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

Annotation of /xvidcore/src/decoder.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.52 - (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 1.51 * Copyright(C) 2002 MinChen <chenm001@163.com>
7 : suxen_drol 1.52 * 2002-2004 Peter Ross <pross@xvid.org>
8 : edgomez 1.42 *
9 : edgomez 1.51 * 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 1.51 * 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 1.51 * 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 1.51 * 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 : suxen_drol 1.52 * $Id: decoder.c,v 1.51 2004/03/22 22:36:23 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 1.51 #include "quant/quant.h"
44 :     #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 : edgomez 1.51 #include "motion/gmc.h"
58 : Isibaar 1.1
59 :     #include "image/image.h"
60 :     #include "image/colorspace.h"
61 : edgomez 1.51 #include "image/postprocessing.h"
62 : Isibaar 1.3 #include "utils/mem_align.h"
63 : Isibaar 1.1
64 : edgomez 1.51 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 1.51 if (dec->last_mbs)
77 : edgomez 1.47 xvid_free(dec->last_mbs);
78 :     if (dec->mbs)
79 :     xvid_free(dec->mbs);
80 : edgomez 1.51 if (dec->qscale)
81 :     xvid_free(dec->qscale);
82 : Isibaar 1.1
83 : edgomez 1.47 /* realloc */
84 : Isibaar 1.1 dec->mb_width = (dec->width + 15) / 16;
85 :     dec->mb_height = (dec->height + 15) / 16;
86 :    
87 :     dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE;
88 :     dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE;
89 : edgomez 1.19
90 :     if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) {
91 : Isibaar 1.3 xvid_free(dec);
92 : Isibaar 1.1 return XVID_ERR_MEMORY;
93 :     }
94 :    
95 : edgomez 1.19 if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) {
96 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
97 : Isibaar 1.3 xvid_free(dec);
98 : Isibaar 1.1 return XVID_ERR_MEMORY;
99 :     }
100 : edgomez 1.47
101 : edgomez 1.48 /* Support B-frame to reference last 2 frame */
102 : edgomez 1.19 if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) {
103 : chenm001 1.11 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 : edgomez 1.47 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);
113 :     return XVID_ERR_MEMORY;
114 :     }
115 :    
116 :     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 :     if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) {
126 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
127 : chenm001 1.14 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 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
131 : chenm001 1.14 xvid_free(dec);
132 :     return XVID_ERR_MEMORY;
133 :     }
134 : Isibaar 1.1
135 : edgomez 1.19 dec->mbs =
136 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
137 :     CACHE_LINE);
138 :     if (dec->mbs == NULL) {
139 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
140 : chenm001 1.14 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
141 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
142 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
143 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
144 : chenm001 1.14 xvid_free(dec);
145 :     return XVID_ERR_MEMORY;
146 :     }
147 : Isibaar 1.20 memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
148 :    
149 : edgomez 1.48 /* For skip MB flag */
150 : edgomez 1.19 dec->last_mbs =
151 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
152 :     CACHE_LINE);
153 :     if (dec->last_mbs == NULL) {
154 : chenm001 1.14 xvid_free(dec->mbs);
155 :     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 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
159 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
160 : Isibaar 1.3 xvid_free(dec);
161 : Isibaar 1.1 return XVID_ERR_MEMORY;
162 :     }
163 : Isibaar 1.20
164 :     memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
165 : Isibaar 1.1
166 : edgomez 1.51 /* 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 : edgomez 1.47 }
175 :    
176 :    
177 :     int
178 : edgomez 1.51 decoder_create(xvid_dec_create_t * create)
179 : edgomez 1.47 {
180 :     DECODER *dec;
181 :    
182 : edgomez 1.51 if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */
183 :     return XVID_ERR_VERSION;
184 :    
185 : edgomez 1.47 dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
186 :     if (dec == NULL) {
187 :     return XVID_ERR_MEMORY;
188 :     }
189 : edgomez 1.51
190 : edgomez 1.47 memset(dec, 0, sizeof(DECODER));
191 :    
192 : edgomez 1.51 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 : edgomez 1.47
200 : edgomez 1.51 dec->width = create->width;
201 :     dec->height = create->height;
202 : edgomez 1.47
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 : edgomez 1.48 /* image based GMC */
210 : edgomez 1.47 image_null(&dec->gmc);
211 :    
212 :     dec->mbs = NULL;
213 :     dec->last_mbs = NULL;
214 : edgomez 1.51 dec->qscale = NULL;
215 : edgomez 1.47
216 : Isibaar 1.1 init_timer();
217 : edgomez 1.51 init_postproc(&dec->postproc);
218 :     init_mpeg_matrix(dec->mpeg_quant_matrices);
219 : Isibaar 1.1
220 : edgomez 1.48 /* For B-frame support (used to save reference frame's time */
221 : edgomez 1.47 dec->frames = 0;
222 : chenm001 1.14 dec->time = dec->time_base = dec->last_time_base = 0;
223 : edgomez 1.47 dec->low_delay = 0;
224 :     dec->packed_mode = 0;
225 :    
226 :     dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
227 : edgomez 1.19
228 : edgomez 1.47 if (dec->fixed_dimensions)
229 :     return decoder_resize(dec);
230 :     else
231 : edgomez 1.51 return 0;
232 : Isibaar 1.1 }
233 :    
234 :    
235 : edgomez 1.19 int
236 :     decoder_destroy(DECODER * dec)
237 : Isibaar 1.1 {
238 : chenm001 1.14 xvid_free(dec->last_mbs);
239 : Isibaar 1.3 xvid_free(dec->mbs);
240 : edgomez 1.51 xvid_free(dec->qscale);
241 : edgomez 1.47
242 : edgomez 1.48 /* image based GMC */
243 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
244 : edgomez 1.47
245 : chenm001 1.11 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
246 : chenm001 1.14 image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
247 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
248 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
249 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
250 : edgomez 1.51 xvid_free(dec->mpeg_quant_matrices);
251 : Isibaar 1.3 xvid_free(dec);
252 : Isibaar 1.1
253 :     write_timer();
254 : edgomez 1.51 return 0;
255 : Isibaar 1.1 }
256 :    
257 : edgomez 1.19 static const int32_t dquant_table[4] = {
258 : Isibaar 1.1 -1, -2, 1, 2
259 :     };
260 :    
261 : edgomez 1.48 /* decode an intra macroblock */
262 : edgomez 1.51 static void
263 : edgomez 1.19 decoder_mbintra(DECODER * dec,
264 :     MACROBLOCK * pMB,
265 :     const uint32_t x_pos,
266 :     const uint32_t y_pos,
267 :     const uint32_t acpred_flag,
268 :     const uint32_t cbp,
269 :     Bitstream * bs,
270 :     const uint32_t quant,
271 : chenm001 1.26 const uint32_t intra_dc_threshold,
272 : edgomez 1.47 const unsigned int bound,
273 :     const int reduced_resolution)
274 : Isibaar 1.1 {
275 : edgomez 1.7
276 :     DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
277 : edgomez 1.19 DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
278 : edgomez 1.7
279 : h 1.8 uint32_t stride = dec->edged_width;
280 :     uint32_t stride2 = stride / 2;
281 :     uint32_t next_block = stride * 8;
282 : h 1.5 uint32_t i;
283 :     uint32_t iQuant = pMB->quant;
284 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
285 :    
286 : edgomez 1.47 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);
292 :     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);
294 :     }
295 : edgomez 1.7
296 : edgomez 1.48 memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */
297 : h 1.5
298 : edgomez 1.19 for (i = 0; i < 6; i++) {
299 : h 1.5 uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);
300 : Isibaar 1.1 int16_t predictors[8];
301 :     int start_coeff;
302 :    
303 :     start_timer();
304 : edgomez 1.19 predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64],
305 : chenm001 1.26 iQuant, iDcScaler, predictors, bound);
306 : edgomez 1.19 if (!acpred_flag) {
307 : h 1.5 pMB->acpred_directions[i] = 0;
308 : Isibaar 1.1 }
309 :     stop_prediction_timer();
310 :    
311 : edgomez 1.19 if (quant < intra_dc_threshold) {
312 : Isibaar 1.1 int dc_size;
313 :     int dc_dif;
314 :    
315 : edgomez 1.19 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;
317 : Isibaar 1.1
318 : edgomez 1.19 if (dc_size > 8) {
319 : edgomez 1.48 BitstreamSkip(bs, 1); /* marker */
320 : Isibaar 1.1 }
321 : edgomez 1.19
322 :     block[i * 64 + 0] = dc_dif;
323 : Isibaar 1.1 start_coeff = 1;
324 : chenm001 1.26
325 : edgomez 1.51 DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif);
326 : edgomez 1.19 } else {
327 : Isibaar 1.1 start_coeff = 0;
328 :     }
329 :    
330 :     start_timer();
331 : edgomez 1.48 if (cbp & (1 << (5 - i))) /* coded */
332 : Isibaar 1.1 {
333 : edgomez 1.47 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 : Isibaar 1.1 }
338 :     stop_coding_timer();
339 :    
340 :     start_timer();
341 : edgomez 1.19 add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors);
342 : Isibaar 1.1 stop_prediction_timer();
343 :    
344 :     start_timer();
345 : edgomez 1.19 if (dec->quant_type == 0) {
346 : edgomez 1.51 dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
347 : edgomez 1.19 } else {
348 : edgomez 1.51 dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
349 : Isibaar 1.1 }
350 :     stop_iquant_timer();
351 :    
352 :     start_timer();
353 : edgomez 1.19 idct(&data[i * 64]);
354 : Isibaar 1.1 stop_idct_timer();
355 : edgomez 1.47
356 : h 1.5 }
357 : Isibaar 1.1
358 : edgomez 1.19 if (dec->interlacing && pMB->field_dct) {
359 : h 1.8 next_block = stride;
360 :     stride *= 2;
361 : Isibaar 1.1 }
362 : h 1.5
363 :     start_timer();
364 : edgomez 1.47
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);
376 :     transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
377 :     transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
378 :     transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
379 :     transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
380 :     transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
381 :     }
382 : h 1.5 stop_transfer_timer();
383 : Isibaar 1.1 }
384 :    
385 : edgomez 1.51 static void
386 :     decoder_mb_decode(DECODER * dec,
387 :     const uint32_t cbp,
388 :     Bitstream * bs,
389 :     uint8_t * pY_Cur,
390 :     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);
396 :     DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
397 :    
398 :     int stride = dec->edged_width;
399 :     int next_block = stride * (reduced_resolution ? 16 : 8);
400 :     const int stride2 = stride/2;
401 :     int i;
402 :     const uint32_t iQuant = pMB->quant;
403 :     const int direction = dec->alternate_vertical_scan ? 2 : 0;
404 :     const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter;
405 :    
406 :     for (i = 0; i < 6; i++) {
407 :    
408 :     if (cbp & (1 << (5 - i))) { /* coded */
409 :    
410 :     memset(block, 0, 64 * sizeof(int16_t)); /* clear */
411 :    
412 :     start_timer();
413 :     get_inter_block(bs, block, direction);
414 :     stop_coding_timer();
415 :    
416 :     start_timer();
417 :     dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices);
418 :     stop_iquant_timer();
419 :    
420 :     start_timer();
421 :     idct(&data[i * 64]);
422 :     stop_idct_timer();
423 :     }
424 :     }
425 : Isibaar 1.1
426 : edgomez 1.51 if (dec->interlacing && pMB->field_dct) {
427 :     next_block = stride;
428 :     stride *= 2;
429 :     }
430 : Isibaar 1.1
431 : edgomez 1.51 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)
447 :     transfer_16to8add(pY_Cur, &data[0 * 64], stride);
448 :     if (cbp & 16)
449 :     transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride);
450 :     if (cbp & 8)
451 :     transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride);
452 :     if (cbp & 4)
453 :     transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride);
454 :     if (cbp & 2)
455 :     transfer_16to8add(pU_Cur, &data[4 * 64], stride2);
456 :     if (cbp & 1)
457 :     transfer_16to8add(pV_Cur, &data[5 * 64], stride2);
458 :     }
459 :     stop_transfer_timer();
460 :     }
461 : Isibaar 1.1
462 : edgomez 1.48 /* decode an inter macroblock */
463 : edgomez 1.51 static void
464 : edgomez 1.19 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 : edgomez 1.47 const uint32_t rounding,
471 : edgomez 1.51 const int reduced_resolution,
472 :     const int ref)
473 : Isibaar 1.1 {
474 : h 1.8 uint32_t stride = dec->edged_width;
475 :     uint32_t stride2 = stride / 2;
476 : edgomez 1.7 uint32_t i;
477 : edgomez 1.51
478 : h 1.5 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
479 : edgomez 1.47
480 : Isibaar 1.1 int uv_dx, uv_dy;
481 : edgomez 1.47 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 : edgomez 1.51 start_timer();
500 :    
501 :     if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
502 :    
503 :     uv_dx = mv[0].x;
504 :     uv_dy = mv[0].y;
505 :     if (dec->quarterpel) {
506 :     uv_dx /= 2;
507 :     uv_dy /= 2;
508 :     }
509 : edgomez 1.47 uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
510 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
511 :    
512 :     if (reduced_resolution)
513 :     interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
514 : edgomez 1.51 mv[0].x, mv[0].y, stride, rounding);
515 :     else if (dec->quarterpel)
516 :     interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
517 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
518 :     mv[0].x, mv[0].y, stride, rounding);
519 : edgomez 1.47 else
520 : edgomez 1.51 interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
521 :     mv[0].x, mv[0].y, stride, rounding);
522 : Isibaar 1.1
523 : edgomez 1.47 } else { /* MODE_INTER4V */
524 :    
525 : edgomez 1.51 if(dec->quarterpel) {
526 :     uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
527 :     uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
528 :     } else {
529 :     uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
530 :     uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
531 :     }
532 : edgomez 1.47
533 : edgomez 1.51 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
534 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
535 : edgomez 1.47
536 : edgomez 1.51 if (reduced_resolution) {
537 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
538 : edgomez 1.51 mv[0].x, mv[0].y, stride, rounding);
539 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos,
540 : edgomez 1.51 mv[1].x, mv[1].y, stride, rounding);
541 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16,
542 : edgomez 1.51 mv[2].x, mv[2].y, stride, rounding);
543 :     interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16,
544 :     mv[3].x, mv[3].y, stride, rounding);
545 : edgomez 1.47 interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos,
546 : edgomez 1.51 uv_dx, uv_dy, stride2, rounding);
547 : edgomez 1.47 interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos,
548 : edgomez 1.51 uv_dx, uv_dy, stride2, rounding);
549 : edgomez 1.47
550 : edgomez 1.51 } else if (dec->quarterpel) {
551 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
552 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
553 :     mv[0].x, mv[0].y, stride, rounding);
554 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
555 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
556 :     mv[1].x, mv[1].y, stride, rounding);
557 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
558 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
559 :     mv[2].x, mv[2].y, stride, rounding);
560 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
561 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
562 :     mv[3].x, mv[3].y, stride, rounding);
563 :     } else {
564 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
565 :     mv[0].x, mv[0].y, stride, rounding);
566 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
567 :     mv[1].x, mv[1].y, stride, rounding);
568 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
569 :     mv[2].x, mv[2].y, stride, rounding);
570 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
571 :     mv[3].x, mv[3].y, stride, rounding);
572 : edgomez 1.47 }
573 :     }
574 :    
575 : edgomez 1.51 /* chroma */
576 :     if (reduced_resolution) {
577 :     interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos,
578 :     uv_dx, uv_dy, stride2, rounding);
579 :     interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos,
580 :     uv_dx, uv_dy, stride2, rounding);
581 :     } else {
582 :     interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
583 :     uv_dx, uv_dy, stride2, rounding);
584 :     interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
585 :     uv_dx, uv_dy, stride2, rounding);
586 : edgomez 1.47 }
587 : Isibaar 1.1
588 : edgomez 1.51 stop_comp_timer();
589 : edgomez 1.47
590 : edgomez 1.51 if (cbp)
591 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur,
592 :     reduced_resolution, pMB);
593 : edgomez 1.47 }
594 :    
595 :     static void
596 :     decoder_mbgmc(DECODER * dec,
597 :     MACROBLOCK * const pMB,
598 :     const uint32_t x_pos,
599 :     const uint32_t y_pos,
600 :     const uint32_t fcode,
601 :     const uint32_t cbp,
602 :     Bitstream * bs,
603 : edgomez 1.51 const uint32_t rounding)
604 : edgomez 1.47 {
605 :     const uint32_t stride = dec->edged_width;
606 :     const uint32_t stride2 = stride / 2;
607 : edgomez 1.51
608 : edgomez 1.47 uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
609 :     uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
610 :     uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
611 :    
612 : edgomez 1.51 NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
613 :    
614 : edgomez 1.47 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
615 : Isibaar 1.1
616 :     start_timer();
617 : edgomez 1.51
618 : edgomez 1.47 /* this is where the calculations are done */
619 :    
620 : edgomez 1.51 gmc_data->predict_16x16(gmc_data,
621 :     dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
622 :     stride, stride, x_pos, y_pos, rounding);
623 :    
624 :     gmc_data->predict_8x8(gmc_data,
625 :     dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
626 :     dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
627 :     stride2, stride2, x_pos, y_pos, rounding);
628 : edgomez 1.47
629 : edgomez 1.51 gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
630 : Isibaar 1.1
631 : edgomez 1.51 pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
632 :     pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
633 : edgomez 1.47
634 : edgomez 1.51 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
635 : Isibaar 1.1
636 : edgomez 1.51 stop_transfer_timer();
637 : Isibaar 1.1
638 : edgomez 1.51 if (cbp)
639 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
640 : Isibaar 1.1
641 :     }
642 :    
643 :    
644 : edgomez 1.51 static void
645 : edgomez 1.19 decoder_iframe(DECODER * dec,
646 : edgomez 1.51 Bitstream * bs,
647 :     int reduced_resolution,
648 :     int quant,
649 :     int intra_dc_threshold)
650 : Isibaar 1.1 {
651 : chenm001 1.26 uint32_t bound;
652 :     uint32_t x, y;
653 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
654 :     uint32_t mb_height = dec->mb_height;
655 : edgomez 1.51
656 :     if (reduced_resolution) {
657 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
658 :     mb_height = (dec->height + 31) / 32;
659 :     }
660 : chenm001 1.25
661 : chenm001 1.26 bound = 0;
662 : edgomez 1.7
663 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
664 :     for (x = 0; x < mb_width; x++) {
665 : chenm001 1.26 MACROBLOCK *mb;
666 : Isibaar 1.1 uint32_t mcbpc;
667 :     uint32_t cbpc;
668 :     uint32_t acpred_flag;
669 :     uint32_t cbpy;
670 :     uint32_t cbp;
671 :    
672 : chenm001 1.26 while (BitstreamShowBits(bs, 9) == 1)
673 :     BitstreamSkip(bs, 9);
674 :    
675 :     if (check_resync_marker(bs, 0))
676 :     {
677 : edgomez 1.51 bound = read_video_packet_header(bs, dec, 0,
678 : edgomez 1.47 &quant, NULL, NULL, &intra_dc_threshold);
679 :     x = bound % mb_width;
680 :     y = bound / mb_width;
681 : chenm001 1.26 }
682 :     mb = &dec->mbs[y * dec->mb_width + x];
683 :    
684 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
685 : chenm001 1.26
686 : Isibaar 1.1 mcbpc = get_mcbpc_intra(bs);
687 :     mb->mode = mcbpc & 7;
688 :     cbpc = (mcbpc >> 4);
689 :    
690 :     acpred_flag = BitstreamGetBit(bs);
691 :    
692 :     cbpy = get_cbpy(bs, 1);
693 :     cbp = (cbpy << 2) | cbpc;
694 :    
695 : edgomez 1.19 if (mb->mode == MODE_INTRA_Q) {
696 :     quant += dquant_table[BitstreamGetBits(bs, 2)];
697 :     if (quant > 31) {
698 : Isibaar 1.1 quant = 31;
699 : edgomez 1.19 } else if (quant < 1) {
700 : Isibaar 1.1 quant = 1;
701 :     }
702 :     }
703 :     mb->quant = quant;
704 : chenm001 1.30 mb->mvs[0].x = mb->mvs[0].y =
705 :     mb->mvs[1].x = mb->mvs[1].y =
706 :     mb->mvs[2].x = mb->mvs[2].y =
707 :     mb->mvs[3].x = mb->mvs[3].y =0;
708 : h 1.5
709 : edgomez 1.19 if (dec->interlacing) {
710 : h 1.5 mb->field_dct = BitstreamGetBit(bs);
711 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
712 : h 1.5 }
713 : Isibaar 1.1
714 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
715 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
716 :    
717 : Isibaar 1.1 }
718 : albeu 1.32 if(dec->out_frm)
719 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
720 : Isibaar 1.1 }
721 : edgomez 1.7
722 : Isibaar 1.1 }
723 :    
724 :    
725 : edgomez 1.51 static void
726 : edgomez 1.19 get_motion_vector(DECODER * dec,
727 : edgomez 1.51 Bitstream * bs,
728 :     int x,
729 :     int y,
730 :     int k,
731 :     VECTOR * ret_mv,
732 :     int fcode,
733 :     const int bound)
734 : Isibaar 1.1 {
735 : edgomez 1.7
736 : edgomez 1.51 const int scale_fac = 1 << (fcode - 1);
737 :     const int high = (32 * scale_fac) - 1;
738 :     const int low = ((-32) * scale_fac);
739 :     const int range = (64 * scale_fac);
740 : Isibaar 1.1
741 : edgomez 1.51 const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
742 : edgomez 1.47 VECTOR mv;
743 : Isibaar 1.1
744 : edgomez 1.47 mv.x = get_mv(bs, fcode);
745 :     mv.y = get_mv(bs, fcode);
746 : edgomez 1.19
747 : edgomez 1.51 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);
748 : chenm001 1.26
749 : edgomez 1.47 mv.x += pmv.x;
750 :     mv.y += pmv.y;
751 : Isibaar 1.1
752 : edgomez 1.47 if (mv.x < low) {
753 :     mv.x += range;
754 :     } else if (mv.x > high) {
755 :     mv.x -= range;
756 : Isibaar 1.1 }
757 :    
758 : edgomez 1.47 if (mv.y < low) {
759 :     mv.y += range;
760 :     } else if (mv.y > high) {
761 :     mv.y -= range;
762 : Isibaar 1.1 }
763 :    
764 : edgomez 1.51 /* clip to valid range */
765 : edgomez 1.47
766 : edgomez 1.51 if (mv.x > ((int)(dec->mb_width - x) << (5 + dec->quarterpel)) )
767 :     mv.x = (int)(dec->mb_width - x) << (5 + dec->quarterpel);
768 :    
769 :     else if (mv.x < (int)(-x-1) << (5 + dec->quarterpel))
770 :     mv.x = (int)(-x-1) << (5 + dec->quarterpel);
771 : edgomez 1.47
772 : edgomez 1.51 if (mv.y > ((int)(dec->mb_height - y) << (5 + dec->quarterpel)) )
773 :     mv.y = (int)(dec->mb_height - y) << (5 + dec->quarterpel);
774 : Isibaar 1.1
775 : edgomez 1.51 else if (mv.y < ((int)(-y-1)) << (5 + dec->quarterpel) )
776 :     mv.y = (int)(-y-1) << (5 + dec->quarterpel);
777 : Isibaar 1.1
778 : edgomez 1.51 ret_mv->x = mv.x;
779 :     ret_mv->y = mv.y;
780 :     }
781 : Isibaar 1.1
782 : edgomez 1.47 /* for P_VOP set gmc_warp to NULL */
783 : edgomez 1.51 static void
784 : edgomez 1.19 decoder_pframe(DECODER * dec,
785 : edgomez 1.51 Bitstream * bs,
786 :     int rounding,
787 :     int reduced_resolution,
788 :     int quant,
789 :     int fcode,
790 :     int intra_dc_threshold,
791 :     const WARPPOINTS *const gmc_warp)
792 : Isibaar 1.1 {
793 :     uint32_t x, y;
794 : chenm001 1.26 uint32_t bound;
795 : albeu 1.32 int cp_mb, st_mb;
796 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
797 :     uint32_t mb_height = dec->mb_height;
798 : edgomez 1.51
799 :     if (reduced_resolution) {
800 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
801 :     mb_height = (dec->height + 31) / 32;
802 :     }
803 : Isibaar 1.1
804 :     start_timer();
805 : edgomez 1.19 image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
806 : edgomez 1.51 dec->width, dec->height, dec->bs_version);
807 : Isibaar 1.1 stop_edges_timer();
808 :    
809 : edgomez 1.51 if (gmc_warp) {
810 :     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
811 :     generate_GMCparameters( dec->sprite_warping_points,
812 :     dec->sprite_warping_accuracy, gmc_warp,
813 :     dec->width, dec->height, &dec->new_gmc_data);
814 : edgomez 1.47
815 : edgomez 1.51 /* image warping is done block-based in decoder_mbgmc(), now */
816 : edgomez 1.47 }
817 :    
818 : chenm001 1.26 bound = 0;
819 :    
820 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
821 : albeu 1.32 cp_mb = st_mb = 0;
822 : edgomez 1.47 for (x = 0; x < mb_width; x++) {
823 : chenm001 1.26 MACROBLOCK *mb;
824 :    
825 : edgomez 1.48 /* skip stuffing */
826 : chenm001 1.26 while (BitstreamShowBits(bs, 10) == 1)
827 :     BitstreamSkip(bs, 10);
828 :    
829 : edgomez 1.51 if (check_resync_marker(bs, fcode - 1)) {
830 :     bound = read_video_packet_header(bs, dec, fcode - 1,
831 : edgomez 1.47 &quant, &fcode, NULL, &intra_dc_threshold);
832 :     x = bound % mb_width;
833 :     y = bound / mb_width;
834 : chenm001 1.26 }
835 :     mb = &dec->mbs[y * dec->mb_width + x];
836 :    
837 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
838 : suxen_drol 1.23
839 : edgomez 1.51 if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
840 :     uint32_t mcbpc, cbpc, cbpy, cbp;
841 :     uint32_t intra, acpred_flag = 0;
842 : edgomez 1.48 int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */
843 : Isibaar 1.1
844 : albeu 1.32 cp_mb++;
845 : Isibaar 1.1 mcbpc = get_mcbpc_inter(bs);
846 :     mb->mode = mcbpc & 7;
847 :     cbpc = (mcbpc >> 4);
848 : edgomez 1.51
849 :     DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
850 :     DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
851 : Isibaar 1.1
852 :     intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
853 : edgomez 1.19
854 : edgomez 1.47 if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
855 :     mcsel = BitstreamGetBit(bs);
856 : edgomez 1.51 else if (intra)
857 :     acpred_flag = BitstreamGetBit(bs);
858 : edgomez 1.47
859 : chenm001 1.26 cbpy = get_cbpy(bs, intra);
860 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
861 : chenm001 1.25
862 : Isibaar 1.1 cbp = (cbpy << 2) | cbpc;
863 :    
864 : edgomez 1.19 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
865 : chenm001 1.26 int dquant = dquant_table[BitstreamGetBits(bs, 2)];
866 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
867 : chenm001 1.26 quant += dquant;
868 : edgomez 1.19 if (quant > 31) {
869 : Isibaar 1.1 quant = 31;
870 : chenm001 1.26 } else if (quant < 1) {
871 : Isibaar 1.1 quant = 1;
872 :     }
873 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
874 : Isibaar 1.1 }
875 :     mb->quant = quant;
876 : h 1.5
877 : edgomez 1.19 if (dec->interlacing) {
878 : h 1.37 if (cbp || intra) {
879 :     mb->field_dct = BitstreamGetBit(bs);
880 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
881 : h 1.37 }
882 : h 1.5
883 : edgomez 1.51 if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) {
884 : h 1.5 mb->field_pred = BitstreamGetBit(bs);
885 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
886 : h 1.5
887 : edgomez 1.19 if (mb->field_pred) {
888 : h 1.5 mb->field_for_top = BitstreamGetBit(bs);
889 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
890 : h 1.5 mb->field_for_bot = BitstreamGetBit(bs);
891 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
892 : h 1.5 }
893 :     }
894 :     }
895 : edgomez 1.51
896 : edgomez 1.47 if (mcsel) {
897 : edgomez 1.51 decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
898 : edgomez 1.47 continue;
899 :    
900 :     } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
901 : h 1.5
902 : edgomez 1.19 if (dec->interlacing && mb->field_pred) {
903 : edgomez 1.51 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
904 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound);
905 : edgomez 1.19 } else {
906 : edgomez 1.51 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
907 : edgomez 1.47 mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
908 : h 1.5 }
909 : chl 1.35 } else if (mb->mode == MODE_INTER4V ) {
910 : chenm001 1.26 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
911 :     get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
912 :     get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
913 :     get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
914 : edgomez 1.51 } else { /* MODE_INTRA, MODE_INTRA_Q */
915 :     mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
916 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
917 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
918 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
919 : Isibaar 1.1 continue;
920 :     }
921 :    
922 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, cbp, bs,
923 :     rounding, reduced_resolution, 0);
924 : edgomez 1.47
925 : edgomez 1.51 } else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */
926 : edgomez 1.47 mb->mode = MODE_NOT_CODED_GMC;
927 : edgomez 1.51 mb->quant = quant;
928 :     decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
929 : edgomez 1.39
930 : edgomez 1.47 if(dec->out_frm && cp_mb > 0) {
931 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
932 :     cp_mb = 0;
933 : edgomez 1.47 }
934 :     st_mb = x+1;
935 : edgomez 1.51 } else { /* not coded P_VOP macroblock */
936 : chenm001 1.14 mb->mode = MODE_NOT_CODED;
937 : edgomez 1.51 mb->quant = quant;
938 : edgomez 1.47
939 : Isibaar 1.1 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
940 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
941 :    
942 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, 0, bs,
943 :     rounding, reduced_resolution, 0);
944 : edgomez 1.47
945 : albeu 1.32 if(dec->out_frm && cp_mb > 0) {
946 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
947 :     cp_mb = 0;
948 : albeu 1.32 }
949 :     st_mb = x+1;
950 : Isibaar 1.1 }
951 :     }
952 : edgomez 1.51
953 : albeu 1.32 if(dec->out_frm && cp_mb > 0)
954 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
955 : Isibaar 1.1 }
956 :     }
957 :    
958 : edgomez 1.47
959 : edgomez 1.48 /* decode B-frame motion vector */
960 : edgomez 1.51 static void
961 :     get_b_motion_vector(Bitstream * bs,
962 : edgomez 1.47 VECTOR * mv,
963 :     int fcode,
964 : edgomez 1.51 const VECTOR pmv,
965 :     const DECODER * const dec,
966 :     const int x, const int y)
967 : chenm001 1.14 {
968 : edgomez 1.51 const int scale_fac = 1 << (fcode - 1);
969 :     const int high = (32 * scale_fac) - 1;
970 :     const int low = ((-32) * scale_fac);
971 :     const int range = (64 * scale_fac);
972 : edgomez 1.7
973 : edgomez 1.51 int mv_x = get_mv(bs, fcode);
974 :     int mv_y = get_mv(bs, fcode);
975 : Isibaar 1.1
976 : edgomez 1.51 mv_x += pmv.x;
977 :     mv_y += pmv.y;
978 : albeu 1.32
979 : edgomez 1.51 if (mv_x < low)
980 : edgomez 1.47 mv_x += range;
981 : edgomez 1.51 else if (mv_x > high)
982 : edgomez 1.47 mv_x -= range;
983 :    
984 : edgomez 1.51 if (mv_y < low)
985 : edgomez 1.47 mv_y += range;
986 : edgomez 1.51 else if (mv_y > high)
987 : edgomez 1.47 mv_y -= range;
988 :    
989 :    
990 : edgomez 1.51 /* clip to valid range */
991 :     if (mv_x > ((int)(dec->mb_width - x) << (5 + dec->quarterpel)) )
992 :     mv_x = (int)(dec->mb_width - x) << (5 + dec->quarterpel);
993 :    
994 :     else if (mv_x < (int)(-x-1) << (5 + dec->quarterpel))
995 :     mv_x = (int)(-x-1) << (5 + dec->quarterpel);
996 : edgomez 1.47
997 : edgomez 1.51 if (mv_y > ((int)(dec->mb_height - y) << (5 + dec->quarterpel)) )
998 :     mv_y = (int)(dec->mb_height - y) << (5 + dec->quarterpel);
999 : edgomez 1.47
1000 : edgomez 1.51 else if (mv_y < ((int)(-y-1)) << (5 + dec->quarterpel) )
1001 :     mv_y = (int)(-y-1) << (5 + dec->quarterpel);
1002 : edgomez 1.47
1003 : edgomez 1.51 mv->x = mv_x;
1004 :     mv->y = mv_y;
1005 : edgomez 1.47 }
1006 :    
1007 : edgomez 1.51 /* decode an B-frame direct & interpolate macroblock */
1008 :     static void
1009 : edgomez 1.47 decoder_bf_interpolate_mbinter(DECODER * dec,
1010 : edgomez 1.51 IMAGE forward,
1011 :     IMAGE backward,
1012 :     const MACROBLOCK * pMB,
1013 :     const uint32_t x_pos,
1014 :     const uint32_t y_pos,
1015 :     Bitstream * bs,
1016 :     const int direct)
1017 : edgomez 1.47 {
1018 :     uint32_t stride = dec->edged_width;
1019 :     uint32_t stride2 = stride / 2;
1020 :     int uv_dx, uv_dy;
1021 :     int b_uv_dx, b_uv_dy;
1022 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
1023 : edgomez 1.51 const uint32_t cbp = pMB->cbp;
1024 : edgomez 1.47
1025 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
1026 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
1027 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
1028 :    
1029 : edgomez 1.51 if (!direct) {
1030 : edgomez 1.47 uv_dx = pMB->mvs[0].x;
1031 :     uv_dy = pMB->mvs[0].y;
1032 :    
1033 :     b_uv_dx = pMB->b_mvs[0].x;
1034 :     b_uv_dy = pMB->b_mvs[0].y;
1035 :    
1036 : edgomez 1.51 if (dec->quarterpel) {
1037 : edgomez 1.47 uv_dx /= 2;
1038 :     uv_dy /= 2;
1039 :     b_uv_dx /= 2;
1040 :     b_uv_dy /= 2;
1041 :     }
1042 :    
1043 :     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
1044 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
1045 :    
1046 :     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
1047 :     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
1048 : edgomez 1.51
1049 : edgomez 1.47 } else {
1050 : edgomez 1.51 if(dec->quarterpel) {
1051 :     uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
1052 :     uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
1053 :     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);
1054 :     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);
1055 :     } else {
1056 :     uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
1057 :     uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
1058 :     b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
1059 :     b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
1060 :     }
1061 : edgomez 1.47
1062 : edgomez 1.51 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
1063 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
1064 :     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
1065 :     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
1066 : edgomez 1.47 }
1067 :    
1068 :     start_timer();
1069 :     if(dec->quarterpel) {
1070 : edgomez 1.51 if(!direct) {
1071 : edgomez 1.47 interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1072 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1073 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1074 :     } else {
1075 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1076 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1077 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1078 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1079 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1080 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1081 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1082 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1083 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1084 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1085 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1086 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1087 : edgomez 1.47 }
1088 : edgomez 1.51 } else {
1089 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
1090 : edgomez 1.51 pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1091 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
1092 : edgomez 1.51 pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1093 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
1094 : edgomez 1.51 pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1095 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8,
1096 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1097 : edgomez 1.47 }
1098 :    
1099 :     interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
1100 : edgomez 1.51 uv_dy, stride2, 0);
1101 : edgomez 1.47 interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
1102 : edgomez 1.51 uv_dy, stride2, 0);
1103 : edgomez 1.47
1104 :    
1105 :     if(dec->quarterpel) {
1106 : edgomez 1.51 if(!direct) {
1107 : edgomez 1.47 interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1108 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1109 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1110 :     } else {
1111 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1112 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1113 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1114 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1115 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1116 :     pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1117 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1118 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1119 :     pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1120 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1121 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1122 :     pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1123 : edgomez 1.47 }
1124 : edgomez 1.51 } else {
1125 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos,
1126 : edgomez 1.51 pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1127 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1128 : edgomez 1.51 16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1129 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos,
1130 : edgomez 1.51 16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1131 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1132 : edgomez 1.51 16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1133 : edgomez 1.47 }
1134 :    
1135 :     interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos,
1136 : edgomez 1.51 b_uv_dx, b_uv_dy, stride2, 0);
1137 : edgomez 1.47 interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos,
1138 : edgomez 1.51 b_uv_dx, b_uv_dy, stride2, 0);
1139 : edgomez 1.47
1140 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1141 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1142 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos,
1143 :     stride, 1, 8);
1144 :    
1145 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1146 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1147 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1148 :     stride, 1, 8);
1149 :    
1150 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1151 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1152 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1153 :     stride, 1, 8);
1154 :    
1155 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1156 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1157 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1158 :     stride, 1, 8);
1159 :    
1160 :     interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1161 :     dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1162 :     dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos,
1163 :     stride2, 1, 8);
1164 :    
1165 :     interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1166 :     dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1167 :     dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos,
1168 :     stride2, 1, 8);
1169 :    
1170 :     stop_comp_timer();
1171 :    
1172 : edgomez 1.51 if (cbp)
1173 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
1174 : edgomez 1.47 }
1175 :    
1176 : edgomez 1.48 /* for decode B-frame dbquant */
1177 : edgomez 1.51 static __inline int32_t
1178 : edgomez 1.47 get_dbquant(Bitstream * bs)
1179 :     {
1180 : edgomez 1.51 if (!BitstreamGetBit(bs)) /* '0' */
1181 : edgomez 1.47 return (0);
1182 : edgomez 1.51 else if (!BitstreamGetBit(bs)) /* '10' */
1183 : edgomez 1.47 return (-2);
1184 : edgomez 1.51 else /* '11' */
1185 : edgomez 1.48 return (2);
1186 : edgomez 1.47 }
1187 :    
1188 : edgomez 1.48 /*
1189 : edgomez 1.51 * decode B-frame mb_type
1190 :     * bit ret_value
1191 :     * 1 0
1192 :     * 01 1
1193 :     * 001 2
1194 :     * 0001 3
1195 : edgomez 1.48 */
1196 : edgomez 1.51 static int32_t __inline
1197 : edgomez 1.47 get_mbtype(Bitstream * bs)
1198 :     {
1199 :     int32_t mb_type;
1200 :    
1201 : edgomez 1.51 for (mb_type = 0; mb_type <= 3; mb_type++)
1202 : edgomez 1.47 if (BitstreamGetBit(bs))
1203 : edgomez 1.51 return (mb_type);
1204 : edgomez 1.47
1205 : edgomez 1.51 return -1;
1206 : edgomez 1.47 }
1207 :    
1208 : edgomez 1.51 static void
1209 : edgomez 1.47 decoder_bframe(DECODER * dec,
1210 : edgomez 1.51 Bitstream * bs,
1211 :     int quant,
1212 :     int fcode_forward,
1213 :     int fcode_backward)
1214 : edgomez 1.47 {
1215 :     uint32_t x, y;
1216 :     VECTOR mv;
1217 :     const VECTOR zeromv = {0,0};
1218 : edgomez 1.51 const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp;
1219 :     int i;
1220 : edgomez 1.47
1221 :     start_timer();
1222 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
1223 : edgomez 1.51 dec->width, dec->height, dec->bs_version);
1224 : edgomez 1.47 image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
1225 : edgomez 1.51 dec->width, dec->height, dec->bs_version);
1226 : edgomez 1.47 stop_edges_timer();
1227 :    
1228 :     for (y = 0; y < dec->mb_height; y++) {
1229 : edgomez 1.48 /* Initialize Pred Motion Vector */
1230 : edgomez 1.47 dec->p_fmv = dec->p_bmv = zeromv;
1231 :     for (x = 0; x < dec->mb_width; x++) {
1232 :     MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
1233 :     MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
1234 : edgomez 1.51 const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward;
1235 :     uint32_t intra_dc_threshold; /* fake variable */
1236 :    
1237 :     if (check_resync_marker(bs, fcode_max - 1)) {
1238 :     int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant,
1239 :     &fcode_forward, &fcode_backward, &intra_dc_threshold);
1240 :     x = bound % dec->mb_width;
1241 :     y = bound / dec->mb_width;
1242 :     /* reset predicted macroblocks */
1243 :     dec->p_fmv = dec->p_bmv = zeromv;
1244 :     }
1245 : edgomez 1.47
1246 :     mv =
1247 :     mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
1248 :     mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
1249 : edgomez 1.51 mb->quant = quant;
1250 : edgomez 1.47
1251 : edgomez 1.48 /*
1252 : edgomez 1.51 * skip if the co-located P_VOP macroblock is not coded
1253 : edgomez 1.48 * if not codec in co-located S_VOP macroblock is _not_
1254 :     * automatically skipped
1255 :     */
1256 : edgomez 1.47
1257 :     if (last_mb->mode == MODE_NOT_CODED) {
1258 :     mb->cbp = 0;
1259 : edgomez 1.51 mb->mode = MODE_FORWARD;
1260 :     decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1261 : edgomez 1.47 continue;
1262 :     }
1263 :    
1264 : edgomez 1.48 if (!BitstreamGetBit(bs)) { /* modb=='0' */
1265 : edgomez 1.47 const uint8_t modb2 = BitstreamGetBit(bs);
1266 :    
1267 : edgomez 1.51 mb->mode = get_mbtype(bs);
1268 : edgomez 1.47
1269 : edgomez 1.51 if (!modb2) /* modb=='00' */
1270 : edgomez 1.47 mb->cbp = BitstreamGetBits(bs, 6);
1271 : edgomez 1.51 else
1272 : edgomez 1.47 mb->cbp = 0;
1273 : edgomez 1.51
1274 :     if (mb->mode && mb->cbp) {
1275 : edgomez 1.47 quant += get_dbquant(bs);
1276 : edgomez 1.51 if (quant > 31)
1277 : edgomez 1.47 quant = 31;
1278 : edgomez 1.51 else if (quant < 1)
1279 : edgomez 1.47 quant = 1;
1280 : edgomez 1.51 }
1281 :     mb->quant = quant;
1282 :    
1283 :     if (dec->interlacing) {
1284 :     if (mb->cbp) {
1285 :     mb->field_dct = BitstreamGetBit(bs);
1286 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1287 :     }
1288 :    
1289 :     if (mb->mode) {
1290 :     mb->field_pred = BitstreamGetBit(bs);
1291 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1292 :    
1293 :     if (mb->field_pred) {
1294 :     mb->field_for_top = BitstreamGetBit(bs);
1295 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1296 :     mb->field_for_bot = BitstreamGetBit(bs);
1297 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1298 :     }
1299 : edgomez 1.47 }
1300 :     }
1301 : edgomez 1.51
1302 : edgomez 1.47 } else {
1303 : edgomez 1.51 mb->mode = MODE_DIRECT_NONE_MV;
1304 : edgomez 1.47 mb->cbp = 0;
1305 :     }
1306 :    
1307 : edgomez 1.51 switch (mb->mode) {
1308 : edgomez 1.47 case MODE_DIRECT:
1309 : edgomez 1.51 get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y);
1310 : edgomez 1.47
1311 :     case MODE_DIRECT_NONE_MV:
1312 : edgomez 1.51 for (i = 0; i < 4; i++) {
1313 :     mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x);
1314 :     mb->b_mvs[i].x = (int32_t) ((mv.x == 0)
1315 :     ? ((TRB - TRD) * last_mb->mvs[i].x) / TRD
1316 :     : mb->mvs[i].x - last_mb->mvs[i].x);
1317 :     mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y);
1318 :     mb->b_mvs[i].y = (int32_t) ((mv.y == 0)
1319 :     ? ((TRB - TRD) * last_mb->mvs[i].y) / TRD
1320 :     : mb->mvs[i].y - last_mb->mvs[i].y);
1321 : edgomez 1.47 }
1322 : edgomez 1.51
1323 : edgomez 1.47 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1324 : edgomez 1.51 mb, x, y, bs, 1);
1325 : edgomez 1.47 break;
1326 :    
1327 :     case MODE_INTERPOLATE:
1328 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1329 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1330 :    
1331 : edgomez 1.51 get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1332 :     dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
1333 : edgomez 1.47
1334 :     decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1335 : edgomez 1.51 mb, x, y, bs, 0);
1336 : edgomez 1.47 break;
1337 :    
1338 :     case MODE_BACKWARD:
1339 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1340 : edgomez 1.47 dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1341 :    
1342 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0);
1343 : edgomez 1.47 break;
1344 :    
1345 :     case MODE_FORWARD:
1346 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1347 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1348 :    
1349 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1350 : edgomez 1.47 break;
1351 :    
1352 :     default:
1353 : edgomez 1.51 DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
1354 : edgomez 1.47 }
1355 : edgomez 1.48 } /* End of for */
1356 :     }
1357 : edgomez 1.47 }
1358 :    
1359 :     /* perform post processing if necessary, and output the image */
1360 : edgomez 1.51 void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
1361 :     xvid_dec_frame_t * frame, xvid_dec_stats_t * stats,
1362 :     int coding_type, int quant)
1363 : edgomez 1.47 {
1364 : edgomez 1.51 if (dec->cartoon_mode)
1365 :     frame->general &= ~XVID_FILMEFFECT;
1366 : edgomez 1.47
1367 : suxen_drol 1.52 if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || frame->brightness!=0)
1368 :     && mbs != NULL) /* post process */
1369 : edgomez 1.47 {
1370 :     /* note: image is stored to tmp */
1371 :     image_copy(&dec->tmp, img, dec->edged_width, dec->height);
1372 : edgomez 1.51 image_postproc(&dec->postproc, &dec->tmp, dec->edged_width,
1373 :     mbs, dec->mb_width, dec->mb_height, dec->mb_width,
1374 : suxen_drol 1.52 frame->general, frame->brightness, dec->frames, (coding_type == B_VOP));
1375 : edgomez 1.47 img = &dec->tmp;
1376 :     }
1377 :    
1378 :     image_output(img, dec->width, dec->height,
1379 : edgomez 1.51 dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
1380 :     frame->output.csp, dec->interlacing);
1381 :    
1382 :     if (stats) {
1383 :     stats->type = coding2type(coding_type);
1384 :     stats->data.vop.time_base = (int)dec->time_base;
1385 :     stats->data.vop.time_increment = 0; /* XXX: todo */
1386 :     stats->data.vop.qscale_stride = dec->mb_width;
1387 :     stats->data.vop.qscale = dec->qscale;
1388 :     if (stats->data.vop.qscale != NULL && mbs != NULL) {
1389 :     int i;
1390 :     for (i = 0; i < dec->mb_width*dec->mb_height; i++)
1391 :     stats->data.vop.qscale[i] = mbs[i].quant;
1392 :     } else
1393 :     stats->data.vop.qscale = NULL;
1394 :     }
1395 : edgomez 1.47 }
1396 :    
1397 :    
1398 :     int
1399 :     decoder_decode(DECODER * dec,
1400 : edgomez 1.51 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
1401 : edgomez 1.47 {
1402 :    
1403 :     Bitstream bs;
1404 :     uint32_t rounding;
1405 :     uint32_t reduced_resolution;
1406 : edgomez 1.51 uint32_t quant = 2;
1407 : edgomez 1.47 uint32_t fcode_forward;
1408 :     uint32_t fcode_backward;
1409 :     uint32_t intra_dc_threshold;
1410 :     WARPPOINTS gmc_warp;
1411 : edgomez 1.51 int coding_type;
1412 :     int success, output, seen_something;
1413 :    
1414 :     if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */
1415 :     return XVID_ERR_VERSION;
1416 : edgomez 1.47
1417 :     start_global_timer();
1418 :    
1419 : edgomez 1.51 dec->low_delay_default = (frame->general & XVID_LOWDELAY);
1420 :     if ((frame->general & XVID_DISCONTINUITY))
1421 : edgomez 1.47 dec->frames = 0;
1422 : edgomez 1.51 dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
1423 : edgomez 1.47
1424 : edgomez 1.51 if (frame->length < 0) { /* decoder flush */
1425 :     int ret;
1426 :     /* if not decoding "low_delay/packed", and this isn't low_delay and
1427 :     we have a reference frame, then outout the reference frame */
1428 :     if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
1429 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1430 :     dec->frames = 0;
1431 :     ret = 0;
1432 :     } else {
1433 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1434 :     ret = XVID_ERR_END;
1435 : edgomez 1.47 }
1436 :    
1437 :     emms();
1438 :     stop_global_timer();
1439 : edgomez 1.51 return ret;
1440 : edgomez 1.47 }
1441 : edgomez 1.19
1442 : Isibaar 1.1 BitstreamInit(&bs, frame->bitstream, frame->length);
1443 :    
1444 : edgomez 1.48 /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
1445 : edgomez 1.47 if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
1446 :     {
1447 :     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
1448 : edgomez 1.51 (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
1449 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1450 : edgomez 1.47 emms();
1451 : edgomez 1.51 return 1; /* one byte consumed */
1452 : edgomez 1.47 }
1453 :    
1454 : edgomez 1.51 success = 0;
1455 :     output = 0;
1456 :     seen_something = 0;
1457 :    
1458 : edgomez 1.47 repeat:
1459 :    
1460 : edgomez 1.51 coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution,
1461 : edgomez 1.47 &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
1462 :    
1463 : edgomez 1.51 DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n",
1464 :     coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
1465 : Isibaar 1.1
1466 : edgomez 1.51 if (coding_type == -1) { /* nothing */
1467 : edgomez 1.47 if (success) goto done;
1468 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1469 : edgomez 1.47 emms();
1470 : edgomez 1.51 return BitstreamPos(&bs)/8;
1471 : Isibaar 1.1 }
1472 :    
1473 : edgomez 1.51 if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
1474 :    
1475 :     if (coding_type == -3)
1476 : edgomez 1.47 decoder_resize(dec);
1477 : edgomez 1.51
1478 :     if (stats) {
1479 :     stats->type = XVID_TYPE_VOL;
1480 : edgomez 1.47 stats->data.vol.general = 0;
1481 : edgomez 1.51 /*XXX: if (dec->interlacing)
1482 :     stats->data.vol.general |= ++INTERLACING; */
1483 : edgomez 1.47 stats->data.vol.width = dec->width;
1484 :     stats->data.vol.height = dec->height;
1485 : edgomez 1.51 stats->data.vol.par = dec->aspect_ratio;
1486 : edgomez 1.47 stats->data.vol.par_width = dec->par_width;
1487 :     stats->data.vol.par_height = dec->par_height;
1488 :     emms();
1489 : edgomez 1.51 return BitstreamPos(&bs)/8; /* number of bytes consumed */
1490 : edgomez 1.47 }
1491 :     goto repeat;
1492 : edgomez 1.51 }
1493 : edgomez 1.47
1494 : 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 */
1495 : edgomez 1.47
1496 :     /* packed_mode: special-N_VOP treament */
1497 : edgomez 1.51 if (dec->packed_mode && coding_type == N_VOP) {
1498 :     if (dec->low_delay_default && dec->frames > 0) {
1499 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1500 : edgomez 1.47 output = 1;
1501 :     }
1502 :     /* ignore otherwise */
1503 : edgomez 1.51 } else if (coding_type != B_VOP) {
1504 :     switch(coding_type) {
1505 : edgomez 1.47 case I_VOP :
1506 :     decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold);
1507 :     break;
1508 :     case P_VOP :
1509 : edgomez 1.51 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1510 : edgomez 1.47 fcode_forward, intra_dc_threshold, NULL);
1511 :     break;
1512 :     case S_VOP :
1513 : edgomez 1.51 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1514 : edgomez 1.47 fcode_forward, intra_dc_threshold, &gmc_warp);
1515 :     break;
1516 :     case N_VOP :
1517 : edgomez 1.51 /* XXX: not_coded vops are not used for forward prediction */
1518 :     /* we should not swap(last_mbs,mbs) */
1519 : edgomez 1.47 image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
1520 : edgomez 1.51 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */
1521 : edgomez 1.47 break;
1522 :     }
1523 :    
1524 : edgomez 1.51 if (reduced_resolution) {
1525 : edgomez 1.47 image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs,
1526 :     (dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width,
1527 : edgomez 1.51 16, 0);
1528 : edgomez 1.47 }
1529 : Isibaar 1.1
1530 : edgomez 1.47 /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
1531 : edgomez 1.51 if (!(dec->low_delay_default && dec->packed_mode)) {
1532 :     if (dec->low_delay) {
1533 :     decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1534 : edgomez 1.47 output = 1;
1535 : edgomez 1.51 } else if (dec->frames > 0) { /* is the reference frame valid? */
1536 : edgomez 1.47 /* output the reference frame */
1537 : edgomez 1.51 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1538 : edgomez 1.47 output = 1;
1539 :     }
1540 :     }
1541 : Isibaar 1.18
1542 : chenm001 1.14 image_swap(&dec->refn[0], &dec->refn[1]);
1543 :     image_swap(&dec->cur, &dec->refn[0]);
1544 : edgomez 1.51 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
1545 : edgomez 1.47 dec->last_reduced_resolution = reduced_resolution;
1546 : edgomez 1.51 dec->last_coding_type = coding_type;
1547 : edgomez 1.47
1548 :     dec->frames++;
1549 :     seen_something = 1;
1550 : chenm001 1.31
1551 : edgomez 1.51 } else { /* B_VOP */
1552 : edgomez 1.47
1553 : edgomez 1.51 if (dec->low_delay) {
1554 :     DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
1555 : edgomez 1.47 dec->low_delay = 1;
1556 :     }
1557 :    
1558 : edgomez 1.51 if (dec->frames < 2) {
1559 : edgomez 1.47 /* attemping to decode a bvop without atleast 2 reference frames */
1560 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1561 :     "broken b-frame, mising ref frames");
1562 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1563 :     } else if (dec->time_pp <= dec->time_bp) {
1564 :     /* this occurs when dx50_bvop_compatibility==0 sequences are
1565 : edgomez 1.47 decoded in vfw. */
1566 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1567 :     "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
1568 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1569 :     } else {
1570 : edgomez 1.47 decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
1571 : edgomez 1.51 decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1572 : edgomez 1.47 }
1573 :    
1574 :     output = 1;
1575 :     dec->frames++;
1576 : chenm001 1.14 }
1577 :    
1578 : edgomez 1.51 #if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */
1579 :     BitstreamByteAlign(&bs);
1580 :     #endif
1581 : edgomez 1.47
1582 :     /* low_delay_default mode: repeat in packed_mode */
1583 : edgomez 1.51 if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
1584 : edgomez 1.47 success = 1;
1585 :     goto repeat;
1586 :     }
1587 :    
1588 :     done :
1589 :    
1590 :     /* low_delay_default mode: if we've gotten here without outputting anything,
1591 :     then output the recently decoded frame, or print an error message */
1592 : edgomez 1.51 if (dec->low_delay_default && output == 0) {
1593 :     if (dec->packed_mode && seen_something) {
1594 : edgomez 1.47 /* output the recently decoded frame */
1595 : edgomez 1.51 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1596 :     } else {
1597 : edgomez 1.47 image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
1598 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1599 :     "warning: nothing to output");
1600 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64,
1601 :     "bframe decoder lag");
1602 :    
1603 : edgomez 1.51 decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant);
1604 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1605 : edgomez 1.47 }
1606 :     }
1607 : Isibaar 1.50
1608 : Isibaar 1.1 emms();
1609 :     stop_global_timer();
1610 :    
1611 : edgomez 1.51 return (BitstreamPos(&bs) + 7) / 8; /* number of bytes consumed */
1612 : Isibaar 1.1 }

No admin address has been configured
ViewVC Help
Powered by ViewVC 1.0.4