1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder 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 |
* 2002-2003 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 program is free software; you can redistribute it and/or modify |
10 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
20 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
21 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
23 |
*************************************************************************/ |
* $Id: decoder.c,v 1.49.2.9 2003-06-28 15:48:39 chl Exp $ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
|
* |
|
|
* 08.05.2002 add low_delay support for B_VOP decode |
|
|
* MinChen <chenm001@163.com> |
|
|
* 05.05.2002 fix some B-frame decode problem |
|
|
* 02.05.2002 add B-frame decode support(have some problem); |
|
|
* MinChen <chenm001@163.com> |
|
|
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
|
|
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
|
|
* reconstructing blocks, thus artifacts |
|
|
* interlacing speedup - used transfers to re-interlace |
|
|
* 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> |
|
|
* |
|
|
* $Id: decoder.c,v 1.20 2002-06-14 13:21:13 Isibaar Exp $ |
|
24 |
* |
* |
25 |
*************************************************************************/ |
****************************************************************************/ |
26 |
|
|
27 |
|
#include <stdio.h> |
28 |
#include <stdlib.h> |
#include <stdlib.h> |
29 |
#include <string.h> |
#include <string.h> |
30 |
|
|
31 |
|
#ifdef BFRAMES_DEC_DEBUG |
32 |
|
#define BFRAMES_DEC |
33 |
|
#endif |
34 |
|
|
35 |
#include "xvid.h" |
#include "xvid.h" |
36 |
#include "portab.h" |
#include "portab.h" |
37 |
|
#include "global.h" |
38 |
|
|
39 |
#include "decoder.h" |
#include "decoder.h" |
40 |
#include "bitstream/bitstream.h" |
#include "bitstream/bitstream.h" |
46 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
47 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
48 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
49 |
|
#include "image/reduced.h" |
50 |
|
#include "image/font.h" |
51 |
|
|
52 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
53 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
54 |
#include "utils/timer.h" |
#include "utils/timer.h" |
55 |
#include "utils/emms.h" |
#include "utils/emms.h" |
56 |
|
#include "motion/motion.h" |
57 |
|
#include "motion/gmc.h" |
58 |
|
|
59 |
#include "image/image.h" |
#include "image/image.h" |
60 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
61 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
62 |
|
|
63 |
int |
int |
64 |
decoder_create(XVID_DEC_PARAM * param) |
decoder_resize(DECODER * dec) |
65 |
{ |
{ |
66 |
DECODER *dec; |
/* free existing */ |
67 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
68 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
69 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
70 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
71 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
72 |
|
|
73 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
|
if (dec == NULL) { |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
param->handle = dec; |
|
74 |
|
|
75 |
dec->width = param->width; |
if (dec->last_mbs) |
76 |
dec->height = param->height; |
xvid_free(dec->last_mbs); |
77 |
|
if (dec->mbs) |
78 |
|
xvid_free(dec->mbs); |
79 |
|
|
80 |
|
/* realloc */ |
81 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
82 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
83 |
|
|
84 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
85 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
|
dec->low_delay = 0; |
|
86 |
|
|
87 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
88 |
xvid_free(dec); |
xvid_free(dec); |
94 |
xvid_free(dec); |
xvid_free(dec); |
95 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
96 |
} |
} |
97 |
// add by chenm001 <chenm001@163.com> |
|
98 |
// for support B-frame to reference last 2 frame |
/* Support B-frame to reference last 2 frame */ |
99 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
100 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
101 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
102 |
xvid_free(dec); |
xvid_free(dec); |
103 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
104 |
} |
} |
105 |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
106 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
107 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
108 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
110 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
111 |
} |
} |
112 |
|
|
113 |
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
114 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
115 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
116 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
117 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
118 |
|
xvid_free(dec); |
119 |
|
return XVID_ERR_MEMORY; |
120 |
|
} |
121 |
|
|
122 |
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
123 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
124 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
125 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
126 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
127 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
128 |
|
xvid_free(dec); |
129 |
|
return XVID_ERR_MEMORY; |
130 |
|
} |
131 |
|
|
132 |
dec->mbs = |
dec->mbs = |
133 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
134 |
CACHE_LINE); |
CACHE_LINE); |
136 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
137 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
138 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
139 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
140 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
141 |
xvid_free(dec); |
xvid_free(dec); |
142 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
143 |
} |
} |
|
|
|
144 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
145 |
|
|
146 |
// add by chenm001 <chenm001@163.com> |
/* For skip MB flag */ |
|
// for skip MB flag |
|
147 |
dec->last_mbs = |
dec->last_mbs = |
148 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
149 |
CACHE_LINE); |
CACHE_LINE); |
152 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
153 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
154 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
155 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
156 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
157 |
xvid_free(dec); |
xvid_free(dec); |
158 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
159 |
} |
} |
160 |
|
|
161 |
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); |
162 |
|
|
163 |
|
return 0; |
164 |
|
} |
165 |
|
|
166 |
|
|
167 |
|
int |
168 |
|
decoder_create(xvid_dec_create_t * create) |
169 |
|
{ |
170 |
|
DECODER *dec; |
171 |
|
|
172 |
|
if (XVID_MAJOR(create->version) != 1) /* v1.x.x */ |
173 |
|
return XVID_ERR_VERSION; |
174 |
|
|
175 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
176 |
|
if (dec == NULL) { |
177 |
|
return XVID_ERR_MEMORY; |
178 |
|
} |
179 |
|
memset(dec, 0, sizeof(DECODER)); |
180 |
|
|
181 |
|
create->handle = dec; |
182 |
|
|
183 |
|
dec->width = create->width; |
184 |
|
dec->height = create->height; |
185 |
|
|
186 |
|
image_null(&dec->cur); |
187 |
|
image_null(&dec->refn[0]); |
188 |
|
image_null(&dec->refn[1]); |
189 |
|
image_null(&dec->tmp); |
190 |
|
image_null(&dec->qtmp); |
191 |
|
|
192 |
|
/* image based GMC */ |
193 |
|
image_null(&dec->gmc); |
194 |
|
|
195 |
|
|
196 |
|
dec->mbs = NULL; |
197 |
|
dec->last_mbs = NULL; |
198 |
|
|
199 |
init_timer(); |
init_timer(); |
200 |
|
|
201 |
// add by chenm001 <chenm001@163.com> |
/* For B-frame support (used to save reference frame's time */ |
202 |
// for support B-frame to save reference frame's time |
dec->frames = 0; |
|
dec->frames = -1; |
|
203 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
204 |
|
dec->low_delay = 0; |
205 |
|
dec->packed_mode = 0; |
206 |
|
|
207 |
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
208 |
|
|
209 |
return XVID_ERR_OK; |
if (dec->fixed_dimensions) |
210 |
|
return decoder_resize(dec); |
211 |
|
else |
212 |
|
return 0; |
213 |
} |
} |
214 |
|
|
215 |
|
|
218 |
{ |
{ |
219 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
220 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
221 |
|
|
222 |
|
/* image based GMC */ |
223 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
224 |
|
|
225 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
226 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
227 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
228 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
229 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
230 |
xvid_free(dec); |
xvid_free(dec); |
231 |
|
|
232 |
write_timer(); |
write_timer(); |
233 |
return XVID_ERR_OK; |
return 0; |
234 |
} |
} |
235 |
|
|
236 |
|
|
242 |
|
|
243 |
|
|
244 |
|
|
245 |
// decode an intra macroblock |
/* decode an intra macroblock */ |
|
|
|
246 |
void |
void |
247 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
248 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
252 |
const uint32_t cbp, |
const uint32_t cbp, |
253 |
Bitstream * bs, |
Bitstream * bs, |
254 |
const uint32_t quant, |
const uint32_t quant, |
255 |
const uint32_t intra_dc_threshold) |
const uint32_t intra_dc_threshold, |
256 |
|
const unsigned int bound, |
257 |
|
const int reduced_resolution) |
258 |
{ |
{ |
259 |
|
|
260 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
267 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
268 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
269 |
|
|
270 |
|
if (reduced_resolution) { |
271 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
272 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
273 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
274 |
|
}else{ |
275 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
276 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
277 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
278 |
|
} |
279 |
|
|
280 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
281 |
|
|
282 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
283 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
286 |
|
|
287 |
start_timer(); |
start_timer(); |
288 |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
289 |
iQuant, iDcScaler, predictors); |
iQuant, iDcScaler, predictors, bound); |
290 |
if (!acpred_flag) { |
if (!acpred_flag) { |
291 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
292 |
} |
} |
300 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
301 |
|
|
302 |
if (dc_size > 8) { |
if (dc_size > 8) { |
303 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); /* marker */ |
304 |
} |
} |
305 |
|
|
306 |
block[i * 64 + 0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
307 |
start_coeff = 1; |
start_coeff = 1; |
308 |
|
|
309 |
|
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif); |
310 |
} else { |
} else { |
311 |
start_coeff = 0; |
start_coeff = 0; |
312 |
} |
} |
313 |
|
|
314 |
start_timer(); |
start_timer(); |
315 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
316 |
{ |
{ |
317 |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
int direction = dec->alternate_vertical_scan ? |
318 |
start_coeff); |
2 : pMB->acpred_directions[i]; |
319 |
|
|
320 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
321 |
} |
} |
322 |
stop_coding_timer(); |
stop_coding_timer(); |
323 |
|
|
336 |
start_timer(); |
start_timer(); |
337 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
338 |
stop_idct_timer(); |
stop_idct_timer(); |
339 |
|
|
340 |
} |
} |
341 |
|
|
342 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
345 |
} |
} |
346 |
|
|
347 |
start_timer(); |
start_timer(); |
348 |
|
|
349 |
|
if (reduced_resolution) |
350 |
|
{ |
351 |
|
next_block*=2; |
352 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
353 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
354 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
355 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
356 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
357 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
358 |
|
}else{ |
359 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
360 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
361 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
362 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
363 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
364 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
365 |
|
} |
366 |
stop_transfer_timer(); |
stop_transfer_timer(); |
367 |
} |
} |
368 |
|
|
369 |
|
|
370 |
|
|
371 |
|
|
372 |
|
/* decode an inter macroblock */ |
|
#define SIGN(X) (((X)>0)?1:-1) |
|
|
#define ABS(X) (((X)>0)?(X):-(X)) |
|
|
static const uint32_t roundtab[16] = |
|
|
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
|
|
|
|
|
|
|
|
// decode an inter macroblock |
|
|
|
|
373 |
void |
void |
374 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
375 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
376 |
const uint32_t x_pos, |
const uint32_t x_pos, |
377 |
const uint32_t y_pos, |
const uint32_t y_pos, |
378 |
const uint32_t acpred_flag, |
const uint32_t fcode, |
379 |
const uint32_t cbp, |
const uint32_t cbp, |
380 |
Bitstream * bs, |
Bitstream * bs, |
381 |
const uint32_t quant, |
const uint32_t quant, |
382 |
const uint32_t rounding) |
const uint32_t rounding, |
383 |
|
const int reduced_resolution) |
384 |
{ |
{ |
385 |
|
|
386 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
388 |
|
|
389 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
390 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
391 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
392 |
uint32_t i; |
uint32_t i; |
393 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
394 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
395 |
|
|
396 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
397 |
|
VECTOR mv[4]; /* local copy of mvs */ |
398 |
|
|
399 |
|
if (reduced_resolution) { |
400 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
401 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
402 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
403 |
|
for (i = 0; i < 4; i++) { |
404 |
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
405 |
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
406 |
|
} |
407 |
|
} else { |
408 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
409 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
410 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
411 |
|
for (i = 0; i < 4; i++) |
412 |
|
mv[i] = pMB->mvs[i]; |
413 |
|
} |
414 |
|
|
415 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
|
uv_dx = pMB->mvs[0].x; |
|
|
uv_dy = pMB->mvs[0].y; |
|
416 |
|
|
417 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = mv[0].x / (1 + dec->quarterpel); |
418 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = mv[0].y / (1 + dec->quarterpel); |
|
} else { |
|
|
int sum; |
|
419 |
|
|
420 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
421 |
uv_dx = |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
422 |
|
|
423 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
start_timer(); |
424 |
uv_dy = |
if (reduced_resolution) |
425 |
(sum == |
{ |
426 |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
427 |
(ABS(sum) / 16) * 2)); |
mv[0].x, mv[0].y, stride, rounding); |
428 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
429 |
|
uv_dx, uv_dy, stride2, rounding); |
430 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
431 |
|
uv_dx, uv_dy, stride2, rounding); |
432 |
|
|
433 |
|
} |
434 |
|
else |
435 |
|
{ |
436 |
|
if(dec->quarterpel) { |
437 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
438 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
439 |
|
mv[0].x, mv[0].y, stride, rounding); |
440 |
|
} |
441 |
|
else { |
442 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
443 |
|
mv[0].x, mv[0].y, stride, rounding); |
444 |
} |
} |
445 |
|
|
446 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
447 |
|
uv_dx, uv_dy, stride2, rounding); |
448 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
449 |
|
uv_dx, uv_dy, stride2, rounding); |
450 |
|
} |
451 |
|
stop_comp_timer(); |
452 |
|
|
453 |
|
} else { /* MODE_INTER4V */ |
454 |
|
int sum; |
455 |
|
|
456 |
|
if(dec->quarterpel) |
457 |
|
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
458 |
|
else |
459 |
|
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
460 |
|
|
461 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
462 |
|
|
463 |
|
if(dec->quarterpel) |
464 |
|
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
465 |
|
else |
466 |
|
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
467 |
|
|
468 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
469 |
|
|
470 |
start_timer(); |
start_timer(); |
471 |
|
if (reduced_resolution) |
472 |
|
{ |
473 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
474 |
|
mv[0].x, mv[0].y, stride, rounding); |
475 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
476 |
|
mv[1].x, mv[1].y, stride, rounding); |
477 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
478 |
|
mv[2].x, mv[2].y, stride, rounding); |
479 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
480 |
|
mv[3].x, mv[3].y, stride, rounding); |
481 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
482 |
|
uv_dx, uv_dy, stride2, rounding); |
483 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
484 |
|
uv_dx, uv_dy, stride2, rounding); |
485 |
|
|
486 |
|
/* set_block(pY_Cur, stride, 32, 32, 127); */ |
487 |
|
} |
488 |
|
else |
489 |
|
{ |
490 |
|
if(dec->quarterpel) { |
491 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
492 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
493 |
|
mv[0].x, mv[0].y, stride, rounding); |
494 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
495 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
496 |
|
mv[1].x, mv[1].y, stride, rounding); |
497 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
498 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
499 |
|
mv[2].x, mv[2].y, stride, rounding); |
500 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
501 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
502 |
|
mv[3].x, mv[3].y, stride, rounding); |
503 |
|
} |
504 |
|
else { |
505 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, 16 * y_pos, |
506 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
507 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, |
508 |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, |
mv[1].x, mv[1].y, stride, rounding); |
509 |
rounding); |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8, |
510 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, |
mv[2].x, mv[2].y, stride, rounding); |
511 |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8, |
512 |
rounding); |
mv[3].x, mv[3].y, stride, rounding); |
513 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos + 8, |
} |
514 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
|
|
rounding); |
|
515 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
516 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
517 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
518 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
519 |
|
} |
520 |
stop_comp_timer(); |
stop_comp_timer(); |
521 |
|
} |
522 |
|
|
523 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
524 |
if (cbp & (1 << (5 - i))) // coded |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
525 |
|
|
526 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
527 |
{ |
{ |
528 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
529 |
|
|
530 |
start_timer(); |
start_timer(); |
531 |
get_inter_block(bs, &block[i * 64]); |
get_inter_block(bs, &block[i * 64], direction); |
532 |
stop_coding_timer(); |
stop_coding_timer(); |
533 |
|
|
534 |
start_timer(); |
start_timer(); |
551 |
} |
} |
552 |
|
|
553 |
start_timer(); |
start_timer(); |
554 |
|
if (reduced_resolution) |
555 |
|
{ |
556 |
|
if (cbp & 32) |
557 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
558 |
|
if (cbp & 16) |
559 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
560 |
|
if (cbp & 8) |
561 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
562 |
|
if (cbp & 4) |
563 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
564 |
|
if (cbp & 2) |
565 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
566 |
|
if (cbp & 1) |
567 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
568 |
|
} |
569 |
|
else |
570 |
|
{ |
571 |
|
if (cbp & 32) |
572 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
573 |
|
if (cbp & 16) |
574 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
575 |
|
if (cbp & 8) |
576 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
577 |
|
if (cbp & 4) |
578 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
579 |
|
if (cbp & 2) |
580 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
581 |
|
if (cbp & 1) |
582 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
583 |
|
} |
584 |
|
stop_transfer_timer(); |
585 |
|
} |
586 |
|
|
587 |
|
|
588 |
|
static void |
589 |
|
decoder_mbgmc(DECODER * dec, |
590 |
|
MACROBLOCK * const pMB, |
591 |
|
const uint32_t x_pos, |
592 |
|
const uint32_t y_pos, |
593 |
|
const uint32_t fcode, |
594 |
|
const uint32_t cbp, |
595 |
|
Bitstream * bs, |
596 |
|
const uint32_t quant, |
597 |
|
const uint32_t rounding, |
598 |
|
const int reduced_resolution) /* no reduced res support */ |
599 |
|
{ |
600 |
|
|
601 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
602 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
603 |
|
|
604 |
|
const uint32_t stride = dec->edged_width; |
605 |
|
const uint32_t stride2 = stride / 2; |
606 |
|
const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
607 |
|
uint32_t i; |
608 |
|
const uint32_t iQuant = pMB->quant; |
609 |
|
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
610 |
|
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
611 |
|
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
612 |
|
|
613 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
614 |
|
|
615 |
|
start_timer(); |
616 |
|
|
617 |
|
/* this is where the calculations are done */ |
618 |
|
|
619 |
|
{ NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
620 |
|
|
621 |
|
gmc_data->predict_16x16(gmc_data, |
622 |
|
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
623 |
|
stride, stride, x_pos, y_pos, rounding); |
624 |
|
|
625 |
|
gmc_data->predict_8x8(gmc_data, |
626 |
|
dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u, |
627 |
|
dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v, |
628 |
|
stride2, stride2, x_pos, y_pos, rounding); |
629 |
|
|
630 |
|
gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel); |
631 |
|
|
632 |
|
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
633 |
|
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
634 |
|
} |
635 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
636 |
|
|
637 |
|
/* |
638 |
|
transfer16x16_copy(pY_Cur, dec->gmc.y + (y_pos << 4)*stride + (x_pos << 4), stride); |
639 |
|
transfer8x8_copy(pU_Cur, dec->gmc.u + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
640 |
|
transfer8x8_copy(pV_Cur, dec->gmc.v + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
641 |
|
*/ |
642 |
|
|
643 |
|
|
644 |
|
stop_transfer_timer(); |
645 |
|
|
646 |
|
if (!cbp) return; |
647 |
|
|
648 |
|
for (i = 0; i < 6; i++) { |
649 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
650 |
|
|
651 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
652 |
|
{ |
653 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
654 |
|
|
655 |
|
start_timer(); |
656 |
|
get_inter_block(bs, &block[i * 64], direction); |
657 |
|
stop_coding_timer(); |
658 |
|
|
659 |
|
start_timer(); |
660 |
|
if (dec->quant_type == 0) { |
661 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
662 |
|
} else { |
663 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
664 |
|
} |
665 |
|
stop_iquant_timer(); |
666 |
|
|
667 |
|
start_timer(); |
668 |
|
idct(&data[i * 64]); |
669 |
|
stop_idct_timer(); |
670 |
|
} |
671 |
|
} |
672 |
|
|
673 |
|
/* interlace + GMC is this possible ??? */ |
674 |
|
/* |
675 |
|
if (dec->interlacing && pMB->field_dct) { |
676 |
|
next_block = stride; |
677 |
|
stride *= 2; |
678 |
|
} |
679 |
|
*/ |
680 |
|
start_timer(); |
681 |
if (cbp & 32) |
if (cbp & 32) |
682 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
683 |
if (cbp & 16) |
if (cbp & 16) |
697 |
void |
void |
698 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
699 |
Bitstream * bs, |
Bitstream * bs, |
700 |
|
int reduced_resolution, |
701 |
int quant, |
int quant, |
702 |
int intra_dc_threshold) |
int intra_dc_threshold) |
703 |
{ |
{ |
704 |
|
uint32_t bound; |
705 |
uint32_t x, y; |
uint32_t x, y; |
706 |
|
uint32_t mb_width = dec->mb_width; |
707 |
|
uint32_t mb_height = dec->mb_height; |
708 |
|
|
709 |
for (y = 0; y < dec->mb_height; y++) { |
if (reduced_resolution) |
710 |
for (x = 0; x < dec->mb_width; x++) { |
{ |
711 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
mb_width = (dec->width + 31) / 32; |
712 |
|
mb_height = (dec->height + 31) / 32; |
713 |
|
} |
714 |
|
|
715 |
|
bound = 0; |
716 |
|
|
717 |
|
for (y = 0; y < mb_height; y++) { |
718 |
|
for (x = 0; x < mb_width; x++) { |
719 |
|
MACROBLOCK *mb; |
720 |
uint32_t mcbpc; |
uint32_t mcbpc; |
721 |
uint32_t cbpc; |
uint32_t cbpc; |
722 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
723 |
uint32_t cbpy; |
uint32_t cbpy; |
724 |
uint32_t cbp; |
uint32_t cbp; |
725 |
|
|
726 |
|
while (BitstreamShowBits(bs, 9) == 1) |
727 |
|
BitstreamSkip(bs, 9); |
728 |
|
|
729 |
|
if (check_resync_marker(bs, 0)) |
730 |
|
{ |
731 |
|
bound = read_video_packet_header(bs, dec, 0, |
732 |
|
&quant, NULL, NULL, &intra_dc_threshold); |
733 |
|
x = bound % mb_width; |
734 |
|
y = bound / mb_width; |
735 |
|
} |
736 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
737 |
|
|
738 |
|
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
739 |
|
|
740 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
741 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
742 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
743 |
|
|
744 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
745 |
|
|
|
if (mb->mode == MODE_STUFFING) { |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
746 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
747 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
748 |
|
|
755 |
} |
} |
756 |
} |
} |
757 |
mb->quant = quant; |
mb->quant = quant; |
758 |
|
mb->mvs[0].x = mb->mvs[0].y = |
759 |
|
mb->mvs[1].x = mb->mvs[1].y = |
760 |
|
mb->mvs[2].x = mb->mvs[2].y = |
761 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
762 |
|
|
763 |
if (dec->interlacing) { |
if (dec->interlacing) { |
764 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
765 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct); |
766 |
} |
} |
767 |
|
|
768 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
769 |
intra_dc_threshold); |
intra_dc_threshold, bound, reduced_resolution); |
770 |
|
|
771 |
} |
} |
772 |
|
if(dec->out_frm) |
773 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
774 |
} |
} |
775 |
|
|
776 |
} |
} |
782 |
int x, |
int x, |
783 |
int y, |
int y, |
784 |
int k, |
int k, |
785 |
VECTOR * mv, |
VECTOR * ret_mv, |
786 |
int fcode) |
int fcode, |
787 |
|
const int bound) |
788 |
{ |
{ |
789 |
|
|
790 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
792 |
int low = ((-32) * scale_fac); |
int low = ((-32) * scale_fac); |
793 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
794 |
|
|
795 |
VECTOR pmv[4]; |
VECTOR pmv; |
796 |
int32_t psad[4]; |
VECTOR mv; |
797 |
|
|
798 |
int mv_x, mv_y; |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
|
int pmv_x, pmv_y; |
|
799 |
|
|
800 |
|
mv.x = get_mv(bs, fcode); |
801 |
|
mv.y = get_mv(bs, fcode); |
802 |
|
|
803 |
get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad); |
DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y); |
804 |
|
|
805 |
pmv_x = pmv[0].x; |
mv.x += pmv.x; |
806 |
pmv_y = pmv[0].y; |
mv.y += pmv.y; |
807 |
|
|
808 |
mv_x = get_mv(bs, fcode); |
if (mv.x < low) { |
809 |
mv_y = get_mv(bs, fcode); |
mv.x += range; |
810 |
|
} else if (mv.x > high) { |
811 |
mv_x += pmv_x; |
mv.x -= range; |
812 |
mv_y += pmv_y; |
} |
813 |
|
|
814 |
if (mv_x < low) { |
if (mv.y < low) { |
815 |
mv_x += range; |
mv.y += range; |
816 |
} else if (mv_x > high) { |
} else if (mv.y > high) { |
817 |
mv_x -= range; |
mv.y -= range; |
818 |
} |
} |
819 |
|
|
820 |
if (mv_y < low) { |
ret_mv->x = mv.x; |
821 |
mv_y += range; |
ret_mv->y = mv.y; |
|
} else if (mv_y > high) { |
|
|
mv_y -= range; |
|
822 |
} |
} |
823 |
|
|
|
mv->x = mv_x; |
|
|
mv->y = mv_y; |
|
824 |
|
|
|
} |
|
825 |
|
|
826 |
|
|
827 |
|
|
828 |
|
/* for P_VOP set gmc_warp to NULL */ |
829 |
void |
void |
830 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
831 |
Bitstream * bs, |
Bitstream * bs, |
832 |
int rounding, |
int rounding, |
833 |
|
int reduced_resolution, |
834 |
int quant, |
int quant, |
835 |
int fcode, |
int fcode, |
836 |
int intra_dc_threshold) |
int intra_dc_threshold, |
837 |
|
const WARPPOINTS *const gmc_warp) |
838 |
{ |
{ |
839 |
|
|
840 |
uint32_t x, y; |
uint32_t x, y; |
841 |
|
uint32_t bound; |
842 |
|
int cp_mb, st_mb; |
843 |
|
uint32_t mb_width = dec->mb_width; |
844 |
|
uint32_t mb_height = dec->mb_height; |
845 |
|
|
846 |
|
if (reduced_resolution) |
847 |
|
{ |
848 |
|
mb_width = (dec->width + 31) / 32; |
849 |
|
mb_height = (dec->height + 31) / 32; |
850 |
|
} |
851 |
|
|
852 |
start_timer(); |
start_timer(); |
853 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
854 |
dec->width, dec->height, dec->interlacing); |
dec->width, dec->height); |
855 |
stop_edges_timer(); |
stop_edges_timer(); |
856 |
|
|
857 |
for (y = 0; y < dec->mb_height; y++) { |
if (gmc_warp) |
858 |
for (x = 0; x < dec->mb_width; x++) { |
{ |
859 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
|
860 |
|
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
861 |
|
/* { |
862 |
|
fprintf(stderr,"GMC parameters acc=%d(-> 1/%d), %d pts!!!\n", |
863 |
|
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
864 |
|
dec->sprite_warping_points); |
865 |
|
}*/ |
866 |
|
|
867 |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
generate_GMCparameters( dec->sprite_warping_points, |
868 |
if (!(BitstreamGetBit(bs))) // not_coded |
dec->sprite_warping_accuracy, gmc_warp, |
869 |
|
dec->width, dec->height, &dec->new_gmc_data); |
870 |
|
|
871 |
|
/* image warping is done block-based in decoder_mbgmc(), now */ |
872 |
|
} |
873 |
|
|
874 |
|
bound = 0; |
875 |
|
|
876 |
|
for (y = 0; y < mb_height; y++) { |
877 |
|
cp_mb = st_mb = 0; |
878 |
|
for (x = 0; x < mb_width; x++) { |
879 |
|
MACROBLOCK *mb; |
880 |
|
|
881 |
|
/* skip stuffing */ |
882 |
|
while (BitstreamShowBits(bs, 10) == 1) |
883 |
|
BitstreamSkip(bs, 10); |
884 |
|
|
885 |
|
if (check_resync_marker(bs, fcode - 1)) |
886 |
|
{ |
887 |
|
bound = read_video_packet_header(bs, dec, fcode - 1, |
888 |
|
&quant, &fcode, NULL, &intra_dc_threshold); |
889 |
|
x = bound % mb_width; |
890 |
|
y = bound / mb_width; |
891 |
|
} |
892 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
893 |
|
|
894 |
|
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
895 |
|
|
896 |
|
/* if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) */ /* not_coded */ |
897 |
|
if (!(BitstreamGetBit(bs))) /* block _is_ coded */ |
898 |
{ |
{ |
899 |
uint32_t mcbpc; |
uint32_t mcbpc; |
900 |
uint32_t cbpc; |
uint32_t cbpc; |
902 |
uint32_t cbpy; |
uint32_t cbpy; |
903 |
uint32_t cbp; |
uint32_t cbp; |
904 |
uint32_t intra; |
uint32_t intra; |
905 |
|
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
906 |
|
|
907 |
|
cp_mb++; |
908 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
909 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
910 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
911 |
|
|
912 |
|
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
913 |
|
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
914 |
acpred_flag = 0; |
acpred_flag = 0; |
915 |
|
|
916 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
919 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
920 |
} |
} |
921 |
|
|
922 |
if (mb->mode == MODE_STUFFING) { |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
923 |
DEBUG("-- STUFFING ?"); |
{ |
924 |
continue; |
mcsel = BitstreamGetBit(bs); |
925 |
} |
} |
926 |
|
|
927 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
928 |
|
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
929 |
|
|
930 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
931 |
|
|
932 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
933 |
quant += dquant_table[BitstreamGetBits(bs, 2)]; |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
934 |
|
DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant); |
935 |
|
quant += dquant; |
936 |
if (quant > 31) { |
if (quant > 31) { |
937 |
quant = 31; |
quant = 31; |
938 |
} else if (mb->quant < 1) { |
} else if (quant < 1) { |
939 |
quant = 1; |
quant = 1; |
940 |
} |
} |
941 |
|
DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant); |
942 |
} |
} |
943 |
mb->quant = quant; |
mb->quant = quant; |
944 |
|
|
945 |
if (dec->interlacing) { |
if (dec->interlacing) { |
946 |
|
if (cbp || intra) { |
947 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
948 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
949 |
|
} |
950 |
|
|
951 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
952 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
953 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
954 |
|
|
955 |
if (mb->field_pred) { |
if (mb->field_pred) { |
956 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
957 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
958 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
959 |
DEBUG1("decp: field_for_bot: ", mb->field_for_bot); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
960 |
} |
} |
961 |
} |
} |
962 |
} |
} |
963 |
|
|
964 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mcsel) { |
965 |
|
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, |
966 |
|
rounding, reduced_resolution); |
967 |
|
continue; |
968 |
|
|
969 |
|
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
970 |
|
|
971 |
if (dec->interlacing && mb->field_pred) { |
if (dec->interlacing && mb->field_pred) { |
972 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
973 |
fcode); |
fcode, bound); |
974 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
975 |
fcode); |
fcode, bound); |
976 |
} else { |
} else { |
977 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
978 |
fcode); |
fcode, bound); |
979 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
980 |
mb->mvs[0].x; |
} |
981 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
} else if (mb->mode == MODE_INTER4V ) { |
982 |
mb->mvs[0].y; |
|
983 |
} |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
984 |
} else if (mb->mode == |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
985 |
MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */ ) { |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
986 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
987 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
} else /* MODE_INTRA, MODE_INTRA_Q */ |
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
|
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
|
|
} else // MODE_INTRA, MODE_INTRA_Q |
|
988 |
{ |
{ |
989 |
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 = |
990 |
0; |
0; |
991 |
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 = |
992 |
0; |
0; |
993 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
994 |
intra_dc_threshold); |
intra_dc_threshold, bound, reduced_resolution); |
995 |
continue; |
continue; |
996 |
} |
} |
997 |
|
|
998 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, |
999 |
rounding); |
rounding, reduced_resolution); |
1000 |
} else // not coded |
|
1001 |
|
} |
1002 |
|
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
1003 |
|
{ |
1004 |
|
mb->mode = MODE_NOT_CODED_GMC; |
1005 |
|
|
1006 |
|
start_timer(); |
1007 |
|
|
1008 |
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, quant, |
1009 |
|
rounding, reduced_resolution); |
1010 |
|
|
1011 |
|
stop_transfer_timer(); |
1012 |
|
|
1013 |
|
if(dec->out_frm && cp_mb > 0) { |
1014 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1015 |
|
cp_mb = 0; |
1016 |
|
} |
1017 |
|
st_mb = x+1; |
1018 |
|
} |
1019 |
|
else /* not coded P_VOP macroblock */ |
1020 |
{ |
{ |
|
//DEBUG2("P-frame MB at (X,Y)=",x,y); |
|
1021 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
1022 |
|
|
1023 |
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; |
1024 |
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; |
1025 |
|
/* copy macroblock directly from ref to cur */ |
|
// copy macroblock directly from ref to cur |
|
1026 |
|
|
1027 |
start_timer(); |
start_timer(); |
1028 |
|
|
1029 |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
if (reduced_resolution) |
1030 |
(16 * x), |
{ |
1031 |
dec->refn[0].y + (16 * y) * dec->edged_width + |
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
1032 |
(16 * x), dec->edged_width); |
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
1033 |
|
dec->edged_width); |
1034 |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
|
1035 |
(16 * x + 8), |
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
1036 |
dec->refn[0].y + (16 * y) * dec->edged_width + |
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
1037 |
(16 * x + 8), dec->edged_width); |
dec->edged_width/2); |
1038 |
|
|
1039 |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
1040 |
(16 * x), |
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
1041 |
dec->refn[0].y + (16 * y + |
dec->edged_width/2); |
1042 |
8) * dec->edged_width + |
} |
1043 |
(16 * x), dec->edged_width); |
else |
1044 |
|
{ |
1045 |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
1046 |
(16 * x + 8), |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
1047 |
dec->refn[0].y + (16 * y + |
dec->edged_width); |
1048 |
8) * dec->edged_width + |
|
1049 |
(16 * x + 8), dec->edged_width); |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
1050 |
|
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
|
|
(8 * x), |
|
|
dec->refn[0].u + |
|
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
|
1051 |
dec->edged_width / 2); |
dec->edged_width / 2); |
1052 |
|
|
1053 |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
1054 |
(8 * x), |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
|
dec->refn[0].v + |
|
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
|
1055 |
dec->edged_width / 2); |
dec->edged_width / 2); |
1056 |
|
} |
1057 |
|
|
1058 |
stop_transfer_timer(); |
stop_transfer_timer(); |
1059 |
|
|
1060 |
|
if(dec->out_frm && cp_mb > 0) { |
1061 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1062 |
|
cp_mb = 0; |
1063 |
} |
} |
1064 |
|
st_mb = x+1; |
1065 |
} |
} |
1066 |
} |
} |
1067 |
|
if(dec->out_frm && cp_mb > 0) |
1068 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1069 |
|
} |
1070 |
} |
} |
1071 |
|
|
1072 |
|
|
1073 |
// add by MinChen <chenm001@163.com> |
/* decode B-frame motion vector */ |
|
// decode B-frame motion vector |
|
1074 |
void |
void |
1075 |
get_b_motion_vector(DECODER * dec, |
get_b_motion_vector(DECODER * dec, |
1076 |
Bitstream * bs, |
Bitstream * bs, |
1114 |
} |
} |
1115 |
|
|
1116 |
|
|
1117 |
// add by MinChen <chenm001@163.com> |
/* decode an B-frame forward & backward inter macroblock */ |
|
// decode an B-frame forward & backward inter macroblock |
|
1118 |
void |
void |
1119 |
decoder_bf_mbinter(DECODER * dec, |
decoder_bf_mbinter(DECODER * dec, |
1120 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
1141 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1142 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1143 |
|
|
1144 |
|
|
1145 |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1146 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1147 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
1148 |
|
|
1149 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
if (dec->quarterpel) |
1150 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
{ |
1151 |
|
uv_dx /= 2; |
1152 |
|
uv_dy /= 2; |
1153 |
|
} |
1154 |
|
|
1155 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1156 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1157 |
} else { |
} else { |
1158 |
int sum; |
int sum; |
1159 |
|
|
1160 |
|
if(dec->quarterpel) |
1161 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1162 |
|
else |
1163 |
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; |
|
uv_dx = |
|
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1164 |
|
|
1165 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1166 |
|
|
1167 |
|
if(dec->quarterpel) |
1168 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1169 |
|
else |
1170 |
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; |
1171 |
uv_dy = |
|
1172 |
(sum == |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1173 |
} |
} |
1174 |
|
|
1175 |
start_timer(); |
start_timer(); |
1176 |
|
if(dec->quarterpel) { |
1177 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
1178 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1179 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1180 |
|
} |
1181 |
|
else { |
1182 |
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, |
1183 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1184 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
1185 |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1186 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
1187 |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1188 |
0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
1189 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1190 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
} |
1191 |
0); |
|
1192 |
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, |
1193 |
uv_dx, uv_dy, stride2, 0); |
uv_dx, uv_dy, stride2, 0); |
1194 |
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, |
1196 |
stop_comp_timer(); |
stop_comp_timer(); |
1197 |
|
|
1198 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
1199 |
if (cbp & (1 << (5 - i))) // coded |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1200 |
|
|
1201 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
1202 |
{ |
{ |
1203 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
1204 |
|
|
1205 |
start_timer(); |
start_timer(); |
1206 |
get_inter_block(bs, &block[i * 64]); |
get_inter_block(bs, &block[i * 64], direction); |
1207 |
stop_coding_timer(); |
stop_coding_timer(); |
1208 |
|
|
1209 |
start_timer(); |
start_timer(); |
1241 |
stop_transfer_timer(); |
stop_transfer_timer(); |
1242 |
} |
} |
1243 |
|
|
1244 |
|
/* decode an B-frame direct & inter macroblock */ |
|
// add by MinChen <chenm001@163.com> |
|
|
// decode an B-frame direct & inter macroblock |
|
1245 |
void |
void |
1246 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
1247 |
IMAGE forward, |
IMAGE forward, |
1249 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
1250 |
const uint32_t x_pos, |
const uint32_t x_pos, |
1251 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t cbp, |
|
1252 |
Bitstream * bs) |
Bitstream * bs) |
1253 |
{ |
{ |
1254 |
|
|
1263 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
1264 |
uint32_t i; |
uint32_t i; |
1265 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1266 |
|
const uint32_t cbp = pMB->cbp; |
1267 |
|
|
1268 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1269 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1270 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1271 |
|
|
1272 |
|
|
1273 |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1274 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1275 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
1276 |
|
|
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
|
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
|
|
|
1277 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
1278 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1279 |
|
|
1280 |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
if (dec->quarterpel) |
1281 |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
{ |
1282 |
|
uv_dx /= 2; |
1283 |
|
uv_dy /= 2; |
1284 |
|
|
1285 |
|
b_uv_dx /= 2; |
1286 |
|
b_uv_dy /= 2; |
1287 |
|
} |
1288 |
|
|
1289 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1290 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1291 |
|
|
1292 |
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1293 |
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1294 |
} else { |
} else { |
1295 |
int sum; |
int sum; |
1296 |
|
|
1297 |
|
if(dec->quarterpel) |
1298 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1299 |
|
else |
1300 |
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; |
|
uv_dx = |
|
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1301 |
|
|
1302 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1303 |
|
|
1304 |
|
if(dec->quarterpel) |
1305 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1306 |
|
else |
1307 |
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; |
1308 |
uv_dy = |
|
1309 |
(sum == |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1310 |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
1311 |
(ABS(sum) / 16) * 2)); |
|
1312 |
|
if(dec->quarterpel) |
1313 |
sum = |
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); |
1314 |
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
else |
1315 |
pMB->b_mvs[3].x; |
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1316 |
b_uv_dx = |
|
1317 |
(sum == |
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1318 |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
1319 |
(ABS(sum) / 16) * 2)); |
if(dec->quarterpel) |
1320 |
|
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); |
1321 |
sum = |
else |
1322 |
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1323 |
pMB->b_mvs[3].y; |
|
1324 |
b_uv_dy = |
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1325 |
} |
} |
1326 |
|
|
1327 |
|
|
1328 |
start_timer(); |
start_timer(); |
1329 |
|
if(dec->quarterpel) { |
1330 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1331 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1332 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1333 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1334 |
|
else { |
1335 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1336 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1337 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1338 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1339 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1340 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1341 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1342 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1343 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1344 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1345 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1346 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1347 |
|
} |
1348 |
|
} |
1349 |
|
else { |
1350 |
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, |
1351 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1352 |
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, |
1356 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1357 |
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, |
1358 |
0); |
0); |
1359 |
|
} |
1360 |
|
|
1361 |
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, |
1362 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1363 |
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, |
1364 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1365 |
|
|
1366 |
|
|
1367 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
if(dec->quarterpel) { |
1368 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1369 |
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1370 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1371 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1372 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
else { |
1373 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1374 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1375 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1376 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1377 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1378 |
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1379 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1380 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1381 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1382 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1383 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1384 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1385 |
|
} |
1386 |
|
} |
1387 |
|
else { |
1388 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1389 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1390 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1391 |
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, |
1392 |
0); |
0); |
1393 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1394 |
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, |
1395 |
stride, 0); |
stride, 0); |
1396 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1397 |
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, |
1398 |
stride, 0); |
stride, 0); |
1399 |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
} |
1400 |
|
|
1401 |
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1402 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1403 |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
1404 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1405 |
|
|
1406 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1407 |
stride); |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1408 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1409 |
stride); |
stride, 1, 8); |
1410 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, |
|
1411 |
stride); |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1412 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1413 |
16 * y_pos + 8, stride); |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1414 |
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, |
stride, 1, 8); |
1415 |
stride2); |
|
1416 |
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1417 |
stride2); |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1418 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1419 |
|
stride, 1, 8); |
1420 |
|
|
1421 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1422 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1423 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1424 |
|
stride, 1, 8); |
1425 |
|
|
1426 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1427 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1428 |
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1429 |
|
stride2, 1, 8); |
1430 |
|
|
1431 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1432 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1433 |
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1434 |
|
stride2, 1, 8); |
1435 |
|
|
1436 |
stop_comp_timer(); |
stop_comp_timer(); |
1437 |
|
|
1438 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
1439 |
if (cbp & (1 << (5 - i))) // coded |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1440 |
|
|
1441 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
1442 |
{ |
{ |
1443 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
1444 |
|
|
1445 |
start_timer(); |
start_timer(); |
1446 |
get_inter_block(bs, &block[i * 64]); |
get_inter_block(bs, &block[i * 64], direction); |
1447 |
stop_coding_timer(); |
stop_coding_timer(); |
1448 |
|
|
1449 |
start_timer(); |
start_timer(); |
1482 |
} |
} |
1483 |
|
|
1484 |
|
|
1485 |
// add by MinChen <chenm001@163.com> |
/* for decode B-frame dbquant */ |
|
// for decode B-frame dbquant |
|
1486 |
int32_t __inline |
int32_t __inline |
1487 |
get_dbquant(Bitstream * bs) |
get_dbquant(Bitstream * bs) |
1488 |
{ |
{ |
1489 |
if (!BitstreamGetBit(bs)) // '0' |
if (!BitstreamGetBit(bs)) /* '0' */ |
1490 |
return (0); |
return (0); |
1491 |
else if (!BitstreamGetBit(bs)) // '10' |
else if (!BitstreamGetBit(bs)) /* '10' */ |
1492 |
return (-2); |
return (-2); |
1493 |
else |
else /* '11' */ |
1494 |
return (2); // '11' |
return (2); |
1495 |
} |
} |
1496 |
|
|
1497 |
// add by MinChen <chenm001@163.com> |
/* |
1498 |
// for decode B-frame mb_type |
* For decode B-frame mb_type |
1499 |
// bit ret_value |
* bit ret_value |
1500 |
// 1 0 |
* 1 0 |
1501 |
// 01 1 |
* 01 1 |
1502 |
// 001 2 |
* 001 2 |
1503 |
// 0001 3 |
* 0001 3 |
1504 |
|
*/ |
1505 |
int32_t __inline |
int32_t __inline |
1506 |
get_mbtype(Bitstream * bs) |
get_mbtype(Bitstream * bs) |
1507 |
{ |
{ |
1525 |
int fcode_forward, |
int fcode_forward, |
1526 |
int fcode_backward) |
int fcode_backward) |
1527 |
{ |
{ |
|
|
|
1528 |
uint32_t x, y; |
uint32_t x, y; |
1529 |
VECTOR mv, zeromv; |
VECTOR mv; |
1530 |
|
const VECTOR zeromv = {0,0}; |
1531 |
|
#ifdef BFRAMES_DEC_DEBUG |
1532 |
|
FILE *fp; |
1533 |
|
static char first=0; |
1534 |
|
#define BFRAME_DEBUG if (!first && fp){ \ |
1535 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
1536 |
|
} |
1537 |
|
#endif |
1538 |
|
|
1539 |
start_timer(); |
start_timer(); |
1540 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1541 |
dec->width, dec->height, dec->interlacing); |
dec->width, dec->height); |
1542 |
//image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1543 |
|
dec->width, dec->height); |
1544 |
stop_edges_timer(); |
stop_edges_timer(); |
1545 |
|
|
1546 |
|
#ifdef BFRAMES_DEC_DEBUG |
1547 |
|
if (!first){ |
1548 |
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
1549 |
|
} |
1550 |
|
#endif |
1551 |
|
|
1552 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1553 |
// Initialize Pred Motion Vector |
/* Initialize Pred Motion Vector */ |
1554 |
dec->p_fmv.x = dec->p_fmv.y = dec->p_bmv.x = dec->p_bmv.y = 0; |
dec->p_fmv = dec->p_bmv = zeromv; |
1555 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1556 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1557 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1558 |
|
|
1559 |
mb->mvs[0].x = mb->mvs[0].y = zeromv.x = zeromv.y = mv.x = mv.y = |
mv = |
1560 |
0; |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1561 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1562 |
|
|
1563 |
|
/* |
1564 |
|
* skip if the co-located P_VOP macroblock is not coded |
1565 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1566 |
|
* automatically skipped |
1567 |
|
*/ |
1568 |
|
|
|
// the last P_VOP is skip macroblock ? |
|
1569 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1570 |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
/* DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); */ |
|
mb->mb_type = MODE_FORWARD; |
|
1571 |
mb->cbp = 0; |
mb->cbp = 0; |
1572 |
|
#ifdef BFRAMES_DEC_DEBUG |
1573 |
|
mb->mb_type = MODE_NOT_CODED; |
1574 |
|
BFRAME_DEBUG |
1575 |
|
#endif |
1576 |
|
mb->mb_type = MODE_FORWARD; |
1577 |
|
mb->quant = last_mb->quant; |
1578 |
|
/* |
1579 |
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; |
1580 |
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; |
1581 |
mb->quant = 8; |
*/ |
1582 |
|
|
1583 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1584 |
continue; |
continue; |
1585 |
} |
} |
|
//t=BitstreamShowBits(bs,32); |
|
1586 |
|
|
1587 |
if (!BitstreamGetBit(bs)) { // modb=='0' |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1588 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1589 |
|
|
1590 |
mb->mb_type = get_mbtype(bs); |
mb->mb_type = get_mbtype(bs); |
1591 |
|
|
1592 |
if (!modb2) { // modb=='00' |
if (!modb2) { /* modb=='00' */ |
1593 |
mb->cbp = BitstreamGetBits(bs, 6); |
mb->cbp = BitstreamGetBits(bs, 6); |
1594 |
} else { |
} else { |
1595 |
mb->cbp = 0; |
mb->cbp = 0; |
1599 |
|
|
1600 |
if (quant > 31) { |
if (quant > 31) { |
1601 |
quant = 31; |
quant = 31; |
1602 |
} else if (mb->quant < 1) { |
} else if (quant < 1) { |
1603 |
quant = 1; |
quant = 1; |
1604 |
} |
} |
|
} else { |
|
|
quant = 8; |
|
1605 |
} |
} |
|
mb->quant = quant; |
|
1606 |
} else { |
} else { |
1607 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mb_type = MODE_DIRECT_NONE_MV; |
1608 |
mb->cbp = 0; |
mb->cbp = 0; |
1609 |
} |
} |
1610 |
|
|
1611 |
mb->mode = MODE_INTER; |
mb->quant = quant; |
1612 |
//DEBUG1("Switch bm_type=",mb->mb_type); |
mb->mode = MODE_INTER4V; |
1613 |
|
/* DEBUG1("Switch bm_type=",mb->mb_type); */ |
1614 |
|
|
1615 |
|
#ifdef BFRAMES_DEC_DEBUG |
1616 |
|
BFRAME_DEBUG |
1617 |
|
#endif |
1618 |
|
|
1619 |
switch (mb->mb_type) { |
switch (mb->mb_type) { |
1620 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1621 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], 1, zeromv); |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1622 |
|
|
1623 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
1624 |
{ // Because this file is a C file not C++ so I use '{' to define var |
{ |
1625 |
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = |
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
|
dec->time_pp; |
|
1626 |
int i; |
int i; |
1627 |
|
|
1628 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1629 |
mb->mvs[i].x = |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
1630 |
(int32_t) ((TRB * last_mb->mvs[i].x) / TRD + |
/ TRD + mv.x); |
1631 |
mb->mvs[0].x); |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1632 |
mb->b_mvs[i].x = |
? ((TRB - TRD) * last_mb->mvs[i].x) |
1633 |
(int32_t) ((mb->mvs[0].x == |
/ TRD |
1634 |
0) ? ((TRB - |
: mb->mvs[i].x - last_mb->mvs[i].x); |
1635 |
TRD) * last_mb->mvs[i].x) / |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
1636 |
TRD : mb->mvs[i].x - last_mb->mvs[i].x); |
/ TRD + mv.y); |
1637 |
mb->mvs[i].y = |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1638 |
(int32_t) ((TRB * last_mb->mvs[i].y) / TRD + |
? ((TRB - TRD) * last_mb->mvs[i].y) |
1639 |
mb->mvs[0].y); |
/ TRD |
1640 |
mb->b_mvs[i].y = |
: mb->mvs[i].y - last_mb->mvs[i].y); |
|
(int32_t) ((mb->mvs[0].y == |
|
|
0) ? ((TRB - |
|
|
TRD) * last_mb->mvs[i].y) / |
|
|
TRD : mb->mvs[i].y - last_mb->mvs[i].y); |
|
1641 |
} |
} |
1642 |
//DEBUG("B-frame Direct!\n"); |
/* DEBUG("B-frame Direct!\n"); */ |
1643 |
} |
} |
|
mb->mode = MODE_INTER4V; |
|
1644 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1645 |
mb, x, y, mb->cbp, bs); |
mb, x, y, bs); |
1646 |
break; |
break; |
1647 |
|
|
1648 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1649 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1650 |
dec->p_fmv); |
dec->p_fmv); |
1651 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
mb->mvs[0].x; |
|
|
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
mb->mvs[0].y; |
|
1652 |
|
|
1653 |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1654 |
fcode_backward, dec->p_bmv); |
fcode_backward, dec->p_bmv); |
1655 |
dec->p_bmv.x = mb->b_mvs[1].x = mb->b_mvs[2].x = |
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
1656 |
mb->b_mvs[3].x = mb->b_mvs[0].x; |
mb->b_mvs[3] = mb->b_mvs[0]; |
|
dec->p_bmv.y = mb->b_mvs[1].y = mb->b_mvs[2].y = |
|
|
mb->b_mvs[3].y = mb->b_mvs[0].y; |
|
1657 |
|
|
1658 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1659 |
mb, x, y, mb->cbp, bs); |
mb, x, y, bs); |
1660 |
//DEBUG("B-frame Bidir!\n"); |
/* DEBUG("B-frame Bidir!\n"); */ |
1661 |
break; |
break; |
1662 |
|
|
1663 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1664 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1665 |
dec->p_bmv); |
dec->p_bmv); |
1666 |
dec->p_bmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
mb->mvs[0].x; |
|
|
dec->p_bmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
mb->mvs[0].y; |
|
1667 |
|
|
1668 |
|
mb->mode = MODE_INTER; |
1669 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1670 |
//DEBUG("B-frame Backward!\n"); |
/* DEBUG("B-frame Backward!\n"); */ |
1671 |
break; |
break; |
1672 |
|
|
1673 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1674 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1675 |
dec->p_fmv); |
dec->p_fmv); |
1676 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
mb->mvs[0].x; |
|
|
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
mb->mvs[0].y; |
|
1677 |
|
|
1678 |
|
mb->mode = MODE_INTER; |
1679 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1680 |
//DEBUG("B-frame Forward!\n"); |
/* DEBUG("B-frame Forward!\n"); */ |
1681 |
break; |
break; |
1682 |
|
|
1683 |
default: |
default: |
1684 |
DEBUG1("Not support B-frame mb_type =", mb->mb_type); |
DPRINTF(XVID_DEBUG_ERROR,"Not support B-frame mb_type = %i\n", mb->mb_type); |
1685 |
|
} |
1686 |
|
} /* End of for */ |
1687 |
} |
} |
1688 |
|
|
1689 |
} // end of FOR |
#ifdef BFRAMES_DEC_DEBUG |
1690 |
|
if (!first){ |
1691 |
|
first=1; |
1692 |
|
if (fp) |
1693 |
|
fclose(fp); |
1694 |
} |
} |
1695 |
|
#endif |
1696 |
} |
} |
1697 |
|
|
1698 |
// swap two MACROBLOCK array |
|
1699 |
void |
|
1700 |
mb_swap(MACROBLOCK ** mb1, |
/* perform post processing if necessary, and output the image */ |
1701 |
MACROBLOCK ** mb2) |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1702 |
|
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type) |
1703 |
{ |
{ |
|
MACROBLOCK *temp = *mb1; |
|
1704 |
|
|
1705 |
*mb1 = *mb2; |
|
1706 |
*mb2 = temp; |
image_output(img, dec->width, dec->height, |
1707 |
|
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1708 |
|
frame->output.csp, dec->interlacing); |
1709 |
|
|
1710 |
|
if (stats) |
1711 |
|
{ |
1712 |
|
stats->type = coding2type(coding_type); |
1713 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1714 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1715 |
} |
} |
1716 |
|
} |
1717 |
|
|
1718 |
|
|
1719 |
int |
int |
1720 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1721 |
XVID_DEC_FRAME * frame) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1722 |
{ |
{ |
1723 |
|
|
1724 |
Bitstream bs; |
Bitstream bs; |
1725 |
uint32_t rounding; |
uint32_t rounding; |
1726 |
|
uint32_t reduced_resolution; |
1727 |
uint32_t quant; |
uint32_t quant; |
1728 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1729 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1730 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1731 |
uint32_t vop_type; |
WARPPOINTS gmc_warp; |
1732 |
|
int coding_type; |
1733 |
|
int success, output, seen_something; |
1734 |
|
idctFuncPtr idct_save; |
1735 |
|
|
1736 |
|
if (XVID_MAJOR(frame->version) != 1 || (stats && XVID_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1737 |
|
return XVID_ERR_VERSION; |
1738 |
|
|
1739 |
start_global_timer(); |
start_global_timer(); |
1740 |
|
|
1741 |
|
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1742 |
|
if ((frame->general & XVID_DISCONTINUITY)) |
1743 |
|
dec->frames = 0; |
1744 |
|
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1745 |
|
|
1746 |
|
if (frame->length < 0) /* decoder flush */ |
1747 |
|
{ |
1748 |
|
int ret; |
1749 |
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1750 |
|
we have a reference frame, then outout the reference frame */ |
1751 |
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1752 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1753 |
|
dec->frames = 0; |
1754 |
|
ret = 0; |
1755 |
|
}else{ |
1756 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1757 |
|
ret = XVID_ERR_END; |
1758 |
|
} |
1759 |
|
|
1760 |
|
emms(); |
1761 |
|
stop_global_timer(); |
1762 |
|
return ret; |
1763 |
|
} |
1764 |
|
|
1765 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1766 |
|
|
1767 |
// add by chenm001 <chenm001@163.com> |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1768 |
// for support B-frame to reference last 2 frame |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1769 |
dec->frames++; |
{ |
1770 |
vop_type = |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1771 |
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
(uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing); |
1772 |
&fcode_backward, &intra_dc_threshold); |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1773 |
|
emms(); |
1774 |
|
return 1; /* one byte consumed */ |
1775 |
|
} |
1776 |
|
|
1777 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
success = 0; |
1778 |
|
output = 0; |
1779 |
|
seen_something = 0; |
1780 |
|
idct_save = idct; |
1781 |
|
|
1782 |
switch (vop_type) { |
repeat: |
|
case P_VOP: |
|
|
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
|
|
intra_dc_threshold); |
|
|
DEBUG1("P_VOP Time=", dec->time); |
|
|
break; |
|
1783 |
|
|
1784 |
case I_VOP: |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1785 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
|
DEBUG1("I_VOP Time=", dec->time); |
|
|
break; |
|
1786 |
|
|
1787 |
case B_VOP: |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
1788 |
#ifdef BFRAMES |
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1789 |
if (dec->time_pp > dec->time_bp) { |
|
1790 |
DEBUG1("B_VOP Time=", dec->time); |
if (coding_type == -1) /* nothing */ |
1791 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
{ |
1792 |
} else { |
if (success) goto done; |
1793 |
DEBUG("broken B-frame!"); |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1794 |
|
emms(); |
1795 |
|
return BitstreamPos(&bs)/8; |
1796 |
} |
} |
|
#endif |
|
|
break; |
|
1797 |
|
|
1798 |
case N_VOP: // vop not coded |
if (coding_type == -2 || coding_type == -3) /* vol and/or resize */ |
1799 |
break; |
{ |
1800 |
|
if (coding_type == -3) |
1801 |
|
decoder_resize(dec); |
1802 |
|
|
1803 |
default: |
if (stats) |
1804 |
return XVID_ERR_FAIL; |
{ |
1805 |
|
stats->type = XVID_TYPE_VOL; |
1806 |
|
stats->data.vol.general = 0; |
1807 |
|
/*XXX: if (dec->interlacing) |
1808 |
|
stats->data.vol.general |= ++INTERLACING; */ |
1809 |
|
stats->data.vol.width = dec->width; |
1810 |
|
stats->data.vol.height = dec->height; |
1811 |
|
stats->data.vol.par = dec->aspect_ratio; |
1812 |
|
stats->data.vol.par_width = dec->par_width; |
1813 |
|
stats->data.vol.par_height = dec->par_height; |
1814 |
|
emms(); |
1815 |
|
return BitstreamPos(&bs)/8; /* number of bytes consumed */ |
1816 |
|
} |
1817 |
|
goto repeat; |
1818 |
} |
} |
1819 |
|
|
1820 |
frame->length = BitstreamPos(&bs) / 8; |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1821 |
|
|
1822 |
#ifdef BFRAMES |
#if defined(ARCH_IS_IA32) |
1823 |
// test if no B_VOP |
/* |
1824 |
if (dec->low_delay) { |
* /!\ Ugly hack /!\ |
1825 |
|
* IA32: Prior to xvid bitstream 10, we were using Walten's mmx/xmm idct |
1826 |
|
*/ |
1827 |
|
if((idct == simple_idct_mmx) && (dec->bs_version < 10)) |
1828 |
|
idct = idct_mmx; |
1829 |
#endif |
#endif |
1830 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
|
1831 |
frame->image, frame->stride, frame->colorspace); |
/* packed_mode: special-N_VOP treament */ |
1832 |
#ifdef BFRAMES |
if (dec->packed_mode && coding_type == N_VOP) |
1833 |
} else { |
{ |
1834 |
if (dec->frames >= 1) { |
if (dec->low_delay_default && dec->frames > 0) |
1835 |
start_timer(); |
{ |
1836 |
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1837 |
image_output(&dec->refn[0], dec->width, dec->height, |
output = 1; |
|
dec->edged_width, frame->image, frame->stride, |
|
|
frame->colorspace); |
|
|
} else if (vop_type == B_VOP) { |
|
|
image_output(&dec->cur, dec->width, dec->height, |
|
|
dec->edged_width, frame->image, frame->stride, |
|
|
frame->colorspace); |
|
1838 |
} |
} |
1839 |
stop_conv_timer(); |
/* ignore otherwise */ |
1840 |
|
} |
1841 |
|
else if (coding_type != B_VOP) |
1842 |
|
{ |
1843 |
|
switch(coding_type) |
1844 |
|
{ |
1845 |
|
case I_VOP : |
1846 |
|
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1847 |
|
break; |
1848 |
|
case P_VOP : |
1849 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1850 |
|
fcode_forward, intra_dc_threshold, NULL); |
1851 |
|
break; |
1852 |
|
case S_VOP : |
1853 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1854 |
|
fcode_forward, intra_dc_threshold, &gmc_warp); |
1855 |
|
break; |
1856 |
|
case N_VOP : |
1857 |
|
/* XXX: not_coded vops are not used for forward prediction */ |
1858 |
|
/* we should not swap(last_mbs,mbs) */ |
1859 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1860 |
|
break; |
1861 |
|
} |
1862 |
|
|
1863 |
|
if (reduced_resolution) |
1864 |
|
{ |
1865 |
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1866 |
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1867 |
|
16, 0); |
1868 |
|
} |
1869 |
|
|
1870 |
|
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1871 |
|
if (!(dec->low_delay_default && dec->packed_mode)) |
1872 |
|
{ |
1873 |
|
if (dec->low_delay) |
1874 |
|
{ |
1875 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1876 |
|
output = 1; |
1877 |
|
} |
1878 |
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
1879 |
|
{ |
1880 |
|
/* output the reference frame */ |
1881 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1882 |
|
output = 1; |
1883 |
} |
} |
1884 |
} |
} |
|
#endif |
|
1885 |
|
|
|
if (vop_type == I_VOP || vop_type == P_VOP) { |
|
1886 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1887 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1888 |
// swap MACROBLOCK |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
1889 |
if (dec->low_delay && vop_type == P_VOP) |
dec->last_reduced_resolution = reduced_resolution; |
1890 |
mb_swap(&dec->mbs, &dec->last_mbs); |
dec->last_coding_type = coding_type; |
1891 |
|
|
1892 |
|
dec->frames++; |
1893 |
|
seen_something = 1; |
1894 |
|
|
1895 |
|
}else{ /* B_VOP */ |
1896 |
|
|
1897 |
|
if (dec->low_delay) |
1898 |
|
{ |
1899 |
|
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
1900 |
|
dec->low_delay = 1; |
1901 |
} |
} |
1902 |
|
|
1903 |
emms(); |
if (dec->frames < 2) |
1904 |
|
{ |
1905 |
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
1906 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1907 |
|
"broken b-frame, mising ref frames"); |
1908 |
|
}else if (dec->time_pp <= dec->time_bp) { |
1909 |
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1910 |
|
decoded in vfw. */ |
1911 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1912 |
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1913 |
|
}else{ |
1914 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1915 |
|
} |
1916 |
|
|
1917 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1918 |
|
output = 1; |
1919 |
|
dec->frames++; |
1920 |
|
} |
1921 |
|
|
1922 |
|
BitstreamByteAlign(&bs); |
1923 |
|
|
1924 |
|
/* low_delay_default mode: repeat in packed_mode */ |
1925 |
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
1926 |
|
{ |
1927 |
|
success = 1; |
1928 |
|
goto repeat; |
1929 |
|
} |
1930 |
|
|
1931 |
|
done : |
1932 |
|
|
1933 |
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
1934 |
|
then output the recently decoded frame, or print an error message */ |
1935 |
|
if (dec->low_delay_default && output == 0) |
1936 |
|
{ |
1937 |
|
if (dec->packed_mode && seen_something) |
1938 |
|
{ |
1939 |
|
/* output the recently decoded frame */ |
1940 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1941 |
|
} |
1942 |
|
else |
1943 |
|
{ |
1944 |
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1945 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1946 |
|
"warning: nothing to output"); |
1947 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1948 |
|
"bframe decoder lag"); |
1949 |
|
|
1950 |
|
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
1951 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1952 |
|
|
1953 |
|
} |
1954 |
|
} |
1955 |
|
|
1956 |
|
emms(); |
1957 |
stop_global_timer(); |
stop_global_timer(); |
1958 |
|
|
1959 |
return XVID_ERR_OK; |
idct = idct_save; |
1960 |
|
|
1961 |
|
return BitstreamPos(&bs) / 8; /* number of bytes consumed */ |
1962 |
} |
} |