[cvs] / xvidcore / src / decoder.c Repository:
ViewVC logotype

Annotation of /xvidcore/src/decoder.c

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1.49.2.25 - (view) (download)

1 : edgomez 1.48 /*****************************************************************************
2 : Isibaar 1.1 *
3 : edgomez 1.12 * XVID MPEG-4 VIDEO CODEC
4 : edgomez 1.48 * - Decoder Module -
5 : Isibaar 1.1 *
6 : edgomez 1.49.2.7 * Copyright(C) 2002 MinChen <chenm001@163.com>
7 :     * 2002-2003 Peter Ross <pross@xvid.org>
8 : edgomez 1.42 *
9 : edgomez 1.49.2.7 * This program is free software ; you can redistribute it and/or modify
10 : edgomez 1.47 * it under the terms of the GNU General Public License as published by
11 : edgomez 1.49.2.7 * the Free Software Foundation ; either version 2 of the License, or
12 : edgomez 1.12 * (at your option) any later version.
13 :     *
14 :     * This program is distributed in the hope that it will be useful,
15 : edgomez 1.49.2.7 * but WITHOUT ANY WARRANTY ; without even the implied warranty of
16 : edgomez 1.12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 :     * GNU General Public License for more details.
18 :     *
19 :     * You should have received a copy of the GNU General Public License
20 : edgomez 1.49.2.7 * along with this program ; if not, write to the Free Software
21 : edgomez 1.12 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 : Isibaar 1.1 *
23 : syskin 1.49.2.25 * $Id: decoder.c,v 1.49.2.24 2003/12/10 15:07:42 edgomez Exp $
24 : edgomez 1.43 *
25 : edgomez 1.48 ****************************************************************************/
26 : Isibaar 1.1
27 : edgomez 1.47 #include <stdio.h>
28 : Isibaar 1.1 #include <stdlib.h>
29 : edgomez 1.12 #include <string.h>
30 : Isibaar 1.1
31 : chenm001 1.27 #ifdef BFRAMES_DEC_DEBUG
32 :     #define BFRAMES_DEC
33 :     #endif
34 :    
35 : Isibaar 1.1 #include "xvid.h"
36 :     #include "portab.h"
37 : edgomez 1.47 #include "global.h"
38 : Isibaar 1.1
39 :     #include "decoder.h"
40 :     #include "bitstream/bitstream.h"
41 :     #include "bitstream/mbcoding.h"
42 :    
43 : edgomez 1.49.2.15 #include "quant/quant.h"
44 : edgomez 1.49.2.20 #include "quant/quant_matrix.h"
45 : Isibaar 1.1 #include "dct/idct.h"
46 :     #include "dct/fdct.h"
47 :     #include "utils/mem_transfer.h"
48 :     #include "image/interpolate8x8.h"
49 : edgomez 1.47 #include "image/reduced.h"
50 :     #include "image/font.h"
51 : Isibaar 1.1
52 :     #include "bitstream/mbcoding.h"
53 :     #include "prediction/mbprediction.h"
54 :     #include "utils/timer.h"
55 :     #include "utils/emms.h"
56 : edgomez 1.47 #include "motion/motion.h"
57 : chl 1.49.2.9 #include "motion/gmc.h"
58 : Isibaar 1.1
59 :     #include "image/image.h"
60 :     #include "image/colorspace.h"
61 : edgomez 1.49.2.24 #include "image/postprocessing.h"
62 : Isibaar 1.3 #include "utils/mem_align.h"
63 : Isibaar 1.1
64 : syskin 1.49.2.17 static int
65 : edgomez 1.47 decoder_resize(DECODER * dec)
66 : Isibaar 1.1 {
67 : edgomez 1.47 /* free existing */
68 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
69 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
70 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
71 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
72 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
73 :    
74 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
75 : Isibaar 1.1
76 : edgomez 1.49.2.13 if (dec->last_mbs)
77 : edgomez 1.47 xvid_free(dec->last_mbs);
78 :     if (dec->mbs)
79 :     xvid_free(dec->mbs);
80 : Isibaar 1.1
81 : edgomez 1.47 /* realloc */
82 : Isibaar 1.1 dec->mb_width = (dec->width + 15) / 16;
83 :     dec->mb_height = (dec->height + 15) / 16;
84 :    
85 :     dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE;
86 :     dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE;
87 : edgomez 1.19
88 :     if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) {
89 : Isibaar 1.3 xvid_free(dec);
90 : Isibaar 1.1 return XVID_ERR_MEMORY;
91 :     }
92 :    
93 : edgomez 1.19 if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) {
94 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
95 : Isibaar 1.3 xvid_free(dec);
96 : Isibaar 1.1 return XVID_ERR_MEMORY;
97 :     }
98 : edgomez 1.47
99 : edgomez 1.48 /* Support B-frame to reference last 2 frame */
100 : edgomez 1.19 if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) {
101 : chenm001 1.11 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
102 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
103 :     xvid_free(dec);
104 :     return XVID_ERR_MEMORY;
105 :     }
106 : edgomez 1.47 if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) {
107 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
108 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
109 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
110 :     xvid_free(dec);
111 :     return XVID_ERR_MEMORY;
112 :     }
113 :    
114 :     if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) {
115 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
116 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
117 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
118 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
119 :     xvid_free(dec);
120 :     return XVID_ERR_MEMORY;
121 :     }
122 :    
123 :     if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) {
124 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
125 : chenm001 1.14 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
126 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
127 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
128 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
129 : chenm001 1.14 xvid_free(dec);
130 :     return XVID_ERR_MEMORY;
131 :     }
132 : Isibaar 1.1
133 : edgomez 1.19 dec->mbs =
134 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
135 :     CACHE_LINE);
136 :     if (dec->mbs == NULL) {
137 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
138 : chenm001 1.14 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
139 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
140 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
141 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
142 : chenm001 1.14 xvid_free(dec);
143 :     return XVID_ERR_MEMORY;
144 :     }
145 : Isibaar 1.20 memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
146 :    
147 : edgomez 1.48 /* For skip MB flag */
148 : edgomez 1.19 dec->last_mbs =
149 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
150 :     CACHE_LINE);
151 :     if (dec->last_mbs == NULL) {
152 : chenm001 1.14 xvid_free(dec->mbs);
153 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
154 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
155 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
156 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
157 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
158 : Isibaar 1.3 xvid_free(dec);
159 : Isibaar 1.1 return XVID_ERR_MEMORY;
160 :     }
161 : Isibaar 1.20
162 :     memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
163 : Isibaar 1.1
164 : suxen_drol 1.49.2.1 return 0;
165 : edgomez 1.47 }
166 :    
167 :    
168 :     int
169 : suxen_drol 1.49.2.1 decoder_create(xvid_dec_create_t * create)
170 : edgomez 1.47 {
171 :     DECODER *dec;
172 :    
173 : edgomez 1.49.2.10 if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */
174 : suxen_drol 1.49.2.1 return XVID_ERR_VERSION;
175 :    
176 : edgomez 1.47 dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
177 :     if (dec == NULL) {
178 :     return XVID_ERR_MEMORY;
179 :     }
180 : edgomez 1.49.2.20
181 : edgomez 1.47 memset(dec, 0, sizeof(DECODER));
182 :    
183 : edgomez 1.49.2.20 dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE);
184 :     if (dec->mpeg_quant_matrices == NULL) {
185 :     xvid_free(dec);
186 :     return XVID_ERR_MEMORY;
187 :     }
188 :    
189 : suxen_drol 1.49.2.1 create->handle = dec;
190 : edgomez 1.47
191 : suxen_drol 1.49.2.1 dec->width = create->width;
192 :     dec->height = create->height;
193 : edgomez 1.47
194 :     image_null(&dec->cur);
195 :     image_null(&dec->refn[0]);
196 :     image_null(&dec->refn[1]);
197 :     image_null(&dec->tmp);
198 :     image_null(&dec->qtmp);
199 :    
200 : edgomez 1.48 /* image based GMC */
201 : edgomez 1.47 image_null(&dec->gmc);
202 :    
203 :    
204 :     dec->mbs = NULL;
205 :     dec->last_mbs = NULL;
206 :    
207 : Isibaar 1.1 init_timer();
208 : edgomez 1.49.2.20 init_mpeg_matrix(dec->mpeg_quant_matrices);
209 : Isibaar 1.1
210 : edgomez 1.48 /* For B-frame support (used to save reference frame's time */
211 : edgomez 1.47 dec->frames = 0;
212 : chenm001 1.14 dec->time = dec->time_base = dec->last_time_base = 0;
213 : edgomez 1.47 dec->low_delay = 0;
214 :     dec->packed_mode = 0;
215 :    
216 :     dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
217 : edgomez 1.19
218 : edgomez 1.47 if (dec->fixed_dimensions)
219 :     return decoder_resize(dec);
220 :     else
221 : edgomez 1.49.2.13 return 0;
222 : Isibaar 1.1 }
223 :    
224 :    
225 : edgomez 1.19 int
226 :     decoder_destroy(DECODER * dec)
227 : Isibaar 1.1 {
228 : chenm001 1.14 xvid_free(dec->last_mbs);
229 : Isibaar 1.3 xvid_free(dec->mbs);
230 : edgomez 1.47
231 : edgomez 1.48 /* image based GMC */
232 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
233 : edgomez 1.47
234 : chenm001 1.11 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
235 : chenm001 1.14 image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
236 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
237 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
238 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
239 : edgomez 1.49.2.20 xvid_free(dec->mpeg_quant_matrices);
240 : Isibaar 1.3 xvid_free(dec);
241 : Isibaar 1.1
242 :     write_timer();
243 : suxen_drol 1.49.2.1 return 0;
244 : Isibaar 1.1 }
245 :    
246 : edgomez 1.19 static const int32_t dquant_table[4] = {
247 : Isibaar 1.1 -1, -2, 1, 2
248 :     };
249 :    
250 : edgomez 1.48 /* decode an intra macroblock */
251 : syskin 1.49.2.17 static void
252 : edgomez 1.19 decoder_mbintra(DECODER * dec,
253 :     MACROBLOCK * pMB,
254 :     const uint32_t x_pos,
255 :     const uint32_t y_pos,
256 :     const uint32_t acpred_flag,
257 :     const uint32_t cbp,
258 :     Bitstream * bs,
259 :     const uint32_t quant,
260 : chenm001 1.26 const uint32_t intra_dc_threshold,
261 : edgomez 1.47 const unsigned int bound,
262 :     const int reduced_resolution)
263 : Isibaar 1.1 {
264 : edgomez 1.7
265 :     DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
266 : edgomez 1.19 DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
267 : edgomez 1.7
268 : h 1.8 uint32_t stride = dec->edged_width;
269 :     uint32_t stride2 = stride / 2;
270 :     uint32_t next_block = stride * 8;
271 : h 1.5 uint32_t i;
272 :     uint32_t iQuant = pMB->quant;
273 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
274 :    
275 : edgomez 1.47 if (reduced_resolution) {
276 :     pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5);
277 :     pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4);
278 :     pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4);
279 :     }else{
280 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
281 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
282 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
283 :     }
284 : edgomez 1.7
285 : edgomez 1.48 memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */
286 : h 1.5
287 : edgomez 1.19 for (i = 0; i < 6; i++) {
288 : h 1.5 uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);
289 : Isibaar 1.1 int16_t predictors[8];
290 :     int start_coeff;
291 :    
292 :     start_timer();
293 : edgomez 1.19 predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64],
294 : chenm001 1.26 iQuant, iDcScaler, predictors, bound);
295 : edgomez 1.19 if (!acpred_flag) {
296 : h 1.5 pMB->acpred_directions[i] = 0;
297 : Isibaar 1.1 }
298 :     stop_prediction_timer();
299 :    
300 : edgomez 1.19 if (quant < intra_dc_threshold) {
301 : Isibaar 1.1 int dc_size;
302 :     int dc_dif;
303 :    
304 : edgomez 1.19 dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs);
305 :     dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0;
306 : Isibaar 1.1
307 : edgomez 1.19 if (dc_size > 8) {
308 : edgomez 1.48 BitstreamSkip(bs, 1); /* marker */
309 : Isibaar 1.1 }
310 : edgomez 1.19
311 :     block[i * 64 + 0] = dc_dif;
312 : Isibaar 1.1 start_coeff = 1;
313 : chenm001 1.26
314 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif);
315 : edgomez 1.19 } else {
316 : Isibaar 1.1 start_coeff = 0;
317 :     }
318 :    
319 :     start_timer();
320 : edgomez 1.48 if (cbp & (1 << (5 - i))) /* coded */
321 : Isibaar 1.1 {
322 : edgomez 1.47 int direction = dec->alternate_vertical_scan ?
323 :     2 : pMB->acpred_directions[i];
324 :    
325 :     get_intra_block(bs, &block[i * 64], direction, start_coeff);
326 : Isibaar 1.1 }
327 :     stop_coding_timer();
328 :    
329 :     start_timer();
330 : edgomez 1.19 add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors);
331 : Isibaar 1.1 stop_prediction_timer();
332 :    
333 :     start_timer();
334 : edgomez 1.19 if (dec->quant_type == 0) {
335 : edgomez 1.49.2.20 dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
336 : edgomez 1.19 } else {
337 : edgomez 1.49.2.20 dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
338 : Isibaar 1.1 }
339 :     stop_iquant_timer();
340 :    
341 :     start_timer();
342 : edgomez 1.19 idct(&data[i * 64]);
343 : Isibaar 1.1 stop_idct_timer();
344 : edgomez 1.47
345 : h 1.5 }
346 : Isibaar 1.1
347 : edgomez 1.19 if (dec->interlacing && pMB->field_dct) {
348 : h 1.8 next_block = stride;
349 :     stride *= 2;
350 : Isibaar 1.1 }
351 : h 1.5
352 :     start_timer();
353 : edgomez 1.47
354 :     if (reduced_resolution)
355 :     {
356 :     next_block*=2;
357 :     copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
358 :     copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
359 :     copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
360 :     copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
361 :     copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
362 :     copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
363 :     }else{
364 :     transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
365 :     transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
366 :     transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
367 :     transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
368 :     transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
369 :     transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
370 :     }
371 : h 1.5 stop_transfer_timer();
372 : Isibaar 1.1 }
373 :    
374 : syskin 1.49.2.17 static void
375 :     decoder_mb_decode(DECODER * dec,
376 :     const uint32_t cbp,
377 :     Bitstream * bs,
378 :     uint8_t * pY_Cur,
379 :     uint8_t * pU_Cur,
380 :     uint8_t * pV_Cur,
381 :     const int reduced_resolution,
382 :     const MACROBLOCK * pMB)
383 :     {
384 :     DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE);
385 :     DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
386 : Isibaar 1.1
387 : edgomez 1.49.2.18 int stride = dec->edged_width;
388 :     int next_block = stride * (reduced_resolution ? 16 : 8);
389 : syskin 1.49.2.17 const int stride2 = stride/2;
390 :     int i;
391 :     const uint32_t iQuant = pMB->quant;
392 :     const int direction = dec->alternate_vertical_scan ? 2 : 0;
393 :     const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter;
394 : Isibaar 1.1
395 : syskin 1.49.2.17 for (i = 0; i < 6; i++) {
396 :    
397 :     if (cbp & (1 << (5 - i))) { /* coded */
398 :    
399 :     memset(block, 0, 64 * sizeof(int16_t)); /* clear */
400 :    
401 :     start_timer();
402 :     get_inter_block(bs, block, direction);
403 :     stop_coding_timer();
404 :    
405 :     start_timer();
406 : edgomez 1.49.2.20 dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices);
407 : syskin 1.49.2.17 stop_iquant_timer();
408 :    
409 :     start_timer();
410 :     idct(&data[i * 64]);
411 :     stop_idct_timer();
412 :     }
413 :     }
414 :    
415 :     if (dec->interlacing && pMB->field_dct) {
416 :     next_block = stride;
417 :     stride *= 2;
418 :     }
419 :    
420 :     start_timer();
421 :     if (reduced_resolution) {
422 :     if (cbp & 32)
423 :     add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
424 :     if (cbp & 16)
425 :     add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
426 :     if (cbp & 8)
427 :     add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
428 :     if (cbp & 4)
429 :     add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
430 :     if (cbp & 2)
431 :     add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
432 :     if (cbp & 1)
433 :     add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
434 :     } else {
435 :     if (cbp & 32)
436 :     transfer_16to8add(pY_Cur, &data[0 * 64], stride);
437 :     if (cbp & 16)
438 :     transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride);
439 :     if (cbp & 8)
440 :     transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride);
441 :     if (cbp & 4)
442 :     transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride);
443 :     if (cbp & 2)
444 :     transfer_16to8add(pU_Cur, &data[4 * 64], stride2);
445 :     if (cbp & 1)
446 :     transfer_16to8add(pV_Cur, &data[5 * 64], stride2);
447 :     }
448 :     stop_transfer_timer();
449 :     }
450 : Isibaar 1.1
451 : edgomez 1.48 /* decode an inter macroblock */
452 : syskin 1.49.2.17 static void
453 : edgomez 1.19 decoder_mbinter(DECODER * dec,
454 :     const MACROBLOCK * pMB,
455 :     const uint32_t x_pos,
456 :     const uint32_t y_pos,
457 :     const uint32_t cbp,
458 :     Bitstream * bs,
459 : edgomez 1.47 const uint32_t rounding,
460 : syskin 1.49.2.17 const int reduced_resolution,
461 :     const int ref)
462 : Isibaar 1.1 {
463 : h 1.8 uint32_t stride = dec->edged_width;
464 :     uint32_t stride2 = stride / 2;
465 : edgomez 1.7 uint32_t i;
466 : syskin 1.49.2.17
467 : h 1.5 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
468 : edgomez 1.47
469 : Isibaar 1.1 int uv_dx, uv_dy;
470 : edgomez 1.47 VECTOR mv[4]; /* local copy of mvs */
471 :    
472 :     if (reduced_resolution) {
473 :     pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5);
474 :     pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4);
475 :     pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4);
476 :     for (i = 0; i < 4; i++) {
477 :     mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x);
478 :     mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y);
479 :     }
480 :     } else {
481 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
482 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
483 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
484 :     for (i = 0; i < 4; i++)
485 :     mv[i] = pMB->mvs[i];
486 :     }
487 : edgomez 1.49.2.13
488 : syskin 1.49.2.17 start_timer();
489 : edgomez 1.47
490 : syskin 1.49.2.17 if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
491 : edgomez 1.49.2.13
492 : syskin 1.49.2.17 uv_dx = mv[0].x;
493 :     uv_dy = mv[0].y;
494 :     if (dec->quarterpel) {
495 :     uv_dx /= 2;
496 :     uv_dy /= 2;
497 :     }
498 : edgomez 1.47 uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
499 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
500 :    
501 :     if (reduced_resolution)
502 :     interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
503 : syskin 1.49.2.17 mv[0].x, mv[0].y, stride, rounding);
504 :     else if (dec->quarterpel)
505 :     interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
506 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
507 : edgomez 1.47 mv[0].x, mv[0].y, stride, rounding);
508 :     else
509 : syskin 1.49.2.17 interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
510 :     mv[0].x, mv[0].y, stride, rounding);
511 : edgomez 1.47
512 : syskin 1.49.2.17 } else { /* MODE_INTER4V */
513 : edgomez 1.47
514 : syskin 1.49.2.17 if(dec->quarterpel) {
515 :     uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
516 :     uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
517 :     } else {
518 :     uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
519 :     uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
520 :     }
521 : edgomez 1.47
522 : syskin 1.49.2.17 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
523 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
524 : edgomez 1.47
525 : syskin 1.49.2.17 if (reduced_resolution) {
526 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
527 : syskin 1.49.2.17 mv[0].x, mv[0].y, stride, rounding);
528 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos,
529 : syskin 1.49.2.17 mv[1].x, mv[1].y, stride, rounding);
530 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16,
531 : syskin 1.49.2.17 mv[2].x, mv[2].y, stride, rounding);
532 : edgomez 1.49.2.13 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16,
533 : syskin 1.49.2.17 mv[3].x, mv[3].y, stride, rounding);
534 : edgomez 1.47 interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos,
535 : syskin 1.49.2.17 uv_dx, uv_dy, stride2, rounding);
536 : edgomez 1.47 interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos,
537 : syskin 1.49.2.17 uv_dx, uv_dy, stride2, rounding);
538 : edgomez 1.47
539 : syskin 1.49.2.17 } else if (dec->quarterpel) {
540 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
541 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
542 :     mv[0].x, mv[0].y, stride, rounding);
543 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
544 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
545 :     mv[1].x, mv[1].y, stride, rounding);
546 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
547 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
548 :     mv[2].x, mv[2].y, stride, rounding);
549 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
550 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
551 :     mv[3].x, mv[3].y, stride, rounding);
552 :     } else {
553 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
554 :     mv[0].x, mv[0].y, stride, rounding);
555 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
556 :     mv[1].x, mv[1].y, stride, rounding);
557 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
558 :     mv[2].x, mv[2].y, stride, rounding);
559 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
560 :     mv[3].x, mv[3].y, stride, rounding);
561 : edgomez 1.47 }
562 :     }
563 :    
564 : syskin 1.49.2.17 /* chroma */
565 :     if (reduced_resolution) {
566 :     interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos,
567 :     uv_dx, uv_dy, stride2, rounding);
568 :     interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos,
569 :     uv_dx, uv_dy, stride2, rounding);
570 :     } else {
571 :     interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
572 :     uv_dx, uv_dy, stride2, rounding);
573 :     interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
574 :     uv_dx, uv_dy, stride2, rounding);
575 : edgomez 1.47 }
576 : Isibaar 1.1
577 : syskin 1.49.2.17 stop_comp_timer();
578 : Isibaar 1.34
579 : syskin 1.49.2.17 if (cbp)
580 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur,
581 :     reduced_resolution, pMB);
582 : edgomez 1.47 }
583 :    
584 :     static void
585 :     decoder_mbgmc(DECODER * dec,
586 :     MACROBLOCK * const pMB,
587 :     const uint32_t x_pos,
588 :     const uint32_t y_pos,
589 :     const uint32_t fcode,
590 :     const uint32_t cbp,
591 :     Bitstream * bs,
592 : syskin 1.49.2.17 const uint32_t rounding)
593 : edgomez 1.47 {
594 :     const uint32_t stride = dec->edged_width;
595 :     const uint32_t stride2 = stride / 2;
596 : syskin 1.49.2.17
597 : edgomez 1.47 uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
598 :     uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
599 :     uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
600 :    
601 : syskin 1.49.2.17 NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
602 :    
603 : edgomez 1.47 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
604 : Isibaar 1.1
605 :     start_timer();
606 : edgomez 1.49.2.13
607 : edgomez 1.47 /* this is where the calculations are done */
608 : edgomez 1.49.2.13
609 : syskin 1.49.2.17 gmc_data->predict_16x16(gmc_data,
610 :     dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
611 :     stride, stride, x_pos, y_pos, rounding);
612 :    
613 :     gmc_data->predict_8x8(gmc_data,
614 :     dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
615 :     dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
616 :     stride2, stride2, x_pos, y_pos, rounding);
617 : chl 1.49.2.9
618 : syskin 1.49.2.17 gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
619 : chl 1.49.2.9
620 : syskin 1.49.2.17 pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
621 :     pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
622 : chl 1.49.2.9
623 : edgomez 1.47 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
624 : edgomez 1.49.2.13
625 : edgomez 1.47 stop_transfer_timer();
626 : edgomez 1.49.2.13
627 : syskin 1.49.2.17 if (cbp)
628 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
629 : Isibaar 1.1
630 :     }
631 :    
632 :    
633 : syskin 1.49.2.17 static void
634 : edgomez 1.19 decoder_iframe(DECODER * dec,
635 : syskin 1.49.2.17 Bitstream * bs,
636 :     int reduced_resolution,
637 :     int quant,
638 :     int intra_dc_threshold)
639 : Isibaar 1.1 {
640 : chenm001 1.26 uint32_t bound;
641 :     uint32_t x, y;
642 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
643 :     uint32_t mb_height = dec->mb_height;
644 : edgomez 1.49.2.13
645 : syskin 1.49.2.17 if (reduced_resolution) {
646 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
647 :     mb_height = (dec->height + 31) / 32;
648 :     }
649 : chenm001 1.25
650 : chenm001 1.26 bound = 0;
651 : edgomez 1.7
652 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
653 :     for (x = 0; x < mb_width; x++) {
654 : chenm001 1.26 MACROBLOCK *mb;
655 : Isibaar 1.1 uint32_t mcbpc;
656 :     uint32_t cbpc;
657 :     uint32_t acpred_flag;
658 :     uint32_t cbpy;
659 :     uint32_t cbp;
660 :    
661 : chenm001 1.26 while (BitstreamShowBits(bs, 9) == 1)
662 :     BitstreamSkip(bs, 9);
663 :    
664 :     if (check_resync_marker(bs, 0))
665 :     {
666 : edgomez 1.49.2.13 bound = read_video_packet_header(bs, dec, 0,
667 : edgomez 1.47 &quant, NULL, NULL, &intra_dc_threshold);
668 :     x = bound % mb_width;
669 :     y = bound / mb_width;
670 : chenm001 1.26 }
671 :     mb = &dec->mbs[y * dec->mb_width + x];
672 :    
673 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
674 : chenm001 1.26
675 : Isibaar 1.1 mcbpc = get_mcbpc_intra(bs);
676 :     mb->mode = mcbpc & 7;
677 :     cbpc = (mcbpc >> 4);
678 :    
679 :     acpred_flag = BitstreamGetBit(bs);
680 :    
681 :     cbpy = get_cbpy(bs, 1);
682 :     cbp = (cbpy << 2) | cbpc;
683 :    
684 : edgomez 1.19 if (mb->mode == MODE_INTRA_Q) {
685 :     quant += dquant_table[BitstreamGetBits(bs, 2)];
686 :     if (quant > 31) {
687 : Isibaar 1.1 quant = 31;
688 : edgomez 1.19 } else if (quant < 1) {
689 : Isibaar 1.1 quant = 1;
690 :     }
691 :     }
692 :     mb->quant = quant;
693 : chenm001 1.30 mb->mvs[0].x = mb->mvs[0].y =
694 :     mb->mvs[1].x = mb->mvs[1].y =
695 :     mb->mvs[2].x = mb->mvs[2].y =
696 :     mb->mvs[3].x = mb->mvs[3].y =0;
697 : h 1.5
698 : edgomez 1.19 if (dec->interlacing) {
699 : h 1.5 mb->field_dct = BitstreamGetBit(bs);
700 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
701 : h 1.5 }
702 : Isibaar 1.1
703 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
704 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
705 :    
706 : Isibaar 1.1 }
707 : albeu 1.32 if(dec->out_frm)
708 : syskin 1.49.2.17 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
709 : Isibaar 1.1 }
710 : edgomez 1.7
711 : Isibaar 1.1 }
712 :    
713 :    
714 : syskin 1.49.2.17 static void
715 : edgomez 1.19 get_motion_vector(DECODER * dec,
716 : syskin 1.49.2.17 Bitstream * bs,
717 :     int x,
718 :     int y,
719 :     int k,
720 :     VECTOR * ret_mv,
721 :     int fcode,
722 :     const int bound)
723 : Isibaar 1.1 {
724 : edgomez 1.7
725 : syskin 1.49.2.17 const int scale_fac = 1 << (fcode - 1);
726 :     const int high = (32 * scale_fac) - 1;
727 :     const int low = ((-32) * scale_fac);
728 :     const int range = (64 * scale_fac);
729 : Isibaar 1.1
730 : syskin 1.49.2.17 const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
731 : edgomez 1.47 VECTOR mv;
732 : Isibaar 1.1
733 : edgomez 1.47 mv.x = get_mv(bs, fcode);
734 :     mv.y = get_mv(bs, fcode);
735 : edgomez 1.19
736 : edgomez 1.49.2.5 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);
737 : chenm001 1.26
738 : edgomez 1.47 mv.x += pmv.x;
739 :     mv.y += pmv.y;
740 : Isibaar 1.1
741 : edgomez 1.47 if (mv.x < low) {
742 :     mv.x += range;
743 :     } else if (mv.x > high) {
744 :     mv.x -= range;
745 : Isibaar 1.1 }
746 :    
747 : edgomez 1.47 if (mv.y < low) {
748 :     mv.y += range;
749 :     } else if (mv.y > high) {
750 :     mv.y -= range;
751 : Isibaar 1.1 }
752 :    
753 : edgomez 1.47 ret_mv->x = mv.x;
754 :     ret_mv->y = mv.y;
755 :     }
756 :    
757 :     /* for P_VOP set gmc_warp to NULL */
758 : syskin 1.49.2.17 static void
759 : edgomez 1.19 decoder_pframe(DECODER * dec,
760 : syskin 1.49.2.17 Bitstream * bs,
761 :     int rounding,
762 :     int reduced_resolution,
763 :     int quant,
764 :     int fcode,
765 :     int intra_dc_threshold,
766 :     const WARPPOINTS *const gmc_warp)
767 : Isibaar 1.1 {
768 :     uint32_t x, y;
769 : chenm001 1.26 uint32_t bound;
770 : albeu 1.32 int cp_mb, st_mb;
771 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
772 :     uint32_t mb_height = dec->mb_height;
773 : edgomez 1.49.2.13
774 : syskin 1.49.2.17 if (reduced_resolution) {
775 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
776 :     mb_height = (dec->height + 31) / 32;
777 :     }
778 : Isibaar 1.1
779 :     start_timer();
780 : edgomez 1.19 image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
781 : syskin 1.49.2.17 dec->width, dec->height);
782 : Isibaar 1.1 stop_edges_timer();
783 :    
784 : syskin 1.49.2.17 if (gmc_warp) {
785 :     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
786 : edgomez 1.49.2.13 generate_GMCparameters( dec->sprite_warping_points,
787 :     dec->sprite_warping_accuracy, gmc_warp,
788 : chl 1.49.2.9 dec->width, dec->height, &dec->new_gmc_data);
789 : edgomez 1.47
790 : syskin 1.49.2.17 /* image warping is done block-based in decoder_mbgmc(), now */
791 : edgomez 1.47 }
792 :    
793 : chenm001 1.26 bound = 0;
794 :    
795 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
796 : albeu 1.32 cp_mb = st_mb = 0;
797 : edgomez 1.47 for (x = 0; x < mb_width; x++) {
798 : chenm001 1.26 MACROBLOCK *mb;
799 :    
800 : edgomez 1.48 /* skip stuffing */
801 : chenm001 1.26 while (BitstreamShowBits(bs, 10) == 1)
802 :     BitstreamSkip(bs, 10);
803 :    
804 : syskin 1.49.2.17 if (check_resync_marker(bs, fcode - 1)) {
805 : edgomez 1.49.2.13 bound = read_video_packet_header(bs, dec, fcode - 1,
806 : edgomez 1.47 &quant, &fcode, NULL, &intra_dc_threshold);
807 :     x = bound % mb_width;
808 :     y = bound / mb_width;
809 : chenm001 1.26 }
810 :     mb = &dec->mbs[y * dec->mb_width + x];
811 :    
812 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
813 : suxen_drol 1.23
814 : syskin 1.49.2.17 if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
815 :     uint32_t mcbpc, cbpc, cbpy, cbp;
816 :     uint32_t intra, acpred_flag = 0;
817 : edgomez 1.48 int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */
818 : Isibaar 1.1
819 : albeu 1.32 cp_mb++;
820 : Isibaar 1.1 mcbpc = get_mcbpc_inter(bs);
821 :     mb->mode = mcbpc & 7;
822 :     cbpc = (mcbpc >> 4);
823 : edgomez 1.49.2.13
824 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
825 :     DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
826 : Isibaar 1.1
827 :     intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
828 : edgomez 1.19
829 : edgomez 1.47 if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
830 :     mcsel = BitstreamGetBit(bs);
831 : syskin 1.49.2.17 else if (intra)
832 : edgomez 1.49.2.12 acpred_flag = BitstreamGetBit(bs);
833 : edgomez 1.47
834 : chenm001 1.26 cbpy = get_cbpy(bs, intra);
835 : syskin 1.49.2.17 DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
836 : chenm001 1.25
837 : Isibaar 1.1 cbp = (cbpy << 2) | cbpc;
838 :    
839 : edgomez 1.19 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
840 : chenm001 1.26 int dquant = dquant_table[BitstreamGetBits(bs, 2)];
841 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
842 : chenm001 1.26 quant += dquant;
843 : edgomez 1.19 if (quant > 31) {
844 : Isibaar 1.1 quant = 31;
845 : chenm001 1.26 } else if (quant < 1) {
846 : Isibaar 1.1 quant = 1;
847 :     }
848 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
849 : Isibaar 1.1 }
850 :     mb->quant = quant;
851 : h 1.5
852 : edgomez 1.19 if (dec->interlacing) {
853 : syskin 1.49.2.14 if ((cbp || intra) && !mcsel) {
854 : h 1.37 mb->field_dct = BitstreamGetBit(bs);
855 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
856 : h 1.37 }
857 : h 1.5
858 : edgomez 1.19 if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
859 : h 1.5 mb->field_pred = BitstreamGetBit(bs);
860 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
861 : h 1.5
862 : edgomez 1.19 if (mb->field_pred) {
863 : h 1.5 mb->field_for_top = BitstreamGetBit(bs);
864 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
865 : h 1.5 mb->field_for_bot = BitstreamGetBit(bs);
866 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
867 : h 1.5 }
868 :     }
869 :     }
870 : edgomez 1.49.2.13
871 : edgomez 1.47 if (mcsel) {
872 : syskin 1.49.2.17 decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
873 : edgomez 1.47 continue;
874 :    
875 :     } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
876 : h 1.5
877 : edgomez 1.19 if (dec->interlacing && mb->field_pred) {
878 : syskin 1.49.2.17 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
879 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound);
880 : edgomez 1.19 } else {
881 : syskin 1.49.2.17 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
882 : edgomez 1.47 mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
883 : h 1.5 }
884 : chl 1.35 } else if (mb->mode == MODE_INTER4V ) {
885 : chenm001 1.26 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
886 :     get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
887 :     get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
888 :     get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
889 : syskin 1.49.2.17 } else { /* MODE_INTRA, MODE_INTRA_Q */
890 :     mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
891 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
892 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
893 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
894 : Isibaar 1.1 continue;
895 :     }
896 :    
897 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, cbp, bs,
898 :     rounding, reduced_resolution, 0);
899 : edgomez 1.47
900 : syskin 1.49.2.17 } else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */
901 : edgomez 1.47 mb->mode = MODE_NOT_CODED_GMC;
902 : syskin 1.49.2.17 decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
903 : edgomez 1.39
904 : edgomez 1.47 if(dec->out_frm && cp_mb > 0) {
905 : syskin 1.49.2.17 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
906 :     cp_mb = 0;
907 : edgomez 1.47 }
908 :     st_mb = x+1;
909 : syskin 1.49.2.17 } else { /* not coded P_VOP macroblock */
910 : chenm001 1.14 mb->mode = MODE_NOT_CODED;
911 : edgomez 1.47
912 : Isibaar 1.1 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
913 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
914 : edgomez 1.49.2.13
915 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, 0, bs,
916 :     rounding, reduced_resolution, 0);
917 : edgomez 1.47
918 : albeu 1.32 if(dec->out_frm && cp_mb > 0) {
919 : syskin 1.49.2.17 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
920 :     cp_mb = 0;
921 : albeu 1.32 }
922 :     st_mb = x+1;
923 : Isibaar 1.1 }
924 :     }
925 : syskin 1.49.2.17
926 : albeu 1.32 if(dec->out_frm && cp_mb > 0)
927 : syskin 1.49.2.17 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
928 : Isibaar 1.1 }
929 :     }
930 :    
931 : edgomez 1.47
932 : edgomez 1.48 /* decode B-frame motion vector */
933 : syskin 1.49.2.17 static void
934 :     get_b_motion_vector(Bitstream * bs,
935 : edgomez 1.47 VECTOR * mv,
936 :     int fcode,
937 :     const VECTOR pmv)
938 : chenm001 1.14 {
939 : syskin 1.49.2.17 const int scale_fac = 1 << (fcode - 1);
940 :     const int high = (32 * scale_fac) - 1;
941 :     const int low = ((-32) * scale_fac);
942 :     const int range = (64 * scale_fac);
943 : edgomez 1.7
944 : syskin 1.49.2.17 int mv_x = get_mv(bs, fcode);
945 :     int mv_y = get_mv(bs, fcode);
946 : Isibaar 1.1
947 : syskin 1.49.2.17 mv_x += pmv.x;
948 :     mv_y += pmv.y;
949 : albeu 1.32
950 : syskin 1.49.2.17 if (mv_x < low)
951 : edgomez 1.47 mv_x += range;
952 : syskin 1.49.2.17 else if (mv_x > high)
953 : edgomez 1.47 mv_x -= range;
954 :    
955 : syskin 1.49.2.17 if (mv_y < low)
956 : edgomez 1.47 mv_y += range;
957 : syskin 1.49.2.17 else if (mv_y > high)
958 : edgomez 1.47 mv_y -= range;
959 :    
960 :     mv->x = mv_x;
961 :     mv->y = mv_y;
962 :     }
963 :    
964 : syskin 1.49.2.16 /* decode an B-frame direct & interpolate macroblock */
965 : syskin 1.49.2.17 static void
966 : edgomez 1.47 decoder_bf_interpolate_mbinter(DECODER * dec,
967 : syskin 1.49.2.17 IMAGE forward,
968 :     IMAGE backward,
969 :     const MACROBLOCK * pMB,
970 :     const uint32_t x_pos,
971 :     const uint32_t y_pos,
972 :     Bitstream * bs,
973 :     const int direct)
974 : edgomez 1.47 {
975 :     uint32_t stride = dec->edged_width;
976 :     uint32_t stride2 = stride / 2;
977 :     int uv_dx, uv_dy;
978 :     int b_uv_dx, b_uv_dy;
979 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
980 : syskin 1.49.2.17 const uint32_t cbp = pMB->cbp;
981 : edgomez 1.47
982 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
983 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
984 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
985 :    
986 : syskin 1.49.2.16 if (!direct) {
987 : edgomez 1.47 uv_dx = pMB->mvs[0].x;
988 :     uv_dy = pMB->mvs[0].y;
989 :    
990 :     b_uv_dx = pMB->b_mvs[0].x;
991 :     b_uv_dy = pMB->b_mvs[0].y;
992 :    
993 : syskin 1.49.2.17 if (dec->quarterpel) {
994 : edgomez 1.47 uv_dx /= 2;
995 :     uv_dy /= 2;
996 :     b_uv_dx /= 2;
997 :     b_uv_dy /= 2;
998 :     }
999 :    
1000 :     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
1001 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
1002 :    
1003 :     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
1004 :     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
1005 :    
1006 : syskin 1.49.2.17 } else {
1007 :     if(dec->quarterpel) {
1008 :     uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
1009 :     uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
1010 :     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);
1011 :     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);
1012 :     } else {
1013 :     uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
1014 :     uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
1015 :     b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
1016 :     b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
1017 :     }
1018 : edgomez 1.47
1019 : syskin 1.49.2.17 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
1020 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
1021 :     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
1022 :     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
1023 : edgomez 1.47 }
1024 :    
1025 :     start_timer();
1026 :     if(dec->quarterpel) {
1027 : syskin 1.49.2.17 if(!direct) {
1028 : edgomez 1.47 interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1029 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1030 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1031 :     } else {
1032 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1033 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1034 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1035 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1036 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1037 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1038 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1039 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1040 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1041 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1042 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1043 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1044 : edgomez 1.47 }
1045 : syskin 1.49.2.17 } else {
1046 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
1047 : syskin 1.49.2.17 pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1048 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
1049 : syskin 1.49.2.17 pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1050 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
1051 : syskin 1.49.2.17 pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1052 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8,
1053 : syskin 1.49.2.17 16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1054 : edgomez 1.47 }
1055 :    
1056 :     interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
1057 : syskin 1.49.2.17 uv_dy, stride2, 0);
1058 : edgomez 1.47 interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
1059 : syskin 1.49.2.17 uv_dy, stride2, 0);
1060 : edgomez 1.47
1061 :    
1062 :     if(dec->quarterpel) {
1063 : syskin 1.49.2.17 if(!direct) {
1064 : edgomez 1.47 interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1065 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1066 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1067 :     } else {
1068 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1069 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1070 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1071 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1072 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1073 :     pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1074 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1075 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1076 :     pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1077 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1078 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1079 :     pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1080 : edgomez 1.47 }
1081 : syskin 1.49.2.17 } else {
1082 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos,
1083 : syskin 1.49.2.17 pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1084 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1085 : syskin 1.49.2.17 16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1086 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos,
1087 : syskin 1.49.2.17 16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1088 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1089 : syskin 1.49.2.17 16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1090 : edgomez 1.47 }
1091 :    
1092 :     interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos,
1093 : syskin 1.49.2.17 b_uv_dx, b_uv_dy, stride2, 0);
1094 : edgomez 1.47 interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos,
1095 : syskin 1.49.2.17 b_uv_dx, b_uv_dy, stride2, 0);
1096 : edgomez 1.47
1097 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1098 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1099 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos,
1100 :     stride, 1, 8);
1101 :    
1102 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1103 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1104 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1105 :     stride, 1, 8);
1106 :    
1107 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1108 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1109 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1110 :     stride, 1, 8);
1111 :    
1112 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1113 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1114 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1115 :     stride, 1, 8);
1116 :    
1117 :     interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1118 :     dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1119 :     dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos,
1120 :     stride2, 1, 8);
1121 :    
1122 :     interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1123 :     dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1124 :     dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos,
1125 :     stride2, 1, 8);
1126 :    
1127 :     stop_comp_timer();
1128 :    
1129 : syskin 1.49.2.17 if (cbp)
1130 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
1131 : edgomez 1.47 }
1132 :    
1133 : edgomez 1.48 /* for decode B-frame dbquant */
1134 : syskin 1.49.2.17 static __inline int32_t
1135 : edgomez 1.47 get_dbquant(Bitstream * bs)
1136 :     {
1137 : syskin 1.49.2.17 if (!BitstreamGetBit(bs)) /* '0' */
1138 : edgomez 1.47 return (0);
1139 : syskin 1.49.2.17 else if (!BitstreamGetBit(bs)) /* '10' */
1140 : edgomez 1.47 return (-2);
1141 : syskin 1.49.2.17 else /* '11' */
1142 : edgomez 1.48 return (2);
1143 : edgomez 1.47 }
1144 :    
1145 : edgomez 1.48 /*
1146 : syskin 1.49.2.17 * decode B-frame mb_type
1147 :     * bit ret_value
1148 :     * 1 0
1149 :     * 01 1
1150 :     * 001 2
1151 :     * 0001 3
1152 : edgomez 1.48 */
1153 : syskin 1.49.2.17 static int32_t __inline
1154 : edgomez 1.47 get_mbtype(Bitstream * bs)
1155 :     {
1156 :     int32_t mb_type;
1157 :    
1158 : syskin 1.49.2.17 for (mb_type = 0; mb_type <= 3; mb_type++)
1159 : edgomez 1.47 if (BitstreamGetBit(bs))
1160 : syskin 1.49.2.17 return (mb_type);
1161 : edgomez 1.47
1162 : syskin 1.49.2.17 return -1;
1163 : edgomez 1.47 }
1164 :    
1165 : syskin 1.49.2.17 static void
1166 : edgomez 1.47 decoder_bframe(DECODER * dec,
1167 : syskin 1.49.2.17 Bitstream * bs,
1168 :     int quant,
1169 :     int fcode_forward,
1170 :     int fcode_backward)
1171 : edgomez 1.47 {
1172 :     uint32_t x, y;
1173 :     VECTOR mv;
1174 :     const VECTOR zeromv = {0,0};
1175 : syskin 1.49.2.17 const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp;
1176 :     int i;
1177 :    
1178 : edgomez 1.47 start_timer();
1179 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
1180 : syskin 1.49.2.17 dec->width, dec->height);
1181 : edgomez 1.47 image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
1182 : syskin 1.49.2.17 dec->width, dec->height);
1183 : edgomez 1.47 stop_edges_timer();
1184 :    
1185 :     for (y = 0; y < dec->mb_height; y++) {
1186 : edgomez 1.48 /* Initialize Pred Motion Vector */
1187 : edgomez 1.47 dec->p_fmv = dec->p_bmv = zeromv;
1188 :     for (x = 0; x < dec->mb_width; x++) {
1189 :     MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
1190 :     MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
1191 : edgomez 1.49.2.19 const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward;
1192 :     uint32_t intra_dc_threshold; /* fake variable */
1193 :    
1194 :     if (check_resync_marker(bs, fcode_max - 1)) {
1195 :     int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant,
1196 :     &fcode_forward, &fcode_backward, &intra_dc_threshold);
1197 :     x = bound % dec->mb_width;
1198 :     y = bound / dec->mb_width;
1199 :     /* reset predicted macroblocks */
1200 :     dec->p_fmv = dec->p_bmv = zeromv;
1201 :     }
1202 : edgomez 1.47
1203 :     mv =
1204 :     mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
1205 :     mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
1206 : syskin 1.49.2.17 mb->quant = quant;
1207 : edgomez 1.47
1208 : edgomez 1.48 /*
1209 : edgomez 1.49.2.13 * skip if the co-located P_VOP macroblock is not coded
1210 : edgomez 1.48 * if not codec in co-located S_VOP macroblock is _not_
1211 :     * automatically skipped
1212 :     */
1213 : edgomez 1.47
1214 :     if (last_mb->mode == MODE_NOT_CODED) {
1215 :     mb->cbp = 0;
1216 : syskin 1.49.2.16 mb->mode = MODE_FORWARD;
1217 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1218 : edgomez 1.47 continue;
1219 :     }
1220 :    
1221 : edgomez 1.48 if (!BitstreamGetBit(bs)) { /* modb=='0' */
1222 : edgomez 1.47 const uint8_t modb2 = BitstreamGetBit(bs);
1223 :    
1224 : syskin 1.49.2.16 mb->mode = get_mbtype(bs);
1225 : edgomez 1.47
1226 : syskin 1.49.2.17 if (!modb2) /* modb=='00' */
1227 : edgomez 1.47 mb->cbp = BitstreamGetBits(bs, 6);
1228 : syskin 1.49.2.17 else
1229 : edgomez 1.47 mb->cbp = 0;
1230 : syskin 1.49.2.17
1231 : syskin 1.49.2.16 if (mb->mode && mb->cbp) {
1232 : edgomez 1.47 quant += get_dbquant(bs);
1233 : syskin 1.49.2.17 if (quant > 31)
1234 : edgomez 1.47 quant = 31;
1235 : syskin 1.49.2.17 else if (quant < 1)
1236 : edgomez 1.47 quant = 1;
1237 :     }
1238 : syskin 1.49.2.17 mb->quant = quant;
1239 : syskin 1.49.2.14
1240 :     if (dec->interlacing) {
1241 :     if (mb->cbp) {
1242 :     mb->field_dct = BitstreamGetBit(bs);
1243 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1244 :     }
1245 :    
1246 : syskin 1.49.2.16 if (mb->mode) {
1247 : syskin 1.49.2.14 mb->field_pred = BitstreamGetBit(bs);
1248 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1249 :    
1250 :     if (mb->field_pred) {
1251 :     mb->field_for_top = BitstreamGetBit(bs);
1252 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1253 :     mb->field_for_bot = BitstreamGetBit(bs);
1254 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1255 :     }
1256 :     }
1257 :     }
1258 :    
1259 : edgomez 1.47 } else {
1260 : syskin 1.49.2.16 mb->mode = MODE_DIRECT_NONE_MV;
1261 : edgomez 1.47 mb->cbp = 0;
1262 :     }
1263 :    
1264 : syskin 1.49.2.16 switch (mb->mode) {
1265 : edgomez 1.47 case MODE_DIRECT:
1266 : syskin 1.49.2.17 get_b_motion_vector(bs, &mv, 1, zeromv);
1267 : edgomez 1.47
1268 :     case MODE_DIRECT_NONE_MV:
1269 : syskin 1.49.2.17 for (i = 0; i < 4; i++) {
1270 :     mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x);
1271 :     mb->b_mvs[i].x = (int32_t) ((mv.x == 0)
1272 :     ? ((TRB - TRD) * last_mb->mvs[i].x) / TRD
1273 :     : mb->mvs[i].x - last_mb->mvs[i].x);
1274 :     mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y);
1275 :     mb->b_mvs[i].y = (int32_t) ((mv.y == 0)
1276 :     ? ((TRB - TRD) * last_mb->mvs[i].y) / TRD
1277 :     : mb->mvs[i].y - last_mb->mvs[i].y);
1278 : edgomez 1.47 }
1279 : syskin 1.49.2.17
1280 : edgomez 1.47 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1281 : syskin 1.49.2.17 mb, x, y, bs, 1);
1282 : edgomez 1.47 break;
1283 :    
1284 :     case MODE_INTERPOLATE:
1285 : syskin 1.49.2.17 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv);
1286 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1287 :    
1288 : syskin 1.49.2.17 get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv);
1289 :     dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
1290 : edgomez 1.47
1291 :     decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1292 : syskin 1.49.2.17 mb, x, y, bs, 0);
1293 : edgomez 1.47 break;
1294 :    
1295 :     case MODE_BACKWARD:
1296 : syskin 1.49.2.17 get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv);
1297 : edgomez 1.47 dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1298 :    
1299 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0);
1300 : edgomez 1.47 break;
1301 :    
1302 :     case MODE_FORWARD:
1303 : syskin 1.49.2.17 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv);
1304 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1305 :    
1306 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1307 : edgomez 1.47 break;
1308 :    
1309 :     default:
1310 : syskin 1.49.2.16 DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
1311 : edgomez 1.47 }
1312 : edgomez 1.48 } /* End of for */
1313 :     }
1314 : edgomez 1.47 }
1315 :    
1316 :     /* perform post processing if necessary, and output the image */
1317 : edgomez 1.49.2.13 void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
1318 : suxen_drol 1.49.2.1 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type)
1319 : edgomez 1.47 {
1320 : syskin 1.49.2.25 if (frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV) && mbs != NULL) /* post process */
1321 : Isibaar 1.49.2.22 {
1322 :     /* note: image is stored to tmp */
1323 :     image_copy(&dec->tmp, img, dec->edged_width, dec->height);
1324 :     image_deblock(&dec->tmp, dec->edged_width,
1325 :     mbs, dec->mb_width, dec->mb_height, dec->mb_width,
1326 :     frame->general);
1327 :     img = &dec->tmp;
1328 :     }
1329 :    
1330 : edgomez 1.47 image_output(img, dec->width, dec->height,
1331 : suxen_drol 1.49.2.1 dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
1332 :     frame->output.csp, dec->interlacing);
1333 :    
1334 : syskin 1.49.2.17 if (stats) {
1335 : suxen_drol 1.49.2.1 stats->type = coding2type(coding_type);
1336 :     stats->data.vop.time_base = (int)dec->time_base;
1337 : edgomez 1.49.2.6 stats->data.vop.time_increment = 0; /* XXX: todo */
1338 : suxen_drol 1.49.2.1 }
1339 : edgomez 1.47 }
1340 :    
1341 :    
1342 :     int
1343 :     decoder_decode(DECODER * dec,
1344 : syskin 1.49.2.17 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
1345 : edgomez 1.47 {
1346 :    
1347 :     Bitstream bs;
1348 :     uint32_t rounding;
1349 :     uint32_t reduced_resolution;
1350 :     uint32_t quant;
1351 :     uint32_t fcode_forward;
1352 :     uint32_t fcode_backward;
1353 :     uint32_t intra_dc_threshold;
1354 :     WARPPOINTS gmc_warp;
1355 : suxen_drol 1.49.2.1 int coding_type;
1356 :     int success, output, seen_something;
1357 : edgomez 1.47
1358 : edgomez 1.49.2.10 if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */
1359 : suxen_drol 1.49.2.1 return XVID_ERR_VERSION;
1360 : edgomez 1.47
1361 : suxen_drol 1.49.2.1 start_global_timer();
1362 : edgomez 1.47
1363 : suxen_drol 1.49.2.1 dec->low_delay_default = (frame->general & XVID_LOWDELAY);
1364 :     if ((frame->general & XVID_DISCONTINUITY))
1365 : edgomez 1.47 dec->frames = 0;
1366 : suxen_drol 1.49.2.1 dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
1367 : edgomez 1.47
1368 : syskin 1.49.2.17 if (frame->length < 0) { /* decoder flush */
1369 :     int ret;
1370 :     /* if not decoding "low_delay/packed", and this isn't low_delay and
1371 :     we have a reference frame, then outout the reference frame */
1372 : suxen_drol 1.49.2.3 if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
1373 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type);
1374 : syskin 1.49.2.17 dec->frames = 0;
1375 :     ret = 0;
1376 :     } else {
1377 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1378 :     ret = XVID_ERR_END;
1379 :     }
1380 : edgomez 1.47
1381 :     emms();
1382 :     stop_global_timer();
1383 : suxen_drol 1.49.2.3 return ret;
1384 : edgomez 1.47 }
1385 : edgomez 1.19
1386 : Isibaar 1.1 BitstreamInit(&bs, frame->bitstream, frame->length);
1387 :    
1388 : edgomez 1.48 /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
1389 : edgomez 1.47 if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
1390 :     {
1391 :     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
1392 : suxen_drol 1.49.2.1 (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
1393 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1394 : edgomez 1.47 emms();
1395 : syskin 1.49.2.17 return 1; /* one byte consumed */
1396 : edgomez 1.47 }
1397 :    
1398 : suxen_drol 1.49.2.1 success = 0;
1399 :     output = 0;
1400 :     seen_something = 0;
1401 :    
1402 : edgomez 1.47 repeat:
1403 :    
1404 : syskin 1.49.2.17 coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution,
1405 : edgomez 1.47 &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
1406 :    
1407 : edgomez 1.49.2.13 DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n",
1408 : suxen_drol 1.49.2.1 coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
1409 : Isibaar 1.1
1410 : syskin 1.49.2.17 if (coding_type == -1) { /* nothing */
1411 : edgomez 1.47 if (success) goto done;
1412 : syskin 1.49.2.17 if (stats) stats->type = XVID_TYPE_NOTHING;
1413 : edgomez 1.47 emms();
1414 : syskin 1.49.2.17 return BitstreamPos(&bs)/8;
1415 : Isibaar 1.1 }
1416 :    
1417 : syskin 1.49.2.17 if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
1418 :    
1419 : suxen_drol 1.49.2.1 if (coding_type == -3)
1420 : edgomez 1.47 decoder_resize(dec);
1421 : edgomez 1.49.2.13
1422 : syskin 1.49.2.17 if (stats) {
1423 : suxen_drol 1.49.2.1 stats->type = XVID_TYPE_VOL;
1424 : edgomez 1.47 stats->data.vol.general = 0;
1425 : suxen_drol 1.49.2.1 /*XXX: if (dec->interlacing)
1426 :     stats->data.vol.general |= ++INTERLACING; */
1427 : edgomez 1.47 stats->data.vol.width = dec->width;
1428 :     stats->data.vol.height = dec->height;
1429 : suxen_drol 1.49.2.1 stats->data.vol.par = dec->aspect_ratio;
1430 : edgomez 1.47 stats->data.vol.par_width = dec->par_width;
1431 :     stats->data.vol.par_height = dec->par_height;
1432 :     emms();
1433 : suxen_drol 1.49.2.1 return BitstreamPos(&bs)/8; /* number of bytes consumed */
1434 : edgomez 1.47 }
1435 :     goto repeat;
1436 : edgomez 1.49.2.13 }
1437 : edgomez 1.47
1438 : edgomez 1.48 dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */
1439 : edgomez 1.47
1440 :     /* packed_mode: special-N_VOP treament */
1441 : syskin 1.49.2.17 if (dec->packed_mode && coding_type == N_VOP) {
1442 :     if (dec->low_delay_default && dec->frames > 0) {
1443 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type);
1444 : edgomez 1.47 output = 1;
1445 :     }
1446 :     /* ignore otherwise */
1447 : syskin 1.49.2.17 } else if (coding_type != B_VOP) {
1448 :     switch(coding_type) {
1449 : edgomez 1.47 case I_VOP :
1450 :     decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold);
1451 :     break;
1452 :     case P_VOP :
1453 : edgomez 1.49.2.13 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1454 : edgomez 1.47 fcode_forward, intra_dc_threshold, NULL);
1455 :     break;
1456 :     case S_VOP :
1457 : edgomez 1.49.2.13 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1458 : edgomez 1.47 fcode_forward, intra_dc_threshold, &gmc_warp);
1459 :     break;
1460 :     case N_VOP :
1461 : edgomez 1.49.2.6 /* XXX: not_coded vops are not used for forward prediction */
1462 :     /* we should not swap(last_mbs,mbs) */
1463 : edgomez 1.47 image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
1464 : syskin 1.49.2.23 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */
1465 : edgomez 1.47 break;
1466 :     }
1467 :    
1468 : syskin 1.49.2.17 if (reduced_resolution) {
1469 : edgomez 1.47 image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs,
1470 :     (dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width,
1471 : suxen_drol 1.49.2.1 16, 0);
1472 : edgomez 1.47 }
1473 : Isibaar 1.1
1474 : edgomez 1.47 /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
1475 : syskin 1.49.2.17 if (!(dec->low_delay_default && dec->packed_mode)) {
1476 :     if (dec->low_delay) {
1477 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type);
1478 : edgomez 1.47 output = 1;
1479 : syskin 1.49.2.17 } else if (dec->frames > 0) { /* is the reference frame valid? */
1480 : edgomez 1.47 /* output the reference frame */
1481 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type);
1482 : edgomez 1.47 output = 1;
1483 :     }
1484 :     }
1485 : Isibaar 1.18
1486 : chenm001 1.14 image_swap(&dec->refn[0], &dec->refn[1]);
1487 :     image_swap(&dec->cur, &dec->refn[0]);
1488 : syskin 1.49.2.17 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
1489 : edgomez 1.47 dec->last_reduced_resolution = reduced_resolution;
1490 : syskin 1.49.2.17 dec->last_coding_type = coding_type;
1491 : edgomez 1.47
1492 :     dec->frames++;
1493 :     seen_something = 1;
1494 : chenm001 1.31
1495 : syskin 1.49.2.17 } else { /* B_VOP */
1496 : edgomez 1.47
1497 : syskin 1.49.2.17 if (dec->low_delay) {
1498 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
1499 : edgomez 1.47 dec->low_delay = 1;
1500 :     }
1501 :    
1502 : syskin 1.49.2.17 if (dec->frames < 2) {
1503 : edgomez 1.47 /* attemping to decode a bvop without atleast 2 reference frames */
1504 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1505 :     "broken b-frame, mising ref frames");
1506 : syskin 1.49.2.21 stats->type = XVID_TYPE_NOTHING;
1507 : syskin 1.49.2.17 } else if (dec->time_pp <= dec->time_bp) {
1508 : edgomez 1.49.2.13 /* this occurs when dx50_bvop_compatibility==0 sequences are
1509 : edgomez 1.47 decoded in vfw. */
1510 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1511 :     "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
1512 : syskin 1.49.2.21 stats->type = XVID_TYPE_NOTHING;
1513 : syskin 1.49.2.17 } else {
1514 : edgomez 1.47 decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
1515 : syskin 1.49.2.21 decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type);
1516 : edgomez 1.47 }
1517 :    
1518 :     output = 1;
1519 :     dec->frames++;
1520 : chenm001 1.14 }
1521 :    
1522 : edgomez 1.47 BitstreamByteAlign(&bs);
1523 :    
1524 :     /* low_delay_default mode: repeat in packed_mode */
1525 : syskin 1.49.2.17 if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
1526 : edgomez 1.47 success = 1;
1527 :     goto repeat;
1528 :     }
1529 :    
1530 :     done :
1531 :    
1532 :     /* low_delay_default mode: if we've gotten here without outputting anything,
1533 :     then output the recently decoded frame, or print an error message */
1534 : syskin 1.49.2.17 if (dec->low_delay_default && output == 0) {
1535 :     if (dec->packed_mode && seen_something) {
1536 : edgomez 1.47 /* output the recently decoded frame */
1537 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type);
1538 : syskin 1.49.2.17 } else {
1539 : edgomez 1.47 image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
1540 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1541 :     "warning: nothing to output");
1542 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64,
1543 :     "bframe decoder lag");
1544 :    
1545 : edgomez 1.49.2.13 decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP);
1546 : suxen_drol 1.49.2.1 if (stats) stats->type = XVID_TYPE_NOTHING;
1547 : edgomez 1.47 }
1548 :     }
1549 :    
1550 : Isibaar 1.1 emms();
1551 :     stop_global_timer();
1552 :    
1553 : suxen_drol 1.49.2.1 return BitstreamPos(&bs) / 8; /* number of bytes consumed */
1554 : Isibaar 1.1 }
No admin address has been configured
 ViewVC Help
Powered by ViewVC 1.0.4