Statistics
| Branch: | Revision:

ffmpeg / libavcodec / vp8dsp.c @ 3ae079a3

History | View | Annotate | Download (16.3 KB)

1
/**
2
 * VP8 compatible video decoder
3
 *
4
 * Copyright (C) 2010 David Conrad
5
 * Copyright (C) 2010 Ronald S. Bultje
6
 *
7
 * This file is part of FFmpeg.
8
 *
9
 * FFmpeg is free software; you can redistribute it and/or
10
 * modify it under the terms of the GNU Lesser General Public
11
 * License as published by the Free Software Foundation; either
12
 * version 2.1 of the License, or (at your option) any later version.
13
 *
14
 * FFmpeg is distributed in the hope that it will be useful,
15
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
17
 * Lesser General Public License for more details.
18
 *
19
 * You should have received a copy of the GNU Lesser General Public
20
 * License along with FFmpeg; if not, write to the Free Software
21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22
 */
23

    
24
#include "dsputil.h"
25
#include "vp8dsp.h"
26

    
27
// TODO: Maybe add dequant
28
static void vp8_luma_dc_wht_c(DCTELEM block[4][4][16], DCTELEM dc[16])
29
{
30
    int i, t0, t1, t2, t3;
31

    
32
    for (i = 0; i < 4; i++) {
33
        t0 = dc[0*4+i] + dc[3*4+i];
34
        t1 = dc[1*4+i] + dc[2*4+i];
35
        t2 = dc[1*4+i] - dc[2*4+i];
36
        t3 = dc[0*4+i] - dc[3*4+i];
37

    
38
        dc[0*4+i] = t0 + t1;
39
        dc[1*4+i] = t3 + t2;
40
        dc[2*4+i] = t0 - t1;
41
        dc[3*4+i] = t3 - t2;
42
    }
43

    
44
    for (i = 0; i < 4; i++) {
45
        t0 = dc[i*4+0] + dc[i*4+3] + 3; // rounding
46
        t1 = dc[i*4+1] + dc[i*4+2];
47
        t2 = dc[i*4+1] - dc[i*4+2];
48
        t3 = dc[i*4+0] - dc[i*4+3] + 3; // rounding
49

    
50
        *block[i][0] = (t0 + t1) >> 3;
51
        *block[i][1] = (t3 + t2) >> 3;
52
        *block[i][2] = (t0 - t1) >> 3;
53
        *block[i][3] = (t3 - t2) >> 3;
54
    }
55
}
56

    
57

    
58
#define MUL_20091(a) ((((a)*20091) >> 16) + (a))
59
#define MUL_35468(a)  (((a)*35468) >> 16)
60

    
61
static void vp8_idct_add_c(uint8_t *dst, DCTELEM block[16], int stride)
62
{
63
    int i, t0, t1, t2, t3;
64
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
65
    DCTELEM tmp[16];
66

    
67
    for (i = 0; i < 4; i++) {
68
        t0 = block[0*4+i] + block[2*4+i];
69
        t1 = block[0*4+i] - block[2*4+i];
70
        t2 = MUL_35468(block[1*4+i]) - MUL_20091(block[3*4+i]);
71
        t3 = MUL_20091(block[1*4+i]) + MUL_35468(block[3*4+i]);
72
        block[0*4+i] = 0;
73
        block[1*4+i] = 0;
74
        block[2*4+i] = 0;
75
        block[3*4+i] = 0;
76

    
77
        tmp[i*4+0] = t0 + t3;
78
        tmp[i*4+1] = t1 + t2;
79
        tmp[i*4+2] = t1 - t2;
80
        tmp[i*4+3] = t0 - t3;
81
    }
82

    
83
    for (i = 0; i < 4; i++) {
84
        t0 = tmp[0*4+i] + tmp[2*4+i];
85
        t1 = tmp[0*4+i] - tmp[2*4+i];
86
        t2 = MUL_35468(tmp[1*4+i]) - MUL_20091(tmp[3*4+i]);
87
        t3 = MUL_20091(tmp[1*4+i]) + MUL_35468(tmp[3*4+i]);
88

    
89
        dst[0] = cm[dst[0] + ((t0 + t3 + 4) >> 3)];
90
        dst[1] = cm[dst[1] + ((t1 + t2 + 4) >> 3)];
91
        dst[2] = cm[dst[2] + ((t1 - t2 + 4) >> 3)];
92
        dst[3] = cm[dst[3] + ((t0 - t3 + 4) >> 3)];
93
        dst += stride;
94
    }
95
}
96

    
97
static void vp8_idct_dc_add_c(uint8_t *dst, DCTELEM block[16], int stride)
98
{
99
    int i, dc = (block[0] + 4) >> 3;
100
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP + dc;
101
    block[0] = 0;
102

    
103
    for (i = 0; i < 4; i++) {
104
        dst[0] = cm[dst[0]];
105
        dst[1] = cm[dst[1]];
106
        dst[2] = cm[dst[2]];
107
        dst[3] = cm[dst[3]];
108
        dst += stride;
109
    }
110
}
111

    
112
static void vp8_idct_dc_add4uv_c(uint8_t *dst, DCTELEM block[4][16], int stride)
113
{
114
    vp8_idct_dc_add_c(dst+stride*0+0, block[0], stride);
115
    vp8_idct_dc_add_c(dst+stride*0+4, block[1], stride);
116
    vp8_idct_dc_add_c(dst+stride*4+0, block[2], stride);
117
    vp8_idct_dc_add_c(dst+stride*4+4, block[3], stride);
118
}
119

    
120
static void vp8_idct_dc_add4y_c(uint8_t *dst, DCTELEM block[4][16], int stride)
121
{
122
    vp8_idct_dc_add_c(dst+ 0, block[0], stride);
123
    vp8_idct_dc_add_c(dst+ 4, block[1], stride);
124
    vp8_idct_dc_add_c(dst+ 8, block[2], stride);
125
    vp8_idct_dc_add_c(dst+12, block[3], stride);
126
}
127

    
128
// because I like only having two parameters to pass functions...
129
#define LOAD_PIXELS\
130
    int av_unused p3 = p[-4*stride];\
131
    int av_unused p2 = p[-3*stride];\
132
    int av_unused p1 = p[-2*stride];\
133
    int av_unused p0 = p[-1*stride];\
134
    int av_unused q0 = p[ 0*stride];\
135
    int av_unused q1 = p[ 1*stride];\
136
    int av_unused q2 = p[ 2*stride];\
137
    int av_unused q3 = p[ 3*stride];
138

    
139
#define clip_int8(n) (cm[n+0x80]-0x80)
140

    
141
static av_always_inline void filter_common(uint8_t *p, int stride, int is4tap)
142
{
143
    LOAD_PIXELS
144
    int a, f1, f2;
145
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
146

    
147
    a = 3*(q0 - p0);
148

    
149
    if (is4tap)
150
        a += clip_int8(p1 - q1);
151

    
152
    a = clip_int8(a);
153

    
154
    // We deviate from the spec here with c(a+3) >> 3
155
    // since that's what libvpx does.
156
    f1 = FFMIN(a+4, 127) >> 3;
157
    f2 = FFMIN(a+3, 127) >> 3;
158

    
159
    // Despite what the spec says, we do need to clamp here to
160
    // be bitexact with libvpx.
161
    p[-1*stride] = cm[p0 + f2];
162
    p[ 0*stride] = cm[q0 - f1];
163

    
164
    // only used for _inner on blocks without high edge variance
165
    if (!is4tap) {
166
        a = (f1+1)>>1;
167
        p[-2*stride] = cm[p1 + a];
168
        p[ 1*stride] = cm[q1 - a];
169
    }
170
}
171

    
172
static av_always_inline int simple_limit(uint8_t *p, int stride, int flim)
173
{
174
    LOAD_PIXELS
175
    return 2*FFABS(p0-q0) + (FFABS(p1-q1) >> 1) <= flim;
176
}
177

    
178
/**
179
 * E - limit at the macroblock edge
180
 * I - limit for interior difference
181
 */
182
static av_always_inline int normal_limit(uint8_t *p, int stride, int E, int I)
183
{
184
    LOAD_PIXELS
185
    return simple_limit(p, stride, E)
186
        && FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I
187
        && FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I;
188
}
189

    
190
// high edge variance
191
static av_always_inline int hev(uint8_t *p, int stride, int thresh)
192
{
193
    LOAD_PIXELS
194
    return FFABS(p1-p0) > thresh || FFABS(q1-q0) > thresh;
195
}
196

    
197
static av_always_inline void filter_mbedge(uint8_t *p, int stride)
198
{
199
    int a0, a1, a2, w;
200
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
201

    
202
    LOAD_PIXELS
203

    
204
    w = clip_int8(p1-q1);
205
    w = clip_int8(w + 3*(q0-p0));
206

    
207
    a0 = (27*w + 63) >> 7;
208
    a1 = (18*w + 63) >> 7;
209
    a2 = ( 9*w + 63) >> 7;
210

    
211
    p[-3*stride] = cm[p2 + a2];
212
    p[-2*stride] = cm[p1 + a1];
213
    p[-1*stride] = cm[p0 + a0];
214
    p[ 0*stride] = cm[q0 - a0];
215
    p[ 1*stride] = cm[q1 - a1];
216
    p[ 2*stride] = cm[q2 - a2];
217
}
218

    
219
#define LOOP_FILTER(dir, size, stridea, strideb, maybe_inline) \
220
static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, int stride,\
221
                                     int flim_E, int flim_I, int hev_thresh)\
222
{\
223
    int i;\
224
\
225
    for (i = 0; i < size; i++)\
226
        if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
227
            if (hev(dst+i*stridea, strideb, hev_thresh))\
228
                filter_common(dst+i*stridea, strideb, 1);\
229
            else\
230
                filter_mbedge(dst+i*stridea, strideb);\
231
        }\
232
}\
233
\
234
static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, int stride,\
235
                                      int flim_E, int flim_I, int hev_thresh)\
236
{\
237
    int i;\
238
\
239
    for (i = 0; i < size; i++)\
240
        if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
241
            int hv = hev(dst+i*stridea, strideb, hev_thresh);\
242
            if (hv) \
243
                filter_common(dst+i*stridea, strideb, 1);\
244
            else \
245
                filter_common(dst+i*stridea, strideb, 0);\
246
        }\
247
}
248

    
249
LOOP_FILTER(v, 16, 1, stride,)
250
LOOP_FILTER(h, 16, stride, 1,)
251

    
252
#define UV_LOOP_FILTER(dir, stridea, strideb) \
253
LOOP_FILTER(dir, 8, stridea, strideb, av_always_inline) \
254
static void vp8_ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, int stride,\
255
                                      int fE, int fI, int hev_thresh)\
256
{\
257
  vp8_ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh);\
258
  vp8_ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh);\
259
}\
260
static void vp8_ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, uint8_t *dstV, int stride,\
261
                                      int fE, int fI, int hev_thresh)\
262
{\
263
  vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);\
264
  vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);\
265
}
266

    
267
UV_LOOP_FILTER(v, 1, stride)
268
UV_LOOP_FILTER(h, stride, 1)
269

    
270
static void vp8_v_loop_filter_simple_c(uint8_t *dst, int stride, int flim)
271
{
272
    int i;
273

    
274
    for (i = 0; i < 16; i++)
275
        if (simple_limit(dst+i, stride, flim))
276
            filter_common(dst+i, stride, 1);
277
}
278

    
279
static void vp8_h_loop_filter_simple_c(uint8_t *dst, int stride, int flim)
280
{
281
    int i;
282

    
283
    for (i = 0; i < 16; i++)
284
        if (simple_limit(dst+i*stride, 1, flim))
285
            filter_common(dst+i*stride, 1, 1);
286
}
287

    
288
static const uint8_t subpel_filters[7][6] = {
289
    { 0,   6, 123,  12,   1,   0 },
290
    { 2,  11, 108,  36,   8,   1 },
291
    { 0,   9,  93,  50,   6,   0 },
292
    { 3,  16,  77,  77,  16,   3 },
293
    { 0,   6,  50,  93,   9,   0 },
294
    { 1,   8,  36, 108,  11,   2 },
295
    { 0,   1,  12, 123,   6,   0 },
296
};
297

    
298
#define PUT_PIXELS(WIDTH) \
299
static void put_vp8_pixels ## WIDTH ##_c(uint8_t *dst, int dststride, uint8_t *src, int srcstride, int h, int x, int y) { \
300
    int i; \
301
    for (i = 0; i < h; i++, dst+= dststride, src+= srcstride) { \
302
        memcpy(dst, src, WIDTH); \
303
    } \
304
}
305

    
306
PUT_PIXELS(16)
307
PUT_PIXELS(8)
308
PUT_PIXELS(4)
309

    
310
#define FILTER_6TAP(src, F, stride) \
311
    cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + F[0]*src[x-2*stride] + \
312
        F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + F[5]*src[x+3*stride] + 64) >> 7]
313

    
314
#define FILTER_4TAP(src, F, stride) \
315
    cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + \
316
        F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + 64) >> 7]
317

    
318
#define VP8_EPEL_H(SIZE, FILTER, FILTERNAME) \
319
static void put_vp8_epel ## SIZE ## _ ## FILTERNAME ## _c(uint8_t *dst, int dststride, uint8_t *src, int srcstride, int h, int mx, int my) \
320
{ \
321
    const uint8_t *filter = subpel_filters[mx-1]; \
322
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
323
    int x, y; \
324
\
325
    for (y = 0; y < h; y++) { \
326
        for (x = 0; x < SIZE; x++) \
327
            dst[x] = FILTER(src, filter, 1); \
328
        dst += dststride; \
329
        src += srcstride; \
330
    } \
331
}
332
#define VP8_EPEL_V(SIZE, FILTER, FILTERNAME) \
333
static void put_vp8_epel ## SIZE ## _ ## FILTERNAME ## _c(uint8_t *dst, int dststride, uint8_t *src, int srcstride, int h, int mx, int my) \
334
{ \
335
    const uint8_t *filter = subpel_filters[my-1]; \
336
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
337
    int x, y; \
338
\
339
    for (y = 0; y < h; y++) { \
340
        for (x = 0; x < SIZE; x++) \
341
            dst[x] = FILTER(src, filter, srcstride); \
342
        dst += dststride; \
343
        src += srcstride; \
344
    } \
345
}
346
#define VP8_EPEL_HV(SIZE, FILTERX, FILTERY, FILTERNAME) \
347
static void put_vp8_epel ## SIZE ## _ ## FILTERNAME ## _c(uint8_t *dst, int dststride, uint8_t *src, int srcstride, int h, int mx, int my) \
348
{ \
349
    const uint8_t *filter = subpel_filters[mx-1]; \
350
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
351
    int x, y; \
352
    uint8_t tmp_array[(2*SIZE+5)*SIZE]; \
353
    uint8_t *tmp = tmp_array; \
354
    src -= 2*srcstride; \
355
\
356
    for (y = 0; y < h+5; y++) { \
357
        for (x = 0; x < SIZE; x++) \
358
            tmp[x] = FILTERX(src, filter, 1); \
359
        tmp += SIZE; \
360
        src += srcstride; \
361
    } \
362
\
363
    tmp = tmp_array + 2*SIZE; \
364
    filter = subpel_filters[my-1]; \
365
\
366
    for (y = 0; y < h; y++) { \
367
        for (x = 0; x < SIZE; x++) \
368
            dst[x] = FILTERY(tmp, filter, SIZE); \
369
        dst += dststride; \
370
        tmp += SIZE; \
371
    } \
372
}
373

    
374
VP8_EPEL_H(16, FILTER_4TAP, h4)
375
VP8_EPEL_H(8,  FILTER_4TAP, h4)
376
VP8_EPEL_H(4,  FILTER_4TAP, h4)
377
VP8_EPEL_H(16, FILTER_6TAP, h6)
378
VP8_EPEL_H(8,  FILTER_6TAP, h6)
379
VP8_EPEL_H(4,  FILTER_6TAP, h6)
380
VP8_EPEL_V(16, FILTER_4TAP, v4)
381
VP8_EPEL_V(8,  FILTER_4TAP, v4)
382
VP8_EPEL_V(4,  FILTER_4TAP, v4)
383
VP8_EPEL_V(16, FILTER_6TAP, v6)
384
VP8_EPEL_V(8,  FILTER_6TAP, v6)
385
VP8_EPEL_V(4,  FILTER_6TAP, v6)
386
VP8_EPEL_HV(16, FILTER_4TAP, FILTER_4TAP, h4v4)
387
VP8_EPEL_HV(8,  FILTER_4TAP, FILTER_4TAP, h4v4)
388
VP8_EPEL_HV(4,  FILTER_4TAP, FILTER_4TAP, h4v4)
389
VP8_EPEL_HV(16, FILTER_4TAP, FILTER_6TAP, h4v6)
390
VP8_EPEL_HV(8,  FILTER_4TAP, FILTER_6TAP, h4v6)
391
VP8_EPEL_HV(4,  FILTER_4TAP, FILTER_6TAP, h4v6)
392
VP8_EPEL_HV(16, FILTER_6TAP, FILTER_4TAP, h6v4)
393
VP8_EPEL_HV(8,  FILTER_6TAP, FILTER_4TAP, h6v4)
394
VP8_EPEL_HV(4,  FILTER_6TAP, FILTER_4TAP, h6v4)
395
VP8_EPEL_HV(16, FILTER_6TAP, FILTER_6TAP, h6v6)
396
VP8_EPEL_HV(8,  FILTER_6TAP, FILTER_6TAP, h6v6)
397
VP8_EPEL_HV(4,  FILTER_6TAP, FILTER_6TAP, h6v6)
398

    
399
#define VP8_BILINEAR(SIZE) \
400
static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, int stride, uint8_t *src, int s2, int h, int mx, int my) \
401
{ \
402
    int a = 8-mx, b = mx; \
403
    int x, y; \
404
\
405
    for (y = 0; y < h; y++) { \
406
        for (x = 0; x < SIZE; x++) \
407
            dst[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
408
        dst += stride; \
409
        src += stride; \
410
    } \
411
} \
412
static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, int stride, uint8_t *src, int s2, int h, int mx, int my) \
413
{ \
414
    int c = 8-my, d = my; \
415
    int x, y; \
416
\
417
    for (y = 0; y < h; y++) { \
418
        for (x = 0; x < SIZE; x++) \
419
            dst[x] = (c*src[x] + d*src[x+stride] + 4) >> 3; \
420
        dst += stride; \
421
        src += stride; \
422
    } \
423
} \
424
\
425
static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst, int stride, uint8_t *src, int s2, int h, int mx, int my) \
426
{ \
427
    int a = 8-mx, b = mx; \
428
    int c = 8-my, d = my; \
429
    int x, y; \
430
    uint8_t tmp_array[(2*SIZE+1)*SIZE]; \
431
    uint8_t *tmp = tmp_array; \
432
\
433
    for (y = 0; y < h+1; y++) { \
434
        for (x = 0; x < SIZE; x++) \
435
            tmp[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
436
        tmp += SIZE; \
437
        src += stride; \
438
    } \
439
\
440
    tmp = tmp_array; \
441
\
442
    for (y = 0; y < h; y++) { \
443
        for (x = 0; x < SIZE; x++) \
444
            dst[x] = (c*tmp[x] + d*tmp[x+SIZE] + 4) >> 3; \
445
        dst += stride; \
446
        tmp += SIZE; \
447
    } \
448
}
449

    
450
VP8_BILINEAR(16)
451
VP8_BILINEAR(8)
452
VP8_BILINEAR(4)
453

    
454
#define VP8_MC_FUNC(IDX, SIZE) \
455
    dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
456
    dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \
457
    dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c; \
458
    dsp->put_vp8_epel_pixels_tab[IDX][1][0] = put_vp8_epel ## SIZE ## _v4_c; \
459
    dsp->put_vp8_epel_pixels_tab[IDX][1][1] = put_vp8_epel ## SIZE ## _h4v4_c; \
460
    dsp->put_vp8_epel_pixels_tab[IDX][1][2] = put_vp8_epel ## SIZE ## _h6v4_c; \
461
    dsp->put_vp8_epel_pixels_tab[IDX][2][0] = put_vp8_epel ## SIZE ## _v6_c; \
462
    dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \
463
    dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c
464

    
465
#define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \
466
    dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
467
    dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \
468
    dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \
469
    dsp->put_vp8_bilinear_pixels_tab[IDX][1][0] = put_vp8_bilinear ## SIZE ## _v_c; \
470
    dsp->put_vp8_bilinear_pixels_tab[IDX][1][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
471
    dsp->put_vp8_bilinear_pixels_tab[IDX][1][2] = put_vp8_bilinear ## SIZE ## _hv_c; \
472
    dsp->put_vp8_bilinear_pixels_tab[IDX][2][0] = put_vp8_bilinear ## SIZE ## _v_c; \
473
    dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
474
    dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c
475

    
476
av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
477
{
478
    dsp->vp8_luma_dc_wht    = vp8_luma_dc_wht_c;
479
    dsp->vp8_idct_add       = vp8_idct_add_c;
480
    dsp->vp8_idct_dc_add    = vp8_idct_dc_add_c;
481
    dsp->vp8_idct_dc_add4y  = vp8_idct_dc_add4y_c;
482
    dsp->vp8_idct_dc_add4uv = vp8_idct_dc_add4uv_c;
483

    
484
    dsp->vp8_v_loop_filter16y = vp8_v_loop_filter16_c;
485
    dsp->vp8_h_loop_filter16y = vp8_h_loop_filter16_c;
486
    dsp->vp8_v_loop_filter8uv = vp8_v_loop_filter8uv_c;
487
    dsp->vp8_h_loop_filter8uv = vp8_h_loop_filter8uv_c;
488

    
489
    dsp->vp8_v_loop_filter16y_inner = vp8_v_loop_filter16_inner_c;
490
    dsp->vp8_h_loop_filter16y_inner = vp8_h_loop_filter16_inner_c;
491
    dsp->vp8_v_loop_filter8uv_inner = vp8_v_loop_filter8uv_inner_c;
492
    dsp->vp8_h_loop_filter8uv_inner = vp8_h_loop_filter8uv_inner_c;
493

    
494
    dsp->vp8_v_loop_filter_simple = vp8_v_loop_filter_simple_c;
495
    dsp->vp8_h_loop_filter_simple = vp8_h_loop_filter_simple_c;
496

    
497
    VP8_MC_FUNC(0, 16);
498
    VP8_MC_FUNC(1, 8);
499
    VP8_MC_FUNC(2, 4);
500

    
501
    VP8_BILINEAR_MC_FUNC(0, 16);
502
    VP8_BILINEAR_MC_FUNC(1, 8);
503
    VP8_BILINEAR_MC_FUNC(2, 4);
504

    
505
    if (HAVE_MMX)
506
        ff_vp8dsp_init_x86(dsp);
507
    if (HAVE_ALTIVEC)
508
        ff_vp8dsp_init_altivec(dsp);
509
}