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

Annotation of /xvidcore/src/decoder.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.63 - (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 : edgomez 1.61 * $Id$
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 : edgomez 1.62 #include "image/qpel.h"
52 : Isibaar 1.1
53 :     #include "bitstream/mbcoding.h"
54 :     #include "prediction/mbprediction.h"
55 :     #include "utils/timer.h"
56 :     #include "utils/emms.h"
57 : edgomez 1.47 #include "motion/motion.h"
58 : edgomez 1.51 #include "motion/gmc.h"
59 : Isibaar 1.1
60 :     #include "image/image.h"
61 :     #include "image/colorspace.h"
62 : edgomez 1.51 #include "image/postprocessing.h"
63 : Isibaar 1.3 #include "utils/mem_align.h"
64 : Isibaar 1.1
65 : edgomez 1.62 #ifdef ARCH_IS_IA32
66 :     #define interpolate16x16_quarterpel new_interpolate16x16_quarterpel
67 :     #define interpolate8x8_quarterpel new_interpolate8x8_quarterpel
68 :     #endif
69 :    
70 : edgomez 1.51 static int
71 : edgomez 1.47 decoder_resize(DECODER * dec)
72 : Isibaar 1.1 {
73 : edgomez 1.47 /* free existing */
74 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
75 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
76 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
77 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
78 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
79 :    
80 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
81 : Isibaar 1.1
82 : edgomez 1.51 if (dec->last_mbs)
83 : edgomez 1.47 xvid_free(dec->last_mbs);
84 :     if (dec->mbs)
85 :     xvid_free(dec->mbs);
86 : edgomez 1.51 if (dec->qscale)
87 :     xvid_free(dec->qscale);
88 : Isibaar 1.1
89 : edgomez 1.47 /* realloc */
90 : Isibaar 1.1 dec->mb_width = (dec->width + 15) / 16;
91 :     dec->mb_height = (dec->height + 15) / 16;
92 :    
93 :     dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE;
94 :     dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE;
95 : edgomez 1.19
96 :     if (image_create(&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 :    
101 : edgomez 1.19 if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) {
102 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
103 : Isibaar 1.3 xvid_free(dec);
104 : Isibaar 1.1 return XVID_ERR_MEMORY;
105 :     }
106 : edgomez 1.47
107 : edgomez 1.48 /* Support B-frame to reference last 2 frame */
108 : edgomez 1.19 if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) {
109 : chenm001 1.11 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
110 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
111 :     xvid_free(dec);
112 :     return XVID_ERR_MEMORY;
113 :     }
114 : edgomez 1.47 if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) {
115 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
116 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
117 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
118 :     xvid_free(dec);
119 :     return XVID_ERR_MEMORY;
120 :     }
121 :    
122 :     if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) {
123 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
124 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
125 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
126 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
127 :     xvid_free(dec);
128 :     return XVID_ERR_MEMORY;
129 :     }
130 :    
131 :     if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) {
132 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
133 : chenm001 1.14 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
134 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
135 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
136 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
137 : chenm001 1.14 xvid_free(dec);
138 :     return XVID_ERR_MEMORY;
139 :     }
140 : Isibaar 1.1
141 : edgomez 1.19 dec->mbs =
142 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
143 :     CACHE_LINE);
144 :     if (dec->mbs == NULL) {
145 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
146 : chenm001 1.14 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
147 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
148 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
149 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
150 : chenm001 1.14 xvid_free(dec);
151 :     return XVID_ERR_MEMORY;
152 :     }
153 : Isibaar 1.20 memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
154 :    
155 : edgomez 1.48 /* For skip MB flag */
156 : edgomez 1.19 dec->last_mbs =
157 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
158 :     CACHE_LINE);
159 :     if (dec->last_mbs == NULL) {
160 : chenm001 1.14 xvid_free(dec->mbs);
161 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
162 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
163 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
164 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
165 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
166 : Isibaar 1.3 xvid_free(dec);
167 : Isibaar 1.1 return XVID_ERR_MEMORY;
168 :     }
169 : Isibaar 1.20
170 :     memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
171 : Isibaar 1.1
172 : edgomez 1.51 /* nothing happens if that fails */
173 :     dec->qscale =
174 :     xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE);
175 :    
176 :     if (dec->qscale)
177 :     memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height);
178 :    
179 :     return 0;
180 : edgomez 1.47 }
181 :    
182 :    
183 :     int
184 : edgomez 1.51 decoder_create(xvid_dec_create_t * create)
185 : edgomez 1.47 {
186 :     DECODER *dec;
187 :    
188 : edgomez 1.51 if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */
189 :     return XVID_ERR_VERSION;
190 :    
191 : edgomez 1.47 dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
192 :     if (dec == NULL) {
193 :     return XVID_ERR_MEMORY;
194 :     }
195 : edgomez 1.51
196 : edgomez 1.47 memset(dec, 0, sizeof(DECODER));
197 :    
198 : edgomez 1.51 dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE);
199 :     if (dec->mpeg_quant_matrices == NULL) {
200 :     xvid_free(dec);
201 :     return XVID_ERR_MEMORY;
202 :     }
203 :    
204 :     create->handle = dec;
205 : edgomez 1.47
206 : edgomez 1.51 dec->width = create->width;
207 :     dec->height = create->height;
208 : edgomez 1.47
209 :     image_null(&dec->cur);
210 :     image_null(&dec->refn[0]);
211 :     image_null(&dec->refn[1]);
212 :     image_null(&dec->tmp);
213 :     image_null(&dec->qtmp);
214 :    
215 : edgomez 1.48 /* image based GMC */
216 : edgomez 1.47 image_null(&dec->gmc);
217 :    
218 :     dec->mbs = NULL;
219 :     dec->last_mbs = NULL;
220 : edgomez 1.51 dec->qscale = NULL;
221 : edgomez 1.47
222 : Isibaar 1.1 init_timer();
223 : edgomez 1.51 init_postproc(&dec->postproc);
224 :     init_mpeg_matrix(dec->mpeg_quant_matrices);
225 : Isibaar 1.1
226 : edgomez 1.48 /* For B-frame support (used to save reference frame's time */
227 : edgomez 1.47 dec->frames = 0;
228 : chenm001 1.14 dec->time = dec->time_base = dec->last_time_base = 0;
229 : edgomez 1.47 dec->low_delay = 0;
230 :     dec->packed_mode = 0;
231 : syskin 1.56 dec->time_inc_resolution = 1; /* until VOL header says otherwise */
232 : edgomez 1.47
233 :     dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
234 : edgomez 1.19
235 : edgomez 1.47 if (dec->fixed_dimensions)
236 :     return decoder_resize(dec);
237 :     else
238 : edgomez 1.51 return 0;
239 : Isibaar 1.1 }
240 :    
241 :    
242 : edgomez 1.19 int
243 :     decoder_destroy(DECODER * dec)
244 : Isibaar 1.1 {
245 : chenm001 1.14 xvid_free(dec->last_mbs);
246 : Isibaar 1.3 xvid_free(dec->mbs);
247 : edgomez 1.51 xvid_free(dec->qscale);
248 : edgomez 1.47
249 : edgomez 1.48 /* image based GMC */
250 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
251 : edgomez 1.47
252 : chenm001 1.11 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
253 : chenm001 1.14 image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
254 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
255 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
256 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
257 : edgomez 1.51 xvid_free(dec->mpeg_quant_matrices);
258 : Isibaar 1.3 xvid_free(dec);
259 : Isibaar 1.1
260 :     write_timer();
261 : edgomez 1.51 return 0;
262 : Isibaar 1.1 }
263 :    
264 : edgomez 1.19 static const int32_t dquant_table[4] = {
265 : Isibaar 1.1 -1, -2, 1, 2
266 :     };
267 :    
268 : edgomez 1.48 /* decode an intra macroblock */
269 : edgomez 1.51 static void
270 : edgomez 1.19 decoder_mbintra(DECODER * dec,
271 :     MACROBLOCK * pMB,
272 :     const uint32_t x_pos,
273 :     const uint32_t y_pos,
274 :     const uint32_t acpred_flag,
275 :     const uint32_t cbp,
276 :     Bitstream * bs,
277 :     const uint32_t quant,
278 : chenm001 1.26 const uint32_t intra_dc_threshold,
279 : edgomez 1.47 const unsigned int bound,
280 :     const int reduced_resolution)
281 : Isibaar 1.1 {
282 : edgomez 1.7
283 :     DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
284 : edgomez 1.19 DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
285 : edgomez 1.7
286 : h 1.8 uint32_t stride = dec->edged_width;
287 :     uint32_t stride2 = stride / 2;
288 :     uint32_t next_block = stride * 8;
289 : h 1.5 uint32_t i;
290 :     uint32_t iQuant = pMB->quant;
291 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
292 :    
293 : edgomez 1.47 if (reduced_resolution) {
294 :     pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5);
295 :     pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4);
296 :     pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4);
297 :     }else{
298 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
299 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
300 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
301 :     }
302 : edgomez 1.7
303 : edgomez 1.48 memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */
304 : h 1.5
305 : edgomez 1.19 for (i = 0; i < 6; i++) {
306 : h 1.5 uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);
307 : Isibaar 1.1 int16_t predictors[8];
308 :     int start_coeff;
309 :    
310 :     start_timer();
311 : edgomez 1.19 predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64],
312 : edgomez 1.59 iQuant, iDcScaler, predictors, bound);
313 : edgomez 1.19 if (!acpred_flag) {
314 : h 1.5 pMB->acpred_directions[i] = 0;
315 : Isibaar 1.1 }
316 :     stop_prediction_timer();
317 :    
318 : edgomez 1.19 if (quant < intra_dc_threshold) {
319 : Isibaar 1.1 int dc_size;
320 :     int dc_dif;
321 :    
322 : edgomez 1.19 dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs);
323 :     dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0;
324 : Isibaar 1.1
325 : edgomez 1.19 if (dc_size > 8) {
326 : edgomez 1.48 BitstreamSkip(bs, 1); /* marker */
327 : Isibaar 1.1 }
328 : edgomez 1.19
329 :     block[i * 64 + 0] = dc_dif;
330 : Isibaar 1.1 start_coeff = 1;
331 : chenm001 1.26
332 : edgomez 1.51 DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif);
333 : edgomez 1.19 } else {
334 : Isibaar 1.1 start_coeff = 0;
335 :     }
336 :    
337 :     start_timer();
338 : edgomez 1.48 if (cbp & (1 << (5 - i))) /* coded */
339 : Isibaar 1.1 {
340 : edgomez 1.47 int direction = dec->alternate_vertical_scan ?
341 :     2 : pMB->acpred_directions[i];
342 :    
343 :     get_intra_block(bs, &block[i * 64], direction, start_coeff);
344 : Isibaar 1.1 }
345 :     stop_coding_timer();
346 :    
347 :     start_timer();
348 : edgomez 1.59 add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version);
349 : Isibaar 1.1 stop_prediction_timer();
350 :    
351 :     start_timer();
352 : edgomez 1.19 if (dec->quant_type == 0) {
353 : edgomez 1.51 dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
354 : edgomez 1.19 } else {
355 : edgomez 1.51 dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
356 : Isibaar 1.1 }
357 :     stop_iquant_timer();
358 :    
359 :     start_timer();
360 : edgomez 1.19 idct(&data[i * 64]);
361 : Isibaar 1.1 stop_idct_timer();
362 : edgomez 1.47
363 : h 1.5 }
364 : Isibaar 1.1
365 : edgomez 1.19 if (dec->interlacing && pMB->field_dct) {
366 : h 1.8 next_block = stride;
367 :     stride *= 2;
368 : Isibaar 1.1 }
369 : h 1.5
370 :     start_timer();
371 : edgomez 1.47
372 :     if (reduced_resolution)
373 :     {
374 :     next_block*=2;
375 :     copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
376 :     copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
377 :     copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
378 :     copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
379 :     copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
380 :     copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
381 :     }else{
382 :     transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
383 :     transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
384 :     transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
385 :     transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
386 :     transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
387 :     transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
388 :     }
389 : h 1.5 stop_transfer_timer();
390 : Isibaar 1.1 }
391 :    
392 : edgomez 1.51 static void
393 :     decoder_mb_decode(DECODER * dec,
394 :     const uint32_t cbp,
395 :     Bitstream * bs,
396 :     uint8_t * pY_Cur,
397 :     uint8_t * pU_Cur,
398 :     uint8_t * pV_Cur,
399 :     const int reduced_resolution,
400 :     const MACROBLOCK * pMB)
401 :     {
402 :     DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
403 :    
404 :     int stride = dec->edged_width;
405 :     int next_block = stride * (reduced_resolution ? 16 : 8);
406 :     const int stride2 = stride/2;
407 :     int i;
408 :     const uint32_t iQuant = pMB->quant;
409 :     const int direction = dec->alternate_vertical_scan ? 2 : 0;
410 : edgomez 1.61 typedef void (*get_inter_block_function_t)(
411 :     Bitstream * bs,
412 :     int16_t * block,
413 :     int direction,
414 :     const int quant,
415 :     const uint16_t *matrix);
416 :    
417 :     const get_inter_block_function_t get_inter_block = (dec->quant_type == 0)
418 :     ? get_inter_block_h263
419 :     : get_inter_block_mpeg;
420 :    
421 :     memset(&data[0], 0, 6*64*sizeof(int16_t)); /* clear */
422 : edgomez 1.51
423 :     for (i = 0; i < 6; i++) {
424 :    
425 :     if (cbp & (1 << (5 - i))) { /* coded */
426 :    
427 :    
428 : edgomez 1.61 /* Decode coeffs and dequantize on the fly */
429 : edgomez 1.51 start_timer();
430 : edgomez 1.61 get_inter_block(bs, &data[i*64], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices));
431 : edgomez 1.51 stop_coding_timer();
432 :    
433 :     start_timer();
434 :     idct(&data[i * 64]);
435 :     stop_idct_timer();
436 :     }
437 :     }
438 : Isibaar 1.1
439 : edgomez 1.51 if (dec->interlacing && pMB->field_dct) {
440 :     next_block = stride;
441 :     stride *= 2;
442 :     }
443 : Isibaar 1.1
444 : edgomez 1.51 start_timer();
445 :     if (reduced_resolution) {
446 :     if (cbp & 32)
447 :     add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
448 :     if (cbp & 16)
449 :     add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
450 :     if (cbp & 8)
451 :     add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
452 :     if (cbp & 4)
453 :     add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
454 :     if (cbp & 2)
455 :     add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
456 :     if (cbp & 1)
457 :     add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
458 :     } else {
459 :     if (cbp & 32)
460 :     transfer_16to8add(pY_Cur, &data[0 * 64], stride);
461 :     if (cbp & 16)
462 :     transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride);
463 :     if (cbp & 8)
464 :     transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride);
465 :     if (cbp & 4)
466 :     transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride);
467 :     if (cbp & 2)
468 :     transfer_16to8add(pU_Cur, &data[4 * 64], stride2);
469 :     if (cbp & 1)
470 :     transfer_16to8add(pV_Cur, &data[5 * 64], stride2);
471 :     }
472 :     stop_transfer_timer();
473 :     }
474 : Isibaar 1.1
475 : edgomez 1.48 /* decode an inter macroblock */
476 : edgomez 1.51 static void
477 : edgomez 1.19 decoder_mbinter(DECODER * dec,
478 :     const MACROBLOCK * pMB,
479 :     const uint32_t x_pos,
480 :     const uint32_t y_pos,
481 :     const uint32_t cbp,
482 :     Bitstream * bs,
483 : edgomez 1.47 const uint32_t rounding,
484 : edgomez 1.51 const int reduced_resolution,
485 :     const int ref)
486 : Isibaar 1.1 {
487 : h 1.8 uint32_t stride = dec->edged_width;
488 :     uint32_t stride2 = stride / 2;
489 : edgomez 1.7 uint32_t i;
490 : edgomez 1.51
491 : h 1.5 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
492 : edgomez 1.47
493 : Isibaar 1.1 int uv_dx, uv_dy;
494 : edgomez 1.47 VECTOR mv[4]; /* local copy of mvs */
495 :    
496 :     if (reduced_resolution) {
497 :     pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5);
498 :     pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4);
499 :     pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4);
500 :     for (i = 0; i < 4; i++) {
501 :     mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x);
502 :     mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y);
503 :     }
504 :     } else {
505 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
506 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
507 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
508 :     for (i = 0; i < 4; i++)
509 :     mv[i] = pMB->mvs[i];
510 :     }
511 :    
512 : syskin 1.54 for (i = 0; i < 4; i++) {
513 :     /* clip to valid range */
514 :     int border = (int)(dec->mb_width - x_pos) << (5 + dec->quarterpel);
515 :     if (mv[i].x > border) {
516 :     DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", mv[i].x, border, x_pos, y_pos);
517 :     mv[i].x = border;
518 :     } else {
519 :     border = (-(int)x_pos-1) << (5 + dec->quarterpel);
520 :     if (mv[i].x < border) {
521 :     DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", mv[i].x, border, x_pos, y_pos);
522 :     mv[i].x = border;
523 :     }
524 :     }
525 :    
526 :     border = (int)(dec->mb_height - y_pos) << (5 + dec->quarterpel);
527 :     if (mv[i].y > border) {
528 :     DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", mv[i].y, border, x_pos, y_pos);
529 :     mv[i].y = border;
530 :     } else {
531 :     border = (-(int)y_pos-1) << (5 + dec->quarterpel);
532 :     if (mv[i].y < border) {
533 :     DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", mv[i].y, border, x_pos, y_pos);
534 :     mv[i].y = border;
535 :     }
536 :     }
537 :     }
538 :    
539 : edgomez 1.51 start_timer();
540 :    
541 :     if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
542 :    
543 :     uv_dx = mv[0].x;
544 :     uv_dy = mv[0].y;
545 :     if (dec->quarterpel) {
546 :     uv_dx /= 2;
547 :     uv_dy /= 2;
548 :     }
549 : edgomez 1.47 uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
550 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
551 :    
552 :     if (reduced_resolution)
553 :     interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
554 : edgomez 1.51 mv[0].x, mv[0].y, stride, rounding);
555 :     else if (dec->quarterpel)
556 :     interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
557 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
558 :     mv[0].x, mv[0].y, stride, rounding);
559 : edgomez 1.47 else
560 : edgomez 1.51 interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
561 :     mv[0].x, mv[0].y, stride, rounding);
562 : Isibaar 1.1
563 : edgomez 1.47 } else { /* MODE_INTER4V */
564 :    
565 : edgomez 1.51 if(dec->quarterpel) {
566 :     uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
567 :     uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
568 :     } else {
569 :     uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
570 :     uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
571 :     }
572 : edgomez 1.47
573 : edgomez 1.51 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
574 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
575 : edgomez 1.47
576 : edgomez 1.51 if (reduced_resolution) {
577 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
578 : edgomez 1.51 mv[0].x, mv[0].y, stride, rounding);
579 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos,
580 : edgomez 1.51 mv[1].x, mv[1].y, stride, rounding);
581 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16,
582 : edgomez 1.51 mv[2].x, mv[2].y, stride, rounding);
583 :     interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16,
584 :     mv[3].x, mv[3].y, stride, rounding);
585 : edgomez 1.47 interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos,
586 : edgomez 1.51 uv_dx, uv_dy, stride2, rounding);
587 : edgomez 1.47 interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos,
588 : edgomez 1.51 uv_dx, uv_dy, stride2, rounding);
589 : edgomez 1.47
590 : edgomez 1.51 } else if (dec->quarterpel) {
591 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
592 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
593 :     mv[0].x, mv[0].y, stride, rounding);
594 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
595 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
596 :     mv[1].x, mv[1].y, stride, rounding);
597 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
598 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
599 :     mv[2].x, mv[2].y, stride, rounding);
600 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
601 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
602 :     mv[3].x, mv[3].y, stride, rounding);
603 :     } else {
604 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
605 :     mv[0].x, mv[0].y, stride, rounding);
606 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
607 :     mv[1].x, mv[1].y, stride, rounding);
608 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
609 :     mv[2].x, mv[2].y, stride, rounding);
610 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
611 :     mv[3].x, mv[3].y, stride, rounding);
612 : edgomez 1.47 }
613 :     }
614 :    
615 : edgomez 1.51 /* chroma */
616 :     if (reduced_resolution) {
617 :     interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos,
618 :     uv_dx, uv_dy, stride2, rounding);
619 :     interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos,
620 :     uv_dx, uv_dy, stride2, rounding);
621 :     } else {
622 :     interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
623 :     uv_dx, uv_dy, stride2, rounding);
624 :     interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
625 :     uv_dx, uv_dy, stride2, rounding);
626 : edgomez 1.47 }
627 : Isibaar 1.1
628 : edgomez 1.51 stop_comp_timer();
629 : edgomez 1.47
630 : edgomez 1.51 if (cbp)
631 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur,
632 :     reduced_resolution, pMB);
633 : edgomez 1.47 }
634 :    
635 :     static void
636 :     decoder_mbgmc(DECODER * dec,
637 :     MACROBLOCK * const pMB,
638 :     const uint32_t x_pos,
639 :     const uint32_t y_pos,
640 :     const uint32_t fcode,
641 :     const uint32_t cbp,
642 :     Bitstream * bs,
643 : edgomez 1.51 const uint32_t rounding)
644 : edgomez 1.47 {
645 :     const uint32_t stride = dec->edged_width;
646 :     const uint32_t stride2 = stride / 2;
647 : edgomez 1.51
648 : edgomez 1.47 uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
649 :     uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
650 :     uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
651 :    
652 : edgomez 1.51 NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
653 :    
654 : edgomez 1.47 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
655 : Isibaar 1.1
656 :     start_timer();
657 : edgomez 1.51
658 : edgomez 1.47 /* this is where the calculations are done */
659 :    
660 : edgomez 1.51 gmc_data->predict_16x16(gmc_data,
661 :     dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
662 :     stride, stride, x_pos, y_pos, rounding);
663 :    
664 :     gmc_data->predict_8x8(gmc_data,
665 :     dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
666 :     dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
667 :     stride2, stride2, x_pos, y_pos, rounding);
668 : edgomez 1.47
669 : edgomez 1.51 gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
670 : Isibaar 1.1
671 : edgomez 1.51 pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
672 :     pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
673 : edgomez 1.47
674 : edgomez 1.51 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
675 : Isibaar 1.1
676 : edgomez 1.51 stop_transfer_timer();
677 : Isibaar 1.1
678 : edgomez 1.51 if (cbp)
679 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
680 : Isibaar 1.1
681 :     }
682 :    
683 :    
684 : edgomez 1.51 static void
685 : edgomez 1.19 decoder_iframe(DECODER * dec,
686 : edgomez 1.51 Bitstream * bs,
687 :     int reduced_resolution,
688 :     int quant,
689 :     int intra_dc_threshold)
690 : Isibaar 1.1 {
691 : chenm001 1.26 uint32_t bound;
692 :     uint32_t x, y;
693 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
694 :     uint32_t mb_height = dec->mb_height;
695 : edgomez 1.51
696 :     if (reduced_resolution) {
697 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
698 :     mb_height = (dec->height + 31) / 32;
699 :     }
700 : chenm001 1.25
701 : chenm001 1.26 bound = 0;
702 : edgomez 1.7
703 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
704 :     for (x = 0; x < mb_width; x++) {
705 : chenm001 1.26 MACROBLOCK *mb;
706 : Isibaar 1.1 uint32_t mcbpc;
707 :     uint32_t cbpc;
708 :     uint32_t acpred_flag;
709 :     uint32_t cbpy;
710 :     uint32_t cbp;
711 :    
712 : chenm001 1.26 while (BitstreamShowBits(bs, 9) == 1)
713 :     BitstreamSkip(bs, 9);
714 :    
715 :     if (check_resync_marker(bs, 0))
716 :     {
717 : edgomez 1.51 bound = read_video_packet_header(bs, dec, 0,
718 : edgomez 1.47 &quant, NULL, NULL, &intra_dc_threshold);
719 :     x = bound % mb_width;
720 :     y = bound / mb_width;
721 : chenm001 1.26 }
722 :     mb = &dec->mbs[y * dec->mb_width + x];
723 :    
724 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
725 : chenm001 1.26
726 : Isibaar 1.1 mcbpc = get_mcbpc_intra(bs);
727 :     mb->mode = mcbpc & 7;
728 :     cbpc = (mcbpc >> 4);
729 :    
730 :     acpred_flag = BitstreamGetBit(bs);
731 :    
732 :     cbpy = get_cbpy(bs, 1);
733 :     cbp = (cbpy << 2) | cbpc;
734 :    
735 : edgomez 1.19 if (mb->mode == MODE_INTRA_Q) {
736 :     quant += dquant_table[BitstreamGetBits(bs, 2)];
737 :     if (quant > 31) {
738 : Isibaar 1.1 quant = 31;
739 : edgomez 1.19 } else if (quant < 1) {
740 : Isibaar 1.1 quant = 1;
741 :     }
742 :     }
743 :     mb->quant = quant;
744 : chenm001 1.30 mb->mvs[0].x = mb->mvs[0].y =
745 :     mb->mvs[1].x = mb->mvs[1].y =
746 :     mb->mvs[2].x = mb->mvs[2].y =
747 :     mb->mvs[3].x = mb->mvs[3].y =0;
748 : h 1.5
749 : edgomez 1.19 if (dec->interlacing) {
750 : h 1.5 mb->field_dct = BitstreamGetBit(bs);
751 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
752 : h 1.5 }
753 : Isibaar 1.1
754 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
755 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
756 :    
757 : Isibaar 1.1 }
758 : albeu 1.32 if(dec->out_frm)
759 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
760 : Isibaar 1.1 }
761 : edgomez 1.7
762 : Isibaar 1.1 }
763 :    
764 :    
765 : edgomez 1.51 static void
766 : edgomez 1.19 get_motion_vector(DECODER * dec,
767 : edgomez 1.51 Bitstream * bs,
768 :     int x,
769 :     int y,
770 :     int k,
771 :     VECTOR * ret_mv,
772 :     int fcode,
773 :     const int bound)
774 : Isibaar 1.1 {
775 : edgomez 1.7
776 : edgomez 1.51 const int scale_fac = 1 << (fcode - 1);
777 :     const int high = (32 * scale_fac) - 1;
778 :     const int low = ((-32) * scale_fac);
779 :     const int range = (64 * scale_fac);
780 : Isibaar 1.1
781 : edgomez 1.51 const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
782 : edgomez 1.47 VECTOR mv;
783 : Isibaar 1.1
784 : edgomez 1.47 mv.x = get_mv(bs, fcode);
785 :     mv.y = get_mv(bs, fcode);
786 : edgomez 1.19
787 : 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);
788 : chenm001 1.26
789 : edgomez 1.47 mv.x += pmv.x;
790 :     mv.y += pmv.y;
791 : Isibaar 1.1
792 : edgomez 1.47 if (mv.x < low) {
793 :     mv.x += range;
794 :     } else if (mv.x > high) {
795 :     mv.x -= range;
796 : Isibaar 1.1 }
797 :    
798 : edgomez 1.47 if (mv.y < low) {
799 :     mv.y += range;
800 :     } else if (mv.y > high) {
801 :     mv.y -= range;
802 : Isibaar 1.1 }
803 :    
804 : edgomez 1.51 ret_mv->x = mv.x;
805 :     ret_mv->y = mv.y;
806 :     }
807 : Isibaar 1.1
808 : edgomez 1.47 /* for P_VOP set gmc_warp to NULL */
809 : edgomez 1.51 static void
810 : edgomez 1.19 decoder_pframe(DECODER * dec,
811 : edgomez 1.51 Bitstream * bs,
812 :     int rounding,
813 :     int reduced_resolution,
814 :     int quant,
815 :     int fcode,
816 :     int intra_dc_threshold,
817 :     const WARPPOINTS *const gmc_warp)
818 : Isibaar 1.1 {
819 :     uint32_t x, y;
820 : chenm001 1.26 uint32_t bound;
821 : albeu 1.32 int cp_mb, st_mb;
822 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
823 :     uint32_t mb_height = dec->mb_height;
824 : edgomez 1.51
825 :     if (reduced_resolution) {
826 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
827 :     mb_height = (dec->height + 31) / 32;
828 :     }
829 : Isibaar 1.1
830 : edgomez 1.57 if (!dec->is_edged[0]) {
831 :     start_timer();
832 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
833 :     dec->width, dec->height, dec->bs_version);
834 :     dec->is_edged[0] = 1;
835 :     stop_edges_timer();
836 :     }
837 : Isibaar 1.1
838 : edgomez 1.51 if (gmc_warp) {
839 :     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
840 :     generate_GMCparameters( dec->sprite_warping_points,
841 :     dec->sprite_warping_accuracy, gmc_warp,
842 :     dec->width, dec->height, &dec->new_gmc_data);
843 : edgomez 1.47
844 : edgomez 1.51 /* image warping is done block-based in decoder_mbgmc(), now */
845 : edgomez 1.47 }
846 :    
847 : chenm001 1.26 bound = 0;
848 :    
849 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
850 : albeu 1.32 cp_mb = st_mb = 0;
851 : edgomez 1.47 for (x = 0; x < mb_width; x++) {
852 : chenm001 1.26 MACROBLOCK *mb;
853 :    
854 : edgomez 1.48 /* skip stuffing */
855 : chenm001 1.26 while (BitstreamShowBits(bs, 10) == 1)
856 :     BitstreamSkip(bs, 10);
857 :    
858 : edgomez 1.51 if (check_resync_marker(bs, fcode - 1)) {
859 :     bound = read_video_packet_header(bs, dec, fcode - 1,
860 : edgomez 1.47 &quant, &fcode, NULL, &intra_dc_threshold);
861 :     x = bound % mb_width;
862 :     y = bound / mb_width;
863 : chenm001 1.26 }
864 :     mb = &dec->mbs[y * dec->mb_width + x];
865 :    
866 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
867 : suxen_drol 1.23
868 : edgomez 1.51 if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
869 :     uint32_t mcbpc, cbpc, cbpy, cbp;
870 :     uint32_t intra, acpred_flag = 0;
871 : edgomez 1.48 int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */
872 : Isibaar 1.1
873 : albeu 1.32 cp_mb++;
874 : Isibaar 1.1 mcbpc = get_mcbpc_inter(bs);
875 :     mb->mode = mcbpc & 7;
876 :     cbpc = (mcbpc >> 4);
877 : edgomez 1.51
878 :     DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
879 :     DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
880 : Isibaar 1.1
881 :     intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
882 : edgomez 1.19
883 : edgomez 1.47 if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
884 :     mcsel = BitstreamGetBit(bs);
885 : edgomez 1.51 else if (intra)
886 :     acpred_flag = BitstreamGetBit(bs);
887 : edgomez 1.47
888 : chenm001 1.26 cbpy = get_cbpy(bs, intra);
889 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
890 : chenm001 1.25
891 : Isibaar 1.1 cbp = (cbpy << 2) | cbpc;
892 :    
893 : edgomez 1.19 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
894 : chenm001 1.26 int dquant = dquant_table[BitstreamGetBits(bs, 2)];
895 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
896 : chenm001 1.26 quant += dquant;
897 : edgomez 1.19 if (quant > 31) {
898 : Isibaar 1.1 quant = 31;
899 : chenm001 1.26 } else if (quant < 1) {
900 : Isibaar 1.1 quant = 1;
901 :     }
902 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
903 : Isibaar 1.1 }
904 :     mb->quant = quant;
905 : h 1.5
906 : edgomez 1.19 if (dec->interlacing) {
907 : h 1.37 if (cbp || intra) {
908 :     mb->field_dct = BitstreamGetBit(bs);
909 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
910 : h 1.37 }
911 : h 1.5
912 : edgomez 1.51 if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) {
913 : h 1.5 mb->field_pred = BitstreamGetBit(bs);
914 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
915 : h 1.5
916 : edgomez 1.19 if (mb->field_pred) {
917 : h 1.5 mb->field_for_top = BitstreamGetBit(bs);
918 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
919 : h 1.5 mb->field_for_bot = BitstreamGetBit(bs);
920 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
921 : h 1.5 }
922 :     }
923 :     }
924 : edgomez 1.51
925 : edgomez 1.47 if (mcsel) {
926 : edgomez 1.51 decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
927 : edgomez 1.47 continue;
928 :    
929 :     } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
930 : h 1.5
931 : edgomez 1.19 if (dec->interlacing && mb->field_pred) {
932 : edgomez 1.51 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
933 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound);
934 : edgomez 1.19 } else {
935 : edgomez 1.51 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
936 : edgomez 1.47 mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
937 : h 1.5 }
938 : chl 1.35 } else if (mb->mode == MODE_INTER4V ) {
939 : chenm001 1.26 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
940 :     get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
941 :     get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
942 :     get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
943 : edgomez 1.51 } else { /* MODE_INTRA, MODE_INTRA_Q */
944 :     mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
945 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
946 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
947 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
948 : Isibaar 1.1 continue;
949 :     }
950 :    
951 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, cbp, bs,
952 :     rounding, reduced_resolution, 0);
953 : edgomez 1.47
954 : edgomez 1.51 } else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */
955 : edgomez 1.47 mb->mode = MODE_NOT_CODED_GMC;
956 : edgomez 1.51 mb->quant = quant;
957 :     decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
958 : edgomez 1.39
959 : edgomez 1.47 if(dec->out_frm && cp_mb > 0) {
960 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
961 :     cp_mb = 0;
962 : edgomez 1.47 }
963 :     st_mb = x+1;
964 : edgomez 1.51 } else { /* not coded P_VOP macroblock */
965 : chenm001 1.14 mb->mode = MODE_NOT_CODED;
966 : edgomez 1.51 mb->quant = quant;
967 : edgomez 1.47
968 : Isibaar 1.1 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
969 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
970 :    
971 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, 0, bs,
972 :     rounding, reduced_resolution, 0);
973 : edgomez 1.47
974 : albeu 1.32 if(dec->out_frm && cp_mb > 0) {
975 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
976 :     cp_mb = 0;
977 : albeu 1.32 }
978 :     st_mb = x+1;
979 : Isibaar 1.1 }
980 :     }
981 : edgomez 1.51
982 : albeu 1.32 if(dec->out_frm && cp_mb > 0)
983 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
984 : Isibaar 1.1 }
985 :     }
986 :    
987 : edgomez 1.47
988 : edgomez 1.48 /* decode B-frame motion vector */
989 : edgomez 1.51 static void
990 :     get_b_motion_vector(Bitstream * bs,
991 : edgomez 1.47 VECTOR * mv,
992 :     int fcode,
993 : edgomez 1.51 const VECTOR pmv,
994 :     const DECODER * const dec,
995 :     const int x, const int y)
996 : chenm001 1.14 {
997 : edgomez 1.51 const int scale_fac = 1 << (fcode - 1);
998 :     const int high = (32 * scale_fac) - 1;
999 :     const int low = ((-32) * scale_fac);
1000 :     const int range = (64 * scale_fac);
1001 : edgomez 1.7
1002 : edgomez 1.51 int mv_x = get_mv(bs, fcode);
1003 :     int mv_y = get_mv(bs, fcode);
1004 : Isibaar 1.1
1005 : edgomez 1.51 mv_x += pmv.x;
1006 :     mv_y += pmv.y;
1007 : albeu 1.32
1008 : edgomez 1.51 if (mv_x < low)
1009 : edgomez 1.47 mv_x += range;
1010 : edgomez 1.51 else if (mv_x > high)
1011 : edgomez 1.47 mv_x -= range;
1012 :    
1013 : edgomez 1.51 if (mv_y < low)
1014 : edgomez 1.47 mv_y += range;
1015 : edgomez 1.51 else if (mv_y > high)
1016 : edgomez 1.47 mv_y -= range;
1017 :    
1018 : edgomez 1.51 mv->x = mv_x;
1019 :     mv->y = mv_y;
1020 : edgomez 1.47 }
1021 :    
1022 : edgomez 1.51 /* decode an B-frame direct & interpolate macroblock */
1023 :     static void
1024 : edgomez 1.47 decoder_bf_interpolate_mbinter(DECODER * dec,
1025 : edgomez 1.51 IMAGE forward,
1026 :     IMAGE backward,
1027 :     const MACROBLOCK * pMB,
1028 :     const uint32_t x_pos,
1029 :     const uint32_t y_pos,
1030 :     Bitstream * bs,
1031 :     const int direct)
1032 : edgomez 1.47 {
1033 :     uint32_t stride = dec->edged_width;
1034 :     uint32_t stride2 = stride / 2;
1035 :     int uv_dx, uv_dy;
1036 :     int b_uv_dx, b_uv_dy;
1037 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
1038 : edgomez 1.51 const uint32_t cbp = pMB->cbp;
1039 : edgomez 1.47
1040 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
1041 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
1042 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
1043 :    
1044 : edgomez 1.51 if (!direct) {
1045 : edgomez 1.47 uv_dx = pMB->mvs[0].x;
1046 :     uv_dy = pMB->mvs[0].y;
1047 :     b_uv_dx = pMB->b_mvs[0].x;
1048 :     b_uv_dy = pMB->b_mvs[0].y;
1049 :    
1050 : edgomez 1.51 if (dec->quarterpel) {
1051 : edgomez 1.47 uv_dx /= 2;
1052 :     uv_dy /= 2;
1053 :     b_uv_dx /= 2;
1054 :     b_uv_dy /= 2;
1055 :     }
1056 :    
1057 :     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
1058 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
1059 :     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
1060 :     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
1061 : edgomez 1.51
1062 : edgomez 1.47 } else {
1063 : edgomez 1.62 uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
1064 :     uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
1065 :     b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
1066 :     b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
1067 :    
1068 :     if (dec->quarterpel) {
1069 :     uv_dx /= 2;
1070 :     uv_dy /= 2;
1071 :     b_uv_dx /= 2;
1072 :     b_uv_dy /= 2;
1073 : edgomez 1.51 }
1074 : edgomez 1.47
1075 : edgomez 1.51 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
1076 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
1077 :     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
1078 :     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
1079 : edgomez 1.47 }
1080 :    
1081 :     start_timer();
1082 :     if(dec->quarterpel) {
1083 : edgomez 1.51 if(!direct) {
1084 : edgomez 1.47 interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1085 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1086 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1087 :     } else {
1088 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1089 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1090 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1091 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1092 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1093 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1094 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1095 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1096 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1097 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1098 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1099 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1100 : edgomez 1.47 }
1101 : edgomez 1.51 } else {
1102 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
1103 : edgomez 1.51 pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1104 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
1105 : edgomez 1.51 pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1106 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
1107 : edgomez 1.51 pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1108 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8,
1109 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1110 : edgomez 1.47 }
1111 :    
1112 :     interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
1113 : edgomez 1.51 uv_dy, stride2, 0);
1114 : edgomez 1.47 interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
1115 : edgomez 1.51 uv_dy, stride2, 0);
1116 : edgomez 1.47
1117 :    
1118 :     if(dec->quarterpel) {
1119 : edgomez 1.51 if(!direct) {
1120 : edgomez 1.47 interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1121 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1122 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1123 :     } else {
1124 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1125 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1126 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1127 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1128 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1129 :     pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1130 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1131 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1132 :     pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1133 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1134 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1135 :     pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1136 : edgomez 1.47 }
1137 : edgomez 1.51 } else {
1138 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos,
1139 : edgomez 1.51 pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1140 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1141 : edgomez 1.51 16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1142 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos,
1143 : edgomez 1.51 16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1144 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1145 : edgomez 1.51 16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1146 : edgomez 1.47 }
1147 :    
1148 :     interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos,
1149 : edgomez 1.51 b_uv_dx, b_uv_dy, stride2, 0);
1150 : edgomez 1.47 interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos,
1151 : edgomez 1.51 b_uv_dx, b_uv_dy, stride2, 0);
1152 : edgomez 1.47
1153 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1154 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1155 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos,
1156 : edgomez 1.63 stride, 0, 8);
1157 : edgomez 1.47
1158 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1159 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1160 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1161 : edgomez 1.63 stride, 0, 8);
1162 : edgomez 1.47
1163 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1164 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1165 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1166 : edgomez 1.63 stride, 0, 8);
1167 : edgomez 1.47
1168 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1169 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1170 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1171 : edgomez 1.63 stride, 0, 8);
1172 : edgomez 1.47
1173 :     interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1174 :     dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1175 :     dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos,
1176 : edgomez 1.63 stride2, 0, 8);
1177 : edgomez 1.47
1178 :     interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1179 :     dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1180 :     dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos,
1181 : edgomez 1.63 stride2, 0, 8);
1182 : edgomez 1.47
1183 :     stop_comp_timer();
1184 :    
1185 : edgomez 1.51 if (cbp)
1186 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
1187 : edgomez 1.47 }
1188 :    
1189 : edgomez 1.48 /* for decode B-frame dbquant */
1190 : edgomez 1.51 static __inline int32_t
1191 : edgomez 1.47 get_dbquant(Bitstream * bs)
1192 :     {
1193 : edgomez 1.51 if (!BitstreamGetBit(bs)) /* '0' */
1194 : edgomez 1.47 return (0);
1195 : edgomez 1.51 else if (!BitstreamGetBit(bs)) /* '10' */
1196 : edgomez 1.47 return (-2);
1197 : edgomez 1.51 else /* '11' */
1198 : edgomez 1.48 return (2);
1199 : edgomez 1.47 }
1200 :    
1201 : edgomez 1.48 /*
1202 : edgomez 1.51 * decode B-frame mb_type
1203 :     * bit ret_value
1204 :     * 1 0
1205 :     * 01 1
1206 :     * 001 2
1207 :     * 0001 3
1208 : edgomez 1.48 */
1209 : edgomez 1.51 static int32_t __inline
1210 : edgomez 1.47 get_mbtype(Bitstream * bs)
1211 :     {
1212 :     int32_t mb_type;
1213 :    
1214 : edgomez 1.51 for (mb_type = 0; mb_type <= 3; mb_type++)
1215 : edgomez 1.47 if (BitstreamGetBit(bs))
1216 : edgomez 1.51 return (mb_type);
1217 : edgomez 1.47
1218 : edgomez 1.51 return -1;
1219 : edgomez 1.47 }
1220 :    
1221 : edgomez 1.51 static void
1222 : edgomez 1.47 decoder_bframe(DECODER * dec,
1223 : edgomez 1.51 Bitstream * bs,
1224 :     int quant,
1225 :     int fcode_forward,
1226 :     int fcode_backward)
1227 : edgomez 1.47 {
1228 :     uint32_t x, y;
1229 :     VECTOR mv;
1230 :     const VECTOR zeromv = {0,0};
1231 : edgomez 1.51 int i;
1232 : edgomez 1.47
1233 : edgomez 1.57 if (!dec->is_edged[0]) {
1234 :     start_timer();
1235 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
1236 :     dec->width, dec->height, dec->bs_version);
1237 :     dec->is_edged[0] = 1;
1238 :     stop_edges_timer();
1239 :     }
1240 :    
1241 :     if (!dec->is_edged[1]) {
1242 :     start_timer();
1243 :     image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
1244 :     dec->width, dec->height, dec->bs_version);
1245 :     dec->is_edged[1] = 1;
1246 :     stop_edges_timer();
1247 :     }
1248 : edgomez 1.47
1249 :     for (y = 0; y < dec->mb_height; y++) {
1250 : edgomez 1.48 /* Initialize Pred Motion Vector */
1251 : edgomez 1.47 dec->p_fmv = dec->p_bmv = zeromv;
1252 :     for (x = 0; x < dec->mb_width; x++) {
1253 :     MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
1254 :     MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
1255 : edgomez 1.51 const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward;
1256 :     uint32_t intra_dc_threshold; /* fake variable */
1257 :    
1258 :     if (check_resync_marker(bs, fcode_max - 1)) {
1259 :     int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant,
1260 :     &fcode_forward, &fcode_backward, &intra_dc_threshold);
1261 :     x = bound % dec->mb_width;
1262 :     y = bound / dec->mb_width;
1263 :     /* reset predicted macroblocks */
1264 :     dec->p_fmv = dec->p_bmv = zeromv;
1265 :     }
1266 : edgomez 1.47
1267 :     mv =
1268 :     mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
1269 :     mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
1270 : edgomez 1.51 mb->quant = quant;
1271 : edgomez 1.47
1272 : edgomez 1.48 /*
1273 : edgomez 1.51 * skip if the co-located P_VOP macroblock is not coded
1274 : edgomez 1.48 * if not codec in co-located S_VOP macroblock is _not_
1275 :     * automatically skipped
1276 :     */
1277 : edgomez 1.47
1278 :     if (last_mb->mode == MODE_NOT_CODED) {
1279 :     mb->cbp = 0;
1280 : edgomez 1.51 mb->mode = MODE_FORWARD;
1281 :     decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1282 : edgomez 1.47 continue;
1283 :     }
1284 :    
1285 : edgomez 1.48 if (!BitstreamGetBit(bs)) { /* modb=='0' */
1286 : edgomez 1.47 const uint8_t modb2 = BitstreamGetBit(bs);
1287 :    
1288 : edgomez 1.51 mb->mode = get_mbtype(bs);
1289 : edgomez 1.47
1290 : edgomez 1.51 if (!modb2) /* modb=='00' */
1291 : edgomez 1.47 mb->cbp = BitstreamGetBits(bs, 6);
1292 : edgomez 1.51 else
1293 : edgomez 1.47 mb->cbp = 0;
1294 : edgomez 1.51
1295 :     if (mb->mode && mb->cbp) {
1296 : edgomez 1.47 quant += get_dbquant(bs);
1297 : edgomez 1.51 if (quant > 31)
1298 : edgomez 1.47 quant = 31;
1299 : edgomez 1.51 else if (quant < 1)
1300 : edgomez 1.47 quant = 1;
1301 : edgomez 1.51 }
1302 :     mb->quant = quant;
1303 :    
1304 :     if (dec->interlacing) {
1305 :     if (mb->cbp) {
1306 :     mb->field_dct = BitstreamGetBit(bs);
1307 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1308 :     }
1309 :    
1310 :     if (mb->mode) {
1311 :     mb->field_pred = BitstreamGetBit(bs);
1312 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1313 :    
1314 :     if (mb->field_pred) {
1315 :     mb->field_for_top = BitstreamGetBit(bs);
1316 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1317 :     mb->field_for_bot = BitstreamGetBit(bs);
1318 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1319 :     }
1320 : edgomez 1.47 }
1321 :     }
1322 : edgomez 1.51
1323 : edgomez 1.47 } else {
1324 : edgomez 1.51 mb->mode = MODE_DIRECT_NONE_MV;
1325 : edgomez 1.47 mb->cbp = 0;
1326 :     }
1327 :    
1328 : edgomez 1.51 switch (mb->mode) {
1329 : edgomez 1.47 case MODE_DIRECT:
1330 : edgomez 1.51 get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y);
1331 : edgomez 1.47
1332 :     case MODE_DIRECT_NONE_MV:
1333 : edgomez 1.51 for (i = 0; i < 4; i++) {
1334 : edgomez 1.58 mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x;
1335 :     mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y;
1336 :    
1337 :     mb->b_mvs[i].x = (mv.x)
1338 :     ? mb->mvs[i].x - last_mb->mvs[i].x
1339 :     : last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp;
1340 :     mb->b_mvs[i].y = (mv.y)
1341 :     ? mb->mvs[i].y - last_mb->mvs[i].y
1342 :     : last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp;
1343 : edgomez 1.47 }
1344 : edgomez 1.51
1345 : edgomez 1.47 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1346 : edgomez 1.51 mb, x, y, bs, 1);
1347 : edgomez 1.47 break;
1348 :    
1349 :     case MODE_INTERPOLATE:
1350 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1351 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1352 :    
1353 : edgomez 1.51 get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1354 :     dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
1355 : edgomez 1.47
1356 :     decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1357 : edgomez 1.51 mb, x, y, bs, 0);
1358 : edgomez 1.47 break;
1359 :    
1360 :     case MODE_BACKWARD:
1361 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1362 : edgomez 1.47 dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1363 :    
1364 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0);
1365 : edgomez 1.47 break;
1366 :    
1367 :     case MODE_FORWARD:
1368 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1369 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1370 :    
1371 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1372 : edgomez 1.47 break;
1373 :    
1374 :     default:
1375 : edgomez 1.51 DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
1376 : edgomez 1.47 }
1377 : edgomez 1.48 } /* End of for */
1378 :     }
1379 : edgomez 1.47 }
1380 :    
1381 :     /* perform post processing if necessary, and output the image */
1382 : edgomez 1.51 void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
1383 :     xvid_dec_frame_t * frame, xvid_dec_stats_t * stats,
1384 :     int coding_type, int quant)
1385 : edgomez 1.47 {
1386 : suxen_drol 1.55 const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0;
1387 :    
1388 : edgomez 1.51 if (dec->cartoon_mode)
1389 :     frame->general &= ~XVID_FILMEFFECT;
1390 : edgomez 1.47
1391 : suxen_drol 1.55 if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0)
1392 : suxen_drol 1.52 && mbs != NULL) /* post process */
1393 : edgomez 1.47 {
1394 :     /* note: image is stored to tmp */
1395 :     image_copy(&dec->tmp, img, dec->edged_width, dec->height);
1396 : edgomez 1.51 image_postproc(&dec->postproc, &dec->tmp, dec->edged_width,
1397 :     mbs, dec->mb_width, dec->mb_height, dec->mb_width,
1398 : suxen_drol 1.55 frame->general, brightness, dec->frames, (coding_type == B_VOP));
1399 : edgomez 1.47 img = &dec->tmp;
1400 :     }
1401 :    
1402 :     image_output(img, dec->width, dec->height,
1403 : edgomez 1.51 dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
1404 :     frame->output.csp, dec->interlacing);
1405 :    
1406 :     if (stats) {
1407 :     stats->type = coding2type(coding_type);
1408 :     stats->data.vop.time_base = (int)dec->time_base;
1409 :     stats->data.vop.time_increment = 0; /* XXX: todo */
1410 :     stats->data.vop.qscale_stride = dec->mb_width;
1411 :     stats->data.vop.qscale = dec->qscale;
1412 :     if (stats->data.vop.qscale != NULL && mbs != NULL) {
1413 :     int i;
1414 :     for (i = 0; i < dec->mb_width*dec->mb_height; i++)
1415 :     stats->data.vop.qscale[i] = mbs[i].quant;
1416 :     } else
1417 :     stats->data.vop.qscale = NULL;
1418 :     }
1419 : edgomez 1.47 }
1420 :    
1421 :    
1422 :     int
1423 :     decoder_decode(DECODER * dec,
1424 : edgomez 1.51 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
1425 : edgomez 1.47 {
1426 :    
1427 :     Bitstream bs;
1428 :     uint32_t rounding;
1429 :     uint32_t reduced_resolution;
1430 : edgomez 1.51 uint32_t quant = 2;
1431 : edgomez 1.47 uint32_t fcode_forward;
1432 :     uint32_t fcode_backward;
1433 :     uint32_t intra_dc_threshold;
1434 :     WARPPOINTS gmc_warp;
1435 : edgomez 1.51 int coding_type;
1436 :     int success, output, seen_something;
1437 :    
1438 :     if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */
1439 :     return XVID_ERR_VERSION;
1440 : edgomez 1.47
1441 :     start_global_timer();
1442 :    
1443 : edgomez 1.51 dec->low_delay_default = (frame->general & XVID_LOWDELAY);
1444 :     if ((frame->general & XVID_DISCONTINUITY))
1445 : edgomez 1.47 dec->frames = 0;
1446 : edgomez 1.51 dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
1447 : edgomez 1.47
1448 : edgomez 1.51 if (frame->length < 0) { /* decoder flush */
1449 :     int ret;
1450 :     /* if not decoding "low_delay/packed", and this isn't low_delay and
1451 :     we have a reference frame, then outout the reference frame */
1452 :     if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
1453 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1454 :     dec->frames = 0;
1455 :     ret = 0;
1456 :     } else {
1457 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1458 :     ret = XVID_ERR_END;
1459 : edgomez 1.47 }
1460 :    
1461 :     emms();
1462 :     stop_global_timer();
1463 : edgomez 1.51 return ret;
1464 : edgomez 1.47 }
1465 : edgomez 1.19
1466 : Isibaar 1.1 BitstreamInit(&bs, frame->bitstream, frame->length);
1467 :    
1468 : edgomez 1.48 /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
1469 : edgomez 1.47 if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
1470 :     {
1471 :     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
1472 : edgomez 1.51 (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
1473 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1474 : edgomez 1.47 emms();
1475 : edgomez 1.51 return 1; /* one byte consumed */
1476 : edgomez 1.47 }
1477 :    
1478 : edgomez 1.51 success = 0;
1479 :     output = 0;
1480 :     seen_something = 0;
1481 :    
1482 : edgomez 1.47 repeat:
1483 :    
1484 : edgomez 1.51 coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution,
1485 : edgomez 1.47 &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
1486 :    
1487 : edgomez 1.51 DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n",
1488 :     coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
1489 : Isibaar 1.1
1490 : edgomez 1.51 if (coding_type == -1) { /* nothing */
1491 : edgomez 1.47 if (success) goto done;
1492 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1493 : edgomez 1.47 emms();
1494 : edgomez 1.51 return BitstreamPos(&bs)/8;
1495 : Isibaar 1.1 }
1496 :    
1497 : edgomez 1.51 if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
1498 :    
1499 :     if (coding_type == -3)
1500 : edgomez 1.47 decoder_resize(dec);
1501 : edgomez 1.51
1502 :     if (stats) {
1503 :     stats->type = XVID_TYPE_VOL;
1504 : edgomez 1.47 stats->data.vol.general = 0;
1505 : edgomez 1.51 /*XXX: if (dec->interlacing)
1506 :     stats->data.vol.general |= ++INTERLACING; */
1507 : edgomez 1.47 stats->data.vol.width = dec->width;
1508 :     stats->data.vol.height = dec->height;
1509 : edgomez 1.51 stats->data.vol.par = dec->aspect_ratio;
1510 : edgomez 1.47 stats->data.vol.par_width = dec->par_width;
1511 :     stats->data.vol.par_height = dec->par_height;
1512 :     emms();
1513 : edgomez 1.51 return BitstreamPos(&bs)/8; /* number of bytes consumed */
1514 : edgomez 1.47 }
1515 : syskin 1.56 goto repeat;
1516 :     }
1517 :    
1518 :     if(dec->frames == 0 && coding_type != I_VOP) {
1519 :     /* 1st frame is not an i-vop */
1520 : edgomez 1.47 goto repeat;
1521 : edgomez 1.51 }
1522 : edgomez 1.47
1523 : 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 */
1524 : edgomez 1.47
1525 :     /* packed_mode: special-N_VOP treament */
1526 : edgomez 1.51 if (dec->packed_mode && coding_type == N_VOP) {
1527 :     if (dec->low_delay_default && dec->frames > 0) {
1528 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1529 : edgomez 1.47 output = 1;
1530 :     }
1531 :     /* ignore otherwise */
1532 : edgomez 1.51 } else if (coding_type != B_VOP) {
1533 :     switch(coding_type) {
1534 : edgomez 1.47 case I_VOP :
1535 :     decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold);
1536 :     break;
1537 :     case P_VOP :
1538 : edgomez 1.51 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1539 : edgomez 1.47 fcode_forward, intra_dc_threshold, NULL);
1540 :     break;
1541 :     case S_VOP :
1542 : edgomez 1.51 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1543 : edgomez 1.47 fcode_forward, intra_dc_threshold, &gmc_warp);
1544 :     break;
1545 :     case N_VOP :
1546 : edgomez 1.51 /* XXX: not_coded vops are not used for forward prediction */
1547 :     /* we should not swap(last_mbs,mbs) */
1548 : edgomez 1.47 image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
1549 : edgomez 1.51 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */
1550 : edgomez 1.47 break;
1551 :     }
1552 :    
1553 : edgomez 1.51 if (reduced_resolution) {
1554 : edgomez 1.47 image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs,
1555 :     (dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width,
1556 : edgomez 1.51 16, 0);
1557 : edgomez 1.47 }
1558 : Isibaar 1.1
1559 : edgomez 1.47 /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
1560 : edgomez 1.51 if (!(dec->low_delay_default && dec->packed_mode)) {
1561 :     if (dec->low_delay) {
1562 :     decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1563 : edgomez 1.47 output = 1;
1564 : edgomez 1.51 } else if (dec->frames > 0) { /* is the reference frame valid? */
1565 : edgomez 1.47 /* output the reference frame */
1566 : edgomez 1.51 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1567 : edgomez 1.47 output = 1;
1568 :     }
1569 :     }
1570 : Isibaar 1.18
1571 : chenm001 1.14 image_swap(&dec->refn[0], &dec->refn[1]);
1572 : edgomez 1.57 dec->is_edged[1] = dec->is_edged[0];
1573 : chenm001 1.14 image_swap(&dec->cur, &dec->refn[0]);
1574 : edgomez 1.57 dec->is_edged[0] = 0;
1575 : edgomez 1.51 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
1576 : edgomez 1.47 dec->last_reduced_resolution = reduced_resolution;
1577 : edgomez 1.51 dec->last_coding_type = coding_type;
1578 : edgomez 1.47
1579 :     dec->frames++;
1580 :     seen_something = 1;
1581 : chenm001 1.31
1582 : edgomez 1.51 } else { /* B_VOP */
1583 : edgomez 1.47
1584 : edgomez 1.51 if (dec->low_delay) {
1585 :     DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
1586 : syskin 1.60 dec->low_delay = 0;
1587 : edgomez 1.47 }
1588 :    
1589 : edgomez 1.51 if (dec->frames < 2) {
1590 : edgomez 1.47 /* attemping to decode a bvop without atleast 2 reference frames */
1591 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1592 :     "broken b-frame, mising ref frames");
1593 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1594 :     } else if (dec->time_pp <= dec->time_bp) {
1595 :     /* this occurs when dx50_bvop_compatibility==0 sequences are
1596 : edgomez 1.47 decoded in vfw. */
1597 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1598 :     "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
1599 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1600 :     } else {
1601 : edgomez 1.47 decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
1602 : edgomez 1.51 decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1603 : edgomez 1.47 }
1604 :    
1605 :     output = 1;
1606 :     dec->frames++;
1607 : chenm001 1.14 }
1608 :    
1609 : edgomez 1.51 #if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */
1610 :     BitstreamByteAlign(&bs);
1611 :     #endif
1612 : edgomez 1.47
1613 :     /* low_delay_default mode: repeat in packed_mode */
1614 : edgomez 1.51 if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
1615 : edgomez 1.47 success = 1;
1616 :     goto repeat;
1617 :     }
1618 :    
1619 :     done :
1620 :    
1621 :     /* low_delay_default mode: if we've gotten here without outputting anything,
1622 :     then output the recently decoded frame, or print an error message */
1623 : edgomez 1.51 if (dec->low_delay_default && output == 0) {
1624 :     if (dec->packed_mode && seen_something) {
1625 : edgomez 1.47 /* output the recently decoded frame */
1626 : edgomez 1.51 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1627 :     } else {
1628 : edgomez 1.47 image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
1629 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1630 :     "warning: nothing to output");
1631 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64,
1632 :     "bframe decoder lag");
1633 :    
1634 : edgomez 1.51 decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant);
1635 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1636 : edgomez 1.47 }
1637 :     }
1638 : Isibaar 1.50
1639 : Isibaar 1.1 emms();
1640 :     stop_global_timer();
1641 :    
1642 : edgomez 1.51 return (BitstreamPos(&bs) + 7) / 8; /* number of bytes consumed */
1643 : Isibaar 1.1 }

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