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"
38 #include "vdpau_internal.h"
43 #define MB_INTRA_VLC_BITS 9
46 static const uint16_t table_mb_intra[64][2];
50 * Init VC-1 specific tables and VC1Context members
51 * @param v The VC1Context to initialize
54 static int vc1_init_common(VC1Context *v)
59 v->hrd_rate = v->hrd_buffer = NULL;
65 init_vlc(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,
66 ff_vc1_bfraction_bits, 1, 1,
67 ff_vc1_bfraction_codes, 1, 1, 1);
68 init_vlc(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,
69 ff_vc1_norm2_bits, 1, 1,
70 ff_vc1_norm2_codes, 1, 1, 1);
71 init_vlc(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,
72 ff_vc1_norm6_bits, 1, 1,
73 ff_vc1_norm6_codes, 2, 2, 1);
74 init_vlc(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,
75 ff_vc1_imode_bits, 1, 1,
76 ff_vc1_imode_codes, 1, 1, 1);
79 init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,
80 ff_vc1_ttmb_bits[i], 1, 1,
81 ff_vc1_ttmb_codes[i], 2, 2, 1);
82 init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,
83 ff_vc1_ttblk_bits[i], 1, 1,
84 ff_vc1_ttblk_codes[i], 1, 1, 1);
85 init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,
86 ff_vc1_subblkpat_bits[i], 1, 1,
87 ff_vc1_subblkpat_codes[i], 1, 1, 1);
91 init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,
92 ff_vc1_4mv_block_pattern_bits[i], 1, 1,
93 ff_vc1_4mv_block_pattern_codes[i], 1, 1, 1);
94 init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,
95 ff_vc1_cbpcy_p_bits[i], 1, 1,
96 ff_vc1_cbpcy_p_codes[i], 2, 2, 1);
97 init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,
98 ff_vc1_mv_diff_bits[i], 1, 1,
99 ff_vc1_mv_diff_codes[i], 2, 2, 1);
102 init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],
103 &vc1_ac_tables[i][0][1], 8, 4,
104 &vc1_ac_tables[i][0][0], 8, 4, 1);
105 init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,
106 &ff_msmp4_mb_i_table[0][1], 4, 2,
107 &ff_msmp4_mb_i_table[0][0], 4, 2, 1);
112 v->mvrange = 0; /* 7.1.1.18, p80 */
117 /***********************************************************************/
119 * @defgroup bitplane VC9 Bitplane decoding
124 /** @addtogroup bitplane
137 /** @} */ //imode defines
139 /** Decode rows by checking if they are skipped
140 * @param plane Buffer to store decoded bits
141 * @param[in] width Width of this buffer
142 * @param[in] height Height of this buffer
143 * @param[in] stride of this buffer
145 static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
148 for (y=0; y<height; y++){
149 if (!get_bits1(gb)) //rowskip
150 memset(plane, 0, width);
152 for (x=0; x<width; x++)
153 plane[x] = get_bits1(gb);
158 /** Decode columns by checking if they are skipped
159 * @param plane Buffer to store decoded bits
160 * @param[in] width Width of this buffer
161 * @param[in] height Height of this buffer
162 * @param[in] stride of this buffer
163 * @todo FIXME: Optimize
165 static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
168 for (x=0; x<width; x++){
169 if (!get_bits1(gb)) //colskip
170 for (y=0; y<height; y++)
173 for (y=0; y<height; y++)
174 plane[y*stride] = get_bits1(gb);
179 /** Decode a bitplane's bits
180 * @param bp Bitplane where to store the decode bits
181 * @param v VC-1 context for bit reading and logging
183 * @todo FIXME: Optimize
185 static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)
187 GetBitContext *gb = &v->s.gb;
189 int imode, x, y, code, offset;
190 uint8_t invert, *planep = data;
191 int width, height, stride;
193 width = v->s.mb_width;
194 height = v->s.mb_height;
195 stride = v->s.mb_stride;
196 invert = get_bits1(gb);
197 imode = get_vlc2(gb, ff_vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);
203 //Data is actually read in the MB layer (same for all tests == "raw")
204 *raw_flag = 1; //invert ignored
208 if ((height * width) & 1)
210 *planep++ = get_bits1(gb);
214 // decode bitplane as one long line
215 for (y = offset; y < height * width; y += 2) {
216 code = get_vlc2(gb, ff_vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
217 *planep++ = code & 1;
219 if(offset == width) {
221 planep += stride - width;
223 *planep++ = code >> 1;
225 if(offset == width) {
227 planep += stride - width;
233 if(!(height % 3) && (width % 3)) { // use 2x3 decoding
234 for(y = 0; y < height; y+= 3) {
235 for(x = width & 1; x < width; x += 2) {
236 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
238 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
241 planep[x + 0] = (code >> 0) & 1;
242 planep[x + 1] = (code >> 1) & 1;
243 planep[x + 0 + stride] = (code >> 2) & 1;
244 planep[x + 1 + stride] = (code >> 3) & 1;
245 planep[x + 0 + stride * 2] = (code >> 4) & 1;
246 planep[x + 1 + stride * 2] = (code >> 5) & 1;
248 planep += stride * 3;
250 if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);
252 planep += (height & 1) * stride;
253 for(y = height & 1; y < height; y += 2) {
254 for(x = width % 3; x < width; x += 3) {
255 code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
257 av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
260 planep[x + 0] = (code >> 0) & 1;
261 planep[x + 1] = (code >> 1) & 1;
262 planep[x + 2] = (code >> 2) & 1;
263 planep[x + 0 + stride] = (code >> 3) & 1;
264 planep[x + 1 + stride] = (code >> 4) & 1;
265 planep[x + 2 + stride] = (code >> 5) & 1;
267 planep += stride * 2;
270 if(x) decode_colskip(data , x, height , stride, &v->s.gb);
271 if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);
275 decode_rowskip(data, width, height, stride, &v->s.gb);
278 decode_colskip(data, width, height, stride, &v->s.gb);
283 /* Applying diff operator */
284 if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
288 for (x=1; x<width; x++)
289 planep[x] ^= planep[x-1];
290 for (y=1; y<height; y++)
293 planep[0] ^= planep[-stride];
294 for (x=1; x<width; x++)
296 if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;
297 else planep[x] ^= planep[x-1];
304 for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride
306 return (imode<<1) + invert;
309 /** @} */ //Bitplane group
311 #define FILTSIGN(a) ((a) >= 0 ? 1 : -1)
313 * VC-1 in-loop deblocking filter for one line
314 * @param src source block type
315 * @param pq block quantizer
316 * @return whether other 3 pairs should be filtered or not
319 static int av_always_inline vc1_filter_line(uint8_t* src, int stride, int pq){
320 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
322 int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
323 int a0_sign = a0 >> 31; /* Store sign */
324 a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */
326 int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3);
327 int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3);
328 if(a1 < a0 || a2 < a0){
329 int clip = src[-1*stride] - src[ 0*stride];
330 int clip_sign = clip >> 31;
331 clip = ((clip ^ clip_sign) - clip_sign)>>1;
333 int a3 = FFMIN(a1, a2);
334 int d = 5 * (a3 - a0);
335 int d_sign = (d >> 31);
336 d = ((d ^ d_sign) - d_sign) >> 3;
339 if( d_sign ^ clip_sign )
343 d = (d ^ d_sign) - d_sign; /* Restore sign */
344 src[-1*stride] = cm[src[-1*stride] - d];
345 src[ 0*stride] = cm[src[ 0*stride] + d];
355 * VC-1 in-loop deblocking filter
356 * @param src source block type
357 * @param len edge length to filter (4 or 8 pixels)
358 * @param pq block quantizer
361 static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
366 for(i = 0; i < len; i += 4){
367 filt3 = vc1_filter_line(src + 2*step, stride, pq);
369 vc1_filter_line(src + 0*step, stride, pq);
370 vc1_filter_line(src + 1*step, stride, pq);
371 vc1_filter_line(src + 3*step, stride, pq);
377 static void vc1_loop_filter_iblk(MpegEncContext *s, int pq)
380 if(!s->first_slice_line)
381 vc1_loop_filter(s->dest[0], 1, s->linesize, 16, pq);
382 vc1_loop_filter(s->dest[0] + 8*s->linesize, 1, s->linesize, 16, pq);
383 for(i = !s->mb_x*8; i < 16; i += 8)
384 vc1_loop_filter(s->dest[0] + i, s->linesize, 1, 16, pq);
385 for(j = 0; j < 2; j++){
386 if(!s->first_slice_line)
387 vc1_loop_filter(s->dest[j+1], 1, s->uvlinesize, 8, pq);
389 vc1_loop_filter(s->dest[j+1], s->uvlinesize, 1, 8, pq);
393 /***********************************************************************/
394 /** VOP Dquant decoding
395 * @param v VC-1 Context
397 static int vop_dquant_decoding(VC1Context *v)
399 GetBitContext *gb = &v->s.gb;
405 pqdiff = get_bits(gb, 3);
406 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
407 else v->altpq = v->pq + pqdiff + 1;
411 v->dquantfrm = get_bits1(gb);
414 v->dqprofile = get_bits(gb, 2);
415 switch (v->dqprofile)
417 case DQPROFILE_SINGLE_EDGE:
418 case DQPROFILE_DOUBLE_EDGES:
419 v->dqsbedge = get_bits(gb, 2);
421 case DQPROFILE_ALL_MBS:
422 v->dqbilevel = get_bits1(gb);
425 default: break; //Forbidden ?
427 if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)
429 pqdiff = get_bits(gb, 3);
430 if (pqdiff == 7) v->altpq = get_bits(gb, 5);
431 else v->altpq = v->pq + pqdiff + 1;
438 /** Put block onto picture
440 static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
444 DSPContext *dsp = &v->s.dsp;
448 for(k = 0; k < 6; k++)
449 for(j = 0; j < 8; j++)
450 for(i = 0; i < 8; i++)
451 block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;
454 ys = v->s.current_picture.linesize[0];
455 us = v->s.current_picture.linesize[1];
456 vs = v->s.current_picture.linesize[2];
459 dsp->put_pixels_clamped(block[0], Y, ys);
460 dsp->put_pixels_clamped(block[1], Y + 8, ys);
462 dsp->put_pixels_clamped(block[2], Y, ys);
463 dsp->put_pixels_clamped(block[3], Y + 8, ys);
465 if(!(v->s.flags & CODEC_FLAG_GRAY)) {
466 dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
467 dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
471 /** Do motion compensation over 1 macroblock
472 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
474 static void vc1_mc_1mv(VC1Context *v, int dir)
476 MpegEncContext *s = &v->s;
477 DSPContext *dsp = &v->s.dsp;
478 uint8_t *srcY, *srcU, *srcV;
479 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
481 if(!v->s.last_picture.data[0])return;
483 mx = s->mv[dir][0][0];
484 my = s->mv[dir][0][1];
486 // store motion vectors for further use in B frames
487 if(s->pict_type == FF_P_TYPE) {
488 s->current_picture.motion_val[1][s->block_index[0]][0] = mx;
489 s->current_picture.motion_val[1][s->block_index[0]][1] = my;
491 uvmx = (mx + ((mx & 3) == 3)) >> 1;
492 uvmy = (my + ((my & 3) == 3)) >> 1;
494 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
495 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
498 srcY = s->last_picture.data[0];
499 srcU = s->last_picture.data[1];
500 srcV = s->last_picture.data[2];
502 srcY = s->next_picture.data[0];
503 srcU = s->next_picture.data[1];
504 srcV = s->next_picture.data[2];
507 src_x = s->mb_x * 16 + (mx >> 2);
508 src_y = s->mb_y * 16 + (my >> 2);
509 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
510 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
512 if(v->profile != PROFILE_ADVANCED){
513 src_x = av_clip( src_x, -16, s->mb_width * 16);
514 src_y = av_clip( src_y, -16, s->mb_height * 16);
515 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
516 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
518 src_x = av_clip( src_x, -17, s->avctx->coded_width);
519 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
520 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
521 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
524 srcY += src_y * s->linesize + src_x;
525 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
526 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
528 /* for grayscale we should not try to read from unknown area */
529 if(s->flags & CODEC_FLAG_GRAY) {
530 srcU = s->edge_emu_buffer + 18 * s->linesize;
531 srcV = s->edge_emu_buffer + 18 * s->linesize;
534 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
535 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3
536 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){
537 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
539 srcY -= s->mspel * (1 + s->linesize);
540 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
541 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
542 srcY = s->edge_emu_buffer;
543 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
544 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
545 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
546 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
549 /* if we deal with range reduction we need to scale source blocks */
555 for(j = 0; j < 17 + s->mspel*2; j++) {
556 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
559 src = srcU; src2 = srcV;
560 for(j = 0; j < 9; j++) {
561 for(i = 0; i < 9; i++) {
562 src[i] = ((src[i] - 128) >> 1) + 128;
563 src2[i] = ((src2[i] - 128) >> 1) + 128;
565 src += s->uvlinesize;
566 src2 += s->uvlinesize;
569 /* if we deal with intensity compensation we need to scale source blocks */
570 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
575 for(j = 0; j < 17 + s->mspel*2; j++) {
576 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];
579 src = srcU; src2 = srcV;
580 for(j = 0; j < 9; j++) {
581 for(i = 0; i < 9; i++) {
582 src[i] = v->lutuv[src[i]];
583 src2[i] = v->lutuv[src2[i]];
585 src += s->uvlinesize;
586 src2 += s->uvlinesize;
589 srcY += s->mspel * (1 + s->linesize);
593 dxy = ((my & 3) << 2) | (mx & 3);
594 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] , srcY , s->linesize, v->rnd);
595 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
596 srcY += s->linesize * 8;
597 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize , srcY , s->linesize, v->rnd);
598 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
599 } else { // hpel mc - always used for luma
600 dxy = (my & 2) | ((mx & 2) >> 1);
603 dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
605 dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
608 if(s->flags & CODEC_FLAG_GRAY) return;
609 /* Chroma MC always uses qpel bilinear */
610 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
614 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
615 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
617 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
618 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
622 /** Do motion compensation for 4-MV macroblock - luminance block
624 static void vc1_mc_4mv_luma(VC1Context *v, int n)
626 MpegEncContext *s = &v->s;
627 DSPContext *dsp = &v->s.dsp;
629 int dxy, mx, my, src_x, src_y;
632 if(!v->s.last_picture.data[0])return;
635 srcY = s->last_picture.data[0];
637 off = s->linesize * 4 * (n&2) + (n&1) * 8;
639 src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);
640 src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);
642 if(v->profile != PROFILE_ADVANCED){
643 src_x = av_clip( src_x, -16, s->mb_width * 16);
644 src_y = av_clip( src_y, -16, s->mb_height * 16);
646 src_x = av_clip( src_x, -17, s->avctx->coded_width);
647 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
650 srcY += src_y * s->linesize + src_x;
652 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
653 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2
654 || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){
655 srcY -= s->mspel * (1 + s->linesize);
656 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,
657 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
658 srcY = s->edge_emu_buffer;
659 /* if we deal with range reduction we need to scale source blocks */
665 for(j = 0; j < 9 + s->mspel*2; j++) {
666 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
670 /* if we deal with intensity compensation we need to scale source blocks */
671 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
676 for(j = 0; j < 9 + s->mspel*2; j++) {
677 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];
681 srcY += s->mspel * (1 + s->linesize);
685 dxy = ((my & 3) << 2) | (mx & 3);
686 dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);
687 } else { // hpel mc - always used for luma
688 dxy = (my & 2) | ((mx & 2) >> 1);
690 dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
692 dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
696 static inline int median4(int a, int b, int c, int d)
699 if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
700 else return (FFMIN(b, c) + FFMAX(a, d)) / 2;
702 if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
703 else return (FFMIN(a, c) + FFMAX(b, d)) / 2;
708 /** Do motion compensation for 4-MV macroblock - both chroma blocks
710 static void vc1_mc_4mv_chroma(VC1Context *v)
712 MpegEncContext *s = &v->s;
713 DSPContext *dsp = &v->s.dsp;
714 uint8_t *srcU, *srcV;
715 int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;
716 int i, idx, tx = 0, ty = 0;
717 int mvx[4], mvy[4], intra[4];
718 static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
720 if(!v->s.last_picture.data[0])return;
721 if(s->flags & CODEC_FLAG_GRAY) return;
723 for(i = 0; i < 4; i++) {
724 mvx[i] = s->mv[0][i][0];
725 mvy[i] = s->mv[0][i][1];
726 intra[i] = v->mb_type[0][s->block_index[i]];
729 /* calculate chroma MV vector from four luma MVs */
730 idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];
731 if(!idx) { // all blocks are inter
732 tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
733 ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
734 } else if(count[idx] == 1) { // 3 inter blocks
737 tx = mid_pred(mvx[1], mvx[2], mvx[3]);
738 ty = mid_pred(mvy[1], mvy[2], mvy[3]);
741 tx = mid_pred(mvx[0], mvx[2], mvx[3]);
742 ty = mid_pred(mvy[0], mvy[2], mvy[3]);
745 tx = mid_pred(mvx[0], mvx[1], mvx[3]);
746 ty = mid_pred(mvy[0], mvy[1], mvy[3]);
749 tx = mid_pred(mvx[0], mvx[1], mvx[2]);
750 ty = mid_pred(mvy[0], mvy[1], mvy[2]);
753 } else if(count[idx] == 2) {
755 for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}
756 for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}
757 tx = (mvx[t1] + mvx[t2]) / 2;
758 ty = (mvy[t1] + mvy[t2]) / 2;
760 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
761 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
762 return; //no need to do MC for inter blocks
765 s->current_picture.motion_val[1][s->block_index[0]][0] = tx;
766 s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
767 uvmx = (tx + ((tx&3) == 3)) >> 1;
768 uvmy = (ty + ((ty&3) == 3)) >> 1;
770 uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
771 uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
774 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
775 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
777 if(v->profile != PROFILE_ADVANCED){
778 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
779 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
781 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
782 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
785 srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
786 srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
787 if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
788 || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
789 || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){
790 ff_emulated_edge_mc(s->edge_emu_buffer , srcU, s->uvlinesize, 8+1, 8+1,
791 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
792 ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,
793 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
794 srcU = s->edge_emu_buffer;
795 srcV = s->edge_emu_buffer + 16;
797 /* if we deal with range reduction we need to scale source blocks */
802 src = srcU; src2 = srcV;
803 for(j = 0; j < 9; j++) {
804 for(i = 0; i < 9; i++) {
805 src[i] = ((src[i] - 128) >> 1) + 128;
806 src2[i] = ((src2[i] - 128) >> 1) + 128;
808 src += s->uvlinesize;
809 src2 += s->uvlinesize;
812 /* if we deal with intensity compensation we need to scale source blocks */
813 if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
817 src = srcU; src2 = srcV;
818 for(j = 0; j < 9; j++) {
819 for(i = 0; i < 9; i++) {
820 src[i] = v->lutuv[src[i]];
821 src2[i] = v->lutuv[src2[i]];
823 src += s->uvlinesize;
824 src2 += s->uvlinesize;
829 /* Chroma MC always uses qpel bilinear */
830 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
834 dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
835 dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
837 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
838 dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
842 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);
845 * Decode Simple/Main Profiles sequence header
846 * @see Figure 7-8, p16-17
847 * @param avctx Codec context
848 * @param gb GetBit context initialized from Codec context extra_data
851 static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)
853 VC1Context *v = avctx->priv_data;
855 av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));
856 v->profile = get_bits(gb, 2);
857 if (v->profile == PROFILE_COMPLEX)
859 av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");
862 if (v->profile == PROFILE_ADVANCED)
864 v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz;
865 v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz;
866 return decode_sequence_header_adv(v, gb);
870 v->zz_8x4 = wmv2_scantableA;
871 v->zz_4x8 = wmv2_scantableB;
872 v->res_sm = get_bits(gb, 2); //reserved
875 av_log(avctx, AV_LOG_ERROR,
876 "Reserved RES_SM=%i is forbidden\n", v->res_sm);
882 v->frmrtq_postproc = get_bits(gb, 3); //common
883 // (bitrate-32kbps)/64kbps
884 v->bitrtq_postproc = get_bits(gb, 5); //common
885 v->s.loop_filter = get_bits1(gb); //common
886 if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)
888 av_log(avctx, AV_LOG_ERROR,
889 "LOOPFILTER shell not be enabled in simple profile\n");
891 if(v->s.avctx->skip_loop_filter >= AVDISCARD_ALL)
892 v->s.loop_filter = 0;
894 v->res_x8 = get_bits1(gb); //reserved
895 v->multires = get_bits1(gb);
896 v->res_fasttx = get_bits1(gb);
899 v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct;
900 v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;
901 v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;
902 v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;
905 v->fastuvmc = get_bits1(gb); //common
906 if (!v->profile && !v->fastuvmc)
908 av_log(avctx, AV_LOG_ERROR,
909 "FASTUVMC unavailable in Simple Profile\n");
912 v->extended_mv = get_bits1(gb); //common
913 if (!v->profile && v->extended_mv)
915 av_log(avctx, AV_LOG_ERROR,
916 "Extended MVs unavailable in Simple Profile\n");
919 v->dquant = get_bits(gb, 2); //common
920 v->vstransform = get_bits1(gb); //common
922 v->res_transtab = get_bits1(gb);
925 av_log(avctx, AV_LOG_ERROR,
926 "1 for reserved RES_TRANSTAB is forbidden\n");
930 v->overlap = get_bits1(gb); //common
932 v->s.resync_marker = get_bits1(gb);
933 v->rangered = get_bits1(gb);
934 if (v->rangered && v->profile == PROFILE_SIMPLE)
936 av_log(avctx, AV_LOG_INFO,
937 "RANGERED should be set to 0 in simple profile\n");
940 v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common
941 v->quantizer_mode = get_bits(gb, 2); //common
943 v->finterpflag = get_bits1(gb); //common
944 v->res_rtm_flag = get_bits1(gb); //reserved
945 if (!v->res_rtm_flag)
947 // av_log(avctx, AV_LOG_ERROR,
948 // "0 for reserved RES_RTM_FLAG is forbidden\n");
949 av_log(avctx, AV_LOG_ERROR,
950 "Old WMV3 version detected, only I-frames will be decoded\n");
953 //TODO: figure out what they mean (always 0x402F)
954 if(!v->res_fasttx) skip_bits(gb, 16);
955 av_log(avctx, AV_LOG_DEBUG,
956 "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
957 "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"
958 "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"
959 "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",
960 v->profile, v->frmrtq_postproc, v->bitrtq_postproc,
961 v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,
962 v->rangered, v->vstransform, v->overlap, v->s.resync_marker,
963 v->dquant, v->quantizer_mode, avctx->max_b_frames
968 static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)
971 v->level = get_bits(gb, 3);
974 av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);
976 v->chromaformat = get_bits(gb, 2);
977 if (v->chromaformat != 1)
979 av_log(v->s.avctx, AV_LOG_ERROR,
980 "Only 4:2:0 chroma format supported\n");
985 v->frmrtq_postproc = get_bits(gb, 3); //common
986 // (bitrate-32kbps)/64kbps
987 v->bitrtq_postproc = get_bits(gb, 5); //common
988 v->postprocflag = get_bits1(gb); //common
990 v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;
991 v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;
992 v->s.avctx->width = v->s.avctx->coded_width;
993 v->s.avctx->height = v->s.avctx->coded_height;
994 v->broadcast = get_bits1(gb);
995 v->interlace = get_bits1(gb);
996 v->tfcntrflag = get_bits1(gb);
997 v->finterpflag = get_bits1(gb);
998 skip_bits1(gb); // reserved
1000 v->s.h_edge_pos = v->s.avctx->coded_width;
1001 v->s.v_edge_pos = v->s.avctx->coded_height;
1003 av_log(v->s.avctx, AV_LOG_DEBUG,
1004 "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
1005 "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"
1006 "TFCTRflag=%i, FINTERPflag=%i\n",
1007 v->level, v->frmrtq_postproc, v->bitrtq_postproc,
1008 v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,
1009 v->tfcntrflag, v->finterpflag
1012 v->psf = get_bits1(gb);
1013 if(v->psf) { //PsF, 6.1.13
1014 av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");
1017 v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
1018 if(get_bits1(gb)) { //Display Info - decoding is not affected by it
1020 av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
1021 v->s.avctx->coded_width = w = get_bits(gb, 14) + 1;
1022 v->s.avctx->coded_height = h = get_bits(gb, 14) + 1;
1023 av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
1025 ar = get_bits(gb, 4);
1027 v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
1029 w = get_bits(gb, 8);
1030 h = get_bits(gb, 8);
1031 v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
1033 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);
1035 if(get_bits1(gb)){ //framerate stuff
1037 v->s.avctx->time_base.num = 32;
1038 v->s.avctx->time_base.den = get_bits(gb, 16) + 1;
1041 nr = get_bits(gb, 8);
1042 dr = get_bits(gb, 4);
1043 if(nr && nr < 8 && dr && dr < 3){
1044 v->s.avctx->time_base.num = ff_vc1_fps_dr[dr - 1];
1045 v->s.avctx->time_base.den = ff_vc1_fps_nr[nr - 1] * 1000;
1051 v->color_prim = get_bits(gb, 8);
1052 v->transfer_char = get_bits(gb, 8);
1053 v->matrix_coef = get_bits(gb, 8);
1057 v->hrd_param_flag = get_bits1(gb);
1058 if(v->hrd_param_flag) {
1060 v->hrd_num_leaky_buckets = get_bits(gb, 5);
1061 skip_bits(gb, 4); //bitrate exponent
1062 skip_bits(gb, 4); //buffer size exponent
1063 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1064 skip_bits(gb, 16); //hrd_rate[n]
1065 skip_bits(gb, 16); //hrd_buffer[n]
1071 static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb)
1073 VC1Context *v = avctx->priv_data;
1074 int i, blink, clentry;
1076 av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32));
1077 blink = get_bits1(gb); // broken link
1078 clentry = get_bits1(gb); // closed entry
1079 v->panscanflag = get_bits1(gb);
1080 v->refdist_flag = get_bits1(gb);
1081 v->s.loop_filter = get_bits1(gb);
1082 v->fastuvmc = get_bits1(gb);
1083 v->extended_mv = get_bits1(gb);
1084 v->dquant = get_bits(gb, 2);
1085 v->vstransform = get_bits1(gb);
1086 v->overlap = get_bits1(gb);
1087 v->quantizer_mode = get_bits(gb, 2);
1089 if(v->hrd_param_flag){
1090 for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1091 skip_bits(gb, 8); //hrd_full[n]
1096 avctx->coded_width = (get_bits(gb, 12)+1)<<1;
1097 avctx->coded_height = (get_bits(gb, 12)+1)<<1;
1100 v->extended_dmv = get_bits1(gb);
1101 if((v->range_mapy_flag = get_bits1(gb))) {
1102 av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n");
1103 v->range_mapy = get_bits(gb, 3);
1105 if((v->range_mapuv_flag = get_bits1(gb))) {
1106 av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n");
1107 v->range_mapuv = get_bits(gb, 3);
1110 av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n"
1111 "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n"
1112 "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n"
1113 "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n",
1114 blink, clentry, v->panscanflag, v->refdist_flag, v->s.loop_filter,
1115 v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode);
1120 static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)
1122 int pqindex, lowquant, status;
1124 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1125 skip_bits(gb, 2); //framecnt unused
1127 if (v->rangered) v->rangeredfrm = get_bits1(gb);
1128 v->s.pict_type = get_bits1(gb);
1129 if (v->s.avctx->max_b_frames) {
1130 if (!v->s.pict_type) {
1131 if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE;
1132 else v->s.pict_type = FF_B_TYPE;
1133 } else v->s.pict_type = FF_P_TYPE;
1134 } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE;
1137 if(v->s.pict_type == FF_B_TYPE) {
1138 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1139 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1140 if(v->bfraction == 0) {
1141 v->s.pict_type = FF_BI_TYPE;
1144 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1145 skip_bits(gb, 7); // skip buffer fullness
1148 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1150 if(v->s.pict_type == FF_P_TYPE)
1153 /* Quantizer stuff */
1154 pqindex = get_bits(gb, 5);
1155 if(!pqindex) return -1;
1156 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1157 v->pq = ff_vc1_pquant_table[0][pqindex];
1159 v->pq = ff_vc1_pquant_table[1][pqindex];
1162 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1163 v->pquantizer = pqindex < 9;
1164 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1166 v->pqindex = pqindex;
1167 if (pqindex < 9) v->halfpq = get_bits1(gb);
1169 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1170 v->pquantizer = get_bits1(gb);
1172 if (v->extended_mv == 1) v->mvrange = get_unary(gb, 0, 3);
1173 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1174 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1175 v->range_x = 1 << (v->k_x - 1);
1176 v->range_y = 1 << (v->k_y - 1);
1177 if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2);
1179 if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){
1180 v->x8_type = get_bits1(gb);
1181 }else v->x8_type = 0;
1182 //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",
1183 // (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);
1185 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1187 switch(v->s.pict_type) {
1189 if (v->pq < 5) v->tt_index = 0;
1190 else if(v->pq < 13) v->tt_index = 1;
1191 else v->tt_index = 2;
1193 lowquant = (v->pq > 12) ? 0 : 1;
1194 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1195 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1197 int scale, shift, i;
1198 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1199 v->lumscale = get_bits(gb, 6);
1200 v->lumshift = get_bits(gb, 6);
1202 /* fill lookup tables for intensity compensation */
1205 shift = (255 - v->lumshift * 2) << 6;
1206 if(v->lumshift > 31)
1209 scale = v->lumscale + 32;
1210 if(v->lumshift > 31)
1211 shift = (v->lumshift - 64) << 6;
1213 shift = v->lumshift << 6;
1215 for(i = 0; i < 256; i++) {
1216 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1217 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1220 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1221 v->s.quarter_sample = 0;
1222 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1223 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1224 v->s.quarter_sample = 0;
1226 v->s.quarter_sample = 1;
1228 v->s.quarter_sample = 1;
1229 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));
1231 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1232 v->mv_mode2 == MV_PMODE_MIXED_MV)
1233 || v->mv_mode == MV_PMODE_MIXED_MV)
1235 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1236 if (status < 0) return -1;
1237 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1238 "Imode: %i, Invert: %i\n", status>>1, status&1);
1240 v->mv_type_is_raw = 0;
1241 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1243 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1244 if (status < 0) return -1;
1245 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1246 "Imode: %i, Invert: %i\n", status>>1, status&1);
1248 /* Hopefully this is correct for P frames */
1249 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1250 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1254 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1255 vop_dquant_decoding(v);
1258 v->ttfrm = 0; //FIXME Is that so ?
1261 v->ttmbf = get_bits1(gb);
1264 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1272 if (v->pq < 5) v->tt_index = 0;
1273 else if(v->pq < 13) v->tt_index = 1;
1274 else v->tt_index = 2;
1276 lowquant = (v->pq > 12) ? 0 : 1;
1277 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1278 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1279 v->s.mspel = v->s.quarter_sample;
1281 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1282 if (status < 0) return -1;
1283 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1284 "Imode: %i, Invert: %i\n", status>>1, status&1);
1285 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1286 if (status < 0) return -1;
1287 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1288 "Imode: %i, Invert: %i\n", status>>1, status&1);
1290 v->s.mv_table_index = get_bits(gb, 2);
1291 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1295 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1296 vop_dquant_decoding(v);
1302 v->ttmbf = get_bits1(gb);
1305 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1317 v->c_ac_table_index = decode012(gb);
1318 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1320 v->y_ac_table_index = decode012(gb);
1323 v->s.dc_table_index = get_bits1(gb);
1326 if(v->s.pict_type == FF_BI_TYPE) {
1327 v->s.pict_type = FF_B_TYPE;
1333 static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)
1335 int pqindex, lowquant;
1338 v->p_frame_skipped = 0;
1341 v->fcm = decode012(gb);
1342 if(v->fcm) return -1; // interlaced frames/fields are not implemented
1344 switch(get_unary(gb, 0, 4)) {
1346 v->s.pict_type = FF_P_TYPE;
1349 v->s.pict_type = FF_B_TYPE;
1352 v->s.pict_type = FF_I_TYPE;
1355 v->s.pict_type = FF_BI_TYPE;
1358 v->s.pict_type = FF_P_TYPE; // skipped pic
1359 v->p_frame_skipped = 1;
1365 if(!v->interlace || v->psf) {
1366 v->rptfrm = get_bits(gb, 2);
1368 v->tff = get_bits1(gb);
1369 v->rptfrm = get_bits1(gb);
1372 if(v->panscanflag) {
1375 v->rnd = get_bits1(gb);
1377 v->uvsamp = get_bits1(gb);
1378 if(v->finterpflag) v->interpfrm = get_bits1(gb);
1379 if(v->s.pict_type == FF_B_TYPE) {
1380 v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1381 v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1382 if(v->bfraction == 0) {
1383 v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */
1386 pqindex = get_bits(gb, 5);
1387 if(!pqindex) return -1;
1388 v->pqindex = pqindex;
1389 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1390 v->pq = ff_vc1_pquant_table[0][pqindex];
1392 v->pq = ff_vc1_pquant_table[1][pqindex];
1395 if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1396 v->pquantizer = pqindex < 9;
1397 if (v->quantizer_mode == QUANT_NON_UNIFORM)
1399 v->pqindex = pqindex;
1400 if (pqindex < 9) v->halfpq = get_bits1(gb);
1402 if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1403 v->pquantizer = get_bits1(gb);
1405 v->postproc = get_bits(gb, 2);
1407 if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1409 switch(v->s.pict_type) {
1412 status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v);
1413 if (status < 0) return -1;
1414 av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: "
1415 "Imode: %i, Invert: %i\n", status>>1, status&1);
1416 v->condover = CONDOVER_NONE;
1417 if(v->overlap && v->pq <= 8) {
1418 v->condover = decode012(gb);
1419 if(v->condover == CONDOVER_SELECT) {
1420 status = bitplane_decoding(v->over_flags_plane, &v->overflg_is_raw, v);
1421 if (status < 0) return -1;
1422 av_log(v->s.avctx, AV_LOG_DEBUG, "CONDOVER plane encoding: "
1423 "Imode: %i, Invert: %i\n", status>>1, status&1);
1428 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1429 else v->mvrange = 0;
1430 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1431 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1432 v->range_x = 1 << (v->k_x - 1);
1433 v->range_y = 1 << (v->k_y - 1);
1435 if (v->pq < 5) v->tt_index = 0;
1436 else if(v->pq < 13) v->tt_index = 1;
1437 else v->tt_index = 2;
1439 lowquant = (v->pq > 12) ? 0 : 1;
1440 v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1441 if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1443 int scale, shift, i;
1444 v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1445 v->lumscale = get_bits(gb, 6);
1446 v->lumshift = get_bits(gb, 6);
1447 /* fill lookup tables for intensity compensation */
1450 shift = (255 - v->lumshift * 2) << 6;
1451 if(v->lumshift > 31)
1454 scale = v->lumscale + 32;
1455 if(v->lumshift > 31)
1456 shift = (v->lumshift - 64) << 6;
1458 shift = v->lumshift << 6;
1460 for(i = 0; i < 256; i++) {
1461 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1462 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1466 if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1467 v->s.quarter_sample = 0;
1468 else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1469 if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1470 v->s.quarter_sample = 0;
1472 v->s.quarter_sample = 1;
1474 v->s.quarter_sample = 1;
1475 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));
1477 if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1478 v->mv_mode2 == MV_PMODE_MIXED_MV)
1479 || v->mv_mode == MV_PMODE_MIXED_MV)
1481 status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1482 if (status < 0) return -1;
1483 av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1484 "Imode: %i, Invert: %i\n", status>>1, status&1);
1486 v->mv_type_is_raw = 0;
1487 memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1489 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1490 if (status < 0) return -1;
1491 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1492 "Imode: %i, Invert: %i\n", status>>1, status&1);
1494 /* Hopefully this is correct for P frames */
1495 v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1496 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1499 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1500 vop_dquant_decoding(v);
1503 v->ttfrm = 0; //FIXME Is that so ?
1506 v->ttmbf = get_bits1(gb);
1509 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1517 if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1518 else v->mvrange = 0;
1519 v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1520 v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1521 v->range_x = 1 << (v->k_x - 1);
1522 v->range_y = 1 << (v->k_y - 1);
1524 if (v->pq < 5) v->tt_index = 0;
1525 else if(v->pq < 13) v->tt_index = 1;
1526 else v->tt_index = 2;
1528 lowquant = (v->pq > 12) ? 0 : 1;
1529 v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1530 v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1531 v->s.mspel = v->s.quarter_sample;
1533 status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1534 if (status < 0) return -1;
1535 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1536 "Imode: %i, Invert: %i\n", status>>1, status&1);
1537 status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1538 if (status < 0) return -1;
1539 av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1540 "Imode: %i, Invert: %i\n", status>>1, status&1);
1542 v->s.mv_table_index = get_bits(gb, 2);
1543 v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1547 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1548 vop_dquant_decoding(v);
1554 v->ttmbf = get_bits1(gb);
1557 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1567 v->c_ac_table_index = decode012(gb);
1568 if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1570 v->y_ac_table_index = decode012(gb);
1573 v->s.dc_table_index = get_bits1(gb);
1574 if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) {
1575 av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1576 vop_dquant_decoding(v);
1580 if(v->s.pict_type == FF_BI_TYPE) {
1581 v->s.pict_type = FF_B_TYPE;
1587 /***********************************************************************/
1589 * @defgroup block VC-1 Block-level functions
1590 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
1596 * @brief Get macroblock-level quantizer scale
1598 #define GET_MQUANT() \
1602 if (v->dqprofile == DQPROFILE_ALL_MBS) \
1606 mquant = (get_bits1(gb)) ? v->altpq : v->pq; \
1610 mqdiff = get_bits(gb, 3); \
1611 if (mqdiff != 7) mquant = v->pq + mqdiff; \
1612 else mquant = get_bits(gb, 5); \
1615 if(v->dqprofile == DQPROFILE_SINGLE_EDGE) \
1616 edges = 1 << v->dqsbedge; \
1617 else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES) \
1618 edges = (3 << v->dqsbedge) % 15; \
1619 else if(v->dqprofile == DQPROFILE_FOUR_EDGES) \
1621 if((edges&1) && !s->mb_x) \
1622 mquant = v->altpq; \
1623 if((edges&2) && s->first_slice_line) \
1624 mquant = v->altpq; \
1625 if((edges&4) && s->mb_x == (s->mb_width - 1)) \
1626 mquant = v->altpq; \
1627 if((edges&8) && s->mb_y == (s->mb_height - 1)) \
1628 mquant = v->altpq; \
1632 * @def GET_MVDATA(_dmv_x, _dmv_y)
1633 * @brief Get MV differentials
1634 * @see MVDATA decoding from 8.3.5.2, p(1)20
1635 * @param _dmv_x Horizontal differential for decoded MV
1636 * @param _dmv_y Vertical differential for decoded MV
1638 #define GET_MVDATA(_dmv_x, _dmv_y) \
1639 index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table,\
1640 VC1_MV_DIFF_VLC_BITS, 2); \
1643 mb_has_coeffs = 1; \
1646 else mb_has_coeffs = 0; \
1648 if (!index) { _dmv_x = _dmv_y = 0; } \
1649 else if (index == 35) \
1651 _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample); \
1652 _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample); \
1654 else if (index == 36) \
1663 if (!s->quarter_sample && index1 == 5) val = 1; \
1665 if(size_table[index1] - val > 0) \
1666 val = get_bits(gb, size_table[index1] - val); \
1668 sign = 0 - (val&1); \
1669 _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1672 if (!s->quarter_sample && index1 == 5) val = 1; \
1674 if(size_table[index1] - val > 0) \
1675 val = get_bits(gb, size_table[index1] - val); \
1677 sign = 0 - (val&1); \
1678 _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign; \
1681 /** Predict and set motion vector
1683 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)
1685 int xy, wrap, off = 0;
1690 /* scale MV difference to be quad-pel */
1691 dmv_x <<= 1 - s->quarter_sample;
1692 dmv_y <<= 1 - s->quarter_sample;
1694 wrap = s->b8_stride;
1695 xy = s->block_index[n];
1698 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
1699 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
1700 s->current_picture.motion_val[1][xy][0] = 0;
1701 s->current_picture.motion_val[1][xy][1] = 0;
1702 if(mv1) { /* duplicate motion data for 1-MV block */
1703 s->current_picture.motion_val[0][xy + 1][0] = 0;
1704 s->current_picture.motion_val[0][xy + 1][1] = 0;
1705 s->current_picture.motion_val[0][xy + wrap][0] = 0;
1706 s->current_picture.motion_val[0][xy + wrap][1] = 0;
1707 s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
1708 s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
1709 s->current_picture.motion_val[1][xy + 1][0] = 0;
1710 s->current_picture.motion_val[1][xy + 1][1] = 0;
1711 s->current_picture.motion_val[1][xy + wrap][0] = 0;
1712 s->current_picture.motion_val[1][xy + wrap][1] = 0;
1713 s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
1714 s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
1719 C = s->current_picture.motion_val[0][xy - 1];
1720 A = s->current_picture.motion_val[0][xy - wrap];
1722 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
1724 //in 4-MV mode different blocks have different B predictor position
1727 off = (s->mb_x > 0) ? -1 : 1;
1730 off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
1739 B = s->current_picture.motion_val[0][xy - wrap + off];
1741 if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds
1742 if(s->mb_width == 1) {
1746 px = mid_pred(A[0], B[0], C[0]);
1747 py = mid_pred(A[1], B[1], C[1]);
1749 } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds
1755 /* Pullback MV as specified in 8.3.5.3.4 */
1758 qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);
1759 qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);
1760 X = (s->mb_width << 6) - 4;
1761 Y = (s->mb_height << 6) - 4;
1763 if(qx + px < -60) px = -60 - qx;
1764 if(qy + py < -60) py = -60 - qy;
1766 if(qx + px < -28) px = -28 - qx;
1767 if(qy + py < -28) py = -28 - qy;
1769 if(qx + px > X) px = X - qx;
1770 if(qy + py > Y) py = Y - qy;
1772 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
1773 if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {
1774 if(is_intra[xy - wrap])
1775 sum = FFABS(px) + FFABS(py);
1777 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
1779 if(get_bits1(&s->gb)) {
1787 if(is_intra[xy - 1])
1788 sum = FFABS(px) + FFABS(py);
1790 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
1792 if(get_bits1(&s->gb)) {
1802 /* store MV using signed modulus of MV range defined in 4.11 */
1803 s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
1804 s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;
1805 if(mv1) { /* duplicate motion data for 1-MV block */
1806 s->current_picture.motion_val[0][xy + 1][0] = s->current_picture.motion_val[0][xy][0];
1807 s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];
1808 s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];
1809 s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];
1810 s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];
1811 s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];
1815 /** Motion compensation for direct or interpolated blocks in B-frames
1817 static void vc1_interp_mc(VC1Context *v)
1819 MpegEncContext *s = &v->s;
1820 DSPContext *dsp = &v->s.dsp;
1821 uint8_t *srcY, *srcU, *srcV;
1822 int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
1824 if(!v->s.next_picture.data[0])return;
1826 mx = s->mv[1][0][0];
1827 my = s->mv[1][0][1];
1828 uvmx = (mx + ((mx & 3) == 3)) >> 1;
1829 uvmy = (my + ((my & 3) == 3)) >> 1;
1831 uvmx = uvmx + ((uvmx<0)?-(uvmx&1):(uvmx&1));
1832 uvmy = uvmy + ((uvmy<0)?-(uvmy&1):(uvmy&1));
1834 srcY = s->next_picture.data[0];
1835 srcU = s->next_picture.data[1];
1836 srcV = s->next_picture.data[2];
1838 src_x = s->mb_x * 16 + (mx >> 2);
1839 src_y = s->mb_y * 16 + (my >> 2);
1840 uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
1841 uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
1843 if(v->profile != PROFILE_ADVANCED){
1844 src_x = av_clip( src_x, -16, s->mb_width * 16);
1845 src_y = av_clip( src_y, -16, s->mb_height * 16);
1846 uvsrc_x = av_clip(uvsrc_x, -8, s->mb_width * 8);
1847 uvsrc_y = av_clip(uvsrc_y, -8, s->mb_height * 8);
1849 src_x = av_clip( src_x, -17, s->avctx->coded_width);
1850 src_y = av_clip( src_y, -18, s->avctx->coded_height + 1);
1851 uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1);
1852 uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);
1855 srcY += src_y * s->linesize + src_x;
1856 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
1857 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
1859 /* for grayscale we should not try to read from unknown area */
1860 if(s->flags & CODEC_FLAG_GRAY) {
1861 srcU = s->edge_emu_buffer + 18 * s->linesize;
1862 srcV = s->edge_emu_buffer + 18 * s->linesize;
1866 || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16
1867 || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){
1868 uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
1870 srcY -= s->mspel * (1 + s->linesize);
1871 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
1872 src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
1873 srcY = s->edge_emu_buffer;
1874 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8+1, 8+1,
1875 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1876 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
1877 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1880 /* if we deal with range reduction we need to scale source blocks */
1881 if(v->rangeredfrm) {
1883 uint8_t *src, *src2;
1886 for(j = 0; j < 17 + s->mspel*2; j++) {
1887 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
1890 src = srcU; src2 = srcV;
1891 for(j = 0; j < 9; j++) {
1892 for(i = 0; i < 9; i++) {
1893 src[i] = ((src[i] - 128) >> 1) + 128;
1894 src2[i] = ((src2[i] - 128) >> 1) + 128;
1896 src += s->uvlinesize;
1897 src2 += s->uvlinesize;
1900 srcY += s->mspel * (1 + s->linesize);
1905 dxy = ((my & 1) << 1) | (mx & 1);
1907 dsp->avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
1909 if(s->flags & CODEC_FLAG_GRAY) return;
1910 /* Chroma MC always uses qpel blilinear */
1911 uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
1914 dsp->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
1915 dsp->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
1918 static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)
1922 #if B_FRACTION_DEN==256
1926 return 2 * ((value * n + 255) >> 9);
1927 return (value * n + 128) >> 8;
1930 n -= B_FRACTION_DEN;
1932 return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));
1933 return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;
1937 /** Reconstruct motion vector for B-frame and do motion compensation
1939 static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode)
1942 v->mv_mode2 = v->mv_mode;
1943 v->mv_mode = MV_PMODE_INTENSITY_COMP;
1948 if(v->use_ic) v->mv_mode = v->mv_mode2;
1951 if(mode == BMV_TYPE_INTERPOLATED) {
1954 if(v->use_ic) v->mv_mode = v->mv_mode2;
1958 if(v->use_ic && (mode == BMV_TYPE_BACKWARD)) v->mv_mode = v->mv_mode2;
1959 vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));
1960 if(v->use_ic) v->mv_mode = v->mv_mode2;
1963 static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mvtype)
1965 MpegEncContext *s = &v->s;
1966 int xy, wrap, off = 0;
1971 const uint8_t *is_intra = v->mb_type[0];
1975 /* scale MV difference to be quad-pel */
1976 dmv_x[0] <<= 1 - s->quarter_sample;
1977 dmv_y[0] <<= 1 - s->quarter_sample;
1978 dmv_x[1] <<= 1 - s->quarter_sample;
1979 dmv_y[1] <<= 1 - s->quarter_sample;
1981 wrap = s->b8_stride;
1982 xy = s->block_index[0];
1985 s->current_picture.motion_val[0][xy][0] =
1986 s->current_picture.motion_val[0][xy][1] =
1987 s->current_picture.motion_val[1][xy][0] =
1988 s->current_picture.motion_val[1][xy][1] = 0;
1991 s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);
1992 s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);
1993 s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);
1994 s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);
1996 /* Pullback predicted motion vectors as specified in 8.4.5.4 */
1997 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));
1998 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));
1999 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));
2000 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));
2002 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2003 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2004 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2005 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2009 if((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2010 C = s->current_picture.motion_val[0][xy - 2];
2011 A = s->current_picture.motion_val[0][xy - wrap*2];
2012 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2013 B = s->current_picture.motion_val[0][xy - wrap*2 + off];
2015 if(!s->mb_x) C[0] = C[1] = 0;
2016 if(!s->first_slice_line) { // predictor A is not out of bounds
2017 if(s->mb_width == 1) {
2021 px = mid_pred(A[0], B[0], C[0]);
2022 py = mid_pred(A[1], B[1], C[1]);
2024 } else if(s->mb_x) { // predictor C is not out of bounds
2030 /* Pullback MV as specified in 8.3.5.3.4 */
2033 if(v->profile < PROFILE_ADVANCED) {
2034 qx = (s->mb_x << 5);
2035 qy = (s->mb_y << 5);
2036 X = (s->mb_width << 5) - 4;
2037 Y = (s->mb_height << 5) - 4;
2038 if(qx + px < -28) px = -28 - qx;
2039 if(qy + py < -28) py = -28 - qy;
2040 if(qx + px > X) px = X - qx;
2041 if(qy + py > Y) py = Y - qy;
2043 qx = (s->mb_x << 6);
2044 qy = (s->mb_y << 6);
2045 X = (s->mb_width << 6) - 4;
2046 Y = (s->mb_height << 6) - 4;
2047 if(qx + px < -60) px = -60 - qx;
2048 if(qy + py < -60) py = -60 - qy;
2049 if(qx + px > X) px = X - qx;
2050 if(qy + py > Y) py = Y - qy;
2053 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2054 if(0 && !s->first_slice_line && s->mb_x) {
2055 if(is_intra[xy - wrap])
2056 sum = FFABS(px) + FFABS(py);
2058 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2060 if(get_bits1(&s->gb)) {
2068 if(is_intra[xy - 2])
2069 sum = FFABS(px) + FFABS(py);
2071 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2073 if(get_bits1(&s->gb)) {
2083 /* store MV using signed modulus of MV range defined in 4.11 */
2084 s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;
2085 s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;
2087 if((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2088 C = s->current_picture.motion_val[1][xy - 2];
2089 A = s->current_picture.motion_val[1][xy - wrap*2];
2090 off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2091 B = s->current_picture.motion_val[1][xy - wrap*2 + off];
2093 if(!s->mb_x) C[0] = C[1] = 0;
2094 if(!s->first_slice_line) { // predictor A is not out of bounds
2095 if(s->mb_width == 1) {
2099 px = mid_pred(A[0], B[0], C[0]);
2100 py = mid_pred(A[1], B[1], C[1]);
2102 } else if(s->mb_x) { // predictor C is not out of bounds
2108 /* Pullback MV as specified in 8.3.5.3.4 */
2111 if(v->profile < PROFILE_ADVANCED) {
2112 qx = (s->mb_x << 5);
2113 qy = (s->mb_y << 5);
2114 X = (s->mb_width << 5) - 4;
2115 Y = (s->mb_height << 5) - 4;
2116 if(qx + px < -28) px = -28 - qx;
2117 if(qy + py < -28) py = -28 - qy;
2118 if(qx + px > X) px = X - qx;
2119 if(qy + py > Y) py = Y - qy;
2121 qx = (s->mb_x << 6);
2122 qy = (s->mb_y << 6);
2123 X = (s->mb_width << 6) - 4;
2124 Y = (s->mb_height << 6) - 4;
2125 if(qx + px < -60) px = -60 - qx;
2126 if(qy + py < -60) py = -60 - qy;
2127 if(qx + px > X) px = X - qx;
2128 if(qy + py > Y) py = Y - qy;
2131 /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2132 if(0 && !s->first_slice_line && s->mb_x) {
2133 if(is_intra[xy - wrap])
2134 sum = FFABS(px) + FFABS(py);
2136 sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2138 if(get_bits1(&s->gb)) {
2146 if(is_intra[xy - 2])
2147 sum = FFABS(px) + FFABS(py);
2149 sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2151 if(get_bits1(&s->gb)) {
2161 /* store MV using signed modulus of MV range defined in 4.11 */
2163 s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;
2164 s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;
2166 s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2167 s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2168 s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2169 s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2172 /** Get predicted DC value for I-frames only
2173 * prediction dir: left=0, top=1
2174 * @param s MpegEncContext
2175 * @param[in] n block index in the current MB
2176 * @param dc_val_ptr Pointer to DC predictor
2177 * @param dir_ptr Prediction direction for use in AC prediction
2179 static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2180 int16_t **dc_val_ptr, int *dir_ptr)
2182 int a, b, c, wrap, pred, scale;
2184 static const uint16_t dcpred[32] = {
2185 -1, 1024, 512, 341, 256, 205, 171, 146, 128,
2186 114, 102, 93, 85, 79, 73, 68, 64,
2187 60, 57, 54, 51, 49, 47, 45, 43,
2188 41, 39, 38, 37, 35, 34, 33
2191 /* find prediction - wmv3_dc_scale always used here in fact */
2192 if (n < 4) scale = s->y_dc_scale;
2193 else scale = s->c_dc_scale;
2195 wrap = s->block_wrap[n];
2196 dc_val= s->dc_val[0] + s->block_index[n];
2202 b = dc_val[ - 1 - wrap];
2203 a = dc_val[ - wrap];
2205 if (pq < 9 || !overlap)
2207 /* Set outer values */
2208 if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];
2209 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];
2213 /* Set outer values */
2214 if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;
2215 if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;
2218 if (abs(a - b) <= abs(b - c)) {
2226 /* update predictor */
2227 *dc_val_ptr = &dc_val[0];
2232 /** Get predicted DC value
2233 * prediction dir: left=0, top=1
2234 * @param s MpegEncContext
2235 * @param[in] n block index in the current MB
2236 * @param dc_val_ptr Pointer to DC predictor
2237 * @param dir_ptr Prediction direction for use in AC prediction
2239 static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2240 int a_avail, int c_avail,
2241 int16_t **dc_val_ptr, int *dir_ptr)
2243 int a, b, c, wrap, pred, scale;
2245 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2248 /* find prediction - wmv3_dc_scale always used here in fact */
2249 if (n < 4) scale = s->y_dc_scale;
2250 else scale = s->c_dc_scale;
2252 wrap = s->block_wrap[n];
2253 dc_val= s->dc_val[0] + s->block_index[n];
2259 b = dc_val[ - 1 - wrap];
2260 a = dc_val[ - wrap];
2261 /* scale predictors if needed */
2262 q1 = s->current_picture.qscale_table[mb_pos];
2263 if(c_avail && (n!= 1 && n!=3)) {
2264 q2 = s->current_picture.qscale_table[mb_pos - 1];
2266 c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2268 if(a_avail && (n!= 2 && n!=3)) {
2269 q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2271 a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2273 if(a_avail && c_avail && (n!=3)) {
2276 if(n != 2) off -= s->mb_stride;
2277 q2 = s->current_picture.qscale_table[off];
2279 b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2282 if(a_avail && c_avail) {
2283 if(abs(a - b) <= abs(b - c)) {
2290 } else if(a_avail) {
2293 } else if(c_avail) {
2301 /* update predictor */
2302 *dc_val_ptr = &dc_val[0];
2308 * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles
2309 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
2313 static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)
2315 int xy, wrap, pred, a, b, c;
2317 xy = s->block_index[n];
2318 wrap = s->b8_stride;
2323 a = s->coded_block[xy - 1 ];
2324 b = s->coded_block[xy - 1 - wrap];
2325 c = s->coded_block[xy - wrap];
2334 *coded_block_ptr = &s->coded_block[xy];
2340 * Decode one AC coefficient
2341 * @param v The VC1 context
2342 * @param last Last coefficient
2343 * @param skip How much zero coefficients to skip
2344 * @param value Decoded AC coefficient value
2347 static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)
2349 GetBitContext *gb = &v->s.gb;
2350 int index, escape, run = 0, level = 0, lst = 0;
2352 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2353 if (index != vc1_ac_sizes[codingset] - 1) {
2354 run = vc1_index_decode_table[codingset][index][0];
2355 level = vc1_index_decode_table[codingset][index][1];
2356 lst = index >= vc1_last_decode_table[codingset];
2360 escape = decode210(gb);
2362 index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2363 run = vc1_index_decode_table[codingset][index][0];
2364 level = vc1_index_decode_table[codingset][index][1];
2365 lst = index >= vc1_last_decode_table[codingset];
2368 level += vc1_last_delta_level_table[codingset][run];
2370 level += vc1_delta_level_table[codingset][run];
2373 run += vc1_last_delta_run_table[codingset][level] + 1;
2375 run += vc1_delta_run_table[codingset][level] + 1;
2381 lst = get_bits1(gb);
2382 if(v->s.esc3_level_length == 0) {
2383 if(v->pq < 8 || v->dquantfrm) { // table 59
2384 v->s.esc3_level_length = get_bits(gb, 3);
2385 if(!v->s.esc3_level_length)
2386 v->s.esc3_level_length = get_bits(gb, 2) + 8;
2388 v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;
2390 v->s.esc3_run_length = 3 + get_bits(gb, 2);
2392 run = get_bits(gb, v->s.esc3_run_length);
2393 sign = get_bits1(gb);
2394 level = get_bits(gb, v->s.esc3_level_length);
2405 /** Decode intra block in intra frames - should be faster than decode_intra_block
2406 * @param v VC1Context
2407 * @param block block to decode
2408 * @param coded are AC coeffs present or not
2409 * @param codingset set of VLC to decode data
2411 static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)
2413 GetBitContext *gb = &v->s.gb;
2414 MpegEncContext *s = &v->s;
2415 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2418 int16_t *ac_val, *ac_val2;
2421 /* Get DC differential */
2423 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2425 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2428 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2433 if (dcdiff == 119 /* ESC index value */)
2435 /* TODO: Optimize */
2436 if (v->pq == 1) dcdiff = get_bits(gb, 10);
2437 else if (v->pq == 2) dcdiff = get_bits(gb, 9);
2438 else dcdiff = get_bits(gb, 8);
2443 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2444 else if (v->pq == 2)
2445 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2452 dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);
2455 /* Store the quantized DC coeff, used for prediction */
2457 block[0] = dcdiff * s->y_dc_scale;
2459 block[0] = dcdiff * s->c_dc_scale;
2472 int last = 0, skip, value;
2473 const int8_t *zz_table;
2477 scale = v->pq * 2 + v->halfpq;
2481 zz_table = wmv1_scantable[2];
2483 zz_table = wmv1_scantable[3];
2485 zz_table = wmv1_scantable[1];
2487 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2489 if(dc_pred_dir) //left
2492 ac_val -= 16 * s->block_wrap[n];
2495 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2499 block[zz_table[i++]] = value;
2502 /* apply AC prediction if needed */
2504 if(dc_pred_dir) { //left
2505 for(k = 1; k < 8; k++)
2506 block[k << 3] += ac_val[k];
2508 for(k = 1; k < 8; k++)
2509 block[k] += ac_val[k + 8];
2512 /* save AC coeffs for further prediction */
2513 for(k = 1; k < 8; k++) {
2514 ac_val2[k] = block[k << 3];
2515 ac_val2[k + 8] = block[k];
2518 /* scale AC coeffs */
2519 for(k = 1; k < 64; k++)
2523 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2526 if(s->ac_pred) i = 63;
2532 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2535 scale = v->pq * 2 + v->halfpq;
2536 memset(ac_val2, 0, 16 * 2);
2537 if(dc_pred_dir) {//left
2540 memcpy(ac_val2, ac_val, 8 * 2);
2542 ac_val -= 16 * s->block_wrap[n];
2544 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2547 /* apply AC prediction if needed */
2549 if(dc_pred_dir) { //left
2550 for(k = 1; k < 8; k++) {
2551 block[k << 3] = ac_val[k] * scale;
2552 if(!v->pquantizer && block[k << 3])
2553 block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;
2556 for(k = 1; k < 8; k++) {
2557 block[k] = ac_val[k + 8] * scale;
2558 if(!v->pquantizer && block[k])
2559 block[k] += (block[k] < 0) ? -v->pq : v->pq;
2565 s->block_last_index[n] = i;
2570 /** Decode intra block in intra frames - should be faster than decode_intra_block
2571 * @param v VC1Context
2572 * @param block block to decode
2573 * @param coded are AC coeffs present or not
2574 * @param codingset set of VLC to decode data
2576 static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant)
2578 GetBitContext *gb = &v->s.gb;
2579 MpegEncContext *s = &v->s;
2580 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2583 int16_t *ac_val, *ac_val2;
2585 int a_avail = v->a_avail, c_avail = v->c_avail;
2586 int use_pred = s->ac_pred;
2589 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2591 /* Get DC differential */
2593 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2595 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2598 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2603 if (dcdiff == 119 /* ESC index value */)
2605 /* TODO: Optimize */
2606 if (mquant == 1) dcdiff = get_bits(gb, 10);
2607 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2608 else dcdiff = get_bits(gb, 8);
2613 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2614 else if (mquant == 2)
2615 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2622 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);
2625 /* Store the quantized DC coeff, used for prediction */
2627 block[0] = dcdiff * s->y_dc_scale;
2629 block[0] = dcdiff * s->c_dc_scale;
2638 /* check if AC is needed at all */
2639 if(!a_avail && !c_avail) use_pred = 0;
2640 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2643 scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);
2645 if(dc_pred_dir) //left
2648 ac_val -= 16 * s->block_wrap[n];
2650 q1 = s->current_picture.qscale_table[mb_pos];
2651 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2652 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2653 if(dc_pred_dir && n==1) q2 = q1;
2654 if(!dc_pred_dir && n==2) q2 = q1;
2658 int last = 0, skip, value;
2659 const int8_t *zz_table;
2664 zz_table = wmv1_scantable[2];
2666 zz_table = wmv1_scantable[3];
2668 zz_table = wmv1_scantable[1];
2671 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2675 block[zz_table[i++]] = value;
2678 /* apply AC prediction if needed */
2680 /* scale predictors if needed*/
2682 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2683 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2685 if(dc_pred_dir) { //left
2686 for(k = 1; k < 8; k++)
2687 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2689 for(k = 1; k < 8; k++)
2690 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2693 if(dc_pred_dir) { //left
2694 for(k = 1; k < 8; k++)
2695 block[k << 3] += ac_val[k];
2697 for(k = 1; k < 8; k++)
2698 block[k] += ac_val[k + 8];
2702 /* save AC coeffs for further prediction */
2703 for(k = 1; k < 8; k++) {
2704 ac_val2[k] = block[k << 3];
2705 ac_val2[k + 8] = block[k];
2708 /* scale AC coeffs */
2709 for(k = 1; k < 64; k++)
2713 block[k] += (block[k] < 0) ? -mquant : mquant;
2716 if(use_pred) i = 63;
2717 } else { // no AC coeffs
2720 memset(ac_val2, 0, 16 * 2);
2721 if(dc_pred_dir) {//left
2723 memcpy(ac_val2, ac_val, 8 * 2);
2725 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2726 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2727 for(k = 1; k < 8; k++)
2728 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2733 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2735 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2736 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2737 for(k = 1; k < 8; k++)
2738 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2743 /* apply AC prediction if needed */
2745 if(dc_pred_dir) { //left
2746 for(k = 1; k < 8; k++) {
2747 block[k << 3] = ac_val2[k] * scale;
2748 if(!v->pquantizer && block[k << 3])
2749 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2752 for(k = 1; k < 8; k++) {
2753 block[k] = ac_val2[k + 8] * scale;
2754 if(!v->pquantizer && block[k])
2755 block[k] += (block[k] < 0) ? -mquant : mquant;
2761 s->block_last_index[n] = i;
2766 /** Decode intra block in inter frames - more generic version than vc1_decode_i_block
2767 * @param v VC1Context
2768 * @param block block to decode
2769 * @param coded are AC coeffs present or not
2770 * @param mquant block quantizer
2771 * @param codingset set of VLC to decode data
2773 static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)
2775 GetBitContext *gb = &v->s.gb;
2776 MpegEncContext *s = &v->s;
2777 int dc_pred_dir = 0; /* Direction of the DC prediction used */
2780 int16_t *ac_val, *ac_val2;
2782 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2783 int a_avail = v->a_avail, c_avail = v->c_avail;
2784 int use_pred = s->ac_pred;
2788 /* XXX: Guard against dumb values of mquant */
2789 mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );
2791 /* Set DC scale - y and c use the same */
2792 s->y_dc_scale = s->y_dc_scale_table[mquant];
2793 s->c_dc_scale = s->c_dc_scale_table[mquant];
2795 /* Get DC differential */
2797 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2799 dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2802 av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2807 if (dcdiff == 119 /* ESC index value */)
2809 /* TODO: Optimize */
2810 if (mquant == 1) dcdiff = get_bits(gb, 10);
2811 else if (mquant == 2) dcdiff = get_bits(gb, 9);
2812 else dcdiff = get_bits(gb, 8);
2817 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2818 else if (mquant == 2)
2819 dcdiff = (dcdiff<<1) + get_bits1(gb) - 1;
2826 dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);
2829 /* Store the quantized DC coeff, used for prediction */
2832 block[0] = dcdiff * s->y_dc_scale;
2834 block[0] = dcdiff * s->c_dc_scale;
2843 /* check if AC is needed at all and adjust direction if needed */
2844 if(!a_avail) dc_pred_dir = 1;
2845 if(!c_avail) dc_pred_dir = 0;
2846 if(!a_avail && !c_avail) use_pred = 0;
2847 ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2850 scale = mquant * 2 + v->halfpq;
2852 if(dc_pred_dir) //left
2855 ac_val -= 16 * s->block_wrap[n];
2857 q1 = s->current_picture.qscale_table[mb_pos];
2858 if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2859 if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2860 if(dc_pred_dir && n==1) q2 = q1;
2861 if(!dc_pred_dir && n==2) q2 = q1;
2865 int last = 0, skip, value;
2866 const int8_t *zz_table;
2869 zz_table = wmv1_scantable[0];
2872 vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2876 block[zz_table[i++]] = value;
2879 /* apply AC prediction if needed */
2881 /* scale predictors if needed*/
2883 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2884 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2886 if(dc_pred_dir) { //left
2887 for(k = 1; k < 8; k++)
2888 block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2890 for(k = 1; k < 8; k++)
2891 block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2894 if(dc_pred_dir) { //left
2895 for(k = 1; k < 8; k++)
2896 block[k << 3] += ac_val[k];
2898 for(k = 1; k < 8; k++)
2899 block[k] += ac_val[k + 8];
2903 /* save AC coeffs for further prediction */
2904 for(k = 1; k < 8; k++) {
2905 ac_val2[k] = block[k << 3];
2906 ac_val2[k + 8] = block[k];
2909 /* scale AC coeffs */
2910 for(k = 1; k < 64; k++)
2914 block[k] += (block[k] < 0) ? -mquant : mquant;
2917 if(use_pred) i = 63;
2918 } else { // no AC coeffs
2921 memset(ac_val2, 0, 16 * 2);
2922 if(dc_pred_dir) {//left
2924 memcpy(ac_val2, ac_val, 8 * 2);
2926 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2927 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2928 for(k = 1; k < 8; k++)
2929 ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2934 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2936 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2937 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2938 for(k = 1; k < 8; k++)
2939 ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2944 /* apply AC prediction if needed */
2946 if(dc_pred_dir) { //left
2947 for(k = 1; k < 8; k++) {
2948 block[k << 3] = ac_val2[k] * scale;
2949 if(!v->pquantizer && block[k << 3])
2950 block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2953 for(k = 1; k < 8; k++) {
2954 block[k] = ac_val2[k + 8] * scale;
2955 if(!v->pquantizer && block[k])
2956 block[k] += (block[k] < 0) ? -mquant : mquant;
2962 s->block_last_index[n] = i;
2969 static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,
2970 uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left)
2972 MpegEncContext *s = &v->s;
2973 GetBitContext *gb = &s->gb;
2976 int scale, off, idx, last, skip, value;
2977 int ttblk = ttmb & 7;
2981 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)];
2983 if(ttblk == TT_4X4) {
2984 subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);
2986 if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {
2987 subblkpat = decode012(gb);
2988 if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits
2989 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;
2990 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;
2992 scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);
2994 // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT
2995 if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {
2996 subblkpat = 2 - (ttblk == TT_8X4_TOP);
2999 if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
3000 subblkpat = 2 - (ttblk == TT_4X8_LEFT);
3009 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3013 idx = wmv1_scantable[0][i++];
3014 block[idx] = value * scale;
3016 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3019 s->dsp.vc1_inv_trans_8x8(block);
3020 s->dsp.add_pixels_clamped(block, dst, linesize);
3021 if(apply_filter && cbp_top & 0xC)
3022 vc1_loop_filter(dst, 1, linesize, 8, mquant);
3023 if(apply_filter && cbp_left & 0xA)
3024 vc1_loop_filter(dst, linesize, 1, 8, mquant);
3028 pat = ~subblkpat & 0xF;
3029 for(j = 0; j < 4; j++) {
3030 last = subblkpat & (1 << (3 - j));
3032 off = (j & 1) * 4 + (j & 2) * 16;
3034 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3038 idx = ff_vc1_simple_progressive_4x4_zz[i++];
3039 block[idx + off] = value * scale;
3041 block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;
3043 if(!(subblkpat & (1 << (3 - j))) && !skip_block){
3044 s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
3045 if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))
3046 vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, 1, linesize, 4, mquant);
3047 if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))
3048 vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, linesize, 1, 4, mquant);
3053 pat = ~((subblkpat & 2)*6 + (subblkpat & 1)*3) & 0xF;
3054 for(j = 0; j < 2; j++) {
3055 last = subblkpat & (1 << (1 - j));
3059 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3063 idx = v->zz_8x4[i++]+off;
3064 block[idx] = value * scale;
3066 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3068 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3069 s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
3070 if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))
3071 vc1_loop_filter(dst + j*4*linesize, 1, linesize, 8, mquant);
3072 if(apply_filter && cbp_left & (2 << j))
3073 vc1_loop_filter(dst + j*4*linesize, linesize, 1, 4, mquant);
3078 pat = ~(subblkpat*5) & 0xF;
3079 for(j = 0; j < 2; j++) {
3080 last = subblkpat & (1 << (1 - j));
3084 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3088 idx = v->zz_4x8[i++]+off;
3089 block[idx] = value * scale;
3091 block[idx] += (block[idx] < 0) ? -mquant : mquant;
3093 if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3094 s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
3095 if(apply_filter && cbp_top & (2 << j))
3096 vc1_loop_filter(dst + j*4, 1, linesize, 4, mquant);
3097 if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))
3098 vc1_loop_filter(dst + j*4, linesize, 1, 8, mquant);
3106 static const int size_table [6] = { 0, 2, 3, 4, 5, 8 };
3107 static const int offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3109 /** Decode one P-frame MB (in Simple/Main profile)
3111 static int vc1_decode_p_mb(VC1Context *v)
3113 MpegEncContext *s = &v->s;
3114 GetBitContext *gb = &s->gb;
3116 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3117 int cbp; /* cbp decoding stuff */
3118 int mqdiff, mquant; /* MB quantization */
3119 int ttmb = v->ttfrm; /* MB Transform type */
3121 int mb_has_coeffs = 1; /* last_flag */
3122 int dmv_x, dmv_y; /* Differential MV components */
3123 int index, index1; /* LUT indexes */
3124 int val, sign; /* temp values */
3125 int first_block = 1;
3127 int skipped, fourmv;
3128 int block_cbp = 0, pat;
3129 int apply_loop_filter;
3131 mquant = v->pq; /* Loosy initialization */
3133 if (v->mv_type_is_raw)
3134 fourmv = get_bits1(gb);
3136 fourmv = v->mv_type_mb_plane[mb_pos];
3138 skipped = get_bits1(gb);
3140 skipped = v->s.mbskip_table[mb_pos];
3142 s->dsp.clear_blocks(s->block[0]);
3144 apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);
3145 if (!fourmv) /* 1MV mode */
3149 GET_MVDATA(dmv_x, dmv_y);
3152 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3153 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3155 s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
3156 vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
3158 /* FIXME Set DC val for inter block ? */
3159 if (s->mb_intra && !mb_has_coeffs)
3162 s->ac_pred = get_bits1(gb);
3165 else if (mb_has_coeffs)
3167 if (s->mb_intra) s->ac_pred = get_bits1(gb);
3168 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3176 s->current_picture.qscale_table[mb_pos] = mquant;
3178 if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3179 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,
3180 VC1_TTMB_VLC_BITS, 2);
3181 if(!s->mb_intra) vc1_mc_1mv(v, 0);
3185 s->dc_val[0][s->block_index[i]] = 0;
3187 val = ((cbp >> (5 - i)) & 1);
3188 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3189 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3191 /* check if prediction blocks A and C are available */
3192 v->a_avail = v->c_avail = 0;
3193 if(i == 2 || i == 3 || !s->first_slice_line)
3194 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3195 if(i == 1 || i == 3 || s->mb_x)
3196 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3198 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3199 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3200 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3201 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3202 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3203 if(v->pq >= 9 && v->overlap) {
3205 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3207 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3209 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3210 int left_cbp, top_cbp;
3212 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3213 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3215 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3216 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3219 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3221 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3223 block_cbp |= 0xF << (i << 2);
3225 int left_cbp = 0, top_cbp = 0, filter = 0;
3226 if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3229 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3230 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3232 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3233 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3236 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3238 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3240 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);
3241 block_cbp |= pat << (i << 2);
3242 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3250 for(i = 0; i < 6; i++) {
3251 v->mb_type[0][s->block_index[i]] = 0;
3252 s->dc_val[0][s->block_index[i]] = 0;
3254 s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;
3255 s->current_picture.qscale_table[mb_pos] = 0;
3256 vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
3263 if (!skipped /* unskipped MB */)
3265 int intra_count = 0, coded_inter = 0;
3266 int is_intra[6], is_coded[6];
3268 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3271 val = ((cbp >> (5 - i)) & 1);
3272 s->dc_val[0][s->block_index[i]] = 0;
3279 GET_MVDATA(dmv_x, dmv_y);
3281 vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
3282 if(!s->mb_intra) vc1_mc_4mv_luma(v, i);
3283 intra_count += s->mb_intra;
3284 is_intra[i] = s->mb_intra;
3285 is_coded[i] = mb_has_coeffs;
3288 is_intra[i] = (intra_count >= 3);
3291 if(i == 4) vc1_mc_4mv_chroma(v);
3292 v->mb_type[0][s->block_index[i]] = is_intra[i];
3293 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];
3295 // if there are no coded blocks then don't do anything more
3296 if(!intra_count && !coded_inter) return 0;
3299 s->current_picture.qscale_table[mb_pos] = mquant;
3300 /* test if block is intra and has pred */
3305 if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])
3306 || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {
3311 if(intrapred)s->ac_pred = get_bits1(gb);
3312 else s->ac_pred = 0;
3314 if (!v->ttmbf && coded_inter)
3315 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3319 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3320 s->mb_intra = is_intra[i];
3322 /* check if prediction blocks A and C are available */
3323 v->a_avail = v->c_avail = 0;
3324 if(i == 2 || i == 3 || !s->first_slice_line)
3325 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3326 if(i == 1 || i == 3 || s->mb_x)
3327 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3329 vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);
3330 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3331 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3332 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3333 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
3334 if(v->pq >= 9 && v->overlap) {
3336 s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3338 s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3340 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)){
3341 int left_cbp, top_cbp;
3343 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3344 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3346 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3347 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3350 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3352 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3354 block_cbp |= 0xF << (i << 2);
3355 } else if(is_coded[i]) {
3356 int left_cbp = 0, top_cbp = 0, filter = 0;
3357 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)){
3360 left_cbp = v->cbp[s->mb_x - 1] >> (i * 4);
3361 top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3363 left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4));
3364 top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3367 vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3369 vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3371 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);
3372 block_cbp |= pat << (i << 2);
3373 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3382 s->current_picture.qscale_table[mb_pos] = 0;
3383 for (i=0; i<6; i++) {
3384 v->mb_type[0][s->block_index[i]] = 0;
3385 s->dc_val[0][s->block_index[i]] = 0;
3389 vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
3390 vc1_mc_4mv_luma(v, i);
3392 vc1_mc_4mv_chroma(v);
3393 s->current_picture.qscale_table[mb_pos] = 0;
3397 v->cbp[s->mb_x] = block_cbp;
3399 /* Should never happen */
3403 /** Decode one B-frame MB (in Main profile)
3405 static void vc1_decode_b_mb(VC1Context *v)
3407 MpegEncContext *s = &v->s;
3408 GetBitContext *gb = &s->gb;
3410 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3411 int cbp = 0; /* cbp decoding stuff */
3412 int mqdiff, mquant; /* MB quantization */
3413 int ttmb = v->ttfrm; /* MB Transform type */
3414 int mb_has_coeffs = 0; /* last_flag */
3415 int index, index1; /* LUT indexes */
3416 int val, sign; /* temp values */
3417 int first_block = 1;
3419 int skipped, direct;
3420 int dmv_x[2], dmv_y[2];
3421 int bmvtype = BMV_TYPE_BACKWARD;
3423 mquant = v->pq; /* Loosy initialization */
3427 direct = get_bits1(gb);
3429 direct = v->direct_mb_plane[mb_pos];
3431 skipped = get_bits1(gb);
3433 skipped = v->s.mbskip_table[mb_pos];
3435 s->dsp.clear_blocks(s->block[0]);
3436 dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
3437 for(i = 0; i < 6; i++) {
3438 v->mb_type[0][s->block_index[i]] = 0;
3439 s->dc_val[0][s->block_index[i]] = 0;
3441 s->current_picture.qscale_table[mb_pos] = 0;
3445 GET_MVDATA(dmv_x[0], dmv_y[0]);
3446 dmv_x[1] = dmv_x[0];
3447 dmv_y[1] = dmv_y[0];
3449 if(skipped || !s->mb_intra) {
3450 bmvtype = decode012(gb);
3453 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
3456 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
3459 bmvtype = BMV_TYPE_INTERPOLATED;
3460 dmv_x[0] = dmv_y[0] = 0;
3464 for(i = 0; i < 6; i++)
3465 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3468 if(direct) bmvtype = BMV_TYPE_INTERPOLATED;
3469 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3470 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3474 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3478 s->current_picture.qscale_table[mb_pos] = mquant;
3480 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3481 dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;
3482 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3483 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3485 if(!mb_has_coeffs && !s->mb_intra) {
3486 /* no coded blocks - effectively skipped */
3487 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3488 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3491 if(s->mb_intra && !mb_has_coeffs) {
3493 s->current_picture.qscale_table[mb_pos] = mquant;
3494 s->ac_pred = get_bits1(gb);
3496 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3498 if(bmvtype == BMV_TYPE_INTERPOLATED) {
3499 GET_MVDATA(dmv_x[0], dmv_y[0]);
3500 if(!mb_has_coeffs) {
3501 /* interpolated skipped block */
3502 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3503 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3507 vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3509 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3512 s->ac_pred = get_bits1(gb);
3513 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3515 s->current_picture.qscale_table[mb_pos] = mquant;
3516 if(!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3517 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3523 s->dc_val[0][s->block_index[i]] = 0;
3525 val = ((cbp >> (5 - i)) & 1);
3526 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3527 v->mb_type[0][s->block_index[i]] = s->mb_intra;
3529 /* check if prediction blocks A and C are available */
3530 v->a_avail = v->c_avail = 0;
3531 if(i == 2 || i == 3 || !s->first_slice_line)
3532 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3533 if(i == 1 || i == 3 || s->mb_x)
3534 v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3536 vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3537 if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3538 s->dsp.vc1_inv_trans_8x8(s->block[i]);
3539 if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3540 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3542 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);
3543 if(!v->ttmbf && ttmb < 8) ttmb = -1;
3549 /** Decode blocks of I-frame
3551 static void vc1_decode_i_blocks(VC1Context *v)
3554 MpegEncContext *s = &v->s;
3559 /* select codingmode used for VLC tables selection */
3560 switch(v->y_ac_table_index){
3562 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3565 v->codingset = CS_HIGH_MOT_INTRA;
3568 v->codingset = CS_MID_RATE_INTRA;
3572 switch(v->c_ac_table_index){
3574 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3577 v->codingset2 = CS_HIGH_MOT_INTER;
3580 v->codingset2 = CS_MID_RATE_INTER;
3584 /* Set DC scale - y and c use the same */
3585 s->y_dc_scale = s->y_dc_scale_table[v->pq];
3586 s->c_dc_scale = s->c_dc_scale_table[v->pq];
3589 s->mb_x = s->mb_y = 0;
3591 s->first_slice_line = 1;
3592 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3593 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3594 ff_init_block_index(s);
3595 ff_update_block_index(s);
3596 s->dsp.clear_blocks(s->block[0]);
3597 mb_pos = s->mb_x + s->mb_y * s->mb_width;
3598 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3599 s->current_picture.qscale_table[mb_pos] = v->pq;
3600 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3601 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3603 // do actual MB decoding and displaying
3604 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3605 v->s.ac_pred = get_bits1(&v->s.gb);
3607 for(k = 0; k < 6; k++) {
3608 val = ((cbp >> (5 - k)) & 1);
3611 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3615 cbp |= val << (5 - k);
3617 vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);
3619 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3620 if(v->pq >= 9 && v->overlap) {
3621 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3625 vc1_put_block(v, s->block);
3626 if(v->pq >= 9 && v->overlap) {
3628 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3629 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3630 if(!(s->flags & CODEC_FLAG_GRAY)) {
3631 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3632 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3635 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3636 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3637 if(!s->first_slice_line) {
3638 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3639 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3640 if(!(s->flags & CODEC_FLAG_GRAY)) {
3641 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3642 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3645 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3646 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3648 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3650 if(get_bits_count(&s->gb) > v->bits) {
3651 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3652 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3656 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3657 s->first_slice_line = 0;
3659 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3662 /** Decode blocks of I-frame for advanced profile
3664 static void vc1_decode_i_blocks_adv(VC1Context *v)
3667 MpegEncContext *s = &v->s;
3674 GetBitContext *gb = &s->gb;
3676 /* select codingmode used for VLC tables selection */
3677 switch(v->y_ac_table_index){
3679 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3682 v->codingset = CS_HIGH_MOT_INTRA;
3685 v->codingset = CS_MID_RATE_INTRA;
3689 switch(v->c_ac_table_index){
3691 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3694 v->codingset2 = CS_HIGH_MOT_INTER;
3697 v->codingset2 = CS_MID_RATE_INTER;
3702 s->mb_x = s->mb_y = 0;
3704 s->first_slice_line = 1;
3705 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3706 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3707 ff_init_block_index(s);
3708 ff_update_block_index(s);
3709 s->dsp.clear_blocks(s->block[0]);
3710 mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3711 s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3712 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3713 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3715 // do actual MB decoding and displaying
3716 cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3717 if(v->acpred_is_raw)
3718 v->s.ac_pred = get_bits1(&v->s.gb);
3720 v->s.ac_pred = v->acpred_plane[mb_pos];
3722 if(v->condover == CONDOVER_SELECT) {
3723 if(v->overflg_is_raw)
3724 overlap = get_bits1(&v->s.gb);
3726 overlap = v->over_flags_plane[mb_pos];
3728 overlap = (v->condover == CONDOVER_ALL);
3732 s->current_picture.qscale_table[mb_pos] = mquant;
3733 /* Set DC scale - y and c use the same */
3734 s->y_dc_scale = s->y_dc_scale_table[mquant];
3735 s->c_dc_scale = s->c_dc_scale_table[mquant];
3737 for(k = 0; k < 6; k++) {
3738 val = ((cbp >> (5 - k)) & 1);
3741 int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3745 cbp |= val << (5 - k);
3747 v->a_avail = !s->first_slice_line || (k==2 || k==3);
3748 v->c_avail = !!s->mb_x || (k==1 || k==3);
3750 vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);
3752 s->dsp.vc1_inv_trans_8x8(s->block[k]);
3753 for(j = 0; j < 64; j++) s->block[k][j] += 128;
3756 vc1_put_block(v, s->block);
3759 s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3760 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3761 if(!(s->flags & CODEC_FLAG_GRAY)) {
3762 s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3763 s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3766 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3767 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3768 if(!s->first_slice_line) {
3769 s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3770 s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3771 if(!(s->flags & CODEC_FLAG_GRAY)) {
3772 s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3773 s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3776 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3777 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3779 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3781 if(get_bits_count(&s->gb) > v->bits) {
3782 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3783 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3787 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3788 s->first_slice_line = 0;
3790 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3793 static void vc1_decode_p_blocks(VC1Context *v)
3795 MpegEncContext *s = &v->s;
3797 /* select codingmode used for VLC tables selection */
3798 switch(v->c_ac_table_index){
3800 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3803 v->codingset = CS_HIGH_MOT_INTRA;
3806 v->codingset = CS_MID_RATE_INTRA;
3810 switch(v->c_ac_table_index){
3812 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3815 v->codingset2 = CS_HIGH_MOT_INTER;
3818 v->codingset2 = CS_MID_RATE_INTER;
3822 s->first_slice_line = 1;
3823 memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);
3824 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3825 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3826 ff_init_block_index(s);
3827 ff_update_block_index(s);
3828 s->dsp.clear_blocks(s->block[0]);
3831 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3832 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3833 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);
3837 memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride);
3838 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3839 s->first_slice_line = 0;
3841 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3844 static void vc1_decode_b_blocks(VC1Context *v)
3846 MpegEncContext *s = &v->s;
3848 /* select codingmode used for VLC tables selection */
3849 switch(v->c_ac_table_index){
3851 v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3854 v->codingset = CS_HIGH_MOT_INTRA;
3857 v->codingset = CS_MID_RATE_INTRA;
3861 switch(v->c_ac_table_index){
3863 v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3866 v->codingset2 = CS_HIGH_MOT_INTER;
3869 v->codingset2 = CS_MID_RATE_INTER;
3873 s->first_slice_line = 1;
3874 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3875 for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3876 ff_init_block_index(s);
3877 ff_update_block_index(s);
3878 s->dsp.clear_blocks(s->block[0]);
3881 if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3882 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3883 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);
3886 if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[s->mb_x + s->mb_y *s->mb_stride]);
3888 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3889 s->first_slice_line = 0;
3891 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3894 static void vc1_decode_skip_blocks(VC1Context *v)
3896 MpegEncContext *s = &v->s;
3898 ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3899 s->first_slice_line = 1;
3900 for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3902 ff_init_block_index(s);
3903 ff_update_block_index(s);
3904 memcpy(s->dest[0], s->last_picture.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16);
3905 memcpy(s->dest[1], s->last_picture.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3906 memcpy(s->dest[2], s->last_picture.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3907 ff_draw_horiz_band(s, s->mb_y * 16, 16);
3908 s->first_slice_line = 0;
3910 s->pict_type = FF_P_TYPE;
3913 static void vc1_decode_blocks(VC1Context *v)
3916 v->s.esc3_level_length = 0;
3918 ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );
3921 switch(v->s.pict_type) {
3923 if(v->profile == PROFILE_ADVANCED)
3924 vc1_decode_i_blocks_adv(v);
3926 vc1_decode_i_blocks(v);
3929 if(v->p_frame_skipped)
3930 vc1_decode_skip_blocks(v);
3932 vc1_decode_p_blocks(v);
3936 if(v->profile == PROFILE_ADVANCED)
3937 vc1_decode_i_blocks_adv(v);
3939 vc1_decode_i_blocks(v);
3941 vc1_decode_b_blocks(v);
3947 /** Find VC-1 marker in buffer
3948 * @return position where next marker starts or end of buffer if no marker found
3950 static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)
3952 uint32_t mrk = 0xFFFFFFFF;
3954 if(end-src < 4) return end;
3956 mrk = (mrk << 8) | *src++;
3963 static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)
3968 for(dsize = 0; dsize < size; dsize++) *dst++ = *src++;
3971 for(i = 0; i < size; i++, src++) {
3972 if(src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {
3973 dst[dsize++] = src[1];
3977 dst[dsize++] = *src;
3982 /** Initialize a VC1/WMV3 decoder
3983 * @todo TODO: Handle VC-1 IDUs (Transport level?)
3984 * @todo TODO: Decypher remaining bits in extra_data
3986 static av_cold int vc1_decode_init(AVCodecContext *avctx)
3988 VC1Context *v = avctx->priv_data;
3989 MpegEncContext *s = &v->s;
3992 if (!avctx->extradata_size || !avctx->extradata) return -1;
3993 if (!(avctx->flags & CODEC_FLAG_GRAY))
3994 avctx->pix_fmt = PIX_FMT_YUV420P;
3996 avctx->pix_fmt = PIX_FMT_GRAY8;
3998 avctx->flags |= CODEC_FLAG_EMU_EDGE;
3999 v->s.flags |= CODEC_FLAG_EMU_EDGE;
4001 if(avctx->idct_algo==FF_IDCT_AUTO){
4002 avctx->idct_algo=FF_IDCT_WMV2;
4005 if(ff_h263_decode_init(avctx) < 0)
4007 if (vc1_init_common(v) < 0) return -1;
4009 avctx->coded_width = avctx->width;
4010 avctx->coded_height = avctx->height;
4011 if (avctx->codec_id == CODEC_ID_WMV3)
4015 // looks like WMV3 has a sequence header stored in the extradata
4016 // advanced sequence header may be before the first frame
4017 // the last byte of the extradata is a version number, 1 for the
4018 // samples we can decode
4020 init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);
4022 if (decode_sequence_header(avctx, &gb) < 0)
4025 count = avctx->extradata_size*8 - get_bits_count(&gb);
4028 av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",
4029 count, get_bits(&gb, count));
4033 av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);
4035 } else { // VC1/WVC1
4036 const uint8_t *start = avctx->extradata;
4037 uint8_t *end = avctx->extradata + avctx->extradata_size;
4038 const uint8_t *next;
4039 int size, buf2_size;
4040 uint8_t *buf2 = NULL;
4041 int seq_initialized = 0, ep_initialized = 0;
4043 if(avctx->extradata_size < 16) {
4044 av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);
4048 buf2 = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
4049 if(start[0]) start++; // in WVC1 extradata first byte is its size
4051 for(; next < end; start = next){
4052 next = find_next_marker(start + 4, end);
4053 size = next - start - 4;
4054 if(size <= 0) continue;
4055 buf2_size = vc1_unescape_buffer(start + 4, size, buf2);
4056 init_get_bits(&gb, buf2, buf2_size * 8);
4057 switch(AV_RB32(start)){
4058 case VC1_CODE_SEQHDR:
4059 if(decode_sequence_header(avctx, &gb) < 0){
4063 seq_initialized = 1;
4065 case VC1_CODE_ENTRYPOINT:
4066 if(decode_entry_point(avctx, &gb) < 0){
4075 if(!seq_initialized || !ep_initialized){
4076 av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");
4080 avctx->has_b_frames= !!(avctx->max_b_frames);
4081 s->low_delay = !avctx->has_b_frames;
4083 s->mb_width = (avctx->coded_width+15)>>4;
4084 s->mb_height = (avctx->coded_height+15)>>4;
4086 /* Allocate mb bitplanes */
4087 v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4088 v->direct_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4089 v->acpred_plane = av_malloc(s->mb_stride * s->mb_height);
4090 v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height);
4092 v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);
4093 v->cbp = v->cbp_base + s->mb_stride;
4095 /* allocate block type info in that way so it could be used with s->block_index[] */
4096 v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);
4097 v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;
4098 v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;
4099 v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);
4101 /* Init coded blocks info */
4102 if (v->profile == PROFILE_ADVANCED)
4104 // if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
4106 // if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
4110 ff_intrax8_common_init(&v->x8,s);
4115 /** Decode a VC1/WMV3 frame
4116 * @todo TODO: Handle VC-1 IDUs (Transport level?)
4118 static int vc1_decode_frame(AVCodecContext *avctx,
4119 void *data, int *data_size,
4120 const uint8_t *buf, int buf_size)
4122 VC1Context *v = avctx->priv_data;
4123 MpegEncContext *s = &v->s;
4124 AVFrame *pict = data;
4125 uint8_t *buf2 = NULL;
4126 const uint8_t *buf_vdpau = buf;
4128 /* no supplementary picture */
4129 if (buf_size == 0) {
4130 /* special case for last picture */
4131 if (s->low_delay==0 && s->next_picture_ptr) {
4132 *pict= *(AVFrame*)s->next_picture_ptr;
4133 s->next_picture_ptr= NULL;
4135 *data_size = sizeof(AVFrame);
4141 /* We need to set current_picture_ptr before reading the header,
4142 * otherwise we cannot store anything in there. */
4143 if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){
4144 int i= ff_find_unused_picture(s, 0);
4145 s->current_picture_ptr= &s->picture[i];
4148 if (s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU){
4149 if (v->profile < PROFILE_ADVANCED)
4150 avctx->pix_fmt = PIX_FMT_VDPAU_WMV3;
4152 avctx->pix_fmt = PIX_FMT_VDPAU_VC1;
4155 //for advanced profile we may need to parse and unescape data
4156 if (avctx->codec_id == CODEC_ID_VC1) {
4158 buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
4160 if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */
4161 const uint8_t *start, *end, *next;
4165 for(start = buf, end = buf + buf_size; next < end; start = next){
4166 next = find_next_marker(start + 4, end);
4167 size = next - start - 4;
4168 if(size <= 0) continue;
4169 switch(AV_RB32(start)){
4170 case VC1_CODE_FRAME:
4171 if (s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
4173 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4175 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */
4176 buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4177 init_get_bits(&s->gb, buf2, buf_size2*8);
4178 decode_entry_point(avctx, &s->gb);
4180 case VC1_CODE_SLICE:
4181 av_log(avctx, AV_LOG_ERROR, "Sliced decoding is not implemented (yet)\n");
4186 }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */
4187 const uint8_t *divider;
4189 divider = find_next_marker(buf, buf + buf_size);
4190 if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){
4191 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");
4196 buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);
4198 av_free(buf2);return -1;
4200 buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);
4202 init_get_bits(&s->gb, buf2, buf_size2*8);
4204 init_get_bits(&s->gb, buf, buf_size*8);
4205 // do parse frame header
4206 if(v->profile < PROFILE_ADVANCED) {
4207 if(vc1_parse_frame_header(v, &s->gb) == -1) {
4212 if(vc1_parse_frame_header_adv(v, &s->gb) == -1) {
4218 if(s->pict_type != FF_I_TYPE && !v->res_rtm_flag){
4224 s->current_picture.pict_type= s->pict_type;
4225 s->current_picture.key_frame= s->pict_type == FF_I_TYPE;
4227 /* skip B-frames if we don't have reference frames */
4228 if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){
4230 return -1;//buf_size;
4232 /* skip b frames if we are in a hurry */
4233 if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;
4234 if( (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
4235 || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
4236 || avctx->skip_frame >= AVDISCARD_ALL) {
4240 /* skip everything if we are in a hurry>=5 */
4241 if(avctx->hurry_up>=5) {
4243 return -1;//buf_size;
4246 if(s->next_p_frame_damaged){
4247 if(s->pict_type==FF_B_TYPE)
4250 s->next_p_frame_damaged=0;
4253 if(MPV_frame_start(s, avctx) < 0) {
4258 s->me.qpel_put= s->dsp.put_qpel_pixels_tab;
4259 s->me.qpel_avg= s->dsp.avg_qpel_pixels_tab;
4261 if ((CONFIG_VC1_VDPAU_DECODER || CONFIG_WMV3_VDPAU_DECODER)
4262 &&s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
4263 ff_vdpau_vc1_decode_picture(s, buf_vdpau, (buf + buf_size) - buf_vdpau);
4265 ff_er_frame_start(s);
4267 v->bits = buf_size * 8;
4268 vc1_decode_blocks(v);
4269 //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);
4270 // if(get_bits_count(&s->gb) > buf_size * 8)
4277 assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);
4278 assert(s->current_picture.pict_type == s->pict_type);
4279 if (s->pict_type == FF_B_TYPE || s->low_delay) {
4280 *pict= *(AVFrame*)s->current_picture_ptr;
4281 } else if (s->last_picture_ptr != NULL) {
4282 *pict= *(AVFrame*)s->last_picture_ptr;
4285 if(s->last_picture_ptr || s->low_delay){
4286 *data_size = sizeof(AVFrame);
4287 ff_print_debug_info(s, pict);
4290 /* Return the Picture timestamp as the frame number */
4291 /* we subtract 1 because it is added on utils.c */
4292 avctx->frame_number = s->picture_number - 1;
4299 /** Close a VC1/WMV3 decoder
4300 * @warning Initial try at using MpegEncContext stuff
4302 static av_cold int vc1_decode_end(AVCodecContext *avctx)
4304 VC1Context *v = avctx->priv_data;
4306 av_freep(&v->hrd_rate);
4307 av_freep(&v->hrd_buffer);
4308 MPV_common_end(&v->s);
4309 av_freep(&v->mv_type_mb_plane);
4310 av_freep(&v->direct_mb_plane);
4311 av_freep(&v->acpred_plane);
4312 av_freep(&v->over_flags_plane);
4313 av_freep(&v->mb_type_base);
4314 av_freep(&v->cbp_base);
4315 ff_intrax8_common_end(&v->x8);
4320 AVCodec vc1_decoder = {
4331 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),
4334 AVCodec wmv3_decoder = {
4345 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),
4348 #if CONFIG_WMV3_VDPAU_DECODER
4349 AVCodec wmv3_vdpau_decoder = {
4358 CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,
4360 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 VDPAU"),
4364 #if CONFIG_VC1_VDPAU_DECODER
4365 AVCodec vc1_vdpau_decoder = {
4374 CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,
4376 .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1 VDPAU"),