mozilla
/
aom
зеркало из
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

Merge "Restructure multi-threaded decoder"

This commit is contained in:
Yunqing Wang 2010-09-21 05:00:30 -07:00 коммит произвёл Code Review
Родитель b7dc9398f2 f857a85088
Коммит a23ccf8f8c
8 изменённых файлов: 1624 добавлений и 474 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

980
vp8/common/reconintra_mt.c Normal file
Просмотреть файл

@ -0,0 +1,980 @@
/*
* 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/config.h"
#include "recon.h"
#include "reconintra.h"
#include "vpx_mem/vpx_mem.h"
#include "onyxd_int.h"
// For skip_recon_mb(), add vp8_build_intra_predictors_mby_s(MACROBLOCKD *x) and
// vp8_build_intra_predictors_mbuv_s(MACROBLOCKD *x).
void vp8mt_build_intra_predictors_mby(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col)
{
#if CONFIG_MULTITHREAD
unsigned char *yabove_row; // = x->dst.y_buffer - x->dst.y_stride;
unsigned char *yleft_col;
unsigned char yleft_buf[16];
unsigned char ytop_left; // = yabove_row[-1];
unsigned char *ypred_ptr = x->predictor;
int r, c, i;
if (pbi->common.filter_level)
{
yabove_row = pbi->mt_yabove_row[mb_row] + mb_col*16 +32;
yleft_col = pbi->mt_yleft_col[mb_row];
} else
{
yabove_row = x->dst.y_buffer - x->dst.y_stride;
for (i = 0; i < 16; i++)
yleft_buf[i] = x->dst.y_buffer [i* x->dst.y_stride -1];
yleft_col = yleft_buf;
}
ytop_left = yabove_row[-1];
// for Y
switch (x->mode_info_context->mbmi.mode)
{
case DC_PRED:
{
int expected_dc;
int i;
int shift;
int average = 0;
if (x->up_available || x->left_available)
{
if (x->up_available)
{
for (i = 0; i < 16; i++)
{
average += yabove_row[i];
}
}
if (x->left_available)
{
for (i = 0; i < 16; i++)
{
average += yleft_col[i];
}
}
shift = 3 + x->up_available + x->left_available;
expected_dc = (average + (1 << (shift - 1))) >> shift;
}
else
{
expected_dc = 128;
}
vpx_memset(ypred_ptr, expected_dc, 256);
}
break;
case V_PRED:
{
for (r = 0; r < 16; r++)
{
((int *)ypred_ptr)[0] = ((int *)yabove_row)[0];
((int *)ypred_ptr)[1] = ((int *)yabove_row)[1];
((int *)ypred_ptr)[2] = ((int *)yabove_row)[2];
((int *)ypred_ptr)[3] = ((int *)yabove_row)[3];
ypred_ptr += 16;
}
}
break;
case H_PRED:
{
for (r = 0; r < 16; r++)
{
vpx_memset(ypred_ptr, yleft_col[r], 16);
ypred_ptr += 16;
}
}
break;
case TM_PRED:
{
for (r = 0; r < 16; r++)
{
for (c = 0; c < 16; 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 += 16;
}
}
break;
case B_PRED:
case NEARESTMV:
case NEARMV:
case ZEROMV:
case NEWMV:
case SPLITMV:
case MB_MODE_COUNT:
break;
}
#else
(void) pbi;
(void) x;
(void) mb_row;
(void) mb_col;
#endif
}
void vp8mt_build_intra_predictors_mby_s(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col)
{
#if CONFIG_MULTITHREAD
unsigned char *yabove_row; // = x->dst.y_buffer - x->dst.y_stride;
unsigned char *yleft_col;
unsigned char yleft_buf[16];
unsigned char ytop_left; // = yabove_row[-1];
unsigned char *ypred_ptr = x->predictor;
int r, c, i;
int y_stride = x->dst.y_stride;
ypred_ptr = x->dst.y_buffer; //x->predictor;
if (pbi->common.filter_level)
{
yabove_row = pbi->mt_yabove_row[mb_row] + mb_col*16 +32;
yleft_col = pbi->mt_yleft_col[mb_row];
} else
{
yabove_row = x->dst.y_buffer - x->dst.y_stride;
for (i = 0; i < 16; i++)
yleft_buf[i] = x->dst.y_buffer [i* x->dst.y_stride -1];
yleft_col = yleft_buf;
}
ytop_left = yabove_row[-1];
// for Y
switch (x->mode_info_context->mbmi.mode)
{
case DC_PRED:
{
int expected_dc;
int i;
int shift;
int average = 0;
if (x->up_available || x->left_available)
{
if (x->up_available)
{
for (i = 0; i < 16; i++)
{
average += yabove_row[i];
}
}
if (x->left_available)
{
for (i = 0; i < 16; i++)
{
average += yleft_col[i];
}
}
shift = 3 + x->up_available + x->left_available;
expected_dc = (average + (1 << (shift - 1))) >> shift;
}
else
{
expected_dc = 128;
}
//vpx_memset(ypred_ptr, expected_dc, 256);
for (r = 0; r < 16; r++)
{
vpx_memset(ypred_ptr, expected_dc, 16);
ypred_ptr += y_stride; //16;
}
}
break;
case V_PRED:
{
for (r = 0; r < 16; r++)
{
((int *)ypred_ptr)[0] = ((int *)yabove_row)[0];
((int *)ypred_ptr)[1] = ((int *)yabove_row)[1];
((int *)ypred_ptr)[2] = ((int *)yabove_row)[2];
((int *)ypred_ptr)[3] = ((int *)yabove_row)[3];
ypred_ptr += y_stride; //16;
}
}
break;
case H_PRED:
{
for (r = 0; r < 16; r++)
{
vpx_memset(ypred_ptr, yleft_col[r], 16);
ypred_ptr += y_stride; //16;
}
}
break;
case TM_PRED:
{
for (r = 0; r < 16; r++)
{
for (c = 0; c < 16; 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 += y_stride; //16;
}
}
break;
case B_PRED:
case NEARESTMV:
case NEARMV:
case ZEROMV:
case NEWMV:
case SPLITMV:
case MB_MODE_COUNT:
break;
}
#else
(void) pbi;
(void) x;
(void) mb_row;
(void) mb_col;
#endif
}
void vp8mt_build_intra_predictors_mbuv(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col)
{
#if CONFIG_MULTITHREAD
unsigned char *uabove_row; // = x->dst.u_buffer - x->dst.uv_stride;
unsigned char *uleft_col; //[16];
unsigned char uleft_buf[8];
unsigned char utop_left; // = uabove_row[-1];
unsigned char *vabove_row; // = x->dst.v_buffer - x->dst.uv_stride;
unsigned char *vleft_col; //[20];
unsigned char vleft_buf[8];
unsigned char vtop_left; // = vabove_row[-1];
unsigned char *upred_ptr = &x->predictor[256];
unsigned char *vpred_ptr = &x->predictor[320];
int i, j;
if (pbi->common.filter_level)
{
uabove_row = pbi->mt_uabove_row[mb_row] + mb_col*8 +16;
vabove_row = pbi->mt_vabove_row[mb_row] + mb_col*8 +16;
uleft_col = pbi->mt_uleft_col[mb_row];
vleft_col = pbi->mt_vleft_col[mb_row];
} else
{
uabove_row = x->dst.u_buffer - x->dst.uv_stride;
vabove_row = x->dst.v_buffer - x->dst.uv_stride;
for (i = 0; i < 8; i++)
{
uleft_buf[i] = x->dst.u_buffer [i* x->dst.uv_stride -1];
vleft_buf[i] = x->dst.v_buffer [i* x->dst.uv_stride -1];
}
uleft_col = uleft_buf;
vleft_col = vleft_buf;
}
utop_left = uabove_row[-1];
vtop_left = vabove_row[-1];
switch (x->mode_info_context->mbmi.uv_mode)
{
case DC_PRED:
{
int expected_udc;
int expected_vdc;
int i;
int shift;
int Uaverage = 0;
int Vaverage = 0;
if (x->up_available)
{
for (i = 0; i < 8; i++)
{
Uaverage += uabove_row[i];
Vaverage += vabove_row[i];
}
}
if (x->left_available)
{
for (i = 0; i < 8; i++)
{
Uaverage += uleft_col[i];
Vaverage += vleft_col[i];
}
}
if (!x->up_available && !x->left_available)
{
expected_udc = 128;
expected_vdc = 128;
}
else
{
shift = 2 + x->up_available + x->left_available;
expected_udc = (Uaverage + (1 << (shift - 1))) >> shift;
expected_vdc = (Vaverage + (1 << (shift - 1))) >> shift;
}
vpx_memset(upred_ptr, expected_udc, 64);
vpx_memset(vpred_ptr, expected_vdc, 64);
}
break;
case V_PRED:
{
int i;
for (i = 0; i < 8; i++)
{
vpx_memcpy(upred_ptr, uabove_row, 8);
vpx_memcpy(vpred_ptr, vabove_row, 8);
upred_ptr += 8;
vpred_ptr += 8;
}
}
break;
case H_PRED:
{
int i;
for (i = 0; i < 8; i++)
{
vpx_memset(upred_ptr, uleft_col[i], 8);
vpx_memset(vpred_ptr, vleft_col[i], 8);
upred_ptr += 8;
vpred_ptr += 8;
}
}
break;
case TM_PRED:
{
int i;
for (i = 0; i < 8; i++)
{
for (j = 0; j < 8; j++)
{
int predu = uleft_col[i] + uabove_row[j] - utop_left;
int predv = vleft_col[i] + vabove_row[j] - vtop_left;
if (predu < 0)
predu = 0;
if (predu > 255)
predu = 255;
if (predv < 0)
predv = 0;
if (predv > 255)
predv = 255;
upred_ptr[j] = predu;
vpred_ptr[j] = predv;
}
upred_ptr += 8;
vpred_ptr += 8;
}
}
break;
case B_PRED:
case NEARESTMV:
case NEARMV:
case ZEROMV:
case NEWMV:
case SPLITMV:
case MB_MODE_COUNT:
break;
}
#else
(void) pbi;
(void) x;
(void) mb_row;
(void) mb_col;
#endif
}
void vp8mt_build_intra_predictors_mbuv_s(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col)
{
#if CONFIG_MULTITHREAD
unsigned char *uabove_row; // = x->dst.u_buffer - x->dst.uv_stride;
unsigned char *uleft_col; //[16];
unsigned char uleft_buf[8];
unsigned char utop_left; // = uabove_row[-1];
unsigned char *vabove_row; // = x->dst.v_buffer - x->dst.uv_stride;
unsigned char *vleft_col; //[20];
unsigned char vleft_buf[8];
unsigned char vtop_left; // = vabove_row[-1];
unsigned char *upred_ptr = x->dst.u_buffer; //&x->predictor[256];
unsigned char *vpred_ptr = x->dst.v_buffer; //&x->predictor[320];
int uv_stride = x->dst.uv_stride;
int i, j;
if (pbi->common.filter_level)
{
uabove_row = pbi->mt_uabove_row[mb_row] + mb_col*8 +16;
vabove_row = pbi->mt_vabove_row[mb_row] + mb_col*8 +16;
uleft_col = pbi->mt_uleft_col[mb_row];
vleft_col = pbi->mt_vleft_col[mb_row];
} else
{
uabove_row = x->dst.u_buffer - x->dst.uv_stride;
vabove_row = x->dst.v_buffer - x->dst.uv_stride;
for (i = 0; i < 8; i++)
{
uleft_buf[i] = x->dst.u_buffer [i* x->dst.uv_stride -1];
vleft_buf[i] = x->dst.v_buffer [i* x->dst.uv_stride -1];
}
uleft_col = uleft_buf;
vleft_col = vleft_buf;
}
utop_left = uabove_row[-1];
vtop_left = vabove_row[-1];
switch (x->mode_info_context->mbmi.uv_mode)
{
case DC_PRED:
{
int expected_udc;
int expected_vdc;
int i;
int shift;
int Uaverage = 0;
int Vaverage = 0;
if (x->up_available)
{
for (i = 0; i < 8; i++)
{
Uaverage += uabove_row[i];
Vaverage += vabove_row[i];
}
}
if (x->left_available)
{
for (i = 0; i < 8; i++)
{
Uaverage += uleft_col[i];
Vaverage += vleft_col[i];
}
}
if (!x->up_available && !x->left_available)
{
expected_udc = 128;
expected_vdc = 128;
}
else
{
shift = 2 + x->up_available + x->left_available;
expected_udc = (Uaverage + (1 << (shift - 1))) >> shift;
expected_vdc = (Vaverage + (1 << (shift - 1))) >> shift;
}
//vpx_memset(upred_ptr,expected_udc,64);
//vpx_memset(vpred_ptr,expected_vdc,64);
for (i = 0; i < 8; i++)
{
vpx_memset(upred_ptr, expected_udc, 8);
vpx_memset(vpred_ptr, expected_vdc, 8);
upred_ptr += uv_stride; //8;
vpred_ptr += uv_stride; //8;
}
}
break;
case V_PRED:
{
int i;
for (i = 0; i < 8; i++)
{
vpx_memcpy(upred_ptr, uabove_row, 8);
vpx_memcpy(vpred_ptr, vabove_row, 8);
upred_ptr += uv_stride; //8;
vpred_ptr += uv_stride; //8;
}
}
break;
case H_PRED:
{
int i;
for (i = 0; i < 8; i++)
{
vpx_memset(upred_ptr, uleft_col[i], 8);
vpx_memset(vpred_ptr, vleft_col[i], 8);
upred_ptr += uv_stride; //8;
vpred_ptr += uv_stride; //8;
}
}
break;
case TM_PRED:
{
int i;
for (i = 0; i < 8; i++)
{
for (j = 0; j < 8; j++)
{
int predu = uleft_col[i] + uabove_row[j] - utop_left;
int predv = vleft_col[i] + vabove_row[j] - vtop_left;
if (predu < 0)
predu = 0;
if (predu > 255)
predu = 255;
if (predv < 0)
predv = 0;
if (predv > 255)
predv = 255;
upred_ptr[j] = predu;
vpred_ptr[j] = predv;
}
upred_ptr += uv_stride; //8;
vpred_ptr += uv_stride; //8;
}
}
break;
case B_PRED:
case NEARESTMV:
case NEARMV:
case ZEROMV:
case NEWMV:
case SPLITMV:
case MB_MODE_COUNT:
break;
}
#else
(void) pbi;
(void) x;
(void) mb_row;
(void) mb_col;
#endif
}
void vp8mt_predict_intra4x4(VP8D_COMP *pbi,
MACROBLOCKD *xd,
int b_mode,
unsigned char *predictor,
int mb_row,
int mb_col,
int num)
{
#if CONFIG_MULTITHREAD
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];
BLOCKD *x = &xd->block[num];
//Caution: For some b_mode, it needs 8 pixels (4 above + 4 above-right).
if (num < 4 && pbi->common.filter_level)
Above = pbi->mt_yabove_row[mb_row] + mb_col*16 + num*4 + 32;
else
Above = *(x->base_dst) + x->dst - x->dst_stride;
if (num%4==0 && pbi->common.filter_level)
{
for (i=0; i<4; i++)
Left[i] = pbi->mt_yleft_col[mb_row][num + i];
}else
{
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 ((num==4 || num==8 || num==12) && pbi->common.filter_level)
top_left = pbi->mt_yleft_col[mb_row][num-1];
else
top_left = Above[-1];
switch (b_mode)
{
case B_DC_PRED:
{
int expected_dc = 0;
for (i = 0; i < 4; i++)
{
expected_dc += Above[i];
expected_dc += Left[i];
}
expected_dc = (expected_dc + 4) >> 3;
for (r = 0; r < 4; r++)
{
for (c = 0; c < 4; c++)
{
predictor[c] = expected_dc;
}
predictor += 16;
}
}
break;
case B_TM_PRED:
{
// 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 += 16;
}
}
break;
case B_VE_PRED:
{
unsigned int ap[4];
ap[0] = (top_left + 2 * Above[0] + Above[1] + 2) >> 2;
ap[1] = (Above[0] + 2 * Above[1] + Above[2] + 2) >> 2;
ap[2] = (Above[1] + 2 * Above[2] + Above[3] + 2) >> 2;
ap[3] = (Above[2] + 2 * Above[3] + Above[4] + 2) >> 2;
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] = (top_left + 2 * Left[0] + Left[1] + 2) >> 2;
lp[1] = (Left[0] + 2 * Left[1] + Left[2] + 2) >> 2;
lp[2] = (Left[1] + 2 * Left[2] + Left[3] + 2) >> 2;
lp[3] = (Left[2] + 2 * Left[3] + Left[3] + 2) >> 2;
for (r = 0; r < 4; r++)
{
for (c = 0; c < 4; c++)
{
predictor[c] = lp[r];
}
predictor += 16;
}
}
break;
case B_LD_PRED:
{
unsigned char *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;
predictor[0 * 16 + 2] =
predictor[1 * 16 + 1] =
predictor[2 * 16 + 0] = (ptr[2] + ptr[3] * 2 + ptr[4] + 2) >> 2;
predictor[0 * 16 + 3] =
predictor[1 * 16 + 2] =
predictor[2 * 16 + 1] =
predictor[3 * 16 + 0] = (ptr[3] + ptr[4] * 2 + ptr[5] + 2) >> 2;
predictor[1 * 16 + 3] =
predictor[2 * 16 + 2] =
predictor[3 * 16 + 1] = (ptr[4] + ptr[5] * 2 + ptr[6] + 2) >> 2;
predictor[2 * 16 + 3] =
predictor[3 * 16 + 2] = (ptr[5] + ptr[6] * 2 + ptr[7] + 2) >> 2;
predictor[3 * 16 + 3] = (ptr[6] + ptr[7] * 2 + ptr[7] + 2) >> 2;
}
break;
case B_RD_PRED:
{
unsigned char 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];
predictor[3 * 16 + 0] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
predictor[3 * 16 + 1] =
predictor[2 * 16 + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
predictor[3 * 16 + 2] =
predictor[2 * 16 + 1] =
predictor[1 * 16 + 0] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
predictor[3 * 16 + 3] =
predictor[2 * 16 + 2] =
predictor[1 * 16 + 1] =
predictor[0 * 16 + 0] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
predictor[2 * 16 + 3] =
predictor[1 * 16 + 2] =
predictor[0 * 16 + 1] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
predictor[1 * 16 + 3] =
predictor[0 * 16 + 2] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
predictor[0 * 16 + 3] = (pp[6] + pp[7] * 2 + pp[8] + 2) >> 2;
}
break;
case B_VR_PRED:
{
unsigned char 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];
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;
predictor[3 * 16 + 1] =
predictor[1 * 16 + 0] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
predictor[2 * 16 + 1] =
predictor[0 * 16 + 0] = (pp[4] + pp[5] + 1) >> 1;
predictor[3 * 16 + 2] =
predictor[1 * 16 + 1] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
predictor[2 * 16 + 2] =
predictor[0 * 16 + 1] = (pp[5] + pp[6] + 1) >> 1;
predictor[3 * 16 + 3] =
predictor[1 * 16 + 2] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
predictor[2 * 16 + 3] =
predictor[0 * 16 + 2] = (pp[6] + pp[7] + 1) >> 1;
predictor[1 * 16 + 3] = (pp[6] + pp[7] * 2 + pp[8] + 2) >> 2;
predictor[0 * 16 + 3] = (pp[7] + pp[8] + 1) >> 1;
}
break;
case B_VL_PRED:
{
unsigned char *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;
predictor[2 * 16 + 0] =
predictor[0 * 16 + 1] = (pp[1] + pp[2] + 1) >> 1;
predictor[1 * 16 + 1] =
predictor[3 * 16 + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
predictor[2 * 16 + 1] =
predictor[0 * 16 + 2] = (pp[2] + pp[3] + 1) >> 1;
predictor[3 * 16 + 1] =
predictor[1 * 16 + 2] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
predictor[0 * 16 + 3] =
predictor[2 * 16 + 2] = (pp[3] + pp[4] + 1) >> 1;
predictor[1 * 16 + 3] =
predictor[3 * 16 + 2] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
predictor[2 * 16 + 3] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
predictor[3 * 16 + 3] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
}
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];
pp[4] = top_left;
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;
predictor[3 * 16 + 1] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
predictor[2 * 16 + 0] =
predictor[3 * 16 + 2] = (pp[1] + pp[2] + 1) >> 1;
predictor[2 * 16 + 1] =
predictor[3 * 16 + 3] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
predictor[2 * 16 + 2] =
predictor[1 * 16 + 0] = (pp[2] + pp[3] + 1) >> 1;
predictor[2 * 16 + 3] =
predictor[1 * 16 + 1] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
predictor[1 * 16 + 2] =
predictor[0 * 16 + 0] = (pp[3] + pp[4] + 1) >> 1;
predictor[1 * 16 + 3] =
predictor[0 * 16 + 1] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
predictor[0 * 16 + 2] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
predictor[0 * 16 + 3] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
}
break;
case B_HU_PRED:
{
unsigned char *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] =
predictor[1 * 16 + 0] = (pp[1] + pp[2] + 1) >> 1;
predictor[0 * 16 + 3] =
predictor[1 * 16 + 1] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
predictor[1 * 16 + 2] =
predictor[2 * 16 + 0] = (pp[2] + pp[3] + 1) >> 1;
predictor[1 * 16 + 3] =
predictor[2 * 16 + 1] = (pp[2] + pp[3] * 2 + pp[3] + 2) >> 2;
predictor[2 * 16 + 2] =
predictor[2 * 16 + 3] =
predictor[3 * 16 + 0] =
predictor[3 * 16 + 1] =
predictor[3 * 16 + 2] =
predictor[3 * 16 + 3] = pp[3];
}
break;
}
#else
(void) pbi;
(void) xd;
(void) b_mode;
(void) predictor;
(void) mb_row;
(void) mb_col;
(void) num;
#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 vp8mt_intra_prediction_down_copy(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col)
{
#if CONFIG_MULTITHREAD
unsigned char *above_right; // = *(x->block[0].base_dst) + x->block[0].dst - x->block[0].dst_stride + 16;
unsigned int *src_ptr;
unsigned int *dst_ptr0;
unsigned int *dst_ptr1;
unsigned int *dst_ptr2;
if (pbi->common.filter_level)
above_right = pbi->mt_yabove_row[mb_row] + mb_col*16 + 32 +16;
else
above_right = *(x->block[0].base_dst) + x->block[0].dst - x->block[0].dst_stride + 16;
src_ptr = (unsigned int *)above_right;
//dst_ptr0 = (unsigned int *)(above_right + 4 * x->block[0].dst_stride);
//dst_ptr1 = (unsigned int *)(above_right + 8 * x->block[0].dst_stride);
//dst_ptr2 = (unsigned int *)(above_right + 12 * x->block[0].dst_stride);
dst_ptr0 = (unsigned int *)(*(x->block[0].base_dst) + x->block[0].dst + 16 + 3 * x->block[0].dst_stride);
dst_ptr1 = (unsigned int *)(*(x->block[0].base_dst) + x->block[0].dst + 16 + 7 * x->block[0].dst_stride);
dst_ptr2 = (unsigned int *)(*(x->block[0].base_dst) + x->block[0].dst + 16 + 11 * x->block[0].dst_stride);
*dst_ptr0 = *src_ptr;
*dst_ptr1 = *src_ptr;
*dst_ptr2 = *src_ptr;
#else
(void) pbi;
(void) x;
(void) mb_row;
(void) mb_col;
#endif
}

26
vp8/common/reconintra_mt.h Normal file
Просмотреть файл

@ -0,0 +1,26 @@
/*
* 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 __INC_RECONINTRA_MT_H
#define __INC_RECONINTRA_MT_H
// reconintra functions used in multi-threaded decoder
#if CONFIG_MULTITHREAD
extern void vp8mt_build_intra_predictors_mby(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col);
extern void vp8mt_build_intra_predictors_mby_s(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col);
extern void vp8mt_build_intra_predictors_mbuv(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col);
extern void vp8mt_build_intra_predictors_mbuv_s(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col);
extern void vp8mt_predict_intra4x4(VP8D_COMP *pbi, MACROBLOCKD *x, int b_mode, unsigned char *predictor, int mb_row, int mb_col, int num);
extern void vp8mt_intra_prediction_down_copy(VP8D_COMP *pbi, MACROBLOCKD *x, int mb_row, int mb_col);
#endif
#endif

12
vp8/decoder/decoderthreading.h
Просмотреть файл

@ -15,12 +15,12 @@
#ifndef _DECODER_THREADING_H
#define _DECODER_THREADING_H
extern void vp8_mtdecode_mb_rows(VP8D_COMP *pbi,
MACROBLOCKD *xd);
extern void vp8_mt_loop_filter_frame(VP8D_COMP *pbi);
extern void vp8_stop_lfthread(VP8D_COMP *pbi);
extern void vp8_start_lfthread(VP8D_COMP *pbi);
#if CONFIG_MULTITHREAD
extern void vp8mt_decode_mb_rows(VP8D_COMP *pbi, MACROBLOCKD *xd);
extern void vp8_decoder_remove_threads(VP8D_COMP *pbi);
extern void vp8_decoder_create_threads(VP8D_COMP *pbi);
extern int vp8mt_alloc_temp_buffers(VP8D_COMP *pbi, int width, int prev_mb_rows);
extern void vp8mt_de_alloc_temp_buffers(VP8D_COMP *pbi, int mb_rows);
#endif
#endif

38
vp8/decoder/decodframe.c
Просмотреть файл

@ -21,7 +21,8 @@
#include "alloccommon.h"
#include "entropymode.h"
#include "quant_common.h"
#include "vpx_scale/vpxscale.h"
#include "vpx_scale/yv12extend.h"
#include "setupintrarecon.h"
#include "decodemv.h"
@ -64,7 +65,7 @@ void vp8cx_init_de_quantizer(VP8D_COMP *pbi)
}
}
static void mb_init_dequantizer(VP8D_COMP *pbi, MACROBLOCKD *xd)
void mb_init_dequantizer(VP8D_COMP *pbi, MACROBLOCKD *xd)
{
int i;
int QIndex;
@ -158,7 +159,7 @@ static void clamp_uvmv_to_umv_border(MV *mv, const MACROBLOCKD *xd)
mv->row = (2*mv->row > xd->mb_to_bottom_edge + (18 << 3)) ? (xd->mb_to_bottom_edge + (16 << 3)) >> 1 : mv->row;
}
static void clamp_mvs(MACROBLOCKD *xd)
void clamp_mvs(MACROBLOCKD *xd)
{
if (xd->mode_info_context->mbmi.mode == SPLITMV)
{
@ -294,6 +295,7 @@ void vp8_decode_macroblock(VP8D_COMP *pbi, MACROBLOCKD *xd)
xd->dst.uv_stride, xd->eobs+16);
}
static int get_delta_q(vp8_reader *bc, int prev, int *q_update)
{
int ret_val = 0;
@ -398,7 +400,6 @@ void vp8_decode_mb_row(VP8D_COMP *pbi,
xd->above_context++;
pbi->current_mb_col_main = mb_col;
}
// adjust to the next row of mbs
@ -408,8 +409,6 @@ void vp8_decode_mb_row(VP8D_COMP *pbi,
);
++xd->mode_info_context; /* skip prediction column */
pbi->last_mb_row_decoded = mb_row;
}
@ -601,6 +600,8 @@ int vp8_decode_frame(VP8D_COMP *pbi)
if (Width != pc->Width || Height != pc->Height)
{
int prev_mb_rows = pc->mb_rows;
if (pc->Width <= 0)
{
pc->Width = Width;
@ -618,6 +619,11 @@ int vp8_decode_frame(VP8D_COMP *pbi)
if (vp8_alloc_frame_buffers(pc, pc->Width, pc->Height))
vpx_internal_error(&pc->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate frame buffers");
#if CONFIG_MULTITHREAD
if (pbi->b_multithreaded_rd)
vp8mt_alloc_temp_buffers(pbi, pc->Width, prev_mb_rows);
#endif
}
}
@ -822,7 +828,8 @@ int vp8_decode_frame(VP8D_COMP *pbi)
vpx_memcpy(&xd->dst, &pc->yv12_fb[pc->new_fb_idx], sizeof(YV12_BUFFER_CONFIG));
// set up frame new frame for intra coded blocks
vp8_setup_intra_recon(&pc->yv12_fb[pc->new_fb_idx]);
if (!(pbi->b_multithreaded_rd) || pc->multi_token_partition == ONE_PARTITION || !(pc->filter_level))
vp8_setup_intra_recon(&pc->yv12_fb[pc->new_fb_idx]);
vp8_setup_block_dptrs(xd);
@ -841,13 +848,18 @@ int vp8_decode_frame(VP8D_COMP *pbi)
vpx_memcpy(&xd->block[0].bmi, &xd->mode_info_context->bmi[0], sizeof(B_MODE_INFO));
if (pbi->b_multithreaded_lf && pc->filter_level != 0)
vp8_start_lfthread(pbi);
if (pbi->b_multithreaded_rd && pc->multi_token_partition != ONE_PARTITION)
{
vp8_mtdecode_mb_rows(pbi, xd);
vp8mt_decode_mb_rows(pbi, xd);
if(pbi->common.filter_level)
{
//vp8_mt_loop_filter_frame(pbi); //cm, &pbi->mb, cm->filter_level);
pc->last_frame_type = pc->frame_type;
pc->last_filter_type = pc->filter_type;
pc->last_sharpness_level = pc->sharpness_level;
}
vp8_yv12_extend_frame_borders_ptr(&pc->yv12_fb[pc->new_fb_idx]); //cm->frame_to_show);
}
else
{
@ -869,8 +881,6 @@ int vp8_decode_frame(VP8D_COMP *pbi)
vp8_decode_mb_row(pbi, pc, mb_row, xd);
}
pbi->last_mb_row_decoded = mb_row;
}

84
vp8/decoder/onyxd_if.c
Просмотреть файл

@ -142,6 +142,10 @@ void vp8dx_remove_decompressor(VP8D_PTR ptr)
if (!pbi)
return;
#if CONFIG_MULTITHREAD
if (pbi->b_multithreaded_rd)
vp8mt_de_alloc_temp_buffers(pbi, pbi->common.mb_rows);
#endif
vp8_decoder_remove_threads(pbi);
vp8_remove_common(&pbi->common);
vpx_free(pbi);
@ -355,66 +359,40 @@ int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsign
return retcode;
}
if (pbi->b_multithreaded_lf && pbi->common.filter_level != 0)
vp8_stop_lfthread(pbi);
if (swap_frame_buffers (cm))
if (pbi->b_multithreaded_rd && cm->multi_token_partition != ONE_PARTITION)
{
pbi->common.error.error_code = VPX_CODEC_ERROR;
pbi->common.error.setjmp = 0;
return -1;
}
/*
if (!pbi->b_multithreaded_lf)
if (swap_frame_buffers (cm))
{
pbi->common.error.error_code = VPX_CODEC_ERROR;
pbi->common.error.setjmp = 0;
return -1;
}
} else
{
struct vpx_usec_timer lpftimer;
vpx_usec_timer_start(&lpftimer);
// Apply the loop filter if appropriate.
if (swap_frame_buffers (cm))
{
pbi->common.error.error_code = VPX_CODEC_ERROR;
pbi->common.error.setjmp = 0;
return -1;
}
if(pbi->common.filter_level)
{
struct vpx_usec_timer lpftimer;
vpx_usec_timer_start(&lpftimer);
// Apply the loop filter if appropriate.
if (cm->filter_level > 0)
vp8_loop_filter_frame(cm, &pbi->mb, cm->filter_level);
vpx_usec_timer_mark(&lpftimer);
pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
}else{
struct vpx_usec_timer lpftimer;
vpx_usec_timer_start(&lpftimer);
// Apply the loop filter if appropriate.
vpx_usec_timer_mark(&lpftimer);
pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
if (cm->filter_level > 0)
vp8_mt_loop_filter_frame(cm, &pbi->mb, cm->filter_level);
vpx_usec_timer_mark(&lpftimer);
pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
cm->last_frame_type = cm->frame_type;
cm->last_filter_type = cm->filter_type;
cm->last_sharpness_level = cm->sharpness_level;
}
vp8_yv12_extend_frame_borders_ptr(cm->frame_to_show);
}
if (cm->filter_level > 0) {
cm->last_frame_type = cm->frame_type;
cm->last_filter_type = cm->filter_type;
cm->last_sharpness_level = cm->sharpness_level;
}
*/
if(pbi->common.filter_level)
{
struct vpx_usec_timer lpftimer;
vpx_usec_timer_start(&lpftimer);
// Apply the loop filter if appropriate.
if (pbi->b_multithreaded_lf && cm->multi_token_partition != ONE_PARTITION)
vp8_mt_loop_filter_frame(pbi); //cm, &pbi->mb, cm->filter_level);
else
vp8_loop_filter_frame(cm, &pbi->mb, cm->filter_level);
vpx_usec_timer_mark(&lpftimer);
pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
cm->last_frame_type = cm->frame_type;
cm->last_filter_type = cm->filter_type;
cm->last_sharpness_level = cm->sharpness_level;
}
vp8_yv12_extend_frame_borders_ptr(cm->frame_to_show);
#if 0
// DEBUG code

26
vp8/decoder/onyxd_int.h
Просмотреть файл

@ -88,30 +88,32 @@ typedef struct VP8Decompressor
unsigned int time_loop_filtering;
volatile int b_multithreaded_rd;
volatile int b_multithreaded_lf;
int max_threads;
int last_mb_row_decoded;
int current_mb_col_main;
int decoding_thread_count;
int allocated_decoding_thread_count;
int *current_mb_col; //Each row remembers its already decoded column.
int mt_baseline_filter_level[MAX_MB_SEGMENTS];
// variable for threading
DECLARE_ALIGNED(16, MACROBLOCKD, lpfmb);
#if CONFIG_MULTITHREAD
//pthread_t h_thread_lpf; // thread for postprocessing
sem_t h_event_end_lpf; // Event for post_proc completed
sem_t *h_event_start_lpf;
#endif
int mt_baseline_filter_level[MAX_MB_SEGMENTS];
int *mt_current_mb_col; // Each row remembers its already decoded column.
unsigned char **mt_yabove_row; // mb_rows x width
unsigned char **mt_uabove_row;
unsigned char **mt_vabove_row;
unsigned char **mt_yleft_col; // mb_rows x 16
unsigned char **mt_uleft_col; // mb_rows x 8
unsigned char **mt_vleft_col; // mb_rows x 8
MB_ROW_DEC *mb_row_di;
DECODETHREAD_DATA *de_thread_data;
#if CONFIG_MULTITHREAD
DECODETHREAD_DATA *de_thread_data;
pthread_t *h_decoding_thread;
sem_t *h_event_start_decoding;
sem_t h_event_end_decoding;
sem_t h_event_end_decoding;
// end of threading data
#endif
vp8_reader *mbc;
INT64 last_time_stamp;
int ready_for_new_data;

930
vp8/decoder/threading.c
Просмотреть файл

Разница между файлами не показана из-за своего большого размера Загрузить разницу

2
vp8/vp8_common.mk
Просмотреть файл

@ -81,6 +81,8 @@ VP8_COMMON_SRCS-yes += common/recon.c
VP8_COMMON_SRCS-yes += common/reconinter.c
VP8_COMMON_SRCS-yes += common/reconintra.c
VP8_COMMON_SRCS-yes += common/reconintra4x4.c
VP8_COMMON_SRCS-yes += common/reconintra_mt.h
VP8_COMMON_SRCS-yes += common/reconintra_mt.c
VP8_COMMON_SRCS-yes += common/setupintrarecon.c
VP8_COMMON_SRCS-yes += common/swapyv12buffer.c
VP8_COMMON_SRCS-yes += common/textblit.c