Statistics
| Branch: | Revision:

ffmpeg / libavcodec / vc1dsp.c @ 7919d10c

History | View | Annotate | Download (13.8 KB)

1
/*
2
 * VC-1 and WMV3 decoder - DSP functions
3
 * Copyright (c) 2006 Konstantin Shishkov
4
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg 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
 * FFmpeg 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 FFmpeg; 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 vc1dsp.c
24
 * VC-1 and WMV3 decoder
25
 *
26
 */
27

    
28
#include "dsputil.h"
29

    
30

    
31
/** Apply overlap transform to horizontal edge
32
*/
33
static void vc1_v_overlap_c(uint8_t* src, int stride)
34
{
35
    int i;
36
    int a, b, c, d;
37
    int d1, d2;
38
    int rnd = 1;
39
    for(i = 0; i < 8; i++) {
40
        a = src[-2*stride];
41
        b = src[-stride];
42
        c = src[0];
43
        d = src[stride];
44
        d1 = (a - d + 3 + rnd) >> 3;
45
        d2 = (a - d + b - c + 4 - rnd) >> 3;
46

    
47
        src[-2*stride] = a - d1;
48
        src[-stride] = b - d2;
49
        src[0] = c + d2;
50
        src[stride] = d + d1;
51
        src++;
52
        rnd = !rnd;
53
    }
54
}
55

    
56
/** Apply overlap transform to vertical edge
57
*/
58
static void vc1_h_overlap_c(uint8_t* src, int stride)
59
{
60
    int i;
61
    int a, b, c, d;
62
    int d1, d2;
63
    int rnd = 1;
64
    for(i = 0; i < 8; i++) {
65
        a = src[-2];
66
        b = src[-1];
67
        c = src[0];
68
        d = src[1];
69
        d1 = (a - d + 3 + rnd) >> 3;
70
        d2 = (a - d + b - c + 4 - rnd) >> 3;
71

    
72
        src[-2] = a - d1;
73
        src[-1] = b - d2;
74
        src[0] = c + d2;
75
        src[1] = d + d1;
76
        src += stride;
77
        rnd = !rnd;
78
    }
79
}
80

    
81

    
82
/** Do inverse transform on 8x8 block
83
*/
84
static void vc1_inv_trans_8x8_c(DCTELEM block[64])
85
{
86
    int i;
87
    register int t1,t2,t3,t4,t5,t6,t7,t8;
88
    DCTELEM *src, *dst;
89

    
90
    src = block;
91
    dst = block;
92
    for(i = 0; i < 8; i++){
93
        t1 = 12 * (src[0] + src[4]);
94
        t2 = 12 * (src[0] - src[4]);
95
        t3 = 16 * src[2] +  6 * src[6];
96
        t4 =  6 * src[2] - 16 * src[6];
97

    
98
        t5 = t1 + t3;
99
        t6 = t2 + t4;
100
        t7 = t2 - t4;
101
        t8 = t1 - t3;
102

    
103
        t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
104
        t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
105
        t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
106
        t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
107

    
108
        dst[0] = (t5 + t1 + 4) >> 3;
109
        dst[1] = (t6 + t2 + 4) >> 3;
110
        dst[2] = (t7 + t3 + 4) >> 3;
111
        dst[3] = (t8 + t4 + 4) >> 3;
112
        dst[4] = (t8 - t4 + 4) >> 3;
113
        dst[5] = (t7 - t3 + 4) >> 3;
114
        dst[6] = (t6 - t2 + 4) >> 3;
115
        dst[7] = (t5 - t1 + 4) >> 3;
116

    
117
        src += 8;
118
        dst += 8;
119
    }
120

    
121
    src = block;
122
    dst = block;
123
    for(i = 0; i < 8; i++){
124
        t1 = 12 * (src[ 0] + src[32]);
125
        t2 = 12 * (src[ 0] - src[32]);
126
        t3 = 16 * src[16] +  6 * src[48];
127
        t4 =  6 * src[16] - 16 * src[48];
128

    
129
        t5 = t1 + t3;
130
        t6 = t2 + t4;
131
        t7 = t2 - t4;
132
        t8 = t1 - t3;
133

    
134
        t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
135
        t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
136
        t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
137
        t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
138

    
139
        dst[ 0] = (t5 + t1 + 64) >> 7;
140
        dst[ 8] = (t6 + t2 + 64) >> 7;
141
        dst[16] = (t7 + t3 + 64) >> 7;
142
        dst[24] = (t8 + t4 + 64) >> 7;
143
        dst[32] = (t8 - t4 + 64 + 1) >> 7;
144
        dst[40] = (t7 - t3 + 64 + 1) >> 7;
145
        dst[48] = (t6 - t2 + 64 + 1) >> 7;
146
        dst[56] = (t5 - t1 + 64 + 1) >> 7;
147

    
148
        src++;
149
        dst++;
150
    }
151
}
152

    
153
/** Do inverse transform on 8x4 part of block
154
*/
155
static void vc1_inv_trans_8x4_c(DCTELEM block[64], int n)
156
{
157
    int i;
158
    register int t1,t2,t3,t4,t5,t6,t7,t8;
159
    DCTELEM *src, *dst;
160
    int off;
161

    
162
    off = n * 32;
163
    src = block + off;
164
    dst = block + off;
165
    for(i = 0; i < 4; i++){
166
        t1 = 12 * (src[0] + src[4]);
167
        t2 = 12 * (src[0] - src[4]);
168
        t3 = 16 * src[2] +  6 * src[6];
169
        t4 =  6 * src[2] - 16 * src[6];
170

    
171
        t5 = t1 + t3;
172
        t6 = t2 + t4;
173
        t7 = t2 - t4;
174
        t8 = t1 - t3;
175

    
176
        t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
177
        t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
178
        t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
179
        t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
180

    
181
        dst[0] = (t5 + t1 + 4) >> 3;
182
        dst[1] = (t6 + t2 + 4) >> 3;
183
        dst[2] = (t7 + t3 + 4) >> 3;
184
        dst[3] = (t8 + t4 + 4) >> 3;
185
        dst[4] = (t8 - t4 + 4) >> 3;
186
        dst[5] = (t7 - t3 + 4) >> 3;
187
        dst[6] = (t6 - t2 + 4) >> 3;
188
        dst[7] = (t5 - t1 + 4) >> 3;
189

    
190
        src += 8;
191
        dst += 8;
192
    }
193

    
194
    src = block + off;
195
    dst = block + off;
196
    for(i = 0; i < 8; i++){
197
        t1 = 17 * (src[ 0] + src[16]);
198
        t2 = 17 * (src[ 0] - src[16]);
199
        t3 = 22 * src[ 8];
200
        t4 = 22 * src[24];
201
        t5 = 10 * src[ 8];
202
        t6 = 10 * src[24];
203

    
204
        dst[ 0] = (t1 + t3 + t6 + 64) >> 7;
205
        dst[ 8] = (t2 - t4 + t5 + 64) >> 7;
206
        dst[16] = (t2 + t4 - t5 + 64) >> 7;
207
        dst[24] = (t1 - t3 - t6 + 64) >> 7;
208

    
209
        src ++;
210
        dst ++;
211
    }
212
}
213

    
214
/** Do inverse transform on 4x8 parts of block
215
*/
216
static void vc1_inv_trans_4x8_c(DCTELEM block[64], int n)
217
{
218
    int i;
219
    register int t1,t2,t3,t4,t5,t6,t7,t8;
220
    DCTELEM *src, *dst;
221
    int off;
222

    
223
    off = n * 4;
224
    src = block + off;
225
    dst = block + off;
226
    for(i = 0; i < 8; i++){
227
        t1 = 17 * (src[0] + src[2]);
228
        t2 = 17 * (src[0] - src[2]);
229
        t3 = 22 * src[1];
230
        t4 = 22 * src[3];
231
        t5 = 10 * src[1];
232
        t6 = 10 * src[3];
233

    
234
        dst[0] = (t1 + t3 + t6 + 4) >> 3;
235
        dst[1] = (t2 - t4 + t5 + 4) >> 3;
236
        dst[2] = (t2 + t4 - t5 + 4) >> 3;
237
        dst[3] = (t1 - t3 - t6 + 4) >> 3;
238

    
239
        src += 8;
240
        dst += 8;
241
    }
242

    
243
    src = block + off;
244
    dst = block + off;
245
    for(i = 0; i < 4; i++){
246
        t1 = 12 * (src[ 0] + src[32]);
247
        t2 = 12 * (src[ 0] - src[32]);
248
        t3 = 16 * src[16] +  6 * src[48];
249
        t4 =  6 * src[16] - 16 * src[48];
250

    
251
        t5 = t1 + t3;
252
        t6 = t2 + t4;
253
        t7 = t2 - t4;
254
        t8 = t1 - t3;
255

    
256
        t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
257
        t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
258
        t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
259
        t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
260

    
261
        dst[ 0] = (t5 + t1 + 64) >> 7;
262
        dst[ 8] = (t6 + t2 + 64) >> 7;
263
        dst[16] = (t7 + t3 + 64) >> 7;
264
        dst[24] = (t8 + t4 + 64) >> 7;
265
        dst[32] = (t8 - t4 + 64 + 1) >> 7;
266
        dst[40] = (t7 - t3 + 64 + 1) >> 7;
267
        dst[48] = (t6 - t2 + 64 + 1) >> 7;
268
        dst[56] = (t5 - t1 + 64 + 1) >> 7;
269

    
270
        src++;
271
        dst++;
272
    }
273
}
274

    
275
/** Do inverse transform on 4x4 part of block
276
*/
277
static void vc1_inv_trans_4x4_c(DCTELEM block[64], int n)
278
{
279
    int i;
280
    register int t1,t2,t3,t4,t5,t6;
281
    DCTELEM *src, *dst;
282
    int off;
283

    
284
    off = (n&1) * 4 + (n&2) * 16;
285
    src = block + off;
286
    dst = block + off;
287
    for(i = 0; i < 4; i++){
288
        t1 = 17 * (src[0] + src[2]);
289
        t2 = 17 * (src[0] - src[2]);
290
        t3 = 22 * src[1];
291
        t4 = 22 * src[3];
292
        t5 = 10 * src[1];
293
        t6 = 10 * src[3];
294

    
295
        dst[0] = (t1 + t3 + t6 + 4) >> 3;
296
        dst[1] = (t2 - t4 + t5 + 4) >> 3;
297
        dst[2] = (t2 + t4 - t5 + 4) >> 3;
298
        dst[3] = (t1 - t3 - t6 + 4) >> 3;
299

    
300
        src += 8;
301
        dst += 8;
302
    }
303

    
304
    src = block + off;
305
    dst = block + off;
306
    for(i = 0; i < 4; i++){
307
        t1 = 17 * (src[ 0] + src[16]);
308
        t2 = 17 * (src[ 0] - src[16]);
309
        t3 = 22 * src[ 8];
310
        t4 = 22 * src[24];
311
        t5 = 10 * src[ 8];
312
        t6 = 10 * src[24];
313

    
314
        dst[ 0] = (t1 + t3 + t6 + 64) >> 7;
315
        dst[ 8] = (t2 - t4 + t5 + 64) >> 7;
316
        dst[16] = (t2 + t4 - t5 + 64) >> 7;
317
        dst[24] = (t1 - t3 - t6 + 64) >> 7;
318

    
319
        src ++;
320
        dst ++;
321
    }
322
}
323

    
324
/* motion compensation functions */
325
/** Filter in case of 2 filters */
326
#define VC1_MSPEL_FILTER_16B(DIR, TYPE)                                 \
327
static av_always_inline int vc1_mspel_ ## DIR ## _filter_16bits(const TYPE *src, int stride, int mode) \
328
{                                                                       \
329
    switch(mode){                                                       \
330
    case 0: /* no shift - should not occur */                           \
331
        return 0;                                                       \
332
    case 1: /* 1/4 shift */                                             \
333
        return -4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2]; \
334
    case 2: /* 1/2 shift */                                             \
335
        return -src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2]; \
336
    case 3: /* 3/4 shift */                                             \
337
        return -3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2]; \
338
    }                                                                   \
339
    return 0; /* should not occur */                                    \
340
}
341

    
342
VC1_MSPEL_FILTER_16B(ver, uint8_t);
343
VC1_MSPEL_FILTER_16B(hor, int16_t);
344

    
345

    
346
/** Filter used to interpolate fractional pel values
347
 */
348
static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r)
349
{
350
    switch(mode){
351
    case 0: //no shift
352
        return src[0];
353
    case 1: // 1/4 shift
354
        return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6;
355
    case 2: // 1/2 shift
356
        return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4;
357
    case 3: // 3/4 shift
358
        return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6;
359
    }
360
    return 0; //should not occur
361
}
362

    
363
/** Function used to do motion compensation with bicubic interpolation
364
 */
365
static void vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)
366
{
367
    int     i, j;
368

    
369
    if (vmode) { /* Horizontal filter to apply */
370
        int r;
371

    
372
        if (hmode) { /* Vertical filter to apply, output to tmp */
373
            static const int shift_value[] = { 0, 5, 1, 5 };
374
            int              shift = (shift_value[hmode]+shift_value[vmode])>>1;
375
            int16_t          tmp[11*8], *tptr = tmp;
376

    
377
            r = (1<<(shift-1)) + rnd-1;
378

    
379
            src -= 1;
380
            for(j = 0; j < 8; j++) {
381
                for(i = 0; i < 11; i++)
382
                    tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode)+r)>>shift;
383
                src += stride;
384
                tptr += 11;
385
            }
386

    
387
            r = 64-rnd;
388
            tptr = tmp+1;
389
            for(j = 0; j < 8; j++) {
390
                for(i = 0; i < 8; i++)
391
                    dst[i] = av_clip_uint8((vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode)+r)>>7);
392
                dst += stride;
393
                tptr += 11;
394
            }
395

    
396
            return;
397
        }
398
        else { /* No horizontal filter, output 8 lines to dst */
399
            r = 1-rnd;
400

    
401
            for(j = 0; j < 8; j++) {
402
                for(i = 0; i < 8; i++)
403
                    dst[i] = av_clip_uint8(vc1_mspel_filter(src + i, stride, vmode, r));
404
                src += stride;
405
                dst += stride;
406
            }
407
            return;
408
        }
409
    }
410

    
411
    /* Horizontal mode with no vertical mode */
412
    for(j = 0; j < 8; j++) {
413
        for(i = 0; i < 8; i++)
414
            dst[i] = av_clip_uint8(vc1_mspel_filter(src + i, 1, hmode, rnd));
415
        dst += stride;
416
        src += stride;
417
    }
418
}
419

    
420
/* pixel functions - really are entry points to vc1_mspel_mc */
421

    
422
/* this one is defined in dsputil.c */
423
void ff_put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
424

    
425
#define PUT_VC1_MSPEL(a, b)\
426
static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
427
     vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
428
}
429

    
430
PUT_VC1_MSPEL(1, 0)
431
PUT_VC1_MSPEL(2, 0)
432
PUT_VC1_MSPEL(3, 0)
433

    
434
PUT_VC1_MSPEL(0, 1)
435
PUT_VC1_MSPEL(1, 1)
436
PUT_VC1_MSPEL(2, 1)
437
PUT_VC1_MSPEL(3, 1)
438

    
439
PUT_VC1_MSPEL(0, 2)
440
PUT_VC1_MSPEL(1, 2)
441
PUT_VC1_MSPEL(2, 2)
442
PUT_VC1_MSPEL(3, 2)
443

    
444
PUT_VC1_MSPEL(0, 3)
445
PUT_VC1_MSPEL(1, 3)
446
PUT_VC1_MSPEL(2, 3)
447
PUT_VC1_MSPEL(3, 3)
448

    
449
void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
450
    dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
451
    dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
452
    dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
453
    dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
454
    dsp->vc1_h_overlap = vc1_h_overlap_c;
455
    dsp->vc1_v_overlap = vc1_v_overlap_c;
456

    
457
    dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_vc1_mspel_mc00_c;
458
    dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c;
459
    dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_c;
460
    dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_c;
461
    dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_c;
462
    dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_c;
463
    dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_c;
464
    dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_c;
465
    dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_c;
466
    dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_c;
467
    dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c;
468
    dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c;
469
    dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c;
470
    dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_c;
471
    dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c;
472
    dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c;
473
}