?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'mirror' into vdpau
[FFMpeg-mirror/ffmpeg-vdpau.git] / libavcodec / vc1.c
blob83b9fe818ba7eb2a19f0bb65889f2d7e6b27663d
1 /*
2 * VC-1 and WMV3 decoder
3 * Copyright (c) 2006-2007 Konstantin Shishkov
4 * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 /**
24 * @file vc1.c
25 * VC-1 and WMV3 decoder
26 *
27 */
28 #include "dsputil.h"
29 #include "avcodec.h"
30 #include "mpegvideo.h"
31 #include "vc1.h"
32 #include "vc1data.h"
33 #include "vc1acdata.h"
34 #include "msmpeg4data.h"
35 #include "unary.h"
36 #include "simple_idct.h"
37
38 #undef NDEBUG
39 #include <assert.h>
40
41 #define MB_INTRA_VLC_BITS 9
42 #define DC_VLC_BITS 9
43 #define AC_VLC_BITS 9
44
45 extern int ff_VDPAU_vc1_decode_picture(MpegEncContext *s, AVCodecContext *avctx, VC1Context *v, const uint8_t *buf, int buf_size);
46
47 static const uint16_t table_mb_intra[64][2];
48
49 #ifdef HAVE_VDPAU
50 static const enum PixelFormat pixfmt_vdpau_vc1_simple_420[] = {
51 PIX_FMT_VDPAU_VC1_SIMPLE,
52 PIX_FMT_NONE};
53 static const enum PixelFormat pixfmt_vdpau_vc1_main_420[] = {
54 PIX_FMT_VDPAU_VC1_MAIN,
55 PIX_FMT_NONE};
56 static const enum PixelFormat pixfmt_vdpau_vc1_advanced_420[] = {
57 PIX_FMT_VDPAU_VC1_ADVANCED,
58 PIX_FMT_NONE};
59 #endif
60
61 /**
62 * Init VC-1 specific tables and VC1Context members
63 * @param v The VC1Context to initialize
64 * @return Status
65 */
66 static int vc1_init_common(VC1Context *v)
67 {
68 static int done = 0;
69 int i = 0;
70
71 v->hrd_rate = v->hrd_buffer = NULL;
72
73 /* VLC tables */
74 if(!done)
75 {
76 done = 1;
77 init_vlc(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,
78 ff_vc1_bfraction_bits, 1, 1,
79 ff_vc1_bfraction_codes, 1, 1, 1);
80 init_vlc(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,
81 ff_vc1_norm2_bits, 1, 1,
82 ff_vc1_norm2_codes, 1, 1, 1);
83 init_vlc(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,
84 ff_vc1_norm6_bits, 1, 1,
85 ff_vc1_norm6_codes, 2, 2, 1);
86 init_vlc(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,
87 ff_vc1_imode_bits, 1, 1,
88 ff_vc1_imode_codes, 1, 1, 1);
89 for (i=0; i<3; i++)
90 {
91 init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,
92 ff_vc1_ttmb_bits[i], 1, 1,
93 ff_vc1_ttmb_codes[i], 2, 2, 1);
94 init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,
95 ff_vc1_ttblk_bits[i], 1, 1,
96 ff_vc1_ttblk_codes[i], 1, 1, 1);
97 init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,
98 ff_vc1_subblkpat_bits[i], 1, 1,
99 ff_vc1_subblkpat_codes[i], 1, 1, 1);
100 }
101 for(i=0; i<4; i++)
102 {
103 init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,
104 ff_vc1_4mv_block_pattern_bits[i], 1, 1,
105 ff_vc1_4mv_block_pattern_codes[i], 1, 1, 1);
106 init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,
107 ff_vc1_cbpcy_p_bits[i], 1, 1,
108 ff_vc1_cbpcy_p_codes[i], 2, 2, 1);
109 init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,
110 ff_vc1_mv_diff_bits[i], 1, 1,
111 ff_vc1_mv_diff_codes[i], 2, 2, 1);
112 }
113 for(i=0; i<8; i++)
114 init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],
115 &vc1_ac_tables[i][0][1], 8, 4,
116 &vc1_ac_tables[i][0][0], 8, 4, 1);
117 init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,
118 &ff_msmp4_mb_i_table[0][1], 4, 2,
119 &ff_msmp4_mb_i_table[0][0], 4, 2, 1);
120 }
121
122 /* Other defaults */
123 v->pq = -1;
124 v->mvrange = 0; /* 7.1.1.18, p80 */
125
126 return 0;
127 }
128
129 /***********************************************************************/
130 /**
131 * @defgroup bitplane VC9 Bitplane decoding
132 * @see 8.7, p56
133 * @{
134 */
135
136 /** @addtogroup bitplane
137 * Imode types
138 * @{
139 */
140 enum Imode {
141 IMODE_RAW,
142 IMODE_NORM2,
143 IMODE_DIFF2,
144 IMODE_NORM6,
145 IMODE_DIFF6,
146 IMODE_ROWSKIP,
147 IMODE_COLSKIP
148 };
149 /** @} */ //imode defines
150
151 /** Decode rows by checking if they are skipped
152 * @param plane Buffer to store decoded bits
153 * @param[in] width Width of this buffer
154 * @param[in] height Height of this buffer
155 * @param[in] stride of this buffer
156 */
157 static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
158 int x, y;
159
160 for (y=0; y<height; y++){
161 if (!get_bits1(gb)) //rowskip
162 memset(plane, 0, width);
163 else
164 for (x=0; x<width; x++)
165 plane[x] = get_bits1(gb);
166 plane += stride;
167 }
168 }
169
170 /** Decode columns by checking if they are skipped
171 * @param plane Buffer to store decoded bits
172 * @param[in] width Width of this buffer
173 * @param[in] height Height of this buffer
174 * @param[in] stride of this buffer
175 * @todo FIXME: Optimize
176 */
177 static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
178 int x, y;
179
180 for (x=0; x<width; x++){
181 if (!get_bits1(gb)) //colskip
182 for (y=0; y<height; y++)
183 plane[y*stride] = 0;
184 else
185 for (y=0; y<height; y++)
186 plane[y*stride] = get_bits1(gb);
187 plane ++;
188 }
189 }
190
191 /** Decode a bitplane's bits
192 * @param bp Bitplane where to store the decode bits
193 * @param v VC-1 context for bit reading and logging
194 * @return Status
195 * @todo FIXME: Optimize
196 */
197 static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)
198 {
199 GetBitContext *gb = &v->s.gb;
200
201 int imode, x, y, code, offset;
202 uint8_t invert, *planep = data;
203 int width, height, stride;
204
205 width = v->s.mb_width;
206 height = v->s.mb_height;
207 stride = v->s.mb_stride;
208 invert = get_bits1(gb);
209 imode = get_vlc2(gb, ff_vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);
210
211 *raw_flag = 0;
212 switch (imode)
213 {
214 case IMODE_RAW:
215 //Data is actually read in the MB layer (same for all tests == "raw")
216 *raw_flag = 1; //invert ignored
217 return invert;
218 case IMODE_DIFF2:
219 case IMODE_NORM2:
220 if ((height * width) & 1)
221 {
222 *planep++ = get_bits1(gb);
223 offset = 1;
224 }
225 else offset = 0;
226 // decode bitplane as one long line
227 for (y = offset; y < height * width; y += 2) {
228 code = get_vlc2(gb, ff_vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
229 *planep++ = code & 1;
230 offset++;
231 if(offset == width) {
232 offset = 0;
233 planep += stride - width;
234 }
235 *planep++ = code >> 1;
236 offset++;
237 if(offset == width) {
238 offset = 0;
239 planep += stride - width;
240 }
241 }
242 break;
243 case IMODE_DIFF6:
244 case IMODE_NORM6:
245 if(!(height % 3) && (width % 3)) { // use 2x3 decoding
246 for(y = 0; y < height; y+= 3) {
247 for(x = width & 1; x < width; x += 2) {
248 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
249 if(code < 0){
250 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
251 return -1;
252 }
253 planep[x + 0] = (code >> 0) & 1;
254 planep[x + 1] = (code >> 1) & 1;
255 planep[x + 0 + stride] = (code >> 2) & 1;
256 planep[x + 1 + stride] = (code >> 3) & 1;
257 planep[x + 0 + stride * 2] = (code >> 4) & 1;
258 planep[x + 1 + stride * 2] = (code >> 5) & 1;
259 }
260 planep += stride * 3;
261 }
262 if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);
263 } else { // 3x2
264 planep += (height & 1) * stride;
265 for(y = height & 1; y < height; y += 2) {
266 for(x = width % 3; x < width; x += 3) {
267 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
268 if(code < 0){
269 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
270 return -1;
271 }
272 planep[x + 0] = (code >> 0) & 1;
273 planep[x + 1] = (code >> 1) & 1;
274 planep[x + 2] = (code >> 2) & 1;
275 planep[x + 0 + stride] = (code >> 3) & 1;
276 planep[x + 1 + stride] = (code >> 4) & 1;
277 planep[x + 2 + stride] = (code >> 5) & 1;
278 }
279 planep += stride * 2;
280 }
281 x = width % 3;
282 if(x) decode_colskip(data , x, height , stride, &v->s.gb);
283 if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);
284 }
285 break;
286 case IMODE_ROWSKIP:
287 decode_rowskip(data, width, height, stride, &v->s.gb);
288 break;
289 case IMODE_COLSKIP:
290 decode_colskip(data, width, height, stride, &v->s.gb);
291 break;
292 default: break;
293 }
294
295 /* Applying diff operator */
296 if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
297 {
298 planep = data;
299 planep[0] ^= invert;
300 for (x=1; x<width; x++)
301 planep[x] ^= planep[x-1];
302 for (y=1; y<height; y++)
303 {
304 planep += stride;
305 planep[0] ^= planep[-stride];
306 for (x=1; x<width; x++)
307 {
308 if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;
309 else planep[x] ^= planep[x-1];
310 }
311 }
312 }
313 else if (invert)
314 {
315 planep = data;
316 for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride
317 }
318 return (imode<<1) + invert;
319 }
320
321 /** @} */ //Bitplane group
322
323 #define FILTSIGN(a) ((a) >= 0 ? 1 : -1)
324 /**
325 * VC-1 in-loop deblocking filter for one line
326 * @param src source block type
327 * @param pq block quantizer
328 * @return whether other 3 pairs should be filtered or not
329 * @see 8.6
330 */
331 static int vc1_filter_line(uint8_t* src, int stride, int pq){
332 int a0, a1, a2, a3, d, clip, filt3 = 0;
333 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
334
335 a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
336 if(FFABS(a0) < pq){
337 a1 = (2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3;
338 a2 = (2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3;
339 a3 = FFMIN(FFABS(a1), FFABS(a2));
340 if(a3 < FFABS(a0)){
341 d = 5 * ((a0 >=0 ? a3 : -a3) - a0) / 8;
342 clip = (src[-1*stride] - src[ 0*stride])/2;
343 if(clip){
344 filt3 = 1;
345 if(clip > 0)
346 d = av_clip(d, 0, clip);
347 else
348 d = av_clip(d, clip, 0);
349 src[-1*stride] = cm[src[-1*stride] - d];
350 src[ 0*stride] = cm[src[ 0*stride] + d];
351 }
352 }
353 }
354 return filt3;
355 }
356
357 /**
358 * VC-1 in-loop deblocking filter
359 * @param src source block type
360 * @param len edge length to filter (4 or 8 pixels)
361 * @param pq block quantizer
362 * @see 8.6
363 */
364 static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
365 {
366 int i;
367 int filt3;
368
369 for(i = 0; i < len; i += 4){
370 filt3 = vc1_filter_line(src + 2*step, stride, pq);
371 if(filt3){
372 vc1_filter_line(src + 0*step, stride, pq);
373 vc1_filter_line(src + 1*step, stride, pq);
374 vc1_filter_line(src + 3*step, stride, pq);
375 }
376 src += step * 4;
377 }
378 }
379
380 static void vc1_loop_filter_iblk(MpegEncContext *s, int pq)
381 {
382 int i, j;
383 if(!s->first_slice_line)
384 vc1_loop_filter(s->dest[0], 1, s->linesize, 16, pq);
385 vc1_loop_filter(s->dest[0] + 8*s->linesize, 1, s->linesize, 16, pq);
386 for(i = !s->mb_x*8; i < 16; i += 8)
387 vc1_loop_filter(s->dest[0] + i, s->linesize, 1, 16, pq);
388 for(j = 0; j < 2; j++){
389 if(!s->first_slice_line)
390 vc1_loop_filter(s->dest[j+1], 1, s->uvlinesize, 8, pq);
391 if(s->mb_x)
392 vc1_loop_filter(s->dest[j+1], s->uvlinesize, 1, 8, pq);
393 }
394 }
395
396 /***********************************************************************/
397 /** VOP Dquant decoding
398 * @param v VC-1 Context
399 */
400 static int vop_dquant_decoding(VC1Context *v)
401 {
402 GetBitContext *gb = &v->s.gb;
403 int pqdiff;
404
405 //variable size
406 if (v->dquant == 2)
407 {
408 pqdiff = get_bits(gb, 3);
409 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
410 else v->altpq = v->pq + pqdiff + 1;
411 }
412 else
413 {
414 v->dquantfrm = get_bits1(gb);
415 if ( v->dquantfrm )
416 {
417 v->dqprofile = get_bits(gb, 2);
418 switch (v->dqprofile)
419 {
420 case DQPROFILE_SINGLE_EDGE:
421 case DQPROFILE_DOUBLE_EDGES:
422 v->dqsbedge = get_bits(gb, 2);
423 break;
424 case DQPROFILE_ALL_MBS:
425 v->dqbilevel = get_bits1(gb);
426 if(!v->dqbilevel)
427 v->halfpq = 0;
428 default: break; //Forbidden ?
429 }
430 if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)
431 {
432 pqdiff = get_bits(gb, 3);
433 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
434 else v->altpq = v->pq + pqdiff + 1;
435 }
436 }
437 }
438 return 0;
439 }
440
441 /** Put block onto picture
442 */
443 static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
444 {
445 uint8_t *Y;
446 int ys, us, vs;
447 DSPContext *dsp = &v->s.dsp;
448
449 if(v->rangeredfrm) {
450 int i, j, k;
451 for(k = 0; k < 6; k++)
452 for(j = 0; j < 8; j++)
453 for(i = 0; i < 8; i++)
454 block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;
455
456 }
457 ys = v->s.current_picture.linesize[0];
458 us = v->s.current_picture.linesize[1];
459 vs = v->s.current_picture.linesize[2];
460 Y = v->s.dest[0];
461
462 dsp->put_pixels_clamped(block[0], Y, ys);
463 dsp->put_pixels_clamped(block[1], Y + 8, ys);
464 Y += ys * 8;
465 dsp->put_pixels_clamped(block[2], Y, ys);
466 dsp->put_pixels_clamped(block[3], Y + 8, ys);
467
468 if(!(v->s.flags & CODEC_FLAG_GRAY)) {
469 dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
470 dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
471 }
472 }
473
474 /** Do motion compensation over 1 macroblock
475 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
476 */
477 static void vc1_mc_1mv(VC1Context *v, int dir)
478 {
479 MpegEncContext *s = &v->s;
480 DSPContext *dsp = &v->s.dsp;
481 uint8_t *srcY, *srcU, *srcV;
482 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
483
484 if(!v->s.last_picture.data[0])return;
485
486 mx = s->mv[dir][0][0];
487 my = s->mv[dir][0][1];
488
489 // store motion vectors for further use in B frames
490 if(s->pict_type == FF_P_TYPE) {
491 s->current_picture.motion_val[1][s->block_index[0]][0] = mx;
492 s->current_picture.motion_val[1][s->block_index[0]][1] = my;
493 }
494 uvmx = (mx + ((mx & 3) == 3)) >> 1;
495 uvmy = (my + ((my & 3) == 3)) >> 1;
496 if(v->fastuvmc) {
497 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
498 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
499 }
500 if(!dir) {
501 srcY = s->last_picture.data[0];
502 srcU = s->last_picture.data[1];
503 srcV = s->last_picture.data[2];
504 } else {
505 srcY = s->next_picture.data[0];
506 srcU = s->next_picture.data[1];
507 srcV = s->next_picture.data[2];
508 }
509
510 src_x = s->mb_x * 16 + (mx >> 2);
511 src_y = s->mb_y * 16 + (my >> 2);
512 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
513 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
514
515 if(v->profile != PROFILE_ADVANCED){
516 src_x = av_clip( src_x, -16, s->mb_width * 16);
517 src_y = av_clip( src_y, -16, s->mb_height * 16);
518 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
519 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
520 }else{
521 src_x = av_clip( src_x, -17, s->avctx->coded_width);
522 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
523 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
524 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
525 }
526
527 srcY += src_y * s->linesize + src_x;
528 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
529 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
530
531 /* for grayscale we should not try to read from unknown area */
532 if(s->flags & CODEC_FLAG_GRAY) {
533 srcU = s->edge_emu_buffer + 18 * s->linesize;
534 srcV = s->edge_emu_buffer + 18 * s->linesize;
535 }
536
537 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
538 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3
539 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){
540 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
541
542 srcY -= s->mspel * (1 + s->linesize);
543 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
544 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
545 srcY = s->edge_emu_buffer;
546 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
547 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
548 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
549 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
550 srcU = uvbuf;
551 srcV = uvbuf + 16;
552 /* if we deal with range reduction we need to scale source blocks */
553 if(v->rangeredfrm) {
554 int i, j;
555 uint8_t *src, *src2;
556
557 src = srcY;
558 for(j = 0; j < 17 + s->mspel*2; j++) {
559 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
560 src += s->linesize;
561 }
562 src = srcU; src2 = srcV;
563 for(j = 0; j < 9; j++) {
564 for(i = 0; i < 9; i++) {
565 src[i] = ((src[i] - 128) >> 1) + 128;
566 src2[i] = ((src2[i] - 128) >> 1) + 128;
567 }
568 src += s->uvlinesize;
569 src2 += s->uvlinesize;
570 }
571 }
572 /* if we deal with intensity compensation we need to scale source blocks */
573 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
574 int i, j;
575 uint8_t *src, *src2;
576
577 src = srcY;
578 for(j = 0; j < 17 + s->mspel*2; j++) {
579 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];
580 src += s->linesize;
581 }
582 src = srcU; src2 = srcV;
583 for(j = 0; j < 9; j++) {
584 for(i = 0; i < 9; i++) {
585 src[i] = v->lutuv[src[i]];
586 src2[i] = v->lutuv[src2[i]];
587 }
588 src += s->uvlinesize;
589 src2 += s->uvlinesize;
590 }
591 }
592 srcY += s->mspel * (1 + s->linesize);
593 }
594
595 if(s->mspel) {
596 dxy = ((my & 3) << 2) | (mx & 3);
597 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] , srcY , s->linesize, v->rnd);
598 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
599 srcY += s->linesize * 8;
600 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize , srcY , s->linesize, v->rnd);
601 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
602 } else { // hpel mc - always used for luma
603 dxy = (my & 2) | ((mx & 2) >> 1);
604
605 if(!v->rnd)
606 dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
607 else
608 dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
609 }
610
611 if(s->flags & CODEC_FLAG_GRAY) return;
612 /* Chroma MC always uses qpel bilinear */
613 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
614 uvmx = (uvmx&3)<<1;
615 uvmy = (uvmy&3)<<1;
616 if(!v->rnd){
617 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
618 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
619 }else{
620 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
621 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
622 }
623 }
624
625 /** Do motion compensation for 4-MV macroblock - luminance block
626 */
627 static void vc1_mc_4mv_luma(VC1Context *v, int n)
628 {
629 MpegEncContext *s = &v->s;
630 DSPContext *dsp = &v->s.dsp;
631 uint8_t *srcY;
632 int dxy, mx, my, src_x, src_y;
633 int off;
634
635 if(!v->s.last_picture.data[0])return;
636 mx = s->mv[0][n][0];
637 my = s->mv[0][n][1];
638 srcY = s->last_picture.data[0];
639
640 off = s->linesize * 4 * (n&2) + (n&1) * 8;
641
642 src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);
643 src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);
644
645 if(v->profile != PROFILE_ADVANCED){
646 src_x = av_clip( src_x, -16, s->mb_width * 16);
647 src_y = av_clip( src_y, -16, s->mb_height * 16);
648 }else{
649 src_x = av_clip( src_x, -17, s->avctx->coded_width);
650 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
651 }
652
653 srcY += src_y * s->linesize + src_x;
654
655 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
656 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2
657 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){
658 srcY -= s->mspel * (1 + s->linesize);
659 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,
660 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
661 srcY = s->edge_emu_buffer;
662 /* if we deal with range reduction we need to scale source blocks */
663 if(v->rangeredfrm) {
664 int i, j;
665 uint8_t *src;
666
667 src = srcY;
668 for(j = 0; j < 9 + s->mspel*2; j++) {
669 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
670 src += s->linesize;
671 }
672 }
673 /* if we deal with intensity compensation we need to scale source blocks */
674 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
675 int i, j;
676 uint8_t *src;
677
678 src = srcY;
679 for(j = 0; j < 9 + s->mspel*2; j++) {
680 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];
681 src += s->linesize;
682 }
683 }
684 srcY += s->mspel * (1 + s->linesize);
685 }
686
687 if(s->mspel) {
688 dxy = ((my & 3) << 2) | (mx & 3);
689 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);
690 } else { // hpel mc - always used for luma
691 dxy = (my & 2) | ((mx & 2) >> 1);
692 if(!v->rnd)
693 dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
694 else
695 dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
696 }
697 }
698
699 static inline int median4(int a, int b, int c, int d)
700 {
701 if(a < b) {
702 if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
703 else return (FFMIN(b, c) + FFMAX(a, d)) / 2;
704 } else {
705 if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
706 else return (FFMIN(a, c) + FFMAX(b, d)) / 2;
707 }
708 }
709
710
711 /** Do motion compensation for 4-MV macroblock - both chroma blocks
712 */
713 static void vc1_mc_4mv_chroma(VC1Context *v)
714 {
715 MpegEncContext *s = &v->s;
716 DSPContext *dsp = &v->s.dsp;
717 uint8_t *srcU, *srcV;
718 int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;
719 int i, idx, tx = 0, ty = 0;
720 int mvx[4], mvy[4], intra[4];
721 static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
722
723 if(!v->s.last_picture.data[0])return;
724 if(s->flags & CODEC_FLAG_GRAY) return;
725
726 for(i = 0; i < 4; i++) {
727 mvx[i] = s->mv[0][i][0];
728 mvy[i] = s->mv[0][i][1];
729 intra[i] = v->mb_type[0][s->block_index[i]];
730 }
731
732 /* calculate chroma MV vector from four luma MVs */
733 idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];
734 if(!idx) { // all blocks are inter
735 tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
736 ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
737 } else if(count[idx] == 1) { // 3 inter blocks
738 switch(idx) {
739 case 0x1:
740 tx = mid_pred(mvx[1], mvx[2], mvx[3]);
741 ty = mid_pred(mvy[1], mvy[2], mvy[3]);
742 break;
743 case 0x2:
744 tx = mid_pred(mvx[0], mvx[2], mvx[3]);
745 ty = mid_pred(mvy[0], mvy[2], mvy[3]);
746 break;
747 case 0x4:
748 tx = mid_pred(mvx[0], mvx[1], mvx[3]);
749 ty = mid_pred(mvy[0], mvy[1], mvy[3]);
750 break;
751 case 0x8:
752 tx = mid_pred(mvx[0], mvx[1], mvx[2]);
753 ty = mid_pred(mvy[0], mvy[1], mvy[2]);
754 break;
755 }
756 } else if(count[idx] == 2) {
757 int t1 = 0, t2 = 0;
758 for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}
759 for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}
760 tx = (mvx[t1] + mvx[t2]) / 2;
761 ty = (mvy[t1] + mvy[t2]) / 2;
762 } else {
763 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
764 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
765 return; //no need to do MC for inter blocks
766 }
767
768 s->current_picture.motion_val[1][s->block_index[0]][0] = tx;
769 s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
770 uvmx = (tx + ((tx&3) == 3)) >> 1;
771 uvmy = (ty + ((ty&3) == 3)) >> 1;
772 if(v->fastuvmc) {
773 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
774 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
775 }
776
777 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
778 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
779
780 if(v->profile != PROFILE_ADVANCED){
781 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
782 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
783 }else{
784 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
785 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
786 }
787
788 srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
789 srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
790 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
791 || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
792 || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){
793 ff_emulated_edge_mc(s->edge_emu_buffer , srcU, s->uvlinesize, 8+1, 8+1,
794 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
795 ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,
796 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
797 srcU = s->edge_emu_buffer;
798 srcV = s->edge_emu_buffer + 16;
799
800 /* if we deal with range reduction we need to scale source blocks */
801 if(v->rangeredfrm) {
802 int i, j;
803 uint8_t *src, *src2;
804
805 src = srcU; src2 = srcV;
806 for(j = 0; j < 9; j++) {
807 for(i = 0; i < 9; i++) {
808 src[i] = ((src[i] - 128) >> 1) + 128;
809 src2[i] = ((src2[i] - 128) >> 1) + 128;
810 }
811 src += s->uvlinesize;
812 src2 += s->uvlinesize;
813 }
814 }
815 /* if we deal with intensity compensation we need to scale source blocks */
816 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
817 int i, j;
818 uint8_t *src, *src2;
819
820 src = srcU; src2 = srcV;
821 for(j = 0; j < 9; j++) {
822 for(i = 0; i < 9; i++) {
823 src[i] = v->lutuv[src[i]];
824 src2[i] = v->lutuv[src2[i]];
825 }
826 src += s->uvlinesize;
827 src2 += s->uvlinesize;
828 }
829 }
830 }
831
832 /* Chroma MC always uses qpel bilinear */
833 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
834 uvmx = (uvmx&3)<<1;
835 uvmy = (uvmy&3)<<1;
836 if(!v->rnd){
837 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
838 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
839 }else{
840 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
841 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
842 }
843 }
844
845 #ifdef HAVE_VDPAU
846 static int decode_postinit(VC1Context *v, AVCodecContext *avctx)
847 {
848 if (avctx->pix_fmt != PIX_FMT_NONE){
849 return 0;
850 }
851
852 if (avctx->vdpau_acceleration) { // VC1
853 if (v->profile == 0) {
854 avctx->pix_fmt = avctx->get_format(avctx, pixfmt_vdpau_vc1_simple_420);
855 } else if (v->profile == 1) {
856 avctx->pix_fmt = avctx->get_format(avctx, pixfmt_vdpau_vc1_main_420);
857 } else if (v->profile == 3) {
858 avctx->pix_fmt = avctx->get_format(avctx, pixfmt_vdpau_vc1_advanced_420);
859 } else {
860 return -2;
861 }
862 }
863
864 return 0;
865 }
866 #endif
867
868 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);
869
870 /**
871 * Decode Simple/Main Profiles sequence header
872 * @see Figure 7-8, p16-17
873 * @param avctx Codec context
874 * @param gb GetBit context initialized from Codec context extra_data
875 * @return Status
876 */
877 static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)
878 {
879 VC1Context *v = avctx->priv_data;
880
881 av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));
882 v->profile = get_bits(gb, 2);
883 if (v->profile == PROFILE_COMPLEX)
884 {
885 av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");
886 }
887
888 if (v->profile == PROFILE_ADVANCED)
889 {
890 v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz;
891 v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz;
892 return decode_sequence_header_adv(v, gb);
893 }
894 else
895 {
896 v->zz_8x4 = wmv2_scantableA;
897 v->zz_4x8 = wmv2_scantableB;
898 v->res_sm = get_bits(gb, 2); //reserved
899 if (v->res_sm)
900 {
901 av_log(avctx, AV_LOG_ERROR,
902 "Reserved RES_SM=%i is forbidden\n", v->res_sm);
903 return -1;
904 }
905 }
906
907 // (fps-2)/4 (->30)
908 v->frmrtq_postproc = get_bits(gb, 3); //common
909 // (bitrate-32kbps)/64kbps
910 v->bitrtq_postproc = get_bits(gb, 5); //common
911 v->s.loop_filter = get_bits1(gb); //common
912 if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)
913 {
914 av_log(avctx, AV_LOG_ERROR,
915 "LOOPFILTER shell not be enabled in simple profile\n");
916 }
917 if(v->s.avctx->skip_loop_filter >= AVDISCARD_ALL)
918 v->s.loop_filter = 0;
919
920 v->res_x8 = get_bits1(gb); //reserved
921 v->multires = get_bits1(gb);
922 v->res_fasttx = get_bits1(gb);
923 if (!v->res_fasttx)
924 {
925 v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct;
926 v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;
927 v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;
928 v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;
929 }
930
931 v->fastuvmc = get_bits1(gb); //common
932 if (!v->profile && !v->fastuvmc)
933 {
934 av_log(avctx, AV_LOG_ERROR,
935 "FASTUVMC unavailable in Simple Profile\n");
936 return -1;
937 }
938 v->extended_mv = get_bits1(gb); //common
939 if (!v->profile && v->extended_mv)
940 {
941 av_log(avctx, AV_LOG_ERROR,
942 "Extended MVs unavailable in Simple Profile\n");
943 return -1;
944 }
945 v->dquant = get_bits(gb, 2); //common
946 v->vstransform = get_bits1(gb); //common
947
948 v->res_transtab = get_bits1(gb);
949 if (v->res_transtab)
950 {
951 av_log(avctx, AV_LOG_ERROR,
952 "1 for reserved RES_TRANSTAB is forbidden\n");
953 return -1;
954 }
955
956 v->overlap = get_bits1(gb); //common
957
958 v->s.resync_marker = get_bits1(gb);
959 v->rangered = get_bits1(gb);
960 if (v->rangered && v->profile == PROFILE_SIMPLE)
961 {
962 av_log(avctx, AV_LOG_INFO,
963 "RANGERED should be set to 0 in simple profile\n");
964 }
965
966 v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common
967 v->quantizer_mode = get_bits(gb, 2); //common
968
969 v->finterpflag = get_bits1(gb); //common
970 v->res_rtm_flag = get_bits1(gb); //reserved
971 if (!v->res_rtm_flag)
972 {
973 // av_log(avctx, AV_LOG_ERROR,
974 // "0 for reserved RES_RTM_FLAG is forbidden\n");
975 av_log(avctx, AV_LOG_ERROR,
976 "Old WMV3 version detected, only I-frames will be decoded\n");
977 //return -1;
978 }
979 //TODO: figure out what they mean (always 0x402F)
980 if(!v->res_fasttx) skip_bits(gb, 16);
981 av_log(avctx, AV_LOG_DEBUG,
982 "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
983 "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"
984 "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"
985 "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",
986 v->profile, v->frmrtq_postproc, v->bitrtq_postproc,
987 v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,
988 v->rangered, v->vstransform, v->overlap, v->s.resync_marker,
989 v->dquant, v->quantizer_mode, avctx->max_b_frames
990 );
991 return 0;
992 }
993
994 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)
995 {
996 v->res_rtm_flag = 1;
997 v->level = get_bits(gb, 3);
998 if(v->level >= 5)
999 {
1000 av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);
1001 }
1002 v->chromaformat = get_bits(gb, 2);
1003 if (v->chromaformat != 1)
1004 {
1005 av_log(v->s.avctx, AV_LOG_ERROR,
1006 "Only 4:2:0 chroma format supported\n");
1007 return -1;
1008 }
1009
1010 // (fps-2)/4 (->30)
1011 v->frmrtq_postproc = get_bits(gb, 3); //common
1012 // (bitrate-32kbps)/64kbps
1013 v->bitrtq_postproc = get_bits(gb, 5); //common
1014 v->postprocflag = get_bits1(gb); //common
1015
1016 v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;
1017 v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;
1018 v->s.avctx->width = v->s.avctx->coded_width;
1019 v->s.avctx->height = v->s.avctx->coded_height;
1020 v->broadcast = get_bits1(gb);
1021 v->interlace = get_bits1(gb);
1022 v->tfcntrflag = get_bits1(gb);
1023 v->finterpflag = get_bits1(gb);
1024 skip_bits1(gb); // reserved
1025
1026 v->s.h_edge_pos = v->s.avctx->coded_width;
1027 v->s.v_edge_pos = v->s.avctx->coded_height;
1028
1029 av_log(v->s.avctx, AV_LOG_DEBUG,
1030 "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
1031 "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"
1032 "TFCTRflag=%i, FINTERPflag=%i\n",
1033 v->level, v->frmrtq_postproc, v->bitrtq_postproc,
1034 v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,
1035 v->tfcntrflag, v->finterpflag
1036 );
1037
1038 v->psf = get_bits1(gb);
1039 if(v->psf) { //PsF, 6.1.13
1040 av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");
1041 return -1;
1042 }
1043 v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
1044 if(get_bits1(gb)) { //Display Info - decoding is not affected by it
1045 int w, h, ar = 0;
1046 av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
1047 // FIXME: The w/h parsed here are the *display* width/height, not the
1048 // coded width/height. Ideally, we should make the commented
1049 // assignments below, but that causes problems:
1050 // * The SW decoder in this file experiences errors, because it
1051 // assumes these assigned values are the coded size:
1052 // [vc1 @ 0x86f2130]concealing 150 DC, 150 AC, 150 MV errors
1053 // * VDPAU also assumes these are the coded size, since this is the
1054 // only size passed to vo_vdpau.c:config(). This causes errors
1055 // during the decode process.
1056 // However, simply removing these assignments is not the complete fix,
1057 // because without them, the stream is displayed at its coded size,
1058 // not this requested display size. Ideally, setting:
1059 // sample_aspect_ratio = (AVRational){w, h}
1060 // in the case when ar is not present/set would persuade other modules
1061 // to scale to this requested size. However, sample_aspect_ratio
1062 // appears to be completely ignored elsewhere.
1063 /*v->s.avctx->width = v->s.width =*/ w = get_bits(gb, 14) + 1;
1064 /*v->s.avctx->height = v->s.height =*/ h = get_bits(gb, 14) + 1;
1065 av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
1066 if(get_bits1(gb))
1067 ar = get_bits(gb, 4);
1068 if(ar && ar < 14){
1069 v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
1070 }else if(ar == 15){
1071 w = get_bits(gb, 8);
1072 h = get_bits(gb, 8);
1073 v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
1074 }
1075
1076 if(get_bits1(gb)){ //framerate stuff
1077 if(get_bits1(gb)) {
1078 v->s.avctx->time_base.num = 32;
1079 v->s.avctx->time_base.den = get_bits(gb, 16) + 1;
1080 } else {
1081 int nr, dr;
1082 nr = get_bits(gb, 8);
1083 dr = get_bits(gb, 4);
1084 if(nr && nr < 8 && dr && dr < 3){
1085 v->s.avctx->time_base.num = ff_vc1_fps_dr[dr - 1];
1086 v->s.avctx->time_base.den = ff_vc1_fps_nr[nr - 1] * 1000;
1087 }
1088 }
1089 }
1090
1091 if(get_bits1(gb)){
1092 v->color_prim = get_bits(gb, 8);
1093 v->transfer_char = get_bits(gb, 8);
1094 v->matrix_coef = get_bits(gb, 8);
1095 }
1096 }
1097
1098 v->hrd_param_flag = get_bits1(gb);
1099 if(v->hrd_param_flag) {
1100 int i;
1101 v->hrd_num_leaky_buckets = get_bits(gb, 5);
1102 skip_bits(gb, 4); //bitrate exponent
1103 skip_bits(gb, 4); //buffer size exponent
1104 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1105 skip_bits(gb, 16); //hrd_rate[n]
1106 skip_bits(gb, 16); //hrd_buffer[n]
1107 }
1108 }
1109 return 0;
1110 }
1111
1112 static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb)
1113 {
1114 VC1Context *v = avctx->priv_data;
1115 int i, blink, clentry;
1116
1117 av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32));
1118 blink = get_bits1(gb); // broken link
1119 clentry = get_bits1(gb); // closed entry
1120 v->panscanflag = get_bits1(gb);
1121 v->refdist_flag = get_bits1(gb);
1122 v->s.loop_filter = get_bits1(gb);
1123 v->fastuvmc = get_bits1(gb);
1124 v->extended_mv = get_bits1(gb);
1125 v->dquant = get_bits(gb, 2);
1126 v->vstransform = get_bits1(gb);
1127 v->overlap = get_bits1(gb);
1128 v->quantizer_mode = get_bits(gb, 2);
1129
1130 if(v->hrd_param_flag){
1131 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1132 skip_bits(gb, 8); //hrd_full[n]
1133 }
1134 }
1135
1136 if(get_bits1(gb)){
1137 avctx->coded_width = (get_bits(gb, 12)+1)<<1;
1138 avctx->coded_height = (get_bits(gb, 12)+1)<<1;
1139 }
1140 if(v->extended_mv)
1141 v->extended_dmv = get_bits1(gb);
1142 v->range_mapy_flag = get_bits1(gb);
1143 if(v->range_mapy_flag) {
1144 av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n");
1145 v->range_mapy = get_bits(gb, 3);
1146 }
1147 v->range_mapuv_flag = get_bits1(gb);
1148 if(v->range_mapuv_flag) {
1149 av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n");
1150 v->range_mapuv = get_bits(gb, 3);
1151 }
1152
1153 av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n"
1154 "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n"
1155 "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n"
1156 "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n",
1157 blink, clentry, v->panscanflag, v->refdist_flag, v->s.loop_filter,
1158 v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode);
1159
1160 return 0;
1161 }
1162
1163 static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)
1164 {
1165 int pqindex, lowquant, status;
1166
1167 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1168 skip_bits(gb, 2); //framecnt unused
1169 v->rangeredfrm = 0;
1170 if (v->rangered) v->rangeredfrm = get_bits1(gb);
1171 v->s.pict_type = get_bits1(gb);
1172 if (v->s.avctx->max_b_frames) {
1173 if (!v->s.pict_type) {
1174 if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE;
1175 else v->s.pict_type = FF_B_TYPE;
1176 } else v->s.pict_type = FF_P_TYPE;
1177 } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE;
1178
1179 v->bi_type = 0;
1180 if(v->s.pict_type == FF_B_TYPE) {
1181 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1182 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1183 if(v->bfraction == 0) {
1184 v->s.pict_type = FF_BI_TYPE;
1185 }
1186 }
1187 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1188 skip_bits(gb, 7); // skip buffer fullness
1189
1190 /* calculate RND */
1191 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1192 v->rnd = 1;
1193 if(v->s.pict_type == FF_P_TYPE)
1194 v->rnd ^= 1;
1195
1196 /* Quantizer stuff */
1197 pqindex = get_bits(gb, 5);
1198 if(!pqindex) return -1;
1199 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1200 v->pq = ff_vc1_pquant_table[0][pqindex];
1201 else
1202 v->pq = ff_vc1_pquant_table[1][pqindex];
1203
1204 v->pquantizer = 1;
1205 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1206 v->pquantizer = pqindex < 9;
1207 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1208 v->pquantizer = 0;
1209 v->pqindex = pqindex;
1210 if (pqindex < 9) v->halfpq = get_bits1(gb);
1211 else v->halfpq = 0;
1212 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1213 v->pquantizer = get_bits1(gb);
1214 v->dquantfrm = 0;
1215 if (v->extended_mv == 1) v->mvrange = get_unary(gb, 0, 3);
1216 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1217 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1218 v->range_x = 1 << (v->k_x - 1);
1219 v->range_y = 1 << (v->k_y - 1);
1220 if (v->profile == PROFILE_ADVANCED)
1221 {
1222 if (v->postprocflag) v->postproc = get_bits1(gb);
1223 }
1224 else
1225 if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2);
1226
1227 if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){
1228 v->x8_type = get_bits1(gb);
1229 }else v->x8_type = 0;
1230 //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",
1231 // (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);
1232
1233 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1234
1235 switch(v->s.pict_type) {
1236 case FF_P_TYPE:
1237 if (v->pq < 5) v->tt_index = 0;
1238 else if(v->pq < 13) v->tt_index = 1;
1239 else v->tt_index = 2;
1240
1241 lowquant = (v->pq > 12) ? 0 : 1;
1242 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1243 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1244 {
1245 int scale, shift, i;
1246 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1247 v->lumscale = get_bits(gb, 6);
1248 v->lumshift = get_bits(gb, 6);
1249 v->use_ic = 1;
1250 /* fill lookup tables for intensity compensation */
1251 if(!v->lumscale) {
1252 scale = -64;
1253 shift = (255 - v->lumshift * 2) << 6;
1254 if(v->lumshift > 31)
1255 shift += 128 << 6;
1256 } else {
1257 scale = v->lumscale + 32;
1258 if(v->lumshift > 31)
1259 shift = (v->lumshift - 64) << 6;
1260 else
1261 shift = v->lumshift << 6;
1262 }
1263 for(i = 0; i < 256; i++) {
1264 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1265 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1266 }
1267 }
1268 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1269 v->s.quarter_sample = 0;
1270 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1271 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1272 v->s.quarter_sample = 0;
1273 else
1274 v->s.quarter_sample = 1;
1275 } else
1276 v->s.quarter_sample = 1;
1277 v->s.mspel = !(v->mv_mode == MV_PMODE_1MV_HPEL_BILIN || (v->mv_mode == MV_PMODE_INTENSITY_COMP && v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN));
1278
1279 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1280 v->mv_mode2 == MV_PMODE_MIXED_MV)
1281 || v->mv_mode == MV_PMODE_MIXED_MV)
1282 {
1283 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1284 if (status < 0) return -1;
1285 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1286 "Imode: %i, Invert: %i\n", status>>1, status&1);
1287 } else {
1288 v->mv_type_is_raw = 0;
1289 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1290 }
1291 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1292 if (status < 0) return -1;
1293 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1294 "Imode: %i, Invert: %i\n", status>>1, status&1);
1295
1296 /* Hopefully this is correct for P frames */
1297 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1298 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1299
1300 if (v->dquant)
1301 {
1302 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1303 vop_dquant_decoding(v);
1304 }
1305
1306 v->ttfrm = 0; //FIXME Is that so ?
1307 if (v->vstransform)
1308 {
1309 v->ttmbf = get_bits1(gb);
1310 if (v->ttmbf)
1311 {
1312 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1313 }
1314 } else {
1315 v->ttmbf = 1;
1316 v->ttfrm = TT_8X8;
1317 }
1318 break;
1319 case FF_B_TYPE:
1320 if (v->pq < 5) v->tt_index = 0;
1321 else if(v->pq < 13) v->tt_index = 1;
1322 else v->tt_index = 2;
1323
1324 lowquant = (v->pq > 12) ? 0 : 1;
1325 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1326 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1327 v->s.mspel = v->s.quarter_sample;
1328
1329 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1330 if (status < 0) return -1;
1331 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1332 "Imode: %i, Invert: %i\n", status>>1, status&1);
1333 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1334 if (status < 0) return -1;
1335 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1336 "Imode: %i, Invert: %i\n", status>>1, status&1);
1337
1338 v->s.mv_table_index = get_bits(gb, 2);
1339 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1340
1341 if (v->dquant)
1342 {
1343 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1344 vop_dquant_decoding(v);
1345 }
1346
1347 v->ttfrm = 0;
1348 if (v->vstransform)
1349 {
1350 v->ttmbf = get_bits1(gb);
1351 if (v->ttmbf)
1352 {
1353 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1354 }
1355 } else {
1356 v->ttmbf = 1;
1357 v->ttfrm = TT_8X8;
1358 }
1359 break;
1360 }
1361
1362 if(!v->x8_type)
1363 {
1364 /* AC Syntax */
1365 v->c_ac_table_index = decode012(gb);
1366 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1367 {
1368 v->y_ac_table_index = decode012(gb);
1369 }
1370 /* DC Syntax */
1371 v->s.dc_table_index = get_bits1(gb);
1372 }
1373
1374 if(v->s.pict_type == FF_BI_TYPE) {
1375 v->s.pict_type = FF_B_TYPE;
1376 v->bi_type = 1;
1377 }
1378 return 0;
1379 }
1380
1381 static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)
1382 {
1383 int pqindex, lowquant;
1384 int status;
1385
1386 v->p_frame_skipped = 0;
1387
1388 if(v->interlace){
1389 v->fcm = decode012(gb);
1390 if(v->fcm) return -1; // interlaced frames/fields are not implemented
1391 }
1392 switch(get_unary(gb, 0, 4)) {
1393 case 0:
1394 v->s.pict_type = FF_P_TYPE;
1395 break;
1396 case 1:
1397 v->s.pict_type = FF_B_TYPE;
1398 break;
1399 case 2:
1400 v->s.pict_type = FF_I_TYPE;
1401 break;
1402 case 3:
1403 v->s.pict_type = FF_BI_TYPE;
1404 break;
1405 case 4:
1406 v->s.pict_type = FF_P_TYPE; // skipped pic
1407 v->p_frame_skipped = 1;
1408 return 0;
1409 }
1410 if(v->tfcntrflag)
1411 skip_bits(gb, 8);
1412 if(v->broadcast) {
1413 if(!v->interlace || v->psf) {
1414 v->rptfrm = get_bits(gb, 2);
1415 } else {
1416 v->tff = get_bits1(gb);
1417 v->rptfrm = get_bits1(gb);
1418 }
1419 }
1420 if(v->panscanflag) {
1421 //...
1422 }
1423 v->rnd = get_bits1(gb);
1424 if(v->interlace)
1425 v->uvsamp = get_bits1(gb);
1426 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1427 if(v->s.pict_type == FF_B_TYPE) {
1428 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1429 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1430 if(v->bfraction == 0) {
1431 v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */
1432 }
1433 }
1434 pqindex = get_bits(gb, 5);
1435 if(!pqindex) return -1;
1436 v->pqindex = pqindex;
1437 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1438 v->pq = ff_vc1_pquant_table[0][pqindex];
1439 else
1440 v->pq = ff_vc1_pquant_table[1][pqindex];
1441
1442 v->pquantizer = 1;
1443 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1444 v->pquantizer = pqindex < 9;
1445 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1446 v->pquantizer = 0;
1447 v->pqindex = pqindex;
1448 if (pqindex < 9) v->halfpq = get_bits1(gb);
1449 else v->halfpq = 0;
1450 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1451 v->pquantizer = get_bits1(gb);
1452 if(v->postprocflag)
1453 v->postproc = get_bits1(gb);
1454
1455 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1456
1457 if(v->postprocflag)
1458 v->postproc = get_bits(gb, 2);
1459
1460 switch(v->s.pict_type) {
1461 case FF_I_TYPE:
1462 case FF_BI_TYPE:
1463 status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v);
1464 if (status < 0) return -1;
1465 av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: "
1466 "Imode: %i, Invert: %i\n", status>>1, status&1);
1467 v->condover = CONDOVER_NONE;
1468 if(v->overlap && v->pq <= 8) {
1469 v->condover = decode012(gb);
1470 if(v->condover == CONDOVER_SELECT) {
1471 status = bitplane_decoding(v->over_flags_plane, &v->overflg_is_raw, v);
1472 if (status < 0) return -1;
1473 av_log(v->s.avctx, AV_LOG_DEBUG, "CONDOVER plane encoding: "
1474 "Imode: %i, Invert: %i\n", status>>1, status&1);
1475 }
1476 }
1477 break;
1478 case FF_P_TYPE:
1479 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1480 else v->mvrange = 0;
1481 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1482 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1483 v->range_x = 1 << (v->k_x - 1);
1484 v->range_y = 1 << (v->k_y - 1);
1485
1486 if (v->pq < 5) v->tt_index = 0;
1487 else if(v->pq < 13) v->tt_index = 1;
1488 else v->tt_index = 2;
1489
1490 lowquant = (v->pq > 12) ? 0 : 1;
1491 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1492 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1493 {
1494 int scale, shift, i;
1495 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1496 v->lumscale = get_bits(gb, 6);
1497 v->lumshift = get_bits(gb, 6);
1498 /* fill lookup tables for intensity compensation */
1499 if(!v->lumscale) {
1500 scale = -64;
1501 shift = (255 - v->lumshift * 2) << 6;
1502 if(v->lumshift > 31)
1503 shift += 128 << 6;
1504 } else {
1505 scale = v->lumscale + 32;
1506 if(v->lumshift > 31)
1507 shift = (v->lumshift - 64) << 6;
1508 else
1509 shift = v->lumshift << 6;
1510 }
1511 for(i = 0; i < 256; i++) {
1512 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1513 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1514 }
1515 v->use_ic = 1;
1516 }
1517 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1518 v->s.quarter_sample = 0;
1519 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1520 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1521 v->s.quarter_sample = 0;
1522 else
1523 v->s.quarter_sample = 1;
1524 } else
1525 v->s.quarter_sample = 1;
1526 v->s.mspel = !(v->mv_mode == MV_PMODE_1MV_HPEL_BILIN || (v->mv_mode == MV_PMODE_INTENSITY_COMP && v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN));
1527
1528 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1529 v->mv_mode2 == MV_PMODE_MIXED_MV)
1530 || v->mv_mode == MV_PMODE_MIXED_MV)
1531 {
1532 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1533 if (status < 0) return -1;
1534 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1535 "Imode: %i, Invert: %i\n", status>>1, status&1);
1536 } else {
1537 v->mv_type_is_raw = 0;
1538 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1539 }
1540 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1541 if (status < 0) return -1;
1542 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1543 "Imode: %i, Invert: %i\n", status>>1, status&1);
1544
1545 /* Hopefully this is correct for P frames */
1546 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1547 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1548 if (v->dquant)
1549 {
1550 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1551 vop_dquant_decoding(v);
1552 }
1553
1554 v->ttfrm = 0; //FIXME Is that so ?
1555 if (v->vstransform)
1556 {
1557 v->ttmbf = get_bits1(gb);
1558 if (v->ttmbf)
1559 {
1560 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1561 }
1562 } else {
1563 v->ttmbf = 1;
1564 v->ttfrm = TT_8X8;
1565 }
1566 break;
1567 case FF_B_TYPE:
1568 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1569 else v->mvrange = 0;
1570 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1571 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1572 v->range_x = 1 << (v->k_x - 1);
1573 v->range_y = 1 << (v->k_y - 1);
1574
1575 if (v->pq < 5) v->tt_index = 0;
1576 else if(v->pq < 13) v->tt_index = 1;
1577 else v->tt_index = 2;
1578
1579 lowquant = (v->pq > 12) ? 0 : 1;
1580 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1581 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1582 v->s.mspel = v->s.quarter_sample;
1583
1584 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1585 if (status < 0) return -1;
1586 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1587 "Imode: %i, Invert: %i\n", status>>1, status&1);
1588 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1589 if (status < 0) return -1;
1590 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1591 "Imode: %i, Invert: %i\n", status>>1, status&1);
1592
1593 v->s.mv_table_index = get_bits(gb, 2);
1594 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1595
1596 if (v->dquant)
1597 {
1598 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1599 vop_dquant_decoding(v);
1600 }
1601
1602 v->ttfrm = 0;
1603 if (v->vstransform)
1604 {
1605 v->ttmbf = get_bits1(gb);
1606 if (v->ttmbf)
1607 {
1608 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1609 }
1610 } else {
1611 v->ttmbf = 1;
1612 v->ttfrm = TT_8X8;
1613 }
1614 break;
1615 }
1616
1617 /* AC Syntax */
1618 v->c_ac_table_index = decode012(gb);
1619 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1620 {
1621 v->y_ac_table_index = decode012(gb);
1622 }
1623 /* DC Syntax */
1624 v->s.dc_table_index = get_bits1(gb);
1625 if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) {
1626 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1627 vop_dquant_decoding(v);
1628 }
1629
1630 v->bi_type = 0;
1631 if(v->s.pict_type == FF_BI_TYPE) {
1632 v->s.pict_type = FF_B_TYPE;
1633 v->bi_type = 1;
1634 }
1635 return 0;
1636 }
1637
1638 /***********************************************************************/
1639 /**
1640 * @defgroup block VC-1 Block-level functions
1641 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
1642 * @{
1643 */
1644
1645 /**
1646 * @def GET_MQUANT
1647 * @brief Get macroblock-level quantizer scale
1648 */
1649 #define GET_MQUANT() \
1650 if (v->dquantfrm) \
1651 { \
1652 int edges = 0; \
1653 if (v->dqprofile == DQPROFILE_ALL_MBS) \
1654 { \
1655 if (v->dqbilevel) \
1656 { \
1657 mquant = (get_bits1(gb)) ? v->altpq : v->pq; \
1658 } \
1659 else \
1660 { \
1661 mqdiff = get_bits(gb, 3); \
1662 if (mqdiff != 7) mquant = v->pq + mqdiff; \
1663 else mquant = get_bits(gb, 5); \
1664 } \
1665 } \
1666 if(v->dqprofile == DQPROFILE_SINGLE_EDGE) \
1667 edges = 1 << v->dqsbedge; \
1668 else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES) \
1669 edges = (3 << v->dqsbedge) % 15; \
1670 else if(v->dqprofile == DQPROFILE_FOUR_EDGES) \
1671 edges = 15; \
1672 if((edges&1) && !s->mb_x) \
1673 mquant = v->altpq; \
1674 if((edges&2) && s->first_slice_line) \
1675 mquant = v->altpq; \
1676 if((edges&4) && s->mb_x == (s->mb_width - 1)) \
1677 mquant = v->altpq; \
1678 if((edges&8) && s->mb_y == (s->mb_height - 1)) \
1679 mquant = v->altpq; \
1680 }
1681
1682 /**
1683 * @def GET_MVDATA(_dmv_x, _dmv_y)
1684 * @brief Get MV differentials
1685 * @see MVDATA decoding from 8.3.5.2, p(1)20
1686 * @param _dmv_x Horizontal differential for decoded MV
1687 * @param _dmv_y Vertical differential for decoded MV
1688 */
1689 #define GET_MVDATA(_dmv_x, _dmv_y) \
1690 index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table,\
1691 VC1_MV_DIFF_VLC_BITS, 2); \
1692 if (index > 36) \
1693 { \
1694 mb_has_coeffs = 1; \
1695 index -= 37; \
1696 } \
1697 else mb_has_coeffs = 0; \
1698 s->mb_intra = 0; \
1699 if (!index) { _dmv_x = _dmv_y = 0; } \
1700 else if (index == 35) \
1701 { \
1702 _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample); \
1703 _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample); \
1704 } \
1705 else if (index == 36) \
1706 { \
1707 _dmv_x = 0; \
1708 _dmv_y = 0; \
1709 s->mb_intra = 1; \
1710 } \
1711 else \
1712 { \
1713 index1 = index%6; \
1714 if (!s->quarter_sample && index1 == 5) val = 1; \
1715 else val = 0; \
1716 if(size_table[index1] - val > 0) \
1717 val = get_bits(gb, size_table[index1] - val); \
1718 else val = 0; \
1719 sign = 0 - (val&1); \
1720 _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1721 \
1722 index1 = index/6; \
1723 if (!s->quarter_sample && index1 == 5) val = 1; \
1724 else val = 0; \
1725 if(size_table[index1] - val > 0) \
1726 val = get_bits(gb, size_table[index1] - val); \
1727 else val = 0; \
1728 sign = 0 - (val&1); \
1729 _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1730 }
1731
1732 /** Predict and set motion vector
1733 */
1734 static inline void vc1_pred_mv(MpegEncContext *s, int n, int dmv_x, int dmv_y, int mv1, int r_x, int r_y, uint8_t* is_intra)
1735 {
1736 int xy, wrap, off = 0;
1737 int16_t *A, *B, *C;
1738 int px, py;
1739 int sum;
1740
1741 /* scale MV difference to be quad-pel */
1742 dmv_x <<= 1 - s->quarter_sample;
1743 dmv_y <<= 1 - s->quarter_sample;
1744
1745 wrap = s->b8_stride;
1746 xy = s->block_index[n];
1747
1748 if(s->mb_intra){
1749 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
1750 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
1751 s->current_picture.motion_val[1][xy][0] = 0;
1752 s->current_picture.motion_val[1][xy][1] = 0;
1753 if(mv1) { /* duplicate motion data for 1-MV block */
1754 s->current_picture.motion_val[0][xy + 1][0] = 0;
1755 s->current_picture.motion_val[0][xy + 1][1] = 0;
1756 s->current_picture.motion_val[0][xy + wrap][0] = 0;
1757 s->current_picture.motion_val[0][xy + wrap][1] = 0;
1758 s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
1759 s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
1760 s->current_picture.motion_val[1][xy + 1][0] = 0;
1761 s->current_picture.motion_val[1][xy + 1][1] = 0;
1762 s->current_picture.motion_val[1][xy + wrap][0] = 0;
1763 s->current_picture.motion_val[1][xy + wrap][1] = 0;
1764 s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
1765 s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
1766 }
1767 return;
1768 }
1769
1770 C = s->current_picture.motion_val[0][xy - 1];
1771 A = s->current_picture.motion_val[0][xy - wrap];
1772 if(mv1)
1773 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
1774 else {
1775 //in 4-MV mode different blocks have different B predictor position
1776 switch(n){
1777 case 0:
1778 off = (s->mb_x > 0) ? -1 : 1;
1779 break;
1780 case 1:
1781 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
1782 break;
1783 case 2:
1784 off = 1;
1785 break;
1786 case 3:
1787 off = -1;
1788 }
1789 }
1790 B = s->current_picture.motion_val[0][xy - wrap + off];
1791
1792 if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds
1793 if(s->mb_width == 1) {
1794 px = A[0];
1795 py = A[1];
1796 } else {
1797 px = mid_pred(A[0], B[0], C[0]);
1798 py = mid_pred(A[1], B[1], C[1]);
1799 }
1800 } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds
1801 px = C[0];
1802 py = C[1];
1803 } else {
1804 px = py = 0;
1805 }
1806 /* Pullback MV as specified in 8.3.5.3.4 */
1807 {
1808 int qx, qy, X, Y;
1809 qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);
1810 qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);
1811 X = (s->mb_width << 6) - 4;
1812 Y = (s->mb_height << 6) - 4;
1813 if(mv1) {
1814 if(qx + px < -60) px = -60 - qx;
1815 if(qy + py < -60) py = -60 - qy;
1816 } else {
1817 if(qx + px < -28) px = -28 - qx;
1818 if(qy + py < -28) py = -28 - qy;
1819 }
1820 if(qx + px > X) px = X - qx;
1821 if(qy + py > Y) py = Y - qy;
1822 }
1823 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
1824 if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {
1825 if(is_intra[xy - wrap])
1826 sum = FFABS(px) + FFABS(py);
1827 else
1828 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
1829 if(sum > 32) {
1830 if(get_bits1(&s->gb)) {
1831 px = A[0];
1832 py = A[1];
1833 } else {
1834 px = C[0];
1835 py = C[1];
1836 }
1837 } else {
1838 if(is_intra[xy - 1])
1839 sum = FFABS(px) + FFABS(py);
1840 else
1841 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
1842 if(sum > 32) {
1843 if(get_bits1(&s->gb)) {
1844 px = A[0];
1845 py = A[1];
1846 } else {
1847 px = C[0];
1848 py = C[1];
1849 }
1850 }
1851 }
1852 }
1853 /* store MV using signed modulus of MV range defined in 4.11 */
1854 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
1855 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;
1856 if(mv1) { /* duplicate motion data for 1-MV block */
1857 s->current_picture.motion_val[0][xy + 1][0] = s->current_picture.motion_val[0][xy][0];
1858 s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];
1859 s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];
1860 s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];
1861 s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];
1862 s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];
1863 }
1864 }
1865
1866 /** Motion compensation for direct or interpolated blocks in B-frames
1867 */
1868 static void vc1_interp_mc(VC1Context *v)
1869 {
1870 MpegEncContext *s = &v->s;
1871 DSPContext *dsp = &v->s.dsp;
1872 uint8_t *srcY, *srcU, *srcV;
1873 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
1874
1875 if(!v->s.next_picture.data[0])return;
1876
1877 mx = s->mv[1][0][0];
1878 my = s->mv[1][0][1];
1879 uvmx = (mx + ((mx & 3) == 3)) >> 1;
1880 uvmy = (my + ((my & 3) == 3)) >> 1;
1881 if(v->fastuvmc) {
1882 uvmx = uvmx + ((uvmx<0)?-(uvmx&1):(uvmx&1));
1883 uvmy = uvmy + ((uvmy<0)?-(uvmy&1):(uvmy&1));
1884 }
1885 srcY = s->next_picture.data[0];
1886 srcU = s->next_picture.data[1];
1887 srcV = s->next_picture.data[2];
1888
1889 src_x = s->mb_x * 16 + (mx >> 2);
1890 src_y = s->mb_y * 16 + (my >> 2);
1891 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
1892 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
1893
1894 if(v->profile != PROFILE_ADVANCED){
1895 src_x = av_clip( src_x, -16, s->mb_width * 16);
1896 src_y = av_clip( src_y, -16, s->mb_height * 16);
1897 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
1898 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
1899 }else{
1900 src_x = av_clip( src_x, -17, s->avctx->coded_width);
1901 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
1902 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
1903 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
1904 }
1905
1906 srcY += src_y * s->linesize + src_x;
1907 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
1908 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
1909
1910 /* for grayscale we should not try to read from unknown area */
1911 if(s->flags & CODEC_FLAG_GRAY) {
1912 srcU = s->edge_emu_buffer + 18 * s->linesize;
1913 srcV = s->edge_emu_buffer + 18 * s->linesize;
1914 }
1915
1916 if(v->rangeredfrm
1917 || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16
1918 || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){
1919 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
1920
1921 srcY -= s->mspel * (1 + s->linesize);
1922 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
1923 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
1924 srcY = s->edge_emu_buffer;
1925 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
1926 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1927 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
1928 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1929 srcU = uvbuf;
1930 srcV = uvbuf + 16;
1931 /* if we deal with range reduction we need to scale source blocks */
1932 if(v->rangeredfrm) {
1933 int i, j;
1934 uint8_t *src, *src2;
1935
1936 src = srcY;
1937 for(j = 0; j < 17 + s->mspel*2; j++) {
1938 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
1939 src += s->linesize;
1940 }
1941 src = srcU; src2 = srcV;
1942 for(j = 0; j < 9; j++) {
1943 for(i = 0; i < 9; i++) {
1944 src[i] = ((src[i] - 128) >> 1) + 128;
1945 src2[i] = ((src2[i] - 128) >> 1) + 128;
1946 }
1947 src += s->uvlinesize;
1948 src2 += s->uvlinesize;
1949 }
1950 }
1951 srcY += s->mspel * (1 + s->linesize);
1952 }
1953
1954 mx >>= 1;
1955 my >>= 1;
1956 dxy = ((my & 1) << 1) | (mx & 1);
1957
1958 dsp->avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
1959
1960 if(s->flags & CODEC_FLAG_GRAY) return;
1961 /* Chroma MC always uses qpel blilinear */
1962 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
1963 uvmx = (uvmx&3)<<1;
1964 uvmy = (uvmy&3)<<1;
1965 dsp->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
1966 dsp->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
1967 }
1968
1969 static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)
1970 {
1971 int n = bfrac;
1972
1973 #if B_FRACTION_DEN==256
1974 if(inv)
1975 n -= 256;
1976 if(!qs)
1977 return 2 * ((value * n + 255) >> 9);
1978 return (value * n + 128) >> 8;
1979 #else
1980 if(inv)
1981 n -= B_FRACTION_DEN;
1982 if(!qs)
1983 return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));
1984 return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;
1985 #endif
1986 }
1987
1988 /** Reconstruct motion vector for B-frame and do motion compensation
1989 */
1990 static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode)
1991 {
1992 if(v->use_ic) {
1993 v->mv_mode2 = v->mv_mode;
1994 v->mv_mode = MV_PMODE_INTENSITY_COMP;
1995 }
1996 if(direct) {
1997 vc1_mc_1mv(v, 0);
1998 vc1_interp_mc(v);
1999 if(v->use_ic) v->mv_mode = v->mv_mode2;
2000 return;
2001 }
2002 if(mode == BMV_TYPE_INTERPOLATED) {
2003 vc1_mc_1mv(v, 0);
2004 vc1_interp_mc(v);
2005 if(v->use_ic) v->mv_mode = v->mv_mode2;
2006 return;
2007 }
2008
2009 if(v->use_ic && (mode == BMV_TYPE_BACKWARD)) v->mv_mode = v->mv_mode2;
2010 vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));
2011 if(v->use_ic) v->mv_mode = v->mv_mode2;
2012 }
2013
2014 static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mvtype)
2015 {
2016 MpegEncContext *s = &v->s;
2017 int xy, wrap, off = 0;
2018 int16_t *A, *B, *C;
2019 int px, py;
2020 int sum;
2021 int r_x, r_y;
2022 const uint8_t *is_intra = v->mb_type[0];
2023
2024 r_x = v->range_x;
2025 r_y = v->range_y;
2026 /* scale MV difference to be quad-pel */
2027 dmv_x[0] <<= 1 - s->quarter_sample;
2028 dmv_y[0] <<= 1 - s->quarter_sample;
2029 dmv_x[1] <<= 1 - s->quarter_sample;
2030 dmv_y[1] <<= 1 - s->quarter_sample;
2031
2032 wrap = s->b8_stride;
2033 xy = s->block_index[0];
2034
2035 if(s->mb_intra) {
2036 s->current_picture.motion_val[0][xy][0] =
2037 s->current_picture.motion_val[0][xy][1] =
2038 s->current_picture.motion_val[1][xy][0] =
2039 s->current_picture.motion_val[1][xy][1] = 0;
2040 return;
2041 }
2042 s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);
2043 s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);
2044 s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);
2045 s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);
2046
2047 /* Pullback predicted motion vectors as specified in 8.4.5.4 */
2048 s->mv[0][0][0] = av_clip(s->mv[0][0][0], -60 - (s->mb_x << 6), (s->mb_width << 6) - 4 - (s->mb_x << 6));
2049 s->mv[0][0][1] = av_clip(s->mv[0][0][1], -60 - (s->mb_y << 6), (s->mb_height << 6) - 4 - (s->mb_y << 6));
2050 s->mv[1][0][0] = av_clip(s->mv[1][0][0], -60 - (s->mb_x << 6), (s->mb_width << 6) - 4 - (s->mb_x << 6));
2051 s->mv[1][0][1] = av_clip(s->mv[1][0][1], -60 - (s->mb_y << 6), (s->mb_height << 6) - 4 - (s->mb_y << 6));
2052 if(direct) {
2053 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2054 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2055 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2056 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2057 return;
2058 }
2059
2060 if((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2061 C = s->current_picture.motion_val[0][xy - 2];
2062 A = s->current_picture.motion_val[0][xy - wrap*2];
2063 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2064 B = s->current_picture.motion_val[0][xy - wrap*2 + off];
2065
2066 if(!s->mb_x) C[0] = C[1] = 0;
2067 if(!s->first_slice_line) { // predictor A is not out of bounds
2068 if(s->mb_width == 1) {
2069 px = A[0];
2070 py = A[1];
2071 } else {
2072 px = mid_pred(A[0], B[0], C[0]);
2073 py = mid_pred(A[1], B[1], C[1]);
2074 }
2075 } else if(s->mb_x) { // predictor C is not out of bounds
2076 px = C[0];
2077 py = C[1];
2078 } else {
2079 px = py = 0;
2080 }
2081 /* Pullback MV as specified in 8.3.5.3.4 */
2082 {
2083 int qx, qy, X, Y;
2084 if(v->profile < PROFILE_ADVANCED) {
2085 qx = (s->mb_x << 5);
2086 qy = (s->mb_y << 5);
2087 X = (s->mb_width << 5) - 4;
2088 Y = (s->mb_height << 5) - 4;
2089 if(qx + px < -28) px = -28 - qx;
2090 if(qy + py < -28) py = -28 - qy;
2091 if(qx + px > X) px = X - qx;
2092 if(qy + py > Y) py = Y - qy;
2093 } else {
2094 qx = (s->mb_x << 6);
2095 qy = (s->mb_y << 6);
2096 X = (s->mb_width << 6) - 4;
2097 Y = (s->mb_height << 6) - 4;
2098 if(qx + px < -60) px = -60 - qx;
2099 if(qy + py < -60) py = -60 - qy;
2100 if(qx + px > X) px = X - qx;
2101 if(qy + py > Y) py = Y - qy;
2102 }
2103 }
2104 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2105 if(0 && !s->first_slice_line && s->mb_x) {
2106 if(is_intra[xy - wrap])
2107 sum = FFABS(px) + FFABS(py);
2108 else
2109 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2110 if(sum > 32) {
2111 if(get_bits1(&s->gb)) {
2112 px = A[0];
2113 py = A[1];
2114 } else {
2115 px = C[0];
2116 py = C[1];
2117 }
2118 } else {
2119 if(is_intra[xy - 2])
2120 sum = FFABS(px) + FFABS(py);
2121 else
2122 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2123 if(sum > 32) {
2124 if(get_bits1(&s->gb)) {
2125 px = A[0];
2126 py = A[1];
2127 } else {
2128 px = C[0];
2129 py = C[1];
2130 }
2131 }
2132 }
2133 }
2134 /* store MV using signed modulus of MV range defined in 4.11 */
2135 s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;
2136 s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;
2137 }
2138 if((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2139 C = s->current_picture.motion_val[1][xy - 2];
2140 A = s->current_picture.motion_val[1][xy - wrap*2];
2141 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2142 B = s->current_picture.motion_val[1][xy - wrap*2 + off];
2143
2144 if(!s->mb_x) C[0] = C[1] = 0;
2145 if(!s->first_slice_line) { // predictor A is not out of bounds
2146 if(s->mb_width == 1) {
2147 px = A[0];
2148 py = A[1];
2149 } else {
2150 px = mid_pred(A[0], B[0], C[0]);
2151 py = mid_pred(A[1], B[1], C[1]);
2152 }
2153 } else if(s->mb_x) { // predictor C is not out of bounds
2154 px = C[0];
2155 py = C[1];
2156 } else {
2157 px = py = 0;
2158 }
2159 /* Pullback MV as specified in 8.3.5.3.4 */
2160 {
2161 int qx, qy, X, Y;
2162 if(v->profile < PROFILE_ADVANCED) {
2163 qx = (s->mb_x << 5);
2164 qy = (s->mb_y << 5);
2165 X = (s->mb_width << 5) - 4;
2166 Y = (s->mb_height << 5) - 4;
2167 if(qx + px < -28) px = -28 - qx;
2168 if(qy + py < -28) py = -28 - qy;
2169 if(qx + px > X) px = X - qx;
2170 if(qy + py > Y) py = Y - qy;
2171 } else {
2172 qx = (s->mb_x << 6);
2173 qy = (s->mb_y << 6);
2174 X = (s->mb_width << 6) - 4;
2175 Y = (s->mb_height << 6) - 4;
2176 if(qx + px < -60) px = -60 - qx;
2177 if(qy + py < -60) py = -60 - qy;
2178 if(qx + px > X) px = X - qx;
2179 if(qy + py > Y) py = Y - qy;
2180 }
2181 }
2182 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2183 if(0 && !s->first_slice_line && s->mb_x) {
2184 if(is_intra[xy - wrap])
2185 sum = FFABS(px) + FFABS(py);
2186 else
2187 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2188 if(sum > 32) {
2189 if(get_bits1(&s->gb)) {
2190 px = A[0];
2191 py = A[1];
2192 } else {
2193 px = C[0];
2194 py = C[1];
2195 }
2196 } else {
2197 if(is_intra[xy - 2])
2198 sum = FFABS(px) + FFABS(py);
2199 else
2200 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2201 if(sum > 32) {
2202 if(get_bits1(&s->gb)) {
2203 px = A[0];
2204 py = A[1];
2205 } else {
2206 px = C[0];
2207 py = C[1];
2208 }
2209 }
2210 }
2211 }
2212 /* store MV using signed modulus of MV range defined in 4.11 */
2213
2214 s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;
2215 s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;
2216 }
2217 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2218 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2219 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2220 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2221 }
2222
2223 /** Get predicted DC value for I-frames only
2224 * prediction dir: left=0, top=1
2225 * @param s MpegEncContext
2226 * @param[in] n block index in the current MB
2227 * @param dc_val_ptr Pointer to DC predictor
2228 * @param dir_ptr Prediction direction for use in AC prediction
2229 */
2230 static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2231 int16_t **dc_val_ptr, int *dir_ptr)
2232 {
2233 int a, b, c, wrap, pred, scale;
2234 int16_t *dc_val;
2235 static const uint16_t dcpred[32] = {
2236 -1, 1024, 512, 341, 256, 205, 171, 146, 128,
2237 114, 102, 93, 85, 79, 73, 68, 64,
2238 60, 57, 54, 51, 49, 47, 45, 43,
2239 41, 39, 38, 37, 35, 34, 33
2240 };
2241
2242 /* find prediction - wmv3_dc_scale always used here in fact */
2243 if (n < 4) scale = s->y_dc_scale;
2244 else scale = s->c_dc_scale;
2245
2246 wrap = s->block_wrap[n];
2247 dc_val= s->dc_val[0] + s->block_index[n];
2248
2249 /* B A
2250 * C X
2251 */
2252 c = dc_val[ - 1];
2253 b = dc_val[ - 1 - wrap];
2254 a = dc_val[ - wrap];
2255
2256 if (pq < 9 || !overlap)
2257 {
2258 /* Set outer values */
2259 if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];
2260 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];
2261 }
2262 else
2263 {
2264 /* Set outer values */
2265 if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;
2266 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;
2267 }
2268
2269 if (abs(a - b) <= abs(b - c)) {
2270 pred = c;
2271 *dir_ptr = 1;//left
2272 } else {
2273 pred = a;
2274 *dir_ptr = 0;//top
2275 }
2276
2277 /* update predictor */
2278 *dc_val_ptr = &dc_val[0];
2279 return pred;
2280 }
2281
2282
2283 /** Get predicted DC value
2284 * prediction dir: left=0, top=1
2285 * @param s MpegEncContext
2286 * @param[in] n block index in the current MB
2287 * @param dc_val_ptr Pointer to DC predictor
2288 * @param dir_ptr Prediction direction for use in AC prediction
2289 */
2290 static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2291 int a_avail, int c_avail,
2292 int16_t **dc_val_ptr, int *dir_ptr)
2293 {
2294 int a, b, c, wrap, pred, scale;
2295 int16_t *dc_val;
2296 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2297 int q1, q2 = 0;
2298
2299 /* find prediction - wmv3_dc_scale always used here in fact */
2300 if (n < 4) scale = s->y_dc_scale;
2301 else scale = s->c_dc_scale;
2302
2303 wrap = s->block_wrap[n];
2304 dc_val= s->dc_val[0] + s->block_index[n];
2305
2306 /* B A
2307 * C X
2308 */
2309 c = dc_val[ - 1];
2310 b = dc_val[ - 1 - wrap];
2311 a = dc_val[ - wrap];
2312 /* scale predictors if needed */
2313 q1 = s->current_picture.qscale_table[mb_pos];
2314 if(c_avail && (n!= 1 && n!=3)) {
2315 q2 = s->current_picture.qscale_table[mb_pos - 1];
2316 if(q2 && q2 != q1)
2317 c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2318 }
2319 if(a_avail && (n!= 2 && n!=3)) {
2320 q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2321 if(q2 && q2 != q1)
2322 a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2323 }
2324 if(a_avail && c_avail && (n!=3)) {
2325 int off = mb_pos;
2326 if(n != 1) off--;
2327 if(n != 2) off -= s->mb_stride;
2328 q2 = s->current_picture.qscale_table[off];
2329 if(q2 && q2 != q1)
2330 b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2331 }
2332
2333 if(a_avail && c_avail) {
2334 if(abs(a - b) <= abs(b - c)) {
2335 pred = c;
2336 *dir_ptr = 1;//left
2337 } else {
2338 pred = a;
2339 *dir_ptr = 0;//top
2340 }
2341 } else if(a_avail) {
2342 pred = a;
2343 *dir_ptr = 0;//top
2344 } else if(c_avail) {
2345 pred = c;
2346 *dir_ptr = 1;//left
2347 } else {
2348 pred = 0;
2349 *dir_ptr = 1;//left
2350 }
2351
2352 /* update predictor */
2353 *dc_val_ptr = &dc_val[0];
2354 return pred;
2355 }
2356
2357
2358 /**
2359 * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles
2360 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
2361 * @{
2362 */
2363
2364 static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)
2365 {
2366 int xy, wrap, pred, a, b, c;
2367
2368 xy = s->block_index[n];
2369 wrap = s->b8_stride;
2370
2371 /* B C
2372 * A X
2373 */
2374 a = s->coded_block[xy - 1 ];
2375 b = s->coded_block[xy - 1 - wrap];
2376 c = s->coded_block[xy - wrap];
2377
2378 if (b == c) {
2379 pred = a;
2380 } else {
2381 pred = c;
2382 }
2383
2384 /* store value */
2385 *coded_block_ptr = &s->coded_block[xy];
2386
2387 return pred;
2388 }
2389
2390 /**
2391 * Decode one AC coefficient
2392 * @param v The VC1 context
2393 * @param last Last coefficient
2394 * @param skip How much zero coefficients to skip
2395 * @param value Decoded AC coefficient value
2396 * @see 8.1.3.4
2397 */
2398 static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)
2399 {
2400 GetBitContext *gb = &v->s.gb;
2401 int index, escape, run = 0, level = 0, lst = 0;
2402
2403 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2404 if (index != vc1_ac_sizes[codingset] - 1) {
2405 run = vc1_index_decode_table[codingset][index][0];
2406 level = vc1_index_decode_table[codingset][index][1];
2407 lst = index >= vc1_last_decode_table[codingset];
2408 if(get_bits1(gb))
2409 level = -level;
2410 } else {
2411 escape = decode210(gb);
2412 if (escape != 2) {
2413 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2414 run = vc1_index_decode_table[codingset][index][0];
2415 level = vc1_index_decode_table[codingset][index][1];
2416 lst = index >= vc1_last_decode_table[codingset];
2417 if(escape == 0) {
2418 if(lst)
2419 level += vc1_last_delta_level_table[codingset][run];
2420 else
2421 level += vc1_delta_level_table[codingset][run];
2422 } else {
2423 if(lst)
2424 run += vc1_last_delta_run_table[codingset][level] + 1;
2425 else
2426 run += vc1_delta_run_table[codingset][level] + 1;
2427 }
2428 if(get_bits1(gb))
2429 level = -level;
2430 } else {
2431 int sign;
2432 lst = get_bits1(gb);
2433 if(v->s.esc3_level_length == 0) {
2434 if(v->pq < 8 || v->dquantfrm) { // table 59
2435 v->s.esc3_level_length = get_bits(gb, 3);
2436 if(!v->s.esc3_level_length)
2437 v->s.esc3_level_length = get_bits(gb, 2) + 8;
2438 } else { //table 60
2439 v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;
2440 }
2441 v->s.esc3_run_length = 3 + get_bits(gb, 2);
2442 }
2443 run = get_bits(gb, v->s.esc3_run_length);
2444 sign = get_bits1(gb);
2445 level = get_bits(gb, v->s.esc3_level_length);
2446 if(sign)
2447 level = -level;
2448 }
2449 }
2450
2451 *last = lst;
2452 *skip = run;
2453 *value = level;
2454 }
2455
2456 /** Decode intra block in intra frames - should be faster than decode_intra_block
2457 * @param v VC1Context
2458 * @param block block to decode
2459 * @param coded are AC coeffs present or not
2460 * @param codingset set of VLC to decode data
2461 */
2462 static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)
2463 {
2464 GetBitContext *gb = &v->s.gb;
2465 MpegEncContext *s = &v->s;
2466 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2467 int run_diff, i;
2468 int16_t *dc_val;
2469 int16_t *ac_val, *ac_val2;
2470 int dcdiff;
2471
2472 /* Get DC differential */
2473 if (n < 4) {
2474 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2475 } else {
2476 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2477 }
2478 if (dcdiff < 0){
2479 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2480 return -1;
2481 }
2482 if (dcdiff)
2483 {
2484 if (dcdiff == 119 /* ESC index value */)
2485 {
2486 /* TODO: Optimize */
2487 if (v->pq == 1) dcdiff = get_bits(gb, 10);
2488 else if (v->pq == 2) dcdiff = get_bits(gb, 9);
2489 else dcdiff = get_bits(gb, 8);
2490 }
2491 else
2492 {
2493 if (v->pq == 1)
2494 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2495 else if (v->pq == 2)
2496 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2497 }
2498 if (get_bits1(gb))
2499 dcdiff = -dcdiff;
2500 }
2501
2502 /* Prediction */
2503 dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);
2504 *dc_val = dcdiff;
2505
2506 /* Store the quantized DC coeff, used for prediction */
2507 if (n < 4) {
2508 block[0] = dcdiff * s->y_dc_scale;
2509 } else {
2510 block[0] = dcdiff * s->c_dc_scale;
2511 }
2512 /* Skip ? */
2513 run_diff = 0;
2514 i = 0;
2515 if (!coded) {
2516 goto not_coded;
2517 }
2518
2519 //AC Decoding
2520 i = 1;
2521
2522 {
2523 int last = 0, skip, value;
2524 const int8_t *zz_table;
2525 int scale;
2526 int k;
2527
2528 scale = v->pq * 2 + v->halfpq;
2529
2530 if(v->s.ac_pred) {
2531 if(!dc_pred_dir)
2532 zz_table = wmv1_scantable[2];
2533 else
2534 zz_table = wmv1_scantable[3];
2535 } else
2536 zz_table = wmv1_scantable[1];
2537
2538 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2539 ac_val2 = ac_val;
2540 if(dc_pred_dir) //left
2541 ac_val -= 16;
2542 else //top
2543 ac_val -= 16 * s->block_wrap[n];
2544
2545 while (!last) {
2546 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2547 i += skip;
2548 if(i > 63)
2549 break;
2550 block[zz_table[i++]] = value;
2551 }
2552
2553 /* apply AC prediction if needed */
2554 if(s->ac_pred) {
2555 if(dc_pred_dir) { //left
2556 for(k = 1; k < 8; k++)
2557 block[k << 3] += ac_val[k];
2558 } else { //top
2559 for(k = 1; k < 8; k++)
2560 block[k] += ac_val[k + 8];
2561 }
2562 }
2563 /* save AC coeffs for further prediction */
2564 for(k = 1; k < 8; k++) {
2565 ac_val2[k] = block[k << 3];
2566 ac_val2[k + 8] = block[k];
2567 }
2568
2569 /* scale AC coeffs */
2570 for(k = 1; k < 64; k++)
2571 if(block[k]) {
2572 block[k] *= scale;
2573 if(!v->pquantizer)
2574 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2575 }
2576
2577 if(s->ac_pred) i = 63;
2578 }
2579
2580 not_coded:
2581 if(!coded) {
2582 int k, scale;
2583 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2584 ac_val2 = ac_val;
2585
2586 scale = v->pq * 2 + v->halfpq;
2587 memset(ac_val2, 0, 16 * 2);
2588 if(dc_pred_dir) {//left
2589 ac_val -= 16;
2590 if(s->ac_pred)
2591 memcpy(ac_val2, ac_val, 8 * 2);
2592 } else {//top
2593 ac_val -= 16 * s->block_wrap[n];
2594 if(s->ac_pred)
2595 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2596 }
2597
2598 /* apply AC prediction if needed */
2599 if(s->ac_pred) {
2600 if(dc_pred_dir) { //left
2601 for(k = 1; k < 8; k++) {
2602 block[k << 3] = ac_val[k] * scale;
2603 if(!v->pquantizer && block[k << 3])
2604 block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;
2605 }
2606 } else { //top
2607 for(k = 1; k < 8; k++) {
2608 block[k] = ac_val[k + 8] * scale;
2609 if(!v->pquantizer && block[k])
2610 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2611 }
2612 }
2613 i = 63;
2614 }
2615 }
2616 s->block_last_index[n] = i;
2617
2618 return 0;
2619 }
2620
2621 /** Decode intra block in intra frames - should be faster than decode_intra_block
2622 * @param v VC1Context
2623 * @param block block to decode
2624 * @param coded are AC coeffs present or not
2625 * @param codingset set of VLC to decode data
2626 */
2627 static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant)
2628 {
2629 GetBitContext *gb = &v->s.gb;
2630 MpegEncContext *s = &v->s;
2631 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2632 int run_diff, i;
2633 int16_t *dc_val;
2634 int16_t *ac_val, *ac_val2;
2635 int dcdiff;
2636 int a_avail = v->a_avail, c_avail = v->c_avail;
2637 int use_pred = s->ac_pred;
2638 int scale;
2639 int q1, q2 = 0;
2640 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2641
2642 /* Get DC differential */
2643 if (n < 4) {
2644 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2645 } else {
2646 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2647 }
2648 if (dcdiff < 0){
2649 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2650 return -1;
2651 }
2652 if (dcdiff)
2653 {
2654 if (dcdiff == 119 /* ESC index value */)
2655 {
2656 /* TODO: Optimize */
2657 if (mquant == 1) dcdiff = get_bits(gb, 10);
2658 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2659 else dcdiff = get_bits(gb, 8);
2660 }
2661 else
2662 {
2663 if (mquant == 1)
2664 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2665 else if (mquant == 2)
2666 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2667 }
2668 if (get_bits1(gb))
2669 dcdiff = -dcdiff;
2670 }
2671
2672 /* Prediction */
2673 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);
2674 *dc_val = dcdiff;
2675
2676 /* Store the quantized DC coeff, used for prediction */
2677 if (n < 4) {
2678 block[0] = dcdiff * s->y_dc_scale;
2679 } else {
2680 block[0] = dcdiff * s->c_dc_scale;
2681 }
2682 /* Skip ? */
2683 run_diff = 0;
2684 i = 0;
2685
2686 //AC Decoding
2687 i = 1;
2688
2689 /* check if AC is needed at all */
2690 if(!a_avail && !c_avail) use_pred = 0;
2691 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2692 ac_val2 = ac_val;
2693
2694 scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);
2695
2696 if(dc_pred_dir) //left
2697 ac_val -= 16;
2698 else //top
2699 ac_val -= 16 * s->block_wrap[n];
2700
2701 q1 = s->current_picture.qscale_table[mb_pos];
2702 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2703 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2704 if(dc_pred_dir && n==1) q2 = q1;
2705 if(!dc_pred_dir && n==2) q2 = q1;
2706 if(n==3) q2 = q1;
2707
2708 if(coded) {
2709 int last = 0, skip, value;
2710 const int8_t *zz_table;
2711 int k;
2712
2713 if(v->s.ac_pred) {
2714 if(!dc_pred_dir)
2715 zz_table = wmv1_scantable[2];
2716 else
2717 zz_table = wmv1_scantable[3];
2718 } else
2719 zz_table = wmv1_scantable[1];
2720
2721 while (!last) {
2722 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2723 i += skip;
2724 if(i > 63)
2725 break;
2726 block[zz_table[i++]] = value;
2727 }
2728
2729 /* apply AC prediction if needed */
2730 if(use_pred) {
2731 /* scale predictors if needed*/
2732 if(q2 && q1!=q2) {
2733 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2734 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2735
2736 if(dc_pred_dir) { //left
2737 for(k = 1; k < 8; k++)
2738 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2739 } else { //top
2740 for(k = 1; k < 8; k++)
2741 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2742 }
2743 } else {
2744 if(dc_pred_dir) { //left
2745 for(k = 1; k < 8; k++)
2746 block[k << 3] += ac_val[k];
2747 } else { //top
2748 for(k = 1; k < 8; k++)
2749 block[k] += ac_val[k + 8];
2750 }
2751 }
2752 }
2753 /* save AC coeffs for further prediction */
2754 for(k = 1; k < 8; k++) {
2755 ac_val2[k] = block[k << 3];
2756 ac_val2[k + 8] = block[k];
2757 }
2758
2759 /* scale AC coeffs */
2760 for(k = 1; k < 64; k++)
2761 if(block[k]) {
2762 block[k] *= scale;
2763 if(!v->pquantizer)
2764 block[k] += (block[k] < 0) ? -mquant : mquant;
2765 }
2766
2767 if(use_pred) i = 63;
2768 } else { // no AC coeffs
2769 int k;
2770
2771 memset(ac_val2, 0, 16 * 2);
2772 if(dc_pred_dir) {//left
2773 if(use_pred) {
2774 memcpy(ac_val2, ac_val, 8 * 2);
2775 if(q2 && q1!=q2) {
2776 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2777 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2778 for(k = 1; k < 8; k++)
2779 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2780 }
2781 }
2782 } else {//top
2783 if(use_pred) {
2784 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2785 if(q2 && q1!=q2) {
2786 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2787 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2788 for(k = 1; k < 8; k++)
2789 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2790 }
2791 }
2792 }
2793
2794 /* apply AC prediction if needed */
2795 if(use_pred) {
2796 if(dc_pred_dir) { //left
2797 for(k = 1; k < 8; k++) {
2798 block[k << 3] = ac_val2[k] * scale;
2799 if(!v->pquantizer && block[k << 3])
2800 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2801 }
2802 } else { //top
2803 for(k = 1; k < 8; k++) {
2804 block[k] = ac_val2[k + 8] * scale;
2805 if(!v->pquantizer && block[k])
2806 block[k] += (block[k] < 0) ? -mquant : mquant;
2807 }
2808 }
2809 i = 63;
2810 }
2811 }
2812 s->block_last_index[n] = i;
2813
2814 return 0;
2815 }
2816
2817 /** Decode intra block in inter frames - more generic version than vc1_decode_i_block
2818 * @param v VC1Context
2819 * @param block block to decode
2820 * @param coded are AC coeffs present or not
2821 * @param mquant block quantizer
2822 * @param codingset set of VLC to decode data
2823 */
2824 static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)
2825 {
2826 GetBitContext *gb = &v->s.gb;
2827 MpegEncContext *s = &v->s;
2828 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2829 int run_diff, i;
2830 int16_t *dc_val;
2831 int16_t *ac_val, *ac_val2;
2832 int dcdiff;
2833 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2834 int a_avail = v->a_avail, c_avail = v->c_avail;
2835 int use_pred = s->ac_pred;
2836 int scale;
2837 int q1, q2 = 0;
2838
2839 /* XXX: Guard against dumb values of mquant */
2840 mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );
2841
2842 /* Set DC scale - y and c use the same */
2843 s->y_dc_scale = s->y_dc_scale_table[mquant];
2844 s->c_dc_scale = s->c_dc_scale_table[mquant];
2845
2846 /* Get DC differential */
2847 if (n < 4) {
2848 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2849 } else {
2850 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2851 }
2852 if (dcdiff < 0){
2853 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2854 return -1;
2855 }
2856 if (dcdiff)
2857 {
2858 if (dcdiff == 119 /* ESC index value */)
2859 {
2860 /* TODO: Optimize */
2861 if (mquant == 1) dcdiff = get_bits(gb, 10);
2862 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2863 else dcdiff = get_bits(gb, 8);
2864 }
2865 else
2866 {
2867 if (mquant == 1)
2868 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2869 else if (mquant == 2)
2870 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2871 }
2872 if (get_bits1(gb))
2873 dcdiff = -dcdiff;
2874 }
2875
2876 /* Prediction */
2877 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);
2878 *dc_val = dcdiff;
2879
2880 /* Store the quantized DC coeff, used for prediction */
2881
2882 if (n < 4) {
2883 block[0] = dcdiff * s->y_dc_scale;
2884 } else {
2885 block[0] = dcdiff * s->c_dc_scale;
2886 }
2887 /* Skip ? */
2888 run_diff = 0;
2889 i = 0;
2890
2891 //AC Decoding
2892 i = 1;
2893
2894 /* check if AC is needed at all and adjust direction if needed */
2895 if(!a_avail) dc_pred_dir = 1;
2896 if(!c_avail) dc_pred_dir = 0;
2897 if(!a_avail && !c_avail) use_pred = 0;
2898 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2899 ac_val2 = ac_val;
2900
2901 scale = mquant * 2 + v->halfpq;
2902
2903 if(dc_pred_dir) //left
2904 ac_val -= 16;
2905 else //top
2906 ac_val -= 16 * s->block_wrap[n];
2907
2908 q1 = s->current_picture.qscale_table[mb_pos];
2909 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2910 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2911 if(dc_pred_dir && n==1) q2 = q1;
2912 if(!dc_pred_dir && n==2) q2 = q1;
2913 if(n==3) q2 = q1;
2914
2915 if(coded) {
2916 int last = 0, skip, value;
2917 const int8_t *zz_table;
2918 int k;
2919
2920 zz_table = wmv1_scantable[0];
2921
2922 while (!last) {
2923 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2924 i += skip;
2925 if(i > 63)
2926 break;
2927 block[zz_table[i++]] = value;
2928 }
2929
2930 /* apply AC prediction if needed */
2931 if(use_pred) {
2932 /* scale predictors if needed*/
2933 if(q2 && q1!=q2) {
2934 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2935 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2936
2937 if(dc_pred_dir) { //left
2938 for(k = 1; k < 8; k++)
2939 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2940 } else { //top
2941 for(k = 1; k < 8; k++)
2942 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2943 }
2944 } else {
2945 if(dc_pred_dir) { //left
2946 for(k = 1; k < 8; k++)
2947 block[k << 3] += ac_val[k];
2948 } else { //top
2949 for(k = 1; k < 8; k++)
2950 block[k] += ac_val[k + 8];
2951 }
2952 }
2953 }
2954 /* save AC coeffs for further prediction */
2955 for(k = 1; k < 8; k++) {
2956 ac_val2[k] = block[k << 3];
2957 ac_val2[k + 8] = block[k];
2958 }
2959
2960 /* scale AC coeffs */
2961 for(k = 1; k < 64; k++)
2962 if(block[k]) {
2963 block[k] *= scale;
2964 if(!v->pquantizer)
2965 block[k] += (block[k] < 0) ? -mquant : mquant;
2966 }
2967
2968 if(use_pred) i = 63;
2969 } else { // no AC coeffs
2970 int k;
2971
2972 memset(ac_val2, 0, 16 * 2);
2973 if(dc_pred_dir) {//left
2974 if(use_pred) {
2975 memcpy(ac_val2, ac_val, 8 * 2);
2976 if(q2 && q1!=q2) {
2977 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2978 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2979 for(k = 1; k < 8; k++)
2980 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2981 }
2982 }
2983 } else {//top
2984 if(use_pred) {
2985 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2986 if(q2 && q1!=q2) {
2987 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2988 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2989 for(k = 1; k < 8; k++)
2990 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2991 }
2992 }
2993 }
2994
2995 /* apply AC prediction if needed */
2996 if(use_pred) {
2997 if(dc_pred_dir) { //left
2998 for(k = 1; k < 8; k++) {
2999 block[k << 3] = ac_val2[k] * scale;
3000 if(!v->pquantizer && block[k << 3])
3001 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
3002 }
3003 } else { //top
3004 for(k = 1; k < 8; k++) {
3005 block[k] = ac_val2[k + 8] * scale;
3006 if(!v->pquantizer && block[k])
3007 block[k] += (block[k] < 0) ? -mquant : mquant;
3008 }
3009 }
3010 i = 63;
3011 }
3012 }
3013 s->block_last_index[n] = i;
3014
3015 return 0;
3016 }
3017
3018 /** Decode P block
3019 */
3020 static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,
3021 uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left)
3022 {
3023 MpegEncContext *s = &v->s;
3024 GetBitContext *gb = &s->gb;
3025 int i, j;
3026 int subblkpat = 0;
3027 int scale, off, idx, last, skip, value;
3028 int ttblk = ttmb & 7;
3029 int pat = 0;
3030
3031 if(ttmb == -1) {
3032 ttblk = ff_vc1_ttblk_to_tt[v->tt_index][get_vlc2(gb, ff_vc1_ttblk_vlc[v->tt_index].table, VC1_TTBLK_VLC_BITS, 1)];
3033 }
3034 if(ttblk == TT_4X4) {
3035 subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);
3036 }
3037 if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {
3038 subblkpat = decode012(gb);
3039 if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits
3040 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;
3041 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;
3042 }
3043 scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);
3044
3045 // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT
3046 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {
3047 subblkpat = 2 - (ttblk == TT_8X4_TOP);
3048 ttblk = TT_8X4;
3049 }
3050 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
3051 subblkpat = 2 - (ttblk == TT_4X8_LEFT);
3052 ttblk = TT_4X8;
3053 }
3054 switch(ttblk) {
3055 case TT_8X8:
3056 pat = 0xF;
3057 i = 0;
3058 last = 0;
3059 while (!last) {
3060 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3061 i += skip;
3062 if(i > 63)
3063 break;
3064 idx = wmv1_scantable[0][i++];
3065 block[idx] = value * scale;
3066 if(!v->pquantizer)
3067 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3068 }
3069 if(!skip_block){
3070 s->dsp.vc1_inv_trans_8x8(block);
3071 s->dsp.add_pixels_clamped(block, dst, linesize);
3072 if(apply_filter && cbp_top & 0xC)
3073 vc1_loop_filter(dst, 1, linesize, 8, mquant);
3074 if(apply_filter && cbp_left & 0xA)
3075 vc1_loop_filter(dst, linesize, 1, 8, mquant);
3076 }
3077 break;
3078 case TT_4X4:
3079 pat = ~subblkpat & 0xF;
3080 for(j = 0; j < 4; j++) {
3081 last = subblkpat & (1 << (3 - j));
3082 i = 0;
3083 off = (j & 1) * 4 + (j & 2) * 16;
3084 while (!last) {
3085 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3086 i += skip;
3087 if(i > 15)
3088 break;
3089 idx = ff_vc1_simple_progressive_4x4_zz[i++];
3090 block[idx + off] = value * scale;
3091 if(!v->pquantizer)
3092 block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;
3093 }
3094 if(!(subblkpat & (1 << (3 - j))) && !skip_block){
3095 s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
3096 if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))
3097 vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, 1, linesize, 4, mquant);
3098 if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))
3099 vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, linesize, 1, 4, mquant);
3100 }
3101 }
3102 break;
3103 case TT_8X4:
3104 pat = ~((subblkpat & 2)*6 + (subblkpat & 1)*3) & 0xF;
3105 for(j = 0; j < 2; j++) {
3106 last = subblkpat & (1 << (1 - j));
3107 i = 0;
3108 off = j * 32;
3109 while (!last) {
3110 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3111 i += skip;
3112 if(i > 31)
3113 break;
3114 idx = v->zz_8x4[i++]+off;
3115 block[idx] = value * scale;
3116 if(!v->pquantizer)
3117 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3118 }
3119 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3120 s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
3121 if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))
3122 vc1_loop_filter(dst + j*4*linesize, 1, linesize, 8, mquant);
3123 if(apply_filter && cbp_left & (2 << j))
3124 vc1_loop_filter(dst + j*4*linesize, linesize, 1, 4, mquant);
3125 }
3126 }
3127 break;
3128 case TT_4X8:
3129 pat = ~(subblkpat*5) & 0xF;
3130 for(j = 0; j < 2; j++) {
3131 last = subblkpat & (1 << (1 - j));
3132 i = 0;
3133 off = j * 4;
3134 while (!last) {
3135 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3136 i += skip;
3137 if(i > 31)
3138 break;
3139 idx = v->zz_4x8[i++]+off;
3140 block[idx] = value * scale;
3141 if(!v->pquantizer)
3142 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3143 }
3144 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3145 s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
3146 if(apply_filter && cbp_top & (2 << j))
3147 vc1_loop_filter(dst + j*4, 1, linesize, 4, mquant);
3148 if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))
3149 vc1_loop_filter(dst + j*4, linesize, 1, 8, mquant);
3150 }
3151 }
3152 break;
3153 }
3154 return pat;
3155 }
3156
3157
3158 /** Decode one P-frame MB (in Simple/Main profile)
3159 */
3160 static int vc1_decode_p_mb(VC1Context *v)
3161 {
3162 MpegEncContext *s = &v->s;
3163 GetBitContext *gb = &s->gb;
3164 int i, j;
3165 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3166 int cbp; /* cbp decoding stuff */
3167 int mqdiff, mquant; /* MB quantization */
3168 int ttmb = v->ttfrm; /* MB Transform type */
3169
3170 static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3171 offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3172 int mb_has_coeffs = 1; /* last_flag */
3173 int dmv_x, dmv_y; /* Differential MV components */
3174 int index, index1; /* LUT indexes */
3175 int val, sign; /* temp values */
3176 int first_block = 1;
3177 int dst_idx, off;
3178 int skipped, fourmv;
3179 int block_cbp = 0, pat;
3180 int apply_loop_filter;
3181
3182 mquant = v->pq; /* Loosy initialization */
3183
3184 if (v->mv_type_is_raw)
3185 fourmv = get_bits1(gb);
3186 else
3187 fourmv = v->mv_type_mb_plane[mb_pos];
3188 if (v->skip_is_raw)
3189 skipped = get_bits1(gb);
3190 else
3191 skipped = v->s.mbskip_table[mb_pos];
3192
3193 s->dsp.clear_blocks(s->block[0]);
3194
3195 apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);
3196 if (!fourmv) /* 1MV mode */
3197 {
3198 if (!skipped)
3199 {
3200 GET_MVDATA(dmv_x, dmv_y);
3201
3202 if (s->mb_intra) {
3203 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3204 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3205 }
3206 s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
3207 vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
3208
3209 /* FIXME Set DC val for inter block ? */
3210 if (s->mb_intra && !mb_has_coeffs)
3211 {
3212 GET_MQUANT();
3213 s->ac_pred = get_bits1(gb);
3214 cbp = 0;
3215 }
3216 else if (mb_has_coeffs)
3217 {
3218 if (s->mb_intra) s->ac_pred = get_bits1(gb);
3219 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3220 GET_MQUANT();
3221 }
3222 else
3223 {
3224 mquant = v->pq;
3225 cbp = 0;
3226 }
3227 s->current_picture.qscale_table[mb_pos] = mquant;
3228
3229 if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3230 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,
3231 VC1_TTMB_VLC_BITS, 2);
3232 if(!s->mb_intra) vc1_mc_1mv(v, 0);
3233 dst_idx = 0;
3234 for (i=0; i<6; i++)
3235 {
3236 s->dc_val[0][s->block_index[i]] = 0;
3237 dst_idx += i >> 2;
3238 val = ((cbp >> (5 - i)) & 1);
3239 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3240 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3241 if(s->mb_intra) {
3242 /* check if prediction blocks A and C are available */
3243 v->a_avail = v->c_avail = 0;
3244 if(i == 2 || i == 3 || !s->first_slice_line)
3245 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3246 if(i == 1 || i == 3 || s->mb_x)
3247 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3248
3249 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3250 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3251 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3252 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3253 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3254 if(v->pq >= 9 && v->overlap) {
3255 if(v->c_avail)
3256 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3257 if(v->a_avail)
3258 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3259 }
3260 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3261 int left_cbp, top_cbp;
3262 if(i & 4){
3263 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3264 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3265 }else{
3266 left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3267 top_cbp = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3268 }
3269 if(left_cbp & 0xC)
3270 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3271 if(top_cbp & 0xA)
3272 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3273 }
3274 block_cbp |= 0xF << (i << 2);
3275 } else if(val) {
3276 int left_cbp = 0, top_cbp = 0, filter = 0;
3277 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3278 filter = 1;
3279 if(i & 4){
3280 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3281 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3282 }else{
3283 left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3284 top_cbp = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3285 }
3286 }
3287 pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);
3288 block_cbp |= pat << (i << 2);
3289 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3290 first_block = 0;
3291 }
3292 }
3293 }
3294 else //Skipped
3295 {
3296 s->mb_intra = 0;
3297 for(i = 0; i < 6; i++) {
3298 v->mb_type[0][s->block_index[i]] = 0;
3299 s->dc_val[0][s->block_index[i]] = 0;
3300 }
3301 s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;
3302 s->current_picture.qscale_table[mb_pos] = 0;
3303 vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
3304 vc1_mc_1mv(v, 0);
3305 return 0;
3306 }
3307 } //1MV mode
3308 else //4MV mode
3309 {
3310 if (!skipped /* unskipped MB */)
3311 {
3312 int intra_count = 0, coded_inter = 0;
3313 int is_intra[6], is_coded[6];
3314 /* Get CBPCY */
3315 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3316 for (i=0; i<6; i++)
3317 {
3318 val = ((cbp >> (5 - i)) & 1);
3319 s->dc_val[0][s->block_index[i]] = 0;
3320 s->mb_intra = 0;
3321 if(i < 4) {
3322 dmv_x = dmv_y = 0;
3323 s->mb_intra = 0;
3324 mb_has_coeffs = 0;
3325 if(val) {
3326 GET_MVDATA(dmv_x, dmv_y);
3327 }
3328 vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
3329 if(!s->mb_intra) vc1_mc_4mv_luma(v, i);
3330 intra_count += s->mb_intra;
3331 is_intra[i] = s->mb_intra;
3332 is_coded[i] = mb_has_coeffs;
3333 }
3334 if(i&4){
3335 is_intra[i] = (intra_count >= 3);
3336 is_coded[i] = val;
3337 }
3338 if(i == 4) vc1_mc_4mv_chroma(v);
3339 v->mb_type[0][s->block_index[i]] = is_intra[i];
3340 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];
3341 }
3342 // if there are no coded blocks then don't do anything more
3343 if(!intra_count && !coded_inter) return 0;
3344 dst_idx = 0;
3345 GET_MQUANT();
3346 s->current_picture.qscale_table[mb_pos] = mquant;
3347 /* test if block is intra and has pred */
3348 {
3349 int intrapred = 0;
3350 for(i=0; i<6; i++)
3351 if(is_intra[i]) {
3352 if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])
3353 || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {
3354 intrapred = 1;
3355 break;
3356 }
3357 }
3358 if(intrapred)s->ac_pred = get_bits1(gb);
3359 else s->ac_pred = 0;
3360 }
3361 if (!v->ttmbf && coded_inter)
3362 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3363 for (i=0; i<6; i++)
3364 {
3365 dst_idx += i >> 2;
3366 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3367 s->mb_intra = is_intra[i];
3368 if (is_intra[i]) {
3369 /* check if prediction blocks A and C are available */
3370 v->a_avail = v->c_avail = 0;
3371 if(i == 2 || i == 3 || !s->first_slice_line)
3372 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3373 if(i == 1 || i == 3 || s->mb_x)
3374 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3375
3376 vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);
3377 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3378 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3379 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3380 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
3381 if(v->pq >= 9 && v->overlap) {
3382 if(v->c_avail)
3383 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3384 if(v->a_avail)
3385 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3386 }
3387 if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3388 int left_cbp, top_cbp;
3389 if(i & 4){
3390 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3391 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3392 }else{
3393 left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3394 top_cbp = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3395 }
3396 if(left_cbp & 0xC)
3397 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3398 if(top_cbp & 0xA)
3399 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3400 }
3401 block_cbp |= 0xF << (i << 2);
3402 } else if(is_coded[i]) {
3403 int left_cbp = 0, top_cbp = 0, filter = 0;
3404 if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3405 filter = 1;
3406 if(i & 4){
3407 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3408 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3409 }else{
3410 left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3411 top_cbp = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3412 }
3413 }
3414 pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);
3415 block_cbp |= pat << (i << 2);
3416 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3417 first_block = 0;
3418 }
3419 }
3420 return 0;
3421 }
3422 else //Skipped MB
3423 {
3424 s->mb_intra = 0;
3425 s->current_picture.qscale_table[mb_pos] = 0;
3426 for (i=0; i<6; i++) {
3427 v->mb_type[0][s->block_index[i]] = 0;
3428 s->dc_val[0][s->block_index[i]] = 0;
3429 }
3430 for (i=0; i<4; i++)
3431 {
3432 vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
3433 vc1_mc_4mv_luma(v, i);
3434 }
3435 vc1_mc_4mv_chroma(v);
3436 s->current_picture.qscale_table[mb_pos] = 0;
3437 return 0;
3438 }
3439 }
3440 v->cbp[s->mb_x] = block_cbp;
3441
3442 /* Should never happen */
3443 return -1;
3444 }
3445
3446 /** Decode one B-frame MB (in Main profile)
3447 */
3448 static void vc1_decode_b_mb(VC1Context *v)
3449 {
3450 MpegEncContext *s = &v->s;
3451 GetBitContext *gb = &s->gb;
3452 int i, j;
3453 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3454 int cbp = 0; /* cbp decoding stuff */
3455 int mqdiff, mquant; /* MB quantization */
3456 int ttmb = v->ttfrm; /* MB Transform type */
3457
3458 static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3459 offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3460 int mb_has_coeffs = 0; /* last_flag */
3461 int index, index1; /* LUT indexes */
3462 int val, sign; /* temp values */
3463 int first_block = 1;
3464 int dst_idx, off;
3465 int skipped, direct;
3466 int dmv_x[2], dmv_y[2];
3467 int bmvtype = BMV_TYPE_BACKWARD;
3468
3469 mquant = v->pq; /* Loosy initialization */
3470 s->mb_intra = 0;
3471
3472 if (v->dmb_is_raw)
3473 direct = get_bits1(gb);
3474 else
3475 direct = v->direct_mb_plane[mb_pos];
3476 if (v->skip_is_raw)
3477 skipped = get_bits1(gb);
3478 else
3479 skipped = v->s.mbskip_table[mb_pos];
3480
3481 s->dsp.clear_blocks(s->block[0]);
3482 dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
3483 for(i = 0; i < 6; i++) {
3484 v->mb_type[0][s->block_index[i]] = 0;
3485 s->dc_val[0][s->block_index[i]] = 0;
3486 }
3487 s->current_picture.qscale_table[mb_pos] = 0;
3488
3489 if (!direct) {
3490 if (!skipped) {
3491 GET_MVDATA(dmv_x[0], dmv_y[0]);
3492 dmv_x[1] = dmv_x[0];
3493 dmv_y[1] = dmv_y[0];
3494 }
3495 if(skipped || !s->mb_intra) {
3496 bmvtype = decode012(gb);
3497 switch(bmvtype) {
3498 case 0:
3499 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
3500 break;
3501 case 1:
3502 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
3503 break;
3504 case 2:
3505 bmvtype = BMV_TYPE_INTERPOLATED;
3506 dmv_x[0] = dmv_y[0] = 0;
3507 }
3508 }
3509 }
3510 for(i = 0; i < 6; i++)
3511 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3512
3513 if (skipped) {
3514 if(direct) bmvtype = BMV_TYPE_INTERPOLATED;
3515 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3516 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3517 return;
3518 }
3519 if (direct) {
3520 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3521 GET_MQUANT();
3522 s->mb_intra = 0;
3523 mb_has_coeffs = 0;
3524 s->current_picture.qscale_table[mb_pos] = mquant;
3525 if(!v->ttmbf)
3526 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3527 dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;
3528 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3529 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3530 } else {
3531 if(!mb_has_coeffs && !s->mb_intra) {
3532 /* no coded blocks - effectively skipped */
3533 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3534 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3535 return;
3536 }
3537 if(s->mb_intra && !mb_has_coeffs) {
3538 GET_MQUANT();
3539 s->current_picture.qscale_table[mb_pos] = mquant;
3540 s->ac_pred = get_bits1(gb);
3541 cbp = 0;
3542 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3543 } else {
3544 if(bmvtype == BMV_TYPE_INTERPOLATED) {
3545 GET_MVDATA(dmv_x[0], dmv_y[0]);
3546 if(!mb_has_coeffs) {
3547 /* interpolated skipped block */
3548 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3549 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3550 return;
3551 }
3552 }
3553 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3554 if(!s->mb_intra) {
3555 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3556 }
3557 if(s->mb_intra)
3558 s->ac_pred = get_bits1(gb);
3559 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3560 GET_MQUANT();
3561 s->current_picture.qscale_table[mb_pos] = mquant;
3562 if(!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3563 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3564 }
3565 }
3566 dst_idx = 0;
3567 for (i=0; i<6; i++)
3568 {
3569 s->dc_val[0][s->block_index[i]] = 0;
3570 dst_idx += i >> 2;
3571 val = ((cbp >> (5 - i)) & 1);
3572 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3573 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3574 if(s->mb_intra) {
3575 /* check if prediction blocks A and C are available */
3576 v->a_avail = v->c_avail = 0;
3577 if(i == 2 || i == 3 || !s->first_slice_line)
3578 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3579 if(i == 1 || i == 3 || s->mb_x)
3580 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3581
3582 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3583 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3584 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3585 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3586 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3587 } else if(val) {
3588 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), 0, 0, 0);
3589 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3590 first_block = 0;
3591 }
3592 }
3593 }
3594
3595 /** Decode blocks of I-frame
3596 */
3597 static void vc1_decode_i_blocks(VC1Context *v)
3598 {
3599 int k, j;
3600 MpegEncContext *s = &v->s;
3601 int cbp, val;
3602 uint8_t *coded_val;
3603 int mb_pos;
3604
3605 /* select codingmode used for VLC tables selection */
3606 switch(v->y_ac_table_index){
3607 case 0:
3608 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3609 break;
3610 case 1:
3611 v->codingset = CS_HIGH_MOT_INTRA;
3612 break;
3613 case 2:
3614 v->codingset = CS_MID_RATE_INTRA;
3615 break;
3616 }
3617
3618 switch(v->c_ac_table_index){
3619 case 0:
3620 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3621 break;
3622 case 1:
3623 v->codingset2 = CS_HIGH_MOT_INTER;
3624 break;
3625 case 2:
3626 v->codingset2 = CS_MID_RATE_INTER;
3627 break;
3628 }
3629
3630 /* Set DC scale - y and c use the same */
3631 s->y_dc_scale = s->y_dc_scale_table[v->pq];
3632 s->c_dc_scale = s->c_dc_scale_table[v->pq];
3633
3634 //do frame decode
3635 s->mb_x = s->mb_y = 0;
3636 s->mb_intra = 1;
3637 s->first_slice_line = 1;
3638 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3639 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3640 ff_init_block_index(s);
3641 ff_update_block_index(s);
3642 s->dsp.clear_blocks(s->block[0]);
3643 mb_pos = s->mb_x + s->mb_y * s->mb_width;
3644 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3645 s->current_picture.qscale_table[mb_pos] = v->pq;
3646 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3647 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3648
3649 // do actual MB decoding and displaying
3650 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3651 v->s.ac_pred = get_bits1(&v->s.gb);
3652
3653 for(k = 0; k < 6; k++) {
3654 val = ((cbp >> (5 - k)) & 1);
3655
3656 if (k < 4) {
3657 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3658 val = val ^ pred;
3659 *coded_val = val;
3660 }
3661 cbp |= val << (5 - k);
3662
3663 vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);
3664
3665 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3666 if(v->pq >= 9 && v->overlap) {
3667 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3668 }
3669 }
3670
3671 vc1_put_block(v, s->block);
3672 if(v->pq >= 9 && v->overlap) {
3673 if(s->mb_x) {
3674 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3675 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3676 if(!(s->flags & CODEC_FLAG_GRAY)) {
3677 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3678 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3679 }
3680 }
3681 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3682 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3683 if(!s->first_slice_line) {
3684 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3685 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3686 if(!(s->flags & CODEC_FLAG_GRAY)) {
3687 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3688 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3689 }
3690 }
3691 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3692 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3693 }
3694 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3695
3696 if(get_bits_count(&s->gb) > v->bits) {
3697 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3698 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3699 return;
3700 }
3701 }
3702 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3703 s->first_slice_line = 0;
3704 }
3705 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3706 }
3707
3708 /** Decode blocks of I-frame for advanced profile
3709 */
3710 static void vc1_decode_i_blocks_adv(VC1Context *v)
3711 {
3712 int k, j;
3713 MpegEncContext *s = &v->s;
3714 int cbp, val;
3715 uint8_t *coded_val;
3716 int mb_pos;
3717 int mquant = v->pq;
3718 int mqdiff;
3719 int overlap;
3720 GetBitContext *gb = &s->gb;
3721
3722 /* select codingmode used for VLC tables selection */
3723 switch(v->y_ac_table_index){
3724 case 0:
3725 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3726 break;
3727 case 1:
3728 v->codingset = CS_HIGH_MOT_INTRA;
3729 break;
3730 case 2:
3731 v->codingset = CS_MID_RATE_INTRA;
3732 break;
3733 }
3734
3735 switch(v->c_ac_table_index){
3736 case 0:
3737 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3738 break;
3739 case 1:
3740 v->codingset2 = CS_HIGH_MOT_INTER;
3741 break;
3742 case 2:
3743 v->codingset2 = CS_MID_RATE_INTER;
3744 break;
3745 }
3746
3747 //do frame decode
3748 s->mb_x = s->mb_y = 0;
3749 s->mb_intra = 1;
3750 s->first_slice_line = 1;
3751 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3752 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3753 ff_init_block_index(s);
3754 ff_update_block_index(s);
3755 s->dsp.clear_blocks(s->block[0]);
3756 mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3757 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3758 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3759 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3760
3761 // do actual MB decoding and displaying
3762 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3763 if(v->acpred_is_raw)
3764 v->s.ac_pred = get_bits1(&v->s.gb);
3765 else
3766 v->s.ac_pred = v->acpred_plane[mb_pos];
3767
3768 if(v->condover == CONDOVER_SELECT) {
3769 if(v->overflg_is_raw)
3770 overlap = get_bits1(&v->s.gb);
3771 else
3772 overlap = v->over_flags_plane[mb_pos];
3773 } else
3774 overlap = (v->condover == CONDOVER_ALL);
3775
3776 GET_MQUANT();
3777
3778 s->current_picture.qscale_table[mb_pos] = mquant;
3779 /* Set DC scale - y and c use the same */
3780 s->y_dc_scale = s->y_dc_scale_table[mquant];
3781 s->c_dc_scale = s->c_dc_scale_table[mquant];
3782
3783 for(k = 0; k < 6; k++) {
3784 val = ((cbp >> (5 - k)) & 1);
3785
3786 if (k < 4) {
3787 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3788 val = val ^ pred;
3789 *coded_val = val;
3790 }
3791 cbp |= val << (5 - k);
3792
3793 v->a_avail = !s->first_slice_line || (k==2 || k==3);
3794 v->c_avail = !!s->mb_x || (k==1 || k==3);
3795
3796 vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);
3797
3798 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3799 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3800 }
3801
3802 vc1_put_block(v, s->block);
3803 if(overlap) {
3804 if(s->mb_x) {
3805 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3806 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3807 if(!(s->flags & CODEC_FLAG_GRAY)) {
3808 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3809 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3810 }
3811 }
3812 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3813 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3814 if(!s->first_slice_line) {
3815 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3816 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3817 if(!(s->flags & CODEC_FLAG_GRAY)) {
3818 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3819 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3820 }
3821 }
3822 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3823 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3824 }
3825 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3826
3827 if(get_bits_count(&s->gb) > v->bits) {
3828 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3829 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3830 return;
3831 }
3832 }
3833 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3834 s->first_slice_line = 0;
3835 }
3836 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3837 }
3838
3839 static void vc1_decode_p_blocks(VC1Context *v)
3840 {
3841 MpegEncContext *s = &v->s;
3842
3843 /* select codingmode used for VLC tables selection */
3844 switch(v->c_ac_table_index){
3845 case 0:
3846 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3847 break;
3848 case 1:
3849 v->codingset = CS_HIGH_MOT_INTRA;
3850 break;
3851 case 2:
3852 v->codingset = CS_MID_RATE_INTRA;
3853 break;
3854 }
3855
3856 switch(v->c_ac_table_index){
3857 case 0:
3858 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3859 break;
3860 case 1:
3861 v->codingset2 = CS_HIGH_MOT_INTER;
3862 break;
3863 case 2:
3864 v->codingset2 = CS_MID_RATE_INTER;
3865 break;
3866 }
3867
3868 s->first_slice_line = 1;
3869 memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);
3870 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3871 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3872 ff_init_block_index(s);
3873 ff_update_block_index(s);
3874 s->dsp.clear_blocks(s->block[0]);
3875
3876 vc1_decode_p_mb(v);
3877 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3878 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3879 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits,s->mb_x,s->mb_y);
3880 return;
3881 }
3882 }
3883 memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride);
3884 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3885 s->first_slice_line = 0;
3886 }
3887 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3888 }
3889
3890 static void vc1_decode_b_blocks(VC1Context *v)
3891 {
3892 MpegEncContext *s = &v->s;
3893
3894 /* select codingmode used for VLC tables selection */
3895 switch(v->c_ac_table_index){
3896 case 0:
3897 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3898 break;
3899 case 1:
3900 v->codingset = CS_HIGH_MOT_INTRA;
3901 break;
3902 case 2:
3903 v->codingset = CS_MID_RATE_INTRA;
3904 break;
3905 }
3906
3907 switch(v->c_ac_table_index){
3908 case 0:
3909 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3910 break;
3911 case 1:
3912 v->codingset2 = CS_HIGH_MOT_INTER;
3913 break;
3914 case 2:
3915 v->codingset2 = CS_MID_RATE_INTER;
3916 break;
3917 }
3918
3919 s->first_slice_line = 1;
3920 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3921 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3922 ff_init_block_index(s);
3923 ff_update_block_index(s);
3924 s->dsp.clear_blocks(s->block[0]);
3925
3926 vc1_decode_b_mb(v);
3927 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3928 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3929 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits,s->mb_x,s->mb_y);
3930 return;
3931 }
3932 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[s->mb_x + s->mb_y *s->mb_stride]);
3933 }
3934 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3935 s->first_slice_line = 0;
3936 }
3937 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3938 }
3939
3940 static void vc1_decode_skip_blocks(VC1Context *v)
3941 {
3942 MpegEncContext *s = &v->s;
3943
3944 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3945 s->first_slice_line = 1;
3946 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3947 s->mb_x = 0;
3948 ff_init_block_index(s);
3949 ff_update_block_index(s);
3950 memcpy(s->dest[0], s->last_picture.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16);
3951 memcpy(s->dest[1], s->last_picture.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3952 memcpy(s->dest[2], s->last_picture.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3953 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3954 s->first_slice_line = 0;
3955 }
3956 s->pict_type = FF_P_TYPE;
3957 }
3958
3959 static void vc1_decode_blocks(VC1Context *v)
3960 {
3961
3962 v->s.esc3_level_length = 0;
3963 if(v->x8_type){
3964 ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );
3965 }else{
3966
3967 switch(v->s.pict_type) {
3968 case FF_I_TYPE:
3969 if(v->profile == PROFILE_ADVANCED)
3970 vc1_decode_i_blocks_adv(v);
3971 else
3972 vc1_decode_i_blocks(v);
3973 break;
3974 case FF_P_TYPE:
3975 if(v->p_frame_skipped)
3976 vc1_decode_skip_blocks(v);
3977 else
3978 vc1_decode_p_blocks(v);
3979 break;
3980 case FF_B_TYPE:
3981 if(v->bi_type){
3982 if(v->profile == PROFILE_ADVANCED)
3983 vc1_decode_i_blocks_adv(v);
3984 else
3985 vc1_decode_i_blocks(v);
3986 }else
3987 vc1_decode_b_blocks(v);
3988 break;
3989 }
3990 }
3991 }
3992
3993 /** Find VC-1 marker in buffer
3994 * @return position where next marker starts or end of buffer if no marker found
3995 */
3996 static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)
3997 {
3998 uint32_t mrk = 0xFFFFFFFF;
3999
4000 if(end-src < 4) return end;
4001 while(src < end){
4002 mrk = (mrk << 8) | *src++;
4003 if(IS_MARKER(mrk))
4004 return src-4;
4005 }
4006 return end;
4007 }
4008
4009 static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)
4010 {
4011 int dsize = 0, i;
4012
4013 if(size < 4){
4014 for(dsize = 0; dsize < size; dsize++) *dst++ = *src++;
4015 return size;
4016 }
4017 for(i = 0; i < size; i++, src++) {
4018 if(src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {
4019 dst[dsize++] = src[1];
4020 src++;
4021 i++;
4022 } else
4023 dst[dsize++] = *src;
4024 }
4025 return dsize;
4026 }
4027
4028 /** Initialize a VC1/WMV3 decoder
4029 * @todo TODO: Handle VC-1 IDUs (Transport level?)
4030 * @todo TODO: Decypher remaining bits in extra_data
4031 */
4032 static av_cold int vc1_decode_init(AVCodecContext *avctx)
4033 {
4034 VC1Context *v = avctx->priv_data;
4035 MpegEncContext *s = &v->s;
4036 GetBitContext gb;
4037
4038 if (!avctx->extradata_size || !avctx->extradata) return -1;
4039 if (!(avctx->flags & CODEC_FLAG_GRAY))
4040 avctx->pix_fmt = PIX_FMT_YUV420P;
4041 else
4042 avctx->pix_fmt = PIX_FMT_GRAY8;
4043 v->s.avctx = avctx;
4044 avctx->flags |= CODEC_FLAG_EMU_EDGE;
4045 v->s.flags |= CODEC_FLAG_EMU_EDGE;
4046
4047 if(avctx->idct_algo==FF_IDCT_AUTO){
4048 avctx->idct_algo=FF_IDCT_WMV2;
4049 }
4050
4051 if(ff_h263_decode_init(avctx) < 0)
4052 return -1;
4053 if (vc1_init_common(v) < 0) return -1;
4054
4055 avctx->coded_width = avctx->width;
4056 avctx->coded_height = avctx->height;
4057 if ((avctx->codec_id == CODEC_ID_WMV3) || (avctx->codec_id == CODEC_ID_WMV3_VDPAU))
4058 {
4059 int count = 0;
4060
4061 // looks like WMV3 has a sequence header stored in the extradata
4062 // advanced sequence header may be before the first frame
4063 // the last byte of the extradata is a version number, 1 for the
4064 // samples we can decode
4065
4066 init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);
4067
4068 if (decode_sequence_header(avctx, &gb) < 0)
4069 return -1;
4070
4071 count = avctx->extradata_size*8 - get_bits_count(&gb);
4072 if (count>0)
4073 {
4074 av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",
4075 count, get_bits(&gb, count));
4076 }
4077 else if (count < 0)
4078 {
4079 av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);
4080 }
4081 } else { // VC1/WVC1
4082 const uint8_t *start = avctx->extradata;
4083 uint8_t *end = avctx->extradata + avctx->extradata_size;
4084 const uint8_t *next;
4085 int size, buf2_size;
4086 uint8_t *buf2 = NULL;
4087 int seq_initialized = 0, ep_initialized = 0;
4088
4089 if(avctx->extradata_size < 16) {
4090 av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);
4091 return -1;
4092 }
4093
4094 buf2 = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
4095 if(start[0]) start++; // in WVC1 extradata first byte is its size
4096 next = start;
4097 for(; next < end; start = next){
4098 next = find_next_marker(start + 4, end);
4099 size = next - start - 4;
4100 if(size <= 0) continue;
4101 buf2_size = vc1_unescape_buffer(start + 4, size, buf2);
4102 init_get_bits(&gb, buf2, buf2_size * 8);
4103 switch(AV_RB32(start)){
4104 case VC1_CODE_SEQHDR:
4105 if(decode_sequence_header(avctx, &gb) < 0){
4106 av_free(buf2);
4107 return -1;
4108 }
4109 seq_initialized = 1;
4110 break;
4111 case VC1_CODE_ENTRYPOINT:
4112 if(decode_entry_point(avctx, &gb) < 0){
4113 av_free(buf2);
4114 return -1;
4115 }
4116 ep_initialized = 1;
4117 break;
4118 }
4119 }
4120 av_free(buf2);
4121 if(!seq_initialized || !ep_initialized){
4122 av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");
4123 return -1;
4124 }
4125 }
4126 avctx->has_b_frames= !!(avctx->max_b_frames);
4127 s->low_delay = !avctx->has_b_frames;
4128
4129 s->mb_width = (avctx->coded_width+15)>>4;
4130 s->mb_height = (avctx->coded_height+15)>>4;
4131
4132 /* Allocate mb bitplanes */
4133 v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4134 v->direct_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4135 v->acpred_plane = av_malloc(s->mb_stride * s->mb_height);
4136 v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height);
4137
4138 v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);
4139 v->cbp = v->cbp_base + s->mb_stride;
4140
4141 /* allocate block type info in that way so it could be used with s->block_index[] */
4142 v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);
4143 v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;
4144 v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;
4145 v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);
4146
4147 /* Init coded blocks info */
4148 if (v->profile == PROFILE_ADVANCED)
4149 {
4150 // if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
4151 // return -1;
4152 // if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
4153 // return -1;
4154 }
4155
4156 ff_intrax8_common_init(&v->x8,s);
4157 return 0;
4158 }
4159
4160
4161 /** Decode a VC1/WMV3 frame
4162 * @todo TODO: Handle VC-1 IDUs (Transport level?)
4163 */
4164 static int vc1_decode_frame(AVCodecContext *avctx,
4165 void *data, int *data_size,
4166 const uint8_t *buf, int buf_size)
4167 {
4168 VC1Context *v = avctx->priv_data;
4169 MpegEncContext *s = &v->s;
4170 AVFrame *pict = data;
4171 uint8_t *buf2 = NULL;
4172 #ifdef HAVE_VDPAU
4173 const uint8_t *buf_vdpau = buf;
4174 #endif
4175
4176 /* no supplementary picture */
4177 if (buf_size == 0) {
4178 /* special case for last picture */
4179 if (s->low_delay==0 && s->next_picture_ptr) {
4180 *pict= *(AVFrame*)s->next_picture_ptr;
4181 s->next_picture_ptr= NULL;
4182
4183 *data_size = sizeof(AVFrame);
4184 }
4185
4186 return 0;
4187 }
4188
4189 /* We need to set current_picture_ptr before reading the header,
4190 * otherwise we cannot store anything in there. */
4191 if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){
4192 int i= ff_find_unused_picture(s, 0);
4193 s->current_picture_ptr= &s->picture[i];
4194 }
4195
4196 #ifdef HAVE_VDPAU
4197 // pxt_fmt calculation for VDPAU.
4198 if (decode_postinit(v, avctx) < 0)
4199 return -1;
4200 #endif
4201
4202 //for advanced profile we may need to parse and unescape data
4203 if ((avctx->codec_id == CODEC_ID_VC1) || (avctx->codec_id == CODEC_ID_VC1_VDPAU)) {
4204 int buf_size2 = 0;
4205 buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
4206
4207 if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */
4208 const uint8_t *start, *end, *next;
4209 int size;
4210
4211 next = buf;
4212 for(start = buf, end = buf + buf_size; next < end; start = next){
4213 next = find_next_marker(start + 4, end);
4214 size = next - start - 4;
4215 if(size <= 0) continue;
4216 switch(AV_RB32(start)){
4217 case VC1_CODE_FRAME:
4218 #ifdef HAVE_VDPAU
4219 buf_vdpau = start;
4220 #endif
4221 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4222 break;
4223 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */
4224 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4225 init_get_bits(&s->gb, buf2, buf_size2*8);
4226 decode_entry_point(avctx, &s->gb);
4227 break;
4228 case VC1_CODE_SLICE:
4229 av_log(avctx, AV_LOG_ERROR, "Sliced decoding is not implemented (yet)\n");
4230 av_free(buf2);
4231 return -1;
4232 }
4233 }
4234 }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */
4235 const uint8_t *divider;
4236
4237 divider = find_next_marker(buf, buf + buf_size);
4238 if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){
4239 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");
4240 av_free(buf2);
4241 return -1;
4242 }
4243
4244 buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);
4245 // TODO
4246 av_free(buf2);return -1;
4247 }else{
4248 buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);
4249 }
4250 init_get_bits(&s->gb, buf2, buf_size2*8);
4251 } else
4252 init_get_bits(&s->gb, buf, buf_size*8);
4253 // do parse frame header
4254 if(v->profile < PROFILE_ADVANCED) {
4255 if(vc1_parse_frame_header(v, &s->gb) == -1) {
4256 av_free(buf2);
4257 return -1;
4258 }
4259 } else {
4260 if(vc1_parse_frame_header_adv(v, &s->gb) == -1) {
4261 av_free(buf2);
4262 return -1;
4263 }
4264 }
4265
4266 if(s->pict_type != FF_I_TYPE && !v->res_rtm_flag){
4267 av_free(buf2);
4268 return -1;
4269 }
4270
4271 // for hurry_up==5
4272 s->current_picture.pict_type= s->pict_type;
4273 s->current_picture.key_frame= s->pict_type == FF_I_TYPE;
4274
4275 /* skip B-frames if we don't have reference frames */
4276 if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){
4277 av_free(buf2);
4278 return -1;//buf_size;
4279 }
4280 /* skip b frames if we are in a hurry */
4281 if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;
4282 if( (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
4283 || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
4284 || avctx->skip_frame >= AVDISCARD_ALL) {
4285 av_free(buf2);
4286 return buf_size;
4287 }
4288 /* skip everything if we are in a hurry>=5 */
4289 if(avctx->hurry_up>=5) {
4290 av_free(buf2);
4291 return -1;//buf_size;
4292 }
4293
4294 if(s->next_p_frame_damaged){
4295 if(s->pict_type==FF_B_TYPE)
4296 return buf_size;
4297 else
4298 s->next_p_frame_damaged=0;
4299 }
4300
4301 if(MPV_frame_start(s, avctx) < 0) {
4302 av_free(buf2);
4303 return -1;
4304 }
4305
4306 #ifdef HAVE_VDPAU
4307 // MPV_frame_start() calls to get_buffer/videoSurfaces. Now we call
4308 // VDPAU_vc1_field_start where picture-parameters are filled.
4309 // VDPAU_vc1_picture_complete calls to vdpau_decoder_render.
4310
4311 if (avctx->vdpau_acceleration) {
4312 if (ff_VDPAU_vc1_decode_picture(s, avctx, v, buf_vdpau, (buf + buf_size) - buf_vdpau) < 0) {
4313 av_free(buf2);
4314 return -1;
4315 }
4316 }
4317 #endif
4318
4319 s->me.qpel_put= s->dsp.put_qpel_pixels_tab;
4320 s->me.qpel_avg= s->dsp.avg_qpel_pixels_tab;
4321
4322 #ifdef HAVE_VDPAU
4323 if (!avctx->vdpau_acceleration) {
4324 #endif
4325 ff_er_frame_start(s);
4326
4327 v->bits = buf_size * 8;
4328 vc1_decode_blocks(v);
4329 //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);
4330 // if(get_bits_count(&s->gb) > buf_size * 8)
4331 // return -1;
4332 ff_er_frame_end(s);
4333 #ifdef HAVE_VDPAU
4334 }
4335 #endif
4336
4337 MPV_frame_end(s);
4338
4339 assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);
4340 assert(s->current_picture.pict_type == s->pict_type);
4341 if (s->pict_type == FF_B_TYPE || s->low_delay) {
4342 *pict= *(AVFrame*)s->current_picture_ptr;
4343 } else if (s->last_picture_ptr != NULL) {
4344 *pict= *(AVFrame*)s->last_picture_ptr;
4345 }
4346
4347 if(s->last_picture_ptr || s->low_delay){
4348 *data_size = sizeof(AVFrame);
4349 ff_print_debug_info(s, pict);
4350 }
4351
4352 /* Return the Picture timestamp as the frame number */
4353 /* we subtract 1 because it is added on utils.c */
4354 avctx->frame_number = s->picture_number - 1;
4355
4356 av_free(buf2);
4357 return buf_size;
4358 }
4359
4360
4361 /** Close a VC1/WMV3 decoder
4362 * @warning Initial try at using MpegEncContext stuff
4363 */
4364 static av_cold int vc1_decode_end(AVCodecContext *avctx)
4365 {
4366 VC1Context *v = avctx->priv_data;
4367
4368 av_freep(&v->hrd_rate);
4369 av_freep(&v->hrd_buffer);
4370 MPV_common_end(&v->s);
4371 av_freep(&v->mv_type_mb_plane);
4372 av_freep(&v->direct_mb_plane);
4373 av_freep(&v->acpred_plane);
4374 av_freep(&v->over_flags_plane);
4375 av_freep(&v->mb_type_base);
4376 av_freep(&v->cbp_base);
4377 ff_intrax8_common_end(&v->x8);
4378 return 0;
4379 }
4380
4381
4382 AVCodec vc1_decoder = {
4383 "vc1",
4384 CODEC_TYPE_VIDEO,
4385 CODEC_ID_VC1,
4386 sizeof(VC1Context),
4387 vc1_decode_init,
4388 NULL,
4389 vc1_decode_end,
4390 vc1_decode_frame,
4391 CODEC_CAP_DELAY,
4392 NULL,
4393 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),
4394 };
4395
4396 AVCodec wmv3_decoder = {
4397 "wmv3",
4398 CODEC_TYPE_VIDEO,
4399 CODEC_ID_WMV3,
4400 sizeof(VC1Context),
4401 vc1_decode_init,
4402 NULL,
4403 vc1_decode_end,
4404 vc1_decode_frame,
4405 CODEC_CAP_DELAY,
4406 NULL,
4407 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),
4408 };
4409
4410 #ifdef HAVE_VDPAU
4411 static av_cold int vc1_vdpau_decode_init(AVCodecContext *avctx){
4412 if( avctx->thread_count > 1)
4413 return -1;
4414 if( !(avctx->slice_flags & SLICE_FLAG_CODED_ORDER) )
4415 return -1;
4416 if( !(avctx->slice_flags & SLICE_FLAG_ALLOW_FIELD) ){
4417 dprintf(avctx, "vc1.c: VDPAU decoder does not set SLICE_FLAG_ALLOW_FIELD\n");
4418 }
4419 avctx->vdpau_acceleration = 1;
4420 vc1_decode_init(avctx);
4421 avctx->pix_fmt = PIX_FMT_NONE;
4422
4423 return 0;
4424 }
4425
4426 AVCodec wmv3_vdpau_decoder = {
4427 "wmv3_vdpau",
4428 CODEC_TYPE_VIDEO,
4429 CODEC_ID_WMV3_VDPAU,
4430 sizeof(VC1Context),
4431 vc1_vdpau_decode_init,
4432 NULL,
4433 vc1_decode_end,
4434 vc1_decode_frame,
4435 CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,
4436 NULL,
4437 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 VDPAU"),
4438 };
4439
4440 AVCodec vc1_vdpau_decoder = {
4441 "vc1_vdpau",
4442 CODEC_TYPE_VIDEO,
4443 CODEC_ID_VC1_VDPAU,
4444 sizeof(VC1Context),
4445 vc1_vdpau_decode_init,
4446 NULL,
4447 vc1_decode_end,
4448 vc1_decode_frame,
4449 CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,
4450 NULL,
4451 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1 VDPAU"),
4452 };
4453 #endif
VDPAU for FFmpeg