Statistics
| Branch: | Revision:

ffmpeg / libavcodec / rv40.c @ 2912e87a

History | View | Annotate | Download (26.5 KB)

1
/*
2
 * RV40 decoder
3
 * Copyright (c) 2007 Konstantin Shishkov
4
 *
5
 * This file is part of Libav.
6
 *
7
 * Libav is free software; you can redistribute it and/or
8
 * modify it under the terms of the GNU Lesser General Public
9
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
11
 *
12
 * Libav is distributed in the hope that it will be useful,
13
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15
 * Lesser General Public License for more details.
16
 *
17
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with Libav; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
 */
21

    
22
/**
23
 * @file
24
 * RV40 decoder
25
 */
26

    
27
#include "libavutil/imgutils.h"
28

    
29
#include "avcodec.h"
30
#include "dsputil.h"
31
#include "mpegvideo.h"
32
#include "golomb.h"
33

    
34
#include "rv34.h"
35
#include "rv40vlc2.h"
36
#include "rv40data.h"
37

    
38
static VLC aic_top_vlc;
39
static VLC aic_mode1_vlc[AIC_MODE1_NUM], aic_mode2_vlc[AIC_MODE2_NUM];
40
static VLC ptype_vlc[NUM_PTYPE_VLCS], btype_vlc[NUM_BTYPE_VLCS];
41

    
42
static const int16_t mode2_offs[] = {
43
       0,  614, 1222, 1794, 2410,  3014,  3586,  4202,  4792, 5382, 5966, 6542,
44
    7138, 7716, 8292, 8864, 9444, 10030, 10642, 11212, 11814
45
};
46

    
47
/**
48
 * Initialize all tables.
49
 */
50
static av_cold void rv40_init_tables(void)
51
{
52
    int i;
53
    static VLC_TYPE aic_table[1 << AIC_TOP_BITS][2];
54
    static VLC_TYPE aic_mode1_table[AIC_MODE1_NUM << AIC_MODE1_BITS][2];
55
    static VLC_TYPE aic_mode2_table[11814][2];
56
    static VLC_TYPE ptype_table[NUM_PTYPE_VLCS << PTYPE_VLC_BITS][2];
57
    static VLC_TYPE btype_table[NUM_BTYPE_VLCS << BTYPE_VLC_BITS][2];
58

    
59
    aic_top_vlc.table = aic_table;
60
    aic_top_vlc.table_allocated = 1 << AIC_TOP_BITS;
61
    init_vlc(&aic_top_vlc, AIC_TOP_BITS, AIC_TOP_SIZE,
62
             rv40_aic_top_vlc_bits,  1, 1,
63
             rv40_aic_top_vlc_codes, 1, 1, INIT_VLC_USE_NEW_STATIC);
64
    for(i = 0; i < AIC_MODE1_NUM; i++){
65
        // Every tenth VLC table is empty
66
        if((i % 10) == 9) continue;
67
        aic_mode1_vlc[i].table = &aic_mode1_table[i << AIC_MODE1_BITS];
68
        aic_mode1_vlc[i].table_allocated = 1 << AIC_MODE1_BITS;
69
        init_vlc(&aic_mode1_vlc[i], AIC_MODE1_BITS, AIC_MODE1_SIZE,
70
                 aic_mode1_vlc_bits[i],  1, 1,
71
                 aic_mode1_vlc_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);
72
    }
73
    for(i = 0; i < AIC_MODE2_NUM; i++){
74
        aic_mode2_vlc[i].table = &aic_mode2_table[mode2_offs[i]];
75
        aic_mode2_vlc[i].table_allocated = mode2_offs[i + 1] - mode2_offs[i];
76
        init_vlc(&aic_mode2_vlc[i], AIC_MODE2_BITS, AIC_MODE2_SIZE,
77
                 aic_mode2_vlc_bits[i],  1, 1,
78
                 aic_mode2_vlc_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
79
    }
80
    for(i = 0; i < NUM_PTYPE_VLCS; i++){
81
        ptype_vlc[i].table = &ptype_table[i << PTYPE_VLC_BITS];
82
        ptype_vlc[i].table_allocated = 1 << PTYPE_VLC_BITS;
83
        init_vlc_sparse(&ptype_vlc[i], PTYPE_VLC_BITS, PTYPE_VLC_SIZE,
84
                         ptype_vlc_bits[i],  1, 1,
85
                         ptype_vlc_codes[i], 1, 1,
86
                         ptype_vlc_syms,     1, 1, INIT_VLC_USE_NEW_STATIC);
87
    }
88
    for(i = 0; i < NUM_BTYPE_VLCS; i++){
89
        btype_vlc[i].table = &btype_table[i << BTYPE_VLC_BITS];
90
        btype_vlc[i].table_allocated = 1 << BTYPE_VLC_BITS;
91
        init_vlc_sparse(&btype_vlc[i], BTYPE_VLC_BITS, BTYPE_VLC_SIZE,
92
                         btype_vlc_bits[i],  1, 1,
93
                         btype_vlc_codes[i], 1, 1,
94
                         btype_vlc_syms,     1, 1, INIT_VLC_USE_NEW_STATIC);
95
    }
96
}
97

    
98
/**
99
 * Get stored dimension from bitstream.
100
 *
101
 * If the width/height is the standard one then it's coded as a 3-bit index.
102
 * Otherwise it is coded as escaped 8-bit portions.
103
 */
104
static int get_dimension(GetBitContext *gb, const int *dim)
105
{
106
    int t   = get_bits(gb, 3);
107
    int val = dim[t];
108
    if(val < 0)
109
        val = dim[get_bits1(gb) - val];
110
    if(!val){
111
        do{
112
            t = get_bits(gb, 8);
113
            val += t << 2;
114
        }while(t == 0xFF);
115
    }
116
    return val;
117
}
118

    
119
/**
120
 * Get encoded picture size - usually this is called from rv40_parse_slice_header.
121
 */
122
static void rv40_parse_picture_size(GetBitContext *gb, int *w, int *h)
123
{
124
    *w = get_dimension(gb, rv40_standard_widths);
125
    *h = get_dimension(gb, rv40_standard_heights);
126
}
127

    
128
static int rv40_parse_slice_header(RV34DecContext *r, GetBitContext *gb, SliceInfo *si)
129
{
130
    int mb_bits;
131
    int w = r->s.width, h = r->s.height;
132
    int mb_size;
133

    
134
    memset(si, 0, sizeof(SliceInfo));
135
    if(get_bits1(gb))
136
        return -1;
137
    si->type = get_bits(gb, 2);
138
    if(si->type == 1) si->type = 0;
139
    si->quant = get_bits(gb, 5);
140
    if(get_bits(gb, 2))
141
        return -1;
142
    si->vlc_set = get_bits(gb, 2);
143
    skip_bits1(gb);
144
    si->pts = get_bits(gb, 13);
145
    if(!si->type || !get_bits1(gb))
146
        rv40_parse_picture_size(gb, &w, &h);
147
    if(av_image_check_size(w, h, 0, r->s.avctx) < 0)
148
        return -1;
149
    si->width  = w;
150
    si->height = h;
151
    mb_size = ((w + 15) >> 4) * ((h + 15) >> 4);
152
    mb_bits = ff_rv34_get_start_offset(gb, mb_size);
153
    si->start = get_bits(gb, mb_bits);
154

    
155
    return 0;
156
}
157

    
158
/**
159
 * Decode 4x4 intra types array.
160
 */
161
static int rv40_decode_intra_types(RV34DecContext *r, GetBitContext *gb, int8_t *dst)
162
{
163
    MpegEncContext *s = &r->s;
164
    int i, j, k, v;
165
    int A, B, C;
166
    int pattern;
167
    int8_t *ptr;
168

    
169
    for(i = 0; i < 4; i++, dst += r->intra_types_stride){
170
        if(!i && s->first_slice_line){
171
            pattern = get_vlc2(gb, aic_top_vlc.table, AIC_TOP_BITS, 1);
172
            dst[0] = (pattern >> 2) & 2;
173
            dst[1] = (pattern >> 1) & 2;
174
            dst[2] =  pattern       & 2;
175
            dst[3] = (pattern << 1) & 2;
176
            continue;
177
        }
178
        ptr = dst;
179
        for(j = 0; j < 4; j++){
180
            /* Coefficients are read using VLC chosen by the prediction pattern
181
             * The first one (used for retrieving a pair of coefficients) is
182
             * constructed from the top, top right and left coefficients
183
             * The second one (used for retrieving only one coefficient) is
184
             * top + 10 * left.
185
             */
186
            A = ptr[-r->intra_types_stride + 1]; // it won't be used for the last coefficient in a row
187
            B = ptr[-r->intra_types_stride];
188
            C = ptr[-1];
189
            pattern = A + (B << 4) + (C << 8);
190
            for(k = 0; k < MODE2_PATTERNS_NUM; k++)
191
                if(pattern == rv40_aic_table_index[k])
192
                    break;
193
            if(j < 3 && k < MODE2_PATTERNS_NUM){ //pattern is found, decoding 2 coefficients
194
                v = get_vlc2(gb, aic_mode2_vlc[k].table, AIC_MODE2_BITS, 2);
195
                *ptr++ = v/9;
196
                *ptr++ = v%9;
197
                j++;
198
            }else{
199
                if(B != -1 && C != -1)
200
                    v = get_vlc2(gb, aic_mode1_vlc[B + C*10].table, AIC_MODE1_BITS, 1);
201
                else{ // tricky decoding
202
                    v = 0;
203
                    switch(C){
204
                    case -1: // code 0 -> 1, 1 -> 0
205
                        if(B < 2)
206
                            v = get_bits1(gb) ^ 1;
207
                        break;
208
                    case  0:
209
                    case  2: // code 0 -> 2, 1 -> 0
210
                        v = (get_bits1(gb) ^ 1) << 1;
211
                        break;
212
                    }
213
                }
214
                *ptr++ = v;
215
            }
216
        }
217
    }
218
    return 0;
219
}
220

    
221
/**
222
 * Decode macroblock information.
223
 */
224
static int rv40_decode_mb_info(RV34DecContext *r)
225
{
226
    MpegEncContext *s = &r->s;
227
    GetBitContext *gb = &s->gb;
228
    int q, i;
229
    int prev_type = 0;
230
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
231
    int blocks[RV34_MB_TYPES] = {0};
232
    int count = 0;
233

    
234
    if(!r->s.mb_skip_run)
235
        r->s.mb_skip_run = svq3_get_ue_golomb(gb) + 1;
236

    
237
    if(--r->s.mb_skip_run)
238
         return RV34_MB_SKIP;
239

    
240
    if(r->avail_cache[6-1])
241
        blocks[r->mb_type[mb_pos - 1]]++;
242
    if(r->avail_cache[6-4]){
243
        blocks[r->mb_type[mb_pos - s->mb_stride]]++;
244
        if(r->avail_cache[6-2])
245
            blocks[r->mb_type[mb_pos - s->mb_stride + 1]]++;
246
        if(r->avail_cache[6-5])
247
            blocks[r->mb_type[mb_pos - s->mb_stride - 1]]++;
248
    }
249

    
250
    for(i = 0; i < RV34_MB_TYPES; i++){
251
        if(blocks[i] > count){
252
            count = blocks[i];
253
            prev_type = i;
254
        }
255
    }
256
    if(s->pict_type == FF_P_TYPE){
257
        prev_type = block_num_to_ptype_vlc_num[prev_type];
258
        q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1);
259
        if(q < PBTYPE_ESCAPE)
260
            return q;
261
        q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1);
262
        av_log(s->avctx, AV_LOG_ERROR, "Dquant for P-frame\n");
263
    }else{
264
        prev_type = block_num_to_btype_vlc_num[prev_type];
265
        q = get_vlc2(gb, btype_vlc[prev_type].table, BTYPE_VLC_BITS, 1);
266
        if(q < PBTYPE_ESCAPE)
267
            return q;
268
        q = get_vlc2(gb, btype_vlc[prev_type].table, BTYPE_VLC_BITS, 1);
269
        av_log(s->avctx, AV_LOG_ERROR, "Dquant for B-frame\n");
270
    }
271
    return 0;
272
}
273

    
274
#define CLIP_SYMM(a, b) av_clip(a, -(b), b)
275
/**
276
 * weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1
277
 */
278
static inline void rv40_weak_loop_filter(uint8_t *src, const int step,
279
                                         const int filter_p1, const int filter_q1,
280
                                         const int alpha, const int beta,
281
                                         const int lim_p0q0,
282
                                         const int lim_q1, const int lim_p1,
283
                                         const int diff_p1p0, const int diff_q1q0,
284
                                         const int diff_p1p2, const int diff_q1q2)
285
{
286
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
287
    int t, u, diff;
288

    
289
    t = src[0*step] - src[-1*step];
290
    if(!t)
291
        return;
292
    u = (alpha * FFABS(t)) >> 7;
293
    if(u > 3 - (filter_p1 && filter_q1))
294
        return;
295

    
296
    t <<= 2;
297
    if(filter_p1 && filter_q1)
298
        t += src[-2*step] - src[1*step];
299
    diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);
300
    src[-1*step] = cm[src[-1*step] + diff];
301
    src[ 0*step] = cm[src[ 0*step] - diff];
302
    if(FFABS(diff_p1p2) <= beta && filter_p1){
303
        t = (diff_p1p0 + diff_p1p2 - diff) >> 1;
304
        src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];
305
    }
306
    if(FFABS(diff_q1q2) <= beta && filter_q1){
307
        t = (diff_q1q0 + diff_q1q2 + diff) >> 1;
308
        src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];
309
    }
310
}
311

    
312
static av_always_inline void rv40_adaptive_loop_filter(uint8_t *src, const int step,
313
                                             const int stride, const int dmode,
314
                                             const int lim_q1, const int lim_p1,
315
                                             const int alpha,
316
                                             const int beta, const int beta2,
317
                                             const int chroma, const int edge)
318
{
319
    int diff_p1p0[4], diff_q1q0[4], diff_p1p2[4], diff_q1q2[4];
320
    int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;
321
    uint8_t *ptr;
322
    int flag_strong0 = 1, flag_strong1 = 1;
323
    int filter_p1, filter_q1;
324
    int i;
325
    int lims;
326

    
327
    for(i = 0, ptr = src; i < 4; i++, ptr += stride){
328
        diff_p1p0[i] = ptr[-2*step] - ptr[-1*step];
329
        diff_q1q0[i] = ptr[ 1*step] - ptr[ 0*step];
330
        sum_p1p0 += diff_p1p0[i];
331
        sum_q1q0 += diff_q1q0[i];
332
    }
333
    filter_p1 = FFABS(sum_p1p0) < (beta<<2);
334
    filter_q1 = FFABS(sum_q1q0) < (beta<<2);
335
    if(!filter_p1 && !filter_q1)
336
        return;
337

    
338
    for(i = 0, ptr = src; i < 4; i++, ptr += stride){
339
        diff_p1p2[i] = ptr[-2*step] - ptr[-3*step];
340
        diff_q1q2[i] = ptr[ 1*step] - ptr[ 2*step];
341
        sum_p1p2 += diff_p1p2[i];
342
        sum_q1q2 += diff_q1q2[i];
343
    }
344

    
345
    if(edge){
346
        flag_strong0 = filter_p1 && (FFABS(sum_p1p2) < beta2);
347
        flag_strong1 = filter_q1 && (FFABS(sum_q1q2) < beta2);
348
    }else{
349
        flag_strong0 = flag_strong1 = 0;
350
    }
351

    
352
    lims = filter_p1 + filter_q1 + ((lim_q1 + lim_p1) >> 1) + 1;
353
    if(flag_strong0 && flag_strong1){ /* strong filtering */
354
        for(i = 0; i < 4; i++, src += stride){
355
            int sflag, p0, q0, p1, q1;
356
            int t = src[0*step] - src[-1*step];
357

    
358
            if(!t) continue;
359
            sflag = (alpha * FFABS(t)) >> 7;
360
            if(sflag > 1) continue;
361

    
362
            p0 = (25*src[-3*step] + 26*src[-2*step]
363
                + 26*src[-1*step]
364
                + 26*src[ 0*step] + 25*src[ 1*step] + rv40_dither_l[dmode + i]) >> 7;
365
            q0 = (25*src[-2*step] + 26*src[-1*step]
366
                + 26*src[ 0*step]
367
                + 26*src[ 1*step] + 25*src[ 2*step] + rv40_dither_r[dmode + i]) >> 7;
368
            if(sflag){
369
                p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);
370
                q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);
371
            }
372
            p1 = (25*src[-4*step] + 26*src[-3*step]
373
                + 26*src[-2*step]
374
                + 26*p0           + 25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;
375
            q1 = (25*src[-1*step] + 26*q0
376
                + 26*src[ 1*step]
377
                + 26*src[ 2*step] + 25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;
378
            if(sflag){
379
                p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);
380
                q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);
381
            }
382
            src[-2*step] = p1;
383
            src[-1*step] = p0;
384
            src[ 0*step] = q0;
385
            src[ 1*step] = q1;
386
            if(!chroma){
387
                src[-3*step] = (25*src[-1*step] + 26*src[-2*step] + 51*src[-3*step] + 26*src[-4*step] + 64) >> 7;
388
                src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] + 51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;
389
            }
390
        }
391
    }else if(filter_p1 && filter_q1){
392
        for(i = 0; i < 4; i++, src += stride)
393
            rv40_weak_loop_filter(src, step, 1, 1, alpha, beta, lims, lim_q1, lim_p1,
394
                                  diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]);
395
    }else{
396
        for(i = 0; i < 4; i++, src += stride)
397
            rv40_weak_loop_filter(src, step, filter_p1, filter_q1,
398
                                  alpha, beta, lims>>1, lim_q1>>1, lim_p1>>1,
399
                                  diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]);
400
    }
401
}
402

    
403
static void rv40_v_loop_filter(uint8_t *src, int stride, int dmode,
404
                               int lim_q1, int lim_p1,
405
                               int alpha, int beta, int beta2, int chroma, int edge){
406
    rv40_adaptive_loop_filter(src, 1, stride, dmode, lim_q1, lim_p1,
407
                              alpha, beta, beta2, chroma, edge);
408
}
409
static void rv40_h_loop_filter(uint8_t *src, int stride, int dmode,
410
                               int lim_q1, int lim_p1,
411
                               int alpha, int beta, int beta2, int chroma, int edge){
412
    rv40_adaptive_loop_filter(src, stride, 1, dmode, lim_q1, lim_p1,
413
                              alpha, beta, beta2, chroma, edge);
414
}
415

    
416
enum RV40BlockPos{
417
    POS_CUR,
418
    POS_TOP,
419
    POS_LEFT,
420
    POS_BOTTOM,
421
};
422

    
423
#define MASK_CUR          0x0001
424
#define MASK_RIGHT        0x0008
425
#define MASK_BOTTOM       0x0010
426
#define MASK_TOP          0x1000
427
#define MASK_Y_TOP_ROW    0x000F
428
#define MASK_Y_LAST_ROW   0xF000
429
#define MASK_Y_LEFT_COL   0x1111
430
#define MASK_Y_RIGHT_COL  0x8888
431
#define MASK_C_TOP_ROW    0x0003
432
#define MASK_C_LAST_ROW   0x000C
433
#define MASK_C_LEFT_COL   0x0005
434
#define MASK_C_RIGHT_COL  0x000A
435

    
436
static const int neighbour_offs_x[4] = { 0,  0, -1, 0 };
437
static const int neighbour_offs_y[4] = { 0, -1,  0, 1 };
438

    
439
/**
440
 * RV40 loop filtering function
441
 */
442
static void rv40_loop_filter(RV34DecContext *r, int row)
443
{
444
    MpegEncContext *s = &r->s;
445
    int mb_pos, mb_x;
446
    int i, j, k;
447
    uint8_t *Y, *C;
448
    int alpha, beta, betaY, betaC;
449
    int q;
450
    int mbtype[4];   ///< current macroblock and its neighbours types
451
    /**
452
     * flags indicating that macroblock can be filtered with strong filter
453
     * it is set only for intra coded MB and MB with DCs coded separately
454
     */
455
    int mb_strong[4];
456
    int clip[4];     ///< MB filter clipping value calculated from filtering strength
457
    /**
458
     * coded block patterns for luma part of current macroblock and its neighbours
459
     * Format:
460
     * LSB corresponds to the top left block,
461
     * each nibble represents one row of subblocks.
462
     */
463
    int cbp[4];
464
    /**
465
     * coded block patterns for chroma part of current macroblock and its neighbours
466
     * Format is the same as for luma with two subblocks in a row.
467
     */
468
    int uvcbp[4][2];
469
    /**
470
     * This mask represents the pattern of luma subblocks that should be filtered
471
     * in addition to the coded ones because because they lie at the edge of
472
     * 8x8 block with different enough motion vectors
473
     */
474
    int mvmasks[4];
475

    
476
    mb_pos = row * s->mb_stride;
477
    for(mb_x = 0; mb_x < s->mb_width; mb_x++, mb_pos++){
478
        int mbtype = s->current_picture_ptr->mb_type[mb_pos];
479
        if(IS_INTRA(mbtype) || IS_SEPARATE_DC(mbtype))
480
            r->cbp_luma  [mb_pos] = r->deblock_coefs[mb_pos] = 0xFFFF;
481
        if(IS_INTRA(mbtype))
482
            r->cbp_chroma[mb_pos] = 0xFF;
483
    }
484
    mb_pos = row * s->mb_stride;
485
    for(mb_x = 0; mb_x < s->mb_width; mb_x++, mb_pos++){
486
        int y_h_deblock, y_v_deblock;
487
        int c_v_deblock[2], c_h_deblock[2];
488
        int clip_left;
489
        int avail[4];
490
        int y_to_deblock, c_to_deblock[2];
491

    
492
        q = s->current_picture_ptr->qscale_table[mb_pos];
493
        alpha = rv40_alpha_tab[q];
494
        beta  = rv40_beta_tab [q];
495
        betaY = betaC = beta * 3;
496
        if(s->width * s->height <= 176*144)
497
            betaY += beta;
498

    
499
        avail[0] = 1;
500
        avail[1] = row;
501
        avail[2] = mb_x;
502
        avail[3] = row < s->mb_height - 1;
503
        for(i = 0; i < 4; i++){
504
            if(avail[i]){
505
                int pos = mb_pos + neighbour_offs_x[i] + neighbour_offs_y[i]*s->mb_stride;
506
                mvmasks[i] = r->deblock_coefs[pos];
507
                mbtype [i] = s->current_picture_ptr->mb_type[pos];
508
                cbp    [i] = r->cbp_luma[pos];
509
                uvcbp[i][0] = r->cbp_chroma[pos] & 0xF;
510
                uvcbp[i][1] = r->cbp_chroma[pos] >> 4;
511
            }else{
512
                mvmasks[i] = 0;
513
                mbtype [i] = mbtype[0];
514
                cbp    [i] = 0;
515
                uvcbp[i][0] = uvcbp[i][1] = 0;
516
            }
517
            mb_strong[i] = IS_INTRA(mbtype[i]) || IS_SEPARATE_DC(mbtype[i]);
518
            clip[i] = rv40_filter_clip_tbl[mb_strong[i] + 1][q];
519
        }
520
        y_to_deblock =  mvmasks[POS_CUR]
521
                     | (mvmasks[POS_BOTTOM] << 16);
522
        /* This pattern contains bits signalling that horizontal edges of
523
         * the current block can be filtered.
524
         * That happens when either of adjacent subblocks is coded or lies on
525
         * the edge of 8x8 blocks with motion vectors differing by more than
526
         * 3/4 pel in any component (any edge orientation for some reason).
527
         */
528
        y_h_deblock =   y_to_deblock
529
                    | ((cbp[POS_CUR]                           <<  4) & ~MASK_Y_TOP_ROW)
530
                    | ((cbp[POS_TOP]        & MASK_Y_LAST_ROW) >> 12);
531
        /* This pattern contains bits signalling that vertical edges of
532
         * the current block can be filtered.
533
         * That happens when either of adjacent subblocks is coded or lies on
534
         * the edge of 8x8 blocks with motion vectors differing by more than
535
         * 3/4 pel in any component (any edge orientation for some reason).
536
         */
537
        y_v_deblock =   y_to_deblock
538
                    | ((cbp[POS_CUR]                      << 1) & ~MASK_Y_LEFT_COL)
539
                    | ((cbp[POS_LEFT] & MASK_Y_RIGHT_COL) >> 3);
540
        if(!mb_x)
541
            y_v_deblock &= ~MASK_Y_LEFT_COL;
542
        if(!row)
543
            y_h_deblock &= ~MASK_Y_TOP_ROW;
544
        if(row == s->mb_height - 1 || (mb_strong[POS_CUR] || mb_strong[POS_BOTTOM]))
545
            y_h_deblock &= ~(MASK_Y_TOP_ROW << 16);
546
        /* Calculating chroma patterns is similar and easier since there is
547
         * no motion vector pattern for them.
548
         */
549
        for(i = 0; i < 2; i++){
550
            c_to_deblock[i] = (uvcbp[POS_BOTTOM][i] << 4) | uvcbp[POS_CUR][i];
551
            c_v_deblock[i] =   c_to_deblock[i]
552
                           | ((uvcbp[POS_CUR] [i]                       << 1) & ~MASK_C_LEFT_COL)
553
                           | ((uvcbp[POS_LEFT][i]   & MASK_C_RIGHT_COL) >> 1);
554
            c_h_deblock[i] =   c_to_deblock[i]
555
                           | ((uvcbp[POS_TOP][i]    & MASK_C_LAST_ROW)  >> 2)
556
                           |  (uvcbp[POS_CUR][i]                        << 2);
557
            if(!mb_x)
558
                c_v_deblock[i] &= ~MASK_C_LEFT_COL;
559
            if(!row)
560
                c_h_deblock[i] &= ~MASK_C_TOP_ROW;
561
            if(row == s->mb_height - 1 || mb_strong[POS_CUR] || mb_strong[POS_BOTTOM])
562
                c_h_deblock[i] &= ~(MASK_C_TOP_ROW << 4);
563
        }
564

    
565
        for(j = 0; j < 16; j += 4){
566
            Y = s->current_picture_ptr->data[0] + mb_x*16 + (row*16 + j) * s->linesize;
567
            for(i = 0; i < 4; i++, Y += 4){
568
                int ij = i + j;
569
                int clip_cur = y_to_deblock & (MASK_CUR << ij) ? clip[POS_CUR] : 0;
570
                int dither = j ? ij : i*4;
571

    
572
                // if bottom block is coded then we can filter its top edge
573
                // (or bottom edge of this block, which is the same)
574
                if(y_h_deblock & (MASK_BOTTOM << ij)){
575
                    rv40_h_loop_filter(Y+4*s->linesize, s->linesize, dither,
576
                                       y_to_deblock & (MASK_BOTTOM << ij) ? clip[POS_CUR] : 0,
577
                                       clip_cur,
578
                                       alpha, beta, betaY, 0, 0);
579
                }
580
                // filter left block edge in ordinary mode (with low filtering strength)
581
                if(y_v_deblock & (MASK_CUR << ij) && (i || !(mb_strong[POS_CUR] || mb_strong[POS_LEFT]))){
582
                    if(!i)
583
                        clip_left = mvmasks[POS_LEFT] & (MASK_RIGHT << j) ? clip[POS_LEFT] : 0;
584
                    else
585
                        clip_left = y_to_deblock & (MASK_CUR << (ij-1)) ? clip[POS_CUR] : 0;
586
                    rv40_v_loop_filter(Y, s->linesize, dither,
587
                                       clip_cur,
588
                                       clip_left,
589
                                       alpha, beta, betaY, 0, 0);
590
                }
591
                // filter top edge of the current macroblock when filtering strength is high
592
                if(!j && y_h_deblock & (MASK_CUR << i) && (mb_strong[POS_CUR] || mb_strong[POS_TOP])){
593
                    rv40_h_loop_filter(Y, s->linesize, dither,
594
                                       clip_cur,
595
                                       mvmasks[POS_TOP] & (MASK_TOP << i) ? clip[POS_TOP] : 0,
596
                                       alpha, beta, betaY, 0, 1);
597
                }
598
                // filter left block edge in edge mode (with high filtering strength)
599
                if(y_v_deblock & (MASK_CUR << ij) && !i && (mb_strong[POS_CUR] || mb_strong[POS_LEFT])){
600
                    clip_left = mvmasks[POS_LEFT] & (MASK_RIGHT << j) ? clip[POS_LEFT] : 0;
601
                    rv40_v_loop_filter(Y, s->linesize, dither,
602
                                       clip_cur,
603
                                       clip_left,
604
                                       alpha, beta, betaY, 0, 1);
605
                }
606
            }
607
        }
608
        for(k = 0; k < 2; k++){
609
            for(j = 0; j < 2; j++){
610
                C = s->current_picture_ptr->data[k+1] + mb_x*8 + (row*8 + j*4) * s->uvlinesize;
611
                for(i = 0; i < 2; i++, C += 4){
612
                    int ij = i + j*2;
613
                    int clip_cur = c_to_deblock[k] & (MASK_CUR << ij) ? clip[POS_CUR] : 0;
614
                    if(c_h_deblock[k] & (MASK_CUR << (ij+2))){
615
                        int clip_bot = c_to_deblock[k] & (MASK_CUR << (ij+2)) ? clip[POS_CUR] : 0;
616
                        rv40_h_loop_filter(C+4*s->uvlinesize, s->uvlinesize, i*8,
617
                                           clip_bot,
618
                                           clip_cur,
619
                                           alpha, beta, betaC, 1, 0);
620
                    }
621
                    if((c_v_deblock[k] & (MASK_CUR << ij)) && (i || !(mb_strong[POS_CUR] || mb_strong[POS_LEFT]))){
622
                        if(!i)
623
                            clip_left = uvcbp[POS_LEFT][k] & (MASK_CUR << (2*j+1)) ? clip[POS_LEFT] : 0;
624
                        else
625
                            clip_left = c_to_deblock[k]    & (MASK_CUR << (ij-1))  ? clip[POS_CUR]  : 0;
626
                        rv40_v_loop_filter(C, s->uvlinesize, j*8,
627
                                           clip_cur,
628
                                           clip_left,
629
                                           alpha, beta, betaC, 1, 0);
630
                    }
631
                    if(!j && c_h_deblock[k] & (MASK_CUR << ij) && (mb_strong[POS_CUR] || mb_strong[POS_TOP])){
632
                        int clip_top = uvcbp[POS_TOP][k] & (MASK_CUR << (ij+2)) ? clip[POS_TOP] : 0;
633
                        rv40_h_loop_filter(C, s->uvlinesize, i*8,
634
                                           clip_cur,
635
                                           clip_top,
636
                                           alpha, beta, betaC, 1, 1);
637
                    }
638
                    if(c_v_deblock[k] & (MASK_CUR << ij) && !i && (mb_strong[POS_CUR] || mb_strong[POS_LEFT])){
639
                        clip_left = uvcbp[POS_LEFT][k] & (MASK_CUR << (2*j+1)) ? clip[POS_LEFT] : 0;
640
                        rv40_v_loop_filter(C, s->uvlinesize, j*8,
641
                                           clip_cur,
642
                                           clip_left,
643
                                           alpha, beta, betaC, 1, 1);
644
                    }
645
                }
646
            }
647
        }
648
    }
649
}
650

    
651
/**
652
 * Initialize decoder.
653
 */
654
static av_cold int rv40_decode_init(AVCodecContext *avctx)
655
{
656
    RV34DecContext *r = avctx->priv_data;
657

    
658
    r->rv30 = 0;
659
    ff_rv34_decode_init(avctx);
660
    if(!aic_top_vlc.bits)
661
        rv40_init_tables();
662
    r->parse_slice_header = rv40_parse_slice_header;
663
    r->decode_intra_types = rv40_decode_intra_types;
664
    r->decode_mb_info     = rv40_decode_mb_info;
665
    r->loop_filter        = rv40_loop_filter;
666
    r->luma_dc_quant_i = rv40_luma_dc_quant[0];
667
    r->luma_dc_quant_p = rv40_luma_dc_quant[1];
668
    return 0;
669
}
670

    
671
AVCodec ff_rv40_decoder = {
672
    "rv40",
673
    AVMEDIA_TYPE_VIDEO,
674
    CODEC_ID_RV40,
675
    sizeof(RV34DecContext),
676
    rv40_decode_init,
677
    NULL,
678
    ff_rv34_decode_end,
679
    ff_rv34_decode_frame,
680
    CODEC_CAP_DR1 | CODEC_CAP_DELAY,
681
    .flush = ff_mpeg_flush,
682
    .long_name = NULL_IF_CONFIG_SMALL("RealVideo 4.0"),
683
    .pix_fmts= ff_pixfmt_list_420,
684
};