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
6 * This file is part of FFmpeg.
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.
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.
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
24 * @file libavcodec/vc1.c
25 * VC-1 and WMV3 decoder
31 #include "mpegvideo.h"
34 #include "vc1acdata.h"
35 #include "msmpeg4data.h"
37 #include "simple_idct.h"
39 #include "vdpau_internal.h"
44 #define MB_INTRA_VLC_BITS 9
47 static const uint16_t table_mb_intra[64][2];
50 static const uint16_t vlc_offs[] = {
51 0, 520, 552, 616, 1128, 1160, 1224, 1740, 1772, 1836, 1900, 2436,
52 2986, 3050, 3610, 4154, 4218, 4746, 5326, 5390, 5902, 6554, 7658, 8620,
53 9262, 10202, 10756, 11310, 12228, 15078
57 * Init VC-1 specific tables and VC1Context members
58 * @param v The VC1Context to initialize
61 static int vc1_init_common(VC1Context *v)
65 static VLC_TYPE vlc_table[15078][2];
67 v->hrd_rate = v->hrd_buffer = NULL;
72 INIT_VLC_STATIC(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,
73 ff_vc1_bfraction_bits, 1, 1,
74 ff_vc1_bfraction_codes, 1, 1, 1 << VC1_BFRACTION_VLC_BITS);
75 INIT_VLC_STATIC(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,
76 ff_vc1_norm2_bits, 1, 1,
77 ff_vc1_norm2_codes, 1, 1, 1 << VC1_NORM2_VLC_BITS);
78 INIT_VLC_STATIC(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,
79 ff_vc1_norm6_bits, 1, 1,
80 ff_vc1_norm6_codes, 2, 2, 556);
81 INIT_VLC_STATIC(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,
82 ff_vc1_imode_bits, 1, 1,
83 ff_vc1_imode_codes, 1, 1, 1 << VC1_IMODE_VLC_BITS);
86 ff_vc1_ttmb_vlc[i].table = &vlc_table[vlc_offs[i*3+0]];
87 ff_vc1_ttmb_vlc[i].table_allocated = vlc_offs[i*3+1] - vlc_offs[i*3+0];
88 init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,
89 ff_vc1_ttmb_bits[i], 1, 1,
90 ff_vc1_ttmb_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
91 ff_vc1_ttblk_vlc[i].table = &vlc_table[vlc_offs[i*3+1]];
92 ff_vc1_ttblk_vlc[i].table_allocated = vlc_offs[i*3+2] - vlc_offs[i*3+1];
93 init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,
94 ff_vc1_ttblk_bits[i], 1, 1,
95 ff_vc1_ttblk_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
96 ff_vc1_subblkpat_vlc[i].table = &vlc_table[vlc_offs[i*3+2]];
97 ff_vc1_subblkpat_vlc[i].table_allocated = vlc_offs[i*3+3] - vlc_offs[i*3+2];
98 init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,
99 ff_vc1_subblkpat_bits[i], 1, 1,
100 ff_vc1_subblkpat_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
104 ff_vc1_4mv_block_pattern_vlc[i].table = &vlc_table[vlc_offs[i*3+9]];
105 ff_vc1_4mv_block_pattern_vlc[i].table_allocated = vlc_offs[i*3+10] - vlc_offs[i*3+9];
106 init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,
107 ff_vc1_4mv_block_pattern_bits[i], 1, 1,
108 ff_vc1_4mv_block_pattern_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
109 ff_vc1_cbpcy_p_vlc[i].table = &vlc_table[vlc_offs[i*3+10]];
110 ff_vc1_cbpcy_p_vlc[i].table_allocated = vlc_offs[i*3+11] - vlc_offs[i*3+10];
111 init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,
112 ff_vc1_cbpcy_p_bits[i], 1, 1,
113 ff_vc1_cbpcy_p_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
114 ff_vc1_mv_diff_vlc[i].table = &vlc_table[vlc_offs[i*3+11]];
115 ff_vc1_mv_diff_vlc[i].table_allocated = vlc_offs[i*3+12] - vlc_offs[i*3+11];
116 init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,
117 ff_vc1_mv_diff_bits[i], 1, 1,
118 ff_vc1_mv_diff_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
121 ff_vc1_ac_coeff_table[i].table = &vlc_table[vlc_offs[i+21]];
122 ff_vc1_ac_coeff_table[i].table_allocated = vlc_offs[i+22] - vlc_offs[i+21];
123 init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],
124 &vc1_ac_tables[i][0][1], 8, 4,
125 &vc1_ac_tables[i][0][0], 8, 4, INIT_VLC_USE_NEW_STATIC);
127 //FIXME: switching to INIT_VLC_STATIC() results in incorrect decoding
128 init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,
129 &ff_msmp4_mb_i_table[0][1], 4, 2,
130 &ff_msmp4_mb_i_table[0][0], 4, 2, INIT_VLC_USE_STATIC);
136 v->mvrange = 0; /* 7.1.1.18, p80 */
141 /***********************************************************************/
143 * @defgroup vc1bitplane VC-1 Bitplane decoding
161 /** @} */ //imode defines
163 /** Decode rows by checking if they are skipped
164 * @param plane Buffer to store decoded bits
165 * @param[in] width Width of this buffer
166 * @param[in] height Height of this buffer
167 * @param[in] stride of this buffer
169 static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
172 for (y=0; y<height; y++){
173 if (!get_bits1(gb)) //rowskip
174 memset(plane, 0, width);
176 for (x=0; x<width; x++)
177 plane[x] = get_bits1(gb);
182 /** Decode columns by checking if they are skipped
183 * @param plane Buffer to store decoded bits
184 * @param[in] width Width of this buffer
185 * @param[in] height Height of this buffer
186 * @param[in] stride of this buffer
187 * @todo FIXME: Optimize
189 static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
192 for (x=0; x<width; x++){
193 if (!get_bits1(gb)) //colskip
194 for (y=0; y<height; y++)
197 for (y=0; y<height; y++)
198 plane[y*stride] = get_bits1(gb);
203 /** Decode a bitplane's bits
204 * @param data bitplane where to store the decode bits
205 * @param[out] raw_flag pointer to the flag indicating that this bitplane is not coded explicitly
206 * @param v VC-1 context for bit reading and logging
208 * @todo FIXME: Optimize
210 static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)
212 GetBitContext *gb = &v->s.gb;
214 int imode, x, y, code, offset;
215 uint8_t invert, *planep = data;
216 int width, height, stride;
218 width = v->s.mb_width;
219 height = v->s.mb_height;
220 stride = v->s.mb_stride;
221 invert = get_bits1(gb);
222 imode = get_vlc2(gb, ff_vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);
228 //Data is actually read in the MB layer (same for all tests == "raw")
229 *raw_flag = 1; //invert ignored
233 if ((height * width) & 1)
235 *planep++ = get_bits1(gb);
239 // decode bitplane as one long line
240 for (y = offset; y < height * width; y += 2) {
241 code = get_vlc2(gb, ff_vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
242 *planep++ = code & 1;
244 if(offset == width) {
246 planep += stride - width;
248 *planep++ = code >> 1;
250 if(offset == width) {
252 planep += stride - width;
258 if(!(height % 3) && (width % 3)) { // use 2x3 decoding
259 for(y = 0; y < height; y+= 3) {
260 for(x = width & 1; x < width; x += 2) {
261 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
263 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
266 planep[x + 0] = (code >> 0) & 1;
267 planep[x + 1] = (code >> 1) & 1;
268 planep[x + 0 + stride] = (code >> 2) & 1;
269 planep[x + 1 + stride] = (code >> 3) & 1;
270 planep[x + 0 + stride * 2] = (code >> 4) & 1;
271 planep[x + 1 + stride * 2] = (code >> 5) & 1;
273 planep += stride * 3;
275 if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);
277 planep += (height & 1) * stride;
278 for(y = height & 1; y < height; y += 2) {
279 for(x = width % 3; x < width; x += 3) {
280 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
282 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
285 planep[x + 0] = (code >> 0) & 1;
286 planep[x + 1] = (code >> 1) & 1;
287 planep[x + 2] = (code >> 2) & 1;
288 planep[x + 0 + stride] = (code >> 3) & 1;
289 planep[x + 1 + stride] = (code >> 4) & 1;
290 planep[x + 2 + stride] = (code >> 5) & 1;
292 planep += stride * 2;
295 if(x) decode_colskip(data , x, height , stride, &v->s.gb);
296 if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);
300 decode_rowskip(data, width, height, stride, &v->s.gb);
303 decode_colskip(data, width, height, stride, &v->s.gb);
308 /* Applying diff operator */
309 if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
313 for (x=1; x<width; x++)
314 planep[x] ^= planep[x-1];
315 for (y=1; y<height; y++)
318 planep[0] ^= planep[-stride];
319 for (x=1; x<width; x++)
321 if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;
322 else planep[x] ^= planep[x-1];
329 for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride
331 return (imode<<1) + invert;
334 /** @} */ //Bitplane group
336 static void vc1_loop_filter_iblk(MpegEncContext *s, int pq)
339 if(!s->first_slice_line)
340 s->dsp.vc1_v_loop_filter16(s->dest[0], s->linesize, pq);
341 s->dsp.vc1_v_loop_filter16(s->dest[0] + 8*s->linesize, s->linesize, pq);
342 for(i = !s->mb_x*8; i < 16; i += 8)
343 s->dsp.vc1_h_loop_filter16(s->dest[0] + i, s->linesize, pq);
344 for(j = 0; j < 2; j++){
345 if(!s->first_slice_line)
346 s->dsp.vc1_v_loop_filter8(s->dest[j+1], s->uvlinesize, pq);
348 s->dsp.vc1_h_loop_filter8(s->dest[j+1], s->uvlinesize, pq);
352 /***********************************************************************/
353 /** VOP Dquant decoding
354 * @param v VC-1 Context
356 static int vop_dquant_decoding(VC1Context *v)
358 GetBitContext *gb = &v->s.gb;
364 pqdiff = get_bits(gb, 3);
365 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
366 else v->altpq = v->pq + pqdiff + 1;
370 v->dquantfrm = get_bits1(gb);
373 v->dqprofile = get_bits(gb, 2);
374 switch (v->dqprofile)
376 case DQPROFILE_SINGLE_EDGE:
377 case DQPROFILE_DOUBLE_EDGES:
378 v->dqsbedge = get_bits(gb, 2);
380 case DQPROFILE_ALL_MBS:
381 v->dqbilevel = get_bits1(gb);
384 default: break; //Forbidden ?
386 if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)
388 pqdiff = get_bits(gb, 3);
389 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
390 else v->altpq = v->pq + pqdiff + 1;
397 /** Put block onto picture
399 static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
403 DSPContext *dsp = &v->s.dsp;
407 for(k = 0; k < 6; k++)
408 for(j = 0; j < 8; j++)
409 for(i = 0; i < 8; i++)
410 block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;
413 ys = v->s.current_picture.linesize[0];
414 us = v->s.current_picture.linesize[1];
415 vs = v->s.current_picture.linesize[2];
418 dsp->put_pixels_clamped(block[0], Y, ys);
419 dsp->put_pixels_clamped(block[1], Y + 8, ys);
421 dsp->put_pixels_clamped(block[2], Y, ys);
422 dsp->put_pixels_clamped(block[3], Y + 8, ys);
424 if(!(v->s.flags & CODEC_FLAG_GRAY)) {
425 dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
426 dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
430 /** Do motion compensation over 1 macroblock
431 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
433 static void vc1_mc_1mv(VC1Context *v, int dir)
435 MpegEncContext *s = &v->s;
436 DSPContext *dsp = &v->s.dsp;
437 uint8_t *srcY, *srcU, *srcV;
438 int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
440 if(!v->s.last_picture.data[0])return;
442 mx = s->mv[dir][0][0];
443 my = s->mv[dir][0][1];
445 // store motion vectors for further use in B frames
446 if(s->pict_type == FF_P_TYPE) {
447 s->current_picture.motion_val[1][s->block_index[0]][0] = mx;
448 s->current_picture.motion_val[1][s->block_index[0]][1] = my;
450 uvmx = (mx + ((mx & 3) == 3)) >> 1;
451 uvmy = (my + ((my & 3) == 3)) >> 1;
453 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
454 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
457 srcY = s->last_picture.data[0];
458 srcU = s->last_picture.data[1];
459 srcV = s->last_picture.data[2];
461 srcY = s->next_picture.data[0];
462 srcU = s->next_picture.data[1];
463 srcV = s->next_picture.data[2];
466 src_x = s->mb_x * 16 + (mx >> 2);
467 src_y = s->mb_y * 16 + (my >> 2);
468 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
469 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
471 if(v->profile != PROFILE_ADVANCED){
472 src_x = av_clip( src_x, -16, s->mb_width * 16);
473 src_y = av_clip( src_y, -16, s->mb_height * 16);
474 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
475 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
477 src_x = av_clip( src_x, -17, s->avctx->coded_width);
478 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
479 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
480 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
483 srcY += src_y * s->linesize + src_x;
484 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
485 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
487 /* for grayscale we should not try to read from unknown area */
488 if(s->flags & CODEC_FLAG_GRAY) {
489 srcU = s->edge_emu_buffer + 18 * s->linesize;
490 srcV = s->edge_emu_buffer + 18 * s->linesize;
493 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
494 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3
495 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){
496 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
498 srcY -= s->mspel * (1 + s->linesize);
499 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
500 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
501 srcY = s->edge_emu_buffer;
502 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
503 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
504 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
505 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
508 /* if we deal with range reduction we need to scale source blocks */
514 for(j = 0; j < 17 + s->mspel*2; j++) {
515 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
518 src = srcU; src2 = srcV;
519 for(j = 0; j < 9; j++) {
520 for(i = 0; i < 9; i++) {
521 src[i] = ((src[i] - 128) >> 1) + 128;
522 src2[i] = ((src2[i] - 128) >> 1) + 128;
524 src += s->uvlinesize;
525 src2 += s->uvlinesize;
528 /* if we deal with intensity compensation we need to scale source blocks */
529 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
534 for(j = 0; j < 17 + s->mspel*2; j++) {
535 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];
538 src = srcU; src2 = srcV;
539 for(j = 0; j < 9; j++) {
540 for(i = 0; i < 9; i++) {
541 src[i] = v->lutuv[src[i]];
542 src2[i] = v->lutuv[src2[i]];
544 src += s->uvlinesize;
545 src2 += s->uvlinesize;
548 srcY += s->mspel * (1 + s->linesize);
552 dxy = ((my & 3) << 2) | (mx & 3);
553 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] , srcY , s->linesize, v->rnd);
554 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
555 srcY += s->linesize * 8;
556 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize , srcY , s->linesize, v->rnd);
557 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
558 } else { // hpel mc - always used for luma
559 dxy = (my & 2) | ((mx & 2) >> 1);
562 dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
564 dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
567 if(s->flags & CODEC_FLAG_GRAY) return;
568 /* Chroma MC always uses qpel bilinear */
572 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
573 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
575 dsp->put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
576 dsp->put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
580 /** Do motion compensation for 4-MV macroblock - luminance block
582 static void vc1_mc_4mv_luma(VC1Context *v, int n)
584 MpegEncContext *s = &v->s;
585 DSPContext *dsp = &v->s.dsp;
587 int dxy, mx, my, src_x, src_y;
590 if(!v->s.last_picture.data[0])return;
593 srcY = s->last_picture.data[0];
595 off = s->linesize * 4 * (n&2) + (n&1) * 8;
597 src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);
598 src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);
600 if(v->profile != PROFILE_ADVANCED){
601 src_x = av_clip( src_x, -16, s->mb_width * 16);
602 src_y = av_clip( src_y, -16, s->mb_height * 16);
604 src_x = av_clip( src_x, -17, s->avctx->coded_width);
605 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
608 srcY += src_y * s->linesize + src_x;
610 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
611 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2
612 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){
613 srcY -= s->mspel * (1 + s->linesize);
614 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,
615 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
616 srcY = s->edge_emu_buffer;
617 /* if we deal with range reduction we need to scale source blocks */
623 for(j = 0; j < 9 + s->mspel*2; j++) {
624 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
628 /* if we deal with intensity compensation we need to scale source blocks */
629 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
634 for(j = 0; j < 9 + s->mspel*2; j++) {
635 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];
639 srcY += s->mspel * (1 + s->linesize);
643 dxy = ((my & 3) << 2) | (mx & 3);
644 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);
645 } else { // hpel mc - always used for luma
646 dxy = (my & 2) | ((mx & 2) >> 1);
648 dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
650 dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
654 static inline int median4(int a, int b, int c, int d)
657 if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
658 else return (FFMIN(b, c) + FFMAX(a, d)) / 2;
660 if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
661 else return (FFMIN(a, c) + FFMAX(b, d)) / 2;
666 /** Do motion compensation for 4-MV macroblock - both chroma blocks
668 static void vc1_mc_4mv_chroma(VC1Context *v)
670 MpegEncContext *s = &v->s;
671 DSPContext *dsp = &v->s.dsp;
672 uint8_t *srcU, *srcV;
673 int uvmx, uvmy, uvsrc_x, uvsrc_y;
674 int i, idx, tx = 0, ty = 0;
675 int mvx[4], mvy[4], intra[4];
676 static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
678 if(!v->s.last_picture.data[0])return;
679 if(s->flags & CODEC_FLAG_GRAY) return;
681 for(i = 0; i < 4; i++) {
682 mvx[i] = s->mv[0][i][0];
683 mvy[i] = s->mv[0][i][1];
684 intra[i] = v->mb_type[0][s->block_index[i]];
687 /* calculate chroma MV vector from four luma MVs */
688 idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];
689 if(!idx) { // all blocks are inter
690 tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
691 ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
692 } else if(count[idx] == 1) { // 3 inter blocks
695 tx = mid_pred(mvx[1], mvx[2], mvx[3]);
696 ty = mid_pred(mvy[1], mvy[2], mvy[3]);
699 tx = mid_pred(mvx[0], mvx[2], mvx[3]);
700 ty = mid_pred(mvy[0], mvy[2], mvy[3]);
703 tx = mid_pred(mvx[0], mvx[1], mvx[3]);
704 ty = mid_pred(mvy[0], mvy[1], mvy[3]);
707 tx = mid_pred(mvx[0], mvx[1], mvx[2]);
708 ty = mid_pred(mvy[0], mvy[1], mvy[2]);
711 } else if(count[idx] == 2) {
713 for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}
714 for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}
715 tx = (mvx[t1] + mvx[t2]) / 2;
716 ty = (mvy[t1] + mvy[t2]) / 2;
718 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
719 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
720 return; //no need to do MC for inter blocks
723 s->current_picture.motion_val[1][s->block_index[0]][0] = tx;
724 s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
725 uvmx = (tx + ((tx&3) == 3)) >> 1;
726 uvmy = (ty + ((ty&3) == 3)) >> 1;
728 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
729 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
732 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
733 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
735 if(v->profile != PROFILE_ADVANCED){
736 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
737 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
739 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
740 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
743 srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
744 srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
745 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
746 || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
747 || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){
748 ff_emulated_edge_mc(s->edge_emu_buffer , srcU, s->uvlinesize, 8+1, 8+1,
749 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
750 ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,
751 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
752 srcU = s->edge_emu_buffer;
753 srcV = s->edge_emu_buffer + 16;
755 /* if we deal with range reduction we need to scale source blocks */
760 src = srcU; src2 = srcV;
761 for(j = 0; j < 9; j++) {
762 for(i = 0; i < 9; i++) {
763 src[i] = ((src[i] - 128) >> 1) + 128;
764 src2[i] = ((src2[i] - 128) >> 1) + 128;
766 src += s->uvlinesize;
767 src2 += s->uvlinesize;
770 /* if we deal with intensity compensation we need to scale source blocks */
771 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
775 src = srcU; src2 = srcV;
776 for(j = 0; j < 9; j++) {
777 for(i = 0; i < 9; i++) {
778 src[i] = v->lutuv[src[i]];
779 src2[i] = v->lutuv[src2[i]];
781 src += s->uvlinesize;
782 src2 += s->uvlinesize;
787 /* Chroma MC always uses qpel bilinear */
791 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
792 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
794 dsp->put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
795 dsp->put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
799 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);
802 * Decode Simple/Main Profiles sequence header
803 * @see Figure 7-8, p16-17
804 * @param avctx Codec context
805 * @param gb GetBit context initialized from Codec context extra_data
808 static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)
810 VC1Context *v = avctx->priv_data;
812 av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));
813 v->profile = get_bits(gb, 2);
814 if (v->profile == PROFILE_COMPLEX)
816 av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");
819 if (v->profile == PROFILE_ADVANCED)
821 v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz;
822 v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz;
823 return decode_sequence_header_adv(v, gb);
827 v->zz_8x4 = wmv2_scantableA;
828 v->zz_4x8 = wmv2_scantableB;
829 v->res_sm = get_bits(gb, 2); //reserved
832 av_log(avctx, AV_LOG_ERROR,
833 "Reserved RES_SM=%i is forbidden\n", v->res_sm);
839 v->frmrtq_postproc = get_bits(gb, 3); //common
840 // (bitrate-32kbps)/64kbps
841 v->bitrtq_postproc = get_bits(gb, 5); //common
842 v->s.loop_filter = get_bits1(gb); //common
843 if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)
845 av_log(avctx, AV_LOG_ERROR,
846 "LOOPFILTER shell not be enabled in simple profile\n");
848 if(v->s.avctx->skip_loop_filter >= AVDISCARD_ALL)
849 v->s.loop_filter = 0;
851 v->res_x8 = get_bits1(gb); //reserved
852 v->multires = get_bits1(gb);
853 v->res_fasttx = get_bits1(gb);
856 v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct;
857 v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;
858 v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;
859 v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;
862 v->fastuvmc = get_bits1(gb); //common
863 if (!v->profile && !v->fastuvmc)
865 av_log(avctx, AV_LOG_ERROR,
866 "FASTUVMC unavailable in Simple Profile\n");
869 v->extended_mv = get_bits1(gb); //common
870 if (!v->profile && v->extended_mv)
872 av_log(avctx, AV_LOG_ERROR,
873 "Extended MVs unavailable in Simple Profile\n");
876 v->dquant = get_bits(gb, 2); //common
877 v->vstransform = get_bits1(gb); //common
879 v->res_transtab = get_bits1(gb);
882 av_log(avctx, AV_LOG_ERROR,
883 "1 for reserved RES_TRANSTAB is forbidden\n");
887 v->overlap = get_bits1(gb); //common
889 v->s.resync_marker = get_bits1(gb);
890 v->rangered = get_bits1(gb);
891 if (v->rangered && v->profile == PROFILE_SIMPLE)
893 av_log(avctx, AV_LOG_INFO,
894 "RANGERED should be set to 0 in simple profile\n");
897 v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common
898 v->quantizer_mode = get_bits(gb, 2); //common
900 v->finterpflag = get_bits1(gb); //common
901 v->res_rtm_flag = get_bits1(gb); //reserved
902 if (!v->res_rtm_flag)
904 // av_log(avctx, AV_LOG_ERROR,
905 // "0 for reserved RES_RTM_FLAG is forbidden\n");
906 av_log(avctx, AV_LOG_ERROR,
907 "Old WMV3 version detected, only I-frames will be decoded\n");
910 //TODO: figure out what they mean (always 0x402F)
911 if(!v->res_fasttx) skip_bits(gb, 16);
912 av_log(avctx, AV_LOG_DEBUG,
913 "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
914 "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"
915 "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"
916 "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",
917 v->profile, v->frmrtq_postproc, v->bitrtq_postproc,
918 v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,
919 v->rangered, v->vstransform, v->overlap, v->s.resync_marker,
920 v->dquant, v->quantizer_mode, avctx->max_b_frames
925 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)
928 v->level = get_bits(gb, 3);
931 av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);
933 v->chromaformat = get_bits(gb, 2);
934 if (v->chromaformat != 1)
936 av_log(v->s.avctx, AV_LOG_ERROR,
937 "Only 4:2:0 chroma format supported\n");
942 v->frmrtq_postproc = get_bits(gb, 3); //common
943 // (bitrate-32kbps)/64kbps
944 v->bitrtq_postproc = get_bits(gb, 5); //common
945 v->postprocflag = get_bits1(gb); //common
947 v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;
948 v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;
949 v->s.avctx->width = v->s.avctx->coded_width;
950 v->s.avctx->height = v->s.avctx->coded_height;
951 v->broadcast = get_bits1(gb);
952 v->interlace = get_bits1(gb);
953 v->tfcntrflag = get_bits1(gb);
954 v->finterpflag = get_bits1(gb);
955 skip_bits1(gb); // reserved
957 v->s.h_edge_pos = v->s.avctx->coded_width;
958 v->s.v_edge_pos = v->s.avctx->coded_height;
960 av_log(v->s.avctx, AV_LOG_DEBUG,
961 "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
962 "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"
963 "TFCTRflag=%i, FINTERPflag=%i\n",
964 v->level, v->frmrtq_postproc, v->bitrtq_postproc,
965 v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,
966 v->tfcntrflag, v->finterpflag
969 v->psf = get_bits1(gb);
970 if(v->psf) { //PsF, 6.1.13
971 av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");
974 v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
975 if(get_bits1(gb)) { //Display Info - decoding is not affected by it
977 av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
978 v->s.avctx->width = w = get_bits(gb, 14) + 1;
979 v->s.avctx->height = h = get_bits(gb, 14) + 1;
980 av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
982 ar = get_bits(gb, 4);
984 v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
988 v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
990 av_log(v->s.avctx, AV_LOG_DEBUG, "Aspect: %i:%i\n", v->s.avctx->sample_aspect_ratio.num, v->s.avctx->sample_aspect_ratio.den);
992 if(get_bits1(gb)){ //framerate stuff
994 v->s.avctx->time_base.num = 32;
995 v->s.avctx->time_base.den = get_bits(gb, 16) + 1;
998 nr = get_bits(gb, 8);
999 dr = get_bits(gb, 4);
1000 if(nr && nr < 8 && dr && dr < 3){
1001 v->s.avctx->time_base.num = ff_vc1_fps_dr[dr - 1];
1002 v->s.avctx->time_base.den = ff_vc1_fps_nr[nr - 1] * 1000;
1008 v->color_prim = get_bits(gb, 8);
1009 v->transfer_char = get_bits(gb, 8);
1010 v->matrix_coef = get_bits(gb, 8);
1014 v->hrd_param_flag = get_bits1(gb);
1015 if(v->hrd_param_flag) {
1017 v->hrd_num_leaky_buckets = get_bits(gb, 5);
1018 skip_bits(gb, 4); //bitrate exponent
1019 skip_bits(gb, 4); //buffer size exponent
1020 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1021 skip_bits(gb, 16); //hrd_rate[n]
1022 skip_bits(gb, 16); //hrd_buffer[n]
1028 static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb)
1030 VC1Context *v = avctx->priv_data;
1033 av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32));
1034 v->broken_link = get_bits1(gb);
1035 v->closed_entry = get_bits1(gb);
1036 v->panscanflag = get_bits1(gb);
1037 v->refdist_flag = get_bits1(gb);
1038 v->s.loop_filter = get_bits1(gb);
1039 v->fastuvmc = get_bits1(gb);
1040 v->extended_mv = get_bits1(gb);
1041 v->dquant = get_bits(gb, 2);
1042 v->vstransform = get_bits1(gb);
1043 v->overlap = get_bits1(gb);
1044 v->quantizer_mode = get_bits(gb, 2);
1046 if(v->hrd_param_flag){
1047 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1048 skip_bits(gb, 8); //hrd_full[n]
1053 avctx->coded_width = (get_bits(gb, 12)+1)<<1;
1054 avctx->coded_height = (get_bits(gb, 12)+1)<<1;
1057 v->extended_dmv = get_bits1(gb);
1058 if((v->range_mapy_flag = get_bits1(gb))) {
1059 av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n");
1060 v->range_mapy = get_bits(gb, 3);
1062 if((v->range_mapuv_flag = get_bits1(gb))) {
1063 av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n");
1064 v->range_mapuv = get_bits(gb, 3);
1067 av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n"
1068 "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n"
1069 "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n"
1070 "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n",
1071 v->broken_link, v->closed_entry, v->panscanflag, v->refdist_flag, v->s.loop_filter,
1072 v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode);
1077 static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)
1079 int pqindex, lowquant, status;
1081 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1082 skip_bits(gb, 2); //framecnt unused
1084 if (v->rangered) v->rangeredfrm = get_bits1(gb);
1085 v->s.pict_type = get_bits1(gb);
1086 if (v->s.avctx->max_b_frames) {
1087 if (!v->s.pict_type) {
1088 if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE;
1089 else v->s.pict_type = FF_B_TYPE;
1090 } else v->s.pict_type = FF_P_TYPE;
1091 } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE;
1094 if(v->s.pict_type == FF_B_TYPE) {
1095 v->bfraction_lut_index = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1096 v->bfraction = ff_vc1_bfraction_lut[v->bfraction_lut_index];
1097 if(v->bfraction == 0) {
1098 v->s.pict_type = FF_BI_TYPE;
1101 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1102 skip_bits(gb, 7); // skip buffer fullness
1105 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1107 if(v->s.pict_type == FF_P_TYPE)
1110 /* Quantizer stuff */
1111 pqindex = get_bits(gb, 5);
1112 if(!pqindex) return -1;
1113 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1114 v->pq = ff_vc1_pquant_table[0][pqindex];
1116 v->pq = ff_vc1_pquant_table[1][pqindex];
1119 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1120 v->pquantizer = pqindex < 9;
1121 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1123 v->pqindex = pqindex;
1124 if (pqindex < 9) v->halfpq = get_bits1(gb);
1126 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1127 v->pquantizer = get_bits1(gb);
1129 if (v->extended_mv == 1) v->mvrange = get_unary(gb, 0, 3);
1130 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1131 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1132 v->range_x = 1 << (v->k_x - 1);
1133 v->range_y = 1 << (v->k_y - 1);
1134 if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2);
1136 if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){
1137 v->x8_type = get_bits1(gb);
1138 }else v->x8_type = 0;
1139 //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",
1140 // (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);
1142 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1144 switch(v->s.pict_type) {
1146 if (v->pq < 5) v->tt_index = 0;
1147 else if(v->pq < 13) v->tt_index = 1;
1148 else v->tt_index = 2;
1150 lowquant = (v->pq > 12) ? 0 : 1;
1151 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1152 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1154 int scale, shift, i;
1155 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1156 v->lumscale = get_bits(gb, 6);
1157 v->lumshift = get_bits(gb, 6);
1159 /* fill lookup tables for intensity compensation */
1162 shift = (255 - v->lumshift * 2) << 6;
1163 if(v->lumshift > 31)
1166 scale = v->lumscale + 32;
1167 if(v->lumshift > 31)
1168 shift = (v->lumshift - 64) << 6;
1170 shift = v->lumshift << 6;
1172 for(i = 0; i < 256; i++) {
1173 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1174 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1177 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1178 v->s.quarter_sample = 0;
1179 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1180 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1181 v->s.quarter_sample = 0;
1183 v->s.quarter_sample = 1;
1185 v->s.quarter_sample = 1;
1186 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));
1188 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1189 v->mv_mode2 == MV_PMODE_MIXED_MV)
1190 || v->mv_mode == MV_PMODE_MIXED_MV)
1192 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1193 if (status < 0) return -1;
1194 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1195 "Imode: %i, Invert: %i\n", status>>1, status&1);
1197 v->mv_type_is_raw = 0;
1198 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1200 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1201 if (status < 0) return -1;
1202 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1203 "Imode: %i, Invert: %i\n", status>>1, status&1);
1205 /* Hopefully this is correct for P frames */
1206 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1207 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1211 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1212 vop_dquant_decoding(v);
1215 v->ttfrm = 0; //FIXME Is that so ?
1218 v->ttmbf = get_bits1(gb);
1221 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1229 if (v->pq < 5) v->tt_index = 0;
1230 else if(v->pq < 13) v->tt_index = 1;
1231 else v->tt_index = 2;
1233 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1234 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1235 v->s.mspel = v->s.quarter_sample;
1237 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1238 if (status < 0) return -1;
1239 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1240 "Imode: %i, Invert: %i\n", status>>1, status&1);
1241 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1242 if (status < 0) return -1;
1243 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1244 "Imode: %i, Invert: %i\n", status>>1, status&1);
1246 v->s.mv_table_index = get_bits(gb, 2);
1247 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1251 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1252 vop_dquant_decoding(v);
1258 v->ttmbf = get_bits1(gb);
1261 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1273 v->c_ac_table_index = decode012(gb);
1274 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1276 v->y_ac_table_index = decode012(gb);
1279 v->s.dc_table_index = get_bits1(gb);
1282 if(v->s.pict_type == FF_BI_TYPE) {
1283 v->s.pict_type = FF_B_TYPE;
1289 static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)
1291 int pqindex, lowquant;
1294 v->p_frame_skipped = 0;
1297 v->fcm = decode012(gb);
1298 if(v->fcm) return -1; // interlaced frames/fields are not implemented
1300 switch(get_unary(gb, 0, 4)) {
1302 v->s.pict_type = FF_P_TYPE;
1305 v->s.pict_type = FF_B_TYPE;
1308 v->s.pict_type = FF_I_TYPE;
1311 v->s.pict_type = FF_BI_TYPE;
1314 v->s.pict_type = FF_P_TYPE; // skipped pic
1315 v->p_frame_skipped = 1;
1321 if(!v->interlace || v->psf) {
1322 v->rptfrm = get_bits(gb, 2);
1324 v->tff = get_bits1(gb);
1325 v->rptfrm = get_bits1(gb);
1328 if(v->panscanflag) {
1331 v->rnd = get_bits1(gb);
1333 v->uvsamp = get_bits1(gb);
1334 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1335 if(v->s.pict_type == FF_B_TYPE) {
1336 v->bfraction_lut_index = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1337 v->bfraction = ff_vc1_bfraction_lut[v->bfraction_lut_index];
1338 if(v->bfraction == 0) {
1339 v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */
1342 pqindex = get_bits(gb, 5);
1343 if(!pqindex) return -1;
1344 v->pqindex = pqindex;
1345 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1346 v->pq = ff_vc1_pquant_table[0][pqindex];
1348 v->pq = ff_vc1_pquant_table[1][pqindex];
1351 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1352 v->pquantizer = pqindex < 9;
1353 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1355 v->pqindex = pqindex;
1356 if (pqindex < 9) v->halfpq = get_bits1(gb);
1358 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1359 v->pquantizer = get_bits1(gb);
1361 v->postproc = get_bits(gb, 2);
1363 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1365 switch(v->s.pict_type) {
1368 status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v);
1369 if (status < 0) return -1;
1370 av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: "
1371 "Imode: %i, Invert: %i\n", status>>1, status&1);
1372 v->condover = CONDOVER_NONE;
1373 if(v->overlap && v->pq <= 8) {
1374 v->condover = decode012(gb);
1375 if(v->condover == CONDOVER_SELECT) {
1376 status = bitplane_decoding(v->over_flags_plane, &v->overflg_is_raw, v);
1377 if (status < 0) return -1;
1378 av_log(v->s.avctx, AV_LOG_DEBUG, "CONDOVER plane encoding: "
1379 "Imode: %i, Invert: %i\n", status>>1, status&1);
1384 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1385 else v->mvrange = 0;
1386 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1387 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1388 v->range_x = 1 << (v->k_x - 1);
1389 v->range_y = 1 << (v->k_y - 1);
1391 if (v->pq < 5) v->tt_index = 0;
1392 else if(v->pq < 13) v->tt_index = 1;
1393 else v->tt_index = 2;
1395 lowquant = (v->pq > 12) ? 0 : 1;
1396 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1397 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1399 int scale, shift, i;
1400 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1401 v->lumscale = get_bits(gb, 6);
1402 v->lumshift = get_bits(gb, 6);
1403 /* fill lookup tables for intensity compensation */
1406 shift = (255 - v->lumshift * 2) << 6;
1407 if(v->lumshift > 31)
1410 scale = v->lumscale + 32;
1411 if(v->lumshift > 31)
1412 shift = (v->lumshift - 64) << 6;
1414 shift = v->lumshift << 6;
1416 for(i = 0; i < 256; i++) {
1417 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1418 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1422 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1423 v->s.quarter_sample = 0;
1424 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1425 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1426 v->s.quarter_sample = 0;
1428 v->s.quarter_sample = 1;
1430 v->s.quarter_sample = 1;
1431 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));
1433 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1434 v->mv_mode2 == MV_PMODE_MIXED_MV)
1435 || v->mv_mode == MV_PMODE_MIXED_MV)
1437 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1438 if (status < 0) return -1;
1439 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1440 "Imode: %i, Invert: %i\n", status>>1, status&1);
1442 v->mv_type_is_raw = 0;
1443 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1445 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1446 if (status < 0) return -1;
1447 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1448 "Imode: %i, Invert: %i\n", status>>1, status&1);
1450 /* Hopefully this is correct for P frames */
1451 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1452 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1455 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1456 vop_dquant_decoding(v);
1459 v->ttfrm = 0; //FIXME Is that so ?
1462 v->ttmbf = get_bits1(gb);
1465 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1473 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1474 else v->mvrange = 0;
1475 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1476 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1477 v->range_x = 1 << (v->k_x - 1);
1478 v->range_y = 1 << (v->k_y - 1);
1480 if (v->pq < 5) v->tt_index = 0;
1481 else if(v->pq < 13) v->tt_index = 1;
1482 else v->tt_index = 2;
1484 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1485 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1486 v->s.mspel = v->s.quarter_sample;
1488 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1489 if (status < 0) return -1;
1490 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1491 "Imode: %i, Invert: %i\n", status>>1, status&1);
1492 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1493 if (status < 0) return -1;
1494 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1495 "Imode: %i, Invert: %i\n", status>>1, status&1);
1497 v->s.mv_table_index = get_bits(gb, 2);
1498 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1502 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1503 vop_dquant_decoding(v);
1509 v->ttmbf = get_bits1(gb);
1512 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1522 v->c_ac_table_index = decode012(gb);
1523 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1525 v->y_ac_table_index = decode012(gb);
1528 v->s.dc_table_index = get_bits1(gb);
1529 if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) {
1530 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1531 vop_dquant_decoding(v);
1535 if(v->s.pict_type == FF_BI_TYPE) {
1536 v->s.pict_type = FF_B_TYPE;
1542 /***********************************************************************/
1544 * @defgroup vc1block VC-1 Block-level functions
1545 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
1551 * @brief Get macroblock-level quantizer scale
1553 #define GET_MQUANT() \
1557 if (v->dqprofile == DQPROFILE_ALL_MBS) \
1561 mquant = (get_bits1(gb)) ? v->altpq : v->pq; \
1565 mqdiff = get_bits(gb, 3); \
1566 if (mqdiff != 7) mquant = v->pq + mqdiff; \
1567 else mquant = get_bits(gb, 5); \
1570 if(v->dqprofile == DQPROFILE_SINGLE_EDGE) \
1571 edges = 1 << v->dqsbedge; \
1572 else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES) \
1573 edges = (3 << v->dqsbedge) % 15; \
1574 else if(v->dqprofile == DQPROFILE_FOUR_EDGES) \
1576 if((edges&1) && !s->mb_x) \
1577 mquant = v->altpq; \
1578 if((edges&2) && s->first_slice_line) \
1579 mquant = v->altpq; \
1580 if((edges&4) && s->mb_x == (s->mb_width - 1)) \
1581 mquant = v->altpq; \
1582 if((edges&8) && s->mb_y == (s->mb_height - 1)) \
1583 mquant = v->altpq; \
1587 * @def GET_MVDATA(_dmv_x, _dmv_y)
1588 * @brief Get MV differentials
1589 * @see MVDATA decoding from 8.3.5.2, p(1)20
1590 * @param _dmv_x Horizontal differential for decoded MV
1591 * @param _dmv_y Vertical differential for decoded MV
1593 #define GET_MVDATA(_dmv_x, _dmv_y) \
1594 index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table,\
1595 VC1_MV_DIFF_VLC_BITS, 2); \
1598 mb_has_coeffs = 1; \
1601 else mb_has_coeffs = 0; \
1603 if (!index) { _dmv_x = _dmv_y = 0; } \
1604 else if (index == 35) \
1606 _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample); \
1607 _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample); \
1609 else if (index == 36) \
1618 if (!s->quarter_sample && index1 == 5) val = 1; \
1620 if(size_table[index1] - val > 0) \
1621 val = get_bits(gb, size_table[index1] - val); \
1623 sign = 0 - (val&1); \
1624 _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1627 if (!s->quarter_sample && index1 == 5) val = 1; \
1629 if(size_table[index1] - val > 0) \
1630 val = get_bits(gb, size_table[index1] - val); \
1632 sign = 0 - (val&1); \
1633 _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1636 /** Predict and set motion vector
1638 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)
1640 int xy, wrap, off = 0;
1645 /* scale MV difference to be quad-pel */
1646 dmv_x <<= 1 - s->quarter_sample;
1647 dmv_y <<= 1 - s->quarter_sample;
1649 wrap = s->b8_stride;
1650 xy = s->block_index[n];
1653 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
1654 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
1655 s->current_picture.motion_val[1][xy][0] = 0;
1656 s->current_picture.motion_val[1][xy][1] = 0;
1657 if(mv1) { /* duplicate motion data for 1-MV block */
1658 s->current_picture.motion_val[0][xy + 1][0] = 0;
1659 s->current_picture.motion_val[0][xy + 1][1] = 0;
1660 s->current_picture.motion_val[0][xy + wrap][0] = 0;
1661 s->current_picture.motion_val[0][xy + wrap][1] = 0;
1662 s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
1663 s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
1664 s->current_picture.motion_val[1][xy + 1][0] = 0;
1665 s->current_picture.motion_val[1][xy + 1][1] = 0;
1666 s->current_picture.motion_val[1][xy + wrap][0] = 0;
1667 s->current_picture.motion_val[1][xy + wrap][1] = 0;
1668 s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
1669 s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
1674 C = s->current_picture.motion_val[0][xy - 1];
1675 A = s->current_picture.motion_val[0][xy - wrap];
1677 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
1679 //in 4-MV mode different blocks have different B predictor position
1682 off = (s->mb_x > 0) ? -1 : 1;
1685 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
1694 B = s->current_picture.motion_val[0][xy - wrap + off];
1696 if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds
1697 if(s->mb_width == 1) {
1701 px = mid_pred(A[0], B[0], C[0]);
1702 py = mid_pred(A[1], B[1], C[1]);
1704 } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds
1710 /* Pullback MV as specified in 8.3.5.3.4 */
1713 qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);
1714 qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);
1715 X = (s->mb_width << 6) - 4;
1716 Y = (s->mb_height << 6) - 4;
1718 if(qx + px < -60) px = -60 - qx;
1719 if(qy + py < -60) py = -60 - qy;
1721 if(qx + px < -28) px = -28 - qx;
1722 if(qy + py < -28) py = -28 - qy;
1724 if(qx + px > X) px = X - qx;
1725 if(qy + py > Y) py = Y - qy;
1727 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
1728 if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {
1729 if(is_intra[xy - wrap])
1730 sum = FFABS(px) + FFABS(py);
1732 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
1734 if(get_bits1(&s->gb)) {
1742 if(is_intra[xy - 1])
1743 sum = FFABS(px) + FFABS(py);
1745 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
1747 if(get_bits1(&s->gb)) {
1757 /* store MV using signed modulus of MV range defined in 4.11 */
1758 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
1759 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;
1760 if(mv1) { /* duplicate motion data for 1-MV block */
1761 s->current_picture.motion_val[0][xy + 1][0] = s->current_picture.motion_val[0][xy][0];
1762 s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];
1763 s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];
1764 s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];
1765 s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];
1766 s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];
1770 /** Motion compensation for direct or interpolated blocks in B-frames
1772 static void vc1_interp_mc(VC1Context *v)
1774 MpegEncContext *s = &v->s;
1775 DSPContext *dsp = &v->s.dsp;
1776 uint8_t *srcY, *srcU, *srcV;
1777 int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
1779 if(!v->s.next_picture.data[0])return;
1781 mx = s->mv[1][0][0];
1782 my = s->mv[1][0][1];
1783 uvmx = (mx + ((mx & 3) == 3)) >> 1;
1784 uvmy = (my + ((my & 3) == 3)) >> 1;
1786 uvmx = uvmx + ((uvmx<0)?-(uvmx&1):(uvmx&1));
1787 uvmy = uvmy + ((uvmy<0)?-(uvmy&1):(uvmy&1));
1789 srcY = s->next_picture.data[0];
1790 srcU = s->next_picture.data[1];
1791 srcV = s->next_picture.data[2];
1793 src_x = s->mb_x * 16 + (mx >> 2);
1794 src_y = s->mb_y * 16 + (my >> 2);
1795 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
1796 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
1798 if(v->profile != PROFILE_ADVANCED){
1799 src_x = av_clip( src_x, -16, s->mb_width * 16);
1800 src_y = av_clip( src_y, -16, s->mb_height * 16);
1801 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
1802 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
1804 src_x = av_clip( src_x, -17, s->avctx->coded_width);
1805 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
1806 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
1807 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
1810 srcY += src_y * s->linesize + src_x;
1811 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
1812 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
1814 /* for grayscale we should not try to read from unknown area */
1815 if(s->flags & CODEC_FLAG_GRAY) {
1816 srcU = s->edge_emu_buffer + 18 * s->linesize;
1817 srcV = s->edge_emu_buffer + 18 * s->linesize;
1821 || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16
1822 || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){
1823 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
1825 srcY -= s->mspel * (1 + s->linesize);
1826 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
1827 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
1828 srcY = s->edge_emu_buffer;
1829 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
1830 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1831 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
1832 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1835 /* if we deal with range reduction we need to scale source blocks */
1836 if(v->rangeredfrm) {
1838 uint8_t *src, *src2;
1841 for(j = 0; j < 17 + s->mspel*2; j++) {
1842 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
1845 src = srcU; src2 = srcV;
1846 for(j = 0; j < 9; j++) {
1847 for(i = 0; i < 9; i++) {
1848 src[i] = ((src[i] - 128) >> 1) + 128;
1849 src2[i] = ((src2[i] - 128) >> 1) + 128;
1851 src += s->uvlinesize;
1852 src2 += s->uvlinesize;
1855 srcY += s->mspel * (1 + s->linesize);
1859 dxy = ((my & 3) << 2) | (mx & 3);
1860 dsp->avg_vc1_mspel_pixels_tab[dxy](s->dest[0] , srcY , s->linesize, v->rnd);
1861 dsp->avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
1862 srcY += s->linesize * 8;
1863 dsp->avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize , srcY , s->linesize, v->rnd);
1864 dsp->avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
1866 dxy = (my & 2) | ((mx & 2) >> 1);
1869 dsp->avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
1871 dsp->avg_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
1874 if(s->flags & CODEC_FLAG_GRAY) return;
1875 /* Chroma MC always uses qpel blilinear */
1879 dsp->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
1880 dsp->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
1882 dsp->avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
1883 dsp->avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
1887 static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)
1891 #if B_FRACTION_DEN==256
1895 return 2 * ((value * n + 255) >> 9);
1896 return (value * n + 128) >> 8;
1899 n -= B_FRACTION_DEN;
1901 return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));
1902 return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;
1906 /** Reconstruct motion vector for B-frame and do motion compensation
1908 static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode)
1911 v->mv_mode2 = v->mv_mode;
1912 v->mv_mode = MV_PMODE_INTENSITY_COMP;
1917 if(v->use_ic) v->mv_mode = v->mv_mode2;
1920 if(mode == BMV_TYPE_INTERPOLATED) {
1923 if(v->use_ic) v->mv_mode = v->mv_mode2;
1927 if(v->use_ic && (mode == BMV_TYPE_BACKWARD)) v->mv_mode = v->mv_mode2;
1928 vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));
1929 if(v->use_ic) v->mv_mode = v->mv_mode2;
1932 static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mvtype)
1934 MpegEncContext *s = &v->s;
1935 int xy, wrap, off = 0;
1940 const uint8_t *is_intra = v->mb_type[0];
1944 /* scale MV difference to be quad-pel */
1945 dmv_x[0] <<= 1 - s->quarter_sample;
1946 dmv_y[0] <<= 1 - s->quarter_sample;
1947 dmv_x[1] <<= 1 - s->quarter_sample;
1948 dmv_y[1] <<= 1 - s->quarter_sample;
1950 wrap = s->b8_stride;
1951 xy = s->block_index[0];
1954 s->current_picture.motion_val[0][xy][0] =
1955 s->current_picture.motion_val[0][xy][1] =
1956 s->current_picture.motion_val[1][xy][0] =
1957 s->current_picture.motion_val[1][xy][1] = 0;
1960 s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);
1961 s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);
1962 s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);
1963 s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);
1965 /* Pullback predicted motion vectors as specified in 8.4.5.4 */
1966 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));
1967 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));
1968 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));
1969 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));
1971 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
1972 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
1973 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
1974 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
1978 if((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
1979 C = s->current_picture.motion_val[0][xy - 2];
1980 A = s->current_picture.motion_val[0][xy - wrap*2];
1981 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
1982 B = s->current_picture.motion_val[0][xy - wrap*2 + off];
1984 if(!s->mb_x) C[0] = C[1] = 0;
1985 if(!s->first_slice_line) { // predictor A is not out of bounds
1986 if(s->mb_width == 1) {
1990 px = mid_pred(A[0], B[0], C[0]);
1991 py = mid_pred(A[1], B[1], C[1]);
1993 } else if(s->mb_x) { // predictor C is not out of bounds
1999 /* Pullback MV as specified in 8.3.5.3.4 */
2002 if(v->profile < PROFILE_ADVANCED) {
2003 qx = (s->mb_x << 5);
2004 qy = (s->mb_y << 5);
2005 X = (s->mb_width << 5) - 4;
2006 Y = (s->mb_height << 5) - 4;
2007 if(qx + px < -28) px = -28 - qx;
2008 if(qy + py < -28) py = -28 - qy;
2009 if(qx + px > X) px = X - qx;
2010 if(qy + py > Y) py = Y - qy;
2012 qx = (s->mb_x << 6);
2013 qy = (s->mb_y << 6);
2014 X = (s->mb_width << 6) - 4;
2015 Y = (s->mb_height << 6) - 4;
2016 if(qx + px < -60) px = -60 - qx;
2017 if(qy + py < -60) py = -60 - qy;
2018 if(qx + px > X) px = X - qx;
2019 if(qy + py > Y) py = Y - qy;
2022 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2023 if(0 && !s->first_slice_line && s->mb_x) {
2024 if(is_intra[xy - wrap])
2025 sum = FFABS(px) + FFABS(py);
2027 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2029 if(get_bits1(&s->gb)) {
2037 if(is_intra[xy - 2])
2038 sum = FFABS(px) + FFABS(py);
2040 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2042 if(get_bits1(&s->gb)) {
2052 /* store MV using signed modulus of MV range defined in 4.11 */
2053 s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;
2054 s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;
2056 if((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2057 C = s->current_picture.motion_val[1][xy - 2];
2058 A = s->current_picture.motion_val[1][xy - wrap*2];
2059 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2060 B = s->current_picture.motion_val[1][xy - wrap*2 + off];
2062 if(!s->mb_x) C[0] = C[1] = 0;
2063 if(!s->first_slice_line) { // predictor A is not out of bounds
2064 if(s->mb_width == 1) {
2068 px = mid_pred(A[0], B[0], C[0]);
2069 py = mid_pred(A[1], B[1], C[1]);
2071 } else if(s->mb_x) { // predictor C is not out of bounds
2077 /* Pullback MV as specified in 8.3.5.3.4 */
2080 if(v->profile < PROFILE_ADVANCED) {
2081 qx = (s->mb_x << 5);
2082 qy = (s->mb_y << 5);
2083 X = (s->mb_width << 5) - 4;
2084 Y = (s->mb_height << 5) - 4;
2085 if(qx + px < -28) px = -28 - qx;
2086 if(qy + py < -28) py = -28 - qy;
2087 if(qx + px > X) px = X - qx;
2088 if(qy + py > Y) py = Y - qy;
2090 qx = (s->mb_x << 6);
2091 qy = (s->mb_y << 6);
2092 X = (s->mb_width << 6) - 4;
2093 Y = (s->mb_height << 6) - 4;
2094 if(qx + px < -60) px = -60 - qx;
2095 if(qy + py < -60) py = -60 - qy;
2096 if(qx + px > X) px = X - qx;
2097 if(qy + py > Y) py = Y - qy;
2100 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2101 if(0 && !s->first_slice_line && s->mb_x) {
2102 if(is_intra[xy - wrap])
2103 sum = FFABS(px) + FFABS(py);
2105 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2107 if(get_bits1(&s->gb)) {
2115 if(is_intra[xy - 2])
2116 sum = FFABS(px) + FFABS(py);
2118 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2120 if(get_bits1(&s->gb)) {
2130 /* store MV using signed modulus of MV range defined in 4.11 */
2132 s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;
2133 s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;
2135 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2136 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2137 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2138 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2141 /** Get predicted DC value for I-frames only
2142 * prediction dir: left=0, top=1
2143 * @param s MpegEncContext
2144 * @param overlap flag indicating that overlap filtering is used
2145 * @param pq integer part of picture quantizer
2146 * @param[in] n block index in the current MB
2147 * @param dc_val_ptr Pointer to DC predictor
2148 * @param dir_ptr Prediction direction for use in AC prediction
2150 static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2151 int16_t **dc_val_ptr, int *dir_ptr)
2153 int a, b, c, wrap, pred, scale;
2155 static const uint16_t dcpred[32] = {
2156 -1, 1024, 512, 341, 256, 205, 171, 146, 128,
2157 114, 102, 93, 85, 79, 73, 68, 64,
2158 60, 57, 54, 51, 49, 47, 45, 43,
2159 41, 39, 38, 37, 35, 34, 33
2162 /* find prediction - wmv3_dc_scale always used here in fact */
2163 if (n < 4) scale = s->y_dc_scale;
2164 else scale = s->c_dc_scale;
2166 wrap = s->block_wrap[n];
2167 dc_val= s->dc_val[0] + s->block_index[n];
2173 b = dc_val[ - 1 - wrap];
2174 a = dc_val[ - wrap];
2176 if (pq < 9 || !overlap)
2178 /* Set outer values */
2179 if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];
2180 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];
2184 /* Set outer values */
2185 if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;
2186 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;
2189 if (abs(a - b) <= abs(b - c)) {
2197 /* update predictor */
2198 *dc_val_ptr = &dc_val[0];
2203 /** Get predicted DC value
2204 * prediction dir: left=0, top=1
2205 * @param s MpegEncContext
2206 * @param overlap flag indicating that overlap filtering is used
2207 * @param pq integer part of picture quantizer
2208 * @param[in] n block index in the current MB
2209 * @param a_avail flag indicating top block availability
2210 * @param c_avail flag indicating left block availability
2211 * @param dc_val_ptr Pointer to DC predictor
2212 * @param dir_ptr Prediction direction for use in AC prediction
2214 static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2215 int a_avail, int c_avail,
2216 int16_t **dc_val_ptr, int *dir_ptr)
2218 int a, b, c, wrap, pred;
2220 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2223 wrap = s->block_wrap[n];
2224 dc_val= s->dc_val[0] + s->block_index[n];
2230 b = dc_val[ - 1 - wrap];
2231 a = dc_val[ - wrap];
2232 /* scale predictors if needed */
2233 q1 = s->current_picture.qscale_table[mb_pos];
2234 if(c_avail && (n!= 1 && n!=3)) {
2235 q2 = s->current_picture.qscale_table[mb_pos - 1];
2237 c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2239 if(a_avail && (n!= 2 && n!=3)) {
2240 q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2242 a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2244 if(a_avail && c_avail && (n!=3)) {
2247 if(n != 2) off -= s->mb_stride;
2248 q2 = s->current_picture.qscale_table[off];
2250 b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2253 if(a_avail && c_avail) {
2254 if(abs(a - b) <= abs(b - c)) {
2261 } else if(a_avail) {
2264 } else if(c_avail) {
2272 /* update predictor */
2273 *dc_val_ptr = &dc_val[0];
2277 /** @} */ // Block group
2280 * @defgroup vc1_std_mb VC1 Macroblock-level functions in Simple/Main Profiles
2281 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
2285 static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)
2287 int xy, wrap, pred, a, b, c;
2289 xy = s->block_index[n];
2290 wrap = s->b8_stride;
2295 a = s->coded_block[xy - 1 ];
2296 b = s->coded_block[xy - 1 - wrap];
2297 c = s->coded_block[xy - wrap];
2306 *coded_block_ptr = &s->coded_block[xy];
2312 * Decode one AC coefficient
2313 * @param v The VC1 context
2314 * @param last Last coefficient
2315 * @param skip How much zero coefficients to skip
2316 * @param value Decoded AC coefficient value
2317 * @param codingset set of VLC to decode data
2320 static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)
2322 GetBitContext *gb = &v->s.gb;
2323 int index, escape, run = 0, level = 0, lst = 0;
2325 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2326 if (index != vc1_ac_sizes[codingset] - 1) {
2327 run = vc1_index_decode_table[codingset][index][0];
2328 level = vc1_index_decode_table[codingset][index][1];
2329 lst = index >= vc1_last_decode_table[codingset];
2333 escape = decode210(gb);
2335 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2336 run = vc1_index_decode_table[codingset][index][0];
2337 level = vc1_index_decode_table[codingset][index][1];
2338 lst = index >= vc1_last_decode_table[codingset];
2341 level += vc1_last_delta_level_table[codingset][run];
2343 level += vc1_delta_level_table[codingset][run];
2346 run += vc1_last_delta_run_table[codingset][level] + 1;
2348 run += vc1_delta_run_table[codingset][level] + 1;
2354 lst = get_bits1(gb);
2355 if(v->s.esc3_level_length == 0) {
2356 if(v->pq < 8 || v->dquantfrm) { // table 59
2357 v->s.esc3_level_length = get_bits(gb, 3);
2358 if(!v->s.esc3_level_length)
2359 v->s.esc3_level_length = get_bits(gb, 2) + 8;
2361 v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;
2363 v->s.esc3_run_length = 3 + get_bits(gb, 2);
2365 run = get_bits(gb, v->s.esc3_run_length);
2366 sign = get_bits1(gb);
2367 level = get_bits(gb, v->s.esc3_level_length);
2378 /** Decode intra block in intra frames - should be faster than decode_intra_block
2379 * @param v VC1Context
2380 * @param block block to decode
2381 * @param[in] n subblock index
2382 * @param coded are AC coeffs present or not
2383 * @param codingset set of VLC to decode data
2385 static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)
2387 GetBitContext *gb = &v->s.gb;
2388 MpegEncContext *s = &v->s;
2389 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2392 int16_t *ac_val, *ac_val2;
2395 /* Get DC differential */
2397 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2399 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2402 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2407 if (dcdiff == 119 /* ESC index value */)
2409 /* TODO: Optimize */
2410 if (v->pq == 1) dcdiff = get_bits(gb, 10);
2411 else if (v->pq == 2) dcdiff = get_bits(gb, 9);
2412 else dcdiff = get_bits(gb, 8);
2417 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2418 else if (v->pq == 2)
2419 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2426 dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);
2429 /* Store the quantized DC coeff, used for prediction */
2431 block[0] = dcdiff * s->y_dc_scale;
2433 block[0] = dcdiff * s->c_dc_scale;
2445 int last = 0, skip, value;
2446 const int8_t *zz_table;
2450 scale = v->pq * 2 + v->halfpq;
2454 zz_table = wmv1_scantable[2];
2456 zz_table = wmv1_scantable[3];
2458 zz_table = wmv1_scantable[1];
2460 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2462 if(dc_pred_dir) //left
2465 ac_val -= 16 * s->block_wrap[n];
2468 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2472 block[zz_table[i++]] = value;
2475 /* apply AC prediction if needed */
2477 if(dc_pred_dir) { //left
2478 for(k = 1; k < 8; k++)
2479 block[k << 3] += ac_val[k];
2481 for(k = 1; k < 8; k++)
2482 block[k] += ac_val[k + 8];
2485 /* save AC coeffs for further prediction */
2486 for(k = 1; k < 8; k++) {
2487 ac_val2[k] = block[k << 3];
2488 ac_val2[k + 8] = block[k];
2491 /* scale AC coeffs */
2492 for(k = 1; k < 64; k++)
2496 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2499 if(s->ac_pred) i = 63;
2505 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2509 scale = v->pq * 2 + v->halfpq;
2510 memset(ac_val2, 0, 16 * 2);
2511 if(dc_pred_dir) {//left
2514 memcpy(ac_val2, ac_val, 8 * 2);
2516 ac_val -= 16 * s->block_wrap[n];
2518 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2521 /* apply AC prediction if needed */
2523 if(dc_pred_dir) { //left
2524 for(k = 1; k < 8; k++) {
2525 block[k << 3] = ac_val[k] * scale;
2526 if(!v->pquantizer && block[k << 3])
2527 block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;
2530 for(k = 1; k < 8; k++) {
2531 block[k] = ac_val[k + 8] * scale;
2532 if(!v->pquantizer && block[k])
2533 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2539 s->block_last_index[n] = i;
2544 /** Decode intra block in intra frames - should be faster than decode_intra_block
2545 * @param v VC1Context
2546 * @param block block to decode
2547 * @param[in] n subblock number
2548 * @param coded are AC coeffs present or not
2549 * @param codingset set of VLC to decode data
2550 * @param mquant quantizer value for this macroblock
2552 static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant)
2554 GetBitContext *gb = &v->s.gb;
2555 MpegEncContext *s = &v->s;
2556 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2559 int16_t *ac_val, *ac_val2;
2561 int a_avail = v->a_avail, c_avail = v->c_avail;
2562 int use_pred = s->ac_pred;
2565 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2567 /* Get DC differential */
2569 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2571 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2574 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2579 if (dcdiff == 119 /* ESC index value */)
2581 /* TODO: Optimize */
2582 if (mquant == 1) dcdiff = get_bits(gb, 10);
2583 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2584 else dcdiff = get_bits(gb, 8);
2589 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2590 else if (mquant == 2)
2591 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2598 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);
2601 /* Store the quantized DC coeff, used for prediction */
2603 block[0] = dcdiff * s->y_dc_scale;
2605 block[0] = dcdiff * s->c_dc_scale;
2613 /* check if AC is needed at all */
2614 if(!a_avail && !c_avail) use_pred = 0;
2615 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2618 scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);
2620 if(dc_pred_dir) //left
2623 ac_val -= 16 * s->block_wrap[n];
2625 q1 = s->current_picture.qscale_table[mb_pos];
2626 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2627 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2628 if(dc_pred_dir && n==1) q2 = q1;
2629 if(!dc_pred_dir && n==2) q2 = q1;
2633 int last = 0, skip, value;
2634 const int8_t *zz_table;
2639 zz_table = wmv1_scantable[2];
2641 zz_table = wmv1_scantable[3];
2643 zz_table = wmv1_scantable[1];
2646 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2650 block[zz_table[i++]] = value;
2653 /* apply AC prediction if needed */
2655 /* scale predictors if needed*/
2657 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2658 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2660 if(dc_pred_dir) { //left
2661 for(k = 1; k < 8; k++)
2662 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2664 for(k = 1; k < 8; k++)
2665 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2668 if(dc_pred_dir) { //left
2669 for(k = 1; k < 8; k++)
2670 block[k << 3] += ac_val[k];
2672 for(k = 1; k < 8; k++)
2673 block[k] += ac_val[k + 8];
2677 /* save AC coeffs for further prediction */
2678 for(k = 1; k < 8; k++) {
2679 ac_val2[k] = block[k << 3];
2680 ac_val2[k + 8] = block[k];
2683 /* scale AC coeffs */
2684 for(k = 1; k < 64; k++)
2688 block[k] += (block[k] < 0) ? -mquant : mquant;
2691 if(use_pred) i = 63;
2692 } else { // no AC coeffs
2695 memset(ac_val2, 0, 16 * 2);
2696 if(dc_pred_dir) {//left
2698 memcpy(ac_val2, ac_val, 8 * 2);
2700 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2701 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2702 for(k = 1; k < 8; k++)
2703 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2708 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2710 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2711 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2712 for(k = 1; k < 8; k++)
2713 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2718 /* apply AC prediction if needed */
2720 if(dc_pred_dir) { //left
2721 for(k = 1; k < 8; k++) {
2722 block[k << 3] = ac_val2[k] * scale;
2723 if(!v->pquantizer && block[k << 3])
2724 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2727 for(k = 1; k < 8; k++) {
2728 block[k] = ac_val2[k + 8] * scale;
2729 if(!v->pquantizer && block[k])
2730 block[k] += (block[k] < 0) ? -mquant : mquant;
2736 s->block_last_index[n] = i;
2741 /** Decode intra block in inter frames - more generic version than vc1_decode_i_block
2742 * @param v VC1Context
2743 * @param block block to decode
2744 * @param[in] n subblock index
2745 * @param coded are AC coeffs present or not
2746 * @param mquant block quantizer
2747 * @param codingset set of VLC to decode data
2749 static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)
2751 GetBitContext *gb = &v->s.gb;
2752 MpegEncContext *s = &v->s;
2753 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2756 int16_t *ac_val, *ac_val2;
2758 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2759 int a_avail = v->a_avail, c_avail = v->c_avail;
2760 int use_pred = s->ac_pred;
2764 /* XXX: Guard against dumb values of mquant */
2765 mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );
2767 /* Set DC scale - y and c use the same */
2768 s->y_dc_scale = s->y_dc_scale_table[mquant];
2769 s->c_dc_scale = s->c_dc_scale_table[mquant];
2771 /* Get DC differential */
2773 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2775 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2778 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2783 if (dcdiff == 119 /* ESC index value */)
2785 /* TODO: Optimize */
2786 if (mquant == 1) dcdiff = get_bits(gb, 10);
2787 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2788 else dcdiff = get_bits(gb, 8);
2793 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2794 else if (mquant == 2)
2795 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2802 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);
2805 /* Store the quantized DC coeff, used for prediction */
2808 block[0] = dcdiff * s->y_dc_scale;
2810 block[0] = dcdiff * s->c_dc_scale;
2818 /* check if AC is needed at all and adjust direction if needed */
2819 if(!a_avail) dc_pred_dir = 1;
2820 if(!c_avail) dc_pred_dir = 0;
2821 if(!a_avail && !c_avail) use_pred = 0;
2822 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2825 scale = mquant * 2 + v->halfpq;
2827 if(dc_pred_dir) //left
2830 ac_val -= 16 * s->block_wrap[n];
2832 q1 = s->current_picture.qscale_table[mb_pos];
2833 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2834 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2835 if(dc_pred_dir && n==1) q2 = q1;
2836 if(!dc_pred_dir && n==2) q2 = q1;
2840 int last = 0, skip, value;
2841 const int8_t *zz_table;
2844 zz_table = wmv1_scantable[0];
2847 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2851 block[zz_table[i++]] = value;
2854 /* apply AC prediction if needed */
2856 /* scale predictors if needed*/
2858 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2859 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2861 if(dc_pred_dir) { //left
2862 for(k = 1; k < 8; k++)
2863 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2865 for(k = 1; k < 8; k++)
2866 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2869 if(dc_pred_dir) { //left
2870 for(k = 1; k < 8; k++)
2871 block[k << 3] += ac_val[k];
2873 for(k = 1; k < 8; k++)
2874 block[k] += ac_val[k + 8];
2878 /* save AC coeffs for further prediction */
2879 for(k = 1; k < 8; k++) {
2880 ac_val2[k] = block[k << 3];
2881 ac_val2[k + 8] = block[k];
2884 /* scale AC coeffs */
2885 for(k = 1; k < 64; k++)
2889 block[k] += (block[k] < 0) ? -mquant : mquant;
2892 if(use_pred) i = 63;
2893 } else { // no AC coeffs
2896 memset(ac_val2, 0, 16 * 2);
2897 if(dc_pred_dir) {//left
2899 memcpy(ac_val2, ac_val, 8 * 2);
2901 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2902 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2903 for(k = 1; k < 8; k++)
2904 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2909 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2911 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2912 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2913 for(k = 1; k < 8; k++)
2914 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2919 /* apply AC prediction if needed */
2921 if(dc_pred_dir) { //left
2922 for(k = 1; k < 8; k++) {
2923 block[k << 3] = ac_val2[k] * scale;
2924 if(!v->pquantizer && block[k << 3])
2925 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2928 for(k = 1; k < 8; k++) {
2929 block[k] = ac_val2[k + 8] * scale;
2930 if(!v->pquantizer && block[k])
2931 block[k] += (block[k] < 0) ? -mquant : mquant;
2937 s->block_last_index[n] = i;
2944 static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,
2945 uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left)
2947 MpegEncContext *s = &v->s;
2948 GetBitContext *gb = &s->gb;
2951 int scale, off, idx, last, skip, value;
2952 int ttblk = ttmb & 7;
2956 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)];
2958 if(ttblk == TT_4X4) {
2959 subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);
2961 if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {
2962 subblkpat = decode012(gb);
2963 if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits
2964 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;
2965 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;
2967 scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);
2969 // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT
2970 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {
2971 subblkpat = 2 - (ttblk == TT_8X4_TOP);
2974 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
2975 subblkpat = 2 - (ttblk == TT_4X8_LEFT);
2984 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
2988 idx = wmv1_scantable[0][i++];
2989 block[idx] = value * scale;
2991 block[idx] += (block[idx] < 0) ? -mquant : mquant;
2994 s->dsp.vc1_inv_trans_8x8(block);
2995 s->dsp.add_pixels_clamped(block, dst, linesize);
2996 if(apply_filter && cbp_top & 0xC)
2997 s->dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
2998 if(apply_filter && cbp_left & 0xA)
2999 s->dsp.vc1_h_loop_filter8(dst, linesize, v->pq);
3003 pat = ~subblkpat & 0xF;
3004 for(j = 0; j < 4; j++) {
3005 last = subblkpat & (1 << (3 - j));
3007 off = (j & 1) * 4 + (j & 2) * 16;
3009 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3013 idx = ff_vc1_simple_progressive_4x4_zz[i++];
3014 block[idx + off] = value * scale;
3016 block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;
3018 if(!(subblkpat & (1 << (3 - j))) && !skip_block){
3019 s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
3020 if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))
3021 s->dsp.vc1_v_loop_filter4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, v->pq);
3022 if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))
3023 s->dsp.vc1_h_loop_filter4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, v->pq);
3028 pat = ~((subblkpat & 2)*6 + (subblkpat & 1)*3) & 0xF;
3029 for(j = 0; j < 2; j++) {
3030 last = subblkpat & (1 << (1 - j));
3034 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3038 idx = v->zz_8x4[i++]+off;
3039 block[idx] = value * scale;
3041 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3043 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3044 s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
3045 if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))
3046 s->dsp.vc1_v_loop_filter8(dst + j*4*linesize, linesize, v->pq);
3047 if(apply_filter && cbp_left & (2 << j))
3048 s->dsp.vc1_h_loop_filter4(dst + j*4*linesize, linesize, v->pq);
3053 pat = ~(subblkpat*5) & 0xF;
3054 for(j = 0; j < 2; j++) {
3055 last = subblkpat & (1 << (1 - j));
3059 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3063 idx = v->zz_4x8[i++]+off;
3064 block[idx] = value * scale;
3066 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3068 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3069 s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
3070 if(apply_filter && cbp_top & (2 << j))
3071 s->dsp.vc1_v_loop_filter4(dst + j*4, linesize, v->pq);
3072 if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))
3073 s->dsp.vc1_h_loop_filter8(dst + j*4, linesize, v->pq);
3081 /** @} */ // Macroblock group
3083 static const int size_table [6] = { 0, 2, 3, 4, 5, 8 };
3084 static const int offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3086 /** Decode one P-frame MB (in Simple/Main profile)
3088 static int vc1_decode_p_mb(VC1Context *v)
3090 MpegEncContext *s = &v->s;
3091 GetBitContext *gb = &s->gb;
3093 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3094 int cbp; /* cbp decoding stuff */
3095 int mqdiff, mquant; /* MB quantization */
3096 int ttmb = v->ttfrm; /* MB Transform type */
3098 int mb_has_coeffs = 1; /* last_flag */
3099 int dmv_x, dmv_y; /* Differential MV components */
3100 int index, index1; /* LUT indexes */
3101 int val, sign; /* temp values */
3102 int first_block = 1;
3104 int skipped, fourmv;
3105 int block_cbp = 0, pat;
3106 int apply_loop_filter;
3108 mquant = v->pq; /* Loosy initialization */
3110 if (v->mv_type_is_raw)
3111 fourmv = get_bits1(gb);
3113 fourmv = v->mv_type_mb_plane[mb_pos];
3115 skipped = get_bits1(gb);
3117 skipped = v->s.mbskip_table[mb_pos];
3119 s->dsp.clear_blocks(s->block[0]);
3121 apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);
3122 if (!fourmv) /* 1MV mode */
3126 GET_MVDATA(dmv_x, dmv_y);
3129 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3130 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3132 s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
3133 vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
3135 /* FIXME Set DC val for inter block ? */
3136 if (s->mb_intra && !mb_has_coeffs)
3139 s->ac_pred = get_bits1(gb);
3142 else if (mb_has_coeffs)
3144 if (s->mb_intra) s->ac_pred = get_bits1(gb);
3145 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3153 s->current_picture.qscale_table[mb_pos] = mquant;
3155 if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3156 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,
3157 VC1_TTMB_VLC_BITS, 2);
3158 if(!s->mb_intra) vc1_mc_1mv(v, 0);
3162 s->dc_val[0][s->block_index[i]] = 0;
3164 val = ((cbp >> (5 - i)) & 1);
3165 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3166 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3168 /* check if prediction blocks A and C are available */
3169 v->a_avail = v->c_avail = 0;
3170 if(i == 2 || i == 3 || !s->first_slice_line)
3171 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3172 if(i == 1 || i == 3 || s->mb_x)
3173 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3175 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3176 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3177 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3178 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3179 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3180 if(v->pq >= 9 && v->overlap) {
3182 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3184 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3186 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3187 int left_cbp, top_cbp;
3189 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3190 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3192 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3193 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3196 s->dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
3198 s->dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
3200 block_cbp |= 0xF << (i << 2);
3202 int left_cbp = 0, top_cbp = 0, filter = 0;
3203 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3206 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3207 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3209 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3210 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3213 s->dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
3215 s->dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
3217 pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);
3218 block_cbp |= pat << (i << 2);
3219 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3227 for(i = 0; i < 6; i++) {
3228 v->mb_type[0][s->block_index[i]] = 0;
3229 s->dc_val[0][s->block_index[i]] = 0;
3231 s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;
3232 s->current_picture.qscale_table[mb_pos] = 0;
3233 vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
3240 if (!skipped /* unskipped MB */)
3242 int intra_count = 0, coded_inter = 0;
3243 int is_intra[6], is_coded[6];
3245 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3248 val = ((cbp >> (5 - i)) & 1);
3249 s->dc_val[0][s->block_index[i]] = 0;
3256 GET_MVDATA(dmv_x, dmv_y);
3258 vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
3259 if(!s->mb_intra) vc1_mc_4mv_luma(v, i);
3260 intra_count += s->mb_intra;
3261 is_intra[i] = s->mb_intra;
3262 is_coded[i] = mb_has_coeffs;
3265 is_intra[i] = (intra_count >= 3);
3268 if(i == 4) vc1_mc_4mv_chroma(v);
3269 v->mb_type[0][s->block_index[i]] = is_intra[i];
3270 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];
3272 // if there are no coded blocks then don't do anything more
3273 if(!intra_count && !coded_inter) return 0;
3276 s->current_picture.qscale_table[mb_pos] = mquant;
3277 /* test if block is intra and has pred */
3282 if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])
3283 || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {
3288 if(intrapred)s->ac_pred = get_bits1(gb);
3289 else s->ac_pred = 0;
3291 if (!v->ttmbf && coded_inter)
3292 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3296 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3297 s->mb_intra = is_intra[i];
3299 /* check if prediction blocks A and C are available */
3300 v->a_avail = v->c_avail = 0;
3301 if(i == 2 || i == 3 || !s->first_slice_line)
3302 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3303 if(i == 1 || i == 3 || s->mb_x)
3304 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3306 vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);
3307 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3308 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3309 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3310 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
3311 if(v->pq >= 9 && v->overlap) {
3313 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3315 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3317 if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3318 int left_cbp, top_cbp;
3320 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3321 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3323 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3324 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3327 s->dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
3329 s->dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
3331 block_cbp |= 0xF << (i << 2);
3332 } else if(is_coded[i]) {
3333 int left_cbp = 0, top_cbp = 0, filter = 0;
3334 if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3337 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3338 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3340 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3341 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3344 s->dsp.vc1_v_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
3346 s->dsp.vc1_h_loop_filter8(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, v->pq);
3348 pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);
3349 block_cbp |= pat << (i << 2);
3350 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3359 s->current_picture.qscale_table[mb_pos] = 0;
3360 for (i=0; i<6; i++) {
3361 v->mb_type[0][s->block_index[i]] = 0;
3362 s->dc_val[0][s->block_index[i]] = 0;
3366 vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
3367 vc1_mc_4mv_luma(v, i);
3369 vc1_mc_4mv_chroma(v);
3370 s->current_picture.qscale_table[mb_pos] = 0;
3374 v->cbp[s->mb_x] = block_cbp;
3376 /* Should never happen */
3380 /** Decode one B-frame MB (in Main profile)
3382 static void vc1_decode_b_mb(VC1Context *v)
3384 MpegEncContext *s = &v->s;
3385 GetBitContext *gb = &s->gb;
3387 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3388 int cbp = 0; /* cbp decoding stuff */
3389 int mqdiff, mquant; /* MB quantization */
3390 int ttmb = v->ttfrm; /* MB Transform type */
3391 int mb_has_coeffs = 0; /* last_flag */
3392 int index, index1; /* LUT indexes */
3393 int val, sign; /* temp values */
3394 int first_block = 1;
3396 int skipped, direct;
3397 int dmv_x[2], dmv_y[2];
3398 int bmvtype = BMV_TYPE_BACKWARD;
3400 mquant = v->pq; /* Loosy initialization */
3404 direct = get_bits1(gb);
3406 direct = v->direct_mb_plane[mb_pos];
3408 skipped = get_bits1(gb);
3410 skipped = v->s.mbskip_table[mb_pos];
3412 s->dsp.clear_blocks(s->block[0]);
3413 dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
3414 for(i = 0; i < 6; i++) {
3415 v->mb_type[0][s->block_index[i]] = 0;
3416 s->dc_val[0][s->block_index[i]] = 0;
3418 s->current_picture.qscale_table[mb_pos] = 0;
3422 GET_MVDATA(dmv_x[0], dmv_y[0]);
3423 dmv_x[1] = dmv_x[0];
3424 dmv_y[1] = dmv_y[0];
3426 if(skipped || !s->mb_intra) {
3427 bmvtype = decode012(gb);
3430 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
3433 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
3436 bmvtype = BMV_TYPE_INTERPOLATED;
3437 dmv_x[0] = dmv_y[0] = 0;
3441 for(i = 0; i < 6; i++)
3442 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3445 if(direct) bmvtype = BMV_TYPE_INTERPOLATED;
3446 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3447 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3451 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3454 s->current_picture.qscale_table[mb_pos] = mquant;
3456 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3457 dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;
3458 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3459 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3461 if(!mb_has_coeffs && !s->mb_intra) {
3462 /* no coded blocks - effectively skipped */
3463 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3464 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3467 if(s->mb_intra && !mb_has_coeffs) {
3469 s->current_picture.qscale_table[mb_pos] = mquant;
3470 s->ac_pred = get_bits1(gb);
3472 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3474 if(bmvtype == BMV_TYPE_INTERPOLATED) {
3475 GET_MVDATA(dmv_x[0], dmv_y[0]);
3476 if(!mb_has_coeffs) {
3477 /* interpolated skipped block */
3478 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3479 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3483 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3485 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3488 s->ac_pred = get_bits1(gb);
3489 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3491 s->current_picture.qscale_table[mb_pos] = mquant;
3492 if(!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3493 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3499 s->dc_val[0][s->block_index[i]] = 0;
3501 val = ((cbp >> (5 - i)) & 1);
3502 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3503 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3505 /* check if prediction blocks A and C are available */
3506 v->a_avail = v->c_avail = 0;
3507 if(i == 2 || i == 3 || !s->first_slice_line)
3508 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3509 if(i == 1 || i == 3 || s->mb_x)
3510 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3512 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3513 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3514 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3515 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3516 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3518 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), 0, 0, 0);
3519 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3525 /** Decode blocks of I-frame
3527 static void vc1_decode_i_blocks(VC1Context *v)
3530 MpegEncContext *s = &v->s;
3535 /* select codingmode used for VLC tables selection */
3536 switch(v->y_ac_table_index){
3538 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3541 v->codingset = CS_HIGH_MOT_INTRA;
3544 v->codingset = CS_MID_RATE_INTRA;
3548 switch(v->c_ac_table_index){
3550 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3553 v->codingset2 = CS_HIGH_MOT_INTER;
3556 v->codingset2 = CS_MID_RATE_INTER;
3560 /* Set DC scale - y and c use the same */
3561 s->y_dc_scale = s->y_dc_scale_table[v->pq];
3562 s->c_dc_scale = s->c_dc_scale_table[v->pq];
3565 s->mb_x = s->mb_y = 0;
3567 s->first_slice_line = 1;
3568 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3569 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3570 ff_init_block_index(s);
3571 ff_update_block_index(s);
3572 s->dsp.clear_blocks(s->block[0]);
3573 mb_pos = s->mb_x + s->mb_y * s->mb_width;
3574 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3575 s->current_picture.qscale_table[mb_pos] = v->pq;
3576 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3577 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3579 // do actual MB decoding and displaying
3580 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3581 v->s.ac_pred = get_bits1(&v->s.gb);
3583 for(k = 0; k < 6; k++) {
3584 val = ((cbp >> (5 - k)) & 1);
3587 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3591 cbp |= val << (5 - k);
3593 vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);
3595 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3596 if(v->pq >= 9 && v->overlap) {
3597 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3601 vc1_put_block(v, s->block);
3602 if(v->pq >= 9 && v->overlap) {
3604 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3605 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3606 if(!(s->flags & CODEC_FLAG_GRAY)) {
3607 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3608 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3611 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3612 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3613 if(!s->first_slice_line) {
3614 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3615 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3616 if(!(s->flags & CODEC_FLAG_GRAY)) {
3617 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3618 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3621 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3622 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3624 if(v->s.loop_filter) vc1_loop_filter_iblk(s, v->pq);
3626 if(get_bits_count(&s->gb) > v->bits) {
3627 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3628 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3632 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3633 s->first_slice_line = 0;
3635 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3638 /** Decode blocks of I-frame for advanced profile
3640 static void vc1_decode_i_blocks_adv(VC1Context *v)
3643 MpegEncContext *s = &v->s;
3650 GetBitContext *gb = &s->gb;
3652 /* select codingmode used for VLC tables selection */
3653 switch(v->y_ac_table_index){
3655 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3658 v->codingset = CS_HIGH_MOT_INTRA;
3661 v->codingset = CS_MID_RATE_INTRA;
3665 switch(v->c_ac_table_index){
3667 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3670 v->codingset2 = CS_HIGH_MOT_INTER;
3673 v->codingset2 = CS_MID_RATE_INTER;
3678 s->mb_x = s->mb_y = 0;
3680 s->first_slice_line = 1;
3681 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3682 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3683 ff_init_block_index(s);
3684 ff_update_block_index(s);
3685 s->dsp.clear_blocks(s->block[0]);
3686 mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3687 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3688 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3689 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3691 // do actual MB decoding and displaying
3692 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3693 if(v->acpred_is_raw)
3694 v->s.ac_pred = get_bits1(&v->s.gb);
3696 v->s.ac_pred = v->acpred_plane[mb_pos];
3698 if(v->condover == CONDOVER_SELECT) {
3699 if(v->overflg_is_raw)
3700 overlap = get_bits1(&v->s.gb);
3702 overlap = v->over_flags_plane[mb_pos];
3704 overlap = (v->condover == CONDOVER_ALL);
3708 s->current_picture.qscale_table[mb_pos] = mquant;
3709 /* Set DC scale - y and c use the same */
3710 s->y_dc_scale = s->y_dc_scale_table[mquant];
3711 s->c_dc_scale = s->c_dc_scale_table[mquant];
3713 for(k = 0; k < 6; k++) {
3714 val = ((cbp >> (5 - k)) & 1);
3717 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3721 cbp |= val << (5 - k);
3723 v->a_avail = !s->first_slice_line || (k==2 || k==3);
3724 v->c_avail = !!s->mb_x || (k==1 || k==3);
3726 vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);
3728 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3729 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3732 vc1_put_block(v, s->block);
3735 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3736 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3737 if(!(s->flags & CODEC_FLAG_GRAY)) {
3738 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3739 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3742 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3743 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3744 if(!s->first_slice_line) {
3745 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3746 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3747 if(!(s->flags & CODEC_FLAG_GRAY)) {
3748 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3749 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3752 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3753 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3755 if(v->s.loop_filter) vc1_loop_filter_iblk(s, v->pq);
3757 if(get_bits_count(&s->gb) > v->bits) {
3758 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3759 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3763 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3764 s->first_slice_line = 0;
3766 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3769 static void vc1_decode_p_blocks(VC1Context *v)
3771 MpegEncContext *s = &v->s;
3773 /* select codingmode used for VLC tables selection */
3774 switch(v->c_ac_table_index){
3776 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3779 v->codingset = CS_HIGH_MOT_INTRA;
3782 v->codingset = CS_MID_RATE_INTRA;
3786 switch(v->c_ac_table_index){
3788 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3791 v->codingset2 = CS_HIGH_MOT_INTER;
3794 v->codingset2 = CS_MID_RATE_INTER;
3798 s->first_slice_line = 1;
3799 memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);
3800 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3801 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3802 ff_init_block_index(s);
3803 ff_update_block_index(s);
3804 s->dsp.clear_blocks(s->block[0]);
3807 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3808 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3809 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);
3813 memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride);
3814 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3815 s->first_slice_line = 0;
3817 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3820 static void vc1_decode_b_blocks(VC1Context *v)
3822 MpegEncContext *s = &v->s;
3824 /* select codingmode used for VLC tables selection */
3825 switch(v->c_ac_table_index){
3827 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3830 v->codingset = CS_HIGH_MOT_INTRA;
3833 v->codingset = CS_MID_RATE_INTRA;
3837 switch(v->c_ac_table_index){
3839 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3842 v->codingset2 = CS_HIGH_MOT_INTER;
3845 v->codingset2 = CS_MID_RATE_INTER;
3849 s->first_slice_line = 1;
3850 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3851 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3852 ff_init_block_index(s);
3853 ff_update_block_index(s);
3854 s->dsp.clear_blocks(s->block[0]);
3857 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3858 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3859 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);
3862 if(v->s.loop_filter) vc1_loop_filter_iblk(s, v->pq);
3864 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3865 s->first_slice_line = 0;
3867 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3870 static void vc1_decode_skip_blocks(VC1Context *v)
3872 MpegEncContext *s = &v->s;
3874 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3875 s->first_slice_line = 1;
3876 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3878 ff_init_block_index(s);
3879 ff_update_block_index(s);
3880 memcpy(s->dest[0], s->last_picture.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16);
3881 memcpy(s->dest[1], s->last_picture.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3882 memcpy(s->dest[2], s->last_picture.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3883 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3884 s->first_slice_line = 0;
3886 s->pict_type = FF_P_TYPE;
3889 static void vc1_decode_blocks(VC1Context *v)
3892 v->s.esc3_level_length = 0;
3894 ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );
3897 switch(v->s.pict_type) {
3899 if(v->profile == PROFILE_ADVANCED)
3900 vc1_decode_i_blocks_adv(v);
3902 vc1_decode_i_blocks(v);
3905 if(v->p_frame_skipped)
3906 vc1_decode_skip_blocks(v);
3908 vc1_decode_p_blocks(v);
3912 if(v->profile == PROFILE_ADVANCED)
3913 vc1_decode_i_blocks_adv(v);
3915 vc1_decode_i_blocks(v);
3917 vc1_decode_b_blocks(v);
3923 /** Find VC-1 marker in buffer
3924 * @return position where next marker starts or end of buffer if no marker found
3926 static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)
3928 uint32_t mrk = 0xFFFFFFFF;
3930 if(end-src < 4) return end;
3932 mrk = (mrk << 8) | *src++;
3939 static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)
3944 for(dsize = 0; dsize < size; dsize++) *dst++ = *src++;
3947 for(i = 0; i < size; i++, src++) {
3948 if(src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {
3949 dst[dsize++] = src[1];
3953 dst[dsize++] = *src;
3958 /** Initialize a VC1/WMV3 decoder
3959 * @todo TODO: Handle VC-1 IDUs (Transport level?)
3960 * @todo TODO: Decypher remaining bits in extra_data
3962 static av_cold int vc1_decode_init(AVCodecContext *avctx)
3964 VC1Context *v = avctx->priv_data;
3965 MpegEncContext *s = &v->s;
3968 if (!avctx->extradata_size || !avctx->extradata) return -1;
3969 if (!(avctx->flags & CODEC_FLAG_GRAY))
3970 avctx->pix_fmt = avctx->get_format(avctx, avctx->codec->pix_fmts);
3972 avctx->pix_fmt = PIX_FMT_GRAY8;
3973 avctx->hwaccel = ff_find_hwaccel(avctx->codec->id, avctx->pix_fmt);
3975 avctx->flags |= CODEC_FLAG_EMU_EDGE;
3976 v->s.flags |= CODEC_FLAG_EMU_EDGE;
3978 if(avctx->idct_algo==FF_IDCT_AUTO){
3979 avctx->idct_algo=FF_IDCT_WMV2;
3982 if(ff_h263_decode_init(avctx) < 0)
3984 if (vc1_init_common(v) < 0) return -1;
3986 avctx->coded_width = avctx->width;
3987 avctx->coded_height = avctx->height;
3988 if (avctx->codec_id == CODEC_ID_WMV3)
3992 // looks like WMV3 has a sequence header stored in the extradata
3993 // advanced sequence header may be before the first frame
3994 // the last byte of the extradata is a version number, 1 for the
3995 // samples we can decode
3997 init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);
3999 if (decode_sequence_header(avctx, &gb) < 0)
4002 count = avctx->extradata_size*8 - get_bits_count(&gb);
4005 av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",
4006 count, get_bits(&gb, count));
4010 av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);
4012 } else { // VC1/WVC1
4013 const uint8_t *start = avctx->extradata;
4014 uint8_t *end = avctx->extradata + avctx->extradata_size;
4015 const uint8_t *next;
4016 int size, buf2_size;
4017 uint8_t *buf2 = NULL;
4018 int seq_initialized = 0, ep_initialized = 0;
4020 if(avctx->extradata_size < 16) {
4021 av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);
4025 buf2 = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
4026 if(start[0]) start++; // in WVC1 extradata first byte is its size
4028 for(; next < end; start = next){
4029 next = find_next_marker(start + 4, end);
4030 size = next - start - 4;
4031 if(size <= 0) continue;
4032 buf2_size = vc1_unescape_buffer(start + 4, size, buf2);
4033 init_get_bits(&gb, buf2, buf2_size * 8);
4034 switch(AV_RB32(start)){
4035 case VC1_CODE_SEQHDR:
4036 if(decode_sequence_header(avctx, &gb) < 0){
4040 seq_initialized = 1;
4042 case VC1_CODE_ENTRYPOINT:
4043 if(decode_entry_point(avctx, &gb) < 0){
4052 if(!seq_initialized || !ep_initialized){
4053 av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");
4057 avctx->has_b_frames= !!(avctx->max_b_frames);
4058 s->low_delay = !avctx->has_b_frames;
4060 s->mb_width = (avctx->coded_width+15)>>4;
4061 s->mb_height = (avctx->coded_height+15)>>4;
4063 /* Allocate mb bitplanes */
4064 v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4065 v->direct_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4066 v->acpred_plane = av_malloc(s->mb_stride * s->mb_height);
4067 v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height);
4069 v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);
4070 v->cbp = v->cbp_base + s->mb_stride;
4072 /* allocate block type info in that way so it could be used with s->block_index[] */
4073 v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);
4074 v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;
4075 v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;
4076 v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);
4078 /* Init coded blocks info */
4079 if (v->profile == PROFILE_ADVANCED)
4081 // if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
4083 // if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
4087 ff_intrax8_common_init(&v->x8,s);
4092 /** Decode a VC1/WMV3 frame
4093 * @todo TODO: Handle VC-1 IDUs (Transport level?)
4095 static int vc1_decode_frame(AVCodecContext *avctx,
4096 void *data, int *data_size,
4099 const uint8_t *buf = avpkt->data;
4100 int buf_size = avpkt->size;
4101 VC1Context *v = avctx->priv_data;
4102 MpegEncContext *s = &v->s;
4103 AVFrame *pict = data;
4104 uint8_t *buf2 = NULL;
4105 const uint8_t *buf_start = buf;
4107 /* no supplementary picture */
4108 if (buf_size == 0) {
4109 /* special case for last picture */
4110 if (s->low_delay==0 && s->next_picture_ptr) {
4111 *pict= *(AVFrame*)s->next_picture_ptr;
4112 s->next_picture_ptr= NULL;
4114 *data_size = sizeof(AVFrame);
4120 /* We need to set current_picture_ptr before reading the header,
4121 * otherwise we cannot store anything in there. */
4122 if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){
4123 int i= ff_find_unused_picture(s, 0);
4124 s->current_picture_ptr= &s->picture[i];
4127 if (s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU){
4128 if (v->profile < PROFILE_ADVANCED)
4129 avctx->pix_fmt = PIX_FMT_VDPAU_WMV3;
4131 avctx->pix_fmt = PIX_FMT_VDPAU_VC1;
4134 //for advanced profile we may need to parse and unescape data
4135 if (avctx->codec_id == CODEC_ID_VC1) {
4137 buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
4139 if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */
4140 const uint8_t *start, *end, *next;
4144 for(start = buf, end = buf + buf_size; next < end; start = next){
4145 next = find_next_marker(start + 4, end);
4146 size = next - start - 4;
4147 if(size <= 0) continue;
4148 switch(AV_RB32(start)){
4149 case VC1_CODE_FRAME:
4150 if (avctx->hwaccel ||
4151 s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
4153 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4155 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */
4156 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4157 init_get_bits(&s->gb, buf2, buf_size2*8);
4158 decode_entry_point(avctx, &s->gb);
4160 case VC1_CODE_SLICE:
4161 av_log(avctx, AV_LOG_ERROR, "Sliced decoding is not implemented (yet)\n");
4166 }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */
4167 const uint8_t *divider;
4169 divider = find_next_marker(buf, buf + buf_size);
4170 if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){
4171 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");
4176 buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);
4178 av_free(buf2);return -1;
4180 buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);
4182 init_get_bits(&s->gb, buf2, buf_size2*8);
4184 init_get_bits(&s->gb, buf, buf_size*8);
4185 // do parse frame header
4186 if(v->profile < PROFILE_ADVANCED) {
4187 if(vc1_parse_frame_header(v, &s->gb) == -1) {
4192 if(vc1_parse_frame_header_adv(v, &s->gb) == -1) {
4198 if(s->pict_type != FF_I_TYPE && !v->res_rtm_flag){
4204 s->current_picture.pict_type= s->pict_type;
4205 s->current_picture.key_frame= s->pict_type == FF_I_TYPE;
4207 /* skip B-frames if we don't have reference frames */
4208 if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){
4210 return -1;//buf_size;
4212 /* skip b frames if we are in a hurry */
4213 if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;
4214 if( (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
4215 || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
4216 || avctx->skip_frame >= AVDISCARD_ALL) {
4220 /* skip everything if we are in a hurry>=5 */
4221 if(avctx->hurry_up>=5) {
4223 return -1;//buf_size;
4226 if(s->next_p_frame_damaged){
4227 if(s->pict_type==FF_B_TYPE)
4230 s->next_p_frame_damaged=0;
4233 if(MPV_frame_start(s, avctx) < 0) {
4238 s->me.qpel_put= s->dsp.put_qpel_pixels_tab;
4239 s->me.qpel_avg= s->dsp.avg_qpel_pixels_tab;
4241 if ((CONFIG_VC1_VDPAU_DECODER || CONFIG_WMV3_VDPAU_DECODER)
4242 &&s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
4243 ff_vdpau_vc1_decode_picture(s, buf_start, (buf + buf_size) - buf_start);
4244 else if (avctx->hwaccel) {
4245 if (avctx->hwaccel->start_frame(avctx, buf, buf_size) < 0)
4247 if (avctx->hwaccel->decode_slice(avctx, buf_start, (buf + buf_size) - buf_start) < 0)
4249 if (avctx->hwaccel->end_frame(avctx) < 0)
4252 ff_er_frame_start(s);
4254 v->bits = buf_size * 8;
4255 vc1_decode_blocks(v);
4256 //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);
4257 // if(get_bits_count(&s->gb) > buf_size * 8)
4264 assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);
4265 assert(s->current_picture.pict_type == s->pict_type);
4266 if (s->pict_type == FF_B_TYPE || s->low_delay) {
4267 *pict= *(AVFrame*)s->current_picture_ptr;
4268 } else if (s->last_picture_ptr != NULL) {
4269 *pict= *(AVFrame*)s->last_picture_ptr;
4272 if(s->last_picture_ptr || s->low_delay){
4273 *data_size = sizeof(AVFrame);
4274 ff_print_debug_info(s, pict);
4277 /* Return the Picture timestamp as the frame number */
4278 /* we subtract 1 because it is added on utils.c */
4279 avctx->frame_number = s->picture_number - 1;
4286 /** Close a VC1/WMV3 decoder
4287 * @warning Initial try at using MpegEncContext stuff
4289 static av_cold int vc1_decode_end(AVCodecContext *avctx)
4291 VC1Context *v = avctx->priv_data;
4293 av_freep(&v->hrd_rate);
4294 av_freep(&v->hrd_buffer);
4295 MPV_common_end(&v->s);
4296 av_freep(&v->mv_type_mb_plane);
4297 av_freep(&v->direct_mb_plane);
4298 av_freep(&v->acpred_plane);
4299 av_freep(&v->over_flags_plane);
4300 av_freep(&v->mb_type_base);
4301 av_freep(&v->cbp_base);
4302 ff_intrax8_common_end(&v->x8);
4307 AVCodec vc1_decoder = {
4318 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),
4319 .pix_fmts = ff_hwaccel_pixfmt_list_420
4322 AVCodec wmv3_decoder = {
4333 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),
4334 .pix_fmts = ff_hwaccel_pixfmt_list_420
4337 #if CONFIG_WMV3_VDPAU_DECODER
4338 AVCodec wmv3_vdpau_decoder = {
4347 CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,
4349 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 VDPAU"),
4350 .pix_fmts = (enum PixelFormat[]){PIX_FMT_VDPAU_WMV3, PIX_FMT_NONE}
4354 #if CONFIG_VC1_VDPAU_DECODER
4355 AVCodec vc1_vdpau_decoder = {
4364 CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,
4366 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1 VDPAU"),
4367 .pix_fmts = (enum PixelFormat[]){PIX_FMT_VDPAU_VC1, PIX_FMT_NONE}