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 xvid_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 |
11 |
* the xvid_free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation ; either version 2 of the License, or |
12 |
* (at your option) any later version. |
* (at your option) any later version. |
13 |
* |
* |
14 |
* This program is distributed in the hope that it will be useful, |
* This program is distributed in the hope that it will be useful, |
17 |
* GNU General Public License for more details. |
* GNU General Public License for more details. |
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 xvid_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 1.54 2004-04-11 09:41:27 syskin Exp $ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
24 |
* |
* |
25 |
* 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 block based interpolation |
|
|
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
|
*************************************************************************/ |
|
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 |
|
|
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 "image/postprocessing.h" |
62 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
63 |
|
|
64 |
int decoder_create(XVID_DEC_PARAM * param) |
static int |
65 |
{ |
decoder_resize(DECODER * dec) |
|
DECODER * dec; |
|
|
|
|
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
|
|
if (dec == NULL) |
|
66 |
{ |
{ |
67 |
return XVID_ERR_MEMORY; |
/* free existing */ |
68 |
} |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
69 |
param->handle = dec; |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
70 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
71 |
dec->width = param->width; |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
72 |
dec->height = param->height; |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
73 |
|
|
74 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
75 |
|
|
76 |
|
if (dec->last_mbs) |
77 |
|
xvid_free(dec->last_mbs); |
78 |
|
if (dec->mbs) |
79 |
|
xvid_free(dec->mbs); |
80 |
|
if (dec->qscale) |
81 |
|
xvid_free(dec->qscale); |
82 |
|
|
83 |
|
/* realloc */ |
84 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
85 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
86 |
|
|
87 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
88 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
89 |
|
|
90 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
|
{ |
|
91 |
xvid_free(dec); |
xvid_free(dec); |
92 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
93 |
} |
} |
94 |
|
|
95 |
if (image_create(&dec->refn, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|
{ |
|
96 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
97 |
xvid_free(dec); |
xvid_free(dec); |
98 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
99 |
} |
} |
100 |
|
|
101 |
dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
/* Support B-frame to reference last 2 frame */ |
102 |
if (dec->mbs == NULL) |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
|
{ |
|
103 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
104 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
105 |
|
xvid_free(dec); |
106 |
|
return XVID_ERR_MEMORY; |
107 |
|
} |
108 |
|
if (image_create(&dec->tmp, 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 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
112 |
xvid_free(dec); |
xvid_free(dec); |
113 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
114 |
} |
} |
115 |
|
|
116 |
init_timer(); |
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
117 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
118 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
119 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
120 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
121 |
|
xvid_free(dec); |
122 |
|
return XVID_ERR_MEMORY; |
123 |
|
} |
124 |
|
|
125 |
return XVID_ERR_OK; |
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
126 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
127 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
128 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
129 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
130 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
131 |
|
xvid_free(dec); |
132 |
|
return XVID_ERR_MEMORY; |
133 |
} |
} |
134 |
|
|
135 |
|
dec->mbs = |
136 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
137 |
|
CACHE_LINE); |
138 |
|
if (dec->mbs == NULL) { |
139 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
140 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
141 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
142 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
143 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
144 |
|
xvid_free(dec); |
145 |
|
return XVID_ERR_MEMORY; |
146 |
|
} |
147 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
148 |
|
|
149 |
int decoder_destroy(DECODER * dec) |
/* For skip MB flag */ |
150 |
{ |
dec->last_mbs = |
151 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
152 |
|
CACHE_LINE); |
153 |
|
if (dec->last_mbs == NULL) { |
154 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
|
image_destroy(&dec->refn, dec->edged_width, dec->edged_height); |
|
155 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
156 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
157 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
158 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
159 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
160 |
xvid_free(dec); |
xvid_free(dec); |
161 |
|
return XVID_ERR_MEMORY; |
162 |
|
} |
163 |
|
|
164 |
write_timer(); |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
165 |
return XVID_ERR_OK; |
|
166 |
|
/* nothing happens if that fails */ |
167 |
|
dec->qscale = |
168 |
|
xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE); |
169 |
|
|
170 |
|
if (dec->qscale) |
171 |
|
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
172 |
|
|
173 |
|
return 0; |
174 |
|
} |
175 |
|
|
176 |
|
|
177 |
|
int |
178 |
|
decoder_create(xvid_dec_create_t * create) |
179 |
|
{ |
180 |
|
DECODER *dec; |
181 |
|
|
182 |
|
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
183 |
|
return XVID_ERR_VERSION; |
184 |
|
|
185 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
186 |
|
if (dec == NULL) { |
187 |
|
return XVID_ERR_MEMORY; |
188 |
|
} |
189 |
|
|
190 |
|
memset(dec, 0, sizeof(DECODER)); |
191 |
|
|
192 |
|
dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE); |
193 |
|
if (dec->mpeg_quant_matrices == NULL) { |
194 |
|
xvid_free(dec); |
195 |
|
return XVID_ERR_MEMORY; |
196 |
} |
} |
197 |
|
|
198 |
|
create->handle = dec; |
199 |
|
|
200 |
|
dec->width = create->width; |
201 |
|
dec->height = create->height; |
202 |
|
|
203 |
|
image_null(&dec->cur); |
204 |
|
image_null(&dec->refn[0]); |
205 |
|
image_null(&dec->refn[1]); |
206 |
|
image_null(&dec->tmp); |
207 |
|
image_null(&dec->qtmp); |
208 |
|
|
209 |
|
/* image based GMC */ |
210 |
|
image_null(&dec->gmc); |
211 |
|
|
212 |
|
dec->mbs = NULL; |
213 |
|
dec->last_mbs = NULL; |
214 |
|
dec->qscale = NULL; |
215 |
|
|
216 |
|
init_timer(); |
217 |
|
init_postproc(&dec->postproc); |
218 |
|
init_mpeg_matrix(dec->mpeg_quant_matrices); |
219 |
|
|
220 |
|
/* For B-frame support (used to save reference frame's time */ |
221 |
|
dec->frames = 0; |
222 |
|
dec->time = dec->time_base = dec->last_time_base = 0; |
223 |
|
dec->low_delay = 0; |
224 |
|
dec->packed_mode = 0; |
225 |
|
|
226 |
static const int32_t dquant_table[4] = |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
227 |
|
|
228 |
|
if (dec->fixed_dimensions) |
229 |
|
return decoder_resize(dec); |
230 |
|
else |
231 |
|
return 0; |
232 |
|
} |
233 |
|
|
234 |
|
|
235 |
|
int |
236 |
|
decoder_destroy(DECODER * dec) |
237 |
{ |
{ |
238 |
-1, -2, 1, 2 |
xvid_free(dec->last_mbs); |
239 |
}; |
xvid_free(dec->mbs); |
240 |
|
xvid_free(dec->qscale); |
241 |
|
|
242 |
|
/* image based GMC */ |
243 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
244 |
|
|
245 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
246 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
247 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
248 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
249 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
250 |
|
xvid_free(dec->mpeg_quant_matrices); |
251 |
|
xvid_free(dec); |
252 |
|
|
253 |
|
write_timer(); |
254 |
|
return 0; |
255 |
|
} |
256 |
|
|
257 |
// decode an intra macroblock |
static const int32_t dquant_table[4] = { |
258 |
|
-1, -2, 1, 2 |
259 |
|
}; |
260 |
|
|
261 |
void decoder_mbintra(DECODER * dec, |
/* decode an intra macroblock */ |
262 |
|
static void |
263 |
|
decoder_mbintra(DECODER * dec, |
264 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
265 |
const uint32_t x_pos, |
const uint32_t x_pos, |
266 |
const uint32_t y_pos, |
const uint32_t y_pos, |
268 |
const uint32_t cbp, |
const uint32_t cbp, |
269 |
Bitstream * bs, |
Bitstream * bs, |
270 |
const uint32_t quant, |
const uint32_t quant, |
271 |
const uint32_t intra_dc_threshold) |
const uint32_t intra_dc_threshold, |
272 |
|
const unsigned int bound, |
273 |
|
const int reduced_resolution) |
274 |
{ |
{ |
275 |
|
|
276 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
283 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
284 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
285 |
|
|
286 |
|
if (reduced_resolution) { |
287 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
288 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
289 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
290 |
|
}else{ |
291 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
292 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
293 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
294 |
|
} |
295 |
|
|
296 |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
297 |
|
|
298 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
|
{ |
|
299 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
300 |
int16_t predictors[8]; |
int16_t predictors[8]; |
301 |
int start_coeff; |
int start_coeff; |
302 |
|
|
303 |
start_timer(); |
start_timer(); |
304 |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i*64], iQuant, iDcScaler, predictors); |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
305 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors, bound); |
306 |
{ |
if (!acpred_flag) { |
307 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
308 |
} |
} |
309 |
stop_prediction_timer(); |
stop_prediction_timer(); |
310 |
|
|
311 |
if (quant < intra_dc_threshold) |
if (quant < intra_dc_threshold) { |
|
{ |
|
312 |
int dc_size; |
int dc_size; |
313 |
int dc_dif; |
int dc_dif; |
314 |
|
|
315 |
dc_size = i < 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); |
316 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
317 |
|
|
318 |
if (dc_size > 8) |
if (dc_size > 8) { |
319 |
{ |
BitstreamSkip(bs, 1); /* marker */ |
|
BitstreamSkip(bs, 1); // marker |
|
320 |
} |
} |
321 |
|
|
322 |
block[i*64 + 0] = dc_dif; |
block[i*64 + 0] = dc_dif; |
323 |
start_coeff = 1; |
start_coeff = 1; |
324 |
} |
|
325 |
else |
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif); |
326 |
{ |
} else { |
327 |
start_coeff = 0; |
start_coeff = 0; |
328 |
} |
} |
329 |
|
|
330 |
start_timer(); |
start_timer(); |
331 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
332 |
{ |
{ |
333 |
get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); |
int direction = dec->alternate_vertical_scan ? |
334 |
|
2 : pMB->acpred_directions[i]; |
335 |
|
|
336 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
337 |
} |
} |
338 |
stop_coding_timer(); |
stop_coding_timer(); |
339 |
|
|
342 |
stop_prediction_timer(); |
stop_prediction_timer(); |
343 |
|
|
344 |
start_timer(); |
start_timer(); |
345 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
346 |
{ |
dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
347 |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
} else { |
348 |
} |
dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
|
else |
|
|
{ |
|
|
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
|
349 |
} |
} |
350 |
stop_iquant_timer(); |
stop_iquant_timer(); |
351 |
|
|
352 |
start_timer(); |
start_timer(); |
353 |
idct(&data[i*64]); |
idct(&data[i*64]); |
354 |
stop_idct_timer(); |
stop_idct_timer(); |
355 |
|
|
356 |
} |
} |
357 |
|
|
358 |
if (pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
359 |
next_block = stride; |
next_block = stride; |
360 |
stride *= 2; |
stride *= 2; |
361 |
} |
} |
362 |
|
|
363 |
start_timer(); |
start_timer(); |
364 |
|
|
365 |
|
if (reduced_resolution) |
366 |
|
{ |
367 |
|
next_block*=2; |
368 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
369 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
370 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
371 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
372 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
373 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
374 |
|
}else{ |
375 |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
376 |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
377 |
transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); |
378 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3*64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3*64], stride); |
379 |
transfer_16to8copy(pU_Cur, &data[4*64], stride2); |
transfer_16to8copy(pU_Cur, &data[4*64], stride2); |
380 |
transfer_16to8copy(pV_Cur, &data[5*64], stride2); |
transfer_16to8copy(pV_Cur, &data[5*64], stride2); |
381 |
|
} |
382 |
stop_transfer_timer(); |
stop_transfer_timer(); |
383 |
} |
} |
384 |
|
|
385 |
|
static void |
386 |
|
decoder_mb_decode(DECODER * dec, |
|
|
|
|
|
|
|
#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 |
|
|
|
|
|
void decoder_mbinter(DECODER * dec, |
|
|
const MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
const uint32_t acpred_flag, |
|
387 |
const uint32_t cbp, |
const uint32_t cbp, |
388 |
Bitstream * bs, |
Bitstream * bs, |
389 |
const uint32_t quant, |
uint8_t * pY_Cur, |
390 |
const uint32_t rounding) |
uint8_t * pU_Cur, |
391 |
|
uint8_t * pV_Cur, |
392 |
|
const int reduced_resolution, |
393 |
|
const MACROBLOCK * pMB) |
394 |
{ |
{ |
395 |
|
DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); |
|
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
|
396 |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
397 |
|
|
398 |
uint32_t stride = dec->edged_width; |
int stride = dec->edged_width; |
399 |
uint32_t stride2 = stride / 2; |
int next_block = stride * (reduced_resolution ? 16 : 8); |
400 |
uint32_t next_block = stride * 8; |
const int stride2 = stride/2; |
401 |
uint32_t i; |
int i; |
402 |
uint32_t iQuant = pMB->quant; |
const uint32_t iQuant = pMB->quant; |
403 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
404 |
int uv_dx, uv_dy; |
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
|
|
|
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
|
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
405 |
|
|
406 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
for (i = 0; i < 6; i++) { |
|
{ |
|
|
uv_dx = pMB->mvs[0].x; |
|
|
uv_dy = pMB->mvs[0].y; |
|
407 |
|
|
408 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
if (cbp & (1 << (5 - i))) { /* coded */ |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
|
} |
|
|
else |
|
|
{ |
|
|
int sum; |
|
|
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) ); |
|
409 |
|
|
410 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
|
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
|
|
} |
|
411 |
|
|
412 |
start_timer(); |
start_timer(); |
413 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
get_inter_block(bs, block, direction); |
|
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.u, dec->refn.u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate8x8_switch(dec->cur.v, dec->refn.v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
|
|
stop_comp_timer(); |
|
|
|
|
|
for (i = 0; i < 6; i++) |
|
|
{ |
|
|
if (cbp & (1 << (5-i))) // coded |
|
|
{ |
|
|
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
|
|
|
|
|
start_timer(); |
|
|
get_inter_block(bs, &block[i*64]); |
|
414 |
stop_coding_timer(); |
stop_coding_timer(); |
415 |
|
|
416 |
start_timer(); |
start_timer(); |
417 |
if (dec->quant_type == 0) |
dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); |
|
{ |
|
|
dequant_inter(&data[i*64], &block[i*64], iQuant); |
|
|
} |
|
|
else |
|
|
{ |
|
|
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
|
|
} |
|
418 |
stop_iquant_timer(); |
stop_iquant_timer(); |
419 |
|
|
420 |
start_timer(); |
start_timer(); |
423 |
} |
} |
424 |
} |
} |
425 |
|
|
426 |
if (pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
427 |
next_block = stride; |
next_block = stride; |
428 |
stride *= 2; |
stride *= 2; |
429 |
} |
} |
430 |
|
|
431 |
start_timer(); |
start_timer(); |
432 |
|
if (reduced_resolution) { |
433 |
|
if (cbp & 32) |
434 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
435 |
|
if (cbp & 16) |
436 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
437 |
|
if (cbp & 8) |
438 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
439 |
|
if (cbp & 4) |
440 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
441 |
|
if (cbp & 2) |
442 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
443 |
|
if (cbp & 1) |
444 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
445 |
|
} else { |
446 |
if (cbp & 32) |
if (cbp & 32) |
447 |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
448 |
if (cbp & 16) |
if (cbp & 16) |
455 |
transfer_16to8add(pU_Cur, &data[4*64], stride2); |
transfer_16to8add(pU_Cur, &data[4*64], stride2); |
456 |
if (cbp & 1) |
if (cbp & 1) |
457 |
transfer_16to8add(pV_Cur, &data[5*64], stride2); |
transfer_16to8add(pV_Cur, &data[5*64], stride2); |
458 |
|
} |
459 |
stop_transfer_timer(); |
stop_transfer_timer(); |
460 |
} |
} |
461 |
|
|
462 |
|
/* decode an inter macroblock */ |
463 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
static void |
464 |
|
decoder_mbinter(DECODER * dec, |
465 |
|
const MACROBLOCK * pMB, |
466 |
|
const uint32_t x_pos, |
467 |
|
const uint32_t y_pos, |
468 |
|
const uint32_t cbp, |
469 |
|
Bitstream * bs, |
470 |
|
const uint32_t rounding, |
471 |
|
const int reduced_resolution, |
472 |
|
const int ref) |
473 |
{ |
{ |
474 |
|
uint32_t stride = dec->edged_width; |
475 |
|
uint32_t stride2 = stride / 2; |
476 |
|
uint32_t i; |
477 |
|
|
478 |
uint32_t x, y; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
479 |
|
|
480 |
|
int uv_dx, uv_dy; |
481 |
|
VECTOR mv[4]; /* local copy of mvs */ |
482 |
|
|
483 |
|
if (reduced_resolution) { |
484 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
485 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
486 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
487 |
|
for (i = 0; i < 4; i++) { |
488 |
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
489 |
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
490 |
|
} |
491 |
|
} else { |
492 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
493 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
494 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
495 |
|
for (i = 0; i < 4; i++) |
496 |
|
mv[i] = pMB->mvs[i]; |
497 |
|
} |
498 |
|
|
499 |
|
for (i = 0; i < 4; i++) { |
500 |
|
/* clip to valid range */ |
501 |
|
int border = (int)(dec->mb_width - x_pos) << (5 + dec->quarterpel); |
502 |
|
if (mv[i].x > border) { |
503 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); |
504 |
|
mv[i].x = border; |
505 |
|
} else { |
506 |
|
border = (-(int)x_pos-1) << (5 + dec->quarterpel); |
507 |
|
if (mv[i].x < border) { |
508 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); |
509 |
|
mv[i].x = border; |
510 |
|
} |
511 |
|
} |
512 |
|
|
513 |
|
border = (int)(dec->mb_height - y_pos) << (5 + dec->quarterpel); |
514 |
|
if (mv[i].y > border) { |
515 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); |
516 |
|
mv[i].y = border; |
517 |
|
} else { |
518 |
|
border = (-(int)y_pos-1) << (5 + dec->quarterpel); |
519 |
|
if (mv[i].y < border) { |
520 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); |
521 |
|
mv[i].y = border; |
522 |
|
} |
523 |
|
} |
524 |
|
} |
525 |
|
|
526 |
for (y = 0; y < dec->mb_height; y++) |
start_timer(); |
527 |
|
|
528 |
|
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
529 |
|
|
530 |
|
uv_dx = mv[0].x; |
531 |
|
uv_dy = mv[0].y; |
532 |
|
if (dec->quarterpel) { |
533 |
|
uv_dx /= 2; |
534 |
|
uv_dy /= 2; |
535 |
|
} |
536 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
537 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
538 |
|
|
539 |
|
if (reduced_resolution) |
540 |
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
541 |
|
mv[0].x, mv[0].y, stride, rounding); |
542 |
|
else if (dec->quarterpel) |
543 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
544 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
545 |
|
mv[0].x, mv[0].y, stride, rounding); |
546 |
|
else |
547 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
548 |
|
mv[0].x, mv[0].y, stride, rounding); |
549 |
|
|
550 |
|
} else { /* MODE_INTER4V */ |
551 |
|
|
552 |
|
if(dec->quarterpel) { |
553 |
|
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
554 |
|
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
555 |
|
} else { |
556 |
|
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
557 |
|
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
558 |
|
} |
559 |
|
|
560 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
561 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
562 |
|
|
563 |
|
if (reduced_resolution) { |
564 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
565 |
|
mv[0].x, mv[0].y, stride, rounding); |
566 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
567 |
|
mv[1].x, mv[1].y, stride, rounding); |
568 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
569 |
|
mv[2].x, mv[2].y, stride, rounding); |
570 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
571 |
|
mv[3].x, mv[3].y, stride, rounding); |
572 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
573 |
|
uv_dx, uv_dy, stride2, rounding); |
574 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
575 |
|
uv_dx, uv_dy, stride2, rounding); |
576 |
|
|
577 |
|
} else if (dec->quarterpel) { |
578 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
579 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
580 |
|
mv[0].x, mv[0].y, stride, rounding); |
581 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
582 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
583 |
|
mv[1].x, mv[1].y, stride, rounding); |
584 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
585 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
586 |
|
mv[2].x, mv[2].y, stride, rounding); |
587 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
588 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
589 |
|
mv[3].x, mv[3].y, stride, rounding); |
590 |
|
} else { |
591 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos, |
592 |
|
mv[0].x, mv[0].y, stride, rounding); |
593 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, |
594 |
|
mv[1].x, mv[1].y, stride, rounding); |
595 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8, |
596 |
|
mv[2].x, mv[2].y, stride, rounding); |
597 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8, |
598 |
|
mv[3].x, mv[3].y, stride, rounding); |
599 |
|
} |
600 |
|
} |
601 |
|
|
602 |
|
/* chroma */ |
603 |
|
if (reduced_resolution) { |
604 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
605 |
|
uv_dx, uv_dy, stride2, rounding); |
606 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
607 |
|
uv_dx, uv_dy, stride2, rounding); |
608 |
|
} else { |
609 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
610 |
|
uv_dx, uv_dy, stride2, rounding); |
611 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
612 |
|
uv_dx, uv_dy, stride2, rounding); |
613 |
|
} |
614 |
|
|
615 |
|
stop_comp_timer(); |
616 |
|
|
617 |
|
if (cbp) |
618 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
619 |
|
reduced_resolution, pMB); |
620 |
|
} |
621 |
|
|
622 |
|
static void |
623 |
|
decoder_mbgmc(DECODER * dec, |
624 |
|
MACROBLOCK * const pMB, |
625 |
|
const uint32_t x_pos, |
626 |
|
const uint32_t y_pos, |
627 |
|
const uint32_t fcode, |
628 |
|
const uint32_t cbp, |
629 |
|
Bitstream * bs, |
630 |
|
const uint32_t rounding) |
631 |
{ |
{ |
632 |
for (x = 0; x < dec->mb_width; x++) |
const uint32_t stride = dec->edged_width; |
633 |
|
const uint32_t stride2 = stride / 2; |
634 |
|
|
635 |
|
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
636 |
|
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
637 |
|
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
638 |
|
|
639 |
|
NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
640 |
|
|
641 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
642 |
|
|
643 |
|
start_timer(); |
644 |
|
|
645 |
|
/* this is where the calculations are done */ |
646 |
|
|
647 |
|
gmc_data->predict_16x16(gmc_data, |
648 |
|
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
649 |
|
stride, stride, x_pos, y_pos, rounding); |
650 |
|
|
651 |
|
gmc_data->predict_8x8(gmc_data, |
652 |
|
dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u, |
653 |
|
dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v, |
654 |
|
stride2, stride2, x_pos, y_pos, rounding); |
655 |
|
|
656 |
|
gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel); |
657 |
|
|
658 |
|
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
659 |
|
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
660 |
|
|
661 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
662 |
|
|
663 |
|
stop_transfer_timer(); |
664 |
|
|
665 |
|
if (cbp) |
666 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
667 |
|
|
668 |
|
} |
669 |
|
|
670 |
|
|
671 |
|
static void |
672 |
|
decoder_iframe(DECODER * dec, |
673 |
|
Bitstream * bs, |
674 |
|
int reduced_resolution, |
675 |
|
int quant, |
676 |
|
int intra_dc_threshold) |
677 |
{ |
{ |
678 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
uint32_t bound; |
679 |
|
uint32_t x, y; |
680 |
|
uint32_t mb_width = dec->mb_width; |
681 |
|
uint32_t mb_height = dec->mb_height; |
682 |
|
|
683 |
|
if (reduced_resolution) { |
684 |
|
mb_width = (dec->width + 31) / 32; |
685 |
|
mb_height = (dec->height + 31) / 32; |
686 |
|
} |
687 |
|
|
688 |
|
bound = 0; |
689 |
|
|
690 |
|
for (y = 0; y < mb_height; y++) { |
691 |
|
for (x = 0; x < mb_width; x++) { |
692 |
|
MACROBLOCK *mb; |
693 |
uint32_t mcbpc; |
uint32_t mcbpc; |
694 |
uint32_t cbpc; |
uint32_t cbpc; |
695 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
696 |
uint32_t cbpy; |
uint32_t cbpy; |
697 |
uint32_t cbp; |
uint32_t cbp; |
698 |
|
|
699 |
|
while (BitstreamShowBits(bs, 9) == 1) |
700 |
|
BitstreamSkip(bs, 9); |
701 |
|
|
702 |
|
if (check_resync_marker(bs, 0)) |
703 |
|
{ |
704 |
|
bound = read_video_packet_header(bs, dec, 0, |
705 |
|
&quant, NULL, NULL, &intra_dc_threshold); |
706 |
|
x = bound % mb_width; |
707 |
|
y = bound / mb_width; |
708 |
|
} |
709 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
710 |
|
|
711 |
|
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
712 |
|
|
713 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
714 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
715 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
716 |
|
|
717 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
718 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
719 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
720 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
721 |
|
|
722 |
if (mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
723 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
724 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
725 |
quant = 31; |
quant = 31; |
726 |
} |
} else if (quant < 1) { |
|
else if (quant < 1) |
|
|
{ |
|
727 |
quant = 1; |
quant = 1; |
728 |
} |
} |
729 |
} |
} |
730 |
mb->quant = quant; |
mb->quant = quant; |
731 |
|
mb->mvs[0].x = mb->mvs[0].y = |
732 |
|
mb->mvs[1].x = mb->mvs[1].y = |
733 |
|
mb->mvs[2].x = mb->mvs[2].y = |
734 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
735 |
|
|
736 |
if (dec->interlacing) |
if (dec->interlacing) { |
|
{ |
|
737 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
738 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct); |
739 |
} |
} |
740 |
|
|
741 |
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, |
742 |
|
intra_dc_threshold, bound, reduced_resolution); |
743 |
|
|
744 |
} |
} |
745 |
|
if(dec->out_frm) |
746 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
747 |
} |
} |
748 |
|
|
749 |
} |
} |
750 |
|
|
751 |
|
|
752 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
static void |
753 |
|
get_motion_vector(DECODER * dec, |
754 |
|
Bitstream * bs, |
755 |
|
int x, |
756 |
|
int y, |
757 |
|
int k, |
758 |
|
VECTOR * ret_mv, |
759 |
|
int fcode, |
760 |
|
const int bound) |
761 |
{ |
{ |
762 |
|
|
763 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
764 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
765 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
766 |
int range = (64 * scale_fac); |
const int range = (64 * scale_fac); |
|
|
|
|
VECTOR pmv[4]; |
|
|
uint32_t psad[4]; |
|
767 |
|
|
768 |
int mv_x, mv_y; |
const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
769 |
int pmv_x, pmv_y; |
VECTOR mv; |
770 |
|
|
771 |
|
mv.x = get_mv(bs, fcode); |
772 |
|
mv.y = get_mv(bs, fcode); |
773 |
|
|
774 |
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); |
775 |
|
|
776 |
pmv_x = pmv[0].x; |
mv.x += pmv.x; |
777 |
pmv_y = pmv[0].y; |
mv.y += pmv.y; |
778 |
|
|
779 |
mv_x = get_mv(bs, fcode); |
if (mv.x < low) { |
780 |
mv_y = get_mv(bs, fcode); |
mv.x += range; |
781 |
|
} else if (mv.x > high) { |
782 |
mv_x += pmv_x; |
mv.x -= range; |
|
mv_y += pmv_y; |
|
|
|
|
|
if (mv_x < low) |
|
|
{ |
|
|
mv_x += range; |
|
783 |
} |
} |
784 |
else if (mv_x > high) |
|
785 |
{ |
if (mv.y < low) { |
786 |
mv_x -= range; |
mv.y += range; |
787 |
|
} else if (mv.y > high) { |
788 |
|
mv.y -= range; |
789 |
} |
} |
790 |
|
|
791 |
if (mv_y < low) |
ret_mv->x = mv.x; |
792 |
{ |
ret_mv->y = mv.y; |
|
mv_y += range; |
|
793 |
} |
} |
794 |
else if (mv_y > high) |
|
795 |
|
/* for P_VOP set gmc_warp to NULL */ |
796 |
|
static void |
797 |
|
decoder_pframe(DECODER * dec, |
798 |
|
Bitstream * bs, |
799 |
|
int rounding, |
800 |
|
int reduced_resolution, |
801 |
|
int quant, |
802 |
|
int fcode, |
803 |
|
int intra_dc_threshold, |
804 |
|
const WARPPOINTS *const gmc_warp) |
805 |
{ |
{ |
806 |
mv_y -= range; |
uint32_t x, y; |
807 |
|
uint32_t bound; |
808 |
|
int cp_mb, st_mb; |
809 |
|
uint32_t mb_width = dec->mb_width; |
810 |
|
uint32_t mb_height = dec->mb_height; |
811 |
|
|
812 |
|
if (reduced_resolution) { |
813 |
|
mb_width = (dec->width + 31) / 32; |
814 |
|
mb_height = (dec->height + 31) / 32; |
815 |
} |
} |
816 |
|
|
817 |
mv->x = mv_x; |
start_timer(); |
818 |
mv->y = mv_y; |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
819 |
|
dec->width, dec->height, dec->bs_version); |
820 |
|
stop_edges_timer(); |
821 |
|
|
822 |
} |
if (gmc_warp) { |
823 |
|
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
824 |
|
generate_GMCparameters( dec->sprite_warping_points, |
825 |
|
dec->sprite_warping_accuracy, gmc_warp, |
826 |
|
dec->width, dec->height, &dec->new_gmc_data); |
827 |
|
|
828 |
|
/* image warping is done block-based in decoder_mbgmc(), now */ |
829 |
|
} |
830 |
|
|
831 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
bound = 0; |
|
{ |
|
832 |
|
|
833 |
uint32_t x, y; |
for (y = 0; y < mb_height; y++) { |
834 |
|
cp_mb = st_mb = 0; |
835 |
|
for (x = 0; x < mb_width; x++) { |
836 |
|
MACROBLOCK *mb; |
837 |
|
|
838 |
image_swap(&dec->cur, &dec->refn); |
/* skip stuffing */ |
839 |
|
while (BitstreamShowBits(bs, 10) == 1) |
840 |
|
BitstreamSkip(bs, 10); |
841 |
|
|
842 |
start_timer(); |
if (check_resync_marker(bs, fcode - 1)) { |
843 |
image_setedges(&dec->refn, dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
bound = read_video_packet_header(bs, dec, fcode - 1, |
844 |
stop_edges_timer(); |
&quant, &fcode, NULL, &intra_dc_threshold); |
845 |
|
x = bound % mb_width; |
846 |
|
y = bound / mb_width; |
847 |
|
} |
848 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
849 |
|
|
850 |
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]; |
|
851 |
|
|
852 |
if (!BitstreamGetBit(bs)) // not_coded |
if (!(BitstreamGetBit(bs))) { /* block _is_ coded */ |
853 |
{ |
uint32_t mcbpc, cbpc, cbpy, cbp; |
854 |
uint32_t mcbpc; |
uint32_t intra, acpred_flag = 0; |
855 |
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; |
|
856 |
|
|
857 |
|
cp_mb++; |
858 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
859 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
860 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
861 |
acpred_flag = 0; |
|
862 |
|
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
863 |
|
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
864 |
|
|
865 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
866 |
|
|
867 |
if (intra) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
868 |
{ |
mcsel = BitstreamGetBit(bs); |
869 |
|
else if (intra) |
870 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
|
} |
|
|
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
871 |
|
|
872 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
873 |
|
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
874 |
|
|
875 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
876 |
|
|
877 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
878 |
{ |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
879 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant); |
880 |
if (quant > 31) |
quant += dquant; |
881 |
{ |
if (quant > 31) { |
882 |
quant = 31; |
quant = 31; |
883 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
884 |
quant = 1; |
quant = 1; |
885 |
} |
} |
886 |
|
DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant); |
887 |
} |
} |
888 |
mb->quant = quant; |
mb->quant = quant; |
889 |
|
|
890 |
if (dec->interlacing) |
if (dec->interlacing) { |
891 |
{ |
if (cbp || intra) { |
892 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
893 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
894 |
|
} |
895 |
|
|
896 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) { |
|
{ |
|
897 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
898 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
899 |
|
|
900 |
if (mb->field_pred) |
if (mb->field_pred) { |
|
{ |
|
901 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
902 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
903 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
904 |
DEBUG1("decp: field_for_bot: ", mb->field_for_bot); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
905 |
} |
} |
906 |
} |
} |
907 |
} |
} |
908 |
|
|
909 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mcsel) { |
910 |
{ |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding); |
911 |
if (dec->interlacing && mb->field_pred) |
continue; |
912 |
|
|
913 |
|
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
914 |
|
|
915 |
|
if (dec->interlacing && mb->field_pred) { |
916 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
917 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound); |
918 |
|
} else { |
919 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
920 |
|
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
921 |
|
} |
922 |
|
} else if (mb->mode == MODE_INTER4V ) { |
923 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
924 |
|
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
925 |
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
926 |
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
927 |
|
} else { /* MODE_INTRA, MODE_INTRA_Q */ |
928 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
929 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
930 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
931 |
|
intra_dc_threshold, bound, reduced_resolution); |
932 |
|
continue; |
933 |
|
} |
934 |
|
|
935 |
|
decoder_mbinter(dec, mb, x, y, cbp, bs, |
936 |
|
rounding, reduced_resolution, 0); |
937 |
|
|
938 |
|
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
939 |
|
mb->mode = MODE_NOT_CODED_GMC; |
940 |
|
mb->quant = quant; |
941 |
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding); |
942 |
|
|
943 |
|
if(dec->out_frm && cp_mb > 0) { |
944 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
945 |
|
cp_mb = 0; |
946 |
|
} |
947 |
|
st_mb = x+1; |
948 |
|
} else { /* not coded P_VOP macroblock */ |
949 |
|
mb->mode = MODE_NOT_CODED; |
950 |
|
mb->quant = quant; |
951 |
|
|
952 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
953 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
954 |
|
|
955 |
|
decoder_mbinter(dec, mb, x, y, 0, bs, |
956 |
|
rounding, reduced_resolution, 0); |
957 |
|
|
958 |
|
if(dec->out_frm && cp_mb > 0) { |
959 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
960 |
|
cp_mb = 0; |
961 |
|
} |
962 |
|
st_mb = x+1; |
963 |
|
} |
964 |
|
} |
965 |
|
|
966 |
|
if(dec->out_frm && cp_mb > 0) |
967 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
968 |
|
} |
969 |
|
} |
970 |
|
|
971 |
|
|
972 |
|
/* decode B-frame motion vector */ |
973 |
|
static void |
974 |
|
get_b_motion_vector(Bitstream * bs, |
975 |
|
VECTOR * mv, |
976 |
|
int fcode, |
977 |
|
const VECTOR pmv, |
978 |
|
const DECODER * const dec, |
979 |
|
const int x, const int y) |
980 |
{ |
{ |
981 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
const int scale_fac = 1 << (fcode - 1); |
982 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode); |
const int high = (32 * scale_fac) - 1; |
983 |
|
const int low = ((-32) * scale_fac); |
984 |
|
const int range = (64 * scale_fac); |
985 |
|
|
986 |
|
int mv_x = get_mv(bs, fcode); |
987 |
|
int mv_y = get_mv(bs, fcode); |
988 |
|
|
989 |
|
mv_x += pmv.x; |
990 |
|
mv_y += pmv.y; |
991 |
|
|
992 |
|
if (mv_x < low) |
993 |
|
mv_x += range; |
994 |
|
else if (mv_x > high) |
995 |
|
mv_x -= range; |
996 |
|
|
997 |
|
if (mv_y < low) |
998 |
|
mv_y += range; |
999 |
|
else if (mv_y > high) |
1000 |
|
mv_y -= range; |
1001 |
|
|
1002 |
|
mv->x = mv_x; |
1003 |
|
mv->y = mv_y; |
1004 |
} |
} |
1005 |
else |
|
1006 |
|
/* decode an B-frame direct & interpolate macroblock */ |
1007 |
|
static void |
1008 |
|
decoder_bf_interpolate_mbinter(DECODER * dec, |
1009 |
|
IMAGE forward, |
1010 |
|
IMAGE backward, |
1011 |
|
const MACROBLOCK * pMB, |
1012 |
|
const uint32_t x_pos, |
1013 |
|
const uint32_t y_pos, |
1014 |
|
Bitstream * bs, |
1015 |
|
const int direct) |
1016 |
{ |
{ |
1017 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
uint32_t stride = dec->edged_width; |
1018 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
uint32_t stride2 = stride / 2; |
1019 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
int uv_dx, uv_dy; |
1020 |
|
int b_uv_dx, b_uv_dy; |
1021 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1022 |
|
const uint32_t cbp = pMB->cbp; |
1023 |
|
|
1024 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1025 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1026 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1027 |
|
|
1028 |
|
if (!direct) { |
1029 |
|
uv_dx = pMB->mvs[0].x; |
1030 |
|
uv_dy = pMB->mvs[0].y; |
1031 |
|
|
1032 |
|
b_uv_dx = pMB->b_mvs[0].x; |
1033 |
|
b_uv_dy = pMB->b_mvs[0].y; |
1034 |
|
|
1035 |
|
if (dec->quarterpel) { |
1036 |
|
uv_dx /= 2; |
1037 |
|
uv_dy /= 2; |
1038 |
|
b_uv_dx /= 2; |
1039 |
|
b_uv_dy /= 2; |
1040 |
|
} |
1041 |
|
|
1042 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1043 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1044 |
|
|
1045 |
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1046 |
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1047 |
|
|
1048 |
|
} else { |
1049 |
|
if(dec->quarterpel) { |
1050 |
|
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1051 |
|
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1052 |
|
b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
1053 |
|
b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
1054 |
|
} else { |
1055 |
|
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1056 |
|
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1057 |
|
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1058 |
|
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1059 |
|
} |
1060 |
|
|
1061 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1062 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
1063 |
|
b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf]; |
1064 |
|
b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf]; |
1065 |
|
} |
1066 |
|
|
1067 |
|
start_timer(); |
1068 |
|
if(dec->quarterpel) { |
1069 |
|
if(!direct) { |
1070 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1071 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1072 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1073 |
|
} else { |
1074 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1075 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1076 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1077 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1078 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1079 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1080 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1081 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1082 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1083 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1084 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1085 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1086 |
|
} |
1087 |
|
} else { |
1088 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1089 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1090 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1091 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1092 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1093 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1094 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8, |
1095 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1096 |
|
} |
1097 |
|
|
1098 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1099 |
|
uv_dy, stride2, 0); |
1100 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1101 |
|
uv_dy, stride2, 0); |
1102 |
|
|
1103 |
|
|
1104 |
|
if(dec->quarterpel) { |
1105 |
|
if(!direct) { |
1106 |
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1107 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1108 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1109 |
|
} else { |
1110 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1111 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1112 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1113 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1114 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1115 |
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1116 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1117 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1118 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1119 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1120 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1121 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1122 |
|
} |
1123 |
|
} else { |
1124 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1125 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1126 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1127 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1128 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1129 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1130 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1131 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1132 |
|
} |
1133 |
|
|
1134 |
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1135 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1136 |
|
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
1137 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1138 |
|
|
1139 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1140 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1141 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1142 |
|
stride, 1, 8); |
1143 |
|
|
1144 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1145 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1146 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1147 |
|
stride, 1, 8); |
1148 |
|
|
1149 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1150 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1151 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1152 |
|
stride, 1, 8); |
1153 |
|
|
1154 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1155 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1156 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1157 |
|
stride, 1, 8); |
1158 |
|
|
1159 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1160 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1161 |
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1162 |
|
stride2, 1, 8); |
1163 |
|
|
1164 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1165 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1166 |
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1167 |
|
stride2, 1, 8); |
1168 |
|
|
1169 |
|
stop_comp_timer(); |
1170 |
|
|
1171 |
|
if (cbp) |
1172 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
1173 |
} |
} |
1174 |
|
|
1175 |
|
/* for decode B-frame dbquant */ |
1176 |
|
static __inline int32_t |
1177 |
|
get_dbquant(Bitstream * bs) |
1178 |
|
{ |
1179 |
|
if (!BitstreamGetBit(bs)) /* '0' */ |
1180 |
|
return (0); |
1181 |
|
else if (!BitstreamGetBit(bs)) /* '10' */ |
1182 |
|
return (-2); |
1183 |
|
else /* '11' */ |
1184 |
|
return (2); |
1185 |
} |
} |
1186 |
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
|
1187 |
|
/* |
1188 |
|
* decode B-frame mb_type |
1189 |
|
* bit ret_value |
1190 |
|
* 1 0 |
1191 |
|
* 01 1 |
1192 |
|
* 001 2 |
1193 |
|
* 0001 3 |
1194 |
|
*/ |
1195 |
|
static int32_t __inline |
1196 |
|
get_mbtype(Bitstream * bs) |
1197 |
{ |
{ |
1198 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
int32_t mb_type; |
1199 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
|
1200 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
for (mb_type = 0; mb_type <= 3; mb_type++) |
1201 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
if (BitstreamGetBit(bs)) |
1202 |
|
return (mb_type); |
1203 |
|
|
1204 |
|
return -1; |
1205 |
} |
} |
1206 |
else // MODE_INTRA, MODE_INTRA_Q |
|
1207 |
|
static void |
1208 |
|
decoder_bframe(DECODER * dec, |
1209 |
|
Bitstream * bs, |
1210 |
|
int quant, |
1211 |
|
int fcode_forward, |
1212 |
|
int fcode_backward) |
1213 |
{ |
{ |
1214 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
uint32_t x, y; |
1215 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
VECTOR mv; |
1216 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
const VECTOR zeromv = {0,0}; |
1217 |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1218 |
|
int i; |
1219 |
|
|
1220 |
|
start_timer(); |
1221 |
|
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1222 |
|
dec->width, dec->height, dec->bs_version); |
1223 |
|
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1224 |
|
dec->width, dec->height, dec->bs_version); |
1225 |
|
stop_edges_timer(); |
1226 |
|
|
1227 |
|
for (y = 0; y < dec->mb_height; y++) { |
1228 |
|
/* Initialize Pred Motion Vector */ |
1229 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1230 |
|
for (x = 0; x < dec->mb_width; x++) { |
1231 |
|
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1232 |
|
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1233 |
|
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1234 |
|
uint32_t intra_dc_threshold; /* fake variable */ |
1235 |
|
|
1236 |
|
if (check_resync_marker(bs, fcode_max - 1)) { |
1237 |
|
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1238 |
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1239 |
|
x = bound % dec->mb_width; |
1240 |
|
y = bound / dec->mb_width; |
1241 |
|
/* reset predicted macroblocks */ |
1242 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1243 |
|
} |
1244 |
|
|
1245 |
|
mv = |
1246 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1247 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1248 |
|
mb->quant = quant; |
1249 |
|
|
1250 |
|
/* |
1251 |
|
* skip if the co-located P_VOP macroblock is not coded |
1252 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1253 |
|
* automatically skipped |
1254 |
|
*/ |
1255 |
|
|
1256 |
|
if (last_mb->mode == MODE_NOT_CODED) { |
1257 |
|
mb->cbp = 0; |
1258 |
|
mb->mode = MODE_FORWARD; |
1259 |
|
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
1260 |
continue; |
continue; |
1261 |
} |
} |
1262 |
|
|
1263 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1264 |
|
const uint8_t modb2 = BitstreamGetBit(bs); |
1265 |
|
|
1266 |
|
mb->mode = get_mbtype(bs); |
1267 |
|
|
1268 |
|
if (!modb2) /* modb=='00' */ |
1269 |
|
mb->cbp = BitstreamGetBits(bs, 6); |
1270 |
|
else |
1271 |
|
mb->cbp = 0; |
1272 |
|
|
1273 |
|
if (mb->mode && mb->cbp) { |
1274 |
|
quant += get_dbquant(bs); |
1275 |
|
if (quant > 31) |
1276 |
|
quant = 31; |
1277 |
|
else if (quant < 1) |
1278 |
|
quant = 1; |
1279 |
} |
} |
1280 |
else // not coded |
mb->quant = quant; |
|
{ |
|
1281 |
|
|
1282 |
mb->mode = MODE_NOT_CODED; |
if (dec->interlacing) { |
1283 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
if (mb->cbp) { |
1284 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
mb->field_dct = BitstreamGetBit(bs); |
1285 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1286 |
|
} |
1287 |
|
|
1288 |
// copy macroblock directly from ref to cur |
if (mb->mode) { |
1289 |
|
mb->field_pred = BitstreamGetBit(bs); |
1290 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1291 |
|
|
1292 |
start_timer(); |
if (mb->field_pred) { |
1293 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1294 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1295 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1296 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1297 |
|
} |
1298 |
|
} |
1299 |
|
} |
1300 |
|
|
1301 |
|
} else { |
1302 |
|
mb->mode = MODE_DIRECT_NONE_MV; |
1303 |
|
mb->cbp = 0; |
1304 |
|
} |
1305 |
|
|
1306 |
|
switch (mb->mode) { |
1307 |
|
case MODE_DIRECT: |
1308 |
|
get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y); |
1309 |
|
|
1310 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
case MODE_DIRECT_NONE_MV: |
1311 |
dec->refn.y + (16*y)*dec->edged_width + (16*x), |
for (i = 0; i < 4; i++) { |
1312 |
dec->edged_width); |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x); |
1313 |
|
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1314 |
|
? ((TRB - TRD) * last_mb->mvs[i].x) / TRD |
1315 |
|
: mb->mvs[i].x - last_mb->mvs[i].x); |
1316 |
|
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y); |
1317 |
|
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1318 |
|
? ((TRB - TRD) * last_mb->mvs[i].y) / TRD |
1319 |
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
1320 |
|
} |
1321 |
|
|
1322 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1323 |
dec->refn.y + (16*y)*dec->edged_width + (16*x+8), |
mb, x, y, bs, 1); |
1324 |
dec->edged_width); |
break; |
1325 |
|
|
1326 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
case MODE_INTERPOLATE: |
1327 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x), |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1328 |
dec->edged_width); |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1329 |
|
|
1330 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1331 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x+8), |
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0]; |
|
dec->edged_width); |
|
1332 |
|
|
1333 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1334 |
dec->refn.u + (8*y)*dec->edged_width/2 + (8*x), |
mb, x, y, bs, 0); |
1335 |
dec->edged_width/2); |
break; |
1336 |
|
|
1337 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
case MODE_BACKWARD: |
1338 |
dec->refn.v + (8*y)*dec->edged_width/2 + (8*x), |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1339 |
dec->edged_width/2); |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1340 |
|
|
1341 |
stop_transfer_timer(); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0); |
1342 |
|
break; |
1343 |
|
|
1344 |
|
case MODE_FORWARD: |
1345 |
|
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1346 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1347 |
|
|
1348 |
|
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
1349 |
|
break; |
1350 |
|
|
1351 |
|
default: |
1352 |
|
DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode); |
1353 |
|
} |
1354 |
|
} /* End of for */ |
1355 |
} |
} |
1356 |
} |
} |
1357 |
|
|
1358 |
|
/* perform post processing if necessary, and output the image */ |
1359 |
|
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1360 |
|
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1361 |
|
int coding_type, int quant) |
1362 |
|
{ |
1363 |
|
if (dec->cartoon_mode) |
1364 |
|
frame->general &= ~XVID_FILMEFFECT; |
1365 |
|
|
1366 |
|
if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || frame->brightness!=0) |
1367 |
|
&& mbs != NULL) /* post process */ |
1368 |
|
{ |
1369 |
|
/* note: image is stored to tmp */ |
1370 |
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1371 |
|
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
1372 |
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1373 |
|
frame->general, frame->brightness, dec->frames, (coding_type == B_VOP)); |
1374 |
|
img = &dec->tmp; |
1375 |
|
} |
1376 |
|
|
1377 |
|
image_output(img, dec->width, dec->height, |
1378 |
|
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1379 |
|
frame->output.csp, dec->interlacing); |
1380 |
|
|
1381 |
|
if (stats) { |
1382 |
|
stats->type = coding2type(coding_type); |
1383 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1384 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1385 |
|
stats->data.vop.qscale_stride = dec->mb_width; |
1386 |
|
stats->data.vop.qscale = dec->qscale; |
1387 |
|
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1388 |
|
int i; |
1389 |
|
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1390 |
|
stats->data.vop.qscale[i] = mbs[i].quant; |
1391 |
|
} else |
1392 |
|
stats->data.vop.qscale = NULL; |
1393 |
} |
} |
1394 |
} |
} |
1395 |
|
|
1396 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
|
1397 |
|
int |
1398 |
|
decoder_decode(DECODER * dec, |
1399 |
|
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1400 |
{ |
{ |
1401 |
|
|
1402 |
Bitstream bs; |
Bitstream bs; |
1403 |
uint32_t rounding; |
uint32_t rounding; |
1404 |
uint32_t quant; |
uint32_t reduced_resolution; |
1405 |
uint32_t fcode; |
uint32_t quant = 2; |
1406 |
|
uint32_t fcode_forward; |
1407 |
|
uint32_t fcode_backward; |
1408 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1409 |
|
WARPPOINTS gmc_warp; |
1410 |
|
int coding_type; |
1411 |
|
int success, output, seen_something; |
1412 |
|
|
1413 |
|
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1414 |
|
return XVID_ERR_VERSION; |
1415 |
|
|
1416 |
start_global_timer(); |
start_global_timer(); |
1417 |
|
|
1418 |
|
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1419 |
|
if ((frame->general & XVID_DISCONTINUITY)) |
1420 |
|
dec->frames = 0; |
1421 |
|
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1422 |
|
|
1423 |
|
if (frame->length < 0) { /* decoder flush */ |
1424 |
|
int ret; |
1425 |
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1426 |
|
we have a reference frame, then outout the reference frame */ |
1427 |
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1428 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1429 |
|
dec->frames = 0; |
1430 |
|
ret = 0; |
1431 |
|
} else { |
1432 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1433 |
|
ret = XVID_ERR_END; |
1434 |
|
} |
1435 |
|
|
1436 |
|
emms(); |
1437 |
|
stop_global_timer(); |
1438 |
|
return ret; |
1439 |
|
} |
1440 |
|
|
1441 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1442 |
|
|
1443 |
switch (BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold)) |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1444 |
|
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1445 |
{ |
{ |
1446 |
case P_VOP : |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1447 |
decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold); |
(uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing); |
1448 |
break; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1449 |
|
emms(); |
1450 |
|
return 1; /* one byte consumed */ |
1451 |
|
} |
1452 |
|
|
1453 |
case I_VOP : |
success = 0; |
1454 |
//DEBUG1("",intra_dc_threshold); |
output = 0; |
1455 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
seen_something = 0; |
|
break; |
|
1456 |
|
|
1457 |
case B_VOP : // ignore |
repeat: |
|
break; |
|
1458 |
|
|
1459 |
case N_VOP : // vop not coded |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1460 |
break; |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1461 |
|
|
1462 |
default : |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
1463 |
return XVID_ERR_FAIL; |
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1464 |
|
|
1465 |
|
if (coding_type == -1) { /* nothing */ |
1466 |
|
if (success) goto done; |
1467 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1468 |
|
emms(); |
1469 |
|
return BitstreamPos(&bs)/8; |
1470 |
} |
} |
1471 |
|
|
1472 |
frame->length = BitstreamPos(&bs) / 8; |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1473 |
|
|
1474 |
start_timer(); |
if (coding_type == -3) |
1475 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
decoder_resize(dec); |
|
frame->image, frame->stride, frame->colorspace); |
|
|
stop_conv_timer(); |
|
1476 |
|
|
1477 |
|
if (stats) { |
1478 |
|
stats->type = XVID_TYPE_VOL; |
1479 |
|
stats->data.vol.general = 0; |
1480 |
|
/*XXX: if (dec->interlacing) |
1481 |
|
stats->data.vol.general |= ++INTERLACING; */ |
1482 |
|
stats->data.vol.width = dec->width; |
1483 |
|
stats->data.vol.height = dec->height; |
1484 |
|
stats->data.vol.par = dec->aspect_ratio; |
1485 |
|
stats->data.vol.par_width = dec->par_width; |
1486 |
|
stats->data.vol.par_height = dec->par_height; |
1487 |
emms(); |
emms(); |
1488 |
|
return BitstreamPos(&bs)/8; /* number of bytes consumed */ |
1489 |
|
} |
1490 |
|
goto repeat; |
1491 |
|
} |
1492 |
|
|
1493 |
stop_global_timer(); |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1494 |
|
|
1495 |
|
/* packed_mode: special-N_VOP treament */ |
1496 |
|
if (dec->packed_mode && coding_type == N_VOP) { |
1497 |
|
if (dec->low_delay_default && dec->frames > 0) { |
1498 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1499 |
|
output = 1; |
1500 |
|
} |
1501 |
|
/* ignore otherwise */ |
1502 |
|
} else if (coding_type != B_VOP) { |
1503 |
|
switch(coding_type) { |
1504 |
|
case I_VOP : |
1505 |
|
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1506 |
|
break; |
1507 |
|
case P_VOP : |
1508 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1509 |
|
fcode_forward, intra_dc_threshold, NULL); |
1510 |
|
break; |
1511 |
|
case S_VOP : |
1512 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1513 |
|
fcode_forward, intra_dc_threshold, &gmc_warp); |
1514 |
|
break; |
1515 |
|
case N_VOP : |
1516 |
|
/* XXX: not_coded vops are not used for forward prediction */ |
1517 |
|
/* we should not swap(last_mbs,mbs) */ |
1518 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1519 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */ |
1520 |
|
break; |
1521 |
|
} |
1522 |
|
|
1523 |
|
if (reduced_resolution) { |
1524 |
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1525 |
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1526 |
|
16, 0); |
1527 |
|
} |
1528 |
|
|
1529 |
|
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1530 |
|
if (!(dec->low_delay_default && dec->packed_mode)) { |
1531 |
|
if (dec->low_delay) { |
1532 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
1533 |
|
output = 1; |
1534 |
|
} else if (dec->frames > 0) { /* is the reference frame valid? */ |
1535 |
|
/* output the reference frame */ |
1536 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1537 |
|
output = 1; |
1538 |
|
} |
1539 |
|
} |
1540 |
|
|
1541 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1542 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1543 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
1544 |
|
dec->last_reduced_resolution = reduced_resolution; |
1545 |
|
dec->last_coding_type = coding_type; |
1546 |
|
|
1547 |
|
dec->frames++; |
1548 |
|
seen_something = 1; |
1549 |
|
|
1550 |
|
} else { /* B_VOP */ |
1551 |
|
|
1552 |
|
if (dec->low_delay) { |
1553 |
|
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
1554 |
|
dec->low_delay = 1; |
1555 |
|
} |
1556 |
|
|
1557 |
|
if (dec->frames < 2) { |
1558 |
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
1559 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1560 |
|
"broken b-frame, mising ref frames"); |
1561 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1562 |
|
} else if (dec->time_pp <= dec->time_bp) { |
1563 |
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1564 |
|
decoded in vfw. */ |
1565 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1566 |
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1567 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1568 |
|
} else { |
1569 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1570 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
1571 |
|
} |
1572 |
|
|
1573 |
|
output = 1; |
1574 |
|
dec->frames++; |
1575 |
|
} |
1576 |
|
|
1577 |
|
#if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */ |
1578 |
|
BitstreamByteAlign(&bs); |
1579 |
|
#endif |
1580 |
|
|
1581 |
|
/* low_delay_default mode: repeat in packed_mode */ |
1582 |
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) { |
1583 |
|
success = 1; |
1584 |
|
goto repeat; |
1585 |
|
} |
1586 |
|
|
1587 |
|
done : |
1588 |
|
|
1589 |
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
1590 |
|
then output the recently decoded frame, or print an error message */ |
1591 |
|
if (dec->low_delay_default && output == 0) { |
1592 |
|
if (dec->packed_mode && seen_something) { |
1593 |
|
/* output the recently decoded frame */ |
1594 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1595 |
|
} else { |
1596 |
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1597 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1598 |
|
"warning: nothing to output"); |
1599 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1600 |
|
"bframe decoder lag"); |
1601 |
|
|
1602 |
return XVID_ERR_OK; |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1603 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1604 |
|
} |
1605 |
|
} |
1606 |
|
|
1607 |
|
emms(); |
1608 |
|
stop_global_timer(); |
1609 |
|
|
1610 |
|
return (BitstreamPos(&bs) + 7) / 8; /* number of bytes consumed */ |
1611 |
} |
} |