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

Annotation of /xvidcore/src/decoder.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.68 - (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 : syskin 1.68 * $Id: decoder.c,v 1.67 2004/08/16 22:38:06 edgomez Exp $
24 : edgomez 1.43 *
25 : edgomez 1.48 ****************************************************************************/
26 : Isibaar 1.1
27 : edgomez 1.47 #include <stdio.h>
28 : Isibaar 1.1 #include <stdlib.h>
29 : edgomez 1.12 #include <string.h>
30 : Isibaar 1.1
31 : chenm001 1.27 #ifdef BFRAMES_DEC_DEBUG
32 :     #define BFRAMES_DEC
33 :     #endif
34 :    
35 : Isibaar 1.1 #include "xvid.h"
36 :     #include "portab.h"
37 : edgomez 1.47 #include "global.h"
38 : Isibaar 1.1
39 :     #include "decoder.h"
40 :     #include "bitstream/bitstream.h"
41 :     #include "bitstream/mbcoding.h"
42 :    
43 : edgomez 1.51 #include "quant/quant.h"
44 :     #include "quant/quant_matrix.h"
45 : Isibaar 1.1 #include "dct/idct.h"
46 :     #include "dct/fdct.h"
47 :     #include "utils/mem_transfer.h"
48 :     #include "image/interpolate8x8.h"
49 : edgomez 1.47 #include "image/font.h"
50 : edgomez 1.62 #include "image/qpel.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 : syskin 1.56 dec->time_inc_resolution = 1; /* until VOL header says otherwise */
226 : edgomez 1.47
227 :     dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
228 : edgomez 1.19
229 : edgomez 1.47 if (dec->fixed_dimensions)
230 :     return decoder_resize(dec);
231 :     else
232 : edgomez 1.51 return 0;
233 : Isibaar 1.1 }
234 :    
235 :    
236 : edgomez 1.19 int
237 :     decoder_destroy(DECODER * dec)
238 : Isibaar 1.1 {
239 : chenm001 1.14 xvid_free(dec->last_mbs);
240 : Isibaar 1.3 xvid_free(dec->mbs);
241 : edgomez 1.51 xvid_free(dec->qscale);
242 : edgomez 1.47
243 : edgomez 1.48 /* image based GMC */
244 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
245 : edgomez 1.47
246 : chenm001 1.11 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
247 : chenm001 1.14 image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
248 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
249 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
250 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
251 : edgomez 1.51 xvid_free(dec->mpeg_quant_matrices);
252 : Isibaar 1.3 xvid_free(dec);
253 : Isibaar 1.1
254 :     write_timer();
255 : edgomez 1.51 return 0;
256 : Isibaar 1.1 }
257 :    
258 : edgomez 1.19 static const int32_t dquant_table[4] = {
259 : Isibaar 1.1 -1, -2, 1, 2
260 :     };
261 :    
262 : edgomez 1.48 /* decode an intra macroblock */
263 : edgomez 1.51 static void
264 : edgomez 1.19 decoder_mbintra(DECODER * dec,
265 :     MACROBLOCK * pMB,
266 :     const uint32_t x_pos,
267 :     const uint32_t y_pos,
268 :     const uint32_t acpred_flag,
269 :     const uint32_t cbp,
270 :     Bitstream * bs,
271 :     const uint32_t quant,
272 : chenm001 1.26 const uint32_t intra_dc_threshold,
273 : syskin 1.68 const unsigned int bound)
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 : syskin 1.68 pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
287 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
288 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
289 : edgomez 1.7
290 : edgomez 1.48 memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */
291 : h 1.5
292 : edgomez 1.19 for (i = 0; i < 6; i++) {
293 : h 1.5 uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);
294 : Isibaar 1.1 int16_t predictors[8];
295 :     int start_coeff;
296 :    
297 :     start_timer();
298 : edgomez 1.19 predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64],
299 : edgomez 1.59 iQuant, iDcScaler, predictors, bound);
300 : edgomez 1.19 if (!acpred_flag) {
301 : h 1.5 pMB->acpred_directions[i] = 0;
302 : Isibaar 1.1 }
303 :     stop_prediction_timer();
304 :    
305 : edgomez 1.19 if (quant < intra_dc_threshold) {
306 : Isibaar 1.1 int dc_size;
307 :     int dc_dif;
308 :    
309 : edgomez 1.19 dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs);
310 :     dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0;
311 : Isibaar 1.1
312 : edgomez 1.19 if (dc_size > 8) {
313 : edgomez 1.48 BitstreamSkip(bs, 1); /* marker */
314 : Isibaar 1.1 }
315 : edgomez 1.19
316 :     block[i * 64 + 0] = dc_dif;
317 : Isibaar 1.1 start_coeff = 1;
318 : chenm001 1.26
319 : edgomez 1.51 DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif);
320 : edgomez 1.19 } else {
321 : Isibaar 1.1 start_coeff = 0;
322 :     }
323 :    
324 :     start_timer();
325 : edgomez 1.48 if (cbp & (1 << (5 - i))) /* coded */
326 : Isibaar 1.1 {
327 : edgomez 1.47 int direction = dec->alternate_vertical_scan ?
328 :     2 : pMB->acpred_directions[i];
329 :    
330 :     get_intra_block(bs, &block[i * 64], direction, start_coeff);
331 : Isibaar 1.1 }
332 :     stop_coding_timer();
333 :    
334 :     start_timer();
335 : edgomez 1.59 add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version);
336 : Isibaar 1.1 stop_prediction_timer();
337 :    
338 :     start_timer();
339 : edgomez 1.19 if (dec->quant_type == 0) {
340 : edgomez 1.51 dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
341 : edgomez 1.19 } else {
342 : edgomez 1.51 dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
343 : Isibaar 1.1 }
344 :     stop_iquant_timer();
345 :    
346 :     start_timer();
347 : edgomez 1.19 idct(&data[i * 64]);
348 : Isibaar 1.1 stop_idct_timer();
349 : edgomez 1.47
350 : h 1.5 }
351 : Isibaar 1.1
352 : edgomez 1.19 if (dec->interlacing && pMB->field_dct) {
353 : h 1.8 next_block = stride;
354 :     stride *= 2;
355 : Isibaar 1.1 }
356 : h 1.5
357 :     start_timer();
358 : syskin 1.68 transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
359 :     transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
360 :     transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
361 :     transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
362 :     transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
363 :     transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
364 : h 1.5 stop_transfer_timer();
365 : Isibaar 1.1 }
366 :    
367 : edgomez 1.51 static void
368 :     decoder_mb_decode(DECODER * dec,
369 :     const uint32_t cbp,
370 :     Bitstream * bs,
371 :     uint8_t * pY_Cur,
372 :     uint8_t * pU_Cur,
373 :     uint8_t * pV_Cur,
374 :     const MACROBLOCK * pMB)
375 :     {
376 : edgomez 1.64 DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE);
377 : edgomez 1.51
378 :     int stride = dec->edged_width;
379 : syskin 1.68 int next_block = stride * 8;
380 : edgomez 1.51 int i;
381 :     const uint32_t iQuant = pMB->quant;
382 :     const int direction = dec->alternate_vertical_scan ? 2 : 0;
383 : edgomez 1.61 typedef void (*get_inter_block_function_t)(
384 :     Bitstream * bs,
385 :     int16_t * block,
386 :     int direction,
387 :     const int quant,
388 :     const uint16_t *matrix);
389 : edgomez 1.64 typedef void (*add_residual_function_t)(
390 :     uint8_t *predicted_block,
391 :     const int16_t *residual,
392 :     int stride);
393 : edgomez 1.61
394 :     const get_inter_block_function_t get_inter_block = (dec->quant_type == 0)
395 : edgomez 1.64 ? (get_inter_block_function_t)get_inter_block_h263
396 :     : (get_inter_block_function_t)get_inter_block_mpeg;
397 :    
398 :     uint8_t *dst[6];
399 :     int strides[6];
400 : edgomez 1.61
401 : edgomez 1.51
402 : edgomez 1.64 if (dec->interlacing && pMB->field_dct) {
403 :     next_block = stride;
404 :     stride *= 2;
405 :     }
406 :    
407 :     dst[0] = pY_Cur;
408 :     dst[2] = pY_Cur + next_block;
409 : syskin 1.68 dst[1] = dst[0] + 8;
410 :     dst[3] = dst[2] + 8;
411 : edgomez 1.64 dst[4] = pU_Cur;
412 :     dst[5] = pV_Cur;
413 :     strides[0] = strides[1] = strides[2] = strides[3] = stride;
414 :     strides[4] = stride/2;
415 :     strides[5] = stride/2;
416 :    
417 : edgomez 1.51 for (i = 0; i < 6; i++) {
418 : edgomez 1.64 /* Process only coded blocks */
419 :     if (cbp & (1 << (5 - i))) {
420 : edgomez 1.51
421 : edgomez 1.64 /* Clear the block */
422 :     memset(&data[0], 0, 64*sizeof(int16_t));
423 : edgomez 1.51
424 : edgomez 1.61 /* Decode coeffs and dequantize on the fly */
425 : edgomez 1.51 start_timer();
426 : edgomez 1.64 get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices));
427 : edgomez 1.51 stop_coding_timer();
428 :    
429 : edgomez 1.64 /* iDCT */
430 : edgomez 1.51 start_timer();
431 : edgomez 1.64 idct(&data[0]);
432 : edgomez 1.51 stop_idct_timer();
433 : edgomez 1.64
434 :     /* Add this residual to the predicted block */
435 :     start_timer();
436 : syskin 1.68 transfer_16to8add(dst[i], &data[0], strides[i]);
437 : edgomez 1.64 stop_transfer_timer();
438 : edgomez 1.51 }
439 :     }
440 :     }
441 : Isibaar 1.1
442 : edgomez 1.67 static void __inline
443 : syskin 1.66 validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec)
444 :     {
445 :     /* clip a vector to valid range
446 :     prevents crashes if bitstream is broken
447 :     */
448 : edgomez 1.67 int shift = 5 + dec->quarterpel;
449 :     int xborder_high = (int)(dec->mb_width - x_pos) << shift;
450 :     int xborder_low = (-(int)x_pos-1) << shift;
451 :     int yborder_high = (int)(dec->mb_height - y_pos) << shift;
452 :     int yborder_low = (-(int)y_pos-1) << shift;
453 :    
454 :     #define CHECK_MV(mv) \
455 :     do { \
456 :     if ((mv).x > xborder_high) { \
457 :     DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \
458 :     (mv).x = xborder_high; \
459 :     } else if ((mv).x < xborder_low) { \
460 :     DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \
461 :     (mv).x = xborder_low; \
462 :     } \
463 :     if ((mv).y > yborder_high) { \
464 :     DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \
465 :     (mv).y = yborder_high; \
466 :     } else if ((mv).y < yborder_low) { \
467 :     DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \
468 :     (mv).y = yborder_low; \
469 :     } \
470 :     } while (0)
471 :    
472 :     CHECK_MV(mv[0]);
473 :     CHECK_MV(mv[1]);
474 :     CHECK_MV(mv[2]);
475 :     CHECK_MV(mv[3]);
476 : syskin 1.66 }
477 :    
478 : edgomez 1.48 /* decode an inter macroblock */
479 : edgomez 1.51 static void
480 : edgomez 1.19 decoder_mbinter(DECODER * dec,
481 :     const MACROBLOCK * pMB,
482 :     const uint32_t x_pos,
483 :     const uint32_t y_pos,
484 :     const uint32_t cbp,
485 :     Bitstream * bs,
486 : edgomez 1.47 const uint32_t rounding,
487 : edgomez 1.51 const int ref)
488 : Isibaar 1.1 {
489 : h 1.8 uint32_t stride = dec->edged_width;
490 :     uint32_t stride2 = stride / 2;
491 : edgomez 1.7 uint32_t i;
492 : edgomez 1.51
493 : h 1.5 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
494 : edgomez 1.47
495 : Isibaar 1.1 int uv_dx, uv_dy;
496 : edgomez 1.47 VECTOR mv[4]; /* local copy of mvs */
497 :    
498 : syskin 1.68 pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
499 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
500 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
501 :     for (i = 0; i < 4; i++)
502 :     mv[i] = pMB->mvs[i];
503 : edgomez 1.47
504 : syskin 1.66 validate_vector(mv, x_pos, y_pos, dec);
505 : syskin 1.54
506 : edgomez 1.51 start_timer();
507 :    
508 :     if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
509 :    
510 :     uv_dx = mv[0].x;
511 :     uv_dy = mv[0].y;
512 :     if (dec->quarterpel) {
513 :     uv_dx /= 2;
514 :     uv_dy /= 2;
515 :     }
516 : edgomez 1.47 uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
517 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
518 :    
519 : syskin 1.68 if (dec->quarterpel)
520 : edgomez 1.51 interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
521 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
522 :     mv[0].x, mv[0].y, stride, rounding);
523 : edgomez 1.47 else
524 : edgomez 1.51 interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
525 :     mv[0].x, mv[0].y, stride, rounding);
526 : Isibaar 1.1
527 : edgomez 1.47 } else { /* MODE_INTER4V */
528 :    
529 : edgomez 1.51 if(dec->quarterpel) {
530 :     uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
531 :     uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
532 :     } else {
533 :     uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
534 :     uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
535 :     }
536 : edgomez 1.47
537 : edgomez 1.51 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
538 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
539 : edgomez 1.47
540 : syskin 1.68 if (dec->quarterpel) {
541 : edgomez 1.51 interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
542 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
543 :     mv[0].x, mv[0].y, stride, rounding);
544 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
545 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
546 :     mv[1].x, mv[1].y, stride, rounding);
547 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
548 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
549 :     mv[2].x, mv[2].y, stride, rounding);
550 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
551 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
552 :     mv[3].x, mv[3].y, stride, rounding);
553 :     } else {
554 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
555 :     mv[0].x, mv[0].y, stride, rounding);
556 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
557 :     mv[1].x, mv[1].y, stride, rounding);
558 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
559 :     mv[2].x, mv[2].y, stride, rounding);
560 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
561 :     mv[3].x, mv[3].y, stride, rounding);
562 : edgomez 1.47 }
563 :     }
564 :    
565 : edgomez 1.51 /* chroma */
566 : syskin 1.68 interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
567 :     uv_dx, uv_dy, stride2, rounding);
568 :     interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
569 :     uv_dx, uv_dy, stride2, rounding);
570 : Isibaar 1.1
571 : edgomez 1.51 stop_comp_timer();
572 : edgomez 1.47
573 : edgomez 1.51 if (cbp)
574 : syskin 1.68 decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
575 : edgomez 1.47 }
576 :    
577 :     static void
578 :     decoder_mbgmc(DECODER * dec,
579 :     MACROBLOCK * const pMB,
580 :     const uint32_t x_pos,
581 :     const uint32_t y_pos,
582 :     const uint32_t fcode,
583 :     const uint32_t cbp,
584 :     Bitstream * bs,
585 : edgomez 1.51 const uint32_t rounding)
586 : edgomez 1.47 {
587 :     const uint32_t stride = dec->edged_width;
588 :     const uint32_t stride2 = stride / 2;
589 : edgomez 1.51
590 : edgomez 1.47 uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
591 :     uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
592 :     uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
593 :    
594 : edgomez 1.51 NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
595 :    
596 : edgomez 1.47 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
597 : Isibaar 1.1
598 :     start_timer();
599 : edgomez 1.51
600 : edgomez 1.47 /* this is where the calculations are done */
601 :    
602 : edgomez 1.51 gmc_data->predict_16x16(gmc_data,
603 :     dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
604 :     stride, stride, x_pos, y_pos, rounding);
605 :    
606 :     gmc_data->predict_8x8(gmc_data,
607 :     dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
608 :     dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
609 :     stride2, stride2, x_pos, y_pos, rounding);
610 : edgomez 1.47
611 : edgomez 1.51 gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
612 : Isibaar 1.1
613 : edgomez 1.51 pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
614 :     pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
615 : edgomez 1.47
616 : edgomez 1.51 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
617 : Isibaar 1.1
618 : edgomez 1.51 stop_transfer_timer();
619 : Isibaar 1.1
620 : edgomez 1.51 if (cbp)
621 : syskin 1.68 decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
622 : Isibaar 1.1
623 :     }
624 :    
625 :    
626 : edgomez 1.51 static void
627 : edgomez 1.19 decoder_iframe(DECODER * dec,
628 : edgomez 1.51 Bitstream * bs,
629 :     int quant,
630 :     int intra_dc_threshold)
631 : Isibaar 1.1 {
632 : chenm001 1.26 uint32_t bound;
633 :     uint32_t x, y;
634 : syskin 1.68 const uint32_t mb_width = dec->mb_width;
635 :     const uint32_t mb_height = dec->mb_height;
636 : chenm001 1.25
637 : chenm001 1.26 bound = 0;
638 : edgomez 1.7
639 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
640 :     for (x = 0; x < mb_width; x++) {
641 : chenm001 1.26 MACROBLOCK *mb;
642 : Isibaar 1.1 uint32_t mcbpc;
643 :     uint32_t cbpc;
644 :     uint32_t acpred_flag;
645 :     uint32_t cbpy;
646 :     uint32_t cbp;
647 :    
648 : chenm001 1.26 while (BitstreamShowBits(bs, 9) == 1)
649 :     BitstreamSkip(bs, 9);
650 :    
651 :     if (check_resync_marker(bs, 0))
652 :     {
653 : edgomez 1.51 bound = read_video_packet_header(bs, dec, 0,
654 : edgomez 1.47 &quant, NULL, NULL, &intra_dc_threshold);
655 :     x = bound % mb_width;
656 :     y = bound / mb_width;
657 : chenm001 1.26 }
658 :     mb = &dec->mbs[y * dec->mb_width + x];
659 :    
660 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
661 : chenm001 1.26
662 : Isibaar 1.1 mcbpc = get_mcbpc_intra(bs);
663 :     mb->mode = mcbpc & 7;
664 :     cbpc = (mcbpc >> 4);
665 :    
666 :     acpred_flag = BitstreamGetBit(bs);
667 :    
668 :     cbpy = get_cbpy(bs, 1);
669 :     cbp = (cbpy << 2) | cbpc;
670 :    
671 : edgomez 1.19 if (mb->mode == MODE_INTRA_Q) {
672 :     quant += dquant_table[BitstreamGetBits(bs, 2)];
673 :     if (quant > 31) {
674 : Isibaar 1.1 quant = 31;
675 : edgomez 1.19 } else if (quant < 1) {
676 : Isibaar 1.1 quant = 1;
677 :     }
678 :     }
679 :     mb->quant = quant;
680 : chenm001 1.30 mb->mvs[0].x = mb->mvs[0].y =
681 :     mb->mvs[1].x = mb->mvs[1].y =
682 :     mb->mvs[2].x = mb->mvs[2].y =
683 :     mb->mvs[3].x = mb->mvs[3].y =0;
684 : h 1.5
685 : edgomez 1.19 if (dec->interlacing) {
686 : h 1.5 mb->field_dct = BitstreamGetBit(bs);
687 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
688 : h 1.5 }
689 : Isibaar 1.1
690 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
691 : syskin 1.68 intra_dc_threshold, bound);
692 : edgomez 1.47
693 : Isibaar 1.1 }
694 : albeu 1.32 if(dec->out_frm)
695 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
696 : Isibaar 1.1 }
697 : edgomez 1.7
698 : Isibaar 1.1 }
699 :    
700 :    
701 : edgomez 1.51 static void
702 : edgomez 1.19 get_motion_vector(DECODER * dec,
703 : edgomez 1.51 Bitstream * bs,
704 :     int x,
705 :     int y,
706 :     int k,
707 :     VECTOR * ret_mv,
708 :     int fcode,
709 :     const int bound)
710 : Isibaar 1.1 {
711 : edgomez 1.7
712 : edgomez 1.51 const int scale_fac = 1 << (fcode - 1);
713 :     const int high = (32 * scale_fac) - 1;
714 :     const int low = ((-32) * scale_fac);
715 :     const int range = (64 * scale_fac);
716 : Isibaar 1.1
717 : edgomez 1.51 const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
718 : edgomez 1.47 VECTOR mv;
719 : Isibaar 1.1
720 : edgomez 1.47 mv.x = get_mv(bs, fcode);
721 :     mv.y = get_mv(bs, fcode);
722 : edgomez 1.19
723 : 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);
724 : chenm001 1.26
725 : edgomez 1.47 mv.x += pmv.x;
726 :     mv.y += pmv.y;
727 : Isibaar 1.1
728 : edgomez 1.47 if (mv.x < low) {
729 :     mv.x += range;
730 :     } else if (mv.x > high) {
731 :     mv.x -= range;
732 : Isibaar 1.1 }
733 :    
734 : edgomez 1.47 if (mv.y < low) {
735 :     mv.y += range;
736 :     } else if (mv.y > high) {
737 :     mv.y -= range;
738 : Isibaar 1.1 }
739 :    
740 : edgomez 1.51 ret_mv->x = mv.x;
741 :     ret_mv->y = mv.y;
742 :     }
743 : Isibaar 1.1
744 : edgomez 1.47 /* for P_VOP set gmc_warp to NULL */
745 : edgomez 1.51 static void
746 : edgomez 1.19 decoder_pframe(DECODER * dec,
747 : edgomez 1.51 Bitstream * bs,
748 :     int rounding,
749 :     int quant,
750 :     int fcode,
751 :     int intra_dc_threshold,
752 :     const WARPPOINTS *const gmc_warp)
753 : Isibaar 1.1 {
754 :     uint32_t x, y;
755 : chenm001 1.26 uint32_t bound;
756 : albeu 1.32 int cp_mb, st_mb;
757 : syskin 1.68 const uint32_t mb_width = dec->mb_width;
758 :     const uint32_t mb_height = dec->mb_height;
759 : Isibaar 1.1
760 : edgomez 1.57 if (!dec->is_edged[0]) {
761 :     start_timer();
762 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
763 :     dec->width, dec->height, dec->bs_version);
764 :     dec->is_edged[0] = 1;
765 :     stop_edges_timer();
766 :     }
767 : Isibaar 1.1
768 : edgomez 1.51 if (gmc_warp) {
769 :     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
770 :     generate_GMCparameters( dec->sprite_warping_points,
771 :     dec->sprite_warping_accuracy, gmc_warp,
772 :     dec->width, dec->height, &dec->new_gmc_data);
773 : edgomez 1.47
774 : edgomez 1.51 /* image warping is done block-based in decoder_mbgmc(), now */
775 : edgomez 1.47 }
776 :    
777 : chenm001 1.26 bound = 0;
778 :    
779 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
780 : albeu 1.32 cp_mb = st_mb = 0;
781 : edgomez 1.47 for (x = 0; x < mb_width; x++) {
782 : chenm001 1.26 MACROBLOCK *mb;
783 :    
784 : edgomez 1.48 /* skip stuffing */
785 : chenm001 1.26 while (BitstreamShowBits(bs, 10) == 1)
786 :     BitstreamSkip(bs, 10);
787 :    
788 : edgomez 1.51 if (check_resync_marker(bs, fcode - 1)) {
789 :     bound = read_video_packet_header(bs, dec, fcode - 1,
790 : edgomez 1.47 &quant, &fcode, NULL, &intra_dc_threshold);
791 :     x = bound % mb_width;
792 :     y = bound / mb_width;
793 : chenm001 1.26 }
794 :     mb = &dec->mbs[y * dec->mb_width + x];
795 :    
796 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
797 : suxen_drol 1.23
798 : edgomez 1.51 if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
799 :     uint32_t mcbpc, cbpc, cbpy, cbp;
800 :     uint32_t intra, acpred_flag = 0;
801 : edgomez 1.48 int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */
802 : Isibaar 1.1
803 : albeu 1.32 cp_mb++;
804 : Isibaar 1.1 mcbpc = get_mcbpc_inter(bs);
805 :     mb->mode = mcbpc & 7;
806 :     cbpc = (mcbpc >> 4);
807 : edgomez 1.51
808 :     DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
809 :     DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
810 : Isibaar 1.1
811 :     intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
812 : edgomez 1.19
813 : edgomez 1.47 if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
814 :     mcsel = BitstreamGetBit(bs);
815 : edgomez 1.51 else if (intra)
816 :     acpred_flag = BitstreamGetBit(bs);
817 : edgomez 1.47
818 : chenm001 1.26 cbpy = get_cbpy(bs, intra);
819 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
820 : chenm001 1.25
821 : Isibaar 1.1 cbp = (cbpy << 2) | cbpc;
822 :    
823 : edgomez 1.19 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
824 : chenm001 1.26 int dquant = dquant_table[BitstreamGetBits(bs, 2)];
825 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
826 : chenm001 1.26 quant += dquant;
827 : edgomez 1.19 if (quant > 31) {
828 : Isibaar 1.1 quant = 31;
829 : chenm001 1.26 } else if (quant < 1) {
830 : Isibaar 1.1 quant = 1;
831 :     }
832 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
833 : Isibaar 1.1 }
834 :     mb->quant = quant;
835 : h 1.5
836 : edgomez 1.19 if (dec->interlacing) {
837 : h 1.37 if (cbp || intra) {
838 :     mb->field_dct = BitstreamGetBit(bs);
839 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
840 : h 1.37 }
841 : h 1.5
842 : edgomez 1.51 if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) {
843 : h 1.5 mb->field_pred = BitstreamGetBit(bs);
844 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
845 : h 1.5
846 : edgomez 1.19 if (mb->field_pred) {
847 : h 1.5 mb->field_for_top = BitstreamGetBit(bs);
848 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
849 : h 1.5 mb->field_for_bot = BitstreamGetBit(bs);
850 : edgomez 1.51 DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
851 : h 1.5 }
852 :     }
853 :     }
854 : edgomez 1.51
855 : edgomez 1.47 if (mcsel) {
856 : edgomez 1.51 decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
857 : edgomez 1.47 continue;
858 :    
859 :     } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
860 : h 1.5
861 : edgomez 1.19 if (dec->interlacing && mb->field_pred) {
862 : edgomez 1.51 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
863 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound);
864 : edgomez 1.19 } else {
865 : edgomez 1.51 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
866 : edgomez 1.47 mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
867 : h 1.5 }
868 : chl 1.35 } else if (mb->mode == MODE_INTER4V ) {
869 : chenm001 1.26 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
870 :     get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
871 :     get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
872 :     get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
873 : edgomez 1.51 } else { /* MODE_INTRA, MODE_INTRA_Q */
874 :     mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
875 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
876 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
877 : syskin 1.68 intra_dc_threshold, bound);
878 : Isibaar 1.1 continue;
879 :     }
880 :    
881 : syskin 1.68 decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0);
882 : edgomez 1.47
883 : edgomez 1.51 } else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */
884 : edgomez 1.47 mb->mode = MODE_NOT_CODED_GMC;
885 : edgomez 1.51 mb->quant = quant;
886 :     decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
887 : edgomez 1.39
888 : edgomez 1.47 if(dec->out_frm && cp_mb > 0) {
889 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
890 :     cp_mb = 0;
891 : edgomez 1.47 }
892 :     st_mb = x+1;
893 : edgomez 1.51 } else { /* not coded P_VOP macroblock */
894 : chenm001 1.14 mb->mode = MODE_NOT_CODED;
895 : edgomez 1.51 mb->quant = quant;
896 : edgomez 1.47
897 : Isibaar 1.1 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
898 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
899 :    
900 : edgomez 1.51 decoder_mbinter(dec, mb, x, y, 0, bs,
901 : syskin 1.68 rounding, 0);
902 : edgomez 1.47
903 : albeu 1.32 if(dec->out_frm && cp_mb > 0) {
904 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
905 :     cp_mb = 0;
906 : albeu 1.32 }
907 :     st_mb = x+1;
908 : Isibaar 1.1 }
909 :     }
910 : edgomez 1.51
911 : albeu 1.32 if(dec->out_frm && cp_mb > 0)
912 : edgomez 1.51 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
913 : Isibaar 1.1 }
914 :     }
915 :    
916 : edgomez 1.47
917 : edgomez 1.48 /* decode B-frame motion vector */
918 : edgomez 1.51 static void
919 :     get_b_motion_vector(Bitstream * bs,
920 : edgomez 1.47 VECTOR * mv,
921 :     int fcode,
922 : edgomez 1.51 const VECTOR pmv,
923 :     const DECODER * const dec,
924 :     const int x, const int y)
925 : chenm001 1.14 {
926 : edgomez 1.51 const int scale_fac = 1 << (fcode - 1);
927 :     const int high = (32 * scale_fac) - 1;
928 :     const int low = ((-32) * scale_fac);
929 :     const int range = (64 * scale_fac);
930 : edgomez 1.7
931 : edgomez 1.51 int mv_x = get_mv(bs, fcode);
932 :     int mv_y = get_mv(bs, fcode);
933 : Isibaar 1.1
934 : edgomez 1.51 mv_x += pmv.x;
935 :     mv_y += pmv.y;
936 : albeu 1.32
937 : edgomez 1.51 if (mv_x < low)
938 : edgomez 1.47 mv_x += range;
939 : edgomez 1.51 else if (mv_x > high)
940 : edgomez 1.47 mv_x -= range;
941 :    
942 : edgomez 1.51 if (mv_y < low)
943 : edgomez 1.47 mv_y += range;
944 : edgomez 1.51 else if (mv_y > high)
945 : edgomez 1.47 mv_y -= range;
946 :    
947 : edgomez 1.51 mv->x = mv_x;
948 :     mv->y = mv_y;
949 : edgomez 1.47 }
950 :    
951 : edgomez 1.51 /* decode an B-frame direct & interpolate macroblock */
952 :     static void
953 : edgomez 1.47 decoder_bf_interpolate_mbinter(DECODER * dec,
954 : edgomez 1.51 IMAGE forward,
955 :     IMAGE backward,
956 : syskin 1.66 MACROBLOCK * pMB,
957 : edgomez 1.51 const uint32_t x_pos,
958 :     const uint32_t y_pos,
959 :     Bitstream * bs,
960 :     const int direct)
961 : edgomez 1.47 {
962 :     uint32_t stride = dec->edged_width;
963 :     uint32_t stride2 = stride / 2;
964 :     int uv_dx, uv_dy;
965 :     int b_uv_dx, b_uv_dy;
966 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
967 : edgomez 1.51 const uint32_t cbp = pMB->cbp;
968 : edgomez 1.47
969 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
970 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
971 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
972 :    
973 : syskin 1.66 validate_vector(pMB->mvs, x_pos, y_pos, dec);
974 :     validate_vector(pMB->b_mvs, x_pos, y_pos, dec);
975 :    
976 : edgomez 1.51 if (!direct) {
977 : edgomez 1.47 uv_dx = pMB->mvs[0].x;
978 :     uv_dy = pMB->mvs[0].y;
979 :     b_uv_dx = pMB->b_mvs[0].x;
980 :     b_uv_dy = pMB->b_mvs[0].y;
981 :    
982 : edgomez 1.51 if (dec->quarterpel) {
983 : edgomez 1.47 uv_dx /= 2;
984 :     uv_dy /= 2;
985 :     b_uv_dx /= 2;
986 :     b_uv_dy /= 2;
987 :     }
988 :    
989 :     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
990 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
991 :     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
992 :     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
993 : edgomez 1.51
994 : edgomez 1.47 } else {
995 : edgomez 1.62 uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
996 :     uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
997 :     b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
998 :     b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
999 :    
1000 :     if (dec->quarterpel) {
1001 :     uv_dx /= 2;
1002 :     uv_dy /= 2;
1003 :     b_uv_dx /= 2;
1004 :     b_uv_dy /= 2;
1005 : edgomez 1.51 }
1006 : edgomez 1.47
1007 : edgomez 1.51 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
1008 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
1009 :     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
1010 :     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
1011 : edgomez 1.47 }
1012 :    
1013 :     start_timer();
1014 :     if(dec->quarterpel) {
1015 : edgomez 1.51 if(!direct) {
1016 : edgomez 1.47 interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1017 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1018 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1019 :     } else {
1020 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1021 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1022 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1023 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1024 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1025 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1026 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1027 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1028 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1029 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1030 : edgomez 1.51 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1031 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1032 : edgomez 1.47 }
1033 : edgomez 1.51 } else {
1034 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
1035 : edgomez 1.51 pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1036 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
1037 : edgomez 1.51 pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1038 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
1039 : edgomez 1.51 pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1040 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8,
1041 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1042 : edgomez 1.47 }
1043 :    
1044 :     interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
1045 : edgomez 1.51 uv_dy, stride2, 0);
1046 : edgomez 1.47 interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
1047 : edgomez 1.51 uv_dy, stride2, 0);
1048 : edgomez 1.47
1049 :    
1050 :     if(dec->quarterpel) {
1051 : edgomez 1.51 if(!direct) {
1052 : edgomez 1.65 interpolate16x16_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1053 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1054 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1055 : edgomez 1.51 } else {
1056 : edgomez 1.65 interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1057 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1058 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1059 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1060 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1061 :     pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1062 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1063 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1064 :     pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1065 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1066 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1067 :     pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1068 : edgomez 1.47 }
1069 : edgomez 1.51 } else {
1070 : edgomez 1.65 interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos,
1071 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1072 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
1073 :     16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1074 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos,
1075 :     16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1076 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
1077 :     16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1078 :     }
1079 :    
1080 :     interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos,
1081 :     b_uv_dx, b_uv_dy, stride2, 0);
1082 :     interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos,
1083 :     b_uv_dx, b_uv_dy, stride2, 0);
1084 : edgomez 1.47
1085 :     stop_comp_timer();
1086 :    
1087 : edgomez 1.51 if (cbp)
1088 : syskin 1.68 decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
1089 : edgomez 1.47 }
1090 :    
1091 : edgomez 1.48 /* for decode B-frame dbquant */
1092 : edgomez 1.51 static __inline int32_t
1093 : edgomez 1.47 get_dbquant(Bitstream * bs)
1094 :     {
1095 : edgomez 1.51 if (!BitstreamGetBit(bs)) /* '0' */
1096 : edgomez 1.47 return (0);
1097 : edgomez 1.51 else if (!BitstreamGetBit(bs)) /* '10' */
1098 : edgomez 1.47 return (-2);
1099 : edgomez 1.51 else /* '11' */
1100 : edgomez 1.48 return (2);
1101 : edgomez 1.47 }
1102 :    
1103 : edgomez 1.48 /*
1104 : edgomez 1.51 * decode B-frame mb_type
1105 :     * bit ret_value
1106 :     * 1 0
1107 :     * 01 1
1108 :     * 001 2
1109 :     * 0001 3
1110 : edgomez 1.48 */
1111 : edgomez 1.51 static int32_t __inline
1112 : edgomez 1.47 get_mbtype(Bitstream * bs)
1113 :     {
1114 :     int32_t mb_type;
1115 :    
1116 : edgomez 1.51 for (mb_type = 0; mb_type <= 3; mb_type++)
1117 : edgomez 1.47 if (BitstreamGetBit(bs))
1118 : edgomez 1.51 return (mb_type);
1119 : edgomez 1.47
1120 : edgomez 1.51 return -1;
1121 : edgomez 1.47 }
1122 :    
1123 : edgomez 1.51 static void
1124 : edgomez 1.47 decoder_bframe(DECODER * dec,
1125 : edgomez 1.51 Bitstream * bs,
1126 :     int quant,
1127 :     int fcode_forward,
1128 :     int fcode_backward)
1129 : edgomez 1.47 {
1130 :     uint32_t x, y;
1131 :     VECTOR mv;
1132 :     const VECTOR zeromv = {0,0};
1133 : edgomez 1.51 int i;
1134 : edgomez 1.47
1135 : edgomez 1.57 if (!dec->is_edged[0]) {
1136 :     start_timer();
1137 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
1138 :     dec->width, dec->height, dec->bs_version);
1139 :     dec->is_edged[0] = 1;
1140 :     stop_edges_timer();
1141 :     }
1142 :    
1143 :     if (!dec->is_edged[1]) {
1144 :     start_timer();
1145 :     image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
1146 :     dec->width, dec->height, dec->bs_version);
1147 :     dec->is_edged[1] = 1;
1148 :     stop_edges_timer();
1149 :     }
1150 : edgomez 1.47
1151 :     for (y = 0; y < dec->mb_height; y++) {
1152 : edgomez 1.48 /* Initialize Pred Motion Vector */
1153 : edgomez 1.47 dec->p_fmv = dec->p_bmv = zeromv;
1154 :     for (x = 0; x < dec->mb_width; x++) {
1155 :     MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
1156 :     MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
1157 : edgomez 1.51 const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward;
1158 :     uint32_t intra_dc_threshold; /* fake variable */
1159 :    
1160 :     if (check_resync_marker(bs, fcode_max - 1)) {
1161 :     int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant,
1162 :     &fcode_forward, &fcode_backward, &intra_dc_threshold);
1163 :     x = bound % dec->mb_width;
1164 :     y = bound / dec->mb_width;
1165 :     /* reset predicted macroblocks */
1166 :     dec->p_fmv = dec->p_bmv = zeromv;
1167 :     }
1168 : edgomez 1.47
1169 :     mv =
1170 :     mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
1171 :     mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
1172 : edgomez 1.51 mb->quant = quant;
1173 : edgomez 1.47
1174 : edgomez 1.48 /*
1175 : edgomez 1.51 * skip if the co-located P_VOP macroblock is not coded
1176 : edgomez 1.48 * if not codec in co-located S_VOP macroblock is _not_
1177 :     * automatically skipped
1178 :     */
1179 : edgomez 1.47
1180 :     if (last_mb->mode == MODE_NOT_CODED) {
1181 :     mb->cbp = 0;
1182 : edgomez 1.51 mb->mode = MODE_FORWARD;
1183 : syskin 1.68 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1);
1184 : edgomez 1.47 continue;
1185 :     }
1186 :    
1187 : edgomez 1.48 if (!BitstreamGetBit(bs)) { /* modb=='0' */
1188 : edgomez 1.47 const uint8_t modb2 = BitstreamGetBit(bs);
1189 :    
1190 : edgomez 1.51 mb->mode = get_mbtype(bs);
1191 : edgomez 1.47
1192 : edgomez 1.51 if (!modb2) /* modb=='00' */
1193 : edgomez 1.47 mb->cbp = BitstreamGetBits(bs, 6);
1194 : edgomez 1.51 else
1195 : edgomez 1.47 mb->cbp = 0;
1196 : edgomez 1.51
1197 :     if (mb->mode && mb->cbp) {
1198 : edgomez 1.47 quant += get_dbquant(bs);
1199 : edgomez 1.51 if (quant > 31)
1200 : edgomez 1.47 quant = 31;
1201 : edgomez 1.51 else if (quant < 1)
1202 : edgomez 1.47 quant = 1;
1203 : edgomez 1.51 }
1204 :     mb->quant = quant;
1205 :    
1206 :     if (dec->interlacing) {
1207 :     if (mb->cbp) {
1208 :     mb->field_dct = BitstreamGetBit(bs);
1209 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1210 :     }
1211 :    
1212 :     if (mb->mode) {
1213 :     mb->field_pred = BitstreamGetBit(bs);
1214 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1215 :    
1216 :     if (mb->field_pred) {
1217 :     mb->field_for_top = BitstreamGetBit(bs);
1218 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1219 :     mb->field_for_bot = BitstreamGetBit(bs);
1220 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1221 :     }
1222 : edgomez 1.47 }
1223 :     }
1224 : edgomez 1.51
1225 : edgomez 1.47 } else {
1226 : edgomez 1.51 mb->mode = MODE_DIRECT_NONE_MV;
1227 : edgomez 1.47 mb->cbp = 0;
1228 :     }
1229 :    
1230 : edgomez 1.51 switch (mb->mode) {
1231 : edgomez 1.47 case MODE_DIRECT:
1232 : edgomez 1.51 get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y);
1233 : edgomez 1.47
1234 :     case MODE_DIRECT_NONE_MV:
1235 : edgomez 1.51 for (i = 0; i < 4; i++) {
1236 : edgomez 1.58 mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x;
1237 :     mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y;
1238 :    
1239 :     mb->b_mvs[i].x = (mv.x)
1240 :     ? mb->mvs[i].x - last_mb->mvs[i].x
1241 :     : last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp;
1242 :     mb->b_mvs[i].y = (mv.y)
1243 :     ? mb->mvs[i].y - last_mb->mvs[i].y
1244 :     : last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp;
1245 : edgomez 1.47 }
1246 : edgomez 1.51
1247 : edgomez 1.47 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1248 : edgomez 1.51 mb, x, y, bs, 1);
1249 : edgomez 1.47 break;
1250 :    
1251 :     case MODE_INTERPOLATE:
1252 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1253 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1254 :    
1255 : edgomez 1.51 get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1256 :     dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
1257 : edgomez 1.47
1258 :     decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1259 : edgomez 1.51 mb, x, y, bs, 0);
1260 : edgomez 1.47 break;
1261 :    
1262 :     case MODE_BACKWARD:
1263 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1264 : edgomez 1.47 dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1265 :    
1266 : syskin 1.68 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0);
1267 : edgomez 1.47 break;
1268 :    
1269 :     case MODE_FORWARD:
1270 : edgomez 1.51 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1271 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1272 :    
1273 : syskin 1.68 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1);
1274 : edgomez 1.47 break;
1275 :    
1276 :     default:
1277 : edgomez 1.51 DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
1278 : edgomez 1.47 }
1279 : edgomez 1.48 } /* End of for */
1280 :     }
1281 : edgomez 1.47 }
1282 :    
1283 :     /* perform post processing if necessary, and output the image */
1284 : edgomez 1.51 void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
1285 :     xvid_dec_frame_t * frame, xvid_dec_stats_t * stats,
1286 :     int coding_type, int quant)
1287 : edgomez 1.47 {
1288 : suxen_drol 1.55 const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0;
1289 :    
1290 : edgomez 1.51 if (dec->cartoon_mode)
1291 :     frame->general &= ~XVID_FILMEFFECT;
1292 : edgomez 1.47
1293 : suxen_drol 1.55 if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0)
1294 : suxen_drol 1.52 && mbs != NULL) /* post process */
1295 : edgomez 1.47 {
1296 :     /* note: image is stored to tmp */
1297 :     image_copy(&dec->tmp, img, dec->edged_width, dec->height);
1298 : edgomez 1.51 image_postproc(&dec->postproc, &dec->tmp, dec->edged_width,
1299 :     mbs, dec->mb_width, dec->mb_height, dec->mb_width,
1300 : suxen_drol 1.55 frame->general, brightness, dec->frames, (coding_type == B_VOP));
1301 : edgomez 1.47 img = &dec->tmp;
1302 :     }
1303 :    
1304 :     image_output(img, dec->width, dec->height,
1305 : edgomez 1.51 dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
1306 :     frame->output.csp, dec->interlacing);
1307 :    
1308 :     if (stats) {
1309 :     stats->type = coding2type(coding_type);
1310 :     stats->data.vop.time_base = (int)dec->time_base;
1311 :     stats->data.vop.time_increment = 0; /* XXX: todo */
1312 :     stats->data.vop.qscale_stride = dec->mb_width;
1313 :     stats->data.vop.qscale = dec->qscale;
1314 :     if (stats->data.vop.qscale != NULL && mbs != NULL) {
1315 :     int i;
1316 :     for (i = 0; i < dec->mb_width*dec->mb_height; i++)
1317 :     stats->data.vop.qscale[i] = mbs[i].quant;
1318 :     } else
1319 :     stats->data.vop.qscale = NULL;
1320 :     }
1321 : edgomez 1.47 }
1322 :    
1323 :     int
1324 :     decoder_decode(DECODER * dec,
1325 : edgomez 1.51 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
1326 : edgomez 1.47 {
1327 :    
1328 :     Bitstream bs;
1329 :     uint32_t rounding;
1330 : edgomez 1.51 uint32_t quant = 2;
1331 : edgomez 1.47 uint32_t fcode_forward;
1332 :     uint32_t fcode_backward;
1333 :     uint32_t intra_dc_threshold;
1334 :     WARPPOINTS gmc_warp;
1335 : edgomez 1.51 int coding_type;
1336 :     int success, output, seen_something;
1337 :    
1338 :     if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */
1339 :     return XVID_ERR_VERSION;
1340 : edgomez 1.47
1341 :     start_global_timer();
1342 :    
1343 : edgomez 1.51 dec->low_delay_default = (frame->general & XVID_LOWDELAY);
1344 :     if ((frame->general & XVID_DISCONTINUITY))
1345 : edgomez 1.47 dec->frames = 0;
1346 : edgomez 1.51 dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
1347 : edgomez 1.47
1348 : edgomez 1.51 if (frame->length < 0) { /* decoder flush */
1349 :     int ret;
1350 :     /* if not decoding "low_delay/packed", and this isn't low_delay and
1351 :     we have a reference frame, then outout the reference frame */
1352 :     if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
1353 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1354 :     dec->frames = 0;
1355 :     ret = 0;
1356 :     } else {
1357 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1358 :     ret = XVID_ERR_END;
1359 : edgomez 1.47 }
1360 :    
1361 :     emms();
1362 :     stop_global_timer();
1363 : edgomez 1.51 return ret;
1364 : edgomez 1.47 }
1365 : edgomez 1.19
1366 : Isibaar 1.1 BitstreamInit(&bs, frame->bitstream, frame->length);
1367 :    
1368 : edgomez 1.48 /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
1369 : edgomez 1.47 if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
1370 :     {
1371 :     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
1372 : edgomez 1.51 (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
1373 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1374 : edgomez 1.47 emms();
1375 : edgomez 1.51 return 1; /* one byte consumed */
1376 : edgomez 1.47 }
1377 :    
1378 : edgomez 1.51 success = 0;
1379 :     output = 0;
1380 :     seen_something = 0;
1381 :    
1382 : edgomez 1.47 repeat:
1383 :    
1384 : syskin 1.68 coding_type = BitstreamReadHeaders(&bs, dec, &rounding,
1385 : edgomez 1.47 &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
1386 :    
1387 : edgomez 1.51 DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n",
1388 :     coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
1389 : Isibaar 1.1
1390 : edgomez 1.51 if (coding_type == -1) { /* nothing */
1391 : edgomez 1.47 if (success) goto done;
1392 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1393 : edgomez 1.47 emms();
1394 : edgomez 1.51 return BitstreamPos(&bs)/8;
1395 : Isibaar 1.1 }
1396 :    
1397 : edgomez 1.51 if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
1398 :    
1399 :     if (coding_type == -3)
1400 : edgomez 1.47 decoder_resize(dec);
1401 : edgomez 1.51
1402 :     if (stats) {
1403 :     stats->type = XVID_TYPE_VOL;
1404 : edgomez 1.47 stats->data.vol.general = 0;
1405 : edgomez 1.51 /*XXX: if (dec->interlacing)
1406 :     stats->data.vol.general |= ++INTERLACING; */
1407 : edgomez 1.47 stats->data.vol.width = dec->width;
1408 :     stats->data.vol.height = dec->height;
1409 : edgomez 1.51 stats->data.vol.par = dec->aspect_ratio;
1410 : edgomez 1.47 stats->data.vol.par_width = dec->par_width;
1411 :     stats->data.vol.par_height = dec->par_height;
1412 :     emms();
1413 : edgomez 1.51 return BitstreamPos(&bs)/8; /* number of bytes consumed */
1414 : edgomez 1.47 }
1415 : syskin 1.56 goto repeat;
1416 :     }
1417 :    
1418 :     if(dec->frames == 0 && coding_type != I_VOP) {
1419 :     /* 1st frame is not an i-vop */
1420 : edgomez 1.47 goto repeat;
1421 : edgomez 1.51 }
1422 : edgomez 1.47
1423 : 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 */
1424 : edgomez 1.47
1425 :     /* packed_mode: special-N_VOP treament */
1426 : edgomez 1.51 if (dec->packed_mode && coding_type == N_VOP) {
1427 :     if (dec->low_delay_default && dec->frames > 0) {
1428 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1429 : edgomez 1.47 output = 1;
1430 :     }
1431 :     /* ignore otherwise */
1432 : edgomez 1.51 } else if (coding_type != B_VOP) {
1433 :     switch(coding_type) {
1434 : edgomez 1.47 case I_VOP :
1435 : syskin 1.68 decoder_iframe(dec, &bs, quant, intra_dc_threshold);
1436 : edgomez 1.47 break;
1437 :     case P_VOP :
1438 : syskin 1.68 decoder_pframe(dec, &bs, rounding, quant,
1439 : edgomez 1.47 fcode_forward, intra_dc_threshold, NULL);
1440 :     break;
1441 :     case S_VOP :
1442 : syskin 1.68 decoder_pframe(dec, &bs, rounding, quant,
1443 : edgomez 1.47 fcode_forward, intra_dc_threshold, &gmc_warp);
1444 :     break;
1445 :     case N_VOP :
1446 : edgomez 1.51 /* XXX: not_coded vops are not used for forward prediction */
1447 :     /* we should not swap(last_mbs,mbs) */
1448 : edgomez 1.47 image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
1449 : edgomez 1.51 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */
1450 : edgomez 1.47 break;
1451 :     }
1452 :    
1453 :     /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
1454 : edgomez 1.51 if (!(dec->low_delay_default && dec->packed_mode)) {
1455 :     if (dec->low_delay) {
1456 :     decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1457 : edgomez 1.47 output = 1;
1458 : edgomez 1.51 } else if (dec->frames > 0) { /* is the reference frame valid? */
1459 : edgomez 1.47 /* output the reference frame */
1460 : edgomez 1.51 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1461 : edgomez 1.47 output = 1;
1462 :     }
1463 :     }
1464 : Isibaar 1.18
1465 : chenm001 1.14 image_swap(&dec->refn[0], &dec->refn[1]);
1466 : edgomez 1.57 dec->is_edged[1] = dec->is_edged[0];
1467 : chenm001 1.14 image_swap(&dec->cur, &dec->refn[0]);
1468 : edgomez 1.57 dec->is_edged[0] = 0;
1469 : edgomez 1.51 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
1470 :     dec->last_coding_type = coding_type;
1471 : edgomez 1.47
1472 :     dec->frames++;
1473 :     seen_something = 1;
1474 : chenm001 1.31
1475 : edgomez 1.51 } else { /* B_VOP */
1476 : edgomez 1.47
1477 : edgomez 1.51 if (dec->low_delay) {
1478 :     DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
1479 : syskin 1.60 dec->low_delay = 0;
1480 : edgomez 1.47 }
1481 :    
1482 : edgomez 1.51 if (dec->frames < 2) {
1483 : edgomez 1.47 /* attemping to decode a bvop without atleast 2 reference frames */
1484 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1485 :     "broken b-frame, mising ref frames");
1486 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1487 :     } else if (dec->time_pp <= dec->time_bp) {
1488 :     /* this occurs when dx50_bvop_compatibility==0 sequences are
1489 : edgomez 1.47 decoded in vfw. */
1490 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1491 :     "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
1492 : edgomez 1.51 if (stats) stats->type = XVID_TYPE_NOTHING;
1493 :     } else {
1494 : edgomez 1.47 decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
1495 : edgomez 1.51 decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1496 : edgomez 1.47 }
1497 :    
1498 :     output = 1;
1499 :     dec->frames++;
1500 : chenm001 1.14 }
1501 :    
1502 : edgomez 1.51 #if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */
1503 :     BitstreamByteAlign(&bs);
1504 :     #endif
1505 : edgomez 1.47
1506 :     /* low_delay_default mode: repeat in packed_mode */
1507 : edgomez 1.51 if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
1508 : edgomez 1.47 success = 1;
1509 :     goto repeat;
1510 :     }
1511 :    
1512 :     done :
1513 :    
1514 :     /* low_delay_default mode: if we've gotten here without outputting anything,
1515 :     then output the recently decoded frame, or print an error message */
1516 : edgomez 1.51 if (dec->low_delay_default && output == 0) {
1517 :     if (dec->packed_mode && seen_something) {
1518 : edgomez 1.47 /* output the recently decoded frame */
1519 : edgomez 1.51 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1520 :     } else {
1521 : edgomez 1.47 image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
1522 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1523 :     "warning: nothing to output");
1524 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64,
1525 :     "bframe decoder lag");
1526 :    
1527 : edgomez 1.51 decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant);
1528 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1529 : edgomez 1.47 }
1530 :     }
1531 : Isibaar 1.50
1532 : Isibaar 1.1 emms();
1533 :     stop_global_timer();
1534 :    
1535 : edgomez 1.51 return (BitstreamPos(&bs) + 7) / 8; /* number of bytes consumed */
1536 : Isibaar 1.1 }

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