[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 1669 - (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 : Isibaar 1669 * $Id: decoder.c,v 1.76 2005-12-24 01:06:20 Isibaar 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 : edgomez 195
199 : suxen_drol 1632 dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
200 : edgomez 851
201 : suxen_drol 1632 if (dec->fixed_dimensions)
202 :     return decoder_resize(dec);
203 :     else
204 :     return 0;
205 : Isibaar 3 }
206 :    
207 :    
208 : edgomez 195 int
209 :     decoder_destroy(DECODER * dec)
210 : Isibaar 3 {
211 : suxen_drol 1632 xvid_free(dec->last_mbs);
212 :     xvid_free(dec->mbs);
213 :     xvid_free(dec->qscale);
214 : edgomez 851
215 : suxen_drol 1632 /* image based GMC */
216 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
217 : edgomez 851
218 : suxen_drol 1632 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
219 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
220 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
221 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
222 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
223 :     xvid_free(dec->mpeg_quant_matrices);
224 :     xvid_free(dec);
225 : Isibaar 3
226 : suxen_drol 1632 write_timer();
227 :     return 0;
228 : Isibaar 3 }
229 :    
230 : edgomez 195 static const int32_t dquant_table[4] = {
231 : suxen_drol 1632 -1, -2, 1, 2
232 : Isibaar 3 };
233 :    
234 : edgomez 866 /* decode an intra macroblock */
235 : edgomez 1382 static void
236 : edgomez 195 decoder_mbintra(DECODER * dec,
237 : suxen_drol 1632 MACROBLOCK * pMB,
238 :     const uint32_t x_pos,
239 :     const uint32_t y_pos,
240 :     const uint32_t acpred_flag,
241 :     const uint32_t cbp,
242 :     Bitstream * bs,
243 :     const uint32_t quant,
244 :     const uint32_t intra_dc_threshold,
245 :     const unsigned int bound)
246 : Isibaar 3 {
247 : edgomez 78
248 : suxen_drol 1632 DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
249 :     DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
250 : edgomez 78
251 : suxen_drol 1632 uint32_t stride = dec->edged_width;
252 :     uint32_t stride2 = stride / 2;
253 :     uint32_t next_block = stride * 8;
254 :     uint32_t i;
255 :     uint32_t iQuant = pMB->quant;
256 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
257 : h 69
258 : suxen_drol 1632 pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
259 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
260 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
261 : h 69
262 : suxen_drol 1632 memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */
263 : h 69
264 : suxen_drol 1632 for (i = 0; i < 6; i++) {
265 :     uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);
266 :     int16_t predictors[8];
267 :     int start_coeff;
268 : Isibaar 3
269 : suxen_drol 1632 start_timer();
270 :     predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64],
271 :     iQuant, iDcScaler, predictors, bound);
272 :     if (!acpred_flag) {
273 :     pMB->acpred_directions[i] = 0;
274 :     }
275 :     stop_prediction_timer();
276 : Isibaar 3
277 : suxen_drol 1632 if (quant < intra_dc_threshold) {
278 :     int dc_size;
279 :     int dc_dif;
280 : Isibaar 3
281 : suxen_drol 1632 dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs);
282 :     dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0;
283 : Isibaar 3
284 : suxen_drol 1632 if (dc_size > 8) {
285 :     BitstreamSkip(bs, 1); /* marker */
286 :     }
287 : edgomez 195
288 : suxen_drol 1632 block[i * 64 + 0] = dc_dif;
289 :     start_coeff = 1;
290 : chenm001 272
291 : suxen_drol 1632 DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif);
292 :     } else {
293 :     start_coeff = 0;
294 :     }
295 : Isibaar 3
296 : suxen_drol 1632 start_timer();
297 :     if (cbp & (1 << (5 - i))) /* coded */
298 :     {
299 :     int direction = dec->alternate_vertical_scan ?
300 :     2 : pMB->acpred_directions[i];
301 : edgomez 851
302 : suxen_drol 1632 get_intra_block(bs, &block[i * 64], direction, start_coeff);
303 :     }
304 :     stop_coding_timer();
305 : Isibaar 3
306 : suxen_drol 1632 start_timer();
307 :     add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version);
308 :     stop_prediction_timer();
309 : Isibaar 3
310 : suxen_drol 1632 start_timer();
311 :     if (dec->quant_type == 0) {
312 :     dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
313 :     } else {
314 :     dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
315 :     }
316 :     stop_iquant_timer();
317 : Isibaar 3
318 : suxen_drol 1632 start_timer();
319 : suxen_drol 1653 idct((short * const)&data[i * 64]);
320 : suxen_drol 1632 stop_idct_timer();
321 : edgomez 851
322 : suxen_drol 1632 }
323 : Isibaar 3
324 : suxen_drol 1632 if (dec->interlacing && pMB->field_dct) {
325 :     next_block = stride;
326 :     stride *= 2;
327 :     }
328 : h 69
329 : suxen_drol 1632 start_timer();
330 :     transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
331 :     transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
332 :     transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
333 :     transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
334 :     transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
335 :     transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
336 :     stop_transfer_timer();
337 : Isibaar 3 }
338 :    
339 : edgomez 1382 static void
340 :     decoder_mb_decode(DECODER * dec,
341 : suxen_drol 1632 const uint32_t cbp,
342 :     Bitstream * bs,
343 :     uint8_t * pY_Cur,
344 :     uint8_t * pU_Cur,
345 :     uint8_t * pV_Cur,
346 :     const MACROBLOCK * pMB)
347 : edgomez 1382 {
348 : suxen_drol 1632 DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE);
349 : Isibaar 3
350 : suxen_drol 1632 int stride = dec->edged_width;
351 :     int i;
352 :     const uint32_t iQuant = pMB->quant;
353 :     const int direction = dec->alternate_vertical_scan ? 2 : 0;
354 :     typedef void (*get_inter_block_function_t)(
355 :     Bitstream * bs,
356 :     int16_t * block,
357 :     int direction,
358 :     const int quant,
359 :     const uint16_t *matrix);
360 :     typedef void (*add_residual_function_t)(
361 :     uint8_t *predicted_block,
362 :     const int16_t *residual,
363 :     int stride);
364 : Isibaar 3
365 : suxen_drol 1632 const get_inter_block_function_t get_inter_block = (dec->quant_type == 0)
366 :     ? (get_inter_block_function_t)get_inter_block_h263
367 :     : (get_inter_block_function_t)get_inter_block_mpeg;
368 : edgomez 1524
369 : suxen_drol 1632 uint8_t *dst[6];
370 :     int strides[6];
371 : edgomez 1486
372 : edgomez 1524
373 : suxen_drol 1632 if (dec->interlacing && pMB->field_dct) {
374 :     dst[0] = pY_Cur;
375 :     dst[1] = pY_Cur + 8;
376 :     dst[2] = pY_Cur + stride;
377 :     dst[3] = dst[2] + 8;
378 :     dst[4] = pU_Cur;
379 :     dst[5] = pV_Cur;
380 :     strides[0] = strides[1] = strides[2] = strides[3] = stride*2;
381 :     strides[4] = stride/2;
382 :     strides[5] = stride/2;
383 :     } else {
384 :     dst[0] = pY_Cur;
385 :     dst[1] = pY_Cur + 8;
386 :     dst[2] = pY_Cur + 8*stride;
387 :     dst[3] = dst[2] + 8;
388 :     dst[4] = pU_Cur;
389 :     dst[5] = pV_Cur;
390 :     strides[0] = strides[1] = strides[2] = strides[3] = stride;
391 :     strides[4] = stride/2;
392 :     strides[5] = stride/2;
393 :     }
394 : Isibaar 3
395 : suxen_drol 1632 for (i = 0; i < 6; i++) {
396 :     /* Process only coded blocks */
397 :     if (cbp & (1 << (5 - i))) {
398 : edgomez 1382
399 : suxen_drol 1632 /* Clear the block */
400 :     memset(&data[0], 0, 64*sizeof(int16_t));
401 : edgomez 1382
402 : suxen_drol 1632 /* Decode coeffs and dequantize on the fly */
403 :     start_timer();
404 :     get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices));
405 :     stop_coding_timer();
406 : edgomez 1524
407 : suxen_drol 1632 /* iDCT */
408 :     start_timer();
409 : suxen_drol 1653 idct((short * const)&data[0]);
410 : suxen_drol 1632 stop_idct_timer();
411 :    
412 :     /* Add this residual to the predicted block */
413 :     start_timer();
414 :     transfer_16to8add(dst[i], &data[0], strides[i]);
415 :     stop_transfer_timer();
416 :     }
417 :     }
418 : edgomez 1382 }
419 :    
420 : edgomez 1533 static void __inline
421 : syskin 1532 validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec)
422 :     {
423 : suxen_drol 1632 /* clip a vector to valid range
424 :     prevents crashes if bitstream is broken
425 :     */
426 :     int shift = 5 + dec->quarterpel;
427 :     int xborder_high = (int)(dec->mb_width - x_pos) << shift;
428 :     int xborder_low = (-(int)x_pos-1) << shift;
429 :     int yborder_high = (int)(dec->mb_height - y_pos) << shift;
430 :     int yborder_low = (-(int)y_pos-1) << shift;
431 : syskin 1532
432 : edgomez 1533 #define CHECK_MV(mv) \
433 : suxen_drol 1632 do { \
434 :     if ((mv).x > xborder_high) { \
435 :     DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \
436 :     (mv).x = xborder_high; \
437 :     } else if ((mv).x < xborder_low) { \
438 :     DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \
439 :     (mv).x = xborder_low; \
440 :     } \
441 :     if ((mv).y > yborder_high) { \
442 :     DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \
443 :     (mv).y = yborder_high; \
444 :     } else if ((mv).y < yborder_low) { \
445 :     DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \
446 :     (mv).y = yborder_low; \
447 :     } \
448 :     } while (0)
449 : syskin 1532
450 : suxen_drol 1632 CHECK_MV(mv[0]);
451 :     CHECK_MV(mv[1]);
452 :     CHECK_MV(mv[2]);
453 :     CHECK_MV(mv[3]);
454 : syskin 1532 }
455 :    
456 : Isibaar 1628 /* Up to this version, chroma rounding was wrong with qpel.
457 :     * So we try to be backward compatible to avoid artifacts */
458 :     #define BS_VERSION_BUGGY_CHROMA_ROUNDING 1
459 :    
460 : edgomez 866 /* decode an inter macroblock */
461 : edgomez 1382 static void
462 : edgomez 195 decoder_mbinter(DECODER * dec,
463 : suxen_drol 1632 const MACROBLOCK * pMB,
464 :     const uint32_t x_pos,
465 :     const uint32_t y_pos,
466 :     const uint32_t cbp,
467 :     Bitstream * bs,
468 :     const uint32_t rounding,
469 : Isibaar 1669 const int ref,
470 :     const int bvop)
471 : Isibaar 3 {
472 : suxen_drol 1632 uint32_t stride = dec->edged_width;
473 :     uint32_t stride2 = stride / 2;
474 :     uint32_t i;
475 : edgomez 1382
476 : suxen_drol 1632 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
477 : edgomez 851
478 : suxen_drol 1632 int uv_dx, uv_dy;
479 :     VECTOR mv[4]; /* local copy of mvs */
480 : Isibaar 3
481 : suxen_drol 1632 pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
482 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
483 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
484 :     for (i = 0; i < 4; i++)
485 :     mv[i] = pMB->mvs[i];
486 : Isibaar 3
487 : suxen_drol 1632 validate_vector(mv, x_pos, y_pos, dec);
488 : syskin 1420
489 : suxen_drol 1632 start_timer();
490 : edgomez 1382
491 : Isibaar 1669 if ((pMB->mode != MODE_INTER4V) || (bvop)) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
492 : edgomez 1382
493 : suxen_drol 1632 uv_dx = mv[0].x;
494 :     uv_dy = mv[0].y;
495 :     if (dec->quarterpel) {
496 : Isibaar 1628 if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
497 : Isibaar 1669 uv_dx = (uv_dx>>1) | (uv_dx&1);
498 : Isibaar 1628 uv_dy = (uv_dy>>1) | (uv_dy&1);
499 :     }
500 :     else {
501 : suxen_drol 1632 uv_dx /= 2;
502 :     uv_dy /= 2;
503 :     }
504 :     }
505 :     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
506 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
507 : edgomez 851
508 : suxen_drol 1632 if (dec->quarterpel)
509 :     interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
510 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
511 :     mv[0].x, mv[0].y, stride, rounding);
512 :     else
513 :     interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
514 :     mv[0].x, mv[0].y, stride, rounding);
515 : edgomez 851
516 : suxen_drol 1632 } else { /* MODE_INTER4V */
517 : Isibaar 3
518 : suxen_drol 1632 if(dec->quarterpel) {
519 : Isibaar 1628 if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
520 :     int z;
521 :     uv_dx = 0; uv_dy = 0;
522 :     for (z = 0; z < 4; z++) {
523 :     uv_dx += ((mv[z].x>>1) | (mv[z].x&1));
524 :     uv_dy += ((mv[z].y>>1) | (mv[z].y&1));
525 :     }
526 :     }
527 :     else {
528 : suxen_drol 1632 uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
529 :     uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
530 :     }
531 :     } else {
532 :     uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
533 :     uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
534 :     }
535 : Isibaar 333
536 : suxen_drol 1632 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
537 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
538 : Isibaar 333
539 : suxen_drol 1632 if (dec->quarterpel) {
540 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
541 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
542 :     mv[0].x, mv[0].y, stride, rounding);
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 + 8, 16*y_pos,
545 :     mv[1].x, mv[1].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, 16*y_pos + 8,
548 :     mv[2].x, mv[2].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 + 8, 16*y_pos + 8,
551 :     mv[3].x, mv[3].y, stride, rounding);
552 :     } else {
553 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
554 :     mv[0].x, mv[0].y, stride, rounding);
555 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
556 :     mv[1].x, mv[1].y, stride, rounding);
557 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
558 :     mv[2].x, mv[2].y, stride, rounding);
559 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
560 :     mv[3].x, mv[3].y, stride, rounding);
561 :     }
562 :     }
563 : Isibaar 3
564 : suxen_drol 1632 /* chroma */
565 :     interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
566 :     uv_dx, uv_dy, stride2, rounding);
567 :     interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
568 :     uv_dx, uv_dy, stride2, rounding);
569 : h 69
570 : suxen_drol 1632 stop_comp_timer();
571 : h 69
572 : suxen_drol 1632 if (cbp)
573 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
574 : edgomez 851 }
575 :    
576 : suxen_drol 1632 /* decode an inter macroblock in field mode */
577 : edgomez 851 static void
578 : suxen_drol 1632 decoder_mbinter_field(DECODER * dec,
579 :     const MACROBLOCK * pMB,
580 :     const uint32_t x_pos,
581 :     const uint32_t y_pos,
582 :     const uint32_t cbp,
583 :     Bitstream * bs,
584 :     const uint32_t rounding,
585 : Isibaar 1669 const int ref,
586 :     const int bvop)
587 : suxen_drol 1632 {
588 :     uint32_t stride = dec->edged_width;
589 :     uint32_t stride2 = stride / 2;
590 :    
591 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
592 :    
593 :     int uvtop_dx, uvtop_dy;
594 :     int uvbot_dx, uvbot_dy;
595 :     VECTOR mv[4]; /* local copy of mvs */
596 :    
597 :     /* Get pointer to memory areas */
598 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
599 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
600 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
601 :    
602 :     mv[0] = pMB->mvs[0];
603 :     mv[1] = pMB->mvs[1];
604 :     memset(&mv[2],0,2*sizeof(VECTOR));
605 :    
606 :     validate_vector(mv, x_pos, y_pos, dec);
607 :    
608 :     start_timer();
609 :    
610 : Isibaar 1669 if((pMB->mode!=MODE_INTER4V) || (bvop)) /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
611 : suxen_drol 1632 {
612 :     /* Prepare top field vector */
613 :     uvtop_dx = DIV2ROUND(mv[0].x);
614 :     uvtop_dy = DIV2ROUND(mv[0].y);
615 :    
616 :     /* Prepare bottom field vector */
617 :     uvbot_dx = DIV2ROUND(mv[1].x);
618 :     uvbot_dy = DIV2ROUND(mv[1].y);
619 :    
620 :     if(dec->quarterpel)
621 :     {
622 :     /* NOT supported */
623 :     }
624 :     else
625 :     {
626 :     /* Interpolate top field left part(we use double stride for every 2nd line) */
627 :     interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride,
628 :     16*x_pos,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding);
629 :     /* top field right part */
630 :     interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride,
631 :     16*x_pos+8,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding);
632 :    
633 :     /* Interpolate bottom field left part(we use double stride for every 2nd line) */
634 :     interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride,
635 :     16*x_pos,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding);
636 :     /* Bottom field right part */
637 :     interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride,
638 :     16*x_pos+8,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding);
639 :    
640 :     /* Interpolate field1 U */
641 :     interpolate8x4_switch(dec->cur.u,dec->refn[ref].u+pMB->field_for_top*stride2,
642 :     8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding);
643 :    
644 :     /* Interpolate field1 V */
645 :     interpolate8x4_switch(dec->cur.v,dec->refn[ref].v+pMB->field_for_top*stride2,
646 :     8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding);
647 :    
648 :     /* Interpolate field2 U */
649 :     interpolate8x4_switch(dec->cur.u+stride2,dec->refn[ref].u+pMB->field_for_bot*stride2,
650 :     8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding);
651 :    
652 :     /* Interpolate field2 V */
653 :     interpolate8x4_switch(dec->cur.v+stride2,dec->refn[ref].v+pMB->field_for_bot*stride2,
654 :     8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding);
655 :     }
656 :     }
657 :     else
658 :     {
659 :     /* We don't expect 4 motion vectors in interlaced mode */
660 :     }
661 :    
662 :     stop_comp_timer();
663 :    
664 :     /* Must add error correction? */
665 :     if(cbp)
666 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
667 :     }
668 :    
669 :     static void
670 : edgomez 851 decoder_mbgmc(DECODER * dec,
671 : suxen_drol 1632 MACROBLOCK * const pMB,
672 :     const uint32_t x_pos,
673 :     const uint32_t y_pos,
674 :     const uint32_t fcode,
675 :     const uint32_t cbp,
676 :     Bitstream * bs,
677 :     const uint32_t rounding)
678 : edgomez 851 {
679 : suxen_drol 1632 const uint32_t stride = dec->edged_width;
680 :     const uint32_t stride2 = stride / 2;
681 : edgomez 1382
682 : suxen_drol 1632 uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
683 :     uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
684 :     uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
685 : edgomez 851
686 : suxen_drol 1632 NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
687 : edgomez 1382
688 : suxen_drol 1632 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
689 : edgomez 851
690 : suxen_drol 1632 start_timer();
691 : edgomez 1382
692 : edgomez 851 /* this is where the calculations are done */
693 :    
694 : suxen_drol 1632 gmc_data->predict_16x16(gmc_data,
695 :     dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
696 :     stride, stride, x_pos, y_pos, rounding);
697 : edgomez 851
698 : suxen_drol 1632 gmc_data->predict_8x8(gmc_data,
699 :     dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
700 :     dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
701 :     stride2, stride2, x_pos, y_pos, rounding);
702 : edgomez 851
703 : suxen_drol 1632 gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
704 : edgomez 851
705 : suxen_drol 1632 pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
706 :     pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
707 : edgomez 851
708 : suxen_drol 1632 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
709 : edgomez 851
710 : suxen_drol 1632 stop_transfer_timer();
711 : edgomez 851
712 : suxen_drol 1632 if (cbp)
713 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
714 : edgomez 851
715 : Isibaar 3 }
716 :    
717 :    
718 : edgomez 1382 static void
719 : edgomez 195 decoder_iframe(DECODER * dec,
720 : suxen_drol 1632 Bitstream * bs,
721 :     int quant,
722 :     int intra_dc_threshold)
723 : Isibaar 3 {
724 : suxen_drol 1632 uint32_t bound;
725 :     uint32_t x, y;
726 :     const uint32_t mb_width = dec->mb_width;
727 :     const uint32_t mb_height = dec->mb_height;
728 : edgomez 1382
729 : suxen_drol 1632 bound = 0;
730 : chenm001 272
731 : suxen_drol 1632 for (y = 0; y < mb_height; y++) {
732 :     for (x = 0; x < mb_width; x++) {
733 :     MACROBLOCK *mb;
734 :     uint32_t mcbpc;
735 :     uint32_t cbpc;
736 :     uint32_t acpred_flag;
737 :     uint32_t cbpy;
738 :     uint32_t cbp;
739 : Isibaar 3
740 : suxen_drol 1632 while (BitstreamShowBits(bs, 9) == 1)
741 :     BitstreamSkip(bs, 9);
742 : chenm001 272
743 : suxen_drol 1632 if (check_resync_marker(bs, 0))
744 :     {
745 :     bound = read_video_packet_header(bs, dec, 0,
746 :     &quant, NULL, NULL, &intra_dc_threshold);
747 :     x = bound % mb_width;
748 :     y = bound / mb_width;
749 :     }
750 :     mb = &dec->mbs[y * dec->mb_width + x];
751 : chenm001 272
752 : suxen_drol 1632 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
753 : chenm001 272
754 : suxen_drol 1632 mcbpc = get_mcbpc_intra(bs);
755 :     mb->mode = mcbpc & 7;
756 :     cbpc = (mcbpc >> 4);
757 : Isibaar 3
758 : suxen_drol 1632 acpred_flag = BitstreamGetBit(bs);
759 : Isibaar 3
760 : suxen_drol 1632 cbpy = get_cbpy(bs, 1);
761 :     cbp = (cbpy << 2) | cbpc;
762 : Isibaar 3
763 : suxen_drol 1632 if (mb->mode == MODE_INTRA_Q) {
764 :     quant += dquant_table[BitstreamGetBits(bs, 2)];
765 :     if (quant > 31) {
766 :     quant = 31;
767 :     } else if (quant < 1) {
768 :     quant = 1;
769 :     }
770 :     }
771 :     mb->quant = quant;
772 :     mb->mvs[0].x = mb->mvs[0].y =
773 :     mb->mvs[1].x = mb->mvs[1].y =
774 :     mb->mvs[2].x = mb->mvs[2].y =
775 :     mb->mvs[3].x = mb->mvs[3].y =0;
776 : Isibaar 3
777 : suxen_drol 1632 if (dec->interlacing) {
778 :     mb->field_dct = BitstreamGetBit(bs);
779 :     DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
780 :     }
781 : h 69
782 : suxen_drol 1632 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
783 :     intra_dc_threshold, bound);
784 : edgomez 851
785 : suxen_drol 1632 }
786 :     if(dec->out_frm)
787 :     output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
788 :     }
789 : edgomez 78
790 : Isibaar 3 }
791 :    
792 :    
793 : edgomez 1382 static void
794 : edgomez 195 get_motion_vector(DECODER * dec,
795 : suxen_drol 1632 Bitstream * bs,
796 :     int x,
797 :     int y,
798 :     int k,
799 :     VECTOR * ret_mv,
800 :     int fcode,
801 :     const int bound)
802 : Isibaar 3 {
803 : edgomez 78
804 : suxen_drol 1632 const int scale_fac = 1 << (fcode - 1);
805 :     const int high = (32 * scale_fac) - 1;
806 :     const int low = ((-32) * scale_fac);
807 :     const int range = (64 * scale_fac);
808 : Isibaar 3
809 : suxen_drol 1632 const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
810 :     VECTOR mv;
811 : Isibaar 3
812 : suxen_drol 1632 mv.x = get_mv(bs, fcode);
813 :     mv.y = get_mv(bs, fcode);
814 : edgomez 195
815 : 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);
816 : suxen_drol 252
817 : suxen_drol 1632 mv.x += pmv.x;
818 :     mv.y += pmv.y;
819 : chenm001 272
820 : suxen_drol 1632 if (mv.x < low) {
821 :     mv.x += range;
822 :     } else if (mv.x > high) {
823 :     mv.x -= range;
824 :     }
825 : Isibaar 3
826 : suxen_drol 1632 if (mv.y < low) {
827 :     mv.y += range;
828 :     } else if (mv.y > high) {
829 :     mv.y -= range;
830 :     }
831 : Isibaar 3
832 : suxen_drol 1632 ret_mv->x = mv.x;
833 :     ret_mv->y = mv.y;
834 : edgomez 1382 }
835 : edgomez 851
836 : suxen_drol 1632 /* We use this when decoder runs interlaced -> different prediction */
837 :    
838 :     static void get_motion_vector_interlaced(DECODER * dec,
839 :     Bitstream * bs,
840 :     int x,
841 :     int y,
842 :     int k,
843 :     MACROBLOCK *pMB,
844 :     int fcode,
845 :     const int bound)
846 :     {
847 :     const int scale_fac = 1 << (fcode - 1);
848 :     const int high = (32 * scale_fac) - 1;
849 :     const int low = ((-32) * scale_fac);
850 :     const int range = (64 * scale_fac);
851 :    
852 :     /* Get interlaced prediction */
853 :     const VECTOR pmv=get_pmv2_interlaced(dec->mbs,dec->mb_width,bound,x,y,k);
854 :     VECTOR mv,mvf1,mvf2;
855 :    
856 :     if(!pMB->field_pred)
857 :     {
858 :     mv.x = get_mv(bs,fcode);
859 :     mv.y = get_mv(bs,fcode);
860 :    
861 :     mv.x += pmv.x;
862 :     mv.y += pmv.y;
863 :    
864 :     if(mv.x<low) {
865 :     mv.x += range;
866 :     } else if (mv.x>high) {
867 :     mv.x-=range;
868 :     }
869 :    
870 :     if (mv.y < low) {
871 :     mv.y += range;
872 :     } else if (mv.y > high) {
873 :     mv.y -= range;
874 :     }
875 :    
876 :     pMB->mvs[0]=pMB->mvs[1]=pMB->mvs[2]=pMB->mvs[3]=mv;
877 :     }
878 :     else
879 :     {
880 :     mvf1.x = get_mv(bs, fcode);
881 :     mvf1.y = get_mv(bs, fcode);
882 :    
883 :     mvf1.x += pmv.x;
884 :     mvf1.y = 2*(mvf1.y+pmv.y/2); /* It's multiple of 2 */
885 :    
886 :     if (mvf1.x < low) {
887 :     mvf1.x += range;
888 :     } else if (mvf1.x > high) {
889 :     mvf1.x -= range;
890 :     }
891 :    
892 :     if (mvf1.y < low) {
893 :     mvf1.y += range;
894 :     } else if (mvf1.y > high) {
895 :     mvf1.y -= range;
896 :     }
897 :    
898 :     mvf2.x = get_mv(bs, fcode);
899 :     mvf2.y = get_mv(bs, fcode);
900 :    
901 :     mvf2.x += pmv.x;
902 :     mvf2.y = 2*(mvf2.y+pmv.y/2); /* It's multiple of 2 */
903 :    
904 :     if (mvf2.x < low) {
905 :     mvf2.x += range;
906 :     } else if (mvf2.x > high) {
907 :     mvf2.x -= range;
908 :     }
909 :    
910 :     if (mvf2.y < low) {
911 :     mvf2.y += range;
912 :     } else if (mvf2.y > high) {
913 :     mvf2.y -= range;
914 :     }
915 :    
916 :     pMB->mvs[0]=mvf1;
917 :     pMB->mvs[1]=mvf2;
918 :     pMB->mvs[2].x=pMB->mvs[3].x=0;
919 :     pMB->mvs[2].y=pMB->mvs[3].y=0;
920 :    
921 :     /* Calculate average for as it is field predicted */
922 :     pMB->mvs_avg.x=DIV2ROUND(pMB->mvs[0].x+pMB->mvs[1].x);
923 :     pMB->mvs_avg.y=DIV2ROUND(pMB->mvs[0].y+pMB->mvs[1].y);
924 :     }
925 :     }
926 :    
927 : edgomez 851 /* for P_VOP set gmc_warp to NULL */
928 : edgomez 1382 static void
929 : edgomez 195 decoder_pframe(DECODER * dec,
930 : suxen_drol 1632 Bitstream * bs,
931 :     int rounding,
932 :     int quant,
933 :     int fcode,
934 :     int intra_dc_threshold,
935 :     const WARPPOINTS *const gmc_warp)
936 : Isibaar 3 {
937 : suxen_drol 1632 uint32_t x, y;
938 :     uint32_t bound;
939 :     int cp_mb, st_mb;
940 :     const uint32_t mb_width = dec->mb_width;
941 :     const uint32_t mb_height = dec->mb_height;
942 : edgomez 1382
943 : suxen_drol 1632 if (!dec->is_edged[0]) {
944 :     start_timer();
945 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
946 :     dec->width, dec->height, dec->bs_version);
947 :     dec->is_edged[0] = 1;
948 :     stop_edges_timer();
949 :     }
950 : Isibaar 3
951 : suxen_drol 1632 if (gmc_warp) {
952 :     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
953 :     generate_GMCparameters( dec->sprite_warping_points,
954 :     dec->sprite_warping_accuracy, gmc_warp,
955 :     dec->width, dec->height, &dec->new_gmc_data);
956 : edgomez 851
957 : suxen_drol 1632 /* image warping is done block-based in decoder_mbgmc(), now */
958 :     }
959 : edgomez 851
960 : suxen_drol 1632 bound = 0;
961 : chenm001 272
962 : suxen_drol 1632 for (y = 0; y < mb_height; y++) {
963 :     cp_mb = st_mb = 0;
964 :     for (x = 0; x < mb_width; x++) {
965 :     MACROBLOCK *mb;
966 : Isibaar 3
967 : suxen_drol 1632 /* skip stuffing */
968 :     while (BitstreamShowBits(bs, 10) == 1)
969 :     BitstreamSkip(bs, 10);
970 : chenm001 272
971 : suxen_drol 1632 if (check_resync_marker(bs, fcode - 1)) {
972 :     bound = read_video_packet_header(bs, dec, fcode - 1,
973 :     &quant, &fcode, NULL, &intra_dc_threshold);
974 :     x = bound % mb_width;
975 :     y = bound / mb_width;
976 :     }
977 :     mb = &dec->mbs[y * dec->mb_width + x];
978 : chenm001 272
979 : suxen_drol 1632 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
980 : chenm001 272
981 : suxen_drol 1632 if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
982 :     uint32_t mcbpc, cbpc, cbpy, cbp;
983 :     uint32_t intra, acpred_flag = 0;
984 :     int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */
985 : Isibaar 3
986 : suxen_drol 1632 cp_mb++;
987 :     mcbpc = get_mcbpc_inter(bs);
988 :     mb->mode = mcbpc & 7;
989 :     cbpc = (mcbpc >> 4);
990 : Isibaar 3
991 : suxen_drol 1632 DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
992 :     DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
993 : edgomez 1382
994 : suxen_drol 1632 intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
995 : edgomez 195
996 : suxen_drol 1632 if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
997 :     mcsel = BitstreamGetBit(bs);
998 :     else if (intra)
999 :     acpred_flag = BitstreamGetBit(bs);
1000 : edgomez 851
1001 : suxen_drol 1632 cbpy = get_cbpy(bs, intra);
1002 :     DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
1003 : chenm001 272
1004 : suxen_drol 1632 cbp = (cbpy << 2) | cbpc;
1005 : Isibaar 3
1006 : suxen_drol 1632 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
1007 :     int dquant = dquant_table[BitstreamGetBits(bs, 2)];
1008 :     DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
1009 :     quant += dquant;
1010 :     if (quant > 31) {
1011 :     quant = 31;
1012 :     } else if (quant < 1) {
1013 :     quant = 1;
1014 :     }
1015 :     DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
1016 :     }
1017 :     mb->quant = quant;
1018 : h 69
1019 : suxen_drol 1632 mb->field_pred=0;
1020 :     if (dec->interlacing) {
1021 :     if (cbp || intra) {
1022 :     mb->field_dct = BitstreamGetBit(bs);
1023 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1024 :     }
1025 : Isibaar 3
1026 : suxen_drol 1632 if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) {
1027 :     mb->field_pred = BitstreamGetBit(bs);
1028 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1029 : h 69
1030 : suxen_drol 1632 if (mb->field_pred) {
1031 :     mb->field_for_top = BitstreamGetBit(bs);
1032 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1033 :     mb->field_for_bot = BitstreamGetBit(bs);
1034 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1035 :     }
1036 :     }
1037 :     }
1038 : edgomez 1382
1039 : suxen_drol 1632 if (mcsel) {
1040 :     decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
1041 :     continue;
1042 : h 69
1043 : suxen_drol 1632 } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
1044 : edgomez 851
1045 : suxen_drol 1632 if(dec->interlacing) {
1046 :     /* Get motion vectors interlaced, field_pred is handled there */
1047 :     get_motion_vector_interlaced(dec, bs, x, y, 0, mb, fcode, bound);
1048 :     } else {
1049 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
1050 :     mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1051 :     }
1052 :     } else if (mb->mode == MODE_INTER4V ) {
1053 :     /* interlaced missing here */
1054 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
1055 :     get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
1056 :     get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
1057 :     get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
1058 :     } else { /* MODE_INTRA, MODE_INTRA_Q */
1059 :     mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
1060 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
1061 :     decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
1062 :     intra_dc_threshold, bound);
1063 :     continue;
1064 :     }
1065 : Isibaar 3
1066 : suxen_drol 1632 /* See how to decode */
1067 :     if(!mb->field_pred)
1068 : Isibaar 1669 decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0, 0);
1069 : suxen_drol 1632 else
1070 : Isibaar 1669 decoder_mbinter_field(dec, mb, x, y, cbp, bs, rounding, 0, 0);
1071 : edgomez 851
1072 : suxen_drol 1632 } else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */
1073 :     mb->mode = MODE_NOT_CODED_GMC;
1074 :     mb->quant = quant;
1075 :     decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
1076 : edgomez 514
1077 : suxen_drol 1632 if(dec->out_frm && cp_mb > 0) {
1078 :     output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
1079 :     cp_mb = 0;
1080 :     }
1081 :     st_mb = x+1;
1082 :     } else { /* not coded P_VOP macroblock */
1083 :     mb->mode = MODE_NOT_CODED;
1084 :     mb->quant = quant;
1085 : edgomez 851
1086 : suxen_drol 1632 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
1087 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
1088 :     mb->field_pred=0; /* (!) */
1089 : edgomez 195
1090 : suxen_drol 1632 decoder_mbinter(dec, mb, x, y, 0, bs,
1091 : Isibaar 1669 rounding, 0, 0);
1092 : Isibaar 3
1093 : suxen_drol 1632 if(dec->out_frm && cp_mb > 0) {
1094 :     output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
1095 :     cp_mb = 0;
1096 :     }
1097 :     st_mb = x+1;
1098 :     }
1099 :     }
1100 : edgomez 1382
1101 : suxen_drol 1632 if(dec->out_frm && cp_mb > 0)
1102 :     output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
1103 :     }
1104 : Isibaar 3 }
1105 :    
1106 : edgomez 851
1107 : edgomez 866 /* decode B-frame motion vector */
1108 : edgomez 1382 static void
1109 :     get_b_motion_vector(Bitstream * bs,
1110 : suxen_drol 1632 VECTOR * mv,
1111 :     int fcode,
1112 :     const VECTOR pmv,
1113 :     const DECODER * const dec,
1114 :     const int x, const int y)
1115 : edgomez 851 {
1116 : suxen_drol 1632 const int scale_fac = 1 << (fcode - 1);
1117 :     const int high = (32 * scale_fac) - 1;
1118 :     const int low = ((-32) * scale_fac);
1119 :     const int range = (64 * scale_fac);
1120 : edgomez 851
1121 : suxen_drol 1632 int mv_x = get_mv(bs, fcode);
1122 :     int mv_y = get_mv(bs, fcode);
1123 : edgomez 851
1124 : suxen_drol 1632 mv_x += pmv.x;
1125 :     mv_y += pmv.y;
1126 : edgomez 851
1127 : suxen_drol 1632 if (mv_x < low)
1128 :     mv_x += range;
1129 :     else if (mv_x > high)
1130 :     mv_x -= range;
1131 : edgomez 851
1132 : suxen_drol 1632 if (mv_y < low)
1133 :     mv_y += range;
1134 :     else if (mv_y > high)
1135 :     mv_y -= range;
1136 : edgomez 851
1137 : suxen_drol 1632 mv->x = mv_x;
1138 :     mv->y = mv_y;
1139 : edgomez 851 }
1140 :    
1141 : edgomez 1382 /* decode an B-frame direct & interpolate macroblock */
1142 :     static void
1143 : edgomez 851 decoder_bf_interpolate_mbinter(DECODER * dec,
1144 : suxen_drol 1632 IMAGE forward,
1145 :     IMAGE backward,
1146 :     MACROBLOCK * pMB,
1147 :     const uint32_t x_pos,
1148 :     const uint32_t y_pos,
1149 :     Bitstream * bs,
1150 :     const int direct)
1151 : edgomez 851 {
1152 : suxen_drol 1632 uint32_t stride = dec->edged_width;
1153 :     uint32_t stride2 = stride / 2;
1154 :     int uv_dx, uv_dy;
1155 :     int b_uv_dx, b_uv_dy;
1156 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
1157 :     const uint32_t cbp = pMB->cbp;
1158 : edgomez 851
1159 : suxen_drol 1632 pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
1160 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
1161 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
1162 : edgomez 851
1163 : suxen_drol 1632 validate_vector(pMB->mvs, x_pos, y_pos, dec);
1164 :     validate_vector(pMB->b_mvs, x_pos, y_pos, dec);
1165 : syskin 1532
1166 : suxen_drol 1632 if (!direct) {
1167 :     uv_dx = pMB->mvs[0].x;
1168 :     uv_dy = pMB->mvs[0].y;
1169 :     b_uv_dx = pMB->b_mvs[0].x;
1170 :     b_uv_dy = pMB->b_mvs[0].y;
1171 : edgomez 851
1172 : suxen_drol 1632 if (dec->quarterpel) {
1173 : Isibaar 1628 if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
1174 :     uv_dx = (uv_dx>>1) | (uv_dx&1);
1175 :     uv_dy = (uv_dy>>1) | (uv_dy&1);
1176 :     b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1);
1177 :     b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1);
1178 :     }
1179 :     else {
1180 : suxen_drol 1632 uv_dx /= 2;
1181 :     uv_dy /= 2;
1182 :     b_uv_dx /= 2;
1183 :     b_uv_dy /= 2;
1184 :     }
1185 :     }
1186 : edgomez 851
1187 : suxen_drol 1632 uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
1188 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
1189 :     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
1190 :     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
1191 : edgomez 1382
1192 : suxen_drol 1632 } else {
1193 : 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. */
1194 :     /* TODO: figure out if we ever did it wrong on the encoder side. If yes, add some workaround */
1195 :     if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) {
1196 :     int z;
1197 :     uv_dx = 0; uv_dy = 0;
1198 :     b_uv_dx = 0; b_uv_dy = 0;
1199 :     for (z = 0; z < 4; z++) {
1200 :     uv_dx += ((pMB->mvs[z].x>>1) | (pMB->mvs[z].x&1));
1201 :     uv_dy += ((pMB->mvs[z].y>>1) | (pMB->mvs[z].y&1));
1202 :     b_uv_dx += ((pMB->b_mvs[z].x>>1) | (pMB->b_mvs[z].x&1));
1203 :     b_uv_dy += ((pMB->b_mvs[z].y>>1) | (pMB->b_mvs[z].y&1));
1204 : Isibaar 1628 }
1205 : Isibaar 1669 }
1206 :     else {
1207 :     uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
1208 :     uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
1209 :     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);
1210 :     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);
1211 :     }
1212 :     } else {
1213 :     uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
1214 :     uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
1215 :     b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
1216 :     b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
1217 : suxen_drol 1632 }
1218 : edgomez 851
1219 : suxen_drol 1632 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
1220 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
1221 :     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
1222 :     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
1223 :     }
1224 : edgomez 851
1225 : suxen_drol 1632 start_timer();
1226 :     if(dec->quarterpel) {
1227 :     if(!direct) {
1228 :     interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1229 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1230 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1231 :     } else {
1232 :     interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1233 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1234 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1235 :     interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1236 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1237 :     pMB->mvs[1].x, pMB->mvs[1].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, 16*y_pos + 8,
1240 :     pMB->mvs[2].x, pMB->mvs[2].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 + 8, 16*y_pos + 8,
1243 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1244 :     }
1245 :     } else {
1246 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
1247 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1248 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
1249 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1250 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
1251 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1252 :     interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8,
1253 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1254 :     }
1255 : edgomez 851
1256 : suxen_drol 1632 interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
1257 :     uv_dy, stride2, 0);
1258 :     interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
1259 :     uv_dy, stride2, 0);
1260 : edgomez 851
1261 :    
1262 : suxen_drol 1632 if(dec->quarterpel) {
1263 :     if(!direct) {
1264 :     interpolate16x16_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1265 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1266 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1267 :     } else {
1268 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1269 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1270 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1271 :     interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1272 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1273 :     pMB->b_mvs[1].x, pMB->b_mvs[1].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, 16*y_pos + 8,
1276 :     pMB->b_mvs[2].x, pMB->b_mvs[2].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 + 8, 16*y_pos + 8,
1279 :     pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1280 :     }
1281 :     } else {
1282 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos,
1283 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1284 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
1285 :     16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1286 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos,
1287 :     16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1288 :     interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
1289 :     16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1290 :     }
1291 : edgomez 851
1292 : suxen_drol 1632 interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos,
1293 :     b_uv_dx, b_uv_dy, stride2, 0);
1294 :     interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos,
1295 :     b_uv_dx, b_uv_dy, stride2, 0);
1296 : edgomez 851
1297 : suxen_drol 1632 stop_comp_timer();
1298 : edgomez 851
1299 : suxen_drol 1632 if (cbp)
1300 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
1301 : edgomez 851 }
1302 :    
1303 : edgomez 866 /* for decode B-frame dbquant */
1304 : edgomez 1382 static __inline int32_t
1305 : edgomez 851 get_dbquant(Bitstream * bs)
1306 :     {
1307 : suxen_drol 1632 if (!BitstreamGetBit(bs)) /* '0' */
1308 :     return (0);
1309 :     else if (!BitstreamGetBit(bs)) /* '10' */
1310 :     return (-2);
1311 :     else /* '11' */
1312 :     return (2);
1313 : edgomez 851 }
1314 :    
1315 : edgomez 866 /*
1316 : edgomez 1382 * decode B-frame mb_type
1317 : suxen_drol 1632 * bit ret_value
1318 :     * 1 0
1319 :     * 01 1
1320 :     * 001 2
1321 :     * 0001 3
1322 : edgomez 866 */
1323 : edgomez 1382 static int32_t __inline
1324 : edgomez 851 get_mbtype(Bitstream * bs)
1325 :     {
1326 : suxen_drol 1632 int32_t mb_type;
1327 : edgomez 851
1328 : suxen_drol 1632 for (mb_type = 0; mb_type <= 3; mb_type++)
1329 :     if (BitstreamGetBit(bs))
1330 :     return (mb_type);
1331 : edgomez 851
1332 : suxen_drol 1632 return -1;
1333 : edgomez 851 }
1334 :    
1335 : edgomez 1382 static void
1336 : edgomez 851 decoder_bframe(DECODER * dec,
1337 : suxen_drol 1632 Bitstream * bs,
1338 :     int quant,
1339 :     int fcode_forward,
1340 :     int fcode_backward)
1341 : edgomez 851 {
1342 : suxen_drol 1632 uint32_t x, y;
1343 :     VECTOR mv;
1344 :     const VECTOR zeromv = {0,0};
1345 :     int i;
1346 : edgomez 851
1347 : suxen_drol 1632 if (!dec->is_edged[0]) {
1348 :     start_timer();
1349 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
1350 :     dec->width, dec->height, dec->bs_version);
1351 :     dec->is_edged[0] = 1;
1352 :     stop_edges_timer();
1353 :     }
1354 : edgomez 851
1355 : suxen_drol 1632 if (!dec->is_edged[1]) {
1356 :     start_timer();
1357 :     image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
1358 :     dec->width, dec->height, dec->bs_version);
1359 :     dec->is_edged[1] = 1;
1360 :     stop_edges_timer();
1361 :     }
1362 : edgomez 1451
1363 : suxen_drol 1632 for (y = 0; y < dec->mb_height; y++) {
1364 :     /* Initialize Pred Motion Vector */
1365 :     dec->p_fmv = dec->p_bmv = zeromv;
1366 :     for (x = 0; x < dec->mb_width; x++) {
1367 :     MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
1368 :     MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
1369 :     const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward;
1370 : suxen_drol 1639 int intra_dc_threshold; /* fake variable */
1371 : edgomez 851
1372 : suxen_drol 1632 if (check_resync_marker(bs, fcode_max - 1)) {
1373 :     int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant,
1374 :     &fcode_forward, &fcode_backward, &intra_dc_threshold);
1375 :     x = bound % dec->mb_width;
1376 :     y = bound / dec->mb_width;
1377 :     /* reset predicted macroblocks */
1378 :     dec->p_fmv = dec->p_bmv = zeromv;
1379 :     }
1380 : edgomez 1382
1381 : suxen_drol 1632 mv =
1382 :     mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
1383 :     mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
1384 :     mb->quant = quant;
1385 : edgomez 851
1386 : suxen_drol 1632 /*
1387 :     * skip if the co-located P_VOP macroblock is not coded
1388 :     * if not codec in co-located S_VOP macroblock is _not_
1389 :     * automatically skipped
1390 :     */
1391 : edgomez 851
1392 : suxen_drol 1632 if (last_mb->mode == MODE_NOT_CODED) {
1393 :     mb->cbp = 0;
1394 :     mb->mode = MODE_FORWARD;
1395 : Isibaar 1669 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1);
1396 : suxen_drol 1632 continue;
1397 :     }
1398 : edgomez 851
1399 : suxen_drol 1632 if (!BitstreamGetBit(bs)) { /* modb=='0' */
1400 :     const uint8_t modb2 = BitstreamGetBit(bs);
1401 : edgomez 851
1402 : suxen_drol 1632 mb->mode = get_mbtype(bs);
1403 : edgomez 851
1404 : suxen_drol 1632 if (!modb2) /* modb=='00' */
1405 :     mb->cbp = BitstreamGetBits(bs, 6);
1406 :     else
1407 :     mb->cbp = 0;
1408 : edgomez 851
1409 : suxen_drol 1632 if (mb->mode && mb->cbp) {
1410 :     quant += get_dbquant(bs);
1411 :     if (quant > 31)
1412 :     quant = 31;
1413 :     else if (quant < 1)
1414 :     quant = 1;
1415 :     }
1416 :     mb->quant = quant;
1417 : edgomez 1382
1418 : suxen_drol 1632 if (dec->interlacing) {
1419 :     if (mb->cbp) {
1420 :     mb->field_dct = BitstreamGetBit(bs);
1421 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1422 :     }
1423 : edgomez 1382
1424 : suxen_drol 1632 if (mb->mode) {
1425 :     mb->field_pred = BitstreamGetBit(bs);
1426 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1427 : edgomez 1382
1428 : suxen_drol 1632 if (mb->field_pred) {
1429 :     mb->field_for_top = BitstreamGetBit(bs);
1430 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1431 :     mb->field_for_bot = BitstreamGetBit(bs);
1432 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1433 :     }
1434 :     }
1435 :     }
1436 : edgomez 1382
1437 : suxen_drol 1632 } else {
1438 :     mb->mode = MODE_DIRECT_NONE_MV;
1439 :     mb->cbp = 0;
1440 :     }
1441 : edgomez 851
1442 : suxen_drol 1632 switch (mb->mode) {
1443 :     case MODE_DIRECT:
1444 :     get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y);
1445 : edgomez 851
1446 : suxen_drol 1632 case MODE_DIRECT_NONE_MV:
1447 :     for (i = 0; i < 4; i++) {
1448 :     mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x;
1449 :     mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y;
1450 : edgomez 1382
1451 : suxen_drol 1632 mb->b_mvs[i].x = (mv.x)
1452 :     ? mb->mvs[i].x - last_mb->mvs[i].x
1453 :     : last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp;
1454 :     mb->b_mvs[i].y = (mv.y)
1455 :     ? mb->mvs[i].y - last_mb->mvs[i].y
1456 :     : last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp;
1457 :     }
1458 : edgomez 851
1459 : suxen_drol 1632 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1460 :     mb, x, y, bs, 1);
1461 :     break;
1462 : edgomez 851
1463 : suxen_drol 1632 case MODE_INTERPOLATE:
1464 :     get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1465 :     dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1466 : edgomez 851
1467 : suxen_drol 1632 get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1468 :     dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
1469 : edgomez 851
1470 : suxen_drol 1632 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1471 :     mb, x, y, bs, 0);
1472 :     break;
1473 : edgomez 851
1474 : suxen_drol 1632 case MODE_BACKWARD:
1475 :     get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
1476 :     dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1477 : edgomez 851
1478 : Isibaar 1669 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1479 : suxen_drol 1632 break;
1480 : edgomez 851
1481 : suxen_drol 1632 case MODE_FORWARD:
1482 :     get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
1483 :     dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1484 : edgomez 851
1485 : Isibaar 1669 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1);
1486 : suxen_drol 1632 break;
1487 :    
1488 :     default:
1489 :     DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
1490 :     }
1491 :     } /* End of for */
1492 :     }
1493 : edgomez 851 }
1494 :    
1495 :     /* perform post processing if necessary, and output the image */
1496 : suxen_drol 1653 static void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
1497 : suxen_drol 1632 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats,
1498 :     int coding_type, int quant)
1499 : edgomez 851 {
1500 : suxen_drol 1632 const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0;
1501 : suxen_drol 1431
1502 : suxen_drol 1632 if (dec->cartoon_mode)
1503 :     frame->general &= ~XVID_FILMEFFECT;
1504 : edgomez 851
1505 : suxen_drol 1632 if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0)
1506 :     && mbs != NULL) /* post process */
1507 :     {
1508 :     /* note: image is stored to tmp */
1509 :     image_copy(&dec->tmp, img, dec->edged_width, dec->height);
1510 :     image_postproc(&dec->postproc, &dec->tmp, dec->edged_width,
1511 :     mbs, dec->mb_width, dec->mb_height, dec->mb_width,
1512 :     frame->general, brightness, dec->frames, (coding_type == B_VOP));
1513 :     img = &dec->tmp;
1514 :     }
1515 : edgomez 851
1516 : suxen_drol 1632 image_output(img, dec->width, dec->height,
1517 :     dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
1518 :     frame->output.csp, dec->interlacing);
1519 : edgomez 1382
1520 : suxen_drol 1632 if (stats) {
1521 :     stats->type = coding2type(coding_type);
1522 :     stats->data.vop.time_base = (int)dec->time_base;
1523 :     stats->data.vop.time_increment = 0; /* XXX: todo */
1524 :     stats->data.vop.qscale_stride = dec->mb_width;
1525 :     stats->data.vop.qscale = dec->qscale;
1526 :     if (stats->data.vop.qscale != NULL && mbs != NULL) {
1527 :     unsigned int i;
1528 :     for (i = 0; i < dec->mb_width*dec->mb_height; i++)
1529 :     stats->data.vop.qscale[i] = mbs[i].quant;
1530 :     } else
1531 :     stats->data.vop.qscale = NULL;
1532 :     }
1533 : edgomez 851 }
1534 :    
1535 : edgomez 195 int
1536 :     decoder_decode(DECODER * dec,
1537 : suxen_drol 1632 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
1538 : Isibaar 3 {
1539 : edgomez 78
1540 : suxen_drol 1632 Bitstream bs;
1541 :     uint32_t rounding;
1542 :     uint32_t quant = 2;
1543 :     uint32_t fcode_forward;
1544 :     uint32_t fcode_backward;
1545 :     uint32_t intra_dc_threshold;
1546 :     WARPPOINTS gmc_warp;
1547 :     int coding_type;
1548 :     int success, output, seen_something;
1549 : Isibaar 3
1550 : suxen_drol 1632 if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */
1551 :     return XVID_ERR_VERSION;
1552 : edgomez 1382
1553 : suxen_drol 1632 start_global_timer();
1554 : edgomez 195
1555 : suxen_drol 1632 dec->low_delay_default = (frame->general & XVID_LOWDELAY);
1556 :     if ((frame->general & XVID_DISCONTINUITY))
1557 :     dec->frames = 0;
1558 :     dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
1559 : edgomez 851
1560 : suxen_drol 1632 if(frame->length<0) { /* decoder flush */
1561 :     int ret;
1562 :     /* if not decoding "low_delay/packed", and this isn't low_delay and
1563 :     we have a reference frame, then outout the reference frame */
1564 :     if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
1565 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1566 :     dec->frames = 0;
1567 :     ret = 0;
1568 :     } else {
1569 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1570 :     ret = XVID_ERR_END;
1571 :     }
1572 : edgomez 851
1573 : suxen_drol 1632 emms();
1574 :     stop_global_timer();
1575 :     return ret;
1576 :     }
1577 : edgomez 851
1578 : suxen_drol 1632 BitstreamInit(&bs, frame->bitstream, frame->length);
1579 : Isibaar 3
1580 : suxen_drol 1632 /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
1581 :     if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
1582 :     {
1583 :     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
1584 :     (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
1585 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1586 :     emms();
1587 :     return 1; /* one byte consumed */
1588 :     }
1589 : Isibaar 179
1590 : suxen_drol 1632 success = 0;
1591 :     output = 0;
1592 :     seen_something = 0;
1593 : edgomez 1382
1594 : edgomez 851 repeat:
1595 : edgomez 195
1596 : suxen_drol 1632 coding_type = BitstreamReadHeaders(&bs, dec, &rounding,
1597 :     &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
1598 : Isibaar 3
1599 : suxen_drol 1632 DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%"
1600 : suxen_drol 1607 #if defined(_MSC_VER)
1601 :     "I64"
1602 :     #else
1603 :     "ll"
1604 :     #endif
1605 :     "i, time_pp=%i, time_bp=%i\n",
1606 : suxen_drol 1632 coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
1607 : Isibaar 3
1608 : suxen_drol 1632 if (coding_type == -1) { /* nothing */
1609 :     if (success) goto done;
1610 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1611 :     emms();
1612 :     return BitstreamPos(&bs)/8;
1613 :     }
1614 : Isibaar 3
1615 : suxen_drol 1632 if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
1616 : edgomez 1382
1617 : suxen_drol 1632 if (coding_type == -3)
1618 :     decoder_resize(dec);
1619 : edgomez 1382
1620 : suxen_drol 1632 if(stats) {
1621 :     stats->type = XVID_TYPE_VOL;
1622 :     stats->data.vol.general = 0;
1623 :     /*XXX: if (dec->interlacing)
1624 :     stats->data.vol.general |= ++INTERLACING; */
1625 :     stats->data.vol.width = dec->width;
1626 :     stats->data.vol.height = dec->height;
1627 :     stats->data.vol.par = dec->aspect_ratio;
1628 :     stats->data.vol.par_width = dec->par_width;
1629 :     stats->data.vol.par_height = dec->par_height;
1630 :     emms();
1631 :     return BitstreamPos(&bs)/8; /* number of bytes consumed */
1632 :     }
1633 :     goto repeat;
1634 :     }
1635 : Isibaar 3
1636 : suxen_drol 1632 if(dec->frames == 0 && coding_type != I_VOP) {
1637 :     /* 1st frame is not an i-vop */
1638 :     goto repeat;
1639 :     }
1640 : syskin 1439
1641 : 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 */
1642 : chenm001 290
1643 : suxen_drol 1632 /* packed_mode: special-N_VOP treament */
1644 :     if (dec->packed_mode && coding_type == N_VOP) {
1645 :     if (dec->low_delay_default && dec->frames > 0) {
1646 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1647 :     output = 1;
1648 :     }
1649 :     /* ignore otherwise */
1650 :     } else if (coding_type != B_VOP) {
1651 :     switch(coding_type) {
1652 :     case I_VOP :
1653 :     decoder_iframe(dec, &bs, quant, intra_dc_threshold);
1654 :     break;
1655 :     case P_VOP :
1656 :     decoder_pframe(dec, &bs, rounding, quant,
1657 :     fcode_forward, intra_dc_threshold, NULL);
1658 :     break;
1659 :     case S_VOP :
1660 :     decoder_pframe(dec, &bs, rounding, quant,
1661 :     fcode_forward, intra_dc_threshold, &gmc_warp);
1662 :     break;
1663 :     case N_VOP :
1664 :     /* XXX: not_coded vops are not used for forward prediction */
1665 :     /* we should not swap(last_mbs,mbs) */
1666 :     image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
1667 :     SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */
1668 :     break;
1669 :     }
1670 : edgomez 851
1671 : suxen_drol 1632 /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
1672 :     if (!(dec->low_delay_default && dec->packed_mode)) {
1673 :     if(dec->low_delay) {
1674 :     decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1675 :     output = 1;
1676 :     } else if (dec->frames > 0) { /* is the reference frame valid? */
1677 :     /* output the reference frame */
1678 :     decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1679 :     output = 1;
1680 :     }
1681 :     }
1682 :    
1683 :     image_swap(&dec->refn[0], &dec->refn[1]);
1684 :     dec->is_edged[1] = dec->is_edged[0];
1685 :     image_swap(&dec->cur, &dec->refn[0]);
1686 :     dec->is_edged[0] = 0;
1687 :     SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
1688 :     dec->last_coding_type = coding_type;
1689 : edgomez 851
1690 : suxen_drol 1632 dec->frames++;
1691 :     seen_something = 1;
1692 : chenm001 306
1693 : suxen_drol 1632 } else { /* B_VOP */
1694 : edgomez 851
1695 : suxen_drol 1632 if (dec->low_delay) {
1696 :     DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
1697 :     dec->low_delay = 0;
1698 :     }
1699 : edgomez 851
1700 : suxen_drol 1632 if (dec->frames < 2) {
1701 :     /* attemping to decode a bvop without atleast 2 reference frames */
1702 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1703 :     "broken b-frame, mising ref frames");
1704 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1705 :     } else if (dec->time_pp <= dec->time_bp) {
1706 :     /* this occurs when dx50_bvop_compatibility==0 sequences are
1707 :     decoded in vfw. */
1708 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1709 :     "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
1710 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1711 :     } else {
1712 :     decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
1713 :     decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant);
1714 :     }
1715 : edgomez 851
1716 : suxen_drol 1632 output = 1;
1717 :     dec->frames++;
1718 :     }
1719 : edgomez 851
1720 : edgomez 1382 #if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */
1721 : suxen_drol 1632 BitstreamByteAlign(&bs);
1722 : edgomez 1382 #endif
1723 : edgomez 851
1724 : suxen_drol 1632 /* low_delay_default mode: repeat in packed_mode */
1725 :     if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
1726 :     success = 1;
1727 :     goto repeat;
1728 :     }
1729 : edgomez 851
1730 :     done :
1731 :    
1732 : suxen_drol 1607 /* if we reach here without outputing anything _and_
1733 :     the calling application has specified low_delay_default,
1734 :     we *must* output something.
1735 :     this always occurs on the first call to decode() call
1736 :     when bframes are present in the bitstream. it may also
1737 :     occur if no vops were seen in the bitstream
1738 :    
1739 :     if packed_mode is enabled, then we output the recently
1740 :     decoded frame (the very first ivop). otherwise we have
1741 :     nothing to display, and therefore output a black screen.
1742 :     */
1743 :     if (dec->low_delay_default && output == 0) {
1744 :     if (dec->packed_mode && seen_something) {
1745 : suxen_drol 1632 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
1746 : suxen_drol 1607 } else {
1747 : suxen_drol 1632 image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
1748 :     decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant);
1749 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1750 :     }
1751 :     }
1752 : edgomez 851
1753 : suxen_drol 1632 emms();
1754 :     stop_global_timer();
1755 : Isibaar 3
1756 : suxen_drol 1632 return (BitstreamPos(&bs)+7)/8; /* number of bytes consumed */
1757 : Isibaar 3 }

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