1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* decoder main |
* - 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 |
* 2002-2004 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 |
18 |
* |
* |
19 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
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., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
23 |
*************************************************************************/ |
* $Id: decoder.c,v 1.62 2004-07-14 23:26:06 edgomez Exp $ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
|
* |
|
|
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
|
|
* 22.12.2001 block based interpolation |
|
|
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
24 |
* |
* |
25 |
*************************************************************************/ |
****************************************************************************/ |
26 |
|
|
27 |
|
#include <stdio.h> |
28 |
#include <stdlib.h> |
#include <stdlib.h> |
29 |
#include <string.h> // memset |
#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" |
41 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
42 |
|
|
43 |
#include "quant/quant_h263.h" |
#include "quant/quant.h" |
44 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_matrix.h" |
45 |
#include "dct/idct.h" |
#include "dct/idct.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 |
|
#include "image/qpel.h" |
52 |
|
|
53 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
54 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
55 |
#include "utils/timer.h" |
#include "utils/timer.h" |
56 |
#include "utils/emms.h" |
#include "utils/emms.h" |
57 |
|
#include "motion/motion.h" |
58 |
|
#include "motion/gmc.h" |
59 |
|
|
60 |
#include "image/image.h" |
#include "image/image.h" |
61 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
62 |
|
#include "image/postprocessing.h" |
63 |
|
#include "utils/mem_align.h" |
64 |
|
|
65 |
int decoder_create(XVID_DEC_PARAM * param) |
#ifdef ARCH_IS_IA32 |
66 |
{ |
#define interpolate16x16_quarterpel new_interpolate16x16_quarterpel |
67 |
DECODER * dec; |
#define interpolate8x8_quarterpel new_interpolate8x8_quarterpel |
68 |
|
#endif |
69 |
|
|
70 |
dec = malloc(sizeof(DECODER)); |
static int |
71 |
if (dec == NULL) |
decoder_resize(DECODER * dec) |
72 |
{ |
{ |
73 |
return XVID_ERR_MEMORY; |
/* free existing */ |
74 |
} |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
75 |
param->handle = dec; |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
76 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
77 |
dec->width = param->width; |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
78 |
dec->height = param->height; |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
79 |
|
|
80 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
81 |
|
|
82 |
|
if (dec->last_mbs) |
83 |
|
xvid_free(dec->last_mbs); |
84 |
|
if (dec->mbs) |
85 |
|
xvid_free(dec->mbs); |
86 |
|
if (dec->qscale) |
87 |
|
xvid_free(dec->qscale); |
88 |
|
|
89 |
|
/* realloc */ |
90 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
91 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
92 |
|
|
93 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
94 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
95 |
|
|
96 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
97 |
{ |
xvid_free(dec); |
|
free(dec); |
|
98 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
99 |
} |
} |
100 |
|
|
101 |
if (image_create(&dec->refn, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|
{ |
|
102 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
103 |
free(dec); |
xvid_free(dec); |
104 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
105 |
} |
} |
106 |
|
|
107 |
dec->mbs = malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
/* Support B-frame to reference last 2 frame */ |
108 |
if (dec->mbs == NULL) |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
109 |
{ |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
110 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
111 |
|
xvid_free(dec); |
112 |
|
return XVID_ERR_MEMORY; |
113 |
|
} |
114 |
|
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
115 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
116 |
free(dec); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
117 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
118 |
|
xvid_free(dec); |
119 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
120 |
} |
} |
121 |
|
|
122 |
init_timer(); |
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
123 |
create_vlc_tables(); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
124 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
125 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
126 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
127 |
|
xvid_free(dec); |
128 |
|
return XVID_ERR_MEMORY; |
129 |
|
} |
130 |
|
|
131 |
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
132 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
133 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
134 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
135 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
136 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
137 |
|
xvid_free(dec); |
138 |
|
return XVID_ERR_MEMORY; |
139 |
|
} |
140 |
|
|
141 |
|
dec->mbs = |
142 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
143 |
|
CACHE_LINE); |
144 |
|
if (dec->mbs == NULL) { |
145 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
146 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
147 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
148 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
149 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
150 |
|
xvid_free(dec); |
151 |
|
return XVID_ERR_MEMORY; |
152 |
|
} |
153 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
154 |
|
|
155 |
|
/* For skip MB flag */ |
156 |
|
dec->last_mbs = |
157 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
158 |
|
CACHE_LINE); |
159 |
|
if (dec->last_mbs == NULL) { |
160 |
|
xvid_free(dec->mbs); |
161 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
162 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
163 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
164 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
165 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
166 |
|
xvid_free(dec); |
167 |
|
return XVID_ERR_MEMORY; |
168 |
|
} |
169 |
|
|
170 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
171 |
|
|
172 |
|
/* nothing happens if that fails */ |
173 |
|
dec->qscale = |
174 |
|
xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE); |
175 |
|
|
176 |
|
if (dec->qscale) |
177 |
|
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
178 |
|
|
179 |
return XVID_ERR_OK; |
return 0; |
180 |
} |
} |
181 |
|
|
182 |
|
|
183 |
int decoder_destroy(DECODER * dec) |
int |
184 |
|
decoder_create(xvid_dec_create_t * create) |
185 |
{ |
{ |
186 |
free(dec->mbs); |
DECODER *dec; |
|
image_destroy(&dec->refn, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
free(dec); |
|
187 |
|
|
188 |
destroy_vlc_tables(); |
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
189 |
|
return XVID_ERR_VERSION; |
190 |
|
|
191 |
write_timer(); |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
192 |
return XVID_ERR_OK; |
if (dec == NULL) { |
193 |
|
return XVID_ERR_MEMORY; |
194 |
} |
} |
195 |
|
|
196 |
|
memset(dec, 0, sizeof(DECODER)); |
197 |
|
|
198 |
|
dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE); |
199 |
|
if (dec->mpeg_quant_matrices == NULL) { |
200 |
|
xvid_free(dec); |
201 |
|
return XVID_ERR_MEMORY; |
202 |
|
} |
203 |
|
|
204 |
|
create->handle = dec; |
205 |
|
|
206 |
|
dec->width = create->width; |
207 |
|
dec->height = create->height; |
208 |
|
|
209 |
|
image_null(&dec->cur); |
210 |
|
image_null(&dec->refn[0]); |
211 |
|
image_null(&dec->refn[1]); |
212 |
|
image_null(&dec->tmp); |
213 |
|
image_null(&dec->qtmp); |
214 |
|
|
215 |
|
/* image based GMC */ |
216 |
|
image_null(&dec->gmc); |
217 |
|
|
218 |
|
dec->mbs = NULL; |
219 |
|
dec->last_mbs = NULL; |
220 |
|
dec->qscale = NULL; |
221 |
|
|
222 |
|
init_timer(); |
223 |
|
init_postproc(&dec->postproc); |
224 |
|
init_mpeg_matrix(dec->mpeg_quant_matrices); |
225 |
|
|
226 |
|
/* For B-frame support (used to save reference frame's time */ |
227 |
|
dec->frames = 0; |
228 |
|
dec->time = dec->time_base = dec->last_time_base = 0; |
229 |
|
dec->low_delay = 0; |
230 |
|
dec->packed_mode = 0; |
231 |
|
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
232 |
|
|
233 |
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
234 |
|
|
235 |
|
if (dec->fixed_dimensions) |
236 |
|
return decoder_resize(dec); |
237 |
|
else |
238 |
|
return 0; |
239 |
|
} |
240 |
|
|
241 |
|
|
242 |
static const int32_t dquant_table[4] = |
int |
243 |
|
decoder_destroy(DECODER * dec) |
244 |
{ |
{ |
245 |
|
xvid_free(dec->last_mbs); |
246 |
|
xvid_free(dec->mbs); |
247 |
|
xvid_free(dec->qscale); |
248 |
|
|
249 |
|
/* image based GMC */ |
250 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
251 |
|
|
252 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
253 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
254 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
255 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
256 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
257 |
|
xvid_free(dec->mpeg_quant_matrices); |
258 |
|
xvid_free(dec); |
259 |
|
|
260 |
|
write_timer(); |
261 |
|
return 0; |
262 |
|
} |
263 |
|
|
264 |
|
static const int32_t dquant_table[4] = { |
265 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
266 |
}; |
}; |
267 |
|
|
268 |
|
/* decode an intra macroblock */ |
269 |
|
static void |
270 |
|
decoder_mbintra(DECODER * dec, |
271 |
|
MACROBLOCK * pMB, |
272 |
|
const uint32_t x_pos, |
273 |
|
const uint32_t y_pos, |
274 |
|
const uint32_t acpred_flag, |
275 |
|
const uint32_t cbp, |
276 |
|
Bitstream * bs, |
277 |
|
const uint32_t quant, |
278 |
|
const uint32_t intra_dc_threshold, |
279 |
|
const unsigned int bound, |
280 |
|
const int reduced_resolution) |
281 |
|
{ |
282 |
|
|
283 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
284 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
285 |
|
|
286 |
|
uint32_t stride = dec->edged_width; |
287 |
|
uint32_t stride2 = stride / 2; |
288 |
|
uint32_t next_block = stride * 8; |
289 |
|
uint32_t i; |
290 |
|
uint32_t iQuant = pMB->quant; |
291 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
292 |
|
|
293 |
|
if (reduced_resolution) { |
294 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
295 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
296 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
297 |
|
}else{ |
298 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
299 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
300 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
301 |
|
} |
302 |
|
|
303 |
// decode an intra macroblock |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
|
|
|
|
void decoder_mbintra(DECODER * dec, MACROBLOCK * mb, int x, int y, uint32_t acpred_flag, uint32_t cbp, Bitstream * bs, int quant, int intra_dc_threshold) |
|
|
{ |
|
|
uint32_t k; |
|
304 |
|
|
305 |
for (k = 0; k < 6; k++) |
for (i = 0; i < 6; i++) { |
306 |
{ |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
uint32_t dcscalar; |
|
|
int16_t block[64]; |
|
|
int16_t data[64]; |
|
307 |
int16_t predictors[8]; |
int16_t predictors[8]; |
308 |
int start_coeff; |
int start_coeff; |
309 |
|
|
|
dcscalar = get_dc_scaler(mb->quant, k < 4); |
|
|
|
|
310 |
start_timer(); |
start_timer(); |
311 |
predict_acdc(dec->mbs, x, y, dec->mb_width, k, block, mb->quant, dcscalar, predictors); |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
312 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors, bound); |
313 |
{ |
if (!acpred_flag) { |
314 |
mb->acpred_directions[k] = 0; |
pMB->acpred_directions[i] = 0; |
315 |
} |
} |
316 |
stop_prediction_timer(); |
stop_prediction_timer(); |
317 |
|
|
318 |
memset(block, 0, 64*sizeof(int16_t)); // clear |
if (quant < intra_dc_threshold) { |
|
|
|
|
if (quant < intra_dc_threshold) |
|
|
{ |
|
319 |
int dc_size; |
int dc_size; |
320 |
int dc_dif; |
int dc_dif; |
321 |
|
|
322 |
dc_size = k < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs); |
dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs); |
323 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
324 |
|
|
325 |
if (dc_size > 8) |
if (dc_size > 8) { |
326 |
{ |
BitstreamSkip(bs, 1); /* marker */ |
|
BitstreamSkip(bs, 1); // marker |
|
327 |
} |
} |
328 |
|
|
329 |
block[0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
330 |
start_coeff = 1; |
start_coeff = 1; |
331 |
} |
|
332 |
else |
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif); |
333 |
{ |
} else { |
334 |
start_coeff = 0; |
start_coeff = 0; |
335 |
} |
} |
336 |
|
|
337 |
start_timer(); |
start_timer(); |
338 |
if (cbp & (1 << (5-k))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
339 |
{ |
{ |
340 |
get_intra_block(bs, block, mb->acpred_directions[k], start_coeff); |
int direction = dec->alternate_vertical_scan ? |
341 |
|
2 : pMB->acpred_directions[i]; |
342 |
|
|
343 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
344 |
} |
} |
345 |
stop_coding_timer(); |
stop_coding_timer(); |
346 |
|
|
347 |
start_timer(); |
start_timer(); |
348 |
add_acdc(mb, k, block, dcscalar, predictors); |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version); |
349 |
stop_prediction_timer(); |
stop_prediction_timer(); |
350 |
|
|
351 |
start_timer(); |
start_timer(); |
352 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
353 |
{ |
dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
354 |
dequant_intra(data, block, mb->quant, dcscalar); |
} else { |
355 |
} |
dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
|
else |
|
|
{ |
|
|
dequant4_intra(data, block, mb->quant, dcscalar); |
|
356 |
} |
} |
357 |
stop_iquant_timer(); |
stop_iquant_timer(); |
358 |
|
|
359 |
start_timer(); |
start_timer(); |
360 |
idct(data); |
idct(&data[i * 64]); |
361 |
stop_idct_timer(); |
stop_idct_timer(); |
362 |
|
|
|
start_timer(); |
|
|
if (k < 4) |
|
|
{ |
|
|
transfer_16to8copy(dec->cur.y + (16*y*dec->edged_width) + 16*x + (4*(k&2)*dec->edged_width) + 8*(k&1), data, dec->edged_width); |
|
363 |
} |
} |
364 |
else if (k == 4) |
|
365 |
{ |
if (dec->interlacing && pMB->field_dct) { |
366 |
transfer_16to8copy(dec->cur.u+ 8*y*(dec->edged_width/2) + 8*x, data, (dec->edged_width/2)); |
next_block = stride; |
367 |
|
stride *= 2; |
368 |
} |
} |
369 |
else // if (k == 5) |
|
370 |
|
start_timer(); |
371 |
|
|
372 |
|
if (reduced_resolution) |
373 |
{ |
{ |
374 |
transfer_16to8copy(dec->cur.v + 8*y*(dec->edged_width/2) + 8*x, data, (dec->edged_width/2)); |
next_block*=2; |
375 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
376 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
377 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
378 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
379 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
380 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
381 |
|
}else{ |
382 |
|
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
383 |
|
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
384 |
|
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
385 |
|
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
386 |
|
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
387 |
|
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
388 |
} |
} |
389 |
stop_transfer_timer(); |
stop_transfer_timer(); |
390 |
} |
} |
|
} |
|
391 |
|
|
392 |
|
static void |
393 |
|
decoder_mb_decode(DECODER * dec, |
394 |
|
const uint32_t cbp, |
395 |
|
Bitstream * bs, |
396 |
|
uint8_t * pY_Cur, |
397 |
|
uint8_t * pU_Cur, |
398 |
|
uint8_t * pV_Cur, |
399 |
|
const int reduced_resolution, |
400 |
|
const MACROBLOCK * pMB) |
401 |
|
{ |
402 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
403 |
|
|
404 |
|
int stride = dec->edged_width; |
405 |
|
int next_block = stride * (reduced_resolution ? 16 : 8); |
406 |
|
const int stride2 = stride/2; |
407 |
|
int i; |
408 |
|
const uint32_t iQuant = pMB->quant; |
409 |
|
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
410 |
|
typedef void (*get_inter_block_function_t)( |
411 |
|
Bitstream * bs, |
412 |
|
int16_t * block, |
413 |
|
int direction, |
414 |
|
const int quant, |
415 |
|
const uint16_t *matrix); |
416 |
|
|
417 |
|
const get_inter_block_function_t get_inter_block = (dec->quant_type == 0) |
418 |
|
? get_inter_block_h263 |
419 |
|
: get_inter_block_mpeg; |
420 |
|
|
421 |
|
memset(&data[0], 0, 6*64*sizeof(int16_t)); /* clear */ |
422 |
|
|
423 |
|
for (i = 0; i < 6; i++) { |
424 |
|
|
425 |
|
if (cbp & (1 << (5 - i))) { /* coded */ |
426 |
|
|
|
#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 }; |
|
427 |
|
|
428 |
|
/* Decode coeffs and dequantize on the fly */ |
429 |
|
start_timer(); |
430 |
|
get_inter_block(bs, &data[i*64], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices)); |
431 |
|
stop_coding_timer(); |
432 |
|
|
433 |
// decode an inter macroblock |
start_timer(); |
434 |
|
idct(&data[i * 64]); |
435 |
|
stop_idct_timer(); |
436 |
|
} |
437 |
|
} |
438 |
|
|
439 |
|
if (dec->interlacing && pMB->field_dct) { |
440 |
|
next_block = stride; |
441 |
|
stride *= 2; |
442 |
|
} |
443 |
|
|
444 |
|
start_timer(); |
445 |
|
if (reduced_resolution) { |
446 |
|
if (cbp & 32) |
447 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
448 |
|
if (cbp & 16) |
449 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
450 |
|
if (cbp & 8) |
451 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
452 |
|
if (cbp & 4) |
453 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
454 |
|
if (cbp & 2) |
455 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
456 |
|
if (cbp & 1) |
457 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
458 |
|
} else { |
459 |
|
if (cbp & 32) |
460 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
461 |
|
if (cbp & 16) |
462 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
463 |
|
if (cbp & 8) |
464 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
465 |
|
if (cbp & 4) |
466 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
467 |
|
if (cbp & 2) |
468 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
469 |
|
if (cbp & 1) |
470 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
471 |
|
} |
472 |
|
stop_transfer_timer(); |
473 |
|
} |
474 |
|
|
475 |
|
/* decode an inter macroblock */ |
476 |
|
static void |
477 |
|
decoder_mbinter(DECODER * dec, |
478 |
|
const MACROBLOCK * pMB, |
479 |
|
const uint32_t x_pos, |
480 |
|
const uint32_t y_pos, |
481 |
|
const uint32_t cbp, |
482 |
|
Bitstream * bs, |
483 |
|
const uint32_t rounding, |
484 |
|
const int reduced_resolution, |
485 |
|
const int ref) |
486 |
|
{ |
487 |
|
uint32_t stride = dec->edged_width; |
488 |
|
uint32_t stride2 = stride / 2; |
489 |
|
uint32_t i; |
490 |
|
|
491 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
492 |
|
|
|
void decoder_mbinter(DECODER * dec, MACROBLOCK * mb, int x, int y, uint32_t acpred_flag, uint32_t cbp, Bitstream * bs, int quant, int rounding) |
|
|
{ |
|
|
const uint32_t stride = dec->edged_width; |
|
|
const uint32_t stride2 = dec->edged_width / 2; |
|
493 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
494 |
uint32_t k; |
VECTOR mv[4]; /* local copy of mvs */ |
495 |
|
|
496 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (reduced_resolution) { |
497 |
{ |
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
498 |
uv_dx = mb->mvs[0].x; |
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
499 |
uv_dy = mb->mvs[0].y; |
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
500 |
|
for (i = 0; i < 4; i++) { |
501 |
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
502 |
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
503 |
|
} |
504 |
|
} else { |
505 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
506 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
507 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
508 |
|
for (i = 0; i < 4; i++) |
509 |
|
mv[i] = pMB->mvs[i]; |
510 |
|
} |
511 |
|
|
512 |
|
for (i = 0; i < 4; i++) { |
513 |
|
/* clip to valid range */ |
514 |
|
int border = (int)(dec->mb_width - x_pos) << (5 + dec->quarterpel); |
515 |
|
if (mv[i].x > border) { |
516 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); |
517 |
|
mv[i].x = border; |
518 |
|
} else { |
519 |
|
border = (-(int)x_pos-1) << (5 + dec->quarterpel); |
520 |
|
if (mv[i].x < border) { |
521 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); |
522 |
|
mv[i].x = border; |
523 |
|
} |
524 |
|
} |
525 |
|
|
526 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
border = (int)(dec->mb_height - y_pos) << (5 + dec->quarterpel); |
527 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
if (mv[i].y > border) { |
528 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); |
529 |
|
mv[i].y = border; |
530 |
|
} else { |
531 |
|
border = (-(int)y_pos-1) << (5 + dec->quarterpel); |
532 |
|
if (mv[i].y < border) { |
533 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); |
534 |
|
mv[i].y = border; |
535 |
|
} |
536 |
|
} |
537 |
} |
} |
538 |
|
|
539 |
|
start_timer(); |
540 |
|
|
541 |
|
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
542 |
|
|
543 |
|
uv_dx = mv[0].x; |
544 |
|
uv_dy = mv[0].y; |
545 |
|
if (dec->quarterpel) { |
546 |
|
uv_dx /= 2; |
547 |
|
uv_dy /= 2; |
548 |
|
} |
549 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
550 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
551 |
|
|
552 |
|
if (reduced_resolution) |
553 |
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
554 |
|
mv[0].x, mv[0].y, stride, rounding); |
555 |
|
else if (dec->quarterpel) |
556 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
557 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
558 |
|
mv[0].x, mv[0].y, stride, rounding); |
559 |
else |
else |
560 |
{ |
interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
561 |
int sum; |
mv[0].x, mv[0].y, stride, rounding); |
562 |
sum = mb->mvs[0].x + mb->mvs[1].x + mb->mvs[2].x + mb->mvs[3].x; |
|
563 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
} else { /* MODE_INTER4V */ |
564 |
|
|
565 |
sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y; |
if(dec->quarterpel) { |
566 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
567 |
|
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
568 |
|
} else { |
569 |
|
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
570 |
|
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
571 |
|
} |
572 |
|
|
573 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
574 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
575 |
|
|
576 |
|
if (reduced_resolution) { |
577 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
578 |
|
mv[0].x, mv[0].y, stride, rounding); |
579 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
580 |
|
mv[1].x, mv[1].y, stride, rounding); |
581 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
582 |
|
mv[2].x, mv[2].y, stride, rounding); |
583 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
584 |
|
mv[3].x, mv[3].y, stride, rounding); |
585 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
586 |
|
uv_dx, uv_dy, stride2, rounding); |
587 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
588 |
|
uv_dx, uv_dy, stride2, rounding); |
589 |
|
|
590 |
|
} else if (dec->quarterpel) { |
591 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
592 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
593 |
|
mv[0].x, mv[0].y, stride, rounding); |
594 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
595 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
596 |
|
mv[1].x, mv[1].y, stride, rounding); |
597 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
598 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
599 |
|
mv[2].x, mv[2].y, stride, rounding); |
600 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
601 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
602 |
|
mv[3].x, mv[3].y, stride, rounding); |
603 |
|
} else { |
604 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos, |
605 |
|
mv[0].x, mv[0].y, stride, rounding); |
606 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, |
607 |
|
mv[1].x, mv[1].y, stride, rounding); |
608 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8, |
609 |
|
mv[2].x, mv[2].y, stride, rounding); |
610 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8, |
611 |
|
mv[3].x, mv[3].y, stride, rounding); |
612 |
|
} |
613 |
|
} |
614 |
|
|
615 |
|
/* chroma */ |
616 |
|
if (reduced_resolution) { |
617 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
618 |
|
uv_dx, uv_dy, stride2, rounding); |
619 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
620 |
|
uv_dx, uv_dy, stride2, rounding); |
621 |
|
} else { |
622 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
623 |
|
uv_dx, uv_dy, stride2, rounding); |
624 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
625 |
|
uv_dx, uv_dy, stride2, rounding); |
626 |
} |
} |
627 |
|
|
|
start_timer(); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x, 16*y , mb->mvs[0].x, mb->mvs[0].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x + 8, 16*y , mb->mvs[1].x, mb->mvs[1].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x, 16*y + 8, mb->mvs[2].x, mb->mvs[2].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x + 8, 16*y + 8, mb->mvs[3].x, mb->mvs[3].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.u, dec->refn.u, 8*x, 8*y, uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate8x8_switch(dec->cur.v, dec->refn.v, 8*x, 8*y, uv_dx, uv_dy, stride2, rounding); |
|
628 |
stop_comp_timer(); |
stop_comp_timer(); |
629 |
|
|
630 |
|
if (cbp) |
631 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
632 |
|
reduced_resolution, pMB); |
633 |
|
} |
634 |
|
|
635 |
for (k = 0; k < 6; k++) |
static void |
636 |
|
decoder_mbgmc(DECODER * dec, |
637 |
|
MACROBLOCK * const pMB, |
638 |
|
const uint32_t x_pos, |
639 |
|
const uint32_t y_pos, |
640 |
|
const uint32_t fcode, |
641 |
|
const uint32_t cbp, |
642 |
|
Bitstream * bs, |
643 |
|
const uint32_t rounding) |
644 |
{ |
{ |
645 |
int16_t block[64]; |
const uint32_t stride = dec->edged_width; |
646 |
int16_t data[64]; |
const uint32_t stride2 = stride / 2; |
647 |
|
|
648 |
if (cbp & (1 << (5-k))) // coded |
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
649 |
{ |
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
650 |
memset(block, 0, 64 * sizeof(int16_t)); // clear |
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
651 |
|
|
652 |
start_timer(); |
NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
|
get_inter_block(bs, block); |
|
|
stop_coding_timer(); |
|
653 |
|
|
654 |
start_timer(); |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
|
if (dec->quant_type == 0) |
|
|
{ |
|
|
dequant_inter(data, block, mb->quant); |
|
|
} |
|
|
else |
|
|
{ |
|
|
dequant4_inter(data, block, mb->quant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
655 |
|
|
656 |
start_timer(); |
start_timer(); |
|
idct(data); |
|
|
stop_idct_timer(); |
|
657 |
|
|
658 |
start_timer(); |
/* this is where the calculations are done */ |
659 |
if (k < 4) |
|
660 |
{ |
gmc_data->predict_16x16(gmc_data, |
661 |
transfer_16to8add(dec->cur.y + (16*y + 4*(k&2))*stride + 16*x + 8*(k&1), data, stride); |
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
662 |
} |
stride, stride, x_pos, y_pos, rounding); |
663 |
else if (k == 4) |
|
664 |
{ |
gmc_data->predict_8x8(gmc_data, |
665 |
transfer_16to8add(dec->cur.u + 8*y*stride2 + 8*x, data, stride2); |
dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u, |
666 |
} |
dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v, |
667 |
else // k == 5 |
stride2, stride2, x_pos, y_pos, rounding); |
668 |
{ |
|
669 |
transfer_16to8add(dec->cur.v + 8*y*stride2 + 8*x, data, stride2); |
gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel); |
670 |
} |
|
671 |
|
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
672 |
|
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
673 |
|
|
674 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
675 |
|
|
676 |
stop_transfer_timer(); |
stop_transfer_timer(); |
|
} |
|
|
} |
|
|
} |
|
677 |
|
|
678 |
|
if (cbp) |
679 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
680 |
|
|
681 |
|
} |
682 |
|
|
683 |
|
|
684 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
static void |
685 |
|
decoder_iframe(DECODER * dec, |
686 |
|
Bitstream * bs, |
687 |
|
int reduced_resolution, |
688 |
|
int quant, |
689 |
|
int intra_dc_threshold) |
690 |
{ |
{ |
691 |
|
uint32_t bound; |
692 |
uint32_t x, y; |
uint32_t x, y; |
693 |
|
uint32_t mb_width = dec->mb_width; |
694 |
|
uint32_t mb_height = dec->mb_height; |
695 |
|
|
696 |
for (y = 0; y < dec->mb_height; y++) |
if (reduced_resolution) { |
697 |
{ |
mb_width = (dec->width + 31) / 32; |
698 |
for (x = 0; x < dec->mb_width; x++) |
mb_height = (dec->height + 31) / 32; |
699 |
{ |
} |
700 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
|
701 |
|
bound = 0; |
702 |
|
|
703 |
|
for (y = 0; y < mb_height; y++) { |
704 |
|
for (x = 0; x < mb_width; x++) { |
705 |
|
MACROBLOCK *mb; |
706 |
uint32_t mcbpc; |
uint32_t mcbpc; |
707 |
uint32_t cbpc; |
uint32_t cbpc; |
708 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
709 |
uint32_t cbpy; |
uint32_t cbpy; |
710 |
uint32_t cbp; |
uint32_t cbp; |
711 |
|
|
712 |
|
while (BitstreamShowBits(bs, 9) == 1) |
713 |
|
BitstreamSkip(bs, 9); |
714 |
|
|
715 |
|
if (check_resync_marker(bs, 0)) |
716 |
|
{ |
717 |
|
bound = read_video_packet_header(bs, dec, 0, |
718 |
|
&quant, NULL, NULL, &intra_dc_threshold); |
719 |
|
x = bound % mb_width; |
720 |
|
y = bound / mb_width; |
721 |
|
} |
722 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
723 |
|
|
724 |
|
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
725 |
|
|
726 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
727 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
728 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
729 |
|
|
730 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
731 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
732 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
733 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
734 |
|
|
735 |
if (mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
736 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
737 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
738 |
quant = 31; |
quant = 31; |
739 |
} |
} else if (quant < 1) { |
|
else if (quant < 1) |
|
|
{ |
|
740 |
quant = 1; |
quant = 1; |
741 |
} |
} |
742 |
} |
} |
743 |
mb->quant = quant; |
mb->quant = quant; |
744 |
|
mb->mvs[0].x = mb->mvs[0].y = |
745 |
|
mb->mvs[1].x = mb->mvs[1].y = |
746 |
|
mb->mvs[2].x = mb->mvs[2].y = |
747 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
748 |
|
|
749 |
|
if (dec->interlacing) { |
750 |
|
mb->field_dct = BitstreamGetBit(bs); |
751 |
|
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct); |
752 |
|
} |
753 |
|
|
754 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
755 |
|
intra_dc_threshold, bound, reduced_resolution); |
756 |
|
|
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
|
757 |
} |
} |
758 |
|
if(dec->out_frm) |
759 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
760 |
} |
} |
761 |
|
|
762 |
} |
} |
763 |
|
|
764 |
|
|
765 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
static void |
766 |
|
get_motion_vector(DECODER * dec, |
767 |
|
Bitstream * bs, |
768 |
|
int x, |
769 |
|
int y, |
770 |
|
int k, |
771 |
|
VECTOR * ret_mv, |
772 |
|
int fcode, |
773 |
|
const int bound) |
774 |
{ |
{ |
|
int scale_fac = 1 << (fcode - 1); |
|
|
int high = (32 * scale_fac) - 1; |
|
|
int low = ((-32) * scale_fac); |
|
|
int range = (64 * scale_fac); |
|
|
|
|
|
VECTOR pmv[4]; |
|
|
uint32_t psad[4]; |
|
775 |
|
|
776 |
int mv_x, mv_y; |
const int scale_fac = 1 << (fcode - 1); |
777 |
int pmv_x, pmv_y; |
const int high = (32 * scale_fac) - 1; |
778 |
|
const int low = ((-32) * scale_fac); |
779 |
|
const int range = (64 * scale_fac); |
780 |
|
|
781 |
|
const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
782 |
|
VECTOR mv; |
783 |
|
|
784 |
get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad); |
mv.x = get_mv(bs, fcode); |
785 |
|
mv.y = get_mv(bs, fcode); |
786 |
|
|
787 |
pmv_x = pmv[0].x; |
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); |
|
pmv_y = pmv[0].y; |
|
788 |
|
|
789 |
mv_x = get_mv(bs, fcode); |
mv.x += pmv.x; |
790 |
mv_y = get_mv(bs, fcode); |
mv.y += pmv.y; |
791 |
|
|
792 |
mv_x += pmv_x; |
if (mv.x < low) { |
793 |
mv_y += pmv_y; |
mv.x += range; |
794 |
|
} else if (mv.x > high) { |
795 |
|
mv.x -= range; |
796 |
|
} |
797 |
|
|
798 |
if (mv_x < low) |
if (mv.y < low) { |
799 |
{ |
mv.y += range; |
800 |
mv_x += range; |
} else if (mv.y > high) { |
801 |
|
mv.y -= range; |
802 |
} |
} |
803 |
else if (mv_x > high) |
|
804 |
{ |
ret_mv->x = mv.x; |
805 |
mv_x -= range; |
ret_mv->y = mv.y; |
806 |
} |
} |
807 |
|
|
808 |
if (mv_y < low) |
/* for P_VOP set gmc_warp to NULL */ |
809 |
|
static void |
810 |
|
decoder_pframe(DECODER * dec, |
811 |
|
Bitstream * bs, |
812 |
|
int rounding, |
813 |
|
int reduced_resolution, |
814 |
|
int quant, |
815 |
|
int fcode, |
816 |
|
int intra_dc_threshold, |
817 |
|
const WARPPOINTS *const gmc_warp) |
818 |
{ |
{ |
819 |
mv_y += range; |
uint32_t x, y; |
820 |
|
uint32_t bound; |
821 |
|
int cp_mb, st_mb; |
822 |
|
uint32_t mb_width = dec->mb_width; |
823 |
|
uint32_t mb_height = dec->mb_height; |
824 |
|
|
825 |
|
if (reduced_resolution) { |
826 |
|
mb_width = (dec->width + 31) / 32; |
827 |
|
mb_height = (dec->height + 31) / 32; |
828 |
} |
} |
829 |
else if (mv_y > high) |
|
830 |
{ |
if (!dec->is_edged[0]) { |
831 |
mv_y -= range; |
start_timer(); |
832 |
|
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
833 |
|
dec->width, dec->height, dec->bs_version); |
834 |
|
dec->is_edged[0] = 1; |
835 |
|
stop_edges_timer(); |
836 |
} |
} |
837 |
|
|
838 |
mv->x = mv_x; |
if (gmc_warp) { |
839 |
mv->y = mv_y; |
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
840 |
|
generate_GMCparameters( dec->sprite_warping_points, |
841 |
|
dec->sprite_warping_accuracy, gmc_warp, |
842 |
|
dec->width, dec->height, &dec->new_gmc_data); |
843 |
|
|
844 |
|
/* image warping is done block-based in decoder_mbgmc(), now */ |
845 |
} |
} |
846 |
|
|
847 |
|
bound = 0; |
848 |
|
|
849 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
for (y = 0; y < mb_height; y++) { |
850 |
{ |
cp_mb = st_mb = 0; |
851 |
uint32_t x, y; |
for (x = 0; x < mb_width; x++) { |
852 |
|
MACROBLOCK *mb; |
853 |
|
|
854 |
image_swap(&dec->cur, &dec->refn); |
/* skip stuffing */ |
855 |
|
while (BitstreamShowBits(bs, 10) == 1) |
856 |
|
BitstreamSkip(bs, 10); |
857 |
|
|
858 |
start_timer(); |
if (check_resync_marker(bs, fcode - 1)) { |
859 |
image_setedges(&dec->refn, dec->edged_width, dec->edged_height, dec->width, dec->height); |
bound = read_video_packet_header(bs, dec, fcode - 1, |
860 |
stop_edges_timer(); |
&quant, &fcode, NULL, &intra_dc_threshold); |
861 |
|
x = bound % mb_width; |
862 |
|
y = bound / mb_width; |
863 |
|
} |
864 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
865 |
|
|
866 |
for (y = 0; y < dec->mb_height; y++) |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
|
{ |
|
|
for (x = 0; x < dec->mb_width; x++) |
|
|
{ |
|
|
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
|
867 |
|
|
868 |
if (!BitstreamGetBit(bs)) // not_coded |
if (!(BitstreamGetBit(bs))) { /* block _is_ coded */ |
869 |
{ |
uint32_t mcbpc, cbpc, cbpy, cbp; |
870 |
uint32_t mcbpc; |
uint32_t intra, acpred_flag = 0; |
871 |
uint32_t cbpc; |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
|
uint32_t acpred_flag; |
|
|
uint32_t cbpy; |
|
|
uint32_t cbp; |
|
|
uint32_t intra; |
|
872 |
|
|
873 |
|
cp_mb++; |
874 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
875 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
876 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
877 |
|
|
878 |
|
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
879 |
|
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
880 |
|
|
881 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
882 |
|
|
883 |
if (intra) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
884 |
{ |
mcsel = BitstreamGetBit(bs); |
885 |
|
else if (intra) |
886 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
|
} |
|
|
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
887 |
|
|
888 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
889 |
|
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
890 |
|
|
891 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
892 |
|
|
893 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
894 |
{ |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
895 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant); |
896 |
if (quant > 31) |
quant += dquant; |
897 |
{ |
if (quant > 31) { |
898 |
quant = 31; |
quant = 31; |
899 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
900 |
quant = 1; |
quant = 1; |
901 |
} |
} |
902 |
|
DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant); |
903 |
} |
} |
904 |
mb->quant = quant; |
mb->quant = quant; |
905 |
|
|
906 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (dec->interlacing) { |
907 |
{ |
if (cbp || intra) { |
908 |
|
mb->field_dct = BitstreamGetBit(bs); |
909 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
910 |
|
} |
911 |
|
|
912 |
|
if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) { |
913 |
|
mb->field_pred = BitstreamGetBit(bs); |
914 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
915 |
|
|
916 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
if (mb->field_pred) { |
917 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
mb->field_for_top = BitstreamGetBit(bs); |
918 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
919 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
920 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
921 |
} |
} |
|
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
|
|
{ |
|
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
|
|
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
|
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
|
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
|
922 |
} |
} |
923 |
else // MODE_INTRA, MODE_INTRA_Q |
} |
924 |
{ |
|
925 |
|
if (mcsel) { |
926 |
|
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding); |
927 |
|
continue; |
928 |
|
|
929 |
|
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
930 |
|
|
931 |
|
if (dec->interlacing && mb->field_pred) { |
932 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
933 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound); |
934 |
|
} else { |
935 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
936 |
|
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
937 |
|
} |
938 |
|
} else if (mb->mode == MODE_INTER4V ) { |
939 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
940 |
|
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
941 |
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
942 |
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
943 |
|
} else { /* MODE_INTRA, MODE_INTRA_Q */ |
944 |
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; |
945 |
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; |
946 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
947 |
|
intra_dc_threshold, bound, reduced_resolution); |
948 |
continue; |
continue; |
949 |
} |
} |
950 |
|
|
951 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
952 |
} |
rounding, reduced_resolution, 0); |
|
else // not coded |
|
|
{ |
|
953 |
|
|
954 |
|
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
955 |
|
mb->mode = MODE_NOT_CODED_GMC; |
956 |
|
mb->quant = quant; |
957 |
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding); |
958 |
|
|
959 |
|
if(dec->out_frm && cp_mb > 0) { |
960 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
961 |
|
cp_mb = 0; |
962 |
|
} |
963 |
|
st_mb = x+1; |
964 |
|
} else { /* not coded P_VOP macroblock */ |
965 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
966 |
|
mb->quant = quant; |
967 |
|
|
968 |
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; |
969 |
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; |
970 |
|
|
971 |
// copy macroblock directly from ref to cur |
decoder_mbinter(dec, mb, x, y, 0, bs, |
972 |
|
rounding, reduced_resolution, 0); |
973 |
|
|
974 |
|
if(dec->out_frm && cp_mb > 0) { |
975 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
976 |
|
cp_mb = 0; |
977 |
|
} |
978 |
|
st_mb = x+1; |
979 |
|
} |
980 |
|
} |
981 |
|
|
982 |
|
if(dec->out_frm && cp_mb > 0) |
983 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
984 |
|
} |
985 |
|
} |
986 |
|
|
987 |
|
|
988 |
|
/* decode B-frame motion vector */ |
989 |
|
static void |
990 |
|
get_b_motion_vector(Bitstream * bs, |
991 |
|
VECTOR * mv, |
992 |
|
int fcode, |
993 |
|
const VECTOR pmv, |
994 |
|
const DECODER * const dec, |
995 |
|
const int x, const int y) |
996 |
|
{ |
997 |
|
const int scale_fac = 1 << (fcode - 1); |
998 |
|
const int high = (32 * scale_fac) - 1; |
999 |
|
const int low = ((-32) * scale_fac); |
1000 |
|
const int range = (64 * scale_fac); |
1001 |
|
|
1002 |
|
int mv_x = get_mv(bs, fcode); |
1003 |
|
int mv_y = get_mv(bs, fcode); |
1004 |
|
|
1005 |
|
mv_x += pmv.x; |
1006 |
|
mv_y += pmv.y; |
1007 |
|
|
1008 |
|
if (mv_x < low) |
1009 |
|
mv_x += range; |
1010 |
|
else if (mv_x > high) |
1011 |
|
mv_x -= range; |
1012 |
|
|
1013 |
|
if (mv_y < low) |
1014 |
|
mv_y += range; |
1015 |
|
else if (mv_y > high) |
1016 |
|
mv_y -= range; |
1017 |
|
|
1018 |
|
mv->x = mv_x; |
1019 |
|
mv->y = mv_y; |
1020 |
|
} |
1021 |
|
|
1022 |
|
/* decode an B-frame direct & interpolate macroblock */ |
1023 |
|
static void |
1024 |
|
decoder_bf_interpolate_mbinter(DECODER * dec, |
1025 |
|
IMAGE forward, |
1026 |
|
IMAGE backward, |
1027 |
|
const MACROBLOCK * pMB, |
1028 |
|
const uint32_t x_pos, |
1029 |
|
const uint32_t y_pos, |
1030 |
|
Bitstream * bs, |
1031 |
|
const int direct) |
1032 |
|
{ |
1033 |
|
uint32_t stride = dec->edged_width; |
1034 |
|
uint32_t stride2 = stride / 2; |
1035 |
|
int uv_dx, uv_dy; |
1036 |
|
int b_uv_dx, b_uv_dy; |
1037 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1038 |
|
const uint32_t cbp = pMB->cbp; |
1039 |
|
|
1040 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1041 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1042 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1043 |
|
|
1044 |
|
if (!direct) { |
1045 |
|
uv_dx = pMB->mvs[0].x; |
1046 |
|
uv_dy = pMB->mvs[0].y; |
1047 |
|
b_uv_dx = pMB->b_mvs[0].x; |
1048 |
|
b_uv_dy = pMB->b_mvs[0].y; |
1049 |
|
|
1050 |
|
if (dec->quarterpel) { |
1051 |
|
uv_dx /= 2; |
1052 |
|
uv_dy /= 2; |
1053 |
|
b_uv_dx /= 2; |
1054 |
|
b_uv_dy /= 2; |
1055 |
|
} |
1056 |
|
|
1057 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1058 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1059 |
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1060 |
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1061 |
|
|
1062 |
|
} else { |
1063 |
|
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1064 |
|
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1065 |
|
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1066 |
|
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1067 |
|
|
1068 |
|
if (dec->quarterpel) { |
1069 |
|
uv_dx /= 2; |
1070 |
|
uv_dy /= 2; |
1071 |
|
b_uv_dx /= 2; |
1072 |
|
b_uv_dy /= 2; |
1073 |
|
} |
1074 |
|
|
1075 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1076 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
1077 |
|
b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf]; |
1078 |
|
b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf]; |
1079 |
|
} |
1080 |
|
|
1081 |
start_timer(); |
start_timer(); |
1082 |
|
if(dec->quarterpel) { |
1083 |
|
if(!direct) { |
1084 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1085 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1086 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1087 |
|
} else { |
1088 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1089 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1090 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1091 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1092 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1093 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1094 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1095 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1096 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1097 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1098 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1099 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1100 |
|
} |
1101 |
|
} else { |
1102 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1103 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1104 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1105 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1106 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1107 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1108 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8, |
1109 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1110 |
|
} |
1111 |
|
|
1112 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1113 |
|
uv_dy, stride2, 0); |
1114 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1115 |
|
uv_dy, stride2, 0); |
1116 |
|
|
1117 |
|
|
1118 |
|
if(dec->quarterpel) { |
1119 |
|
if(!direct) { |
1120 |
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1121 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1122 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1123 |
|
} else { |
1124 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1125 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1126 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1127 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1128 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1129 |
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1130 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1131 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1132 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1133 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1134 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1135 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1136 |
|
} |
1137 |
|
} else { |
1138 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1139 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1140 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1141 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1142 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1143 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1144 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1145 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1146 |
|
} |
1147 |
|
|
1148 |
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1149 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1150 |
|
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
1151 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1152 |
|
|
1153 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1154 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1155 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1156 |
|
stride, 1, 8); |
1157 |
|
|
1158 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1159 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1160 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1161 |
|
stride, 1, 8); |
1162 |
|
|
1163 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1164 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1165 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1166 |
|
stride, 1, 8); |
1167 |
|
|
1168 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1169 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1170 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1171 |
|
stride, 1, 8); |
1172 |
|
|
1173 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1174 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1175 |
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1176 |
|
stride2, 1, 8); |
1177 |
|
|
1178 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1179 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1180 |
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1181 |
|
stride2, 1, 8); |
1182 |
|
|
1183 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
stop_comp_timer(); |
|
dec->refn.y + (16*y)*dec->edged_width + (16*x), |
|
|
dec->edged_width); |
|
1184 |
|
|
1185 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
if (cbp) |
1186 |
dec->refn.y + (16*y)*dec->edged_width + (16*x+8), |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
1187 |
dec->edged_width); |
} |
1188 |
|
|
1189 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
/* for decode B-frame dbquant */ |
1190 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x), |
static __inline int32_t |
1191 |
dec->edged_width); |
get_dbquant(Bitstream * bs) |
1192 |
|
{ |
1193 |
|
if (!BitstreamGetBit(bs)) /* '0' */ |
1194 |
|
return (0); |
1195 |
|
else if (!BitstreamGetBit(bs)) /* '10' */ |
1196 |
|
return (-2); |
1197 |
|
else /* '11' */ |
1198 |
|
return (2); |
1199 |
|
} |
1200 |
|
|
1201 |
|
/* |
1202 |
|
* decode B-frame mb_type |
1203 |
|
* bit ret_value |
1204 |
|
* 1 0 |
1205 |
|
* 01 1 |
1206 |
|
* 001 2 |
1207 |
|
* 0001 3 |
1208 |
|
*/ |
1209 |
|
static int32_t __inline |
1210 |
|
get_mbtype(Bitstream * bs) |
1211 |
|
{ |
1212 |
|
int32_t mb_type; |
1213 |
|
|
1214 |
|
for (mb_type = 0; mb_type <= 3; mb_type++) |
1215 |
|
if (BitstreamGetBit(bs)) |
1216 |
|
return (mb_type); |
1217 |
|
|
1218 |
|
return -1; |
1219 |
|
} |
1220 |
|
|
1221 |
|
static void |
1222 |
|
decoder_bframe(DECODER * dec, |
1223 |
|
Bitstream * bs, |
1224 |
|
int quant, |
1225 |
|
int fcode_forward, |
1226 |
|
int fcode_backward) |
1227 |
|
{ |
1228 |
|
uint32_t x, y; |
1229 |
|
VECTOR mv; |
1230 |
|
const VECTOR zeromv = {0,0}; |
1231 |
|
int i; |
1232 |
|
|
1233 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
if (!dec->is_edged[0]) { |
1234 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x+8), |
start_timer(); |
1235 |
dec->edged_width); |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1236 |
|
dec->width, dec->height, dec->bs_version); |
1237 |
|
dec->is_edged[0] = 1; |
1238 |
|
stop_edges_timer(); |
1239 |
|
} |
1240 |
|
|
1241 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
if (!dec->is_edged[1]) { |
1242 |
dec->refn.u + (8*y)*dec->edged_width/2 + (8*x), |
start_timer(); |
1243 |
dec->edged_width/2); |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1244 |
|
dec->width, dec->height, dec->bs_version); |
1245 |
|
dec->is_edged[1] = 1; |
1246 |
|
stop_edges_timer(); |
1247 |
|
} |
1248 |
|
|
1249 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
for (y = 0; y < dec->mb_height; y++) { |
1250 |
dec->refn.v + (8*y)*dec->edged_width/2 + (8*x), |
/* Initialize Pred Motion Vector */ |
1251 |
dec->edged_width/2); |
dec->p_fmv = dec->p_bmv = zeromv; |
1252 |
|
for (x = 0; x < dec->mb_width; x++) { |
1253 |
|
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1254 |
|
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1255 |
|
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1256 |
|
uint32_t intra_dc_threshold; /* fake variable */ |
1257 |
|
|
1258 |
|
if (check_resync_marker(bs, fcode_max - 1)) { |
1259 |
|
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1260 |
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1261 |
|
x = bound % dec->mb_width; |
1262 |
|
y = bound / dec->mb_width; |
1263 |
|
/* reset predicted macroblocks */ |
1264 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1265 |
|
} |
1266 |
|
|
1267 |
|
mv = |
1268 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1269 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1270 |
|
mb->quant = quant; |
1271 |
|
|
1272 |
stop_transfer_timer(); |
/* |
1273 |
|
* skip if the co-located P_VOP macroblock is not coded |
1274 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1275 |
|
* automatically skipped |
1276 |
|
*/ |
1277 |
|
|
1278 |
|
if (last_mb->mode == MODE_NOT_CODED) { |
1279 |
|
mb->cbp = 0; |
1280 |
|
mb->mode = MODE_FORWARD; |
1281 |
|
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
1282 |
|
continue; |
1283 |
|
} |
1284 |
|
|
1285 |
|
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1286 |
|
const uint8_t modb2 = BitstreamGetBit(bs); |
1287 |
|
|
1288 |
|
mb->mode = get_mbtype(bs); |
1289 |
|
|
1290 |
|
if (!modb2) /* modb=='00' */ |
1291 |
|
mb->cbp = BitstreamGetBits(bs, 6); |
1292 |
|
else |
1293 |
|
mb->cbp = 0; |
1294 |
|
|
1295 |
|
if (mb->mode && mb->cbp) { |
1296 |
|
quant += get_dbquant(bs); |
1297 |
|
if (quant > 31) |
1298 |
|
quant = 31; |
1299 |
|
else if (quant < 1) |
1300 |
|
quant = 1; |
1301 |
|
} |
1302 |
|
mb->quant = quant; |
1303 |
|
|
1304 |
|
if (dec->interlacing) { |
1305 |
|
if (mb->cbp) { |
1306 |
|
mb->field_dct = BitstreamGetBit(bs); |
1307 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1308 |
|
} |
1309 |
|
|
1310 |
|
if (mb->mode) { |
1311 |
|
mb->field_pred = BitstreamGetBit(bs); |
1312 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1313 |
|
|
1314 |
|
if (mb->field_pred) { |
1315 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1316 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1317 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1318 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1319 |
|
} |
1320 |
|
} |
1321 |
|
} |
1322 |
|
|
1323 |
|
} else { |
1324 |
|
mb->mode = MODE_DIRECT_NONE_MV; |
1325 |
|
mb->cbp = 0; |
1326 |
} |
} |
1327 |
|
|
1328 |
|
switch (mb->mode) { |
1329 |
|
case MODE_DIRECT: |
1330 |
|
get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y); |
1331 |
|
|
1332 |
|
case MODE_DIRECT_NONE_MV: |
1333 |
|
for (i = 0; i < 4; i++) { |
1334 |
|
mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x; |
1335 |
|
mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y; |
1336 |
|
|
1337 |
|
mb->b_mvs[i].x = (mv.x) |
1338 |
|
? mb->mvs[i].x - last_mb->mvs[i].x |
1339 |
|
: last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp; |
1340 |
|
mb->b_mvs[i].y = (mv.y) |
1341 |
|
? mb->mvs[i].y - last_mb->mvs[i].y |
1342 |
|
: last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp; |
1343 |
} |
} |
1344 |
|
|
1345 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1346 |
|
mb, x, y, bs, 1); |
1347 |
|
break; |
1348 |
|
|
1349 |
|
case MODE_INTERPOLATE: |
1350 |
|
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1351 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1352 |
|
|
1353 |
|
get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1354 |
|
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0]; |
1355 |
|
|
1356 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1357 |
|
mb, x, y, bs, 0); |
1358 |
|
break; |
1359 |
|
|
1360 |
|
case MODE_BACKWARD: |
1361 |
|
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1362 |
|
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1363 |
|
|
1364 |
|
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0); |
1365 |
|
break; |
1366 |
|
|
1367 |
|
case MODE_FORWARD: |
1368 |
|
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1369 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1370 |
|
|
1371 |
|
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
1372 |
|
break; |
1373 |
|
|
1374 |
|
default: |
1375 |
|
DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode); |
1376 |
|
} |
1377 |
|
} /* End of for */ |
1378 |
} |
} |
1379 |
} |
} |
1380 |
|
|
1381 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
/* perform post processing if necessary, and output the image */ |
1382 |
|
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1383 |
|
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1384 |
|
int coding_type, int quant) |
1385 |
{ |
{ |
1386 |
|
const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0; |
1387 |
|
|
1388 |
|
if (dec->cartoon_mode) |
1389 |
|
frame->general &= ~XVID_FILMEFFECT; |
1390 |
|
|
1391 |
|
if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0) |
1392 |
|
&& mbs != NULL) /* post process */ |
1393 |
|
{ |
1394 |
|
/* note: image is stored to tmp */ |
1395 |
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1396 |
|
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
1397 |
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1398 |
|
frame->general, brightness, dec->frames, (coding_type == B_VOP)); |
1399 |
|
img = &dec->tmp; |
1400 |
|
} |
1401 |
|
|
1402 |
|
image_output(img, dec->width, dec->height, |
1403 |
|
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1404 |
|
frame->output.csp, dec->interlacing); |
1405 |
|
|
1406 |
|
if (stats) { |
1407 |
|
stats->type = coding2type(coding_type); |
1408 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1409 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1410 |
|
stats->data.vop.qscale_stride = dec->mb_width; |
1411 |
|
stats->data.vop.qscale = dec->qscale; |
1412 |
|
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1413 |
|
int i; |
1414 |
|
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1415 |
|
stats->data.vop.qscale[i] = mbs[i].quant; |
1416 |
|
} else |
1417 |
|
stats->data.vop.qscale = NULL; |
1418 |
|
} |
1419 |
|
} |
1420 |
|
|
1421 |
|
|
1422 |
|
int |
1423 |
|
decoder_decode(DECODER * dec, |
1424 |
|
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1425 |
|
{ |
1426 |
|
|
1427 |
Bitstream bs; |
Bitstream bs; |
1428 |
uint32_t rounding; |
uint32_t rounding; |
1429 |
uint32_t quant; |
uint32_t reduced_resolution; |
1430 |
uint32_t fcode; |
uint32_t quant = 2; |
1431 |
|
uint32_t fcode_forward; |
1432 |
|
uint32_t fcode_backward; |
1433 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1434 |
|
WARPPOINTS gmc_warp; |
1435 |
|
int coding_type; |
1436 |
|
int success, output, seen_something; |
1437 |
|
|
1438 |
|
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1439 |
|
return XVID_ERR_VERSION; |
1440 |
|
|
1441 |
start_global_timer(); |
start_global_timer(); |
1442 |
|
|
1443 |
|
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1444 |
|
if ((frame->general & XVID_DISCONTINUITY)) |
1445 |
|
dec->frames = 0; |
1446 |
|
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1447 |
|
|
1448 |
|
if (frame->length < 0) { /* decoder flush */ |
1449 |
|
int ret; |
1450 |
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1451 |
|
we have a reference frame, then outout the reference frame */ |
1452 |
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1453 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1454 |
|
dec->frames = 0; |
1455 |
|
ret = 0; |
1456 |
|
} else { |
1457 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1458 |
|
ret = XVID_ERR_END; |
1459 |
|
} |
1460 |
|
|
1461 |
|
emms(); |
1462 |
|
stop_global_timer(); |
1463 |
|
return ret; |
1464 |
|
} |
1465 |
|
|
1466 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1467 |
|
|
1468 |
switch (BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold)) |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1469 |
|
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1470 |
{ |
{ |
1471 |
case P_VOP : |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1472 |
decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold); |
(uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing); |
1473 |
break; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1474 |
|
emms(); |
1475 |
|
return 1; /* one byte consumed */ |
1476 |
|
} |
1477 |
|
|
1478 |
case I_VOP : |
success = 0; |
1479 |
//DEBUG1("",intra_dc_threshold); |
output = 0; |
1480 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
seen_something = 0; |
|
break; |
|
1481 |
|
|
1482 |
case B_VOP : // ignore |
repeat: |
|
break; |
|
1483 |
|
|
1484 |
case N_VOP : // vop not coded |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1485 |
break; |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1486 |
|
|
1487 |
default : |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
1488 |
return XVID_ERR_FAIL; |
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1489 |
|
|
1490 |
|
if (coding_type == -1) { /* nothing */ |
1491 |
|
if (success) goto done; |
1492 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1493 |
|
emms(); |
1494 |
|
return BitstreamPos(&bs)/8; |
1495 |
} |
} |
1496 |
|
|
1497 |
frame->length = BitstreamPos(&bs) / 8; |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1498 |
|
|
1499 |
start_timer(); |
if (coding_type == -3) |
1500 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
decoder_resize(dec); |
|
frame->image, frame->stride, frame->colorspace); |
|
|
stop_conv_timer(); |
|
1501 |
|
|
1502 |
|
if (stats) { |
1503 |
|
stats->type = XVID_TYPE_VOL; |
1504 |
|
stats->data.vol.general = 0; |
1505 |
|
/*XXX: if (dec->interlacing) |
1506 |
|
stats->data.vol.general |= ++INTERLACING; */ |
1507 |
|
stats->data.vol.width = dec->width; |
1508 |
|
stats->data.vol.height = dec->height; |
1509 |
|
stats->data.vol.par = dec->aspect_ratio; |
1510 |
|
stats->data.vol.par_width = dec->par_width; |
1511 |
|
stats->data.vol.par_height = dec->par_height; |
1512 |
emms(); |
emms(); |
1513 |
|
return BitstreamPos(&bs)/8; /* number of bytes consumed */ |
1514 |
|
} |
1515 |
|
goto repeat; |
1516 |
|
} |
1517 |
|
|
1518 |
|
if(dec->frames == 0 && coding_type != I_VOP) { |
1519 |
|
/* 1st frame is not an i-vop */ |
1520 |
|
goto repeat; |
1521 |
|
} |
1522 |
|
|
1523 |
|
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1524 |
|
|
1525 |
|
/* packed_mode: special-N_VOP treament */ |
1526 |
|
if (dec->packed_mode && coding_type == N_VOP) { |
1527 |
|
if (dec->low_delay_default && dec->frames > 0) { |
1528 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1529 |
|
output = 1; |
1530 |
|
} |
1531 |
|
/* ignore otherwise */ |
1532 |
|
} else if (coding_type != B_VOP) { |
1533 |
|
switch(coding_type) { |
1534 |
|
case I_VOP : |
1535 |
|
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1536 |
|
break; |
1537 |
|
case P_VOP : |
1538 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1539 |
|
fcode_forward, intra_dc_threshold, NULL); |
1540 |
|
break; |
1541 |
|
case S_VOP : |
1542 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1543 |
|
fcode_forward, intra_dc_threshold, &gmc_warp); |
1544 |
|
break; |
1545 |
|
case N_VOP : |
1546 |
|
/* XXX: not_coded vops are not used for forward prediction */ |
1547 |
|
/* we should not swap(last_mbs,mbs) */ |
1548 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1549 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */ |
1550 |
|
break; |
1551 |
|
} |
1552 |
|
|
1553 |
|
if (reduced_resolution) { |
1554 |
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1555 |
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1556 |
|
16, 0); |
1557 |
|
} |
1558 |
|
|
1559 |
|
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1560 |
|
if (!(dec->low_delay_default && dec->packed_mode)) { |
1561 |
|
if (dec->low_delay) { |
1562 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
1563 |
|
output = 1; |
1564 |
|
} else if (dec->frames > 0) { /* is the reference frame valid? */ |
1565 |
|
/* output the reference frame */ |
1566 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1567 |
|
output = 1; |
1568 |
|
} |
1569 |
|
} |
1570 |
|
|
1571 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1572 |
|
dec->is_edged[1] = dec->is_edged[0]; |
1573 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1574 |
|
dec->is_edged[0] = 0; |
1575 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
1576 |
|
dec->last_reduced_resolution = reduced_resolution; |
1577 |
|
dec->last_coding_type = coding_type; |
1578 |
|
|
1579 |
|
dec->frames++; |
1580 |
|
seen_something = 1; |
1581 |
|
|
1582 |
|
} else { /* B_VOP */ |
1583 |
|
|
1584 |
|
if (dec->low_delay) { |
1585 |
|
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
1586 |
|
dec->low_delay = 0; |
1587 |
|
} |
1588 |
|
|
1589 |
|
if (dec->frames < 2) { |
1590 |
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
1591 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1592 |
|
"broken b-frame, mising ref frames"); |
1593 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1594 |
|
} else if (dec->time_pp <= dec->time_bp) { |
1595 |
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1596 |
|
decoded in vfw. */ |
1597 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1598 |
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1599 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1600 |
|
} else { |
1601 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1602 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
1603 |
|
} |
1604 |
|
|
1605 |
|
output = 1; |
1606 |
|
dec->frames++; |
1607 |
|
} |
1608 |
|
|
1609 |
|
#if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */ |
1610 |
|
BitstreamByteAlign(&bs); |
1611 |
|
#endif |
1612 |
|
|
1613 |
|
/* low_delay_default mode: repeat in packed_mode */ |
1614 |
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) { |
1615 |
|
success = 1; |
1616 |
|
goto repeat; |
1617 |
|
} |
1618 |
|
|
1619 |
|
done : |
1620 |
|
|
1621 |
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
1622 |
|
then output the recently decoded frame, or print an error message */ |
1623 |
|
if (dec->low_delay_default && output == 0) { |
1624 |
|
if (dec->packed_mode && seen_something) { |
1625 |
|
/* output the recently decoded frame */ |
1626 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1627 |
|
} else { |
1628 |
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1629 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1630 |
|
"warning: nothing to output"); |
1631 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1632 |
|
"bframe decoder lag"); |
1633 |
|
|
1634 |
|
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1635 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1636 |
|
} |
1637 |
|
} |
1638 |
|
|
1639 |
|
emms(); |
1640 |
stop_global_timer(); |
stop_global_timer(); |
1641 |
|
|
1642 |
return XVID_ERR_OK; |
return (BitstreamPos(&bs) + 7) / 8; /* number of bytes consumed */ |
1643 |
} |
} |