Newer
Older
/*
* Copyright (c) 2010 The VP8 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 "encodemb.h"
#include "encodemv.h"
#include "common.h"
#include "onyx_int.h"
#include "extend.h"
#include "entropymode.h"
#include "quant_common.h"
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
#include "setupintrarecon.h"
#include "encodeintra.h"
#include "reconinter.h"
#include "rdopt.h"
#include "pickinter.h"
#include "findnearmv.h"
#include "reconintra.h"
#include <stdio.h>
#include <limits.h>
#include "subpixel.h"
#include "vpx_ports/vpx_timer.h"
#if CONFIG_RUNTIME_CPU_DETECT
#define RTCD(x) &cpi->common.rtcd.x
#define IF_RTCD(x) (x)
#else
#define RTCD(x) NULL
#define IF_RTCD(x) NULL
#endif
extern void vp8_stuff_mb(VP8_COMP *cpi, MACROBLOCKD *x, TOKENEXTRA **t) ;
extern void vp8cx_initialize_me_consts(VP8_COMP *cpi, int QIndex);
extern void vp8_auto_select_speed(VP8_COMP *cpi);
extern void vp8cx_init_mbrthread_data(VP8_COMP *cpi,
MACROBLOCK *x,
MB_ROW_COMP *mbr_ei,
int mb_row,
int count);
void vp8_build_block_offsets(MACROBLOCK *x);
void vp8_setup_block_ptrs(MACROBLOCK *x);
int vp8cx_encode_inter_macroblock(VP8_COMP *cpi, MACROBLOCK *x, TOKENEXTRA **t, int recon_yoffset, int recon_uvoffset);
int vp8cx_encode_intra_macro_block(VP8_COMP *cpi, MACROBLOCK *x, TOKENEXTRA **t);
#ifdef MODE_STATS
unsigned int inter_y_modes[10] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
unsigned int inter_uv_modes[4] = {0, 0, 0, 0};
unsigned int inter_b_modes[15] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
unsigned int y_modes[5] = {0, 0, 0, 0, 0};
unsigned int uv_modes[4] = {0, 0, 0, 0};
unsigned int b_modes[14] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
#endif
static const int qrounding_factors[129] =
{
56, 56, 56, 56, 48, 48, 56, 56,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48,
};
static const int qzbin_factors[129] =
{
72, 72, 72, 72, 80, 80, 72, 72,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80,
};
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
static const int qrounding_factors_y2[129] =
{
56, 56, 56, 56, 48, 48, 56, 56,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48, 48, 48, 48, 48, 48, 48, 48,
48,
};
static const int qzbin_factors_y2[129] =
{
72, 72, 72, 72, 80, 80, 72, 72,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80, 80, 80, 80, 80, 80, 80, 80,
80,
};
//#define EXACT_QUANT
#ifdef EXACT_QUANT
static void vp8cx_invert_quant(short *quant, short *shift, short d)
{
unsigned t;
int l;
t = d;
for(l = 0; t > 1; l++)
t>>=1;
t = 1 + (1<<(16+l))/d;
*quant = (short)(t - (1<<16));
*shift = l;
}
void vp8cx_init_quantizer(VP8_COMP *cpi)
{
int r, c;
int i;
int quant_val;
int Q;
int zbin_boost[16] = {0, 0, 8, 10, 12, 14, 16, 20, 24, 28, 32, 36, 40, 44, 44, 44};
for (Q = 0; Q < QINDEX_RANGE; Q++)
{
// dc values
quant_val = vp8_dc_quant(Q, cpi->common.y1dc_delta_q);
vp8cx_invert_quant(cpi->Y1quant[Q][0] + 0,
cpi->Y1quant_shift[Q][0] + 0, quant_val);
cpi->Y1zbin[Q][0][0] = ((qzbin_factors[Q] * quant_val) + 64) >> 7;
cpi->Y1round[Q][0][0] = (qrounding_factors[Q] * quant_val) >> 7;
cpi->common.Y1dequant[Q][0][0] = quant_val;
cpi->zrun_zbin_boost_y1[Q][0] = (quant_val * zbin_boost[0]) >> 7;
quant_val = vp8_dc2quant(Q, cpi->common.y2dc_delta_q);
vp8cx_invert_quant(cpi->Y2quant[Q][0] + 0,
cpi->Y2quant_shift[Q][0] + 0, quant_val);
cpi->Y2zbin[Q][0][0] = ((qzbin_factors_y2[Q] * quant_val) + 64) >> 7;
cpi->Y2round[Q][0][0] = (qrounding_factors_y2[Q] * quant_val) >> 7;
cpi->common.Y2dequant[Q][0][0] = quant_val;
cpi->zrun_zbin_boost_y2[Q][0] = (quant_val * zbin_boost[0]) >> 7;
quant_val = vp8_dc_uv_quant(Q, cpi->common.uvdc_delta_q);
vp8cx_invert_quant(cpi->UVquant[Q][0] + 0,
cpi->UVquant_shift[Q][0] + 0, quant_val);
cpi->UVzbin[Q][0][0] = ((qzbin_factors[Q] * quant_val) + 64) >> 7;;
cpi->UVround[Q][0][0] = (qrounding_factors[Q] * quant_val) >> 7;
cpi->common.UVdequant[Q][0][0] = quant_val;
cpi->zrun_zbin_boost_uv[Q][0] = (quant_val * zbin_boost[0]) >> 7;
// all the ac values = ;
for (i = 1; i < 16; i++)
{
int rc = vp8_default_zig_zag1d[i];
r = (rc >> 2);
c = (rc & 3);
quant_val = vp8_ac_yquant(Q);
vp8cx_invert_quant(cpi->Y1quant[Q][r] + c,
cpi->Y1quant_shift[Q][r] + c, quant_val);
cpi->Y1zbin[Q][r][c] = ((qzbin_factors[Q] * quant_val) + 64) >> 7;
cpi->Y1round[Q][r][c] = (qrounding_factors[Q] * quant_val) >> 7;
cpi->common.Y1dequant[Q][r][c] = quant_val;
cpi->zrun_zbin_boost_y1[Q][i] = (quant_val * zbin_boost[i]) >> 7;
quant_val = vp8_ac2quant(Q, cpi->common.y2ac_delta_q);
vp8cx_invert_quant(cpi->Y2quant[Q][r] + c,
cpi->Y2quant_shift[Q][r] + c, quant_val);
cpi->Y2zbin[Q][r][c] = ((qzbin_factors_y2[Q] * quant_val) + 64) >> 7;
cpi->Y2round[Q][r][c] = (qrounding_factors_y2[Q] * quant_val) >> 7;
cpi->common.Y2dequant[Q][r][c] = quant_val;
cpi->zrun_zbin_boost_y2[Q][i] = (quant_val * zbin_boost[i]) >> 7;
quant_val = vp8_ac_uv_quant(Q, cpi->common.uvac_delta_q);
vp8cx_invert_quant(cpi->UVquant[Q][r] + c,
cpi->UVquant_shift[Q][r] + c, quant_val);
cpi->UVzbin[Q][r][c] = ((qzbin_factors[Q] * quant_val) + 64) >> 7;
cpi->UVround[Q][r][c] = (qrounding_factors[Q] * quant_val) >> 7;
cpi->common.UVdequant[Q][r][c] = quant_val;
cpi->zrun_zbin_boost_uv[Q][i] = (quant_val * zbin_boost[i]) >> 7;
}
}
}
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
void vp8cx_init_quantizer(VP8_COMP *cpi)
{
int r, c;
int i;
int quant_val;
int Q;
int zbin_boost[16] = {0, 0, 8, 10, 12, 14, 16, 20, 24, 28, 32, 36, 40, 44, 44, 44};
for (Q = 0; Q < QINDEX_RANGE; Q++)
{
// dc values
quant_val = vp8_dc_quant(Q, cpi->common.y1dc_delta_q);
cpi->Y1quant[Q][0][0] = (1 << 16) / quant_val;
cpi->Y1zbin[Q][0][0] = ((qzbin_factors[Q] * quant_val) + 64) >> 7;
cpi->Y1round[Q][0][0] = (qrounding_factors[Q] * quant_val) >> 7;
cpi->common.Y1dequant[Q][0][0] = quant_val;
cpi->zrun_zbin_boost_y1[Q][0] = (quant_val * zbin_boost[0]) >> 7;
quant_val = vp8_dc2quant(Q, cpi->common.y2dc_delta_q);
cpi->Y2quant[Q][0][0] = (1 << 16) / quant_val;
cpi->Y2zbin[Q][0][0] = ((qzbin_factors_y2[Q] * quant_val) + 64) >> 7;
cpi->Y2round[Q][0][0] = (qrounding_factors_y2[Q] * quant_val) >> 7;
cpi->common.Y2dequant[Q][0][0] = quant_val;
cpi->zrun_zbin_boost_y2[Q][0] = (quant_val * zbin_boost[0]) >> 7;
quant_val = vp8_dc_uv_quant(Q, cpi->common.uvdc_delta_q);
cpi->UVquant[Q][0][0] = (1 << 16) / quant_val;
cpi->UVzbin[Q][0][0] = ((qzbin_factors[Q] * quant_val) + 64) >> 7;;
cpi->UVround[Q][0][0] = (qrounding_factors[Q] * quant_val) >> 7;
cpi->common.UVdequant[Q][0][0] = quant_val;
cpi->zrun_zbin_boost_uv[Q][0] = (quant_val * zbin_boost[0]) >> 7;
// all the ac values = ;
for (i = 1; i < 16; i++)
{
int rc = vp8_default_zig_zag1d[i];
r = (rc >> 2);
c = (rc & 3);
quant_val = vp8_ac_yquant(Q);
cpi->Y1quant[Q][r][c] = (1 << 16) / quant_val;
cpi->Y1zbin[Q][r][c] = ((qzbin_factors[Q] * quant_val) + 64) >> 7;
cpi->Y1round[Q][r][c] = (qrounding_factors[Q] * quant_val) >> 7;
cpi->common.Y1dequant[Q][r][c] = quant_val;
cpi->zrun_zbin_boost_y1[Q][i] = (quant_val * zbin_boost[i]) >> 7;
quant_val = vp8_ac2quant(Q, cpi->common.y2ac_delta_q);
cpi->Y2quant[Q][r][c] = (1 << 16) / quant_val;
cpi->Y2zbin[Q][r][c] = ((qzbin_factors_y2[Q] * quant_val) + 64) >> 7;
cpi->Y2round[Q][r][c] = (qrounding_factors_y2[Q] * quant_val) >> 7;
cpi->common.Y2dequant[Q][r][c] = quant_val;
cpi->zrun_zbin_boost_y2[Q][i] = (quant_val * zbin_boost[i]) >> 7;
quant_val = vp8_ac_uv_quant(Q, cpi->common.uvac_delta_q);
cpi->UVquant[Q][r][c] = (1 << 16) / quant_val;
cpi->UVzbin[Q][r][c] = ((qzbin_factors[Q] * quant_val) + 64) >> 7;
cpi->UVround[Q][r][c] = (qrounding_factors[Q] * quant_val) >> 7;
cpi->common.UVdequant[Q][r][c] = quant_val;
cpi->zrun_zbin_boost_uv[Q][i] = (quant_val * zbin_boost[i]) >> 7;
}
}
}
void vp8cx_mb_init_quantizer(VP8_COMP *cpi, MACROBLOCK *x)
{
int i;
int QIndex;
MACROBLOCKD *xd = &x->e_mbd;
int zbin_extra;
// Select the baseline MB Q index.
if (xd->segmentation_enabled)
{
// Abs Value
if (xd->mb_segement_abs_delta == SEGMENT_ABSDATA)
QIndex = xd->segment_feature_data[MB_LVL_ALT_Q][xd->mode_info_context->mbmi.segment_id];
QIndex = cpi->common.base_qindex + xd->segment_feature_data[MB_LVL_ALT_Q][xd->mode_info_context->mbmi.segment_id];
QIndex = (QIndex >= 0) ? ((QIndex <= MAXQ) ? QIndex : MAXQ) : 0; // Clamp to valid range
}
}
else
QIndex = cpi->common.base_qindex;
// Y
zbin_extra = (cpi->common.Y1dequant[QIndex][0][1] * (cpi->zbin_over_quant + cpi->zbin_mode_boost)) >> 7;
for (i = 0; i < 16; i++)
{
x->block[i].quant = cpi->Y1quant[QIndex];
x->block[i].quant_shift = cpi->Y1quant_shift[QIndex];
x->block[i].zbin = cpi->Y1zbin[QIndex];
x->block[i].round = cpi->Y1round[QIndex];
x->e_mbd.block[i].dequant = cpi->common.Y1dequant[QIndex];
x->block[i].zrun_zbin_boost = cpi->zrun_zbin_boost_y1[QIndex];
x->block[i].zbin_extra = (short)zbin_extra;
}
// UV
zbin_extra = (cpi->common.UVdequant[QIndex][0][1] * (cpi->zbin_over_quant + cpi->zbin_mode_boost)) >> 7;
for (i = 16; i < 24; i++)
{
x->block[i].quant = cpi->UVquant[QIndex];
x->block[i].quant_shift = cpi->UVquant_shift[QIndex];
x->block[i].zbin = cpi->UVzbin[QIndex];
x->block[i].round = cpi->UVround[QIndex];
x->e_mbd.block[i].dequant = cpi->common.UVdequant[QIndex];
x->block[i].zrun_zbin_boost = cpi->zrun_zbin_boost_uv[QIndex];
x->block[i].zbin_extra = (short)zbin_extra;
}
// Y2
zbin_extra = (cpi->common.Y2dequant[QIndex][0][1] * ((cpi->zbin_over_quant / 2) + cpi->zbin_mode_boost)) >> 7;
x->block[24].quant = cpi->Y2quant[QIndex];
x->block[24].quant_shift = cpi->Y2quant_shift[QIndex];
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
x->block[24].zbin = cpi->Y2zbin[QIndex];
x->block[24].round = cpi->Y2round[QIndex];
x->e_mbd.block[24].dequant = cpi->common.Y2dequant[QIndex];
x->block[24].zrun_zbin_boost = cpi->zrun_zbin_boost_y2[QIndex];
x->block[24].zbin_extra = (short)zbin_extra;
}
void vp8cx_frame_init_quantizer(VP8_COMP *cpi)
{
// vp8cx_init_quantizer() is first called in vp8_create_compressor(). A check is added here so that vp8cx_init_quantizer() is only called
// when these values are not all zero.
if (cpi->common.y1dc_delta_q | cpi->common.y2dc_delta_q | cpi->common.uvdc_delta_q | cpi->common.y2ac_delta_q | cpi->common.uvac_delta_q)
{
vp8cx_init_quantizer(cpi);
}
// MB level quantizer setup
vp8cx_mb_init_quantizer(cpi, &cpi->mb);
}
static
void encode_mb_row(VP8_COMP *cpi,
VP8_COMMON *cm,
int mb_row,
MACROBLOCK *x,
MACROBLOCKD *xd,
TOKENEXTRA **tp,
int *segment_counts,
int *totalrate)
{
int i;
int recon_yoffset, recon_uvoffset;
int mb_col;
int ref_fb_idx = cm->lst_fb_idx;
int dst_fb_idx = cm->new_fb_idx;
int recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride;
int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
int seg_map_index = (mb_row * cpi->common.mb_cols);
// reset above block coeffs
xd->above_context = cm->above_context;
xd->up_available = (mb_row != 0);
recon_yoffset = (mb_row * recon_y_stride * 16);
recon_uvoffset = (mb_row * recon_uv_stride * 8);
cpi->tplist[mb_row].start = *tp;
//printf("Main mb_row = %d\n", mb_row);
// for each macroblock col in image
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
// Distance of Mb to the various image edges.
// These specified to 8th pel as they are always compared to values that are in 1/8th pel units
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_right_edge = ((cm->mb_cols - 1 - mb_col) * 16) << 3;
xd->mb_to_top_edge = -((mb_row * 16) << 3);
xd->mb_to_bottom_edge = ((cm->mb_rows - 1 - mb_row) * 16) << 3;
// Set up limit values for motion vectors used to prevent them extending outside the UMV borders
x->mv_col_min = -((mb_col * 16) + (VP8BORDERINPIXELS - 16));
x->mv_col_max = ((cm->mb_cols - 1 - mb_col) * 16) + (VP8BORDERINPIXELS - 16);
x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
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;
xd->left_available = (mb_col != 0);
// Is segmentation enabled
// MB level adjutment to quantizer
if (xd->segmentation_enabled)
{
// Code to set segment id in xd->mbmi.segment_id for current MB (with range checking)
if (cpi->segmentation_map[seg_map_index+mb_col] <= 3)
xd->mode_info_context->mbmi.segment_id = cpi->segmentation_map[seg_map_index+mb_col];
xd->mode_info_context->mbmi.segment_id = 0;
vp8cx_mb_init_quantizer(cpi, x);
}
else
xd->mode_info_context->mbmi.segment_id = 0; // Set to Segment 0 by default
x->active_ptr = cpi->active_map + seg_map_index + mb_col;
if (cm->frame_type == KEY_FRAME)
{
*totalrate += vp8cx_encode_intra_macro_block(cpi, x, tp);
#ifdef MODE_STATS
y_modes[xd->mbmi.mode] ++;
#endif
}
else
{
*totalrate += vp8cx_encode_inter_macroblock(cpi, x, tp, recon_yoffset, recon_uvoffset);
#ifdef MODE_STATS
inter_y_modes[xd->mbmi.mode] ++;
if (xd->mbmi.mode == SPLITMV)
{
int b;
for (b = 0; b < xd->mbmi.partition_count; b++)
{
inter_b_modes[x->partition->bmi[b].mode] ++;
}
}
#endif
// Count of last ref frame 0,0 useage
if ((xd->mode_info_context->mbmi.mode == ZEROMV) && (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME))
cpi->inter_zz_count ++;
// Special case code for cyclic refresh
// If cyclic update enabled then copy xd->mbmi.segment_id; (which may have been updated based on mode
// during vp8cx_encode_inter_macroblock()) back into the global sgmentation map
if (cpi->cyclic_refresh_mode_enabled && xd->segmentation_enabled)
{
cpi->segmentation_map[seg_map_index+mb_col] = xd->mode_info_context->mbmi.segment_id;
// If the block has been refreshed mark it as clean (the magnitude of the -ve influences how long it will be before we consider another refresh):
// Else if it was coded (last frame 0,0) and has not already been refreshed then mark it as a candidate for cleanup next time (marked 0)
// else mark it as dirty (1).
if (xd->mode_info_context->mbmi.segment_id)
cpi->cyclic_refresh_map[seg_map_index+mb_col] = -1;
else if ((xd->mode_info_context->mbmi.mode == ZEROMV) && (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME))
{
if (cpi->cyclic_refresh_map[seg_map_index+mb_col] == 1)
cpi->cyclic_refresh_map[seg_map_index+mb_col] = 0;
}
else
cpi->cyclic_refresh_map[seg_map_index+mb_col] = 1;
}
}
cpi->tplist[mb_row].stop = *tp;
x->gf_active_ptr++; // Increment pointer into gf useage flags structure for next mb
for (i = 0; i < 16; i++)
vpx_memcpy(&xd->mode_info_context->bmi[i], &xd->block[i].bmi, sizeof(xd->block[i].bmi));
// adjust to the next column of macroblocks
x->src.y_buffer += 16;
x->src.u_buffer += 8;
x->src.v_buffer += 8;
recon_yoffset += 16;
recon_uvoffset += 8;
// Keep track of segment useage
segment_counts[xd->mode_info_context->mbmi.segment_id] ++;
// skip to next mb
xd->mode_info_context++;
cpi->current_mb_col_main = mb_col;
}
//extend the recon for intra prediction
vp8_extend_mb_row(
&cm->yv12_fb[dst_fb_idx],
xd->dst.y_buffer + 16,
xd->dst.u_buffer + 8,
xd->dst.v_buffer + 8);
// this is to account for the border
xd->mode_info_context++;
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
}
void vp8_encode_frame(VP8_COMP *cpi)
{
int mb_row;
MACROBLOCK *const x = & cpi->mb;
VP8_COMMON *const cm = & cpi->common;
MACROBLOCKD *const xd = & x->e_mbd;
int i;
TOKENEXTRA *tp = cpi->tok;
int segment_counts[MAX_MB_SEGMENTS];
int totalrate;
if (cm->frame_type != KEY_FRAME)
{
if (cm->mcomp_filter_type == SIXTAP)
{
xd->subpixel_predict = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, sixtap4x4);
xd->subpixel_predict8x4 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, sixtap8x4);
xd->subpixel_predict8x8 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, sixtap8x8);
xd->subpixel_predict16x16 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, sixtap16x16);
}
else
{
xd->subpixel_predict = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, bilinear4x4);
xd->subpixel_predict8x4 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, bilinear8x4);
xd->subpixel_predict8x8 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, bilinear8x8);
xd->subpixel_predict16x16 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, bilinear16x16);
}
}
//else // Key Frame
//{
// For key frames make sure the intra ref frame probability value
// is set to "all intra"
//cpi->prob_intra_coded = 255;
//}
x->gf_active_ptr = (signed char *)cpi->gf_active_flags; // Point to base of GF active flags data structure
x->vector_range = 32;
// Count of MBs using the alternate Q if any
cpi->alt_qcount = 0;
// Reset frame count of inter 0,0 motion vector useage.
cpi->inter_zz_count = 0;
vpx_memset(segment_counts, 0, sizeof(segment_counts));
cpi->prediction_error = 0;
cpi->intra_error = 0;
cpi->skip_true_count = 0;
cpi->skip_false_count = 0;
#if 0
// Experimental code
cpi->last_mb_distortion = 0;
#endif
totalrate = 0;
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
xd->mode_info_context = cm->mi;
xd->mode_info_stride = cm->mode_info_stride;
xd->frame_type = cm->frame_type;
xd->frames_since_golden = cm->frames_since_golden;
xd->frames_till_alt_ref_frame = cm->frames_till_alt_ref_frame;
vp8_zero(cpi->MVcount);
// vp8_zero( Contexts)
vp8_zero(cpi->coef_counts);
// reset intra mode contexts
if (cm->frame_type == KEY_FRAME)
vp8_init_mbmode_probs(cm);
vp8cx_frame_init_quantizer(cpi);
if (cpi->compressor_speed == 2)
{
if (cpi->oxcf.cpu_used < 0)
cpi->Speed = -(cpi->oxcf.cpu_used);
else
vp8_auto_select_speed(cpi);
}
vp8_initialize_rd_consts(cpi, vp8_dc_quant(cm->base_qindex, cm->y1dc_delta_q));
//vp8_initialize_rd_consts( cpi, vp8_dc_quant(cpi->avg_frame_qindex, cm->y1dc_delta_q) );
vp8cx_initialize_me_consts(cpi, cm->base_qindex);
//vp8cx_initialize_me_consts( cpi, cpi->avg_frame_qindex);
// Copy data over into macro block data sturctures.
x->src = * cpi->Source;
xd->pre = cm->yv12_fb[cm->lst_fb_idx];
xd->dst = cm->yv12_fb[cm->new_fb_idx];
// set up frame new frame for intra coded blocks
vp8_setup_intra_recon(&cm->yv12_fb[cm->new_fb_idx]);
vp8_build_block_offsets(x);
vp8_setup_block_dptrs(&x->e_mbd);
vp8_setup_block_ptrs(x);
x->rddiv = cpi->RDDIV;
x->rdmult = cpi->RDMULT;
#if 0
// Experimental rd code
// 2 Pass - Possibly set Rdmult based on last frame distortion + this frame target bits or other metrics
// such as cpi->rate_correction_factor that indicate relative complexity.
/*if ( cpi->pass == 2 && (cpi->last_frame_distortion > 0) && (cpi->target_bits_per_mb > 0) )
{
//x->rdmult = ((cpi->last_frame_distortion * 256)/cpi->common.MBs)/ cpi->target_bits_per_mb;
x->rdmult = (int)(cpi->RDMULT * cpi->rate_correction_factor);
}
else
x->rdmult = cpi->RDMULT; */
//x->rdmult = (int)(cpi->RDMULT * pow( (cpi->rate_correction_factor * 2.0), 0.75 ));
#endif
xd->mode_info_context->mbmi.mode = DC_PRED;
xd->mode_info_context->mbmi.uv_mode = DC_PRED;
xd->left_context = &cm->left_context;
vp8_zero(cpi->count_mb_ref_frame_usage)
vp8_zero(cpi->ymode_count)
vp8_zero(cpi->uv_mode_count)
x->mvc = cm->fc.mvc;
vpx_memset(cm->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * cm->mb_cols);
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
{
struct vpx_usec_timer emr_timer;
vpx_usec_timer_start(&emr_timer);
if (!cpi->b_multi_threaded)
{
// for each macroblock row in image
for (mb_row = 0; mb_row < cm->mb_rows; mb_row++)
{
vp8_zero(cm->left_context)
encode_mb_row(cpi, cm, mb_row, x, xd, &tp, segment_counts, &totalrate);
// adjust to the next row of mbs
x->src.y_buffer += 16 * x->src.y_stride - 16 * cm->mb_cols;
x->src.u_buffer += 8 * x->src.uv_stride - 8 * cm->mb_cols;
x->src.v_buffer += 8 * x->src.uv_stride - 8 * cm->mb_cols;
}
cpi->tok_count = tp - cpi->tok;
}
else
{
#if CONFIG_MULTITHREAD
vp8cx_init_mbrthread_data(cpi, x, cpi->mb_row_ei, 1, cpi->encoding_thread_count);
for (mb_row = 0; mb_row < cm->mb_rows; mb_row += (cpi->encoding_thread_count + 1))
{
int i;
cpi->current_mb_col_main = -1;
for (i = 0; i < cpi->encoding_thread_count; i++)
{
if ((mb_row + i + 1) >= cm->mb_rows)
break;
cpi->mb_row_ei[i].mb_row = mb_row + i + 1;
cpi->mb_row_ei[i].tp = cpi->tok + (mb_row + i + 1) * (cm->mb_cols * 16 * 24);
cpi->mb_row_ei[i].current_mb_col = -1;
//SetEvent(cpi->h_event_mbrencoding[i]);
sem_post(&cpi->h_event_mbrencoding[i]);
}
vp8_zero(cm->left_context)
tp = cpi->tok + mb_row * (cm->mb_cols * 16 * 24);
encode_mb_row(cpi, cm, mb_row, x, xd, &tp, segment_counts, &totalrate);
// adjust to the next row of mbs
x->src.y_buffer += 16 * x->src.y_stride * (cpi->encoding_thread_count + 1) - 16 * cm->mb_cols;
x->src.u_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
x->src.v_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
xd->mode_info_context += xd->mode_info_stride * cpi->encoding_thread_count;
x->partition_info += xd->mode_info_stride * cpi->encoding_thread_count;
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
if (mb_row < cm->mb_rows - 1)
//WaitForSingleObject(cpi->h_event_main, INFINITE);
sem_wait(&cpi->h_event_main);
}
/*
for( ;mb_row<cm->mb_rows; mb_row ++)
{
vp8_zero( cm->left_context)
tp = cpi->tok + mb_row * (cm->mb_cols * 16 * 24);
encode_mb_row(cpi, cm, mb_row, x, xd, &tp, segment_counts, &totalrate);
// adjust to the next row of mbs
x->src.y_buffer += 16 * x->src.y_stride - 16 * cm->mb_cols;
x->src.u_buffer += 8 * x->src.uv_stride - 8 * cm->mb_cols;
x->src.v_buffer += 8 * x->src.uv_stride - 8 * cm->mb_cols;
}
*/
cpi->tok_count = 0;
for (mb_row = 0; mb_row < cm->mb_rows; mb_row ++)
{
cpi->tok_count += cpi->tplist[mb_row].stop - cpi->tplist[mb_row].start;
}
if (xd->segmentation_enabled)
{
int i, j;
if (xd->segmentation_enabled)
{
for (i = 0; i < cpi->encoding_thread_count; i++)
{
for (j = 0; j < 4; j++)
segment_counts[j] += cpi->mb_row_ei[i].segment_counts[j];
}
}
}
for (i = 0; i < cpi->encoding_thread_count; i++)
{
totalrate += cpi->mb_row_ei[i].totalrate;
}
#endif
}
vpx_usec_timer_mark(&emr_timer);
cpi->time_encode_mb_row += vpx_usec_timer_elapsed(&emr_timer);
}
// Work out the segment probabilites if segmentation is enabled
if (xd->segmentation_enabled)
{
int tot_count;
int i;
// Set to defaults
vpx_memset(xd->mb_segment_tree_probs, 255 , sizeof(xd->mb_segment_tree_probs));
tot_count = segment_counts[0] + segment_counts[1] + segment_counts[2] + segment_counts[3];
if (tot_count)
{
xd->mb_segment_tree_probs[0] = ((segment_counts[0] + segment_counts[1]) * 255) / tot_count;
tot_count = segment_counts[0] + segment_counts[1];
if (tot_count > 0)
{
xd->mb_segment_tree_probs[1] = (segment_counts[0] * 255) / tot_count;
}
tot_count = segment_counts[2] + segment_counts[3];
if (tot_count > 0)
xd->mb_segment_tree_probs[2] = (segment_counts[2] * 255) / tot_count;
// Zero probabilities not allowed
for (i = 0; i < MB_FEATURE_TREE_PROBS; i ++)
{
if (xd->mb_segment_tree_probs[i] == 0)
xd->mb_segment_tree_probs[i] = 1;
}
}
}
// 256 rate units to the bit
cpi->projected_frame_size = totalrate >> 8; // projected_frame_size in units of BYTES
// Make a note of the percentage MBs coded Intra.
if (cm->frame_type == KEY_FRAME)
{
cpi->this_frame_percent_intra = 100;
}
else
{
int tot_modes;
tot_modes = cpi->count_mb_ref_frame_usage[INTRA_FRAME]
+ cpi->count_mb_ref_frame_usage[LAST_FRAME]
+ cpi->count_mb_ref_frame_usage[GOLDEN_FRAME]
+ cpi->count_mb_ref_frame_usage[ALTREF_FRAME];
if (tot_modes)
cpi->this_frame_percent_intra = cpi->count_mb_ref_frame_usage[INTRA_FRAME] * 100 / tot_modes;
}
#if 0
{
int cnt = 0;
int flag[2] = {0, 0};
for (cnt = 0; cnt < MVPcount; cnt++)
{
if (cm->fc.pre_mvc[0][cnt] != cm->fc.mvc[0][cnt])
{
flag[0] = 1;
vpx_memcpy(cm->fc.pre_mvc[0], cm->fc.mvc[0], MVPcount);
break;
}
}
for (cnt = 0; cnt < MVPcount; cnt++)
{
if (cm->fc.pre_mvc[1][cnt] != cm->fc.mvc[1][cnt])
{
flag[1] = 1;
vpx_memcpy(cm->fc.pre_mvc[1], cm->fc.mvc[1], MVPcount);
break;
}
}
if (flag[0] || flag[1])
vp8_build_component_cost_table(cpi->mb.mvcost, cpi->mb.mvsadcost, (const MV_CONTEXT *) cm->fc.mvc, flag);
}
#endif
// Adjust the projected reference frame useage probability numbers to reflect
// what we have just seen. This may be usefull when we make multiple itterations
// of the recode loop rather than continuing to use values from the previous frame.
if ((cm->frame_type != KEY_FRAME) && !cm->refresh_alt_ref_frame && !cm->refresh_golden_frame)
{
const int *const rfct = cpi->count_mb_ref_frame_usage;
const int rf_intra = rfct[INTRA_FRAME];
const int rf_inter = rfct[LAST_FRAME] + rfct[GOLDEN_FRAME] + rfct[ALTREF_FRAME];
if ((rf_intra + rf_inter) > 0)
{
cpi->prob_intra_coded = (rf_intra * 255) / (rf_intra + rf_inter);
if (cpi->prob_intra_coded < 1)
cpi->prob_intra_coded = 1;
if ((cm->frames_since_golden > 0) || cpi->source_alt_ref_active)
{
cpi->prob_last_coded = rf_inter ? (rfct[LAST_FRAME] * 255) / rf_inter : 128;
if (cpi->prob_last_coded < 1)
cpi->prob_last_coded = 1;
cpi->prob_gf_coded = (rfct[GOLDEN_FRAME] + rfct[ALTREF_FRAME])
? (rfct[GOLDEN_FRAME] * 255) / (rfct[GOLDEN_FRAME] + rfct[ALTREF_FRAME]) : 128;
if (cpi->prob_gf_coded < 1)
cpi->prob_gf_coded = 1;
}
}
}
#if 0
// Keep record of the total distortion this time around for future use
cpi->last_frame_distortion = cpi->frame_distortion;
#endif
}
void vp8_setup_block_ptrs(MACROBLOCK *x)
{
int r, c;
int i;
for (r = 0; r < 4; r++)
{
for (c = 0; c < 4; c++)
{
x->block[r*4+c].src_diff = x->src_diff + r * 4 * 16 + c * 4;
}
}
for (r = 0; r < 2; r++)
{
for (c = 0; c < 2; c++)
{
x->block[16 + r*2+c].src_diff = x->src_diff + 256 + r * 4 * 8 + c * 4;
}
}
for (r = 0; r < 2; r++)
{
for (c = 0; c < 2; c++)
{
x->block[20 + r*2+c].src_diff = x->src_diff + 320 + r * 4 * 8 + c * 4;
}
}
x->block[24].src_diff = x->src_diff + 384;
for (i = 0; i < 25; i++)
{
x->block[i].coeff = x->coeff + i * 16;
}
}
void vp8_build_block_offsets(MACROBLOCK *x)
{
int block = 0;
int br, bc;
vp8_build_block_doffsets(&x->e_mbd);
// y blocks
for (br = 0; br < 4; br++)
{
for (bc = 0; bc < 4; bc++)
{
BLOCK *this_block = &x->block[block];
this_block->base_src = &x->src.y_buffer;
this_block->src_stride = x->src.y_stride;
this_block->src = 4 * br * this_block->src_stride + 4 * bc;
++block;
}
}
// u blocks
for (br = 0; br < 2; br++)
{
for (bc = 0; bc < 2; bc++)
{
BLOCK *this_block = &x->block[block];
this_block->base_src = &x->src.u_buffer;
this_block->src_stride = x->src.uv_stride;
this_block->src = 4 * br * this_block->src_stride + 4 * bc;
++block;
}
}
// v blocks
for (br = 0; br < 2; br++)
{
for (bc = 0; bc < 2; bc++)
{
BLOCK *this_block = &x->block[block];
this_block->base_src = &x->src.v_buffer;