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

Annotation of /trunk/xvidcore/src/decoder.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1772 - (view) (download)

1 : edgomez 866 /*****************************************************************************
2 : Isibaar 3 *
3 : edgomez 146 * XVID MPEG-4 VIDEO CODEC
4 : edgomez 866 * - Decoder Module -
5 : Isibaar 3 *
6 : edgomez 1382 * Copyright(C) 2002 MinChen <chenm001@163.com>
7 : suxen_drol 1397 * 2002-2004 Peter Ross <pross@xvid.org>
8 : edgomez 605 *
9 : edgomez 1382 * This program is free software ; you can redistribute it and/or modify
10 : edgomez 851 * it under the terms of the GNU General Public License as published by
11 : edgomez 1382 * the Free Software Foundation ; either version 2 of the License, or
12 : edgomez 146 * (at your option) any later version.
13 : Isibaar 3 *
14 : edgomez 146 * This program is distributed in the hope that it will be useful,
15 : edgomez 1382 * but WITHOUT ANY WARRANTY ; without even the implied warranty of
16 : edgomez 146 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 :     * GNU General Public License for more details.
18 : Isibaar 3 *
19 : edgomez 146 * You should have received a copy of the GNU General Public License
20 : edgomez 1382 * along with this program ; if not, write to the Free Software
21 : edgomez 146 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 : Isibaar 3 *
23 : Skal 1772 * $Id: decoder.c,v 1.80 2007-04-16 19:01:28 Skal Exp $
24 : Isibaar 3 *
25 : edgomez 866 ****************************************************************************/
26 : Isibaar 3
27 : edgomez 851 #include <stdio.h>
28 : Isibaar 3 #include <stdlib.h>
29 : edgomez 146 #include <string.h>
30 : Isibaar 3
31 : chenm001 290 #ifdef BFRAMES_DEC_DEBUG
32 : suxen_drol 1632 #define BFRAMES_DEC
33 : chenm001 290 #endif
34 :    
35 : Isibaar 3 #include "xvid.h"
36 :     #include "portab.h"
37 : edgomez 851 #include "global.h"
38 : Isibaar 3
39 :     #include "decoder.h"
40 :     #include "bitstream/bitstream.h"
41 :     #include "bitstream/mbcoding.h"
42 :    
43 : edgomez 1382 #include "quant/quant.h"
44 :     #include "quant/quant_matrix.h"
45 : Isibaar 3 #include "dct/idct.h"
46 :     #include "dct/fdct.h"
47 :     #include "utils/mem_transfer.h"
48 :     #include "image/interpolate8x8.h"
49 : edgomez 851 #include "image/font.h"
50 : edgomez 1492 #include "image/qpel.h"
51 : Isibaar 3
52 :     #include "bitstream/mbcoding.h"
53 :     #include "prediction/mbprediction.h"
54 :     #include "utils/timer.h"
55 :     #include "utils/emms.h"
56 : edgomez 851 #include "motion/motion.h"
57 : edgomez 1382 #include "motion/gmc.h"
58 : Isibaar 3
59 :     #include "image/image.h"
60 :     #include "image/colorspace.h"
61 : edgomez 1382 #include "image/postprocessing.h"
62 : Isibaar 41 #include "utils/mem_align.h"
63 : Isibaar 3
64 : suxen_drol 1632 #define DIV2ROUND(n) (((n)>>1)|((n)&1))
65 :     #define DIV2(n) ((n)>>1)
66 :     #define DIVUVMOV(n) (((n) >> 1) + roundtab_79[(n) & 0x3]) //
67 :    
68 : edgomez 1382 static int
69 : edgomez 851 decoder_resize(DECODER * dec)
70 : Isibaar 3 {
71 : edgomez 851 /* free existing */
72 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
73 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
74 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
75 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
76 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
77 : Isibaar 3
78 : edgomez 851 image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
79 : Isibaar 3
80 : Skal 1616 image_null(&dec->cur);
81 :     image_null(&dec->refn[0]);
82 :     image_null(&dec->refn[1]);
83 :     image_null(&dec->tmp);
84 :     image_null(&dec->qtmp);
85 :     image_null(&dec->gmc);
86 : edgomez 851
87 : Skal 1616
88 :     xvid_free(dec->last_mbs);
89 :     xvid_free(dec->mbs);
90 :     xvid_free(dec->qscale);
91 :     dec->last_mbs = NULL;
92 :     dec->mbs = NULL;
93 :     dec->qscale = NULL;
94 :    
95 : edgomez 851 /* realloc */
96 : Isibaar 3 dec->mb_width = (dec->width + 15) / 16;
97 :     dec->mb_height = (dec->height + 15) / 16;
98 :    
99 :     dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE;
100 :     dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE;
101 : edgomez 195
102 : Skal 1616 if ( image_create(&dec->cur, dec->edged_width, dec->edged_height)
103 :     || image_create(&dec->refn[0], dec->edged_width, dec->edged_height)
104 :     || image_create(&dec->refn[1], dec->edged_width, dec->edged_height) /* Support B-frame to reference last 2 frame */
105 :     || image_create(&dec->tmp, dec->edged_width, dec->edged_height)
106 :     || image_create(&dec->qtmp, dec->edged_width, dec->edged_height)
107 :     || image_create(&dec->gmc, dec->edged_width, dec->edged_height) )
108 :     goto memory_error;
109 : Isibaar 3
110 : edgomez 195 dec->mbs =
111 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
112 :     CACHE_LINE);
113 : Skal 1616 if (dec->mbs == NULL)
114 :     goto memory_error;
115 : Isibaar 208 memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
116 :    
117 : edgomez 866 /* For skip MB flag */
118 : edgomez 195 dec->last_mbs =
119 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
120 :     CACHE_LINE);
121 : Skal 1616 if (dec->last_mbs == NULL)
122 :     goto memory_error;
123 : Isibaar 208 memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
124 :    
125 : edgomez 1382 /* nothing happens if that fails */
126 :     dec->qscale =
127 :     xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE);
128 :    
129 :     if (dec->qscale)
130 :     memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height);
131 :    
132 :     return 0;
133 : Skal 1616
134 :     memory_error:
135 :     /* Most structures were deallocated / nullifieded, so it should be safe */
136 :     /* decoder_destroy(dec) minus the write_timer */
137 :     xvid_free(dec->mbs);
138 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
139 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
140 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
141 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
142 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
143 :    
144 :     xvid_free(dec);
145 :     return XVID_ERR_MEMORY;
146 : edgomez 851 }
147 :    
148 :    
149 :     int
150 : edgomez 1382 decoder_create(xvid_dec_create_t * create)
151 : edgomez 851 {
152 : suxen_drol 1632 DECODER *dec;
153 : edgomez 851
154 : suxen_drol 1632 if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */
155 :     return XVID_ERR_VERSION;
156 : edgomez 1382
157 : suxen_drol 1632 dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
158 :     if (dec == NULL) {
159 :     return XVID_ERR_MEMORY;
160 :     }
161 : edgomez 1382
162 : suxen_drol 1632 memset(dec, 0, sizeof(DECODER));
163 : edgomez 851
164 : suxen_drol 1632 dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE);
165 :     if (dec->mpeg_quant_matrices == NULL) {
166 :     xvid_free(dec);
167 :     return XVID_ERR_MEMORY;
168 :     }
169 : edgomez 851
170 : suxen_drol 1632 create->handle = dec;
171 : edgomez 851
172 : suxen_drol 1632 dec->width = create->width;
173 :     dec->height = create->height;
174 : edgomez 1382
175 : suxen_drol 1632 image_null(&dec->cur);
176 :     image_null(&dec->refn[0]);
177 :     image_null(&dec->refn[1]);
178 :     image_null(&dec->tmp);
179 :     image_null(&dec->qtmp);
180 : edgomez 851
181 : suxen_drol 1632 /* image based GMC */
182 :     image_null(&dec->gmc);
183 : edgomez 851
184 : suxen_drol 1632 dec->mbs = NULL;
185 :     dec->last_mbs = NULL;
186 :     dec->qscale = NULL;
187 : edgomez 851
188 : suxen_drol 1632 init_timer();
189 :     init_postproc(&dec->postproc);
190 :     init_mpeg_matrix(dec->mpeg_quant_matrices);
191 : Isibaar 3
192 : suxen_drol 1632 /* For B-frame support (used to save reference frame's time */
193 :     dec->frames = 0;
194 :     dec->time = dec->time_base = dec->last_time_base = 0;
195 :     dec->low_delay = 0;
196 :     dec->packed_mode = 0;
197 :     dec->time_inc_resolution = 1; /* until VOL header says otherwise */
198 : Skal 1696 dec->ver_id = 1;
199 : edgomez 195
200 : Isibaar 1671 dec->bs_version = 0xffff; /* Initialize to very high value -> assume bugfree stream */
201 :    
202 : suxen_drol 1632 dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
203 : edgomez 851
204 : suxen_drol 1632 if (dec->fixed_dimensions)
205 :     return decoder_resize(dec);
206 :     else
207 :     return 0;
208 : Isibaar 3 }
209 :    
210 :    
211 : edgomez 195 int
212 :     decoder_destroy(DECODER * dec)
213 : Isibaar 3 {
214 : suxen_drol 1632 xvid_free(dec->last_mbs);
215 :     xvid_free(dec->mbs);
216 :     xvid_free(dec->qscale);
217 : edgomez 851
218 : suxen_drol 1632 /* image based GMC */
219 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
220 : edgomez 851
221 : suxen_drol 1632 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
222 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
223 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
224 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
225 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
226 :     xvid_free(dec->mpeg_quant_matrices);
227 :     xvid_free(dec);
228 : Isibaar 3
229 : suxen_drol 1632 write_timer();
230 :     return 0;
231 : Isibaar 3 }
232 :    
233 : edgomez 195 static const int32_t dquant_table[4] = {
234 : suxen_drol 1632 -1, -2, 1, 2
235 : Isibaar 3 };
236 :    
237 : edgomez 866 /* decode an intra macroblock */
238 : edgomez 1382 static void
239 : edgomez 195 decoder_mbintra(DECODER * dec,
240 : suxen_drol 1632 MACROBLOCK * pMB,
241 :     const uint32_t x_pos,
242 :     const uint32_t y_pos,
243 :     const uint32_t acpred_flag,
244 :     const uint32_t cbp,
245 :     Bitstream * bs,
246 :     const uint32_t quant,
247 :     const uint32_t intra_dc_threshold,
248 :     const unsigned int bound)
249 : Isibaar 3 {
250 : edgomez 78
251 : suxen_drol 1632 DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
252 :     DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
253 : edgomez 78
254 : suxen_drol 1632 uint32_t stride = dec->edged_width;
255 :     uint32_t stride2 = stride / 2;
256 :     uint32_t next_block = stride * 8;
257 :     uint32_t i;
258 :     uint32_t iQuant = pMB->quant;
259 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
260 : h 69
261 : suxen_drol 1632 pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
262 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
263 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
264 : h 69
265 : suxen_drol 1632 memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */
266 : h 69
267 : suxen_drol 1632 for (i = 0; i < 6; i++) {
268 :     uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);
269 :     int16_t predictors[8];
270 :     int start_coeff;
271 : Isibaar 3
272 : suxen_drol 1632 start_timer();
273 :     predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64],
274 :     iQuant, iDcScaler, predictors, bound);
275 :     if (!acpred_flag) {
276 :     pMB->acpred_directions[i] = 0;
277 :     }
278 :     stop_prediction_timer();
279 : Isibaar 3
280 : suxen_drol 1632 if (quant < intra_dc_threshold) {
281 :     int dc_size;
282 :     int dc_dif;
283 : Isibaar 3
284 : suxen_drol 1632 dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs);
285 :     dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0;
286 : Isibaar 3
287 : suxen_drol 1632 if (dc_size > 8) {
288 :     BitstreamSkip(bs, 1); /* marker */
289 :     }
290 : edgomez 195
291 : suxen_drol 1632 block[i * 64 + 0] = dc_dif;
292 :     start_coeff = 1;
293 : chenm001 272
294 : suxen_drol 1632 DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif);
295 :     } else {
296 :     start_coeff = 0;
297 :     }
298 : Isibaar 3
299 : suxen_drol 1632 start_timer();
300 :     if (cbp & (1 << (5 - i))) /* coded */
301 :     {
302 :     int direction = dec->alternate_vertical_scan ?
303 :     2 : pMB->acpred_directions[i];
304 : edgomez 851
305 : suxen_drol 1632 get_intra_block(bs, &block[i * 64], direction, start_coeff);
306 :     }
307 :     stop_coding_timer();
308 : Isibaar 3
309 : suxen_drol 1632 start_timer();
310 :     add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version);
311 :     stop_prediction_timer();
312 : Isibaar 3
313 : suxen_drol 1632 start_timer();
314 :     if (dec->quant_type == 0) {
315 :     dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
316 :     } else {
317 :     dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
318 :     }
319 :     stop_iquant_timer();
320 : Isibaar 3
321 : suxen_drol 1632 start_timer();
322 : suxen_drol 1653 idct((short * const)&data[i * 64]);
323 : suxen_drol 1632 stop_idct_timer();
324 : edgomez 851
325 : suxen_drol 1632 }
326 : Isibaar 3
327 : suxen_drol 1632 if (dec->interlacing && pMB->field_dct) {
328 :     next_block = stride;
329 :     stride *= 2;
330 :     }
331 : h 69
332 : suxen_drol 1632 start_timer();
333 :     transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
334 :     transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
335 :     transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
336 :     transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
337 :     transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
338 :     transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
339 :     stop_transfer_timer();
340 : Isibaar 3 }
341 :    
342 : edgomez 1382 static void
343 :     decoder_mb_decode(DECODER * dec,
344 : suxen_drol 1632 const uint32_t cbp,
345 :     Bitstream * bs,
346 :     uint8_t * pY_Cur,
347 :     uint8_t * pU_Cur,
348 :     uint8_t * pV_Cur,
349 :     const MACROBLOCK * pMB)
350 : edgomez 1382 {
351 : suxen_drol 1632 DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE);
352 : Isibaar 3
353 : suxen_drol 1632 int stride = dec->edged_width;
354 :     int i;
355 :     const uint32_t iQuant = pMB->quant;
356 :     const int direction = dec->alternate_vertical_scan ? 2 : 0;
357 :     typedef void (*get_inter_block_function_t)(
358 :     Bitstream * bs,
359 :     int16_t * block,
360 :     int direction,
361 :     const int quant,
362 :     const uint16_t *matrix);
363 :     typedef void (*add_residual_function_t)(
364 :     uint8_t *predicted_block,
365 :     const int16_t *residual,
366 :     int stride);
367 : Isibaar 3
368 : suxen_drol 1632 const get_inter_block_function_t get_inter_block = (dec->quant_type == 0)
369 :     ? (get_inter_block_function_t)get_inter_block_h263
370 :     : (get_inter_block_function_t)get_inter_block_mpeg;
371 : edgomez 1524
372 : suxen_drol 1632 uint8_t *dst[6];
373 :     int strides[6];
374 : edgomez 1486
375 : edgomez 1524
376 : suxen_drol 1632 if (dec->interlacing && pMB->field_dct) {
377 :     dst[0] = pY_Cur;
378 :     dst[1] = pY_Cur + 8;
379 :     dst[2] = pY_Cur + stride;
380 :     dst[3] = dst[2] + 8;
381 :     dst[4] = pU_Cur;
382 :     dst[5] = pV_Cur;
383 :     strides[0] = strides[1] = strides[2] = strides[3] = stride*2;
384 :     strides[4] = stride/2;
385 :     strides[5] = stride/2;
386 :     } else {
387 :     dst[0] = pY_Cur;
388 :     dst[1] = pY_Cur + 8;
389 :     dst[2] = pY_Cur + 8*stride;
390 :     dst[3] = dst[2] + 8;
391 :     dst[4] = pU_Cur;
392 :     dst[5] = pV_Cur;
393 :     strides[0] = strides[1] = strides[2] = strides[3] = stride;
394 :     strides[4] = stride/2;
395 :     strides[5] = stride/2;
396 :     }
397 : Isibaar 3
398 : suxen_drol 1632 for (i = 0; i < 6; i++) {
399 :     /* Process only coded blocks */
400 :     if (cbp & (1 << (5 - i))) {
401 : edgomez 1382
402 : suxen_drol 1632 /* Clear the block */
403 :     memset(&data[0], 0, 64*sizeof(int16_t));
404 : edgomez 1382
405 : suxen_drol 1632 /* Decode coeffs and dequantize on the fly */
406 :     start_timer();
407 :     get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices));
408 :     stop_coding_timer();
409 : edgomez 1524
410 : suxen_drol 1632 /* iDCT */
411 :     start_timer();
412 : suxen_drol 1653 idct((short * const)&data[0]);
413 : suxen_drol 1632 stop_idct_timer();
414 :    
415 :     /* Add this residual to the predicted block */
416 :     start_timer();
417 :     transfer_16to8add(dst[i], &data[0], strides[i]);
418 :     stop_transfer_timer();
419 :     }
420 :     }
421 : edgomez 1382 }
422 :    
423 : edgomez 1533 static void __inline
424 : syskin 1532 validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec)
425 :     {
426 : suxen_drol 1632 /* clip a vector to valid range
427 :     prevents crashes if bitstream is broken
428 :     */
429 :     int shift = 5 + dec->quarterpel;
430 :     int xborder_high = (int)(dec->mb_width - x_pos) << shift;
431 :     int xborder_low = (-(int)x_pos-1) << shift;
432 :     int yborder_high = (int)(dec->mb_height - y_pos) << shift;
433 :     int yborder_low = (-(int)y_pos-1) << shift;
434 : syskin 1532
435 : edgomez 1533 #define CHECK_MV(mv) \
436 : suxen_drol 1632 do { \
437 :     if ((mv).x > xborder_high) { \
438 :     DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \
439 :     (mv).x = xborder_high; \
440 :     } else if ((mv).x < xborder_low) { \
441 :     DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \
442 :     (mv).x = xborder_low; \
443 :     } \
444 :     if ((mv).y > yborder_high) { \
445 :     DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \
446 :     (mv).y = yborder_high; \
447 :     } else if ((mv).y < yborder_low) { \
448 :     DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \
449 :     (mv).y = yborder_low; \
450 :     } \
451 :     } while (0)
452 : syskin 1532
453 : suxen_drol 1632 CHECK_MV(mv[0]);
454 :     CHECK_MV(mv[1]);
455 :     CHECK_MV(mv[2]);
456 :     CHECK_MV(mv[3]);
457 : syskin 1532 }
458 :    
459 : Isibaar 1628 /* Up to this version, chroma rounding was wrong with qpel.
460 :     * So we try to be backward compatible to avoid artifacts */
461 :     #define BS_VERSION_BUGGY_CHROMA_ROUNDING 1
462 :    
463 : edgomez 866 /* decode an inter macroblock */
464 : edgomez 1382 static void
465 : edgomez 195 decoder_mbinter(DECODER * dec,
466 : suxen_drol 1632 const MACROBLOCK * pMB,
467 :     const uint32_t x_pos,
468 :     const uint32_t y_pos,
469 :     const uint32_t cbp,
470 :     Bitstream * bs,
471 :     const uint32_t rounding,
472 : Isibaar 1669 const int ref,
473 :     const int bvop)
474 : Isibaar 3 {
475 : suxen_drol 1632 uint32_t stride = dec->edged_width;
476 :     uint32_t stride2 = stride / 2;
477 :     uint32_t i;
478 : edgomez 1382
479 : suxen_drol 1632 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
480 : edgomez 851
481 : suxen_drol 1632 int uv_dx, uv_dy;
482 :     VECTOR mv[4]; /* local copy of mvs */
483 : Isibaar 3
484 : suxen_drol 1632 pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
485 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
486 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
487 :     for (i = 0; i < 4; i++)
488 :     mv[i] = pMB->mvs[i];
489 : Isibaar 3
490 : suxen_drol 1632 validate_vector(mv, x_pos, y_pos, dec);
491 : syskin 1420
492 : suxen_drol 1632 start_timer();
493 : edgomez 1382
494 : Isibaar 1669 if ((pMB->mode != MODE_INTER4V) || (bvop)) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
495 : edgomez 1382
496 : suxen_drol 1632 uv_dx = mv[0].x;
497 :     uv_dy = mv[0].y;
498 :     if (dec->quarterpel) {
499 : Isibaar 1628 if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
500 : Isibaar 1669 uv_dx = (uv_dx>>1) | (uv_dx&1);
501 : Isibaar 1628 uv_dy = (uv_dy>>1) | (uv_dy&1);
502 :     }
503 :     else {
504 : suxen_drol 1632 uv_dx /= 2;
505 :     uv_dy /= 2;
506 :     }
507 :     }
508 :     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
509 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
510 : edgomez 851
511 : suxen_drol 1632 if (dec->quarterpel)
512 :     interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
513 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
514 :     mv[0].x, mv[0].y, stride, rounding);
515 :     else
516 :     interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
517 :     mv[0].x, mv[0].y, stride, rounding);
518 : edgomez 851
519 : suxen_drol 1632 } else { /* MODE_INTER4V */
520 : Isibaar 3
521 : suxen_drol 1632 if(dec->quarterpel) {
522 : Isibaar 1628 if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
523 :     int z;
524 :     uv_dx = 0; uv_dy = 0;
525 :     for (z = 0; z < 4; z++) {
526 :     uv_dx += ((mv[z].x>>1) | (mv[z].x&1));
527 :     uv_dy += ((mv[z].y>>1) | (mv[z].y&1));
528 :     }
529 :     }
530 :     else {
531 : suxen_drol 1632 uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
532 :     uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
533 :     }
534 :     } else {
535 :     uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
536 :     uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
537 :     }
538 : Isibaar 333
539 : suxen_drol 1632 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
540 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
541 : Isibaar 333
542 : suxen_drol 1632 if (dec->quarterpel) {
543 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
544 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
545 :     mv[0].x, mv[0].y, stride, rounding);
546 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
547 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
548 :     mv[1].x, mv[1].y, stride, rounding);
549 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
550 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
551 :     mv[2].x, mv[2].y, stride, rounding);
552 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
553 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
554 :     mv[3].x, mv[3].y, stride, rounding);
555 :     } else {
556 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
557 :     mv[0].x, mv[0].y, stride, rounding);
558 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
559 :     mv[1].x, mv[1].y, stride, rounding);
560 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
561 :     mv[2].x, mv[2].y, stride, rounding);
562 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
563 :     mv[3].x, mv[3].y, stride, rounding);
564 :     }
565 :     }
566 : Isibaar 3
567 : suxen_drol 1632 /* chroma */
568 :     interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
569 :     uv_dx, uv_dy, stride2, rounding);
570 :     interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
571 :     uv_dx, uv_dy, stride2, rounding);
572 : h 69
573 : suxen_drol 1632 stop_comp_timer();
574 : h 69
575 : suxen_drol 1632 if (cbp)
576 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
577 : edgomez 851 }
578 :    
579 : suxen_drol 1632 /* decode an inter macroblock in field mode */
580 : edgomez 851 static void
581 : suxen_drol 1632 decoder_mbinter_field(DECODER * dec,
582 :     const MACROBLOCK * pMB,
583 :     const uint32_t x_pos,
584 :     const uint32_t y_pos,
585 :     const uint32_t cbp,
586 :     Bitstream * bs,
587 :     const uint32_t rounding,
588 : Isibaar 1669 const int ref,
589 :     const int bvop)
590 : suxen_drol 1632 {
591 :     uint32_t stride = dec->edged_width;
592 :     uint32_t stride2 = stride / 2;
593 :    
594 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
595 :    
596 :     int uvtop_dx, uvtop_dy;
597 :     int uvbot_dx, uvbot_dy;
598 :     VECTOR mv[4]; /* local copy of mvs */
599 :    
600 :     /* Get pointer to memory areas */
601 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
602 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
603 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
604 :    
605 :     mv[0] = pMB->mvs[0];
606 :     mv[1] = pMB->mvs[1];
607 :     memset(&mv[2],0,2*sizeof(VECTOR));
608 :    
609 :     validate_vector(mv, x_pos, y_pos, dec);
610 :    
611 :     start_timer();
612 :    
613 : Isibaar 1669 if((pMB->mode!=MODE_INTER4V) || (bvop)) /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
614 : suxen_drol 1632 {
615 :     /* Prepare top field vector */
616 :     uvtop_dx = DIV2ROUND(mv[0].x);
617 :     uvtop_dy = DIV2ROUND(mv[0].y);
618 :    
619 :     /* Prepare bottom field vector */
620 :     uvbot_dx = DIV2ROUND(mv[1].x);
621 :     uvbot_dy = DIV2ROUND(mv[1].y);
622 :    
623 :     if(dec->quarterpel)
624 :     {
625 :     /* NOT supported */
626 :     }
627 :     else
628 :     {
629 :     /* Interpolate top field left part(we use double stride for every 2nd line) */
630 :     interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride,
631 :     16*x_pos,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding);
632 :     /* top field right part */
633 :     interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride,
634 :     16*x_pos+8,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding);
635 :    
636 :     /* Interpolate bottom field left part(we use double stride for every 2nd line) */
637 :     interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride,
638 :     16*x_pos,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding);
639 :     /* Bottom field right part */
640 :     interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride,
641 :     16*x_pos+8,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding);
642 :    
643 :     /* Interpolate field1 U */
644 :     interpolate8x4_switch(dec->cur.u,dec->refn[ref].u+pMB->field_for_top*stride2,
645 :     8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding);
646 :    
647 :     /* Interpolate field1 V */
648 :     interpolate8x4_switch(dec->cur.v,dec->refn[ref].v+pMB->field_for_top*stride2,
649 :     8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding);
650 :    
651 :     /* Interpolate field2 U */
652 :     interpolate8x4_switch(dec->cur.u+stride2,dec->refn[ref].u+pMB->field_for_bot*stride2,
653 :     8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding);
654 :    
655 :     /* Interpolate field2 V */
656 :     interpolate8x4_switch(dec->cur.v+stride2,dec->refn[ref].v+pMB->field_for_bot*stride2,
657 :     8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding);
658 :     }
659 :     }
660 :     else
661 :     {
662 :     /* We don't expect 4 motion vectors in interlaced mode */
663 :     }
664 :    
665 :     stop_comp_timer();
666 :    
667 :     /* Must add error correction? */
668 :     if(cbp)
669 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
670 :     }
671 :    
672 :     static void
673 : edgomez 851 decoder_mbgmc(DECODER * dec,
674 : suxen_drol 1632 MACROBLOCK * const pMB,
675 :     const uint32_t x_pos,
676 :     const uint32_t y_pos,
677 :     const uint32_t fcode,
678 :     const uint32_t cbp,
679 :     Bitstream * bs,
680 :     const uint32_t rounding)
681 : edgomez 851 {
682 : suxen_drol 1632 const uint32_t stride = dec->edged_width;
683 :     const uint32_t stride2 = stride / 2;
684 : edgomez 1382
685 : suxen_drol 1632 uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
686 :     uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
687 :     uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
688 : edgomez 851
689 : suxen_drol 1632 NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
690 : edgomez 1382
691 : suxen_drol 1632 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
692 : edgomez 851
693 : suxen_drol 1632 start_timer();
694 : edgomez 1382
695 : edgomez 851 /* this is where the calculations are done */
696 :    
697 : suxen_drol 1632 gmc_data->predict_16x16(gmc_data,
698 :     dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
699 :     stride, stride, x_pos, y_pos, rounding);
700 : edgomez 851
701 : suxen_drol 1632 gmc_data->predict_8x8(gmc_data,
702 :     dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
703 :     dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
704 :     stride2, stride2, x_pos, y_pos, rounding);
705 : edgomez 851
706 : suxen_drol 1632 gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
707 : edgomez 851
708 : suxen_drol 1632 pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
709 :     pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
710 : edgomez 851
711 : suxen_drol 1632 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
712 : edgomez 851
713 : suxen_drol 1632 stop_transfer_timer();
714 : edgomez 851
715 : suxen_drol 1632 if (cbp)
716 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
717 : edgomez 851
718 : Isibaar 3 }
719 :    
720 :    
721 : edgomez 1382 static void
722 : edgomez 195 decoder_iframe(DECODER * dec,
723 : suxen_drol 1632 Bitstream * bs,
724 :     int quant,
725 :     int intra_dc_threshold)
726 : Isibaar 3 {
727 : suxen_drol 1632 uint32_t bound;
728 :     uint32_t x, y;
729 :     const uint32_t mb_width = dec->mb_width;
730 :     const uint32_t mb_height = dec->mb_height;
731 : edgomez 1382
732 : suxen_drol 1632 bound = 0;
733 : chenm001 272
734 : suxen_drol 1632 for (y = 0; y < mb_height; y++) {
735 :     for (x = 0; x < mb_width; x++) {
736 :     MACROBLOCK *mb;
737 :     uint32_t mcbpc;
738 :     uint32_t cbpc;
739 :     uint32_t acpred_flag;
740 :     uint32_t cbpy;
741 :     uint32_t cbp;
742 : Isibaar 3
743 : suxen_drol 1632 while (BitstreamShowBits(bs, 9) == 1)
744 :     BitstreamSkip(bs, 9);
745 : chenm001 272
746 : suxen_drol 1632 if (check_resync_marker(bs, 0))
747 :     {
748 :     bound = read_video_packet_header(bs, dec, 0,
749 :     &quant, NULL, NULL, &intra_dc_threshold);
750 :     x = bound % mb_width;
751 :     y = bound / mb_width;
752 :     }
753 :     mb = &dec->mbs[y * dec->mb_width + x];
754 : chenm001 272
755 : suxen_drol 1632 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
756 : chenm001 272
757 : suxen_drol 1632 mcbpc = get_mcbpc_intra(bs);
758 :     mb->mode = mcbpc & 7;
759 :     cbpc = (mcbpc >> 4);
760 : Isibaar 3
761 : suxen_drol 1632 acpred_flag = BitstreamGetBit(bs);
762 : Isibaar 3
763 : suxen_drol 1632 cbpy = get_cbpy(bs, 1);
764 :     cbp = (cbpy << 2) | cbpc;
765 : Isibaar 3
766 : suxen_drol 1632 if (mb->mode == MODE_INTRA_Q) {
767 :     quant += dquant_table[BitstreamGetBits(bs, 2)];
768 :     if (quant > 31) {
769 :     quant = 31;
770 :     } else if (quant < 1) {
771 :     quant = 1;
772 :     }
773 :     }
774 :     mb->quant = quant;
775 :     mb->mvs[0].x = mb->mvs[0].y =
776 :     mb->mvs[1].x = mb->mvs[1].y =
777 :     mb->mvs[2].x = mb->mvs[2].y =
778 :     mb->mvs[3].x = mb->mvs[3].y =0;
779 : Isibaar 3
780 : suxen_drol 1632 if (dec->interlacing) {
781 :     mb->field_dct = BitstreamGetBit(bs);
782 :     DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
783 :     }
784 : h 69
785 : suxen_drol 1632 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
786 :     intra_dc_threshold, bound);
787 : edgomez 851
788 : suxen_drol 1632 }
789 :     if(dec->out_frm)
790 :     output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
791 :     }
792 : edgomez 78
793 : Isibaar 3 }
794 :    
795 :    
796 : edgomez 1382 static void
797 : edgomez 195 get_motion_vector(DECODER * dec,
798 : suxen_drol 1632 Bitstream * bs,
799 :     int x,
800 :     int y,
801 :     int k,
802 :     VECTOR * ret_mv,
803 :     int fcode,
804 :     const int bound)
805 : Isibaar 3 {
806 : edgomez 78
807 : suxen_drol 1632 const int scale_fac = 1 << (fcode - 1);
808 :     const int high = (32 * scale_fac) - 1;
809 :     const int low = ((-32) * scale_fac);
810 :     const int range = (64 * scale_fac);
811 : Isibaar 3
812 : suxen_drol 1632 const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
813 :     VECTOR mv;
814 : Isibaar 3
815 : suxen_drol 1632 mv.x = get_mv(bs, fcode);
816 :     mv.y = get_mv(bs, fcode);
817 : edgomez 195
818 : suxen_drol 1632 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);
819 : suxen_drol 252
820 : suxen_drol 1632 mv.x += pmv.x;
821 :     mv.y += pmv.y;
822 : chenm001 272
823 : suxen_drol 1632 if (mv.x < low) {
824 :     mv.x += range;
825 :     } else if (mv.x > high) {
826 :     mv.x -= range;
827 :     }
828 : Isibaar 3
829 : suxen_drol 1632 if (mv.y < low) {
830 :     mv.y += range;
831 :     } else if (mv.y > high) {
832 :     mv.y -= range;
833 :     }
834 : Isibaar 3
835 : suxen_drol 1632 ret_mv->x = mv.x;
836 :     ret_mv->y = mv.y;
837 : edgomez 1382 }
838 : edgomez 851
839 : suxen_drol 1632 /* We use this when decoder runs interlaced -> different prediction */
840 :    
841 :     static void get_motion_vector_interlaced(DECODER * dec,
842 :     Bitstream * bs,
843 :     int x,
844 :     int y,
845 :     int k,
846 :     MACROBLOCK *pMB,
847 :     int fcode,
848 :     const int bound)
849 :     {
850 :     const int scale_fac = 1 << (fcode - 1);
851 :     const int high = (32 * scale_fac) - 1;
852 :     const int low = ((-32) * scale_fac);
853 :     const int range = (64 * scale_fac);
854 :    
855 :     /* Get interlaced prediction */
856 :     const VECTOR pmv=get_pmv2_interlaced(dec->mbs,dec->mb_width,bound,x,y,k);
857 :     VECTOR mv,mvf1,mvf2;
858 :    
859 :     if(!pMB->field_pred)
860 :     {
861 :     mv.x = get_mv(bs,fcode);
862 :     mv.y = get_mv(bs,fcode);
863 :    
864 :     mv.x += pmv.x;
865 :     mv.y += pmv.y;
866 :    
867 :     if(mv.x<low) {
868 :     mv.x += range;
869 :     } else if (mv.x>high) {
870 :     mv.x-=range;
871 :     }
872 :    
873 :     if (mv.y < low) {
874 :     mv.y += range;
875 :     } else if (mv.y > high) {
876 :     mv.y -= range;
877 :     }
878 :    
879 :     pMB->mvs[0]=pMB->mvs[1]=pMB->mvs[2]=pMB->mvs[3]=mv;
880 :     }
881 :     else
882 :     {
883 :     mvf1.x = get_mv(bs, fcode);
884 :     mvf1.y = get_mv(bs, fcode);
885 :    
886 :     mvf1.x += pmv.x;
887 :     mvf1.y = 2*(mvf1.y+pmv.y/2); /* It's multiple of 2 */
888 :    
889 :     if (mvf1.x < low) {
890 :     mvf1.x += range;
891 :     } else if (mvf1.x > high) {
892 :     mvf1.x -= range;
893 :     }
894 :    
895 :     if (mvf1.y < low) {
896 :     mvf1.y += range;
897 :     } else if (mvf1.y > high) {
898 :     mvf1.y -= range;
899 :     }
900 :    
901 :     mvf2.x = get_mv(bs, fcode);
902 :     mvf2.y = get_mv(bs, fcode);
903 :    
904 :     mvf2.x += pmv.x;
905 :     mvf2.y = 2*(mvf2.y+pmv.y/2); /* It's multiple of 2 */
906 :    
907 :     if (mvf2.x < low) {
908 :     mvf2.x += range;
909 :     } else if (mvf2.x > high) {
910 :     mvf2.x -= range;
911 :     }
912 :    
913 :     if (mvf2.y < low) {
914 :     mvf2.y += range;
915 :     } else if (mvf2.y > high) {
916 :     mvf2.y -= range;
917 :     }
918 :    
919 :     pMB->mvs[0]=mvf1;
920 :     pMB->mvs[1]=mvf2;
921 :     pMB->mvs[2].x=pMB->mvs[3].x=0;
922 :     pMB->mvs[2].y=pMB->mvs[3].y=0;
923 :    
924 :     /* Calculate average for as it is field predicted */
925 :     pMB->mvs_avg.x=DIV2ROUND(pMB->mvs[0].x+pMB->mvs[1].x);
926 :     pMB->mvs_avg.y=DIV2ROUND(pMB->mvs[0].y+pMB->mvs[1].y);
927 :     }
928 :     }
929 :    
930 : edgomez 851 /* for P_VOP set gmc_warp to NULL */
931 : edgomez 1382 static void
932 : edgomez 195 decoder_pframe(DECODER * dec,
933 : suxen_drol 1632 Bitstream * bs,
934 :     int rounding,
935 :     int quant,
936 :     int fcode,
937 :     int intra_dc_threshold,
938 :     const WARPPOINTS *const gmc_warp)
939 : Isibaar 3 {
940 : suxen_drol 1632 uint32_t x, y;
941 :     uint32_t bound;
942 :     int cp_mb, st_mb;
943 :     const uint32_t mb_width = dec->mb_width;
944 :     const uint32_t mb_height = dec->mb_height;
945 : edgomez 1382
946 : suxen_drol 1632 if (!dec->is_edged[0]) {
947 :     start_timer();
948 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
949 :     dec->width, dec->height, dec->bs_version);
950 :     dec->is_edged[0] = 1;
951 :     stop_edges_timer();
952 :     }
953 : Isibaar 3
954 : suxen_drol 1632 if (gmc_warp) {
955 :     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
956 :     generate_GMCparameters( dec->sprite_warping_points,
957 :     dec->sprite_warping_accuracy, gmc_warp,
958 :     dec->width, dec->height, &dec->new_gmc_data);
959 : edgomez 851
960 : suxen_drol 1632 /* image warping is done block-based in decoder_mbgmc(), now */
961 :     }
962 : edgomez 851
963 : suxen_drol 1632 bound = 0;
964 : chenm001 272
965 : suxen_drol 1632 for (y = 0; y < mb_height; y++) {
966 :     cp_mb = st_mb = 0;
967 :     for (x = 0; x < mb_width; x++) {
968 :     MACROBLOCK *mb;
969 : Isibaar 3
970 : suxen_drol 1632 /* skip stuffing */
971 :     while (BitstreamShowBits(bs, 10) == 1)
972 :     BitstreamSkip(bs, 10);
973 : chenm001 272
974 : suxen_drol 1632 if (check_resync_marker(bs, fcode - 1)) {
975 :     bound = read_video_packet_header(bs, dec, fcode - 1,
976 :     &quant, &fcode, NULL, &intra_dc_threshold);
977 :     x = bound % mb_width;
978 :     y = bound / mb_width;
979 :     }
980 :     mb = &dec->mbs[y * dec->mb_width + x];
981 : chenm001 272
982 : suxen_drol 1632 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
983 : chenm001 272
984 : suxen_drol 1632 if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
985 :     uint32_t mcbpc, cbpc, cbpy, cbp;
986 :     uint32_t intra, acpred_flag = 0;
987 :     int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */
988 : Isibaar 3
989 : suxen_drol 1632 cp_mb++;
990 :     mcbpc = get_mcbpc_inter(bs);
991 :     mb->mode = mcbpc & 7;
992 :     cbpc = (mcbpc >> 4);
993 : Isibaar 3
994 : suxen_drol 1632 DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
995 :     DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
996 : edgomez 1382
997 : suxen_drol 1632 intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
998 : edgomez 195
999 : suxen_drol 1632 if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
1000 :     mcsel = BitstreamGetBit(bs);
1001 :     else if (intra)
1002 :     acpred_flag = BitstreamGetBit(bs);
1003 : edgomez 851
1004 : suxen_drol 1632 cbpy = get_cbpy(bs, intra);
1005 :     DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
1006 : chenm001 272
1007 : suxen_drol 1632 cbp = (cbpy << 2) | cbpc;
1008 : Isibaar 3
1009 : suxen_drol 1632 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
1010 :     int dquant = dquant_table[BitstreamGetBits(bs, 2)];
1011 :     DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
1012 :     quant += dquant;
1013 :     if (quant > 31) {
1014 :     quant = 31;
1015 :     } else if (quant < 1) {
1016 :     quant = 1;
1017 :     }
1018 :     DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
1019 :     }
1020 :     mb->quant = quant;
1021 : h 69
1022 : suxen_drol 1632 mb->field_pred=0;
1023 :     if (dec->interlacing) {
1024 :     if (cbp || intra) {
1025 :     mb->field_dct = BitstreamGetBit(bs);
1026 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1027 :     }
1028 : Isibaar 3
1029 : suxen_drol 1632 if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) {
1030 :     mb->field_pred = BitstreamGetBit(bs);
1031 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1032 : h 69
1033 : suxen_drol 1632 if (mb->field_pred) {
1034 :     mb->field_for_top = BitstreamGetBit(bs);
1035 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1036 :     mb->field_for_bot = BitstreamGetBit(bs);
1037 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1038 :     }
1039 :     }
1040 :     }
1041 : edgomez 1382
1042 : suxen_drol 1632 if (mcsel) {
1043 :     decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
1044 :     continue;
1045 : h 69
1046 : suxen_drol 1632 } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
1047 : edgomez 851
1048 : suxen_drol 1632 if(dec->interlacing) {
1049 :     /* Get motion vectors interlaced, field_pred is handled there */
1050 :     get_motion_vector_interlaced(dec, bs, x, y, 0, mb, fcode, bound);
1051 :     } else {
1052 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
1053 :     mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1054 :     }
1055 :     } else if (mb->mode == MODE_INTER4V ) {
1056 :     /* interlaced missing here */
1057 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
1058 :     get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
1059 :     get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
1060 :     get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
1061 :     } else { /* MODE_INTRA, MODE_INTRA_Q */
1062 :     mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
1063 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
1064 :     decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
1065 :     intra_dc_threshold, bound);
1066 :     continue;
1067 :     }
1068 : Isibaar 3
1069 : suxen_drol 1632 /* See how to decode */
1070 :     if(!mb->field_pred)
1071 : Isibaar 1669 decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0, 0);
1072 : suxen_drol 1632 else
1073 : Isibaar 1669 decoder_mbinter_field(dec, mb, x, y, cbp, bs, rounding, 0, 0);
1074 : edgomez 851
1075 : suxen_drol 1632 } else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */
1076 :     mb->mode = MODE_NOT_CODED_GMC;
1077 :     mb->quant = quant;
1078 :     decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
1079 : edgomez 514
1080 : suxen_drol 1632 if(dec->out_frm && cp_mb > 0) {
1081 :     output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
1082 :     cp_mb = 0;
1083 :     }
1084 :     st_mb = x+1;
1085 :     } else { /* not coded P_VOP macroblock */
1086 :     mb->mode = MODE_NOT_CODED;
1087 :     mb->quant = quant;
1088 : edgomez 851
1089 : suxen_drol 1632 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
1090 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
1091 :     mb->field_pred=0; /* (!) */
1092 : edgomez 195
1093 : suxen_drol 1632 decoder_mbinter(dec, mb, x, y, 0, bs,
1094 : Isibaar 1669 rounding, 0, 0);
1095 : Isibaar 3
1096 : suxen_drol 1632 if(dec->out_frm && cp_mb > 0) {
1097 :     output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
1098 :     cp_mb = 0;
1099 :     }
1100 :     st_mb = x+1;
1101 :     }
1102 :     }
1103 : edgomez 1382
1104 : suxen_drol 1632 if(dec->out_frm && cp_mb > 0)
1105 :     output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
1106 :     }
1107 : Isibaar 3 }
1108 :    
1109 : edgomez 851
1110 : edgomez 866 /* decode B-frame motion vector */
1111 : edgomez 1382 static void
1112 :     get_b_motion_vector(Bitstream * bs,
1113 : suxen_drol 1632 VECTOR * mv,
1114 :     int fcode,
1115 :     const VECTOR pmv,
1116 :     const DECODER * const dec,
1117 :     const int x, const int y)
1118 : edgomez 851 {
1119 : suxen_drol 1632 const int scale_fac = 1 << (fcode - 1);
1120 :     const int high = (32 * scale_fac) - 1;
1121 :     const int low = ((-32) * scale_fac);
1122 :     const int range = (64 * scale_fac);
1123 : edgomez 851
1124 : suxen_drol 1632 int mv_x = get_mv(bs, fcode);
1125 :     int mv_y = get_mv(bs, fcode);
1126 : edgomez 851
1127 : suxen_drol 1632 mv_x += pmv.x;
1128 :     mv_y += pmv.y;
1129 : edgomez 851
1130 : suxen_drol 1632 if (mv_x < low)
1131 :     mv_x += range;
1132 :     else if (mv_x > high)
1133 :     mv_x -= range;
1134 : edgomez 851
1135 : suxen_drol 1632 if (mv_y < low)
1136 :     mv_y += range;
1137 :     else if (mv_y > high)
1138 :     mv_y -= range;
1139 : edgomez 851
1140 : suxen_drol 1632 mv->x = mv_x;
1141 :     mv->y = mv_y;
1142 : edgomez 851 }
1143 :    
1144 : edgomez 1382 /* decode an B-frame direct & interpolate macroblock */
1145 :     static void
1146 : edgomez 851 decoder_bf_interpolate_mbinter(DECODER * dec,
1147 : suxen_drol 1632 IMAGE forward,
1148 :     IMAGE backward,
1149 :     MACROBLOCK * pMB,
1150 :     const uint32_t x_pos,
1151 :     const uint32_t y_pos,
1152 :     Bitstream * bs,
1153 :     const int direct)
1154 : edgomez 851 {
1155 : suxen_drol 1632 uint32_t stride = dec->edged_width;
1156 :     uint32_t stride2 = stride / 2;
1157 :     int uv_dx, uv_dy;
1158 :     int b_uv_dx, b_uv_dy;
1159 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
1160 :     const uint32_t cbp = pMB->cbp;
1161 : edgomez 851
1162 : suxen_drol 1632 pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
1163 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
1164 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
1165 : edgomez 851
1166 : suxen_drol 1632 validate_vector(pMB->mvs, x_pos, y_pos, dec);
1167 :     validate_vector(pMB->b_mvs, x_pos, y_pos, dec);
1168 : syskin 1532
1169 : suxen_drol 1632 if (!direct) {
1170 :     uv_dx = pMB->mvs[0].x;
1171 :     uv_dy = pMB->mvs[0].y;
1172 :     b_uv_dx = pMB->b_mvs[0].x;
1173 :     b_uv_dy = pMB->b_mvs[0].y;
1174 : edgomez 851
1175 : suxen_drol 1632 if (dec->quarterpel) {
1176 : Isibaar 1628 if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
1177 :     uv_dx = (uv_dx>>1) | (uv_dx&1);
1178 :     uv_dy = (uv_dy>>1) | (uv_dy&1);
1179 :     b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1);
1180 :     b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1);
1181 :     }
1182 :     else {
1183 : suxen_drol 1632 uv_dx /= 2;
1184 :     uv_dy /= 2;
1185 :     b_uv_dx /= 2;
1186 :     b_uv_dy /= 2;
1187 :     }
1188 :     }
1189 : edgomez 851
1190 : suxen_drol 1632 uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
1191 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
1192 :     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
1193 :     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
1194 : edgomez 1382
1195 : suxen_drol 1632 } else {
1196 : Isibaar 1669 if (dec->quarterpel) { /* for qpel the /2 shall be done before summation. We've done it right in the encoder in the past. */
1197 :     /* TODO: figure out if we ever did it wrong on the encoder side. If yes, add some workaround */
1198 :     if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
1199 :     int z;
1200 :     uv_dx = 0; uv_dy = 0;
1201 :     b_uv_dx = 0; b_uv_dy = 0;
1202 :     for (z = 0; z < 4; z++) {
1203 :     uv_dx += ((pMB->mvs[z].x>>1) | (pMB->mvs[z].x&1));
1204 :     uv_dy += ((pMB->mvs[z].y>>1) | (pMB->mvs[z].y&1));
1205 :     b_uv_dx += ((pMB->b_mvs[z].x>>1) | (pMB->b_mvs[z].x&1));
1206 :     b_uv_dy += ((pMB->b_mvs[z].y>>1) | (pMB->b_mvs[z].y&1));
1207 : Isibaar 1628 }
1208 : Isibaar 1669 }
1209 :     else {
1210 :     uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
1211 :     uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
1212 :     b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2);
1213 :     b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2);
1214 :     }
1215 :     } else {
1216 :     uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
1217 :     uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
1218 :     b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
1219 :     b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
1220 : suxen_drol 1632 }
1221 : edgomez 851
1222 : suxen_drol 1632 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
1223 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
1224 :     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
1225 :     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
1226 :     }
1227 : edgomez 851
1228 : suxen_drol 1632 start_timer();
1229 :     if(dec->quarterpel) {
1230 :     if(!direct) {
1231 :     interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1232 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1233 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1234 :     } else {
1235 :     interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1236 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1237 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1238 :     interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1239 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1240 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1241 :     interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1242 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1243 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1244 :     interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1245 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1246 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1247 :     }
1248 :     } else {
1249 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
1250 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1251 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
1252 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1253 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
1254 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1255 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8,
1256 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1257 :     }
1258 : edgomez 851
1259 : suxen_drol 1632 interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
1260 :     uv_dy, stride2, 0);
1261 :     interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
1262 :     uv_dy, stride2, 0);
1263 : edgomez 851
1264 :    
1265 : suxen_drol 1632 if(dec->quarterpel) {
1266 :     if(!direct) {
1267 :     interpolate16x16_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1268 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1269 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1270 :     } else {
1271 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1272 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1273 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1274 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1275 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1276 :     pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1277 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1278 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1279 :     pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1280 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1281 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1282 :     pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1283 :     }
1284 :     } else {
1285 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos,
1286 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1287 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
1288 :     16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1289 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos,
1290 :     16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1291 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
1292 :     16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1293 :     }
1294 : edgomez 851
1295 : suxen_drol 1632 interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos,
1296 :     b_uv_dx, b_uv_dy, stride2, 0);
1297 :     interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos,
1298 :     b_uv_dx, b_uv_dy, stride2, 0);
1299 : edgomez 851
1300 : suxen_drol 1632 stop_comp_timer();
1301 : edgomez 851
1302 : suxen_drol 1632 if (cbp)
1303 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
1304 : edgomez 851 }
1305 :    
1306 : edgomez 866 /* for decode B-frame dbquant */
1307 : edgomez 1382 static __inline int32_t
1308 : edgomez 851 get_dbquant(Bitstream * bs)
1309 :     {
1310 : suxen_drol 1632 if (!BitstreamGetBit(bs)) /* '0' */
1311 :     return (0);
1312 :     else if (!BitstreamGetBit(bs)) /* '10' */
1313 :     return (-2);
1314 :     else /* '11' */
1315 :     return (2);
1316 : edgomez 851 }
1317 :    
1318 : edgomez 866 /*
1319 : edgomez 1382 * decode B-frame mb_type
1320 : suxen_drol 1632 * bit ret_value
1321 :     * 1 0
1322 :     * 01 1
1323 :     * 001 2
1324 :     * 0001 3
1325 : edgomez 866 */
1326 : edgomez 1382 static int32_t __inline
1327 : edgomez 851 get_mbtype(Bitstream * bs)
1328 :     {
1329 : suxen_drol 1632 int32_t mb_type;
1330 : edgomez 851
1331 : suxen_drol 1632 for (mb_type = 0; mb_type <= 3; mb_type++)
1332 :     if (BitstreamGetBit(bs))
1333 :     return (mb_type);
1334 : edgomez 851
1335 : suxen_drol 1632 return -1;
1336 : edgomez 851 }
1337 :    
1338 : Skal 1772 static int __inline get_resync_len_b(const int fcode_backward,
1339 :     const int fcode_forward) {
1340 :     int resync_len = ((fcode_forward>fcode_backward) ? fcode_forward : fcode_backward) - 1;
1341 :     if (resync_len < 1) resync_len = 1;
1342 :     return resync_len;
1343 :     }
1344 :    
1345 : edgomez 1382 static void
1346 : edgomez 851 decoder_bframe(DECODER * dec,
1347 : suxen_drol 1632 Bitstream * bs,
1348 :     int quant,
1349 :     int fcode_forward,
1350 :     int fcode_backward)
1351 : edgomez 851 {
1352 : suxen_drol 1632 uint32_t x, y;
1353 :     VECTOR mv;
1354 :     const VECTOR zeromv = {0,0};
1355 :     int i;
1356 : Skal 1772 int resync_len;
1357 : edgomez 851
1358 : suxen_drol 1632 if (!dec->is_edged[0]) {
1359 :     start_timer();
1360 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
1361 :     dec->width, dec->height, dec->bs_version);
1362 :     dec->is_edged[0] = 1;
1363 :     stop_edges_timer();
1364 :     }
1365 : edgomez 851
1366 : suxen_drol 1632 if (!dec->is_edged[1]) {
1367 :     start_timer();
1368 :     image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
1369 :     dec->width, dec->height, dec->bs_version);
1370 :     dec->is_edged[1] = 1;
1371 :     stop_edges_timer();
1372 :     }
1373 : edgomez 1451
1374 : Skal 1772 resync_len = get_resync_len_b(fcode_backward, fcode_forward);
1375 : suxen_drol 1632 for (y = 0; y < dec->mb_height; y++) {
1376 :     /* Initialize Pred Motion Vector */
1377 :     dec->p_fmv = dec->p_bmv = zeromv;
1378 :     for (x = 0; x < dec->mb_width; x++) {
1379 :     MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
1380 :     MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
1381 : suxen_drol 1639 int intra_dc_threshold; /* fake variable */
1382 : edgomez 851
1383 : Skal 1772 if (check_resync_marker(bs, resync_len)) {
1384 :     int bound = read_video_packet_header(bs, dec, resync_len, &quant,
1385 : suxen_drol 1632 &fcode_forward, &fcode_backward, &intra_dc_threshold);
1386 :     x = bound % dec->mb_width;
1387 :     y = bound / dec->mb_width;
1388 :     /* reset predicted macroblocks */
1389 :     dec->p_fmv = dec->p_bmv = zeromv;
1390 : Skal 1772 /* update resync len with new fcodes */
1391 :     resync_len = get_resync_len_b(fcode_backward, fcode_forward);
1392 : suxen_drol 1632 }
1393 : edgomez 1382
1394 : suxen_drol 1632 mv =
1395 :     mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
1396 :     mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
1397 :     mb->quant = quant;
1398 : edgomez 851
1399 : suxen_drol 1632 /*
1400 :     * skip if the co-located P_VOP macroblock is not coded
1401 :     * if not codec in co-located S_VOP macroblock is _not_
1402 :     * automatically skipped
1403 :     */
1404 : edgomez 851
1405 : suxen_drol 1632 if (last_mb->mode == MODE_NOT_CODED) {
1406 :     mb->cbp = 0;
1407 :     mb->mode = MODE_FORWARD;
1408 : Isibaar 1669 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1);
1409 : suxen_drol 1632 continue;
1410 :     }
1411 : edgomez 851
1412 : suxen_drol 1632 if (!BitstreamGetBit(bs)) { /* modb=='0' */
1413 :     const uint8_t modb2 = BitstreamGetBit(bs);
1414 : edgomez 851
1415 : suxen_drol 1632 mb->mode = get_mbtype(bs);
1416 : edgomez 851
1417 : suxen_drol 1632 if (!modb2) /* modb=='00' */
1418 :     mb->cbp = BitstreamGetBits(bs, 6);
1419 :     else
1420 :     mb->cbp = 0;
1421 : edgomez 851
1422 : suxen_drol 1632 if (mb->mode && mb->cbp) {
1423 :     quant += get_dbquant(bs);
1424 :     if (quant > 31)
1425 :     quant = 31;
1426 :     else if (quant < 1)
1427 :     quant = 1;
1428 :     }
1429 :     mb->quant = quant;
1430 : edgomez 1382
1431 : suxen_drol 1632 if (dec->interlacing) {
1432 :     if (mb->cbp) {
1433 :     mb->field_dct = BitstreamGetBit(bs);
1434 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1435 :     }
1436 : edgomez 1382
1437 : suxen_drol 1632 if (mb->mode) {
1438 :     mb->field_pred = BitstreamGetBit(bs);
1439 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1440 : edgomez 1382
1441 : suxen_drol 1632 if (mb->field_pred) {
1442 :     mb->field_for_top = BitstreamGetBit(bs);
1443 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1444 :     mb->field_for_bot = BitstreamGetBit(bs);
1445 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1446 :     }
1447 :     }
1448 :     }
1449 : edgomez 1382
1450 : suxen_drol 1632 } else {
1451 :     mb->mode = MODE_DIRECT_NONE_MV;
1452 :     mb->cbp = 0;
1453 :     }
1454 : edgomez 851
1455 : suxen_drol 1632 switch (mb->mode) {
1456 :     case MODE_DIRECT:
1457 :     get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y);
1458 : edgomez 851
1459 : suxen_drol 1632 case MODE_DIRECT_NONE_MV:
1460 :     for (i = 0; i < 4; i++) {
1461 :     mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x;
1462 :     mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y;
1463 : edgomez 1382
1464 : suxen_drol 1632 mb->b_mvs[i].x = (mv.x)
1465 :     ? mb->mvs[i].x - last_mb->mvs[i].x
1466 :     : last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp;
1467 :     mb->b_mvs[i].y = (mv.y)
1468 :     ? mb->mvs[i].y - last_mb->mvs[i].y
1469 :     : last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp;
1470 :     }
1471 : edgomez 851
1472 : suxen_drol 1632 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1473 :     mb, x, y, bs, 1);
1474 :     break;
1475 : edgomez 851
1476 : suxen_drol 1632 case MODE_INTERPOLATE:
1477 :     get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1478 :     dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1479 : edgomez 851
1480 : suxen_drol 1632 get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1481 :     dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
1482 : edgomez 851
1483 : suxen_drol 1632 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1484 :     mb, x, y, bs, 0);
1485 :     break;
1486 : edgomez 851
1487 : suxen_drol 1632 case MODE_BACKWARD:
1488 :     get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1489 :     dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1490 : edgomez 851
1491 : Isibaar 1669 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1492 : suxen_drol 1632 break;
1493 : edgomez 851
1494 : suxen_drol 1632 case MODE_FORWARD:
1495 :     get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1496 :     dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1497 : edgomez 851
1498 : Isibaar 1669 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1);
1499 : suxen_drol 1632 break;
1500 :    
1501 :     default:
1502 :     DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
1503 :     }
1504 :     } /* End of for */
1505 :     }
1506 : edgomez 851 }
1507 :    
1508 :     /* perform post processing if necessary, and output the image */
1509 : suxen_drol 1653 static void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
1510 : suxen_drol 1632 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats,
1511 :     int coding_type, int quant)
1512 : edgomez 851 {
1513 : suxen_drol 1632 const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0;
1514 : suxen_drol 1431
1515 : suxen_drol 1632 if (dec->cartoon_mode)
1516 :     frame->general &= ~XVID_FILMEFFECT;
1517 : edgomez 851
1518 : suxen_drol 1632 if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0)
1519 :     && mbs != NULL) /* post process */
1520 :     {
1521 :     /* note: image is stored to tmp */
1522 :     image_copy(&dec->tmp, img, dec->edged_width, dec->height);
1523 :     image_postproc(&dec->postproc, &dec->tmp, dec->edged_width,
1524 :     mbs, dec->mb_width, dec->mb_height, dec->mb_width,
1525 :     frame->general, brightness, dec->frames, (coding_type == B_VOP));
1526 :     img = &dec->tmp;
1527 :     }
1528 : edgomez 851
1529 : suxen_drol 1632 image_output(img, dec->width, dec->height,
1530 :     dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
1531 :     frame->output.csp, dec->interlacing);
1532 : edgomez 1382
1533 : suxen_drol 1632 if (stats) {
1534 :     stats->type = coding2type(coding_type);
1535 :     stats->data.vop.time_base = (int)dec->time_base;
1536 :     stats->data.vop.time_increment = 0; /* XXX: todo */
1537 :     stats->data.vop.qscale_stride = dec->mb_width;
1538 :     stats->data.vop.qscale = dec->qscale;
1539 :     if (stats->data.vop.qscale != NULL && mbs != NULL) {
1540 :     unsigned int i;
1541 :     for (i = 0; i < dec->mb_width*dec->mb_height; i++)
1542 :     stats->data.vop.qscale[i] = mbs[i].quant;
1543 :     } else
1544 :     stats->data.vop.qscale = NULL;
1545 :     }
1546 : edgomez 851 }
1547 :    
1548 : edgomez 195 int
1549 :     decoder_decode(DECODER * dec,
1550 : suxen_drol 1632 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
1551 : Isibaar 3 {
1552 : edgomez 78
1553 : suxen_drol 1632 Bitstream bs;
1554 :     uint32_t rounding;
1555 :     uint32_t quant = 2;
1556 :     uint32_t fcode_forward;
1557 :     uint32_t fcode_backward;
1558 :     uint32_t intra_dc_threshold;
1559 :     WARPPOINTS gmc_warp;
1560 :     int coding_type;
1561 :     int success, output, seen_something;
1562 : Isibaar 3
1563 : suxen_drol 1632 if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */
1564 :     return XVID_ERR_VERSION;
1565 : edgomez 1382
1566 : suxen_drol 1632 start_global_timer();
1567 : edgomez 195
1568 : suxen_drol 1632 dec->low_delay_default = (frame->general & XVID_LOWDELAY);
1569 :     if ((frame->general & XVID_DISCONTINUITY))
1570 :     dec->frames = 0;
1571 :     dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
1572 : edgomez 851
1573 : suxen_drol 1632 if(frame->length<0) { /* decoder flush */
1574 :     int ret;
1575 :     /* if not decoding "low_delay/packed", and this isn't low_delay and
1576 :     we have a reference frame, then outout the reference frame */
1577 :     if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
1578 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1579 :     dec->frames = 0;
1580 :     ret = 0;
1581 :     } else {
1582 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1583 :     ret = XVID_ERR_END;
1584 :     }
1585 : edgomez 851
1586 : suxen_drol 1632 emms();
1587 :     stop_global_timer();
1588 :     return ret;
1589 :     }
1590 : edgomez 851
1591 : suxen_drol 1632 BitstreamInit(&bs, frame->bitstream, frame->length);
1592 : Isibaar 3
1593 : suxen_drol 1632 /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
1594 :     if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
1595 :     {
1596 :     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
1597 :     (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
1598 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1599 :     emms();
1600 :     return 1; /* one byte consumed */
1601 :     }
1602 : Isibaar 179
1603 : suxen_drol 1632 success = 0;
1604 :     output = 0;
1605 :     seen_something = 0;
1606 : edgomez 1382
1607 : edgomez 851 repeat:
1608 : edgomez 195
1609 : suxen_drol 1632 coding_type = BitstreamReadHeaders(&bs, dec, &rounding,
1610 :     &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
1611 : Isibaar 3
1612 : suxen_drol 1632 DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%"
1613 : suxen_drol 1607 #if defined(_MSC_VER)
1614 :     "I64"
1615 :     #else
1616 :     "ll"
1617 :     #endif
1618 :     "i, time_pp=%i, time_bp=%i\n",
1619 : suxen_drol 1632 coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
1620 : Isibaar 3
1621 : suxen_drol 1632 if (coding_type == -1) { /* nothing */
1622 :     if (success) goto done;
1623 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1624 :     emms();
1625 :     return BitstreamPos(&bs)/8;
1626 :     }
1627 : Isibaar 3
1628 : suxen_drol 1632 if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
1629 : edgomez 1382
1630 : suxen_drol 1632 if (coding_type == -3)
1631 : Isibaar 1727 if (decoder_resize(dec)) return XVID_ERR_MEMORY;
1632 : edgomez 1382
1633 : suxen_drol 1632 if(stats) {
1634 :     stats->type = XVID_TYPE_VOL;
1635 :     stats->data.vol.general = 0;
1636 :     /*XXX: if (dec->interlacing)
1637 :     stats->data.vol.general |= ++INTERLACING; */
1638 :     stats->data.vol.width = dec->width;
1639 :     stats->data.vol.height = dec->height;
1640 :     stats->data.vol.par = dec->aspect_ratio;
1641 :     stats->data.vol.par_width = dec->par_width;
1642 :     stats->data.vol.par_height = dec->par_height;
1643 :     emms();
1644 :     return BitstreamPos(&bs)/8; /* number of bytes consumed */
1645 :     }
1646 :     goto repeat;
1647 :     }
1648 : Isibaar 3
1649 : suxen_drol 1632 if(dec->frames == 0 && coding_type != I_VOP) {
1650 :     /* 1st frame is not an i-vop */
1651 :     goto repeat;
1652 :     }
1653 : syskin 1439
1654 : suxen_drol 1632 dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */
1655 : chenm001 290
1656 : suxen_drol 1632 /* packed_mode: special-N_VOP treament */
1657 :     if (dec->packed_mode && coding_type == N_VOP) {
1658 :     if (dec->low_delay_default && dec->frames > 0) {
1659 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1660 :     output = 1;
1661 :     }
1662 :     /* ignore otherwise */
1663 :     } else if (coding_type != B_VOP) {
1664 :     switch(coding_type) {
1665 :     case I_VOP :
1666 :     decoder_iframe(dec, &bs, quant, intra_dc_threshold);
1667 :     break;
1668 :     case P_VOP :
1669 :     decoder_pframe(dec, &bs, rounding, quant,
1670 :     fcode_forward, intra_dc_threshold, NULL);
1671 :     break;
1672 :     case S_VOP :
1673 :     decoder_pframe(dec, &bs, rounding, quant,
1674 :     fcode_forward, intra_dc_threshold, &gmc_warp);
1675 :     break;
1676 :     case N_VOP :
1677 :     /* XXX: not_coded vops are not used for forward prediction */
1678 :     /* we should not swap(last_mbs,mbs) */
1679 :     image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
1680 :     SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */
1681 :     break;
1682 :     }
1683 : edgomez 851
1684 : suxen_drol 1632 /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
1685 :     if (!(dec->low_delay_default && dec->packed_mode)) {
1686 :     if(dec->low_delay) {
1687 :     decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1688 :     output = 1;
1689 :     } else if (dec->frames > 0) { /* is the reference frame valid? */
1690 :     /* output the reference frame */
1691 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1692 :     output = 1;
1693 :     }
1694 :     }
1695 :    
1696 :     image_swap(&dec->refn[0], &dec->refn[1]);
1697 :     dec->is_edged[1] = dec->is_edged[0];
1698 :     image_swap(&dec->cur, &dec->refn[0]);
1699 :     dec->is_edged[0] = 0;
1700 :     SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
1701 :     dec->last_coding_type = coding_type;
1702 : edgomez 851
1703 : suxen_drol 1632 dec->frames++;
1704 :     seen_something = 1;
1705 : chenm001 306
1706 : suxen_drol 1632 } else { /* B_VOP */
1707 : edgomez 851
1708 : suxen_drol 1632 if (dec->low_delay) {
1709 :     DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
1710 :     dec->low_delay = 0;
1711 :     }
1712 : edgomez 851
1713 : suxen_drol 1632 if (dec->frames < 2) {
1714 :     /* attemping to decode a bvop without atleast 2 reference frames */
1715 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1716 :     "broken b-frame, mising ref frames");
1717 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1718 :     } else if (dec->time_pp <= dec->time_bp) {
1719 :     /* this occurs when dx50_bvop_compatibility==0 sequences are
1720 :     decoded in vfw. */
1721 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1722 :     "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
1723 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1724 :     } else {
1725 :     decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
1726 :     decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1727 :     }
1728 : edgomez 851
1729 : suxen_drol 1632 output = 1;
1730 :     dec->frames++;
1731 :     }
1732 : edgomez 851
1733 : edgomez 1382 #if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */
1734 : suxen_drol 1632 BitstreamByteAlign(&bs);
1735 : edgomez 1382 #endif
1736 : edgomez 851
1737 : suxen_drol 1632 /* low_delay_default mode: repeat in packed_mode */
1738 :     if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
1739 :     success = 1;
1740 :     goto repeat;
1741 :     }
1742 : edgomez 851
1743 :     done :
1744 :    
1745 : suxen_drol 1607 /* if we reach here without outputing anything _and_
1746 :     the calling application has specified low_delay_default,
1747 :     we *must* output something.
1748 :     this always occurs on the first call to decode() call
1749 :     when bframes are present in the bitstream. it may also
1750 :     occur if no vops were seen in the bitstream
1751 :    
1752 :     if packed_mode is enabled, then we output the recently
1753 :     decoded frame (the very first ivop). otherwise we have
1754 :     nothing to display, and therefore output a black screen.
1755 :     */
1756 :     if (dec->low_delay_default && output == 0) {
1757 :     if (dec->packed_mode && seen_something) {
1758 : suxen_drol 1632 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1759 : suxen_drol 1607 } else {
1760 : suxen_drol 1632 image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
1761 :     decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant);
1762 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1763 :     }
1764 :     }
1765 : edgomez 851
1766 : suxen_drol 1632 emms();
1767 :     stop_global_timer();
1768 : Isibaar 3
1769 : suxen_drol 1632 return (BitstreamPos(&bs)+7)/8; /* number of bytes consumed */
1770 : Isibaar 3 }

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