?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
More cosmetics
[FFMpeg-mirror/DVCPRO-HD.git] / libavcodec / vc1.c
blob3975d0400e0bf47b86946c584c5e07c42e1a5614
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 static const uint16_t table_mb_intra[64][2];
45
46
47 /**
48 * Init VC-1 specific tables and VC1Context members
49 * @param v The VC1Context to initialize
50 * @return Status
51 */
52 static int vc1_init_common(VC1Context *v)
53 {
54 static int done = 0;
55 int i = 0;
56
57 v->hrd_rate = v->hrd_buffer = NULL;
58
59 /* VLC tables */
60 if(!done)
61 {
62 done = 1;
63 init_vlc(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,
64 ff_vc1_bfraction_bits, 1, 1,
65 ff_vc1_bfraction_codes, 1, 1, 1);
66 init_vlc(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,
67 ff_vc1_norm2_bits, 1, 1,
68 ff_vc1_norm2_codes, 1, 1, 1);
69 init_vlc(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,
70 ff_vc1_norm6_bits, 1, 1,
71 ff_vc1_norm6_codes, 2, 2, 1);
72 init_vlc(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,
73 ff_vc1_imode_bits, 1, 1,
74 ff_vc1_imode_codes, 1, 1, 1);
75 for (i=0; i<3; i++)
76 {
77 init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,
78 ff_vc1_ttmb_bits[i], 1, 1,
79 ff_vc1_ttmb_codes[i], 2, 2, 1);
80 init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,
81 ff_vc1_ttblk_bits[i], 1, 1,
82 ff_vc1_ttblk_codes[i], 1, 1, 1);
83 init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,
84 ff_vc1_subblkpat_bits[i], 1, 1,
85 ff_vc1_subblkpat_codes[i], 1, 1, 1);
86 }
87 for(i=0; i<4; i++)
88 {
89 init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,
90 ff_vc1_4mv_block_pattern_bits[i], 1, 1,
91 ff_vc1_4mv_block_pattern_codes[i], 1, 1, 1);
92 init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,
93 ff_vc1_cbpcy_p_bits[i], 1, 1,
94 ff_vc1_cbpcy_p_codes[i], 2, 2, 1);
95 init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,
96 ff_vc1_mv_diff_bits[i], 1, 1,
97 ff_vc1_mv_diff_codes[i], 2, 2, 1);
98 }
99 for(i=0; i<8; i++)
100 init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],
101 &vc1_ac_tables[i][0][1], 8, 4,
102 &vc1_ac_tables[i][0][0], 8, 4, 1);
103 init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,
104 &ff_msmp4_mb_i_table[0][1], 4, 2,
105 &ff_msmp4_mb_i_table[0][0], 4, 2, 1);
106 }
107
108 /* Other defaults */
109 v->pq = -1;
110 v->mvrange = 0; /* 7.1.1.18, p80 */
111
112 return 0;
113 }
114
115 /***********************************************************************/
116 /**
117 * @defgroup bitplane VC9 Bitplane decoding
118 * @see 8.7, p56
119 * @{
120 */
121
122 /** @addtogroup bitplane
123 * Imode types
124 * @{
125 */
126 enum Imode {
127 IMODE_RAW,
128 IMODE_NORM2,
129 IMODE_DIFF2,
130 IMODE_NORM6,
131 IMODE_DIFF6,
132 IMODE_ROWSKIP,
133 IMODE_COLSKIP
134 };
135 /** @} */ //imode defines
136
137 /** Decode rows by checking if they are skipped
138 * @param plane Buffer to store decoded bits
139 * @param[in] width Width of this buffer
140 * @param[in] height Height of this buffer
141 * @param[in] stride of this buffer
142 */
143 static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
144 int x, y;
145
146 for (y=0; y<height; y++){
147 if (!get_bits1(gb)) //rowskip
148 memset(plane, 0, width);
149 else
150 for (x=0; x<width; x++)
151 plane[x] = get_bits1(gb);
152 plane += stride;
153 }
154 }
155
156 /** Decode columns by checking if they are skipped
157 * @param plane Buffer to store decoded bits
158 * @param[in] width Width of this buffer
159 * @param[in] height Height of this buffer
160 * @param[in] stride of this buffer
161 * @todo FIXME: Optimize
162 */
163 static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
164 int x, y;
165
166 for (x=0; x<width; x++){
167 if (!get_bits1(gb)) //colskip
168 for (y=0; y<height; y++)
169 plane[y*stride] = 0;
170 else
171 for (y=0; y<height; y++)
172 plane[y*stride] = get_bits1(gb);
173 plane ++;
174 }
175 }
176
177 /** Decode a bitplane's bits
178 * @param bp Bitplane where to store the decode bits
179 * @param v VC-1 context for bit reading and logging
180 * @return Status
181 * @todo FIXME: Optimize
182 */
183 static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)
184 {
185 GetBitContext *gb = &v->s.gb;
186
187 int imode, x, y, code, offset;
188 uint8_t invert, *planep = data;
189 int width, height, stride;
190
191 width = v->s.mb_width;
192 height = v->s.mb_height;
193 stride = v->s.mb_stride;
194 invert = get_bits1(gb);
195 imode = get_vlc2(gb, ff_vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);
196
197 *raw_flag = 0;
198 switch (imode)
199 {
200 case IMODE_RAW:
201 //Data is actually read in the MB layer (same for all tests == "raw")
202 *raw_flag = 1; //invert ignored
203 return invert;
204 case IMODE_DIFF2:
205 case IMODE_NORM2:
206 if ((height * width) & 1)
207 {
208 *planep++ = get_bits1(gb);
209 offset = 1;
210 }
211 else offset = 0;
212 // decode bitplane as one long line
213 for (y = offset; y < height * width; y += 2) {
214 code = get_vlc2(gb, ff_vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
215 *planep++ = code & 1;
216 offset++;
217 if(offset == width) {
218 offset = 0;
219 planep += stride - width;
220 }
221 *planep++ = code >> 1;
222 offset++;
223 if(offset == width) {
224 offset = 0;
225 planep += stride - width;
226 }
227 }
228 break;
229 case IMODE_DIFF6:
230 case IMODE_NORM6:
231 if(!(height % 3) && (width % 3)) { // use 2x3 decoding
232 for(y = 0; y < height; y+= 3) {
233 for(x = width & 1; x < width; x += 2) {
234 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
235 if(code < 0){
236 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
237 return -1;
238 }
239 planep[x + 0] = (code >> 0) & 1;
240 planep[x + 1] = (code >> 1) & 1;
241 planep[x + 0 + stride] = (code >> 2) & 1;
242 planep[x + 1 + stride] = (code >> 3) & 1;
243 planep[x + 0 + stride * 2] = (code >> 4) & 1;
244 planep[x + 1 + stride * 2] = (code >> 5) & 1;
245 }
246 planep += stride * 3;
247 }
248 if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);
249 } else { // 3x2
250 planep += (height & 1) * stride;
251 for(y = height & 1; y < height; y += 2) {
252 for(x = width % 3; x < width; x += 3) {
253 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
254 if(code < 0){
255 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
256 return -1;
257 }
258 planep[x + 0] = (code >> 0) & 1;
259 planep[x + 1] = (code >> 1) & 1;
260 planep[x + 2] = (code >> 2) & 1;
261 planep[x + 0 + stride] = (code >> 3) & 1;
262 planep[x + 1 + stride] = (code >> 4) & 1;
263 planep[x + 2 + stride] = (code >> 5) & 1;
264 }
265 planep += stride * 2;
266 }
267 x = width % 3;
268 if(x) decode_colskip(data , x, height , stride, &v->s.gb);
269 if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);
270 }
271 break;
272 case IMODE_ROWSKIP:
273 decode_rowskip(data, width, height, stride, &v->s.gb);
274 break;
275 case IMODE_COLSKIP:
276 decode_colskip(data, width, height, stride, &v->s.gb);
277 break;
278 default: break;
279 }
280
281 /* Applying diff operator */
282 if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
283 {
284 planep = data;
285 planep[0] ^= invert;
286 for (x=1; x<width; x++)
287 planep[x] ^= planep[x-1];
288 for (y=1; y<height; y++)
289 {
290 planep += stride;
291 planep[0] ^= planep[-stride];
292 for (x=1; x<width; x++)
293 {
294 if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;
295 else planep[x] ^= planep[x-1];
296 }
297 }
298 }
299 else if (invert)
300 {
301 planep = data;
302 for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride
303 }
304 return (imode<<1) + invert;
305 }
306
307 /** @} */ //Bitplane group
308
309 /***********************************************************************/
310 /** VOP Dquant decoding
311 * @param v VC-1 Context
312 */
313 static int vop_dquant_decoding(VC1Context *v)
314 {
315 GetBitContext *gb = &v->s.gb;
316 int pqdiff;
317
318 //variable size
319 if (v->dquant == 2)
320 {
321 pqdiff = get_bits(gb, 3);
322 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
323 else v->altpq = v->pq + pqdiff + 1;
324 }
325 else
326 {
327 v->dquantfrm = get_bits1(gb);
328 if ( v->dquantfrm )
329 {
330 v->dqprofile = get_bits(gb, 2);
331 switch (v->dqprofile)
332 {
333 case DQPROFILE_SINGLE_EDGE:
334 case DQPROFILE_DOUBLE_EDGES:
335 v->dqsbedge = get_bits(gb, 2);
336 break;
337 case DQPROFILE_ALL_MBS:
338 v->dqbilevel = get_bits1(gb);
339 if(!v->dqbilevel)
340 v->halfpq = 0;
341 default: break; //Forbidden ?
342 }
343 if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)
344 {
345 pqdiff = get_bits(gb, 3);
346 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
347 else v->altpq = v->pq + pqdiff + 1;
348 }
349 }
350 }
351 return 0;
352 }
353
354 /** Put block onto picture
355 */
356 static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
357 {
358 uint8_t *Y;
359 int ys, us, vs;
360 DSPContext *dsp = &v->s.dsp;
361
362 if(v->rangeredfrm) {
363 int i, j, k;
364 for(k = 0; k < 6; k++)
365 for(j = 0; j < 8; j++)
366 for(i = 0; i < 8; i++)
367 block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;
368
369 }
370 ys = v->s.current_picture.linesize[0];
371 us = v->s.current_picture.linesize[1];
372 vs = v->s.current_picture.linesize[2];
373 Y = v->s.dest[0];
374
375 dsp->put_pixels_clamped(block[0], Y, ys);
376 dsp->put_pixels_clamped(block[1], Y + 8, ys);
377 Y += ys * 8;
378 dsp->put_pixels_clamped(block[2], Y, ys);
379 dsp->put_pixels_clamped(block[3], Y + 8, ys);
380
381 if(!(v->s.flags & CODEC_FLAG_GRAY)) {
382 dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
383 dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
384 }
385 }
386
387 /** Do motion compensation over 1 macroblock
388 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
389 */
390 static void vc1_mc_1mv(VC1Context *v, int dir)
391 {
392 MpegEncContext *s = &v->s;
393 DSPContext *dsp = &v->s.dsp;
394 uint8_t *srcY, *srcU, *srcV;
395 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
396
397 if(!v->s.last_picture.data[0])return;
398
399 mx = s->mv[dir][0][0];
400 my = s->mv[dir][0][1];
401
402 // store motion vectors for further use in B frames
403 if(s->pict_type == FF_P_TYPE) {
404 s->current_picture.motion_val[1][s->block_index[0]][0] = mx;
405 s->current_picture.motion_val[1][s->block_index[0]][1] = my;
406 }
407 uvmx = (mx + ((mx & 3) == 3)) >> 1;
408 uvmy = (my + ((my & 3) == 3)) >> 1;
409 if(v->fastuvmc) {
410 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
411 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
412 }
413 if(!dir) {
414 srcY = s->last_picture.data[0];
415 srcU = s->last_picture.data[1];
416 srcV = s->last_picture.data[2];
417 } else {
418 srcY = s->next_picture.data[0];
419 srcU = s->next_picture.data[1];
420 srcV = s->next_picture.data[2];
421 }
422
423 src_x = s->mb_x * 16 + (mx >> 2);
424 src_y = s->mb_y * 16 + (my >> 2);
425 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
426 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
427
428 if(v->profile != PROFILE_ADVANCED){
429 src_x = av_clip( src_x, -16, s->mb_width * 16);
430 src_y = av_clip( src_y, -16, s->mb_height * 16);
431 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
432 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
433 }else{
434 src_x = av_clip( src_x, -17, s->avctx->coded_width);
435 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
436 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
437 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
438 }
439
440 srcY += src_y * s->linesize + src_x;
441 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
442 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
443
444 /* for grayscale we should not try to read from unknown area */
445 if(s->flags & CODEC_FLAG_GRAY) {
446 srcU = s->edge_emu_buffer + 18 * s->linesize;
447 srcV = s->edge_emu_buffer + 18 * s->linesize;
448 }
449
450 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
451 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3
452 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){
453 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
454
455 srcY -= s->mspel * (1 + s->linesize);
456 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
457 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
458 srcY = s->edge_emu_buffer;
459 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
460 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
461 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
462 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
463 srcU = uvbuf;
464 srcV = uvbuf + 16;
465 /* if we deal with range reduction we need to scale source blocks */
466 if(v->rangeredfrm) {
467 int i, j;
468 uint8_t *src, *src2;
469
470 src = srcY;
471 for(j = 0; j < 17 + s->mspel*2; j++) {
472 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
473 src += s->linesize;
474 }
475 src = srcU; src2 = srcV;
476 for(j = 0; j < 9; j++) {
477 for(i = 0; i < 9; i++) {
478 src[i] = ((src[i] - 128) >> 1) + 128;
479 src2[i] = ((src2[i] - 128) >> 1) + 128;
480 }
481 src += s->uvlinesize;
482 src2 += s->uvlinesize;
483 }
484 }
485 /* if we deal with intensity compensation we need to scale source blocks */
486 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
487 int i, j;
488 uint8_t *src, *src2;
489
490 src = srcY;
491 for(j = 0; j < 17 + s->mspel*2; j++) {
492 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];
493 src += s->linesize;
494 }
495 src = srcU; src2 = srcV;
496 for(j = 0; j < 9; j++) {
497 for(i = 0; i < 9; i++) {
498 src[i] = v->lutuv[src[i]];
499 src2[i] = v->lutuv[src2[i]];
500 }
501 src += s->uvlinesize;
502 src2 += s->uvlinesize;
503 }
504 }
505 srcY += s->mspel * (1 + s->linesize);
506 }
507
508 if(s->mspel) {
509 dxy = ((my & 3) << 2) | (mx & 3);
510 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] , srcY , s->linesize, v->rnd);
511 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
512 srcY += s->linesize * 8;
513 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize , srcY , s->linesize, v->rnd);
514 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
515 } else { // hpel mc - always used for luma
516 dxy = (my & 2) | ((mx & 2) >> 1);
517
518 if(!v->rnd)
519 dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
520 else
521 dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
522 }
523
524 if(s->flags & CODEC_FLAG_GRAY) return;
525 /* Chroma MC always uses qpel bilinear */
526 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
527 uvmx = (uvmx&3)<<1;
528 uvmy = (uvmy&3)<<1;
529 if(!v->rnd){
530 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
531 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
532 }else{
533 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
534 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
535 }
536 }
537
538 /** Do motion compensation for 4-MV macroblock - luminance block
539 */
540 static void vc1_mc_4mv_luma(VC1Context *v, int n)
541 {
542 MpegEncContext *s = &v->s;
543 DSPContext *dsp = &v->s.dsp;
544 uint8_t *srcY;
545 int dxy, mx, my, src_x, src_y;
546 int off;
547
548 if(!v->s.last_picture.data[0])return;
549 mx = s->mv[0][n][0];
550 my = s->mv[0][n][1];
551 srcY = s->last_picture.data[0];
552
553 off = s->linesize * 4 * (n&2) + (n&1) * 8;
554
555 src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);
556 src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);
557
558 if(v->profile != PROFILE_ADVANCED){
559 src_x = av_clip( src_x, -16, s->mb_width * 16);
560 src_y = av_clip( src_y, -16, s->mb_height * 16);
561 }else{
562 src_x = av_clip( src_x, -17, s->avctx->coded_width);
563 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
564 }
565
566 srcY += src_y * s->linesize + src_x;
567
568 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
569 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2
570 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){
571 srcY -= s->mspel * (1 + s->linesize);
572 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,
573 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
574 srcY = s->edge_emu_buffer;
575 /* if we deal with range reduction we need to scale source blocks */
576 if(v->rangeredfrm) {
577 int i, j;
578 uint8_t *src;
579
580 src = srcY;
581 for(j = 0; j < 9 + s->mspel*2; j++) {
582 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
583 src += s->linesize;
584 }
585 }
586 /* if we deal with intensity compensation we need to scale source blocks */
587 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
588 int i, j;
589 uint8_t *src;
590
591 src = srcY;
592 for(j = 0; j < 9 + s->mspel*2; j++) {
593 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];
594 src += s->linesize;
595 }
596 }
597 srcY += s->mspel * (1 + s->linesize);
598 }
599
600 if(s->mspel) {
601 dxy = ((my & 3) << 2) | (mx & 3);
602 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);
603 } else { // hpel mc - always used for luma
604 dxy = (my & 2) | ((mx & 2) >> 1);
605 if(!v->rnd)
606 dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
607 else
608 dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
609 }
610 }
611
612 static inline int median4(int a, int b, int c, int d)
613 {
614 if(a < b) {
615 if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
616 else return (FFMIN(b, c) + FFMAX(a, d)) / 2;
617 } else {
618 if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
619 else return (FFMIN(a, c) + FFMAX(b, d)) / 2;
620 }
621 }
622
623
624 /** Do motion compensation for 4-MV macroblock - both chroma blocks
625 */
626 static void vc1_mc_4mv_chroma(VC1Context *v)
627 {
628 MpegEncContext *s = &v->s;
629 DSPContext *dsp = &v->s.dsp;
630 uint8_t *srcU, *srcV;
631 int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;
632 int i, idx, tx = 0, ty = 0;
633 int mvx[4], mvy[4], intra[4];
634 static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
635
636 if(!v->s.last_picture.data[0])return;
637 if(s->flags & CODEC_FLAG_GRAY) return;
638
639 for(i = 0; i < 4; i++) {
640 mvx[i] = s->mv[0][i][0];
641 mvy[i] = s->mv[0][i][1];
642 intra[i] = v->mb_type[0][s->block_index[i]];
643 }
644
645 /* calculate chroma MV vector from four luma MVs */
646 idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];
647 if(!idx) { // all blocks are inter
648 tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
649 ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
650 } else if(count[idx] == 1) { // 3 inter blocks
651 switch(idx) {
652 case 0x1:
653 tx = mid_pred(mvx[1], mvx[2], mvx[3]);
654 ty = mid_pred(mvy[1], mvy[2], mvy[3]);
655 break;
656 case 0x2:
657 tx = mid_pred(mvx[0], mvx[2], mvx[3]);
658 ty = mid_pred(mvy[0], mvy[2], mvy[3]);
659 break;
660 case 0x4:
661 tx = mid_pred(mvx[0], mvx[1], mvx[3]);
662 ty = mid_pred(mvy[0], mvy[1], mvy[3]);
663 break;
664 case 0x8:
665 tx = mid_pred(mvx[0], mvx[1], mvx[2]);
666 ty = mid_pred(mvy[0], mvy[1], mvy[2]);
667 break;
668 }
669 } else if(count[idx] == 2) {
670 int t1 = 0, t2 = 0;
671 for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}
672 for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}
673 tx = (mvx[t1] + mvx[t2]) / 2;
674 ty = (mvy[t1] + mvy[t2]) / 2;
675 } else {
676 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
677 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
678 return; //no need to do MC for inter blocks
679 }
680
681 s->current_picture.motion_val[1][s->block_index[0]][0] = tx;
682 s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
683 uvmx = (tx + ((tx&3) == 3)) >> 1;
684 uvmy = (ty + ((ty&3) == 3)) >> 1;
685 if(v->fastuvmc) {
686 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
687 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
688 }
689
690 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
691 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
692
693 if(v->profile != PROFILE_ADVANCED){
694 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
695 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
696 }else{
697 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
698 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
699 }
700
701 srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
702 srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
703 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
704 || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
705 || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){
706 ff_emulated_edge_mc(s->edge_emu_buffer , srcU, s->uvlinesize, 8+1, 8+1,
707 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
708 ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,
709 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
710 srcU = s->edge_emu_buffer;
711 srcV = s->edge_emu_buffer + 16;
712
713 /* if we deal with range reduction we need to scale source blocks */
714 if(v->rangeredfrm) {
715 int i, j;
716 uint8_t *src, *src2;
717
718 src = srcU; src2 = srcV;
719 for(j = 0; j < 9; j++) {
720 for(i = 0; i < 9; i++) {
721 src[i] = ((src[i] - 128) >> 1) + 128;
722 src2[i] = ((src2[i] - 128) >> 1) + 128;
723 }
724 src += s->uvlinesize;
725 src2 += s->uvlinesize;
726 }
727 }
728 /* if we deal with intensity compensation we need to scale source blocks */
729 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
730 int i, j;
731 uint8_t *src, *src2;
732
733 src = srcU; src2 = srcV;
734 for(j = 0; j < 9; j++) {
735 for(i = 0; i < 9; i++) {
736 src[i] = v->lutuv[src[i]];
737 src2[i] = v->lutuv[src2[i]];
738 }
739 src += s->uvlinesize;
740 src2 += s->uvlinesize;
741 }
742 }
743 }
744
745 /* Chroma MC always uses qpel bilinear */
746 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
747 uvmx = (uvmx&3)<<1;
748 uvmy = (uvmy&3)<<1;
749 if(!v->rnd){
750 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
751 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
752 }else{
753 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
754 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
755 }
756 }
757
758 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);
759
760 /**
761 * Decode Simple/Main Profiles sequence header
762 * @see Figure 7-8, p16-17
763 * @param avctx Codec context
764 * @param gb GetBit context initialized from Codec context extra_data
765 * @return Status
766 */
767 static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)
768 {
769 VC1Context *v = avctx->priv_data;
770
771 av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));
772 v->profile = get_bits(gb, 2);
773 if (v->profile == PROFILE_COMPLEX)
774 {
775 av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");
776 }
777
778 if (v->profile == PROFILE_ADVANCED)
779 {
780 v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz;
781 v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz;
782 return decode_sequence_header_adv(v, gb);
783 }
784 else
785 {
786 v->zz_8x4 = ff_vc1_simple_progressive_8x4_zz;
787 v->zz_4x8 = ff_vc1_simple_progressive_4x8_zz;
788 v->res_sm = get_bits(gb, 2); //reserved
789 if (v->res_sm)
790 {
791 av_log(avctx, AV_LOG_ERROR,
792 "Reserved RES_SM=%i is forbidden\n", v->res_sm);
793 return -1;
794 }
795 }
796
797 // (fps-2)/4 (->30)
798 v->frmrtq_postproc = get_bits(gb, 3); //common
799 // (bitrate-32kbps)/64kbps
800 v->bitrtq_postproc = get_bits(gb, 5); //common
801 v->s.loop_filter = get_bits1(gb); //common
802 if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)
803 {
804 av_log(avctx, AV_LOG_ERROR,
805 "LOOPFILTER shell not be enabled in simple profile\n");
806 }
807
808 v->res_x8 = get_bits1(gb); //reserved
809 v->multires = get_bits1(gb);
810 v->res_fasttx = get_bits1(gb);
811 if (!v->res_fasttx)
812 {
813 v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct;
814 v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;
815 v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;
816 v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;
817 }
818
819 v->fastuvmc = get_bits1(gb); //common
820 if (!v->profile && !v->fastuvmc)
821 {
822 av_log(avctx, AV_LOG_ERROR,
823 "FASTUVMC unavailable in Simple Profile\n");
824 return -1;
825 }
826 v->extended_mv = get_bits1(gb); //common
827 if (!v->profile && v->extended_mv)
828 {
829 av_log(avctx, AV_LOG_ERROR,
830 "Extended MVs unavailable in Simple Profile\n");
831 return -1;
832 }
833 v->dquant = get_bits(gb, 2); //common
834 v->vstransform = get_bits1(gb); //common
835
836 v->res_transtab = get_bits1(gb);
837 if (v->res_transtab)
838 {
839 av_log(avctx, AV_LOG_ERROR,
840 "1 for reserved RES_TRANSTAB is forbidden\n");
841 return -1;
842 }
843
844 v->overlap = get_bits1(gb); //common
845
846 v->s.resync_marker = get_bits1(gb);
847 v->rangered = get_bits1(gb);
848 if (v->rangered && v->profile == PROFILE_SIMPLE)
849 {
850 av_log(avctx, AV_LOG_INFO,
851 "RANGERED should be set to 0 in simple profile\n");
852 }
853
854 v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common
855 v->quantizer_mode = get_bits(gb, 2); //common
856
857 v->finterpflag = get_bits1(gb); //common
858 v->res_rtm_flag = get_bits1(gb); //reserved
859 if (!v->res_rtm_flag)
860 {
861 // av_log(avctx, AV_LOG_ERROR,
862 // "0 for reserved RES_RTM_FLAG is forbidden\n");
863 av_log(avctx, AV_LOG_ERROR,
864 "Old WMV3 version detected, only I-frames will be decoded\n");
865 //return -1;
866 }
867 //TODO: figure out what they mean (always 0x402F)
868 if(!v->res_fasttx) skip_bits(gb, 16);
869 av_log(avctx, AV_LOG_DEBUG,
870 "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
871 "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"
872 "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"
873 "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",
874 v->profile, v->frmrtq_postproc, v->bitrtq_postproc,
875 v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,
876 v->rangered, v->vstransform, v->overlap, v->s.resync_marker,
877 v->dquant, v->quantizer_mode, avctx->max_b_frames
878 );
879 return 0;
880 }
881
882 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)
883 {
884 v->res_rtm_flag = 1;
885 v->level = get_bits(gb, 3);
886 if(v->level >= 5)
887 {
888 av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);
889 }
890 v->chromaformat = get_bits(gb, 2);
891 if (v->chromaformat != 1)
892 {
893 av_log(v->s.avctx, AV_LOG_ERROR,
894 "Only 4:2:0 chroma format supported\n");
895 return -1;
896 }
897
898 // (fps-2)/4 (->30)
899 v->frmrtq_postproc = get_bits(gb, 3); //common
900 // (bitrate-32kbps)/64kbps
901 v->bitrtq_postproc = get_bits(gb, 5); //common
902 v->postprocflag = get_bits1(gb); //common
903
904 v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;
905 v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;
906 v->s.avctx->width = v->s.avctx->coded_width;
907 v->s.avctx->height = v->s.avctx->coded_height;
908 v->broadcast = get_bits1(gb);
909 v->interlace = get_bits1(gb);
910 v->tfcntrflag = get_bits1(gb);
911 v->finterpflag = get_bits1(gb);
912 skip_bits1(gb); // reserved
913
914 v->s.h_edge_pos = v->s.avctx->coded_width;
915 v->s.v_edge_pos = v->s.avctx->coded_height;
916
917 av_log(v->s.avctx, AV_LOG_DEBUG,
918 "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
919 "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"
920 "TFCTRflag=%i, FINTERPflag=%i\n",
921 v->level, v->frmrtq_postproc, v->bitrtq_postproc,
922 v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,
923 v->tfcntrflag, v->finterpflag
924 );
925
926 v->psf = get_bits1(gb);
927 if(v->psf) { //PsF, 6.1.13
928 av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");
929 return -1;
930 }
931 v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
932 if(get_bits1(gb)) { //Display Info - decoding is not affected by it
933 int w, h, ar = 0;
934 av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
935 v->s.avctx->width = v->s.width = w = get_bits(gb, 14) + 1;
936 v->s.avctx->height = v->s.height = h = get_bits(gb, 14) + 1;
937 av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
938 if(get_bits1(gb))
939 ar = get_bits(gb, 4);
940 if(ar && ar < 14){
941 v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
942 }else if(ar == 15){
943 w = get_bits(gb, 8);
944 h = get_bits(gb, 8);
945 v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
946 }
947
948 if(get_bits1(gb)){ //framerate stuff
949 if(get_bits1(gb)) {
950 v->s.avctx->time_base.num = 32;
951 v->s.avctx->time_base.den = get_bits(gb, 16) + 1;
952 } else {
953 int nr, dr;
954 nr = get_bits(gb, 8);
955 dr = get_bits(gb, 4);
956 if(nr && nr < 8 && dr && dr < 3){
957 v->s.avctx->time_base.num = ff_vc1_fps_dr[dr - 1];
958 v->s.avctx->time_base.den = ff_vc1_fps_nr[nr - 1] * 1000;
959 }
960 }
961 }
962
963 if(get_bits1(gb)){
964 v->color_prim = get_bits(gb, 8);
965 v->transfer_char = get_bits(gb, 8);
966 v->matrix_coef = get_bits(gb, 8);
967 }
968 }
969
970 v->hrd_param_flag = get_bits1(gb);
971 if(v->hrd_param_flag) {
972 int i;
973 v->hrd_num_leaky_buckets = get_bits(gb, 5);
974 skip_bits(gb, 4); //bitrate exponent
975 skip_bits(gb, 4); //buffer size exponent
976 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
977 skip_bits(gb, 16); //hrd_rate[n]
978 skip_bits(gb, 16); //hrd_buffer[n]
979 }
980 }
981 return 0;
982 }
983
984 static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb)
985 {
986 VC1Context *v = avctx->priv_data;
987 int i, blink, clentry, refdist;
988
989 av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32));
990 blink = get_bits1(gb); // broken link
991 clentry = get_bits1(gb); // closed entry
992 v->panscanflag = get_bits1(gb);
993 refdist = get_bits1(gb); // refdist flag
994 v->s.loop_filter = get_bits1(gb);
995 v->fastuvmc = get_bits1(gb);
996 v->extended_mv = get_bits1(gb);
997 v->dquant = get_bits(gb, 2);
998 v->vstransform = get_bits1(gb);
999 v->overlap = get_bits1(gb);
1000 v->quantizer_mode = get_bits(gb, 2);
1001
1002 if(v->hrd_param_flag){
1003 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1004 skip_bits(gb, 8); //hrd_full[n]
1005 }
1006 }
1007
1008 if(get_bits1(gb)){
1009 avctx->coded_width = (get_bits(gb, 12)+1)<<1;
1010 avctx->coded_height = (get_bits(gb, 12)+1)<<1;
1011 }
1012 if(v->extended_mv)
1013 v->extended_dmv = get_bits1(gb);
1014 if(get_bits1(gb)) {
1015 av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n");
1016 skip_bits(gb, 3); // Y range, ignored for now
1017 }
1018 if(get_bits1(gb)) {
1019 av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n");
1020 skip_bits(gb, 3); // UV range, ignored for now
1021 }
1022
1023 av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n"
1024 "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n"
1025 "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n"
1026 "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n",
1027 blink, clentry, v->panscanflag, refdist, v->s.loop_filter,
1028 v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode);
1029
1030 return 0;
1031 }
1032
1033 static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)
1034 {
1035 int pqindex, lowquant, status;
1036
1037 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1038 skip_bits(gb, 2); //framecnt unused
1039 v->rangeredfrm = 0;
1040 if (v->rangered) v->rangeredfrm = get_bits1(gb);
1041 v->s.pict_type = get_bits1(gb);
1042 if (v->s.avctx->max_b_frames) {
1043 if (!v->s.pict_type) {
1044 if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE;
1045 else v->s.pict_type = FF_B_TYPE;
1046 } else v->s.pict_type = FF_P_TYPE;
1047 } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE;
1048
1049 v->bi_type = 0;
1050 if(v->s.pict_type == FF_B_TYPE) {
1051 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1052 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1053 if(v->bfraction == 0) {
1054 v->s.pict_type = FF_BI_TYPE;
1055 }
1056 }
1057 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1058 skip_bits(gb, 7); // skip buffer fullness
1059
1060 /* calculate RND */
1061 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1062 v->rnd = 1;
1063 if(v->s.pict_type == FF_P_TYPE)
1064 v->rnd ^= 1;
1065
1066 /* Quantizer stuff */
1067 pqindex = get_bits(gb, 5);
1068 if(!pqindex) return -1;
1069 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1070 v->pq = ff_vc1_pquant_table[0][pqindex];
1071 else
1072 v->pq = ff_vc1_pquant_table[1][pqindex];
1073
1074 v->pquantizer = 1;
1075 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1076 v->pquantizer = pqindex < 9;
1077 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1078 v->pquantizer = 0;
1079 v->pqindex = pqindex;
1080 if (pqindex < 9) v->halfpq = get_bits1(gb);
1081 else v->halfpq = 0;
1082 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1083 v->pquantizer = get_bits1(gb);
1084 v->dquantfrm = 0;
1085 if (v->extended_mv == 1) v->mvrange = get_unary(gb, 0, 3);
1086 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1087 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1088 v->range_x = 1 << (v->k_x - 1);
1089 v->range_y = 1 << (v->k_y - 1);
1090 if (v->profile == PROFILE_ADVANCED)
1091 {
1092 if (v->postprocflag) v->postproc = get_bits1(gb);
1093 }
1094 else
1095 if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2);
1096
1097 if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){
1098 v->x8_type = get_bits1(gb);
1099 }else v->x8_type = 0;
1100 //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",
1101 // (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);
1102
1103 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1104
1105 switch(v->s.pict_type) {
1106 case FF_P_TYPE:
1107 if (v->pq < 5) v->tt_index = 0;
1108 else if(v->pq < 13) v->tt_index = 1;
1109 else v->tt_index = 2;
1110
1111 lowquant = (v->pq > 12) ? 0 : 1;
1112 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1113 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1114 {
1115 int scale, shift, i;
1116 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1117 v->lumscale = get_bits(gb, 6);
1118 v->lumshift = get_bits(gb, 6);
1119 v->use_ic = 1;
1120 /* fill lookup tables for intensity compensation */
1121 if(!v->lumscale) {
1122 scale = -64;
1123 shift = (255 - v->lumshift * 2) << 6;
1124 if(v->lumshift > 31)
1125 shift += 128 << 6;
1126 } else {
1127 scale = v->lumscale + 32;
1128 if(v->lumshift > 31)
1129 shift = (v->lumshift - 64) << 6;
1130 else
1131 shift = v->lumshift << 6;
1132 }
1133 for(i = 0; i < 256; i++) {
1134 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1135 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1136 }
1137 }
1138 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1139 v->s.quarter_sample = 0;
1140 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1141 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1142 v->s.quarter_sample = 0;
1143 else
1144 v->s.quarter_sample = 1;
1145 } else
1146 v->s.quarter_sample = 1;
1147 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));
1148
1149 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1150 v->mv_mode2 == MV_PMODE_MIXED_MV)
1151 || v->mv_mode == MV_PMODE_MIXED_MV)
1152 {
1153 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1154 if (status < 0) return -1;
1155 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1156 "Imode: %i, Invert: %i\n", status>>1, status&1);
1157 } else {
1158 v->mv_type_is_raw = 0;
1159 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1160 }
1161 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1162 if (status < 0) return -1;
1163 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1164 "Imode: %i, Invert: %i\n", status>>1, status&1);
1165
1166 /* Hopefully this is correct for P frames */
1167 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1168 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1169
1170 if (v->dquant)
1171 {
1172 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1173 vop_dquant_decoding(v);
1174 }
1175
1176 v->ttfrm = 0; //FIXME Is that so ?
1177 if (v->vstransform)
1178 {
1179 v->ttmbf = get_bits1(gb);
1180 if (v->ttmbf)
1181 {
1182 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1183 }
1184 } else {
1185 v->ttmbf = 1;
1186 v->ttfrm = TT_8X8;
1187 }
1188 break;
1189 case FF_B_TYPE:
1190 if (v->pq < 5) v->tt_index = 0;
1191 else if(v->pq < 13) v->tt_index = 1;
1192 else v->tt_index = 2;
1193
1194 lowquant = (v->pq > 12) ? 0 : 1;
1195 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1196 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1197 v->s.mspel = v->s.quarter_sample;
1198
1199 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1200 if (status < 0) return -1;
1201 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1202 "Imode: %i, Invert: %i\n", status>>1, status&1);
1203 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1204 if (status < 0) return -1;
1205 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1206 "Imode: %i, Invert: %i\n", status>>1, status&1);
1207
1208 v->s.mv_table_index = get_bits(gb, 2);
1209 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1210
1211 if (v->dquant)
1212 {
1213 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1214 vop_dquant_decoding(v);
1215 }
1216
1217 v->ttfrm = 0;
1218 if (v->vstransform)
1219 {
1220 v->ttmbf = get_bits1(gb);
1221 if (v->ttmbf)
1222 {
1223 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1224 }
1225 } else {
1226 v->ttmbf = 1;
1227 v->ttfrm = TT_8X8;
1228 }
1229 break;
1230 }
1231
1232 if(!v->x8_type)
1233 {
1234 /* AC Syntax */
1235 v->c_ac_table_index = decode012(gb);
1236 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1237 {
1238 v->y_ac_table_index = decode012(gb);
1239 }
1240 /* DC Syntax */
1241 v->s.dc_table_index = get_bits1(gb);
1242 }
1243
1244 if(v->s.pict_type == FF_BI_TYPE) {
1245 v->s.pict_type = FF_B_TYPE;
1246 v->bi_type = 1;
1247 }
1248 return 0;
1249 }
1250
1251 static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)
1252 {
1253 int pqindex, lowquant;
1254 int status;
1255
1256 v->p_frame_skipped = 0;
1257
1258 if(v->interlace){
1259 v->fcm = decode012(gb);
1260 if(v->fcm) return -1; // interlaced frames/fields are not implemented
1261 }
1262 switch(get_unary(gb, 0, 4)) {
1263 case 0:
1264 v->s.pict_type = FF_P_TYPE;
1265 break;
1266 case 1:
1267 v->s.pict_type = FF_B_TYPE;
1268 break;
1269 case 2:
1270 v->s.pict_type = FF_I_TYPE;
1271 break;
1272 case 3:
1273 v->s.pict_type = FF_BI_TYPE;
1274 break;
1275 case 4:
1276 v->s.pict_type = FF_P_TYPE; // skipped pic
1277 v->p_frame_skipped = 1;
1278 return 0;
1279 }
1280 if(v->tfcntrflag)
1281 skip_bits(gb, 8);
1282 if(v->broadcast) {
1283 if(!v->interlace || v->psf) {
1284 v->rptfrm = get_bits(gb, 2);
1285 } else {
1286 v->tff = get_bits1(gb);
1287 v->rptfrm = get_bits1(gb);
1288 }
1289 }
1290 if(v->panscanflag) {
1291 //...
1292 }
1293 v->rnd = get_bits1(gb);
1294 if(v->interlace)
1295 v->uvsamp = get_bits1(gb);
1296 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1297 if(v->s.pict_type == FF_B_TYPE) {
1298 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1299 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1300 if(v->bfraction == 0) {
1301 v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */
1302 }
1303 }
1304 pqindex = get_bits(gb, 5);
1305 if(!pqindex) return -1;
1306 v->pqindex = pqindex;
1307 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1308 v->pq = ff_vc1_pquant_table[0][pqindex];
1309 else
1310 v->pq = ff_vc1_pquant_table[1][pqindex];
1311
1312 v->pquantizer = 1;
1313 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1314 v->pquantizer = pqindex < 9;
1315 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1316 v->pquantizer = 0;
1317 v->pqindex = pqindex;
1318 if (pqindex < 9) v->halfpq = get_bits1(gb);
1319 else v->halfpq = 0;
1320 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1321 v->pquantizer = get_bits1(gb);
1322
1323 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1324
1325 switch(v->s.pict_type) {
1326 case FF_I_TYPE:
1327 case FF_BI_TYPE:
1328 status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v);
1329 if (status < 0) return -1;
1330 av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: "
1331 "Imode: %i, Invert: %i\n", status>>1, status&1);
1332 v->condover = CONDOVER_NONE;
1333 if(v->overlap && v->pq <= 8) {
1334 v->condover = decode012(gb);
1335 if(v->condover == CONDOVER_SELECT) {
1336 status = bitplane_decoding(v->over_flags_plane, &v->overflg_is_raw, v);
1337 if (status < 0) return -1;
1338 av_log(v->s.avctx, AV_LOG_DEBUG, "CONDOVER plane encoding: "
1339 "Imode: %i, Invert: %i\n", status>>1, status&1);
1340 }
1341 }
1342 break;
1343 case FF_P_TYPE:
1344 if(v->postprocflag)
1345 v->postproc = get_bits1(gb);
1346 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1347 else v->mvrange = 0;
1348 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1349 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1350 v->range_x = 1 << (v->k_x - 1);
1351 v->range_y = 1 << (v->k_y - 1);
1352
1353 if (v->pq < 5) v->tt_index = 0;
1354 else if(v->pq < 13) v->tt_index = 1;
1355 else v->tt_index = 2;
1356
1357 lowquant = (v->pq > 12) ? 0 : 1;
1358 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1359 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1360 {
1361 int scale, shift, i;
1362 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1363 v->lumscale = get_bits(gb, 6);
1364 v->lumshift = get_bits(gb, 6);
1365 /* fill lookup tables for intensity compensation */
1366 if(!v->lumscale) {
1367 scale = -64;
1368 shift = (255 - v->lumshift * 2) << 6;
1369 if(v->lumshift > 31)
1370 shift += 128 << 6;
1371 } else {
1372 scale = v->lumscale + 32;
1373 if(v->lumshift > 31)
1374 shift = (v->lumshift - 64) << 6;
1375 else
1376 shift = v->lumshift << 6;
1377 }
1378 for(i = 0; i < 256; i++) {
1379 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1380 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1381 }
1382 v->use_ic = 1;
1383 }
1384 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1385 v->s.quarter_sample = 0;
1386 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1387 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1388 v->s.quarter_sample = 0;
1389 else
1390 v->s.quarter_sample = 1;
1391 } else
1392 v->s.quarter_sample = 1;
1393 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));
1394
1395 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1396 v->mv_mode2 == MV_PMODE_MIXED_MV)
1397 || v->mv_mode == MV_PMODE_MIXED_MV)
1398 {
1399 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1400 if (status < 0) return -1;
1401 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1402 "Imode: %i, Invert: %i\n", status>>1, status&1);
1403 } else {
1404 v->mv_type_is_raw = 0;
1405 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1406 }
1407 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1408 if (status < 0) return -1;
1409 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1410 "Imode: %i, Invert: %i\n", status>>1, status&1);
1411
1412 /* Hopefully this is correct for P frames */
1413 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1414 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1415 if (v->dquant)
1416 {
1417 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1418 vop_dquant_decoding(v);
1419 }
1420
1421 v->ttfrm = 0; //FIXME Is that so ?
1422 if (v->vstransform)
1423 {
1424 v->ttmbf = get_bits1(gb);
1425 if (v->ttmbf)
1426 {
1427 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1428 }
1429 } else {
1430 v->ttmbf = 1;
1431 v->ttfrm = TT_8X8;
1432 }
1433 break;
1434 case FF_B_TYPE:
1435 if(v->postprocflag)
1436 v->postproc = get_bits1(gb);
1437 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1438 else v->mvrange = 0;
1439 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1440 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1441 v->range_x = 1 << (v->k_x - 1);
1442 v->range_y = 1 << (v->k_y - 1);
1443
1444 if (v->pq < 5) v->tt_index = 0;
1445 else if(v->pq < 13) v->tt_index = 1;
1446 else v->tt_index = 2;
1447
1448 lowquant = (v->pq > 12) ? 0 : 1;
1449 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1450 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1451 v->s.mspel = v->s.quarter_sample;
1452
1453 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1454 if (status < 0) return -1;
1455 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1456 "Imode: %i, Invert: %i\n", status>>1, status&1);
1457 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1458 if (status < 0) return -1;
1459 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1460 "Imode: %i, Invert: %i\n", status>>1, status&1);
1461
1462 v->s.mv_table_index = get_bits(gb, 2);
1463 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1464
1465 if (v->dquant)
1466 {
1467 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1468 vop_dquant_decoding(v);
1469 }
1470
1471 v->ttfrm = 0;
1472 if (v->vstransform)
1473 {
1474 v->ttmbf = get_bits1(gb);
1475 if (v->ttmbf)
1476 {
1477 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1478 }
1479 } else {
1480 v->ttmbf = 1;
1481 v->ttfrm = TT_8X8;
1482 }
1483 break;
1484 }
1485
1486 /* AC Syntax */
1487 v->c_ac_table_index = decode012(gb);
1488 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1489 {
1490 v->y_ac_table_index = decode012(gb);
1491 }
1492 /* DC Syntax */
1493 v->s.dc_table_index = get_bits1(gb);
1494 if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) {
1495 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1496 vop_dquant_decoding(v);
1497 }
1498
1499 v->bi_type = 0;
1500 if(v->s.pict_type == FF_BI_TYPE) {
1501 v->s.pict_type = FF_B_TYPE;
1502 v->bi_type = 1;
1503 }
1504 return 0;
1505 }
1506
1507 /***********************************************************************/
1508 /**
1509 * @defgroup block VC-1 Block-level functions
1510 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
1511 * @{
1512 */
1513
1514 /**
1515 * @def GET_MQUANT
1516 * @brief Get macroblock-level quantizer scale
1517 */
1518 #define GET_MQUANT() \
1519 if (v->dquantfrm) \
1520 { \
1521 int edges = 0; \
1522 if (v->dqprofile == DQPROFILE_ALL_MBS) \
1523 { \
1524 if (v->dqbilevel) \
1525 { \
1526 mquant = (get_bits1(gb)) ? v->altpq : v->pq; \
1527 } \
1528 else \
1529 { \
1530 mqdiff = get_bits(gb, 3); \
1531 if (mqdiff != 7) mquant = v->pq + mqdiff; \
1532 else mquant = get_bits(gb, 5); \
1533 } \
1534 } \
1535 if(v->dqprofile == DQPROFILE_SINGLE_EDGE) \
1536 edges = 1 << v->dqsbedge; \
1537 else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES) \
1538 edges = (3 << v->dqsbedge) % 15; \
1539 else if(v->dqprofile == DQPROFILE_FOUR_EDGES) \
1540 edges = 15; \
1541 if((edges&1) && !s->mb_x) \
1542 mquant = v->altpq; \
1543 if((edges&2) && s->first_slice_line) \
1544 mquant = v->altpq; \
1545 if((edges&4) && s->mb_x == (s->mb_width - 1)) \
1546 mquant = v->altpq; \
1547 if((edges&8) && s->mb_y == (s->mb_height - 1)) \
1548 mquant = v->altpq; \
1549 }
1550
1551 /**
1552 * @def GET_MVDATA(_dmv_x, _dmv_y)
1553 * @brief Get MV differentials
1554 * @see MVDATA decoding from 8.3.5.2, p(1)20
1555 * @param _dmv_x Horizontal differential for decoded MV
1556 * @param _dmv_y Vertical differential for decoded MV
1557 */
1558 #define GET_MVDATA(_dmv_x, _dmv_y) \
1559 index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table,\
1560 VC1_MV_DIFF_VLC_BITS, 2); \
1561 if (index > 36) \
1562 { \
1563 mb_has_coeffs = 1; \
1564 index -= 37; \
1565 } \
1566 else mb_has_coeffs = 0; \
1567 s->mb_intra = 0; \
1568 if (!index) { _dmv_x = _dmv_y = 0; } \
1569 else if (index == 35) \
1570 { \
1571 _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample); \
1572 _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample); \
1573 } \
1574 else if (index == 36) \
1575 { \
1576 _dmv_x = 0; \
1577 _dmv_y = 0; \
1578 s->mb_intra = 1; \
1579 } \
1580 else \
1581 { \
1582 index1 = index%6; \
1583 if (!s->quarter_sample && index1 == 5) val = 1; \
1584 else val = 0; \
1585 if(size_table[index1] - val > 0) \
1586 val = get_bits(gb, size_table[index1] - val); \
1587 else val = 0; \
1588 sign = 0 - (val&1); \
1589 _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1590 \
1591 index1 = index/6; \
1592 if (!s->quarter_sample && index1 == 5) val = 1; \
1593 else val = 0; \
1594 if(size_table[index1] - val > 0) \
1595 val = get_bits(gb, size_table[index1] - val); \
1596 else val = 0; \
1597 sign = 0 - (val&1); \
1598 _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1599 }
1600
1601 /** Predict and set motion vector
1602 */
1603 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)
1604 {
1605 int xy, wrap, off = 0;
1606 int16_t *A, *B, *C;
1607 int px, py;
1608 int sum;
1609
1610 /* scale MV difference to be quad-pel */
1611 dmv_x <<= 1 - s->quarter_sample;
1612 dmv_y <<= 1 - s->quarter_sample;
1613
1614 wrap = s->b8_stride;
1615 xy = s->block_index[n];
1616
1617 if(s->mb_intra){
1618 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
1619 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
1620 s->current_picture.motion_val[1][xy][0] = 0;
1621 s->current_picture.motion_val[1][xy][1] = 0;
1622 if(mv1) { /* duplicate motion data for 1-MV block */
1623 s->current_picture.motion_val[0][xy + 1][0] = 0;
1624 s->current_picture.motion_val[0][xy + 1][1] = 0;
1625 s->current_picture.motion_val[0][xy + wrap][0] = 0;
1626 s->current_picture.motion_val[0][xy + wrap][1] = 0;
1627 s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
1628 s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
1629 s->current_picture.motion_val[1][xy + 1][0] = 0;
1630 s->current_picture.motion_val[1][xy + 1][1] = 0;
1631 s->current_picture.motion_val[1][xy + wrap][0] = 0;
1632 s->current_picture.motion_val[1][xy + wrap][1] = 0;
1633 s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
1634 s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
1635 }
1636 return;
1637 }
1638
1639 C = s->current_picture.motion_val[0][xy - 1];
1640 A = s->current_picture.motion_val[0][xy - wrap];
1641 if(mv1)
1642 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
1643 else {
1644 //in 4-MV mode different blocks have different B predictor position
1645 switch(n){
1646 case 0:
1647 off = (s->mb_x > 0) ? -1 : 1;
1648 break;
1649 case 1:
1650 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
1651 break;
1652 case 2:
1653 off = 1;
1654 break;
1655 case 3:
1656 off = -1;
1657 }
1658 }
1659 B = s->current_picture.motion_val[0][xy - wrap + off];
1660
1661 if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds
1662 if(s->mb_width == 1) {
1663 px = A[0];
1664 py = A[1];
1665 } else {
1666 px = mid_pred(A[0], B[0], C[0]);
1667 py = mid_pred(A[1], B[1], C[1]);
1668 }
1669 } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds
1670 px = C[0];
1671 py = C[1];
1672 } else {
1673 px = py = 0;
1674 }
1675 /* Pullback MV as specified in 8.3.5.3.4 */
1676 {
1677 int qx, qy, X, Y;
1678 qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);
1679 qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);
1680 X = (s->mb_width << 6) - 4;
1681 Y = (s->mb_height << 6) - 4;
1682 if(mv1) {
1683 if(qx + px < -60) px = -60 - qx;
1684 if(qy + py < -60) py = -60 - qy;
1685 } else {
1686 if(qx + px < -28) px = -28 - qx;
1687 if(qy + py < -28) py = -28 - qy;
1688 }
1689 if(qx + px > X) px = X - qx;
1690 if(qy + py > Y) py = Y - qy;
1691 }
1692 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
1693 if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {
1694 if(is_intra[xy - wrap])
1695 sum = FFABS(px) + FFABS(py);
1696 else
1697 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
1698 if(sum > 32) {
1699 if(get_bits1(&s->gb)) {
1700 px = A[0];
1701 py = A[1];
1702 } else {
1703 px = C[0];
1704 py = C[1];
1705 }
1706 } else {
1707 if(is_intra[xy - 1])
1708 sum = FFABS(px) + FFABS(py);
1709 else
1710 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
1711 if(sum > 32) {
1712 if(get_bits1(&s->gb)) {
1713 px = A[0];
1714 py = A[1];
1715 } else {
1716 px = C[0];
1717 py = C[1];
1718 }
1719 }
1720 }
1721 }
1722 /* store MV using signed modulus of MV range defined in 4.11 */
1723 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
1724 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;
1725 if(mv1) { /* duplicate motion data for 1-MV block */
1726 s->current_picture.motion_val[0][xy + 1][0] = s->current_picture.motion_val[0][xy][0];
1727 s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];
1728 s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];
1729 s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];
1730 s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];
1731 s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];
1732 }
1733 }
1734
1735 /** Motion compensation for direct or interpolated blocks in B-frames
1736 */
1737 static void vc1_interp_mc(VC1Context *v)
1738 {
1739 MpegEncContext *s = &v->s;
1740 DSPContext *dsp = &v->s.dsp;
1741 uint8_t *srcY, *srcU, *srcV;
1742 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
1743
1744 if(!v->s.next_picture.data[0])return;
1745
1746 mx = s->mv[1][0][0];
1747 my = s->mv[1][0][1];
1748 uvmx = (mx + ((mx & 3) == 3)) >> 1;
1749 uvmy = (my + ((my & 3) == 3)) >> 1;
1750 if(v->fastuvmc) {
1751 uvmx = uvmx + ((uvmx<0)?-(uvmx&1):(uvmx&1));
1752 uvmy = uvmy + ((uvmy<0)?-(uvmy&1):(uvmy&1));
1753 }
1754 srcY = s->next_picture.data[0];
1755 srcU = s->next_picture.data[1];
1756 srcV = s->next_picture.data[2];
1757
1758 src_x = s->mb_x * 16 + (mx >> 2);
1759 src_y = s->mb_y * 16 + (my >> 2);
1760 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
1761 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
1762
1763 if(v->profile != PROFILE_ADVANCED){
1764 src_x = av_clip( src_x, -16, s->mb_width * 16);
1765 src_y = av_clip( src_y, -16, s->mb_height * 16);
1766 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
1767 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
1768 }else{
1769 src_x = av_clip( src_x, -17, s->avctx->coded_width);
1770 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
1771 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
1772 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
1773 }
1774
1775 srcY += src_y * s->linesize + src_x;
1776 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
1777 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
1778
1779 /* for grayscale we should not try to read from unknown area */
1780 if(s->flags & CODEC_FLAG_GRAY) {
1781 srcU = s->edge_emu_buffer + 18 * s->linesize;
1782 srcV = s->edge_emu_buffer + 18 * s->linesize;
1783 }
1784
1785 if(v->rangeredfrm
1786 || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16
1787 || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){
1788 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
1789
1790 srcY -= s->mspel * (1 + s->linesize);
1791 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
1792 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
1793 srcY = s->edge_emu_buffer;
1794 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
1795 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1796 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
1797 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1798 srcU = uvbuf;
1799 srcV = uvbuf + 16;
1800 /* if we deal with range reduction we need to scale source blocks */
1801 if(v->rangeredfrm) {
1802 int i, j;
1803 uint8_t *src, *src2;
1804
1805 src = srcY;
1806 for(j = 0; j < 17 + s->mspel*2; j++) {
1807 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
1808 src += s->linesize;
1809 }
1810 src = srcU; src2 = srcV;
1811 for(j = 0; j < 9; j++) {
1812 for(i = 0; i < 9; i++) {
1813 src[i] = ((src[i] - 128) >> 1) + 128;
1814 src2[i] = ((src2[i] - 128) >> 1) + 128;
1815 }
1816 src += s->uvlinesize;
1817 src2 += s->uvlinesize;
1818 }
1819 }
1820 srcY += s->mspel * (1 + s->linesize);
1821 }
1822
1823 mx >>= 1;
1824 my >>= 1;
1825 dxy = ((my & 1) << 1) | (mx & 1);
1826
1827 dsp->avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
1828
1829 if(s->flags & CODEC_FLAG_GRAY) return;
1830 /* Chroma MC always uses qpel blilinear */
1831 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
1832 uvmx = (uvmx&3)<<1;
1833 uvmy = (uvmy&3)<<1;
1834 dsp->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
1835 dsp->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
1836 }
1837
1838 static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)
1839 {
1840 int n = bfrac;
1841
1842 #if B_FRACTION_DEN==256
1843 if(inv)
1844 n -= 256;
1845 if(!qs)
1846 return 2 * ((value * n + 255) >> 9);
1847 return (value * n + 128) >> 8;
1848 #else
1849 if(inv)
1850 n -= B_FRACTION_DEN;
1851 if(!qs)
1852 return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));
1853 return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;
1854 #endif
1855 }
1856
1857 /** Reconstruct motion vector for B-frame and do motion compensation
1858 */
1859 static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode)
1860 {
1861 if(v->use_ic) {
1862 v->mv_mode2 = v->mv_mode;
1863 v->mv_mode = MV_PMODE_INTENSITY_COMP;
1864 }
1865 if(direct) {
1866 vc1_mc_1mv(v, 0);
1867 vc1_interp_mc(v);
1868 if(v->use_ic) v->mv_mode = v->mv_mode2;
1869 return;
1870 }
1871 if(mode == BMV_TYPE_INTERPOLATED) {
1872 vc1_mc_1mv(v, 0);
1873 vc1_interp_mc(v);
1874 if(v->use_ic) v->mv_mode = v->mv_mode2;
1875 return;
1876 }
1877
1878 if(v->use_ic && (mode == BMV_TYPE_BACKWARD)) v->mv_mode = v->mv_mode2;
1879 vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));
1880 if(v->use_ic) v->mv_mode = v->mv_mode2;
1881 }
1882
1883 static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mvtype)
1884 {
1885 MpegEncContext *s = &v->s;
1886 int xy, wrap, off = 0;
1887 int16_t *A, *B, *C;
1888 int px, py;
1889 int sum;
1890 int r_x, r_y;
1891 const uint8_t *is_intra = v->mb_type[0];
1892
1893 r_x = v->range_x;
1894 r_y = v->range_y;
1895 /* scale MV difference to be quad-pel */
1896 dmv_x[0] <<= 1 - s->quarter_sample;
1897 dmv_y[0] <<= 1 - s->quarter_sample;
1898 dmv_x[1] <<= 1 - s->quarter_sample;
1899 dmv_y[1] <<= 1 - s->quarter_sample;
1900
1901 wrap = s->b8_stride;
1902 xy = s->block_index[0];
1903
1904 if(s->mb_intra) {
1905 s->current_picture.motion_val[0][xy][0] =
1906 s->current_picture.motion_val[0][xy][1] =
1907 s->current_picture.motion_val[1][xy][0] =
1908 s->current_picture.motion_val[1][xy][1] = 0;
1909 return;
1910 }
1911 s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);
1912 s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);
1913 s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);
1914 s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);
1915
1916 /* Pullback predicted motion vectors as specified in 8.4.5.4 */
1917 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));
1918 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));
1919 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));
1920 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));
1921 if(direct) {
1922 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
1923 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
1924 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
1925 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
1926 return;
1927 }
1928
1929 if((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
1930 C = s->current_picture.motion_val[0][xy - 2];
1931 A = s->current_picture.motion_val[0][xy - wrap*2];
1932 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
1933 B = s->current_picture.motion_val[0][xy - wrap*2 + off];
1934
1935 if(!s->mb_x) C[0] = C[1] = 0;
1936 if(!s->first_slice_line) { // predictor A is not out of bounds
1937 if(s->mb_width == 1) {
1938 px = A[0];
1939 py = A[1];
1940 } else {
1941 px = mid_pred(A[0], B[0], C[0]);
1942 py = mid_pred(A[1], B[1], C[1]);
1943 }
1944 } else if(s->mb_x) { // predictor C is not out of bounds
1945 px = C[0];
1946 py = C[1];
1947 } else {
1948 px = py = 0;
1949 }
1950 /* Pullback MV as specified in 8.3.5.3.4 */
1951 {
1952 int qx, qy, X, Y;
1953 if(v->profile < PROFILE_ADVANCED) {
1954 qx = (s->mb_x << 5);
1955 qy = (s->mb_y << 5);
1956 X = (s->mb_width << 5) - 4;
1957 Y = (s->mb_height << 5) - 4;
1958 if(qx + px < -28) px = -28 - qx;
1959 if(qy + py < -28) py = -28 - qy;
1960 if(qx + px > X) px = X - qx;
1961 if(qy + py > Y) py = Y - qy;
1962 } else {
1963 qx = (s->mb_x << 6);
1964 qy = (s->mb_y << 6);
1965 X = (s->mb_width << 6) - 4;
1966 Y = (s->mb_height << 6) - 4;
1967 if(qx + px < -60) px = -60 - qx;
1968 if(qy + py < -60) py = -60 - qy;
1969 if(qx + px > X) px = X - qx;
1970 if(qy + py > Y) py = Y - qy;
1971 }
1972 }
1973 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
1974 if(0 && !s->first_slice_line && s->mb_x) {
1975 if(is_intra[xy - wrap])
1976 sum = FFABS(px) + FFABS(py);
1977 else
1978 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
1979 if(sum > 32) {
1980 if(get_bits1(&s->gb)) {
1981 px = A[0];
1982 py = A[1];
1983 } else {
1984 px = C[0];
1985 py = C[1];
1986 }
1987 } else {
1988 if(is_intra[xy - 2])
1989 sum = FFABS(px) + FFABS(py);
1990 else
1991 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
1992 if(sum > 32) {
1993 if(get_bits1(&s->gb)) {
1994 px = A[0];
1995 py = A[1];
1996 } else {
1997 px = C[0];
1998 py = C[1];
1999 }
2000 }
2001 }
2002 }
2003 /* store MV using signed modulus of MV range defined in 4.11 */
2004 s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;
2005 s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;
2006 }
2007 if((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2008 C = s->current_picture.motion_val[1][xy - 2];
2009 A = s->current_picture.motion_val[1][xy - wrap*2];
2010 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2011 B = s->current_picture.motion_val[1][xy - wrap*2 + off];
2012
2013 if(!s->mb_x) C[0] = C[1] = 0;
2014 if(!s->first_slice_line) { // predictor A is not out of bounds
2015 if(s->mb_width == 1) {
2016 px = A[0];
2017 py = A[1];
2018 } else {
2019 px = mid_pred(A[0], B[0], C[0]);
2020 py = mid_pred(A[1], B[1], C[1]);
2021 }
2022 } else if(s->mb_x) { // predictor C is not out of bounds
2023 px = C[0];
2024 py = C[1];
2025 } else {
2026 px = py = 0;
2027 }
2028 /* Pullback MV as specified in 8.3.5.3.4 */
2029 {
2030 int qx, qy, X, Y;
2031 if(v->profile < PROFILE_ADVANCED) {
2032 qx = (s->mb_x << 5);
2033 qy = (s->mb_y << 5);
2034 X = (s->mb_width << 5) - 4;
2035 Y = (s->mb_height << 5) - 4;
2036 if(qx + px < -28) px = -28 - qx;
2037 if(qy + py < -28) py = -28 - qy;
2038 if(qx + px > X) px = X - qx;
2039 if(qy + py > Y) py = Y - qy;
2040 } else {
2041 qx = (s->mb_x << 6);
2042 qy = (s->mb_y << 6);
2043 X = (s->mb_width << 6) - 4;
2044 Y = (s->mb_height << 6) - 4;
2045 if(qx + px < -60) px = -60 - qx;
2046 if(qy + py < -60) py = -60 - qy;
2047 if(qx + px > X) px = X - qx;
2048 if(qy + py > Y) py = Y - qy;
2049 }
2050 }
2051 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2052 if(0 && !s->first_slice_line && s->mb_x) {
2053 if(is_intra[xy - wrap])
2054 sum = FFABS(px) + FFABS(py);
2055 else
2056 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2057 if(sum > 32) {
2058 if(get_bits1(&s->gb)) {
2059 px = A[0];
2060 py = A[1];
2061 } else {
2062 px = C[0];
2063 py = C[1];
2064 }
2065 } else {
2066 if(is_intra[xy - 2])
2067 sum = FFABS(px) + FFABS(py);
2068 else
2069 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2070 if(sum > 32) {
2071 if(get_bits1(&s->gb)) {
2072 px = A[0];
2073 py = A[1];
2074 } else {
2075 px = C[0];
2076 py = C[1];
2077 }
2078 }
2079 }
2080 }
2081 /* store MV using signed modulus of MV range defined in 4.11 */
2082
2083 s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;
2084 s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;
2085 }
2086 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2087 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2088 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2089 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2090 }
2091
2092 /** Get predicted DC value for I-frames only
2093 * prediction dir: left=0, top=1
2094 * @param s MpegEncContext
2095 * @param[in] n block index in the current MB
2096 * @param dc_val_ptr Pointer to DC predictor
2097 * @param dir_ptr Prediction direction for use in AC prediction
2098 */
2099 static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2100 int16_t **dc_val_ptr, int *dir_ptr)
2101 {
2102 int a, b, c, wrap, pred, scale;
2103 int16_t *dc_val;
2104 static const uint16_t dcpred[32] = {
2105 -1, 1024, 512, 341, 256, 205, 171, 146, 128,
2106 114, 102, 93, 85, 79, 73, 68, 64,
2107 60, 57, 54, 51, 49, 47, 45, 43,
2108 41, 39, 38, 37, 35, 34, 33
2109 };
2110
2111 /* find prediction - wmv3_dc_scale always used here in fact */
2112 if (n < 4) scale = s->y_dc_scale;
2113 else scale = s->c_dc_scale;
2114
2115 wrap = s->block_wrap[n];
2116 dc_val= s->dc_val[0] + s->block_index[n];
2117
2118 /* B A
2119 * C X
2120 */
2121 c = dc_val[ - 1];
2122 b = dc_val[ - 1 - wrap];
2123 a = dc_val[ - wrap];
2124
2125 if (pq < 9 || !overlap)
2126 {
2127 /* Set outer values */
2128 if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];
2129 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];
2130 }
2131 else
2132 {
2133 /* Set outer values */
2134 if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;
2135 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;
2136 }
2137
2138 if (abs(a - b) <= abs(b - c)) {
2139 pred = c;
2140 *dir_ptr = 1;//left
2141 } else {
2142 pred = a;
2143 *dir_ptr = 0;//top
2144 }
2145
2146 /* update predictor */
2147 *dc_val_ptr = &dc_val[0];
2148 return pred;
2149 }
2150
2151
2152 /** Get predicted DC value
2153 * prediction dir: left=0, top=1
2154 * @param s MpegEncContext
2155 * @param[in] n block index in the current MB
2156 * @param dc_val_ptr Pointer to DC predictor
2157 * @param dir_ptr Prediction direction for use in AC prediction
2158 */
2159 static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2160 int a_avail, int c_avail,
2161 int16_t **dc_val_ptr, int *dir_ptr)
2162 {
2163 int a, b, c, wrap, pred, scale;
2164 int16_t *dc_val;
2165 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2166 int q1, q2 = 0;
2167
2168 /* find prediction - wmv3_dc_scale always used here in fact */
2169 if (n < 4) scale = s->y_dc_scale;
2170 else scale = s->c_dc_scale;
2171
2172 wrap = s->block_wrap[n];
2173 dc_val= s->dc_val[0] + s->block_index[n];
2174
2175 /* B A
2176 * C X
2177 */
2178 c = dc_val[ - 1];
2179 b = dc_val[ - 1 - wrap];
2180 a = dc_val[ - wrap];
2181 /* scale predictors if needed */
2182 q1 = s->current_picture.qscale_table[mb_pos];
2183 if(c_avail && (n!= 1 && n!=3)) {
2184 q2 = s->current_picture.qscale_table[mb_pos - 1];
2185 if(q2 && q2 != q1)
2186 c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2187 }
2188 if(a_avail && (n!= 2 && n!=3)) {
2189 q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2190 if(q2 && q2 != q1)
2191 a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2192 }
2193 if(a_avail && c_avail && (n!=3)) {
2194 int off = mb_pos;
2195 if(n != 1) off--;
2196 if(n != 2) off -= s->mb_stride;
2197 q2 = s->current_picture.qscale_table[off];
2198 if(q2 && q2 != q1)
2199 b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2200 }
2201
2202 if(a_avail && c_avail) {
2203 if(abs(a - b) <= abs(b - c)) {
2204 pred = c;
2205 *dir_ptr = 1;//left
2206 } else {
2207 pred = a;
2208 *dir_ptr = 0;//top
2209 }
2210 } else if(a_avail) {
2211 pred = a;
2212 *dir_ptr = 0;//top
2213 } else if(c_avail) {
2214 pred = c;
2215 *dir_ptr = 1;//left
2216 } else {
2217 pred = 0;
2218 *dir_ptr = 1;//left
2219 }
2220
2221 /* update predictor */
2222 *dc_val_ptr = &dc_val[0];
2223 return pred;
2224 }
2225
2226
2227 /**
2228 * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles
2229 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
2230 * @{
2231 */
2232
2233 static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)
2234 {
2235 int xy, wrap, pred, a, b, c;
2236
2237 xy = s->block_index[n];
2238 wrap = s->b8_stride;
2239
2240 /* B C
2241 * A X
2242 */
2243 a = s->coded_block[xy - 1 ];
2244 b = s->coded_block[xy - 1 - wrap];
2245 c = s->coded_block[xy - wrap];
2246
2247 if (b == c) {
2248 pred = a;
2249 } else {
2250 pred = c;
2251 }
2252
2253 /* store value */
2254 *coded_block_ptr = &s->coded_block[xy];
2255
2256 return pred;
2257 }
2258
2259 /**
2260 * Decode one AC coefficient
2261 * @param v The VC1 context
2262 * @param last Last coefficient
2263 * @param skip How much zero coefficients to skip
2264 * @param value Decoded AC coefficient value
2265 * @see 8.1.3.4
2266 */
2267 static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)
2268 {
2269 GetBitContext *gb = &v->s.gb;
2270 int index, escape, run = 0, level = 0, lst = 0;
2271
2272 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2273 if (index != vc1_ac_sizes[codingset] - 1) {
2274 run = vc1_index_decode_table[codingset][index][0];
2275 level = vc1_index_decode_table[codingset][index][1];
2276 lst = index >= vc1_last_decode_table[codingset];
2277 if(get_bits1(gb))
2278 level = -level;
2279 } else {
2280 escape = decode210(gb);
2281 if (escape != 2) {
2282 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2283 run = vc1_index_decode_table[codingset][index][0];
2284 level = vc1_index_decode_table[codingset][index][1];
2285 lst = index >= vc1_last_decode_table[codingset];
2286 if(escape == 0) {
2287 if(lst)
2288 level += vc1_last_delta_level_table[codingset][run];
2289 else
2290 level += vc1_delta_level_table[codingset][run];
2291 } else {
2292 if(lst)
2293 run += vc1_last_delta_run_table[codingset][level] + 1;
2294 else
2295 run += vc1_delta_run_table[codingset][level] + 1;
2296 }
2297 if(get_bits1(gb))
2298 level = -level;
2299 } else {
2300 int sign;
2301 lst = get_bits1(gb);
2302 if(v->s.esc3_level_length == 0) {
2303 if(v->pq < 8 || v->dquantfrm) { // table 59
2304 v->s.esc3_level_length = get_bits(gb, 3);
2305 if(!v->s.esc3_level_length)
2306 v->s.esc3_level_length = get_bits(gb, 2) + 8;
2307 } else { //table 60
2308 v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;
2309 }
2310 v->s.esc3_run_length = 3 + get_bits(gb, 2);
2311 }
2312 run = get_bits(gb, v->s.esc3_run_length);
2313 sign = get_bits1(gb);
2314 level = get_bits(gb, v->s.esc3_level_length);
2315 if(sign)
2316 level = -level;
2317 }
2318 }
2319
2320 *last = lst;
2321 *skip = run;
2322 *value = level;
2323 }
2324
2325 /** Decode intra block in intra frames - should be faster than decode_intra_block
2326 * @param v VC1Context
2327 * @param block block to decode
2328 * @param coded are AC coeffs present or not
2329 * @param codingset set of VLC to decode data
2330 */
2331 static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)
2332 {
2333 GetBitContext *gb = &v->s.gb;
2334 MpegEncContext *s = &v->s;
2335 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2336 int run_diff, i;
2337 int16_t *dc_val;
2338 int16_t *ac_val, *ac_val2;
2339 int dcdiff;
2340
2341 /* Get DC differential */
2342 if (n < 4) {
2343 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2344 } else {
2345 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2346 }
2347 if (dcdiff < 0){
2348 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2349 return -1;
2350 }
2351 if (dcdiff)
2352 {
2353 if (dcdiff == 119 /* ESC index value */)
2354 {
2355 /* TODO: Optimize */
2356 if (v->pq == 1) dcdiff = get_bits(gb, 10);
2357 else if (v->pq == 2) dcdiff = get_bits(gb, 9);
2358 else dcdiff = get_bits(gb, 8);
2359 }
2360 else
2361 {
2362 if (v->pq == 1)
2363 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2364 else if (v->pq == 2)
2365 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2366 }
2367 if (get_bits1(gb))
2368 dcdiff = -dcdiff;
2369 }
2370
2371 /* Prediction */
2372 dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);
2373 *dc_val = dcdiff;
2374
2375 /* Store the quantized DC coeff, used for prediction */
2376 if (n < 4) {
2377 block[0] = dcdiff * s->y_dc_scale;
2378 } else {
2379 block[0] = dcdiff * s->c_dc_scale;
2380 }
2381 /* Skip ? */
2382 run_diff = 0;
2383 i = 0;
2384 if (!coded) {
2385 goto not_coded;
2386 }
2387
2388 //AC Decoding
2389 i = 1;
2390
2391 {
2392 int last = 0, skip, value;
2393 const int8_t *zz_table;
2394 int scale;
2395 int k;
2396
2397 scale = v->pq * 2 + v->halfpq;
2398
2399 if(v->s.ac_pred) {
2400 if(!dc_pred_dir)
2401 zz_table = ff_vc1_horizontal_zz;
2402 else
2403 zz_table = ff_vc1_vertical_zz;
2404 } else
2405 zz_table = ff_vc1_normal_zz;
2406
2407 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2408 ac_val2 = ac_val;
2409 if(dc_pred_dir) //left
2410 ac_val -= 16;
2411 else //top
2412 ac_val -= 16 * s->block_wrap[n];
2413
2414 while (!last) {
2415 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2416 i += skip;
2417 if(i > 63)
2418 break;
2419 block[zz_table[i++]] = value;
2420 }
2421
2422 /* apply AC prediction if needed */
2423 if(s->ac_pred) {
2424 if(dc_pred_dir) { //left
2425 for(k = 1; k < 8; k++)
2426 block[k << 3] += ac_val[k];
2427 } else { //top
2428 for(k = 1; k < 8; k++)
2429 block[k] += ac_val[k + 8];
2430 }
2431 }
2432 /* save AC coeffs for further prediction */
2433 for(k = 1; k < 8; k++) {
2434 ac_val2[k] = block[k << 3];
2435 ac_val2[k + 8] = block[k];
2436 }
2437
2438 /* scale AC coeffs */
2439 for(k = 1; k < 64; k++)
2440 if(block[k]) {
2441 block[k] *= scale;
2442 if(!v->pquantizer)
2443 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2444 }
2445
2446 if(s->ac_pred) i = 63;
2447 }
2448
2449 not_coded:
2450 if(!coded) {
2451 int k, scale;
2452 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2453 ac_val2 = ac_val;
2454
2455 scale = v->pq * 2 + v->halfpq;
2456 memset(ac_val2, 0, 16 * 2);
2457 if(dc_pred_dir) {//left
2458 ac_val -= 16;
2459 if(s->ac_pred)
2460 memcpy(ac_val2, ac_val, 8 * 2);
2461 } else {//top
2462 ac_val -= 16 * s->block_wrap[n];
2463 if(s->ac_pred)
2464 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2465 }
2466
2467 /* apply AC prediction if needed */
2468 if(s->ac_pred) {
2469 if(dc_pred_dir) { //left
2470 for(k = 1; k < 8; k++) {
2471 block[k << 3] = ac_val[k] * scale;
2472 if(!v->pquantizer && block[k << 3])
2473 block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;
2474 }
2475 } else { //top
2476 for(k = 1; k < 8; k++) {
2477 block[k] = ac_val[k + 8] * scale;
2478 if(!v->pquantizer && block[k])
2479 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2480 }
2481 }
2482 i = 63;
2483 }
2484 }
2485 s->block_last_index[n] = i;
2486
2487 return 0;
2488 }
2489
2490 /** Decode intra block in intra frames - should be faster than decode_intra_block
2491 * @param v VC1Context
2492 * @param block block to decode
2493 * @param coded are AC coeffs present or not
2494 * @param codingset set of VLC to decode data
2495 */
2496 static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant)
2497 {
2498 GetBitContext *gb = &v->s.gb;
2499 MpegEncContext *s = &v->s;
2500 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2501 int run_diff, i;
2502 int16_t *dc_val;
2503 int16_t *ac_val, *ac_val2;
2504 int dcdiff;
2505 int a_avail = v->a_avail, c_avail = v->c_avail;
2506 int use_pred = s->ac_pred;
2507 int scale;
2508 int q1, q2 = 0;
2509 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2510
2511 /* Get DC differential */
2512 if (n < 4) {
2513 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2514 } else {
2515 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2516 }
2517 if (dcdiff < 0){
2518 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2519 return -1;
2520 }
2521 if (dcdiff)
2522 {
2523 if (dcdiff == 119 /* ESC index value */)
2524 {
2525 /* TODO: Optimize */
2526 if (mquant == 1) dcdiff = get_bits(gb, 10);
2527 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2528 else dcdiff = get_bits(gb, 8);
2529 }
2530 else
2531 {
2532 if (mquant == 1)
2533 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2534 else if (mquant == 2)
2535 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2536 }
2537 if (get_bits1(gb))
2538 dcdiff = -dcdiff;
2539 }
2540
2541 /* Prediction */
2542 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);
2543 *dc_val = dcdiff;
2544
2545 /* Store the quantized DC coeff, used for prediction */
2546 if (n < 4) {
2547 block[0] = dcdiff * s->y_dc_scale;
2548 } else {
2549 block[0] = dcdiff * s->c_dc_scale;
2550 }
2551 /* Skip ? */
2552 run_diff = 0;
2553 i = 0;
2554
2555 //AC Decoding
2556 i = 1;
2557
2558 /* check if AC is needed at all */
2559 if(!a_avail && !c_avail) use_pred = 0;
2560 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2561 ac_val2 = ac_val;
2562
2563 scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);
2564
2565 if(dc_pred_dir) //left
2566 ac_val -= 16;
2567 else //top
2568 ac_val -= 16 * s->block_wrap[n];
2569
2570 q1 = s->current_picture.qscale_table[mb_pos];
2571 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2572 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2573 if(dc_pred_dir && n==1) q2 = q1;
2574 if(!dc_pred_dir && n==2) q2 = q1;
2575 if(n==3) q2 = q1;
2576
2577 if(coded) {
2578 int last = 0, skip, value;
2579 const int8_t *zz_table;
2580 int k;
2581
2582 if(v->s.ac_pred) {
2583 if(!dc_pred_dir)
2584 zz_table = ff_vc1_horizontal_zz;
2585 else
2586 zz_table = ff_vc1_vertical_zz;
2587 } else
2588 zz_table = ff_vc1_normal_zz;
2589
2590 while (!last) {
2591 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2592 i += skip;
2593 if(i > 63)
2594 break;
2595 block[zz_table[i++]] = value;
2596 }
2597
2598 /* apply AC prediction if needed */
2599 if(use_pred) {
2600 /* scale predictors if needed*/
2601 if(q2 && q1!=q2) {
2602 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2603 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2604
2605 if(dc_pred_dir) { //left
2606 for(k = 1; k < 8; k++)
2607 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2608 } else { //top
2609 for(k = 1; k < 8; k++)
2610 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2611 }
2612 } else {
2613 if(dc_pred_dir) { //left
2614 for(k = 1; k < 8; k++)
2615 block[k << 3] += ac_val[k];
2616 } else { //top
2617 for(k = 1; k < 8; k++)
2618 block[k] += ac_val[k + 8];
2619 }
2620 }
2621 }
2622 /* save AC coeffs for further prediction */
2623 for(k = 1; k < 8; k++) {
2624 ac_val2[k] = block[k << 3];
2625 ac_val2[k + 8] = block[k];
2626 }
2627
2628 /* scale AC coeffs */
2629 for(k = 1; k < 64; k++)
2630 if(block[k]) {
2631 block[k] *= scale;
2632 if(!v->pquantizer)
2633 block[k] += (block[k] < 0) ? -mquant : mquant;
2634 }
2635
2636 if(use_pred) i = 63;
2637 } else { // no AC coeffs
2638 int k;
2639
2640 memset(ac_val2, 0, 16 * 2);
2641 if(dc_pred_dir) {//left
2642 if(use_pred) {
2643 memcpy(ac_val2, ac_val, 8 * 2);
2644 if(q2 && q1!=q2) {
2645 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2646 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2647 for(k = 1; k < 8; k++)
2648 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2649 }
2650 }
2651 } else {//top
2652 if(use_pred) {
2653 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2654 if(q2 && q1!=q2) {
2655 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2656 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2657 for(k = 1; k < 8; k++)
2658 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2659 }
2660 }
2661 }
2662
2663 /* apply AC prediction if needed */
2664 if(use_pred) {
2665 if(dc_pred_dir) { //left
2666 for(k = 1; k < 8; k++) {
2667 block[k << 3] = ac_val2[k] * scale;
2668 if(!v->pquantizer && block[k << 3])
2669 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2670 }
2671 } else { //top
2672 for(k = 1; k < 8; k++) {
2673 block[k] = ac_val2[k + 8] * scale;
2674 if(!v->pquantizer && block[k])
2675 block[k] += (block[k] < 0) ? -mquant : mquant;
2676 }
2677 }
2678 i = 63;
2679 }
2680 }
2681 s->block_last_index[n] = i;
2682
2683 return 0;
2684 }
2685
2686 /** Decode intra block in inter frames - more generic version than vc1_decode_i_block
2687 * @param v VC1Context
2688 * @param block block to decode
2689 * @param coded are AC coeffs present or not
2690 * @param mquant block quantizer
2691 * @param codingset set of VLC to decode data
2692 */
2693 static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)
2694 {
2695 GetBitContext *gb = &v->s.gb;
2696 MpegEncContext *s = &v->s;
2697 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2698 int run_diff, i;
2699 int16_t *dc_val;
2700 int16_t *ac_val, *ac_val2;
2701 int dcdiff;
2702 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2703 int a_avail = v->a_avail, c_avail = v->c_avail;
2704 int use_pred = s->ac_pred;
2705 int scale;
2706 int q1, q2 = 0;
2707
2708 /* XXX: Guard against dumb values of mquant */
2709 mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );
2710
2711 /* Set DC scale - y and c use the same */
2712 s->y_dc_scale = s->y_dc_scale_table[mquant];
2713 s->c_dc_scale = s->c_dc_scale_table[mquant];
2714
2715 /* Get DC differential */
2716 if (n < 4) {
2717 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2718 } else {
2719 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2720 }
2721 if (dcdiff < 0){
2722 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2723 return -1;
2724 }
2725 if (dcdiff)
2726 {
2727 if (dcdiff == 119 /* ESC index value */)
2728 {
2729 /* TODO: Optimize */
2730 if (mquant == 1) dcdiff = get_bits(gb, 10);
2731 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2732 else dcdiff = get_bits(gb, 8);
2733 }
2734 else
2735 {
2736 if (mquant == 1)
2737 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2738 else if (mquant == 2)
2739 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2740 }
2741 if (get_bits1(gb))
2742 dcdiff = -dcdiff;
2743 }
2744
2745 /* Prediction */
2746 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);
2747 *dc_val = dcdiff;
2748
2749 /* Store the quantized DC coeff, used for prediction */
2750
2751 if (n < 4) {
2752 block[0] = dcdiff * s->y_dc_scale;
2753 } else {
2754 block[0] = dcdiff * s->c_dc_scale;
2755 }
2756 /* Skip ? */
2757 run_diff = 0;
2758 i = 0;
2759
2760 //AC Decoding
2761 i = 1;
2762
2763 /* check if AC is needed at all and adjust direction if needed */
2764 if(!a_avail) dc_pred_dir = 1;
2765 if(!c_avail) dc_pred_dir = 0;
2766 if(!a_avail && !c_avail) use_pred = 0;
2767 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2768 ac_val2 = ac_val;
2769
2770 scale = mquant * 2 + v->halfpq;
2771
2772 if(dc_pred_dir) //left
2773 ac_val -= 16;
2774 else //top
2775 ac_val -= 16 * s->block_wrap[n];
2776
2777 q1 = s->current_picture.qscale_table[mb_pos];
2778 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2779 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2780 if(dc_pred_dir && n==1) q2 = q1;
2781 if(!dc_pred_dir && n==2) q2 = q1;
2782 if(n==3) q2 = q1;
2783
2784 if(coded) {
2785 int last = 0, skip, value;
2786 const int8_t *zz_table;
2787 int k;
2788
2789 zz_table = ff_vc1_simple_progressive_8x8_zz;
2790
2791 while (!last) {
2792 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2793 i += skip;
2794 if(i > 63)
2795 break;
2796 block[zz_table[i++]] = value;
2797 }
2798
2799 /* apply AC prediction if needed */
2800 if(use_pred) {
2801 /* scale predictors if needed*/
2802 if(q2 && q1!=q2) {
2803 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2804 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2805
2806 if(dc_pred_dir) { //left
2807 for(k = 1; k < 8; k++)
2808 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2809 } else { //top
2810 for(k = 1; k < 8; k++)
2811 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2812 }
2813 } else {
2814 if(dc_pred_dir) { //left
2815 for(k = 1; k < 8; k++)
2816 block[k << 3] += ac_val[k];
2817 } else { //top
2818 for(k = 1; k < 8; k++)
2819 block[k] += ac_val[k + 8];
2820 }
2821 }
2822 }
2823 /* save AC coeffs for further prediction */
2824 for(k = 1; k < 8; k++) {
2825 ac_val2[k] = block[k << 3];
2826 ac_val2[k + 8] = block[k];
2827 }
2828
2829 /* scale AC coeffs */
2830 for(k = 1; k < 64; k++)
2831 if(block[k]) {
2832 block[k] *= scale;
2833 if(!v->pquantizer)
2834 block[k] += (block[k] < 0) ? -mquant : mquant;
2835 }
2836
2837 if(use_pred) i = 63;
2838 } else { // no AC coeffs
2839 int k;
2840
2841 memset(ac_val2, 0, 16 * 2);
2842 if(dc_pred_dir) {//left
2843 if(use_pred) {
2844 memcpy(ac_val2, ac_val, 8 * 2);
2845 if(q2 && q1!=q2) {
2846 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2847 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2848 for(k = 1; k < 8; k++)
2849 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2850 }
2851 }
2852 } else {//top
2853 if(use_pred) {
2854 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2855 if(q2 && q1!=q2) {
2856 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2857 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2858 for(k = 1; k < 8; k++)
2859 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2860 }
2861 }
2862 }
2863
2864 /* apply AC prediction if needed */
2865 if(use_pred) {
2866 if(dc_pred_dir) { //left
2867 for(k = 1; k < 8; k++) {
2868 block[k << 3] = ac_val2[k] * scale;
2869 if(!v->pquantizer && block[k << 3])
2870 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2871 }
2872 } else { //top
2873 for(k = 1; k < 8; k++) {
2874 block[k] = ac_val2[k + 8] * scale;
2875 if(!v->pquantizer && block[k])
2876 block[k] += (block[k] < 0) ? -mquant : mquant;
2877 }
2878 }
2879 i = 63;
2880 }
2881 }
2882 s->block_last_index[n] = i;
2883
2884 return 0;
2885 }
2886
2887 /** Decode P block
2888 */
2889 static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,
2890 uint8_t *dst, int linesize, int skip_block)
2891 {
2892 MpegEncContext *s = &v->s;
2893 GetBitContext *gb = &s->gb;
2894 int i, j;
2895 int subblkpat = 0;
2896 int scale, off, idx, last, skip, value;
2897 int ttblk = ttmb & 7;
2898
2899 if(ttmb == -1) {
2900 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)];
2901 }
2902 if(ttblk == TT_4X4) {
2903 subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);
2904 }
2905 if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {
2906 subblkpat = decode012(gb);
2907 if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits
2908 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;
2909 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;
2910 }
2911 scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);
2912
2913 // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT
2914 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {
2915 subblkpat = 2 - (ttblk == TT_8X4_TOP);
2916 ttblk = TT_8X4;
2917 }
2918 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
2919 subblkpat = 2 - (ttblk == TT_4X8_LEFT);
2920 ttblk = TT_4X8;
2921 }
2922 switch(ttblk) {
2923 case TT_8X8:
2924 i = 0;
2925 last = 0;
2926 while (!last) {
2927 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
2928 i += skip;
2929 if(i > 63)
2930 break;
2931 idx = ff_vc1_simple_progressive_8x8_zz[i++];
2932 block[idx] = value * scale;
2933 if(!v->pquantizer)
2934 block[idx] += (block[idx] < 0) ? -mquant : mquant;
2935 }
2936 if(!skip_block){
2937 s->dsp.vc1_inv_trans_8x8(block);
2938 s->dsp.add_pixels_clamped(block, dst, linesize);
2939 }
2940 break;
2941 case TT_4X4:
2942 for(j = 0; j < 4; j++) {
2943 last = subblkpat & (1 << (3 - j));
2944 i = 0;
2945 off = (j & 1) * 4 + (j & 2) * 16;
2946 while (!last) {
2947 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
2948 i += skip;
2949 if(i > 15)
2950 break;
2951 idx = ff_vc1_simple_progressive_4x4_zz[i++];
2952 block[idx + off] = value * scale;
2953 if(!v->pquantizer)
2954 block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;
2955 }
2956 if(!(subblkpat & (1 << (3 - j))) && !skip_block)
2957 s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
2958 }
2959 break;
2960 case TT_8X4:
2961 for(j = 0; j < 2; j++) {
2962 last = subblkpat & (1 << (1 - j));
2963 i = 0;
2964 off = j * 32;
2965 while (!last) {
2966 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
2967 i += skip;
2968 if(i > 31)
2969 break;
2970 idx = v->zz_8x4[i++]+off;
2971 block[idx] = value * scale;
2972 if(!v->pquantizer)
2973 block[idx] += (block[idx] < 0) ? -mquant : mquant;
2974 }
2975 if(!(subblkpat & (1 << (1 - j))) && !skip_block)
2976 s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
2977 }
2978 break;
2979 case TT_4X8:
2980 for(j = 0; j < 2; j++) {
2981 last = subblkpat & (1 << (1 - j));
2982 i = 0;
2983 off = j * 4;
2984 while (!last) {
2985 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
2986 i += skip;
2987 if(i > 31)
2988 break;
2989 idx = v->zz_4x8[i++]+off;
2990 block[idx] = value * scale;
2991 if(!v->pquantizer)
2992 block[idx] += (block[idx] < 0) ? -mquant : mquant;
2993 }
2994 if(!(subblkpat & (1 << (1 - j))) && !skip_block)
2995 s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
2996 }
2997 break;
2998 }
2999 return 0;
3000 }
3001
3002
3003 /** Decode one P-frame MB (in Simple/Main profile)
3004 */
3005 static int vc1_decode_p_mb(VC1Context *v)
3006 {
3007 MpegEncContext *s = &v->s;
3008 GetBitContext *gb = &s->gb;
3009 int i, j;
3010 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3011 int cbp; /* cbp decoding stuff */
3012 int mqdiff, mquant; /* MB quantization */
3013 int ttmb = v->ttfrm; /* MB Transform type */
3014 int status;
3015
3016 static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3017 offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3018 int mb_has_coeffs = 1; /* last_flag */
3019 int dmv_x, dmv_y; /* Differential MV components */
3020 int index, index1; /* LUT indexes */
3021 int val, sign; /* temp values */
3022 int first_block = 1;
3023 int dst_idx, off;
3024 int skipped, fourmv;
3025
3026 mquant = v->pq; /* Loosy initialization */
3027
3028 if (v->mv_type_is_raw)
3029 fourmv = get_bits1(gb);
3030 else
3031 fourmv = v->mv_type_mb_plane[mb_pos];
3032 if (v->skip_is_raw)
3033 skipped = get_bits1(gb);
3034 else
3035 skipped = v->s.mbskip_table[mb_pos];
3036
3037 s->dsp.clear_blocks(s->block[0]);
3038
3039 if (!fourmv) /* 1MV mode */
3040 {
3041 if (!skipped)
3042 {
3043 GET_MVDATA(dmv_x, dmv_y);
3044
3045 if (s->mb_intra) {
3046 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3047 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3048 }
3049 s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
3050 vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
3051
3052 /* FIXME Set DC val for inter block ? */
3053 if (s->mb_intra && !mb_has_coeffs)
3054 {
3055 GET_MQUANT();
3056 s->ac_pred = get_bits1(gb);
3057 cbp = 0;
3058 }
3059 else if (mb_has_coeffs)
3060 {
3061 if (s->mb_intra) s->ac_pred = get_bits1(gb);
3062 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3063 GET_MQUANT();
3064 }
3065 else
3066 {
3067 mquant = v->pq;
3068 cbp = 0;
3069 }
3070 s->current_picture.qscale_table[mb_pos] = mquant;
3071
3072 if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3073 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,
3074 VC1_TTMB_VLC_BITS, 2);
3075 if(!s->mb_intra) vc1_mc_1mv(v, 0);
3076 dst_idx = 0;
3077 for (i=0; i<6; i++)
3078 {
3079 s->dc_val[0][s->block_index[i]] = 0;
3080 dst_idx += i >> 2;
3081 val = ((cbp >> (5 - i)) & 1);
3082 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3083 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3084 if(s->mb_intra) {
3085 /* check if prediction blocks A and C are available */
3086 v->a_avail = v->c_avail = 0;
3087 if(i == 2 || i == 3 || !s->first_slice_line)
3088 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3089 if(i == 1 || i == 3 || s->mb_x)
3090 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3091
3092 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3093 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3094 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3095 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3096 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3097 if(v->pq >= 9 && v->overlap) {
3098 if(v->c_avail)
3099 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3100 if(v->a_avail)
3101 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3102 }
3103 } else if(val) {
3104 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));
3105 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3106 first_block = 0;
3107 }
3108 }
3109 }
3110 else //Skipped
3111 {
3112 s->mb_intra = 0;
3113 for(i = 0; i < 6; i++) {
3114 v->mb_type[0][s->block_index[i]] = 0;
3115 s->dc_val[0][s->block_index[i]] = 0;
3116 }
3117 s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;
3118 s->current_picture.qscale_table[mb_pos] = 0;
3119 vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
3120 vc1_mc_1mv(v, 0);
3121 return 0;
3122 }
3123 } //1MV mode
3124 else //4MV mode
3125 {
3126 if (!skipped /* unskipped MB */)
3127 {
3128 int intra_count = 0, coded_inter = 0;
3129 int is_intra[6], is_coded[6];
3130 /* Get CBPCY */
3131 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3132 for (i=0; i<6; i++)
3133 {
3134 val = ((cbp >> (5 - i)) & 1);
3135 s->dc_val[0][s->block_index[i]] = 0;
3136 s->mb_intra = 0;
3137 if(i < 4) {
3138 dmv_x = dmv_y = 0;
3139 s->mb_intra = 0;
3140 mb_has_coeffs = 0;
3141 if(val) {
3142 GET_MVDATA(dmv_x, dmv_y);
3143 }
3144 vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
3145 if(!s->mb_intra) vc1_mc_4mv_luma(v, i);
3146 intra_count += s->mb_intra;
3147 is_intra[i] = s->mb_intra;
3148 is_coded[i] = mb_has_coeffs;
3149 }
3150 if(i&4){
3151 is_intra[i] = (intra_count >= 3);
3152 is_coded[i] = val;
3153 }
3154 if(i == 4) vc1_mc_4mv_chroma(v);
3155 v->mb_type[0][s->block_index[i]] = is_intra[i];
3156 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];
3157 }
3158 // if there are no coded blocks then don't do anything more
3159 if(!intra_count && !coded_inter) return 0;
3160 dst_idx = 0;
3161 GET_MQUANT();
3162 s->current_picture.qscale_table[mb_pos] = mquant;
3163 /* test if block is intra and has pred */
3164 {
3165 int intrapred = 0;
3166 for(i=0; i<6; i++)
3167 if(is_intra[i]) {
3168 if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])
3169 || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {
3170 intrapred = 1;
3171 break;
3172 }
3173 }
3174 if(intrapred)s->ac_pred = get_bits1(gb);
3175 else s->ac_pred = 0;
3176 }
3177 if (!v->ttmbf && coded_inter)
3178 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3179 for (i=0; i<6; i++)
3180 {
3181 dst_idx += i >> 2;
3182 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3183 s->mb_intra = is_intra[i];
3184 if (is_intra[i]) {
3185 /* check if prediction blocks A and C are available */
3186 v->a_avail = v->c_avail = 0;
3187 if(i == 2 || i == 3 || !s->first_slice_line)
3188 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3189 if(i == 1 || i == 3 || s->mb_x)
3190 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3191
3192 vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);
3193 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3194 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3195 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3196 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
3197 if(v->pq >= 9 && v->overlap) {
3198 if(v->c_avail)
3199 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3200 if(v->a_avail)
3201 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3202 }
3203 } else if(is_coded[i]) {
3204 status = 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));
3205 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3206 first_block = 0;
3207 }
3208 }
3209 return status;
3210 }
3211 else //Skipped MB
3212 {
3213 s->mb_intra = 0;
3214 s->current_picture.qscale_table[mb_pos] = 0;
3215 for (i=0; i<6; i++) {
3216 v->mb_type[0][s->block_index[i]] = 0;
3217 s->dc_val[0][s->block_index[i]] = 0;
3218 }
3219 for (i=0; i<4; i++)
3220 {
3221 vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
3222 vc1_mc_4mv_luma(v, i);
3223 }
3224 vc1_mc_4mv_chroma(v);
3225 s->current_picture.qscale_table[mb_pos] = 0;
3226 return 0;
3227 }
3228 }
3229
3230 /* Should never happen */
3231 return -1;
3232 }
3233
3234 /** Decode one B-frame MB (in Main profile)
3235 */
3236 static void vc1_decode_b_mb(VC1Context *v)
3237 {
3238 MpegEncContext *s = &v->s;
3239 GetBitContext *gb = &s->gb;
3240 int i, j;
3241 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3242 int cbp = 0; /* cbp decoding stuff */
3243 int mqdiff, mquant; /* MB quantization */
3244 int ttmb = v->ttfrm; /* MB Transform type */
3245
3246 static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3247 offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3248 int mb_has_coeffs = 0; /* last_flag */
3249 int index, index1; /* LUT indexes */
3250 int val, sign; /* temp values */
3251 int first_block = 1;
3252 int dst_idx, off;
3253 int skipped, direct;
3254 int dmv_x[2], dmv_y[2];
3255 int bmvtype = BMV_TYPE_BACKWARD;
3256
3257 mquant = v->pq; /* Loosy initialization */
3258 s->mb_intra = 0;
3259
3260 if (v->dmb_is_raw)
3261 direct = get_bits1(gb);
3262 else
3263 direct = v->direct_mb_plane[mb_pos];
3264 if (v->skip_is_raw)
3265 skipped = get_bits1(gb);
3266 else
3267 skipped = v->s.mbskip_table[mb_pos];
3268
3269 s->dsp.clear_blocks(s->block[0]);
3270 dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
3271 for(i = 0; i < 6; i++) {
3272 v->mb_type[0][s->block_index[i]] = 0;
3273 s->dc_val[0][s->block_index[i]] = 0;
3274 }
3275 s->current_picture.qscale_table[mb_pos] = 0;
3276
3277 if (!direct) {
3278 if (!skipped) {
3279 GET_MVDATA(dmv_x[0], dmv_y[0]);
3280 dmv_x[1] = dmv_x[0];
3281 dmv_y[1] = dmv_y[0];
3282 }
3283 if(skipped || !s->mb_intra) {
3284 bmvtype = decode012(gb);
3285 switch(bmvtype) {
3286 case 0:
3287 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
3288 break;
3289 case 1:
3290 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
3291 break;
3292 case 2:
3293 bmvtype = BMV_TYPE_INTERPOLATED;
3294 dmv_x[0] = dmv_y[0] = 0;
3295 }
3296 }
3297 }
3298 for(i = 0; i < 6; i++)
3299 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3300
3301 if (skipped) {
3302 if(direct) bmvtype = BMV_TYPE_INTERPOLATED;
3303 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3304 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3305 return;
3306 }
3307 if (direct) {
3308 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3309 GET_MQUANT();
3310 s->mb_intra = 0;
3311 mb_has_coeffs = 0;
3312 s->current_picture.qscale_table[mb_pos] = mquant;
3313 if(!v->ttmbf)
3314 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3315 dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;
3316 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3317 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3318 } else {
3319 if(!mb_has_coeffs && !s->mb_intra) {
3320 /* no coded blocks - effectively skipped */
3321 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3322 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3323 return;
3324 }
3325 if(s->mb_intra && !mb_has_coeffs) {
3326 GET_MQUANT();
3327 s->current_picture.qscale_table[mb_pos] = mquant;
3328 s->ac_pred = get_bits1(gb);
3329 cbp = 0;
3330 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3331 } else {
3332 if(bmvtype == BMV_TYPE_INTERPOLATED) {
3333 GET_MVDATA(dmv_x[0], dmv_y[0]);
3334 if(!mb_has_coeffs) {
3335 /* interpolated skipped block */
3336 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3337 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3338 return;
3339 }
3340 }
3341 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3342 if(!s->mb_intra) {
3343 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3344 }
3345 if(s->mb_intra)
3346 s->ac_pred = get_bits1(gb);
3347 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3348 GET_MQUANT();
3349 s->current_picture.qscale_table[mb_pos] = mquant;
3350 if(!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3351 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3352 }
3353 }
3354 dst_idx = 0;
3355 for (i=0; i<6; i++)
3356 {
3357 s->dc_val[0][s->block_index[i]] = 0;
3358 dst_idx += i >> 2;
3359 val = ((cbp >> (5 - i)) & 1);
3360 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3361 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3362 if(s->mb_intra) {
3363 /* check if prediction blocks A and C are available */
3364 v->a_avail = v->c_avail = 0;
3365 if(i == 2 || i == 3 || !s->first_slice_line)
3366 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3367 if(i == 1 || i == 3 || s->mb_x)
3368 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3369
3370 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3371 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3372 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3373 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3374 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3375 } else if(val) {
3376 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));
3377 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3378 first_block = 0;
3379 }
3380 }
3381 }
3382
3383 /** Decode blocks of I-frame
3384 */
3385 static void vc1_decode_i_blocks(VC1Context *v)
3386 {
3387 int k, j;
3388 MpegEncContext *s = &v->s;
3389 int cbp, val;
3390 uint8_t *coded_val;
3391 int mb_pos;
3392
3393 /* select codingmode used for VLC tables selection */
3394 switch(v->y_ac_table_index){
3395 case 0:
3396 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3397 break;
3398 case 1:
3399 v->codingset = CS_HIGH_MOT_INTRA;
3400 break;
3401 case 2:
3402 v->codingset = CS_MID_RATE_INTRA;
3403 break;
3404 }
3405
3406 switch(v->c_ac_table_index){
3407 case 0:
3408 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3409 break;
3410 case 1:
3411 v->codingset2 = CS_HIGH_MOT_INTER;
3412 break;
3413 case 2:
3414 v->codingset2 = CS_MID_RATE_INTER;
3415 break;
3416 }
3417
3418 /* Set DC scale - y and c use the same */
3419 s->y_dc_scale = s->y_dc_scale_table[v->pq];
3420 s->c_dc_scale = s->c_dc_scale_table[v->pq];
3421
3422 //do frame decode
3423 s->mb_x = s->mb_y = 0;
3424 s->mb_intra = 1;
3425 s->first_slice_line = 1;
3426 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3427 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3428 ff_init_block_index(s);
3429 ff_update_block_index(s);
3430 s->dsp.clear_blocks(s->block[0]);
3431 mb_pos = s->mb_x + s->mb_y * s->mb_width;
3432 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3433 s->current_picture.qscale_table[mb_pos] = v->pq;
3434 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3435 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3436
3437 // do actual MB decoding and displaying
3438 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3439 v->s.ac_pred = get_bits1(&v->s.gb);
3440
3441 for(k = 0; k < 6; k++) {
3442 val = ((cbp >> (5 - k)) & 1);
3443
3444 if (k < 4) {
3445 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3446 val = val ^ pred;
3447 *coded_val = val;
3448 }
3449 cbp |= val << (5 - k);
3450
3451 vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);
3452
3453 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3454 if(v->pq >= 9 && v->overlap) {
3455 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3456 }
3457 }
3458
3459 vc1_put_block(v, s->block);
3460 if(v->pq >= 9 && v->overlap) {
3461 if(s->mb_x) {
3462 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3463 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3464 if(!(s->flags & CODEC_FLAG_GRAY)) {
3465 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3466 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3467 }
3468 }
3469 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3470 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3471 if(!s->first_slice_line) {
3472 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3473 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3474 if(!(s->flags & CODEC_FLAG_GRAY)) {
3475 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3476 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3477 }
3478 }
3479 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3480 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3481 }
3482
3483 if(get_bits_count(&s->gb) > v->bits) {
3484 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3485 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3486 return;
3487 }
3488 }
3489 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3490 s->first_slice_line = 0;
3491 }
3492 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3493 }
3494
3495 /** Decode blocks of I-frame for advanced profile
3496 */
3497 static void vc1_decode_i_blocks_adv(VC1Context *v)
3498 {
3499 int k, j;
3500 MpegEncContext *s = &v->s;
3501 int cbp, val;
3502 uint8_t *coded_val;
3503 int mb_pos;
3504 int mquant = v->pq;
3505 int mqdiff;
3506 int overlap;
3507 GetBitContext *gb = &s->gb;
3508
3509 /* select codingmode used for VLC tables selection */
3510 switch(v->y_ac_table_index){
3511 case 0:
3512 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3513 break;
3514 case 1:
3515 v->codingset = CS_HIGH_MOT_INTRA;
3516 break;
3517 case 2:
3518 v->codingset = CS_MID_RATE_INTRA;
3519 break;
3520 }
3521
3522 switch(v->c_ac_table_index){
3523 case 0:
3524 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3525 break;
3526 case 1:
3527 v->codingset2 = CS_HIGH_MOT_INTER;
3528 break;
3529 case 2:
3530 v->codingset2 = CS_MID_RATE_INTER;
3531 break;
3532 }
3533
3534 //do frame decode
3535 s->mb_x = s->mb_y = 0;
3536 s->mb_intra = 1;
3537 s->first_slice_line = 1;
3538 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3539 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3540 ff_init_block_index(s);
3541 ff_update_block_index(s);
3542 s->dsp.clear_blocks(s->block[0]);
3543 mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3544 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3545 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3546 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3547
3548 // do actual MB decoding and displaying
3549 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3550 if(v->acpred_is_raw)
3551 v->s.ac_pred = get_bits1(&v->s.gb);
3552 else
3553 v->s.ac_pred = v->acpred_plane[mb_pos];
3554
3555 if(v->condover == CONDOVER_SELECT) {
3556 if(v->overflg_is_raw)
3557 overlap = get_bits1(&v->s.gb);
3558 else
3559 overlap = v->over_flags_plane[mb_pos];
3560 } else
3561 overlap = (v->condover == CONDOVER_ALL);
3562
3563 GET_MQUANT();
3564
3565 s->current_picture.qscale_table[mb_pos] = mquant;
3566 /* Set DC scale - y and c use the same */
3567 s->y_dc_scale = s->y_dc_scale_table[mquant];
3568 s->c_dc_scale = s->c_dc_scale_table[mquant];
3569
3570 for(k = 0; k < 6; k++) {
3571 val = ((cbp >> (5 - k)) & 1);
3572
3573 if (k < 4) {
3574 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3575 val = val ^ pred;
3576 *coded_val = val;
3577 }
3578 cbp |= val << (5 - k);
3579
3580 v->a_avail = !s->first_slice_line || (k==2 || k==3);
3581 v->c_avail = !!s->mb_x || (k==1 || k==3);
3582
3583 vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);
3584
3585 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3586 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3587 }
3588
3589 vc1_put_block(v, s->block);
3590 if(overlap) {
3591 if(s->mb_x) {
3592 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3593 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3594 if(!(s->flags & CODEC_FLAG_GRAY)) {
3595 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3596 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3597 }
3598 }
3599 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3600 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3601 if(!s->first_slice_line) {
3602 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3603 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3604 if(!(s->flags & CODEC_FLAG_GRAY)) {
3605 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3606 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3607 }
3608 }
3609 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3610 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3611 }
3612
3613 if(get_bits_count(&s->gb) > v->bits) {
3614 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3615 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3616 return;
3617 }
3618 }
3619 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3620 s->first_slice_line = 0;
3621 }
3622 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3623 }
3624
3625 static void vc1_decode_p_blocks(VC1Context *v)
3626 {
3627 MpegEncContext *s = &v->s;
3628
3629 /* select codingmode used for VLC tables selection */
3630 switch(v->c_ac_table_index){
3631 case 0:
3632 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3633 break;
3634 case 1:
3635 v->codingset = CS_HIGH_MOT_INTRA;
3636 break;
3637 case 2:
3638 v->codingset = CS_MID_RATE_INTRA;
3639 break;
3640 }
3641
3642 switch(v->c_ac_table_index){
3643 case 0:
3644 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3645 break;
3646 case 1:
3647 v->codingset2 = CS_HIGH_MOT_INTER;
3648 break;
3649 case 2:
3650 v->codingset2 = CS_MID_RATE_INTER;
3651 break;
3652 }
3653
3654 s->first_slice_line = 1;
3655 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3656 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3657 ff_init_block_index(s);
3658 ff_update_block_index(s);
3659 s->dsp.clear_blocks(s->block[0]);
3660
3661 vc1_decode_p_mb(v);
3662 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3663 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3664 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);
3665 return;
3666 }
3667 }
3668 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3669 s->first_slice_line = 0;
3670 }
3671 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3672 }
3673
3674 static void vc1_decode_b_blocks(VC1Context *v)
3675 {
3676 MpegEncContext *s = &v->s;
3677
3678 /* select codingmode used for VLC tables selection */
3679 switch(v->c_ac_table_index){
3680 case 0:
3681 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3682 break;
3683 case 1:
3684 v->codingset = CS_HIGH_MOT_INTRA;
3685 break;
3686 case 2:
3687 v->codingset = CS_MID_RATE_INTRA;
3688 break;
3689 }
3690
3691 switch(v->c_ac_table_index){
3692 case 0:
3693 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3694 break;
3695 case 1:
3696 v->codingset2 = CS_HIGH_MOT_INTER;
3697 break;
3698 case 2:
3699 v->codingset2 = CS_MID_RATE_INTER;
3700 break;
3701 }
3702
3703 s->first_slice_line = 1;
3704 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3705 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3706 ff_init_block_index(s);
3707 ff_update_block_index(s);
3708 s->dsp.clear_blocks(s->block[0]);
3709
3710 vc1_decode_b_mb(v);
3711 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3712 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3713 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);
3714 return;
3715 }
3716 }
3717 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3718 s->first_slice_line = 0;
3719 }
3720 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3721 }
3722
3723 static void vc1_decode_skip_blocks(VC1Context *v)
3724 {
3725 MpegEncContext *s = &v->s;
3726
3727 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3728 s->first_slice_line = 1;
3729 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3730 s->mb_x = 0;
3731 ff_init_block_index(s);
3732 ff_update_block_index(s);
3733 memcpy(s->dest[0], s->last_picture.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16);
3734 memcpy(s->dest[1], s->last_picture.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3735 memcpy(s->dest[2], s->last_picture.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3736 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3737 s->first_slice_line = 0;
3738 }
3739 s->pict_type = FF_P_TYPE;
3740 }
3741
3742 static void vc1_decode_blocks(VC1Context *v)
3743 {
3744
3745 v->s.esc3_level_length = 0;
3746 if(v->x8_type){
3747 ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );
3748 }else{
3749
3750 switch(v->s.pict_type) {
3751 case FF_I_TYPE:
3752 if(v->profile == PROFILE_ADVANCED)
3753 vc1_decode_i_blocks_adv(v);
3754 else
3755 vc1_decode_i_blocks(v);
3756 break;
3757 case FF_P_TYPE:
3758 if(v->p_frame_skipped)
3759 vc1_decode_skip_blocks(v);
3760 else
3761 vc1_decode_p_blocks(v);
3762 break;
3763 case FF_B_TYPE:
3764 if(v->bi_type){
3765 if(v->profile == PROFILE_ADVANCED)
3766 vc1_decode_i_blocks_adv(v);
3767 else
3768 vc1_decode_i_blocks(v);
3769 }else
3770 vc1_decode_b_blocks(v);
3771 break;
3772 }
3773 }
3774 }
3775
3776 /** Find VC-1 marker in buffer
3777 * @return position where next marker starts or end of buffer if no marker found
3778 */
3779 static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)
3780 {
3781 uint32_t mrk = 0xFFFFFFFF;
3782
3783 if(end-src < 4) return end;
3784 while(src < end){
3785 mrk = (mrk << 8) | *src++;
3786 if(IS_MARKER(mrk))
3787 return src-4;
3788 }
3789 return end;
3790 }
3791
3792 static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)
3793 {
3794 int dsize = 0, i;
3795
3796 if(size < 4){
3797 for(dsize = 0; dsize < size; dsize++) *dst++ = *src++;
3798 return size;
3799 }
3800 for(i = 0; i < size; i++, src++) {
3801 if(src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {
3802 dst[dsize++] = src[1];
3803 src++;
3804 i++;
3805 } else
3806 dst[dsize++] = *src;
3807 }
3808 return dsize;
3809 }
3810
3811 /** Initialize a VC1/WMV3 decoder
3812 * @todo TODO: Handle VC-1 IDUs (Transport level?)
3813 * @todo TODO: Decypher remaining bits in extra_data
3814 */
3815 static av_cold int vc1_decode_init(AVCodecContext *avctx)
3816 {
3817 VC1Context *v = avctx->priv_data;
3818 MpegEncContext *s = &v->s;
3819 GetBitContext gb;
3820
3821 if (!avctx->extradata_size || !avctx->extradata) return -1;
3822 if (!(avctx->flags & CODEC_FLAG_GRAY))
3823 avctx->pix_fmt = PIX_FMT_YUV420P;
3824 else
3825 avctx->pix_fmt = PIX_FMT_GRAY8;
3826 v->s.avctx = avctx;
3827 avctx->flags |= CODEC_FLAG_EMU_EDGE;
3828 v->s.flags |= CODEC_FLAG_EMU_EDGE;
3829
3830 if(avctx->idct_algo==FF_IDCT_AUTO){
3831 avctx->idct_algo=FF_IDCT_WMV2;
3832 }
3833
3834 if(ff_h263_decode_init(avctx) < 0)
3835 return -1;
3836 if (vc1_init_common(v) < 0) return -1;
3837
3838 avctx->coded_width = avctx->width;
3839 avctx->coded_height = avctx->height;
3840 if (avctx->codec_id == CODEC_ID_WMV3)
3841 {
3842 int count = 0;
3843
3844 // looks like WMV3 has a sequence header stored in the extradata
3845 // advanced sequence header may be before the first frame
3846 // the last byte of the extradata is a version number, 1 for the
3847 // samples we can decode
3848
3849 init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);
3850
3851 if (decode_sequence_header(avctx, &gb) < 0)
3852 return -1;
3853
3854 count = avctx->extradata_size*8 - get_bits_count(&gb);
3855 if (count>0)
3856 {
3857 av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",
3858 count, get_bits(&gb, count));
3859 }
3860 else if (count < 0)
3861 {
3862 av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);
3863 }
3864 } else { // VC1/WVC1
3865 const uint8_t *start = avctx->extradata;
3866 uint8_t *end = avctx->extradata + avctx->extradata_size;
3867 const uint8_t *next;
3868 int size, buf2_size;
3869 uint8_t *buf2 = NULL;
3870 int seq_initialized = 0, ep_initialized = 0;
3871
3872 if(avctx->extradata_size < 16) {
3873 av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);
3874 return -1;
3875 }
3876
3877 buf2 = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
3878 if(start[0]) start++; // in WVC1 extradata first byte is its size
3879 next = start;
3880 for(; next < end; start = next){
3881 next = find_next_marker(start + 4, end);
3882 size = next - start - 4;
3883 if(size <= 0) continue;
3884 buf2_size = vc1_unescape_buffer(start + 4, size, buf2);
3885 init_get_bits(&gb, buf2, buf2_size * 8);
3886 switch(AV_RB32(start)){
3887 case VC1_CODE_SEQHDR:
3888 if(decode_sequence_header(avctx, &gb) < 0){
3889 av_free(buf2);
3890 return -1;
3891 }
3892 seq_initialized = 1;
3893 break;
3894 case VC1_CODE_ENTRYPOINT:
3895 if(decode_entry_point(avctx, &gb) < 0){
3896 av_free(buf2);
3897 return -1;
3898 }
3899 ep_initialized = 1;
3900 break;
3901 }
3902 }
3903 av_free(buf2);
3904 if(!seq_initialized || !ep_initialized){
3905 av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");
3906 return -1;
3907 }
3908 }
3909 avctx->has_b_frames= !!(avctx->max_b_frames);
3910 s->low_delay = !avctx->has_b_frames;
3911
3912 s->mb_width = (avctx->coded_width+15)>>4;
3913 s->mb_height = (avctx->coded_height+15)>>4;
3914
3915 /* Allocate mb bitplanes */
3916 v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);
3917 v->direct_mb_plane = av_malloc(s->mb_stride * s->mb_height);
3918 v->acpred_plane = av_malloc(s->mb_stride * s->mb_height);
3919 v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height);
3920
3921 /* allocate block type info in that way so it could be used with s->block_index[] */
3922 v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);
3923 v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;
3924 v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;
3925 v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);
3926
3927 /* Init coded blocks info */
3928 if (v->profile == PROFILE_ADVANCED)
3929 {
3930 // if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
3931 // return -1;
3932 // if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
3933 // return -1;
3934 }
3935
3936 ff_intrax8_common_init(&v->x8,s);
3937 return 0;
3938 }
3939
3940
3941 /** Decode a VC1/WMV3 frame
3942 * @todo TODO: Handle VC-1 IDUs (Transport level?)
3943 */
3944 static int vc1_decode_frame(AVCodecContext *avctx,
3945 void *data, int *data_size,
3946 const uint8_t *buf, int buf_size)
3947 {
3948 VC1Context *v = avctx->priv_data;
3949 MpegEncContext *s = &v->s;
3950 AVFrame *pict = data;
3951 uint8_t *buf2 = NULL;
3952
3953 /* no supplementary picture */
3954 if (buf_size == 0) {
3955 /* special case for last picture */
3956 if (s->low_delay==0 && s->next_picture_ptr) {
3957 *pict= *(AVFrame*)s->next_picture_ptr;
3958 s->next_picture_ptr= NULL;
3959
3960 *data_size = sizeof(AVFrame);
3961 }
3962
3963 return 0;
3964 }
3965
3966 /* We need to set current_picture_ptr before reading the header,
3967 * otherwise we cannot store anything in there. */
3968 if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){
3969 int i= ff_find_unused_picture(s, 0);
3970 s->current_picture_ptr= &s->picture[i];
3971 }
3972
3973 //for advanced profile we may need to parse and unescape data
3974 if (avctx->codec_id == CODEC_ID_VC1) {
3975 int buf_size2 = 0;
3976 buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
3977
3978 if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */
3979 const uint8_t *start, *end, *next;
3980 int size;
3981
3982 next = buf;
3983 for(start = buf, end = buf + buf_size; next < end; start = next){
3984 next = find_next_marker(start + 4, end);
3985 size = next - start - 4;
3986 if(size <= 0) continue;
3987 switch(AV_RB32(start)){
3988 case VC1_CODE_FRAME:
3989 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
3990 break;
3991 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */
3992 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
3993 init_get_bits(&s->gb, buf2, buf_size2*8);
3994 decode_entry_point(avctx, &s->gb);
3995 break;
3996 case VC1_CODE_SLICE:
3997 av_log(avctx, AV_LOG_ERROR, "Sliced decoding is not implemented (yet)\n");
3998 av_free(buf2);
3999 return -1;
4000 }
4001 }
4002 }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */
4003 const uint8_t *divider;
4004
4005 divider = find_next_marker(buf, buf + buf_size);
4006 if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){
4007 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");
4008 return -1;
4009 }
4010
4011 buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);
4012 // TODO
4013 av_free(buf2);return -1;
4014 }else{
4015 buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);
4016 }
4017 init_get_bits(&s->gb, buf2, buf_size2*8);
4018 } else
4019 init_get_bits(&s->gb, buf, buf_size*8);
4020 // do parse frame header
4021 if(v->profile < PROFILE_ADVANCED) {
4022 if(vc1_parse_frame_header(v, &s->gb) == -1) {
4023 av_free(buf2);
4024 return -1;
4025 }
4026 } else {
4027 if(vc1_parse_frame_header_adv(v, &s->gb) == -1) {
4028 av_free(buf2);
4029 return -1;
4030 }
4031 }
4032
4033 if(s->pict_type != FF_I_TYPE && !v->res_rtm_flag){
4034 av_free(buf2);
4035 return -1;
4036 }
4037
4038 // for hurry_up==5
4039 s->current_picture.pict_type= s->pict_type;
4040 s->current_picture.key_frame= s->pict_type == FF_I_TYPE;
4041
4042 /* skip B-frames if we don't have reference frames */
4043 if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){
4044 av_free(buf2);
4045 return -1;//buf_size;
4046 }
4047 /* skip b frames if we are in a hurry */
4048 if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;
4049 if( (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
4050 || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
4051 || avctx->skip_frame >= AVDISCARD_ALL) {
4052 av_free(buf2);
4053 return buf_size;
4054 }
4055 /* skip everything if we are in a hurry>=5 */
4056 if(avctx->hurry_up>=5) {
4057 av_free(buf2);
4058 return -1;//buf_size;
4059 }
4060
4061 if(s->next_p_frame_damaged){
4062 if(s->pict_type==FF_B_TYPE)
4063 return buf_size;
4064 else
4065 s->next_p_frame_damaged=0;
4066 }
4067
4068 if(MPV_frame_start(s, avctx) < 0) {
4069 av_free(buf2);
4070 return -1;
4071 }
4072
4073 s->me.qpel_put= s->dsp.put_qpel_pixels_tab;
4074 s->me.qpel_avg= s->dsp.avg_qpel_pixels_tab;
4075
4076 ff_er_frame_start(s);
4077
4078 v->bits = buf_size * 8;
4079 vc1_decode_blocks(v);
4080 //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);
4081 // if(get_bits_count(&s->gb) > buf_size * 8)
4082 // return -1;
4083 ff_er_frame_end(s);
4084
4085 MPV_frame_end(s);
4086
4087 assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);
4088 assert(s->current_picture.pict_type == s->pict_type);
4089 if (s->pict_type == FF_B_TYPE || s->low_delay) {
4090 *pict= *(AVFrame*)s->current_picture_ptr;
4091 } else if (s->last_picture_ptr != NULL) {
4092 *pict= *(AVFrame*)s->last_picture_ptr;
4093 }
4094
4095 if(s->last_picture_ptr || s->low_delay){
4096 *data_size = sizeof(AVFrame);
4097 ff_print_debug_info(s, pict);
4098 }
4099
4100 /* Return the Picture timestamp as the frame number */
4101 /* we subtract 1 because it is added on utils.c */
4102 avctx->frame_number = s->picture_number - 1;
4103
4104 av_free(buf2);
4105 return buf_size;
4106 }
4107
4108
4109 /** Close a VC1/WMV3 decoder
4110 * @warning Initial try at using MpegEncContext stuff
4111 */
4112 static av_cold int vc1_decode_end(AVCodecContext *avctx)
4113 {
4114 VC1Context *v = avctx->priv_data;
4115
4116 av_freep(&v->hrd_rate);
4117 av_freep(&v->hrd_buffer);
4118 MPV_common_end(&v->s);
4119 av_freep(&v->mv_type_mb_plane);
4120 av_freep(&v->direct_mb_plane);
4121 av_freep(&v->acpred_plane);
4122 av_freep(&v->over_flags_plane);
4123 av_freep(&v->mb_type_base);
4124 ff_intrax8_common_end(&v->x8);
4125 return 0;
4126 }
4127
4128
4129 AVCodec vc1_decoder = {
4130 "vc1",
4131 CODEC_TYPE_VIDEO,
4132 CODEC_ID_VC1,
4133 sizeof(VC1Context),
4134 vc1_decode_init,
4135 NULL,
4136 vc1_decode_end,
4137 vc1_decode_frame,
4138 CODEC_CAP_DELAY,
4139 NULL,
4140 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),
4141 };
4142
4143 AVCodec wmv3_decoder = {
4144 "wmv3",
4145 CODEC_TYPE_VIDEO,
4146 CODEC_ID_WMV3,
4147 sizeof(VC1Context),
4148 vc1_decode_init,
4149 NULL,
4150 vc1_decode_end,
4151 vc1_decode_frame,
4152 CODEC_CAP_DELAY,
4153 NULL,
4154 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),
4155 };
An experimental fork of FFmpeg project for DVCPRO-HD developers