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

Annotation of /xvidcore/src/decoder.c

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1.55 - (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 1.55 * $Id: decoder.c,v 1.54 2004/04/11 09:41:27 syskin 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 : syskin 1.54 for (i = 0; i < 4; i++) {
500 :     /* clip to valid range */
501 :     int border = (int)(dec->mb_width - x_pos) << (5 + dec->quarterpel);
502 :     if (mv[i].x > border) {
503 :     DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", mv[i].x, border, x_pos, y_pos);
504 :     mv[i].x = border;
505 :     } else {
506 :     border = (-(int)x_pos-1) << (5 + dec->quarterpel);
507 :     if (mv[i].x < border) {
508 :     DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", mv[i].x, border, x_pos, y_pos);
509 :     mv[i].x = border;
510 :     }
511 :     }
512 :    
513 :     border = (int)(dec->mb_height - y_pos) << (5 + dec->quarterpel);
514 :     if (mv[i].y > border) {
515 :     DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", mv[i].y, border, x_pos, y_pos);
516 :     mv[i].y = border;
517 :     } else {
518 :     border = (-(int)y_pos-1) << (5 + dec->quarterpel);
519 :     if (mv[i].y < border) {
520 :     DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", mv[i].y, border, x_pos, y_pos);
521 :     mv[i].y = border;
522 :     }
523 :     }
524 :     }
525 :    
526 : edgomez 1.51 start_timer();
527 :    
528 :     if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
529 :    
530 :     uv_dx = mv[0].x;
531 :     uv_dy = mv[0].y;
532 :     if (dec->quarterpel) {
533 :     uv_dx /= 2;
534 :     uv_dy /= 2;
535 :     }
536 : edgomez 1.47 uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
537 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
538 :    
539 :     if (reduced_resolution)
540 :     interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
541 : edgomez 1.51 mv[0].x, mv[0].y, stride, rounding);
542 :     else if (dec->quarterpel)
543 :     interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
544 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
545 :     mv[0].x, mv[0].y, stride, rounding);
546 : edgomez 1.47 else
547 : edgomez 1.51 interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
548 :     mv[0].x, mv[0].y, stride, rounding);
549 : Isibaar 1.1
550 : edgomez 1.47 } else { /* MODE_INTER4V */
551 :    
552 : edgomez 1.51 if(dec->quarterpel) {
553 :     uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
554 :     uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
555 :     } else {
556 :     uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
557 :     uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
558 :     }
559 : edgomez 1.47
560 : edgomez 1.51 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
561 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
562 : edgomez 1.47
563 : edgomez 1.51 if (reduced_resolution) {
564 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
565 : edgomez 1.51 mv[0].x, mv[0].y, stride, rounding);
566 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos,
567 : edgomez 1.51 mv[1].x, mv[1].y, stride, rounding);
568 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16,
569 : edgomez 1.51 mv[2].x, mv[2].y, stride, rounding);
570 :     interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16,
571 :     mv[3].x, mv[3].y, stride, rounding);
572 : edgomez 1.47 interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos,
573 : edgomez 1.51 uv_dx, uv_dy, stride2, rounding);
574 : edgomez 1.47 interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos,
575 : edgomez 1.51 uv_dx, uv_dy, stride2, rounding);
576 : edgomez 1.47
577 : edgomez 1.51 } else if (dec->quarterpel) {
578 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
579 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
580 :     mv[0].x, mv[0].y, stride, rounding);
581 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
582 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
583 :     mv[1].x, mv[1].y, stride, rounding);
584 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
585 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
586 :     mv[2].x, mv[2].y, stride, rounding);
587 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
588 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
589 :     mv[3].x, mv[3].y, stride, rounding);
590 :     } else {
591 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
592 :     mv[0].x, mv[0].y, stride, rounding);
593 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
594 :     mv[1].x, mv[1].y, stride, rounding);
595 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
596 :     mv[2].x, mv[2].y, stride, rounding);
597 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
598 :     mv[3].x, mv[3].y, stride, rounding);
599 : edgomez 1.47 }
600 :     }
601 :    
602 : edgomez 1.51 /* chroma */
603 :     if (reduced_resolution) {
604 :     interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos,
605 :     uv_dx, uv_dy, stride2, rounding);
606 :     interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos,
607 :     uv_dx, uv_dy, stride2, rounding);
608 :     } else {
609 :     interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
610 :     uv_dx, uv_dy, stride2, rounding);
611 :     interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
612 :     uv_dx, uv_dy, stride2, rounding);
613 : edgomez 1.47 }
614 : Isibaar 1.1
615 : edgomez 1.51 stop_comp_timer();
616 : edgomez 1.47
617 : edgomez 1.51 if (cbp)
618 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur,
619 :     reduced_resolution, pMB);
620 : edgomez 1.47 }
621 :    
622 :     static void
623 :     decoder_mbgmc(DECODER * dec,
624 :     MACROBLOCK * const pMB,
625 :     const uint32_t x_pos,
626 :     const uint32_t y_pos,
627 :     const uint32_t fcode,
628 :     const uint32_t cbp,
629 :     Bitstream * bs,
630 : edgomez 1.51 const uint32_t rounding)
631 : edgomez 1.47 {
632 :     const uint32_t stride = dec->edged_width;
633 :     const uint32_t stride2 = stride / 2;
634 : edgomez 1.51
635 : edgomez 1.47 uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
636 :     uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
637 :     uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
638 :    
639 : edgomez 1.51 NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
640 :    
641 : edgomez 1.47 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
642 : Isibaar 1.1
643 :     start_timer();
644 : edgomez 1.51
645 : edgomez 1.47 /* this is where the calculations are done */
646 :    
647 : edgomez 1.51 gmc_data->predict_16x16(gmc_data,
648 :     dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
649 :     stride, stride, x_pos, y_pos, rounding);
650 :    
651 :     gmc_data->predict_8x8(gmc_data,
652 :     dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
653 :     dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
654 :     stride2, stride2, x_pos, y_pos, rounding);
655 : edgomez 1.47
656 : edgomez 1.51 gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
657 : Isibaar 1.1
658 : edgomez 1.51 pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
659 :     pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
660 : edgomez 1.47
661 : edgomez 1.51 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
662 : Isibaar 1.1
663 : edgomez 1.51 stop_transfer_timer();
664 : Isibaar 1.1
665 : edgomez 1.51 if (cbp)
666 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
667 : Isibaar 1.1
668 :     }
669 :    
670 :    
671 : edgomez 1.51 static void
672 : edgomez 1.19 decoder_iframe(DECODER * dec,
673 : edgomez 1.51 Bitstream * bs,
674 :     int reduced_resolution,
675 :     int quant,
676 :     int intra_dc_threshold)
677 : Isibaar 1.1 {
678 : chenm001 1.26 uint32_t bound;
679 :     uint32_t x, y;
680 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
681 :     uint32_t mb_height = dec->mb_height;
682 : edgomez 1.51
683 :     if (reduced_resolution) {
684 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
685 :     mb_height = (dec->height + 31) / 32;
686 :     }
687 : chenm001 1.25
688 : chenm001 1.26 bound = 0;
689 : edgomez 1.7
690 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
691 :     for (x = 0; x < mb_width; x++) {
692 : chenm001 1.26 MACROBLOCK *mb;
693 : Isibaar 1.1 uint32_t mcbpc;
694 :     uint32_t cbpc;
695 :     uint32_t acpred_flag;
696 :     uint32_t cbpy;
697 :     uint32_t cbp;
698 :    
699 : chenm001 1.26 while (BitstreamShowBits(bs, 9) == 1)
700 :     BitstreamSkip(bs, 9);
701 :    
702 :     if (check_resync_marker(bs, 0))
703 :     {
704 : edgomez 1.51 bound = read_video_packet_header(bs, dec, 0,
705 : edgomez 1.47 &quant, NULL, NULL, &intra_dc_threshold);
706 :     x = bound % mb_width;
707 :     y = bound / mb_width;
708 : chenm001 1.26 }
709 :     mb = &dec->mbs[y * dec->mb_width + x];
710 :    
711 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
712 : chenm001 1.26
713 : Isibaar 1.1 mcbpc = get_mcbpc_intra(bs);
714 :     mb->mode = mcbpc & 7;
715 :     cbpc = (mcbpc >> 4);
716 :    
717 :     acpred_flag = BitstreamGetBit(bs);
718 :    
719 :     cbpy = get_cbpy(bs, 1);
720 :     cbp = (cbpy << 2) | cbpc;
721 :    
722 : edgomez 1.19 if (mb->mode == MODE_INTRA_Q) {
723 :     quant += dquant_table[BitstreamGetBits(bs, 2)];
724 :     if (quant > 31) {
725 : Isibaar 1.1 quant = 31;
726 : edgomez 1.19 } else if (quant < 1) {
727 : Isibaar 1.1 quant = 1;
728 :     }
729 :     }
730 :     mb->quant = quant;
731 : chenm001 1.30 mb->mvs[0].x = mb->mvs[0].y =
732 :     mb->mvs[1].x = mb->mvs[1].y =
733 :     mb->mvs[2].x = mb->mvs[2].y =
734 :     mb->mvs[3].x = mb->mvs[3].y =0;
735 : h 1.5
736 : edgomez 1.19 if (dec->interlacing) {
737 : h 1.5 mb->field_dct = BitstreamGetBit(bs);
738 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
739 : h 1.5 }
740 : Isibaar 1.1
741 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
742 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
743 :    
744 : Isibaar 1.1 }
745 : albeu 1.32 if(dec->out_frm)
746 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
747 : Isibaar 1.1 }
748 : edgomez 1.7
749 : Isibaar 1.1 }
750 :    
751 :    
752 : edgomez 1.51 static void
753 : edgomez 1.19 get_motion_vector(DECODER * dec,
754 : edgomez 1.51 Bitstream * bs,
755 :     int x,
756 :     int y,
757 :     int k,
758 :     VECTOR * ret_mv,
759 :     int fcode,
760 :     const int bound)
761 : Isibaar 1.1 {
762 : edgomez 1.7
763 : edgomez 1.51 const int scale_fac = 1 << (fcode - 1);
764 :     const int high = (32 * scale_fac) - 1;
765 :     const int low = ((-32) * scale_fac);
766 :     const int range = (64 * scale_fac);
767 : Isibaar 1.1
768 : edgomez 1.51 const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
769 : edgomez 1.47 VECTOR mv;
770 : Isibaar 1.1
771 : edgomez 1.47 mv.x = get_mv(bs, fcode);
772 :     mv.y = get_mv(bs, fcode);
773 : edgomez 1.19
774 : 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);
775 : chenm001 1.26
776 : edgomez 1.47 mv.x += pmv.x;
777 :     mv.y += pmv.y;
778 : Isibaar 1.1
779 : edgomez 1.47 if (mv.x < low) {
780 :     mv.x += range;
781 :     } else if (mv.x > high) {
782 :     mv.x -= range;
783 : Isibaar 1.1 }
784 :    
785 : edgomez 1.47 if (mv.y < low) {
786 :     mv.y += range;
787 :     } else if (mv.y > high) {
788 :     mv.y -= range;
789 : Isibaar 1.1 }
790 :    
791 : edgomez 1.51 ret_mv->x = mv.x;
792 :     ret_mv->y = mv.y;
793 :     }
794 : Isibaar 1.1
795 : edgomez 1.47 /* for P_VOP set gmc_warp to NULL */
796 : edgomez 1.51 static void
797 : edgomez 1.19 decoder_pframe(DECODER * dec,
798 : edgomez 1.51 Bitstream * bs,
799 :     int rounding,
800 :     int reduced_resolution,
801 :     int quant,
802 :     int fcode,
803 :     int intra_dc_threshold,
804 :     const WARPPOINTS *const gmc_warp)
805 : Isibaar 1.1 {
806 :     uint32_t x, y;
807 : chenm001 1.26 uint32_t bound;
808 : albeu 1.32 int cp_mb, st_mb;
809 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
810 :     uint32_t mb_height = dec->mb_height;
811 : edgomez 1.51
812 :     if (reduced_resolution) {
813 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
814 :     mb_height = (dec->height + 31) / 32;
815 :     }
816 : Isibaar 1.1
817 :     start_timer();
818 : edgomez 1.19 image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
819 : edgomez 1.51 dec->width, dec->height, dec->bs_version);
820 : Isibaar 1.1 stop_edges_timer();
821 :    
822 : edgomez 1.51 if (gmc_warp) {
823 :     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
824 :     generate_GMCparameters( dec->sprite_warping_points,
825 :     dec->sprite_warping_accuracy, gmc_warp,
826 :     dec->width, dec->height, &dec->new_gmc_data);
827 : edgomez 1.47
828 : edgomez 1.51 /* image warping is done block-based in decoder_mbgmc(), now */
829 : edgomez 1.47 }
830 :    
831 : chenm001 1.26 bound = 0;
832 :    
833 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
834 : albeu 1.32 cp_mb = st_mb = 0;
835 : edgomez 1.47 for (x = 0; x < mb_width; x++) {
836 : chenm001 1.26 MACROBLOCK *mb;
837 :    
838 : edgomez 1.48 /* skip stuffing */
839 : chenm001 1.26 while (BitstreamShowBits(bs, 10) == 1)
840 :     BitstreamSkip(bs, 10);
841 :    
842 : edgomez 1.51 if (check_resync_marker(bs, fcode - 1)) {
843 :     bound = read_video_packet_header(bs, dec, fcode - 1,
844 : edgomez 1.47 &quant, &fcode, NULL, &intra_dc_threshold);
845 :     x = bound % mb_width;
846 :     y = bound / mb_width;
847 : chenm001 1.26 }
848 :     mb = &dec->mbs[y * dec->mb_width + x];
849 :    
850 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
851 : suxen_drol 1.23
852 : edgomez 1.51 if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
853 :     uint32_t mcbpc, cbpc, cbpy, cbp;
854 :     uint32_t intra, acpred_flag = 0;
855 : edgomez 1.48 int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */
856 : Isibaar 1.1
857 : albeu 1.32 cp_mb++;
858 : Isibaar 1.1 mcbpc = get_mcbpc_inter(bs);
859 :     mb->mode = mcbpc & 7;
860 :     cbpc = (mcbpc >> 4);
861 : edgomez 1.51
862 :     DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
863 :     DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
864 : Isibaar 1.1
865 :     intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
866 : edgomez 1.19
867 : edgomez 1.47 if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
868 :     mcsel = BitstreamGetBit(bs);
869 : edgomez 1.51 else if (intra)
870 :     acpred_flag = BitstreamGetBit(bs);
871 : edgomez 1.47
872 : chenm001 1.26 cbpy = get_cbpy(bs, intra);
873 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
874 : chenm001 1.25
875 : Isibaar 1.1 cbp = (cbpy << 2) | cbpc;
876 :    
877 : edgomez 1.19 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
878 : chenm001 1.26 int dquant = dquant_table[BitstreamGetBits(bs, 2)];
879 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
880 : chenm001 1.26 quant += dquant;
881 : edgomez 1.19 if (quant > 31) {
882 : Isibaar 1.1 quant = 31;
883 : chenm001 1.26 } else if (quant < 1) {
884 : Isibaar 1.1 quant = 1;
885 :     }
886 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
887 : Isibaar 1.1 }
888 :     mb->quant = quant;
889 : h 1.5
890 : edgomez 1.19 if (dec->interlacing) {
891 : h 1.37 if (cbp || intra) {
892 :     mb->field_dct = BitstreamGetBit(bs);
893 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
894 : h 1.37 }
895 : h 1.5
896 : edgomez 1.51 if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) {
897 : h 1.5 mb->field_pred = BitstreamGetBit(bs);
898 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
899 : h 1.5
900 : edgomez 1.19 if (mb->field_pred) {
901 : h 1.5 mb->field_for_top = BitstreamGetBit(bs);
902 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
903 : h 1.5 mb->field_for_bot = BitstreamGetBit(bs);
904 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
905 : h 1.5 }
906 :     }
907 :     }
908 : edgomez 1.51
909 : edgomez 1.47 if (mcsel) {
910 : edgomez 1.51 decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
911 : edgomez 1.47 continue;
912 :    
913 :     } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
914 : h 1.5
915 : edgomez 1.19 if (dec->interlacing && mb->field_pred) {
916 : edgomez 1.51 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
917 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound);
918 : edgomez 1.19 } else {
919 : edgomez 1.51 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
920 : edgomez 1.47 mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
921 : h 1.5 }
922 : chl 1.35 } else if (mb->mode == MODE_INTER4V ) {
923 : chenm001 1.26 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
924 :     get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
925 :     get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
926 :     get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
927 : edgomez 1.51 } else { /* MODE_INTRA, MODE_INTRA_Q */
928 :     mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
929 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
930 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
931 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
932 : Isibaar 1.1 continue;
933 :     }
934 :    
935 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, cbp, bs,
936 :     rounding, reduced_resolution, 0);
937 : edgomez 1.47
938 : edgomez 1.51 } else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */
939 : edgomez 1.47 mb->mode = MODE_NOT_CODED_GMC;
940 : edgomez 1.51 mb->quant = quant;
941 :     decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
942 : edgomez 1.39
943 : edgomez 1.47 if(dec->out_frm && cp_mb > 0) {
944 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
945 :     cp_mb = 0;
946 : edgomez 1.47 }
947 :     st_mb = x+1;
948 : edgomez 1.51 } else { /* not coded P_VOP macroblock */
949 : chenm001 1.14 mb->mode = MODE_NOT_CODED;
950 : edgomez 1.51 mb->quant = quant;
951 : edgomez 1.47
952 : Isibaar 1.1 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
953 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
954 :    
955 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, 0, bs,
956 :     rounding, reduced_resolution, 0);
957 : edgomez 1.47
958 : albeu 1.32 if(dec->out_frm && cp_mb > 0) {
959 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
960 :     cp_mb = 0;
961 : albeu 1.32 }
962 :     st_mb = x+1;
963 : Isibaar 1.1 }
964 :     }
965 : edgomez 1.51
966 : albeu 1.32 if(dec->out_frm && cp_mb > 0)
967 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
968 : Isibaar 1.1 }
969 :     }
970 :    
971 : edgomez 1.47
972 : edgomez 1.48 /* decode B-frame motion vector */
973 : edgomez 1.51 static void
974 :     get_b_motion_vector(Bitstream * bs,
975 : edgomez 1.47 VECTOR * mv,
976 :     int fcode,
977 : edgomez 1.51 const VECTOR pmv,
978 :     const DECODER * const dec,
979 :     const int x, const int y)
980 : chenm001 1.14 {
981 : edgomez 1.51 const int scale_fac = 1 << (fcode - 1);
982 :     const int high = (32 * scale_fac) - 1;
983 :     const int low = ((-32) * scale_fac);
984 :     const int range = (64 * scale_fac);
985 : edgomez 1.7
986 : edgomez 1.51 int mv_x = get_mv(bs, fcode);
987 :     int mv_y = get_mv(bs, fcode);
988 : Isibaar 1.1
989 : edgomez 1.51 mv_x += pmv.x;
990 :     mv_y += pmv.y;
991 : albeu 1.32
992 : edgomez 1.51 if (mv_x < low)
993 : edgomez 1.47 mv_x += range;
994 : edgomez 1.51 else if (mv_x > high)
995 : edgomez 1.47 mv_x -= range;
996 :    
997 : edgomez 1.51 if (mv_y < low)
998 : edgomez 1.47 mv_y += range;
999 : edgomez 1.51 else if (mv_y > high)
1000 : edgomez 1.47 mv_y -= range;
1001 :    
1002 : edgomez 1.51 mv->x = mv_x;
1003 :     mv->y = mv_y;
1004 : edgomez 1.47 }
1005 :    
1006 : edgomez 1.51 /* decode an B-frame direct & interpolate macroblock */
1007 :     static void
1008 : edgomez 1.47 decoder_bf_interpolate_mbinter(DECODER * dec,
1009 : edgomez 1.51 IMAGE forward,
1010 :     IMAGE backward,
1011 :     const MACROBLOCK * pMB,
1012 :     const uint32_t x_pos,
1013 :     const uint32_t y_pos,
1014 :     Bitstream * bs,
1015 :     const int direct)
1016 : edgomez 1.47 {
1017 :     uint32_t stride = dec->edged_width;
1018 :     uint32_t stride2 = stride / 2;
1019 :     int uv_dx, uv_dy;
1020 :     int b_uv_dx, b_uv_dy;
1021 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
1022 : edgomez 1.51 const uint32_t cbp = pMB->cbp;
1023 : edgomez 1.47
1024 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
1025 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
1026 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
1027 :    
1028 : edgomez 1.51 if (!direct) {
1029 : edgomez 1.47 uv_dx = pMB->mvs[0].x;
1030 :     uv_dy = pMB->mvs[0].y;
1031 :    
1032 :     b_uv_dx = pMB->b_mvs[0].x;
1033 :     b_uv_dy = pMB->b_mvs[0].y;
1034 :    
1035 : edgomez 1.51 if (dec->quarterpel) {
1036 : edgomez 1.47 uv_dx /= 2;
1037 :     uv_dy /= 2;
1038 :     b_uv_dx /= 2;
1039 :     b_uv_dy /= 2;
1040 :     }
1041 :    
1042 :     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
1043 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
1044 :    
1045 :     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
1046 :     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
1047 : edgomez 1.51
1048 : edgomez 1.47 } else {
1049 : edgomez 1.51 if(dec->quarterpel) {
1050 :     uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
1051 :     uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
1052 :     b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2);
1053 :     b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2);
1054 :     } else {
1055 :     uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
1056 :     uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
1057 :     b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
1058 :     b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
1059 :     }
1060 : edgomez 1.47
1061 : edgomez 1.51 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
1062 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
1063 :     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
1064 :     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
1065 : edgomez 1.47 }
1066 :    
1067 :     start_timer();
1068 :     if(dec->quarterpel) {
1069 : edgomez 1.51 if(!direct) {
1070 : edgomez 1.47 interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1071 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1072 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1073 :     } else {
1074 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1075 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1076 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1077 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1078 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1079 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1080 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1081 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1082 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1083 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1084 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1085 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1086 : edgomez 1.47 }
1087 : edgomez 1.51 } else {
1088 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
1089 : edgomez 1.51 pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1090 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
1091 : edgomez 1.51 pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1092 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
1093 : edgomez 1.51 pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1094 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8,
1095 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1096 : edgomez 1.47 }
1097 :    
1098 :     interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
1099 : edgomez 1.51 uv_dy, stride2, 0);
1100 : edgomez 1.47 interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
1101 : edgomez 1.51 uv_dy, stride2, 0);
1102 : edgomez 1.47
1103 :    
1104 :     if(dec->quarterpel) {
1105 : edgomez 1.51 if(!direct) {
1106 : edgomez 1.47 interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1107 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1108 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1109 :     } else {
1110 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1111 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1112 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1113 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1114 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1115 :     pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1116 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1117 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1118 :     pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1119 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1120 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1121 :     pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1122 : edgomez 1.47 }
1123 : edgomez 1.51 } else {
1124 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos,
1125 : edgomez 1.51 pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1126 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1127 : edgomez 1.51 16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1128 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos,
1129 : edgomez 1.51 16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1130 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1131 : edgomez 1.51 16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1132 : edgomez 1.47 }
1133 :    
1134 :     interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos,
1135 : edgomez 1.51 b_uv_dx, b_uv_dy, stride2, 0);
1136 : edgomez 1.47 interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos,
1137 : edgomez 1.51 b_uv_dx, b_uv_dy, stride2, 0);
1138 : edgomez 1.47
1139 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1140 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1141 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos,
1142 :     stride, 1, 8);
1143 :    
1144 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1145 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1146 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1147 :     stride, 1, 8);
1148 :    
1149 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1150 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1151 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1152 :     stride, 1, 8);
1153 :    
1154 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1155 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1156 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1157 :     stride, 1, 8);
1158 :    
1159 :     interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1160 :     dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1161 :     dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos,
1162 :     stride2, 1, 8);
1163 :    
1164 :     interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1165 :     dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1166 :     dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos,
1167 :     stride2, 1, 8);
1168 :    
1169 :     stop_comp_timer();
1170 :    
1171 : edgomez 1.51 if (cbp)
1172 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
1173 : edgomez 1.47 }
1174 :    
1175 : edgomez 1.48 /* for decode B-frame dbquant */
1176 : edgomez 1.51 static __inline int32_t
1177 : edgomez 1.47 get_dbquant(Bitstream * bs)
1178 :     {
1179 : edgomez 1.51 if (!BitstreamGetBit(bs)) /* '0' */
1180 : edgomez 1.47 return (0);
1181 : edgomez 1.51 else if (!BitstreamGetBit(bs)) /* '10' */
1182 : edgomez 1.47 return (-2);
1183 : edgomez 1.51 else /* '11' */
1184 : edgomez 1.48 return (2);
1185 : edgomez 1.47 }
1186 :    
1187 : edgomez 1.48 /*
1188 : edgomez 1.51 * decode B-frame mb_type
1189 :     * bit ret_value
1190 :     * 1 0
1191 :     * 01 1
1192 :     * 001 2
1193 :     * 0001 3
1194 : edgomez 1.48 */
1195 : edgomez 1.51 static int32_t __inline
1196 : edgomez 1.47 get_mbtype(Bitstream * bs)
1197 :     {
1198 :     int32_t mb_type;
1199 :    
1200 : edgomez 1.51 for (mb_type = 0; mb_type <= 3; mb_type++)
1201 : edgomez 1.47 if (BitstreamGetBit(bs))
1202 : edgomez 1.51 return (mb_type);
1203 : edgomez 1.47
1204 : edgomez 1.51 return -1;
1205 : edgomez 1.47 }
1206 :    
1207 : edgomez 1.51 static void
1208 : edgomez 1.47 decoder_bframe(DECODER * dec,
1209 : edgomez 1.51 Bitstream * bs,
1210 :     int quant,
1211 :     int fcode_forward,
1212 :     int fcode_backward)
1213 : edgomez 1.47 {
1214 :     uint32_t x, y;
1215 :     VECTOR mv;
1216 :     const VECTOR zeromv = {0,0};
1217 : edgomez 1.51 const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp;
1218 :     int i;
1219 : edgomez 1.47
1220 :     start_timer();
1221 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
1222 : edgomez 1.51 dec->width, dec->height, dec->bs_version);
1223 : edgomez 1.47 image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
1224 : edgomez 1.51 dec->width, dec->height, dec->bs_version);
1225 : edgomez 1.47 stop_edges_timer();
1226 :    
1227 :     for (y = 0; y < dec->mb_height; y++) {
1228 : edgomez 1.48 /* Initialize Pred Motion Vector */
1229 : edgomez 1.47 dec->p_fmv = dec->p_bmv = zeromv;
1230 :     for (x = 0; x < dec->mb_width; x++) {
1231 :     MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
1232 :     MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
1233 : edgomez 1.51 const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward;
1234 :     uint32_t intra_dc_threshold; /* fake variable */
1235 :    
1236 :     if (check_resync_marker(bs, fcode_max - 1)) {
1237 :     int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant,
1238 :     &fcode_forward, &fcode_backward, &intra_dc_threshold);
1239 :     x = bound % dec->mb_width;
1240 :     y = bound / dec->mb_width;
1241 :     /* reset predicted macroblocks */
1242 :     dec->p_fmv = dec->p_bmv = zeromv;
1243 :     }
1244 : edgomez 1.47
1245 :     mv =
1246 :     mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
1247 :     mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
1248 : edgomez 1.51 mb->quant = quant;
1249 : edgomez 1.47
1250 : edgomez 1.48 /*
1251 : edgomez 1.51 * skip if the co-located P_VOP macroblock is not coded
1252 : edgomez 1.48 * if not codec in co-located S_VOP macroblock is _not_
1253 :     * automatically skipped
1254 :     */
1255 : edgomez 1.47
1256 :     if (last_mb->mode == MODE_NOT_CODED) {
1257 :     mb->cbp = 0;
1258 : edgomez 1.51 mb->mode = MODE_FORWARD;
1259 :     decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1260 : edgomez 1.47 continue;
1261 :     }
1262 :    
1263 : edgomez 1.48 if (!BitstreamGetBit(bs)) { /* modb=='0' */
1264 : edgomez 1.47 const uint8_t modb2 = BitstreamGetBit(bs);
1265 :    
1266 : edgomez 1.51 mb->mode = get_mbtype(bs);
1267 : edgomez 1.47
1268 : edgomez 1.51 if (!modb2) /* modb=='00' */
1269 : edgomez 1.47 mb->cbp = BitstreamGetBits(bs, 6);
1270 : edgomez 1.51 else
1271 : edgomez 1.47 mb->cbp = 0;
1272 : edgomez 1.51
1273 :     if (mb->mode && mb->cbp) {
1274 : edgomez 1.47 quant += get_dbquant(bs);
1275 : edgomez 1.51 if (quant > 31)
1276 : edgomez 1.47 quant = 31;
1277 : edgomez 1.51 else if (quant < 1)
1278 : edgomez 1.47 quant = 1;
1279 : edgomez 1.51 }
1280 :     mb->quant = quant;
1281 :    
1282 :     if (dec->interlacing) {
1283 :     if (mb->cbp) {
1284 :     mb->field_dct = BitstreamGetBit(bs);
1285 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1286 :     }
1287 :    
1288 :     if (mb->mode) {
1289 :     mb->field_pred = BitstreamGetBit(bs);
1290 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1291 :    
1292 :     if (mb->field_pred) {
1293 :     mb->field_for_top = BitstreamGetBit(bs);
1294 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1295 :     mb->field_for_bot = BitstreamGetBit(bs);
1296 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1297 :     }
1298 : edgomez 1.47 }
1299 :     }
1300 : edgomez 1.51
1301 : edgomez 1.47 } else {
1302 : edgomez 1.51 mb->mode = MODE_DIRECT_NONE_MV;
1303 : edgomez 1.47 mb->cbp = 0;
1304 :     }
1305 :    
1306 : edgomez 1.51 switch (mb->mode) {
1307 : edgomez 1.47 case MODE_DIRECT:
1308 : edgomez 1.51 get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y);
1309 : edgomez 1.47
1310 :     case MODE_DIRECT_NONE_MV:
1311 : edgomez 1.51 for (i = 0; i < 4; i++) {
1312 :     mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x);
1313 :     mb->b_mvs[i].x = (int32_t) ((mv.x == 0)
1314 :     ? ((TRB - TRD) * last_mb->mvs[i].x) / TRD
1315 :     : mb->mvs[i].x - last_mb->mvs[i].x);
1316 :     mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y);
1317 :     mb->b_mvs[i].y = (int32_t) ((mv.y == 0)
1318 :     ? ((TRB - TRD) * last_mb->mvs[i].y) / TRD
1319 :     : mb->mvs[i].y - last_mb->mvs[i].y);
1320 : edgomez 1.47 }
1321 : edgomez 1.51
1322 : edgomez 1.47 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1323 : edgomez 1.51 mb, x, y, bs, 1);
1324 : edgomez 1.47 break;
1325 :    
1326 :     case MODE_INTERPOLATE:
1327 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1328 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1329 :    
1330 : edgomez 1.51 get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1331 :     dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
1332 : edgomez 1.47
1333 :     decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1334 : edgomez 1.51 mb, x, y, bs, 0);
1335 : edgomez 1.47 break;
1336 :    
1337 :     case MODE_BACKWARD:
1338 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1339 : edgomez 1.47 dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1340 :    
1341 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0);
1342 : edgomez 1.47 break;
1343 :    
1344 :     case MODE_FORWARD:
1345 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1346 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1347 :    
1348 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1349 : edgomez 1.47 break;
1350 :    
1351 :     default:
1352 : edgomez 1.51 DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
1353 : edgomez 1.47 }
1354 : edgomez 1.48 } /* End of for */
1355 :     }
1356 : edgomez 1.47 }
1357 :    
1358 :     /* perform post processing if necessary, and output the image */
1359 : edgomez 1.51 void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
1360 :     xvid_dec_frame_t * frame, xvid_dec_stats_t * stats,
1361 :     int coding_type, int quant)
1362 : edgomez 1.47 {
1363 : suxen_drol 1.55 const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0;
1364 :    
1365 : edgomez 1.51 if (dec->cartoon_mode)
1366 :     frame->general &= ~XVID_FILMEFFECT;
1367 : edgomez 1.47
1368 : suxen_drol 1.55 if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0)
1369 : suxen_drol 1.52 && mbs != NULL) /* post process */
1370 : edgomez 1.47 {
1371 :     /* note: image is stored to tmp */
1372 :     image_copy(&dec->tmp, img, dec->edged_width, dec->height);
1373 : edgomez 1.51 image_postproc(&dec->postproc, &dec->tmp, dec->edged_width,
1374 :     mbs, dec->mb_width, dec->mb_height, dec->mb_width,
1375 : suxen_drol 1.55 frame->general, brightness, dec->frames, (coding_type == B_VOP));
1376 : edgomez 1.47 img = &dec->tmp;
1377 :     }
1378 :    
1379 :     image_output(img, dec->width, dec->height,
1380 : edgomez 1.51 dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
1381 :     frame->output.csp, dec->interlacing);
1382 :    
1383 :     if (stats) {
1384 :     stats->type = coding2type(coding_type);
1385 :     stats->data.vop.time_base = (int)dec->time_base;
1386 :     stats->data.vop.time_increment = 0; /* XXX: todo */
1387 :     stats->data.vop.qscale_stride = dec->mb_width;
1388 :     stats->data.vop.qscale = dec->qscale;
1389 :     if (stats->data.vop.qscale != NULL && mbs != NULL) {
1390 :     int i;
1391 :     for (i = 0; i < dec->mb_width*dec->mb_height; i++)
1392 :     stats->data.vop.qscale[i] = mbs[i].quant;
1393 :     } else
1394 :     stats->data.vop.qscale = NULL;
1395 :     }
1396 : edgomez 1.47 }
1397 :    
1398 :    
1399 :     int
1400 :     decoder_decode(DECODER * dec,
1401 : edgomez 1.51 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
1402 : edgomez 1.47 {
1403 :    
1404 :     Bitstream bs;
1405 :     uint32_t rounding;
1406 :     uint32_t reduced_resolution;
1407 : edgomez 1.51 uint32_t quant = 2;
1408 : edgomez 1.47 uint32_t fcode_forward;
1409 :     uint32_t fcode_backward;
1410 :     uint32_t intra_dc_threshold;
1411 :     WARPPOINTS gmc_warp;
1412 : edgomez 1.51 int coding_type;
1413 :     int success, output, seen_something;
1414 :    
1415 :     if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */
1416 :     return XVID_ERR_VERSION;
1417 : edgomez 1.47
1418 :     start_global_timer();
1419 :    
1420 : edgomez 1.51 dec->low_delay_default = (frame->general & XVID_LOWDELAY);
1421 :     if ((frame->general & XVID_DISCONTINUITY))
1422 : edgomez 1.47 dec->frames = 0;
1423 : edgomez 1.51 dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
1424 : edgomez 1.47
1425 : edgomez 1.51 if (frame->length < 0) { /* decoder flush */
1426 :     int ret;
1427 :     /* if not decoding "low_delay/packed", and this isn't low_delay and
1428 :     we have a reference frame, then outout the reference frame */
1429 :     if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
1430 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1431 :     dec->frames = 0;
1432 :     ret = 0;
1433 :     } else {
1434 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1435 :     ret = XVID_ERR_END;
1436 : edgomez 1.47 }
1437 :    
1438 :     emms();
1439 :     stop_global_timer();
1440 : edgomez 1.51 return ret;
1441 : edgomez 1.47 }
1442 : edgomez 1.19
1443 : Isibaar 1.1 BitstreamInit(&bs, frame->bitstream, frame->length);
1444 :    
1445 : edgomez 1.48 /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
1446 : edgomez 1.47 if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
1447 :     {
1448 :     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
1449 : edgomez 1.51 (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
1450 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1451 : edgomez 1.47 emms();
1452 : edgomez 1.51 return 1; /* one byte consumed */
1453 : edgomez 1.47 }
1454 :    
1455 : edgomez 1.51 success = 0;
1456 :     output = 0;
1457 :     seen_something = 0;
1458 :    
1459 : edgomez 1.47 repeat:
1460 :    
1461 : edgomez 1.51 coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution,
1462 : edgomez 1.47 &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
1463 :    
1464 : edgomez 1.51 DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n",
1465 :     coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
1466 : Isibaar 1.1
1467 : edgomez 1.51 if (coding_type == -1) { /* nothing */
1468 : edgomez 1.47 if (success) goto done;
1469 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1470 : edgomez 1.47 emms();
1471 : edgomez 1.51 return BitstreamPos(&bs)/8;
1472 : Isibaar 1.1 }
1473 :    
1474 : edgomez 1.51 if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
1475 :    
1476 :     if (coding_type == -3)
1477 : edgomez 1.47 decoder_resize(dec);
1478 : edgomez 1.51
1479 :     if (stats) {
1480 :     stats->type = XVID_TYPE_VOL;
1481 : edgomez 1.47 stats->data.vol.general = 0;
1482 : edgomez 1.51 /*XXX: if (dec->interlacing)
1483 :     stats->data.vol.general |= ++INTERLACING; */
1484 : edgomez 1.47 stats->data.vol.width = dec->width;
1485 :     stats->data.vol.height = dec->height;
1486 : edgomez 1.51 stats->data.vol.par = dec->aspect_ratio;
1487 : edgomez 1.47 stats->data.vol.par_width = dec->par_width;
1488 :     stats->data.vol.par_height = dec->par_height;
1489 :     emms();
1490 : edgomez 1.51 return BitstreamPos(&bs)/8; /* number of bytes consumed */
1491 : edgomez 1.47 }
1492 :     goto repeat;
1493 : edgomez 1.51 }
1494 : edgomez 1.47
1495 : 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 */
1496 : edgomez 1.47
1497 :     /* packed_mode: special-N_VOP treament */
1498 : edgomez 1.51 if (dec->packed_mode && coding_type == N_VOP) {
1499 :     if (dec->low_delay_default && dec->frames > 0) {
1500 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1501 : edgomez 1.47 output = 1;
1502 :     }
1503 :     /* ignore otherwise */
1504 : edgomez 1.51 } else if (coding_type != B_VOP) {
1505 :     switch(coding_type) {
1506 : edgomez 1.47 case I_VOP :
1507 :     decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold);
1508 :     break;
1509 :     case P_VOP :
1510 : edgomez 1.51 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1511 : edgomez 1.47 fcode_forward, intra_dc_threshold, NULL);
1512 :     break;
1513 :     case S_VOP :
1514 : edgomez 1.51 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1515 : edgomez 1.47 fcode_forward, intra_dc_threshold, &gmc_warp);
1516 :     break;
1517 :     case N_VOP :
1518 : edgomez 1.51 /* XXX: not_coded vops are not used for forward prediction */
1519 :     /* we should not swap(last_mbs,mbs) */
1520 : edgomez 1.47 image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
1521 : edgomez 1.51 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */
1522 : edgomez 1.47 break;
1523 :     }
1524 :    
1525 : edgomez 1.51 if (reduced_resolution) {
1526 : edgomez 1.47 image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs,
1527 :     (dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width,
1528 : edgomez 1.51 16, 0);
1529 : edgomez 1.47 }
1530 : Isibaar 1.1
1531 : edgomez 1.47 /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
1532 : edgomez 1.51 if (!(dec->low_delay_default && dec->packed_mode)) {
1533 :     if (dec->low_delay) {
1534 :     decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1535 : edgomez 1.47 output = 1;
1536 : edgomez 1.51 } else if (dec->frames > 0) { /* is the reference frame valid? */
1537 : edgomez 1.47 /* output the reference frame */
1538 : edgomez 1.51 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1539 : edgomez 1.47 output = 1;
1540 :     }
1541 :     }
1542 : Isibaar 1.18
1543 : chenm001 1.14 image_swap(&dec->refn[0], &dec->refn[1]);
1544 :     image_swap(&dec->cur, &dec->refn[0]);
1545 : edgomez 1.51 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
1546 : edgomez 1.47 dec->last_reduced_resolution = reduced_resolution;
1547 : edgomez 1.51 dec->last_coding_type = coding_type;
1548 : edgomez 1.47
1549 :     dec->frames++;
1550 :     seen_something = 1;
1551 : chenm001 1.31
1552 : edgomez 1.51 } else { /* B_VOP */
1553 : edgomez 1.47
1554 : edgomez 1.51 if (dec->low_delay) {
1555 :     DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
1556 : edgomez 1.47 dec->low_delay = 1;
1557 :     }
1558 :    
1559 : edgomez 1.51 if (dec->frames < 2) {
1560 : edgomez 1.47 /* attemping to decode a bvop without atleast 2 reference frames */
1561 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1562 :     "broken b-frame, mising ref frames");
1563 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1564 :     } else if (dec->time_pp <= dec->time_bp) {
1565 :     /* this occurs when dx50_bvop_compatibility==0 sequences are
1566 : edgomez 1.47 decoded in vfw. */
1567 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1568 :     "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
1569 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1570 :     } else {
1571 : edgomez 1.47 decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
1572 : edgomez 1.51 decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1573 : edgomez 1.47 }
1574 :    
1575 :     output = 1;
1576 :     dec->frames++;
1577 : chenm001 1.14 }
1578 :    
1579 : edgomez 1.51 #if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */
1580 :     BitstreamByteAlign(&bs);
1581 :     #endif
1582 : edgomez 1.47
1583 :     /* low_delay_default mode: repeat in packed_mode */
1584 : edgomez 1.51 if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
1585 : edgomez 1.47 success = 1;
1586 :     goto repeat;
1587 :     }
1588 :    
1589 :     done :
1590 :    
1591 :     /* low_delay_default mode: if we've gotten here without outputting anything,
1592 :     then output the recently decoded frame, or print an error message */
1593 : edgomez 1.51 if (dec->low_delay_default && output == 0) {
1594 :     if (dec->packed_mode && seen_something) {
1595 : edgomez 1.47 /* output the recently decoded frame */
1596 : edgomez 1.51 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1597 :     } else {
1598 : edgomez 1.47 image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
1599 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1600 :     "warning: nothing to output");
1601 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64,
1602 :     "bframe decoder lag");
1603 :    
1604 : edgomez 1.51 decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant);
1605 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1606 : edgomez 1.47 }
1607 :     }
1608 : Isibaar 1.50
1609 : Isibaar 1.1 emms();
1610 :     stop_global_timer();
1611 :    
1612 : edgomez 1.51 return (BitstreamPos(&bs) + 7) / 8; /* number of bytes consumed */
1613 : Isibaar 1.1 }
No admin address has been configured
 ViewVC Help
Powered by ViewVC 1.0.4