3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder main module - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* 2002 Peter Ross <pross@xvid.org> |
|
* to use this software module in hardware or software products are |
|
|
* advised that its use may infringe existing patents or copyrights, and |
|
|
* any such use would be at such party's own risk. The original |
|
|
* developer of this software module and his/her company, and subsequent |
|
|
* editors and their companies, will have no liability for use of this |
|
|
* software or modifications or derivatives thereof. |
|
8 |
* |
* |
9 |
* This program is free software; you can redistribute it and/or modify |
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
10 |
* it under the terms of the GNU General Public License as published by |
* |
11 |
|
* XviD is free software; you can redistribute it and/or modify it |
12 |
|
* under the terms of the GNU General Public License as published by |
13 |
* the Free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
14 |
* (at your option) any later version. |
* (at your option) any later version. |
15 |
* |
* |
22 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
23 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
24 |
* |
* |
25 |
*************************************************************************/ |
* Under section 8 of the GNU General Public License, the copyright |
26 |
|
* holders of XVID explicitly forbid distribution in the following |
27 |
/************************************************************************** |
* countries: |
28 |
|
* |
29 |
|
* - Japan |
30 |
|
* - United States of America |
31 |
* |
* |
32 |
* History: |
* Linking XviD statically or dynamically with other modules is making a |
33 |
|
* combined work based on XviD. Thus, the terms and conditions of the |
34 |
|
* GNU General Public License cover the whole combination. |
35 |
* |
* |
36 |
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
* As a special exception, the copyright holders of XviD give you |
37 |
* MinChen <chenm001@163.com> |
* permission to link XviD with independent modules that communicate with |
38 |
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
* XviD solely through the VFW1.1 and DShow interfaces, regardless of the |
39 |
* Fix a little bug for low_delay flage |
* license terms of these independent modules, and to copy and distribute |
40 |
* MinChen <chenm001@163.com> |
* the resulting combined work under terms of your choice, provided that |
41 |
* 28.06.2002 added basic resync support to iframe/pframe_decode() |
* every copy of the combined work is accompanied by a complete copy of |
42 |
* 22.06.2002 added primative N_VOP support |
* the source code of XviD (the version of XviD used to produce the |
43 |
* #define BFRAMES_DEC now enables Minchen's bframe decoder |
* combined work), being distributed under the terms of the GNU General |
44 |
* 08.05.2002 add low_delay support for B_VOP decode |
* Public License plus this exception. An independent module is a module |
45 |
* MinChen <chenm001@163.com> |
* which is not derived from or based on XviD. |
46 |
* 05.05.2002 fix some B-frame decode problem |
* |
47 |
* 02.05.2002 add B-frame decode support(have some problem); |
* Note that people who make modified versions of XviD are not obligated |
48 |
* MinChen <chenm001@163.com> |
* to grant this special exception for their modified versions; it is |
49 |
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
* their choice whether to do so. The GNU General Public License gives |
50 |
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
* permission to release a modified version without this exception; this |
51 |
* reconstructing blocks, thus artifacts |
* exception also makes it possible to release a modified version which |
52 |
* interlacing speedup - used transfers to re-interlace |
* carries forward this exception. |
|
* interlaced decoding should be as fast as progressive now |
|
|
* 26.03.2002 interlacing support - moved transfers outside decode loop |
|
|
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
|
|
* 22.12.2001 lock based interpolation |
|
|
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
53 |
* |
* |
54 |
* $Id$ |
* $Id$ |
55 |
* |
* |
56 |
*************************************************************************/ |
*************************************************************************/ |
57 |
|
|
|
#include <stdio.h> |
|
58 |
#include <stdlib.h> |
#include <stdlib.h> |
59 |
#include <string.h> |
#include <string.h> |
60 |
|
|
64 |
|
|
65 |
#include "xvid.h" |
#include "xvid.h" |
66 |
#include "portab.h" |
#include "portab.h" |
|
#include "global.h" |
|
67 |
|
|
68 |
#include "decoder.h" |
#include "decoder.h" |
69 |
#include "bitstream/bitstream.h" |
#include "bitstream/bitstream.h" |
75 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
76 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
77 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "image/reduced.h" |
|
|
#include "image/font.h" |
|
78 |
|
|
79 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
80 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
81 |
#include "utils/timer.h" |
#include "utils/timer.h" |
82 |
#include "utils/emms.h" |
#include "utils/emms.h" |
|
#include "motion/motion.h" |
|
83 |
|
|
84 |
#include "image/image.h" |
#include "image/image.h" |
85 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
86 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
87 |
|
|
88 |
int |
int |
89 |
decoder_resize(DECODER * dec) |
decoder_create(XVID_DEC_PARAM * param) |
90 |
{ |
{ |
91 |
/* free existing */ |
DECODER *dec; |
|
|
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
|
|
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
|
92 |
|
|
93 |
if (dec->last_mbs) |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
94 |
xvid_free(dec->last_mbs); |
if (dec == NULL) { |
95 |
if (dec->mbs) |
return XVID_ERR_MEMORY; |
96 |
xvid_free(dec->mbs); |
} |
97 |
|
param->handle = dec; |
98 |
|
|
99 |
/* realloc */ |
dec->width = param->width; |
100 |
|
dec->height = param->height; |
101 |
|
|
102 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
103 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
104 |
|
|
105 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
106 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
107 |
|
dec->low_delay = 0; |
108 |
|
|
109 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
110 |
xvid_free(dec); |
xvid_free(dec); |
116 |
xvid_free(dec); |
xvid_free(dec); |
117 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
118 |
} |
} |
119 |
|
/* add by chenm001 <chenm001@163.com> */ |
120 |
// add by chenm001 <chenm001@163.com> |
/* for support B-frame to reference last 2 frame */ |
|
// for support B-frame to reference last 2 frame |
|
121 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
122 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
123 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
124 |
xvid_free(dec); |
xvid_free(dec); |
125 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
126 |
} |
} |
127 |
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
|
128 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
129 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
130 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
131 |
xvid_free(dec); |
xvid_free(dec); |
132 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
133 |
} |
} |
139 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
140 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
141 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
142 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
143 |
xvid_free(dec); |
xvid_free(dec); |
144 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
145 |
} |
} |
146 |
|
|
147 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
148 |
|
|
149 |
// add by chenm001 <chenm001@163.com> |
/* add by chenm001 <chenm001@163.com> */ |
150 |
// for skip MB flag |
/* for skip MB flag */ |
151 |
dec->last_mbs = |
dec->last_mbs = |
152 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
153 |
CACHE_LINE); |
CACHE_LINE); |
156 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
157 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
158 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
159 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
160 |
xvid_free(dec); |
xvid_free(dec); |
161 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
162 |
} |
} |
163 |
|
|
164 |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
165 |
|
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
|
|
|
|
int |
|
|
decoder_create(XVID_DEC_PARAM * param) |
|
|
{ |
|
|
DECODER *dec; |
|
|
|
|
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
|
|
if (dec == NULL) { |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
memset(dec, 0, sizeof(DECODER)); |
|
|
|
|
|
param->handle = dec; |
|
|
|
|
|
dec->width = param->width; |
|
|
dec->height = param->height; |
|
|
|
|
|
image_null(&dec->cur); |
|
|
image_null(&dec->refn[0]); |
|
|
image_null(&dec->refn[1]); |
|
|
image_null(&dec->tmp); |
|
|
image_null(&dec->qtmp); |
|
|
|
|
|
/* image based GMC */ |
|
|
image_null(&dec->gmc); |
|
|
|
|
|
|
|
|
dec->mbs = NULL; |
|
|
dec->last_mbs = NULL; |
|
|
|
|
166 |
init_timer(); |
init_timer(); |
167 |
|
|
168 |
// add by chenm001 <chenm001@163.com> |
/* add by chenm001 <chenm001@163.com> */ |
169 |
// for support B-frame to save reference frame's time |
/* for support B-frame to save reference frame's time */ |
170 |
dec->frames = 0; |
dec->frames = -1; |
171 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
|
dec->low_delay = 0; |
|
|
dec->packed_mode = 0; |
|
|
|
|
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
|
172 |
|
|
|
if (dec->fixed_dimensions) |
|
|
return decoder_resize(dec); |
|
|
else |
|
173 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
174 |
} |
} |
175 |
|
|
179 |
{ |
{ |
180 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
181 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
|
|
|
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); /* image based GMC */ |
|
|
|
|
182 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
183 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
184 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
185 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
186 |
xvid_free(dec); |
xvid_free(dec); |
187 |
|
|
198 |
|
|
199 |
|
|
200 |
|
|
201 |
// decode an intra macroblock |
/* decode an intra macroblock */ |
202 |
|
|
203 |
void |
void |
204 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
210 |
Bitstream * bs, |
Bitstream * bs, |
211 |
const uint32_t quant, |
const uint32_t quant, |
212 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
213 |
const unsigned int bound, |
const unsigned int bound) |
|
const int reduced_resolution) |
|
214 |
{ |
{ |
215 |
|
|
216 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
223 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
224 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
225 |
|
|
|
if (reduced_resolution) { |
|
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
}else{ |
|
226 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
227 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
228 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
229 |
|
|
230 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
231 |
|
|
232 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
233 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
250 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
251 |
|
|
252 |
if (dc_size > 8) { |
if (dc_size > 8) { |
253 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); /* marker */ |
254 |
} |
} |
255 |
|
|
256 |
block[i * 64 + 0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
262 |
} |
} |
263 |
|
|
264 |
start_timer(); |
start_timer(); |
265 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
266 |
{ |
{ |
267 |
int direction = dec->alternate_vertical_scan ? |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
268 |
2 : pMB->acpred_directions[i]; |
start_coeff); |
|
|
|
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
|
269 |
} |
} |
270 |
stop_coding_timer(); |
stop_coding_timer(); |
271 |
|
|
284 |
start_timer(); |
start_timer(); |
285 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
286 |
stop_idct_timer(); |
stop_idct_timer(); |
|
|
|
287 |
} |
} |
288 |
|
|
289 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
292 |
} |
} |
293 |
|
|
294 |
start_timer(); |
start_timer(); |
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
next_block*=2; |
|
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
}else{ |
|
295 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
296 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
297 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
298 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
299 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
300 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
301 |
stop_transfer_timer(); |
stop_transfer_timer(); |
302 |
} |
} |
303 |
|
|
304 |
|
|
305 |
|
|
306 |
|
|
307 |
// decode an inter macroblock |
|
308 |
|
#define SIGN(X) (((X)>0)?1:-1) |
309 |
|
#define ABS(X) (((X)>0)?(X):-(X)) |
310 |
|
static const uint32_t roundtab[16] = |
311 |
|
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
312 |
|
|
313 |
|
|
314 |
|
/* decode an inter macroblock */ |
315 |
|
|
316 |
void |
void |
317 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
322 |
const uint32_t cbp, |
const uint32_t cbp, |
323 |
Bitstream * bs, |
Bitstream * bs, |
324 |
const uint32_t quant, |
const uint32_t quant, |
325 |
const uint32_t rounding, |
const uint32_t rounding) |
|
const int reduced_resolution, |
|
|
const int mcsel) |
|
326 |
{ |
{ |
327 |
|
|
328 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
330 |
|
|
331 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
332 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
333 |
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
uint32_t next_block = stride * 8; |
334 |
uint32_t i; |
uint32_t i; |
335 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
336 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
uint8_t *pY_Ref, *pU_Ref, *pV_Ref; /* ref for GMC is _not_ pRef itself */ |
|
|
|
|
337 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
|
VECTOR mv[4]; /* local copy of mvs */ |
|
338 |
|
|
|
if (reduced_resolution) { |
|
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
for (i = 0; i < 4; i++) { |
|
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
|
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
|
|
} |
|
|
} else { |
|
339 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
340 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
341 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
for (i = 0; i < 4; i++) |
|
|
mv[i] = pMB->mvs[i]; |
|
|
} |
|
|
|
|
|
if (mcsel) { |
|
|
mv[0].x = mv[0].y = mv[1].x = mv[1].y = mv[2].x = mv[2].y = mv[3].x = mv[3].y = 0; |
|
|
/* position in ref is same as the block, set vector to (0,0) */ |
|
|
pY_Ref = dec->gmc.y; |
|
|
pU_Ref = dec->gmc.u; |
|
|
pV_Ref = dec->gmc.v; |
|
|
/* but reference itself isn't. It's warped... */ |
|
|
/* Btw., this is too slow! For GMC it should simply be transfer_16to8add() */ |
|
|
} else { |
|
|
pY_Ref = dec->refn[0].y; |
|
|
pU_Ref = dec->refn[0].u; |
|
|
pV_Ref = dec->refn[0].v; |
|
|
} |
|
342 |
|
|
343 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
344 |
|
uv_dx = pMB->mvs[0].x; |
345 |
|
uv_dy = pMB->mvs[0].y; |
346 |
|
|
347 |
uv_dx = mv[0].x / (1 + dec->quarterpel); |
if (dec->quarterpel) |
|
uv_dy = mv[0].y / (1 + dec->quarterpel); |
|
|
|
|
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
|
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
|
|
|
|
start_timer(); |
|
|
if (reduced_resolution) |
|
|
{ |
|
|
interpolate32x32_switch(dec->cur.y, pY_Ref , 32*x_pos, 32*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.u, pU_Ref , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate16x16_switch(dec->cur.v, pV_Ref , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
|
|
|
} |
|
|
else |
|
348 |
{ |
{ |
349 |
if(dec->quarterpel) { |
uv_dx = (uv_dx >> 1) | (uv_dx & 1); |
350 |
interpolate16x16_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
uv_dy = (uv_dy >> 1) | (uv_dy & 1); |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
351 |
} |
} |
|
else { |
|
|
interpolate16x16_switch(dec->cur.y, pY_Ref , 16*x_pos, 16*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
} |
|
|
|
|
|
interpolate8x8_switch(dec->cur.u, pU_Ref , 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate8x8_switch(dec->cur.v, pV_Ref , 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
} |
|
|
stop_comp_timer(); |
|
352 |
|
|
353 |
} else { /* MODE_INTER4V */ |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
354 |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
355 |
|
} else { |
356 |
int sum; |
int sum; |
357 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
358 |
|
|
359 |
if(dec->quarterpel) |
if(dec->quarterpel) |
360 |
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
{ |
361 |
else |
sum /= 2; |
362 |
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
} |
363 |
|
|
364 |
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
365 |
|
|
366 |
if(dec->quarterpel) |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
|
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
|
|
else |
|
|
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
|
|
|
|
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
367 |
|
|
368 |
start_timer(); |
if (dec->quarterpel) |
|
if (reduced_resolution) |
|
369 |
{ |
{ |
370 |
interpolate16x16_switch(dec->cur.y, pY_Ref , 32*x_pos, 32*y_pos, |
sum /= 2; |
371 |
mv[0].x, mv[0].y, stride, rounding); |
} |
|
interpolate16x16_switch(dec->cur.y, pY_Ref , 32*x_pos + 16, 32*y_pos, |
|
|
mv[1].x, mv[1].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, pY_Ref , 32*x_pos, 32*y_pos + 16, |
|
|
mv[2].x, mv[2].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, pY_Ref , 32*x_pos + 16, 32*y_pos + 16, |
|
|
mv[3].x, mv[3].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.u, pU_Ref , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate16x16_switch(dec->cur.v, pV_Ref , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
372 |
|
|
373 |
// set_block(pY_Cur, stride, 32, 32, 127); |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
374 |
} |
} |
375 |
else |
|
376 |
{ |
start_timer(); |
377 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
378 |
interpolate8x8_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
DPRINTF(DPRINTF_DEBUG, "QUARTERPEL\n"); |
379 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
380 |
mv[0].x, mv[0].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
381 |
interpolate8x8_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
382 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
383 |
mv[1].x, mv[1].y, stride, rounding); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
384 |
interpolate8x8_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
385 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
386 |
mv[2].x, mv[2].y, stride, rounding); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
|
interpolate8x8_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
mv[3].x, mv[3].y, stride, rounding); |
|
387 |
} |
} |
388 |
else { |
else { |
389 |
interpolate8x8_switch(dec->cur.y, pY_Ref , 16*x_pos, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
390 |
mv[0].x, mv[0].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
391 |
interpolate8x8_switch(dec->cur.y, pY_Ref , 16*x_pos + 8, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
392 |
mv[1].x, mv[1].y, stride, rounding); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
393 |
interpolate8x8_switch(dec->cur.y, pY_Ref , 16*x_pos, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
394 |
mv[2].x, mv[2].y, stride, rounding); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
395 |
interpolate8x8_switch(dec->cur.y, pY_Ref , 16*x_pos + 8, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
396 |
mv[3].x, mv[3].y, stride, rounding); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
397 |
} |
} |
398 |
|
|
399 |
interpolate8x8_switch(dec->cur.u, pU_Ref , 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
400 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
401 |
interpolate8x8_switch(dec->cur.v, pV_Ref , 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
402 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
|
} |
|
403 |
stop_comp_timer(); |
stop_comp_timer(); |
|
} |
|
404 |
|
|
405 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
406 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
if (cbp & (1 << (5 - i))) /* coded */ |
|
|
|
|
if (cbp & (1 << (5 - i))) // coded |
|
407 |
{ |
{ |
408 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
409 |
|
|
410 |
start_timer(); |
start_timer(); |
411 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
412 |
stop_coding_timer(); |
stop_coding_timer(); |
413 |
|
|
414 |
start_timer(); |
start_timer(); |
431 |
} |
} |
432 |
|
|
433 |
start_timer(); |
start_timer(); |
|
if (reduced_resolution) |
|
|
{ |
|
|
if (cbp & 32) |
|
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
} |
|
|
else |
|
|
{ |
|
434 |
if (cbp & 32) |
if (cbp & 32) |
435 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
436 |
if (cbp & 16) |
if (cbp & 16) |
443 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
444 |
if (cbp & 1) |
if (cbp & 1) |
445 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
446 |
stop_transfer_timer(); |
stop_transfer_timer(); |
447 |
} |
} |
448 |
|
|
450 |
void |
void |
451 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
452 |
Bitstream * bs, |
Bitstream * bs, |
|
int reduced_resolution, |
|
453 |
int quant, |
int quant, |
454 |
int intra_dc_threshold) |
int intra_dc_threshold) |
455 |
{ |
{ |
456 |
uint32_t bound; |
uint32_t bound; |
457 |
uint32_t x, y; |
uint32_t x, y; |
|
uint32_t mb_width = dec->mb_width; |
|
|
uint32_t mb_height = dec->mb_height; |
|
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
458 |
|
|
459 |
bound = 0; |
bound = 0; |
460 |
|
|
461 |
for (y = 0; y < mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
462 |
for (x = 0; x < mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
463 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
464 |
uint32_t mcbpc; |
uint32_t mcbpc; |
465 |
uint32_t cbpc; |
uint32_t cbpc; |
472 |
|
|
473 |
if (check_resync_marker(bs, 0)) |
if (check_resync_marker(bs, 0)) |
474 |
{ |
{ |
475 |
bound = read_video_packet_header(bs, dec, 0, |
bound = read_video_packet_header(bs, 0, &quant); |
476 |
&quant, NULL, NULL, &intra_dc_threshold); |
x = bound % dec->mb_width; |
477 |
x = bound % mb_width; |
y = bound / dec->mb_width; |
|
y = bound / mb_width; |
|
478 |
} |
} |
479 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
480 |
|
|
505 |
|
|
506 |
if (dec->interlacing) { |
if (dec->interlacing) { |
507 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
508 |
DPRINTF(DPRINTF_MB,"deci: field_dct: %i", mb->field_dct); |
DPRINTF(DPRINTF_DEBUG, "deci: field_dct: %d", mb->field_dct); |
509 |
} |
} |
510 |
|
|
511 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
512 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
|
|
|
513 |
} |
} |
514 |
if(dec->out_frm) |
if(dec->out_frm) |
515 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width); |
516 |
|
|
517 |
} |
} |
518 |
|
|
519 |
} |
} |
525 |
int x, |
int x, |
526 |
int y, |
int y, |
527 |
int k, |
int k, |
528 |
VECTOR * ret_mv, |
VECTOR * mv, |
529 |
int fcode, |
int fcode, |
530 |
const int bound) |
const int bound) |
531 |
{ |
{ |
536 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
537 |
|
|
538 |
VECTOR pmv; |
VECTOR pmv; |
539 |
VECTOR mv; |
int mv_x, mv_y; |
540 |
|
|
541 |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
542 |
|
|
543 |
mv.x = get_mv(bs, fcode); |
mv_x = get_mv(bs, fcode); |
544 |
mv.y = get_mv(bs, fcode); |
mv_y = get_mv(bs, fcode); |
|
|
|
|
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y); |
|
545 |
|
|
546 |
mv.x += pmv.x; |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y); |
|
mv.y += pmv.y; |
|
547 |
|
|
548 |
if (mv.x < low) { |
mv_x += pmv.x; |
549 |
mv.x += range; |
mv_y += pmv.y; |
|
} else if (mv.x > high) { |
|
|
mv.x -= range; |
|
|
} |
|
550 |
|
|
551 |
if (mv.y < low) { |
if (mv_x < low) { |
552 |
mv.y += range; |
mv_x += range; |
553 |
} else if (mv.y > high) { |
} else if (mv_x > high) { |
554 |
mv.y -= range; |
mv_x -= range; |
555 |
} |
} |
556 |
|
|
557 |
ret_mv->x = mv.x; |
if (mv_y < low) { |
558 |
ret_mv->y = mv.y; |
mv_y += range; |
559 |
|
} else if (mv_y > high) { |
560 |
|
mv_y -= range; |
561 |
} |
} |
562 |
|
|
563 |
|
mv->x = mv_x; |
564 |
|
mv->y = mv_y; |
565 |
|
|
|
|
|
|
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
|
|
{ |
|
|
int length = 1 << (fcode+4); |
|
|
|
|
|
if (quarterpel) value *= 2; |
|
|
|
|
|
if (value < -length) |
|
|
return -length; |
|
|
else if (value >= length) |
|
|
return length-1; |
|
|
else return value; |
|
566 |
} |
} |
567 |
|
|
568 |
|
|
|
/* for P_VOP set gmc_warp to NULL */ |
|
569 |
void |
void |
570 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
571 |
Bitstream * bs, |
Bitstream * bs, |
572 |
int rounding, |
int rounding, |
|
int reduced_resolution, |
|
573 |
int quant, |
int quant, |
574 |
int fcode, |
int fcode, |
575 |
int intra_dc_threshold, |
int intra_dc_threshold) |
|
const WARPPOINTS *const gmc_warp) |
|
576 |
{ |
{ |
577 |
|
|
578 |
uint32_t x, y; |
uint32_t x, y; |
579 |
uint32_t bound; |
uint32_t bound; |
580 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
|
uint32_t mb_width = dec->mb_width; |
|
|
uint32_t mb_height = dec->mb_height; |
|
|
|
|
|
static int framecount=0; |
|
|
if (reduced_resolution) |
|
|
{ |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
581 |
|
|
582 |
start_timer(); |
start_timer(); |
583 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
584 |
dec->width, dec->height); |
dec->width, dec->height); |
585 |
stop_edges_timer(); |
stop_edges_timer(); |
586 |
|
|
|
if (gmc_warp) |
|
|
{ |
|
|
char filename[80]; |
|
|
sprintf(filename,"dGMC%05d.pgm",framecount); |
|
|
// accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 |
|
|
if ( (dec->sprite_warping_accuracy != 3) || (dec->sprite_warping_points != 2) ) |
|
|
{ |
|
|
fprintf(stderr,"Wrong GMC parameters acc=%d(-> 1/%d), %d!!!\n", |
|
|
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
|
|
dec->sprite_warping_points); |
|
|
} |
|
|
|
|
|
generate_GMCparameters( dec->sprite_warping_points, |
|
|
(2 << dec->sprite_warping_accuracy), gmc_warp, |
|
|
dec->width, dec->height, &dec->gmc_data); |
|
|
|
|
|
generate_GMCimage(&dec->gmc_data, &dec->refn[0], |
|
|
mb_width, mb_height, |
|
|
dec->edged_width, dec->edged_width/2, |
|
|
fcode, 0, 0, |
|
|
rounding, dec->mbs, &dec->gmc); |
|
|
|
|
|
/* |
|
|
sprintf(filename,"dGMC%05d.pgm",framecount); |
|
|
image_dump_yuvpgm(&dec->gmc, |
|
|
dec->edged_width, dec->width, dec->height, filename); |
|
|
|
|
|
sprintf(filename,"dREF%05d.pgm",framecount); |
|
|
image_dump_yuvpgm(&dec->refn[0], |
|
|
dec->edged_width, dec->width, dec->height, filename); |
|
|
sprintf(filename,"dCUR%05d.pgm",framecount); |
|
|
image_dump_yuvpgm(&dec->cur, |
|
|
dec->edged_width, dec->width, dec->height, filename); |
|
|
framecount++; |
|
|
*/ |
|
|
} |
|
|
|
|
587 |
bound = 0; |
bound = 0; |
588 |
|
|
589 |
for (y = 0; y < mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
590 |
cp_mb = st_mb = 0; |
cp_mb = st_mb = 0; |
591 |
for (x = 0; x < mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
592 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
593 |
|
|
594 |
// skip stuffing |
/* skip stuffing */ |
595 |
while (BitstreamShowBits(bs, 10) == 1) |
while (BitstreamShowBits(bs, 10) == 1) |
596 |
BitstreamSkip(bs, 10); |
BitstreamSkip(bs, 10); |
597 |
|
|
598 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) |
599 |
{ |
{ |
600 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
601 |
&quant, &fcode, NULL, &intra_dc_threshold); |
x = bound % dec->mb_width; |
602 |
x = bound % mb_width; |
y = bound / dec->mb_width; |
|
y = bound / mb_width; |
|
603 |
} |
} |
604 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
605 |
|
|
606 |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
607 |
|
|
608 |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
/*if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) not_coded */ |
609 |
if (!(BitstreamGetBit(bs))) // block _is_ coded |
if (!(BitstreamGetBit(bs))) /* not_coded */ |
610 |
{ |
{ |
611 |
uint32_t mcbpc; |
uint32_t mcbpc; |
612 |
uint32_t cbpc; |
uint32_t cbpc; |
614 |
uint32_t cbpy; |
uint32_t cbpy; |
615 |
uint32_t cbp; |
uint32_t cbp; |
616 |
uint32_t intra; |
uint32_t intra; |
|
int mcsel = 0; // mcsel: '0'=local motion, '1'=GMC |
|
617 |
|
|
618 |
cp_mb++; |
cp_mb++; |
619 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
630 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
631 |
} |
} |
632 |
|
|
|
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
|
|
{ |
|
|
mcsel = BitstreamGetBit(bs); |
|
|
} |
|
|
|
|
633 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
634 |
DPRINTF(DPRINTF_MB, "cbpy %i mcsel %i ", cbpy,mcsel); |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
635 |
|
|
636 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
637 |
|
|
651 |
if (dec->interlacing) { |
if (dec->interlacing) { |
652 |
if (cbp || intra) { |
if (cbp || intra) { |
653 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
654 |
DPRINTF(DPRINTF_MB,"decp: field_dct: %i", mb->field_dct); |
DPRINTF(DPRINTF_DEBUG, "decp: field_dct: %d", mb->field_dct); |
655 |
} |
} |
656 |
|
|
657 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
658 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
659 |
DPRINTF(DPRINTF_MB, "decp: field_pred: %i", mb->field_pred); |
DPRINTF(DPRINTF_DEBUG, "decp: field_pred: %d", mb->field_pred); |
660 |
|
|
661 |
if (mb->field_pred) { |
if (mb->field_pred) { |
662 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
663 |
DPRINTF(DPRINTF_MB,"decp: field_for_top: %i", mb->field_for_top); |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_top: %d", mb->field_for_top); |
664 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
665 |
DPRINTF(DPRINTF_MB,"decp: field_for_bot: %i", mb->field_for_bot); |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_bot: %d", mb->field_for_bot); |
666 |
} |
} |
667 |
} |
} |
668 |
} |
} |
669 |
|
|
670 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
671 |
|
if (dec->interlacing && mb->field_pred) { |
|
if (mcsel) |
|
|
{ |
|
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->amv; |
|
|
/* already clipped to fcode */ |
|
|
|
|
|
} else if (dec->interlacing && mb->field_pred) { |
|
672 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
673 |
fcode, bound); |
fcode, bound); |
674 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
676 |
} else { |
} else { |
677 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
678 |
fcode, bound); |
fcode, bound); |
679 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
680 |
|
mb->mvs[0].x; |
681 |
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
682 |
|
mb->mvs[0].y; |
683 |
} |
} |
684 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
685 |
|
|
687 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
688 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
689 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
690 |
} else // MODE_INTRA, MODE_INTRA_Q |
} else /* MODE_INTRA, MODE_INTRA_Q */ |
691 |
{ |
{ |
692 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
693 |
0; |
0; |
694 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
695 |
0; |
0; |
696 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
697 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
698 |
continue; |
continue; |
699 |
} |
} |
700 |
|
|
701 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
702 |
rounding, reduced_resolution, mcsel); |
rounding); |
703 |
|
} else /* not coded */ |
|
} |
|
|
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
|
704 |
{ |
{ |
705 |
mb->mode = MODE_NOT_CODED_GMC; |
DPRINTF(DPRINTF_DEBUG, "P-frame MB at (X,Y)=(%d,%d)", x, y); |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->amv; |
|
|
|
|
|
start_timer(); |
|
706 |
|
|
|
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
|
|
dec->gmc.y + (16*y)*dec->edged_width + (16*x), |
|
|
dec->edged_width); |
|
|
|
|
|
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->gmc.u + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->edged_width/2); |
|
|
|
|
|
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->gmc.v + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->edged_width/2); |
|
|
|
|
|
stop_transfer_timer(); |
|
|
|
|
|
if(dec->out_frm && cp_mb > 0) { |
|
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
|
|
cp_mb = 0; |
|
|
} |
|
|
st_mb = x+1; |
|
|
} |
|
|
else /* not coded P_VOP macroblock */ |
|
|
{ |
|
707 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
|
|
|
708 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
709 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
|
// copy macroblock directly from ref to cur |
|
710 |
|
|
711 |
start_timer(); |
/* copy macroblock directly from ref to cur */ |
712 |
|
|
713 |
if (reduced_resolution) |
start_timer(); |
|
{ |
|
|
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
|
|
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
|
|
dec->edged_width); |
|
|
|
|
|
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->edged_width/2); |
|
|
|
|
|
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->edged_width/2); |
|
|
} |
|
|
else |
|
|
{ |
|
|
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
|
|
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
|
|
dec->edged_width); |
|
714 |
|
|
715 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
716 |
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
(16 * x), |
717 |
|
dec->refn[0].y + (16 * y) * dec->edged_width + |
718 |
|
(16 * x), dec->edged_width); |
719 |
|
|
720 |
|
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
721 |
|
(16 * x + 8), |
722 |
|
dec->refn[0].y + (16 * y) * dec->edged_width + |
723 |
|
(16 * x + 8), dec->edged_width); |
724 |
|
|
725 |
|
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
726 |
|
(16 * x), |
727 |
|
dec->refn[0].y + (16 * y + |
728 |
|
8) * dec->edged_width + |
729 |
|
(16 * x), dec->edged_width); |
730 |
|
|
731 |
|
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
732 |
|
(16 * x + 8), |
733 |
|
dec->refn[0].y + (16 * y + |
734 |
|
8) * dec->edged_width + |
735 |
|
(16 * x + 8), dec->edged_width); |
736 |
|
|
737 |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
738 |
|
(8 * x), |
739 |
|
dec->refn[0].u + |
740 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
741 |
dec->edged_width/2); |
dec->edged_width/2); |
742 |
|
|
743 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
744 |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
(8 * x), |
745 |
|
dec->refn[0].v + |
746 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
747 |
dec->edged_width/2); |
dec->edged_width/2); |
|
} |
|
|
|
|
748 |
stop_transfer_timer(); |
stop_transfer_timer(); |
|
|
|
749 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
750 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
751 |
cp_mb = 0; |
cp_mb = 0; |
759 |
} |
} |
760 |
|
|
761 |
|
|
762 |
// add by MinChen <chenm001@163.com> |
/* add by MinChen <chenm001@163.com> */ |
763 |
// decode B-frame motion vector |
/* decode B-frame motion vector */ |
764 |
void |
void |
765 |
get_b_motion_vector(DECODER * dec, |
get_b_motion_vector(DECODER * dec, |
766 |
Bitstream * bs, |
Bitstream * bs, |
804 |
} |
} |
805 |
|
|
806 |
|
|
807 |
// add by MinChen <chenm001@163.com> |
/* add by MinChen <chenm001@163.com> */ |
808 |
// decode an B-frame forward & backward inter macroblock |
/* decode an B-frame forward & backward inter macroblock */ |
809 |
void |
void |
810 |
decoder_bf_mbinter(DECODER * dec, |
decoder_bf_mbinter(DECODER * dec, |
811 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
837 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
838 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
839 |
|
|
840 |
if (dec->quarterpel) |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
841 |
{ |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
uv_dx /= 2; |
|
|
uv_dy /= 2; |
|
|
} |
|
|
|
|
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
|
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
842 |
} else { |
} else { |
843 |
int sum; |
int sum; |
844 |
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
|
|
else |
|
845 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
846 |
|
uv_dx = |
847 |
|
(sum == |
848 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
849 |
|
(ABS(sum) / 16) * 2)); |
850 |
|
|
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
|
|
else |
|
851 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
852 |
|
uv_dy = |
853 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
854 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
855 |
|
(ABS(sum) / 16) * 2)); |
856 |
} |
} |
857 |
|
|
858 |
start_timer(); |
start_timer(); |
|
if(dec->quarterpel) { |
|
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
} |
|
|
else { |
|
859 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
860 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
861 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
862 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
863 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, |
864 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
865 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
0); |
866 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
867 |
} |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
868 |
|
0); |
869 |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
870 |
uv_dx, uv_dy, stride2, 0); |
uv_dx, uv_dy, stride2, 0); |
871 |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
873 |
stop_comp_timer(); |
stop_comp_timer(); |
874 |
|
|
875 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
876 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
if (cbp & (1 << (5 - i))) /* coded */ |
|
|
|
|
if (cbp & (1 << (5 - i))) // coded |
|
877 |
{ |
{ |
878 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
879 |
|
|
880 |
start_timer(); |
start_timer(); |
881 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
882 |
stop_coding_timer(); |
stop_coding_timer(); |
883 |
|
|
884 |
start_timer(); |
start_timer(); |
916 |
stop_transfer_timer(); |
stop_transfer_timer(); |
917 |
} |
} |
918 |
|
|
919 |
// add by MinChen <chenm001@163.com> |
|
920 |
// decode an B-frame direct & inter macroblock |
/* add by MinChen <chenm001@163.com> */ |
921 |
|
/* decode an B-frame direct & inter macroblock */ |
922 |
void |
void |
923 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
924 |
IMAGE forward, |
IMAGE forward, |
951 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
952 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
953 |
|
|
954 |
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
955 |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
956 |
|
|
957 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
958 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
959 |
|
|
960 |
if (dec->quarterpel) |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
961 |
{ |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
uv_dx /= 2; |
|
|
uv_dy /= 2; |
|
|
|
|
|
b_uv_dx /= 2; |
|
|
b_uv_dy /= 2; |
|
|
} |
|
|
|
|
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
|
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
|
|
|
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
|
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
|
962 |
} else { |
} else { |
963 |
int sum; |
int sum; |
964 |
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
|
|
else |
|
965 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
966 |
|
uv_dx = |
967 |
|
(sum == |
968 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
969 |
|
(ABS(sum) / 16) * 2)); |
970 |
|
|
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
|
|
else |
|
971 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
972 |
|
uv_dy = |
973 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
974 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
975 |
|
(ABS(sum) / 16) * 2)); |
976 |
if(dec->quarterpel) |
|
977 |
sum = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
sum = |
978 |
else |
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
979 |
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
pMB->b_mvs[3].x; |
980 |
|
b_uv_dx = |
981 |
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
982 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
983 |
if(dec->quarterpel) |
(ABS(sum) / 16) * 2)); |
984 |
sum = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
|
985 |
else |
sum = |
986 |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
987 |
|
pMB->b_mvs[3].y; |
988 |
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
b_uv_dy = |
989 |
|
(sum == |
990 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
991 |
|
(ABS(sum) / 16) * 2)); |
992 |
} |
} |
993 |
|
|
994 |
|
|
995 |
start_timer(); |
start_timer(); |
|
if(dec->quarterpel) { |
|
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
else { |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
|
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
|
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
|
|
} |
|
|
} |
|
|
else { |
|
996 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
997 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
998 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1002 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1003 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1004 |
0); |
0); |
|
} |
|
|
|
|
1005 |
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1006 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1007 |
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1008 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1009 |
|
|
1010 |
|
|
1011 |
if(dec->quarterpel) { |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
1012 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1013 |
else { |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
|
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
|
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
|
|
} |
|
|
} |
|
|
else { |
|
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
|
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
|
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
|
1014 |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
1015 |
0); |
0); |
1016 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, |
1017 |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1018 |
stride, 0); |
stride, 0); |
1019 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1020 |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1021 |
stride, 0); |
stride, 0); |
1022 |
} |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
|
|
|
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
|
1023 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1024 |
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1025 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1026 |
|
|
1027 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos, |
1028 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
stride); |
1029 |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, |
1030 |
stride, 1, 8); |
stride); |
1031 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, |
1032 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
stride); |
1033 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, |
1034 |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
16 * y_pos + 8, stride); |
1035 |
stride, 1, 8); |
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, |
1036 |
|
stride2); |
1037 |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos, |
1038 |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
stride2); |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
1039 |
stop_comp_timer(); |
stop_comp_timer(); |
1040 |
|
|
1041 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
1042 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
if (cbp & (1 << (5 - i))) /* coded */ |
|
|
|
|
if (cbp & (1 << (5 - i))) // coded |
|
1043 |
{ |
{ |
1044 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
1045 |
|
|
1046 |
start_timer(); |
start_timer(); |
1047 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
1048 |
stop_coding_timer(); |
stop_coding_timer(); |
1049 |
|
|
1050 |
start_timer(); |
start_timer(); |
1083 |
} |
} |
1084 |
|
|
1085 |
|
|
1086 |
// add by MinChen <chenm001@163.com> |
/* add by MinChen <chenm001@163.com> */ |
1087 |
// for decode B-frame dbquant |
/* for decode B-frame dbquant */ |
1088 |
int32_t __inline |
int32_t __inline |
1089 |
get_dbquant(Bitstream * bs) |
get_dbquant(Bitstream * bs) |
1090 |
{ |
{ |
1091 |
if (!BitstreamGetBit(bs)) // '0' |
if (!BitstreamGetBit(bs)) /* '0' */ |
1092 |
return (0); |
return (0); |
1093 |
else if (!BitstreamGetBit(bs)) // '10' |
else if (!BitstreamGetBit(bs)) /* '10' */ |
1094 |
return (-2); |
return (-2); |
1095 |
else |
else |
1096 |
return (2); // '11' |
return (2); /* '11' */ |
1097 |
} |
} |
1098 |
|
|
1099 |
// add by MinChen <chenm001@163.com> |
/* add by MinChen <chenm001@163.com> */ |
1100 |
// for decode B-frame mb_type |
/* for decode B-frame mb_type */ |
1101 |
// bit ret_value |
/* bit ret_value */ |
1102 |
// 1 0 |
/* 1 0 */ |
1103 |
// 01 1 |
/* 01 1 */ |
1104 |
// 001 2 |
/* 001 2 */ |
1105 |
// 0001 3 |
/* 0001 3 */ |
1106 |
int32_t __inline |
int32_t __inline |
1107 |
get_mbtype(Bitstream * bs) |
get_mbtype(Bitstream * bs) |
1108 |
{ |
{ |
1151 |
#endif |
#endif |
1152 |
|
|
1153 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1154 |
// Initialize Pred Motion Vector |
/* Initialize Pred Motion Vector */ |
1155 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1156 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1157 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1161 |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1162 |
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1163 |
|
|
1164 |
// skip if the co-located P_VOP macroblock is not coded |
/* the last P_VOP is skip macroblock ? */ |
|
// note: gmc+not_coded isn't skipped |
|
|
|
|
1165 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1166 |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
/*DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); */ |
1167 |
mb->cbp = 0; |
mb->cbp = 0; |
1168 |
#ifdef BFRAMES_DEC_DEBUG |
#ifdef BFRAMES_DEC_DEBUG |
1169 |
mb->mb_type = MODE_NOT_CODED; |
mb->mb_type = MODE_NOT_CODED; |
1171 |
#endif |
#endif |
1172 |
mb->mb_type = MODE_FORWARD; |
mb->mb_type = MODE_FORWARD; |
1173 |
mb->quant = last_mb->quant; |
mb->quant = last_mb->quant; |
1174 |
//mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
/*mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; */ |
1175 |
//mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
/*mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; */ |
1176 |
|
|
1177 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1178 |
continue; |
continue; |
1179 |
} |
} |
1180 |
|
|
1181 |
if (!BitstreamGetBit(bs)) { // modb=='0' |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1182 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = (uint8_t)BitstreamGetBit(bs); |
1183 |
|
|
1184 |
mb->mb_type = get_mbtype(bs); |
mb->mb_type = get_mbtype(bs); |
1185 |
|
|
1186 |
if (!modb2) { // modb=='00' |
if (!modb2) { /* modb=='00' */ |
1187 |
mb->cbp = BitstreamGetBits(bs, 6); |
mb->cbp = BitstreamGetBits(bs, 6); |
1188 |
} else { |
} else { |
1189 |
mb->cbp = 0; |
mb->cbp = 0; |
1204 |
|
|
1205 |
mb->quant = quant; |
mb->quant = quant; |
1206 |
mb->mode = MODE_INTER4V; |
mb->mode = MODE_INTER4V; |
1207 |
//DEBUG1("Switch bm_type=",mb->mb_type); |
/*DEBUG1("Switch bm_type=",mb->mb_type); */ |
1208 |
|
|
1209 |
#ifdef BFRAMES_DEC_DEBUG |
#ifdef BFRAMES_DEC_DEBUG |
1210 |
BFRAME_DEBUG |
BFRAME_DEBUG |
1233 |
/ TRD |
/ TRD |
1234 |
: mb->mvs[i].y - last_mb->mvs[i].y); |
: mb->mvs[i].y - last_mb->mvs[i].y); |
1235 |
} |
} |
1236 |
//DEBUG("B-frame Direct!\n"); |
/*DEBUG("B-frame Direct!\n"); */ |
1237 |
} |
} |
1238 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1239 |
mb, x, y, bs); |
mb, x, y, bs); |
1251 |
|
|
1252 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1253 |
mb, x, y, bs); |
mb, x, y, bs); |
1254 |
//DEBUG("B-frame Bidir!\n"); |
/*DEBUG("B-frame Bidir!\n"); */ |
1255 |
break; |
break; |
1256 |
|
|
1257 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1261 |
|
|
1262 |
mb->mode = MODE_INTER; |
mb->mode = MODE_INTER; |
1263 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1264 |
//DEBUG("B-frame Backward!\n"); |
/*DEBUG("B-frame Backward!\n"); */ |
1265 |
break; |
break; |
1266 |
|
|
1267 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1271 |
|
|
1272 |
mb->mode = MODE_INTER; |
mb->mode = MODE_INTER; |
1273 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1274 |
//DEBUG("B-frame Forward!\n"); |
/*DEBUG("B-frame Forward!\n"); */ |
1275 |
break; |
break; |
1276 |
|
|
1277 |
default: |
default: |
1278 |
DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type); |
DPRINTF(DPRINTF_ERROR, "Not support B-frame mb_type = %d", mb->mb_type); |
1279 |
} |
} |
1280 |
|
|
1281 |
} // end of FOR |
} /* end of FOR */ |
1282 |
} |
} |
1283 |
#ifdef BFRAMES_DEC_DEBUG |
#ifdef BFRAMES_DEC_DEBUG |
1284 |
if (!first){ |
if (!first){ |
1289 |
#endif |
#endif |
1290 |
} |
} |
1291 |
|
|
1292 |
// swap two MACROBLOCK array |
/* swap two MACROBLOCK array */ |
1293 |
void |
void |
1294 |
mb_swap(MACROBLOCK ** mb1, |
mb_swap(MACROBLOCK ** mb1, |
1295 |
MACROBLOCK ** mb2) |
MACROBLOCK ** mb2) |
1300 |
*mb2 = temp; |
*mb2 = temp; |
1301 |
} |
} |
1302 |
|
|
|
|
|
|
/* perform post processing if necessary, and output the image */ |
|
|
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
|
|
const XVID_DEC_FRAME * frame, int pp_disable) |
|
|
{ |
|
|
|
|
|
if ((frame->general & (XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV)) && !pp_disable) /* post process */ |
|
|
{ |
|
|
/* note: image is stored to tmp */ |
|
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
|
|
image_deblock_rrv(&dec->tmp, dec->edged_width, |
|
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
|
|
8, frame->general); |
|
|
img = &dec->tmp; |
|
|
} |
|
|
|
|
|
image_output(img, dec->width, dec->height, |
|
|
dec->edged_width, frame->image, frame->stride, |
|
|
frame->colorspace, dec->interlacing); |
|
|
} |
|
|
|
|
|
|
|
1303 |
int |
int |
1304 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1305 |
XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) |
XVID_DEC_FRAME * frame) |
1306 |
{ |
{ |
1307 |
|
|
1308 |
Bitstream bs; |
Bitstream bs; |
1309 |
uint32_t rounding; |
uint32_t rounding; |
|
uint32_t reduced_resolution; |
|
1310 |
uint32_t quant; |
uint32_t quant; |
1311 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1312 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1313 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
|
WARPPOINTS gmc_warp; |
|
1314 |
uint32_t vop_type; |
uint32_t vop_type; |
|
int success = 0; |
|
|
int output = 0; |
|
|
int seen_something = 0; |
|
1315 |
|
|
1316 |
start_global_timer(); |
start_global_timer(); |
1317 |
|
|
|
dec->low_delay_default = (frame->general & XVID_DEC_LOWDELAY); |
|
1318 |
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
1319 |
|
|
|
if ((frame->general & XVID_DEC_DISCONTINUITY)) |
|
|
dec->frames = 0; |
|
|
|
|
|
if (frame->length < 0) /* decoder flush */ |
|
|
{ |
|
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
|
|
we have a reference frame, then outout the reference frame */ |
|
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) |
|
|
{ |
|
|
decoder_output(dec, &dec->refn[0], dec->mbs, frame, dec->last_reduced_resolution); |
|
|
output = 1; |
|
|
} |
|
|
|
|
|
frame->length = 0; |
|
|
if (stats) |
|
|
{ |
|
|
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
|
|
stats->data.vop.time_base = (int)dec->time_base; |
|
|
stats->data.vop.time_increment = 0; //XXX: todo |
|
|
} |
|
|
|
|
|
emms(); |
|
|
|
|
|
stop_global_timer(); |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
1320 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1321 |
|
|
1322 |
// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's |
/* add by chenm001 <chenm001@163.com> */ |
1323 |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
/* for support B-frame to reference last 2 frame */ |
1324 |
{ |
dec->frames++; |
1325 |
if (stats) |
vop_type = |
1326 |
stats->notify = XVID_DEC_VOP; |
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
1327 |
frame->length = 1; |
&fcode_backward, &intra_dc_threshold); |
|
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
|
|
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
|
|
emms(); |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
|
repeat: |
|
|
|
|
|
vop_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
|
|
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
|
|
|
|
|
DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i", |
|
|
vop_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
|
|
|
|
|
if (vop_type == - 1) |
|
|
{ |
|
|
if (success) goto done; |
|
|
emms(); |
|
|
return XVID_ERR_FAIL; |
|
|
} |
|
|
|
|
|
if (vop_type == -2 || vop_type == -3) |
|
|
{ |
|
|
if (vop_type == -3) |
|
|
decoder_resize(dec); |
|
|
|
|
|
if (stats) |
|
|
{ |
|
|
stats->notify = XVID_DEC_VOL; |
|
|
stats->data.vol.general = 0; |
|
|
if (dec->interlacing) |
|
|
stats->data.vol.general |= XVID_INTERLACING; |
|
|
stats->data.vol.width = dec->width; |
|
|
stats->data.vol.height = dec->height; |
|
|
stats->data.vol.aspect_ratio = dec->aspect_ratio; |
|
|
stats->data.vol.par_width = dec->par_width; |
|
|
stats->data.vol.par_height = dec->par_height; |
|
|
frame->length = BitstreamPos(&bs) / 8; |
|
|
emms(); |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
goto repeat; |
|
|
} |
|
1328 |
|
|
1329 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1330 |
|
|
1331 |
|
switch (vop_type) { |
1332 |
|
case P_VOP: |
1333 |
|
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
1334 |
|
intra_dc_threshold); |
1335 |
|
#ifdef BFRAMES_DEC |
1336 |
|
DPRINTF(DPRINTF_DEBUG, "P_VOP Time=%d", dec->time); |
1337 |
|
#endif |
1338 |
|
break; |
1339 |
|
|
|
/* packed_mode: special-N_VOP treament */ |
|
|
if (dec->packed_mode && vop_type == N_VOP) |
|
|
{ |
|
|
if (dec->low_delay_default && dec->frames > 0) |
|
|
{ |
|
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
|
|
output = 1; |
|
|
} |
|
|
/* ignore otherwise */ |
|
|
} |
|
|
else if (vop_type != B_VOP) |
|
|
{ |
|
|
switch(vop_type) |
|
|
{ |
|
1340 |
case I_VOP : |
case I_VOP : |
1341 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1342 |
break; |
#ifdef BFRAMES_DEC |
1343 |
case P_VOP : |
DPRINTF(DPRINTF_DEBUG, "I_VOP Time=%d", dec->time); |
1344 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
#endif |
|
fcode_forward, intra_dc_threshold, NULL); |
|
1345 |
break; |
break; |
1346 |
case S_VOP : |
|
1347 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
case B_VOP: |
1348 |
fcode_forward, intra_dc_threshold, &gmc_warp); |
#ifdef BFRAMES_DEC |
1349 |
|
if (dec->time_pp > dec->time_bp) { |
1350 |
|
DPRINTF(DPRINTF_DEBUG, "B_VOP Time=%d", dec->time); |
1351 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1352 |
|
} else { |
1353 |
|
DPRINTF(DPRINTF_DEBUG, "Broken B_VOP"); |
1354 |
|
} |
1355 |
|
#else |
1356 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1357 |
|
#endif |
1358 |
break; |
break; |
1359 |
case N_VOP : |
|
1360 |
|
case N_VOP: /* vop not coded */ |
1361 |
|
/* when low_delay==0, N_VOP's should interpolate between the past and future frames */ |
1362 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1363 |
break; |
break; |
|
} |
|
1364 |
|
|
1365 |
if (reduced_resolution) |
default: |
1366 |
{ |
return XVID_ERR_FAIL; |
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
|
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
|
|
16, XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV); |
|
1367 |
} |
} |
1368 |
|
|
1369 |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
#ifdef BFRAMES_DEC_DEBUG |
1370 |
if (!(dec->low_delay_default && dec->packed_mode)) |
if (frame->length != BitstreamPos(&bs) / 8){ |
1371 |
{ |
DPRINTF(DPRINTF_DEBUG, "InLen: %d / UseLen: %d", frame->length, BitstreamPos(&bs) / 8); |
|
if (dec->low_delay) |
|
|
{ |
|
|
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
|
|
output = 1; |
|
|
} |
|
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
|
|
{ |
|
|
/* output the reference frame */ |
|
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
|
|
output = 1; |
|
|
} |
|
1372 |
} |
} |
1373 |
|
#endif |
1374 |
|
frame->length = BitstreamPos(&bs) / 8; |
1375 |
|
|
|
image_swap(&dec->refn[0], &dec->refn[1]); |
|
|
image_swap(&dec->cur, &dec->refn[0]); |
|
|
mb_swap(&dec->mbs, &dec->last_mbs); |
|
|
dec->last_reduced_resolution = reduced_resolution; |
|
|
|
|
|
dec->frames++; |
|
|
seen_something = 1; |
|
|
|
|
|
}else{ /* B_VOP */ |
|
1376 |
|
|
1377 |
if (dec->low_delay) |
#ifdef BFRAMES_DEC |
1378 |
{ |
/* test if no B_VOP */ |
1379 |
DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream"); |
if (dec->low_delay || dec->frames == 0) { |
1380 |
dec->low_delay = 1; |
#endif |
1381 |
} |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
1382 |
|
frame->image, frame->stride, frame->colorspace); |
1383 |
|
|
1384 |
if (dec->frames < 2) |
#ifdef BFRAMES_DEC |
|
{ |
|
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
|
|
"broken b-frame, mising ref frames"); |
|
|
}else if (dec->time_pp <= dec->time_bp) { |
|
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
|
|
decoded in vfw. */ |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
|
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
|
1385 |
}else{ |
}else{ |
1386 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
if (dec->frames >= 1) { |
1387 |
} |
start_timer(); |
1388 |
|
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
1389 |
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
image_output(&dec->refn[0], dec->width, dec->height, |
1390 |
output = 1; |
dec->edged_width, frame->image, frame->stride, |
1391 |
dec->frames++; |
frame->colorspace); |
1392 |
} |
} else if (vop_type == B_VOP) { |
1393 |
|
image_output(&dec->cur, dec->width, dec->height, |
1394 |
BitstreamByteAlign(&bs); |
dec->edged_width, frame->image, frame->stride, |
1395 |
|
frame->colorspace); |
|
/* low_delay_default mode: repeat in packed_mode */ |
|
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
|
|
{ |
|
|
success = 1; |
|
|
goto repeat; |
|
|
} |
|
|
|
|
|
done : |
|
|
|
|
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
|
|
then output the recently decoded frame, or print an error message */ |
|
|
if (dec->low_delay_default && output == 0) |
|
|
{ |
|
|
if (dec->packed_mode && seen_something) |
|
|
{ |
|
|
/* output the recently decoded frame */ |
|
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
|
|
output = 1; |
|
1396 |
} |
} |
1397 |
else |
stop_conv_timer(); |
|
{ |
|
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
|
|
"warning: nothing to output"); |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
|
|
"bframe decoder lag"); |
|
|
|
|
|
decoder_output(dec, &dec->cur, NULL, frame, 1 /*disable pp*/); |
|
1398 |
} |
} |
1399 |
} |
} |
1400 |
|
#endif |
1401 |
|
|
1402 |
frame->length = BitstreamPos(&bs) / 8; |
if (vop_type == I_VOP || vop_type == P_VOP) { |
1403 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1404 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1405 |
|
|
1406 |
if (stats) |
/* swap MACROBLOCK */ |
1407 |
{ |
/* the Divx will not set the low_delay flage some times */ |
1408 |
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
/* so follow code will wrong to not swap at that time */ |
1409 |
stats->data.vop.time_base = (int)dec->time_base; |
/* this will broken bitstream! so I'm change it, */ |
1410 |
stats->data.vop.time_increment = 0; //XXX: todo |
/* But that is not the best way! can anyone tell me how */ |
1411 |
|
/* to do another way? */ |
1412 |
|
/* 18-07-2002 MinChen<chenm001@163.com> */ |
1413 |
|
/*if (!dec->low_delay && vop_type == P_VOP) */ |
1414 |
|
if (vop_type == P_VOP) |
1415 |
|
mb_swap(&dec->mbs, &dec->last_mbs); |
1416 |
} |
} |
1417 |
|
|
1418 |
emms(); |
emms(); |