9_subpix_fn_t sppf) {
int r;
unsigned char *ptr_base;
unsigned char *ptr;
unsigned char *pred_ptr = d->predictor;
uint8_t *ptr_base;
uint8_t *ptr;
uint8_t *pred_ptr = d->predictor;
int_mv mv;
ptr_base = *(d->base_second_pre);
@ -251,9 +264,9 @@ void vp9_build_2nd_inter_predictors_b(BLOCKD *d, int pitch,
}
void vp9_build_inter_predictors4b(MACROBLOCKD *xd, BLOCKD *d, int pitch) {
unsigned char *ptr_base;
unsigned char *ptr;
unsigned char *pred_ptr = d->predictor;
uint8_t *ptr_base;
uint8_t *ptr;
uint8_t *pred_ptr = d->predictor;
int_mv mv;
ptr_base = *(d->base_pre);
@ -277,9 +290,9 @@ void vp9_build_inter_predictors4b(MACROBLOCKD *xd, BLOCKD *d, int pitch) {
*/
void vp9_build_2nd_inter_predictors4b(MACROBLOCKD *xd,
BLOCKD *d, int pitch) {
unsigned char *ptr_base;
unsigned char *ptr;
unsigned char *pred_ptr = d->predictor;
uint8_t *ptr_base;
uint8_t *ptr;
uint8_t *pred_ptr = d->predictor;
int_mv mv;
ptr_base = *(d->base_second_pre);
@ -296,9 +309,9 @@ void vp9_build_2nd_inter_predictors4b(MACROBLOCKD *xd,
}
static void build_inter_predictors2b(MACROBLOCKD *xd, BLOCKD *d, int pitch) {
unsigned char *ptr_base;
unsigned char *ptr;
unsigned char *pred_ptr = d->predictor;
uint8_t *ptr_base;
uint8_t *ptr;
uint8_t *pred_ptr = d->predictor;
int_mv mv;
ptr_base = *(d->base_pre);
@ -314,132 +327,6 @@ static void build_inter_predictors2b(MACROBLOCKD *xd, BLOCKD *d, int pitch) {
}
}
/*encoder only*/
#if CONFIG_PRED_FILTER
// Select the thresholded or non-thresholded filter
#define USE_THRESH_FILTER 0
#define PRED_FILT_LEN 5
static const int filt_shift = 4;
static const int pred_filter[PRED_FILT_LEN] = {1, 2, 10, 2, 1};
// Alternative filter {1, 1, 4, 1, 1}
#if !USE_THRESH_FILTER
void filter_mb(unsigned char *src, int src_stride,
unsigned char *dst, int dst_stride,
int width, int height) {
int i, j, k;
unsigned int Temp[32 * 32];
unsigned int *pTmp = Temp;
unsigned char *pSrc = src - (1 + src_stride) * (PRED_FILT_LEN / 2);
// Horizontal
for (i = 0; i < height + PRED_FILT_LEN - 1; i++) {
for (j = 0; j < width; j++) {
int sum = 0;
for (k = 0; k < PRED_FILT_LEN; k++)
sum += pSrc[j + k] * pred_filter[k];
pTmp[j] = sum;
}
pSrc += src_stride;
pTmp += width;
}
// Vertical
pTmp = Temp;
for (i = 0; i < width; i++) {
unsigned char *pDst = dst + i;
for (j = 0; j < height; j++) {
int sum = 0;
for (k = 0; k < PRED_FILT_LEN; k++)
sum += pTmp[(j + k) * width] * pred_filter[k];
// Round
sum = (sum + ((1 << (filt_shift << 1)) >> 1)) >> (filt_shift << 1);
pDst[j * dst_stride] = (sum < 0 ? 0 : sum > 255 ? 255 : sum);
}
++pTmp;
}
}
#else
// Based on vp9_post_proc_down_and_across_c (vp9_postproc.c)
void filter_mb(unsigned char *src, int src_stride,
unsigned char *dst, int dst_stride,
int width, int height) {
unsigned char *pSrc, *pDst;
int row;
int col;
int i;
int v;
unsigned char d[8];
/* TODO flimit should be linked to the quantizer value */
int flimit = 7;
for (row = 0; row < height; row++) {
/* post_proc_down for one row */
pSrc = src;
pDst = dst;
for (col = 0; col < width; col++) {
int kernel = (1 << (filt_shift - 1));
int v = pSrc[col];
for (i = -2; i <= 2; i++) {
if (abs(v - pSrc[col + i * src_stride]) > flimit)
goto down_skip_convolve;
kernel += pred_filter[2 + i] * pSrc[col + i * src_stride];
}
v = (kernel >> filt_shift);
down_skip_convolve:
pDst[col] = v;
}
/* now post_proc_across */
pSrc = dst;
pDst = dst;
for (i = 0; i < 8; i++)
d[i] = pSrc[i];
for (col = 0; col < width; col++) {
int kernel = (1 << (filt_shift - 1));
v = pSrc[col];
d[col & 7] = v;
for (i = -2; i <= 2; i++) {
if (abs(v - pSrc[col + i]) > flimit)
goto across_skip_convolve;
kernel += pred_filter[2 + i] * pSrc[col + i];
}
d[col & 7] = (kernel >> filt_shift);
across_skip_convolve:
if (col >= 2)
pDst[col - 2] = d[(col - 2) & 7];
}
/* handle the last two pixels */
pDst[col - 2] = d[(col - 2) & 7];
pDst[col - 1] = d[(col - 1) & 7];
/* next row */
src += src_stride;
dst += dst_stride;
}
}
#endif // !USE_THRESH_FILTER
#endif // CONFIG_PRED_FILTER
/*encoder only*/
void vp9_build_inter4x4_predictors_mbuv(MACROBLOCKD *xd) {
int i, j;
@ -524,13 +411,13 @@ void vp9_build_inter4x4_predictors_mbuv(MACROBLOCKD *xd) {
if (d0->bmi.as_mv.first.as_int == d1->bmi.as_mv.first.as_int)
build_inter_predictors2b(xd, d0, 8);
else {
vp9_build_inter_predictors_b(d0, 8, xd->subpixel_predict);
vp9_build_inter_predictors_b(d1, 8, xd->subpixel_predict);
vp9_build_inter_predictors_b(d0, 8, xd->subpixel_predict4x4);
vp9_build_inter_predictors_b(d1, 8, xd->subpixel_predict4x4);
}
if (xd->mode_info_context->mbmi.second_ref_frame > 0) {
vp9_build_2nd_inter_predictors_b(d0, 8, xd->subpixel_predict_avg);
vp9_build_2nd_inter_predictors_b(d1, 8, xd->subpixel_predict_avg);
vp9_build_2nd_inter_predictors_b(d0, 8, xd->subpixel_predict_avg4x4);
vp9_build_2nd_inter_predictors_b(d1, 8, xd->subpixel_predict_avg4x4);
}
}
}
@ -573,11 +460,11 @@ static void clamp_uvmv_to_umv_border(MV *mv, const MACROBLOCKD *xd) {
/*encoder only*/
void vp9_build_1st_inter16x16_predictors_mby(MACROBLOCKD *xd,
unsigned char *dst_y,
uint8_t *dst_y,
int dst_ystride,
int clamp_mvs) {
unsigned char *ptr_base = xd->pre.y_buffer;
unsigned char *ptr;
uint8_t *ptr_base = xd->pre.y_buffer;
uint8_t *ptr;
int pre_stride = xd->block[0].pre_stride;
int_mv ymv;
@ -588,29 +475,6 @@ void vp9_build_1st_inter16x16_predictors_mby(MACROBLOCKD *xd,
ptr = ptr_base + (ymv.as_mv.row >> 3) * pre_stride + (ymv.as_mv.col >> 3);
#if CONFIG_PRED_FILTER
if (xd->mode_info_context->mbmi.pred_filter_enabled) {
if ((ymv.as_mv.row | ymv.as_mv.col) & 7) {
// Sub-pel filter needs extended input
int len = 15 + (VP9_INTERP_EXTEND << 1);
unsigned char Temp[32 * 32]; // Data required by sub-pel filter
unsigned char *pTemp = Temp + (VP9_INTERP_EXTEND - 1) * (len + 1);
// Copy extended MB into Temp array, applying the spatial filter
filter_mb(ptr - (VP9_INTERP_EXTEND - 1) * (pre_stride + 1), pre_stride,
Temp, len, len, len);
// Sub-pel interpolation
xd->subpixel_predict16x16(pTemp, len,
(ymv.as_mv.col & 7) << 1,
(ymv.as_mv.row & 7) << 1,
dst_y, dst_ystride);
} else {
// Apply spatial filter to create the prediction directly
filter_mb(ptr, pre_stride, dst_y, dst_ystride, 16, 16);
}
} else
#endif
if ((ymv.as_mv.row | ymv.as_mv.col) & 7) {
xd->subpixel_predict16x16(ptr, pre_stride,
(ymv.as_mv.col & 7) << 1,
@ -622,11 +486,11 @@ void vp9_build_1st_inter16x16_predictors_mby(MACROBLOCKD *xd,
}
void vp9_build_1st_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_uvstride) {
int offset;
unsigned char *uptr, *vptr;
uint8_t *uptr, *vptr;
int pre_stride = xd->block[0].pre_stride;
int_mv _o16x16mv;
int_mv _16x16mv;
@ -659,37 +523,6 @@ void vp9_build_1st_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
uptr = xd->pre.u_buffer + offset;
vptr = xd->pre.v_buffer + offset;
#if CONFIG_PRED_FILTER
if (xd->mode_info_context->mbmi.pred_filter_enabled) {
int i;
unsigned char *pSrc = uptr;
unsigned char *pDst = dst_u;
int len = 7 + (VP9_INTERP_EXTEND << 1);
unsigned char Temp[32 * 32]; // Data required by the sub-pel filter
unsigned char *pTemp = Temp + (VP9_INTERP_EXTEND - 1) * (len + 1);
// U & V
for (i = 0; i < 2; i++) {
if (_o16x16mv.as_int & 0x000f000f) {
// Copy extended MB into Temp array, applying the spatial filter
filter_mb(pSrc - (VP9_INTERP_EXTEND - 1) * (pre_stride + 1), pre_stride,
Temp, len, len, len);
// Sub-pel filter
xd->subpixel_predict8x8(pTemp, len,
_o16x16mv.as_mv.col & 15,
_o16x16mv.as_mv.row & 15,
pDst, dst_uvstride);
} else {
filter_mb(pSrc, pre_stride, pDst, dst_uvstride, 8, 8);
}
// V
pSrc = vptr;
pDst = dst_v;
}
} else
#endif
if (_o16x16mv.as_int & 0x000f000f) {
xd->subpixel_predict8x8(uptr, pre_stride, _o16x16mv.as_mv.col & 15,
_o16x16mv.as_mv.row & 15, dst_u, dst_uvstride);
@ -703,20 +536,19 @@ void vp9_build_1st_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
void vp9_build_1st_inter16x16_predictors_mb(MACROBLOCKD *xd,
unsigned char *dst_y,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride, int dst_uvstride) {
vp9_build_1st_inter16x16_predictors_mby(xd, dst_y, dst_ystride,
xd->mode_info_context->mbmi.need_to_clamp_mvs);
vp9_build_1st_inter16x16_predictors_mbuv(xd, dst_u, dst_v, dst_uvstride);
}
#if CONFIG_SUPERBLOCKS
void vp9_build_inter32x32_predictors_sb(MACROBLOCKD *x,
unsigned char *dst_y,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride) {
uint8_t *y1 = x->pre.y_buffer, *u1 = x->pre.u_buffer, *v1 = x->pre.v_buffer;
@ -781,14 +613,77 @@ void vp9_build_inter32x32_predictors_sb(MACROBLOCKD *x,
}
#endif
}
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) {
uint8_t *y1 = x->pre.y_buffer, *u1 = x->pre.u_buffer, *v1 = x->pre.v_buffer;
uint8_t *y2 = x->second_pre.y_buffer, *u2 = x->second_pre.u_buffer,
*v2 = x->second_pre.v_buffer;
int edge[4], n;
edge[0] = x->mb_to_top_edge;
edge[1] = x->mb_to_bottom_edge;
edge[2] = x->mb_to_left_edge;
edge[3] = x->mb_to_right_edge;
for (n = 0; n < 4; n++) {
const int x_idx = n & 1, y_idx = n >> 1;
x->mb_to_top_edge = edge[0] - ((y_idx * 32) << 3);
x->mb_to_bottom_edge = edge[1] + (((1 - y_idx) * 32) << 3);
x->mb_to_left_edge = edge[2] - ((x_idx * 32) << 3);
x->mb_to_right_edge = edge[3] + (((1 - x_idx) * 32) << 3);
x->pre.y_buffer = y1 + y_idx * 32 * x->pre.y_stride + x_idx * 32;
x->pre.u_buffer = u1 + y_idx * 16 * x->pre.uv_stride + x_idx * 16;
x->pre.v_buffer = v1 + y_idx * 16 * x->pre.uv_stride + x_idx * 16;
if (x->mode_info_context->mbmi.second_ref_frame > 0) {
x->second_pre.y_buffer = y2 + y_idx * 32 * x->pre.y_stride + x_idx * 32;
x->second_pre.u_buffer = u2 + y_idx * 16 * x->pre.uv_stride + x_idx * 16;
x->second_pre.v_buffer = v2 + y_idx * 16 * x->pre.uv_stride + x_idx * 16;
}
vp9_build_inter32x32_predictors_sb(x,
dst_y + y_idx * 32 * dst_ystride + x_idx * 32,
dst_u + y_idx * 16 * dst_uvstride + x_idx * 16,
dst_v + y_idx * 16 * dst_uvstride + x_idx * 16,
dst_ystride, dst_uvstride);
}
x->mb_to_top_edge = edge[0];
x->mb_to_bottom_edge = edge[1];
x->mb_to_left_edge = edge[2];
x->mb_to_right_edge = edge[3];
x->pre.y_buffer = y1;
x->pre.u_buffer = u1;
x->pre.v_buffer = v1;
if (x->mode_info_context->mbmi.second_ref_frame > 0) {
x->second_pre.y_buffer = y2;
x->second_pre.u_buffer = u2;
x->second_pre.v_buffer = v2;
}
#if CONFIG_COMP_INTERINTRA_PRED
if (x->mode_info_context->mbmi.second_ref_frame == INTRA_FRAME) {
vp9_build_interintra_64x64_predictors_sb(x, dst_y, dst_u, dst_v,
dst_ystride, dst_uvstride);
}
#endif
}
/*
* The following functions should be called after an initial
* call to vp9_build_1st_inter16x16_predictors_mb() or _mby()/_mbuv().
* It will run a second sixtap filter on a (different) ref
* It will run a second filter on a (different) ref
* frame and average the result with the output of the
* first sixtap filter. The second reference frame is stored
* first filter. The second reference frame is stored
* in x->second_pre (the reference frame index is in
* x->mode_info_context->mbmi.second_ref_frame). The second
* motion vector is x->mode_info_context->mbmi.second_mv.
@ -798,15 +693,15 @@ void vp9_build_inter32x32_predictors_sb(MACROBLOCKD *x,
* single reference framer.
*/
void vp9_build_2nd_inter16x16_predictors_mby(MACROBLOCKD *xd,
unsigned char *dst_y,
uint8_t *dst_y,
int dst_ystride) {
unsigned char *ptr;
uint8_t *ptr;
int_mv _16x16mv;
int mv_row;
int mv_col;
unsigned char *ptr_base = xd->second_pre.y_buffer;
uint8_t *ptr_base = xd->second_pre.y_buffer;
int pre_stride = xd->block[0].pre_stride;
_16x16mv.as_int = xd->mode_info_context->mbmi.mv[1].as_int;
@ -819,44 +714,20 @@ void vp9_build_2nd_inter16x16_predictors_mby(MACROBLOCKD *xd,
ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
#if CONFIG_PRED_FILTER
if (xd->mode_info_context->mbmi.pred_filter_enabled) {
if ((mv_row | mv_col) & 7) {
// Sub-pel filter needs extended input
int len = 15 + (VP9_INTERP_EXTEND << 1);
unsigned char Temp[32 * 32]; // Data required by sub-pel filter
unsigned char *pTemp = Temp + (VP9_INTERP_EXTEND - 1) * (len + 1);
// Copy extended MB into Temp array, applying the spatial filter
filter_mb(ptr - (VP9_INTERP_EXTEND - 1) * (pre_stride + 1), pre_stride,
Temp, len, len, len);
// Sub-pel filter
xd->subpixel_predict_avg16x16(pTemp, len, (mv_col & 7) << 1,
(mv_row & 7) << 1, dst_y, dst_ystride);
} else {
// TODO Needs to AVERAGE with the dst_y
// For now, do not apply the prediction filter in these cases!
vp9_avg_mem16x16(ptr, pre_stride, dst_y, dst_ystride);
}
} else
#endif // CONFIG_PRED_FILTER
{
if ((mv_row | mv_col) & 7) {
xd->subpixel_predict_avg16x16(ptr, pre_stride, (mv_col & 7) << 1,
(mv_row & 7) << 1, dst_y, dst_ystride);
} else {
vp9_avg_mem16x16(ptr, pre_stride, dst_y, dst_ystride);
}
if ((mv_row | mv_col) & 7) {
xd->subpixel_predict_avg16x16(ptr, pre_stride, (mv_col & 7) << 1,
(mv_row & 7) << 1, dst_y, dst_ystride);
} else {
vp9_avg_mem16x16(ptr, pre_stride, dst_y, dst_ystride);
}
}
void vp9_build_2nd_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_uvstride) {
int offset;
unsigned char *uptr, *vptr;
uint8_t *uptr, *vptr;
int_mv _16x16mv;
int mv_row;
@ -887,37 +758,6 @@ void vp9_build_2nd_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
uptr = xd->second_pre.u_buffer + offset;
vptr = xd->second_pre.v_buffer + offset;
#if CONFIG_PRED_FILTER
if (xd->mode_info_context->mbmi.pred_filter_enabled) {
int i;
int len = 7 + (VP9_INTERP_EXTEND << 1);
unsigned char Temp[32 * 32]; // Data required by sub-pel filter
unsigned char *pTemp = Temp + (VP9_INTERP_EXTEND - 1) * (len + 1);
unsigned char *pSrc = uptr;
unsigned char *pDst = dst_u;
// U & V
for (i = 0; i < 2; i++) {
if ((omv_row | omv_col) & 15) {
// Copy extended MB into Temp array, applying the spatial filter
filter_mb(pSrc - (VP9_INTERP_EXTEND - 1) * (pre_stride + 1), pre_stride,
Temp, len, len, len);
// Sub-pel filter
xd->subpixel_predict_avg8x8(pTemp, len, omv_col & 15,
omv_row & 15, pDst, dst_uvstride);
} else {
// TODO Needs to AVERAGE with the dst_[u|v]
// For now, do not apply the prediction filter here!
vp9_avg_mem8x8(pSrc, pre_stride, pDst, dst_uvstride);
}
// V
pSrc = vptr;
pDst = dst_v;
}
} else
#endif // CONFIG_PRED_FILTER
if ((omv_row | omv_col) & 15) {
xd->subpixel_predict_avg8x8(uptr, pre_stride, omv_col & 15,
omv_row & 15, dst_u, dst_uvstride);
@ -930,9 +770,9 @@ void vp9_build_2nd_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
}
void vp9_build_2nd_inter16x16_predictors_mb(MACROBLOCKD *xd,
unsigned char *dst_y,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride) {
vp9_build_2nd_inter16x16_predictors_mby(xd, dst_y, dst_ystride);
@ -995,13 +835,13 @@ static void build_inter4x4_predictors_mb(MACROBLOCKD *xd) {
if (d0->bmi.as_mv.first.as_int == d1->bmi.as_mv.first.as_int)
build_inter_predictors2b(xd, d0, 16);
else {
vp9_build_inter_predictors_b(d0, 16, xd->subpixel_predict);
vp9_build_inter_predictors_b(d1, 16, xd->subpixel_predict);
vp9_build_inter_predictors_b(d0, 16, xd->subpixel_predict4x4);
vp9_build_inter_predictors_b(d1, 16, xd->subpixel_predict4x4);
}
if (mbmi->second_ref_frame > 0) {
vp9_build_2nd_inter_predictors_b(d0, 16, xd->subpixel_predict_avg);
vp9_build_2nd_inter_predictors_b(d1, 16, xd->subpixel_predict_avg);
vp9_build_2nd_inter_predictors_b(d0, 16, xd->subpixel_predict_avg4x4);
vp9_build_2nd_inter_predictors_b(d1, 16, xd->subpixel_predict_avg4x4);
}
}
}
@ -1013,13 +853,13 @@ static void build_inter4x4_predictors_mb(MACROBLOCKD *xd) {
if (d0->bmi.as_mv.first.as_int == d1->bmi.as_mv.first.as_int)
build_inter_predictors2b(xd, d0, 8);
else {
vp9_build_inter_predictors_b(d0, 8, xd->subpixel_predict);
vp9_build_inter_predictors_b(d1, 8, xd->subpixel_predict);
vp9_build_inter_predictors_b(d0, 8, xd->subpixel_predict4x4);
vp9_build_inter_predictors_b(d1, 8, xd->subpixel_predict4x4);
}
if (mbmi->second_ref_frame > 0) {
vp9_build_2nd_inter_predictors_b(d0, 8, xd->subpixel_predict_avg);
vp9_build_2nd_inter_predictors_b(d1, 8, xd->subpixel_predict_avg);
vp9_build_2nd_inter_predictors_b(d0, 8, xd->subpixel_predict_avg4x4);
vp9_build_2nd_inter_predictors_b(d1, 8, xd->subpixel_predict_avg4x4);
}
}
}

42
vp9/common/vp9_reconinter.h
View File

@ -11,49 +11,55 @@
#ifndef VP9_COMMON_VP9_RECONINTER_H_
#define VP9_COMMON_VP9_RECONINTER_H_
#include "vpx/vpx_integer.h"
#include "vp9/common/vp9_onyxc_int.h"
extern void vp9_build_1st_inter16x16_predictors_mby(MACROBLOCKD *xd,
unsigned char *dst_y,
uint8_t *dst_y,
int dst_ystride,
int clamp_mvs);
extern void vp9_build_1st_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_uvstride);
extern void vp9_build_1st_inter16x16_predictors_mb(MACROBLOCKD *xd,
unsigned char *dst_y,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride);
extern void vp9_build_2nd_inter16x16_predictors_mby(MACROBLOCKD *xd,
unsigned char *dst_y,
uint8_t *dst_y,
int dst_ystride);
extern void vp9_build_2nd_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_uvstride);
extern void vp9_build_2nd_inter16x16_predictors_mb(MACROBLOCKD *xd,
unsigned char *dst_y,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride);
#if CONFIG_SUPERBLOCKS
extern void vp9_build_inter32x32_predictors_sb(MACROBLOCKD *x,
unsigned char *dst_y,
unsigned char *dst_u,
unsigned char *dst_v,
uint8_t *dst_y,
uint8_t *dst_u,
uint8_t *dst_v,
int dst_ystride,
int dst_uvstride);
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);
#endif
extern void vp9_build_inter_predictors_mb(MACROBLOCKD *xd);
@ -75,4 +81,4 @@ extern void vp9_setup_interp_filters(MACROBLOCKD *xd,
INTERPOLATIONFILTERTYPE filter,
VP9_COMMON *cm);
#endif // __INC_RECONINTER_H
#endif // VP9_COMMON_VP9_RECONINTER_H_

248
vp9/common/vp9_reconintra.c
View File

@ -124,18 +124,20 @@ static void d45_predictor(uint8_t *ypred_ptr, int y_stride, int n,
}
}
for (c = 0; c <= r; ++c) {
int yabove_ext = yabove_row[r]; // 2*yabove_row[r] - yabove_row[r-1];
int yleft_ext = yleft_col[r]; // 2*yleft_col[r] - yleft_col[r-1];
yabove_ext = (yabove_ext > 255 ? 255 : (yabove_ext < 0 ? 0 : yabove_ext));
yleft_ext = (yleft_ext > 255 ? 255 : (yleft_ext < 0 ? 0 : yleft_ext));
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);
}
for (r = 1; r < n; ++r) {
for (c = n - r; c < n; ++c)
ypred_ptr[r * y_stride + c] = (ypred_ptr[(r - 1) * y_stride + c] +
ypred_ptr[r * y_stride + c - 1] + 1) >> 1;
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;
}
}
}
@ -196,9 +198,9 @@ static void d153_predictor(uint8_t *ypred_ptr, int y_stride, int n,
}
}
static void corner_predictor(unsigned char *ypred_ptr, int y_stride, int n,
unsigned char *yabove_row,
unsigned char *yleft_col) {
static void corner_predictor(uint8_t *ypred_ptr, int y_stride, int n,
uint8_t *yabove_row,
uint8_t *yleft_col) {
int mh, mv, maxgradh, maxgradv, x, y, nx, ny;
int i, j;
int top_left = yabove_row[-1];
@ -246,14 +248,14 @@ void vp9_recon_intra_mbuv(MACROBLOCKD *xd) {
}
}
void vp9_build_intra_predictors_internal(unsigned char *src, int src_stride,
unsigned char *ypred_ptr,
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) {
unsigned char *yabove_row = src - src_stride;
unsigned char yleft_col[32];
unsigned char ytop_left = yabove_row[-1];
uint8_t *yabove_row = src - src_stride;
uint8_t yleft_col[64];
uint8_t ytop_left = yabove_row[-1];
int r, c, i;
for (i = 0; i < bsize; i++) {
@ -269,15 +271,19 @@ void vp9_build_intra_predictors_internal(unsigned char *src, int src_stride,
int average = 0;
int log2_bsize_minus_1;
assert(bsize == 4 || bsize == 8 || bsize == 16 || bsize == 32);
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 /* bsize == 32 */ {
} else if (bsize == 32) {
log2_bsize_minus_1 = 4;
} else {
assert(bsize == 64);
log2_bsize_minus_1 = 5;
}
if (up_available || left_available) {
@ -321,15 +327,7 @@ void vp9_build_intra_predictors_internal(unsigned char *src, int src_stride,
case TM_PRED: {
for (r = 0; r < bsize; r++) {
for (c = 0; c < bsize; c++) {
int pred = yleft_col[r] + yabove_row[ c] - ytop_left;
if (pred < 0)
pred = 0;
if (pred > 255)
pred = 255;
ypred_ptr[c] = pred;
ypred_ptr[c] = clip_pixel(yleft_col[r] + yabove_row[c] - ytop_left);
}
ypred_ptr += y_stride;
@ -374,9 +372,9 @@ void vp9_build_intra_predictors_internal(unsigned char *src, int src_stride,
#if CONFIG_COMP_INTERINTRA_PRED
static void combine_interintra(MB_PREDICTION_MODE mode,
unsigned char *interpred,
uint8_t *interpred,
int interstride,
unsigned char *intrapred,
uint8_t *intrapred,
int intrastride,
int size) {
// TODO(debargha): Explore different ways of combining predictors
@ -523,16 +521,17 @@ static void combine_interintra(MB_PREDICTION_MODE mode,
71, 70, 70, 70, 69, 69, 69, 68,
68, 68, 68, 68, 67, 67, 67, 67,
};
int size_scale = (size == 32 ? 1 :
int size_scale = (size >= 32 ? 1 :
size == 16 ? 2 :
size == 8 ? 4 : 8);
int size_shift = size == 64 ? 1 : 0;
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];
int scale = weights1d[i * size_scale >> size_shift];
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -545,7 +544,7 @@ 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];
int scale = weights1d[j * size_scale >> size_shift];
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -559,8 +558,9 @@ 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] +
weights1d[i * size_scale]) >> 1;
int scale = (weights2d[(i * size_scale * 32 +
j * size_scale) >> size_shift] +
weights1d[i * size_scale >> size_shift]) >> 1;
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -574,8 +574,9 @@ 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] +
weights1d[j * size_scale]) >> 1;
int scale = (weights2d[(i * size_scale * 32 +
j * size_scale) >> size_shift] +
weights1d[j * size_scale >> size_shift]) >> 1;
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -588,7 +589,8 @@ 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];
int scale = weights2d[(i * size_scale * 32 +
j * size_scale) >> size_shift];
interpred[k] =
((scale_max - scale) * interpred[k] +
scale * intrapred[i * intrastride + j] + scale_round)
@ -613,18 +615,18 @@ static void combine_interintra(MB_PREDICTION_MODE mode,
}
void vp9_build_interintra_16x16_predictors_mb(MACROBLOCKD *xd,
unsigned char *ypred,
unsigned char *upred,
unsigned char *vpred,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride, int uvstride) {
vp9_build_interintra_16x16_predictors_mby(xd, ypred, ystride);
vp9_build_interintra_16x16_predictors_mbuv(xd, upred, vpred, uvstride);
}
void vp9_build_interintra_16x16_predictors_mby(MACROBLOCKD *xd,
unsigned char *ypred,
uint8_t *ypred,
int ystride) {
unsigned char intrapredictor[256];
uint8_t intrapredictor[256];
vp9_build_intra_predictors_internal(
xd->dst.y_buffer, xd->dst.y_stride,
intrapredictor, 16,
@ -635,11 +637,11 @@ void vp9_build_interintra_16x16_predictors_mby(MACROBLOCKD *xd,
}
void vp9_build_interintra_16x16_predictors_mbuv(MACROBLOCKD *xd,
unsigned char *upred,
unsigned char *vpred,
uint8_t *upred,
uint8_t *vpred,
int uvstride) {
unsigned char uintrapredictor[64];
unsigned char vintrapredictor[64];
uint8_t uintrapredictor[64];
uint8_t vintrapredictor[64];
vp9_build_intra_predictors_internal(
xd->dst.u_buffer, xd->dst.uv_stride,
uintrapredictor, 8,
@ -656,11 +658,10 @@ void vp9_build_interintra_16x16_predictors_mbuv(MACROBLOCKD *xd,
vpred, uvstride, vintrapredictor, 8, 8);
}
#if CONFIG_SUPERBLOCKS
void vp9_build_interintra_32x32_predictors_sby(MACROBLOCKD *xd,
unsigned char *ypred,
uint8_t *ypred,
int ystride) {
unsigned char intrapredictor[1024];
uint8_t intrapredictor[1024];
vp9_build_intra_predictors_internal(
xd->dst.y_buffer, xd->dst.y_stride,
intrapredictor, 32,
@ -671,11 +672,11 @@ void vp9_build_interintra_32x32_predictors_sby(MACROBLOCKD *xd,
}
void vp9_build_interintra_32x32_predictors_sbuv(MACROBLOCKD *xd,
unsigned char *upred,
unsigned char *vpred,
uint8_t *upred,
uint8_t *vpred,
int uvstride) {
unsigned char uintrapredictor[256];
unsigned char vintrapredictor[256];
uint8_t uintrapredictor[256];
uint8_t vintrapredictor[256];
vp9_build_intra_predictors_internal(
xd->dst.u_buffer, xd->dst.uv_stride,
uintrapredictor, 16,
@ -693,16 +694,56 @@ void vp9_build_interintra_32x32_predictors_sbuv(MACROBLOCKD *xd,
}
void vp9_build_interintra_32x32_predictors_sb(MACROBLOCKD *xd,
unsigned char *ypred,
unsigned char *upred,
unsigned char *vpred,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride) {
vp9_build_interintra_32x32_predictors_sby(xd, ypred, ystride);
vp9_build_interintra_32x32_predictors_sbuv(xd, upred, vpred, uvstride);
}
#endif
#endif
void vp9_build_interintra_64x64_predictors_sby(MACROBLOCKD *xd,
uint8_t *ypred,
int ystride) {
uint8_t intrapredictor[4096];
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);
combine_interintra(xd->mode_info_context->mbmi.interintra_mode,
ypred, ystride, intrapredictor, 64, 64);
}
void vp9_build_interintra_64x64_predictors_sbuv(MACROBLOCKD *xd,
uint8_t *upred,
uint8_t *vpred,
int uvstride) {
uint8_t uintrapredictor[1024];
uint8_t vintrapredictor[1024];
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);
vp9_build_intra_predictors_internal(xd->dst.v_buffer, xd->dst.uv_stride,
vintrapredictor, 32, mode, 32,
xd->up_available, xd->left_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,
vpred, uvstride, vintrapredictor, 32, 32);
}
void vp9_build_interintra_64x64_predictors_sb(MACROBLOCKD *xd,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride) {
vp9_build_interintra_64x64_predictors_sby(xd, ypred, ystride);
vp9_build_interintra_64x64_predictors_sbuv(xd, upred, vpred, uvstride);
}
#endif // CONFIG_COMP_INTERINTRA_PRED
void vp9_build_intra_predictors_mby(MACROBLOCKD *xd) {
vp9_build_intra_predictors_internal(xd->dst.y_buffer, xd->dst.y_stride,
@ -718,40 +759,23 @@ void vp9_build_intra_predictors_mby_s(MACROBLOCKD *xd) {
xd->up_available, xd->left_available);
}
#if CONFIG_SUPERBLOCKS
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);
}
#endif
#if CONFIG_COMP_INTRA_PRED
void vp9_build_comp_intra_predictors_mby(MACROBLOCKD *xd) {
unsigned char predictor[2][256];
int i;
void vp9_build_intra_predictors_sb64y_s(MACROBLOCKD *xd) {
vp9_build_intra_predictors_internal(xd->dst.y_buffer, xd->dst.y_stride,
predictor[0], 16,
xd->mode_info_context->mbmi.mode,
16, xd->up_available,
xd->left_available);
vp9_build_intra_predictors_internal(xd->dst.y_buffer, xd->dst.y_stride,
predictor[1], 16,
xd->mode_info_context->mbmi.second_mode,
16, xd->up_available,
xd->left_available);
for (i = 0; i < 256; i++) {
xd->predictor[i] = (predictor[0][i] + predictor[1][i] + 1) >> 1;
}
xd->dst.y_buffer, xd->dst.y_stride,
xd->mode_info_context->mbmi.mode, 64,
xd->up_available, xd->left_available);
}
#endif
void vp9_build_intra_predictors_mbuv_internal(MACROBLOCKD *xd,
unsigned char *upred_ptr,
unsigned char *vpred_ptr,
uint8_t *upred_ptr,
uint8_t *vpred_ptr,
int uv_stride,
int mode, int bsize) {
vp9_build_intra_predictors_internal(xd->dst.u_buffer, xd->dst.uv_stride,
@ -777,86 +801,36 @@ void vp9_build_intra_predictors_mbuv_s(MACROBLOCKD *xd) {
8);
}
#if CONFIG_SUPERBLOCKS
void vp9_build_intra_predictors_sbuv_s(MACROBLOCKD *xd) {
vp9_build_intra_predictors_mbuv_internal(xd, xd->dst.u_buffer,
xd->dst.v_buffer, xd->dst.uv_stride,
xd->mode_info_context->mbmi.uv_mode,
16);
}
#endif
#if CONFIG_COMP_INTRA_PRED
void vp9_build_comp_intra_predictors_mbuv(MACROBLOCKD *xd) {
unsigned char predictor[2][2][64];
int i;
vp9_build_intra_predictors_mbuv_internal(
xd, predictor[0][0], predictor[1][0], 8,
xd->mode_info_context->mbmi.uv_mode, 8);
vp9_build_intra_predictors_mbuv_internal(
xd, predictor[0][1], predictor[1][1], 8,
xd->mode_info_context->mbmi.second_uv_mode, 8);
for (i = 0; i < 64; i++) {
xd->predictor[256 + i] = (predictor[0][0][i] + predictor[0][1][i] + 1) >> 1;
xd->predictor[256 + 64 + i] = (predictor[1][0][i] +
predictor[1][1][i] + 1) >> 1;
}
void vp9_build_intra_predictors_sb64uv_s(MACROBLOCKD *xd) {
vp9_build_intra_predictors_mbuv_internal(xd, xd->dst.u_buffer,
xd->dst.v_buffer, xd->dst.uv_stride,
xd->mode_info_context->mbmi.uv_mode,
32);
}
#endif
void vp9_intra8x8_predict(BLOCKD *xd,
int mode,
unsigned char *predictor) {
uint8_t *predictor) {
vp9_build_intra_predictors_internal(*(xd->base_dst) + xd->dst,
xd->dst_stride, predictor, 16,
mode, 8, 1, 1);
}
#if CONFIG_COMP_INTRA_PRED
void vp9_comp_intra8x8_predict(BLOCKD *xd,
int mode, int second_mode,
unsigned char *out_predictor) {
unsigned char predictor[2][8 * 16];
int i, j;
vp9_intra8x8_predict(xd, mode, predictor[0]);
vp9_intra8x8_predict(xd, second_mode, predictor[1]);
for (i = 0; i < 8 * 16; i += 16) {
for (j = i; j < i + 8; j++) {
out_predictor[j] = (predictor[0][j] + predictor[1][j] + 1) >> 1;
}
}
}
#endif
void vp9_intra_uv4x4_predict(BLOCKD *xd,
int mode,
unsigned char *predictor) {
uint8_t *predictor) {
vp9_build_intra_predictors_internal(*(xd->base_dst) + xd->dst,
xd->dst_stride, predictor, 8,
mode, 4, 1, 1);
}
#if CONFIG_COMP_INTRA_PRED
void vp9_comp_intra_uv4x4_predict(BLOCKD *xd,
int mode, int mode2,
unsigned char *out_predictor) {
unsigned char predictor[2][8 * 4];
int i, j;
vp9_intra_uv4x4_predict(xd, mode, predictor[0]);
vp9_intra_uv4x4_predict(xd, mode2, predictor[1]);
for (i = 0; i < 4 * 8; i += 8) {
for (j = i; j < i + 4; j++) {
out_predictor[j] = (predictor[0][j] + predictor[1][j] + 1) >> 1;
}
}
}
#endif
/* TODO: try different ways of use Y-UV mode correlation
Current code assumes that a uv 4x4 block use same mode
as corresponding Y 8x8 area

34
vp9/common/vp9_reconintra.h
View File

@ -11,34 +11,40 @@
#ifndef VP9_COMMON_VP9_RECONINTRA_H_
#define VP9_COMMON_VP9_RECONINTRA_H_
#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(unsigned char *ptr,
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);
#if CONFIG_COMP_INTERINTRA_PRED
extern void vp9_build_interintra_16x16_predictors_mb(MACROBLOCKD *xd,
unsigned char *ypred,
unsigned char *upred,
unsigned char *vpred,
uint8_t *ypred,
uint8_t *upred,
uint8_t *vpred,
int ystride,
int uvstride);
extern void vp9_build_interintra_16x16_predictors_mby(MACROBLOCKD *xd,
unsigned char *ypred,
uint8_t *ypred,
int ystride);
extern void vp9_build_interintra_16x16_predictors_mbuv(MACROBLOCKD *xd,
unsigned char *upred,
unsigned char *vpred,
uint8_t *upred,
uint8_t *vpred,
int uvstride);
#if CONFIG_SUPERBLOCKS
#endif // CONFIG_COMP_INTERINTRA_PRED
extern void vp9_build_interintra_32x32_predictors_sb(MACROBLOCKD *xd,
unsigned char *ypred,
unsigned char *upred,
unsigned char *vpred,
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);
#endif
#endif
#endif // __INC_RECONINTRA_H
#endif // VP9_COMMON_VP9_RECONINTRA_H_

185
vp9/common/vp9_reconintra4x4.c
View File

@ -15,7 +15,7 @@
#include "vp9_rtcd.h"
#if CONFIG_NEWBINTRAMODES
static int find_grad_measure(unsigned char *x, int stride, int n, int t,
static int find_grad_measure(uint8_t *x, int stride, int n, int t,
int dx, int dy) {
int i, j;
int count = 0, gsum = 0, gdiv;
@ -35,8 +35,8 @@ static int find_grad_measure(unsigned char *x, int stride, int n, int t,
}
#if CONTEXT_PRED_REPLACEMENTS == 6
B_PREDICTION_MODE vp9_find_dominant_direction(
unsigned char *ptr, int stride, int n) {
B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
int stride, int n) {
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);
@ -72,8 +72,8 @@ B_PREDICTION_MODE vp9_find_dominant_direction(
}
}
#elif CONTEXT_PRED_REPLACEMENTS == 4
B_PREDICTION_MODE vp9_find_dominant_direction(
unsigned char *ptr, int stride, int n) {
B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
int stride, int n) {
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);
@ -103,8 +103,8 @@ B_PREDICTION_MODE vp9_find_dominant_direction(
}
}
#elif CONTEXT_PRED_REPLACEMENTS == 0
B_PREDICTION_MODE vp9_find_dominant_direction(
unsigned char *ptr, int stride, int n) {
B_PREDICTION_MODE vp9_find_dominant_direction(uint8_t *ptr,
int stride, int n) {
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);
@ -145,7 +145,7 @@ B_PREDICTION_MODE vp9_find_dominant_direction(
#endif
B_PREDICTION_MODE vp9_find_bpred_context(BLOCKD *x) {
unsigned char *ptr = *(x->base_dst) + x->dst;
uint8_t *ptr = *(x->base_dst) + x->dst;
int stride = x->dst_stride;
return vp9_find_dominant_direction(ptr, stride, 4);
}
@ -153,17 +153,17 @@ B_PREDICTION_MODE vp9_find_bpred_context(BLOCKD *x) {
void vp9_intra4x4_predict(BLOCKD *x,
int b_mode,
unsigned char *predictor) {
uint8_t *predictor) {
int i, r, c;
unsigned char *Above = *(x->base_dst) + x->dst - x->dst_stride;
unsigned char Left[4];
unsigned char top_left = Above[-1];
uint8_t *above = *(x->base_dst) + x->dst - x->dst_stride;
uint8_t left[4];
uint8_t top_left = above[-1];
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] = (*(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];
#if CONFIG_NEWBINTRAMODES
if (b_mode == B_CONTEXT_PRED)
@ -175,8 +175,8 @@ void vp9_intra4x4_predict(BLOCKD *x,
int expected_dc = 0;
for (i = 0; i < 4; i++) {
expected_dc += Above[i];
expected_dc += Left[i];
expected_dc += above[i];
expected_dc += left[i];
}
expected_dc = (expected_dc + 4) >> 3;
@ -194,15 +194,7 @@ void vp9_intra4x4_predict(BLOCKD *x,
/* prediction similar to true_motion prediction */
for (r = 0; r < 4; r++) {
for (c = 0; c < 4; c++) {
int pred = Above[c] - top_left + Left[r];
if (pred < 0)
pred = 0;
if (pred > 255)
pred = 255;
predictor[c] = pred;
predictor[c] = clip_pixel(above[c] - top_left + left[r]);
}
predictor += 16;
@ -211,33 +203,30 @@ void vp9_intra4x4_predict(BLOCKD *x,
break;
case B_VE_PRED: {
unsigned int ap[4];
ap[0] = Above[0];
ap[1] = Above[1];
ap[2] = Above[2];
ap[3] = Above[3];
ap[0] = above[0];
ap[1] = above[1];
ap[2] = above[2];
ap[3] = above[3];
for (r = 0; r < 4; r++) {
for (c = 0; c < 4; c++) {
predictor[c] = ap[c];
}
predictor += 16;
}
}
break;
case B_HE_PRED: {
unsigned int lp[4];
lp[0] = Left[0];
lp[1] = Left[1];
lp[2] = Left[2];
lp[3] = Left[3];
lp[0] = left[0];
lp[1] = left[1];
lp[2] = left[2];
lp[3] = left[3];
for (r = 0; r < 4; r++) {
for (c = 0; c < 4; c++) {
@ -249,7 +238,8 @@ void vp9_intra4x4_predict(BLOCKD *x,
}
break;
case B_LD_PRED: {
unsigned char *ptr = Above;
uint8_t *ptr = above;
predictor[0 * 16 + 0] = (ptr[0] + ptr[1] * 2 + ptr[2] + 2) >> 2;
predictor[0 * 16 + 1] =
predictor[1 * 16 + 0] = (ptr[1] + ptr[2] * 2 + ptr[3] + 2) >> 2;
@ -270,18 +260,17 @@ void vp9_intra4x4_predict(BLOCKD *x,
}
break;
case B_RD_PRED: {
uint8_t pp[9];
unsigned char pp[9];
pp[0] = Left[3];
pp[1] = Left[2];
pp[2] = Left[1];
pp[3] = Left[0];
pp[0] = left[3];
pp[1] = left[2];
pp[2] = left[1];
pp[3] = left[0];
pp[4] = top_left;
pp[5] = Above[0];
pp[6] = Above[1];
pp[7] = Above[2];
pp[8] = Above[3];
pp[5] = above[0];
pp[6] = above[1];
pp[7] = above[2];
pp[8] = above[3];
predictor[3 * 16 + 0] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
predictor[3 * 16 + 1] =
@ -303,19 +292,17 @@ void vp9_intra4x4_predict(BLOCKD *x,
}
break;
case B_VR_PRED: {
uint8_t pp[9];
unsigned char pp[9];
pp[0] = Left[3];
pp[1] = Left[2];
pp[2] = Left[1];
pp[3] = Left[0];
pp[0] = left[3];
pp[1] = left[2];
pp[2] = left[1];
pp[3] = left[0];
pp[4] = top_left;
pp[5] = Above[0];
pp[6] = Above[1];
pp[7] = Above[2];
pp[8] = Above[3];
pp[5] = above[0];
pp[6] = above[1];
pp[7] = above[2];
pp[8] = above[3];
predictor[3 * 16 + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
predictor[2 * 16 + 0] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
@ -337,8 +324,7 @@ void vp9_intra4x4_predict(BLOCKD *x,
}
break;
case B_VL_PRED: {
unsigned char *pp = Above;
uint8_t *pp = above;
predictor[0 * 16 + 0] = (pp[0] + pp[1] + 1) >> 1;
predictor[1 * 16 + 0] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
@ -360,16 +346,17 @@ void vp9_intra4x4_predict(BLOCKD *x,
break;
case B_HD_PRED: {
unsigned char pp[9];
pp[0] = Left[3];
pp[1] = Left[2];
pp[2] = Left[1];
pp[3] = Left[0];
uint8_t pp[9];
pp[0] = left[3];
pp[1] = left[2];
pp[2] = left[1];
pp[3] = left[0];
pp[4] = top_left;
pp[5] = Above[0];
pp[6] = Above[1];
pp[7] = Above[2];
pp[8] = Above[3];
pp[5] = above[0];
pp[6] = above[1];
pp[7] = above[2];
pp[8] = above[3];
predictor[3 * 16 + 0] = (pp[0] + pp[1] + 1) >> 1;
@ -393,7 +380,7 @@ void vp9_intra4x4_predict(BLOCKD *x,
case B_HU_PRED: {
unsigned char *pp = Left;
uint8_t *pp = left;
predictor[0 * 16 + 0] = (pp[0] + pp[1] + 1) >> 1;
predictor[0 * 16 + 1] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
predictor[0 * 16 + 2] =
@ -418,48 +405,38 @@ void vp9_intra4x4_predict(BLOCKD *x,
break;
/*
case B_CORNER_PRED:
corner_predictor(predictor, 16, 4, Above, Left);
corner_predictor(predictor, 16, 4, above, left);
break;
*/
#endif
}
}
#if CONFIG_COMP_INTRA_PRED
void vp9_comp_intra4x4_predict_c(BLOCKD *x,
int b_mode, int b_mode2,
unsigned char *out_predictor) {
unsigned char predictor[2][4 * 16];
int i, j;
vp9_intra4x4_predict(x, b_mode, predictor[0]);
vp9_intra4x4_predict(x, b_mode2, predictor[1]);
for (i = 0; i < 16 * 4; i += 16) {
for (j = i; j < i + 4; j++) {
out_predictor[j] = (predictor[0][j] + predictor[1][j] + 1) >> 1;
}
}
}
#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);
unsigned char *above_right = *(xd->block[0].base_dst) + xd->block[0].dst -
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;
unsigned int *src_ptr = (unsigned int *)
(above_right - (xd->mb_index == 3 ? 16 * xd->block[0].dst_stride : 0));
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);
unsigned int *dst_ptr0 = (unsigned int *)above_right;
unsigned int *dst_ptr1 =
(unsigned int *)(above_right + 4 * xd->block[0].dst_stride);
unsigned int *dst_ptr2 =
(unsigned int *)(above_right + 8 * xd->block[0].dst_stride);
unsigned int *dst_ptr3 =
(unsigned int *)(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;

2
vp9/common/vp9_reconintra4x4.h
View File

@ -14,4 +14,4 @@
extern void vp9_intra_prediction_down_copy(MACROBLOCKD *xd);
#endif
#endif // VP9_COMMON_VP9_RECONINTRA4X4_H_

437
vp9/common/vp9_rtcd_defs.sh
View File

@ -4,6 +4,8 @@ cat <<EOF
* VP9
*/
#include "vpx/vpx_integer.h"
struct loop_filter_info;
struct blockd;
struct macroblockd;
@ -21,10 +23,10 @@ EOF
}
forward_decls vp9_common_forward_decls
prototype void vp9_filter_block2d_4x4_8 "const unsigned char *src_ptr, const unsigned int src_stride, const short *HFilter_aligned16, const short *VFilter_aligned16, unsigned char *dst_ptr, unsigned int dst_stride"
prototype void vp9_filter_block2d_8x4_8 "const unsigned char *src_ptr, const unsigned int src_stride, const short *HFilter_aligned16, const short *VFilter_aligned16, unsigned char *dst_ptr, unsigned int dst_stride"
prototype void vp9_filter_block2d_8x8_8 "const unsigned char *src_ptr, const unsigned int src_stride, const short *HFilter_aligned16, const short *VFilter_aligned16, unsigned char *dst_ptr, unsigned int dst_stride"
prototype void vp9_filter_block2d_16x16_8 "const unsigned char *src_ptr, const unsigned int src_stride, const short *HFilter_aligned16, const short *VFilter_aligned16, unsigned char *dst_ptr, unsigned int dst_stride"
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
@ -45,70 +47,76 @@ 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 "short *q, const short *dq, unsigned char *pre, unsigned char *dst, int stride, unsigned short *eobs, const short *dc, struct macroblockd *xd"
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 "short *q, const short *dq, unsigned char *pre, unsigned char *dst, int stride, unsigned short *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, uint16_t *eobs, struct macroblockd *xd"
specialize vp9_dequant_idct_add_y_block_8x8
prototype void vp9_dequant_idct_add_uv_block_8x8 "short *q, const short *dq, unsigned char *pre, unsigned char *dstu, unsigned char *dstv, int stride, unsigned short *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, uint16_t *eobs, struct macroblockd *xd"
specialize vp9_dequant_idct_add_uv_block_8x8
prototype void vp9_dequant_idct_add_16x16 "short *input, const short *dq, unsigned char *pred, unsigned char *dest, int pitch, int stride, unsigned short eobs"
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 "short *input, const short *dq, unsigned char *pred, unsigned char *dest, int pitch, int stride, int dc, unsigned short eobs"
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"
specialize vp9_dequant_idct_add_8x8
prototype void vp9_dequant_idct_add "short *input, const short *dq, unsigned char *pred, unsigned char *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"
specialize vp9_dequant_idct_add
prototype void vp9_dequant_dc_idct_add "short *input, const short *dq, unsigned char *pred, unsigned char *dest, int pitch, int stride, int Dc"
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 "short *q, const short *dq, unsigned char *pre, unsigned char *dst, int stride, unsigned short *eobs, const short *dc"
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 "short *q, const short *dq, unsigned char *pre, unsigned char *dst, int stride, unsigned short *eobs"
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"
specialize vp9_dequant_idct_add_y_block
prototype void vp9_dequant_idct_add_uv_block "short *q, const short *dq, unsigned char *pre, unsigned char *dstu, unsigned char *dstv, int stride, unsigned short *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, uint16_t *eobs"
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"
specialize vp9_dequant_idct_add_uv_block_16x16
#
# RECON
#
prototype void vp9_copy_mem16x16 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
prototype void vp9_copy_mem16x16 "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 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
prototype void vp9_copy_mem8x8 "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 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
prototype void vp9_copy_mem8x4 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
specialize vp9_copy_mem8x4 mmx
prototype void vp9_avg_mem16x16 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
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 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
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 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
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 "unsigned char *pred_ptr, short *diff_ptr, unsigned char *dst_ptr, int stride"
prototype void vp9_recon_b "uint8_t *pred_ptr, int16_t *diff_ptr, uint8_t *dst_ptr, int stride"
specialize vp9_recon_b
prototype void vp9_recon_uv_b "unsigned char *pred_ptr, short *diff_ptr, unsigned char *dst_ptr, int stride"
prototype void vp9_recon_uv_b "uint8_t *pred_ptr, int16_t *diff_ptr, uint8_t *dst_ptr, int stride"
specialize vp9_recon_uv_b
prototype void vp9_recon2b "unsigned char *pred_ptr, short *diff_ptr, unsigned char *dst_ptr, int stride"
prototype void vp9_recon2b "uint8_t *pred_ptr, int16_t *diff_ptr, uint8_t *dst_ptr, int stride"
specialize vp9_recon2b sse2
prototype void vp9_recon4b "unsigned char *pred_ptr, short *diff_ptr, unsigned char *dst_ptr, int stride"
prototype void vp9_recon4b "uint8_t *pred_ptr, int16_t *diff_ptr, uint8_t *dst_ptr, int stride"
specialize vp9_recon4b sse2
prototype void vp9_recon_mb "struct macroblockd *x"
@ -117,12 +125,18 @@ specialize vp9_recon_mb
prototype void vp9_recon_mby "struct macroblockd *x"
specialize vp9_recon_mby
prototype void vp9_recon_mby_s "struct macroblockd *x, unsigned char *dst"
prototype void vp9_recon_mby_s "struct macroblockd *x, uint8_t *dst"
specialize vp9_recon_mby_s
prototype void vp9_recon_mbuv_s "struct macroblockd *x, unsigned char *udst, unsigned char *vdst"
prototype void vp9_recon_mbuv_s "struct macroblockd *x, uint8_t *udst, uint8_t *vdst"
specialize void vp9_recon_mbuv_s
prototype void vp9_recon_sby_s "struct macroblockd *x, uint8_t *dst"
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_build_intra_predictors_mby_s "struct macroblockd *x"
specialize vp9_build_intra_predictors_mby_s
@ -135,9 +149,6 @@ specialize vp9_build_intra_predictors_sbuv_s;
prototype void vp9_build_intra_predictors_mby "struct macroblockd *x"
specialize vp9_build_intra_predictors_mby;
prototype void vp9_build_comp_intra_predictors_mby "struct macroblockd *x"
specialize vp9_build_comp_intra_predictors_mby;
prototype void vp9_build_intra_predictors_mby_s "struct macroblockd *x"
specialize vp9_build_intra_predictors_mby_s;
@ -147,262 +158,283 @@ specialize vp9_build_intra_predictors_mbuv;
prototype void vp9_build_intra_predictors_mbuv_s "struct macroblockd *x"
specialize vp9_build_intra_predictors_mbuv_s;
prototype void vp9_build_comp_intra_predictors_mbuv "struct macroblockd *x"
specialize vp9_build_comp_intra_predictors_mbuv;
prototype void vp9_build_intra_predictors_sb64y_s "struct macroblockd *x"
specialize vp9_build_intra_predictors_sb64y_s;
prototype void vp9_intra4x4_predict "struct blockd *x, int b_mode, unsigned char *predictor"
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"
specialize vp9_intra4x4_predict;
prototype void vp9_comp_intra4x4_predict "struct blockd *x, int b_mode, int second_mode, unsigned char *predictor"
specialize vp9_comp_intra4x4_predict;
prototype void vp9_intra8x8_predict "struct blockd *x, int b_mode, unsigned char *predictor"
prototype void vp9_intra8x8_predict "struct blockd *x, int b_mode, uint8_t *predictor"
specialize vp9_intra8x8_predict;
prototype void vp9_comp_intra8x8_predict "struct blockd *x, int b_mode, int second_mode, unsigned char *predictor"
specialize vp9_comp_intra8x8_predict;
prototype void vp9_intra_uv4x4_predict "struct blockd *x, int b_mode, unsigned char *predictor"
prototype void vp9_intra_uv4x4_predict "struct blockd *x, int b_mode, uint8_t *predictor"
specialize vp9_intra_uv4x4_predict;
prototype void vp9_comp_intra_uv4x4_predict "struct blockd *x, int b_mode, int second_mode, unsigned char *predictor"
specialize vp9_comp_intra_uv4x4_predict;
#
# Loopfilter
#
prototype void vp9_loop_filter_mbv "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
prototype void vp9_loop_filter_mbv "uint8_t *y, uint8_t *u, uint8_t *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
specialize vp9_loop_filter_mbv sse2
prototype void vp9_loop_filter_bv "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
prototype void vp9_loop_filter_bv "uint8_t *y, uint8_t *u, uint8_t *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
specialize vp9_loop_filter_bv sse2
prototype void vp9_loop_filter_bv8x8 "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
prototype void vp9_loop_filter_bv8x8 "uint8_t *y, uint8_t *u, uint8_t *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
specialize vp9_loop_filter_bv8x8 sse2
prototype void vp9_loop_filter_mbh "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
prototype void vp9_loop_filter_mbh "uint8_t *y, uint8_t *u, uint8_t *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
specialize vp9_loop_filter_mbh sse2
prototype void vp9_loop_filter_bh "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
prototype void vp9_loop_filter_bh "uint8_t *y, uint8_t *u, uint8_t *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
specialize vp9_loop_filter_bh sse2
prototype void vp9_loop_filter_bh8x8 "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
prototype void vp9_loop_filter_bh8x8 "uint8_t *y, uint8_t *u, uint8_t *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
specialize vp9_loop_filter_bh8x8 sse2
prototype void vp9_loop_filter_simple_mbv "unsigned char *y, int ystride, const unsigned char *blimit"
prototype void vp9_loop_filter_simple_mbv "uint8_t *y, int ystride, const uint8_t *blimit"
specialize vp9_loop_filter_simple_mbv mmx sse2
vp9_loop_filter_simple_mbv_c=vp9_loop_filter_simple_vertical_edge_c
vp9_loop_filter_simple_mbv_mmx=vp9_loop_filter_simple_vertical_edge_mmx
vp9_loop_filter_simple_mbv_sse2=vp9_loop_filter_simple_vertical_edge_sse2
prototype void vp9_loop_filter_simple_mbh "unsigned char *y, int ystride, const unsigned char *blimit"
prototype void vp9_loop_filter_simple_mbh "uint8_t *y, int ystride, const uint8_t *blimit"
specialize vp9_loop_filter_simple_mbh mmx sse2
vp9_loop_filter_simple_mbh_c=vp9_loop_filter_simple_horizontal_edge_c
vp9_loop_filter_simple_mbh_mmx=vp9_loop_filter_simple_horizontal_edge_mmx
vp9_loop_filter_simple_mbh_sse2=vp9_loop_filter_simple_horizontal_edge_sse2
prototype void vp9_loop_filter_simple_bv "unsigned char *y, int ystride, const unsigned char *blimit"
prototype void vp9_loop_filter_simple_bv "uint8_t *y, int ystride, const uint8_t *blimit"
specialize vp9_loop_filter_simple_bv mmx sse2
vp9_loop_filter_simple_bv_c=vp9_loop_filter_bvs_c
vp9_loop_filter_simple_bv_mmx=vp9_loop_filter_bvs_mmx
vp9_loop_filter_simple_bv_sse2=vp9_loop_filter_bvs_sse2
prototype void vp9_loop_filter_simple_bh "unsigned char *y, int ystride, const unsigned char *blimit"
prototype void vp9_loop_filter_simple_bh "uint8_t *y, int ystride, const uint8_t *blimit"
specialize vp9_loop_filter_simple_bh mmx sse2
vp9_loop_filter_simple_bh_c=vp9_loop_filter_bhs_c
vp9_loop_filter_simple_bh_mmx=vp9_loop_filter_bhs_mmx
vp9_loop_filter_simple_bh_sse2=vp9_loop_filter_bhs_sse2
prototype void vp9_lpf_mbh_w "unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr, int y_stride, int uv_stride, struct loop_filter_info *lfi"
specialize vp9_lpf_mbh_w sse2
prototype void vp9_lpf_mbv_w "unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr, int y_stride, int uv_stride, struct loop_filter_info *lfi"
specialize vp9_lpf_mbv_w sse2
#
# post proc
#
if [ "$CONFIG_POSTPROC" = "yes" ]; then
prototype void vp9_mbpost_proc_down "unsigned char *dst, int pitch, int rows, int cols, int flimit"
prototype void vp9_mbpost_proc_down "uint8_t *dst, int pitch, int rows, int cols, int flimit"
specialize vp9_mbpost_proc_down mmx sse2
vp9_mbpost_proc_down_sse2=vp9_mbpost_proc_down_xmm
prototype void vp9_mbpost_proc_across_ip "unsigned char *src, int pitch, int rows, int cols, int flimit"
prototype void vp9_mbpost_proc_across_ip "uint8_t *src, int pitch, int rows, int cols, int flimit"
specialize vp9_mbpost_proc_across_ip sse2
vp9_mbpost_proc_across_ip_sse2=vp9_mbpost_proc_across_ip_xmm
prototype 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"
prototype void vp9_post_proc_down_and_across "uint8_t *src_ptr, uint8_t *dst_ptr, int src_pixels_per_line, int dst_pixels_per_line, int rows, int cols, int flimit"
specialize vp9_post_proc_down_and_across mmx sse2
vp9_post_proc_down_and_across_sse2=vp9_post_proc_down_and_across_xmm
prototype void vp9_plane_add_noise "unsigned char *Start, char *noise, char blackclamp[16], char whiteclamp[16], char bothclamp[16], unsigned int Width, unsigned int Height, int Pitch"
prototype void vp9_plane_add_noise "uint8_t *Start, char *noise, char blackclamp[16], char whiteclamp[16], char bothclamp[16], unsigned int Width, unsigned int Height, int Pitch"
specialize vp9_plane_add_noise mmx sse2
vp9_plane_add_noise_sse2=vp9_plane_add_noise_wmt
fi
prototype void vp9_blend_mb_inner "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride"
prototype void vp9_blend_mb_inner "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride"
specialize vp9_blend_mb_inner
prototype void vp9_blend_mb_outer "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride"
prototype void vp9_blend_mb_outer "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride"
specialize vp9_blend_mb_outer
prototype void vp9_blend_b "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride"
prototype void vp9_blend_b "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride"
specialize vp9_blend_b
#
# sad 16x3, 3x16
#
prototype unsigned int vp9_sad16x3 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride"
prototype unsigned int vp9_sad16x3 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride"
specialize vp9_sad16x3 sse2
prototype unsigned int vp9_sad3x16 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride"
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"
specialize vp9_sub_pixel_variance16x2 sse2
#
# Sub Pixel Filters
#
prototype void vp9_eighttap_predict16x16 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict8x8 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict_avg16x16 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict_avg8x8 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict_avg4x4 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict8x4 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict
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
prototype void vp9_eighttap_predict16x16_sharp "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict8x8_sharp "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict_avg16x16_sharp "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict_avg8x8_sharp "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict_avg4x4_sharp "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict8x4_sharp "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_eighttap_predict_sharp "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
specialize vp9_eighttap_predict_sharp
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_sixtap_predict16x16 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_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_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_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_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_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_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_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_sixtap_predict8x8 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_sixtap_predict_avg16x16 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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_predict "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict
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_avg "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
specialize vp9_sixtap_predict_avg
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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict16x16 mmx sse2
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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict8x8 mmx 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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict8x4 mmx
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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
specialize vp9_bilinear_predict4x4 mmx
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 "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch"
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
#
# dct
#
prototype void vp9_short_idct4x4llm_1 "short *input, short *output, int pitch"
prototype void vp9_short_idct4x4llm_1 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct4x4llm_1
prototype void vp9_short_idct4x4llm "short *input, short *output, int pitch"
prototype void vp9_short_idct4x4llm "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct4x4llm
prototype void vp9_short_idct8x8 "short *input, short *output, int pitch"
prototype void vp9_short_idct8x8 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct8x8
prototype void vp9_short_idct10_8x8 "short *input, short *output, int pitch"
prototype void vp9_short_idct10_8x8 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct10_8x8
prototype void vp9_short_ihaar2x2 "short *input, short *output, int pitch"
prototype void vp9_short_ihaar2x2 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_ihaar2x2
prototype void vp9_short_idct16x16 "short *input, short *output, int pitch"
prototype void vp9_short_idct16x16 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct16x16
prototype void vp9_short_idct10_16x16 "short *input, short *output, int pitch"
prototype void vp9_short_idct10_16x16 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct10_16x16
prototype void vp9_ihtllm "const short *input, short *output, int pitch, int tx_type, int tx_dim, short eobs"
prototype void vp9_short_idct32x32 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct32x32
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
#
# 2nd order
#
prototype void vp9_short_inv_walsh4x4_1 "short *in, short *out"
prototype void vp9_short_inv_walsh4x4_1 "int16_t *in, int16_t *out"
specialize vp9_short_inv_walsh4x4_1
prototype void vp9_short_inv_walsh4x4 "short *in, short *out"
prototype void vp9_short_inv_walsh4x4 "int16_t *in, int16_t *out"
specialize vp9_short_inv_walsh4x4_
# dct and add
prototype void vp9_dc_only_idct_add_8x8 "short input_dc, unsigned char *pred_ptr, unsigned char *dst_ptr, int pitch, int stride"
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 "short input_dc, unsigned char *pred_ptr, unsigned char *dst_ptr, int pitch, int stride"
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
if [ "$CONFIG_LOSSLESS" = "yes" ]; then
prototype void vp9_short_inv_walsh4x4_1_x8 "short *input, short *output, int pitch"
prototype void vp9_short_inv_walsh4x4_x8 "short *input, short *output, int pitch"
prototype void vp9_dc_only_inv_walsh_add "short input_dc, unsigned char *pred_ptr, unsigned char *dst_ptr, int pitch, int stride"
prototype void vp9_short_inv_walsh4x4_1_lossless "short *in, short *out"
prototype void vp9_short_inv_walsh4x4_lossless "short *in, short *out"
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_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
if [ "$CONFIG_SUPERBLOCKS" = "yes" ]; then
prototype unsigned int vp9_sad32x3 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, int max_sad"
prototype unsigned int vp9_sad32x3 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, int max_sad"
specialize vp9_sad32x3
prototype unsigned int vp9_sad3x32 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, int max_sad"
prototype unsigned int vp9_sad3x32 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, int max_sad"
specialize vp9_sad3x32
fi
#
# Encoder functions below this point.
#
@ -412,154 +444,181 @@ if [ "$CONFIG_VP9_ENCODER" = "yes" ]; then
# variance
[ $arch = "x86_64" ] && mmx_x86_64=mmx && sse2_x86_64=sse2
prototype unsigned int vp9_variance32x32 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance32x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance32x32
prototype unsigned int vp9_variance16x16 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance64x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance64x64
prototype unsigned int vp9_variance16x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance16x16 mmx sse2
vp9_variance16x16_sse2=vp9_variance16x16_wmt
vp9_variance16x16_mmx=vp9_variance16x16_mmx
prototype unsigned int vp9_variance16x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance16x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance16x8 mmx sse2
vp9_variance16x8_sse2=vp9_variance16x8_wmt
vp9_variance16x8_mmx=vp9_variance16x8_mmx
prototype unsigned int vp9_variance8x16 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance8x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance8x16 mmx sse2
vp9_variance8x16_sse2=vp9_variance8x16_wmt
vp9_variance8x16_mmx=vp9_variance8x16_mmx
prototype unsigned int vp9_variance8x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance8x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance8x8 mmx sse2
vp9_variance8x8_sse2=vp9_variance8x8_wmt
vp9_variance8x8_mmx=vp9_variance8x8_mmx
prototype unsigned int vp9_variance4x4 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance4x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance4x4 mmx sse2
vp9_variance4x4_sse2=vp9_variance4x4_wmt
vp9_variance4x4_mmx=vp9_variance4x4_mmx
prototype unsigned int vp9_sub_pixel_variance32x32 "const unsigned char *src_ptr, int source_stride, int xoffset, int yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse"
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_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_variance16x16 "const unsigned char *src_ptr, int source_stride, int xoffset, int yoffset, const unsigned char *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 Refstride, 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 unsigned char *src_ptr, int source_stride, int xoffset, int yoffset, const unsigned char *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 Refstride, 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 unsigned char *src_ptr, int source_stride, int xoffset, int yoffset, const unsigned char *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 Refstride, 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 unsigned char *src_ptr, int source_stride, int xoffset, int yoffset, const unsigned char *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 Refstride, 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 unsigned char *src_ptr, int source_stride, int xoffset, int yoffset, const unsigned char *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 Refstride, unsigned int *sse"
specialize vp9_sub_pixel_variance4x4 sse2 mmx
vp9_sub_pixel_variance4x4_sse2=vp9_sub_pixel_variance4x4_wmt
prototype unsigned int vp9_sad32x32 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
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
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
prototype unsigned int vp9_sad16x16 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
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
prototype unsigned int vp9_sad16x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
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 unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
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 unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
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 unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
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
prototype unsigned int vp9_variance_halfpixvar16x16_h "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *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
vp9_variance_halfpixvar16x16_h_sse2=vp9_variance_halfpixvar16x16_h_wmt
prototype unsigned int vp9_variance_halfpixvar16x16_v "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance_halfpixvar16x16_v "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar16x16_v mmx sse2
vp9_variance_halfpixvar16x16_v_sse2=vp9_variance_halfpixvar16x16_v_wmt
prototype unsigned int vp9_variance_halfpixvar16x16_hv "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance_halfpixvar16x16_hv "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar16x16_hv mmx sse2
vp9_variance_halfpixvar16x16_hv_sse2=vp9_variance_halfpixvar16x16_hv_wmt
prototype unsigned int vp9_variance_halfpixvar32x32_h "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance_halfpixvar64x64_h "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar64x64_h
prototype unsigned int vp9_variance_halfpixvar64x64_v "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar64x64_v
prototype unsigned int vp9_variance_halfpixvar64x64_hv "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar64x64_hv
prototype unsigned int vp9_variance_halfpixvar32x32_h "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar32x32_h
prototype unsigned int vp9_variance_halfpixvar32x32_v "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance_halfpixvar32x32_v "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar32x32_v
prototype unsigned int vp9_variance_halfpixvar32x32_hv "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sse"
prototype unsigned int vp9_variance_halfpixvar32x32_hv "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance_halfpixvar32x32_hv
prototype void vp9_sad32x32x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array"
prototype void vp9_sad64x64x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad64x64x3
prototype void vp9_sad32x32x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad32x32x3
prototype void vp9_sad16x16x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array"
prototype void vp9_sad16x16x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad16x16x3 sse3 ssse3
prototype void vp9_sad16x8x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array"
prototype void vp9_sad16x8x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad16x8x3 sse3 ssse3
prototype void vp9_sad8x16x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array"
prototype void vp9_sad8x16x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad8x16x3 sse3
prototype void vp9_sad8x8x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array"
prototype void vp9_sad8x8x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array"
specialize vp9_sad8x8x3 sse3
prototype void vp9_sad4x4x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array"
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_sad32x32x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array"
prototype void vp9_sad64x64x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_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"
specialize vp9_sad32x32x8
prototype void vp9_sad16x16x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array"
prototype void vp9_sad16x16x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
specialize vp9_sad16x16x8 sse4
prototype void vp9_sad16x8x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array"
prototype void vp9_sad16x8x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
specialize vp9_sad16x8x8 sse4
prototype void vp9_sad8x16x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array"
prototype void vp9_sad8x16x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
specialize vp9_sad8x16x8 sse4
prototype void vp9_sad8x8x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array"
prototype void vp9_sad8x8x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
specialize vp9_sad8x8x8 sse4
prototype void vp9_sad4x4x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array"
prototype void vp9_sad4x4x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint16_t *sad_array"
specialize vp9_sad4x4x8 sse4
prototype void vp9_sad32x32x4d "const unsigned char *src_ptr, int src_stride, const unsigned char **ref_ptr, int ref_stride, unsigned int *sad_array"
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_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_sad16x16x4d "const unsigned char *src_ptr, int src_stride, const unsigned char **ref_ptr, int ref_stride, unsigned int *sad_array"
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_sad16x8x4d "const unsigned char *src_ptr, int src_stride, const unsigned char **ref_ptr, int ref_stride, unsigned int *sad_array"
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_sad8x16x4d "const unsigned char *src_ptr, int src_stride, const unsigned char **ref_ptr, int ref_stride, unsigned int *sad_array"
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_sad8x8x4d "const unsigned char *src_ptr, int src_stride, const unsigned char **ref_ptr, int ref_stride, unsigned int *sad_array"
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 unsigned char *src_ptr, int src_stride, const unsigned char **ref_ptr, int ref_stride, unsigned int *sad_array"
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
#
@ -567,30 +626,33 @@ specialize vp9_sad4x4x4d sse3
#
case $arch in
x86*)
prototype void vp9_copy32xn "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int n"
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 unsigned int vp9_sub_pixel_mse16x16 "const unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, const unsigned char *dst_ptr, int dst_pixels_per_line, unsigned int *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 unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int recon_stride, unsigned int *sse"
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_mse32x32 "const unsigned char *src_ptr, int source_stride, int xoffset, int yoffset, const unsigned char *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 Refstride, 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"
specialize vp9_sub_pixel_mse32x32
prototype unsigned int vp9_get_mb_ss "const short *"
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"
specialize vp9_mbblock_error mmx sse2
vp9_mbblock_error_sse2=vp9_mbblock_error_xmm
prototype int vp9_block_error "short *coeff, short *dqcoeff, int block_size"
prototype int vp9_block_error "int16_t *coeff, int16_t *dqcoeff, int block_size"
specialize vp9_block_error mmx sse2
vp9_block_error_sse2=vp9_block_error_xmm
@ -604,10 +666,10 @@ vp9_mbuverror_sse2=vp9_mbuverror_xmm
prototype void vp9_subtract_b "struct block *be, struct blockd *bd, int pitch"
specialize vp9_subtract_b mmx sse2
prototype void vp9_subtract_mby "short *diff, unsigned char *src, unsigned char *pred, int stride"
prototype void vp9_subtract_mby "int16_t *diff, uint8_t *src, uint8_t *pred, int stride"
specialize vp9_subtract_mby mmx sse2
prototype void vp9_subtract_mbuv "short *diff, unsigned char *usrc, unsigned char *vsrc, unsigned char *pred, int stride"
prototype void vp9_subtract_mbuv "int16_t *diff, uint8_t *usrc, uint8_t *vsrc, uint8_t *pred, int stride"
specialize vp9_subtract_mbuv mmx sse2
#
@ -616,42 +678,45 @@ specialize vp9_subtract_mbuv mmx sse2
if [ "$CONFIG_INTERNAL_STATS" = "yes" ]; then
[ $arch = "x86_64" ] && sse2_on_x86_64=sse2
prototype void vp9_ssim_parms_8x8 "unsigned char *s, int sp, unsigned char *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
prototype void vp9_ssim_parms_8x8 "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
specialize vp9_ssim_parms_8x8 $sse2_on_x86_64
prototype void vp9_ssim_parms_16x16 "unsigned char *s, int sp, unsigned char *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
prototype void vp9_ssim_parms_16x16 "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
specialize vp9_ssim_parms_16x16 $sse2_on_x86_64
fi
# fdct functions
prototype void vp9_fht "const short *input, int pitch, short *output, int tx_type, int tx_dim"
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_fdct8x8 "short *InputData, short *OutputData, int pitch"
prototype void vp9_short_fdct8x8 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct8x8
prototype void vp9_short_fhaar2x2 "short *InputData, short *OutputData, int pitch"
prototype void vp9_short_fhaar2x2 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fhaar2x2
prototype void vp9_short_fdct4x4 "short *InputData, short *OutputData, int pitch"
prototype void vp9_short_fdct4x4 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct4x4
prototype void vp9_short_fdct8x4 "short *InputData, short *OutputData, int pitch"
prototype void vp9_short_fdct8x4 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct8x4
prototype void vp9_short_walsh4x4 "short *InputData, short *OutputData, int pitch"
prototype void vp9_short_walsh4x4 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh4x4
prototype void vp9_short_fdct16x16 "short *InputData, short *OutputData, int pitch"
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
prototype void vp9_short_walsh4x4_lossless "short *InputData, short *OutputData, int pitch"
prototype void vp9_short_walsh4x4_lossless "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh4x4_lossless
prototype void vp9_short_walsh4x4_x8 "short *InputData, short *OutputData, int pitch"
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 "short *InputData, short *OutputData, int pitch"
prototype void vp9_short_walsh8x4_x8 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_walsh8x4_x8
#
@ -670,7 +735,7 @@ prototype int vp9_diamond_search_sad "struct macroblock *x, struct block *b, str
specialize vp9_diamond_search_sad sse3
vp9_diamond_search_sad_sse3=vp9_diamond_search_sadx4
prototype void vp9_temporal_filter_apply "unsigned char *frame1, unsigned int stride, unsigned char *frame2, unsigned int block_size, int strength, int filter_weight, unsigned int *accumulator, unsigned short *count"
prototype void vp9_temporal_filter_apply "uint8_t *frame1, unsigned int stride, uint8_t *frame2, unsigned int block_size, int strength, int filter_weight, unsigned int *accumulator, uint16_t *count"
specialize vp9_temporal_filter_apply sse2
prototype void vp9_yv12_copy_partial_frame "struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc, int fraction"

18
vp9/common/vp9_sadmxn.h
View File

@ -11,14 +11,14 @@
#ifndef VP9_COMMON_VP9_SADMXN_H_
#define VP9_COMMON_VP9_SADMXN_H_
static __inline
unsigned int sad_mx_n_c(
const unsigned char *src_ptr,
int src_stride,
const unsigned char *ref_ptr,
int ref_stride,
int m,
int n) {
#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) {
int r, c;
unsigned int sad = 0;
@ -34,4 +34,4 @@ unsigned int sad_mx_n_c(
return sad;
}
#endif
#endif // VP9_COMMON_VP9_SADMXN_H_

2
vp9/common/vp9_seg_common.c
View File

@ -14,7 +14,7 @@
static const int segfeaturedata_signed[SEG_LVL_MAX] = { 1, 1, 0, 0, 0, 0 };
static const int seg_feature_data_max[SEG_LVL_MAX] =
{ MAXQ, 63, 0xf, MB_MODE_COUNT - 1, 255, TX_SIZE_MAX - 1};
{ MAXQ, 63, 0xf, MB_MODE_COUNT - 1, 255, TX_SIZE_MAX_SB - 1};
// These functions provide access to new segment level features.
// Eventually these function may be "optimized out" but for the moment,

3
vp9/common/vp9_seg_common.h
View File

@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vp9/common/vp9_type_aliases.h"
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/common/vp9_blockd.h"
@ -60,5 +59,5 @@ int vp9_check_segref_inter(MACROBLOCKD *xd, int segment_id);
int vp9_get_seg_tx_type(MACROBLOCKD *xd, int segment_id);
#endif /* __INC_SEG_COMMON_H__ */
#endif // VP9_COMMON_VP9_SEG_COMMON_H_

6
vp9/common/vp9_setupintrarecon.c
View File

@ -18,14 +18,14 @@ void vp9_setup_intra_recon(YV12_BUFFER_CONFIG *ybf) {
/* set up frame new frame for intra coded blocks */
vpx_memset(ybf->y_buffer - 1 - ybf->y_stride, 127, ybf->y_width + 5);
for (i = 0; i < ybf->y_height; i++)
ybf->y_buffer[ybf->y_stride * i - 1] = (unsigned char) 129;
ybf->y_buffer[ybf->y_stride * i - 1] = (uint8_t) 129;
vpx_memset(ybf->u_buffer - 1 - ybf->uv_stride, 127, ybf->uv_width + 5);
for (i = 0; i < ybf->uv_height; i++)
ybf->u_buffer[ybf->uv_stride * i - 1] = (unsigned char) 129;
ybf->u_buffer[ybf->uv_stride * i - 1] = (uint8_t) 129;
vpx_memset(ybf->v_buffer - 1 - ybf->uv_stride, 127, ybf->uv_width + 5);
for (i = 0; i < ybf->uv_height; i++)
ybf->v_buffer[ybf->uv_stride * i - 1] = (unsigned char) 129;
ybf->v_buffer[ybf->uv_stride * i - 1] = (uint8_t) 129;
}

5
vp9/common/vp9_setupintrarecon.h
View File

@ -8,6 +8,11 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_SETUPINTRARECON_H_
#define VP9_COMMON_VP9_SETUPINTRARECON_H_
#include "vpx_scale/yv12config.h"
extern void vp9_setup_intra_recon(YV12_BUFFER_CONFIG *ybf);
#endif // VP9_COMMON_VP9_SETUPINTRARECON_H_

51
vp9/common/vp9_subpelvar.h
View File

@ -8,14 +8,14 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_SUBPELVAR_H_
#define VP9_COMMON_VP9_SUBPELVAR_H_
#include "vp9/common/vp9_filter.h"
static void variance(const unsigned char *src_ptr,
static void variance(const uint8_t *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
const uint8_t *ref_ptr,
int recon_stride,
int w,
int h,
@ -43,14 +43,14 @@ static void variance(const unsigned char *src_ptr,
*
* ROUTINE : filter_block2d_bil_first_pass
*
* INPUTS : UINT8 *src_ptr : Pointer to source block.
* UINT32 src_pixels_per_line : Stride of input block.
* UINT32 pixel_step : Offset between filter input samples (see notes).
* UINT32 output_height : Input block height.
* UINT32 output_width : Input block width.
* INT32 *vp9_filter : Array of 2 bi-linear filter taps.
* INPUTS : uint8_t *src_ptr : Pointer to source block.
* uint32_t src_pixels_per_line : Stride of input block.
* uint32_t pixel_step : Offset between filter input samples (see notes).
* uint32_t output_height : Input block height.
* uint32_t output_width : Input block width.
* int32_t *vp9_filter : Array of 2 bi-linear filter taps.
*
* OUTPUTS : INT32 *output_ptr : Pointer to filtered block.
* OUTPUTS : int32_t *output_ptr : Pointer to filtered block.
*
* RETURNS : void
*
@ -59,7 +59,7 @@ static void variance(const unsigned char *src_ptr,
* filtered output block. Used to implement first-pass
* of 2-D separable filter.
*
* SPECIAL NOTES : Produces INT32 output to retain precision for next pass.
* SPECIAL NOTES : Produces int32_t output to retain precision for next pass.
* Two filter taps should sum to VP9_FILTER_WEIGHT.
* pixel_step defines whether the filter is applied
* horizontally (pixel_step=1) or vertically (pixel_step=stride).
@ -67,13 +67,13 @@ static void variance(const unsigned char *src_ptr,
* to the next.
*
****************************************************************************/
static void var_filter_block2d_bil_first_pass(const unsigned char *src_ptr,
unsigned short *output_ptr,
static void var_filter_block2d_bil_first_pass(const uint8_t *src_ptr,
uint16_t *output_ptr,
unsigned int src_pixels_per_line,
int pixel_step,
unsigned int output_height,
unsigned int output_width,
const short *vp9_filter) {
const int16_t *vp9_filter) {
unsigned int i, j;
for (i = 0; i < output_height; i++) {
@ -95,14 +95,14 @@ static void var_filter_block2d_bil_first_pass(const unsigned char *src_ptr,
*
* ROUTINE : filter_block2d_bil_second_pass
*
* INPUTS : INT32 *src_ptr : Pointer to source block.
* UINT32 src_pixels_per_line : Stride of input block.
* UINT32 pixel_step : Offset between filter input samples (see notes).
* UINT32 output_height : Input block height.
* UINT32 output_width : Input block width.
* INT32 *vp9_filter : Array of 2 bi-linear filter taps.
* INPUTS : int32_t *src_ptr : Pointer to source block.
* uint32_t src_pixels_per_line : Stride of input block.
* uint32_t pixel_step : Offset between filter input samples (see notes).
* uint32_t output_height : Input block height.
* uint32_t output_width : Input block width.
* int32_t *vp9_filter : Array of 2 bi-linear filter taps.
*
* OUTPUTS : UINT16 *output_ptr : Pointer to filtered block.
* OUTPUTS : uint16_t *output_ptr : Pointer to filtered block.
*
* RETURNS : void
*
@ -119,13 +119,13 @@ static void var_filter_block2d_bil_first_pass(const unsigned char *src_ptr,
* to the next.
*
****************************************************************************/
static void var_filter_block2d_bil_second_pass(const unsigned short *src_ptr,
unsigned char *output_ptr,
static void var_filter_block2d_bil_second_pass(const uint16_t *src_ptr,
uint8_t *output_ptr,
unsigned int src_pixels_per_line,
unsigned int pixel_step,
unsigned int output_height,
unsigned int output_width,
const short *vp9_filter) {
const int16_t *vp9_filter) {
unsigned int i, j;
int Temp;
@ -145,3 +145,4 @@ static void var_filter_block2d_bil_second_pass(const unsigned short *src_ptr,
}
}
#endif // VP9_COMMON_VP9_SUBPELVAR_H_

7
vp9/common/vp9_subpixel.h
View File

@ -8,14 +8,13 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_SUBPIXEL_H_
#define VP9_COMMON_VP9_SUBPIXEL_H_
#define prototype_subpixel_predict(sym) \
void sym(unsigned char *src, int src_pitch, int xofst, int yofst, \
unsigned char *dst, int dst_pitch)
void sym(uint8_t *src, int src_pitch, int xofst, int yofst, \
uint8_t *dst, int dst_pitch)
typedef prototype_subpixel_predict((*vp9_subpix_fn_t));
#endif
#endif // VP9_COMMON_VP9_SUBPIXEL_H_

2
vp9/common/vp9_swapyv12buffer.c
View File

@ -12,7 +12,7 @@
void vp9_swap_yv12_buffer(YV12_BUFFER_CONFIG *new_frame,
YV12_BUFFER_CONFIG *last_frame) {
unsigned char *temp;
uint8_t *temp;
temp = last_frame->buffer_alloc;
last_frame->buffer_alloc = new_frame->buffer_alloc;

2
vp9/common/vp9_swapyv12buffer.h
View File

@ -16,4 +16,4 @@
void vp9_swap_yv12_buffer(YV12_BUFFER_CONFIG *new_frame,
YV12_BUFFER_CONFIG *last_frame);
#endif // __SWAPYV12_BUFFER_H
#endif // VP9_COMMON_VP9_SWAPYV12BUFFER_H_

18
vp9/common/vp9_systemdependent.h
View File

@ -7,9 +7,14 @@
* 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_SYSTEMDEPENDENT_H_
#define VP9_COMMON_VP9_SYSTEMDEPENDENT_H_
#ifdef _MSC_VER
#include <math.h>
#endif
#include "./vpx_config.h"
#if ARCH_X86 || ARCH_X86_64
void vpx_reset_mmx_state(void);
@ -18,6 +23,17 @@ void vpx_reset_mmx_state(void);
#define vp9_clear_system_state()
#endif
#ifdef _MSC_VER
// round is not defined in MSVC
static int round(double x) {
if (x < 0)
return (int)ceil(x - 0.5);
else
return (int)floor(x + 0.5);
}
#endif
struct VP9Common;
void vp9_machine_specific_config(struct VP9Common *);
#endif
#endif // VP9_COMMON_VP9_SYSTEMDEPENDENT_H_

2
vp9/common/vp9_textblit.h
View File

@ -16,4 +16,4 @@ extern void vp9_blit_text(const char *msg, unsigned char *address,
extern void vp9_blit_line(int x0, int x1, int y0, int y1,
unsigned char *image, const int pitch);
#endif // __INC_TEXTBLIT_H
#endif // VP9_COMMON_VP9_TEXTBLIT_H_

29
vp9/common/vp9_treecoder.c
View File

@ -100,9 +100,7 @@ void vp9_tree_probs_from_distribution(
vp9_tree tree,
vp9_prob probs [ /* n-1 */ ],
unsigned int branch_ct [ /* n-1 */ ] [2],
const unsigned int num_events[ /* n */ ],
unsigned int Pfac,
int rd
const unsigned int num_events[ /* n */ ]
) {
const int tree_len = n - 1;
int t = 0;
@ -110,29 +108,6 @@ void vp9_tree_probs_from_distribution(
branch_counts(n, tok, tree, branch_ct, num_events);
do {
const unsigned int *const c = branch_ct[t];
const unsigned int tot = c[0] + c[1];
#if CONFIG_DEBUG
assert(tot < (1 << 24)); /* no overflow below */
#endif
if (tot) {
const unsigned int p = ((c[0] * Pfac) + (rd ? tot >> 1 : 0)) / tot;
probs[t] = p < 256 ? (p ? p : 1) : 255; /* agree w/old version for now */
} else
probs[t] = vp9_prob_half;
probs[t] = get_binary_prob(branch_ct[t][0], branch_ct[t][1]);
} while (++t < tree_len);
}
vp9_prob vp9_bin_prob_from_distribution(const unsigned int counts[2]) {
int tot_count = counts[0] + counts[1];
vp9_prob prob;
if (tot_count) {
prob = (counts[0] * 255 + (tot_count >> 1)) / tot_count;
prob += !prob;
} else {
prob = 128;
}
return prob;
}

62
vp9/common/vp9_treecoder.h
View File

@ -8,29 +8,18 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_TREECODER_H_
#define VP9_COMMON_VP9_TREECODER_H_
typedef unsigned char vp9_prob;
#include "vpx/vpx_integer.h"
#define vp9_prob_half ( (vp9_prob) 128)
typedef uint8_t vp9_prob;
typedef signed char vp9_tree_index;
struct bool_coder_spec;
#define vp9_prob_half ((vp9_prob) 128)
typedef struct bool_coder_spec bool_coder_spec;
typedef struct bool_writer bool_writer;
typedef struct bool_reader bool_reader;
typedef const bool_coder_spec c_bool_coder_spec;
typedef const bool_writer c_bool_writer;
typedef const bool_reader c_bool_reader;
# define vp9_complement( x) (255 - x)
typedef int8_t vp9_tree_index;
#define vp9_complement(x) (255 - x)
/* We build coding trees compactly in arrays.
Each node of the tree is a pair of vp9_tree_indices.
@ -41,7 +30,6 @@ typedef const bool_reader c_bool_reader;
typedef const vp9_tree_index vp9_tree[], *vp9_tree_p;
typedef const struct vp9_token_struct {
int value;
int Len;
@ -53,31 +41,33 @@ void vp9_tokens_from_tree(struct vp9_token_struct *, vp9_tree);
void vp9_tokens_from_tree_offset(struct vp9_token_struct *, vp9_tree,
int offset);
/* Convert array of token occurrence counts into a table of probabilities
for the associated binary encoding tree. Also writes count of branches
taken for each node on the tree; this facilitiates decisions as to
probability updates. */
void vp9_tree_probs_from_distribution(
int n, /* n = size of alphabet */
vp9_token tok [ /* n */ ],
vp9_tree tree,
vp9_prob probs [ /* n-1 */ ],
unsigned int branch_ct [ /* n-1 */ ] [2],
const unsigned int num_events[ /* n */ ],
unsigned int Pfactor,
int Round
);
void vp9_tree_probs_from_distribution(int n, /* n = size of alphabet */
vp9_token tok[ /* n */ ],
vp9_tree tree,
vp9_prob probs[ /* n - 1 */ ],
unsigned int branch_ct[ /* n - 1 */ ][2],
const unsigned int num_events[ /* n */ ]);
static __inline int clip_prob(int p) {
if (p > 255)
return 255;
else if (p < 1)
return 1;
return p;
static __inline vp9_prob clip_prob(int p) {
return (p > 255) ? 255u : (p < 1) ? 1u : p;
}
vp9_prob vp9_bin_prob_from_distribution(const unsigned int counts[2]);
static __inline vp9_prob get_prob(int num, int den) {
return (den == 0) ? 128u : clip_prob((num * 256 + (den >> 1)) / den);
}
#endif
static __inline vp9_prob get_binary_prob(int n0, int n1) {
return get_prob(n0, n0 + n1);
}
/* this function assumes prob1 and prob2 are already within [1,255] range */
static __inline vp9_prob weighted_prob(int prob1, int prob2, int factor) {
return (prob1 * (256 - factor) + prob2 * factor + 128) >> 8;
}
#endif // VP9_COMMON_VP9_TREECODER_H_

122
vp9/common/vp9_type_aliases.h
View File

@ -1,122 +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.
*/
/****************************************************************************
*
* Module Title : vp9_type_aliases.h
*
* Description : Standard type aliases
*
****************************************************************************/
#ifndef VP9_COMMON_VP9_TYPE_ALIASES_H_
#define VP9_COMMON_VP9_TYPE_ALIASES_H_
/****************************************************************************
* Macros
****************************************************************************/
#define EXPORT
#define IMPORT extern /* Used to declare imported data & routines */
#define PRIVATE static /* Used to declare & define module-local data */
#define LOCAL static /* Used to define all persistent routine-local data */
#define STD_IN_PATH 0 /* Standard input path */
#define STD_OUT_PATH 1 /* Standard output path */
#define STD_ERR_PATH 2 /* Standard error path */
#define STD_IN_FILE stdin /* Standard input file pointer */
#define STD_OUT_FILE stdout /* Standard output file pointer */
#define STD_ERR_FILE stderr /* Standard error file pointer */
#define max_int 0x7FFFFFFF
#define __export
#define _export
#define CCONV
#ifndef NULL
#ifdef __cplusplus
#define NULL 0
#else
#define NULL ((void *)0)
#endif
#endif
#ifndef FALSE
#define FALSE 0
#endif
#ifndef TRUE
#define TRUE 1
#endif
/****************************************************************************
* Typedefs
****************************************************************************/
#ifndef TYPE_INT8
#define TYPE_INT8
typedef signed char INT8;
#endif
#ifndef TYPE_INT16
/*#define TYPE_INT16*/
typedef signed short INT16;
#endif
#ifndef TYPE_INT32
/*#define TYPE_INT32*/
typedef signed int INT32;
#endif
#ifndef TYPE_UINT8
/*#define TYPE_UINT8*/
typedef unsigned char UINT8;
#endif
#ifndef TYPE_UINT32
/*#define TYPE_UINT32*/
typedef unsigned int UINT32;
#endif
#ifndef TYPE_UINT16
/*#define TYPE_UINT16*/
typedef unsigned short UINT16;
#endif
#ifndef TYPE_BOOL
/*#define TYPE_BOOL*/
typedef int BOOL;
#endif
typedef unsigned char BOOLEAN;
#ifdef _MSC_VER
typedef __int64 INT64;
#if _MSC_VER < 1600
#ifndef INT64_MAX
#define INT64_MAX LLONG_MAX
#endif
#endif
#else
#ifndef TYPE_INT64
#ifdef _TMS320C6X
/* for now we only have 40bits */
typedef long INT64;
#else
typedef long long INT64;
#endif
#endif
#endif
/* Floating point */
typedef double FLOAT64;
typedef float FLOAT32;
#endif

24
vp9/common/x86/vp9_asm_stubs.c
View File

@ -15,8 +15,6 @@
extern const short vp9_six_tap_mmx[8][6 * 8];
extern const short vp9_bilinear_filters_8x_mmx[8][2 * 8];
extern void vp9_filter_block1d_h6_mmx(unsigned char *src_ptr,
unsigned short *output_ptr,
unsigned int src_pixels_per_line,
@ -95,8 +93,6 @@ extern void vp9_filter_block1d8_v6_only_sse2(unsigned char *src_ptr,
unsigned int output_height,
const short *vp9_filter);
extern prototype_subpixel_predict(vp9_bilinear_predict8x8_mmx);
///////////////////////////////////////////////////////////////////////////
// the mmx function that does the bilinear filtering and var calculation //
// int one pass //
@ -232,26 +228,6 @@ void vp9_sixtap_predict8x4_mmx(unsigned char *src_ptr,
vp9_filter_block1dc_v6_mmx(fdata2 + 20, dst_ptr + 4, dst_pitch,
16, 8, 4, 8, vfilter);
}
void vp9_bilinear_predict16x16_mmx(unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
unsigned char *dst_ptr,
int dst_pitch) {
vp9_bilinear_predict8x8_mmx(src_ptr,
src_pixels_per_line, xoffset, yoffset,
dst_ptr, dst_pitch);
vp9_bilinear_predict8x8_mmx(src_ptr + 8,
src_pixels_per_line, xoffset, yoffset,
dst_ptr + 8, dst_pitch);
vp9_bilinear_predict8x8_mmx(src_ptr + 8 * src_pixels_per_line,
src_pixels_per_line, xoffset, yoffset,
dst_ptr + dst_pitch * 8, dst_pitch);
vp9_bilinear_predict8x8_mmx(src_ptr + 8 * src_pixels_per_line + 8,
src_pixels_per_line, xoffset, yoffset,
dst_ptr + dst_pitch * 8 + 8, dst_pitch);
}
#endif
#if HAVE_SSE2

544
vp9/common/x86/vp9_loopfilter_x86.c
View File

@ -85,6 +85,480 @@ void vp9_loop_filter_bvs_mmx(unsigned char *y_ptr, int y_stride,
#endif
#if HAVE_SSE2
void vp9_mb_lpf_horizontal_edge_w_sse2(unsigned char *s,
int p,
const unsigned char *_blimit,
const unsigned char *_limit,
const unsigned char *_thresh) {
DECLARE_ALIGNED(16, unsigned char, flat2_op[7][16]);
DECLARE_ALIGNED(16, unsigned char, flat2_oq[7][16]);
DECLARE_ALIGNED(16, unsigned char, flat_op2[16]);
DECLARE_ALIGNED(16, unsigned char, flat_op1[16]);
DECLARE_ALIGNED(16, unsigned char, flat_op0[16]);
DECLARE_ALIGNED(16, unsigned char, flat_oq0[16]);
DECLARE_ALIGNED(16, unsigned char, flat_oq1[16]);
DECLARE_ALIGNED(16, unsigned char, flat_oq2[16]);
__m128i mask, hev, flat, flat2;
const __m128i zero = _mm_set1_epi16(0);
__m128i p7, p6, p5;
__m128i p4, p3, p2, p1, p0, q0, q1, q2, q3, q4;
__m128i q5, q6, q7;
int i = 0;
const unsigned int extended_thresh = _thresh[0] * 0x01010101u;
const unsigned int extended_limit = _limit[0] * 0x01010101u;
const unsigned int extended_blimit = _blimit[0] * 0x01010101u;
const __m128i thresh =
_mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_thresh), 0);
const __m128i limit =
_mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_limit), 0);
const __m128i blimit =
_mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_blimit), 0);
p4 = _mm_loadu_si128((__m128i *)(s - 5 * p));
p3 = _mm_loadu_si128((__m128i *)(s - 4 * p));
p2 = _mm_loadu_si128((__m128i *)(s - 3 * p));
p1 = _mm_loadu_si128((__m128i *)(s - 2 * p));
p0 = _mm_loadu_si128((__m128i *)(s - 1 * p));
q0 = _mm_loadu_si128((__m128i *)(s - 0 * p));
q1 = _mm_loadu_si128((__m128i *)(s + 1 * p));
q2 = _mm_loadu_si128((__m128i *)(s + 2 * p));
q3 = _mm_loadu_si128((__m128i *)(s + 3 * p));
q4 = _mm_loadu_si128((__m128i *)(s + 4 * p));
{
const __m128i abs_p1p0 = _mm_or_si128(_mm_subs_epu8(p1, p0),
_mm_subs_epu8(p0, p1));
const __m128i abs_q1q0 = _mm_or_si128(_mm_subs_epu8(q1, q0),
_mm_subs_epu8(q0, q1));
const __m128i one = _mm_set1_epi8(1);
const __m128i fe = _mm_set1_epi8(0xfe);
const __m128i ff = _mm_cmpeq_epi8(abs_p1p0, abs_p1p0);
__m128i abs_p0q0 = _mm_or_si128(_mm_subs_epu8(p0, q0),
_mm_subs_epu8(q0, p0));
__m128i abs_p1q1 = _mm_or_si128(_mm_subs_epu8(p1, q1),
_mm_subs_epu8(q1, p1));
__m128i work;
flat = _mm_max_epu8(abs_p1p0, abs_q1q0);
hev = _mm_subs_epu8(flat, thresh);
hev = _mm_xor_si128(_mm_cmpeq_epi8(hev, zero), ff);
abs_p0q0 =_mm_adds_epu8(abs_p0q0, abs_p0q0);
abs_p1q1 = _mm_srli_epi16(_mm_and_si128(abs_p1q1, fe), 1);
mask = _mm_subs_epu8(_mm_adds_epu8(abs_p0q0, abs_p1q1), blimit);
mask = _mm_xor_si128(_mm_cmpeq_epi8(mask, zero), ff);
// mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 > blimit) * -1;
mask = _mm_max_epu8(flat, mask);
// mask |= (abs(p1 - p0) > limit) * -1;
// mask |= (abs(q1 - q0) > limit) * -1;
work = _mm_max_epu8(_mm_or_si128(_mm_subs_epu8(p2, p1),
_mm_subs_epu8(p1, p2)),
_mm_or_si128(_mm_subs_epu8(p3, p2),
_mm_subs_epu8(p2, p3)));
mask = _mm_max_epu8(work, mask);
work = _mm_max_epu8(_mm_or_si128(_mm_subs_epu8(q2, q1),
_mm_subs_epu8(q1, q2)),
_mm_or_si128(_mm_subs_epu8(q3, q2),
_mm_subs_epu8(q2, q3)));
mask = _mm_max_epu8(work, mask);
mask = _mm_subs_epu8(mask, limit);
mask = _mm_cmpeq_epi8(mask, zero);
work = _mm_max_epu8(_mm_or_si128(_mm_subs_epu8(p2, p0),
_mm_subs_epu8(p0, p2)),
_mm_or_si128(_mm_subs_epu8(q2, q0),
_mm_subs_epu8(q0, q2)));
flat = _mm_max_epu8(work, flat);
work = _mm_max_epu8(_mm_or_si128(_mm_subs_epu8(p3, p0),
_mm_subs_epu8(p0, p3)),
_mm_or_si128(_mm_subs_epu8(q3, q0),
_mm_subs_epu8(q0, q3)));
flat = _mm_max_epu8(work, flat);
work = _mm_max_epu8(_mm_or_si128(_mm_subs_epu8(p4, p0),
_mm_subs_epu8(p0, p4)),
_mm_or_si128(_mm_subs_epu8(q4, q0),
_mm_subs_epu8(q0, q4)));
flat = _mm_max_epu8(work, flat);
flat = _mm_subs_epu8(flat, one);
flat = _mm_cmpeq_epi8(flat, zero);
flat = _mm_and_si128(flat, mask);
}
// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
// calculate flat2
p4 = _mm_loadu_si128((__m128i *)(s - 8 * p));
p3 = _mm_loadu_si128((__m128i *)(s - 7 * p));
p2 = _mm_loadu_si128((__m128i *)(s - 6 * p));
p1 = _mm_loadu_si128((__m128i *)(s - 5 * p));
// p0 = _mm_loadu_si128((__m128i *)(s - 1 * p));
// q0 = _mm_loadu_si128((__m128i *)(s - 0 * p));
q1 = _mm_loadu_si128((__m128i *)(s + 4 * p));
q2 = _mm_loadu_si128((__m128i *)(s + 5 * p));
q3 = _mm_loadu_si128((__m128i *)(s + 6 * p));
q4 = _mm_loadu_si128((__m128i *)(s + 7 * p));
{
const __m128i abs_p1p0 = _mm_or_si128(_mm_subs_epu8(p1, p0),
_mm_subs_epu8(p0, p1));
const __m128i abs_q1q0 = _mm_or_si128(_mm_subs_epu8(q1, q0),
_mm_subs_epu8(q0, q1));
const __m128i one = _mm_set1_epi8(1);
__m128i work;
flat2 = _mm_max_epu8(abs_p1p0, abs_q1q0);
work = _mm_max_epu8(_mm_or_si128(_mm_subs_epu8(p2, p0),
_mm_subs_epu8(p0, p2)),
_mm_or_si128(_mm_subs_epu8(q2, q0),
_mm_subs_epu8(q0, q2)));
flat2 = _mm_max_epu8(work, flat2);
work = _mm_max_epu8(_mm_or_si128(_mm_subs_epu8(p3, p0),
_mm_subs_epu8(p0, p3)),
_mm_or_si128(_mm_subs_epu8(q3, q0),
_mm_subs_epu8(q0, q3)));
flat2 = _mm_max_epu8(work, flat2);
work = _mm_max_epu8(_mm_or_si128(_mm_subs_epu8(p4, p0),
_mm_subs_epu8(p0, p4)),
_mm_or_si128(_mm_subs_epu8(q4, q0),
_mm_subs_epu8(q0, q4)));
flat2 = _mm_max_epu8(work, flat2);
flat2 = _mm_subs_epu8(flat2, one);
flat2 = _mm_cmpeq_epi8(flat2, zero);
flat2 = _mm_and_si128(flat2, flat); // flat2 & flat & mask
}
// calculate flat2
// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
{
const __m128i four = _mm_set1_epi16(4);
unsigned char *src = s;
i = 0;
do {
__m128i workp_a, workp_b, workp_shft;
p4 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 5 * p)), zero);
p3 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 4 * p)), zero);
p2 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 3 * p)), zero);
p1 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 2 * p)), zero);
p0 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 1 * p)), zero);
q0 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 0 * p)), zero);
q1 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 1 * p)), zero);
q2 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 2 * p)), zero);
q3 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 3 * p)), zero);
q4 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 4 * p)), zero);
workp_a = _mm_add_epi16(_mm_add_epi16(p4, p3), _mm_add_epi16(p2, p1));
workp_a = _mm_add_epi16(_mm_add_epi16(workp_a, four), p0);
workp_b = _mm_add_epi16(_mm_add_epi16(q0, p2), p4);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
_mm_storel_epi64((__m128i *)&flat_op2[i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_b = _mm_add_epi16(_mm_add_epi16(q0, q1), p1);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
_mm_storel_epi64((__m128i *)&flat_op1[i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p4), q2);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p1), p0);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
_mm_storel_epi64((__m128i *)&flat_op0[i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p3), q3);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p0), q0);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
_mm_storel_epi64((__m128i *)&flat_oq0[i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p2), q4);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q0), q1);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
_mm_storel_epi64((__m128i *)&flat_oq1[i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p1), q4);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q1), q2);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 3);
_mm_storel_epi64((__m128i *)&flat_oq2[i*8],
_mm_packus_epi16(workp_shft, workp_shft));
src += 8;
} while (++i < 2);
}
// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
// wide flat
// TODO(slavarnway): interleave with the flat pixel calculations (see above)
{
const __m128i eight = _mm_set1_epi16(8);
unsigned char *src = s;
int i = 0;
do {
__m128i workp_a, workp_b, workp_shft;
p7 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 8 * p)), zero);
p6 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 7 * p)), zero);
p5 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 6 * p)), zero);
p4 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 5 * p)), zero);
p3 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 4 * p)), zero);
p2 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 3 * p)), zero);
p1 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 2 * p)), zero);
p0 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 1 * p)), zero);
q0 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src - 0 * p)), zero);
q1 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 1 * p)), zero);
q2 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 2 * p)), zero);
q3 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 3 * p)), zero);
q4 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 4 * p)), zero);
q5 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 5 * p)), zero);
q6 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 6 * p)), zero);
q7 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i *)(src + 7 * p)), zero);
workp_a = _mm_sub_epi16(_mm_slli_epi16(p7, 3), p7); // p7 * 7
workp_a = _mm_add_epi16(_mm_slli_epi16(p6, 1), workp_a);
workp_b = _mm_add_epi16(_mm_add_epi16(p5, p4), _mm_add_epi16(p3, p2));
workp_a = _mm_add_epi16(_mm_add_epi16(p1, p0), workp_a);
workp_b = _mm_add_epi16(_mm_add_epi16(q0, eight), workp_b);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_op[6][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p7), p5);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p6), q1);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_op[5][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p7), p4);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p5), q2);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_op[4][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p7), p3);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p4), q3);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_op[3][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p7), p2);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p3), q4);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_op[2][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p7), p1);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p2), q5);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_op[1][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p7), p0);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p1), q6);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_op[0][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p7), q0);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, p0), q7);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_oq[0][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p6), q1);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q0), q7);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_oq[1][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p5), q2);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q1), q7);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_oq[2][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p4), q3);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q2), q7);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_oq[3][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p3), q4);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q3), q7);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_oq[4][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p2), q5);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q4), q7);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_oq[5][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
workp_a = _mm_add_epi16(_mm_sub_epi16(workp_a, p1), q6);
workp_b = _mm_add_epi16(_mm_sub_epi16(workp_b, q5), q7);
workp_shft = _mm_srli_epi16(_mm_add_epi16(workp_a, workp_b), 4);
_mm_storel_epi64((__m128i *)&flat2_oq[6][i*8],
_mm_packus_epi16(workp_shft, workp_shft));
src += 8;
} while (++i < 2);
}
// wide flat
// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
// lp filter
{
const __m128i t4 = _mm_set1_epi8(4);
const __m128i t3 = _mm_set1_epi8(3);
const __m128i t80 = _mm_set1_epi8(0x80);
const __m128i te0 = _mm_set1_epi8(0xe0);
const __m128i t1f = _mm_set1_epi8(0x1f);
const __m128i t1 = _mm_set1_epi8(0x1);
const __m128i t7f = _mm_set1_epi8(0x7f);
__m128i ps1 = _mm_xor_si128(_mm_loadu_si128((__m128i *)(s - 2 * p)),
t80);
__m128i ps0 = _mm_xor_si128(_mm_loadu_si128((__m128i *)(s - 1 * p)),
t80);
__m128i qs0 = _mm_xor_si128(_mm_loadu_si128((__m128i *)(s + 0 * p)),
t80);
__m128i qs1 = _mm_xor_si128(_mm_loadu_si128((__m128i *)(s + 1 * p)),
t80);
__m128i filt;
__m128i work_a;
__m128i filter1, filter2;
filt = _mm_and_si128(_mm_subs_epi8(ps1, qs1), hev);
work_a = _mm_subs_epi8(qs0, ps0);
filt = _mm_adds_epi8(filt, work_a);
filt = _mm_adds_epi8(filt, work_a);
filt = _mm_adds_epi8(filt, work_a);
/* (vp9_filter + 3 * (qs0 - ps0)) & mask */
filt = _mm_and_si128(filt, mask);
filter1 = _mm_adds_epi8(filt, t4);
filter2 = _mm_adds_epi8(filt, t3);
/* Filter1 >> 3 */
work_a = _mm_cmpgt_epi8(zero, filter1);
filter1 = _mm_srli_epi16(filter1, 3);
work_a = _mm_and_si128(work_a, te0);
filter1 = _mm_and_si128(filter1, t1f);
filter1 = _mm_or_si128(filter1, work_a);
/* Filter2 >> 3 */
work_a = _mm_cmpgt_epi8(zero, filter2);
filter2 = _mm_srli_epi16(filter2, 3);
work_a = _mm_and_si128(work_a, te0);
filter2 = _mm_and_si128(filter2, t1f);
filter2 = _mm_or_si128(filter2, work_a);
/* filt >> 1 */
filt = _mm_adds_epi8(filter1, t1);
work_a = _mm_cmpgt_epi8(zero, filt);
filt = _mm_srli_epi16(filt, 1);
work_a = _mm_and_si128(work_a, t80);
filt = _mm_and_si128(filt, t7f);
filt = _mm_or_si128(filt, work_a);
filt = _mm_andnot_si128(hev, filt);
ps0 = _mm_xor_si128(_mm_adds_epi8(ps0, filter2), t80);
ps1 = _mm_xor_si128(_mm_adds_epi8(ps1, filt), t80);
qs0 = _mm_xor_si128(_mm_subs_epi8(qs0, filter1), t80);
qs1 = _mm_xor_si128(_mm_subs_epi8(qs1, filt), t80);
// write out op6 - op3
{
unsigned char *dst = (s - 7 * p);
for (i = 6; i > 2; i--) {
__m128i flat2_output;
work_a = _mm_loadu_si128((__m128i *)dst);
flat2_output = _mm_load_si128((__m128i *)flat2_op[i]);
work_a = _mm_andnot_si128(flat2, work_a);
flat2_output = _mm_and_si128(flat2, flat2_output);
work_a = _mm_or_si128(work_a, flat2_output);
_mm_storeu_si128((__m128i *)dst, work_a);
dst += p;
}
}
work_a = _mm_loadu_si128((__m128i *)(s - 3 * p));
p2 = _mm_load_si128((__m128i *)flat_op2);
work_a = _mm_andnot_si128(flat, work_a);
p2 = _mm_and_si128(flat, p2);
work_a = _mm_or_si128(work_a, p2);
p2 = _mm_load_si128((__m128i *)flat2_op[2]);
work_a = _mm_andnot_si128(flat2, work_a);
p2 = _mm_and_si128(flat2, p2);
p2 = _mm_or_si128(work_a, p2);
_mm_storeu_si128((__m128i *)(s - 3 * p), p2);
p1 = _mm_load_si128((__m128i *)flat_op1);
work_a = _mm_andnot_si128(flat, ps1);
p1 = _mm_and_si128(flat, p1);
work_a = _mm_or_si128(work_a, p1);
p1 = _mm_load_si128((__m128i *)flat2_op[1]);
work_a = _mm_andnot_si128(flat2, work_a);
p1 = _mm_and_si128(flat2, p1);
p1 = _mm_or_si128(work_a, p1);
_mm_storeu_si128((__m128i *)(s - 2 * p), p1);
p0 = _mm_load_si128((__m128i *)flat_op0);
work_a = _mm_andnot_si128(flat, ps0);
p0 = _mm_and_si128(flat, p0);
work_a = _mm_or_si128(work_a, p0);
p0 = _mm_load_si128((__m128i *)flat2_op[0]);
work_a = _mm_andnot_si128(flat2, work_a);
p0 = _mm_and_si128(flat2, p0);
p0 = _mm_or_si128(work_a, p0);
_mm_storeu_si128((__m128i *)(s - 1 * p), p0);
q0 = _mm_load_si128((__m128i *)flat_oq0);
work_a = _mm_andnot_si128(flat, qs0);
q0 = _mm_and_si128(flat, q0);
work_a = _mm_or_si128(work_a, q0);
q0 = _mm_load_si128((__m128i *)flat2_oq[0]);
work_a = _mm_andnot_si128(flat2, work_a);
q0 = _mm_and_si128(flat2, q0);
q0 = _mm_or_si128(work_a, q0);
_mm_storeu_si128((__m128i *)(s - 0 * p), q0);
q1 = _mm_load_si128((__m128i *)flat_oq1);
work_a = _mm_andnot_si128(flat, qs1);
q1 = _mm_and_si128(flat, q1);
work_a = _mm_or_si128(work_a, q1);
q1 = _mm_load_si128((__m128i *)flat2_oq[1]);
work_a = _mm_andnot_si128(flat2, work_a);
q1 = _mm_and_si128(flat2, q1);
q1 = _mm_or_si128(work_a, q1);
_mm_storeu_si128((__m128i *)(s + 1 * p), q1);
work_a = _mm_loadu_si128((__m128i *)(s + 2 * p));
q2 = _mm_load_si128((__m128i *)flat_oq2);
work_a = _mm_andnot_si128(flat, work_a);
q2 = _mm_and_si128(flat, q2);
work_a = _mm_or_si128(work_a, q2);
q2 = _mm_load_si128((__m128i *)flat2_oq[2]);
work_a = _mm_andnot_si128(flat2, work_a);
q2 = _mm_and_si128(flat2, q2);
q2 = _mm_or_si128(work_a, q2);
_mm_storeu_si128((__m128i *)(s + 2 * p), q2);
// write out oq3 - oq7
{
unsigned char *dst = (s + 3 * p);
for (i = 3; i < 7; i++) {
__m128i flat2_output;
work_a = _mm_loadu_si128((__m128i *)dst);
flat2_output = _mm_load_si128((__m128i *)flat2_oq[i]);
work_a = _mm_andnot_si128(flat2, work_a);
flat2_output = _mm_and_si128(flat2, flat2_output);
work_a = _mm_or_si128(work_a, flat2_output);
_mm_storeu_si128((__m128i *)dst, work_a);
dst += p;
}
}
}
}
void vp9_mbloop_filter_horizontal_edge_sse2(unsigned char *s,
int p,
const unsigned char *_blimit,
@ -562,6 +1036,38 @@ void vp9_mbloop_filter_vertical_edge_sse2(unsigned char *s,
transpose(src, 16, dst, p, 2);
}
void vp9_mb_lpf_vertical_edge_w_sse2(unsigned char *s,
int p,
const unsigned char *blimit,
const unsigned char *limit,
const unsigned char *thresh) {
DECLARE_ALIGNED_ARRAY(16, unsigned char, t_dst, 256);
unsigned char *src[4];
unsigned char *dst[4];
/* Transpose 16x16 */
transpose8x16(s - 8, s - 8 + p * 8, p, t_dst, 16);
transpose8x16(s, s + p * 8, p, t_dst + 16 * 8, 16);
/* Loop filtering */
vp9_mb_lpf_horizontal_edge_w_sse2(t_dst + 8 * 16, 16, blimit, limit,
thresh);
src[0] = t_dst;
src[1] = t_dst + 8 * 16;
src[2] = t_dst + 8;
src[3] = t_dst + 8 * 16 + 8;
dst[0] = s - 8;
dst[1] = s - 8 + 8;
dst[2] = s - 8 + p * 8;
dst[3] = s - 8 + p * 8 + 8;
/* Transpose 16x16 */
transpose(src, 16, dst, p, 4);
}
void vp9_mbloop_filter_vertical_edge_uv_sse2(unsigned char *u,
int p,
const unsigned char *blimit,
@ -604,11 +1110,30 @@ void vp9_loop_filter_mbh_sse2(unsigned char *y_ptr,
lfi->lim, lfi->hev_thr, v_ptr);
}
void vp9_lpf_mbh_w_sse2(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mb_lpf_horizontal_edge_w_sse2(y_ptr, y_stride,
lfi->mblim, lfi->lim, lfi->hev_thr);
/* u,v */
if (u_ptr)
vp9_mbloop_filter_horizontal_edge_uv_sse2(u_ptr, uv_stride, lfi->mblim,
lfi->lim, lfi->hev_thr, v_ptr);
}
void vp9_loop_filter_bh8x8_sse2(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_horizontal_edge_sse2(
y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr);
if (u_ptr)
vp9_loop_filter_horizontal_edge_uv_sse2(u_ptr + 4 * uv_stride, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr,
v_ptr + 4 * uv_stride);
}
/* Vertical MB Filtering */
@ -624,11 +1149,30 @@ void vp9_loop_filter_mbv_sse2(unsigned char *y_ptr, unsigned char *u_ptr,
lfi->lim, lfi->hev_thr, v_ptr);
}
void vp9_lpf_mbv_w_sse2(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mb_lpf_vertical_edge_w_sse2(y_ptr, y_stride,
lfi->mblim, lfi->lim, lfi->hev_thr);
/* u,v */
if (u_ptr)
vp9_mbloop_filter_vertical_edge_uv_sse2(u_ptr, uv_stride, lfi->mblim,
lfi->lim, lfi->hev_thr, v_ptr);
}
void vp9_loop_filter_bv8x8_sse2(unsigned char *y_ptr, unsigned char *u_ptr,
unsigned char *v_ptr, int y_stride, int uv_stride,
struct loop_filter_info *lfi) {
vp9_mbloop_filter_vertical_edge_sse2(
y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr);
if (u_ptr)
vp9_loop_filter_vertical_edge_uv_sse2(u_ptr + 4, uv_stride,
lfi->blim, lfi->lim, lfi->hev_thr,
v_ptr + 4);
}
/* Horizontal B Filtering */

645
vp9/common/x86/vp9_subpel_variance_impl_sse2.asm Normal file
View File

@ -0,0 +1,645 @@
;
; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
;
; Use of this source code is governed by a BSD-style license
; that can be found in the LICENSE file in the root of the source
; tree. An additional intellectual property rights grant can be found
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
%include "vpx_ports/x86_abi_support.asm"
%define xmm_filter_shift 7
;void vp9_filter_block2d_bil_var_sse2
;(
; unsigned char *ref_ptr,
; int ref_pixels_per_line,
; unsigned char *src_ptr,
; int src_pixels_per_line,
; unsigned int Height,
; int xoffset,
; int yoffset,
; int *sum,
; unsigned int *sumsquared;;
;
;)
global sym(vp9_filter_block2d_bil_var_sse2) PRIVATE
sym(vp9_filter_block2d_bil_var_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 9
SAVE_XMM 7
GET_GOT rbx
push rsi
push rdi
push rbx
; end prolog
pxor xmm6, xmm6 ;
pxor xmm7, xmm7 ;
lea rsi, [GLOBAL(xmm_bi_rd)] ; rounding
movdqa xmm4, XMMWORD PTR [rsi]
lea rcx, [GLOBAL(bilinear_filters_sse2)]
movsxd rax, dword ptr arg(5) ; xoffset
cmp rax, 0 ; skip first_pass filter if xoffset=0
je filter_block2d_bil_var_sse2_sp_only
shl rax, 5 ; point to filter coeff with xoffset
lea rax, [rax + rcx] ; HFilter
movsxd rdx, dword ptr arg(6) ; yoffset
cmp rdx, 0 ; skip second_pass filter if yoffset=0
je filter_block2d_bil_var_sse2_fp_only
shl rdx, 5
lea rdx, [rdx + rcx] ; VFilter
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
pxor xmm0, xmm0 ;
movq xmm1, QWORD PTR [rsi] ;
movq xmm3, QWORD PTR [rsi+1] ;
punpcklbw xmm1, xmm0 ;
pmullw xmm1, [rax] ;
punpcklbw xmm3, xmm0
pmullw xmm3, [rax+16] ;
paddw xmm1, xmm3 ;
paddw xmm1, xmm4 ;
psraw xmm1, xmm_filter_shift ;
movdqa xmm5, xmm1
movsxd rbx, dword ptr arg(1) ;ref_pixels_per_line
lea rsi, [rsi + rbx]
%if ABI_IS_32BIT=0
movsxd r9, dword ptr arg(3) ;src_pixels_per_line
%endif
filter_block2d_bil_var_sse2_loop:
movq xmm1, QWORD PTR [rsi] ;
movq xmm3, QWORD PTR [rsi+1] ;
punpcklbw xmm1, xmm0 ;
pmullw xmm1, [rax] ;
punpcklbw xmm3, xmm0 ;
pmullw xmm3, [rax+16] ;
paddw xmm1, xmm3 ;
paddw xmm1, xmm4 ;
psraw xmm1, xmm_filter_shift ;
movdqa xmm3, xmm5 ;
movdqa xmm5, xmm1 ;
pmullw xmm3, [rdx] ;
pmullw xmm1, [rdx+16] ;
paddw xmm1, xmm3 ;
paddw xmm1, xmm4 ;
psraw xmm1, xmm_filter_shift ;
movq xmm3, QWORD PTR [rdi] ;
punpcklbw xmm3, xmm0 ;
psubw xmm1, xmm3 ;
paddw xmm6, xmm1 ;
pmaddwd xmm1, xmm1 ;
paddd xmm7, xmm1 ;
lea rsi, [rsi + rbx] ;ref_pixels_per_line
%if ABI_IS_32BIT
add rdi, dword ptr arg(3) ;src_pixels_per_line
%else
lea rdi, [rdi + r9]
%endif
sub rcx, 1 ;
jnz filter_block2d_bil_var_sse2_loop ;
jmp filter_block2d_bil_variance
filter_block2d_bil_var_sse2_sp_only:
movsxd rdx, dword ptr arg(6) ; yoffset
cmp rdx, 0 ; skip all if both xoffset=0 and yoffset=0
je filter_block2d_bil_var_sse2_full_pixel
shl rdx, 5
lea rdx, [rdx + rcx] ; VFilter
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
pxor xmm0, xmm0 ;
movq xmm1, QWORD PTR [rsi] ;
punpcklbw xmm1, xmm0 ;
movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
lea rsi, [rsi + rax]
filter_block2d_bil_sp_only_loop:
movq xmm3, QWORD PTR [rsi] ;
punpcklbw xmm3, xmm0 ;
movdqa xmm5, xmm3
pmullw xmm1, [rdx] ;
pmullw xmm3, [rdx+16] ;
paddw xmm1, xmm3 ;
paddw xmm1, xmm4 ;
psraw xmm1, xmm_filter_shift ;
movq xmm3, QWORD PTR [rdi] ;
punpcklbw xmm3, xmm0 ;
psubw xmm1, xmm3 ;
paddw xmm6, xmm1 ;
pmaddwd xmm1, xmm1 ;
paddd xmm7, xmm1 ;
movdqa xmm1, xmm5 ;
lea rsi, [rsi + rax] ;ref_pixels_per_line
lea rdi, [rdi + rbx] ;src_pixels_per_line
sub rcx, 1 ;
jnz filter_block2d_bil_sp_only_loop ;
jmp filter_block2d_bil_variance
filter_block2d_bil_var_sse2_full_pixel:
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
pxor xmm0, xmm0 ;
filter_block2d_bil_full_pixel_loop:
movq xmm1, QWORD PTR [rsi] ;
punpcklbw xmm1, xmm0 ;
movq xmm2, QWORD PTR [rdi] ;
punpcklbw xmm2, xmm0 ;
psubw xmm1, xmm2 ;
paddw xmm6, xmm1 ;
pmaddwd xmm1, xmm1 ;
paddd xmm7, xmm1 ;
lea rsi, [rsi + rax] ;ref_pixels_per_line
lea rdi, [rdi + rbx] ;src_pixels_per_line
sub rcx, 1 ;
jnz filter_block2d_bil_full_pixel_loop ;
jmp filter_block2d_bil_variance
filter_block2d_bil_var_sse2_fp_only:
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
movsxd rdx, dword ptr arg(1) ;ref_pixels_per_line
pxor xmm0, xmm0 ;
movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
filter_block2d_bil_fp_only_loop:
movq xmm1, QWORD PTR [rsi] ;
movq xmm3, QWORD PTR [rsi+1] ;
punpcklbw xmm1, xmm0 ;
pmullw xmm1, [rax] ;
punpcklbw xmm3, xmm0 ;
pmullw xmm3, [rax+16] ;
paddw xmm1, xmm3 ;
paddw xmm1, xmm4 ;
psraw xmm1, xmm_filter_shift ;
movq xmm3, QWORD PTR [rdi] ;
punpcklbw xmm3, xmm0 ;
psubw xmm1, xmm3 ;
paddw xmm6, xmm1 ;
pmaddwd xmm1, xmm1 ;
paddd xmm7, xmm1 ;
lea rsi, [rsi + rdx]
lea rdi, [rdi + rbx] ;src_pixels_per_line
sub rcx, 1 ;
jnz filter_block2d_bil_fp_only_loop ;
jmp filter_block2d_bil_variance
filter_block2d_bil_variance:
movdq2q mm6, xmm6 ;
movdq2q mm7, xmm7 ;
psrldq xmm6, 8
psrldq xmm7, 8
movdq2q mm2, xmm6
movdq2q mm3, xmm7
paddw mm6, mm2
paddd mm7, mm3
pxor mm3, mm3 ;
pxor mm2, mm2 ;
punpcklwd mm2, mm6 ;
punpckhwd mm3, mm6 ;
paddd mm2, mm3 ;
movq mm6, mm2 ;
psrlq mm6, 32 ;
paddd mm2, mm6 ;
psrad mm2, 16 ;
movq mm4, mm7 ;
psrlq mm4, 32 ;
paddd mm4, mm7 ;
mov rsi, arg(7) ; sum
mov rdi, arg(8) ; sumsquared
movd [rsi], mm2 ; xsum
movd [rdi], mm4 ; xxsum
; begin epilog
pop rbx
pop rdi
pop rsi
RESTORE_GOT
RESTORE_XMM
UNSHADOW_ARGS
pop rbp
ret
;void vp9_half_horiz_vert_variance16x_h_sse2
;(
; unsigned char *ref_ptr,
; int ref_pixels_per_line,
; unsigned char *src_ptr,
; int src_pixels_per_line,
; unsigned int Height,
; int *sum,
; unsigned int *sumsquared
;)
global sym(vp9_half_horiz_vert_variance16x_h_sse2) PRIVATE
sym(vp9_half_horiz_vert_variance16x_h_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
SAVE_XMM 7
GET_GOT rbx
push rsi
push rdi
; end prolog
pxor xmm6, xmm6 ; error accumulator
pxor xmm7, xmm7 ; sse eaccumulator
mov rsi, arg(0) ;ref_ptr ;
mov rdi, arg(2) ;src_ptr ;
movsxd rcx, dword ptr arg(4) ;Height ;
movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
pxor xmm0, xmm0 ;
movdqu xmm5, XMMWORD PTR [rsi]
movdqu xmm3, XMMWORD PTR [rsi+1]
pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3) horizontal line 1
lea rsi, [rsi + rax]
.half_horiz_vert_variance16x_h_1:
movdqu xmm1, XMMWORD PTR [rsi] ;
movdqu xmm2, XMMWORD PTR [rsi+1] ;
pavgb xmm1, xmm2 ; xmm1 = avg(xmm1,xmm3) horizontal line i+1
pavgb xmm5, xmm1 ; xmm = vertical average of the above
movdqa xmm4, xmm5
punpcklbw xmm5, xmm0 ; xmm5 = words of above
punpckhbw xmm4, xmm0
movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d7
punpcklbw xmm3, xmm0 ; xmm3 = words of above
psubw xmm5, xmm3 ; xmm5 -= xmm3
movq xmm3, QWORD PTR [rdi+8]
punpcklbw xmm3, xmm0
psubw xmm4, xmm3
paddw xmm6, xmm5 ; xmm6 += accumulated column differences
paddw xmm6, xmm4
pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
pmaddwd xmm4, xmm4
paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
paddd xmm7, xmm4
movdqa xmm5, xmm1 ; save xmm1 for use on the next row
lea rsi, [rsi + rax]
lea rdi, [rdi + rdx]
sub rcx, 1 ;
jnz .half_horiz_vert_variance16x_h_1 ;
pxor xmm1, xmm1
pxor xmm5, xmm5
punpcklwd xmm0, xmm6
punpckhwd xmm1, xmm6
psrad xmm0, 16
psrad xmm1, 16
paddd xmm0, xmm1
movdqa xmm1, xmm0
movdqa xmm6, xmm7
punpckldq xmm6, xmm5
punpckhdq xmm7, xmm5
paddd xmm6, xmm7
punpckldq xmm0, xmm5
punpckhdq xmm1, xmm5
paddd xmm0, xmm1
movdqa xmm7, xmm6
movdqa xmm1, xmm0
psrldq xmm7, 8
psrldq xmm1, 8
paddd xmm6, xmm7
paddd xmm0, xmm1
mov rsi, arg(5) ;[Sum]
mov rdi, arg(6) ;[SSE]
movd [rsi], xmm0
movd [rdi], xmm6
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
RESTORE_XMM
UNSHADOW_ARGS
pop rbp
ret
;void vp9_half_vert_variance16x_h_sse2
;(
; unsigned char *ref_ptr,
; int ref_pixels_per_line,
; unsigned char *src_ptr,
; int src_pixels_per_line,
; unsigned int Height,
; int *sum,
; unsigned int *sumsquared
;)
global sym(vp9_half_vert_variance16x_h_sse2) PRIVATE
sym(vp9_half_vert_variance16x_h_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
SAVE_XMM 7
GET_GOT rbx
push rsi
push rdi
; end prolog
pxor xmm6, xmm6 ; error accumulator
pxor xmm7, xmm7 ; sse eaccumulator
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
movdqu xmm5, XMMWORD PTR [rsi]
lea rsi, [rsi + rax ]
pxor xmm0, xmm0
.half_vert_variance16x_h_1:
movdqu xmm3, XMMWORD PTR [rsi]
pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
movdqa xmm4, xmm5
punpcklbw xmm5, xmm0
punpckhbw xmm4, xmm0
movq xmm2, QWORD PTR [rdi]
punpcklbw xmm2, xmm0
psubw xmm5, xmm2
movq xmm2, QWORD PTR [rdi+8]
punpcklbw xmm2, xmm0
psubw xmm4, xmm2
paddw xmm6, xmm5 ; xmm6 += accumulated column differences
paddw xmm6, xmm4
pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
pmaddwd xmm4, xmm4
paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
paddd xmm7, xmm4
movdqa xmm5, xmm3
lea rsi, [rsi + rax]
lea rdi, [rdi + rdx]
sub rcx, 1
jnz .half_vert_variance16x_h_1
pxor xmm1, xmm1
pxor xmm5, xmm5
punpcklwd xmm0, xmm6
punpckhwd xmm1, xmm6
psrad xmm0, 16
psrad xmm1, 16
paddd xmm0, xmm1
movdqa xmm1, xmm0
movdqa xmm6, xmm7
punpckldq xmm6, xmm5
punpckhdq xmm7, xmm5
paddd xmm6, xmm7
punpckldq xmm0, xmm5
punpckhdq xmm1, xmm5
paddd xmm0, xmm1
movdqa xmm7, xmm6
movdqa xmm1, xmm0
psrldq xmm7, 8
psrldq xmm1, 8
paddd xmm6, xmm7
paddd xmm0, xmm1
mov rsi, arg(5) ;[Sum]
mov rdi, arg(6) ;[SSE]
movd [rsi], xmm0
movd [rdi], xmm6
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
RESTORE_XMM
UNSHADOW_ARGS
pop rbp
ret
;void vp9_half_horiz_variance16x_h_sse2
;(
; unsigned char *ref_ptr,
; int ref_pixels_per_line,
; unsigned char *src_ptr,
; int src_pixels_per_line,
; unsigned int Height,
; int *sum,
; unsigned int *sumsquared
;)
global sym(vp9_half_horiz_variance16x_h_sse2) PRIVATE
sym(vp9_half_horiz_variance16x_h_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
SAVE_XMM 7
GET_GOT rbx
push rsi
push rdi
; end prolog
pxor xmm6, xmm6 ; error accumulator
pxor xmm7, xmm7 ; sse eaccumulator
mov rsi, arg(0) ;ref_ptr ;
mov rdi, arg(2) ;src_ptr ;
movsxd rcx, dword ptr arg(4) ;Height ;
movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
pxor xmm0, xmm0 ;
.half_horiz_variance16x_h_1:
movdqu xmm5, XMMWORD PTR [rsi] ; xmm5 = s0,s1,s2..s15
movdqu xmm3, XMMWORD PTR [rsi+1] ; xmm3 = s1,s2,s3..s16
pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
movdqa xmm1, xmm5
punpcklbw xmm5, xmm0 ; xmm5 = words of above
punpckhbw xmm1, xmm0
movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d7
punpcklbw xmm3, xmm0 ; xmm3 = words of above
movq xmm2, QWORD PTR [rdi+8]
punpcklbw xmm2, xmm0
psubw xmm5, xmm3 ; xmm5 -= xmm3
psubw xmm1, xmm2
paddw xmm6, xmm5 ; xmm6 += accumulated column differences
paddw xmm6, xmm1
pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
pmaddwd xmm1, xmm1
paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
paddd xmm7, xmm1
lea rsi, [rsi + rax]
lea rdi, [rdi + rdx]
sub rcx, 1 ;
jnz .half_horiz_variance16x_h_1 ;
pxor xmm1, xmm1
pxor xmm5, xmm5
punpcklwd xmm0, xmm6
punpckhwd xmm1, xmm6
psrad xmm0, 16
psrad xmm1, 16
paddd xmm0, xmm1
movdqa xmm1, xmm0
movdqa xmm6, xmm7
punpckldq xmm6, xmm5
punpckhdq xmm7, xmm5
paddd xmm6, xmm7
punpckldq xmm0, xmm5
punpckhdq xmm1, xmm5
paddd xmm0, xmm1
movdqa xmm7, xmm6
movdqa xmm1, xmm0
psrldq xmm7, 8
psrldq xmm1, 8
paddd xmm6, xmm7
paddd xmm0, xmm1
mov rsi, arg(5) ;[Sum]
mov rdi, arg(6) ;[SSE]
movd [rsi], xmm0
movd [rdi], xmm6
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
RESTORE_XMM
UNSHADOW_ARGS
pop rbp
ret
SECTION_RODATA
; short xmm_bi_rd[8] = { 64, 64, 64, 64,64, 64, 64, 64};
align 16
xmm_bi_rd:
times 8 dw 64
align 16
bilinear_filters_sse2:
dw 128, 128, 128, 128, 128, 128, 128, 128, 0, 0, 0, 0, 0, 0, 0, 0
dw 120, 120, 120, 120, 120, 120, 120, 120, 8, 8, 8, 8, 8, 8, 8, 8
dw 112, 112, 112, 112, 112, 112, 112, 112, 16, 16, 16, 16, 16, 16, 16, 16
dw 104, 104, 104, 104, 104, 104, 104, 104, 24, 24, 24, 24, 24, 24, 24, 24
dw 96, 96, 96, 96, 96, 96, 96, 96, 32, 32, 32, 32, 32, 32, 32, 32
dw 88, 88, 88, 88, 88, 88, 88, 88, 40, 40, 40, 40, 40, 40, 40, 40
dw 80, 80, 80, 80, 80, 80, 80, 80, 48, 48, 48, 48, 48, 48, 48, 48
dw 72, 72, 72, 72, 72, 72, 72, 72, 56, 56, 56, 56, 56, 56, 56, 56
dw 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
dw 56, 56, 56, 56, 56, 56, 56, 56, 72, 72, 72, 72, 72, 72, 72, 72
dw 48, 48, 48, 48, 48, 48, 48, 48, 80, 80, 80, 80, 80, 80, 80, 80
dw 40, 40, 40, 40, 40, 40, 40, 40, 88, 88, 88, 88, 88, 88, 88, 88
dw 32, 32, 32, 32, 32, 32, 32, 32, 96, 96, 96, 96, 96, 96, 96, 96
dw 24, 24, 24, 24, 24, 24, 24, 24, 104, 104, 104, 104, 104, 104, 104, 104
dw 16, 16, 16, 16, 16, 16, 16, 16, 112, 112, 112, 112, 112, 112, 112, 112
dw 8, 8, 8, 8, 8, 8, 8, 8, 120, 120, 120, 120, 120, 120, 120, 120

459
vp9/common/x86/vp9_subpixel_mmx.asm
View File

@ -202,438 +202,6 @@ sym(vp9_filter_block1dc_v6_mmx):
pop rbp
ret
;void bilinear_predict8x8_mmx
;(
; unsigned char *src_ptr,
; int src_pixels_per_line,
; int xoffset,
; int yoffset,
; unsigned char *dst_ptr,
; int dst_pitch
;)
global sym(vp9_bilinear_predict8x8_mmx) PRIVATE
sym(vp9_bilinear_predict8x8_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
GET_GOT rbx
push rsi
push rdi
; end prolog
;const short *HFilter = bilinear_filters_mmx[xoffset];
;const short *VFilter = bilinear_filters_mmx[yoffset];
movsxd rax, dword ptr arg(2) ;xoffset
mov rdi, arg(4) ;dst_ptr ;
shl rax, 5 ; offset * 32
lea rcx, [GLOBAL(sym(vp9_bilinear_filters_8x_mmx))]
add rax, rcx ; HFilter
mov rsi, arg(0) ;src_ptr ;
movsxd rdx, dword ptr arg(5) ;dst_pitch
movq mm1, [rax] ;
movq mm2, [rax+16] ;
movsxd rax, dword ptr arg(3) ;yoffset
pxor mm0, mm0 ;
shl rax, 5 ; offset*32
add rax, rcx ; VFilter
lea rcx, [rdi+rdx*8] ;
movsxd rdx, dword ptr arg(1) ;src_pixels_per_line ;
; get the first horizontal line done ;
movq mm3, [rsi] ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
movq mm4, mm3 ; make a copy of current line
punpcklbw mm3, mm0 ; xx 00 01 02 03 04 05 06
punpckhbw mm4, mm0 ;
pmullw mm3, mm1 ;
pmullw mm4, mm1 ;
movq mm5, [rsi+1] ;
movq mm6, mm5 ;
punpcklbw mm5, mm0 ;
punpckhbw mm6, mm0 ;
pmullw mm5, mm2 ;
pmullw mm6, mm2 ;
paddw mm3, mm5 ;
paddw mm4, mm6 ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP9_FILTER_SHIFT ;
movq mm7, mm3 ;
packuswb mm7, mm4 ;
add rsi, rdx ; next line
.next_row_8x8:
movq mm3, [rsi] ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
movq mm4, mm3 ; make a copy of current line
punpcklbw mm3, mm0 ; xx 00 01 02 03 04 05 06
punpckhbw mm4, mm0 ;
pmullw mm3, mm1 ;
pmullw mm4, mm1 ;
movq mm5, [rsi+1] ;
movq mm6, mm5 ;
punpcklbw mm5, mm0 ;
punpckhbw mm6, mm0 ;
pmullw mm5, mm2 ;
pmullw mm6, mm2 ;
paddw mm3, mm5 ;
paddw mm4, mm6 ;
movq mm5, mm7 ;
movq mm6, mm7 ;
punpcklbw mm5, mm0 ;
punpckhbw mm6, mm0
pmullw mm5, [rax] ;
pmullw mm6, [rax] ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP9_FILTER_SHIFT ;
movq mm7, mm3 ;
packuswb mm7, mm4 ;
pmullw mm3, [rax+16] ;
pmullw mm4, [rax+16] ;
paddw mm3, mm5 ;
paddw mm4, mm6 ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP9_FILTER_SHIFT ;
packuswb mm3, mm4
movq [rdi], mm3 ; store the results in the destination
%if ABI_IS_32BIT
add rsi, rdx ; next line
add rdi, dword ptr arg(5) ;dst_pitch ;
%else
movsxd r8, dword ptr arg(5) ;dst_pitch
add rsi, rdx ; next line
add rdi, r8 ;dst_pitch
%endif
cmp rdi, rcx ;
jne .next_row_8x8
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
;void bilinear_predict8x4_mmx
;(
; unsigned char *src_ptr,
; int src_pixels_per_line,
; int xoffset,
; int yoffset,
; unsigned char *dst_ptr,
; int dst_pitch
;)
global sym(vp9_bilinear_predict8x4_mmx) PRIVATE
sym(vp9_bilinear_predict8x4_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
GET_GOT rbx
push rsi
push rdi
; end prolog
;const short *HFilter = bilinear_filters_mmx[xoffset];
;const short *VFilter = bilinear_filters_mmx[yoffset];
movsxd rax, dword ptr arg(2) ;xoffset
mov rdi, arg(4) ;dst_ptr ;
lea rcx, [GLOBAL(sym(vp9_bilinear_filters_8x_mmx))]
shl rax, 5
mov rsi, arg(0) ;src_ptr ;
add rax, rcx
movsxd rdx, dword ptr arg(5) ;dst_pitch
movq mm1, [rax] ;
movq mm2, [rax+16] ;
movsxd rax, dword ptr arg(3) ;yoffset
pxor mm0, mm0 ;
shl rax, 5
add rax, rcx
lea rcx, [rdi+rdx*4] ;
movsxd rdx, dword ptr arg(1) ;src_pixels_per_line ;
; get the first horizontal line done ;
movq mm3, [rsi] ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
movq mm4, mm3 ; make a copy of current line
punpcklbw mm3, mm0 ; xx 00 01 02 03 04 05 06
punpckhbw mm4, mm0 ;
pmullw mm3, mm1 ;
pmullw mm4, mm1 ;
movq mm5, [rsi+1] ;
movq mm6, mm5 ;
punpcklbw mm5, mm0 ;
punpckhbw mm6, mm0 ;
pmullw mm5, mm2 ;
pmullw mm6, mm2 ;
paddw mm3, mm5 ;
paddw mm4, mm6 ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP9_FILTER_SHIFT ;
movq mm7, mm3 ;
packuswb mm7, mm4 ;
add rsi, rdx ; next line
.next_row_8x4:
movq mm3, [rsi] ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
movq mm4, mm3 ; make a copy of current line
punpcklbw mm3, mm0 ; xx 00 01 02 03 04 05 06
punpckhbw mm4, mm0 ;
pmullw mm3, mm1 ;
pmullw mm4, mm1 ;
movq mm5, [rsi+1] ;
movq mm6, mm5 ;
punpcklbw mm5, mm0 ;
punpckhbw mm6, mm0 ;
pmullw mm5, mm2 ;
pmullw mm6, mm2 ;
paddw mm3, mm5 ;
paddw mm4, mm6 ;
movq mm5, mm7 ;
movq mm6, mm7 ;
punpcklbw mm5, mm0 ;
punpckhbw mm6, mm0
pmullw mm5, [rax] ;
pmullw mm6, [rax] ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP9_FILTER_SHIFT ;
movq mm7, mm3 ;
packuswb mm7, mm4 ;
pmullw mm3, [rax+16] ;
pmullw mm4, [rax+16] ;
paddw mm3, mm5 ;
paddw mm4, mm6 ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP9_FILTER_SHIFT ;
packuswb mm3, mm4
movq [rdi], mm3 ; store the results in the destination
%if ABI_IS_32BIT
add rsi, rdx ; next line
add rdi, dword ptr arg(5) ;dst_pitch ;
%else
movsxd r8, dword ptr arg(5) ;dst_pitch
add rsi, rdx ; next line
add rdi, r8
%endif
cmp rdi, rcx ;
jne .next_row_8x4
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
;void bilinear_predict4x4_mmx
;(
; unsigned char *src_ptr,
; int src_pixels_per_line,
; int xoffset,
; int yoffset,
; unsigned char *dst_ptr,
; int dst_pitch
;)
global sym(vp9_bilinear_predict4x4_mmx) PRIVATE
sym(vp9_bilinear_predict4x4_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
GET_GOT rbx
push rsi
push rdi
; end prolog
;const short *HFilter = bilinear_filters_mmx[xoffset];
;const short *VFilter = bilinear_filters_mmx[yoffset];
movsxd rax, dword ptr arg(2) ;xoffset
mov rdi, arg(4) ;dst_ptr ;
lea rcx, [GLOBAL(sym(vp9_bilinear_filters_8x_mmx))]
shl rax, 5
add rax, rcx ; HFilter
mov rsi, arg(0) ;src_ptr ;
movsxd rdx, dword ptr arg(5) ;ldst_pitch
movq mm1, [rax] ;
movq mm2, [rax+16] ;
movsxd rax, dword ptr arg(3) ;yoffset
pxor mm0, mm0 ;
shl rax, 5
add rax, rcx
lea rcx, [rdi+rdx*4] ;
movsxd rdx, dword ptr arg(1) ;src_pixels_per_line ;
; get the first horizontal line done ;
movd mm3, [rsi] ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
punpcklbw mm3, mm0 ; xx 00 01 02 03 04 05 06
pmullw mm3, mm1 ;
movd mm5, [rsi+1] ;
punpcklbw mm5, mm0 ;
pmullw mm5, mm2 ;
paddw mm3, mm5 ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
movq mm7, mm3 ;
packuswb mm7, mm0 ;
add rsi, rdx ; next line
.next_row_4x4:
movd mm3, [rsi] ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
punpcklbw mm3, mm0 ; xx 00 01 02 03 04 05 06
pmullw mm3, mm1 ;
movd mm5, [rsi+1] ;
punpcklbw mm5, mm0 ;
pmullw mm5, mm2 ;
paddw mm3, mm5 ;
movq mm5, mm7 ;
punpcklbw mm5, mm0 ;
pmullw mm5, [rax] ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
movq mm7, mm3 ;
packuswb mm7, mm0 ;
pmullw mm3, [rax+16] ;
paddw mm3, mm5 ;
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; xmm3 /= 128
packuswb mm3, mm0
movd [rdi], mm3 ; store the results in the destination
%if ABI_IS_32BIT
add rsi, rdx ; next line
add rdi, dword ptr arg(5) ;dst_pitch ;
%else
movsxd r8, dword ptr arg(5) ;dst_pitch ;
add rsi, rdx ; next line
add rdi, r8
%endif
cmp rdi, rcx ;
jne .next_row_4x4
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
SECTION_RODATA
align 16
rd:
@ -698,30 +266,3 @@ sym(vp9_six_tap_mmx):
times 8 dw -6
times 8 dw 0
align 16
global HIDDEN_DATA(sym(vp9_bilinear_filters_8x_mmx))
sym(vp9_bilinear_filters_8x_mmx):
times 8 dw 128
times 8 dw 0
times 8 dw 112
times 8 dw 16
times 8 dw 96
times 8 dw 32
times 8 dw 80
times 8 dw 48
times 8 dw 64
times 8 dw 64
times 8 dw 48
times 8 dw 80
times 8 dw 32
times 8 dw 96
times 8 dw 16
times 8 dw 112

90
vp9/common/x86/vp9_subpixel_variance_sse2.c Normal file
View File

@ -0,0 +1,90 @@
/*
* 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.
*/
#define HALFNDX 8
void vp9_half_horiz_variance16x_h_sse2(const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared);
void vp9_half_vert_variance16x_h_sse2(const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared);
void vp9_half_horiz_vert_variance16x_h_sse2(const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared);
void vp9_filter_block2d_bil_var_sse2(const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int xoffset,
int yoffset,
int *sum,
unsigned int *sumsquared);
unsigned int vp9_sub_pixel_variance16x2_sse2(const unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse) {
int xsum0, xsum1;
unsigned int xxsum0, xxsum1;
if (xoffset == HALFNDX && yoffset == 0) {
vp9_half_horiz_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 2,
&xsum0, &xxsum0);
} else if (xoffset == 0 && yoffset == HALFNDX) {
vp9_half_vert_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 2,
&xsum0, &xxsum0);
} else if (xoffset == HALFNDX && yoffset == HALFNDX) {
vp9_half_horiz_vert_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 2,
&xsum0, &xxsum0);
} else {
vp9_filter_block2d_bil_var_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 2,
xoffset, yoffset,
&xsum0, &xxsum0);
vp9_filter_block2d_bil_var_sse2(
src_ptr + 8, src_pixels_per_line,
dst_ptr + 8, dst_pixels_per_line, 2,
xoffset, yoffset,
&xsum1, &xxsum1);
xsum0 += xsum1;
xxsum0 += xxsum1;
}
*sse = xxsum0;
return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 5));
}

13
vp9/common/x86/vp9_subpixel_x86.h
View File

@ -25,10 +25,6 @@ extern prototype_subpixel_predict(vp9_sixtap_predict8x8_mmx);
extern prototype_subpixel_predict(vp9_sixtap_predict8x4_mmx);
extern prototype_subpixel_predict(vp9_sixtap_predict4x4_mmx);
extern prototype_subpixel_predict(vp9_bilinear_predict16x16_mmx);
extern prototype_subpixel_predict(vp9_bilinear_predict8x8_mmx);
extern prototype_subpixel_predict(vp9_bilinear_predict8x4_mmx);
extern prototype_subpixel_predict(vp9_bilinear_predict4x4_mmx);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp9_subpix_sixtap16x16
@ -46,15 +42,6 @@ extern prototype_subpixel_predict(vp9_bilinear_predict4x4_mmx);
#undef vp9_subpix_bilinear16x16
#define vp9_subpix_bilinear16x16 vp9_bilinear_predict16x16_mmx
#undef vp9_subpix_bilinear8x8
#define vp9_subpix_bilinear8x8 vp9_bilinear_predict8x8_mmx
#undef vp9_subpix_bilinear8x4
#define vp9_subpix_bilinear8x4 vp9_bilinear_predict8x4_mmx
#undef vp9_subpix_bilinear4x4
#define vp9_subpix_bilinear4x4 vp9_bilinear_predict4x4_mmx
#endif
#endif

6
vp9/decoder/vp9_dboolhuff.h
View File

@ -8,9 +8,9 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_DECODER_VP9_DBOOLHUFF_H_
#define VP9_DECODER_VP9_DBOOLHUFF_H_
#include <stddef.h>
#include <limits.h>
#include "./vpx_config.h"
@ -33,7 +33,7 @@ typedef struct {
unsigned int range;
} BOOL_DECODER;
DECLARE_ALIGNED(16, extern const unsigned char, vp9_norm[256]);
DECLARE_ALIGNED(16, extern const uint8_t, vp9_norm[256]);
int vp9_start_decode(BOOL_DECODER *br,
const unsigned char *source,
@ -152,4 +152,4 @@ static int bool_error(BOOL_DECODER *br) {
extern int vp9_decode_unsigned_max(BOOL_DECODER *br, int max);
#endif
#endif // VP9_DECODER_VP9_DBOOLHUFF_H_

303
vp9/decoder/vp9_decodemv.c
View File

@ -14,7 +14,7 @@
#include "vp9/common/vp9_entropymode.h"
#include "vp9/decoder/vp9_onyxd_int.h"
#include "vp9/common/vp9_findnearmv.h"
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_seg_common.h"
#include "vp9/common/vp9_pred_common.h"
#include "vp9/common/vp9_entropy.h"
@ -51,7 +51,6 @@ static int read_ymode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_ymode_tree, p);
}
#if CONFIG_SUPERBLOCKS
static int read_sb_ymode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_sb_ymode_tree, p);
}
@ -59,7 +58,6 @@ static int read_sb_ymode(vp9_reader *bc, const vp9_prob *p) {
static int read_kf_sb_ymode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_uv_mode_tree, p);
}
#endif
static int read_kf_mb_ymode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_kf_ymode_tree, p);
@ -122,7 +120,21 @@ static void kfread_modes(VP9D_COMP *pbi,
m->mbmi.segment_id = 0;
if (pbi->mb.update_mb_segmentation_map) {
read_mb_segid(bc, &m->mbmi, &pbi->mb);
pbi->common.last_frame_seg_map[map_index] = m->mbmi.segment_id;
if (m->mbmi.sb_type) {
const int nmbs = 1 << m->mbmi.sb_type;
const int ymbs = MIN(cm->mb_rows - mb_row, nmbs);
const int xmbs = MIN(cm->mb_cols - mb_col, nmbs);
int x, y;
for (y = 0; y < ymbs; y++) {
for (x = 0; x < xmbs; x++) {
cm->last_frame_seg_map[map_index + x + y * cm->mb_cols] =
m->mbmi.segment_id;
}
}
} else {
cm->last_frame_seg_map[map_index] = m->mbmi.segment_id;
}
}
m->mbmi.mb_skip_coeff = 0;
@ -144,25 +156,18 @@ static void kfread_modes(VP9D_COMP *pbi,
m->mbmi.mb_skip_coeff = 0;
}
#if CONFIG_SUPERBLOCKS
if (m->mbmi.encoded_as_sb) {
if (m->mbmi.sb_type) {
y_mode = (MB_PREDICTION_MODE) read_kf_sb_ymode(bc,
pbi->common.sb_kf_ymode_prob[pbi->common.kf_ymode_probs_index]);
} else
#endif
y_mode = (MB_PREDICTION_MODE) read_kf_mb_ymode(bc,
pbi->common.kf_ymode_prob[pbi->common.kf_ymode_probs_index]);
#if CONFIG_COMP_INTRA_PRED
m->mbmi.second_mode = (MB_PREDICTION_MODE)(DC_PRED - 1);
#endif
} else {
y_mode = (MB_PREDICTION_MODE) read_kf_mb_ymode(bc,
pbi->common.kf_ymode_prob[pbi->common.kf_ymode_probs_index]);
}
m->mbmi.ref_frame = INTRA_FRAME;
if ((m->mbmi.mode = y_mode) == B_PRED) {
int i = 0;
#if CONFIG_COMP_INTRA_PRED
int use_comp_pred = vp9_read(bc, DEFAULT_COMP_INTRA_PROB);
#endif
do {
const B_PREDICTION_MODE A = above_block_mode(m, i, mis);
const B_PREDICTION_MODE L = left_block_mode(m, i);
@ -170,15 +175,6 @@ static void kfread_modes(VP9D_COMP *pbi,
m->bmi[i].as_mode.first =
(B_PREDICTION_MODE) read_kf_bmode(
bc, pbi->common.kf_bmode_prob [A] [L]);
#if CONFIG_COMP_INTRA_PRED
if (use_comp_pred) {
m->bmi[i].as_mode.second =
(B_PREDICTION_MODE) read_kf_bmode(
bc, pbi->common.kf_bmode_prob [A] [L]);
} else {
m->bmi[i].as_mode.second = (B_PREDICTION_MODE)(B_DC_PRED - 1);
}
#endif
} while (++i < 16);
}
if ((m->mbmi.mode = y_mode) == I8X8_PRED) {
@ -191,26 +187,22 @@ static void kfread_modes(VP9D_COMP *pbi,
m->bmi[ib + 1].as_mode.first = mode8x8;
m->bmi[ib + 4].as_mode.first = mode8x8;
m->bmi[ib + 5].as_mode.first = mode8x8;
#if CONFIG_COMP_INTRA_PRED
m->bmi[ib + 0].as_mode.second = (MB_PREDICTION_MODE)(DC_PRED - 1);
m->bmi[ib + 1].as_mode.second = (MB_PREDICTION_MODE)(DC_PRED - 1);
m->bmi[ib + 4].as_mode.second = (MB_PREDICTION_MODE)(DC_PRED - 1);
m->bmi[ib + 5].as_mode.second = (MB_PREDICTION_MODE)(DC_PRED - 1);
#endif
}
} else
m->mbmi.uv_mode = (MB_PREDICTION_MODE)read_uv_mode(bc,
pbi->common.kf_uv_mode_prob[m->mbmi.mode]);
#if CONFIG_COMP_INTRA_PRED
m->mbmi.second_uv_mode = (MB_PREDICTION_MODE)(DC_PRED - 1);
#endif
if (cm->txfm_mode == TX_MODE_SELECT && m->mbmi.mb_skip_coeff == 0 &&
m->mbmi.mode <= I8X8_PRED) {
// FIXME(rbultje) code ternary symbol once all experiments are merged
m->mbmi.txfm_size = vp9_read(bc, cm->prob_tx[0]);
if (m->mbmi.txfm_size != TX_4X4 && m->mbmi.mode != I8X8_PRED)
if (m->mbmi.txfm_size != TX_4X4 && m->mbmi.mode != I8X8_PRED) {
m->mbmi.txfm_size += vp9_read(bc, cm->prob_tx[1]);
if (m->mbmi.txfm_size != TX_8X8 && m->mbmi.sb_type)
m->mbmi.txfm_size += vp9_read(bc, cm->prob_tx[2]);
}
} else if (cm->txfm_mode >= ALLOW_32X32 && m->mbmi.sb_type) {
m->mbmi.txfm_size = TX_32X32;
} else if (cm->txfm_mode >= ALLOW_16X16 && m->mbmi.mode <= TM_PRED) {
m->mbmi.txfm_size = TX_16X16;
} else if (cm->txfm_mode >= ALLOW_8X8 && m->mbmi.mode != B_PRED) {
@ -478,11 +470,9 @@ static MV_REFERENCE_FRAME read_ref_frame(VP9D_COMP *pbi,
return (MV_REFERENCE_FRAME)ref_frame;
}
#if CONFIG_SUPERBLOCKS
static MB_PREDICTION_MODE read_sb_mv_ref(vp9_reader *bc, const vp9_prob *p) {
return (MB_PREDICTION_MODE) treed_read(bc, vp9_sb_mv_ref_tree, p);
}
#endif
static MB_PREDICTION_MODE read_mv_ref(vp9_reader *bc, const vp9_prob *p) {
return (MB_PREDICTION_MODE) treed_read(bc, vp9_mv_ref_tree, p);
@ -532,12 +522,6 @@ static void mb_mode_mv_init(VP9D_COMP *pbi, vp9_reader *bc) {
if (!cm->kf_ymode_probs_update)
cm->kf_ymode_probs_index = vp9_read_literal(bc, 3);
} else {
#if CONFIG_PRED_FILTER
cm->pred_filter_mode = (vp9_prob)vp9_read_literal(bc, 2);
if (cm->pred_filter_mode == 2)
cm->prob_pred_filter_off = (vp9_prob)vp9_read_literal(bc, 8);
#endif
if (cm->mcomp_filter_type == SWITCHABLE)
read_switchable_interp_probs(pbi, bc);
#if CONFIG_COMP_INTERINTRA_PRED
@ -572,7 +556,6 @@ static void mb_mode_mv_init(VP9D_COMP *pbi, vp9_reader *bc) {
} while (++i < VP9_YMODES - 1);
}
#if CONFIG_SUPERBLOCKS
if (vp9_read_bit(bc)) {
int i = 0;
@ -580,12 +563,6 @@ static void mb_mode_mv_init(VP9D_COMP *pbi, vp9_reader *bc) {
cm->fc.sb_ymode_prob[i] = (vp9_prob) vp9_read_literal(bc, 8);
} while (++i < VP9_I32X32_MODES - 1);
}
#endif
#if CONFIG_NEW_MVREF
// Temp defaults probabilities for ecnoding the MV ref id signal
vpx_memset(xd->mb_mv_ref_id_probs, 192, sizeof(xd->mb_mv_ref_id_probs));
#endif
read_nmvprobs(bc, nmvc, xd->allow_high_precision_mv);
}
@ -633,38 +610,38 @@ static void read_mb_segment_id(VP9D_COMP *pbi,
else {
read_mb_segid(bc, mbmi, xd);
}
#if CONFIG_SUPERBLOCKS
if (mbmi->encoded_as_sb) {
cm->last_frame_seg_map[index] = mbmi->segment_id;
if (mb_col + 1 < cm->mb_cols)
cm->last_frame_seg_map[index + 1] = mbmi->segment_id;
if (mb_row + 1 < cm->mb_rows) {
cm->last_frame_seg_map[index + cm->mb_cols] = mbmi->segment_id;
if (mb_col + 1 < cm->mb_cols)
cm->last_frame_seg_map[index + cm->mb_cols + 1] = mbmi->segment_id;
if (mbmi->sb_type) {
const int nmbs = 1 << mbmi->sb_type;
const int ymbs = MIN(cm->mb_rows - mb_row, nmbs);
const int xmbs = MIN(cm->mb_cols - mb_col, nmbs);
int x, y;
for (y = 0; y < ymbs; y++) {
for (x = 0; x < xmbs; x++) {
cm->last_frame_seg_map[index + x + y * cm->mb_cols] =
mbmi->segment_id;
}
}
} else
#endif
{
} else {
cm->last_frame_seg_map[index] = mbmi->segment_id;
}
} else {
#if CONFIG_SUPERBLOCKS
if (mbmi->encoded_as_sb) {
mbmi->segment_id = cm->last_frame_seg_map[index];
if (mb_col < cm->mb_cols - 1)
mbmi->segment_id = mbmi->segment_id &&
cm->last_frame_seg_map[index + 1];
if (mb_row < cm->mb_rows - 1) {
mbmi->segment_id = mbmi->segment_id &&
cm->last_frame_seg_map[index + cm->mb_cols];
if (mb_col < cm->mb_cols - 1)
mbmi->segment_id = mbmi->segment_id &&
cm->last_frame_seg_map[index + cm->mb_cols + 1];
if (mbmi->sb_type) {
const int nmbs = 1 << mbmi->sb_type;
const int ymbs = MIN(cm->mb_rows - mb_row, nmbs);
const int xmbs = MIN(cm->mb_cols - mb_col, nmbs);
unsigned segment_id = -1;
int x, y;
for (y = 0; y < ymbs; y++) {
for (x = 0; x < xmbs; x++) {
segment_id = MIN(segment_id,
cm->last_frame_seg_map[index + x +
y * cm->mb_cols]);
}
}
} else
#endif
{
mbmi->segment_id = segment_id;
} else {
mbmi->segment_id = cm->last_frame_seg_map[index];
}
}
@ -689,6 +666,7 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
int mb_to_right_edge;
int mb_to_top_edge;
int mb_to_bottom_edge;
const int mb_size = 1 << mi->mbmi.sb_type;
mb_to_top_edge = xd->mb_to_top_edge;
mb_to_bottom_edge = xd->mb_to_bottom_edge;
@ -703,18 +681,8 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
xd->mb_to_left_edge =
mb_to_left_edge = -((mb_col * 16) << 3);
mb_to_left_edge -= LEFT_TOP_MARGIN;
#if CONFIG_SUPERBLOCKS
if (mi->mbmi.encoded_as_sb) {
xd->mb_to_right_edge =
mb_to_right_edge = ((pbi->common.mb_cols - 2 - mb_col) * 16) << 3;
} else {
#endif
xd->mb_to_right_edge =
mb_to_right_edge = ((pbi->common.mb_cols - 1 - mb_col) * 16) << 3;
#if CONFIG_SUPERBLOCKS
}
#endif
xd->mb_to_right_edge =
mb_to_right_edge = ((pbi->common.mb_cols - mb_size - mb_col) * 16) << 3;
mb_to_right_edge += RIGHT_BOTTOM_MARGIN;
// Make sure the MACROBLOCKD mode info pointer is pointed at the
@ -756,10 +724,10 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
int recon_y_stride, recon_yoffset;
int recon_uv_stride, recon_uvoffset;
MV_REFERENCE_FRAME ref_frame = mbmi->ref_frame;
{
int ref_fb_idx;
MV_REFERENCE_FRAME ref_frame = mbmi->ref_frame;
/* Select the appropriate reference frame for this MB */
if (ref_frame == LAST_FRAME)
@ -788,14 +756,33 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
ref_frame, mbmi->ref_mvs[ref_frame],
cm->ref_frame_sign_bias);
vp9_find_best_ref_mvs(xd,
xd->pre.y_buffer,
recon_y_stride,
mbmi->ref_mvs[ref_frame],
&best_mv, &nearest, &nearby);
vp9_mv_ref_probs(&pbi->common, mv_ref_p,
mbmi->mb_mode_context[ref_frame]);
// Is the segment level mode feature enabled for this segment
if (vp9_segfeature_active(xd, mbmi->segment_id, SEG_LVL_MODE)) {
mbmi->mode =
vp9_get_segdata(xd, mbmi->segment_id, SEG_LVL_MODE);
} else {
if (mbmi->sb_type)
mbmi->mode = read_sb_mv_ref(bc, mv_ref_p);
else
mbmi->mode = read_mv_ref(bc, mv_ref_p);
vp9_accum_mv_refs(&pbi->common, mbmi->mode,
mbmi->mb_mode_context[ref_frame]);
}
if (mbmi->mode != ZEROMV) {
vp9_find_best_ref_mvs(xd,
xd->pre.y_buffer,
recon_y_stride,
mbmi->ref_mvs[ref_frame],
&nearest, &nearby);
best_mv.as_int = (mbmi->ref_mvs[ref_frame][0]).as_int;
}
#ifdef DEC_DEBUG
if (dec_debug)
printf("[D %d %d] %d %d %d %d\n", ref_frame,
@ -804,32 +791,6 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
#endif
}
// Is the segment level mode feature enabled for this segment
if (vp9_segfeature_active(xd, mbmi->segment_id, SEG_LVL_MODE)) {
mbmi->mode =
vp9_get_segdata(xd, mbmi->segment_id, SEG_LVL_MODE);
} else {
#if CONFIG_SUPERBLOCKS
if (mbmi->encoded_as_sb) {
mbmi->mode = read_sb_mv_ref(bc, mv_ref_p);
} else
#endif
mbmi->mode = read_mv_ref(bc, mv_ref_p);
vp9_accum_mv_refs(&pbi->common, mbmi->mode,
mbmi->mb_mode_context[mbmi->ref_frame]);
}
#if CONFIG_PRED_FILTER
if (mbmi->mode >= NEARESTMV && mbmi->mode < SPLITMV) {
// Is the prediction filter enabled
if (cm->pred_filter_mode == 2)
mbmi->pred_filter_enabled =
vp9_read(bc, cm->prob_pred_filter_off);
else
mbmi->pred_filter_enabled = cm->pred_filter_mode;
}
#endif
if (mbmi->mode >= NEARESTMV && mbmi->mode <= SPLITMV)
{
if (cm->mcomp_filter_type == SWITCHABLE) {
@ -877,13 +838,15 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
mbmi->ref_mvs[mbmi->second_ref_frame],
cm->ref_frame_sign_bias);
vp9_find_best_ref_mvs(xd,
xd->second_pre.y_buffer,
recon_y_stride,
mbmi->ref_mvs[mbmi->second_ref_frame],
&best_mv_second,
&nearest_second,
&nearby_second);
if (mbmi->mode != ZEROMV) {
vp9_find_best_ref_mvs(xd,
xd->second_pre.y_buffer,
recon_y_stride,
mbmi->ref_mvs[mbmi->second_ref_frame],
&nearest_second,
&nearby_second);
best_mv_second = mbmi->ref_mvs[mbmi->second_ref_frame][0];
}
}
} else {
@ -916,6 +879,29 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
#endif
}
#if CONFIG_NEW_MVREF
// if ((mbmi->mode == NEWMV) || (mbmi->mode == SPLITMV))
if (mbmi->mode == NEWMV) {
int best_index;
MV_REFERENCE_FRAME ref_frame = mbmi->ref_frame;
// Encode the index of the choice.
best_index =
vp9_read_mv_ref_id(bc, xd->mb_mv_ref_probs[ref_frame]);
best_mv.as_int = mbmi->ref_mvs[ref_frame][best_index].as_int;
if (mbmi->second_ref_frame > 0) {
ref_frame = mbmi->second_ref_frame;
// Encode the index of the choice.
best_index =
vp9_read_mv_ref_id(bc, xd->mb_mv_ref_probs[ref_frame]);
best_mv_second.as_int = mbmi->ref_mvs[ref_frame][best_index].as_int;
}
}
#endif
mbmi->uv_mode = DC_PRED;
switch (mbmi->mode) {
case SPLITMV: {
@ -1072,19 +1058,6 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
case NEWMV:
#if CONFIG_NEW_MVREF
{
int best_index;
MV_REFERENCE_FRAME ref_frame = mbmi->ref_frame;
// Encode the index of the choice.
best_index =
vp9_read_mv_ref_id(bc, xd->mb_mv_ref_id_probs[ref_frame]);
best_mv.as_int = mbmi->ref_mvs[ref_frame][best_index].as_int;
}
#endif
read_nmv(bc, &mv->as_mv, &best_mv.as_mv, nmvc);
read_nmv_fp(bc, &mv->as_mv, &best_mv.as_mv, nmvc,
xd->allow_high_precision_mv);
@ -1106,18 +1079,6 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
mb_to_bottom_edge);
if (mbmi->second_ref_frame > 0) {
#if CONFIG_NEW_MVREF
{
int best_index;
MV_REFERENCE_FRAME ref_frame = mbmi->second_ref_frame;
// Encode the index of the choice.
best_index =
vp9_read_mv_ref_id(bc, xd->mb_mv_ref_id_probs[ref_frame]);
best_mv_second.as_int = mbmi->ref_mvs[ref_frame][best_index].as_int;
}
#endif
read_nmv(bc, &mbmi->mv[1].as_mv, &best_mv_second.as_mv, nmvc);
read_nmv_fp(bc, &mbmi->mv[1].as_mv, &best_mv_second.as_mv, nmvc,
xd->allow_high_precision_mv);
@ -1144,27 +1105,19 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
if (vp9_segfeature_active(xd, mbmi->segment_id, SEG_LVL_MODE)) {
mbmi->mode = (MB_PREDICTION_MODE)
vp9_get_segdata(xd, mbmi->segment_id, SEG_LVL_MODE);
#if CONFIG_SUPERBLOCKS
} else if (mbmi->encoded_as_sb) {
} else if (mbmi->sb_type) {
mbmi->mode = (MB_PREDICTION_MODE)
read_sb_ymode(bc, pbi->common.fc.sb_ymode_prob);
pbi->common.fc.sb_ymode_counts[mbmi->mode]++;
#endif
} else {
mbmi->mode = (MB_PREDICTION_MODE)
read_ymode(bc, pbi->common.fc.ymode_prob);
pbi->common.fc.ymode_counts[mbmi->mode]++;
}
#if CONFIG_COMP_INTRA_PRED
mbmi->second_mode = (MB_PREDICTION_MODE)(DC_PRED - 1);
#endif
// If MB mode is BPRED read the block modes
if (mbmi->mode == B_PRED) {
int j = 0;
#if CONFIG_COMP_INTRA_PRED
int use_comp_pred = vp9_read(bc, DEFAULT_COMP_INTRA_PROB);
#endif
do {
int m;
m = mi->bmi[j].as_mode.first = (B_PREDICTION_MODE)
@ -1173,13 +1126,6 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
if (m == B_CONTEXT_PRED) m -= CONTEXT_PRED_REPLACEMENTS;
#endif
pbi->common.fc.bmode_counts[m]++;
#if CONFIG_COMP_INTRA_PRED
if (use_comp_pred) {
mi->bmi[j].as_mode.second = (B_PREDICTION_MODE)read_bmode(bc, pbi->common.fc.bmode_prob);
} else {
mi->bmi[j].as_mode.second = (B_PREDICTION_MODE)(B_DC_PRED - 1);
}
#endif
} while (++j < 16);
}
@ -1194,22 +1140,12 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
mi->bmi[ib + 4].as_mode.first = mode8x8;
mi->bmi[ib + 5].as_mode.first = mode8x8;
pbi->common.fc.i8x8_mode_counts[mode8x8]++;
#if CONFIG_COMP_INTRA_PRED
mi->bmi[ib + 0].as_mode.second = (MB_PREDICTION_MODE)(DC_PRED - 1);
mi->bmi[ib + 1].as_mode.second = (MB_PREDICTION_MODE)(DC_PRED - 1);
mi->bmi[ib + 4].as_mode.second = (MB_PREDICTION_MODE)(DC_PRED - 1);
mi->bmi[ib + 5].as_mode.second = (MB_PREDICTION_MODE)(DC_PRED - 1);
#endif
}
} else {
mbmi->uv_mode = (MB_PREDICTION_MODE)read_uv_mode(
bc, pbi->common.fc.uv_mode_prob[mbmi->mode]);
pbi->common.fc.uv_mode_counts[mbmi->mode][mbmi->uv_mode]++;
}
#if CONFIG_COMP_INTRA_PRED
mbmi->second_uv_mode = (MB_PREDICTION_MODE)(DC_PRED - 1);
#endif
}
if (cm->txfm_mode == TX_MODE_SELECT && mbmi->mb_skip_coeff == 0 &&
@ -1219,8 +1155,13 @@ static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
// FIXME(rbultje) code ternary symbol once all experiments are merged
mbmi->txfm_size = vp9_read(bc, cm->prob_tx[0]);
if (mbmi->txfm_size != TX_4X4 && mbmi->mode != I8X8_PRED &&
mbmi->mode != SPLITMV)
mbmi->mode != SPLITMV) {
mbmi->txfm_size += vp9_read(bc, cm->prob_tx[1]);
if (mbmi->sb_type && mbmi->txfm_size != TX_8X8)
mbmi->txfm_size += vp9_read(bc, cm->prob_tx[2]);
}
} else if (mbmi->sb_type && cm->txfm_mode >= ALLOW_32X32) {
mbmi->txfm_size = TX_32X32;
} else if (cm->txfm_mode >= ALLOW_16X16 &&
((mbmi->ref_frame == INTRA_FRAME && mbmi->mode <= TM_PRED) ||
(mbmi->ref_frame != INTRA_FRAME && mbmi->mode != SPLITMV))) {

4
vp9/decoder/vp9_decodemv.h
View File

@ -8,6 +8,8 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_DECODER_VP9_DECODEMV_H_
#define VP9_DECODER_VP9_DECODEMV_H_
#include "vp9/decoder/vp9_onyxd_int.h"
@ -17,3 +19,5 @@ void vp9_decode_mb_mode_mv(VP9D_COMP* const pbi,
int mb_col,
BOOL_DECODER* const bc);
void vp9_decode_mode_mvs_init(VP9D_COMP* const pbi, BOOL_DECODER* const bc);
#endif // VP9_DECODER_VP9_DECODEMV_H_

741
vp9/decoder/vp9_decodframe.c
View File

@ -10,17 +10,19 @@
#include "vp9/decoder/vp9_onyxd_int.h"
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_header.h"
#include "vp9/common/vp9_reconintra.h"
#include "vp9/common/vp9_reconintra4x4.h"
#include "vp9/common/vp9_reconinter.h"
#include "vp9/common/vp9_entropy.h"
#include "vp9/decoder/vp9_decodframe.h"
#include "vp9/decoder/vp9_detokenize.h"
#include "vp9/common/vp9_invtrans.h"
#include "vp9/common/vp9_alloccommon.h"
#include "vp9/common/vp9_entropymode.h"
#include "vp9/common/vp9_quant_common.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "vp9/common/vp9_setupintrarecon.h"
#include "vp9/decoder/vp9_decodemv.h"
@ -81,17 +83,17 @@ void vp9_init_de_quantizer(VP9D_COMP *pbi) {
VP9_COMMON *const pc = &pbi->common;
for (Q = 0; Q < QINDEX_RANGE; Q++) {
pc->Y1dequant[Q][0] = (short)vp9_dc_quant(Q, pc->y1dc_delta_q);
pc->Y2dequant[Q][0] = (short)vp9_dc2quant(Q, pc->y2dc_delta_q);
pc->UVdequant[Q][0] = (short)vp9_dc_uv_quant(Q, pc->uvdc_delta_q);
pc->Y1dequant[Q][0] = (int16_t)vp9_dc_quant(Q, pc->y1dc_delta_q);
pc->Y2dequant[Q][0] = (int16_t)vp9_dc2quant(Q, pc->y2dc_delta_q);
pc->UVdequant[Q][0] = (int16_t)vp9_dc_uv_quant(Q, pc->uvdc_delta_q);
/* all the ac values =; */
for (i = 1; i < 16; i++) {
int rc = vp9_default_zig_zag1d[i];
int rc = vp9_default_zig_zag1d_4x4[i];
pc->Y1dequant[Q][rc] = (short)vp9_ac_yquant(Q);
pc->Y2dequant[Q][rc] = (short)vp9_ac2quant(Q, pc->y2ac_delta_q);
pc->UVdequant[Q][rc] = (short)vp9_ac_uv_quant(Q, pc->uvac_delta_q);
pc->Y1dequant[Q][rc] = (int16_t)vp9_ac_yquant(Q);
pc->Y2dequant[Q][rc] = (int16_t)vp9_ac2quant(Q, pc->y2ac_delta_q);
pc->UVdequant[Q][rc] = (int16_t)vp9_ac_uv_quant(Q, pc->uvac_delta_q);
}
}
}
@ -170,20 +172,25 @@ static void mb_init_dequantizer(VP9D_COMP *pbi, MACROBLOCKD *xd) {
*/
static void skip_recon_mb(VP9D_COMP *pbi, MACROBLOCKD *xd) {
if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) {
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB64X64) {
vp9_build_intra_predictors_sb64uv_s(xd);
vp9_build_intra_predictors_sb64y_s(xd);
} else if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB32X32) {
vp9_build_intra_predictors_sbuv_s(xd);
vp9_build_intra_predictors_sby_s(xd);
} else {
#endif
vp9_build_intra_predictors_mbuv_s(xd);
vp9_build_intra_predictors_mby_s(xd);
#if CONFIG_SUPERBLOCKS
vp9_build_intra_predictors_mbuv_s(xd);
vp9_build_intra_predictors_mby_s(xd);
}
#endif
} else {
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB64X64) {
vp9_build_inter64x64_predictors_sb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
} else if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB32X32) {
vp9_build_inter32x32_predictors_sb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
@ -191,35 +198,32 @@ static void skip_recon_mb(VP9D_COMP *pbi, MACROBLOCKD *xd) {
xd->dst.y_stride,
xd->dst.uv_stride);
} else {
#endif
vp9_build_1st_inter16x16_predictors_mb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
if (xd->mode_info_context->mbmi.second_ref_frame > 0) {
vp9_build_2nd_inter16x16_predictors_mb(xd,
vp9_build_1st_inter16x16_predictors_mb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
}
#if CONFIG_COMP_INTERINTRA_PRED
else if (xd->mode_info_context->mbmi.second_ref_frame == INTRA_FRAME) {
vp9_build_interintra_16x16_predictors_mb(xd,
if (xd->mode_info_context->mbmi.second_ref_frame > 0) {
vp9_build_2nd_inter16x16_predictors_mb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
}
}
#if CONFIG_COMP_INTERINTRA_PRED
else if (xd->mode_info_context->mbmi.second_ref_frame == INTRA_FRAME) {
vp9_build_interintra_16x16_predictors_mb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
}
#endif
#if CONFIG_SUPERBLOCKS
}
#endif
}
}
@ -283,10 +287,10 @@ static void decode_8x8(VP9D_COMP *pbi, MACROBLOCKD *xd,
for (i = 0; i < 4; i++) {
int ib = vp9_i8x8_block[i];
int idx = (ib & 0x02) ? (ib + 2) : ib;
short *q = xd->block[idx].qcoeff;
short *dq = xd->block[0].dequant;
unsigned char *pre = xd->block[ib].predictor;
unsigned char *dst = *(xd->block[ib].base_dst) + xd->block[ib].dst;
int16_t *q = xd->block[idx].qcoeff;
int16_t *dq = xd->block[0].dequant;
uint8_t *pre = xd->block[ib].predictor;
uint8_t *dst = *(xd->block[ib].base_dst) + xd->block[ib].dst;
int stride = xd->dst.y_stride;
BLOCKD *b = &xd->block[ib];
if (xd->mode_info_context->mbmi.mode == I8X8_PRED) {
@ -414,9 +418,6 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd,
assert(get_2nd_order_usage(xd) == 0);
for (i = 0; i < 16; i++) {
int b_mode;
#if CONFIG_COMP_INTRA_PRED
int b_mode2;
#endif
BLOCKD *b = &xd->block[i];
b_mode = xd->mode_info_context->bmi[i].as_mode.first;
#if CONFIG_NEWBINTRAMODES
@ -425,17 +426,8 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd,
#endif
if (!xd->mode_info_context->mbmi.mb_skip_coeff)
eobtotal += vp9_decode_coefs_4x4(pbi, xd, bc, PLANE_TYPE_Y_WITH_DC, i);
#if CONFIG_COMP_INTRA_PRED
b_mode2 = xd->mode_info_context->bmi[i].as_mode.second;
if (b_mode2 == (B_PREDICTION_MODE)(B_DC_PRED - 1)) {
#endif
vp9_intra4x4_predict(b, b_mode, b->predictor);
#if CONFIG_COMP_INTRA_PRED
} else {
vp9_comp_intra4x4_predict(b, b_mode, b_mode2, b->predictor);
}
#endif
vp9_intra4x4_predict(b, b_mode, b->predictor);
tx_type = get_tx_type_4x4(xd, b);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_c(tx_type, b->qcoeff,
@ -446,12 +438,12 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd,
vp9_dequant_idct_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 16, b->dst_stride);
}
xd->above_context->y2 = 1;
xd->left_context->y2 = 1;
}
if (!xd->mode_info_context->mbmi.mb_skip_coeff) {
vp9_decode_mb_tokens_4x4_uv(pbi, xd, bc);
}
xd->above_context->y2 = 0;
xd->left_context->y2 = 0;
vp9_build_intra_predictors_mbuv(xd);
pbi->idct_add_uv_block(xd->qcoeff + 16 * 16,
xd->block[16].dequant,
@ -546,10 +538,10 @@ static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd,
}
}
#if CONFIG_SUPERBLOCKS
static void decode_16x16_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc, int n) {
int x_idx = n & 1, y_idx = n >> 1;
BOOL_DECODER* const bc, int n,
int maska, int shiftb) {
int x_idx = n & maska, y_idx = n >> shiftb;
TX_TYPE tx_type = get_tx_type_16x16(xd, &xd->block[0]);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_16x16_c(
@ -573,17 +565,18 @@ static void decode_16x16_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
};
static void decode_8x8_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc, int n) {
BOOL_DECODER* const bc, int n,
int maska, int shiftb) {
int x_idx = n & maska, y_idx = n >> shiftb;
BLOCKD *b = &xd->block[24];
int x_idx = n & 1, y_idx = n >> 1;
TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[0]);
if (tx_type != DCT_DCT) {
int i;
for (i = 0; i < 4; i++) {
int ib = vp9_i8x8_block[i];
int idx = (ib & 0x02) ? (ib + 2) : ib;
short *q = xd->block[idx].qcoeff;
short *dq = xd->block[0].dequant;
int16_t *q = xd->block[idx].qcoeff;
int16_t *dq = xd->block[0].dequant;
int stride = xd->dst.y_stride;
BLOCKD *b = &xd->block[ib];
tx_type = get_tx_type_8x8(xd, &xd->block[ib]);
@ -634,9 +627,10 @@ static void decode_8x8_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
};
static void decode_4x4_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc, int n) {
BOOL_DECODER* const bc, int n,
int maska, int shiftb) {
int x_idx = n & maska, y_idx = n >> shiftb;
BLOCKD *b = &xd->block[24];
int x_idx = n & 1, y_idx = n >> 1;
TX_TYPE tx_type = get_tx_type_4x4(xd, &xd->block[0]);
if (tx_type != DCT_DCT) {
int i;
@ -689,15 +683,142 @@ static void decode_4x4_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
xd->dst.uv_stride, xd->eobs + 16, xd);
};
static void decode_superblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
int mb_row, unsigned int mb_col,
BOOL_DECODER* const bc) {
static void decode_superblock64(VP9D_COMP *pbi, MACROBLOCKD *xd,
int mb_row, unsigned int mb_col,
BOOL_DECODER* const bc) {
int i, n, eobtotal;
TX_SIZE tx_size = xd->mode_info_context->mbmi.txfm_size;
VP9_COMMON *const pc = &pbi->common;
MODE_INFO *orig_mi = xd->mode_info_context;
const int mis = pc->mode_info_stride;
assert(xd->mode_info_context->mbmi.encoded_as_sb);
assert(xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB64X64);
if (pbi->common.frame_type != KEY_FRAME)
vp9_setup_interp_filters(xd, xd->mode_info_context->mbmi.interp_filter, pc);
// re-initialize macroblock dequantizer before detokenization
if (xd->segmentation_enabled)
mb_init_dequantizer(pbi, xd);
if (xd->mode_info_context->mbmi.mb_skip_coeff) {
int n;
vp9_reset_mb_tokens_context(xd);
for (n = 1; n <= 3; n++) {
if (mb_col < pc->mb_cols - n)
xd->above_context += n;
if (mb_row < pc->mb_rows - n)
xd->left_context += n;
vp9_reset_mb_tokens_context(xd);
if (mb_col < pc->mb_cols - n)
xd->above_context -= n;
if (mb_row < pc->mb_rows - n)
xd->left_context -= n;
}
/* Special case: Force the loopfilter to skip when eobtotal and
* mb_skip_coeff are zero.
*/
skip_recon_mb(pbi, xd);
return;
}
/* do prediction */
if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) {
vp9_build_intra_predictors_sb64y_s(xd);
vp9_build_intra_predictors_sb64uv_s(xd);
} else {
vp9_build_inter64x64_predictors_sb(xd, xd->dst.y_buffer,
xd->dst.u_buffer, xd->dst.v_buffer,
xd->dst.y_stride, xd->dst.uv_stride);
}
/* dequantization and idct */
if (xd->mode_info_context->mbmi.txfm_size == TX_32X32) {
for (n = 0; n < 4; n++) {
const int x_idx = n & 1, y_idx = n >> 1;
if (mb_col + x_idx * 2 >= pc->mb_cols ||
mb_row + y_idx * 2 >= pc->mb_rows)
continue;
xd->left_context = pc->left_context + (y_idx << 1);
xd->above_context = pc->above_context + mb_col + (x_idx << 1);
xd->mode_info_context = orig_mi + x_idx * 2 + y_idx * 2 * mis;
eobtotal = vp9_decode_sb_tokens(pbi, xd, bc);
if (eobtotal == 0) { // skip loopfilter
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
if (mb_col + 1 < pc->mb_cols)
xd->mode_info_context[1].mbmi.mb_skip_coeff = 1;
if (mb_row + 1 < pc->mb_rows) {
xd->mode_info_context[mis].mbmi.mb_skip_coeff = 1;
if (mb_col + 1 < pc->mb_cols)
xd->mode_info_context[mis + 1].mbmi.mb_skip_coeff = 1;
}
} else {
vp9_dequant_idct_add_32x32(xd->sb_coeff_data.qcoeff, xd->block[0].dequant,
xd->dst.y_buffer + x_idx * 32 +
xd->dst.y_stride * y_idx * 32,
xd->dst.y_buffer + x_idx * 32 +
xd->dst.y_stride * y_idx * 32,
xd->dst.y_stride, xd->dst.y_stride,
xd->eobs[0]);
vp9_dequant_idct_add_uv_block_16x16_c(xd->sb_coeff_data.qcoeff + 1024,
xd->block[16].dequant,
xd->dst.u_buffer + x_idx * 16 +
xd->dst.uv_stride * y_idx * 16,
xd->dst.v_buffer + x_idx * 16 +
xd->dst.uv_stride * y_idx * 16,
xd->dst.uv_stride, xd->eobs + 16);
}
}
} else {
for (n = 0; n < 16; n++) {
int x_idx = n & 3, y_idx = n >> 2;
if (mb_col + x_idx >= pc->mb_cols || mb_row + y_idx >= pc->mb_rows)
continue;
xd->above_context = pc->above_context + mb_col + x_idx;
xd->left_context = pc->left_context + y_idx;
xd->mode_info_context = orig_mi + x_idx + y_idx * mis;
for (i = 0; i < 25; i++) {
xd->block[i].eob = 0;
xd->eobs[i] = 0;
}
eobtotal = vp9_decode_mb_tokens(pbi, xd, bc);
if (eobtotal == 0) { // skip loopfilter
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
continue;
}
if (tx_size == TX_16X16) {
decode_16x16_sb(pbi, xd, bc, n, 3, 2);
} else if (tx_size == TX_8X8) {
decode_8x8_sb(pbi, xd, bc, n, 3, 2);
} else {
decode_4x4_sb(pbi, xd, bc, n, 3, 2);
}
}
}
xd->above_context = pc->above_context + mb_col;
xd->left_context = pc->left_context;
xd->mode_info_context = orig_mi;
}
static void decode_superblock32(VP9D_COMP *pbi, MACROBLOCKD *xd,
int mb_row, unsigned int mb_col,
BOOL_DECODER* const bc) {
int i, n, eobtotal;
TX_SIZE tx_size = xd->mode_info_context->mbmi.txfm_size;
VP9_COMMON *const pc = &pbi->common;
MODE_INFO *orig_mi = xd->mode_info_context;
const int mis = pc->mode_info_stride;
assert(xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB32X32);
if (pbi->common.frame_type != KEY_FRAME)
vp9_setup_interp_filters(xd, xd->mode_info_context->mbmi.interp_filter, pc);
@ -736,41 +857,62 @@ static void decode_superblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
}
/* dequantization and idct */
for (n = 0; n < 4; n++) {
int x_idx = n & 1, y_idx = n >> 1;
if (mb_col + x_idx >= pc->mb_cols || mb_row + y_idx >= pc->mb_rows)
continue;
xd->above_context = pc->above_context + mb_col + x_idx;
xd->left_context = pc->left_context + y_idx;
xd->mode_info_context = orig_mi + x_idx + y_idx * pc->mode_info_stride;
for (i = 0; i < 25; i++) {
xd->block[i].eob = 0;
xd->eobs[i] = 0;
}
eobtotal = vp9_decode_mb_tokens(pbi, xd, bc);
if (xd->mode_info_context->mbmi.txfm_size == TX_32X32) {
eobtotal = vp9_decode_sb_tokens(pbi, xd, bc);
if (eobtotal == 0) { // skip loopfilter
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
continue;
}
if (tx_size == TX_16X16) {
decode_16x16_sb(pbi, xd, bc, n);
} else if (tx_size == TX_8X8) {
decode_8x8_sb(pbi, xd, bc, n);
if (mb_col + 1 < pc->mb_cols)
xd->mode_info_context[1].mbmi.mb_skip_coeff = 1;
if (mb_row + 1 < pc->mb_rows) {
xd->mode_info_context[mis].mbmi.mb_skip_coeff = 1;
if (mb_col + 1 < pc->mb_cols)
xd->mode_info_context[mis + 1].mbmi.mb_skip_coeff = 1;
}
} else {
decode_4x4_sb(pbi, xd, bc, n);
vp9_dequant_idct_add_32x32(xd->sb_coeff_data.qcoeff, xd->block[0].dequant,
xd->dst.y_buffer, xd->dst.y_buffer,
xd->dst.y_stride, xd->dst.y_stride,
xd->eobs[0]);
vp9_dequant_idct_add_uv_block_16x16_c(xd->sb_coeff_data.qcoeff + 1024,
xd->block[16].dequant,
xd->dst.u_buffer, xd->dst.v_buffer,
xd->dst.uv_stride, xd->eobs + 16);
}
}
} else {
for (n = 0; n < 4; n++) {
int x_idx = n & 1, y_idx = n >> 1;
xd->above_context = pc->above_context + mb_col;
xd->left_context = pc->left_context;
xd->mode_info_context = orig_mi;
if (mb_col + x_idx >= pc->mb_cols || mb_row + y_idx >= pc->mb_rows)
continue;
xd->above_context = pc->above_context + mb_col + x_idx;
xd->left_context = pc->left_context + y_idx + (mb_row & 2);
xd->mode_info_context = orig_mi + x_idx + y_idx * mis;
for (i = 0; i < 25; i++) {
xd->block[i].eob = 0;
xd->eobs[i] = 0;
}
eobtotal = vp9_decode_mb_tokens(pbi, xd, bc);
if (eobtotal == 0) { // skip loopfilter
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
continue;
}
if (tx_size == TX_16X16) {
decode_16x16_sb(pbi, xd, bc, n, 1, 1);
} else if (tx_size == TX_8X8) {
decode_8x8_sb(pbi, xd, bc, n, 1, 1);
} else {
decode_4x4_sb(pbi, xd, bc, n, 1, 1);
}
}
xd->above_context = pc->above_context + mb_col;
xd->left_context = pc->left_context + (mb_row & 2);
xd->mode_info_context = orig_mi;
}
}
#endif
static void decode_macroblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
int mb_row, unsigned int mb_col,
@ -780,9 +922,7 @@ static void decode_macroblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
int i;
int tx_size;
#if CONFIG_SUPERBLOCKS
assert(!xd->mode_info_context->mbmi.encoded_as_sb);
#endif
assert(!xd->mode_info_context->mbmi.sb_type);
// re-initialize macroblock dequantizer before detokenization
if (xd->segmentation_enabled)
@ -904,192 +1044,176 @@ static int get_delta_q(vp9_reader *bc, int prev, int *q_update) {
FILE *vpxlog = 0;
#endif
static void set_offsets(VP9D_COMP *pbi, int block_size,
int mb_row, int mb_col) {
VP9_COMMON *const cm = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
const int mis = cm->mode_info_stride;
const int idx = mis * mb_row + mb_col;
const int dst_fb_idx = cm->new_fb_idx;
const int recon_y_stride = cm->yv12_fb[dst_fb_idx].y_stride;
const int recon_uv_stride = cm->yv12_fb[dst_fb_idx].uv_stride;
const int recon_yoffset = mb_row * 16 * recon_y_stride + 16 * mb_col;
const int recon_uvoffset = mb_row * 8 * recon_uv_stride + 8 * mb_col;
xd->mode_info_context = cm->mi + idx;
xd->mode_info_context->mbmi.sb_type = block_size >> 5;
xd->prev_mode_info_context = cm->prev_mi + idx;
xd->above_context = cm->above_context + mb_col;
xd->left_context = cm->left_context + (mb_row & 3);
/* Distance of Mb to the various image edges.
* These are specified to 8th pel as they are always compared to
* values that are in 1/8th pel units
*/
block_size >>= 4; // in mb units
xd->mb_to_top_edge = -((mb_row * 16)) << 3;
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_bottom_edge = ((cm->mb_rows - block_size - mb_row) * 16) << 3;
xd->mb_to_right_edge = ((cm->mb_cols - block_size - mb_col) * 16) << 3;
xd->up_available = (mb_row != 0);
xd->left_available = (mb_col != 0);
xd->dst.y_buffer = cm->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
xd->dst.u_buffer = cm->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
xd->dst.v_buffer = cm->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
}
static void set_refs(VP9D_COMP *pbi, int block_size,
int mb_row, int mb_col) {
VP9_COMMON *const cm = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
MODE_INFO *mi = xd->mode_info_context;
MB_MODE_INFO *const mbmi = &mi->mbmi;
if (mbmi->ref_frame > INTRA_FRAME) {
int ref_fb_idx, ref_yoffset, ref_uvoffset, ref_y_stride, ref_uv_stride;
/* Select the appropriate reference frame for this MB */
if (mbmi->ref_frame == LAST_FRAME)
ref_fb_idx = cm->lst_fb_idx;
else if (mbmi->ref_frame == GOLDEN_FRAME)
ref_fb_idx = cm->gld_fb_idx;
else
ref_fb_idx = cm->alt_fb_idx;
ref_y_stride = cm->yv12_fb[ref_fb_idx].y_stride;
ref_yoffset = mb_row * 16 * ref_y_stride + 16 * mb_col;
xd->pre.y_buffer = cm->yv12_fb[ref_fb_idx].y_buffer + ref_yoffset;
ref_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
ref_uvoffset = mb_row * 8 * ref_uv_stride + 8 * mb_col;
xd->pre.u_buffer = cm->yv12_fb[ref_fb_idx].u_buffer + ref_uvoffset;
xd->pre.v_buffer = cm->yv12_fb[ref_fb_idx].v_buffer + ref_uvoffset;
/* propagate errors from reference frames */
xd->corrupted |= cm->yv12_fb[ref_fb_idx].corrupted;
if (mbmi->second_ref_frame > INTRA_FRAME) {
int second_ref_fb_idx;
/* Select the appropriate reference frame for this MB */
if (mbmi->second_ref_frame == LAST_FRAME)
second_ref_fb_idx = cm->lst_fb_idx;
else if (mbmi->second_ref_frame == GOLDEN_FRAME)
second_ref_fb_idx = cm->gld_fb_idx;
else
second_ref_fb_idx = cm->alt_fb_idx;
xd->second_pre.y_buffer =
cm->yv12_fb[second_ref_fb_idx].y_buffer + ref_yoffset;
xd->second_pre.u_buffer =
cm->yv12_fb[second_ref_fb_idx].u_buffer + ref_uvoffset;
xd->second_pre.v_buffer =
cm->yv12_fb[second_ref_fb_idx].v_buffer + ref_uvoffset;
/* propagate errors from reference frames */
xd->corrupted |= cm->yv12_fb[second_ref_fb_idx].corrupted;
}
}
if (mbmi->sb_type) {
const int n_mbs = 1 << mbmi->sb_type;
const int y_mbs = MIN(n_mbs, cm->mb_rows - mb_row);
const int x_mbs = MIN(n_mbs, cm->mb_cols - mb_col);
const int mis = cm->mode_info_stride;
int x, y;
for (y = 0; y < y_mbs; y++) {
for (x = !y; x < x_mbs; x++) {
mi[y * mis + x] = *mi;
}
}
}
}
/* Decode a row of Superblocks (2x2 region of MBs) */
static void
decode_sb_row(VP9D_COMP *pbi, VP9_COMMON *pc, int mbrow, MACROBLOCKD *xd,
BOOL_DECODER* const bc) {
int i;
int sb_col;
int mb_row, mb_col;
int recon_yoffset, recon_uvoffset;
int ref_fb_idx = pc->lst_fb_idx;
int dst_fb_idx = pc->new_fb_idx;
int recon_y_stride = pc->yv12_fb[ref_fb_idx].y_stride;
int recon_uv_stride = pc->yv12_fb[ref_fb_idx].uv_stride;
int row_delta[4] = { 0, +1, 0, -1};
int col_delta[4] = { +1, -1, +1, +1};
int sb_cols = (pc->mb_cols + 1) >> 1;
static void decode_sb_row(VP9D_COMP *pbi, VP9_COMMON *pc,
int mb_row, MACROBLOCKD *xd,
BOOL_DECODER* const bc) {
int mb_col;
// For a SB there are 2 left contexts, each pertaining to a MB row within
vpx_memset(pc->left_context, 0, sizeof(pc->left_context));
mb_row = mbrow;
mb_col = 0;
for (sb_col = 0; sb_col < sb_cols; sb_col++) {
#if CONFIG_SUPERBLOCKS
MODE_INFO *mi = xd->mode_info_context;
mi->mbmi.encoded_as_sb = vp9_read(bc, pc->sb_coded);
#endif
// Process the 4 MBs within the SB in the order:
// top-left, top-right, bottom-left, bottom-right
for (i = 0; i < 4; i++) {
int dy = row_delta[i];
int dx = col_delta[i];
int offset_extended = dy * xd->mode_info_stride + dx;
xd->mb_index = i;
#if CONFIG_SUPERBLOCKS
mi = xd->mode_info_context;
#endif
if ((mb_row >= pc->mb_rows) || (mb_col >= pc->mb_cols)) {
// MB lies outside frame, skip on to next
mb_row += dy;
mb_col += dx;
xd->mode_info_context += offset_extended;
xd->prev_mode_info_context += offset_extended;
continue;
}
#if CONFIG_SUPERBLOCKS
if (i)
mi->mbmi.encoded_as_sb = 0;
#endif
// Set above context pointer
xd->above_context = pc->above_context + mb_col;
xd->left_context = pc->left_context + (i >> 1);
/* Distance of Mb to the various image edges.
* These are specified to 8th pel as they are always compared to
* values that are in 1/8th pel units
*/
xd->mb_to_top_edge = -((mb_row * 16)) << 3;
xd->mb_to_left_edge = -((mb_col * 16) << 3);
#if CONFIG_SUPERBLOCKS
if (mi->mbmi.encoded_as_sb) {
xd->mb_to_bottom_edge = ((pc->mb_rows - 2 - mb_row) * 16) << 3;
xd->mb_to_right_edge = ((pc->mb_cols - 2 - mb_col) * 16) << 3;
} else {
#endif
xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3;
xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;
#if CONFIG_SUPERBLOCKS
}
#endif
#ifdef DEC_DEBUG
dec_debug = (pbi->common.current_video_frame == 1 &&
mb_row == 2 && mb_col == 8);
if (dec_debug)
#if CONFIG_SUPERBLOCKS
printf("Enter Debug %d %d sb %d\n", mb_row, mb_col,
mi->mbmi.encoded_as_sb);
#else
printf("Enter Debug %d %d\n", mb_row, mb_col);
#endif
#endif
xd->up_available = (mb_row != 0);
xd->left_available = (mb_col != 0);
recon_yoffset = (mb_row * recon_y_stride * 16) + (mb_col * 16);
recon_uvoffset = (mb_row * recon_uv_stride * 8) + (mb_col * 8);
xd->dst.y_buffer = pc->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
xd->dst.u_buffer = pc->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
xd->dst.v_buffer = pc->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
for (mb_col = 0; mb_col < pc->mb_cols; mb_col += 4) {
if (vp9_read(bc, pc->sb64_coded)) {
set_offsets(pbi, 64, mb_row, mb_col);
vp9_decode_mb_mode_mv(pbi, xd, mb_row, mb_col, bc);
set_refs(pbi, 64, mb_row, mb_col);
decode_superblock64(pbi, xd, mb_row, mb_col, bc);
xd->corrupted |= bool_error(bc);
} else {
int j;
update_blockd_bmi(xd);
#ifdef DEC_DEBUG
if (dec_debug)
printf("Hello\n");
#endif
for (j = 0; j < 4; j++) {
const int x_idx_sb = (j & 1) << 1, y_idx_sb = j & 2;
/* Select the appropriate reference frame for this MB */
if (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME)
ref_fb_idx = pc->lst_fb_idx;
else if (xd->mode_info_context->mbmi.ref_frame == GOLDEN_FRAME)
ref_fb_idx = pc->gld_fb_idx;
else
ref_fb_idx = pc->alt_fb_idx;
if (mb_row + y_idx_sb >= pc->mb_rows ||
mb_col + x_idx_sb >= pc->mb_cols) {
// MB lies outside frame, skip on to next
continue;
}
xd->pre.y_buffer = pc->yv12_fb[ref_fb_idx].y_buffer + recon_yoffset;
xd->pre.u_buffer = pc->yv12_fb[ref_fb_idx].u_buffer + recon_uvoffset;
xd->pre.v_buffer = pc->yv12_fb[ref_fb_idx].v_buffer + recon_uvoffset;
xd->sb_index = j;
if (xd->mode_info_context->mbmi.second_ref_frame > 0) {
int second_ref_fb_idx;
if (vp9_read(bc, pc->sb32_coded)) {
set_offsets(pbi, 32, mb_row + y_idx_sb, mb_col + x_idx_sb);
vp9_decode_mb_mode_mv(pbi,
xd, mb_row + y_idx_sb, mb_col + x_idx_sb, bc);
set_refs(pbi, 32, mb_row + y_idx_sb, mb_col + x_idx_sb);
decode_superblock32(pbi,
xd, mb_row + y_idx_sb, mb_col + x_idx_sb, bc);
xd->corrupted |= bool_error(bc);
} else {
int i;
/* Select the appropriate reference frame for this MB */
if (xd->mode_info_context->mbmi.second_ref_frame == LAST_FRAME)
second_ref_fb_idx = pc->lst_fb_idx;
else if (xd->mode_info_context->mbmi.second_ref_frame ==
GOLDEN_FRAME)
second_ref_fb_idx = pc->gld_fb_idx;
else
second_ref_fb_idx = pc->alt_fb_idx;
// Process the 4 MBs within the SB in the order:
// top-left, top-right, bottom-left, bottom-right
for (i = 0; i < 4; i++) {
const int x_idx = x_idx_sb + (i & 1), y_idx = y_idx_sb + (i >> 1);
xd->second_pre.y_buffer =
pc->yv12_fb[second_ref_fb_idx].y_buffer + recon_yoffset;
xd->second_pre.u_buffer =
pc->yv12_fb[second_ref_fb_idx].u_buffer + recon_uvoffset;
xd->second_pre.v_buffer =
pc->yv12_fb[second_ref_fb_idx].v_buffer + recon_uvoffset;
}
if (mb_row + y_idx >= pc->mb_rows ||
mb_col + x_idx >= pc->mb_cols) {
// MB lies outside frame, skip on to next
continue;
}
if (xd->mode_info_context->mbmi.ref_frame != INTRA_FRAME) {
/* propagate errors from reference frames */
xd->corrupted |= pc->yv12_fb[ref_fb_idx].corrupted;
}
set_offsets(pbi, 16, mb_row + y_idx, mb_col + x_idx);
xd->mb_index = i;
vp9_decode_mb_mode_mv(pbi, xd, mb_row + y_idx, mb_col + x_idx, bc);
update_blockd_bmi(xd);
set_refs(pbi, 16, mb_row + y_idx, mb_col + x_idx);
vp9_intra_prediction_down_copy(xd);
decode_macroblock(pbi, xd, mb_row, mb_col, bc);
#if CONFIG_SUPERBLOCKS
if (xd->mode_info_context->mbmi.encoded_as_sb) {
if (mb_col < pc->mb_cols - 1)
mi[1] = mi[0];
if (mb_row < pc->mb_rows - 1) {
mi[pc->mode_info_stride] = mi[0];
if (mb_col < pc->mb_cols - 1)
mi[pc->mode_info_stride + 1] = mi[0];
/* check if the boolean decoder has suffered an error */
xd->corrupted |= bool_error(bc);
}
}
}
if (xd->mode_info_context->mbmi.encoded_as_sb) {
decode_superblock(pbi, xd, mb_row, mb_col, bc);
} else {
#endif
vp9_intra_prediction_down_copy(xd);
decode_macroblock(pbi, xd, mb_row, mb_col, bc);
#if CONFIG_SUPERBLOCKS
}
#endif
/* check if the boolean decoder has suffered an error */
xd->corrupted |= bool_error(bc);
#if CONFIG_SUPERBLOCKS
if (mi->mbmi.encoded_as_sb) {
assert(!i);
mb_col += 2;
xd->mode_info_context += 2;
xd->prev_mode_info_context += 2;
break;
}
#endif
// skip to next MB
xd->mode_info_context += offset_extended;
xd->prev_mode_info_context += offset_extended;
mb_row += dy;
mb_col += dx;
}
}
/* skip prediction column */
xd->mode_info_context += 1 - (pc->mb_cols & 0x1) + xd->mode_info_stride;
xd->prev_mode_info_context += 1 - (pc->mb_cols & 0x1) + xd->mode_info_stride;
}
static unsigned int read_partition_size(const unsigned char *cx_size) {
@ -1212,14 +1336,13 @@ static void init_frame(VP9D_COMP *pbi) {
}
static void read_coef_probs_common(
BOOL_DECODER* const bc,
vp9_prob coef_probs[BLOCK_TYPES][COEF_BANDS]
[PREV_COEF_CONTEXTS][ENTROPY_NODES]) {
static void read_coef_probs_common(BOOL_DECODER* const bc,
vp9_coeff_probs *coef_probs,
int block_types) {
int i, j, k, l;
if (vp9_read_bit(bc)) {
for (i = 0; i < BLOCK_TYPES; i++) {
for (i = 0; i < block_types; i++) {
for (j = !i; j < COEF_BANDS; j++) {
/* NB: This j loop starts from 1 on block type i == 0 */
for (k = 0; k < PREV_COEF_CONTEXTS; k++) {
@ -1242,16 +1365,20 @@ static void read_coef_probs_common(
static void read_coef_probs(VP9D_COMP *pbi, BOOL_DECODER* const bc) {
VP9_COMMON *const pc = &pbi->common;
read_coef_probs_common(bc, pc->fc.coef_probs);
read_coef_probs_common(bc, pc->fc.hybrid_coef_probs);
read_coef_probs_common(bc, pc->fc.coef_probs_4x4, BLOCK_TYPES_4X4);
read_coef_probs_common(bc, pc->fc.hybrid_coef_probs_4x4, BLOCK_TYPES_4X4);
if (pbi->common.txfm_mode != ONLY_4X4) {
read_coef_probs_common(bc, pc->fc.coef_probs_8x8);
read_coef_probs_common(bc, pc->fc.hybrid_coef_probs_8x8);
read_coef_probs_common(bc, pc->fc.coef_probs_8x8, BLOCK_TYPES_8X8);
read_coef_probs_common(bc, pc->fc.hybrid_coef_probs_8x8, BLOCK_TYPES_8X8);
}
if (pbi->common.txfm_mode > ALLOW_8X8) {
read_coef_probs_common(bc, pc->fc.coef_probs_16x16);
read_coef_probs_common(bc, pc->fc.hybrid_coef_probs_16x16);
read_coef_probs_common(bc, pc->fc.coef_probs_16x16, BLOCK_TYPES_16X16);
read_coef_probs_common(bc, pc->fc.hybrid_coef_probs_16x16,
BLOCK_TYPES_16X16);
}
if (pbi->common.txfm_mode > ALLOW_16X16) {
read_coef_probs_common(bc, pc->fc.coef_probs_32x32, BLOCK_TYPES_32X32);
}
}
@ -1437,15 +1564,17 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
}
}
#if CONFIG_SUPERBLOCKS
pc->sb_coded = vp9_read_literal(&header_bc, 8);
#endif
pc->sb64_coded = vp9_read_literal(&header_bc, 8);
pc->sb32_coded = vp9_read_literal(&header_bc, 8);
/* Read the loop filter level and type */
pc->txfm_mode = vp9_read_literal(&header_bc, 2);
if (pc->txfm_mode == 3)
pc->txfm_mode += vp9_read_bit(&header_bc);
if (pc->txfm_mode == TX_MODE_SELECT) {
pc->prob_tx[0] = vp9_read_literal(&header_bc, 8);
pc->prob_tx[1] = vp9_read_literal(&header_bc, 8);
pc->prob_tx[2] = vp9_read_literal(&header_bc, 8);
}
pc->filter_type = (LOOPFILTERTYPE) vp9_read_bit(&header_bc);
@ -1577,6 +1706,33 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
}
}
#if CONFIG_NEW_MVREF
// If Key frame reset mv ref id probabilities to defaults
if (pc->frame_type == KEY_FRAME) {
// 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));
} else {
// Read any mv_ref index probability updates
int i, j;
for (i = 0; i < MAX_REF_FRAMES; ++i) {
// Skip the dummy entry for intra ref frame.
if (i == INTRA_FRAME) {
continue;
}
// Read any updates to probabilities
for (j = 0; j < MAX_MV_REF_CANDIDATES - 1; ++j) {
if (vp9_read(&header_bc, VP9_MVREF_UPDATE_PROB)) {
xd->mb_mv_ref_probs[i][j] =
(vp9_prob)vp9_read_literal(&header_bc, 8);
}
}
}
}
#endif
if (0) {
FILE *z = fopen("decodestats.stt", "a");
fprintf(z, "%6d F:%d,G:%d,A:%d,L:%d,Q:%d\n",
@ -1589,10 +1745,10 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
fclose(z);
}
vp9_copy(pbi->common.fc.pre_coef_probs,
pbi->common.fc.coef_probs);
vp9_copy(pbi->common.fc.pre_hybrid_coef_probs,
pbi->common.fc.hybrid_coef_probs);
vp9_copy(pbi->common.fc.pre_coef_probs_4x4,
pbi->common.fc.coef_probs_4x4);
vp9_copy(pbi->common.fc.pre_hybrid_coef_probs_4x4,
pbi->common.fc.hybrid_coef_probs_4x4);
vp9_copy(pbi->common.fc.pre_coef_probs_8x8,
pbi->common.fc.coef_probs_8x8);
vp9_copy(pbi->common.fc.pre_hybrid_coef_probs_8x8,
@ -1601,10 +1757,10 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
pbi->common.fc.coef_probs_16x16);
vp9_copy(pbi->common.fc.pre_hybrid_coef_probs_16x16,
pbi->common.fc.hybrid_coef_probs_16x16);
vp9_copy(pbi->common.fc.pre_coef_probs_32x32,
pbi->common.fc.coef_probs_32x32);
vp9_copy(pbi->common.fc.pre_ymode_prob, pbi->common.fc.ymode_prob);
#if CONFIG_SUPERBLOCKS
vp9_copy(pbi->common.fc.pre_sb_ymode_prob, pbi->common.fc.sb_ymode_prob);
#endif
vp9_copy(pbi->common.fc.pre_uv_mode_prob, pbi->common.fc.uv_mode_prob);
vp9_copy(pbi->common.fc.pre_bmode_prob, pbi->common.fc.bmode_prob);
vp9_copy(pbi->common.fc.pre_i8x8_mode_prob, pbi->common.fc.i8x8_mode_prob);
@ -1614,16 +1770,15 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
pbi->common.fc.pre_interintra_prob = pbi->common.fc.interintra_prob;
#endif
pbi->common.fc.pre_nmvc = pbi->common.fc.nmvc;
vp9_zero(pbi->common.fc.coef_counts);
vp9_zero(pbi->common.fc.hybrid_coef_counts);
vp9_zero(pbi->common.fc.coef_counts_4x4);
vp9_zero(pbi->common.fc.hybrid_coef_counts_4x4);
vp9_zero(pbi->common.fc.coef_counts_8x8);
vp9_zero(pbi->common.fc.hybrid_coef_counts_8x8);
vp9_zero(pbi->common.fc.coef_counts_16x16);
vp9_zero(pbi->common.fc.hybrid_coef_counts_16x16);
vp9_zero(pbi->common.fc.coef_counts_32x32);
vp9_zero(pbi->common.fc.ymode_counts);
#if CONFIG_SUPERBLOCKS
vp9_zero(pbi->common.fc.sb_ymode_counts);
#endif
vp9_zero(pbi->common.fc.uv_mode_counts);
vp9_zero(pbi->common.fc.bmode_counts);
vp9_zero(pbi->common.fc.i8x8_mode_counts);
@ -1662,12 +1817,8 @@ int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
vpx_memset(pc->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * pc->mb_cols);
// Resset the macroblock mode info context to the start of the list
xd->mode_info_context = pc->mi;
xd->prev_mode_info_context = pc->prev_mi;
/* Decode a row of superblocks */
for (mb_row = 0; mb_row < pc->mb_rows; mb_row += 2) {
for (mb_row = 0; mb_row < pc->mb_rows; mb_row += 4) {
decode_sb_row(pbi, pc, mb_row, xd, &residual_bc);
}
corrupt_tokens |= xd->corrupted;

2
vp9/decoder/vp9_decodframe.h
View File

@ -16,4 +16,4 @@ struct VP9Decompressor;
extern void vp9_init_de_quantizer(struct VP9Decompressor *pbi);
#endif // __INC_DECODFRAME_H
#endif // VP9_DECODER_VP9_DECODFRAME_H_

60
vp9/decoder/vp9_dequantize.c
View File

@ -13,20 +13,14 @@
#include "vp9/decoder/vp9_dequantize.h"
#include "vpx_mem/vpx_mem.h"
#include "vp9/decoder/vp9_onyxd_int.h"
#include "vp9/common/vp9_common.h"
static void add_residual(const int16_t *diff, const uint8_t *pred, int pitch,
uint8_t *dest, int stride, int width, int height) {
int r, c;
for (r = 0; r < height; r++) {
for (c = 0; c < width; c++) {
int a = diff[c] + pred[c];
if (a < 0)
a = 0;
else if (a > 255)
a = 255;
dest[c] = (uint8_t) a;
dest[c] = clip_pixel(diff[c] + pred[c]);
}
dest += stride;
@ -42,14 +36,7 @@ static void add_constant_residual(const int16_t diff, const uint8_t *pred,
for (r = 0; r < height; r++) {
for (c = 0; c < width; c++) {
int a = diff + pred[c];
if (a < 0)
a = 0;
else if (a > 255)
a = 255;
dest[c] = (uint8_t) a;
dest[c] = clip_pixel(diff + pred[c]);
}
dest += stride;
@ -204,7 +191,7 @@ void vp9_dequantize_b_2x2_c(BLOCKD *d) {
void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq,
uint8_t *pred, uint8_t *dest, int pitch,
int stride, int dc, uint16_t eobs) {
int stride, int dc, int eob) {
int16_t output[64];
int16_t *diff_ptr = output;
int i;
@ -220,10 +207,10 @@ void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq,
* TODO(yunqingwang): "eobs = 1" case is also handled in vp9_short_idct8x8_c.
* Combine that with code here.
*/
if (eobs == 0) {
if (eob == 0) {
/* All 0 DCT coefficient */
vp9_copy_mem8x8(pred, pitch, dest, stride);
} else if (eobs == 1) {
} else if (eob == 1) {
/* DC only DCT coefficient. */
int16_t out;
@ -236,7 +223,7 @@ void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq,
input[0] = 0;
add_constant_residual(out, pred, pitch, dest, stride, 8, 8);
} else if (eobs <= 10) {
} else if (eob <= 10) {
input[1] = input[1] * dq[1];
input[2] = input[2] * dq[1];
input[3] = input[3] * dq[1];
@ -301,17 +288,17 @@ void vp9_ht_dequant_idct_add_16x16_c(TX_TYPE tx_type, int16_t *input,
void vp9_dequant_idct_add_16x16_c(int16_t *input, const int16_t *dq,
uint8_t *pred, uint8_t *dest, int pitch,
int stride, uint16_t eobs) {
int stride, int eob) {
int16_t output[256];
int16_t *diff_ptr = output;
int i;
/* The calculation can be simplified if there are not many non-zero dct
* coefficients. Use eobs to separate different cases. */
if (eobs == 0) {
if (eob == 0) {
/* All 0 DCT coefficient */
vp9_copy_mem16x16(pred, pitch, dest, stride);
} else if (eobs == 1) {
} else if (eob == 1) {
/* DC only DCT coefficient. */
int16_t out;
@ -324,7 +311,7 @@ void vp9_dequant_idct_add_16x16_c(int16_t *input, const int16_t *dq,
input[0] = 0;
add_constant_residual(out, pred, pitch, dest, stride, 16, 16);
} else if (eobs <= 10) {
} else if (eob <= 10) {
input[0]= input[0] * dq[0];
input[1] = input[1] * dq[1];
input[2] = input[2] * dq[1];
@ -360,3 +347,28 @@ void vp9_dequant_idct_add_16x16_c(int16_t *input, const int16_t *dq,
add_residual(diff_ptr, pred, pitch, dest, stride, 16, 16);
}
}
void vp9_dequant_idct_add_32x32_c(int16_t *input, const int16_t *dq,
uint8_t *pred, uint8_t *dest, int pitch,
int stride, int eob) {
int16_t output[1024];
int i;
input[0]= input[0] * dq[0] / 2;
for (i = 1; i < 1024; i++)
input[i] = input[i] * dq[1] / 2;
vp9_short_idct32x32_c(input, output, 64);
vpx_memset(input, 0, 2048);
add_residual(output, pred, pitch, dest, stride, 32, 32);
}
void vp9_dequant_idct_add_uv_block_16x16_c(int16_t *q, const int16_t *dq,
uint8_t *dstu,
uint8_t *dstv,
int stride,
uint16_t *eobs) {
vp9_dequant_idct_add_16x16_c(q, dq, dstu, dstu, stride, stride, eobs[0]);
vp9_dequant_idct_add_16x16_c(q + 256, dq,
dstv, dstv, stride, stride, eobs[4]);
}

70
vp9/decoder/vp9_dequantize.h
View File

@ -14,90 +14,88 @@
#include "vp9/common/vp9_blockd.h"
#if CONFIG_LOSSLESS
extern void vp9_dequant_idct_add_lossless_c(short *input, const short *dq,
extern void vp9_dequant_idct_add_lossless_c(int16_t *input, const int16_t *dq,
unsigned char *pred,
unsigned char *output,
int pitch, int stride);
extern void vp9_dequant_dc_idct_add_lossless_c(short *input, const short *dq,
extern void vp9_dequant_dc_idct_add_lossless_c(int16_t *input, const int16_t *dq,
unsigned char *pred,
unsigned char *output,
int pitch, int stride, int dc);
extern void vp9_dequant_dc_idct_add_y_block_lossless_c(short *q,
const short *dq,
extern void vp9_dequant_dc_idct_add_y_block_lossless_c(int16_t *q,
const int16_t *dq,
unsigned char *pre,
unsigned char *dst,
int stride,
unsigned short *eobs,
const short *dc);
extern void vp9_dequant_idct_add_y_block_lossless_c(short *q, const short *dq,
uint16_t *eobs,
const int16_t *dc);
extern void vp9_dequant_idct_add_y_block_lossless_c(int16_t *q, const int16_t *dq,
unsigned char *pre,
unsigned char *dst,
int stride,
unsigned short *eobs);
extern void vp9_dequant_idct_add_uv_block_lossless_c(short *q, const short *dq,
uint16_t *eobs);
extern void vp9_dequant_idct_add_uv_block_lossless_c(int16_t *q, const int16_t *dq,
unsigned char *pre,
unsigned char *dst_u,
unsigned char *dst_v,
int stride,
unsigned short *eobs);
uint16_t *eobs);
#endif
typedef void (*vp9_dequant_idct_add_fn_t)(short *input, const short *dq,
typedef void (*vp9_dequant_idct_add_fn_t)(int16_t *input, const int16_t *dq,
unsigned char *pred, unsigned char *output, int pitch, int stride);
typedef void(*vp9_dequant_dc_idct_add_fn_t)(short *input, const short *dq,
typedef void(*vp9_dequant_dc_idct_add_fn_t)(int16_t *input, const int16_t *dq,
unsigned char *pred, unsigned char *output, int pitch, int stride, int dc);
typedef void(*vp9_dequant_dc_idct_add_y_block_fn_t)(short *q, const short *dq,
unsigned char *pre, unsigned char *dst, int stride, unsigned short *eobs,
const short *dc);
typedef void(*vp9_dequant_idct_add_y_block_fn_t)(short *q, const short *dq,
unsigned char *pre, unsigned char *dst, int stride, unsigned short *eobs);
typedef void(*vp9_dequant_idct_add_uv_block_fn_t)(short *q, const short *dq,
typedef void(*vp9_dequant_dc_idct_add_y_block_fn_t)(int16_t *q, const int16_t *dq,
unsigned char *pre, unsigned char *dst, int stride, uint16_t *eobs,
const int16_t *dc);
typedef void(*vp9_dequant_idct_add_y_block_fn_t)(int16_t *q, const int16_t *dq,
unsigned char *pre, unsigned char *dst, int stride, uint16_t *eobs);
typedef void(*vp9_dequant_idct_add_uv_block_fn_t)(int16_t *q, const int16_t *dq,
unsigned char *pre, unsigned char *dst_u, unsigned char *dst_v, int stride,
unsigned short *eobs);
uint16_t *eobs);
void vp9_ht_dequant_idct_add_c(TX_TYPE tx_type, short *input, const short *dq,
void vp9_ht_dequant_idct_add_c(TX_TYPE tx_type, int16_t *input, const int16_t *dq,
unsigned char *pred, unsigned char *dest,
int pitch, int stride, uint16_t eobs);
void vp9_ht_dequant_idct_add_8x8_c(TX_TYPE tx_type, short *input,
const short *dq, unsigned char *pred,
void vp9_ht_dequant_idct_add_8x8_c(TX_TYPE tx_type, int16_t *input,
const int16_t *dq, unsigned char *pred,
unsigned char *dest, int pitch, int stride,
uint16_t eobs);
void vp9_ht_dequant_idct_add_16x16_c(TX_TYPE tx_type, short *input,
const short *dq, unsigned char *pred,
void vp9_ht_dequant_idct_add_16x16_c(TX_TYPE tx_type, int16_t *input,
const int16_t *dq, unsigned char *pred,
unsigned char *dest,
int pitch, int stride, uint16_t eobs);
#if CONFIG_SUPERBLOCKS
void vp9_dequant_dc_idct_add_y_block_8x8_inplace_c(short *q, const short *dq,
void vp9_dequant_dc_idct_add_y_block_8x8_inplace_c(int16_t *q, const int16_t *dq,
unsigned char *dst,
int stride,
unsigned short *eobs,
const short *dc,
uint16_t *eobs,
const int16_t *dc,
MACROBLOCKD *xd);
void vp9_dequant_dc_idct_add_y_block_4x4_inplace_c(short *q, const short *dq,
void vp9_dequant_dc_idct_add_y_block_4x4_inplace_c(int16_t *q, const int16_t *dq,
unsigned char *dst,
int stride,
unsigned short *eobs,
const short *dc,
uint16_t *eobs,
const int16_t *dc,
MACROBLOCKD *xd);
void vp9_dequant_idct_add_uv_block_8x8_inplace_c(short *q, const short *dq,
void vp9_dequant_idct_add_uv_block_8x8_inplace_c(int16_t *q, const int16_t *dq,
unsigned char *dstu,
unsigned char *dstv,
int stride,
unsigned short *eobs,
uint16_t *eobs,
MACROBLOCKD *xd);
void vp9_dequant_idct_add_uv_block_4x4_inplace_c(short *q, const short *dq,
void vp9_dequant_idct_add_uv_block_4x4_inplace_c(int16_t *q, const int16_t *dq,
unsigned char *dstu,
unsigned char *dstv,
int stride,
unsigned short *eobs,
uint16_t *eobs,
MACROBLOCKD *xd);
#endif
#endif

391
vp9/decoder/vp9_detokenize.c
View File

@ -9,13 +9,11 @@
*/
#include "vp9/common/vp9_type_aliases.h"
#include "vp9/common/vp9_blockd.h"
#include "vp9/decoder/vp9_onyxd_int.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"
#include "vp9/decoder/vp9_detokenize.h"
#include "vp9/common/vp9_seg_common.h"
#define EOB_CONTEXT_NODE 0
@ -55,59 +53,38 @@
#define CAT5_PROB3 157
#define CAT5_PROB4 180
static const unsigned char cat6_prob[14] =
{ 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129, 0 };
static const vp9_prob cat6_prob[15] = {
254, 254, 254, 252, 249, 243, 230, 196, 177, 153, 140, 133, 130, 129, 0
};
void vp9_reset_mb_tokens_context(MACROBLOCKD* const xd) {
/* Clear entropy contexts */
if ((xd->mode_info_context->mbmi.mode != B_PRED &&
xd->mode_info_context->mbmi.mode != I8X8_PRED &&
xd->mode_info_context->mbmi.mode != SPLITMV)
|| xd->mode_info_context->mbmi.txfm_size == TX_16X16) {
vpx_memset(xd->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
vpx_memset(xd->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
} else {
vpx_memset(xd->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) - 1);
vpx_memset(xd->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) - 1);
xd->above_context->y2 = 1;
xd->left_context->y2 = 1;
}
}
DECLARE_ALIGNED(16, extern const unsigned char, vp9_norm[256]);
DECLARE_ALIGNED(16, extern const uint8_t, vp9_norm[256]);
static int get_signed(BOOL_DECODER *br, int value_to_sign) {
const int split = (br->range + 1) >> 1;
const VP9_BD_VALUE bigsplit = (VP9_BD_VALUE)split << (VP9_BD_VALUE_SIZE - 8);
int v;
if (br->count < 0)
vp9_bool_decoder_fill(br);
if (br->value < bigsplit) {
br->range = split;
v = value_to_sign;
} else {
br->range = br->range - split;
br->value = br->value - bigsplit;
v = -value_to_sign;
}
br->range += br->range;
br->value += br->value;
--br->count;
return v;
return decode_bool(br, 128) ? -value_to_sign : value_to_sign;
}
#if CONFIG_NEWCOEFCONTEXT
#define PT pn
#define INCREMENT_COUNT(token) \
do { \
coef_counts[type][coef_bands[c]][pn][token]++; \
pn = pt = vp9_prev_token_class[token]; \
if (c < seg_eob - 1 && NEWCOEFCONTEXT_BAND_COND(coef_bands[c + 1])) \
pn = vp9_get_coef_neighbor_context( \
qcoeff_ptr, nodc, neighbors, scan[c + 1]); \
} while (0)
#else
#define PT pt
#define INCREMENT_COUNT(token) \
do { \
coef_counts[coef_bands[c]][pt][token]++; \
pt = vp9_prev_token_class[token]; \
coef_counts[type][coef_bands[c]][pt][token]++; \
pt = vp9_prev_token_class[token]; \
} while (0)
#endif /* CONFIG_NEWCOEFCONTEXT */
#define WRITE_COEF_CONTINUE(val, token) \
{ \
qcoeff_ptr[scan[c]] = (INT16) get_signed(br, val); \
qcoeff_ptr[scan[c]] = (int16_t) get_signed(br, val); \
INCREMENT_COUNT(token); \
c++; \
continue; \
@ -116,7 +93,7 @@ static int get_signed(BOOL_DECODER *br, int value_to_sign) {
#define ADJUST_COEF(prob, bits_count) \
do { \
if (vp9_read(br, prob)) \
val += (UINT16)(1 << bits_count);\
val += (uint16_t)(1 << bits_count);\
} while (0);
static int decode_coefs(VP9D_COMP *dx, const MACROBLOCKD *xd,
@ -124,51 +101,65 @@ static int decode_coefs(VP9D_COMP *dx, const MACROBLOCKD *xd,
ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l,
PLANE_TYPE type,
TX_TYPE tx_type,
int seg_eob, INT16 *qcoeff_ptr,
int seg_eob, int16_t *qcoeff_ptr,
const int *const scan, TX_SIZE txfm_size,
const int *coef_bands) {
FRAME_CONTEXT *const fc = &dx->common.fc;
int pt, c = (type == PLANE_TYPE_Y_NO_DC);
vp9_prob (*coef_probs)[PREV_COEF_CONTEXTS][ENTROPY_NODES], *prob;
unsigned int (*coef_counts)[PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS];
#if CONFIG_NEWCOEFCONTEXT
const int *neighbors;
int pn;
#endif
int nodc = (type == PLANE_TYPE_Y_NO_DC);
int pt, c = nodc;
vp9_coeff_probs *coef_probs;
vp9_prob *prob;
vp9_coeff_count *coef_counts;
switch (txfm_size) {
default:
case TX_4X4:
if (tx_type == DCT_DCT) {
coef_probs = fc->coef_probs[type];
coef_counts = fc->coef_counts[type];
coef_probs = fc->coef_probs_4x4;
coef_counts = fc->coef_counts_4x4;
} else {
coef_probs = fc->hybrid_coef_probs[type];
coef_counts = fc->hybrid_coef_counts[type];
coef_probs = fc->hybrid_coef_probs_4x4;
coef_counts = fc->hybrid_coef_counts_4x4;
}
break;
case TX_8X8:
if (tx_type == DCT_DCT) {
coef_probs = fc->coef_probs_8x8[type];
coef_counts = fc->coef_counts_8x8[type];
coef_probs = fc->coef_probs_8x8;
coef_counts = fc->coef_counts_8x8;
} else {
coef_probs = fc->hybrid_coef_probs_8x8[type];
coef_counts = fc->hybrid_coef_counts_8x8[type];
coef_probs = fc->hybrid_coef_probs_8x8;
coef_counts = fc->hybrid_coef_counts_8x8;
}
break;
case TX_16X16:
if (tx_type == DCT_DCT) {
coef_probs = fc->coef_probs_16x16[type];
coef_counts = fc->coef_counts_16x16[type];
coef_probs = fc->coef_probs_16x16;
coef_counts = fc->coef_counts_16x16;
} else {
coef_probs = fc->hybrid_coef_probs_16x16[type];
coef_counts = fc->hybrid_coef_counts_16x16[type];
coef_probs = fc->hybrid_coef_probs_16x16;
coef_counts = fc->hybrid_coef_counts_16x16;
}
break;
case TX_32X32:
coef_probs = fc->coef_probs_32x32;
coef_counts = fc->coef_counts_32x32;
break;
}
VP9_COMBINEENTROPYCONTEXTS(pt, *a, *l);
#if CONFIG_NEWCOEFCONTEXT
pn = pt;
neighbors = vp9_get_coef_neighbors_handle(scan);
#endif
while (1) {
int val;
const uint8_t *cat6 = cat6_prob;
if (c >= seg_eob) break;
prob = coef_probs[coef_bands[c]][pt];
prob = coef_probs[type][coef_bands[c]][PT];
if (!vp9_read(br, prob[EOB_CONTEXT_NODE]))
break;
SKIP_START:
@ -176,7 +167,7 @@ SKIP_START:
if (!vp9_read(br, prob[ZERO_CONTEXT_NODE])) {
INCREMENT_COUNT(ZERO_TOKEN);
++c;
prob = coef_probs[coef_bands[c]][pt];
prob = coef_probs[type][coef_bands[c]][PT];
goto SKIP_START;
}
// ONE_CONTEXT_NODE_0_
@ -240,7 +231,7 @@ SKIP_START:
}
if (c < seg_eob)
coef_counts[coef_bands[c]][pt][DCT_EOB_TOKEN]++;
coef_counts[type][coef_bands[c]][PT][DCT_EOB_TOKEN]++;
a[0] = l[0] = (c > !type);
@ -256,38 +247,120 @@ static int get_eob(MACROBLOCKD* const xd, int segment_id, int eob_max) {
return eob;
}
int vp9_decode_sb_tokens(VP9D_COMP* const pbi,
MACROBLOCKD* const xd,
BOOL_DECODER* const bc) {
ENTROPY_CONTEXT* const A = (ENTROPY_CONTEXT *)xd->above_context;
ENTROPY_CONTEXT* const L = (ENTROPY_CONTEXT *)xd->left_context;
ENTROPY_CONTEXT* const A1 = (ENTROPY_CONTEXT *)(&xd->above_context[1]);
ENTROPY_CONTEXT* const L1 = (ENTROPY_CONTEXT *)(&xd->left_context[1]);
uint16_t *const eobs = xd->eobs;
const int segment_id = xd->mode_info_context->mbmi.segment_id;
int c, i, eobtotal = 0, seg_eob;
// Luma block
#if CONFIG_CNVCONTEXT
ENTROPY_CONTEXT above_ec = (A[0] + A[1] + A[2] + A[3] +
A1[0] + A1[1] + A1[2] + A1[3]) != 0;
ENTROPY_CONTEXT left_ec = (L[0] + L[1] + L[2] + L[3] +
L1[0] + L1[1] + L1[2] + L1[3]) != 0;
#else
ENTROPY_CONTEXT above_ec = A[0];
ENTROPY_CONTEXT left_ec = L[0];
#endif
eobs[0] = c = decode_coefs(pbi, xd, bc, &above_ec, &left_ec,
PLANE_TYPE_Y_WITH_DC,
DCT_DCT, get_eob(xd, segment_id, 1024),
xd->sb_coeff_data.qcoeff,
vp9_default_zig_zag1d_32x32,
TX_32X32, vp9_coef_bands_32x32);
A[1] = A[2] = A[3] = A[0] = above_ec;
L[1] = L[2] = L[3] = L[0] = left_ec;
A1[1] = A1[2] = A1[3] = A1[0] = above_ec;
L1[1] = L1[2] = L1[3] = L1[0] = left_ec;
eobtotal += c;
// 16x16 chroma blocks
seg_eob = get_eob(xd, segment_id, 256);
for (i = 16; i < 24; i += 4) {
ENTROPY_CONTEXT* const a = A + vp9_block2above[TX_16X16][i];
ENTROPY_CONTEXT* const l = L + vp9_block2left[TX_16X16][i];
ENTROPY_CONTEXT* const a1 = A1 + vp9_block2above[TX_16X16][i];
ENTROPY_CONTEXT* const l1 = L1 + vp9_block2left[TX_16X16][i];
#if CONFIG_CNVCONTEXT
above_ec = (a[0] + a[1] + a1[0] + a1[1]) != 0;
left_ec = (l[0] + l[1] + l1[0] + l1[1]) != 0;
#else
above_ec = a[0];
left_ec = l[0];
#endif
eobs[i] = c = decode_coefs(pbi, xd, bc,
&above_ec, &left_ec,
PLANE_TYPE_UV,
DCT_DCT, seg_eob,
xd->sb_coeff_data.qcoeff + 1024 + (i - 16) * 64,
vp9_default_zig_zag1d_16x16,
TX_16X16, vp9_coef_bands_16x16);
a1[1] = a1[0] = a[1] = a[0] = above_ec;
l1[1] = l1[0] = l[1] = l[0] = left_ec;
eobtotal += c;
}
// no Y2 block
A[8] = L[8] = A1[8] = L1[8] = 0;
return eobtotal;
}
static int vp9_decode_mb_tokens_16x16(VP9D_COMP* const pbi,
MACROBLOCKD* const xd,
BOOL_DECODER* const bc) {
ENTROPY_CONTEXT* const A = (ENTROPY_CONTEXT *)xd->above_context;
ENTROPY_CONTEXT* const L = (ENTROPY_CONTEXT *)xd->left_context;
unsigned short* const eobs = xd->eobs;
uint16_t *const eobs = xd->eobs;
const int segment_id = xd->mode_info_context->mbmi.segment_id;
int c, i, eobtotal = 0, seg_eob;
// Luma block
eobs[0] = c = decode_coefs(pbi, xd, bc, A, L, PLANE_TYPE_Y_WITH_DC,
#if CONFIG_CNVCONTEXT
ENTROPY_CONTEXT above_ec = (A[0] + A[1] + A[2] + A[3]) != 0;
ENTROPY_CONTEXT left_ec = (L[0] + L[1] + L[2] + L[3]) != 0;
#else
ENTROPY_CONTEXT above_ec = A[0];
ENTROPY_CONTEXT left_ec = L[0];
#endif
eobs[0] = c = decode_coefs(pbi, xd, bc, &above_ec, &left_ec,
PLANE_TYPE_Y_WITH_DC,
get_tx_type(xd, &xd->block[0]),
get_eob(xd, segment_id, 256),
xd->qcoeff, vp9_default_zig_zag1d_16x16,
TX_16X16, vp9_coef_bands_16x16);
A[1] = A[2] = A[3] = A[0];
L[1] = L[2] = L[3] = L[0];
A[1] = A[2] = A[3] = A[0] = above_ec;
L[1] = L[2] = L[3] = L[0] = left_ec;
eobtotal += c;
// 8x8 chroma blocks
seg_eob = get_eob(xd, segment_id, 64);
for (i = 16; i < 24; i += 4) {
ENTROPY_CONTEXT* const a = A + vp9_block2above_8x8[i];
ENTROPY_CONTEXT* const l = L + vp9_block2left_8x8[i];
eobs[i] = c = decode_coefs(pbi, xd, bc, a, l, PLANE_TYPE_UV,
ENTROPY_CONTEXT* const a = A + vp9_block2above[TX_8X8][i];
ENTROPY_CONTEXT* const l = L + vp9_block2left[TX_8X8][i];
#if CONFIG_CNVCONTEXT
above_ec = (a[0] + a[1]) != 0;
left_ec = (l[0] + l[1]) != 0;
#else
above_ec = a[0];
left_ec = l[0];
#endif
eobs[i] = c = decode_coefs(pbi, xd, bc,
&above_ec, &left_ec,
PLANE_TYPE_UV,
DCT_DCT, seg_eob, xd->block[i].qcoeff,
vp9_default_zig_zag1d_8x8,
TX_8X8, vp9_coef_bands_8x8);
a[1] = a[0];
l[1] = l[0];
a[1] = a[0] = above_ec;
l[1] = l[0] = left_ec;
eobtotal += c;
}
A[8] = 0;
@ -300,7 +373,7 @@ static int vp9_decode_mb_tokens_8x8(VP9D_COMP* const pbi,
BOOL_DECODER* const bc) {
ENTROPY_CONTEXT *const A = (ENTROPY_CONTEXT *)xd->above_context;
ENTROPY_CONTEXT *const L = (ENTROPY_CONTEXT *)xd->left_context;
unsigned short *const eobs = xd->eobs;
uint16_t *const eobs = xd->eobs;
PLANE_TYPE type;
int c, i, eobtotal = 0, seg_eob;
const int segment_id = xd->mode_info_context->mbmi.segment_id;
@ -308,18 +381,19 @@ static int vp9_decode_mb_tokens_8x8(VP9D_COMP* const pbi,
int has_2nd_order = get_2nd_order_usage(xd);
// 2nd order DC block
if (has_2nd_order) {
ENTROPY_CONTEXT *const a = A + vp9_block2above_8x8[24];
ENTROPY_CONTEXT *const l = L + vp9_block2left_8x8[24];
ENTROPY_CONTEXT *const a = A + vp9_block2above[TX_8X8][24];
ENTROPY_CONTEXT *const l = L + vp9_block2left[TX_8X8][24];
eobs[24] = c = decode_coefs(pbi, xd, bc, a, l, PLANE_TYPE_Y2,
DCT_DCT, get_eob(xd, segment_id, 4),
xd->block[24].qcoeff,
vp9_default_zig_zag1d, TX_8X8, vp9_coef_bands);
vp9_default_zig_zag1d_4x4, TX_8X8,
vp9_coef_bands_4x4);
eobtotal += c - 4;
type = PLANE_TYPE_Y_NO_DC;
} else {
xd->above_context->y2 = 1;
xd->left_context->y2 = 1;
xd->above_context->y2 = 0;
xd->left_context->y2 = 0;
eobs[24] = 0;
type = PLANE_TYPE_Y_WITH_DC;
}
@ -327,17 +401,23 @@ static int vp9_decode_mb_tokens_8x8(VP9D_COMP* const pbi,
// luma blocks
seg_eob = get_eob(xd, segment_id, 64);
for (i = 0; i < 16; i += 4) {
ENTROPY_CONTEXT *const a = A + vp9_block2above_8x8[i];
ENTROPY_CONTEXT *const l = L + vp9_block2left_8x8[i];
eobs[i] = c = decode_coefs(pbi, xd, bc, a, l, type,
ENTROPY_CONTEXT *const a = A + vp9_block2above[TX_8X8][i];
ENTROPY_CONTEXT *const l = L + vp9_block2left[TX_8X8][i];
#if CONFIG_CNVCONTEXT
ENTROPY_CONTEXT above_ec = (a[0] + a[1]) != 0;
ENTROPY_CONTEXT left_ec = (l[0] + l[1]) != 0;
#else
ENTROPY_CONTEXT above_ec = a[0];
ENTROPY_CONTEXT left_ec = l[0];
#endif
eobs[i] = c = decode_coefs(pbi, xd, bc, &above_ec, &left_ec, type,
type == PLANE_TYPE_Y_WITH_DC ?
get_tx_type(xd, xd->block + i) : DCT_DCT,
seg_eob, xd->block[i].qcoeff,
vp9_default_zig_zag1d_8x8,
TX_8X8, vp9_coef_bands_8x8);
a[1] = a[0];
l[1] = l[0];
a[1] = a[0] = above_ec;
l[1] = l[0] = left_ec;
eobtotal += c;
}
@ -347,25 +427,34 @@ static int vp9_decode_mb_tokens_8x8(VP9D_COMP* const pbi,
// use 4x4 transform for U, V components in I8X8/splitmv prediction mode
seg_eob = get_eob(xd, segment_id, 16);
for (i = 16; i < 24; i++) {
ENTROPY_CONTEXT *const a = A + vp9_block2above[i];
ENTROPY_CONTEXT *const l = L + vp9_block2left[i];
ENTROPY_CONTEXT *const a = A + vp9_block2above[TX_4X4][i];
ENTROPY_CONTEXT *const l = L + vp9_block2left[TX_4X4][i];
eobs[i] = c = decode_coefs(pbi, xd, bc, a, l, PLANE_TYPE_UV,
DCT_DCT, seg_eob, xd->block[i].qcoeff,
vp9_default_zig_zag1d, TX_4X4, vp9_coef_bands);
vp9_default_zig_zag1d_4x4, TX_4X4,
vp9_coef_bands_4x4);
eobtotal += c;
}
} else {
for (i = 16; i < 24; i += 4) {
ENTROPY_CONTEXT *const a = A + vp9_block2above_8x8[i];
ENTROPY_CONTEXT *const l = L + vp9_block2left_8x8[i];
eobs[i] = c = decode_coefs(pbi, xd, bc, a, l, PLANE_TYPE_UV,
ENTROPY_CONTEXT *const a = A + vp9_block2above[TX_8X8][i];
ENTROPY_CONTEXT *const l = L + vp9_block2left[TX_8X8][i];
#if CONFIG_CNVCONTEXT
ENTROPY_CONTEXT above_ec = (a[0] + a[1]) != 0;
ENTROPY_CONTEXT left_ec = (l[0] + l[1]) != 0;
#else
ENTROPY_CONTEXT above_ec = a[0];
ENTROPY_CONTEXT left_ec = l[0];
#endif
eobs[i] = c = decode_coefs(pbi, xd, bc,
&above_ec, &left_ec,
PLANE_TYPE_UV,
DCT_DCT, seg_eob, xd->block[i].qcoeff,
vp9_default_zig_zag1d_8x8,
TX_8X8, vp9_coef_bands_8x8);
a[1] = a[0];
l[1] = l[0];
a[1] = a[0] = above_ec;
l[1] = l[0] = left_ec;
eobtotal += c;
}
}
@ -373,50 +462,77 @@ static int vp9_decode_mb_tokens_8x8(VP9D_COMP* const pbi,
return eobtotal;
}
static int decode_coefs_4x4(VP9D_COMP *dx, MACROBLOCKD *xd,
BOOL_DECODER* const bc,
PLANE_TYPE type, int i, int seg_eob,
TX_TYPE tx_type, const int *scan) {
ENTROPY_CONTEXT *const A = (ENTROPY_CONTEXT *)xd->above_context;
ENTROPY_CONTEXT *const L = (ENTROPY_CONTEXT *)xd->left_context;
ENTROPY_CONTEXT *const a = A + vp9_block2above[TX_4X4][i];
ENTROPY_CONTEXT *const l = L + vp9_block2left[TX_4X4][i];
uint16_t *const eobs = xd->eobs;
int c;
c = decode_coefs(dx, xd, bc, a, l, type, tx_type, seg_eob,
xd->block[i].qcoeff, scan, TX_4X4, vp9_coef_bands_4x4);
eobs[i] = c;
return c;
}
static int decode_coefs_4x4_y(VP9D_COMP *dx, MACROBLOCKD *xd,
BOOL_DECODER* const bc,
PLANE_TYPE type, int i, int seg_eob) {
const TX_TYPE tx_type = (type == PLANE_TYPE_Y_WITH_DC) ?
get_tx_type(xd, &xd->block[i]) : DCT_DCT;
const int *scan;
switch (tx_type) {
case ADST_DCT:
scan = vp9_row_scan_4x4;
break;
case DCT_ADST:
scan = vp9_col_scan_4x4;
break;
default:
scan = vp9_default_zig_zag1d_4x4;
break;
}
return decode_coefs_4x4(dx, xd, bc, type, i, seg_eob, tx_type, scan);
}
int vp9_decode_coefs_4x4(VP9D_COMP *dx, MACROBLOCKD *xd,
BOOL_DECODER* const bc,
PLANE_TYPE type, int i) {
ENTROPY_CONTEXT *const A = (ENTROPY_CONTEXT *)xd->above_context;
ENTROPY_CONTEXT *const L = (ENTROPY_CONTEXT *)xd->left_context;
ENTROPY_CONTEXT *const a = A + vp9_block2above[i];
ENTROPY_CONTEXT *const l = L + vp9_block2left[i];
INT16 *qcoeff_ptr = &xd->qcoeff[0];
const int *scan = vp9_default_zig_zag1d;
unsigned short *const eobs = xd->eobs;
int segment_id = xd->mode_info_context->mbmi.segment_id;
int c, seg_eob = get_eob(xd, segment_id, 16);
TX_TYPE tx_type = DCT_DCT;
const int segment_id = xd->mode_info_context->mbmi.segment_id;
const int seg_eob = get_eob(xd, segment_id, 16);
if (type == PLANE_TYPE_Y_WITH_DC)
tx_type = get_tx_type_4x4(xd, &xd->block[i]);
switch (tx_type) {
case ADST_DCT :
scan = vp9_row_scan;
break;
return decode_coefs_4x4_y(dx, xd, bc, type, i, seg_eob);
}
case DCT_ADST :
scan = vp9_col_scan;
break;
static int decode_mb_tokens_4x4_uv(VP9D_COMP* const dx,
MACROBLOCKD* const xd,
BOOL_DECODER* const bc,
int seg_eob) {
int eobtotal = 0, i;
default :
scan = vp9_default_zig_zag1d;
break;
// chroma blocks
for (i = 16; i < 24; i++) {
eobtotal += decode_coefs_4x4(dx, xd, bc, PLANE_TYPE_UV, i, seg_eob,
DCT_DCT, vp9_default_zig_zag1d_4x4);
}
eobs[i] = c = decode_coefs(dx, xd, bc, a, l, type,
tx_type, seg_eob, qcoeff_ptr + i * 16,
scan, TX_4X4, vp9_coef_bands);
return c;
return eobtotal;
}
int vp9_decode_mb_tokens_4x4_uv(VP9D_COMP* const dx,
MACROBLOCKD* const xd,
BOOL_DECODER* const bc) {
int eobtotal = 0, i;
const int segment_id = xd->mode_info_context->mbmi.segment_id;
const int seg_eob = get_eob(xd, segment_id, 16);
for (i = 16; i < 24; i++)
eobtotal += vp9_decode_coefs_4x4(dx, xd, bc, PLANE_TYPE_UV, i);
return eobtotal;
return decode_mb_tokens_4x4_uv(dx, xd, bc, seg_eob);
}
static int vp9_decode_mb_tokens_4x4(VP9D_COMP* const dx,
@ -424,24 +540,31 @@ static int vp9_decode_mb_tokens_4x4(VP9D_COMP* const dx,
BOOL_DECODER* const bc) {
int i, eobtotal = 0;
PLANE_TYPE type;
const int segment_id = xd->mode_info_context->mbmi.segment_id;
const int seg_eob = get_eob(xd, segment_id, 16);
const int has_2nd_order = get_2nd_order_usage(xd);
int has_2nd_order = get_2nd_order_usage(xd);
// 2nd order DC block
if (has_2nd_order) {
eobtotal += vp9_decode_coefs_4x4(dx, xd, bc, PLANE_TYPE_Y2, 24) - 16;
eobtotal += decode_coefs_4x4(dx, xd, bc, PLANE_TYPE_Y2, 24, seg_eob,
DCT_DCT, vp9_default_zig_zag1d_4x4) - 16;
type = PLANE_TYPE_Y_NO_DC;
} else {
xd->above_context->y2 = 1;
xd->left_context->y2 = 1;
xd->above_context->y2 = 0;
xd->left_context->y2 = 0;
xd->eobs[24] = 0;
type = PLANE_TYPE_Y_WITH_DC;
}
// luma blocks
for (i = 0; i < 16; ++i) {
eobtotal += vp9_decode_coefs_4x4(dx, xd, bc, type, i);
eobtotal += decode_coefs_4x4_y(dx, xd, bc, type, i, seg_eob);
}
return eobtotal + vp9_decode_mb_tokens_4x4_uv(dx, xd, bc);
// chroma blocks
eobtotal += decode_mb_tokens_4x4_uv(dx, xd, bc, seg_eob);
return eobtotal;
}
int vp9_decode_mb_tokens(VP9D_COMP* const dx,

6
vp9/decoder/vp9_detokenize.h
View File

@ -23,7 +23,11 @@ int vp9_decode_coefs_4x4(VP9D_COMP *dx, MACROBLOCKD *xd,
int vp9_decode_mb_tokens(VP9D_COMP* const, MACROBLOCKD* const,
BOOL_DECODER* const);
int vp9_decode_sb_tokens(VP9D_COMP* const pbi,
MACROBLOCKD* const xd,
BOOL_DECODER* const bc);
int vp9_decode_mb_tokens_4x4_uv(VP9D_COMP* const dx, MACROBLOCKD* const xd,
BOOL_DECODER* const bc);
#endif /* DETOKENIZE_H */
#endif // VP9_DECODER_VP9_DETOKENIZE_H_

128
vp9/decoder/vp9_idct_blk.c
View File

@ -14,11 +14,11 @@
#include "vp9/decoder/vp9_dequantize.h"
#endif
void vp9_dequant_dc_idct_add_y_block_c(short *q, const short *dq,
unsigned char *pre,
unsigned char *dst,
int stride, unsigned short *eobs,
const short *dc) {
void vp9_dequant_dc_idct_add_y_block_c(int16_t *q, const int16_t *dq,
uint8_t *pre,
uint8_t *dst,
int stride, uint16_t *eobs,
const int16_t *dc) {
int i, j;
for (i = 0; i < 4; i++) {
@ -39,12 +39,12 @@ void vp9_dequant_dc_idct_add_y_block_c(short *q, const short *dq,
}
}
#if CONFIG_SUPERBLOCKS
void vp9_dequant_dc_idct_add_y_block_4x4_inplace_c(short *q, const short *dq,
unsigned char *dst,
void vp9_dequant_dc_idct_add_y_block_4x4_inplace_c(int16_t *q,
const int16_t *dq,
uint8_t *dst,
int stride,
unsigned short *eobs,
const short *dc,
uint16_t *eobs,
const int16_t *dc,
MACROBLOCKD *xd) {
int i, j;
@ -63,12 +63,11 @@ void vp9_dequant_dc_idct_add_y_block_4x4_inplace_c(short *q, const short *dq,
dst += 4 * stride - 16;
}
}
#endif
void vp9_dequant_idct_add_y_block_c(short *q, const short *dq,
unsigned char *pre,
unsigned char *dst,
int stride, unsigned short *eobs) {
void vp9_dequant_idct_add_y_block_c(int16_t *q, const int16_t *dq,
uint8_t *pre,
uint8_t *dst,
int stride, uint16_t *eobs) {
int i, j;
for (i = 0; i < 4; i++) {
@ -90,10 +89,10 @@ void vp9_dequant_idct_add_y_block_c(short *q, const short *dq,
}
}
void vp9_dequant_idct_add_uv_block_c(short *q, const short *dq,
unsigned char *pre, unsigned char *dstu,
unsigned char *dstv, int stride,
unsigned short *eobs) {
void vp9_dequant_idct_add_uv_block_c(int16_t *q, const int16_t *dq,
uint8_t *pre, uint8_t *dstu,
uint8_t *dstv, int stride,
uint16_t *eobs) {
int i, j;
for (i = 0; i < 2; i++) {
@ -133,12 +132,11 @@ void vp9_dequant_idct_add_uv_block_c(short *q, const short *dq,
}
}
#if CONFIG_SUPERBLOCKS
void vp9_dequant_idct_add_uv_block_4x4_inplace_c(short *q, const short *dq,
unsigned char *dstu,
unsigned char *dstv,
void vp9_dequant_idct_add_uv_block_4x4_inplace_c(int16_t *q, const int16_t *dq,
uint8_t *dstu,
uint8_t *dstv,
int stride,
unsigned short *eobs,
uint16_t *eobs,
MACROBLOCKD *xd) {
int i, j;
@ -174,13 +172,12 @@ void vp9_dequant_idct_add_uv_block_4x4_inplace_c(short *q, const short *dq,
dstv += 4 * stride - 8;
}
}
#endif
void vp9_dequant_dc_idct_add_y_block_8x8_c(short *q, const short *dq,
unsigned char *pre,
unsigned char *dst,
int stride, unsigned short *eobs,
const short *dc,
void vp9_dequant_dc_idct_add_y_block_8x8_c(int16_t *q, const int16_t *dq,
uint8_t *pre,
uint8_t *dst,
int stride, uint16_t *eobs,
const int16_t *dc,
MACROBLOCKD *xd) {
q[0] = dc[0];
vp9_dequant_idct_add_8x8_c(q, dq, pre, dst, 16, stride, 1, xd->eobs[0]);
@ -199,12 +196,12 @@ void vp9_dequant_dc_idct_add_y_block_8x8_c(short *q, const short *dq,
xd->eobs[12]);
}
#if CONFIG_SUPERBLOCKS
void vp9_dequant_dc_idct_add_y_block_8x8_inplace_c(short *q, const short *dq,
unsigned char *dst,
void vp9_dequant_dc_idct_add_y_block_8x8_inplace_c(int16_t *q,
const int16_t *dq,
uint8_t *dst,
int stride,
unsigned short *eobs,
const short *dc,
uint16_t *eobs,
const int16_t *dc,
MACROBLOCKD *xd) {
q[0] = dc[0];
vp9_dequant_idct_add_8x8_c(q, dq, dst, dst, stride, stride, 1, xd->eobs[0]);
@ -223,15 +220,14 @@ void vp9_dequant_dc_idct_add_y_block_8x8_inplace_c(short *q, const short *dq,
dst + 8 * stride + 8, stride, stride, 1,
xd->eobs[12]);
}
#endif
void vp9_dequant_idct_add_y_block_8x8_c(short *q, const short *dq,
unsigned char *pre,
unsigned char *dst,
int stride, unsigned short *eobs,
void vp9_dequant_idct_add_y_block_8x8_c(int16_t *q, const int16_t *dq,
uint8_t *pre,
uint8_t *dst,
int stride, uint16_t *eobs,
MACROBLOCKD *xd) {
unsigned char *origdest = dst;
unsigned char *origpred = pre;
uint8_t *origdest = dst;
uint8_t *origpred = pre;
vp9_dequant_idct_add_8x8_c(q, dq, pre, dst, 16, stride, 0, xd->eobs[0]);
vp9_dequant_idct_add_8x8_c(&q[64], dq, origpred + 8,
@ -243,11 +239,11 @@ void vp9_dequant_idct_add_y_block_8x8_c(short *q, const short *dq,
xd->eobs[12]);
}
void vp9_dequant_idct_add_uv_block_8x8_c(short *q, const short *dq,
unsigned char *pre,
unsigned char *dstu,
unsigned char *dstv,
int stride, unsigned short *eobs,
void vp9_dequant_idct_add_uv_block_8x8_c(int16_t *q, const int16_t *dq,
uint8_t *pre,
uint8_t *dstu,
uint8_t *dstv,
int stride, uint16_t *eobs,
MACROBLOCKD *xd) {
vp9_dequant_idct_add_8x8_c(q, dq, pre, dstu, 8, stride, 0, xd->eobs[16]);
@ -257,12 +253,11 @@ void vp9_dequant_idct_add_uv_block_8x8_c(short *q, const short *dq,
vp9_dequant_idct_add_8x8_c(q, dq, pre, dstv, 8, stride, 0, xd->eobs[20]);
}
#if CONFIG_SUPERBLOCKS
void vp9_dequant_idct_add_uv_block_8x8_inplace_c(short *q, const short *dq,
unsigned char *dstu,
unsigned char *dstv,
void vp9_dequant_idct_add_uv_block_8x8_inplace_c(int16_t *q, const int16_t *dq,
uint8_t *dstu,
uint8_t *dstv,
int stride,
unsigned short *eobs,
uint16_t *eobs,
MACROBLOCKD *xd) {
vp9_dequant_idct_add_8x8_c(q, dq, dstu, dstu, stride, stride, 0,
xd->eobs[16]);
@ -271,15 +266,14 @@ void vp9_dequant_idct_add_uv_block_8x8_inplace_c(short *q, const short *dq,
vp9_dequant_idct_add_8x8_c(q, dq, dstv, dstv, stride, stride, 0,
xd->eobs[20]);
}
#endif
#if CONFIG_LOSSLESS
void vp9_dequant_dc_idct_add_y_block_lossless_c(short *q, const short *dq,
unsigned char *pre,
unsigned char *dst,
void vp9_dequant_dc_idct_add_y_block_lossless_c(int16_t *q, const int16_t *dq,
uint8_t *pre,
uint8_t *dst,
int stride,
unsigned short *eobs,
const short *dc) {
uint16_t *eobs,
const int16_t *dc) {
int i, j;
for (i = 0; i < 4; i++) {
@ -300,10 +294,10 @@ void vp9_dequant_dc_idct_add_y_block_lossless_c(short *q, const short *dq,
}
}
void vp9_dequant_idct_add_y_block_lossless_c(short *q, const short *dq,
unsigned char *pre,
unsigned char *dst,
int stride, unsigned short *eobs) {
void vp9_dequant_idct_add_y_block_lossless_c(int16_t *q, const int16_t *dq,
uint8_t *pre,
uint8_t *dst,
int stride, uint16_t *eobs) {
int i, j;
for (i = 0; i < 4; i++) {
@ -325,12 +319,12 @@ void vp9_dequant_idct_add_y_block_lossless_c(short *q, const short *dq,
}
}
void vp9_dequant_idct_add_uv_block_lossless_c(short *q, const short *dq,
unsigned char *pre,
unsigned char *dstu,
unsigned char *dstv,
void vp9_dequant_idct_add_uv_block_lossless_c(int16_t *q, const int16_t *dq,
uint8_t *pre,
uint8_t *dstu,
uint8_t *dstv,
int stride,
unsigned short *eobs) {
uint16_t *eobs) {
int i, j;
for (i = 0; i < 2; i++) {

8
vp9/common/vp9_onyxd.h → vp9/decoder/vp9_onyxd.h
View File

@ -8,17 +8,13 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_COMMON_VP9_ONYXD_H_
#define VP9_COMMON_VP9_ONYXD_H_
/* Create/destroy static data structures. */
#ifdef __cplusplus
extern "C"
{
extern "C" {
#endif
#include "vp9/common/vp9_type_aliases.h"
#include "vpx_scale/yv12config.h"
#include "vp9/common/vp9_ppflags.h"
#include "vpx_ports/mem.h"
@ -65,4 +61,4 @@ extern "C"
}
#endif
#endif // __INC_ONYXD_H
#endif // VP9_COMMON_VP9_ONYXD_H_

8
vp9/decoder/vp9_onyxd_if.c
View File

@ -13,7 +13,7 @@
#if CONFIG_POSTPROC
#include "vp9/common/vp9_postproc.h"
#endif
#include "vp9/common/vp9_onyxd.h"
#include "vp9/decoder/vp9_onyxd.h"
#include "vp9/decoder/vp9_onyxd_int.h"
#include "vpx_mem/vpx_mem.h"
#include "vp9/common/vp9_alloccommon.h"
@ -23,7 +23,7 @@
#include <assert.h>
#include "vp9/common/vp9_quant_common.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/vpx_scale.h"
#include "vp9/common/vp9_systemdependent.h"
#include "vpx_ports/vpx_timer.h"
#include "vp9/decoder/vp9_decodframe.h"
@ -37,7 +37,7 @@ static void ref_cnt_fb(int *buf, int *idx, int new_idx);
#if WRITE_RECON_BUFFER == 1
static void recon_write_yuv_frame(char *name, YV12_BUFFER_CONFIG *s) {
FILE *yuv_file = fopen((char *)name, "ab");
unsigned char *src = s->y_buffer;
uint8_t *src = s->y_buffer;
int h = s->y_height;
do {
@ -382,7 +382,7 @@ int vp9_receive_compressed_data(VP9D_PTR ptr, unsigned long size,
if (cm->filter_level) {
/* Apply the loop filter if appropriate. */
vp9_loop_filter_frame(cm, &pbi->mb);
vp9_loop_filter_frame(cm, &pbi->mb, cm->filter_level, 0);
}
vp8_yv12_extend_frame_borders(cm->frame_to_show);
}

17
vp9/decoder/vp9_onyxd_int.h
View File

@ -8,11 +8,10 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef VP9_DECODER_VP9_ONYXD_INT_H_
#define VP9_DECODER_VP9_ONYXD_INT_H_
#include "./vpx_config.h"
#include "vp9/common/vp9_onyxd.h"
#include "vp9/decoder/vp9_onyxd.h"
#include "vp9/decoder/vp9_treereader.h"
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/decoder/vp9_dequantize.h"
@ -35,22 +34,22 @@ typedef struct {
typedef struct {
int const *scan;
int const *scan_8x8;
UINT8 const *ptr_block2leftabove;
uint8_t const *ptr_block2leftabove;
vp9_tree_index const *vp9_coef_tree_ptr;
unsigned char *norm_ptr;
UINT8 *ptr_coef_bands_x;
UINT8 *ptr_coef_bands_x_8x8;
uint8_t *ptr_coef_bands_x;
uint8_t *ptr_coef_bands_x_8x8;
ENTROPY_CONTEXT_PLANES *A;
ENTROPY_CONTEXT_PLANES *L;
INT16 *qcoeff_start_ptr;
int16_t *qcoeff_start_ptr;
vp9_prob const *coef_probs[BLOCK_TYPES];
vp9_prob const *coef_probs_4x4[BLOCK_TYPES_4X4];
vp9_prob const *coef_probs_8x8[BLOCK_TYPES_8X8];
vp9_prob const *coef_probs_16X16[BLOCK_TYPES_16X16];
UINT8 eob[25];
uint8_t eob[25];
} DETOK;
@ -103,4 +102,4 @@ int vp9_decode_frame(VP9D_COMP *cpi, const unsigned char **p_data_end);
} while(0)
#endif
#endif // __INC_ONYXD_INT_H
#endif // VP9_DECODER_VP9_TREEREADER_H_

15
vp9/decoder/vp9_reconintra_mt.h
View File

@ -1,15 +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_DECODER_VP9_RECONINTRA_MT_H_
#define VP9_DECODER_VP9_RECONINTRA_MT_H_
#endif

2
vp9/decoder/vp9_treereader.h
View File

@ -34,4 +34,4 @@ static int treed_read(vp9_reader *const r, /* !!! must return a 0 or 1 !!! */
return -i;
}
#endif /* tree_reader_h */
#endif // VP9_DECODER_VP9_TREEREADER_H_

1605
vp9/encoder/vp9_bitstream.c
View File

File diff suppressed because it is too large Load Diff

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