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
25 * VC-1 and WMV3 decoder
30 #include "mpegvideo.h"
33 #include "vc1acdata.h"
34 #include "msmpeg4data.h"
36 #include "simple_idct.h"
41 #define MB_INTRA_VLC_BITS 9
44 static const uint16_t table_mb_intra[64][2];
48 * Init VC-1 specific tables and VC1Context members
49 * @param v The VC1Context to initialize
52 static int vc1_init_common(VC1Context *v)
57 v->hrd_rate = v->hrd_buffer = NULL;
63 init_vlc(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,
64 ff_vc1_bfraction_bits, 1, 1,
65 ff_vc1_bfraction_codes, 1, 1, 1);
66 init_vlc(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,
67 ff_vc1_norm2_bits, 1, 1,
68 ff_vc1_norm2_codes, 1, 1, 1);
69 init_vlc(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,
70 ff_vc1_norm6_bits, 1, 1,
71 ff_vc1_norm6_codes, 2, 2, 1);
72 init_vlc(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,
73 ff_vc1_imode_bits, 1, 1,
74 ff_vc1_imode_codes, 1, 1, 1);
77 init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,
78 ff_vc1_ttmb_bits[i], 1, 1,
79 ff_vc1_ttmb_codes[i], 2, 2, 1);
80 init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,
81 ff_vc1_ttblk_bits[i], 1, 1,
82 ff_vc1_ttblk_codes[i], 1, 1, 1);
83 init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,
84 ff_vc1_subblkpat_bits[i], 1, 1,
85 ff_vc1_subblkpat_codes[i], 1, 1, 1);
89 init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,
90 ff_vc1_4mv_block_pattern_bits[i], 1, 1,
91 ff_vc1_4mv_block_pattern_codes[i], 1, 1, 1);
92 init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,
93 ff_vc1_cbpcy_p_bits[i], 1, 1,
94 ff_vc1_cbpcy_p_codes[i], 2, 2, 1);
95 init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,
96 ff_vc1_mv_diff_bits[i], 1, 1,
97 ff_vc1_mv_diff_codes[i], 2, 2, 1);
100 init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],
101 &vc1_ac_tables[i][0][1], 8, 4,
102 &vc1_ac_tables[i][0][0], 8, 4, 1);
103 init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,
104 &ff_msmp4_mb_i_table[0][1], 4, 2,
105 &ff_msmp4_mb_i_table[0][0], 4, 2, 1);
110 v->mvrange = 0; /* 7.1.1.18, p80 */
115 /***********************************************************************/
117 * @defgroup bitplane VC9 Bitplane decoding
122 /** @addtogroup bitplane
135 /** @} */ //imode defines
137 /** Decode rows by checking if they are skipped
138 * @param plane Buffer to store decoded bits
139 * @param[in] width Width of this buffer
140 * @param[in] height Height of this buffer
141 * @param[in] stride of this buffer
143 static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
146 for (y=0; y<height; y++){
147 if (!get_bits1(gb)) //rowskip
148 memset(plane, 0, width);
150 for (x=0; x<width; x++)
151 plane[x] = get_bits1(gb);
156 /** Decode columns by checking if they are skipped
157 * @param plane Buffer to store decoded bits
158 * @param[in] width Width of this buffer
159 * @param[in] height Height of this buffer
160 * @param[in] stride of this buffer
161 * @todo FIXME: Optimize
163 static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
166 for (x=0; x<width; x++){
167 if (!get_bits1(gb)) //colskip
168 for (y=0; y<height; y++)
171 for (y=0; y<height; y++)
172 plane[y*stride] = get_bits1(gb);
177 /** Decode a bitplane's bits
178 * @param bp Bitplane where to store the decode bits
179 * @param v VC-1 context for bit reading and logging
181 * @todo FIXME: Optimize
183 static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)
185 GetBitContext *gb = &v->s.gb;
187 int imode, x, y, code, offset;
188 uint8_t invert, *planep = data;
189 int width, height, stride;
191 width = v->s.mb_width;
192 height = v->s.mb_height;
193 stride = v->s.mb_stride;
194 invert = get_bits1(gb);
195 imode = get_vlc2(gb, ff_vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);
201 //Data is actually read in the MB layer (same for all tests == "raw")
202 *raw_flag = 1; //invert ignored
206 if ((height * width) & 1)
208 *planep++ = get_bits1(gb);
212 // decode bitplane as one long line
213 for (y = offset; y < height * width; y += 2) {
214 code = get_vlc2(gb, ff_vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
215 *planep++ = code & 1;
217 if(offset == width) {
219 planep += stride - width;
221 *planep++ = code >> 1;
223 if(offset == width) {
225 planep += stride - width;
231 if(!(height % 3) && (width % 3)) { // use 2x3 decoding
232 for(y = 0; y < height; y+= 3) {
233 for(x = width & 1; x < width; x += 2) {
234 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
236 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
239 planep[x + 0] = (code >> 0) & 1;
240 planep[x + 1] = (code >> 1) & 1;
241 planep[x + 0 + stride] = (code >> 2) & 1;
242 planep[x + 1 + stride] = (code >> 3) & 1;
243 planep[x + 0 + stride * 2] = (code >> 4) & 1;
244 planep[x + 1 + stride * 2] = (code >> 5) & 1;
246 planep += stride * 3;
248 if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);
250 planep += (height & 1) * stride;
251 for(y = height & 1; y < height; y += 2) {
252 for(x = width % 3; x < width; x += 3) {
253 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
255 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
258 planep[x + 0] = (code >> 0) & 1;
259 planep[x + 1] = (code >> 1) & 1;
260 planep[x + 2] = (code >> 2) & 1;
261 planep[x + 0 + stride] = (code >> 3) & 1;
262 planep[x + 1 + stride] = (code >> 4) & 1;
263 planep[x + 2 + stride] = (code >> 5) & 1;
265 planep += stride * 2;
268 if(x) decode_colskip(data , x, height , stride, &v->s.gb);
269 if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);
273 decode_rowskip(data, width, height, stride, &v->s.gb);
276 decode_colskip(data, width, height, stride, &v->s.gb);
281 /* Applying diff operator */
282 if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
286 for (x=1; x<width; x++)
287 planep[x] ^= planep[x-1];
288 for (y=1; y<height; y++)
291 planep[0] ^= planep[-stride];
292 for (x=1; x<width; x++)
294 if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;
295 else planep[x] ^= planep[x-1];
302 for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride
304 return (imode<<1) + invert;
307 /** @} */ //Bitplane group
309 #define FILTSIGN(a) ((a) >= 0 ? 1 : -1)
311 * VC-1 in-loop deblocking filter for one line
312 * @param src source block type
313 * @param pq block quantizer
314 * @return whether other 3 pairs should be filtered or not
317 static int vc1_filter_line(uint8_t* src, int stride, int pq){
318 int a0, a1, a2, a3, d, clip, filt3 = 0;
319 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
321 a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
323 a1 = (2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3;
324 a2 = (2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3;
325 a3 = FFMIN(FFABS(a1), FFABS(a2));
327 d = 5 * ((a0 >=0 ? a3 : -a3) - a0) / 8;
328 clip = (src[-1*stride] - src[ 0*stride])/2;
332 d = av_clip(d, 0, clip);
334 d = av_clip(d, clip, 0);
335 src[-1*stride] = cm[src[-1*stride] - d];
336 src[ 0*stride] = cm[src[ 0*stride] + d];
344 * VC-1 in-loop deblocking filter
345 * @param src source block type
346 * @param len edge length to filter (4 or 8 pixels)
347 * @param pq block quantizer
350 static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
355 for(i = 0; i < len; i += 4){
356 filt3 = vc1_filter_line(src + 2*step, stride, pq);
358 vc1_filter_line(src + 0*step, stride, pq);
359 vc1_filter_line(src + 1*step, stride, pq);
360 vc1_filter_line(src + 3*step, stride, pq);
366 static void vc1_loop_filter_iblk(MpegEncContext *s, int pq)
369 if(!s->first_slice_line)
370 vc1_loop_filter(s->dest[0], 1, s->linesize, 16, pq);
371 vc1_loop_filter(s->dest[0] + 8*s->linesize, 1, s->linesize, 16, pq);
372 for(i = !s->mb_x*8; i < 16; i += 8)
373 vc1_loop_filter(s->dest[0] + i, s->linesize, 1, 16, pq);
374 for(j = 0; j < 2; j++){
375 if(!s->first_slice_line)
376 vc1_loop_filter(s->dest[j+1], 1, s->uvlinesize, 8, pq);
378 vc1_loop_filter(s->dest[j+1], s->uvlinesize, 1, 8, pq);
382 /***********************************************************************/
383 /** VOP Dquant decoding
384 * @param v VC-1 Context
386 static int vop_dquant_decoding(VC1Context *v)
388 GetBitContext *gb = &v->s.gb;
394 pqdiff = get_bits(gb, 3);
395 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
396 else v->altpq = v->pq + pqdiff + 1;
400 v->dquantfrm = get_bits1(gb);
403 v->dqprofile = get_bits(gb, 2);
404 switch (v->dqprofile)
406 case DQPROFILE_SINGLE_EDGE:
407 case DQPROFILE_DOUBLE_EDGES:
408 v->dqsbedge = get_bits(gb, 2);
410 case DQPROFILE_ALL_MBS:
411 v->dqbilevel = get_bits1(gb);
414 default: break; //Forbidden ?
416 if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)
418 pqdiff = get_bits(gb, 3);
419 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
420 else v->altpq = v->pq + pqdiff + 1;
427 /** Put block onto picture
429 static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
433 DSPContext *dsp = &v->s.dsp;
437 for(k = 0; k < 6; k++)
438 for(j = 0; j < 8; j++)
439 for(i = 0; i < 8; i++)
440 block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;
443 ys = v->s.current_picture.linesize[0];
444 us = v->s.current_picture.linesize[1];
445 vs = v->s.current_picture.linesize[2];
448 dsp->put_pixels_clamped(block[0], Y, ys);
449 dsp->put_pixels_clamped(block[1], Y + 8, ys);
451 dsp->put_pixels_clamped(block[2], Y, ys);
452 dsp->put_pixels_clamped(block[3], Y + 8, ys);
454 if(!(v->s.flags & CODEC_FLAG_GRAY)) {
455 dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
456 dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
460 /** Do motion compensation over 1 macroblock
461 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
463 static void vc1_mc_1mv(VC1Context *v, int dir)
465 MpegEncContext *s = &v->s;
466 DSPContext *dsp = &v->s.dsp;
467 uint8_t *srcY, *srcU, *srcV;
468 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
470 if(!v->s.last_picture.data[0])return;
472 mx = s->mv[dir][0][0];
473 my = s->mv[dir][0][1];
475 // store motion vectors for further use in B frames
476 if(s->pict_type == FF_P_TYPE) {
477 s->current_picture.motion_val[1][s->block_index[0]][0] = mx;
478 s->current_picture.motion_val[1][s->block_index[0]][1] = my;
480 uvmx = (mx + ((mx & 3) == 3)) >> 1;
481 uvmy = (my + ((my & 3) == 3)) >> 1;
483 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
484 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
487 srcY = s->last_picture.data[0];
488 srcU = s->last_picture.data[1];
489 srcV = s->last_picture.data[2];
491 srcY = s->next_picture.data[0];
492 srcU = s->next_picture.data[1];
493 srcV = s->next_picture.data[2];
496 src_x = s->mb_x * 16 + (mx >> 2);
497 src_y = s->mb_y * 16 + (my >> 2);
498 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
499 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
501 if(v->profile != PROFILE_ADVANCED){
502 src_x = av_clip( src_x, -16, s->mb_width * 16);
503 src_y = av_clip( src_y, -16, s->mb_height * 16);
504 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
505 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
507 src_x = av_clip( src_x, -17, s->avctx->coded_width);
508 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
509 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
510 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
513 srcY += src_y * s->linesize + src_x;
514 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
515 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
517 /* for grayscale we should not try to read from unknown area */
518 if(s->flags & CODEC_FLAG_GRAY) {
519 srcU = s->edge_emu_buffer + 18 * s->linesize;
520 srcV = s->edge_emu_buffer + 18 * s->linesize;
523 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
524 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3
525 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){
526 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
528 srcY -= s->mspel * (1 + s->linesize);
529 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
530 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
531 srcY = s->edge_emu_buffer;
532 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
533 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
534 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
535 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
538 /* if we deal with range reduction we need to scale source blocks */
544 for(j = 0; j < 17 + s->mspel*2; j++) {
545 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
548 src = srcU; src2 = srcV;
549 for(j = 0; j < 9; j++) {
550 for(i = 0; i < 9; i++) {
551 src[i] = ((src[i] - 128) >> 1) + 128;
552 src2[i] = ((src2[i] - 128) >> 1) + 128;
554 src += s->uvlinesize;
555 src2 += s->uvlinesize;
558 /* if we deal with intensity compensation we need to scale source blocks */
559 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
564 for(j = 0; j < 17 + s->mspel*2; j++) {
565 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];
568 src = srcU; src2 = srcV;
569 for(j = 0; j < 9; j++) {
570 for(i = 0; i < 9; i++) {
571 src[i] = v->lutuv[src[i]];
572 src2[i] = v->lutuv[src2[i]];
574 src += s->uvlinesize;
575 src2 += s->uvlinesize;
578 srcY += s->mspel * (1 + s->linesize);
582 dxy = ((my & 3) << 2) | (mx & 3);
583 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] , srcY , s->linesize, v->rnd);
584 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
585 srcY += s->linesize * 8;
586 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize , srcY , s->linesize, v->rnd);
587 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
588 } else { // hpel mc - always used for luma
589 dxy = (my & 2) | ((mx & 2) >> 1);
592 dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
594 dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
597 if(s->flags & CODEC_FLAG_GRAY) return;
598 /* Chroma MC always uses qpel bilinear */
599 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
603 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
604 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
606 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
607 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
611 /** Do motion compensation for 4-MV macroblock - luminance block
613 static void vc1_mc_4mv_luma(VC1Context *v, int n)
615 MpegEncContext *s = &v->s;
616 DSPContext *dsp = &v->s.dsp;
618 int dxy, mx, my, src_x, src_y;
621 if(!v->s.last_picture.data[0])return;
624 srcY = s->last_picture.data[0];
626 off = s->linesize * 4 * (n&2) + (n&1) * 8;
628 src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);
629 src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);
631 if(v->profile != PROFILE_ADVANCED){
632 src_x = av_clip( src_x, -16, s->mb_width * 16);
633 src_y = av_clip( src_y, -16, s->mb_height * 16);
635 src_x = av_clip( src_x, -17, s->avctx->coded_width);
636 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
639 srcY += src_y * s->linesize + src_x;
641 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
642 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2
643 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){
644 srcY -= s->mspel * (1 + s->linesize);
645 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,
646 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
647 srcY = s->edge_emu_buffer;
648 /* if we deal with range reduction we need to scale source blocks */
654 for(j = 0; j < 9 + s->mspel*2; j++) {
655 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
659 /* if we deal with intensity compensation we need to scale source blocks */
660 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
665 for(j = 0; j < 9 + s->mspel*2; j++) {
666 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];
670 srcY += s->mspel * (1 + s->linesize);
674 dxy = ((my & 3) << 2) | (mx & 3);
675 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);
676 } else { // hpel mc - always used for luma
677 dxy = (my & 2) | ((mx & 2) >> 1);
679 dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
681 dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
685 static inline int median4(int a, int b, int c, int d)
688 if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
689 else return (FFMIN(b, c) + FFMAX(a, d)) / 2;
691 if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
692 else return (FFMIN(a, c) + FFMAX(b, d)) / 2;
697 /** Do motion compensation for 4-MV macroblock - both chroma blocks
699 static void vc1_mc_4mv_chroma(VC1Context *v)
701 MpegEncContext *s = &v->s;
702 DSPContext *dsp = &v->s.dsp;
703 uint8_t *srcU, *srcV;
704 int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;
705 int i, idx, tx = 0, ty = 0;
706 int mvx[4], mvy[4], intra[4];
707 static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
709 if(!v->s.last_picture.data[0])return;
710 if(s->flags & CODEC_FLAG_GRAY) return;
712 for(i = 0; i < 4; i++) {
713 mvx[i] = s->mv[0][i][0];
714 mvy[i] = s->mv[0][i][1];
715 intra[i] = v->mb_type[0][s->block_index[i]];
718 /* calculate chroma MV vector from four luma MVs */
719 idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];
720 if(!idx) { // all blocks are inter
721 tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
722 ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
723 } else if(count[idx] == 1) { // 3 inter blocks
726 tx = mid_pred(mvx[1], mvx[2], mvx[3]);
727 ty = mid_pred(mvy[1], mvy[2], mvy[3]);
730 tx = mid_pred(mvx[0], mvx[2], mvx[3]);
731 ty = mid_pred(mvy[0], mvy[2], mvy[3]);
734 tx = mid_pred(mvx[0], mvx[1], mvx[3]);
735 ty = mid_pred(mvy[0], mvy[1], mvy[3]);
738 tx = mid_pred(mvx[0], mvx[1], mvx[2]);
739 ty = mid_pred(mvy[0], mvy[1], mvy[2]);
742 } else if(count[idx] == 2) {
744 for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}
745 for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}
746 tx = (mvx[t1] + mvx[t2]) / 2;
747 ty = (mvy[t1] + mvy[t2]) / 2;
749 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
750 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
751 return; //no need to do MC for inter blocks
754 s->current_picture.motion_val[1][s->block_index[0]][0] = tx;
755 s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
756 uvmx = (tx + ((tx&3) == 3)) >> 1;
757 uvmy = (ty + ((ty&3) == 3)) >> 1;
759 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
760 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
763 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
764 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
766 if(v->profile != PROFILE_ADVANCED){
767 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
768 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
770 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
771 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
774 srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
775 srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
776 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
777 || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
778 || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){
779 ff_emulated_edge_mc(s->edge_emu_buffer , srcU, s->uvlinesize, 8+1, 8+1,
780 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
781 ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,
782 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
783 srcU = s->edge_emu_buffer;
784 srcV = s->edge_emu_buffer + 16;
786 /* if we deal with range reduction we need to scale source blocks */
791 src = srcU; src2 = srcV;
792 for(j = 0; j < 9; j++) {
793 for(i = 0; i < 9; i++) {
794 src[i] = ((src[i] - 128) >> 1) + 128;
795 src2[i] = ((src2[i] - 128) >> 1) + 128;
797 src += s->uvlinesize;
798 src2 += s->uvlinesize;
801 /* if we deal with intensity compensation we need to scale source blocks */
802 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
806 src = srcU; src2 = srcV;
807 for(j = 0; j < 9; j++) {
808 for(i = 0; i < 9; i++) {
809 src[i] = v->lutuv[src[i]];
810 src2[i] = v->lutuv[src2[i]];
812 src += s->uvlinesize;
813 src2 += s->uvlinesize;
818 /* Chroma MC always uses qpel bilinear */
819 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
823 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
824 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
826 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
827 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
831 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);
834 * Decode Simple/Main Profiles sequence header
835 * @see Figure 7-8, p16-17
836 * @param avctx Codec context
837 * @param gb GetBit context initialized from Codec context extra_data
840 static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)
842 VC1Context *v = avctx->priv_data;
844 av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));
845 v->profile = get_bits(gb, 2);
846 if (v->profile == PROFILE_COMPLEX)
848 av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");
851 if (v->profile == PROFILE_ADVANCED)
853 v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz;
854 v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz;
855 return decode_sequence_header_adv(v, gb);
859 v->zz_8x4 = wmv2_scantableA;
860 v->zz_4x8 = wmv2_scantableB;
861 v->res_sm = get_bits(gb, 2); //reserved
864 av_log(avctx, AV_LOG_ERROR,
865 "Reserved RES_SM=%i is forbidden\n", v->res_sm);
871 v->frmrtq_postproc = get_bits(gb, 3); //common
872 // (bitrate-32kbps)/64kbps
873 v->bitrtq_postproc = get_bits(gb, 5); //common
874 v->s.loop_filter = get_bits1(gb); //common
875 if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)
877 av_log(avctx, AV_LOG_ERROR,
878 "LOOPFILTER shell not be enabled in simple profile\n");
880 if(v->s.avctx->skip_loop_filter >= AVDISCARD_ALL)
881 v->s.loop_filter = 0;
883 v->res_x8 = get_bits1(gb); //reserved
884 v->multires = get_bits1(gb);
885 v->res_fasttx = get_bits1(gb);
888 v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct;
889 v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;
890 v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;
891 v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;
894 v->fastuvmc = get_bits1(gb); //common
895 if (!v->profile && !v->fastuvmc)
897 av_log(avctx, AV_LOG_ERROR,
898 "FASTUVMC unavailable in Simple Profile\n");
901 v->extended_mv = get_bits1(gb); //common
902 if (!v->profile && v->extended_mv)
904 av_log(avctx, AV_LOG_ERROR,
905 "Extended MVs unavailable in Simple Profile\n");
908 v->dquant = get_bits(gb, 2); //common
909 v->vstransform = get_bits1(gb); //common
911 v->res_transtab = get_bits1(gb);
914 av_log(avctx, AV_LOG_ERROR,
915 "1 for reserved RES_TRANSTAB is forbidden\n");
919 v->overlap = get_bits1(gb); //common
921 v->s.resync_marker = get_bits1(gb);
922 v->rangered = get_bits1(gb);
923 if (v->rangered && v->profile == PROFILE_SIMPLE)
925 av_log(avctx, AV_LOG_INFO,
926 "RANGERED should be set to 0 in simple profile\n");
929 v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common
930 v->quantizer_mode = get_bits(gb, 2); //common
932 v->finterpflag = get_bits1(gb); //common
933 v->res_rtm_flag = get_bits1(gb); //reserved
934 if (!v->res_rtm_flag)
936 // av_log(avctx, AV_LOG_ERROR,
937 // "0 for reserved RES_RTM_FLAG is forbidden\n");
938 av_log(avctx, AV_LOG_ERROR,
939 "Old WMV3 version detected, only I-frames will be decoded\n");
942 //TODO: figure out what they mean (always 0x402F)
943 if(!v->res_fasttx) skip_bits(gb, 16);
944 av_log(avctx, AV_LOG_DEBUG,
945 "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
946 "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"
947 "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"
948 "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",
949 v->profile, v->frmrtq_postproc, v->bitrtq_postproc,
950 v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,
951 v->rangered, v->vstransform, v->overlap, v->s.resync_marker,
952 v->dquant, v->quantizer_mode, avctx->max_b_frames
957 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)
960 v->level = get_bits(gb, 3);
963 av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);
965 v->chromaformat = get_bits(gb, 2);
966 if (v->chromaformat != 1)
968 av_log(v->s.avctx, AV_LOG_ERROR,
969 "Only 4:2:0 chroma format supported\n");
974 v->frmrtq_postproc = get_bits(gb, 3); //common
975 // (bitrate-32kbps)/64kbps
976 v->bitrtq_postproc = get_bits(gb, 5); //common
977 v->postprocflag = get_bits1(gb); //common
979 v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;
980 v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;
981 v->s.avctx->width = v->s.avctx->coded_width;
982 v->s.avctx->height = v->s.avctx->coded_height;
983 v->broadcast = get_bits1(gb);
984 v->interlace = get_bits1(gb);
985 v->tfcntrflag = get_bits1(gb);
986 v->finterpflag = get_bits1(gb);
987 skip_bits1(gb); // reserved
989 v->s.h_edge_pos = v->s.avctx->coded_width;
990 v->s.v_edge_pos = v->s.avctx->coded_height;
992 av_log(v->s.avctx, AV_LOG_DEBUG,
993 "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
994 "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"
995 "TFCTRflag=%i, FINTERPflag=%i\n",
996 v->level, v->frmrtq_postproc, v->bitrtq_postproc,
997 v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,
998 v->tfcntrflag, v->finterpflag
1001 v->psf = get_bits1(gb);
1002 if(v->psf) { //PsF, 6.1.13
1003 av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");
1006 v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
1007 if(get_bits1(gb)) { //Display Info - decoding is not affected by it
1009 av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
1010 v->s.avctx->width = v->s.width = w = get_bits(gb, 14) + 1;
1011 v->s.avctx->height = v->s.height = h = get_bits(gb, 14) + 1;
1012 av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
1014 ar = get_bits(gb, 4);
1016 v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
1018 w = get_bits(gb, 8);
1019 h = get_bits(gb, 8);
1020 v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
1023 if(get_bits1(gb)){ //framerate stuff
1025 v->s.avctx->time_base.num = 32;
1026 v->s.avctx->time_base.den = get_bits(gb, 16) + 1;
1029 nr = get_bits(gb, 8);
1030 dr = get_bits(gb, 4);
1031 if(nr && nr < 8 && dr && dr < 3){
1032 v->s.avctx->time_base.num = ff_vc1_fps_dr[dr - 1];
1033 v->s.avctx->time_base.den = ff_vc1_fps_nr[nr - 1] * 1000;
1039 v->color_prim = get_bits(gb, 8);
1040 v->transfer_char = get_bits(gb, 8);
1041 v->matrix_coef = get_bits(gb, 8);
1045 v->hrd_param_flag = get_bits1(gb);
1046 if(v->hrd_param_flag) {
1048 v->hrd_num_leaky_buckets = get_bits(gb, 5);
1049 skip_bits(gb, 4); //bitrate exponent
1050 skip_bits(gb, 4); //buffer size exponent
1051 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1052 skip_bits(gb, 16); //hrd_rate[n]
1053 skip_bits(gb, 16); //hrd_buffer[n]
1059 static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb)
1061 VC1Context *v = avctx->priv_data;
1062 int i, blink, clentry, refdist;
1064 av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32));
1065 blink = get_bits1(gb); // broken link
1066 clentry = get_bits1(gb); // closed entry
1067 v->panscanflag = get_bits1(gb);
1068 refdist = get_bits1(gb); // refdist flag
1069 v->s.loop_filter = get_bits1(gb);
1070 v->fastuvmc = get_bits1(gb);
1071 v->extended_mv = get_bits1(gb);
1072 v->dquant = get_bits(gb, 2);
1073 v->vstransform = get_bits1(gb);
1074 v->overlap = get_bits1(gb);
1075 v->quantizer_mode = get_bits(gb, 2);
1077 if(v->hrd_param_flag){
1078 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1079 skip_bits(gb, 8); //hrd_full[n]
1084 avctx->coded_width = (get_bits(gb, 12)+1)<<1;
1085 avctx->coded_height = (get_bits(gb, 12)+1)<<1;
1088 v->extended_dmv = get_bits1(gb);
1090 av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n");
1091 skip_bits(gb, 3); // Y range, ignored for now
1094 av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n");
1095 skip_bits(gb, 3); // UV range, ignored for now
1098 av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n"
1099 "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n"
1100 "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n"
1101 "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n",
1102 blink, clentry, v->panscanflag, refdist, v->s.loop_filter,
1103 v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode);
1108 static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)
1110 int pqindex, lowquant, status;
1112 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1113 skip_bits(gb, 2); //framecnt unused
1115 if (v->rangered) v->rangeredfrm = get_bits1(gb);
1116 v->s.pict_type = get_bits1(gb);
1117 if (v->s.avctx->max_b_frames) {
1118 if (!v->s.pict_type) {
1119 if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE;
1120 else v->s.pict_type = FF_B_TYPE;
1121 } else v->s.pict_type = FF_P_TYPE;
1122 } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE;
1125 if(v->s.pict_type == FF_B_TYPE) {
1126 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1127 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1128 if(v->bfraction == 0) {
1129 v->s.pict_type = FF_BI_TYPE;
1132 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1133 skip_bits(gb, 7); // skip buffer fullness
1136 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1138 if(v->s.pict_type == FF_P_TYPE)
1141 /* Quantizer stuff */
1142 pqindex = get_bits(gb, 5);
1143 if(!pqindex) return -1;
1144 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1145 v->pq = ff_vc1_pquant_table[0][pqindex];
1147 v->pq = ff_vc1_pquant_table[1][pqindex];
1150 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1151 v->pquantizer = pqindex < 9;
1152 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1154 v->pqindex = pqindex;
1155 if (pqindex < 9) v->halfpq = get_bits1(gb);
1157 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1158 v->pquantizer = get_bits1(gb);
1160 if (v->extended_mv == 1) v->mvrange = get_unary(gb, 0, 3);
1161 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1162 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1163 v->range_x = 1 << (v->k_x - 1);
1164 v->range_y = 1 << (v->k_y - 1);
1165 if (v->profile == PROFILE_ADVANCED)
1167 if (v->postprocflag) v->postproc = get_bits1(gb);
1170 if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2);
1172 if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){
1173 v->x8_type = get_bits1(gb);
1174 }else v->x8_type = 0;
1175 //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",
1176 // (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);
1178 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1180 switch(v->s.pict_type) {
1182 if (v->pq < 5) v->tt_index = 0;
1183 else if(v->pq < 13) v->tt_index = 1;
1184 else v->tt_index = 2;
1186 lowquant = (v->pq > 12) ? 0 : 1;
1187 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1188 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1190 int scale, shift, i;
1191 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1192 v->lumscale = get_bits(gb, 6);
1193 v->lumshift = get_bits(gb, 6);
1195 /* fill lookup tables for intensity compensation */
1198 shift = (255 - v->lumshift * 2) << 6;
1199 if(v->lumshift > 31)
1202 scale = v->lumscale + 32;
1203 if(v->lumshift > 31)
1204 shift = (v->lumshift - 64) << 6;
1206 shift = v->lumshift << 6;
1208 for(i = 0; i < 256; i++) {
1209 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1210 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1213 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1214 v->s.quarter_sample = 0;
1215 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1216 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1217 v->s.quarter_sample = 0;
1219 v->s.quarter_sample = 1;
1221 v->s.quarter_sample = 1;
1222 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));
1224 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1225 v->mv_mode2 == MV_PMODE_MIXED_MV)
1226 || v->mv_mode == MV_PMODE_MIXED_MV)
1228 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1229 if (status < 0) return -1;
1230 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1231 "Imode: %i, Invert: %i\n", status>>1, status&1);
1233 v->mv_type_is_raw = 0;
1234 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1236 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1237 if (status < 0) return -1;
1238 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1239 "Imode: %i, Invert: %i\n", status>>1, status&1);
1241 /* Hopefully this is correct for P frames */
1242 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1243 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1247 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1248 vop_dquant_decoding(v);
1251 v->ttfrm = 0; //FIXME Is that so ?
1254 v->ttmbf = get_bits1(gb);
1257 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1265 if (v->pq < 5) v->tt_index = 0;
1266 else if(v->pq < 13) v->tt_index = 1;
1267 else v->tt_index = 2;
1269 lowquant = (v->pq > 12) ? 0 : 1;
1270 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1271 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1272 v->s.mspel = v->s.quarter_sample;
1274 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1275 if (status < 0) return -1;
1276 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1277 "Imode: %i, Invert: %i\n", status>>1, status&1);
1278 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1279 if (status < 0) return -1;
1280 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1281 "Imode: %i, Invert: %i\n", status>>1, status&1);
1283 v->s.mv_table_index = get_bits(gb, 2);
1284 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1288 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1289 vop_dquant_decoding(v);
1295 v->ttmbf = get_bits1(gb);
1298 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1310 v->c_ac_table_index = decode012(gb);
1311 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1313 v->y_ac_table_index = decode012(gb);
1316 v->s.dc_table_index = get_bits1(gb);
1319 if(v->s.pict_type == FF_BI_TYPE) {
1320 v->s.pict_type = FF_B_TYPE;
1326 static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)
1328 int pqindex, lowquant;
1331 v->p_frame_skipped = 0;
1334 v->fcm = decode012(gb);
1335 if(v->fcm) return -1; // interlaced frames/fields are not implemented
1337 switch(get_unary(gb, 0, 4)) {
1339 v->s.pict_type = FF_P_TYPE;
1342 v->s.pict_type = FF_B_TYPE;
1345 v->s.pict_type = FF_I_TYPE;
1348 v->s.pict_type = FF_BI_TYPE;
1351 v->s.pict_type = FF_P_TYPE; // skipped pic
1352 v->p_frame_skipped = 1;
1358 if(!v->interlace || v->psf) {
1359 v->rptfrm = get_bits(gb, 2);
1361 v->tff = get_bits1(gb);
1362 v->rptfrm = get_bits1(gb);
1365 if(v->panscanflag) {
1368 v->rnd = get_bits1(gb);
1370 v->uvsamp = get_bits1(gb);
1371 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1372 if(v->s.pict_type == FF_B_TYPE) {
1373 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1374 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1375 if(v->bfraction == 0) {
1376 v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */
1379 pqindex = get_bits(gb, 5);
1380 if(!pqindex) return -1;
1381 v->pqindex = pqindex;
1382 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1383 v->pq = ff_vc1_pquant_table[0][pqindex];
1385 v->pq = ff_vc1_pquant_table[1][pqindex];
1388 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1389 v->pquantizer = pqindex < 9;
1390 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1392 v->pqindex = pqindex;
1393 if (pqindex < 9) v->halfpq = get_bits1(gb);
1395 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1396 v->pquantizer = get_bits1(gb);
1398 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1400 switch(v->s.pict_type) {
1403 status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v);
1404 if (status < 0) return -1;
1405 av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: "
1406 "Imode: %i, Invert: %i\n", status>>1, status&1);
1407 v->condover = CONDOVER_NONE;
1408 if(v->overlap && v->pq <= 8) {
1409 v->condover = decode012(gb);
1410 if(v->condover == CONDOVER_SELECT) {
1411 status = bitplane_decoding(v->over_flags_plane, &v->overflg_is_raw, v);
1412 if (status < 0) return -1;
1413 av_log(v->s.avctx, AV_LOG_DEBUG, "CONDOVER plane encoding: "
1414 "Imode: %i, Invert: %i\n", status>>1, status&1);
1420 v->postproc = get_bits1(gb);
1421 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1422 else v->mvrange = 0;
1423 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1424 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1425 v->range_x = 1 << (v->k_x - 1);
1426 v->range_y = 1 << (v->k_y - 1);
1428 if (v->pq < 5) v->tt_index = 0;
1429 else if(v->pq < 13) v->tt_index = 1;
1430 else v->tt_index = 2;
1432 lowquant = (v->pq > 12) ? 0 : 1;
1433 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1434 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1436 int scale, shift, i;
1437 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1438 v->lumscale = get_bits(gb, 6);
1439 v->lumshift = get_bits(gb, 6);
1440 /* fill lookup tables for intensity compensation */
1443 shift = (255 - v->lumshift * 2) << 6;
1444 if(v->lumshift > 31)
1447 scale = v->lumscale + 32;
1448 if(v->lumshift > 31)
1449 shift = (v->lumshift - 64) << 6;
1451 shift = v->lumshift << 6;
1453 for(i = 0; i < 256; i++) {
1454 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1455 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1459 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1460 v->s.quarter_sample = 0;
1461 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1462 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1463 v->s.quarter_sample = 0;
1465 v->s.quarter_sample = 1;
1467 v->s.quarter_sample = 1;
1468 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));
1470 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1471 v->mv_mode2 == MV_PMODE_MIXED_MV)
1472 || v->mv_mode == MV_PMODE_MIXED_MV)
1474 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1475 if (status < 0) return -1;
1476 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1477 "Imode: %i, Invert: %i\n", status>>1, status&1);
1479 v->mv_type_is_raw = 0;
1480 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1482 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1483 if (status < 0) return -1;
1484 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1485 "Imode: %i, Invert: %i\n", status>>1, status&1);
1487 /* Hopefully this is correct for P frames */
1488 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1489 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1492 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1493 vop_dquant_decoding(v);
1496 v->ttfrm = 0; //FIXME Is that so ?
1499 v->ttmbf = get_bits1(gb);
1502 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1511 v->postproc = get_bits1(gb);
1512 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1513 else v->mvrange = 0;
1514 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1515 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1516 v->range_x = 1 << (v->k_x - 1);
1517 v->range_y = 1 << (v->k_y - 1);
1519 if (v->pq < 5) v->tt_index = 0;
1520 else if(v->pq < 13) v->tt_index = 1;
1521 else v->tt_index = 2;
1523 lowquant = (v->pq > 12) ? 0 : 1;
1524 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1525 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1526 v->s.mspel = v->s.quarter_sample;
1528 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1529 if (status < 0) return -1;
1530 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1531 "Imode: %i, Invert: %i\n", status>>1, status&1);
1532 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1533 if (status < 0) return -1;
1534 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1535 "Imode: %i, Invert: %i\n", status>>1, status&1);
1537 v->s.mv_table_index = get_bits(gb, 2);
1538 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1542 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1543 vop_dquant_decoding(v);
1549 v->ttmbf = get_bits1(gb);
1552 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1562 v->c_ac_table_index = decode012(gb);
1563 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1565 v->y_ac_table_index = decode012(gb);
1568 v->s.dc_table_index = get_bits1(gb);
1569 if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) {
1570 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1571 vop_dquant_decoding(v);
1575 if(v->s.pict_type == FF_BI_TYPE) {
1576 v->s.pict_type = FF_B_TYPE;
1582 /***********************************************************************/
1584 * @defgroup block VC-1 Block-level functions
1585 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
1591 * @brief Get macroblock-level quantizer scale
1593 #define GET_MQUANT() \
1597 if (v->dqprofile == DQPROFILE_ALL_MBS) \
1601 mquant = (get_bits1(gb)) ? v->altpq : v->pq; \
1605 mqdiff = get_bits(gb, 3); \
1606 if (mqdiff != 7) mquant = v->pq + mqdiff; \
1607 else mquant = get_bits(gb, 5); \
1610 if(v->dqprofile == DQPROFILE_SINGLE_EDGE) \
1611 edges = 1 << v->dqsbedge; \
1612 else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES) \
1613 edges = (3 << v->dqsbedge) % 15; \
1614 else if(v->dqprofile == DQPROFILE_FOUR_EDGES) \
1616 if((edges&1) && !s->mb_x) \
1617 mquant = v->altpq; \
1618 if((edges&2) && s->first_slice_line) \
1619 mquant = v->altpq; \
1620 if((edges&4) && s->mb_x == (s->mb_width - 1)) \
1621 mquant = v->altpq; \
1622 if((edges&8) && s->mb_y == (s->mb_height - 1)) \
1623 mquant = v->altpq; \
1627 * @def GET_MVDATA(_dmv_x, _dmv_y)
1628 * @brief Get MV differentials
1629 * @see MVDATA decoding from 8.3.5.2, p(1)20
1630 * @param _dmv_x Horizontal differential for decoded MV
1631 * @param _dmv_y Vertical differential for decoded MV
1633 #define GET_MVDATA(_dmv_x, _dmv_y) \
1634 index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table,\
1635 VC1_MV_DIFF_VLC_BITS, 2); \
1638 mb_has_coeffs = 1; \
1641 else mb_has_coeffs = 0; \
1643 if (!index) { _dmv_x = _dmv_y = 0; } \
1644 else if (index == 35) \
1646 _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample); \
1647 _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample); \
1649 else if (index == 36) \
1658 if (!s->quarter_sample && index1 == 5) val = 1; \
1660 if(size_table[index1] - val > 0) \
1661 val = get_bits(gb, size_table[index1] - val); \
1663 sign = 0 - (val&1); \
1664 _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1667 if (!s->quarter_sample && index1 == 5) val = 1; \
1669 if(size_table[index1] - val > 0) \
1670 val = get_bits(gb, size_table[index1] - val); \
1672 sign = 0 - (val&1); \
1673 _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1676 /** Predict and set motion vector
1678 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)
1680 int xy, wrap, off = 0;
1685 /* scale MV difference to be quad-pel */
1686 dmv_x <<= 1 - s->quarter_sample;
1687 dmv_y <<= 1 - s->quarter_sample;
1689 wrap = s->b8_stride;
1690 xy = s->block_index[n];
1693 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
1694 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
1695 s->current_picture.motion_val[1][xy][0] = 0;
1696 s->current_picture.motion_val[1][xy][1] = 0;
1697 if(mv1) { /* duplicate motion data for 1-MV block */
1698 s->current_picture.motion_val[0][xy + 1][0] = 0;
1699 s->current_picture.motion_val[0][xy + 1][1] = 0;
1700 s->current_picture.motion_val[0][xy + wrap][0] = 0;
1701 s->current_picture.motion_val[0][xy + wrap][1] = 0;
1702 s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
1703 s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
1704 s->current_picture.motion_val[1][xy + 1][0] = 0;
1705 s->current_picture.motion_val[1][xy + 1][1] = 0;
1706 s->current_picture.motion_val[1][xy + wrap][0] = 0;
1707 s->current_picture.motion_val[1][xy + wrap][1] = 0;
1708 s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
1709 s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
1714 C = s->current_picture.motion_val[0][xy - 1];
1715 A = s->current_picture.motion_val[0][xy - wrap];
1717 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
1719 //in 4-MV mode different blocks have different B predictor position
1722 off = (s->mb_x > 0) ? -1 : 1;
1725 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
1734 B = s->current_picture.motion_val[0][xy - wrap + off];
1736 if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds
1737 if(s->mb_width == 1) {
1741 px = mid_pred(A[0], B[0], C[0]);
1742 py = mid_pred(A[1], B[1], C[1]);
1744 } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds
1750 /* Pullback MV as specified in 8.3.5.3.4 */
1753 qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);
1754 qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);
1755 X = (s->mb_width << 6) - 4;
1756 Y = (s->mb_height << 6) - 4;
1758 if(qx + px < -60) px = -60 - qx;
1759 if(qy + py < -60) py = -60 - qy;
1761 if(qx + px < -28) px = -28 - qx;
1762 if(qy + py < -28) py = -28 - qy;
1764 if(qx + px > X) px = X - qx;
1765 if(qy + py > Y) py = Y - qy;
1767 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
1768 if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {
1769 if(is_intra[xy - wrap])
1770 sum = FFABS(px) + FFABS(py);
1772 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
1774 if(get_bits1(&s->gb)) {
1782 if(is_intra[xy - 1])
1783 sum = FFABS(px) + FFABS(py);
1785 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
1787 if(get_bits1(&s->gb)) {
1797 /* store MV using signed modulus of MV range defined in 4.11 */
1798 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
1799 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;
1800 if(mv1) { /* duplicate motion data for 1-MV block */
1801 s->current_picture.motion_val[0][xy + 1][0] = s->current_picture.motion_val[0][xy][0];
1802 s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];
1803 s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];
1804 s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];
1805 s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];
1806 s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];
1810 /** Motion compensation for direct or interpolated blocks in B-frames
1812 static void vc1_interp_mc(VC1Context *v)
1814 MpegEncContext *s = &v->s;
1815 DSPContext *dsp = &v->s.dsp;
1816 uint8_t *srcY, *srcU, *srcV;
1817 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
1819 if(!v->s.next_picture.data[0])return;
1821 mx = s->mv[1][0][0];
1822 my = s->mv[1][0][1];
1823 uvmx = (mx + ((mx & 3) == 3)) >> 1;
1824 uvmy = (my + ((my & 3) == 3)) >> 1;
1826 uvmx = uvmx + ((uvmx<0)?-(uvmx&1):(uvmx&1));
1827 uvmy = uvmy + ((uvmy<0)?-(uvmy&1):(uvmy&1));
1829 srcY = s->next_picture.data[0];
1830 srcU = s->next_picture.data[1];
1831 srcV = s->next_picture.data[2];
1833 src_x = s->mb_x * 16 + (mx >> 2);
1834 src_y = s->mb_y * 16 + (my >> 2);
1835 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
1836 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
1838 if(v->profile != PROFILE_ADVANCED){
1839 src_x = av_clip( src_x, -16, s->mb_width * 16);
1840 src_y = av_clip( src_y, -16, s->mb_height * 16);
1841 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
1842 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
1844 src_x = av_clip( src_x, -17, s->avctx->coded_width);
1845 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
1846 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
1847 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
1850 srcY += src_y * s->linesize + src_x;
1851 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
1852 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
1854 /* for grayscale we should not try to read from unknown area */
1855 if(s->flags & CODEC_FLAG_GRAY) {
1856 srcU = s->edge_emu_buffer + 18 * s->linesize;
1857 srcV = s->edge_emu_buffer + 18 * s->linesize;
1861 || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16
1862 || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){
1863 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
1865 srcY -= s->mspel * (1 + s->linesize);
1866 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
1867 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
1868 srcY = s->edge_emu_buffer;
1869 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
1870 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1871 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
1872 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1875 /* if we deal with range reduction we need to scale source blocks */
1876 if(v->rangeredfrm) {
1878 uint8_t *src, *src2;
1881 for(j = 0; j < 17 + s->mspel*2; j++) {
1882 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
1885 src = srcU; src2 = srcV;
1886 for(j = 0; j < 9; j++) {
1887 for(i = 0; i < 9; i++) {
1888 src[i] = ((src[i] - 128) >> 1) + 128;
1889 src2[i] = ((src2[i] - 128) >> 1) + 128;
1891 src += s->uvlinesize;
1892 src2 += s->uvlinesize;
1895 srcY += s->mspel * (1 + s->linesize);
1900 dxy = ((my & 1) << 1) | (mx & 1);
1902 dsp->avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
1904 if(s->flags & CODEC_FLAG_GRAY) return;
1905 /* Chroma MC always uses qpel blilinear */
1906 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
1909 dsp->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
1910 dsp->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
1913 static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)
1917 #if B_FRACTION_DEN==256
1921 return 2 * ((value * n + 255) >> 9);
1922 return (value * n + 128) >> 8;
1925 n -= B_FRACTION_DEN;
1927 return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));
1928 return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;
1932 /** Reconstruct motion vector for B-frame and do motion compensation
1934 static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode)
1937 v->mv_mode2 = v->mv_mode;
1938 v->mv_mode = MV_PMODE_INTENSITY_COMP;
1943 if(v->use_ic) v->mv_mode = v->mv_mode2;
1946 if(mode == BMV_TYPE_INTERPOLATED) {
1949 if(v->use_ic) v->mv_mode = v->mv_mode2;
1953 if(v->use_ic && (mode == BMV_TYPE_BACKWARD)) v->mv_mode = v->mv_mode2;
1954 vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));
1955 if(v->use_ic) v->mv_mode = v->mv_mode2;
1958 static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mvtype)
1960 MpegEncContext *s = &v->s;
1961 int xy, wrap, off = 0;
1966 const uint8_t *is_intra = v->mb_type[0];
1970 /* scale MV difference to be quad-pel */
1971 dmv_x[0] <<= 1 - s->quarter_sample;
1972 dmv_y[0] <<= 1 - s->quarter_sample;
1973 dmv_x[1] <<= 1 - s->quarter_sample;
1974 dmv_y[1] <<= 1 - s->quarter_sample;
1976 wrap = s->b8_stride;
1977 xy = s->block_index[0];
1980 s->current_picture.motion_val[0][xy][0] =
1981 s->current_picture.motion_val[0][xy][1] =
1982 s->current_picture.motion_val[1][xy][0] =
1983 s->current_picture.motion_val[1][xy][1] = 0;
1986 s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);
1987 s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);
1988 s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);
1989 s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);
1991 /* Pullback predicted motion vectors as specified in 8.4.5.4 */
1992 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));
1993 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));
1994 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));
1995 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));
1997 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
1998 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
1999 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2000 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2004 if((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2005 C = s->current_picture.motion_val[0][xy - 2];
2006 A = s->current_picture.motion_val[0][xy - wrap*2];
2007 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2008 B = s->current_picture.motion_val[0][xy - wrap*2 + off];
2010 if(!s->mb_x) C[0] = C[1] = 0;
2011 if(!s->first_slice_line) { // predictor A is not out of bounds
2012 if(s->mb_width == 1) {
2016 px = mid_pred(A[0], B[0], C[0]);
2017 py = mid_pred(A[1], B[1], C[1]);
2019 } else if(s->mb_x) { // predictor C is not out of bounds
2025 /* Pullback MV as specified in 8.3.5.3.4 */
2028 if(v->profile < PROFILE_ADVANCED) {
2029 qx = (s->mb_x << 5);
2030 qy = (s->mb_y << 5);
2031 X = (s->mb_width << 5) - 4;
2032 Y = (s->mb_height << 5) - 4;
2033 if(qx + px < -28) px = -28 - qx;
2034 if(qy + py < -28) py = -28 - qy;
2035 if(qx + px > X) px = X - qx;
2036 if(qy + py > Y) py = Y - qy;
2038 qx = (s->mb_x << 6);
2039 qy = (s->mb_y << 6);
2040 X = (s->mb_width << 6) - 4;
2041 Y = (s->mb_height << 6) - 4;
2042 if(qx + px < -60) px = -60 - qx;
2043 if(qy + py < -60) py = -60 - qy;
2044 if(qx + px > X) px = X - qx;
2045 if(qy + py > Y) py = Y - qy;
2048 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2049 if(0 && !s->first_slice_line && s->mb_x) {
2050 if(is_intra[xy - wrap])
2051 sum = FFABS(px) + FFABS(py);
2053 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2055 if(get_bits1(&s->gb)) {
2063 if(is_intra[xy - 2])
2064 sum = FFABS(px) + FFABS(py);
2066 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2068 if(get_bits1(&s->gb)) {
2078 /* store MV using signed modulus of MV range defined in 4.11 */
2079 s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;
2080 s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;
2082 if((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2083 C = s->current_picture.motion_val[1][xy - 2];
2084 A = s->current_picture.motion_val[1][xy - wrap*2];
2085 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2086 B = s->current_picture.motion_val[1][xy - wrap*2 + off];
2088 if(!s->mb_x) C[0] = C[1] = 0;
2089 if(!s->first_slice_line) { // predictor A is not out of bounds
2090 if(s->mb_width == 1) {
2094 px = mid_pred(A[0], B[0], C[0]);
2095 py = mid_pred(A[1], B[1], C[1]);
2097 } else if(s->mb_x) { // predictor C is not out of bounds
2103 /* Pullback MV as specified in 8.3.5.3.4 */
2106 if(v->profile < PROFILE_ADVANCED) {
2107 qx = (s->mb_x << 5);
2108 qy = (s->mb_y << 5);
2109 X = (s->mb_width << 5) - 4;
2110 Y = (s->mb_height << 5) - 4;
2111 if(qx + px < -28) px = -28 - qx;
2112 if(qy + py < -28) py = -28 - qy;
2113 if(qx + px > X) px = X - qx;
2114 if(qy + py > Y) py = Y - qy;
2116 qx = (s->mb_x << 6);
2117 qy = (s->mb_y << 6);
2118 X = (s->mb_width << 6) - 4;
2119 Y = (s->mb_height << 6) - 4;
2120 if(qx + px < -60) px = -60 - qx;
2121 if(qy + py < -60) py = -60 - qy;
2122 if(qx + px > X) px = X - qx;
2123 if(qy + py > Y) py = Y - qy;
2126 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2127 if(0 && !s->first_slice_line && s->mb_x) {
2128 if(is_intra[xy - wrap])
2129 sum = FFABS(px) + FFABS(py);
2131 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2133 if(get_bits1(&s->gb)) {
2141 if(is_intra[xy - 2])
2142 sum = FFABS(px) + FFABS(py);
2144 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2146 if(get_bits1(&s->gb)) {
2156 /* store MV using signed modulus of MV range defined in 4.11 */
2158 s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;
2159 s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;
2161 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2162 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2163 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2164 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2167 /** Get predicted DC value for I-frames only
2168 * prediction dir: left=0, top=1
2169 * @param s MpegEncContext
2170 * @param[in] n block index in the current MB
2171 * @param dc_val_ptr Pointer to DC predictor
2172 * @param dir_ptr Prediction direction for use in AC prediction
2174 static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2175 int16_t **dc_val_ptr, int *dir_ptr)
2177 int a, b, c, wrap, pred, scale;
2179 static const uint16_t dcpred[32] = {
2180 -1, 1024, 512, 341, 256, 205, 171, 146, 128,
2181 114, 102, 93, 85, 79, 73, 68, 64,
2182 60, 57, 54, 51, 49, 47, 45, 43,
2183 41, 39, 38, 37, 35, 34, 33
2186 /* find prediction - wmv3_dc_scale always used here in fact */
2187 if (n < 4) scale = s->y_dc_scale;
2188 else scale = s->c_dc_scale;
2190 wrap = s->block_wrap[n];
2191 dc_val= s->dc_val[0] + s->block_index[n];
2197 b = dc_val[ - 1 - wrap];
2198 a = dc_val[ - wrap];
2200 if (pq < 9 || !overlap)
2202 /* Set outer values */
2203 if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];
2204 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];
2208 /* Set outer values */
2209 if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;
2210 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;
2213 if (abs(a - b) <= abs(b - c)) {
2221 /* update predictor */
2222 *dc_val_ptr = &dc_val[0];
2227 /** Get predicted DC value
2228 * prediction dir: left=0, top=1
2229 * @param s MpegEncContext
2230 * @param[in] n block index in the current MB
2231 * @param dc_val_ptr Pointer to DC predictor
2232 * @param dir_ptr Prediction direction for use in AC prediction
2234 static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2235 int a_avail, int c_avail,
2236 int16_t **dc_val_ptr, int *dir_ptr)
2238 int a, b, c, wrap, pred, scale;
2240 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2243 /* find prediction - wmv3_dc_scale always used here in fact */
2244 if (n < 4) scale = s->y_dc_scale;
2245 else scale = s->c_dc_scale;
2247 wrap = s->block_wrap[n];
2248 dc_val= s->dc_val[0] + s->block_index[n];
2254 b = dc_val[ - 1 - wrap];
2255 a = dc_val[ - wrap];
2256 /* scale predictors if needed */
2257 q1 = s->current_picture.qscale_table[mb_pos];
2258 if(c_avail && (n!= 1 && n!=3)) {
2259 q2 = s->current_picture.qscale_table[mb_pos - 1];
2261 c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2263 if(a_avail && (n!= 2 && n!=3)) {
2264 q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2266 a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2268 if(a_avail && c_avail && (n!=3)) {
2271 if(n != 2) off -= s->mb_stride;
2272 q2 = s->current_picture.qscale_table[off];
2274 b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2277 if(a_avail && c_avail) {
2278 if(abs(a - b) <= abs(b - c)) {
2285 } else if(a_avail) {
2288 } else if(c_avail) {
2296 /* update predictor */
2297 *dc_val_ptr = &dc_val[0];
2303 * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles
2304 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
2308 static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)
2310 int xy, wrap, pred, a, b, c;
2312 xy = s->block_index[n];
2313 wrap = s->b8_stride;
2318 a = s->coded_block[xy - 1 ];
2319 b = s->coded_block[xy - 1 - wrap];
2320 c = s->coded_block[xy - wrap];
2329 *coded_block_ptr = &s->coded_block[xy];
2335 * Decode one AC coefficient
2336 * @param v The VC1 context
2337 * @param last Last coefficient
2338 * @param skip How much zero coefficients to skip
2339 * @param value Decoded AC coefficient value
2342 static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)
2344 GetBitContext *gb = &v->s.gb;
2345 int index, escape, run = 0, level = 0, lst = 0;
2347 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2348 if (index != vc1_ac_sizes[codingset] - 1) {
2349 run = vc1_index_decode_table[codingset][index][0];
2350 level = vc1_index_decode_table[codingset][index][1];
2351 lst = index >= vc1_last_decode_table[codingset];
2355 escape = decode210(gb);
2357 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2358 run = vc1_index_decode_table[codingset][index][0];
2359 level = vc1_index_decode_table[codingset][index][1];
2360 lst = index >= vc1_last_decode_table[codingset];
2363 level += vc1_last_delta_level_table[codingset][run];
2365 level += vc1_delta_level_table[codingset][run];
2368 run += vc1_last_delta_run_table[codingset][level] + 1;
2370 run += vc1_delta_run_table[codingset][level] + 1;
2376 lst = get_bits1(gb);
2377 if(v->s.esc3_level_length == 0) {
2378 if(v->pq < 8 || v->dquantfrm) { // table 59
2379 v->s.esc3_level_length = get_bits(gb, 3);
2380 if(!v->s.esc3_level_length)
2381 v->s.esc3_level_length = get_bits(gb, 2) + 8;
2383 v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;
2385 v->s.esc3_run_length = 3 + get_bits(gb, 2);
2387 run = get_bits(gb, v->s.esc3_run_length);
2388 sign = get_bits1(gb);
2389 level = get_bits(gb, v->s.esc3_level_length);
2400 /** Decode intra block in intra frames - should be faster than decode_intra_block
2401 * @param v VC1Context
2402 * @param block block to decode
2403 * @param coded are AC coeffs present or not
2404 * @param codingset set of VLC to decode data
2406 static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)
2408 GetBitContext *gb = &v->s.gb;
2409 MpegEncContext *s = &v->s;
2410 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2413 int16_t *ac_val, *ac_val2;
2416 /* Get DC differential */
2418 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2420 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2423 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2428 if (dcdiff == 119 /* ESC index value */)
2430 /* TODO: Optimize */
2431 if (v->pq == 1) dcdiff = get_bits(gb, 10);
2432 else if (v->pq == 2) dcdiff = get_bits(gb, 9);
2433 else dcdiff = get_bits(gb, 8);
2438 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2439 else if (v->pq == 2)
2440 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2447 dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);
2450 /* Store the quantized DC coeff, used for prediction */
2452 block[0] = dcdiff * s->y_dc_scale;
2454 block[0] = dcdiff * s->c_dc_scale;
2467 int last = 0, skip, value;
2468 const int8_t *zz_table;
2472 scale = v->pq * 2 + v->halfpq;
2476 zz_table = wmv1_scantable[2];
2478 zz_table = wmv1_scantable[3];
2480 zz_table = wmv1_scantable[1];
2482 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2484 if(dc_pred_dir) //left
2487 ac_val -= 16 * s->block_wrap[n];
2490 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2494 block[zz_table[i++]] = value;
2497 /* apply AC prediction if needed */
2499 if(dc_pred_dir) { //left
2500 for(k = 1; k < 8; k++)
2501 block[k << 3] += ac_val[k];
2503 for(k = 1; k < 8; k++)
2504 block[k] += ac_val[k + 8];
2507 /* save AC coeffs for further prediction */
2508 for(k = 1; k < 8; k++) {
2509 ac_val2[k] = block[k << 3];
2510 ac_val2[k + 8] = block[k];
2513 /* scale AC coeffs */
2514 for(k = 1; k < 64; k++)
2518 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2521 if(s->ac_pred) i = 63;
2527 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2530 scale = v->pq * 2 + v->halfpq;
2531 memset(ac_val2, 0, 16 * 2);
2532 if(dc_pred_dir) {//left
2535 memcpy(ac_val2, ac_val, 8 * 2);
2537 ac_val -= 16 * s->block_wrap[n];
2539 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2542 /* apply AC prediction if needed */
2544 if(dc_pred_dir) { //left
2545 for(k = 1; k < 8; k++) {
2546 block[k << 3] = ac_val[k] * scale;
2547 if(!v->pquantizer && block[k << 3])
2548 block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;
2551 for(k = 1; k < 8; k++) {
2552 block[k] = ac_val[k + 8] * scale;
2553 if(!v->pquantizer && block[k])
2554 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2560 s->block_last_index[n] = i;
2565 /** Decode intra block in intra frames - should be faster than decode_intra_block
2566 * @param v VC1Context
2567 * @param block block to decode
2568 * @param coded are AC coeffs present or not
2569 * @param codingset set of VLC to decode data
2571 static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant)
2573 GetBitContext *gb = &v->s.gb;
2574 MpegEncContext *s = &v->s;
2575 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2578 int16_t *ac_val, *ac_val2;
2580 int a_avail = v->a_avail, c_avail = v->c_avail;
2581 int use_pred = s->ac_pred;
2584 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2586 /* Get DC differential */
2588 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2590 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2593 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2598 if (dcdiff == 119 /* ESC index value */)
2600 /* TODO: Optimize */
2601 if (mquant == 1) dcdiff = get_bits(gb, 10);
2602 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2603 else dcdiff = get_bits(gb, 8);
2608 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2609 else if (mquant == 2)
2610 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2617 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);
2620 /* Store the quantized DC coeff, used for prediction */
2622 block[0] = dcdiff * s->y_dc_scale;
2624 block[0] = dcdiff * s->c_dc_scale;
2633 /* check if AC is needed at all */
2634 if(!a_avail && !c_avail) use_pred = 0;
2635 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2638 scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);
2640 if(dc_pred_dir) //left
2643 ac_val -= 16 * s->block_wrap[n];
2645 q1 = s->current_picture.qscale_table[mb_pos];
2646 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2647 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2648 if(dc_pred_dir && n==1) q2 = q1;
2649 if(!dc_pred_dir && n==2) q2 = q1;
2653 int last = 0, skip, value;
2654 const int8_t *zz_table;
2659 zz_table = wmv1_scantable[2];
2661 zz_table = wmv1_scantable[3];
2663 zz_table = wmv1_scantable[1];
2666 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2670 block[zz_table[i++]] = value;
2673 /* apply AC prediction if needed */
2675 /* scale predictors if needed*/
2677 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2678 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2680 if(dc_pred_dir) { //left
2681 for(k = 1; k < 8; k++)
2682 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2684 for(k = 1; k < 8; k++)
2685 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2688 if(dc_pred_dir) { //left
2689 for(k = 1; k < 8; k++)
2690 block[k << 3] += ac_val[k];
2692 for(k = 1; k < 8; k++)
2693 block[k] += ac_val[k + 8];
2697 /* save AC coeffs for further prediction */
2698 for(k = 1; k < 8; k++) {
2699 ac_val2[k] = block[k << 3];
2700 ac_val2[k + 8] = block[k];
2703 /* scale AC coeffs */
2704 for(k = 1; k < 64; k++)
2708 block[k] += (block[k] < 0) ? -mquant : mquant;
2711 if(use_pred) i = 63;
2712 } else { // no AC coeffs
2715 memset(ac_val2, 0, 16 * 2);
2716 if(dc_pred_dir) {//left
2718 memcpy(ac_val2, ac_val, 8 * 2);
2720 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2721 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2722 for(k = 1; k < 8; k++)
2723 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2728 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2730 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2731 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2732 for(k = 1; k < 8; k++)
2733 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2738 /* apply AC prediction if needed */
2740 if(dc_pred_dir) { //left
2741 for(k = 1; k < 8; k++) {
2742 block[k << 3] = ac_val2[k] * scale;
2743 if(!v->pquantizer && block[k << 3])
2744 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2747 for(k = 1; k < 8; k++) {
2748 block[k] = ac_val2[k + 8] * scale;
2749 if(!v->pquantizer && block[k])
2750 block[k] += (block[k] < 0) ? -mquant : mquant;
2756 s->block_last_index[n] = i;
2761 /** Decode intra block in inter frames - more generic version than vc1_decode_i_block
2762 * @param v VC1Context
2763 * @param block block to decode
2764 * @param coded are AC coeffs present or not
2765 * @param mquant block quantizer
2766 * @param codingset set of VLC to decode data
2768 static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)
2770 GetBitContext *gb = &v->s.gb;
2771 MpegEncContext *s = &v->s;
2772 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2775 int16_t *ac_val, *ac_val2;
2777 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2778 int a_avail = v->a_avail, c_avail = v->c_avail;
2779 int use_pred = s->ac_pred;
2783 /* XXX: Guard against dumb values of mquant */
2784 mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );
2786 /* Set DC scale - y and c use the same */
2787 s->y_dc_scale = s->y_dc_scale_table[mquant];
2788 s->c_dc_scale = s->c_dc_scale_table[mquant];
2790 /* Get DC differential */
2792 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2794 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2797 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2802 if (dcdiff == 119 /* ESC index value */)
2804 /* TODO: Optimize */
2805 if (mquant == 1) dcdiff = get_bits(gb, 10);
2806 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2807 else dcdiff = get_bits(gb, 8);
2812 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2813 else if (mquant == 2)
2814 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2821 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);
2824 /* Store the quantized DC coeff, used for prediction */
2827 block[0] = dcdiff * s->y_dc_scale;
2829 block[0] = dcdiff * s->c_dc_scale;
2838 /* check if AC is needed at all and adjust direction if needed */
2839 if(!a_avail) dc_pred_dir = 1;
2840 if(!c_avail) dc_pred_dir = 0;
2841 if(!a_avail && !c_avail) use_pred = 0;
2842 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2845 scale = mquant * 2 + v->halfpq;
2847 if(dc_pred_dir) //left
2850 ac_val -= 16 * s->block_wrap[n];
2852 q1 = s->current_picture.qscale_table[mb_pos];
2853 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2854 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2855 if(dc_pred_dir && n==1) q2 = q1;
2856 if(!dc_pred_dir && n==2) q2 = q1;
2860 int last = 0, skip, value;
2861 const int8_t *zz_table;
2864 zz_table = wmv1_scantable[0];
2867 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2871 block[zz_table[i++]] = value;
2874 /* apply AC prediction if needed */
2876 /* scale predictors if needed*/
2878 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2879 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2881 if(dc_pred_dir) { //left
2882 for(k = 1; k < 8; k++)
2883 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2885 for(k = 1; k < 8; k++)
2886 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2889 if(dc_pred_dir) { //left
2890 for(k = 1; k < 8; k++)
2891 block[k << 3] += ac_val[k];
2893 for(k = 1; k < 8; k++)
2894 block[k] += ac_val[k + 8];
2898 /* save AC coeffs for further prediction */
2899 for(k = 1; k < 8; k++) {
2900 ac_val2[k] = block[k << 3];
2901 ac_val2[k + 8] = block[k];
2904 /* scale AC coeffs */
2905 for(k = 1; k < 64; k++)
2909 block[k] += (block[k] < 0) ? -mquant : mquant;
2912 if(use_pred) i = 63;
2913 } else { // no AC coeffs
2916 memset(ac_val2, 0, 16 * 2);
2917 if(dc_pred_dir) {//left
2919 memcpy(ac_val2, ac_val, 8 * 2);
2921 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2922 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2923 for(k = 1; k < 8; k++)
2924 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2929 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2931 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2932 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2933 for(k = 1; k < 8; k++)
2934 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2939 /* apply AC prediction if needed */
2941 if(dc_pred_dir) { //left
2942 for(k = 1; k < 8; k++) {
2943 block[k << 3] = ac_val2[k] * scale;
2944 if(!v->pquantizer && block[k << 3])
2945 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2948 for(k = 1; k < 8; k++) {
2949 block[k] = ac_val2[k + 8] * scale;
2950 if(!v->pquantizer && block[k])
2951 block[k] += (block[k] < 0) ? -mquant : mquant;
2957 s->block_last_index[n] = i;
2964 static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,
2965 uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left)
2967 MpegEncContext *s = &v->s;
2968 GetBitContext *gb = &s->gb;
2971 int scale, off, idx, last, skip, value;
2972 int ttblk = ttmb & 7;
2976 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)];
2978 if(ttblk == TT_4X4) {
2979 subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);
2981 if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {
2982 subblkpat = decode012(gb);
2983 if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits
2984 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;
2985 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;
2987 scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);
2989 // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT
2990 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {
2991 subblkpat = 2 - (ttblk == TT_8X4_TOP);
2994 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
2995 subblkpat = 2 - (ttblk == TT_4X8_LEFT);
3004 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3008 idx = wmv1_scantable[0][i++];
3009 block[idx] = value * scale;
3011 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3014 s->dsp.vc1_inv_trans_8x8(block);
3015 s->dsp.add_pixels_clamped(block, dst, linesize);
3016 if(apply_filter && cbp_top & 0xC)
3017 vc1_loop_filter(dst, 1, linesize, 8, mquant);
3018 if(apply_filter && cbp_left & 0xA)
3019 vc1_loop_filter(dst, linesize, 1, 8, mquant);
3023 pat = ~subblkpat & 0xF;
3024 for(j = 0; j < 4; j++) {
3025 last = subblkpat & (1 << (3 - j));
3027 off = (j & 1) * 4 + (j & 2) * 16;
3029 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3033 idx = ff_vc1_simple_progressive_4x4_zz[i++];
3034 block[idx + off] = value * scale;
3036 block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;
3038 if(!(subblkpat & (1 << (3 - j))) && !skip_block){
3039 s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
3040 if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))
3041 vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, 1, linesize, 4, mquant);
3042 if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))
3043 vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, linesize, 1, 4, mquant);
3048 pat = ~((subblkpat & 2)*6 + (subblkpat & 1)*3) & 0xF;
3049 for(j = 0; j < 2; j++) {
3050 last = subblkpat & (1 << (1 - j));
3054 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3058 idx = v->zz_8x4[i++]+off;
3059 block[idx] = value * scale;
3061 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3063 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3064 s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
3065 if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))
3066 vc1_loop_filter(dst + j*4*linesize, 1, linesize, 8, mquant);
3067 if(apply_filter && cbp_left & (2 << j))
3068 vc1_loop_filter(dst + j*4*linesize, linesize, 1, 4, mquant);
3073 pat = ~(subblkpat*5) & 0xF;
3074 for(j = 0; j < 2; j++) {
3075 last = subblkpat & (1 << (1 - j));
3079 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3083 idx = v->zz_4x8[i++]+off;
3084 block[idx] = value * scale;
3086 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3088 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3089 s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
3090 if(apply_filter && cbp_top & (2 << j))
3091 vc1_loop_filter(dst + j*4, 1, linesize, 4, mquant);
3092 if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))
3093 vc1_loop_filter(dst + j*4, linesize, 1, 8, mquant);
3102 /** Decode one P-frame MB (in Simple/Main profile)
3104 static int vc1_decode_p_mb(VC1Context *v)
3106 MpegEncContext *s = &v->s;
3107 GetBitContext *gb = &s->gb;
3109 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3110 int cbp; /* cbp decoding stuff */
3111 int mqdiff, mquant; /* MB quantization */
3112 int ttmb = v->ttfrm; /* MB Transform type */
3114 static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3115 offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3116 int mb_has_coeffs = 1; /* last_flag */
3117 int dmv_x, dmv_y; /* Differential MV components */
3118 int index, index1; /* LUT indexes */
3119 int val, sign; /* temp values */
3120 int first_block = 1;
3122 int skipped, fourmv;
3123 int block_cbp = 0, pat;
3124 int apply_loop_filter;
3126 mquant = v->pq; /* Loosy initialization */
3128 if (v->mv_type_is_raw)
3129 fourmv = get_bits1(gb);
3131 fourmv = v->mv_type_mb_plane[mb_pos];
3133 skipped = get_bits1(gb);
3135 skipped = v->s.mbskip_table[mb_pos];
3137 s->dsp.clear_blocks(s->block[0]);
3139 apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);
3140 if (!fourmv) /* 1MV mode */
3144 GET_MVDATA(dmv_x, dmv_y);
3147 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3148 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3150 s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
3151 vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
3153 /* FIXME Set DC val for inter block ? */
3154 if (s->mb_intra && !mb_has_coeffs)
3157 s->ac_pred = get_bits1(gb);
3160 else if (mb_has_coeffs)
3162 if (s->mb_intra) s->ac_pred = get_bits1(gb);
3163 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3171 s->current_picture.qscale_table[mb_pos] = mquant;
3173 if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3174 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,
3175 VC1_TTMB_VLC_BITS, 2);
3176 if(!s->mb_intra) vc1_mc_1mv(v, 0);
3180 s->dc_val[0][s->block_index[i]] = 0;
3182 val = ((cbp >> (5 - i)) & 1);
3183 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3184 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3186 /* check if prediction blocks A and C are available */
3187 v->a_avail = v->c_avail = 0;
3188 if(i == 2 || i == 3 || !s->first_slice_line)
3189 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3190 if(i == 1 || i == 3 || s->mb_x)
3191 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3193 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3194 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3195 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3196 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3197 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3198 if(v->pq >= 9 && v->overlap) {
3200 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3202 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3204 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3205 int left_cbp, top_cbp;
3207 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3208 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3210 left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3211 top_cbp = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3214 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3216 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3218 block_cbp |= 0xF << (i << 2);
3220 int left_cbp = 0, top_cbp = 0, filter = 0;
3221 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3224 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3225 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3227 left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3228 top_cbp = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3231 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);
3232 block_cbp |= pat << (i << 2);
3233 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3241 for(i = 0; i < 6; i++) {
3242 v->mb_type[0][s->block_index[i]] = 0;
3243 s->dc_val[0][s->block_index[i]] = 0;
3245 s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;
3246 s->current_picture.qscale_table[mb_pos] = 0;
3247 vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
3254 if (!skipped /* unskipped MB */)
3256 int intra_count = 0, coded_inter = 0;
3257 int is_intra[6], is_coded[6];
3259 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3262 val = ((cbp >> (5 - i)) & 1);
3263 s->dc_val[0][s->block_index[i]] = 0;
3270 GET_MVDATA(dmv_x, dmv_y);
3272 vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
3273 if(!s->mb_intra) vc1_mc_4mv_luma(v, i);
3274 intra_count += s->mb_intra;
3275 is_intra[i] = s->mb_intra;
3276 is_coded[i] = mb_has_coeffs;
3279 is_intra[i] = (intra_count >= 3);
3282 if(i == 4) vc1_mc_4mv_chroma(v);
3283 v->mb_type[0][s->block_index[i]] = is_intra[i];
3284 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];
3286 // if there are no coded blocks then don't do anything more
3287 if(!intra_count && !coded_inter) return 0;
3290 s->current_picture.qscale_table[mb_pos] = mquant;
3291 /* test if block is intra and has pred */
3296 if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])
3297 || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {
3302 if(intrapred)s->ac_pred = get_bits1(gb);
3303 else s->ac_pred = 0;
3305 if (!v->ttmbf && coded_inter)
3306 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3310 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3311 s->mb_intra = is_intra[i];
3313 /* check if prediction blocks A and C are available */
3314 v->a_avail = v->c_avail = 0;
3315 if(i == 2 || i == 3 || !s->first_slice_line)
3316 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3317 if(i == 1 || i == 3 || s->mb_x)
3318 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3320 vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);
3321 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3322 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3323 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3324 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
3325 if(v->pq >= 9 && v->overlap) {
3327 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3329 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3331 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)){
3332 int left_cbp, top_cbp;
3334 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3335 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3337 left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3338 top_cbp = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3341 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3343 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3345 block_cbp |= 0xF << (i << 2);
3346 } else if(is_coded[i]) {
3347 int left_cbp = 0, top_cbp = 0, filter = 0;
3348 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)){
3351 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3352 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3354 left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3355 top_cbp = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3358 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);
3359 block_cbp |= pat << (i << 2);
3360 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3369 s->current_picture.qscale_table[mb_pos] = 0;
3370 for (i=0; i<6; i++) {
3371 v->mb_type[0][s->block_index[i]] = 0;
3372 s->dc_val[0][s->block_index[i]] = 0;
3376 vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
3377 vc1_mc_4mv_luma(v, i);
3379 vc1_mc_4mv_chroma(v);
3380 s->current_picture.qscale_table[mb_pos] = 0;
3384 v->cbp[s->mb_x] = block_cbp;
3386 /* Should never happen */
3390 /** Decode one B-frame MB (in Main profile)
3392 static void vc1_decode_b_mb(VC1Context *v)
3394 MpegEncContext *s = &v->s;
3395 GetBitContext *gb = &s->gb;
3397 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3398 int cbp = 0; /* cbp decoding stuff */
3399 int mqdiff, mquant; /* MB quantization */
3400 int ttmb = v->ttfrm; /* MB Transform type */
3402 static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3403 offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3404 int mb_has_coeffs = 0; /* last_flag */
3405 int index, index1; /* LUT indexes */
3406 int val, sign; /* temp values */
3407 int first_block = 1;
3409 int skipped, direct;
3410 int dmv_x[2], dmv_y[2];
3411 int bmvtype = BMV_TYPE_BACKWARD;
3413 mquant = v->pq; /* Loosy initialization */
3417 direct = get_bits1(gb);
3419 direct = v->direct_mb_plane[mb_pos];
3421 skipped = get_bits1(gb);
3423 skipped = v->s.mbskip_table[mb_pos];
3425 s->dsp.clear_blocks(s->block[0]);
3426 dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
3427 for(i = 0; i < 6; i++) {
3428 v->mb_type[0][s->block_index[i]] = 0;
3429 s->dc_val[0][s->block_index[i]] = 0;
3431 s->current_picture.qscale_table[mb_pos] = 0;
3435 GET_MVDATA(dmv_x[0], dmv_y[0]);
3436 dmv_x[1] = dmv_x[0];
3437 dmv_y[1] = dmv_y[0];
3439 if(skipped || !s->mb_intra) {
3440 bmvtype = decode012(gb);
3443 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
3446 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
3449 bmvtype = BMV_TYPE_INTERPOLATED;
3450 dmv_x[0] = dmv_y[0] = 0;
3454 for(i = 0; i < 6; i++)
3455 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3458 if(direct) bmvtype = BMV_TYPE_INTERPOLATED;
3459 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3460 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3464 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3468 s->current_picture.qscale_table[mb_pos] = mquant;
3470 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3471 dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;
3472 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3473 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3475 if(!mb_has_coeffs && !s->mb_intra) {
3476 /* no coded blocks - effectively skipped */
3477 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3478 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3481 if(s->mb_intra && !mb_has_coeffs) {
3483 s->current_picture.qscale_table[mb_pos] = mquant;
3484 s->ac_pred = get_bits1(gb);
3486 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3488 if(bmvtype == BMV_TYPE_INTERPOLATED) {
3489 GET_MVDATA(dmv_x[0], dmv_y[0]);
3490 if(!mb_has_coeffs) {
3491 /* interpolated skipped block */
3492 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3493 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3497 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3499 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3502 s->ac_pred = get_bits1(gb);
3503 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3505 s->current_picture.qscale_table[mb_pos] = mquant;
3506 if(!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3507 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3513 s->dc_val[0][s->block_index[i]] = 0;
3515 val = ((cbp >> (5 - i)) & 1);
3516 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3517 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3519 /* check if prediction blocks A and C are available */
3520 v->a_avail = v->c_avail = 0;
3521 if(i == 2 || i == 3 || !s->first_slice_line)
3522 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3523 if(i == 1 || i == 3 || s->mb_x)
3524 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3526 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3527 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3528 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3529 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3530 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3532 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);
3533 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3539 /** Decode blocks of I-frame
3541 static void vc1_decode_i_blocks(VC1Context *v)
3544 MpegEncContext *s = &v->s;
3549 /* select codingmode used for VLC tables selection */
3550 switch(v->y_ac_table_index){
3552 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3555 v->codingset = CS_HIGH_MOT_INTRA;
3558 v->codingset = CS_MID_RATE_INTRA;
3562 switch(v->c_ac_table_index){
3564 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3567 v->codingset2 = CS_HIGH_MOT_INTER;
3570 v->codingset2 = CS_MID_RATE_INTER;
3574 /* Set DC scale - y and c use the same */
3575 s->y_dc_scale = s->y_dc_scale_table[v->pq];
3576 s->c_dc_scale = s->c_dc_scale_table[v->pq];
3579 s->mb_x = s->mb_y = 0;
3581 s->first_slice_line = 1;
3582 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3583 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3584 ff_init_block_index(s);
3585 ff_update_block_index(s);
3586 s->dsp.clear_blocks(s->block[0]);
3587 mb_pos = s->mb_x + s->mb_y * s->mb_width;
3588 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3589 s->current_picture.qscale_table[mb_pos] = v->pq;
3590 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3591 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3593 // do actual MB decoding and displaying
3594 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3595 v->s.ac_pred = get_bits1(&v->s.gb);
3597 for(k = 0; k < 6; k++) {
3598 val = ((cbp >> (5 - k)) & 1);
3601 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3605 cbp |= val << (5 - k);
3607 vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);
3609 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3610 if(v->pq >= 9 && v->overlap) {
3611 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3615 vc1_put_block(v, s->block);
3616 if(v->pq >= 9 && v->overlap) {
3618 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3619 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3620 if(!(s->flags & CODEC_FLAG_GRAY)) {
3621 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3622 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3625 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3626 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3627 if(!s->first_slice_line) {
3628 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3629 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3630 if(!(s->flags & CODEC_FLAG_GRAY)) {
3631 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3632 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3635 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3636 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3638 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3640 if(get_bits_count(&s->gb) > v->bits) {
3641 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3642 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3646 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3647 s->first_slice_line = 0;
3649 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3652 /** Decode blocks of I-frame for advanced profile
3654 static void vc1_decode_i_blocks_adv(VC1Context *v)
3657 MpegEncContext *s = &v->s;
3664 GetBitContext *gb = &s->gb;
3666 /* select codingmode used for VLC tables selection */
3667 switch(v->y_ac_table_index){
3669 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3672 v->codingset = CS_HIGH_MOT_INTRA;
3675 v->codingset = CS_MID_RATE_INTRA;
3679 switch(v->c_ac_table_index){
3681 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3684 v->codingset2 = CS_HIGH_MOT_INTER;
3687 v->codingset2 = CS_MID_RATE_INTER;
3692 s->mb_x = s->mb_y = 0;
3694 s->first_slice_line = 1;
3695 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3696 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3697 ff_init_block_index(s);
3698 ff_update_block_index(s);
3699 s->dsp.clear_blocks(s->block[0]);
3700 mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3701 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3702 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3703 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3705 // do actual MB decoding and displaying
3706 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3707 if(v->acpred_is_raw)
3708 v->s.ac_pred = get_bits1(&v->s.gb);
3710 v->s.ac_pred = v->acpred_plane[mb_pos];
3712 if(v->condover == CONDOVER_SELECT) {
3713 if(v->overflg_is_raw)
3714 overlap = get_bits1(&v->s.gb);
3716 overlap = v->over_flags_plane[mb_pos];
3718 overlap = (v->condover == CONDOVER_ALL);
3722 s->current_picture.qscale_table[mb_pos] = mquant;
3723 /* Set DC scale - y and c use the same */
3724 s->y_dc_scale = s->y_dc_scale_table[mquant];
3725 s->c_dc_scale = s->c_dc_scale_table[mquant];
3727 for(k = 0; k < 6; k++) {
3728 val = ((cbp >> (5 - k)) & 1);
3731 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3735 cbp |= val << (5 - k);
3737 v->a_avail = !s->first_slice_line || (k==2 || k==3);
3738 v->c_avail = !!s->mb_x || (k==1 || k==3);
3740 vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);
3742 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3743 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3746 vc1_put_block(v, s->block);
3749 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3750 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3751 if(!(s->flags & CODEC_FLAG_GRAY)) {
3752 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3753 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3756 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3757 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3758 if(!s->first_slice_line) {
3759 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3760 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3761 if(!(s->flags & CODEC_FLAG_GRAY)) {
3762 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3763 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3766 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3767 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3769 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3771 if(get_bits_count(&s->gb) > v->bits) {
3772 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3773 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3777 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3778 s->first_slice_line = 0;
3780 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3783 static void vc1_decode_p_blocks(VC1Context *v)
3785 MpegEncContext *s = &v->s;
3787 /* select codingmode used for VLC tables selection */
3788 switch(v->c_ac_table_index){
3790 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3793 v->codingset = CS_HIGH_MOT_INTRA;
3796 v->codingset = CS_MID_RATE_INTRA;
3800 switch(v->c_ac_table_index){
3802 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3805 v->codingset2 = CS_HIGH_MOT_INTER;
3808 v->codingset2 = CS_MID_RATE_INTER;
3812 s->first_slice_line = 1;
3813 memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);
3814 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3815 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3816 ff_init_block_index(s);
3817 ff_update_block_index(s);
3818 s->dsp.clear_blocks(s->block[0]);
3821 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3822 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3823 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);
3827 memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride);
3828 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3829 s->first_slice_line = 0;
3831 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3834 static void vc1_decode_b_blocks(VC1Context *v)
3836 MpegEncContext *s = &v->s;
3838 /* select codingmode used for VLC tables selection */
3839 switch(v->c_ac_table_index){
3841 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3844 v->codingset = CS_HIGH_MOT_INTRA;
3847 v->codingset = CS_MID_RATE_INTRA;
3851 switch(v->c_ac_table_index){
3853 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3856 v->codingset2 = CS_HIGH_MOT_INTER;
3859 v->codingset2 = CS_MID_RATE_INTER;
3863 s->first_slice_line = 1;
3864 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3865 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3866 ff_init_block_index(s);
3867 ff_update_block_index(s);
3868 s->dsp.clear_blocks(s->block[0]);
3871 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3872 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3873 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);
3876 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[s->mb_x + s->mb_y *s->mb_stride]);
3878 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3879 s->first_slice_line = 0;
3881 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3884 static void vc1_decode_skip_blocks(VC1Context *v)
3886 MpegEncContext *s = &v->s;
3888 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3889 s->first_slice_line = 1;
3890 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3892 ff_init_block_index(s);
3893 ff_update_block_index(s);
3894 memcpy(s->dest[0], s->last_picture.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16);
3895 memcpy(s->dest[1], s->last_picture.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3896 memcpy(s->dest[2], s->last_picture.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3897 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3898 s->first_slice_line = 0;
3900 s->pict_type = FF_P_TYPE;
3903 static void vc1_decode_blocks(VC1Context *v)
3906 v->s.esc3_level_length = 0;
3908 ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );
3911 switch(v->s.pict_type) {
3913 if(v->profile == PROFILE_ADVANCED)
3914 vc1_decode_i_blocks_adv(v);
3916 vc1_decode_i_blocks(v);
3919 if(v->p_frame_skipped)
3920 vc1_decode_skip_blocks(v);
3922 vc1_decode_p_blocks(v);
3926 if(v->profile == PROFILE_ADVANCED)
3927 vc1_decode_i_blocks_adv(v);
3929 vc1_decode_i_blocks(v);
3931 vc1_decode_b_blocks(v);
3937 /** Find VC-1 marker in buffer
3938 * @return position where next marker starts or end of buffer if no marker found
3940 static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)
3942 uint32_t mrk = 0xFFFFFFFF;
3944 if(end-src < 4) return end;
3946 mrk = (mrk << 8) | *src++;
3953 static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)
3958 for(dsize = 0; dsize < size; dsize++) *dst++ = *src++;
3961 for(i = 0; i < size; i++, src++) {
3962 if(src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {
3963 dst[dsize++] = src[1];
3967 dst[dsize++] = *src;
3972 /** Initialize a VC1/WMV3 decoder
3973 * @todo TODO: Handle VC-1 IDUs (Transport level?)
3974 * @todo TODO: Decypher remaining bits in extra_data
3976 static av_cold int vc1_decode_init(AVCodecContext *avctx)
3978 VC1Context *v = avctx->priv_data;
3979 MpegEncContext *s = &v->s;
3982 if (!avctx->extradata_size || !avctx->extradata) return -1;
3983 if (!(avctx->flags & CODEC_FLAG_GRAY))
3984 avctx->pix_fmt = PIX_FMT_YUV420P;
3986 avctx->pix_fmt = PIX_FMT_GRAY8;
3988 avctx->flags |= CODEC_FLAG_EMU_EDGE;
3989 v->s.flags |= CODEC_FLAG_EMU_EDGE;
3991 if(avctx->idct_algo==FF_IDCT_AUTO){
3992 avctx->idct_algo=FF_IDCT_WMV2;
3995 if(ff_h263_decode_init(avctx) < 0)
3997 if (vc1_init_common(v) < 0) return -1;
3999 avctx->coded_width = avctx->width;
4000 avctx->coded_height = avctx->height;
4001 if (avctx->codec_id == CODEC_ID_WMV3)
4005 // looks like WMV3 has a sequence header stored in the extradata
4006 // advanced sequence header may be before the first frame
4007 // the last byte of the extradata is a version number, 1 for the
4008 // samples we can decode
4010 init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);
4012 if (decode_sequence_header(avctx, &gb) < 0)
4015 count = avctx->extradata_size*8 - get_bits_count(&gb);
4018 av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",
4019 count, get_bits(&gb, count));
4023 av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);
4025 } else { // VC1/WVC1
4026 const uint8_t *start = avctx->extradata;
4027 uint8_t *end = avctx->extradata + avctx->extradata_size;
4028 const uint8_t *next;
4029 int size, buf2_size;
4030 uint8_t *buf2 = NULL;
4031 int seq_initialized = 0, ep_initialized = 0;
4033 if(avctx->extradata_size < 16) {
4034 av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);
4038 buf2 = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
4039 if(start[0]) start++; // in WVC1 extradata first byte is its size
4041 for(; next < end; start = next){
4042 next = find_next_marker(start + 4, end);
4043 size = next - start - 4;
4044 if(size <= 0) continue;
4045 buf2_size = vc1_unescape_buffer(start + 4, size, buf2);
4046 init_get_bits(&gb, buf2, buf2_size * 8);
4047 switch(AV_RB32(start)){
4048 case VC1_CODE_SEQHDR:
4049 if(decode_sequence_header(avctx, &gb) < 0){
4053 seq_initialized = 1;
4055 case VC1_CODE_ENTRYPOINT:
4056 if(decode_entry_point(avctx, &gb) < 0){
4065 if(!seq_initialized || !ep_initialized){
4066 av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");
4070 avctx->has_b_frames= !!(avctx->max_b_frames);
4071 s->low_delay = !avctx->has_b_frames;
4073 s->mb_width = (avctx->coded_width+15)>>4;
4074 s->mb_height = (avctx->coded_height+15)>>4;
4076 /* Allocate mb bitplanes */
4077 v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4078 v->direct_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4079 v->acpred_plane = av_malloc(s->mb_stride * s->mb_height);
4080 v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height);
4082 v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);
4083 v->cbp = v->cbp_base + s->mb_stride;
4085 /* allocate block type info in that way so it could be used with s->block_index[] */
4086 v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);
4087 v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;
4088 v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;
4089 v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);
4091 /* Init coded blocks info */
4092 if (v->profile == PROFILE_ADVANCED)
4094 // if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
4096 // if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
4100 ff_intrax8_common_init(&v->x8,s);
4105 /** Decode a VC1/WMV3 frame
4106 * @todo TODO: Handle VC-1 IDUs (Transport level?)
4108 static int vc1_decode_frame(AVCodecContext *avctx,
4109 void *data, int *data_size,
4110 const uint8_t *buf, int buf_size)
4112 VC1Context *v = avctx->priv_data;
4113 MpegEncContext *s = &v->s;
4114 AVFrame *pict = data;
4115 uint8_t *buf2 = NULL;
4117 /* no supplementary picture */
4118 if (buf_size == 0) {
4119 /* special case for last picture */
4120 if (s->low_delay==0 && s->next_picture_ptr) {
4121 *pict= *(AVFrame*)s->next_picture_ptr;
4122 s->next_picture_ptr= NULL;
4124 *data_size = sizeof(AVFrame);
4130 /* We need to set current_picture_ptr before reading the header,
4131 * otherwise we cannot store anything in there. */
4132 if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){
4133 int i= ff_find_unused_picture(s, 0);
4134 s->current_picture_ptr= &s->picture[i];
4137 //for advanced profile we may need to parse and unescape data
4138 if (avctx->codec_id == CODEC_ID_VC1) {
4140 buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
4142 if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */
4143 const uint8_t *start, *end, *next;
4147 for(start = buf, end = buf + buf_size; next < end; start = next){
4148 next = find_next_marker(start + 4, end);
4149 size = next - start - 4;
4150 if(size <= 0) continue;
4151 switch(AV_RB32(start)){
4152 case VC1_CODE_FRAME:
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 ff_er_frame_start(s);
4243 v->bits = buf_size * 8;
4244 vc1_decode_blocks(v);
4245 //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);
4246 // if(get_bits_count(&s->gb) > buf_size * 8)
4252 assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);
4253 assert(s->current_picture.pict_type == s->pict_type);
4254 if (s->pict_type == FF_B_TYPE || s->low_delay) {
4255 *pict= *(AVFrame*)s->current_picture_ptr;
4256 } else if (s->last_picture_ptr != NULL) {
4257 *pict= *(AVFrame*)s->last_picture_ptr;
4260 if(s->last_picture_ptr || s->low_delay){
4261 *data_size = sizeof(AVFrame);
4262 ff_print_debug_info(s, pict);
4265 /* Return the Picture timestamp as the frame number */
4266 /* we subtract 1 because it is added on utils.c */
4267 avctx->frame_number = s->picture_number - 1;
4274 /** Close a VC1/WMV3 decoder
4275 * @warning Initial try at using MpegEncContext stuff
4277 static av_cold int vc1_decode_end(AVCodecContext *avctx)
4279 VC1Context *v = avctx->priv_data;
4281 av_freep(&v->hrd_rate);
4282 av_freep(&v->hrd_buffer);
4283 MPV_common_end(&v->s);
4284 av_freep(&v->mv_type_mb_plane);
4285 av_freep(&v->direct_mb_plane);
4286 av_freep(&v->acpred_plane);
4287 av_freep(&v->over_flags_plane);
4288 av_freep(&v->mb_type_base);
4289 av_freep(&v->cbp_base);
4290 ff_intrax8_common_end(&v->x8);
4295 AVCodec vc1_decoder = {
4306 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),
4309 AVCodec wmv3_decoder = {
4320 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),