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"
42 #define MB_INTRA_VLC_BITS 9
45 static const uint16_t table_mb_intra[64][2];
49 * Init VC-1 specific tables and VC1Context members
50 * @param v The VC1Context to initialize
53 static int vc1_init_common(VC1Context *v)
58 v->hrd_rate = v->hrd_buffer = NULL;
64 init_vlc(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,
65 ff_vc1_bfraction_bits, 1, 1,
66 ff_vc1_bfraction_codes, 1, 1, 1);
67 init_vlc(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,
68 ff_vc1_norm2_bits, 1, 1,
69 ff_vc1_norm2_codes, 1, 1, 1);
70 init_vlc(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,
71 ff_vc1_norm6_bits, 1, 1,
72 ff_vc1_norm6_codes, 2, 2, 1);
73 init_vlc(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,
74 ff_vc1_imode_bits, 1, 1,
75 ff_vc1_imode_codes, 1, 1, 1);
78 init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,
79 ff_vc1_ttmb_bits[i], 1, 1,
80 ff_vc1_ttmb_codes[i], 2, 2, 1);
81 init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,
82 ff_vc1_ttblk_bits[i], 1, 1,
83 ff_vc1_ttblk_codes[i], 1, 1, 1);
84 init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,
85 ff_vc1_subblkpat_bits[i], 1, 1,
86 ff_vc1_subblkpat_codes[i], 1, 1, 1);
90 init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,
91 ff_vc1_4mv_block_pattern_bits[i], 1, 1,
92 ff_vc1_4mv_block_pattern_codes[i], 1, 1, 1);
93 init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,
94 ff_vc1_cbpcy_p_bits[i], 1, 1,
95 ff_vc1_cbpcy_p_codes[i], 2, 2, 1);
96 init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,
97 ff_vc1_mv_diff_bits[i], 1, 1,
98 ff_vc1_mv_diff_codes[i], 2, 2, 1);
101 init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],
102 &vc1_ac_tables[i][0][1], 8, 4,
103 &vc1_ac_tables[i][0][0], 8, 4, 1);
104 init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,
105 &ff_msmp4_mb_i_table[0][1], 4, 2,
106 &ff_msmp4_mb_i_table[0][0], 4, 2, 1);
111 v->mvrange = 0; /* 7.1.1.18, p80 */
116 /***********************************************************************/
118 * @defgroup bitplane VC9 Bitplane decoding
123 /** @addtogroup bitplane
136 /** @} */ //imode defines
138 /** Decode rows by checking if they are skipped
139 * @param plane Buffer to store decoded bits
140 * @param[in] width Width of this buffer
141 * @param[in] height Height of this buffer
142 * @param[in] stride of this buffer
144 static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
147 for (y=0; y<height; y++){
148 if (!get_bits1(gb)) //rowskip
149 memset(plane, 0, width);
151 for (x=0; x<width; x++)
152 plane[x] = get_bits1(gb);
157 /** Decode columns by checking if they are skipped
158 * @param plane Buffer to store decoded bits
159 * @param[in] width Width of this buffer
160 * @param[in] height Height of this buffer
161 * @param[in] stride of this buffer
162 * @todo FIXME: Optimize
164 static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
167 for (x=0; x<width; x++){
168 if (!get_bits1(gb)) //colskip
169 for (y=0; y<height; y++)
172 for (y=0; y<height; y++)
173 plane[y*stride] = get_bits1(gb);
178 /** Decode a bitplane's bits
179 * @param bp Bitplane where to store the decode bits
180 * @param v VC-1 context for bit reading and logging
182 * @todo FIXME: Optimize
184 static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)
186 GetBitContext *gb = &v->s.gb;
188 int imode, x, y, code, offset;
189 uint8_t invert, *planep = data;
190 int width, height, stride;
192 width = v->s.mb_width;
193 height = v->s.mb_height;
194 stride = v->s.mb_stride;
195 invert = get_bits1(gb);
196 imode = get_vlc2(gb, ff_vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);
202 //Data is actually read in the MB layer (same for all tests == "raw")
203 *raw_flag = 1; //invert ignored
207 if ((height * width) & 1)
209 *planep++ = get_bits1(gb);
213 // decode bitplane as one long line
214 for (y = offset; y < height * width; y += 2) {
215 code = get_vlc2(gb, ff_vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
216 *planep++ = code & 1;
218 if(offset == width) {
220 planep += stride - width;
222 *planep++ = code >> 1;
224 if(offset == width) {
226 planep += stride - width;
232 if(!(height % 3) && (width % 3)) { // use 2x3 decoding
233 for(y = 0; y < height; y+= 3) {
234 for(x = width & 1; x < width; x += 2) {
235 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
237 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
240 planep[x + 0] = (code >> 0) & 1;
241 planep[x + 1] = (code >> 1) & 1;
242 planep[x + 0 + stride] = (code >> 2) & 1;
243 planep[x + 1 + stride] = (code >> 3) & 1;
244 planep[x + 0 + stride * 2] = (code >> 4) & 1;
245 planep[x + 1 + stride * 2] = (code >> 5) & 1;
247 planep += stride * 3;
249 if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);
251 planep += (height & 1) * stride;
252 for(y = height & 1; y < height; y += 2) {
253 for(x = width % 3; x < width; x += 3) {
254 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
256 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
259 planep[x + 0] = (code >> 0) & 1;
260 planep[x + 1] = (code >> 1) & 1;
261 planep[x + 2] = (code >> 2) & 1;
262 planep[x + 0 + stride] = (code >> 3) & 1;
263 planep[x + 1 + stride] = (code >> 4) & 1;
264 planep[x + 2 + stride] = (code >> 5) & 1;
266 planep += stride * 2;
269 if(x) decode_colskip(data , x, height , stride, &v->s.gb);
270 if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);
274 decode_rowskip(data, width, height, stride, &v->s.gb);
277 decode_colskip(data, width, height, stride, &v->s.gb);
282 /* Applying diff operator */
283 if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
287 for (x=1; x<width; x++)
288 planep[x] ^= planep[x-1];
289 for (y=1; y<height; y++)
292 planep[0] ^= planep[-stride];
293 for (x=1; x<width; x++)
295 if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;
296 else planep[x] ^= planep[x-1];
303 for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride
305 return (imode<<1) + invert;
308 /** @} */ //Bitplane group
310 #define FILTSIGN(a) ((a) >= 0 ? 1 : -1)
312 * VC-1 in-loop deblocking filter for one line
313 * @param src source block type
314 * @param pq block quantizer
315 * @return whether other 3 pairs should be filtered or not
318 static int av_always_inline vc1_filter_line(uint8_t* src, int stride, int pq){
319 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
321 int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
322 int a0_sign = a0 >> 31; /* Store sign */
323 a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */
325 int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3);
326 int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3);
327 if(a1 < a0 || a2 < a0){
328 int clip = src[-1*stride] - src[ 0*stride];
329 int clip_sign = clip >> 31;
330 clip = ((clip ^ clip_sign) - clip_sign)>>1;
332 int a3 = FFMIN(a1, a2);
333 int d = 5 * (a3 - a0);
334 int d_sign = (d >> 31);
335 d = ((d ^ d_sign) - d_sign) >> 3;
338 if( d_sign ^ clip_sign )
342 d = (d ^ d_sign) - d_sign; /* Restore sign */
343 src[-1*stride] = cm[src[-1*stride] - d];
344 src[ 0*stride] = cm[src[ 0*stride] + d];
354 * VC-1 in-loop deblocking filter
355 * @param src source block type
356 * @param len edge length to filter (4 or 8 pixels)
357 * @param pq block quantizer
360 static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
365 for(i = 0; i < len; i += 4){
366 filt3 = vc1_filter_line(src + 2*step, stride, pq);
368 vc1_filter_line(src + 0*step, stride, pq);
369 vc1_filter_line(src + 1*step, stride, pq);
370 vc1_filter_line(src + 3*step, stride, pq);
376 static void vc1_loop_filter_iblk(MpegEncContext *s, int pq)
379 if(!s->first_slice_line)
380 vc1_loop_filter(s->dest[0], 1, s->linesize, 16, pq);
381 vc1_loop_filter(s->dest[0] + 8*s->linesize, 1, s->linesize, 16, pq);
382 for(i = !s->mb_x*8; i < 16; i += 8)
383 vc1_loop_filter(s->dest[0] + i, s->linesize, 1, 16, pq);
384 for(j = 0; j < 2; j++){
385 if(!s->first_slice_line)
386 vc1_loop_filter(s->dest[j+1], 1, s->uvlinesize, 8, pq);
388 vc1_loop_filter(s->dest[j+1], s->uvlinesize, 1, 8, pq);
392 /***********************************************************************/
393 /** VOP Dquant decoding
394 * @param v VC-1 Context
396 static int vop_dquant_decoding(VC1Context *v)
398 GetBitContext *gb = &v->s.gb;
404 pqdiff = get_bits(gb, 3);
405 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
406 else v->altpq = v->pq + pqdiff + 1;
410 v->dquantfrm = get_bits1(gb);
413 v->dqprofile = get_bits(gb, 2);
414 switch (v->dqprofile)
416 case DQPROFILE_SINGLE_EDGE:
417 case DQPROFILE_DOUBLE_EDGES:
418 v->dqsbedge = get_bits(gb, 2);
420 case DQPROFILE_ALL_MBS:
421 v->dqbilevel = get_bits1(gb);
424 default: break; //Forbidden ?
426 if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)
428 pqdiff = get_bits(gb, 3);
429 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
430 else v->altpq = v->pq + pqdiff + 1;
437 /** Put block onto picture
439 static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
443 DSPContext *dsp = &v->s.dsp;
447 for(k = 0; k < 6; k++)
448 for(j = 0; j < 8; j++)
449 for(i = 0; i < 8; i++)
450 block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;
453 ys = v->s.current_picture.linesize[0];
454 us = v->s.current_picture.linesize[1];
455 vs = v->s.current_picture.linesize[2];
458 dsp->put_pixels_clamped(block[0], Y, ys);
459 dsp->put_pixels_clamped(block[1], Y + 8, ys);
461 dsp->put_pixels_clamped(block[2], Y, ys);
462 dsp->put_pixels_clamped(block[3], Y + 8, ys);
464 if(!(v->s.flags & CODEC_FLAG_GRAY)) {
465 dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
466 dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
470 /** Do motion compensation over 1 macroblock
471 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
473 static void vc1_mc_1mv(VC1Context *v, int dir)
475 MpegEncContext *s = &v->s;
476 DSPContext *dsp = &v->s.dsp;
477 uint8_t *srcY, *srcU, *srcV;
478 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
480 if(!v->s.last_picture.data[0])return;
482 mx = s->mv[dir][0][0];
483 my = s->mv[dir][0][1];
485 // store motion vectors for further use in B frames
486 if(s->pict_type == FF_P_TYPE) {
487 s->current_picture.motion_val[1][s->block_index[0]][0] = mx;
488 s->current_picture.motion_val[1][s->block_index[0]][1] = my;
490 uvmx = (mx + ((mx & 3) == 3)) >> 1;
491 uvmy = (my + ((my & 3) == 3)) >> 1;
493 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
494 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
497 srcY = s->last_picture.data[0];
498 srcU = s->last_picture.data[1];
499 srcV = s->last_picture.data[2];
501 srcY = s->next_picture.data[0];
502 srcU = s->next_picture.data[1];
503 srcV = s->next_picture.data[2];
506 src_x = s->mb_x * 16 + (mx >> 2);
507 src_y = s->mb_y * 16 + (my >> 2);
508 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
509 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
511 if(v->profile != PROFILE_ADVANCED){
512 src_x = av_clip( src_x, -16, s->mb_width * 16);
513 src_y = av_clip( src_y, -16, s->mb_height * 16);
514 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
515 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
517 src_x = av_clip( src_x, -17, s->avctx->coded_width);
518 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
519 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
520 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
523 srcY += src_y * s->linesize + src_x;
524 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
525 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
527 /* for grayscale we should not try to read from unknown area */
528 if(s->flags & CODEC_FLAG_GRAY) {
529 srcU = s->edge_emu_buffer + 18 * s->linesize;
530 srcV = s->edge_emu_buffer + 18 * s->linesize;
533 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
534 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3
535 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){
536 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
538 srcY -= s->mspel * (1 + s->linesize);
539 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
540 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
541 srcY = s->edge_emu_buffer;
542 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
543 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
544 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
545 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
548 /* if we deal with range reduction we need to scale source blocks */
554 for(j = 0; j < 17 + s->mspel*2; j++) {
555 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
558 src = srcU; src2 = srcV;
559 for(j = 0; j < 9; j++) {
560 for(i = 0; i < 9; i++) {
561 src[i] = ((src[i] - 128) >> 1) + 128;
562 src2[i] = ((src2[i] - 128) >> 1) + 128;
564 src += s->uvlinesize;
565 src2 += s->uvlinesize;
568 /* if we deal with intensity compensation we need to scale source blocks */
569 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
574 for(j = 0; j < 17 + s->mspel*2; j++) {
575 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];
578 src = srcU; src2 = srcV;
579 for(j = 0; j < 9; j++) {
580 for(i = 0; i < 9; i++) {
581 src[i] = v->lutuv[src[i]];
582 src2[i] = v->lutuv[src2[i]];
584 src += s->uvlinesize;
585 src2 += s->uvlinesize;
588 srcY += s->mspel * (1 + s->linesize);
592 dxy = ((my & 3) << 2) | (mx & 3);
593 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] , srcY , s->linesize, v->rnd);
594 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
595 srcY += s->linesize * 8;
596 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize , srcY , s->linesize, v->rnd);
597 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
598 } else { // hpel mc - always used for luma
599 dxy = (my & 2) | ((mx & 2) >> 1);
602 dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
604 dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
607 if(s->flags & CODEC_FLAG_GRAY) return;
608 /* Chroma MC always uses qpel bilinear */
609 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
613 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
614 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
616 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
617 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
621 /** Do motion compensation for 4-MV macroblock - luminance block
623 static void vc1_mc_4mv_luma(VC1Context *v, int n)
625 MpegEncContext *s = &v->s;
626 DSPContext *dsp = &v->s.dsp;
628 int dxy, mx, my, src_x, src_y;
631 if(!v->s.last_picture.data[0])return;
634 srcY = s->last_picture.data[0];
636 off = s->linesize * 4 * (n&2) + (n&1) * 8;
638 src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);
639 src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);
641 if(v->profile != PROFILE_ADVANCED){
642 src_x = av_clip( src_x, -16, s->mb_width * 16);
643 src_y = av_clip( src_y, -16, s->mb_height * 16);
645 src_x = av_clip( src_x, -17, s->avctx->coded_width);
646 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
649 srcY += src_y * s->linesize + src_x;
651 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
652 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2
653 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){
654 srcY -= s->mspel * (1 + s->linesize);
655 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,
656 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
657 srcY = s->edge_emu_buffer;
658 /* if we deal with range reduction we need to scale source blocks */
664 for(j = 0; j < 9 + s->mspel*2; j++) {
665 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
669 /* if we deal with intensity compensation we need to scale source blocks */
670 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
675 for(j = 0; j < 9 + s->mspel*2; j++) {
676 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];
680 srcY += s->mspel * (1 + s->linesize);
684 dxy = ((my & 3) << 2) | (mx & 3);
685 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);
686 } else { // hpel mc - always used for luma
687 dxy = (my & 2) | ((mx & 2) >> 1);
689 dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
691 dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
695 static inline int median4(int a, int b, int c, int d)
698 if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
699 else return (FFMIN(b, c) + FFMAX(a, d)) / 2;
701 if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
702 else return (FFMIN(a, c) + FFMAX(b, d)) / 2;
707 /** Do motion compensation for 4-MV macroblock - both chroma blocks
709 static void vc1_mc_4mv_chroma(VC1Context *v)
711 MpegEncContext *s = &v->s;
712 DSPContext *dsp = &v->s.dsp;
713 uint8_t *srcU, *srcV;
714 int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;
715 int i, idx, tx = 0, ty = 0;
716 int mvx[4], mvy[4], intra[4];
717 static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
719 if(!v->s.last_picture.data[0])return;
720 if(s->flags & CODEC_FLAG_GRAY) return;
722 for(i = 0; i < 4; i++) {
723 mvx[i] = s->mv[0][i][0];
724 mvy[i] = s->mv[0][i][1];
725 intra[i] = v->mb_type[0][s->block_index[i]];
728 /* calculate chroma MV vector from four luma MVs */
729 idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];
730 if(!idx) { // all blocks are inter
731 tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
732 ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
733 } else if(count[idx] == 1) { // 3 inter blocks
736 tx = mid_pred(mvx[1], mvx[2], mvx[3]);
737 ty = mid_pred(mvy[1], mvy[2], mvy[3]);
740 tx = mid_pred(mvx[0], mvx[2], mvx[3]);
741 ty = mid_pred(mvy[0], mvy[2], mvy[3]);
744 tx = mid_pred(mvx[0], mvx[1], mvx[3]);
745 ty = mid_pred(mvy[0], mvy[1], mvy[3]);
748 tx = mid_pred(mvx[0], mvx[1], mvx[2]);
749 ty = mid_pred(mvy[0], mvy[1], mvy[2]);
752 } else if(count[idx] == 2) {
754 for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}
755 for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}
756 tx = (mvx[t1] + mvx[t2]) / 2;
757 ty = (mvy[t1] + mvy[t2]) / 2;
759 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
760 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
761 return; //no need to do MC for inter blocks
764 s->current_picture.motion_val[1][s->block_index[0]][0] = tx;
765 s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
766 uvmx = (tx + ((tx&3) == 3)) >> 1;
767 uvmy = (ty + ((ty&3) == 3)) >> 1;
769 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
770 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
773 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
774 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
776 if(v->profile != PROFILE_ADVANCED){
777 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
778 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
780 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
781 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
784 srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
785 srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
786 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
787 || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
788 || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){
789 ff_emulated_edge_mc(s->edge_emu_buffer , srcU, s->uvlinesize, 8+1, 8+1,
790 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
791 ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,
792 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
793 srcU = s->edge_emu_buffer;
794 srcV = s->edge_emu_buffer + 16;
796 /* if we deal with range reduction we need to scale source blocks */
801 src = srcU; src2 = srcV;
802 for(j = 0; j < 9; j++) {
803 for(i = 0; i < 9; i++) {
804 src[i] = ((src[i] - 128) >> 1) + 128;
805 src2[i] = ((src2[i] - 128) >> 1) + 128;
807 src += s->uvlinesize;
808 src2 += s->uvlinesize;
811 /* if we deal with intensity compensation we need to scale source blocks */
812 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
816 src = srcU; src2 = srcV;
817 for(j = 0; j < 9; j++) {
818 for(i = 0; i < 9; i++) {
819 src[i] = v->lutuv[src[i]];
820 src2[i] = v->lutuv[src2[i]];
822 src += s->uvlinesize;
823 src2 += s->uvlinesize;
828 /* Chroma MC always uses qpel bilinear */
829 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
833 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
834 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
836 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
837 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
841 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);
844 * Decode Simple/Main Profiles sequence header
845 * @see Figure 7-8, p16-17
846 * @param avctx Codec context
847 * @param gb GetBit context initialized from Codec context extra_data
850 static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)
852 VC1Context *v = avctx->priv_data;
854 av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));
855 v->profile = get_bits(gb, 2);
856 if (v->profile == PROFILE_COMPLEX)
858 av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");
861 if (v->profile == PROFILE_ADVANCED)
863 v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz;
864 v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz;
865 return decode_sequence_header_adv(v, gb);
869 v->zz_8x4 = wmv2_scantableA;
870 v->zz_4x8 = wmv2_scantableB;
871 v->res_sm = get_bits(gb, 2); //reserved
874 av_log(avctx, AV_LOG_ERROR,
875 "Reserved RES_SM=%i is forbidden\n", v->res_sm);
881 v->frmrtq_postproc = get_bits(gb, 3); //common
882 // (bitrate-32kbps)/64kbps
883 v->bitrtq_postproc = get_bits(gb, 5); //common
884 v->s.loop_filter = get_bits1(gb); //common
885 if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)
887 av_log(avctx, AV_LOG_ERROR,
888 "LOOPFILTER shell not be enabled in simple profile\n");
890 if(v->s.avctx->skip_loop_filter >= AVDISCARD_ALL)
891 v->s.loop_filter = 0;
893 v->res_x8 = get_bits1(gb); //reserved
894 v->multires = get_bits1(gb);
895 v->res_fasttx = get_bits1(gb);
898 v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct;
899 v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;
900 v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;
901 v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;
904 v->fastuvmc = get_bits1(gb); //common
905 if (!v->profile && !v->fastuvmc)
907 av_log(avctx, AV_LOG_ERROR,
908 "FASTUVMC unavailable in Simple Profile\n");
911 v->extended_mv = get_bits1(gb); //common
912 if (!v->profile && v->extended_mv)
914 av_log(avctx, AV_LOG_ERROR,
915 "Extended MVs unavailable in Simple Profile\n");
918 v->dquant = get_bits(gb, 2); //common
919 v->vstransform = get_bits1(gb); //common
921 v->res_transtab = get_bits1(gb);
924 av_log(avctx, AV_LOG_ERROR,
925 "1 for reserved RES_TRANSTAB is forbidden\n");
929 v->overlap = get_bits1(gb); //common
931 v->s.resync_marker = get_bits1(gb);
932 v->rangered = get_bits1(gb);
933 if (v->rangered && v->profile == PROFILE_SIMPLE)
935 av_log(avctx, AV_LOG_INFO,
936 "RANGERED should be set to 0 in simple profile\n");
939 v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common
940 v->quantizer_mode = get_bits(gb, 2); //common
942 v->finterpflag = get_bits1(gb); //common
943 v->res_rtm_flag = get_bits1(gb); //reserved
944 if (!v->res_rtm_flag)
946 // av_log(avctx, AV_LOG_ERROR,
947 // "0 for reserved RES_RTM_FLAG is forbidden\n");
948 av_log(avctx, AV_LOG_ERROR,
949 "Old WMV3 version detected, only I-frames will be decoded\n");
952 //TODO: figure out what they mean (always 0x402F)
953 if(!v->res_fasttx) skip_bits(gb, 16);
954 av_log(avctx, AV_LOG_DEBUG,
955 "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
956 "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"
957 "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"
958 "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",
959 v->profile, v->frmrtq_postproc, v->bitrtq_postproc,
960 v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,
961 v->rangered, v->vstransform, v->overlap, v->s.resync_marker,
962 v->dquant, v->quantizer_mode, avctx->max_b_frames
967 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)
970 v->level = get_bits(gb, 3);
973 av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);
975 v->chromaformat = get_bits(gb, 2);
976 if (v->chromaformat != 1)
978 av_log(v->s.avctx, AV_LOG_ERROR,
979 "Only 4:2:0 chroma format supported\n");
984 v->frmrtq_postproc = get_bits(gb, 3); //common
985 // (bitrate-32kbps)/64kbps
986 v->bitrtq_postproc = get_bits(gb, 5); //common
987 v->postprocflag = get_bits1(gb); //common
989 v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;
990 v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;
991 v->s.avctx->width = v->s.avctx->coded_width;
992 v->s.avctx->height = v->s.avctx->coded_height;
993 v->broadcast = get_bits1(gb);
994 v->interlace = get_bits1(gb);
995 v->tfcntrflag = get_bits1(gb);
996 v->finterpflag = get_bits1(gb);
997 skip_bits1(gb); // reserved
999 v->s.h_edge_pos = v->s.avctx->coded_width;
1000 v->s.v_edge_pos = v->s.avctx->coded_height;
1002 av_log(v->s.avctx, AV_LOG_DEBUG,
1003 "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
1004 "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"
1005 "TFCTRflag=%i, FINTERPflag=%i\n",
1006 v->level, v->frmrtq_postproc, v->bitrtq_postproc,
1007 v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,
1008 v->tfcntrflag, v->finterpflag
1011 v->psf = get_bits1(gb);
1012 if(v->psf) { //PsF, 6.1.13
1013 av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");
1016 v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
1017 if(get_bits1(gb)) { //Display Info - decoding is not affected by it
1019 av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
1020 v->s.avctx->coded_width = w = get_bits(gb, 14) + 1;
1021 v->s.avctx->coded_height = h = get_bits(gb, 14) + 1;
1022 av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
1024 ar = get_bits(gb, 4);
1026 v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
1028 w = get_bits(gb, 8);
1029 h = get_bits(gb, 8);
1030 v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
1032 av_log(v->s.avctx, AV_LOG_DEBUG, "Aspect: %i:%i\n", v->s.avctx->sample_aspect_ratio.num, v->s.avctx->sample_aspect_ratio.den);
1034 if(get_bits1(gb)){ //framerate stuff
1036 v->s.avctx->time_base.num = 32;
1037 v->s.avctx->time_base.den = get_bits(gb, 16) + 1;
1040 nr = get_bits(gb, 8);
1041 dr = get_bits(gb, 4);
1042 if(nr && nr < 8 && dr && dr < 3){
1043 v->s.avctx->time_base.num = ff_vc1_fps_dr[dr - 1];
1044 v->s.avctx->time_base.den = ff_vc1_fps_nr[nr - 1] * 1000;
1050 v->color_prim = get_bits(gb, 8);
1051 v->transfer_char = get_bits(gb, 8);
1052 v->matrix_coef = get_bits(gb, 8);
1056 v->hrd_param_flag = get_bits1(gb);
1057 if(v->hrd_param_flag) {
1059 v->hrd_num_leaky_buckets = get_bits(gb, 5);
1060 skip_bits(gb, 4); //bitrate exponent
1061 skip_bits(gb, 4); //buffer size exponent
1062 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1063 skip_bits(gb, 16); //hrd_rate[n]
1064 skip_bits(gb, 16); //hrd_buffer[n]
1070 static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb)
1072 VC1Context *v = avctx->priv_data;
1073 int i, blink, clentry;
1075 av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32));
1076 blink = get_bits1(gb); // broken link
1077 clentry = get_bits1(gb); // closed entry
1078 v->panscanflag = get_bits1(gb);
1079 v->refdist_flag = get_bits1(gb);
1080 v->s.loop_filter = get_bits1(gb);
1081 v->fastuvmc = get_bits1(gb);
1082 v->extended_mv = get_bits1(gb);
1083 v->dquant = get_bits(gb, 2);
1084 v->vstransform = get_bits1(gb);
1085 v->overlap = get_bits1(gb);
1086 v->quantizer_mode = get_bits(gb, 2);
1088 if(v->hrd_param_flag){
1089 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1090 skip_bits(gb, 8); //hrd_full[n]
1095 avctx->coded_width = (get_bits(gb, 12)+1)<<1;
1096 avctx->coded_height = (get_bits(gb, 12)+1)<<1;
1099 v->extended_dmv = get_bits1(gb);
1100 if((v->range_mapy_flag = get_bits1(gb))) {
1101 av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n");
1102 v->range_mapy = get_bits(gb, 3);
1104 if((v->range_mapuv_flag = get_bits1(gb))) {
1105 av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n");
1106 v->range_mapuv = get_bits(gb, 3);
1109 av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n"
1110 "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n"
1111 "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n"
1112 "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n",
1113 blink, clentry, v->panscanflag, v->refdist_flag, v->s.loop_filter,
1114 v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode);
1119 static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)
1121 int pqindex, lowquant, status;
1123 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1124 skip_bits(gb, 2); //framecnt unused
1126 if (v->rangered) v->rangeredfrm = get_bits1(gb);
1127 v->s.pict_type = get_bits1(gb);
1128 if (v->s.avctx->max_b_frames) {
1129 if (!v->s.pict_type) {
1130 if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE;
1131 else v->s.pict_type = FF_B_TYPE;
1132 } else v->s.pict_type = FF_P_TYPE;
1133 } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE;
1136 if(v->s.pict_type == FF_B_TYPE) {
1137 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1138 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1139 if(v->bfraction == 0) {
1140 v->s.pict_type = FF_BI_TYPE;
1143 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1144 skip_bits(gb, 7); // skip buffer fullness
1147 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1149 if(v->s.pict_type == FF_P_TYPE)
1152 /* Quantizer stuff */
1153 pqindex = get_bits(gb, 5);
1154 if(!pqindex) return -1;
1155 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1156 v->pq = ff_vc1_pquant_table[0][pqindex];
1158 v->pq = ff_vc1_pquant_table[1][pqindex];
1161 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1162 v->pquantizer = pqindex < 9;
1163 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1165 v->pqindex = pqindex;
1166 if (pqindex < 9) v->halfpq = get_bits1(gb);
1168 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1169 v->pquantizer = get_bits1(gb);
1171 if (v->extended_mv == 1) v->mvrange = get_unary(gb, 0, 3);
1172 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1173 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1174 v->range_x = 1 << (v->k_x - 1);
1175 v->range_y = 1 << (v->k_y - 1);
1176 if (v->profile == PROFILE_ADVANCED)
1178 if (v->postprocflag) v->postproc = get_bits1(gb);
1181 if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2);
1183 if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){
1184 v->x8_type = get_bits1(gb);
1185 }else v->x8_type = 0;
1186 //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",
1187 // (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);
1189 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1191 switch(v->s.pict_type) {
1193 if (v->pq < 5) v->tt_index = 0;
1194 else if(v->pq < 13) v->tt_index = 1;
1195 else v->tt_index = 2;
1197 lowquant = (v->pq > 12) ? 0 : 1;
1198 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1199 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1201 int scale, shift, i;
1202 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1203 v->lumscale = get_bits(gb, 6);
1204 v->lumshift = get_bits(gb, 6);
1206 /* fill lookup tables for intensity compensation */
1209 shift = (255 - v->lumshift * 2) << 6;
1210 if(v->lumshift > 31)
1213 scale = v->lumscale + 32;
1214 if(v->lumshift > 31)
1215 shift = (v->lumshift - 64) << 6;
1217 shift = v->lumshift << 6;
1219 for(i = 0; i < 256; i++) {
1220 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1221 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1224 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1225 v->s.quarter_sample = 0;
1226 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1227 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1228 v->s.quarter_sample = 0;
1230 v->s.quarter_sample = 1;
1232 v->s.quarter_sample = 1;
1233 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));
1235 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1236 v->mv_mode2 == MV_PMODE_MIXED_MV)
1237 || v->mv_mode == MV_PMODE_MIXED_MV)
1239 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1240 if (status < 0) return -1;
1241 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1242 "Imode: %i, Invert: %i\n", status>>1, status&1);
1244 v->mv_type_is_raw = 0;
1245 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1247 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1248 if (status < 0) return -1;
1249 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1250 "Imode: %i, Invert: %i\n", status>>1, status&1);
1252 /* Hopefully this is correct for P frames */
1253 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1254 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1258 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1259 vop_dquant_decoding(v);
1262 v->ttfrm = 0; //FIXME Is that so ?
1265 v->ttmbf = get_bits1(gb);
1268 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1276 if (v->pq < 5) v->tt_index = 0;
1277 else if(v->pq < 13) v->tt_index = 1;
1278 else v->tt_index = 2;
1280 lowquant = (v->pq > 12) ? 0 : 1;
1281 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1282 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1283 v->s.mspel = v->s.quarter_sample;
1285 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1286 if (status < 0) return -1;
1287 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1288 "Imode: %i, Invert: %i\n", status>>1, status&1);
1289 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1290 if (status < 0) return -1;
1291 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1292 "Imode: %i, Invert: %i\n", status>>1, status&1);
1294 v->s.mv_table_index = get_bits(gb, 2);
1295 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1299 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1300 vop_dquant_decoding(v);
1306 v->ttmbf = get_bits1(gb);
1309 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1321 v->c_ac_table_index = decode012(gb);
1322 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1324 v->y_ac_table_index = decode012(gb);
1327 v->s.dc_table_index = get_bits1(gb);
1330 if(v->s.pict_type == FF_BI_TYPE) {
1331 v->s.pict_type = FF_B_TYPE;
1337 static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)
1339 int pqindex, lowquant;
1342 v->p_frame_skipped = 0;
1345 v->fcm = decode012(gb);
1346 if(v->fcm) return -1; // interlaced frames/fields are not implemented
1348 switch(get_unary(gb, 0, 4)) {
1350 v->s.pict_type = FF_P_TYPE;
1353 v->s.pict_type = FF_B_TYPE;
1356 v->s.pict_type = FF_I_TYPE;
1359 v->s.pict_type = FF_BI_TYPE;
1362 v->s.pict_type = FF_P_TYPE; // skipped pic
1363 v->p_frame_skipped = 1;
1369 if(!v->interlace || v->psf) {
1370 v->rptfrm = get_bits(gb, 2);
1372 v->tff = get_bits1(gb);
1373 v->rptfrm = get_bits1(gb);
1376 if(v->panscanflag) {
1379 v->rnd = get_bits1(gb);
1381 v->uvsamp = get_bits1(gb);
1382 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1383 if(v->s.pict_type == FF_B_TYPE) {
1384 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1385 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1386 if(v->bfraction == 0) {
1387 v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */
1390 pqindex = get_bits(gb, 5);
1391 if(!pqindex) return -1;
1392 v->pqindex = pqindex;
1393 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1394 v->pq = ff_vc1_pquant_table[0][pqindex];
1396 v->pq = ff_vc1_pquant_table[1][pqindex];
1399 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1400 v->pquantizer = pqindex < 9;
1401 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1403 v->pqindex = pqindex;
1404 if (pqindex < 9) v->halfpq = get_bits1(gb);
1406 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1407 v->pquantizer = get_bits1(gb);
1409 v->postproc = get_bits1(gb);
1411 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1413 switch(v->s.pict_type) {
1416 status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v);
1417 if (status < 0) return -1;
1418 av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: "
1419 "Imode: %i, Invert: %i\n", status>>1, status&1);
1420 v->condover = CONDOVER_NONE;
1421 if(v->overlap && v->pq <= 8) {
1422 v->condover = decode012(gb);
1423 if(v->condover == CONDOVER_SELECT) {
1424 status = bitplane_decoding(v->over_flags_plane, &v->overflg_is_raw, v);
1425 if (status < 0) return -1;
1426 av_log(v->s.avctx, AV_LOG_DEBUG, "CONDOVER plane encoding: "
1427 "Imode: %i, Invert: %i\n", status>>1, status&1);
1432 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1433 else v->mvrange = 0;
1434 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1435 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1436 v->range_x = 1 << (v->k_x - 1);
1437 v->range_y = 1 << (v->k_y - 1);
1439 if (v->pq < 5) v->tt_index = 0;
1440 else if(v->pq < 13) v->tt_index = 1;
1441 else v->tt_index = 2;
1443 lowquant = (v->pq > 12) ? 0 : 1;
1444 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1445 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1447 int scale, shift, i;
1448 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1449 v->lumscale = get_bits(gb, 6);
1450 v->lumshift = get_bits(gb, 6);
1451 /* fill lookup tables for intensity compensation */
1454 shift = (255 - v->lumshift * 2) << 6;
1455 if(v->lumshift > 31)
1458 scale = v->lumscale + 32;
1459 if(v->lumshift > 31)
1460 shift = (v->lumshift - 64) << 6;
1462 shift = v->lumshift << 6;
1464 for(i = 0; i < 256; i++) {
1465 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1466 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1470 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1471 v->s.quarter_sample = 0;
1472 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1473 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1474 v->s.quarter_sample = 0;
1476 v->s.quarter_sample = 1;
1478 v->s.quarter_sample = 1;
1479 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));
1481 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1482 v->mv_mode2 == MV_PMODE_MIXED_MV)
1483 || v->mv_mode == MV_PMODE_MIXED_MV)
1485 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1486 if (status < 0) return -1;
1487 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1488 "Imode: %i, Invert: %i\n", status>>1, status&1);
1490 v->mv_type_is_raw = 0;
1491 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1493 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1494 if (status < 0) return -1;
1495 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1496 "Imode: %i, Invert: %i\n", status>>1, status&1);
1498 /* Hopefully this is correct for P frames */
1499 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1500 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1503 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1504 vop_dquant_decoding(v);
1507 v->ttfrm = 0; //FIXME Is that so ?
1510 v->ttmbf = get_bits1(gb);
1513 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1521 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1522 else v->mvrange = 0;
1523 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1524 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1525 v->range_x = 1 << (v->k_x - 1);
1526 v->range_y = 1 << (v->k_y - 1);
1528 if (v->pq < 5) v->tt_index = 0;
1529 else if(v->pq < 13) v->tt_index = 1;
1530 else v->tt_index = 2;
1532 lowquant = (v->pq > 12) ? 0 : 1;
1533 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1534 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1535 v->s.mspel = v->s.quarter_sample;
1537 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1538 if (status < 0) return -1;
1539 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1540 "Imode: %i, Invert: %i\n", status>>1, status&1);
1541 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1542 if (status < 0) return -1;
1543 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1544 "Imode: %i, Invert: %i\n", status>>1, status&1);
1546 v->s.mv_table_index = get_bits(gb, 2);
1547 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1551 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1552 vop_dquant_decoding(v);
1558 v->ttmbf = get_bits1(gb);
1561 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1571 v->c_ac_table_index = decode012(gb);
1572 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1574 v->y_ac_table_index = decode012(gb);
1577 v->s.dc_table_index = get_bits1(gb);
1578 if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) {
1579 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1580 vop_dquant_decoding(v);
1584 if(v->s.pict_type == FF_BI_TYPE) {
1585 v->s.pict_type = FF_B_TYPE;
1591 /***********************************************************************/
1593 * @defgroup block VC-1 Block-level functions
1594 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
1600 * @brief Get macroblock-level quantizer scale
1602 #define GET_MQUANT() \
1606 if (v->dqprofile == DQPROFILE_ALL_MBS) \
1610 mquant = (get_bits1(gb)) ? v->altpq : v->pq; \
1614 mqdiff = get_bits(gb, 3); \
1615 if (mqdiff != 7) mquant = v->pq + mqdiff; \
1616 else mquant = get_bits(gb, 5); \
1619 if(v->dqprofile == DQPROFILE_SINGLE_EDGE) \
1620 edges = 1 << v->dqsbedge; \
1621 else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES) \
1622 edges = (3 << v->dqsbedge) % 15; \
1623 else if(v->dqprofile == DQPROFILE_FOUR_EDGES) \
1625 if((edges&1) && !s->mb_x) \
1626 mquant = v->altpq; \
1627 if((edges&2) && s->first_slice_line) \
1628 mquant = v->altpq; \
1629 if((edges&4) && s->mb_x == (s->mb_width - 1)) \
1630 mquant = v->altpq; \
1631 if((edges&8) && s->mb_y == (s->mb_height - 1)) \
1632 mquant = v->altpq; \
1636 * @def GET_MVDATA(_dmv_x, _dmv_y)
1637 * @brief Get MV differentials
1638 * @see MVDATA decoding from 8.3.5.2, p(1)20
1639 * @param _dmv_x Horizontal differential for decoded MV
1640 * @param _dmv_y Vertical differential for decoded MV
1642 #define GET_MVDATA(_dmv_x, _dmv_y) \
1643 index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table,\
1644 VC1_MV_DIFF_VLC_BITS, 2); \
1647 mb_has_coeffs = 1; \
1650 else mb_has_coeffs = 0; \
1652 if (!index) { _dmv_x = _dmv_y = 0; } \
1653 else if (index == 35) \
1655 _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample); \
1656 _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample); \
1658 else if (index == 36) \
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_x = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1676 if (!s->quarter_sample && index1 == 5) val = 1; \
1678 if(size_table[index1] - val > 0) \
1679 val = get_bits(gb, size_table[index1] - val); \
1681 sign = 0 - (val&1); \
1682 _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1685 /** Predict and set motion vector
1687 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)
1689 int xy, wrap, off = 0;
1694 /* scale MV difference to be quad-pel */
1695 dmv_x <<= 1 - s->quarter_sample;
1696 dmv_y <<= 1 - s->quarter_sample;
1698 wrap = s->b8_stride;
1699 xy = s->block_index[n];
1702 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
1703 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
1704 s->current_picture.motion_val[1][xy][0] = 0;
1705 s->current_picture.motion_val[1][xy][1] = 0;
1706 if(mv1) { /* duplicate motion data for 1-MV block */
1707 s->current_picture.motion_val[0][xy + 1][0] = 0;
1708 s->current_picture.motion_val[0][xy + 1][1] = 0;
1709 s->current_picture.motion_val[0][xy + wrap][0] = 0;
1710 s->current_picture.motion_val[0][xy + wrap][1] = 0;
1711 s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
1712 s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
1713 s->current_picture.motion_val[1][xy + 1][0] = 0;
1714 s->current_picture.motion_val[1][xy + 1][1] = 0;
1715 s->current_picture.motion_val[1][xy + wrap][0] = 0;
1716 s->current_picture.motion_val[1][xy + wrap][1] = 0;
1717 s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
1718 s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
1723 C = s->current_picture.motion_val[0][xy - 1];
1724 A = s->current_picture.motion_val[0][xy - wrap];
1726 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
1728 //in 4-MV mode different blocks have different B predictor position
1731 off = (s->mb_x > 0) ? -1 : 1;
1734 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
1743 B = s->current_picture.motion_val[0][xy - wrap + off];
1745 if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds
1746 if(s->mb_width == 1) {
1750 px = mid_pred(A[0], B[0], C[0]);
1751 py = mid_pred(A[1], B[1], C[1]);
1753 } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds
1759 /* Pullback MV as specified in 8.3.5.3.4 */
1762 qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);
1763 qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);
1764 X = (s->mb_width << 6) - 4;
1765 Y = (s->mb_height << 6) - 4;
1767 if(qx + px < -60) px = -60 - qx;
1768 if(qy + py < -60) py = -60 - qy;
1770 if(qx + px < -28) px = -28 - qx;
1771 if(qy + py < -28) py = -28 - qy;
1773 if(qx + px > X) px = X - qx;
1774 if(qy + py > Y) py = Y - qy;
1776 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
1777 if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {
1778 if(is_intra[xy - wrap])
1779 sum = FFABS(px) + FFABS(py);
1781 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
1783 if(get_bits1(&s->gb)) {
1791 if(is_intra[xy - 1])
1792 sum = FFABS(px) + FFABS(py);
1794 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
1796 if(get_bits1(&s->gb)) {
1806 /* store MV using signed modulus of MV range defined in 4.11 */
1807 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
1808 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;
1809 if(mv1) { /* duplicate motion data for 1-MV block */
1810 s->current_picture.motion_val[0][xy + 1][0] = s->current_picture.motion_val[0][xy][0];
1811 s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];
1812 s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];
1813 s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];
1814 s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];
1815 s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];
1819 /** Motion compensation for direct or interpolated blocks in B-frames
1821 static void vc1_interp_mc(VC1Context *v)
1823 MpegEncContext *s = &v->s;
1824 DSPContext *dsp = &v->s.dsp;
1825 uint8_t *srcY, *srcU, *srcV;
1826 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
1828 if(!v->s.next_picture.data[0])return;
1830 mx = s->mv[1][0][0];
1831 my = s->mv[1][0][1];
1832 uvmx = (mx + ((mx & 3) == 3)) >> 1;
1833 uvmy = (my + ((my & 3) == 3)) >> 1;
1835 uvmx = uvmx + ((uvmx<0)?-(uvmx&1):(uvmx&1));
1836 uvmy = uvmy + ((uvmy<0)?-(uvmy&1):(uvmy&1));
1838 srcY = s->next_picture.data[0];
1839 srcU = s->next_picture.data[1];
1840 srcV = s->next_picture.data[2];
1842 src_x = s->mb_x * 16 + (mx >> 2);
1843 src_y = s->mb_y * 16 + (my >> 2);
1844 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
1845 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
1847 if(v->profile != PROFILE_ADVANCED){
1848 src_x = av_clip( src_x, -16, s->mb_width * 16);
1849 src_y = av_clip( src_y, -16, s->mb_height * 16);
1850 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
1851 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
1853 src_x = av_clip( src_x, -17, s->avctx->coded_width);
1854 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
1855 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
1856 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
1859 srcY += src_y * s->linesize + src_x;
1860 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
1861 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
1863 /* for grayscale we should not try to read from unknown area */
1864 if(s->flags & CODEC_FLAG_GRAY) {
1865 srcU = s->edge_emu_buffer + 18 * s->linesize;
1866 srcV = s->edge_emu_buffer + 18 * s->linesize;
1870 || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16
1871 || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){
1872 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
1874 srcY -= s->mspel * (1 + s->linesize);
1875 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
1876 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
1877 srcY = s->edge_emu_buffer;
1878 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
1879 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1880 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
1881 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1884 /* if we deal with range reduction we need to scale source blocks */
1885 if(v->rangeredfrm) {
1887 uint8_t *src, *src2;
1890 for(j = 0; j < 17 + s->mspel*2; j++) {
1891 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
1894 src = srcU; src2 = srcV;
1895 for(j = 0; j < 9; j++) {
1896 for(i = 0; i < 9; i++) {
1897 src[i] = ((src[i] - 128) >> 1) + 128;
1898 src2[i] = ((src2[i] - 128) >> 1) + 128;
1900 src += s->uvlinesize;
1901 src2 += s->uvlinesize;
1904 srcY += s->mspel * (1 + s->linesize);
1909 dxy = ((my & 1) << 1) | (mx & 1);
1911 dsp->avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
1913 if(s->flags & CODEC_FLAG_GRAY) return;
1914 /* Chroma MC always uses qpel blilinear */
1915 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
1918 dsp->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
1919 dsp->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
1922 static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)
1926 #if B_FRACTION_DEN==256
1930 return 2 * ((value * n + 255) >> 9);
1931 return (value * n + 128) >> 8;
1934 n -= B_FRACTION_DEN;
1936 return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));
1937 return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;
1941 /** Reconstruct motion vector for B-frame and do motion compensation
1943 static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode)
1946 v->mv_mode2 = v->mv_mode;
1947 v->mv_mode = MV_PMODE_INTENSITY_COMP;
1952 if(v->use_ic) v->mv_mode = v->mv_mode2;
1955 if(mode == BMV_TYPE_INTERPOLATED) {
1958 if(v->use_ic) v->mv_mode = v->mv_mode2;
1962 if(v->use_ic && (mode == BMV_TYPE_BACKWARD)) v->mv_mode = v->mv_mode2;
1963 vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));
1964 if(v->use_ic) v->mv_mode = v->mv_mode2;
1967 static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mvtype)
1969 MpegEncContext *s = &v->s;
1970 int xy, wrap, off = 0;
1975 const uint8_t *is_intra = v->mb_type[0];
1979 /* scale MV difference to be quad-pel */
1980 dmv_x[0] <<= 1 - s->quarter_sample;
1981 dmv_y[0] <<= 1 - s->quarter_sample;
1982 dmv_x[1] <<= 1 - s->quarter_sample;
1983 dmv_y[1] <<= 1 - s->quarter_sample;
1985 wrap = s->b8_stride;
1986 xy = s->block_index[0];
1989 s->current_picture.motion_val[0][xy][0] =
1990 s->current_picture.motion_val[0][xy][1] =
1991 s->current_picture.motion_val[1][xy][0] =
1992 s->current_picture.motion_val[1][xy][1] = 0;
1995 s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);
1996 s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);
1997 s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);
1998 s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);
2000 /* Pullback predicted motion vectors as specified in 8.4.5.4 */
2001 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));
2002 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));
2003 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));
2004 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));
2006 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2007 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2008 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2009 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2013 if((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2014 C = s->current_picture.motion_val[0][xy - 2];
2015 A = s->current_picture.motion_val[0][xy - wrap*2];
2016 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2017 B = s->current_picture.motion_val[0][xy - wrap*2 + off];
2019 if(!s->mb_x) C[0] = C[1] = 0;
2020 if(!s->first_slice_line) { // predictor A is not out of bounds
2021 if(s->mb_width == 1) {
2025 px = mid_pred(A[0], B[0], C[0]);
2026 py = mid_pred(A[1], B[1], C[1]);
2028 } else if(s->mb_x) { // predictor C is not out of bounds
2034 /* Pullback MV as specified in 8.3.5.3.4 */
2037 if(v->profile < PROFILE_ADVANCED) {
2038 qx = (s->mb_x << 5);
2039 qy = (s->mb_y << 5);
2040 X = (s->mb_width << 5) - 4;
2041 Y = (s->mb_height << 5) - 4;
2042 if(qx + px < -28) px = -28 - qx;
2043 if(qy + py < -28) py = -28 - qy;
2044 if(qx + px > X) px = X - qx;
2045 if(qy + py > Y) py = Y - qy;
2047 qx = (s->mb_x << 6);
2048 qy = (s->mb_y << 6);
2049 X = (s->mb_width << 6) - 4;
2050 Y = (s->mb_height << 6) - 4;
2051 if(qx + px < -60) px = -60 - qx;
2052 if(qy + py < -60) py = -60 - qy;
2053 if(qx + px > X) px = X - qx;
2054 if(qy + py > Y) py = Y - qy;
2057 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2058 if(0 && !s->first_slice_line && s->mb_x) {
2059 if(is_intra[xy - wrap])
2060 sum = FFABS(px) + FFABS(py);
2062 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2064 if(get_bits1(&s->gb)) {
2072 if(is_intra[xy - 2])
2073 sum = FFABS(px) + FFABS(py);
2075 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2077 if(get_bits1(&s->gb)) {
2087 /* store MV using signed modulus of MV range defined in 4.11 */
2088 s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;
2089 s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;
2091 if((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2092 C = s->current_picture.motion_val[1][xy - 2];
2093 A = s->current_picture.motion_val[1][xy - wrap*2];
2094 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2095 B = s->current_picture.motion_val[1][xy - wrap*2 + off];
2097 if(!s->mb_x) C[0] = C[1] = 0;
2098 if(!s->first_slice_line) { // predictor A is not out of bounds
2099 if(s->mb_width == 1) {
2103 px = mid_pred(A[0], B[0], C[0]);
2104 py = mid_pred(A[1], B[1], C[1]);
2106 } else if(s->mb_x) { // predictor C is not out of bounds
2112 /* Pullback MV as specified in 8.3.5.3.4 */
2115 if(v->profile < PROFILE_ADVANCED) {
2116 qx = (s->mb_x << 5);
2117 qy = (s->mb_y << 5);
2118 X = (s->mb_width << 5) - 4;
2119 Y = (s->mb_height << 5) - 4;
2120 if(qx + px < -28) px = -28 - qx;
2121 if(qy + py < -28) py = -28 - qy;
2122 if(qx + px > X) px = X - qx;
2123 if(qy + py > Y) py = Y - qy;
2125 qx = (s->mb_x << 6);
2126 qy = (s->mb_y << 6);
2127 X = (s->mb_width << 6) - 4;
2128 Y = (s->mb_height << 6) - 4;
2129 if(qx + px < -60) px = -60 - qx;
2130 if(qy + py < -60) py = -60 - qy;
2131 if(qx + px > X) px = X - qx;
2132 if(qy + py > Y) py = Y - qy;
2135 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2136 if(0 && !s->first_slice_line && s->mb_x) {
2137 if(is_intra[xy - wrap])
2138 sum = FFABS(px) + FFABS(py);
2140 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2142 if(get_bits1(&s->gb)) {
2150 if(is_intra[xy - 2])
2151 sum = FFABS(px) + FFABS(py);
2153 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2155 if(get_bits1(&s->gb)) {
2165 /* store MV using signed modulus of MV range defined in 4.11 */
2167 s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;
2168 s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;
2170 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2171 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2172 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2173 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2176 /** Get predicted DC value for I-frames only
2177 * prediction dir: left=0, top=1
2178 * @param s MpegEncContext
2179 * @param[in] n block index in the current MB
2180 * @param dc_val_ptr Pointer to DC predictor
2181 * @param dir_ptr Prediction direction for use in AC prediction
2183 static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2184 int16_t **dc_val_ptr, int *dir_ptr)
2186 int a, b, c, wrap, pred, scale;
2188 static const uint16_t dcpred[32] = {
2189 -1, 1024, 512, 341, 256, 205, 171, 146, 128,
2190 114, 102, 93, 85, 79, 73, 68, 64,
2191 60, 57, 54, 51, 49, 47, 45, 43,
2192 41, 39, 38, 37, 35, 34, 33
2195 /* find prediction - wmv3_dc_scale always used here in fact */
2196 if (n < 4) scale = s->y_dc_scale;
2197 else scale = s->c_dc_scale;
2199 wrap = s->block_wrap[n];
2200 dc_val= s->dc_val[0] + s->block_index[n];
2206 b = dc_val[ - 1 - wrap];
2207 a = dc_val[ - wrap];
2209 if (pq < 9 || !overlap)
2211 /* Set outer values */
2212 if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];
2213 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];
2217 /* Set outer values */
2218 if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;
2219 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;
2222 if (abs(a - b) <= abs(b - c)) {
2230 /* update predictor */
2231 *dc_val_ptr = &dc_val[0];
2236 /** Get predicted DC value
2237 * prediction dir: left=0, top=1
2238 * @param s MpegEncContext
2239 * @param[in] n block index in the current MB
2240 * @param dc_val_ptr Pointer to DC predictor
2241 * @param dir_ptr Prediction direction for use in AC prediction
2243 static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2244 int a_avail, int c_avail,
2245 int16_t **dc_val_ptr, int *dir_ptr)
2247 int a, b, c, wrap, pred, scale;
2249 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2252 /* find prediction - wmv3_dc_scale always used here in fact */
2253 if (n < 4) scale = s->y_dc_scale;
2254 else scale = s->c_dc_scale;
2256 wrap = s->block_wrap[n];
2257 dc_val= s->dc_val[0] + s->block_index[n];
2263 b = dc_val[ - 1 - wrap];
2264 a = dc_val[ - wrap];
2265 /* scale predictors if needed */
2266 q1 = s->current_picture.qscale_table[mb_pos];
2267 if(c_avail && (n!= 1 && n!=3)) {
2268 q2 = s->current_picture.qscale_table[mb_pos - 1];
2270 c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2272 if(a_avail && (n!= 2 && n!=3)) {
2273 q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2275 a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2277 if(a_avail && c_avail && (n!=3)) {
2280 if(n != 2) off -= s->mb_stride;
2281 q2 = s->current_picture.qscale_table[off];
2283 b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2286 if(a_avail && c_avail) {
2287 if(abs(a - b) <= abs(b - c)) {
2294 } else if(a_avail) {
2297 } else if(c_avail) {
2305 /* update predictor */
2306 *dc_val_ptr = &dc_val[0];
2312 * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles
2313 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
2317 static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)
2319 int xy, wrap, pred, a, b, c;
2321 xy = s->block_index[n];
2322 wrap = s->b8_stride;
2327 a = s->coded_block[xy - 1 ];
2328 b = s->coded_block[xy - 1 - wrap];
2329 c = s->coded_block[xy - wrap];
2338 *coded_block_ptr = &s->coded_block[xy];
2344 * Decode one AC coefficient
2345 * @param v The VC1 context
2346 * @param last Last coefficient
2347 * @param skip How much zero coefficients to skip
2348 * @param value Decoded AC coefficient value
2351 static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)
2353 GetBitContext *gb = &v->s.gb;
2354 int index, escape, run = 0, level = 0, lst = 0;
2356 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2357 if (index != vc1_ac_sizes[codingset] - 1) {
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];
2364 escape = decode210(gb);
2366 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2367 run = vc1_index_decode_table[codingset][index][0];
2368 level = vc1_index_decode_table[codingset][index][1];
2369 lst = index >= vc1_last_decode_table[codingset];
2372 level += vc1_last_delta_level_table[codingset][run];
2374 level += vc1_delta_level_table[codingset][run];
2377 run += vc1_last_delta_run_table[codingset][level] + 1;
2379 run += vc1_delta_run_table[codingset][level] + 1;
2385 lst = get_bits1(gb);
2386 if(v->s.esc3_level_length == 0) {
2387 if(v->pq < 8 || v->dquantfrm) { // table 59
2388 v->s.esc3_level_length = get_bits(gb, 3);
2389 if(!v->s.esc3_level_length)
2390 v->s.esc3_level_length = get_bits(gb, 2) + 8;
2392 v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;
2394 v->s.esc3_run_length = 3 + get_bits(gb, 2);
2396 run = get_bits(gb, v->s.esc3_run_length);
2397 sign = get_bits1(gb);
2398 level = get_bits(gb, v->s.esc3_level_length);
2409 /** Decode intra block in intra frames - should be faster than decode_intra_block
2410 * @param v VC1Context
2411 * @param block block to decode
2412 * @param coded are AC coeffs present or not
2413 * @param codingset set of VLC to decode data
2415 static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)
2417 GetBitContext *gb = &v->s.gb;
2418 MpegEncContext *s = &v->s;
2419 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2422 int16_t *ac_val, *ac_val2;
2425 /* Get DC differential */
2427 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2429 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2432 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2437 if (dcdiff == 119 /* ESC index value */)
2439 /* TODO: Optimize */
2440 if (v->pq == 1) dcdiff = get_bits(gb, 10);
2441 else if (v->pq == 2) dcdiff = get_bits(gb, 9);
2442 else dcdiff = get_bits(gb, 8);
2447 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2448 else if (v->pq == 2)
2449 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2456 dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);
2459 /* Store the quantized DC coeff, used for prediction */
2461 block[0] = dcdiff * s->y_dc_scale;
2463 block[0] = dcdiff * s->c_dc_scale;
2476 int last = 0, skip, value;
2477 const int8_t *zz_table;
2481 scale = v->pq * 2 + v->halfpq;
2485 zz_table = wmv1_scantable[2];
2487 zz_table = wmv1_scantable[3];
2489 zz_table = wmv1_scantable[1];
2491 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2493 if(dc_pred_dir) //left
2496 ac_val -= 16 * s->block_wrap[n];
2499 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2503 block[zz_table[i++]] = value;
2506 /* apply AC prediction if needed */
2508 if(dc_pred_dir) { //left
2509 for(k = 1; k < 8; k++)
2510 block[k << 3] += ac_val[k];
2512 for(k = 1; k < 8; k++)
2513 block[k] += ac_val[k + 8];
2516 /* save AC coeffs for further prediction */
2517 for(k = 1; k < 8; k++) {
2518 ac_val2[k] = block[k << 3];
2519 ac_val2[k + 8] = block[k];
2522 /* scale AC coeffs */
2523 for(k = 1; k < 64; k++)
2527 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2530 if(s->ac_pred) i = 63;
2536 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2539 scale = v->pq * 2 + v->halfpq;
2540 memset(ac_val2, 0, 16 * 2);
2541 if(dc_pred_dir) {//left
2544 memcpy(ac_val2, ac_val, 8 * 2);
2546 ac_val -= 16 * s->block_wrap[n];
2548 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2551 /* apply AC prediction if needed */
2553 if(dc_pred_dir) { //left
2554 for(k = 1; k < 8; k++) {
2555 block[k << 3] = ac_val[k] * scale;
2556 if(!v->pquantizer && block[k << 3])
2557 block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;
2560 for(k = 1; k < 8; k++) {
2561 block[k] = ac_val[k + 8] * scale;
2562 if(!v->pquantizer && block[k])
2563 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2569 s->block_last_index[n] = i;
2574 /** Decode intra block in intra frames - should be faster than decode_intra_block
2575 * @param v VC1Context
2576 * @param block block to decode
2577 * @param coded are AC coeffs present or not
2578 * @param codingset set of VLC to decode data
2580 static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant)
2582 GetBitContext *gb = &v->s.gb;
2583 MpegEncContext *s = &v->s;
2584 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2587 int16_t *ac_val, *ac_val2;
2589 int a_avail = v->a_avail, c_avail = v->c_avail;
2590 int use_pred = s->ac_pred;
2593 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2595 /* Get DC differential */
2597 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2599 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2602 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2607 if (dcdiff == 119 /* ESC index value */)
2609 /* TODO: Optimize */
2610 if (mquant == 1) dcdiff = get_bits(gb, 10);
2611 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2612 else dcdiff = get_bits(gb, 8);
2617 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2618 else if (mquant == 2)
2619 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2626 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);
2629 /* Store the quantized DC coeff, used for prediction */
2631 block[0] = dcdiff * s->y_dc_scale;
2633 block[0] = dcdiff * s->c_dc_scale;
2642 /* check if AC is needed at all */
2643 if(!a_avail && !c_avail) use_pred = 0;
2644 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2647 scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);
2649 if(dc_pred_dir) //left
2652 ac_val -= 16 * s->block_wrap[n];
2654 q1 = s->current_picture.qscale_table[mb_pos];
2655 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2656 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2657 if(dc_pred_dir && n==1) q2 = q1;
2658 if(!dc_pred_dir && n==2) q2 = q1;
2662 int last = 0, skip, value;
2663 const int8_t *zz_table;
2668 zz_table = wmv1_scantable[2];
2670 zz_table = wmv1_scantable[3];
2672 zz_table = wmv1_scantable[1];
2675 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2679 block[zz_table[i++]] = value;
2682 /* apply AC prediction if needed */
2684 /* scale predictors if needed*/
2686 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2687 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2689 if(dc_pred_dir) { //left
2690 for(k = 1; k < 8; k++)
2691 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2693 for(k = 1; k < 8; k++)
2694 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2697 if(dc_pred_dir) { //left
2698 for(k = 1; k < 8; k++)
2699 block[k << 3] += ac_val[k];
2701 for(k = 1; k < 8; k++)
2702 block[k] += ac_val[k + 8];
2706 /* save AC coeffs for further prediction */
2707 for(k = 1; k < 8; k++) {
2708 ac_val2[k] = block[k << 3];
2709 ac_val2[k + 8] = block[k];
2712 /* scale AC coeffs */
2713 for(k = 1; k < 64; k++)
2717 block[k] += (block[k] < 0) ? -mquant : mquant;
2720 if(use_pred) i = 63;
2721 } else { // no AC coeffs
2724 memset(ac_val2, 0, 16 * 2);
2725 if(dc_pred_dir) {//left
2727 memcpy(ac_val2, ac_val, 8 * 2);
2729 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2730 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2731 for(k = 1; k < 8; k++)
2732 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2737 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2739 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2740 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2741 for(k = 1; k < 8; k++)
2742 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2747 /* apply AC prediction if needed */
2749 if(dc_pred_dir) { //left
2750 for(k = 1; k < 8; k++) {
2751 block[k << 3] = ac_val2[k] * scale;
2752 if(!v->pquantizer && block[k << 3])
2753 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2756 for(k = 1; k < 8; k++) {
2757 block[k] = ac_val2[k + 8] * scale;
2758 if(!v->pquantizer && block[k])
2759 block[k] += (block[k] < 0) ? -mquant : mquant;
2765 s->block_last_index[n] = i;
2770 /** Decode intra block in inter frames - more generic version than vc1_decode_i_block
2771 * @param v VC1Context
2772 * @param block block to decode
2773 * @param coded are AC coeffs present or not
2774 * @param mquant block quantizer
2775 * @param codingset set of VLC to decode data
2777 static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)
2779 GetBitContext *gb = &v->s.gb;
2780 MpegEncContext *s = &v->s;
2781 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2784 int16_t *ac_val, *ac_val2;
2786 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2787 int a_avail = v->a_avail, c_avail = v->c_avail;
2788 int use_pred = s->ac_pred;
2792 /* XXX: Guard against dumb values of mquant */
2793 mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );
2795 /* Set DC scale - y and c use the same */
2796 s->y_dc_scale = s->y_dc_scale_table[mquant];
2797 s->c_dc_scale = s->c_dc_scale_table[mquant];
2799 /* Get DC differential */
2801 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2803 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2806 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2811 if (dcdiff == 119 /* ESC index value */)
2813 /* TODO: Optimize */
2814 if (mquant == 1) dcdiff = get_bits(gb, 10);
2815 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2816 else dcdiff = get_bits(gb, 8);
2821 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2822 else if (mquant == 2)
2823 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2830 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);
2833 /* Store the quantized DC coeff, used for prediction */
2836 block[0] = dcdiff * s->y_dc_scale;
2838 block[0] = dcdiff * s->c_dc_scale;
2847 /* check if AC is needed at all and adjust direction if needed */
2848 if(!a_avail) dc_pred_dir = 1;
2849 if(!c_avail) dc_pred_dir = 0;
2850 if(!a_avail && !c_avail) use_pred = 0;
2851 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2854 scale = mquant * 2 + v->halfpq;
2856 if(dc_pred_dir) //left
2859 ac_val -= 16 * s->block_wrap[n];
2861 q1 = s->current_picture.qscale_table[mb_pos];
2862 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2863 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2864 if(dc_pred_dir && n==1) q2 = q1;
2865 if(!dc_pred_dir && n==2) q2 = q1;
2869 int last = 0, skip, value;
2870 const int8_t *zz_table;
2873 zz_table = wmv1_scantable[0];
2876 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2880 block[zz_table[i++]] = value;
2883 /* apply AC prediction if needed */
2885 /* scale predictors if needed*/
2887 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2888 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2890 if(dc_pred_dir) { //left
2891 for(k = 1; k < 8; k++)
2892 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2894 for(k = 1; k < 8; k++)
2895 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2898 if(dc_pred_dir) { //left
2899 for(k = 1; k < 8; k++)
2900 block[k << 3] += ac_val[k];
2902 for(k = 1; k < 8; k++)
2903 block[k] += ac_val[k + 8];
2907 /* save AC coeffs for further prediction */
2908 for(k = 1; k < 8; k++) {
2909 ac_val2[k] = block[k << 3];
2910 ac_val2[k + 8] = block[k];
2913 /* scale AC coeffs */
2914 for(k = 1; k < 64; k++)
2918 block[k] += (block[k] < 0) ? -mquant : mquant;
2921 if(use_pred) i = 63;
2922 } else { // no AC coeffs
2925 memset(ac_val2, 0, 16 * 2);
2926 if(dc_pred_dir) {//left
2928 memcpy(ac_val2, ac_val, 8 * 2);
2930 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2931 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2932 for(k = 1; k < 8; k++)
2933 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2938 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2940 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2941 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2942 for(k = 1; k < 8; k++)
2943 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2948 /* apply AC prediction if needed */
2950 if(dc_pred_dir) { //left
2951 for(k = 1; k < 8; k++) {
2952 block[k << 3] = ac_val2[k] * scale;
2953 if(!v->pquantizer && block[k << 3])
2954 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2957 for(k = 1; k < 8; k++) {
2958 block[k] = ac_val2[k + 8] * scale;
2959 if(!v->pquantizer && block[k])
2960 block[k] += (block[k] < 0) ? -mquant : mquant;
2966 s->block_last_index[n] = i;
2973 static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,
2974 uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left)
2976 MpegEncContext *s = &v->s;
2977 GetBitContext *gb = &s->gb;
2980 int scale, off, idx, last, skip, value;
2981 int ttblk = ttmb & 7;
2985 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)];
2987 if(ttblk == TT_4X4) {
2988 subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);
2990 if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {
2991 subblkpat = decode012(gb);
2992 if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits
2993 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;
2994 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;
2996 scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);
2998 // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT
2999 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {
3000 subblkpat = 2 - (ttblk == TT_8X4_TOP);
3003 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
3004 subblkpat = 2 - (ttblk == TT_4X8_LEFT);
3013 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3017 idx = wmv1_scantable[0][i++];
3018 block[idx] = value * scale;
3020 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3023 s->dsp.vc1_inv_trans_8x8(block);
3024 s->dsp.add_pixels_clamped(block, dst, linesize);
3025 if(apply_filter && cbp_top & 0xC)
3026 vc1_loop_filter(dst, 1, linesize, 8, mquant);
3027 if(apply_filter && cbp_left & 0xA)
3028 vc1_loop_filter(dst, linesize, 1, 8, mquant);
3032 pat = ~subblkpat & 0xF;
3033 for(j = 0; j < 4; j++) {
3034 last = subblkpat & (1 << (3 - j));
3036 off = (j & 1) * 4 + (j & 2) * 16;
3038 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3042 idx = ff_vc1_simple_progressive_4x4_zz[i++];
3043 block[idx + off] = value * scale;
3045 block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;
3047 if(!(subblkpat & (1 << (3 - j))) && !skip_block){
3048 s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
3049 if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))
3050 vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, 1, linesize, 4, mquant);
3051 if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))
3052 vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, linesize, 1, 4, mquant);
3057 pat = ~((subblkpat & 2)*6 + (subblkpat & 1)*3) & 0xF;
3058 for(j = 0; j < 2; j++) {
3059 last = subblkpat & (1 << (1 - j));
3063 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3067 idx = v->zz_8x4[i++]+off;
3068 block[idx] = value * scale;
3070 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3072 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3073 s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
3074 if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))
3075 vc1_loop_filter(dst + j*4*linesize, 1, linesize, 8, mquant);
3076 if(apply_filter && cbp_left & (2 << j))
3077 vc1_loop_filter(dst + j*4*linesize, linesize, 1, 4, mquant);
3082 pat = ~(subblkpat*5) & 0xF;
3083 for(j = 0; j < 2; j++) {
3084 last = subblkpat & (1 << (1 - j));
3088 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3092 idx = v->zz_4x8[i++]+off;
3093 block[idx] = value * scale;
3095 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3097 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3098 s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
3099 if(apply_filter && cbp_top & (2 << j))
3100 vc1_loop_filter(dst + j*4, 1, linesize, 4, mquant);
3101 if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))
3102 vc1_loop_filter(dst + j*4, linesize, 1, 8, mquant);
3111 /** Decode one P-frame MB (in Simple/Main profile)
3113 static int vc1_decode_p_mb(VC1Context *v)
3115 MpegEncContext *s = &v->s;
3116 GetBitContext *gb = &s->gb;
3118 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3119 int cbp; /* cbp decoding stuff */
3120 int mqdiff, mquant; /* MB quantization */
3121 int ttmb = v->ttfrm; /* MB Transform type */
3123 static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3124 offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3125 int mb_has_coeffs = 1; /* last_flag */
3126 int dmv_x, dmv_y; /* Differential MV components */
3127 int index, index1; /* LUT indexes */
3128 int val, sign; /* temp values */
3129 int first_block = 1;
3131 int skipped, fourmv;
3132 int block_cbp = 0, pat;
3133 int apply_loop_filter;
3135 mquant = v->pq; /* Loosy initialization */
3137 if (v->mv_type_is_raw)
3138 fourmv = get_bits1(gb);
3140 fourmv = v->mv_type_mb_plane[mb_pos];
3142 skipped = get_bits1(gb);
3144 skipped = v->s.mbskip_table[mb_pos];
3146 s->dsp.clear_blocks(s->block[0]);
3148 apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);
3149 if (!fourmv) /* 1MV mode */
3153 GET_MVDATA(dmv_x, dmv_y);
3156 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3157 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3159 s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
3160 vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
3162 /* FIXME Set DC val for inter block ? */
3163 if (s->mb_intra && !mb_has_coeffs)
3166 s->ac_pred = get_bits1(gb);
3169 else if (mb_has_coeffs)
3171 if (s->mb_intra) s->ac_pred = get_bits1(gb);
3172 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3180 s->current_picture.qscale_table[mb_pos] = mquant;
3182 if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3183 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,
3184 VC1_TTMB_VLC_BITS, 2);
3185 if(!s->mb_intra) vc1_mc_1mv(v, 0);
3189 s->dc_val[0][s->block_index[i]] = 0;
3191 val = ((cbp >> (5 - i)) & 1);
3192 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3193 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3195 /* check if prediction blocks A and C are available */
3196 v->a_avail = v->c_avail = 0;
3197 if(i == 2 || i == 3 || !s->first_slice_line)
3198 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3199 if(i == 1 || i == 3 || s->mb_x)
3200 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3202 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3203 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3204 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3205 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3206 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3207 if(v->pq >= 9 && v->overlap) {
3209 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3211 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3213 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3214 int left_cbp, top_cbp;
3216 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3217 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3219 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3220 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3223 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3225 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3227 block_cbp |= 0xF << (i << 2);
3229 int left_cbp = 0, top_cbp = 0, filter = 0;
3230 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3233 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3234 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3236 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3237 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3240 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3242 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3244 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);
3245 block_cbp |= pat << (i << 2);
3246 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3254 for(i = 0; i < 6; i++) {
3255 v->mb_type[0][s->block_index[i]] = 0;
3256 s->dc_val[0][s->block_index[i]] = 0;
3258 s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;
3259 s->current_picture.qscale_table[mb_pos] = 0;
3260 vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
3267 if (!skipped /* unskipped MB */)
3269 int intra_count = 0, coded_inter = 0;
3270 int is_intra[6], is_coded[6];
3272 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3275 val = ((cbp >> (5 - i)) & 1);
3276 s->dc_val[0][s->block_index[i]] = 0;
3283 GET_MVDATA(dmv_x, dmv_y);
3285 vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
3286 if(!s->mb_intra) vc1_mc_4mv_luma(v, i);
3287 intra_count += s->mb_intra;
3288 is_intra[i] = s->mb_intra;
3289 is_coded[i] = mb_has_coeffs;
3292 is_intra[i] = (intra_count >= 3);
3295 if(i == 4) vc1_mc_4mv_chroma(v);
3296 v->mb_type[0][s->block_index[i]] = is_intra[i];
3297 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];
3299 // if there are no coded blocks then don't do anything more
3300 if(!intra_count && !coded_inter) return 0;
3303 s->current_picture.qscale_table[mb_pos] = mquant;
3304 /* test if block is intra and has pred */
3309 if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])
3310 || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {
3315 if(intrapred)s->ac_pred = get_bits1(gb);
3316 else s->ac_pred = 0;
3318 if (!v->ttmbf && coded_inter)
3319 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3323 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3324 s->mb_intra = is_intra[i];
3326 /* check if prediction blocks A and C are available */
3327 v->a_avail = v->c_avail = 0;
3328 if(i == 2 || i == 3 || !s->first_slice_line)
3329 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3330 if(i == 1 || i == 3 || s->mb_x)
3331 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3333 vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);
3334 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3335 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3336 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3337 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
3338 if(v->pq >= 9 && v->overlap) {
3340 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3342 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3344 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)){
3345 int left_cbp, top_cbp;
3347 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3348 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3350 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3351 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3354 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3356 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3358 block_cbp |= 0xF << (i << 2);
3359 } else if(is_coded[i]) {
3360 int left_cbp = 0, top_cbp = 0, filter = 0;
3361 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)){
3364 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3365 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3367 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3368 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3371 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3373 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3375 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);
3376 block_cbp |= pat << (i << 2);
3377 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3386 s->current_picture.qscale_table[mb_pos] = 0;
3387 for (i=0; i<6; i++) {
3388 v->mb_type[0][s->block_index[i]] = 0;
3389 s->dc_val[0][s->block_index[i]] = 0;
3393 vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
3394 vc1_mc_4mv_luma(v, i);
3396 vc1_mc_4mv_chroma(v);
3397 s->current_picture.qscale_table[mb_pos] = 0;
3401 v->cbp[s->mb_x] = block_cbp;
3403 /* Should never happen */
3407 /** Decode one B-frame MB (in Main profile)
3409 static void vc1_decode_b_mb(VC1Context *v)
3411 MpegEncContext *s = &v->s;
3412 GetBitContext *gb = &s->gb;
3414 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3415 int cbp = 0; /* cbp decoding stuff */
3416 int mqdiff, mquant; /* MB quantization */
3417 int ttmb = v->ttfrm; /* MB Transform type */
3419 static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3420 offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3421 int mb_has_coeffs = 0; /* last_flag */
3422 int index, index1; /* LUT indexes */
3423 int val, sign; /* temp values */
3424 int first_block = 1;
3426 int skipped, direct;
3427 int dmv_x[2], dmv_y[2];
3428 int bmvtype = BMV_TYPE_BACKWARD;
3430 mquant = v->pq; /* Loosy initialization */
3434 direct = get_bits1(gb);
3436 direct = v->direct_mb_plane[mb_pos];
3438 skipped = get_bits1(gb);
3440 skipped = v->s.mbskip_table[mb_pos];
3442 s->dsp.clear_blocks(s->block[0]);
3443 dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
3444 for(i = 0; i < 6; i++) {
3445 v->mb_type[0][s->block_index[i]] = 0;
3446 s->dc_val[0][s->block_index[i]] = 0;
3448 s->current_picture.qscale_table[mb_pos] = 0;
3452 GET_MVDATA(dmv_x[0], dmv_y[0]);
3453 dmv_x[1] = dmv_x[0];
3454 dmv_y[1] = dmv_y[0];
3456 if(skipped || !s->mb_intra) {
3457 bmvtype = decode012(gb);
3460 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
3463 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
3466 bmvtype = BMV_TYPE_INTERPOLATED;
3467 dmv_x[0] = dmv_y[0] = 0;
3471 for(i = 0; i < 6; i++)
3472 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3475 if(direct) bmvtype = BMV_TYPE_INTERPOLATED;
3476 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3477 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3481 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3485 s->current_picture.qscale_table[mb_pos] = mquant;
3487 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3488 dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;
3489 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3490 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3492 if(!mb_has_coeffs && !s->mb_intra) {
3493 /* no coded blocks - effectively skipped */
3494 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3495 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3498 if(s->mb_intra && !mb_has_coeffs) {
3500 s->current_picture.qscale_table[mb_pos] = mquant;
3501 s->ac_pred = get_bits1(gb);
3503 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3505 if(bmvtype == BMV_TYPE_INTERPOLATED) {
3506 GET_MVDATA(dmv_x[0], dmv_y[0]);
3507 if(!mb_has_coeffs) {
3508 /* interpolated skipped block */
3509 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3510 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3514 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3516 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3519 s->ac_pred = get_bits1(gb);
3520 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3522 s->current_picture.qscale_table[mb_pos] = mquant;
3523 if(!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3524 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3530 s->dc_val[0][s->block_index[i]] = 0;
3532 val = ((cbp >> (5 - i)) & 1);
3533 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3534 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3536 /* check if prediction blocks A and C are available */
3537 v->a_avail = v->c_avail = 0;
3538 if(i == 2 || i == 3 || !s->first_slice_line)
3539 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3540 if(i == 1 || i == 3 || s->mb_x)
3541 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3543 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3544 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3545 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3546 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3547 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3549 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);
3550 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3556 /** Decode blocks of I-frame
3558 static void vc1_decode_i_blocks(VC1Context *v)
3561 MpegEncContext *s = &v->s;
3566 /* select codingmode used for VLC tables selection */
3567 switch(v->y_ac_table_index){
3569 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3572 v->codingset = CS_HIGH_MOT_INTRA;
3575 v->codingset = CS_MID_RATE_INTRA;
3579 switch(v->c_ac_table_index){
3581 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3584 v->codingset2 = CS_HIGH_MOT_INTER;
3587 v->codingset2 = CS_MID_RATE_INTER;
3591 /* Set DC scale - y and c use the same */
3592 s->y_dc_scale = s->y_dc_scale_table[v->pq];
3593 s->c_dc_scale = s->c_dc_scale_table[v->pq];
3596 s->mb_x = s->mb_y = 0;
3598 s->first_slice_line = 1;
3599 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3600 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3601 ff_init_block_index(s);
3602 ff_update_block_index(s);
3603 s->dsp.clear_blocks(s->block[0]);
3604 mb_pos = s->mb_x + s->mb_y * s->mb_width;
3605 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3606 s->current_picture.qscale_table[mb_pos] = v->pq;
3607 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3608 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3610 // do actual MB decoding and displaying
3611 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3612 v->s.ac_pred = get_bits1(&v->s.gb);
3614 for(k = 0; k < 6; k++) {
3615 val = ((cbp >> (5 - k)) & 1);
3618 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3622 cbp |= val << (5 - k);
3624 vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);
3626 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3627 if(v->pq >= 9 && v->overlap) {
3628 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3632 vc1_put_block(v, s->block);
3633 if(v->pq >= 9 && v->overlap) {
3635 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3636 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3637 if(!(s->flags & CODEC_FLAG_GRAY)) {
3638 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3639 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3642 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3643 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3644 if(!s->first_slice_line) {
3645 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3646 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3647 if(!(s->flags & CODEC_FLAG_GRAY)) {
3648 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3649 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3652 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3653 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3655 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3657 if(get_bits_count(&s->gb) > v->bits) {
3658 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3659 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3663 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3664 s->first_slice_line = 0;
3666 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3669 /** Decode blocks of I-frame for advanced profile
3671 static void vc1_decode_i_blocks_adv(VC1Context *v)
3674 MpegEncContext *s = &v->s;
3681 GetBitContext *gb = &s->gb;
3683 /* select codingmode used for VLC tables selection */
3684 switch(v->y_ac_table_index){
3686 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3689 v->codingset = CS_HIGH_MOT_INTRA;
3692 v->codingset = CS_MID_RATE_INTRA;
3696 switch(v->c_ac_table_index){
3698 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3701 v->codingset2 = CS_HIGH_MOT_INTER;
3704 v->codingset2 = CS_MID_RATE_INTER;
3709 s->mb_x = s->mb_y = 0;
3711 s->first_slice_line = 1;
3712 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3713 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3714 ff_init_block_index(s);
3715 ff_update_block_index(s);
3716 s->dsp.clear_blocks(s->block[0]);
3717 mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3718 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3719 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3720 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3722 // do actual MB decoding and displaying
3723 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3724 if(v->acpred_is_raw)
3725 v->s.ac_pred = get_bits1(&v->s.gb);
3727 v->s.ac_pred = v->acpred_plane[mb_pos];
3729 if(v->condover == CONDOVER_SELECT) {
3730 if(v->overflg_is_raw)
3731 overlap = get_bits1(&v->s.gb);
3733 overlap = v->over_flags_plane[mb_pos];
3735 overlap = (v->condover == CONDOVER_ALL);
3739 s->current_picture.qscale_table[mb_pos] = mquant;
3740 /* Set DC scale - y and c use the same */
3741 s->y_dc_scale = s->y_dc_scale_table[mquant];
3742 s->c_dc_scale = s->c_dc_scale_table[mquant];
3744 for(k = 0; k < 6; k++) {
3745 val = ((cbp >> (5 - k)) & 1);
3748 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3752 cbp |= val << (5 - k);
3754 v->a_avail = !s->first_slice_line || (k==2 || k==3);
3755 v->c_avail = !!s->mb_x || (k==1 || k==3);
3757 vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);
3759 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3760 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3763 vc1_put_block(v, s->block);
3766 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3767 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3768 if(!(s->flags & CODEC_FLAG_GRAY)) {
3769 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3770 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3773 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3774 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3775 if(!s->first_slice_line) {
3776 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3777 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3778 if(!(s->flags & CODEC_FLAG_GRAY)) {
3779 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3780 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3783 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3784 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3786 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3788 if(get_bits_count(&s->gb) > v->bits) {
3789 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3790 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3794 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3795 s->first_slice_line = 0;
3797 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3800 static void vc1_decode_p_blocks(VC1Context *v)
3802 MpegEncContext *s = &v->s;
3804 /* select codingmode used for VLC tables selection */
3805 switch(v->c_ac_table_index){
3807 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3810 v->codingset = CS_HIGH_MOT_INTRA;
3813 v->codingset = CS_MID_RATE_INTRA;
3817 switch(v->c_ac_table_index){
3819 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3822 v->codingset2 = CS_HIGH_MOT_INTER;
3825 v->codingset2 = CS_MID_RATE_INTER;
3829 s->first_slice_line = 1;
3830 memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);
3831 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3832 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3833 ff_init_block_index(s);
3834 ff_update_block_index(s);
3835 s->dsp.clear_blocks(s->block[0]);
3838 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3839 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3840 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);
3844 memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride);
3845 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3846 s->first_slice_line = 0;
3848 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3851 static void vc1_decode_b_blocks(VC1Context *v)
3853 MpegEncContext *s = &v->s;
3855 /* select codingmode used for VLC tables selection */
3856 switch(v->c_ac_table_index){
3858 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3861 v->codingset = CS_HIGH_MOT_INTRA;
3864 v->codingset = CS_MID_RATE_INTRA;
3868 switch(v->c_ac_table_index){
3870 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3873 v->codingset2 = CS_HIGH_MOT_INTER;
3876 v->codingset2 = CS_MID_RATE_INTER;
3880 s->first_slice_line = 1;
3881 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3882 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3883 ff_init_block_index(s);
3884 ff_update_block_index(s);
3885 s->dsp.clear_blocks(s->block[0]);
3888 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3889 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3890 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);
3893 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[s->mb_x + s->mb_y *s->mb_stride]);
3895 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3896 s->first_slice_line = 0;
3898 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3901 static void vc1_decode_skip_blocks(VC1Context *v)
3903 MpegEncContext *s = &v->s;
3905 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3906 s->first_slice_line = 1;
3907 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3909 ff_init_block_index(s);
3910 ff_update_block_index(s);
3911 memcpy(s->dest[0], s->last_picture.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16);
3912 memcpy(s->dest[1], s->last_picture.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3913 memcpy(s->dest[2], s->last_picture.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3914 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3915 s->first_slice_line = 0;
3917 s->pict_type = FF_P_TYPE;
3920 static void vc1_decode_blocks(VC1Context *v)
3923 v->s.esc3_level_length = 0;
3925 ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );
3928 switch(v->s.pict_type) {
3930 if(v->profile == PROFILE_ADVANCED)
3931 vc1_decode_i_blocks_adv(v);
3933 vc1_decode_i_blocks(v);
3936 if(v->p_frame_skipped)
3937 vc1_decode_skip_blocks(v);
3939 vc1_decode_p_blocks(v);
3943 if(v->profile == PROFILE_ADVANCED)
3944 vc1_decode_i_blocks_adv(v);
3946 vc1_decode_i_blocks(v);
3948 vc1_decode_b_blocks(v);
3954 /** Find VC-1 marker in buffer
3955 * @return position where next marker starts or end of buffer if no marker found
3957 static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)
3959 uint32_t mrk = 0xFFFFFFFF;
3961 if(end-src < 4) return end;
3963 mrk = (mrk << 8) | *src++;
3970 static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)
3975 for(dsize = 0; dsize < size; dsize++) *dst++ = *src++;
3978 for(i = 0; i < size; i++, src++) {
3979 if(src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {
3980 dst[dsize++] = src[1];
3984 dst[dsize++] = *src;
3989 /** Initialize a VC1/WMV3 decoder
3990 * @todo TODO: Handle VC-1 IDUs (Transport level?)
3991 * @todo TODO: Decypher remaining bits in extra_data
3993 static av_cold int vc1_decode_init(AVCodecContext *avctx)
3995 VC1Context *v = avctx->priv_data;
3996 MpegEncContext *s = &v->s;
3999 if (!avctx->extradata_size || !avctx->extradata) return -1;
4000 if (!(avctx->flags & CODEC_FLAG_GRAY))
4001 avctx->pix_fmt = PIX_FMT_YUV420P;
4003 avctx->pix_fmt = PIX_FMT_GRAY8;
4005 avctx->flags |= CODEC_FLAG_EMU_EDGE;
4006 v->s.flags |= CODEC_FLAG_EMU_EDGE;
4008 if(avctx->idct_algo==FF_IDCT_AUTO){
4009 avctx->idct_algo=FF_IDCT_WMV2;
4012 if(ff_h263_decode_init(avctx) < 0)
4014 if (vc1_init_common(v) < 0) return -1;
4016 avctx->coded_width = avctx->width;
4017 avctx->coded_height = avctx->height;
4018 if (avctx->codec_id == CODEC_ID_WMV3)
4022 // looks like WMV3 has a sequence header stored in the extradata
4023 // advanced sequence header may be before the first frame
4024 // the last byte of the extradata is a version number, 1 for the
4025 // samples we can decode
4027 init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);
4029 if (decode_sequence_header(avctx, &gb) < 0)
4032 count = avctx->extradata_size*8 - get_bits_count(&gb);
4035 av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",
4036 count, get_bits(&gb, count));
4040 av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);
4042 } else { // VC1/WVC1
4043 const uint8_t *start = avctx->extradata;
4044 uint8_t *end = avctx->extradata + avctx->extradata_size;
4045 const uint8_t *next;
4046 int size, buf2_size;
4047 uint8_t *buf2 = NULL;
4048 int seq_initialized = 0, ep_initialized = 0;
4050 if(avctx->extradata_size < 16) {
4051 av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);
4055 buf2 = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
4056 if(start[0]) start++; // in WVC1 extradata first byte is its size
4058 for(; next < end; start = next){
4059 next = find_next_marker(start + 4, end);
4060 size = next - start - 4;
4061 if(size <= 0) continue;
4062 buf2_size = vc1_unescape_buffer(start + 4, size, buf2);
4063 init_get_bits(&gb, buf2, buf2_size * 8);
4064 switch(AV_RB32(start)){
4065 case VC1_CODE_SEQHDR:
4066 if(decode_sequence_header(avctx, &gb) < 0){
4070 seq_initialized = 1;
4072 case VC1_CODE_ENTRYPOINT:
4073 if(decode_entry_point(avctx, &gb) < 0){
4082 if(!seq_initialized || !ep_initialized){
4083 av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");
4087 avctx->has_b_frames= !!(avctx->max_b_frames);
4088 s->low_delay = !avctx->has_b_frames;
4090 s->mb_width = (avctx->coded_width+15)>>4;
4091 s->mb_height = (avctx->coded_height+15)>>4;
4093 /* Allocate mb bitplanes */
4094 v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4095 v->direct_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4096 v->acpred_plane = av_malloc(s->mb_stride * s->mb_height);
4097 v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height);
4099 v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);
4100 v->cbp = v->cbp_base + s->mb_stride;
4102 /* allocate block type info in that way so it could be used with s->block_index[] */
4103 v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);
4104 v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;
4105 v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;
4106 v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);
4108 /* Init coded blocks info */
4109 if (v->profile == PROFILE_ADVANCED)
4111 // if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
4113 // if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
4117 ff_intrax8_common_init(&v->x8,s);
4122 /** Decode a VC1/WMV3 frame
4123 * @todo TODO: Handle VC-1 IDUs (Transport level?)
4125 static int vc1_decode_frame(AVCodecContext *avctx,
4126 void *data, int *data_size,
4127 const uint8_t *buf, int buf_size)
4129 VC1Context *v = avctx->priv_data;
4130 MpegEncContext *s = &v->s;
4131 AVFrame *pict = data;
4132 uint8_t *buf2 = NULL;
4134 /* no supplementary picture */
4135 if (buf_size == 0) {
4136 /* special case for last picture */
4137 if (s->low_delay==0 && s->next_picture_ptr) {
4138 *pict= *(AVFrame*)s->next_picture_ptr;
4139 s->next_picture_ptr= NULL;
4141 *data_size = sizeof(AVFrame);
4147 /* We need to set current_picture_ptr before reading the header,
4148 * otherwise we cannot store anything in there. */
4149 if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){
4150 int i= ff_find_unused_picture(s, 0);
4151 s->current_picture_ptr= &s->picture[i];
4154 //for advanced profile we may need to parse and unescape data
4155 if (avctx->codec_id == CODEC_ID_VC1) {
4157 buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
4159 if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */
4160 const uint8_t *start, *end, *next;
4164 for(start = buf, end = buf + buf_size; next < end; start = next){
4165 next = find_next_marker(start + 4, end);
4166 size = next - start - 4;
4167 if(size <= 0) continue;
4168 switch(AV_RB32(start)){
4169 case VC1_CODE_FRAME:
4170 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4172 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */
4173 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4174 init_get_bits(&s->gb, buf2, buf_size2*8);
4175 decode_entry_point(avctx, &s->gb);
4177 case VC1_CODE_SLICE:
4178 av_log(avctx, AV_LOG_ERROR, "Sliced decoding is not implemented (yet)\n");
4183 }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */
4184 const uint8_t *divider;
4186 divider = find_next_marker(buf, buf + buf_size);
4187 if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){
4188 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");
4193 buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);
4195 av_free(buf2);return -1;
4197 buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);
4199 init_get_bits(&s->gb, buf2, buf_size2*8);
4201 init_get_bits(&s->gb, buf, buf_size*8);
4202 // do parse frame header
4203 if(v->profile < PROFILE_ADVANCED) {
4204 if(vc1_parse_frame_header(v, &s->gb) == -1) {
4209 if(vc1_parse_frame_header_adv(v, &s->gb) == -1) {
4215 if(s->pict_type != FF_I_TYPE && !v->res_rtm_flag){
4221 s->current_picture.pict_type= s->pict_type;
4222 s->current_picture.key_frame= s->pict_type == FF_I_TYPE;
4224 /* skip B-frames if we don't have reference frames */
4225 if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){
4227 return -1;//buf_size;
4229 /* skip b frames if we are in a hurry */
4230 if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;
4231 if( (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
4232 || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
4233 || avctx->skip_frame >= AVDISCARD_ALL) {
4237 /* skip everything if we are in a hurry>=5 */
4238 if(avctx->hurry_up>=5) {
4240 return -1;//buf_size;
4243 if(s->next_p_frame_damaged){
4244 if(s->pict_type==FF_B_TYPE)
4247 s->next_p_frame_damaged=0;
4250 if(MPV_frame_start(s, avctx) < 0) {
4255 s->me.qpel_put= s->dsp.put_qpel_pixels_tab;
4256 s->me.qpel_avg= s->dsp.avg_qpel_pixels_tab;
4258 ff_er_frame_start(s);
4260 v->bits = buf_size * 8;
4261 vc1_decode_blocks(v);
4262 //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);
4263 // if(get_bits_count(&s->gb) > buf_size * 8)
4269 assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);
4270 assert(s->current_picture.pict_type == s->pict_type);
4271 if (s->pict_type == FF_B_TYPE || s->low_delay) {
4272 *pict= *(AVFrame*)s->current_picture_ptr;
4273 } else if (s->last_picture_ptr != NULL) {
4274 *pict= *(AVFrame*)s->last_picture_ptr;
4277 if(s->last_picture_ptr || s->low_delay){
4278 *data_size = sizeof(AVFrame);
4279 ff_print_debug_info(s, pict);
4282 /* Return the Picture timestamp as the frame number */
4283 /* we subtract 1 because it is added on utils.c */
4284 avctx->frame_number = s->picture_number - 1;
4291 /** Close a VC1/WMV3 decoder
4292 * @warning Initial try at using MpegEncContext stuff
4294 static av_cold int vc1_decode_end(AVCodecContext *avctx)
4296 VC1Context *v = avctx->priv_data;
4298 av_freep(&v->hrd_rate);
4299 av_freep(&v->hrd_buffer);
4300 MPV_common_end(&v->s);
4301 av_freep(&v->mv_type_mb_plane);
4302 av_freep(&v->direct_mb_plane);
4303 av_freep(&v->acpred_plane);
4304 av_freep(&v->over_flags_plane);
4305 av_freep(&v->mb_type_base);
4306 av_freep(&v->cbp_base);
4307 ff_intrax8_common_end(&v->x8);
4312 AVCodec vc1_decoder = {
4323 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),
4326 AVCodec wmv3_decoder = {
4337 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),