Statistics
| Branch: | Revision:

ffmpeg / libavcodec / vc1.c @ 820089cb

History | View | Annotate | Download (152 KB)

1
/*
2
 * VC-1 and WMV3 decoder
3
 * Copyright (c) 2006-2007 Konstantin Shishkov
4
 * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
5
 *
6
 * This file is part of FFmpeg.
7
 *
8
 * FFmpeg is free software; you can redistribute it and/or
9
 * modify it under the terms of the GNU Lesser General Public
10
 * License as published by the Free Software Foundation; either
11
 * version 2.1 of the License, or (at your option) any later version.
12
 *
13
 * FFmpeg is distributed in the hope that it will be useful,
14
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
16
 * Lesser General Public License for more details.
17
 *
18
 * You should have received a copy of the GNU Lesser General Public
19
 * License along with FFmpeg; if not, write to the Free Software
20
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21
 */
22

    
23
/**
24
 * @file vc1.c
25
 * VC-1 and WMV3 decoder
26
 *
27
 */
28
#include "dsputil.h"
29
#include "avcodec.h"
30
#include "mpegvideo.h"
31
#include "vc1.h"
32
#include "vc1data.h"
33
#include "vc1acdata.h"
34
#include "msmpeg4data.h"
35
#include "unary.h"
36
#include "simple_idct.h"
37

    
38
#undef NDEBUG
39
#include <assert.h>
40

    
41
#define MB_INTRA_VLC_BITS 9
42
#define DC_VLC_BITS 9
43
#define AC_VLC_BITS 9
44
static const uint16_t table_mb_intra[64][2];
45

    
46

    
47
/**
48
 * Init VC-1 specific tables and VC1Context members
49
 * @param v The VC1Context to initialize
50
 * @return Status
51
 */
52
static int vc1_init_common(VC1Context *v)
53
{
54
    static int done = 0;
55
    int i = 0;
56

    
57
    v->hrd_rate = v->hrd_buffer = NULL;
58

    
59
    /* VLC tables */
60
    if(!done)
61
    {
62
        done = 1;
63
        init_vlc(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,
64
                 ff_vc1_bfraction_bits, 1, 1,
65
                 ff_vc1_bfraction_codes, 1, 1, 1);
66
        init_vlc(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,
67
                 ff_vc1_norm2_bits, 1, 1,
68
                 ff_vc1_norm2_codes, 1, 1, 1);
69
        init_vlc(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,
70
                 ff_vc1_norm6_bits, 1, 1,
71
                 ff_vc1_norm6_codes, 2, 2, 1);
72
        init_vlc(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,
73
                 ff_vc1_imode_bits, 1, 1,
74
                 ff_vc1_imode_codes, 1, 1, 1);
75
        for (i=0; i<3; i++)
76
        {
77
            init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,
78
                     ff_vc1_ttmb_bits[i], 1, 1,
79
                     ff_vc1_ttmb_codes[i], 2, 2, 1);
80
            init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,
81
                     ff_vc1_ttblk_bits[i], 1, 1,
82
                     ff_vc1_ttblk_codes[i], 1, 1, 1);
83
            init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,
84
                     ff_vc1_subblkpat_bits[i], 1, 1,
85
                     ff_vc1_subblkpat_codes[i], 1, 1, 1);
86
        }
87
        for(i=0; i<4; i++)
88
        {
89
            init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,
90
                     ff_vc1_4mv_block_pattern_bits[i], 1, 1,
91
                     ff_vc1_4mv_block_pattern_codes[i], 1, 1, 1);
92
            init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,
93
                     ff_vc1_cbpcy_p_bits[i], 1, 1,
94
                     ff_vc1_cbpcy_p_codes[i], 2, 2, 1);
95
            init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,
96
                     ff_vc1_mv_diff_bits[i], 1, 1,
97
                     ff_vc1_mv_diff_codes[i], 2, 2, 1);
98
        }
99
        for(i=0; i<8; i++)
100
            init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],
101
                     &vc1_ac_tables[i][0][1], 8, 4,
102
                     &vc1_ac_tables[i][0][0], 8, 4, 1);
103
        init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,
104
                 &ff_msmp4_mb_i_table[0][1], 4, 2,
105
                 &ff_msmp4_mb_i_table[0][0], 4, 2, 1);
106
    }
107

    
108
    /* Other defaults */
109
    v->pq = -1;
110
    v->mvrange = 0; /* 7.1.1.18, p80 */
111

    
112
    return 0;
113
}
114

    
115
/***********************************************************************/
116
/**
117
 * @defgroup bitplane VC9 Bitplane decoding
118
 * @see 8.7, p56
119
 * @{
120
 */
121

    
122
/** @addtogroup bitplane
123
 * Imode types
124
 * @{
125
 */
126
enum Imode {
127
    IMODE_RAW,
128
    IMODE_NORM2,
129
    IMODE_DIFF2,
130
    IMODE_NORM6,
131
    IMODE_DIFF6,
132
    IMODE_ROWSKIP,
133
    IMODE_COLSKIP
134
};
135
/** @} */ //imode defines
136

    
137
/** Decode rows by checking if they are skipped
138
 * @param plane Buffer to store decoded bits
139
 * @param[in] width Width of this buffer
140
 * @param[in] height Height of this buffer
141
 * @param[in] stride of this buffer
142
 */
143
static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
144
    int x, y;
145

    
146
    for (y=0; y<height; y++){
147
        if (!get_bits1(gb)) //rowskip
148
            memset(plane, 0, width);
149
        else
150
            for (x=0; x<width; x++)
151
                plane[x] = get_bits1(gb);
152
        plane += stride;
153
    }
154
}
155

    
156
/** Decode columns by checking if they are skipped
157
 * @param plane Buffer to store decoded bits
158
 * @param[in] width Width of this buffer
159
 * @param[in] height Height of this buffer
160
 * @param[in] stride of this buffer
161
 * @todo FIXME: Optimize
162
 */
163
static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
164
    int x, y;
165

    
166
    for (x=0; x<width; x++){
167
        if (!get_bits1(gb)) //colskip
168
            for (y=0; y<height; y++)
169
                plane[y*stride] = 0;
170
        else
171
            for (y=0; y<height; y++)
172
                plane[y*stride] = get_bits1(gb);
173
        plane ++;
174
    }
175
}
176

    
177
/** Decode a bitplane's bits
178
 * @param bp Bitplane where to store the decode bits
179
 * @param v VC-1 context for bit reading and logging
180
 * @return Status
181
 * @todo FIXME: Optimize
182
 */
183
static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)
184
{
185
    GetBitContext *gb = &v->s.gb;
186

    
187
    int imode, x, y, code, offset;
188
    uint8_t invert, *planep = data;
189
    int width, height, stride;
190

    
191
    width = v->s.mb_width;
192
    height = v->s.mb_height;
193
    stride = v->s.mb_stride;
194
    invert = get_bits1(gb);
195
    imode = get_vlc2(gb, ff_vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);
196

    
197
    *raw_flag = 0;
198
    switch (imode)
199
    {
200
    case IMODE_RAW:
201
        //Data is actually read in the MB layer (same for all tests == "raw")
202
        *raw_flag = 1; //invert ignored
203
        return invert;
204
    case IMODE_DIFF2:
205
    case IMODE_NORM2:
206
        if ((height * width) & 1)
207
        {
208
            *planep++ = get_bits1(gb);
209
            offset = 1;
210
        }
211
        else offset = 0;
212
        // decode bitplane as one long line
213
        for (y = offset; y < height * width; y += 2) {
214
            code = get_vlc2(gb, ff_vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
215
            *planep++ = code & 1;
216
            offset++;
217
            if(offset == width) {
218
                offset = 0;
219
                planep += stride - width;
220
            }
221
            *planep++ = code >> 1;
222
            offset++;
223
            if(offset == width) {
224
                offset = 0;
225
                planep += stride - width;
226
            }
227
        }
228
        break;
229
    case IMODE_DIFF6:
230
    case IMODE_NORM6:
231
        if(!(height % 3) && (width % 3)) { // use 2x3 decoding
232
            for(y = 0; y < height; y+= 3) {
233
                for(x = width & 1; x < width; x += 2) {
234
                    code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
235
                    if(code < 0){
236
                        av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
237
                        return -1;
238
                    }
239
                    planep[x + 0] = (code >> 0) & 1;
240
                    planep[x + 1] = (code >> 1) & 1;
241
                    planep[x + 0 + stride] = (code >> 2) & 1;
242
                    planep[x + 1 + stride] = (code >> 3) & 1;
243
                    planep[x + 0 + stride * 2] = (code >> 4) & 1;
244
                    planep[x + 1 + stride * 2] = (code >> 5) & 1;
245
                }
246
                planep += stride * 3;
247
            }
248
            if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);
249
        } else { // 3x2
250
            planep += (height & 1) * stride;
251
            for(y = height & 1; y < height; y += 2) {
252
                for(x = width % 3; x < width; x += 3) {
253
                    code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
254
                    if(code < 0){
255
                        av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
256
                        return -1;
257
                    }
258
                    planep[x + 0] = (code >> 0) & 1;
259
                    planep[x + 1] = (code >> 1) & 1;
260
                    planep[x + 2] = (code >> 2) & 1;
261
                    planep[x + 0 + stride] = (code >> 3) & 1;
262
                    planep[x + 1 + stride] = (code >> 4) & 1;
263
                    planep[x + 2 + stride] = (code >> 5) & 1;
264
                }
265
                planep += stride * 2;
266
            }
267
            x = width % 3;
268
            if(x) decode_colskip(data  ,             x, height    , stride, &v->s.gb);
269
            if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);
270
        }
271
        break;
272
    case IMODE_ROWSKIP:
273
        decode_rowskip(data, width, height, stride, &v->s.gb);
274
        break;
275
    case IMODE_COLSKIP:
276
        decode_colskip(data, width, height, stride, &v->s.gb);
277
        break;
278
    default: break;
279
    }
280

    
281
    /* Applying diff operator */
282
    if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
283
    {
284
        planep = data;
285
        planep[0] ^= invert;
286
        for (x=1; x<width; x++)
287
            planep[x] ^= planep[x-1];
288
        for (y=1; y<height; y++)
289
        {
290
            planep += stride;
291
            planep[0] ^= planep[-stride];
292
            for (x=1; x<width; x++)
293
            {
294
                if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;
295
                else                                 planep[x] ^= planep[x-1];
296
            }
297
        }
298
    }
299
    else if (invert)
300
    {
301
        planep = data;
302
        for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride
303
    }
304
    return (imode<<1) + invert;
305
}
306

    
307
/** @} */ //Bitplane group
308

    
309
#define FILTSIGN(a) ((a) >= 0 ? 1 : -1)
310
/**
311
 * VC-1 in-loop deblocking filter for one line
312
 * @param src source block type
313
 * @param pq block quantizer
314
 * @return whether other 3 pairs should be filtered or not
315
 * @see 8.6
316
 */
317
static int vc1_filter_line(uint8_t* src, int stride, int pq){
318
    int a0, a1, a2, a3, d, clip, filt3 = 0;
319
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
320

    
321
    a0     = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
322
    if(FFABS(a0) < pq){
323
        a1 = (2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3;
324
        a2 = (2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3;
325
        a3 = FFMIN(FFABS(a1), FFABS(a2));
326
        if(a3 < FFABS(a0)){
327
            d = 5 * ((a0 >=0 ? a3 : -a3) - a0) / 8;
328
            clip = (src[-1*stride] - src[ 0*stride])/2;
329
            if(clip){
330
                filt3 = 1;
331
                if(clip > 0)
332
                    d = av_clip(d, 0, clip);
333
                else
334
                    d = av_clip(d, clip, 0);
335
                src[-1*stride] = cm[src[-1*stride] - d];
336
                src[ 0*stride] = cm[src[ 0*stride] + d];
337
            }
338
        }
339
    }
340
    return filt3;
341
}
342

    
343
/**
344
 * VC-1 in-loop deblocking filter
345
 * @param src source block type
346
 * @param len edge length to filter (4 or 8 pixels)
347
 * @param pq block quantizer
348
 * @see 8.6
349
 */
350
static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
351
{
352
    int i;
353
    int filt3;
354

    
355
    for(i = 0; i < len; i += 4){
356
        filt3 = vc1_filter_line(src + 2*step, stride, pq);
357
        if(filt3){
358
            vc1_filter_line(src + 0*step, stride, pq);
359
            vc1_filter_line(src + 1*step, stride, pq);
360
            vc1_filter_line(src + 3*step, stride, pq);
361
        }
362
        src += step * 4;
363
    }
364
}
365

    
366
static void vc1_loop_filter_iblk(MpegEncContext *s, int pq)
367
{
368
    int i, j;
369
    if(!s->first_slice_line)
370
        vc1_loop_filter(s->dest[0], 1, s->linesize, 16, pq);
371
    vc1_loop_filter(s->dest[0] + 8*s->linesize, 1, s->linesize, 16, pq);
372
    for(i = !s->mb_x*8; i < 16; i += 8)
373
        vc1_loop_filter(s->dest[0] + i, s->linesize, 1, 16, pq);
374
    for(j = 0; j < 2; j++){
375
        if(!s->first_slice_line)
376
            vc1_loop_filter(s->dest[j+1], 1, s->uvlinesize, 8, pq);
377
        if(s->mb_x)
378
            vc1_loop_filter(s->dest[j+1], s->uvlinesize, 1, 8, pq);
379
    }
380
}
381

    
382
/***********************************************************************/
383
/** VOP Dquant decoding
384
 * @param v VC-1 Context
385
 */
386
static int vop_dquant_decoding(VC1Context *v)
387
{
388
    GetBitContext *gb = &v->s.gb;
389
    int pqdiff;
390

    
391
    //variable size
392
    if (v->dquant == 2)
393
    {
394
        pqdiff = get_bits(gb, 3);
395
        if (pqdiff == 7) v->altpq = get_bits(gb, 5);
396
        else v->altpq = v->pq + pqdiff + 1;
397
    }
398
    else
399
    {
400
        v->dquantfrm = get_bits1(gb);
401
        if ( v->dquantfrm )
402
        {
403
            v->dqprofile = get_bits(gb, 2);
404
            switch (v->dqprofile)
405
            {
406
            case DQPROFILE_SINGLE_EDGE:
407
            case DQPROFILE_DOUBLE_EDGES:
408
                v->dqsbedge = get_bits(gb, 2);
409
                break;
410
            case DQPROFILE_ALL_MBS:
411
                v->dqbilevel = get_bits1(gb);
412
                if(!v->dqbilevel)
413
                    v->halfpq = 0;
414
            default: break; //Forbidden ?
415
            }
416
            if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)
417
            {
418
                pqdiff = get_bits(gb, 3);
419
                if (pqdiff == 7) v->altpq = get_bits(gb, 5);
420
                else v->altpq = v->pq + pqdiff + 1;
421
            }
422
        }
423
    }
424
    return 0;
425
}
426

    
427
/** Put block onto picture
428
 */
429
static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
430
{
431
    uint8_t *Y;
432
    int ys, us, vs;
433
    DSPContext *dsp = &v->s.dsp;
434

    
435
    if(v->rangeredfrm) {
436
        int i, j, k;
437
        for(k = 0; k < 6; k++)
438
            for(j = 0; j < 8; j++)
439
                for(i = 0; i < 8; i++)
440
                    block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;
441

    
442
    }
443
    ys = v->s.current_picture.linesize[0];
444
    us = v->s.current_picture.linesize[1];
445
    vs = v->s.current_picture.linesize[2];
446
    Y = v->s.dest[0];
447

    
448
    dsp->put_pixels_clamped(block[0], Y, ys);
449
    dsp->put_pixels_clamped(block[1], Y + 8, ys);
450
    Y += ys * 8;
451
    dsp->put_pixels_clamped(block[2], Y, ys);
452
    dsp->put_pixels_clamped(block[3], Y + 8, ys);
453

    
454
    if(!(v->s.flags & CODEC_FLAG_GRAY)) {
455
        dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
456
        dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
457
    }
458
}
459

    
460
/** Do motion compensation over 1 macroblock
461
 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
462
 */
463
static void vc1_mc_1mv(VC1Context *v, int dir)
464
{
465
    MpegEncContext *s = &v->s;
466
    DSPContext *dsp = &v->s.dsp;
467
    uint8_t *srcY, *srcU, *srcV;
468
    int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
469

    
470
    if(!v->s.last_picture.data[0])return;
471

    
472
    mx = s->mv[dir][0][0];
473
    my = s->mv[dir][0][1];
474

    
475
    // store motion vectors for further use in B frames
476
    if(s->pict_type == FF_P_TYPE) {
477
        s->current_picture.motion_val[1][s->block_index[0]][0] = mx;
478
        s->current_picture.motion_val[1][s->block_index[0]][1] = my;
479
    }
480
    uvmx = (mx + ((mx & 3) == 3)) >> 1;
481
    uvmy = (my + ((my & 3) == 3)) >> 1;
482
    if(v->fastuvmc) {
483
        uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
484
        uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
485
    }
486
    if(!dir) {
487
        srcY = s->last_picture.data[0];
488
        srcU = s->last_picture.data[1];
489
        srcV = s->last_picture.data[2];
490
    } else {
491
        srcY = s->next_picture.data[0];
492
        srcU = s->next_picture.data[1];
493
        srcV = s->next_picture.data[2];
494
    }
495

    
496
    src_x = s->mb_x * 16 + (mx >> 2);
497
    src_y = s->mb_y * 16 + (my >> 2);
498
    uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
499
    uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
500

    
501
    if(v->profile != PROFILE_ADVANCED){
502
        src_x   = av_clip(  src_x, -16, s->mb_width  * 16);
503
        src_y   = av_clip(  src_y, -16, s->mb_height * 16);
504
        uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);
505
        uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);
506
    }else{
507
        src_x   = av_clip(  src_x, -17, s->avctx->coded_width);
508
        src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);
509
        uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);
510
        uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);
511
    }
512

    
513
    srcY += src_y * s->linesize + src_x;
514
    srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
515
    srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
516

    
517
    /* for grayscale we should not try to read from unknown area */
518
    if(s->flags & CODEC_FLAG_GRAY) {
519
        srcU = s->edge_emu_buffer + 18 * s->linesize;
520
        srcV = s->edge_emu_buffer + 18 * s->linesize;
521
    }
522

    
523
    if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
524
       || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3
525
       || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){
526
        uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
527

    
528
        srcY -= s->mspel * (1 + s->linesize);
529
        ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
530
                            src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
531
        srcY = s->edge_emu_buffer;
532
        ff_emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8+1, 8+1,
533
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
534
        ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
535
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
536
        srcU = uvbuf;
537
        srcV = uvbuf + 16;
538
        /* if we deal with range reduction we need to scale source blocks */
539
        if(v->rangeredfrm) {
540
            int i, j;
541
            uint8_t *src, *src2;
542

    
543
            src = srcY;
544
            for(j = 0; j < 17 + s->mspel*2; j++) {
545
                for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
546
                src += s->linesize;
547
            }
548
            src = srcU; src2 = srcV;
549
            for(j = 0; j < 9; j++) {
550
                for(i = 0; i < 9; i++) {
551
                    src[i] = ((src[i] - 128) >> 1) + 128;
552
                    src2[i] = ((src2[i] - 128) >> 1) + 128;
553
                }
554
                src += s->uvlinesize;
555
                src2 += s->uvlinesize;
556
            }
557
        }
558
        /* if we deal with intensity compensation we need to scale source blocks */
559
        if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
560
            int i, j;
561
            uint8_t *src, *src2;
562

    
563
            src = srcY;
564
            for(j = 0; j < 17 + s->mspel*2; j++) {
565
                for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];
566
                src += s->linesize;
567
            }
568
            src = srcU; src2 = srcV;
569
            for(j = 0; j < 9; j++) {
570
                for(i = 0; i < 9; i++) {
571
                    src[i] = v->lutuv[src[i]];
572
                    src2[i] = v->lutuv[src2[i]];
573
                }
574
                src += s->uvlinesize;
575
                src2 += s->uvlinesize;
576
            }
577
        }
578
        srcY += s->mspel * (1 + s->linesize);
579
    }
580

    
581
    if(s->mspel) {
582
        dxy = ((my & 3) << 2) | (mx & 3);
583
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0]    , srcY    , s->linesize, v->rnd);
584
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
585
        srcY += s->linesize * 8;
586
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize    , srcY    , s->linesize, v->rnd);
587
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
588
    } else { // hpel mc - always used for luma
589
        dxy = (my & 2) | ((mx & 2) >> 1);
590

    
591
        if(!v->rnd)
592
            dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
593
        else
594
            dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
595
    }
596

    
597
    if(s->flags & CODEC_FLAG_GRAY) return;
598
    /* Chroma MC always uses qpel bilinear */
599
    uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
600
    uvmx = (uvmx&3)<<1;
601
    uvmy = (uvmy&3)<<1;
602
    if(!v->rnd){
603
        dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
604
        dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
605
    }else{
606
        dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
607
        dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
608
    }
609
}
610

    
611
/** Do motion compensation for 4-MV macroblock - luminance block
612
 */
613
static void vc1_mc_4mv_luma(VC1Context *v, int n)
614
{
615
    MpegEncContext *s = &v->s;
616
    DSPContext *dsp = &v->s.dsp;
617
    uint8_t *srcY;
618
    int dxy, mx, my, src_x, src_y;
619
    int off;
620

    
621
    if(!v->s.last_picture.data[0])return;
622
    mx = s->mv[0][n][0];
623
    my = s->mv[0][n][1];
624
    srcY = s->last_picture.data[0];
625

    
626
    off = s->linesize * 4 * (n&2) + (n&1) * 8;
627

    
628
    src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);
629
    src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);
630

    
631
    if(v->profile != PROFILE_ADVANCED){
632
        src_x   = av_clip(  src_x, -16, s->mb_width  * 16);
633
        src_y   = av_clip(  src_y, -16, s->mb_height * 16);
634
    }else{
635
        src_x   = av_clip(  src_x, -17, s->avctx->coded_width);
636
        src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);
637
    }
638

    
639
    srcY += src_y * s->linesize + src_x;
640

    
641
    if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
642
       || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2
643
       || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){
644
        srcY -= s->mspel * (1 + s->linesize);
645
        ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,
646
                            src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
647
        srcY = s->edge_emu_buffer;
648
        /* if we deal with range reduction we need to scale source blocks */
649
        if(v->rangeredfrm) {
650
            int i, j;
651
            uint8_t *src;
652

    
653
            src = srcY;
654
            for(j = 0; j < 9 + s->mspel*2; j++) {
655
                for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
656
                src += s->linesize;
657
            }
658
        }
659
        /* if we deal with intensity compensation we need to scale source blocks */
660
        if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
661
            int i, j;
662
            uint8_t *src;
663

    
664
            src = srcY;
665
            for(j = 0; j < 9 + s->mspel*2; j++) {
666
                for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];
667
                src += s->linesize;
668
            }
669
        }
670
        srcY += s->mspel * (1 + s->linesize);
671
    }
672

    
673
    if(s->mspel) {
674
        dxy = ((my & 3) << 2) | (mx & 3);
675
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);
676
    } else { // hpel mc - always used for luma
677
        dxy = (my & 2) | ((mx & 2) >> 1);
678
        if(!v->rnd)
679
            dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
680
        else
681
            dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
682
    }
683
}
684

    
685
static inline int median4(int a, int b, int c, int d)
686
{
687
    if(a < b) {
688
        if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
689
        else      return (FFMIN(b, c) + FFMAX(a, d)) / 2;
690
    } else {
691
        if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
692
        else      return (FFMIN(a, c) + FFMAX(b, d)) / 2;
693
    }
694
}
695

    
696

    
697
/** Do motion compensation for 4-MV macroblock - both chroma blocks
698
 */
699
static void vc1_mc_4mv_chroma(VC1Context *v)
700
{
701
    MpegEncContext *s = &v->s;
702
    DSPContext *dsp = &v->s.dsp;
703
    uint8_t *srcU, *srcV;
704
    int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;
705
    int i, idx, tx = 0, ty = 0;
706
    int mvx[4], mvy[4], intra[4];
707
    static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
708

    
709
    if(!v->s.last_picture.data[0])return;
710
    if(s->flags & CODEC_FLAG_GRAY) return;
711

    
712
    for(i = 0; i < 4; i++) {
713
        mvx[i] = s->mv[0][i][0];
714
        mvy[i] = s->mv[0][i][1];
715
        intra[i] = v->mb_type[0][s->block_index[i]];
716
    }
717

    
718
    /* calculate chroma MV vector from four luma MVs */
719
    idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];
720
    if(!idx) { // all blocks are inter
721
        tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
722
        ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
723
    } else if(count[idx] == 1) { // 3 inter blocks
724
        switch(idx) {
725
        case 0x1:
726
            tx = mid_pred(mvx[1], mvx[2], mvx[3]);
727
            ty = mid_pred(mvy[1], mvy[2], mvy[3]);
728
            break;
729
        case 0x2:
730
            tx = mid_pred(mvx[0], mvx[2], mvx[3]);
731
            ty = mid_pred(mvy[0], mvy[2], mvy[3]);
732
            break;
733
        case 0x4:
734
            tx = mid_pred(mvx[0], mvx[1], mvx[3]);
735
            ty = mid_pred(mvy[0], mvy[1], mvy[3]);
736
            break;
737
        case 0x8:
738
            tx = mid_pred(mvx[0], mvx[1], mvx[2]);
739
            ty = mid_pred(mvy[0], mvy[1], mvy[2]);
740
            break;
741
        }
742
    } else if(count[idx] == 2) {
743
        int t1 = 0, t2 = 0;
744
        for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}
745
        for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}
746
        tx = (mvx[t1] + mvx[t2]) / 2;
747
        ty = (mvy[t1] + mvy[t2]) / 2;
748
    } else {
749
        s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
750
        s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
751
        return; //no need to do MC for inter blocks
752
    }
753

    
754
    s->current_picture.motion_val[1][s->block_index[0]][0] = tx;
755
    s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
756
    uvmx = (tx + ((tx&3) == 3)) >> 1;
757
    uvmy = (ty + ((ty&3) == 3)) >> 1;
758
    if(v->fastuvmc) {
759
        uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
760
        uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
761
    }
762

    
763
    uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
764
    uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
765

    
766
    if(v->profile != PROFILE_ADVANCED){
767
        uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);
768
        uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);
769
    }else{
770
        uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);
771
        uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);
772
    }
773

    
774
    srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
775
    srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;
776
    if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
777
       || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
778
       || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){
779
        ff_emulated_edge_mc(s->edge_emu_buffer     , srcU, s->uvlinesize, 8+1, 8+1,
780
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
781
        ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,
782
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
783
        srcU = s->edge_emu_buffer;
784
        srcV = s->edge_emu_buffer + 16;
785

    
786
        /* if we deal with range reduction we need to scale source blocks */
787
        if(v->rangeredfrm) {
788
            int i, j;
789
            uint8_t *src, *src2;
790

    
791
            src = srcU; src2 = srcV;
792
            for(j = 0; j < 9; j++) {
793
                for(i = 0; i < 9; i++) {
794
                    src[i] = ((src[i] - 128) >> 1) + 128;
795
                    src2[i] = ((src2[i] - 128) >> 1) + 128;
796
                }
797
                src += s->uvlinesize;
798
                src2 += s->uvlinesize;
799
            }
800
        }
801
        /* if we deal with intensity compensation we need to scale source blocks */
802
        if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
803
            int i, j;
804
            uint8_t *src, *src2;
805

    
806
            src = srcU; src2 = srcV;
807
            for(j = 0; j < 9; j++) {
808
                for(i = 0; i < 9; i++) {
809
                    src[i] = v->lutuv[src[i]];
810
                    src2[i] = v->lutuv[src2[i]];
811
                }
812
                src += s->uvlinesize;
813
                src2 += s->uvlinesize;
814
            }
815
        }
816
    }
817

    
818
    /* Chroma MC always uses qpel bilinear */
819
    uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
820
    uvmx = (uvmx&3)<<1;
821
    uvmy = (uvmy&3)<<1;
822
    if(!v->rnd){
823
        dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
824
        dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
825
    }else{
826
        dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
827
        dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
828
    }
829
}
830

    
831
static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);
832

    
833
/**
834
 * Decode Simple/Main Profiles sequence header
835
 * @see Figure 7-8, p16-17
836
 * @param avctx Codec context
837
 * @param gb GetBit context initialized from Codec context extra_data
838
 * @return Status
839
 */
840
static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)
841
{
842
    VC1Context *v = avctx->priv_data;
843

    
844
    av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));
845
    v->profile = get_bits(gb, 2);
846
    if (v->profile == PROFILE_COMPLEX)
847
    {
848
        av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");
849
    }
850

    
851
    if (v->profile == PROFILE_ADVANCED)
852
    {
853
        v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz;
854
        v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz;
855
        return decode_sequence_header_adv(v, gb);
856
    }
857
    else
858
    {
859
        v->zz_8x4 = wmv2_scantableA;
860
        v->zz_4x8 = wmv2_scantableB;
861
        v->res_sm = get_bits(gb, 2); //reserved
862
        if (v->res_sm)
863
        {
864
            av_log(avctx, AV_LOG_ERROR,
865
                   "Reserved RES_SM=%i is forbidden\n", v->res_sm);
866
            return -1;
867
        }
868
    }
869

    
870
    // (fps-2)/4 (->30)
871
    v->frmrtq_postproc = get_bits(gb, 3); //common
872
    // (bitrate-32kbps)/64kbps
873
    v->bitrtq_postproc = get_bits(gb, 5); //common
874
    v->s.loop_filter = get_bits1(gb); //common
875
    if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)
876
    {
877
        av_log(avctx, AV_LOG_ERROR,
878
               "LOOPFILTER shell not be enabled in simple profile\n");
879
    }
880
    if(v->s.avctx->skip_loop_filter >= AVDISCARD_ALL)
881
        v->s.loop_filter = 0;
882

    
883
    v->res_x8 = get_bits1(gb); //reserved
884
    v->multires = get_bits1(gb);
885
    v->res_fasttx = get_bits1(gb);
886
    if (!v->res_fasttx)
887
    {
888
        v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct;
889
        v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;
890
        v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;
891
        v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;
892
    }
893

    
894
    v->fastuvmc =  get_bits1(gb); //common
895
    if (!v->profile && !v->fastuvmc)
896
    {
897
        av_log(avctx, AV_LOG_ERROR,
898
               "FASTUVMC unavailable in Simple Profile\n");
899
        return -1;
900
    }
901
    v->extended_mv =  get_bits1(gb); //common
902
    if (!v->profile && v->extended_mv)
903
    {
904
        av_log(avctx, AV_LOG_ERROR,
905
               "Extended MVs unavailable in Simple Profile\n");
906
        return -1;
907
    }
908
    v->dquant =  get_bits(gb, 2); //common
909
    v->vstransform =  get_bits1(gb); //common
910

    
911
    v->res_transtab = get_bits1(gb);
912
    if (v->res_transtab)
913
    {
914
        av_log(avctx, AV_LOG_ERROR,
915
               "1 for reserved RES_TRANSTAB is forbidden\n");
916
        return -1;
917
    }
918

    
919
    v->overlap = get_bits1(gb); //common
920

    
921
    v->s.resync_marker = get_bits1(gb);
922
    v->rangered = get_bits1(gb);
923
    if (v->rangered && v->profile == PROFILE_SIMPLE)
924
    {
925
        av_log(avctx, AV_LOG_INFO,
926
               "RANGERED should be set to 0 in simple profile\n");
927
    }
928

    
929
    v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common
930
    v->quantizer_mode = get_bits(gb, 2); //common
931

    
932
    v->finterpflag = get_bits1(gb); //common
933
    v->res_rtm_flag = get_bits1(gb); //reserved
934
    if (!v->res_rtm_flag)
935
    {
936
//            av_log(avctx, AV_LOG_ERROR,
937
//                   "0 for reserved RES_RTM_FLAG is forbidden\n");
938
        av_log(avctx, AV_LOG_ERROR,
939
               "Old WMV3 version detected, only I-frames will be decoded\n");
940
        //return -1;
941
    }
942
    //TODO: figure out what they mean (always 0x402F)
943
    if(!v->res_fasttx) skip_bits(gb, 16);
944
    av_log(avctx, AV_LOG_DEBUG,
945
               "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
946
               "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"
947
               "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"
948
               "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",
949
               v->profile, v->frmrtq_postproc, v->bitrtq_postproc,
950
               v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,
951
               v->rangered, v->vstransform, v->overlap, v->s.resync_marker,
952
               v->dquant, v->quantizer_mode, avctx->max_b_frames
953
               );
954
    return 0;
955
}
956

    
957
static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)
958
{
959
    v->res_rtm_flag = 1;
960
    v->level = get_bits(gb, 3);
961
    if(v->level >= 5)
962
    {
963
        av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);
964
    }
965
    v->chromaformat = get_bits(gb, 2);
966
    if (v->chromaformat != 1)
967
    {
968
        av_log(v->s.avctx, AV_LOG_ERROR,
969
               "Only 4:2:0 chroma format supported\n");
970
        return -1;
971
    }
972

    
973
    // (fps-2)/4 (->30)
974
    v->frmrtq_postproc = get_bits(gb, 3); //common
975
    // (bitrate-32kbps)/64kbps
976
    v->bitrtq_postproc = get_bits(gb, 5); //common
977
    v->postprocflag = get_bits1(gb); //common
978

    
979
    v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;
980
    v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;
981
    v->s.avctx->width = v->s.avctx->coded_width;
982
    v->s.avctx->height = v->s.avctx->coded_height;
983
    v->broadcast = get_bits1(gb);
984
    v->interlace = get_bits1(gb);
985
    v->tfcntrflag = get_bits1(gb);
986
    v->finterpflag = get_bits1(gb);
987
    skip_bits1(gb); // reserved
988

    
989
    v->s.h_edge_pos = v->s.avctx->coded_width;
990
    v->s.v_edge_pos = v->s.avctx->coded_height;
991

    
992
    av_log(v->s.avctx, AV_LOG_DEBUG,
993
               "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
994
               "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"
995
               "TFCTRflag=%i, FINTERPflag=%i\n",
996
               v->level, v->frmrtq_postproc, v->bitrtq_postproc,
997
               v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,
998
               v->tfcntrflag, v->finterpflag
999
               );
1000

    
1001
    v->psf = get_bits1(gb);
1002
    if(v->psf) { //PsF, 6.1.13
1003
        av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");
1004
        return -1;
1005
    }
1006
    v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
1007
    if(get_bits1(gb)) { //Display Info - decoding is not affected by it
1008
        int w, h, ar = 0;
1009
        av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
1010
        v->s.avctx->width  = v->s.width  = w = get_bits(gb, 14) + 1;
1011
        v->s.avctx->height = v->s.height = h = get_bits(gb, 14) + 1;
1012
        av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
1013
        if(get_bits1(gb))
1014
            ar = get_bits(gb, 4);
1015
        if(ar && ar < 14){
1016
            v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
1017
        }else if(ar == 15){
1018
            w = get_bits(gb, 8);
1019
            h = get_bits(gb, 8);
1020
            v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
1021
        }
1022

    
1023
        if(get_bits1(gb)){ //framerate stuff
1024
            if(get_bits1(gb)) {
1025
                v->s.avctx->time_base.num = 32;
1026
                v->s.avctx->time_base.den = get_bits(gb, 16) + 1;
1027
            } else {
1028
                int nr, dr;
1029
                nr = get_bits(gb, 8);
1030
                dr = get_bits(gb, 4);
1031
                if(nr && nr < 8 && dr && dr < 3){
1032
                    v->s.avctx->time_base.num = ff_vc1_fps_dr[dr - 1];
1033
                    v->s.avctx->time_base.den = ff_vc1_fps_nr[nr - 1] * 1000;
1034
                }
1035
            }
1036
        }
1037

    
1038
        if(get_bits1(gb)){
1039
            v->color_prim = get_bits(gb, 8);
1040
            v->transfer_char = get_bits(gb, 8);
1041
            v->matrix_coef = get_bits(gb, 8);
1042
        }
1043
    }
1044

    
1045
    v->hrd_param_flag = get_bits1(gb);
1046
    if(v->hrd_param_flag) {
1047
        int i;
1048
        v->hrd_num_leaky_buckets = get_bits(gb, 5);
1049
        skip_bits(gb, 4); //bitrate exponent
1050
        skip_bits(gb, 4); //buffer size exponent
1051
        for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1052
            skip_bits(gb, 16); //hrd_rate[n]
1053
            skip_bits(gb, 16); //hrd_buffer[n]
1054
        }
1055
    }
1056
    return 0;
1057
}
1058

    
1059
static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb)
1060
{
1061
    VC1Context *v = avctx->priv_data;
1062
    int i, blink, clentry;
1063

    
1064
    av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32));
1065
    blink = get_bits1(gb); // broken link
1066
    clentry = get_bits1(gb); // closed entry
1067
    v->panscanflag = get_bits1(gb);
1068
    v->refdist_flag = get_bits1(gb);
1069
    v->s.loop_filter = get_bits1(gb);
1070
    v->fastuvmc = get_bits1(gb);
1071
    v->extended_mv = get_bits1(gb);
1072
    v->dquant = get_bits(gb, 2);
1073
    v->vstransform = get_bits1(gb);
1074
    v->overlap = get_bits1(gb);
1075
    v->quantizer_mode = get_bits(gb, 2);
1076

    
1077
    if(v->hrd_param_flag){
1078
        for(i = 0; i < v->hrd_num_leaky_buckets; i++) {
1079
            skip_bits(gb, 8); //hrd_full[n]
1080
        }
1081
    }
1082

    
1083
    if(get_bits1(gb)){
1084
        avctx->coded_width = (get_bits(gb, 12)+1)<<1;
1085
        avctx->coded_height = (get_bits(gb, 12)+1)<<1;
1086
    }
1087
    if(v->extended_mv)
1088
        v->extended_dmv = get_bits1(gb);
1089
    if((v->range_mapy_flag = get_bits1(gb))) {
1090
        av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n");
1091
        v->range_mapy = get_bits(gb, 3);
1092
    }
1093
    if((v->range_mapuv_flag = get_bits1(gb))) {
1094
        av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n");
1095
        v->range_mapuv = get_bits(gb, 3);
1096
    }
1097

    
1098
    av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n"
1099
        "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n"
1100
        "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n"
1101
        "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n",
1102
        blink, clentry, v->panscanflag, v->refdist_flag, v->s.loop_filter,
1103
        v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode);
1104

    
1105
    return 0;
1106
}
1107

    
1108
static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)
1109
{
1110
    int pqindex, lowquant, status;
1111

    
1112
    if(v->finterpflag) v->interpfrm = get_bits1(gb);
1113
    skip_bits(gb, 2); //framecnt unused
1114
    v->rangeredfrm = 0;
1115
    if (v->rangered) v->rangeredfrm = get_bits1(gb);
1116
    v->s.pict_type = get_bits1(gb);
1117
    if (v->s.avctx->max_b_frames) {
1118
        if (!v->s.pict_type) {
1119
            if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE;
1120
            else v->s.pict_type = FF_B_TYPE;
1121
        } else v->s.pict_type = FF_P_TYPE;
1122
    } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE;
1123

    
1124
    v->bi_type = 0;
1125
    if(v->s.pict_type == FF_B_TYPE) {
1126
        v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1127
        v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1128
        if(v->bfraction == 0) {
1129
            v->s.pict_type = FF_BI_TYPE;
1130
        }
1131
    }
1132
    if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1133
        skip_bits(gb, 7); // skip buffer fullness
1134

    
1135
    /* calculate RND */
1136
    if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1137
        v->rnd = 1;
1138
    if(v->s.pict_type == FF_P_TYPE)
1139
        v->rnd ^= 1;
1140

    
1141
    /* Quantizer stuff */
1142
    pqindex = get_bits(gb, 5);
1143
    if(!pqindex) return -1;
1144
    if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1145
        v->pq = ff_vc1_pquant_table[0][pqindex];
1146
    else
1147
        v->pq = ff_vc1_pquant_table[1][pqindex];
1148

    
1149
    v->pquantizer = 1;
1150
    if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1151
        v->pquantizer = pqindex < 9;
1152
    if (v->quantizer_mode == QUANT_NON_UNIFORM)
1153
        v->pquantizer = 0;
1154
    v->pqindex = pqindex;
1155
    if (pqindex < 9) v->halfpq = get_bits1(gb);
1156
    else v->halfpq = 0;
1157
    if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1158
        v->pquantizer = get_bits1(gb);
1159
    v->dquantfrm = 0;
1160
    if (v->extended_mv == 1) v->mvrange = get_unary(gb, 0, 3);
1161
    v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1162
    v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1163
    v->range_x = 1 << (v->k_x - 1);
1164
    v->range_y = 1 << (v->k_y - 1);
1165
    if (v->profile == PROFILE_ADVANCED)
1166
    {
1167
        if (v->postprocflag) v->postproc = get_bits1(gb);
1168
    }
1169
    else
1170
        if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2);
1171

    
1172
    if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){
1173
        v->x8_type = get_bits1(gb);
1174
    }else v->x8_type = 0;
1175
//av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",
1176
//        (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);
1177

    
1178
    if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1179

    
1180
    switch(v->s.pict_type) {
1181
    case FF_P_TYPE:
1182
        if (v->pq < 5) v->tt_index = 0;
1183
        else if(v->pq < 13) v->tt_index = 1;
1184
        else v->tt_index = 2;
1185

    
1186
        lowquant = (v->pq > 12) ? 0 : 1;
1187
        v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1188
        if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1189
        {
1190
            int scale, shift, i;
1191
            v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1192
            v->lumscale = get_bits(gb, 6);
1193
            v->lumshift = get_bits(gb, 6);
1194
            v->use_ic = 1;
1195
            /* fill lookup tables for intensity compensation */
1196
            if(!v->lumscale) {
1197
                scale = -64;
1198
                shift = (255 - v->lumshift * 2) << 6;
1199
                if(v->lumshift > 31)
1200
                    shift += 128 << 6;
1201
            } else {
1202
                scale = v->lumscale + 32;
1203
                if(v->lumshift > 31)
1204
                    shift = (v->lumshift - 64) << 6;
1205
                else
1206
                    shift = v->lumshift << 6;
1207
            }
1208
            for(i = 0; i < 256; i++) {
1209
                v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1210
                v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1211
            }
1212
        }
1213
        if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1214
            v->s.quarter_sample = 0;
1215
        else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1216
            if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1217
                v->s.quarter_sample = 0;
1218
            else
1219
                v->s.quarter_sample = 1;
1220
        } else
1221
            v->s.quarter_sample = 1;
1222
        v->s.mspel = !(v->mv_mode == MV_PMODE_1MV_HPEL_BILIN || (v->mv_mode == MV_PMODE_INTENSITY_COMP && v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN));
1223

    
1224
        if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1225
                 v->mv_mode2 == MV_PMODE_MIXED_MV)
1226
                || v->mv_mode == MV_PMODE_MIXED_MV)
1227
        {
1228
            status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1229
            if (status < 0) return -1;
1230
            av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1231
                   "Imode: %i, Invert: %i\n", status>>1, status&1);
1232
        } else {
1233
            v->mv_type_is_raw = 0;
1234
            memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1235
        }
1236
        status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1237
        if (status < 0) return -1;
1238
        av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1239
               "Imode: %i, Invert: %i\n", status>>1, status&1);
1240

    
1241
        /* Hopefully this is correct for P frames */
1242
        v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1243
        v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1244

    
1245
        if (v->dquant)
1246
        {
1247
            av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1248
            vop_dquant_decoding(v);
1249
        }
1250

    
1251
        v->ttfrm = 0; //FIXME Is that so ?
1252
        if (v->vstransform)
1253
        {
1254
            v->ttmbf = get_bits1(gb);
1255
            if (v->ttmbf)
1256
            {
1257
                v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1258
            }
1259
        } else {
1260
            v->ttmbf = 1;
1261
            v->ttfrm = TT_8X8;
1262
        }
1263
        break;
1264
    case FF_B_TYPE:
1265
        if (v->pq < 5) v->tt_index = 0;
1266
        else if(v->pq < 13) v->tt_index = 1;
1267
        else v->tt_index = 2;
1268

    
1269
        lowquant = (v->pq > 12) ? 0 : 1;
1270
        v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1271
        v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1272
        v->s.mspel = v->s.quarter_sample;
1273

    
1274
        status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1275
        if (status < 0) return -1;
1276
        av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1277
               "Imode: %i, Invert: %i\n", status>>1, status&1);
1278
        status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1279
        if (status < 0) return -1;
1280
        av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1281
               "Imode: %i, Invert: %i\n", status>>1, status&1);
1282

    
1283
        v->s.mv_table_index = get_bits(gb, 2);
1284
        v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1285

    
1286
        if (v->dquant)
1287
        {
1288
            av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1289
            vop_dquant_decoding(v);
1290
        }
1291

    
1292
        v->ttfrm = 0;
1293
        if (v->vstransform)
1294
        {
1295
            v->ttmbf = get_bits1(gb);
1296
            if (v->ttmbf)
1297
            {
1298
                v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1299
            }
1300
        } else {
1301
            v->ttmbf = 1;
1302
            v->ttfrm = TT_8X8;
1303
        }
1304
        break;
1305
    }
1306

    
1307
    if(!v->x8_type)
1308
    {
1309
        /* AC Syntax */
1310
        v->c_ac_table_index = decode012(gb);
1311
        if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1312
        {
1313
            v->y_ac_table_index = decode012(gb);
1314
        }
1315
        /* DC Syntax */
1316
        v->s.dc_table_index = get_bits1(gb);
1317
    }
1318

    
1319
    if(v->s.pict_type == FF_BI_TYPE) {
1320
        v->s.pict_type = FF_B_TYPE;
1321
        v->bi_type = 1;
1322
    }
1323
    return 0;
1324
}
1325

    
1326
static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)
1327
{
1328
    int pqindex, lowquant;
1329
    int status;
1330

    
1331
    v->p_frame_skipped = 0;
1332

    
1333
    if(v->interlace){
1334
        v->fcm = decode012(gb);
1335
        if(v->fcm) return -1; // interlaced frames/fields are not implemented
1336
    }
1337
    switch(get_unary(gb, 0, 4)) {
1338
    case 0:
1339
        v->s.pict_type = FF_P_TYPE;
1340
        break;
1341
    case 1:
1342
        v->s.pict_type = FF_B_TYPE;
1343
        break;
1344
    case 2:
1345
        v->s.pict_type = FF_I_TYPE;
1346
        break;
1347
    case 3:
1348
        v->s.pict_type = FF_BI_TYPE;
1349
        break;
1350
    case 4:
1351
        v->s.pict_type = FF_P_TYPE; // skipped pic
1352
        v->p_frame_skipped = 1;
1353
        return 0;
1354
    }
1355
    if(v->tfcntrflag)
1356
        skip_bits(gb, 8);
1357
    if(v->broadcast) {
1358
        if(!v->interlace || v->psf) {
1359
            v->rptfrm = get_bits(gb, 2);
1360
        } else {
1361
            v->tff = get_bits1(gb);
1362
            v->rptfrm = get_bits1(gb);
1363
        }
1364
    }
1365
    if(v->panscanflag) {
1366
        //...
1367
    }
1368
    v->rnd = get_bits1(gb);
1369
    if(v->interlace)
1370
        v->uvsamp = get_bits1(gb);
1371
    if(v->finterpflag) v->interpfrm = get_bits1(gb);
1372
    if(v->s.pict_type == FF_B_TYPE) {
1373
        v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);
1374
        v->bfraction = ff_vc1_bfraction_lut[v->bfraction];
1375
        if(v->bfraction == 0) {
1376
            v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */
1377
        }
1378
    }
1379
    pqindex = get_bits(gb, 5);
1380
    if(!pqindex) return -1;
1381
    v->pqindex = pqindex;
1382
    if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1383
        v->pq = ff_vc1_pquant_table[0][pqindex];
1384
    else
1385
        v->pq = ff_vc1_pquant_table[1][pqindex];
1386

    
1387
    v->pquantizer = 1;
1388
    if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)
1389
        v->pquantizer = pqindex < 9;
1390
    if (v->quantizer_mode == QUANT_NON_UNIFORM)
1391
        v->pquantizer = 0;
1392
    v->pqindex = pqindex;
1393
    if (pqindex < 9) v->halfpq = get_bits1(gb);
1394
    else v->halfpq = 0;
1395
    if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)
1396
        v->pquantizer = get_bits1(gb);
1397
    if(v->postprocflag)
1398
        v->postproc = get_bits1(gb);
1399

    
1400
    if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;
1401

    
1402
    switch(v->s.pict_type) {
1403
    case FF_I_TYPE:
1404
    case FF_BI_TYPE:
1405
        status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v);
1406
        if (status < 0) return -1;
1407
        av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: "
1408
                "Imode: %i, Invert: %i\n", status>>1, status&1);
1409
        v->condover = CONDOVER_NONE;
1410
        if(v->overlap && v->pq <= 8) {
1411
            v->condover = decode012(gb);
1412
            if(v->condover == CONDOVER_SELECT) {
1413
                status = bitplane_decoding(v->over_flags_plane, &v->overflg_is_raw, v);
1414
                if (status < 0) return -1;
1415
                av_log(v->s.avctx, AV_LOG_DEBUG, "CONDOVER plane encoding: "
1416
                        "Imode: %i, Invert: %i\n", status>>1, status&1);
1417
            }
1418
        }
1419
        break;
1420
    case FF_P_TYPE:
1421
        if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1422
        else v->mvrange = 0;
1423
        v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1424
        v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1425
        v->range_x = 1 << (v->k_x - 1);
1426
        v->range_y = 1 << (v->k_y - 1);
1427

    
1428
        if (v->pq < 5) v->tt_index = 0;
1429
        else if(v->pq < 13) v->tt_index = 1;
1430
        else v->tt_index = 2;
1431

    
1432
        lowquant = (v->pq > 12) ? 0 : 1;
1433
        v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];
1434
        if (v->mv_mode == MV_PMODE_INTENSITY_COMP)
1435
        {
1436
            int scale, shift, i;
1437
            v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];
1438
            v->lumscale = get_bits(gb, 6);
1439
            v->lumshift = get_bits(gb, 6);
1440
            /* fill lookup tables for intensity compensation */
1441
            if(!v->lumscale) {
1442
                scale = -64;
1443
                shift = (255 - v->lumshift * 2) << 6;
1444
                if(v->lumshift > 31)
1445
                    shift += 128 << 6;
1446
            } else {
1447
                scale = v->lumscale + 32;
1448
                if(v->lumshift > 31)
1449
                    shift = (v->lumshift - 64) << 6;
1450
                else
1451
                    shift = v->lumshift << 6;
1452
            }
1453
            for(i = 0; i < 256; i++) {
1454
                v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);
1455
                v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);
1456
            }
1457
            v->use_ic = 1;
1458
        }
1459
        if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
1460
            v->s.quarter_sample = 0;
1461
        else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
1462
            if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)
1463
                v->s.quarter_sample = 0;
1464
            else
1465
                v->s.quarter_sample = 1;
1466
        } else
1467
            v->s.quarter_sample = 1;
1468
        v->s.mspel = !(v->mv_mode == MV_PMODE_1MV_HPEL_BILIN || (v->mv_mode == MV_PMODE_INTENSITY_COMP && v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN));
1469

    
1470
        if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
1471
                 v->mv_mode2 == MV_PMODE_MIXED_MV)
1472
                || v->mv_mode == MV_PMODE_MIXED_MV)
1473
        {
1474
            status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
1475
            if (status < 0) return -1;
1476
            av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
1477
                   "Imode: %i, Invert: %i\n", status>>1, status&1);
1478
        } else {
1479
            v->mv_type_is_raw = 0;
1480
            memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
1481
        }
1482
        status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1483
        if (status < 0) return -1;
1484
        av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1485
               "Imode: %i, Invert: %i\n", status>>1, status&1);
1486

    
1487
        /* Hopefully this is correct for P frames */
1488
        v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables
1489
        v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1490
        if (v->dquant)
1491
        {
1492
            av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1493
            vop_dquant_decoding(v);
1494
        }
1495

    
1496
        v->ttfrm = 0; //FIXME Is that so ?
1497
        if (v->vstransform)
1498
        {
1499
            v->ttmbf = get_bits1(gb);
1500
            if (v->ttmbf)
1501
            {
1502
                v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1503
            }
1504
        } else {
1505
            v->ttmbf = 1;
1506
            v->ttfrm = TT_8X8;
1507
        }
1508
        break;
1509
    case FF_B_TYPE:
1510
        if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);
1511
        else v->mvrange = 0;
1512
        v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13
1513
        v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11
1514
        v->range_x = 1 << (v->k_x - 1);
1515
        v->range_y = 1 << (v->k_y - 1);
1516

    
1517
        if (v->pq < 5) v->tt_index = 0;
1518
        else if(v->pq < 13) v->tt_index = 1;
1519
        else v->tt_index = 2;
1520

    
1521
        lowquant = (v->pq > 12) ? 0 : 1;
1522
        v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;
1523
        v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);
1524
        v->s.mspel = v->s.quarter_sample;
1525

    
1526
        status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);
1527
        if (status < 0) return -1;
1528
        av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "
1529
               "Imode: %i, Invert: %i\n", status>>1, status&1);
1530
        status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);
1531
        if (status < 0) return -1;
1532
        av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
1533
               "Imode: %i, Invert: %i\n", status>>1, status&1);
1534

    
1535
        v->s.mv_table_index = get_bits(gb, 2);
1536
        v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];
1537

    
1538
        if (v->dquant)
1539
        {
1540
            av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1541
            vop_dquant_decoding(v);
1542
        }
1543

    
1544
        v->ttfrm = 0;
1545
        if (v->vstransform)
1546
        {
1547
            v->ttmbf = get_bits1(gb);
1548
            if (v->ttmbf)
1549
            {
1550
                v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];
1551
            }
1552
        } else {
1553
            v->ttmbf = 1;
1554
            v->ttfrm = TT_8X8;
1555
        }
1556
        break;
1557
    }
1558

    
1559
    /* AC Syntax */
1560
    v->c_ac_table_index = decode012(gb);
1561
    if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)
1562
    {
1563
        v->y_ac_table_index = decode012(gb);
1564
    }
1565
    /* DC Syntax */
1566
    v->s.dc_table_index = get_bits1(gb);
1567
    if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) {
1568
        av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");
1569
        vop_dquant_decoding(v);
1570
    }
1571

    
1572
    v->bi_type = 0;
1573
    if(v->s.pict_type == FF_BI_TYPE) {
1574
        v->s.pict_type = FF_B_TYPE;
1575
        v->bi_type = 1;
1576
    }
1577
    return 0;
1578
}
1579

    
1580
/***********************************************************************/
1581
/**
1582
 * @defgroup block VC-1 Block-level functions
1583
 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
1584
 * @{
1585
 */
1586

    
1587
/**
1588
 * @def GET_MQUANT
1589
 * @brief Get macroblock-level quantizer scale
1590
 */
1591
#define GET_MQUANT()                                           \
1592
  if (v->dquantfrm)                                            \
1593
  {                                                            \
1594
    int edges = 0;                                             \
1595
    if (v->dqprofile == DQPROFILE_ALL_MBS)                     \
1596
    {                                                          \
1597
      if (v->dqbilevel)                                        \
1598
      {                                                        \
1599
        mquant = (get_bits1(gb)) ? v->altpq : v->pq;           \
1600
      }                                                        \
1601
      else                                                     \
1602
      {                                                        \
1603
        mqdiff = get_bits(gb, 3);                              \
1604
        if (mqdiff != 7) mquant = v->pq + mqdiff;              \
1605
        else mquant = get_bits(gb, 5);                         \
1606
      }                                                        \
1607
    }                                                          \
1608
    if(v->dqprofile == DQPROFILE_SINGLE_EDGE)                  \
1609
        edges = 1 << v->dqsbedge;                              \
1610
    else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES)            \
1611
        edges = (3 << v->dqsbedge) % 15;                       \
1612
    else if(v->dqprofile == DQPROFILE_FOUR_EDGES)              \
1613
        edges = 15;                                            \
1614
    if((edges&1) && !s->mb_x)                                  \
1615
        mquant = v->altpq;                                     \
1616
    if((edges&2) && s->first_slice_line)                       \
1617
        mquant = v->altpq;                                     \
1618
    if((edges&4) && s->mb_x == (s->mb_width - 1))              \
1619
        mquant = v->altpq;                                     \
1620
    if((edges&8) && s->mb_y == (s->mb_height - 1))             \
1621
        mquant = v->altpq;                                     \
1622
  }
1623

    
1624
/**
1625
 * @def GET_MVDATA(_dmv_x, _dmv_y)
1626
 * @brief Get MV differentials
1627
 * @see MVDATA decoding from 8.3.5.2, p(1)20
1628
 * @param _dmv_x Horizontal differential for decoded MV
1629
 * @param _dmv_y Vertical differential for decoded MV
1630
 */
1631
#define GET_MVDATA(_dmv_x, _dmv_y)                                  \
1632
  index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table,\
1633
                       VC1_MV_DIFF_VLC_BITS, 2);                    \
1634
  if (index > 36)                                                   \
1635
  {                                                                 \
1636
    mb_has_coeffs = 1;                                              \
1637
    index -= 37;                                                    \
1638
  }                                                                 \
1639
  else mb_has_coeffs = 0;                                           \
1640
  s->mb_intra = 0;                                                  \
1641
  if (!index) { _dmv_x = _dmv_y = 0; }                              \
1642
  else if (index == 35)                                             \
1643
  {                                                                 \
1644
    _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \
1645
    _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \
1646
  }                                                                 \
1647
  else if (index == 36)                                             \
1648
  {                                                                 \
1649
    _dmv_x = 0;                                                     \
1650
    _dmv_y = 0;                                                     \
1651
    s->mb_intra = 1;                                                \
1652
  }                                                                 \
1653
  else                                                              \
1654
  {                                                                 \
1655
    index1 = index%6;                                               \
1656
    if (!s->quarter_sample && index1 == 5) val = 1;                 \
1657
    else                                   val = 0;                 \
1658
    if(size_table[index1] - val > 0)                                \
1659
        val = get_bits(gb, size_table[index1] - val);               \
1660
    else                                   val = 0;                 \
1661
    sign = 0 - (val&1);                                             \
1662
    _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \
1663
                                                                    \
1664
    index1 = index/6;                                               \
1665
    if (!s->quarter_sample && index1 == 5) val = 1;                 \
1666
    else                                   val = 0;                 \
1667
    if(size_table[index1] - val > 0)                                \
1668
        val = get_bits(gb, size_table[index1] - val);               \
1669
    else                                   val = 0;                 \
1670
    sign = 0 - (val&1);                                             \
1671
    _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \
1672
  }
1673

    
1674
/** Predict and set motion vector
1675
 */
1676
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)
1677
{
1678
    int xy, wrap, off = 0;
1679
    int16_t *A, *B, *C;
1680
    int px, py;
1681
    int sum;
1682

    
1683
    /* scale MV difference to be quad-pel */
1684
    dmv_x <<= 1 - s->quarter_sample;
1685
    dmv_y <<= 1 - s->quarter_sample;
1686

    
1687
    wrap = s->b8_stride;
1688
    xy = s->block_index[n];
1689

    
1690
    if(s->mb_intra){
1691
        s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;
1692
        s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;
1693
        s->current_picture.motion_val[1][xy][0] = 0;
1694
        s->current_picture.motion_val[1][xy][1] = 0;
1695
        if(mv1) { /* duplicate motion data for 1-MV block */
1696
            s->current_picture.motion_val[0][xy + 1][0] = 0;
1697
            s->current_picture.motion_val[0][xy + 1][1] = 0;
1698
            s->current_picture.motion_val[0][xy + wrap][0] = 0;
1699
            s->current_picture.motion_val[0][xy + wrap][1] = 0;
1700
            s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;
1701
            s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;
1702
            s->current_picture.motion_val[1][xy + 1][0] = 0;
1703
            s->current_picture.motion_val[1][xy + 1][1] = 0;
1704
            s->current_picture.motion_val[1][xy + wrap][0] = 0;
1705
            s->current_picture.motion_val[1][xy + wrap][1] = 0;
1706
            s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;
1707
            s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;
1708
        }
1709
        return;
1710
    }
1711

    
1712
    C = s->current_picture.motion_val[0][xy - 1];
1713
    A = s->current_picture.motion_val[0][xy - wrap];
1714
    if(mv1)
1715
        off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;
1716
    else {
1717
        //in 4-MV mode different blocks have different B predictor position
1718
        switch(n){
1719
        case 0:
1720
            off = (s->mb_x > 0) ? -1 : 1;
1721
            break;
1722
        case 1:
1723
            off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;
1724
            break;
1725
        case 2:
1726
            off = 1;
1727
            break;
1728
        case 3:
1729
            off = -1;
1730
        }
1731
    }
1732
    B = s->current_picture.motion_val[0][xy - wrap + off];
1733

    
1734
    if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds
1735
        if(s->mb_width == 1) {
1736
            px = A[0];
1737
            py = A[1];
1738
        } else {
1739
            px = mid_pred(A[0], B[0], C[0]);
1740
            py = mid_pred(A[1], B[1], C[1]);
1741
        }
1742
    } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds
1743
        px = C[0];
1744
        py = C[1];
1745
    } else {
1746
        px = py = 0;
1747
    }
1748
    /* Pullback MV as specified in 8.3.5.3.4 */
1749
    {
1750
        int qx, qy, X, Y;
1751
        qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);
1752
        qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);
1753
        X = (s->mb_width << 6) - 4;
1754
        Y = (s->mb_height << 6) - 4;
1755
        if(mv1) {
1756
            if(qx + px < -60) px = -60 - qx;
1757
            if(qy + py < -60) py = -60 - qy;
1758
        } else {
1759
            if(qx + px < -28) px = -28 - qx;
1760
            if(qy + py < -28) py = -28 - qy;
1761
        }
1762
        if(qx + px > X) px = X - qx;
1763
        if(qy + py > Y) py = Y - qy;
1764
    }
1765
    /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
1766
    if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {
1767
        if(is_intra[xy - wrap])
1768
            sum = FFABS(px) + FFABS(py);
1769
        else
1770
            sum = FFABS(px - A[0]) + FFABS(py - A[1]);
1771
        if(sum > 32) {
1772
            if(get_bits1(&s->gb)) {
1773
                px = A[0];
1774
                py = A[1];
1775
            } else {
1776
                px = C[0];
1777
                py = C[1];
1778
            }
1779
        } else {
1780
            if(is_intra[xy - 1])
1781
                sum = FFABS(px) + FFABS(py);
1782
            else
1783
                sum = FFABS(px - C[0]) + FFABS(py - C[1]);
1784
            if(sum > 32) {
1785
                if(get_bits1(&s->gb)) {
1786
                    px = A[0];
1787
                    py = A[1];
1788
                } else {
1789
                    px = C[0];
1790
                    py = C[1];
1791
                }
1792
            }
1793
        }
1794
    }
1795
    /* store MV using signed modulus of MV range defined in 4.11 */
1796
    s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;
1797
    s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;
1798
    if(mv1) { /* duplicate motion data for 1-MV block */
1799
        s->current_picture.motion_val[0][xy + 1][0] = s->current_picture.motion_val[0][xy][0];
1800
        s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];
1801
        s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];
1802
        s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];
1803
        s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];
1804
        s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];
1805
    }
1806
}
1807

    
1808
/** Motion compensation for direct or interpolated blocks in B-frames
1809
 */
1810
static void vc1_interp_mc(VC1Context *v)
1811
{
1812
    MpegEncContext *s = &v->s;
1813
    DSPContext *dsp = &v->s.dsp;
1814
    uint8_t *srcY, *srcU, *srcV;
1815
    int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
1816

    
1817
    if(!v->s.next_picture.data[0])return;
1818

    
1819
    mx = s->mv[1][0][0];
1820
    my = s->mv[1][0][1];
1821
    uvmx = (mx + ((mx & 3) == 3)) >> 1;
1822
    uvmy = (my + ((my & 3) == 3)) >> 1;
1823
    if(v->fastuvmc) {
1824
        uvmx = uvmx + ((uvmx<0)?-(uvmx&1):(uvmx&1));
1825
        uvmy = uvmy + ((uvmy<0)?-(uvmy&1):(uvmy&1));
1826
    }
1827
    srcY = s->next_picture.data[0];
1828
    srcU = s->next_picture.data[1];
1829
    srcV = s->next_picture.data[2];
1830

    
1831
    src_x = s->mb_x * 16 + (mx >> 2);
1832
    src_y = s->mb_y * 16 + (my >> 2);
1833
    uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
1834
    uvsrc_y = s->mb_y * 8 + (uvmy >> 2);
1835

    
1836
    if(v->profile != PROFILE_ADVANCED){
1837
        src_x   = av_clip(  src_x, -16, s->mb_width  * 16);
1838
        src_y   = av_clip(  src_y, -16, s->mb_height * 16);
1839
        uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);
1840
        uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);
1841
    }else{
1842
        src_x   = av_clip(  src_x, -17, s->avctx->coded_width);
1843
        src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);
1844
        uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);
1845
        uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);
1846
    }
1847

    
1848
    srcY += src_y * s->linesize + src_x;
1849
    srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
1850
    srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
1851

    
1852
    /* for grayscale we should not try to read from unknown area */
1853
    if(s->flags & CODEC_FLAG_GRAY) {
1854
        srcU = s->edge_emu_buffer + 18 * s->linesize;
1855
        srcV = s->edge_emu_buffer + 18 * s->linesize;
1856
    }
1857

    
1858
    if(v->rangeredfrm
1859
       || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16
1860
       || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){
1861
        uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;
1862

    
1863
        srcY -= s->mspel * (1 + s->linesize);
1864
        ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
1865
                            src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
1866
        srcY = s->edge_emu_buffer;
1867
        ff_emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8+1, 8+1,
1868
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1869
        ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
1870
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
1871
        srcU = uvbuf;
1872
        srcV = uvbuf + 16;
1873
        /* if we deal with range reduction we need to scale source blocks */
1874
        if(v->rangeredfrm) {
1875
            int i, j;
1876
            uint8_t *src, *src2;
1877

    
1878
            src = srcY;
1879
            for(j = 0; j < 17 + s->mspel*2; j++) {
1880
                for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
1881
                src += s->linesize;
1882
            }
1883
            src = srcU; src2 = srcV;
1884
            for(j = 0; j < 9; j++) {
1885
                for(i = 0; i < 9; i++) {
1886
                    src[i] = ((src[i] - 128) >> 1) + 128;
1887
                    src2[i] = ((src2[i] - 128) >> 1) + 128;
1888
                }
1889
                src += s->uvlinesize;
1890
                src2 += s->uvlinesize;
1891
            }
1892
        }
1893
        srcY += s->mspel * (1 + s->linesize);
1894
    }
1895

    
1896
    mx >>= 1;
1897
    my >>= 1;
1898
    dxy = ((my & 1) << 1) | (mx & 1);
1899

    
1900
    dsp->avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
1901

    
1902
    if(s->flags & CODEC_FLAG_GRAY) return;
1903
    /* Chroma MC always uses qpel blilinear */
1904
    uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
1905
    uvmx = (uvmx&3)<<1;
1906
    uvmy = (uvmy&3)<<1;
1907
    dsp->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
1908
    dsp->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
1909
}
1910

    
1911
static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)
1912
{
1913
    int n = bfrac;
1914

    
1915
#if B_FRACTION_DEN==256
1916
    if(inv)
1917
        n -= 256;
1918
    if(!qs)
1919
        return 2 * ((value * n + 255) >> 9);
1920
    return (value * n + 128) >> 8;
1921
#else
1922
    if(inv)
1923
        n -= B_FRACTION_DEN;
1924
    if(!qs)
1925
        return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));
1926
    return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;
1927
#endif
1928
}
1929

    
1930
/** Reconstruct motion vector for B-frame and do motion compensation
1931
 */
1932
static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode)
1933
{
1934
    if(v->use_ic) {
1935
        v->mv_mode2 = v->mv_mode;
1936
        v->mv_mode = MV_PMODE_INTENSITY_COMP;
1937
    }
1938
    if(direct) {
1939
        vc1_mc_1mv(v, 0);
1940
        vc1_interp_mc(v);
1941
        if(v->use_ic) v->mv_mode = v->mv_mode2;
1942
        return;
1943
    }
1944
    if(mode == BMV_TYPE_INTERPOLATED) {
1945
        vc1_mc_1mv(v, 0);
1946
        vc1_interp_mc(v);
1947
        if(v->use_ic) v->mv_mode = v->mv_mode2;
1948
        return;
1949
    }
1950

    
1951
    if(v->use_ic && (mode == BMV_TYPE_BACKWARD)) v->mv_mode = v->mv_mode2;
1952
    vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));
1953
    if(v->use_ic) v->mv_mode = v->mv_mode2;
1954
}
1955

    
1956
static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mvtype)
1957
{
1958
    MpegEncContext *s = &v->s;
1959
    int xy, wrap, off = 0;
1960
    int16_t *A, *B, *C;
1961
    int px, py;
1962
    int sum;
1963
    int r_x, r_y;
1964
    const uint8_t *is_intra = v->mb_type[0];
1965

    
1966
    r_x = v->range_x;
1967
    r_y = v->range_y;
1968
    /* scale MV difference to be quad-pel */
1969
    dmv_x[0] <<= 1 - s->quarter_sample;
1970
    dmv_y[0] <<= 1 - s->quarter_sample;
1971
    dmv_x[1] <<= 1 - s->quarter_sample;
1972
    dmv_y[1] <<= 1 - s->quarter_sample;
1973

    
1974
    wrap = s->b8_stride;
1975
    xy = s->block_index[0];
1976

    
1977
    if(s->mb_intra) {
1978
        s->current_picture.motion_val[0][xy][0] =
1979
        s->current_picture.motion_val[0][xy][1] =
1980
        s->current_picture.motion_val[1][xy][0] =
1981
        s->current_picture.motion_val[1][xy][1] = 0;
1982
        return;
1983
    }
1984
    s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);
1985
    s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);
1986
    s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);
1987
    s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);
1988

    
1989
    /* Pullback predicted motion vectors as specified in 8.4.5.4 */
1990
    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));
1991
    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));
1992
    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));
1993
    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));
1994
    if(direct) {
1995
        s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
1996
        s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
1997
        s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
1998
        s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
1999
        return;
2000
    }
2001

    
2002
    if((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2003
        C = s->current_picture.motion_val[0][xy - 2];
2004
        A = s->current_picture.motion_val[0][xy - wrap*2];
2005
        off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2006
        B = s->current_picture.motion_val[0][xy - wrap*2 + off];
2007

    
2008
        if(!s->mb_x) C[0] = C[1] = 0;
2009
        if(!s->first_slice_line) { // predictor A is not out of bounds
2010
            if(s->mb_width == 1) {
2011
                px = A[0];
2012
                py = A[1];
2013
            } else {
2014
                px = mid_pred(A[0], B[0], C[0]);
2015
                py = mid_pred(A[1], B[1], C[1]);
2016
            }
2017
        } else if(s->mb_x) { // predictor C is not out of bounds
2018
            px = C[0];
2019
            py = C[1];
2020
        } else {
2021
            px = py = 0;
2022
        }
2023
        /* Pullback MV as specified in 8.3.5.3.4 */
2024
        {
2025
            int qx, qy, X, Y;
2026
            if(v->profile < PROFILE_ADVANCED) {
2027
                qx = (s->mb_x << 5);
2028
                qy = (s->mb_y << 5);
2029
                X = (s->mb_width << 5) - 4;
2030
                Y = (s->mb_height << 5) - 4;
2031
                if(qx + px < -28) px = -28 - qx;
2032
                if(qy + py < -28) py = -28 - qy;
2033
                if(qx + px > X) px = X - qx;
2034
                if(qy + py > Y) py = Y - qy;
2035
            } else {
2036
                qx = (s->mb_x << 6);
2037
                qy = (s->mb_y << 6);
2038
                X = (s->mb_width << 6) - 4;
2039
                Y = (s->mb_height << 6) - 4;
2040
                if(qx + px < -60) px = -60 - qx;
2041
                if(qy + py < -60) py = -60 - qy;
2042
                if(qx + px > X) px = X - qx;
2043
                if(qy + py > Y) py = Y - qy;
2044
            }
2045
        }
2046
        /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2047
        if(0 && !s->first_slice_line && s->mb_x) {
2048
            if(is_intra[xy - wrap])
2049
                sum = FFABS(px) + FFABS(py);
2050
            else
2051
                sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2052
            if(sum > 32) {
2053
                if(get_bits1(&s->gb)) {
2054
                    px = A[0];
2055
                    py = A[1];
2056
                } else {
2057
                    px = C[0];
2058
                    py = C[1];
2059
                }
2060
            } else {
2061
                if(is_intra[xy - 2])
2062
                    sum = FFABS(px) + FFABS(py);
2063
                else
2064
                    sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2065
                if(sum > 32) {
2066
                    if(get_bits1(&s->gb)) {
2067
                        px = A[0];
2068
                        py = A[1];
2069
                    } else {
2070
                        px = C[0];
2071
                        py = C[1];
2072
                    }
2073
                }
2074
            }
2075
        }
2076
        /* store MV using signed modulus of MV range defined in 4.11 */
2077
        s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;
2078
        s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;
2079
    }
2080
    if((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {
2081
        C = s->current_picture.motion_val[1][xy - 2];
2082
        A = s->current_picture.motion_val[1][xy - wrap*2];
2083
        off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;
2084
        B = s->current_picture.motion_val[1][xy - wrap*2 + off];
2085

    
2086
        if(!s->mb_x) C[0] = C[1] = 0;
2087
        if(!s->first_slice_line) { // predictor A is not out of bounds
2088
            if(s->mb_width == 1) {
2089
                px = A[0];
2090
                py = A[1];
2091
            } else {
2092
                px = mid_pred(A[0], B[0], C[0]);
2093
                py = mid_pred(A[1], B[1], C[1]);
2094
            }
2095
        } else if(s->mb_x) { // predictor C is not out of bounds
2096
            px = C[0];
2097
            py = C[1];
2098
        } else {
2099
            px = py = 0;
2100
        }
2101
        /* Pullback MV as specified in 8.3.5.3.4 */
2102
        {
2103
            int qx, qy, X, Y;
2104
            if(v->profile < PROFILE_ADVANCED) {
2105
                qx = (s->mb_x << 5);
2106
                qy = (s->mb_y << 5);
2107
                X = (s->mb_width << 5) - 4;
2108
                Y = (s->mb_height << 5) - 4;
2109
                if(qx + px < -28) px = -28 - qx;
2110
                if(qy + py < -28) py = -28 - qy;
2111
                if(qx + px > X) px = X - qx;
2112
                if(qy + py > Y) py = Y - qy;
2113
            } else {
2114
                qx = (s->mb_x << 6);
2115
                qy = (s->mb_y << 6);
2116
                X = (s->mb_width << 6) - 4;
2117
                Y = (s->mb_height << 6) - 4;
2118
                if(qx + px < -60) px = -60 - qx;
2119
                if(qy + py < -60) py = -60 - qy;
2120
                if(qx + px > X) px = X - qx;
2121
                if(qy + py > Y) py = Y - qy;
2122
            }
2123
        }
2124
        /* Calculate hybrid prediction as specified in 8.3.5.3.5 */
2125
        if(0 && !s->first_slice_line && s->mb_x) {
2126
            if(is_intra[xy - wrap])
2127
                sum = FFABS(px) + FFABS(py);
2128
            else
2129
                sum = FFABS(px - A[0]) + FFABS(py - A[1]);
2130
            if(sum > 32) {
2131
                if(get_bits1(&s->gb)) {
2132
                    px = A[0];
2133
                    py = A[1];
2134
                } else {
2135
                    px = C[0];
2136
                    py = C[1];
2137
                }
2138
            } else {
2139
                if(is_intra[xy - 2])
2140
                    sum = FFABS(px) + FFABS(py);
2141
                else
2142
                    sum = FFABS(px - C[0]) + FFABS(py - C[1]);
2143
                if(sum > 32) {
2144
                    if(get_bits1(&s->gb)) {
2145
                        px = A[0];
2146
                        py = A[1];
2147
                    } else {
2148
                        px = C[0];
2149
                        py = C[1];
2150
                    }
2151
                }
2152
            }
2153
        }
2154
        /* store MV using signed modulus of MV range defined in 4.11 */
2155

    
2156
        s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;
2157
        s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;
2158
    }
2159
    s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];
2160
    s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];
2161
    s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];
2162
    s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];
2163
}
2164

    
2165
/** Get predicted DC value for I-frames only
2166
 * prediction dir: left=0, top=1
2167
 * @param s MpegEncContext
2168
 * @param[in] n block index in the current MB
2169
 * @param dc_val_ptr Pointer to DC predictor
2170
 * @param dir_ptr Prediction direction for use in AC prediction
2171
 */
2172
static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2173
                              int16_t **dc_val_ptr, int *dir_ptr)
2174
{
2175
    int a, b, c, wrap, pred, scale;
2176
    int16_t *dc_val;
2177
    static const uint16_t dcpred[32] = {
2178
    -1, 1024,  512,  341,  256,  205,  171,  146,  128,
2179
         114,  102,   93,   85,   79,   73,   68,   64,
2180
          60,   57,   54,   51,   49,   47,   45,   43,
2181
          41,   39,   38,   37,   35,   34,   33
2182
    };
2183

    
2184
    /* find prediction - wmv3_dc_scale always used here in fact */
2185
    if (n < 4)     scale = s->y_dc_scale;
2186
    else           scale = s->c_dc_scale;
2187

    
2188
    wrap = s->block_wrap[n];
2189
    dc_val= s->dc_val[0] + s->block_index[n];
2190

    
2191
    /* B A
2192
     * C X
2193
     */
2194
    c = dc_val[ - 1];
2195
    b = dc_val[ - 1 - wrap];
2196
    a = dc_val[ - wrap];
2197

    
2198
    if (pq < 9 || !overlap)
2199
    {
2200
        /* Set outer values */
2201
        if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];
2202
        if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];
2203
    }
2204
    else
2205
    {
2206
        /* Set outer values */
2207
        if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;
2208
        if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;
2209
    }
2210

    
2211
    if (abs(a - b) <= abs(b - c)) {
2212
        pred = c;
2213
        *dir_ptr = 1;//left
2214
    } else {
2215
        pred = a;
2216
        *dir_ptr = 0;//top
2217
    }
2218

    
2219
    /* update predictor */
2220
    *dc_val_ptr = &dc_val[0];
2221
    return pred;
2222
}
2223

    
2224

    
2225
/** Get predicted DC value
2226
 * prediction dir: left=0, top=1
2227
 * @param s MpegEncContext
2228
 * @param[in] n block index in the current MB
2229
 * @param dc_val_ptr Pointer to DC predictor
2230
 * @param dir_ptr Prediction direction for use in AC prediction
2231
 */
2232
static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,
2233
                              int a_avail, int c_avail,
2234
                              int16_t **dc_val_ptr, int *dir_ptr)
2235
{
2236
    int a, b, c, wrap, pred, scale;
2237
    int16_t *dc_val;
2238
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2239
    int q1, q2 = 0;
2240

    
2241
    /* find prediction - wmv3_dc_scale always used here in fact */
2242
    if (n < 4)     scale = s->y_dc_scale;
2243
    else           scale = s->c_dc_scale;
2244

    
2245
    wrap = s->block_wrap[n];
2246
    dc_val= s->dc_val[0] + s->block_index[n];
2247

    
2248
    /* B A
2249
     * C X
2250
     */
2251
    c = dc_val[ - 1];
2252
    b = dc_val[ - 1 - wrap];
2253
    a = dc_val[ - wrap];
2254
    /* scale predictors if needed */
2255
    q1 = s->current_picture.qscale_table[mb_pos];
2256
    if(c_avail && (n!= 1 && n!=3)) {
2257
        q2 = s->current_picture.qscale_table[mb_pos - 1];
2258
        if(q2 && q2 != q1)
2259
            c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2260
    }
2261
    if(a_avail && (n!= 2 && n!=3)) {
2262
        q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2263
        if(q2 && q2 != q1)
2264
            a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2265
    }
2266
    if(a_avail && c_avail && (n!=3)) {
2267
        int off = mb_pos;
2268
        if(n != 1) off--;
2269
        if(n != 2) off -= s->mb_stride;
2270
        q2 = s->current_picture.qscale_table[off];
2271
        if(q2 && q2 != q1)
2272
            b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;
2273
    }
2274

    
2275
    if(a_avail && c_avail) {
2276
        if(abs(a - b) <= abs(b - c)) {
2277
            pred = c;
2278
            *dir_ptr = 1;//left
2279
        } else {
2280
            pred = a;
2281
            *dir_ptr = 0;//top
2282
        }
2283
    } else if(a_avail) {
2284
        pred = a;
2285
        *dir_ptr = 0;//top
2286
    } else if(c_avail) {
2287
        pred = c;
2288
        *dir_ptr = 1;//left
2289
    } else {
2290
        pred = 0;
2291
        *dir_ptr = 1;//left
2292
    }
2293

    
2294
    /* update predictor */
2295
    *dc_val_ptr = &dc_val[0];
2296
    return pred;
2297
}
2298

    
2299

    
2300
/**
2301
 * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles
2302
 * @see 7.1.4, p91 and 8.1.1.7, p(1)04
2303
 * @{
2304
 */
2305

    
2306
static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)
2307
{
2308
    int xy, wrap, pred, a, b, c;
2309

    
2310
    xy = s->block_index[n];
2311
    wrap = s->b8_stride;
2312

    
2313
    /* B C
2314
     * A X
2315
     */
2316
    a = s->coded_block[xy - 1       ];
2317
    b = s->coded_block[xy - 1 - wrap];
2318
    c = s->coded_block[xy     - wrap];
2319

    
2320
    if (b == c) {
2321
        pred = a;
2322
    } else {
2323
        pred = c;
2324
    }
2325

    
2326
    /* store value */
2327
    *coded_block_ptr = &s->coded_block[xy];
2328

    
2329
    return pred;
2330
}
2331

    
2332
/**
2333
 * Decode one AC coefficient
2334
 * @param v The VC1 context
2335
 * @param last Last coefficient
2336
 * @param skip How much zero coefficients to skip
2337
 * @param value Decoded AC coefficient value
2338
 * @see 8.1.3.4
2339
 */
2340
static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)
2341
{
2342
    GetBitContext *gb = &v->s.gb;
2343
    int index, escape, run = 0, level = 0, lst = 0;
2344

    
2345
    index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2346
    if (index != vc1_ac_sizes[codingset] - 1) {
2347
        run = vc1_index_decode_table[codingset][index][0];
2348
        level = vc1_index_decode_table[codingset][index][1];
2349
        lst = index >= vc1_last_decode_table[codingset];
2350
        if(get_bits1(gb))
2351
            level = -level;
2352
    } else {
2353
        escape = decode210(gb);
2354
        if (escape != 2) {
2355
            index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
2356
            run = vc1_index_decode_table[codingset][index][0];
2357
            level = vc1_index_decode_table[codingset][index][1];
2358
            lst = index >= vc1_last_decode_table[codingset];
2359
            if(escape == 0) {
2360
                if(lst)
2361
                    level += vc1_last_delta_level_table[codingset][run];
2362
                else
2363
                    level += vc1_delta_level_table[codingset][run];
2364
            } else {
2365
                if(lst)
2366
                    run += vc1_last_delta_run_table[codingset][level] + 1;
2367
                else
2368
                    run += vc1_delta_run_table[codingset][level] + 1;
2369
            }
2370
            if(get_bits1(gb))
2371
                level = -level;
2372
        } else {
2373
            int sign;
2374
            lst = get_bits1(gb);
2375
            if(v->s.esc3_level_length == 0) {
2376
                if(v->pq < 8 || v->dquantfrm) { // table 59
2377
                    v->s.esc3_level_length = get_bits(gb, 3);
2378
                    if(!v->s.esc3_level_length)
2379
                        v->s.esc3_level_length = get_bits(gb, 2) + 8;
2380
                } else { //table 60
2381
                    v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;
2382
                }
2383
                v->s.esc3_run_length = 3 + get_bits(gb, 2);
2384
            }
2385
            run = get_bits(gb, v->s.esc3_run_length);
2386
            sign = get_bits1(gb);
2387
            level = get_bits(gb, v->s.esc3_level_length);
2388
            if(sign)
2389
                level = -level;
2390
        }
2391
    }
2392

    
2393
    *last = lst;
2394
    *skip = run;
2395
    *value = level;
2396
}
2397

    
2398
/** Decode intra block in intra frames - should be faster than decode_intra_block
2399
 * @param v VC1Context
2400
 * @param block block to decode
2401
 * @param coded are AC coeffs present or not
2402
 * @param codingset set of VLC to decode data
2403
 */
2404
static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)
2405
{
2406
    GetBitContext *gb = &v->s.gb;
2407
    MpegEncContext *s = &v->s;
2408
    int dc_pred_dir = 0; /* Direction of the DC prediction used */
2409
    int run_diff, i;
2410
    int16_t *dc_val;
2411
    int16_t *ac_val, *ac_val2;
2412
    int dcdiff;
2413

    
2414
    /* Get DC differential */
2415
    if (n < 4) {
2416
        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2417
    } else {
2418
        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2419
    }
2420
    if (dcdiff < 0){
2421
        av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2422
        return -1;
2423
    }
2424
    if (dcdiff)
2425
    {
2426
        if (dcdiff == 119 /* ESC index value */)
2427
        {
2428
            /* TODO: Optimize */
2429
            if (v->pq == 1) dcdiff = get_bits(gb, 10);
2430
            else if (v->pq == 2) dcdiff = get_bits(gb, 9);
2431
            else dcdiff = get_bits(gb, 8);
2432
        }
2433
        else
2434
        {
2435
            if (v->pq == 1)
2436
                dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2437
            else if (v->pq == 2)
2438
                dcdiff = (dcdiff<<1) + get_bits1(gb)   - 1;
2439
        }
2440
        if (get_bits1(gb))
2441
            dcdiff = -dcdiff;
2442
    }
2443

    
2444
    /* Prediction */
2445
    dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);
2446
    *dc_val = dcdiff;
2447

    
2448
    /* Store the quantized DC coeff, used for prediction */
2449
    if (n < 4) {
2450
        block[0] = dcdiff * s->y_dc_scale;
2451
    } else {
2452
        block[0] = dcdiff * s->c_dc_scale;
2453
    }
2454
    /* Skip ? */
2455
    run_diff = 0;
2456
    i = 0;
2457
    if (!coded) {
2458
        goto not_coded;
2459
    }
2460

    
2461
    //AC Decoding
2462
    i = 1;
2463

    
2464
    {
2465
        int last = 0, skip, value;
2466
        const int8_t *zz_table;
2467
        int scale;
2468
        int k;
2469

    
2470
        scale = v->pq * 2 + v->halfpq;
2471

    
2472
        if(v->s.ac_pred) {
2473
            if(!dc_pred_dir)
2474
                zz_table = wmv1_scantable[2];
2475
            else
2476
                zz_table = wmv1_scantable[3];
2477
        } else
2478
            zz_table = wmv1_scantable[1];
2479

    
2480
        ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2481
        ac_val2 = ac_val;
2482
        if(dc_pred_dir) //left
2483
            ac_val -= 16;
2484
        else //top
2485
            ac_val -= 16 * s->block_wrap[n];
2486

    
2487
        while (!last) {
2488
            vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2489
            i += skip;
2490
            if(i > 63)
2491
                break;
2492
            block[zz_table[i++]] = value;
2493
        }
2494

    
2495
        /* apply AC prediction if needed */
2496
        if(s->ac_pred) {
2497
            if(dc_pred_dir) { //left
2498
                for(k = 1; k < 8; k++)
2499
                    block[k << 3] += ac_val[k];
2500
            } else { //top
2501
                for(k = 1; k < 8; k++)
2502
                    block[k] += ac_val[k + 8];
2503
            }
2504
        }
2505
        /* save AC coeffs for further prediction */
2506
        for(k = 1; k < 8; k++) {
2507
            ac_val2[k] = block[k << 3];
2508
            ac_val2[k + 8] = block[k];
2509
        }
2510

    
2511
        /* scale AC coeffs */
2512
        for(k = 1; k < 64; k++)
2513
            if(block[k]) {
2514
                block[k] *= scale;
2515
                if(!v->pquantizer)
2516
                    block[k] += (block[k] < 0) ? -v->pq : v->pq;
2517
            }
2518

    
2519
        if(s->ac_pred) i = 63;
2520
    }
2521

    
2522
not_coded:
2523
    if(!coded) {
2524
        int k, scale;
2525
        ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2526
        ac_val2 = ac_val;
2527

    
2528
        scale = v->pq * 2 + v->halfpq;
2529
        memset(ac_val2, 0, 16 * 2);
2530
        if(dc_pred_dir) {//left
2531
            ac_val -= 16;
2532
            if(s->ac_pred)
2533
                memcpy(ac_val2, ac_val, 8 * 2);
2534
        } else {//top
2535
            ac_val -= 16 * s->block_wrap[n];
2536
            if(s->ac_pred)
2537
                memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2538
        }
2539

    
2540
        /* apply AC prediction if needed */
2541
        if(s->ac_pred) {
2542
            if(dc_pred_dir) { //left
2543
                for(k = 1; k < 8; k++) {
2544
                    block[k << 3] = ac_val[k] * scale;
2545
                    if(!v->pquantizer && block[k << 3])
2546
                        block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;
2547
                }
2548
            } else { //top
2549
                for(k = 1; k < 8; k++) {
2550
                    block[k] = ac_val[k + 8] * scale;
2551
                    if(!v->pquantizer && block[k])
2552
                        block[k] += (block[k] < 0) ? -v->pq : v->pq;
2553
                }
2554
            }
2555
            i = 63;
2556
        }
2557
    }
2558
    s->block_last_index[n] = i;
2559

    
2560
    return 0;
2561
}
2562

    
2563
/** Decode intra block in intra frames - should be faster than decode_intra_block
2564
 * @param v VC1Context
2565
 * @param block block to decode
2566
 * @param coded are AC coeffs present or not
2567
 * @param codingset set of VLC to decode data
2568
 */
2569
static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant)
2570
{
2571
    GetBitContext *gb = &v->s.gb;
2572
    MpegEncContext *s = &v->s;
2573
    int dc_pred_dir = 0; /* Direction of the DC prediction used */
2574
    int run_diff, i;
2575
    int16_t *dc_val;
2576
    int16_t *ac_val, *ac_val2;
2577
    int dcdiff;
2578
    int a_avail = v->a_avail, c_avail = v->c_avail;
2579
    int use_pred = s->ac_pred;
2580
    int scale;
2581
    int q1, q2 = 0;
2582
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2583

    
2584
    /* Get DC differential */
2585
    if (n < 4) {
2586
        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2587
    } else {
2588
        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2589
    }
2590
    if (dcdiff < 0){
2591
        av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2592
        return -1;
2593
    }
2594
    if (dcdiff)
2595
    {
2596
        if (dcdiff == 119 /* ESC index value */)
2597
        {
2598
            /* TODO: Optimize */
2599
            if (mquant == 1) dcdiff = get_bits(gb, 10);
2600
            else if (mquant == 2) dcdiff = get_bits(gb, 9);
2601
            else dcdiff = get_bits(gb, 8);
2602
        }
2603
        else
2604
        {
2605
            if (mquant == 1)
2606
                dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2607
            else if (mquant == 2)
2608
                dcdiff = (dcdiff<<1) + get_bits1(gb)   - 1;
2609
        }
2610
        if (get_bits1(gb))
2611
            dcdiff = -dcdiff;
2612
    }
2613

    
2614
    /* Prediction */
2615
    dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);
2616
    *dc_val = dcdiff;
2617

    
2618
    /* Store the quantized DC coeff, used for prediction */
2619
    if (n < 4) {
2620
        block[0] = dcdiff * s->y_dc_scale;
2621
    } else {
2622
        block[0] = dcdiff * s->c_dc_scale;
2623
    }
2624
    /* Skip ? */
2625
    run_diff = 0;
2626
    i = 0;
2627

    
2628
    //AC Decoding
2629
    i = 1;
2630

    
2631
    /* check if AC is needed at all */
2632
    if(!a_avail && !c_avail) use_pred = 0;
2633
    ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2634
    ac_val2 = ac_val;
2635

    
2636
    scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);
2637

    
2638
    if(dc_pred_dir) //left
2639
        ac_val -= 16;
2640
    else //top
2641
        ac_val -= 16 * s->block_wrap[n];
2642

    
2643
    q1 = s->current_picture.qscale_table[mb_pos];
2644
    if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2645
    if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2646
    if(dc_pred_dir && n==1) q2 = q1;
2647
    if(!dc_pred_dir && n==2) q2 = q1;
2648
    if(n==3) q2 = q1;
2649

    
2650
    if(coded) {
2651
        int last = 0, skip, value;
2652
        const int8_t *zz_table;
2653
        int k;
2654

    
2655
        if(v->s.ac_pred) {
2656
            if(!dc_pred_dir)
2657
                zz_table = wmv1_scantable[2];
2658
            else
2659
                zz_table = wmv1_scantable[3];
2660
        } else
2661
            zz_table = wmv1_scantable[1];
2662

    
2663
        while (!last) {
2664
            vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2665
            i += skip;
2666
            if(i > 63)
2667
                break;
2668
            block[zz_table[i++]] = value;
2669
        }
2670

    
2671
        /* apply AC prediction if needed */
2672
        if(use_pred) {
2673
            /* scale predictors if needed*/
2674
            if(q2 && q1!=q2) {
2675
                q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2676
                q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2677

    
2678
                if(dc_pred_dir) { //left
2679
                    for(k = 1; k < 8; k++)
2680
                        block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2681
                } else { //top
2682
                    for(k = 1; k < 8; k++)
2683
                        block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2684
                }
2685
            } else {
2686
                if(dc_pred_dir) { //left
2687
                    for(k = 1; k < 8; k++)
2688
                        block[k << 3] += ac_val[k];
2689
                } else { //top
2690
                    for(k = 1; k < 8; k++)
2691
                        block[k] += ac_val[k + 8];
2692
                }
2693
            }
2694
        }
2695
        /* save AC coeffs for further prediction */
2696
        for(k = 1; k < 8; k++) {
2697
            ac_val2[k] = block[k << 3];
2698
            ac_val2[k + 8] = block[k];
2699
        }
2700

    
2701
        /* scale AC coeffs */
2702
        for(k = 1; k < 64; k++)
2703
            if(block[k]) {
2704
                block[k] *= scale;
2705
                if(!v->pquantizer)
2706
                    block[k] += (block[k] < 0) ? -mquant : mquant;
2707
            }
2708

    
2709
        if(use_pred) i = 63;
2710
    } else { // no AC coeffs
2711
        int k;
2712

    
2713
        memset(ac_val2, 0, 16 * 2);
2714
        if(dc_pred_dir) {//left
2715
            if(use_pred) {
2716
                memcpy(ac_val2, ac_val, 8 * 2);
2717
                if(q2 && q1!=q2) {
2718
                    q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2719
                    q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2720
                    for(k = 1; k < 8; k++)
2721
                        ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2722
                }
2723
            }
2724
        } else {//top
2725
            if(use_pred) {
2726
                memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2727
                if(q2 && q1!=q2) {
2728
                    q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2729
                    q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2730
                    for(k = 1; k < 8; k++)
2731
                        ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2732
                }
2733
            }
2734
        }
2735

    
2736
        /* apply AC prediction if needed */
2737
        if(use_pred) {
2738
            if(dc_pred_dir) { //left
2739
                for(k = 1; k < 8; k++) {
2740
                    block[k << 3] = ac_val2[k] * scale;
2741
                    if(!v->pquantizer && block[k << 3])
2742
                        block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2743
                }
2744
            } else { //top
2745
                for(k = 1; k < 8; k++) {
2746
                    block[k] = ac_val2[k + 8] * scale;
2747
                    if(!v->pquantizer && block[k])
2748
                        block[k] += (block[k] < 0) ? -mquant : mquant;
2749
                }
2750
            }
2751
            i = 63;
2752
        }
2753
    }
2754
    s->block_last_index[n] = i;
2755

    
2756
    return 0;
2757
}
2758

    
2759
/** Decode intra block in inter frames - more generic version than vc1_decode_i_block
2760
 * @param v VC1Context
2761
 * @param block block to decode
2762
 * @param coded are AC coeffs present or not
2763
 * @param mquant block quantizer
2764
 * @param codingset set of VLC to decode data
2765
 */
2766
static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)
2767
{
2768
    GetBitContext *gb = &v->s.gb;
2769
    MpegEncContext *s = &v->s;
2770
    int dc_pred_dir = 0; /* Direction of the DC prediction used */
2771
    int run_diff, i;
2772
    int16_t *dc_val;
2773
    int16_t *ac_val, *ac_val2;
2774
    int dcdiff;
2775
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
2776
    int a_avail = v->a_avail, c_avail = v->c_avail;
2777
    int use_pred = s->ac_pred;
2778
    int scale;
2779
    int q1, q2 = 0;
2780

    
2781
    /* XXX: Guard against dumb values of mquant */
2782
    mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );
2783

    
2784
    /* Set DC scale - y and c use the same */
2785
    s->y_dc_scale = s->y_dc_scale_table[mquant];
2786
    s->c_dc_scale = s->c_dc_scale_table[mquant];
2787

    
2788
    /* Get DC differential */
2789
    if (n < 4) {
2790
        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2791
    } else {
2792
        dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);
2793
    }
2794
    if (dcdiff < 0){
2795
        av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");
2796
        return -1;
2797
    }
2798
    if (dcdiff)
2799
    {
2800
        if (dcdiff == 119 /* ESC index value */)
2801
        {
2802
            /* TODO: Optimize */
2803
            if (mquant == 1) dcdiff = get_bits(gb, 10);
2804
            else if (mquant == 2) dcdiff = get_bits(gb, 9);
2805
            else dcdiff = get_bits(gb, 8);
2806
        }
2807
        else
2808
        {
2809
            if (mquant == 1)
2810
                dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;
2811
            else if (mquant == 2)
2812
                dcdiff = (dcdiff<<1) + get_bits1(gb)   - 1;
2813
        }
2814
        if (get_bits1(gb))
2815
            dcdiff = -dcdiff;
2816
    }
2817

    
2818
    /* Prediction */
2819
    dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);
2820
    *dc_val = dcdiff;
2821

    
2822
    /* Store the quantized DC coeff, used for prediction */
2823

    
2824
    if (n < 4) {
2825
        block[0] = dcdiff * s->y_dc_scale;
2826
    } else {
2827
        block[0] = dcdiff * s->c_dc_scale;
2828
    }
2829
    /* Skip ? */
2830
    run_diff = 0;
2831
    i = 0;
2832

    
2833
    //AC Decoding
2834
    i = 1;
2835

    
2836
    /* check if AC is needed at all and adjust direction if needed */
2837
    if(!a_avail) dc_pred_dir = 1;
2838
    if(!c_avail) dc_pred_dir = 0;
2839
    if(!a_avail && !c_avail) use_pred = 0;
2840
    ac_val = s->ac_val[0][0] + s->block_index[n] * 16;
2841
    ac_val2 = ac_val;
2842

    
2843
    scale = mquant * 2 + v->halfpq;
2844

    
2845
    if(dc_pred_dir) //left
2846
        ac_val -= 16;
2847
    else //top
2848
        ac_val -= 16 * s->block_wrap[n];
2849

    
2850
    q1 = s->current_picture.qscale_table[mb_pos];
2851
    if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];
2852
    if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];
2853
    if(dc_pred_dir && n==1) q2 = q1;
2854
    if(!dc_pred_dir && n==2) q2 = q1;
2855
    if(n==3) q2 = q1;
2856

    
2857
    if(coded) {
2858
        int last = 0, skip, value;
2859
        const int8_t *zz_table;
2860
        int k;
2861

    
2862
        zz_table = wmv1_scantable[0];
2863

    
2864
        while (!last) {
2865
            vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);
2866
            i += skip;
2867
            if(i > 63)
2868
                break;
2869
            block[zz_table[i++]] = value;
2870
        }
2871

    
2872
        /* apply AC prediction if needed */
2873
        if(use_pred) {
2874
            /* scale predictors if needed*/
2875
            if(q2 && q1!=q2) {
2876
                q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2877
                q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2878

    
2879
                if(dc_pred_dir) { //left
2880
                    for(k = 1; k < 8; k++)
2881
                        block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2882
                } else { //top
2883
                    for(k = 1; k < 8; k++)
2884
                        block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2885
                }
2886
            } else {
2887
                if(dc_pred_dir) { //left
2888
                    for(k = 1; k < 8; k++)
2889
                        block[k << 3] += ac_val[k];
2890
                } else { //top
2891
                    for(k = 1; k < 8; k++)
2892
                        block[k] += ac_val[k + 8];
2893
                }
2894
            }
2895
        }
2896
        /* save AC coeffs for further prediction */
2897
        for(k = 1; k < 8; k++) {
2898
            ac_val2[k] = block[k << 3];
2899
            ac_val2[k + 8] = block[k];
2900
        }
2901

    
2902
        /* scale AC coeffs */
2903
        for(k = 1; k < 64; k++)
2904
            if(block[k]) {
2905
                block[k] *= scale;
2906
                if(!v->pquantizer)
2907
                    block[k] += (block[k] < 0) ? -mquant : mquant;
2908
            }
2909

    
2910
        if(use_pred) i = 63;
2911
    } else { // no AC coeffs
2912
        int k;
2913

    
2914
        memset(ac_val2, 0, 16 * 2);
2915
        if(dc_pred_dir) {//left
2916
            if(use_pred) {
2917
                memcpy(ac_val2, ac_val, 8 * 2);
2918
                if(q2 && q1!=q2) {
2919
                    q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2920
                    q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2921
                    for(k = 1; k < 8; k++)
2922
                        ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2923
                }
2924
            }
2925
        } else {//top
2926
            if(use_pred) {
2927
                memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
2928
                if(q2 && q1!=q2) {
2929
                    q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;
2930
                    q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;
2931
                    for(k = 1; k < 8; k++)
2932
                        ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;
2933
                }
2934
            }
2935
        }
2936

    
2937
        /* apply AC prediction if needed */
2938
        if(use_pred) {
2939
            if(dc_pred_dir) { //left
2940
                for(k = 1; k < 8; k++) {
2941
                    block[k << 3] = ac_val2[k] * scale;
2942
                    if(!v->pquantizer && block[k << 3])
2943
                        block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;
2944
                }
2945
            } else { //top
2946
                for(k = 1; k < 8; k++) {
2947
                    block[k] = ac_val2[k + 8] * scale;
2948
                    if(!v->pquantizer && block[k])
2949
                        block[k] += (block[k] < 0) ? -mquant : mquant;
2950
                }
2951
            }
2952
            i = 63;
2953
        }
2954
    }
2955
    s->block_last_index[n] = i;
2956

    
2957
    return 0;
2958
}
2959

    
2960
/** Decode P block
2961
 */
2962
static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,
2963
                              uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left)
2964
{
2965
    MpegEncContext *s = &v->s;
2966
    GetBitContext *gb = &s->gb;
2967
    int i, j;
2968
    int subblkpat = 0;
2969
    int scale, off, idx, last, skip, value;
2970
    int ttblk = ttmb & 7;
2971
    int pat = 0;
2972

    
2973
    if(ttmb == -1) {
2974
        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)];
2975
    }
2976
    if(ttblk == TT_4X4) {
2977
        subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);
2978
    }
2979
    if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {
2980
        subblkpat = decode012(gb);
2981
        if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits
2982
        if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;
2983
        if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;
2984
    }
2985
    scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);
2986

    
2987
    // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT
2988
    if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {
2989
        subblkpat = 2 - (ttblk == TT_8X4_TOP);
2990
        ttblk = TT_8X4;
2991
    }
2992
    if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
2993
        subblkpat = 2 - (ttblk == TT_4X8_LEFT);
2994
        ttblk = TT_4X8;
2995
    }
2996
    switch(ttblk) {
2997
    case TT_8X8:
2998
        pat = 0xF;
2999
        i = 0;
3000
        last = 0;
3001
        while (!last) {
3002
            vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3003
            i += skip;
3004
            if(i > 63)
3005
                break;
3006
            idx = wmv1_scantable[0][i++];
3007
            block[idx] = value * scale;
3008
            if(!v->pquantizer)
3009
                block[idx] += (block[idx] < 0) ? -mquant : mquant;
3010
        }
3011
        if(!skip_block){
3012
            s->dsp.vc1_inv_trans_8x8(block);
3013
            s->dsp.add_pixels_clamped(block, dst, linesize);
3014
            if(apply_filter && cbp_top  & 0xC)
3015
                vc1_loop_filter(dst, 1, linesize, 8, mquant);
3016
            if(apply_filter && cbp_left & 0xA)
3017
                vc1_loop_filter(dst, linesize, 1, 8, mquant);
3018
        }
3019
        break;
3020
    case TT_4X4:
3021
        pat = ~subblkpat & 0xF;
3022
        for(j = 0; j < 4; j++) {
3023
            last = subblkpat & (1 << (3 - j));
3024
            i = 0;
3025
            off = (j & 1) * 4 + (j & 2) * 16;
3026
            while (!last) {
3027
                vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3028
                i += skip;
3029
                if(i > 15)
3030
                    break;
3031
                idx = ff_vc1_simple_progressive_4x4_zz[i++];
3032
                block[idx + off] = value * scale;
3033
                if(!v->pquantizer)
3034
                    block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;
3035
            }
3036
            if(!(subblkpat & (1 << (3 - j))) && !skip_block){
3037
                s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);
3038
                if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))
3039
                    vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, 1, linesize, 4, mquant);
3040
                if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))
3041
                    vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, linesize, 1, 4, mquant);
3042
            }
3043
        }
3044
        break;
3045
    case TT_8X4:
3046
        pat = ~((subblkpat & 2)*6 + (subblkpat & 1)*3) & 0xF;
3047
        for(j = 0; j < 2; j++) {
3048
            last = subblkpat & (1 << (1 - j));
3049
            i = 0;
3050
            off = j * 32;
3051
            while (!last) {
3052
                vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3053
                i += skip;
3054
                if(i > 31)
3055
                    break;
3056
                idx = v->zz_8x4[i++]+off;
3057
                block[idx] = value * scale;
3058
                if(!v->pquantizer)
3059
                    block[idx] += (block[idx] < 0) ? -mquant : mquant;
3060
            }
3061
            if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3062
                s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);
3063
                if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))
3064
                    vc1_loop_filter(dst + j*4*linesize, 1, linesize, 8, mquant);
3065
                if(apply_filter && cbp_left & (2 << j))
3066
                    vc1_loop_filter(dst + j*4*linesize, linesize, 1, 4, mquant);
3067
            }
3068
        }
3069
        break;
3070
    case TT_4X8:
3071
        pat = ~(subblkpat*5) & 0xF;
3072
        for(j = 0; j < 2; j++) {
3073
            last = subblkpat & (1 << (1 - j));
3074
            i = 0;
3075
            off = j * 4;
3076
            while (!last) {
3077
                vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
3078
                i += skip;
3079
                if(i > 31)
3080
                    break;
3081
                idx = v->zz_4x8[i++]+off;
3082
                block[idx] = value * scale;
3083
                if(!v->pquantizer)
3084
                    block[idx] += (block[idx] < 0) ? -mquant : mquant;
3085
            }
3086
            if(!(subblkpat & (1 << (1 - j))) && !skip_block){
3087
                s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);
3088
                if(apply_filter && cbp_top & (2 << j))
3089
                    vc1_loop_filter(dst + j*4, 1, linesize, 4, mquant);
3090
                if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))
3091
                    vc1_loop_filter(dst + j*4, linesize, 1, 8, mquant);
3092
            }
3093
        }
3094
        break;
3095
    }
3096
    return pat;
3097
}
3098

    
3099

    
3100
/** Decode one P-frame MB (in Simple/Main profile)
3101
 */
3102
static int vc1_decode_p_mb(VC1Context *v)
3103
{
3104
    MpegEncContext *s = &v->s;
3105
    GetBitContext *gb = &s->gb;
3106
    int i, j;
3107
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3108
    int cbp; /* cbp decoding stuff */
3109
    int mqdiff, mquant; /* MB quantization */
3110
    int ttmb = v->ttfrm; /* MB Transform type */
3111

    
3112
    static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3113
      offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3114
    int mb_has_coeffs = 1; /* last_flag */
3115
    int dmv_x, dmv_y; /* Differential MV components */
3116
    int index, index1; /* LUT indexes */
3117
    int val, sign; /* temp values */
3118
    int first_block = 1;
3119
    int dst_idx, off;
3120
    int skipped, fourmv;
3121
    int block_cbp = 0, pat;
3122
    int apply_loop_filter;
3123

    
3124
    mquant = v->pq; /* Loosy initialization */
3125

    
3126
    if (v->mv_type_is_raw)
3127
        fourmv = get_bits1(gb);
3128
    else
3129
        fourmv = v->mv_type_mb_plane[mb_pos];
3130
    if (v->skip_is_raw)
3131
        skipped = get_bits1(gb);
3132
    else
3133
        skipped = v->s.mbskip_table[mb_pos];
3134

    
3135
    s->dsp.clear_blocks(s->block[0]);
3136

    
3137
    apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);
3138
    if (!fourmv) /* 1MV mode */
3139
    {
3140
        if (!skipped)
3141
        {
3142
            GET_MVDATA(dmv_x, dmv_y);
3143

    
3144
            if (s->mb_intra) {
3145
                s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3146
                s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3147
            }
3148
            s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
3149
            vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);
3150

    
3151
            /* FIXME Set DC val for inter block ? */
3152
            if (s->mb_intra && !mb_has_coeffs)
3153
            {
3154
                GET_MQUANT();
3155
                s->ac_pred = get_bits1(gb);
3156
                cbp = 0;
3157
            }
3158
            else if (mb_has_coeffs)
3159
            {
3160
                if (s->mb_intra) s->ac_pred = get_bits1(gb);
3161
                cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3162
                GET_MQUANT();
3163
            }
3164
            else
3165
            {
3166
                mquant = v->pq;
3167
                cbp = 0;
3168
            }
3169
            s->current_picture.qscale_table[mb_pos] = mquant;
3170

    
3171
            if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3172
                ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,
3173
                                VC1_TTMB_VLC_BITS, 2);
3174
            if(!s->mb_intra) vc1_mc_1mv(v, 0);
3175
            dst_idx = 0;
3176
            for (i=0; i<6; i++)
3177
            {
3178
                s->dc_val[0][s->block_index[i]] = 0;
3179
                dst_idx += i >> 2;
3180
                val = ((cbp >> (5 - i)) & 1);
3181
                off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3182
                v->mb_type[0][s->block_index[i]] = s->mb_intra;
3183
                if(s->mb_intra) {
3184
                    /* check if prediction blocks A and C are available */
3185
                    v->a_avail = v->c_avail = 0;
3186
                    if(i == 2 || i == 3 || !s->first_slice_line)
3187
                        v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3188
                    if(i == 1 || i == 3 || s->mb_x)
3189
                        v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3190

    
3191
                    vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3192
                    if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3193
                    s->dsp.vc1_inv_trans_8x8(s->block[i]);
3194
                    if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3195
                    s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3196
                    if(v->pq >= 9 && v->overlap) {
3197
                        if(v->c_avail)
3198
                            s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3199
                        if(v->a_avail)
3200
                            s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3201
                    }
3202
                    if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3203
                        int left_cbp, top_cbp;
3204
                        if(i & 4){
3205
                            left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);
3206
                            top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3207
                        }else{
3208
                            left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));
3209
                            top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3210
                        }
3211
                        if(left_cbp & 0xC)
3212
                            vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3213
                        if(top_cbp  & 0xA)
3214
                            vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3215
                    }
3216
                    block_cbp |= 0xF << (i << 2);
3217
                } else if(val) {
3218
                    int left_cbp = 0, top_cbp = 0, filter = 0;
3219
                    if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3220
                        filter = 1;
3221
                        if(i & 4){
3222
                            left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);
3223
                            top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3224
                        }else{
3225
                            left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));
3226
                            top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3227
                        }
3228
                        if(left_cbp & 0xC)
3229
                            vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3230
                        if(top_cbp  & 0xA)
3231
                            vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3232
                    }
3233
                    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);
3234
                    block_cbp |= pat << (i << 2);
3235
                    if(!v->ttmbf && ttmb < 8) ttmb = -1;
3236
                    first_block = 0;
3237
                }
3238
            }
3239
        }
3240
        else //Skipped
3241
        {
3242
            s->mb_intra = 0;
3243
            for(i = 0; i < 6; i++) {
3244
                v->mb_type[0][s->block_index[i]] = 0;
3245
                s->dc_val[0][s->block_index[i]] = 0;
3246
            }
3247
            s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;
3248
            s->current_picture.qscale_table[mb_pos] = 0;
3249
            vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);
3250
            vc1_mc_1mv(v, 0);
3251
            return 0;
3252
        }
3253
    } //1MV mode
3254
    else //4MV mode
3255
    {
3256
        if (!skipped /* unskipped MB */)
3257
        {
3258
            int intra_count = 0, coded_inter = 0;
3259
            int is_intra[6], is_coded[6];
3260
            /* Get CBPCY */
3261
            cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3262
            for (i=0; i<6; i++)
3263
            {
3264
                val = ((cbp >> (5 - i)) & 1);
3265
                s->dc_val[0][s->block_index[i]] = 0;
3266
                s->mb_intra = 0;
3267
                if(i < 4) {
3268
                    dmv_x = dmv_y = 0;
3269
                    s->mb_intra = 0;
3270
                    mb_has_coeffs = 0;
3271
                    if(val) {
3272
                        GET_MVDATA(dmv_x, dmv_y);
3273
                    }
3274
                    vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);
3275
                    if(!s->mb_intra) vc1_mc_4mv_luma(v, i);
3276
                    intra_count += s->mb_intra;
3277
                    is_intra[i] = s->mb_intra;
3278
                    is_coded[i] = mb_has_coeffs;
3279
                }
3280
                if(i&4){
3281
                    is_intra[i] = (intra_count >= 3);
3282
                    is_coded[i] = val;
3283
                }
3284
                if(i == 4) vc1_mc_4mv_chroma(v);
3285
                v->mb_type[0][s->block_index[i]] = is_intra[i];
3286
                if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];
3287
            }
3288
            // if there are no coded blocks then don't do anything more
3289
            if(!intra_count && !coded_inter) return 0;
3290
            dst_idx = 0;
3291
            GET_MQUANT();
3292
            s->current_picture.qscale_table[mb_pos] = mquant;
3293
            /* test if block is intra and has pred */
3294
            {
3295
                int intrapred = 0;
3296
                for(i=0; i<6; i++)
3297
                    if(is_intra[i]) {
3298
                        if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])
3299
                            || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {
3300
                            intrapred = 1;
3301
                            break;
3302
                        }
3303
                    }
3304
                if(intrapred)s->ac_pred = get_bits1(gb);
3305
                else s->ac_pred = 0;
3306
            }
3307
            if (!v->ttmbf && coded_inter)
3308
                ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3309
            for (i=0; i<6; i++)
3310
            {
3311
                dst_idx += i >> 2;
3312
                off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3313
                s->mb_intra = is_intra[i];
3314
                if (is_intra[i]) {
3315
                    /* check if prediction blocks A and C are available */
3316
                    v->a_avail = v->c_avail = 0;
3317
                    if(i == 2 || i == 3 || !s->first_slice_line)
3318
                        v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3319
                    if(i == 1 || i == 3 || s->mb_x)
3320
                        v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3321

    
3322
                    vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);
3323
                    if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3324
                    s->dsp.vc1_inv_trans_8x8(s->block[i]);
3325
                    if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3326
                    s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
3327
                    if(v->pq >= 9 && v->overlap) {
3328
                        if(v->c_avail)
3329
                            s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3330
                        if(v->a_avail)
3331
                            s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3332
                    }
3333
                    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)){
3334
                        int left_cbp, top_cbp;
3335
                        if(i & 4){
3336
                            left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);
3337
                            top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3338
                        }else{
3339
                            left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));
3340
                            top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3341
                        }
3342
                        if(left_cbp & 0xC)
3343
                            vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3344
                        if(top_cbp  & 0xA)
3345
                            vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3346
                    }
3347
                    block_cbp |= 0xF << (i << 2);
3348
                } else if(is_coded[i]) {
3349
                    int left_cbp = 0, top_cbp = 0, filter = 0;
3350
                    if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){
3351
                        filter = 1;
3352
                        if(i & 4){
3353
                            left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);
3354
                            top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);
3355
                        }else{
3356
                            left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));
3357
                            top_cbp  = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));
3358
                        }
3359
                        if(left_cbp & 0xC)
3360
                            vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);
3361
                        if(top_cbp  & 0xA)
3362
                            vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);
3363
                    }
3364
                    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);
3365
                    block_cbp |= pat << (i << 2);
3366
                    if(!v->ttmbf && ttmb < 8) ttmb = -1;
3367
                    first_block = 0;
3368
                }
3369
            }
3370
            return 0;
3371
        }
3372
        else //Skipped MB
3373
        {
3374
            s->mb_intra = 0;
3375
            s->current_picture.qscale_table[mb_pos] = 0;
3376
            for (i=0; i<6; i++) {
3377
                v->mb_type[0][s->block_index[i]] = 0;
3378
                s->dc_val[0][s->block_index[i]] = 0;
3379
            }
3380
            for (i=0; i<4; i++)
3381
            {
3382
                vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);
3383
                vc1_mc_4mv_luma(v, i);
3384
            }
3385
            vc1_mc_4mv_chroma(v);
3386
            s->current_picture.qscale_table[mb_pos] = 0;
3387
            return 0;
3388
        }
3389
    }
3390
    v->cbp[s->mb_x] = block_cbp;
3391

    
3392
    /* Should never happen */
3393
    return -1;
3394
}
3395

    
3396
/** Decode one B-frame MB (in Main profile)
3397
 */
3398
static void vc1_decode_b_mb(VC1Context *v)
3399
{
3400
    MpegEncContext *s = &v->s;
3401
    GetBitContext *gb = &s->gb;
3402
    int i, j;
3403
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3404
    int cbp = 0; /* cbp decoding stuff */
3405
    int mqdiff, mquant; /* MB quantization */
3406
    int ttmb = v->ttfrm; /* MB Transform type */
3407

    
3408
    static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },
3409
      offset_table[6] = { 0, 1, 3, 7, 15, 31 };
3410
    int mb_has_coeffs = 0; /* last_flag */
3411
    int index, index1; /* LUT indexes */
3412
    int val, sign; /* temp values */
3413
    int first_block = 1;
3414
    int dst_idx, off;
3415
    int skipped, direct;
3416
    int dmv_x[2], dmv_y[2];
3417
    int bmvtype = BMV_TYPE_BACKWARD;
3418

    
3419
    mquant = v->pq; /* Loosy initialization */
3420
    s->mb_intra = 0;
3421

    
3422
    if (v->dmb_is_raw)
3423
        direct = get_bits1(gb);
3424
    else
3425
        direct = v->direct_mb_plane[mb_pos];
3426
    if (v->skip_is_raw)
3427
        skipped = get_bits1(gb);
3428
    else
3429
        skipped = v->s.mbskip_table[mb_pos];
3430

    
3431
    s->dsp.clear_blocks(s->block[0]);
3432
    dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;
3433
    for(i = 0; i < 6; i++) {
3434
        v->mb_type[0][s->block_index[i]] = 0;
3435
        s->dc_val[0][s->block_index[i]] = 0;
3436
    }
3437
    s->current_picture.qscale_table[mb_pos] = 0;
3438

    
3439
    if (!direct) {
3440
        if (!skipped) {
3441
            GET_MVDATA(dmv_x[0], dmv_y[0]);
3442
            dmv_x[1] = dmv_x[0];
3443
            dmv_y[1] = dmv_y[0];
3444
        }
3445
        if(skipped || !s->mb_intra) {
3446
            bmvtype = decode012(gb);
3447
            switch(bmvtype) {
3448
            case 0:
3449
                bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;
3450
                break;
3451
            case 1:
3452
                bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;
3453
                break;
3454
            case 2:
3455
                bmvtype = BMV_TYPE_INTERPOLATED;
3456
                dmv_x[0] = dmv_y[0] = 0;
3457
            }
3458
        }
3459
    }
3460
    for(i = 0; i < 6; i++)
3461
        v->mb_type[0][s->block_index[i]] = s->mb_intra;
3462

    
3463
    if (skipped) {
3464
        if(direct) bmvtype = BMV_TYPE_INTERPOLATED;
3465
        vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3466
        vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3467
        return;
3468
    }
3469
    if (direct) {
3470
        cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3471
        GET_MQUANT();
3472
        s->mb_intra = 0;
3473
        mb_has_coeffs = 0;
3474
        s->current_picture.qscale_table[mb_pos] = mquant;
3475
        if(!v->ttmbf)
3476
            ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3477
        dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;
3478
        vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3479
        vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3480
    } else {
3481
        if(!mb_has_coeffs && !s->mb_intra) {
3482
            /* no coded blocks - effectively skipped */
3483
            vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3484
            vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3485
            return;
3486
        }
3487
        if(s->mb_intra && !mb_has_coeffs) {
3488
            GET_MQUANT();
3489
            s->current_picture.qscale_table[mb_pos] = mquant;
3490
            s->ac_pred = get_bits1(gb);
3491
            cbp = 0;
3492
            vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3493
        } else {
3494
            if(bmvtype == BMV_TYPE_INTERPOLATED) {
3495
                GET_MVDATA(dmv_x[0], dmv_y[0]);
3496
                if(!mb_has_coeffs) {
3497
                    /* interpolated skipped block */
3498
                    vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3499
                    vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3500
                    return;
3501
                }
3502
            }
3503
            vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);
3504
            if(!s->mb_intra) {
3505
                vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);
3506
            }
3507
            if(s->mb_intra)
3508
                s->ac_pred = get_bits1(gb);
3509
            cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
3510
            GET_MQUANT();
3511
            s->current_picture.qscale_table[mb_pos] = mquant;
3512
            if(!v->ttmbf && !s->mb_intra && mb_has_coeffs)
3513
                ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
3514
        }
3515
    }
3516
    dst_idx = 0;
3517
    for (i=0; i<6; i++)
3518
    {
3519
        s->dc_val[0][s->block_index[i]] = 0;
3520
        dst_idx += i >> 2;
3521
        val = ((cbp >> (5 - i)) & 1);
3522
        off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);
3523
        v->mb_type[0][s->block_index[i]] = s->mb_intra;
3524
        if(s->mb_intra) {
3525
            /* check if prediction blocks A and C are available */
3526
            v->a_avail = v->c_avail = 0;
3527
            if(i == 2 || i == 3 || !s->first_slice_line)
3528
                v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];
3529
            if(i == 1 || i == 3 || s->mb_x)
3530
                v->c_avail = v->mb_type[0][s->block_index[i] - 1];
3531

    
3532
            vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
3533
            if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
3534
            s->dsp.vc1_inv_trans_8x8(s->block[i]);
3535
            if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
3536
            s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
3537
        } else if(val) {
3538
            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);
3539
            if(!v->ttmbf && ttmb < 8) ttmb = -1;
3540
            first_block = 0;
3541
        }
3542
    }
3543
}
3544

    
3545
/** Decode blocks of I-frame
3546
 */
3547
static void vc1_decode_i_blocks(VC1Context *v)
3548
{
3549
    int k, j;
3550
    MpegEncContext *s = &v->s;
3551
    int cbp, val;
3552
    uint8_t *coded_val;
3553
    int mb_pos;
3554

    
3555
    /* select codingmode used for VLC tables selection */
3556
    switch(v->y_ac_table_index){
3557
    case 0:
3558
        v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3559
        break;
3560
    case 1:
3561
        v->codingset = CS_HIGH_MOT_INTRA;
3562
        break;
3563
    case 2:
3564
        v->codingset = CS_MID_RATE_INTRA;
3565
        break;
3566
    }
3567

    
3568
    switch(v->c_ac_table_index){
3569
    case 0:
3570
        v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3571
        break;
3572
    case 1:
3573
        v->codingset2 = CS_HIGH_MOT_INTER;
3574
        break;
3575
    case 2:
3576
        v->codingset2 = CS_MID_RATE_INTER;
3577
        break;
3578
    }
3579

    
3580
    /* Set DC scale - y and c use the same */
3581
    s->y_dc_scale = s->y_dc_scale_table[v->pq];
3582
    s->c_dc_scale = s->c_dc_scale_table[v->pq];
3583

    
3584
    //do frame decode
3585
    s->mb_x = s->mb_y = 0;
3586
    s->mb_intra = 1;
3587
    s->first_slice_line = 1;
3588
    for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3589
        for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3590
            ff_init_block_index(s);
3591
            ff_update_block_index(s);
3592
            s->dsp.clear_blocks(s->block[0]);
3593
            mb_pos = s->mb_x + s->mb_y * s->mb_width;
3594
            s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3595
            s->current_picture.qscale_table[mb_pos] = v->pq;
3596
            s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3597
            s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3598

    
3599
            // do actual MB decoding and displaying
3600
            cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3601
            v->s.ac_pred = get_bits1(&v->s.gb);
3602

    
3603
            for(k = 0; k < 6; k++) {
3604
                val = ((cbp >> (5 - k)) & 1);
3605

    
3606
                if (k < 4) {
3607
                    int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3608
                    val = val ^ pred;
3609
                    *coded_val = val;
3610
                }
3611
                cbp |= val << (5 - k);
3612

    
3613
                vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);
3614

    
3615
                s->dsp.vc1_inv_trans_8x8(s->block[k]);
3616
                if(v->pq >= 9 && v->overlap) {
3617
                    for(j = 0; j < 64; j++) s->block[k][j] += 128;
3618
                }
3619
            }
3620

    
3621
            vc1_put_block(v, s->block);
3622
            if(v->pq >= 9 && v->overlap) {
3623
                if(s->mb_x) {
3624
                    s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3625
                    s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3626
                    if(!(s->flags & CODEC_FLAG_GRAY)) {
3627
                        s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3628
                        s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3629
                    }
3630
                }
3631
                s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3632
                s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3633
                if(!s->first_slice_line) {
3634
                    s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3635
                    s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3636
                    if(!(s->flags & CODEC_FLAG_GRAY)) {
3637
                        s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3638
                        s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3639
                    }
3640
                }
3641
                s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3642
                s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3643
            }
3644
            if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3645

    
3646
            if(get_bits_count(&s->gb) > v->bits) {
3647
                ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3648
                av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3649
                return;
3650
            }
3651
        }
3652
        ff_draw_horiz_band(s, s->mb_y * 16, 16);
3653
        s->first_slice_line = 0;
3654
    }
3655
    ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3656
}
3657

    
3658
/** Decode blocks of I-frame for advanced profile
3659
 */
3660
static void vc1_decode_i_blocks_adv(VC1Context *v)
3661
{
3662
    int k, j;
3663
    MpegEncContext *s = &v->s;
3664
    int cbp, val;
3665
    uint8_t *coded_val;
3666
    int mb_pos;
3667
    int mquant = v->pq;
3668
    int mqdiff;
3669
    int overlap;
3670
    GetBitContext *gb = &s->gb;
3671

    
3672
    /* select codingmode used for VLC tables selection */
3673
    switch(v->y_ac_table_index){
3674
    case 0:
3675
        v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3676
        break;
3677
    case 1:
3678
        v->codingset = CS_HIGH_MOT_INTRA;
3679
        break;
3680
    case 2:
3681
        v->codingset = CS_MID_RATE_INTRA;
3682
        break;
3683
    }
3684

    
3685
    switch(v->c_ac_table_index){
3686
    case 0:
3687
        v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3688
        break;
3689
    case 1:
3690
        v->codingset2 = CS_HIGH_MOT_INTER;
3691
        break;
3692
    case 2:
3693
        v->codingset2 = CS_MID_RATE_INTER;
3694
        break;
3695
    }
3696

    
3697
    //do frame decode
3698
    s->mb_x = s->mb_y = 0;
3699
    s->mb_intra = 1;
3700
    s->first_slice_line = 1;
3701
    for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3702
        for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3703
            ff_init_block_index(s);
3704
            ff_update_block_index(s);
3705
            s->dsp.clear_blocks(s->block[0]);
3706
            mb_pos = s->mb_x + s->mb_y * s->mb_stride;
3707
            s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
3708
            s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
3709
            s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
3710

    
3711
            // do actual MB decoding and displaying
3712
            cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);
3713
            if(v->acpred_is_raw)
3714
                v->s.ac_pred = get_bits1(&v->s.gb);
3715
            else
3716
                v->s.ac_pred = v->acpred_plane[mb_pos];
3717

    
3718
            if(v->condover == CONDOVER_SELECT) {
3719
                if(v->overflg_is_raw)
3720
                    overlap = get_bits1(&v->s.gb);
3721
                else
3722
                    overlap = v->over_flags_plane[mb_pos];
3723
            } else
3724
                overlap = (v->condover == CONDOVER_ALL);
3725

    
3726
            GET_MQUANT();
3727

    
3728
            s->current_picture.qscale_table[mb_pos] = mquant;
3729
            /* Set DC scale - y and c use the same */
3730
            s->y_dc_scale = s->y_dc_scale_table[mquant];
3731
            s->c_dc_scale = s->c_dc_scale_table[mquant];
3732

    
3733
            for(k = 0; k < 6; k++) {
3734
                val = ((cbp >> (5 - k)) & 1);
3735

    
3736
                if (k < 4) {
3737
                    int pred = vc1_coded_block_pred(&v->s, k, &coded_val);
3738
                    val = val ^ pred;
3739
                    *coded_val = val;
3740
                }
3741
                cbp |= val << (5 - k);
3742

    
3743
                v->a_avail = !s->first_slice_line || (k==2 || k==3);
3744
                v->c_avail = !!s->mb_x || (k==1 || k==3);
3745

    
3746
                vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);
3747

    
3748
                s->dsp.vc1_inv_trans_8x8(s->block[k]);
3749
                for(j = 0; j < 64; j++) s->block[k][j] += 128;
3750
            }
3751

    
3752
            vc1_put_block(v, s->block);
3753
            if(overlap) {
3754
                if(s->mb_x) {
3755
                    s->dsp.vc1_h_overlap(s->dest[0], s->linesize);
3756
                    s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3757
                    if(!(s->flags & CODEC_FLAG_GRAY)) {
3758
                        s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);
3759
                        s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);
3760
                    }
3761
                }
3762
                s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);
3763
                s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3764
                if(!s->first_slice_line) {
3765
                    s->dsp.vc1_v_overlap(s->dest[0], s->linesize);
3766
                    s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);
3767
                    if(!(s->flags & CODEC_FLAG_GRAY)) {
3768
                        s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);
3769
                        s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);
3770
                    }
3771
                }
3772
                s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);
3773
                s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);
3774
            }
3775
            if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);
3776

    
3777
            if(get_bits_count(&s->gb) > v->bits) {
3778
                ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3779
                av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
3780
                return;
3781
            }
3782
        }
3783
        ff_draw_horiz_band(s, s->mb_y * 16, 16);
3784
        s->first_slice_line = 0;
3785
    }
3786
    ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3787
}
3788

    
3789
static void vc1_decode_p_blocks(VC1Context *v)
3790
{
3791
    MpegEncContext *s = &v->s;
3792

    
3793
    /* select codingmode used for VLC tables selection */
3794
    switch(v->c_ac_table_index){
3795
    case 0:
3796
        v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3797
        break;
3798
    case 1:
3799
        v->codingset = CS_HIGH_MOT_INTRA;
3800
        break;
3801
    case 2:
3802
        v->codingset = CS_MID_RATE_INTRA;
3803
        break;
3804
    }
3805

    
3806
    switch(v->c_ac_table_index){
3807
    case 0:
3808
        v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3809
        break;
3810
    case 1:
3811
        v->codingset2 = CS_HIGH_MOT_INTER;
3812
        break;
3813
    case 2:
3814
        v->codingset2 = CS_MID_RATE_INTER;
3815
        break;
3816
    }
3817

    
3818
    s->first_slice_line = 1;
3819
    memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);
3820
    for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3821
        for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3822
            ff_init_block_index(s);
3823
            ff_update_block_index(s);
3824
            s->dsp.clear_blocks(s->block[0]);
3825

    
3826
            vc1_decode_p_mb(v);
3827
            if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3828
                ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3829
                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);
3830
                return;
3831
            }
3832
        }
3833
        memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride);
3834
        ff_draw_horiz_band(s, s->mb_y * 16, 16);
3835
        s->first_slice_line = 0;
3836
    }
3837
    ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3838
}
3839

    
3840
static void vc1_decode_b_blocks(VC1Context *v)
3841
{
3842
    MpegEncContext *s = &v->s;
3843

    
3844
    /* select codingmode used for VLC tables selection */
3845
    switch(v->c_ac_table_index){
3846
    case 0:
3847
        v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;
3848
        break;
3849
    case 1:
3850
        v->codingset = CS_HIGH_MOT_INTRA;
3851
        break;
3852
    case 2:
3853
        v->codingset = CS_MID_RATE_INTRA;
3854
        break;
3855
    }
3856

    
3857
    switch(v->c_ac_table_index){
3858
    case 0:
3859
        v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;
3860
        break;
3861
    case 1:
3862
        v->codingset2 = CS_HIGH_MOT_INTER;
3863
        break;
3864
    case 2:
3865
        v->codingset2 = CS_MID_RATE_INTER;
3866
        break;
3867
    }
3868

    
3869
    s->first_slice_line = 1;
3870
    for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3871
        for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
3872
            ff_init_block_index(s);
3873
            ff_update_block_index(s);
3874
            s->dsp.clear_blocks(s->block[0]);
3875

    
3876
            vc1_decode_b_mb(v);
3877
            if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {
3878
                ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));
3879
                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);
3880
                return;
3881
            }
3882
            if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[s->mb_x + s->mb_y *s->mb_stride]);
3883
        }
3884
        ff_draw_horiz_band(s, s->mb_y * 16, 16);
3885
        s->first_slice_line = 0;
3886
    }
3887
    ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3888
}
3889

    
3890
static void vc1_decode_skip_blocks(VC1Context *v)
3891
{
3892
    MpegEncContext *s = &v->s;
3893

    
3894
    ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));
3895
    s->first_slice_line = 1;
3896
    for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
3897
        s->mb_x = 0;
3898
        ff_init_block_index(s);
3899
        ff_update_block_index(s);
3900
        memcpy(s->dest[0], s->last_picture.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16);
3901
        memcpy(s->dest[1], s->last_picture.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3902
        memcpy(s->dest[2], s->last_picture.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);
3903
        ff_draw_horiz_band(s, s->mb_y * 16, 16);
3904
        s->first_slice_line = 0;
3905
    }
3906
    s->pict_type = FF_P_TYPE;
3907
}
3908

    
3909
static void vc1_decode_blocks(VC1Context *v)
3910
{
3911

    
3912
    v->s.esc3_level_length = 0;
3913
    if(v->x8_type){
3914
        ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );
3915
    }else{
3916

    
3917
        switch(v->s.pict_type) {
3918
        case FF_I_TYPE:
3919
            if(v->profile == PROFILE_ADVANCED)
3920
                vc1_decode_i_blocks_adv(v);
3921
            else
3922
                vc1_decode_i_blocks(v);
3923
            break;
3924
        case FF_P_TYPE:
3925
            if(v->p_frame_skipped)
3926
                vc1_decode_skip_blocks(v);
3927
            else
3928
                vc1_decode_p_blocks(v);
3929
            break;
3930
        case FF_B_TYPE:
3931
            if(v->bi_type){
3932
                if(v->profile == PROFILE_ADVANCED)
3933
                    vc1_decode_i_blocks_adv(v);
3934
                else
3935
                    vc1_decode_i_blocks(v);
3936
            }else
3937
                vc1_decode_b_blocks(v);
3938
            break;
3939
        }
3940
    }
3941
}
3942

    
3943
/** Find VC-1 marker in buffer
3944
 * @return position where next marker starts or end of buffer if no marker found
3945
 */
3946
static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)
3947
{
3948
    uint32_t mrk = 0xFFFFFFFF;
3949

    
3950
    if(end-src < 4) return end;
3951
    while(src < end){
3952
        mrk = (mrk << 8) | *src++;
3953
        if(IS_MARKER(mrk))
3954
            return src-4;
3955
    }
3956
    return end;
3957
}
3958

    
3959
static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)
3960
{
3961
    int dsize = 0, i;
3962

    
3963
    if(size < 4){
3964
        for(dsize = 0; dsize < size; dsize++) *dst++ = *src++;
3965
        return size;
3966
    }
3967
    for(i = 0; i < size; i++, src++) {
3968
        if(src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {
3969
            dst[dsize++] = src[1];
3970
            src++;
3971
            i++;
3972
        } else
3973
            dst[dsize++] = *src;
3974
    }
3975
    return dsize;
3976
}
3977

    
3978
/** Initialize a VC1/WMV3 decoder
3979
 * @todo TODO: Handle VC-1 IDUs (Transport level?)
3980
 * @todo TODO: Decypher remaining bits in extra_data
3981
 */
3982
static av_cold int vc1_decode_init(AVCodecContext *avctx)
3983
{
3984
    VC1Context *v = avctx->priv_data;
3985
    MpegEncContext *s = &v->s;
3986
    GetBitContext gb;
3987

    
3988
    if (!avctx->extradata_size || !avctx->extradata) return -1;
3989
    if (!(avctx->flags & CODEC_FLAG_GRAY))
3990
        avctx->pix_fmt = PIX_FMT_YUV420P;
3991
    else
3992
        avctx->pix_fmt = PIX_FMT_GRAY8;
3993
    v->s.avctx = avctx;
3994
    avctx->flags |= CODEC_FLAG_EMU_EDGE;
3995
    v->s.flags |= CODEC_FLAG_EMU_EDGE;
3996

    
3997
    if(avctx->idct_algo==FF_IDCT_AUTO){
3998
        avctx->idct_algo=FF_IDCT_WMV2;
3999
    }
4000

    
4001
    if(ff_h263_decode_init(avctx) < 0)
4002
        return -1;
4003
    if (vc1_init_common(v) < 0) return -1;
4004

    
4005
    avctx->coded_width = avctx->width;
4006
    avctx->coded_height = avctx->height;
4007
    if (avctx->codec_id == CODEC_ID_WMV3)
4008
    {
4009
        int count = 0;
4010

    
4011
        // looks like WMV3 has a sequence header stored in the extradata
4012
        // advanced sequence header may be before the first frame
4013
        // the last byte of the extradata is a version number, 1 for the
4014
        // samples we can decode
4015

    
4016
        init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);
4017

    
4018
        if (decode_sequence_header(avctx, &gb) < 0)
4019
          return -1;
4020

    
4021
        count = avctx->extradata_size*8 - get_bits_count(&gb);
4022
        if (count>0)
4023
        {
4024
            av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",
4025
                   count, get_bits(&gb, count));
4026
        }
4027
        else if (count < 0)
4028
        {
4029
            av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);
4030
        }
4031
    } else { // VC1/WVC1
4032
        const uint8_t *start = avctx->extradata;
4033
        uint8_t *end = avctx->extradata + avctx->extradata_size;
4034
        const uint8_t *next;
4035
        int size, buf2_size;
4036
        uint8_t *buf2 = NULL;
4037
        int seq_initialized = 0, ep_initialized = 0;
4038

    
4039
        if(avctx->extradata_size < 16) {
4040
            av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);
4041
            return -1;
4042
        }
4043

    
4044
        buf2 = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
4045
        if(start[0]) start++; // in WVC1 extradata first byte is its size
4046
        next = start;
4047
        for(; next < end; start = next){
4048
            next = find_next_marker(start + 4, end);
4049
            size = next - start - 4;
4050
            if(size <= 0) continue;
4051
            buf2_size = vc1_unescape_buffer(start + 4, size, buf2);
4052
            init_get_bits(&gb, buf2, buf2_size * 8);
4053
            switch(AV_RB32(start)){
4054
            case VC1_CODE_SEQHDR:
4055
                if(decode_sequence_header(avctx, &gb) < 0){
4056
                    av_free(buf2);
4057
                    return -1;
4058
                }
4059
                seq_initialized = 1;
4060
                break;
4061
            case VC1_CODE_ENTRYPOINT:
4062
                if(decode_entry_point(avctx, &gb) < 0){
4063
                    av_free(buf2);
4064
                    return -1;
4065
                }
4066
                ep_initialized = 1;
4067
                break;
4068
            }
4069
        }
4070
        av_free(buf2);
4071
        if(!seq_initialized || !ep_initialized){
4072
            av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");
4073
            return -1;
4074
        }
4075
    }
4076
    avctx->has_b_frames= !!(avctx->max_b_frames);
4077
    s->low_delay = !avctx->has_b_frames;
4078

    
4079
    s->mb_width = (avctx->coded_width+15)>>4;
4080
    s->mb_height = (avctx->coded_height+15)>>4;
4081

    
4082
    /* Allocate mb bitplanes */
4083
    v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4084
    v->direct_mb_plane = av_malloc(s->mb_stride * s->mb_height);
4085
    v->acpred_plane = av_malloc(s->mb_stride * s->mb_height);
4086
    v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height);
4087

    
4088
    v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);
4089
    v->cbp = v->cbp_base + s->mb_stride;
4090

    
4091
    /* allocate block type info in that way so it could be used with s->block_index[] */
4092
    v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);
4093
    v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;
4094
    v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;
4095
    v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);
4096

    
4097
    /* Init coded blocks info */
4098
    if (v->profile == PROFILE_ADVANCED)
4099
    {
4100
//        if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
4101
//            return -1;
4102
//        if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
4103
//            return -1;
4104
    }
4105

    
4106
    ff_intrax8_common_init(&v->x8,s);
4107
    return 0;
4108
}
4109

    
4110

    
4111
/** Decode a VC1/WMV3 frame
4112
 * @todo TODO: Handle VC-1 IDUs (Transport level?)
4113
 */
4114
static int vc1_decode_frame(AVCodecContext *avctx,
4115
                            void *data, int *data_size,
4116
                            const uint8_t *buf, int buf_size)
4117
{
4118
    VC1Context *v = avctx->priv_data;
4119
    MpegEncContext *s = &v->s;
4120
    AVFrame *pict = data;
4121
    uint8_t *buf2 = NULL;
4122

    
4123
    /* no supplementary picture */
4124
    if (buf_size == 0) {
4125
        /* special case for last picture */
4126
        if (s->low_delay==0 && s->next_picture_ptr) {
4127
            *pict= *(AVFrame*)s->next_picture_ptr;
4128
            s->next_picture_ptr= NULL;
4129

    
4130
            *data_size = sizeof(AVFrame);
4131
        }
4132

    
4133
        return 0;
4134
    }
4135

    
4136
    /* We need to set current_picture_ptr before reading the header,
4137
     * otherwise we cannot store anything in there. */
4138
    if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){
4139
        int i= ff_find_unused_picture(s, 0);
4140
        s->current_picture_ptr= &s->picture[i];
4141
    }
4142

    
4143
    //for advanced profile we may need to parse and unescape data
4144
    if (avctx->codec_id == CODEC_ID_VC1) {
4145
        int buf_size2 = 0;
4146
        buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);
4147

    
4148
        if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */
4149
            const uint8_t *start, *end, *next;
4150
            int size;
4151

    
4152
            next = buf;
4153
            for(start = buf, end = buf + buf_size; next < end; start = next){
4154
                next = find_next_marker(start + 4, end);
4155
                size = next - start - 4;
4156
                if(size <= 0) continue;
4157
                switch(AV_RB32(start)){
4158
                case VC1_CODE_FRAME:
4159
                    buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4160
                    break;
4161
                case VC1_CODE_ENTRYPOINT: /* it should be before frame data */
4162
                    buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);
4163
                    init_get_bits(&s->gb, buf2, buf_size2*8);
4164
                    decode_entry_point(avctx, &s->gb);
4165
                    break;
4166
                case VC1_CODE_SLICE:
4167
                    av_log(avctx, AV_LOG_ERROR, "Sliced decoding is not implemented (yet)\n");
4168
                    av_free(buf2);
4169
                    return -1;
4170
                }
4171
            }
4172
        }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */
4173
            const uint8_t *divider;
4174

    
4175
            divider = find_next_marker(buf, buf + buf_size);
4176
            if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){
4177
                av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");
4178
                av_free(buf2);
4179
                return -1;
4180
            }
4181

    
4182
            buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);
4183
            // TODO
4184
            av_free(buf2);return -1;
4185
        }else{
4186
            buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);
4187
        }
4188
        init_get_bits(&s->gb, buf2, buf_size2*8);
4189
    } else
4190
        init_get_bits(&s->gb, buf, buf_size*8);
4191
    // do parse frame header
4192
    if(v->profile < PROFILE_ADVANCED) {
4193
        if(vc1_parse_frame_header(v, &s->gb) == -1) {
4194
            av_free(buf2);
4195
            return -1;
4196
        }
4197
    } else {
4198
        if(vc1_parse_frame_header_adv(v, &s->gb) == -1) {
4199
            av_free(buf2);
4200
            return -1;
4201
        }
4202
    }
4203

    
4204
    if(s->pict_type != FF_I_TYPE && !v->res_rtm_flag){
4205
        av_free(buf2);
4206
        return -1;
4207
    }
4208

    
4209
    // for hurry_up==5
4210
    s->current_picture.pict_type= s->pict_type;
4211
    s->current_picture.key_frame= s->pict_type == FF_I_TYPE;
4212

    
4213
    /* skip B-frames if we don't have reference frames */
4214
    if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){
4215
        av_free(buf2);
4216
        return -1;//buf_size;
4217
    }
4218
    /* skip b frames if we are in a hurry */
4219
    if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;
4220
    if(   (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
4221
       || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
4222
       ||  avctx->skip_frame >= AVDISCARD_ALL) {
4223
        av_free(buf2);
4224
        return buf_size;
4225
    }
4226
    /* skip everything if we are in a hurry>=5 */
4227
    if(avctx->hurry_up>=5) {
4228
        av_free(buf2);
4229
        return -1;//buf_size;
4230
    }
4231

    
4232
    if(s->next_p_frame_damaged){
4233
        if(s->pict_type==FF_B_TYPE)
4234
            return buf_size;
4235
        else
4236
            s->next_p_frame_damaged=0;
4237
    }
4238

    
4239
    if(MPV_frame_start(s, avctx) < 0) {
4240
        av_free(buf2);
4241
        return -1;
4242
    }
4243

    
4244
    s->me.qpel_put= s->dsp.put_qpel_pixels_tab;
4245
    s->me.qpel_avg= s->dsp.avg_qpel_pixels_tab;
4246

    
4247
    ff_er_frame_start(s);
4248

    
4249
    v->bits = buf_size * 8;
4250
    vc1_decode_blocks(v);
4251
//av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);
4252
//  if(get_bits_count(&s->gb) > buf_size * 8)
4253
//      return -1;
4254
    ff_er_frame_end(s);
4255

    
4256
    MPV_frame_end(s);
4257

    
4258
assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);
4259
assert(s->current_picture.pict_type == s->pict_type);
4260
    if (s->pict_type == FF_B_TYPE || s->low_delay) {
4261
        *pict= *(AVFrame*)s->current_picture_ptr;
4262
    } else if (s->last_picture_ptr != NULL) {
4263
        *pict= *(AVFrame*)s->last_picture_ptr;
4264
    }
4265

    
4266
    if(s->last_picture_ptr || s->low_delay){
4267
        *data_size = sizeof(AVFrame);
4268
        ff_print_debug_info(s, pict);
4269
    }
4270

    
4271
    /* Return the Picture timestamp as the frame number */
4272
    /* we subtract 1 because it is added on utils.c     */
4273
    avctx->frame_number = s->picture_number - 1;
4274

    
4275
    av_free(buf2);
4276
    return buf_size;
4277
}
4278

    
4279

    
4280
/** Close a VC1/WMV3 decoder
4281
 * @warning Initial try at using MpegEncContext stuff
4282
 */
4283
static av_cold int vc1_decode_end(AVCodecContext *avctx)
4284
{
4285
    VC1Context *v = avctx->priv_data;
4286

    
4287
    av_freep(&v->hrd_rate);
4288
    av_freep(&v->hrd_buffer);
4289
    MPV_common_end(&v->s);
4290
    av_freep(&v->mv_type_mb_plane);
4291
    av_freep(&v->direct_mb_plane);
4292
    av_freep(&v->acpred_plane);
4293
    av_freep(&v->over_flags_plane);
4294
    av_freep(&v->mb_type_base);
4295
    av_freep(&v->cbp_base);
4296
    ff_intrax8_common_end(&v->x8);
4297
    return 0;
4298
}
4299

    
4300

    
4301
AVCodec vc1_decoder = {
4302
    "vc1",
4303
    CODEC_TYPE_VIDEO,
4304
    CODEC_ID_VC1,
4305
    sizeof(VC1Context),
4306
    vc1_decode_init,
4307
    NULL,
4308
    vc1_decode_end,
4309
    vc1_decode_frame,
4310
    CODEC_CAP_DELAY,
4311
    NULL,
4312
    .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),
4313
};
4314

    
4315
AVCodec wmv3_decoder = {
4316
    "wmv3",
4317
    CODEC_TYPE_VIDEO,
4318
    CODEC_ID_WMV3,
4319
    sizeof(VC1Context),
4320
    vc1_decode_init,
4321
    NULL,
4322
    vc1_decode_end,
4323
    vc1_decode_frame,
4324
    CODEC_CAP_DELAY,
4325
    NULL,
4326
    .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),
4327
};