FFmpeg
Loading...
Searching...
No Matches
vp8dsp.c
Go to the documentation of this file.
1/*
2 * Copyright (C) 2010 David Conrad
3 * Copyright (C) 2010 Ronald S. Bultje
4 * Copyright (C) 2014 Peter Ross
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
25 * VP8 compatible video decoder
26 */
27
28#include "config_components.h"
29
30#include "libavutil/common.h"
31#include "libavutil/intreadwrite.h"
32
33#include "mathops.h"
34#include "vp8dsp.h"
35
36#define MK_IDCT_DC_ADD4_C(name) \
37static void name ## _idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], \
38 ptrdiff_t stride) \
39{ \
40 name ## _idct_dc_add_c(dst + stride * 0 + 0, block[0], stride); \
41 name ## _idct_dc_add_c(dst + stride * 0 + 4, block[1], stride); \
42 name ## _idct_dc_add_c(dst + stride * 4 + 0, block[2], stride); \
43 name ## _idct_dc_add_c(dst + stride * 4 + 4, block[3], stride); \
44} \
45 \
46static void name ## _idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], \
47 ptrdiff_t stride) \
48{ \
49 name ## _idct_dc_add_c(dst + 0, block[0], stride); \
50 name ## _idct_dc_add_c(dst + 4, block[1], stride); \
51 name ## _idct_dc_add_c(dst + 8, block[2], stride); \
52 name ## _idct_dc_add_c(dst + 12, block[3], stride); \
53}
54
55#if CONFIG_VP7_DECODER
56static void vp7_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
57{
58 int i;
59 unsigned a1, b1, c1, d1;
60 int16_t tmp[16];
61
62 for (i = 0; i < 4; i++) {
63 a1 = (dc[i * 4 + 0] + dc[i * 4 + 2]) * 23170;
64 b1 = (dc[i * 4 + 0] - dc[i * 4 + 2]) * 23170;
65 c1 = dc[i * 4 + 1] * 12540 - dc[i * 4 + 3] * 30274;
66 d1 = dc[i * 4 + 1] * 30274 + dc[i * 4 + 3] * 12540;
67 tmp[i * 4 + 0] = (int)(a1 + d1) >> 14;
68 tmp[i * 4 + 3] = (int)(a1 - d1) >> 14;
69 tmp[i * 4 + 1] = (int)(b1 + c1) >> 14;
70 tmp[i * 4 + 2] = (int)(b1 - c1) >> 14;
71 }
72
73 for (i = 0; i < 4; i++) {
74 a1 = (tmp[i + 0] + tmp[i + 8]) * 23170;
75 b1 = (tmp[i + 0] - tmp[i + 8]) * 23170;
76 c1 = tmp[i + 4] * 12540 - tmp[i + 12] * 30274;
77 d1 = tmp[i + 4] * 30274 + tmp[i + 12] * 12540;
78 AV_ZERO64(dc + i * 4);
79 block[0][i][0] = (int)(a1 + d1 + 0x20000) >> 18;
80 block[3][i][0] = (int)(a1 - d1 + 0x20000) >> 18;
81 block[1][i][0] = (int)(b1 + c1 + 0x20000) >> 18;
82 block[2][i][0] = (int)(b1 - c1 + 0x20000) >> 18;
83 }
84}
85
86static void vp7_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])
87{
88 int i, val = (23170 * (23170 * dc[0] >> 14) + 0x20000) >> 18;
89 dc[0] = 0;
90
91 for (i = 0; i < 4; i++) {
92 block[i][0][0] = val;
93 block[i][1][0] = val;
94 block[i][2][0] = val;
95 block[i][3][0] = val;
96 }
97}
98
99static void vp7_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
100{
101 int i;
102 unsigned a1, b1, c1, d1;
103 int16_t tmp[16];
104
105 for (i = 0; i < 4; i++) {
106 a1 = (block[i * 4 + 0] + block[i * 4 + 2]) * 23170;
107 b1 = (block[i * 4 + 0] - block[i * 4 + 2]) * 23170;
108 c1 = block[i * 4 + 1] * 12540 - block[i * 4 + 3] * 30274;
109 d1 = block[i * 4 + 1] * 30274 + block[i * 4 + 3] * 12540;
110 AV_ZERO64(block + i * 4);
111 tmp[i * 4 + 0] = (int)(a1 + d1) >> 14;
112 tmp[i * 4 + 3] = (int)(a1 - d1) >> 14;
113 tmp[i * 4 + 1] = (int)(b1 + c1) >> 14;
114 tmp[i * 4 + 2] = (int)(b1 - c1) >> 14;
115 }
116
117 for (i = 0; i < 4; i++) {
118 a1 = (tmp[i + 0] + tmp[i + 8]) * 23170;
119 b1 = (tmp[i + 0] - tmp[i + 8]) * 23170;
120 c1 = tmp[i + 4] * 12540 - tmp[i + 12] * 30274;
121 d1 = tmp[i + 4] * 30274 + tmp[i + 12] * 12540;
122 dst[0 * stride + i] = av_clip_uint8(dst[0 * stride + i] +
123 ((int)(a1 + d1 + 0x20000) >> 18));
124 dst[3 * stride + i] = av_clip_uint8(dst[3 * stride + i] +
125 ((int)(a1 - d1 + 0x20000) >> 18));
126 dst[1 * stride + i] = av_clip_uint8(dst[1 * stride + i] +
127 ((int)(b1 + c1 + 0x20000) >> 18));
128 dst[2 * stride + i] = av_clip_uint8(dst[2 * stride + i] +
129 ((int)(b1 - c1 + 0x20000) >> 18));
130 }
131}
132
133static void vp7_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
134{
135 int i, dc = (23170 * (23170 * block[0] >> 14) + 0x20000) >> 18;
136 block[0] = 0;
137
138 for (i = 0; i < 4; i++) {
139 dst[0] = av_clip_uint8(dst[0] + dc);
140 dst[1] = av_clip_uint8(dst[1] + dc);
141 dst[2] = av_clip_uint8(dst[2] + dc);
142 dst[3] = av_clip_uint8(dst[3] + dc);
143 dst += stride;
144 }
145}
146
147MK_IDCT_DC_ADD4_C(vp7)
148#endif /* CONFIG_VP7_DECODER */
149
150// TODO: Maybe add dequant
151#if CONFIG_VP8_DECODER
152static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
153{
154 int i, t0, t1, t2, t3;
155
156 for (i = 0; i < 4; i++) {
157 t0 = dc[0 * 4 + i] + dc[3 * 4 + i];
158 t1 = dc[1 * 4 + i] + dc[2 * 4 + i];
159 t2 = dc[1 * 4 + i] - dc[2 * 4 + i];
160 t3 = dc[0 * 4 + i] - dc[3 * 4 + i];
161
162 dc[0 * 4 + i] = t0 + t1;
163 dc[1 * 4 + i] = t3 + t2;
164 dc[2 * 4 + i] = t0 - t1;
165 dc[3 * 4 + i] = t3 - t2;
166 }
167
168 for (i = 0; i < 4; i++) {
169 t0 = dc[i * 4 + 0] + dc[i * 4 + 3] + 3; // rounding
170 t1 = dc[i * 4 + 1] + dc[i * 4 + 2];
171 t2 = dc[i * 4 + 1] - dc[i * 4 + 2];
172 t3 = dc[i * 4 + 0] - dc[i * 4 + 3] + 3; // rounding
173 AV_ZERO64(dc + i * 4);
174
175 block[i][0][0] = (t0 + t1) >> 3;
176 block[i][1][0] = (t3 + t2) >> 3;
177 block[i][2][0] = (t0 - t1) >> 3;
178 block[i][3][0] = (t3 - t2) >> 3;
179 }
180}
181
182static void vp8_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])
183{
184 int i, val = (dc[0] + 3) >> 3;
185 dc[0] = 0;
186
187 for (i = 0; i < 4; i++) {
188 block[i][0][0] = val;
189 block[i][1][0] = val;
190 block[i][2][0] = val;
191 block[i][3][0] = val;
192 }
193}
194
195#define MUL_20091(a) ((((a) * 20091) >> 16) + (a))
196#define MUL_35468(a) (((a) * 35468) >> 16)
197
198static void vp8_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
199{
200 int i, t0, t1, t2, t3;
201 int16_t tmp[16];
202
203 for (i = 0; i < 4; i++) {
204 t0 = block[0 * 4 + i] + block[2 * 4 + i];
205 t1 = block[0 * 4 + i] - block[2 * 4 + i];
206 t2 = MUL_35468(block[1 * 4 + i]) - MUL_20091(block[3 * 4 + i]);
207 t3 = MUL_20091(block[1 * 4 + i]) + MUL_35468(block[3 * 4 + i]);
208 block[0 * 4 + i] = 0;
209 block[1 * 4 + i] = 0;
210 block[2 * 4 + i] = 0;
211 block[3 * 4 + i] = 0;
212
213 tmp[i * 4 + 0] = t0 + t3;
214 tmp[i * 4 + 1] = t1 + t2;
215 tmp[i * 4 + 2] = t1 - t2;
216 tmp[i * 4 + 3] = t0 - t3;
217 }
218
219 for (i = 0; i < 4; i++) {
220 t0 = tmp[0 * 4 + i] + tmp[2 * 4 + i];
221 t1 = tmp[0 * 4 + i] - tmp[2 * 4 + i];
222 t2 = MUL_35468(tmp[1 * 4 + i]) - MUL_20091(tmp[3 * 4 + i]);
223 t3 = MUL_20091(tmp[1 * 4 + i]) + MUL_35468(tmp[3 * 4 + i]);
224
225 dst[0] = av_clip_uint8(dst[0] + ((t0 + t3 + 4) >> 3));
226 dst[1] = av_clip_uint8(dst[1] + ((t1 + t2 + 4) >> 3));
227 dst[2] = av_clip_uint8(dst[2] + ((t1 - t2 + 4) >> 3));
228 dst[3] = av_clip_uint8(dst[3] + ((t0 - t3 + 4) >> 3));
229 dst += stride;
230 }
231}
232
233static void vp8_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
234{
235 int i, dc = (block[0] + 4) >> 3;
236 block[0] = 0;
237
238 for (i = 0; i < 4; i++) {
239 dst[0] = av_clip_uint8(dst[0] + dc);
240 dst[1] = av_clip_uint8(dst[1] + dc);
241 dst[2] = av_clip_uint8(dst[2] + dc);
242 dst[3] = av_clip_uint8(dst[3] + dc);
243 dst += stride;
244 }
245}
246
247MK_IDCT_DC_ADD4_C(vp8)
248#endif /* CONFIG_VP8_DECODER */
249
250// because I like only having two parameters to pass functions...
251#define LOAD_PIXELS \
252 av_unused int p3 = p[-4 * stride]; \
253 av_unused int p2 = p[-3 * stride]; \
254 av_unused int p1 = p[-2 * stride]; \
255 av_unused int p0 = p[-1 * stride]; \
256 av_unused int q0 = p[ 0 * stride]; \
257 av_unused int q1 = p[ 1 * stride]; \
258 av_unused int q2 = p[ 2 * stride]; \
259 av_unused int q3 = p[ 3 * stride];
260
261#define clip_int8(n) (cm[(n) + 0x80] - 0x80)
262
263static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride,
264 int is4tap, int is_vp7)
265{
266 LOAD_PIXELS
267 int a, f1, f2;
268 const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
269
270 a = 3 * (q0 - p0);
271
272 if (is4tap)
273 a += clip_int8(p1 - q1);
274
275 a = clip_int8(a);
276
277 // We deviate from the spec here with c(a+3) >> 3
278 // since that's what libvpx does.
279 f1 = FFMIN(a + 4, 127) >> 3;
280
281 if (is_vp7)
282 f2 = f1 - ((a & 7) == 4);
283 else
284 f2 = FFMIN(a + 3, 127) >> 3;
285
286 // Despite what the spec says, we do need to clamp here to
287 // be bitexact with libvpx.
288 p[-1 * stride] = cm[p0 + f2];
289 p[ 0 * stride] = cm[q0 - f1];
290
291 // only used for _inner on blocks without high edge variance
292 if (!is4tap) {
293 a = (f1 + 1) >> 1;
294 p[-2 * stride] = cm[p1 + a];
295 p[ 1 * stride] = cm[q1 - a];
296 }
297}
298
299static av_always_inline void vp7_filter_common(uint8_t *p, ptrdiff_t stride,
300 int is4tap)
301{
302 filter_common(p, stride, is4tap, IS_VP7);
303}
304
305static av_always_inline void vp8_filter_common(uint8_t *p, ptrdiff_t stride,
306 int is4tap)
307{
308 filter_common(p, stride, is4tap, IS_VP8);
309}
310
311static av_always_inline int vp7_simple_limit(uint8_t *p, ptrdiff_t stride,
312 int flim)
313{
314 LOAD_PIXELS
315 return FFABS(p0 - q0) <= flim;
316}
317
318static av_always_inline int vp8_simple_limit(uint8_t *p, ptrdiff_t stride,
319 int flim)
320{
321 LOAD_PIXELS
322 return 2 * FFABS(p0 - q0) + (FFABS(p1 - q1) >> 1) <= flim;
323}
324
325/**
326 * E - limit at the macroblock edge
327 * I - limit for interior difference
328 */
329#define NORMAL_LIMIT(vpn) \
330static av_always_inline int vp ## vpn ## _normal_limit(uint8_t *p, \
331 ptrdiff_t stride, \
332 int E, int I) \
333{ \
334 LOAD_PIXELS \
335 return vp ## vpn ## _simple_limit(p, stride, E) && \
336 FFABS(p3 - p2) <= I && FFABS(p2 - p1) <= I && \
337 FFABS(p1 - p0) <= I && FFABS(q3 - q2) <= I && \
338 FFABS(q2 - q1) <= I && FFABS(q1 - q0) <= I; \
339}
340
341NORMAL_LIMIT(7)
342NORMAL_LIMIT(8)
343
344// high edge variance
345static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)
346{
347 LOAD_PIXELS
348 return FFABS(p1 - p0) > thresh || FFABS(q1 - q0) > thresh;
349}
350
351static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
352{
353 int a0, a1, a2, w;
354 const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
355
356 LOAD_PIXELS
357
358 w = clip_int8(p1 - q1);
359 w = clip_int8(w + 3 * (q0 - p0));
360
361 a0 = (27 * w + 63) >> 7;
362 a1 = (18 * w + 63) >> 7;
363 a2 = (9 * w + 63) >> 7;
364
365 p[-3 * stride] = cm[p2 + a2];
366 p[-2 * stride] = cm[p1 + a1];
367 p[-1 * stride] = cm[p0 + a0];
368 p[ 0 * stride] = cm[q0 - a0];
369 p[ 1 * stride] = cm[q1 - a1];
370 p[ 2 * stride] = cm[q2 - a2];
371}
372
373#define LOOP_FILTER(vpn, dir, size, stridea, strideb, maybe_inline) \
374static maybe_inline \
375void vpn ## _ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, \
376 ptrdiff_t stride, \
377 int flim_E, int flim_I, \
378 int hev_thresh) \
379{ \
380 int i; \
381 for (i = 0; i < size; i++) \
382 if (vpn ## _normal_limit(dst + i * stridea, strideb, \
383 flim_E, flim_I)) { \
384 if (hev(dst + i * stridea, strideb, hev_thresh)) \
385 vpn ## _filter_common(dst + i * stridea, strideb, 1); \
386 else \
387 filter_mbedge(dst + i * stridea, strideb); \
388 } \
389} \
390 \
391static maybe_inline \
392void vpn ## _ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, \
393 ptrdiff_t stride, \
394 int flim_E, \
395 int flim_I, \
396 int hev_thresh) \
397{ \
398 int i; \
399 for (i = 0; i < size; i++) \
400 if (vpn ## _normal_limit(dst + i * stridea, strideb, \
401 flim_E, flim_I)) { \
402 int hv = hev(dst + i * stridea, strideb, hev_thresh); \
403 if (hv) \
404 vpn ## _filter_common(dst + i * stridea, strideb, 1); \
405 else \
406 vpn ## _filter_common(dst + i * stridea, strideb, 0); \
407 } \
408}
409
410#define UV_LOOP_FILTER(vpn, dir, stridea, strideb) \
411LOOP_FILTER(vpn, dir, 8, stridea, strideb, av_always_inline) \
412static void vpn ## _ ## dir ## _loop_filter8uv_c(uint8_t *dstU, \
413 uint8_t *dstV, \
414 ptrdiff_t stride, int fE, \
415 int fI, int hev_thresh) \
416{ \
417 vpn ## _ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh); \
418 vpn ## _ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh); \
419} \
420 \
421static void vpn ## _ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, \
422 uint8_t *dstV, \
423 ptrdiff_t stride, \
424 int fE, int fI, \
425 int hev_thresh) \
426{ \
427 vpn ## _ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, \
428 hev_thresh); \
429 vpn ## _ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, \
430 hev_thresh); \
431}
432
433#define LOOP_FILTER_SIMPLE(vpn) \
434static void vpn ## _v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, \
435 int flim) \
436{ \
437 int i; \
438 for (i = 0; i < 16; i++) \
439 if (vpn ## _simple_limit(dst + i, stride, flim)) \
440 vpn ## _filter_common(dst + i, stride, 1); \
441} \
442 \
443static void vpn ## _h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, \
444 int flim) \
445{ \
446 int i; \
447 for (i = 0; i < 16; i++) \
448 if (vpn ## _simple_limit(dst + i * stride, 1, flim)) \
449 vpn ## _filter_common(dst + i * stride, 1, 1); \
450}
451
452#define LOOP_FILTERS(vpn) \
453 LOOP_FILTER(vpn, v, 16, 1, stride, ) \
454 LOOP_FILTER(vpn, h, 16, stride, 1, ) \
455 UV_LOOP_FILTER(vpn, v, 1, stride) \
456 UV_LOOP_FILTER(vpn, h, stride, 1) \
457 LOOP_FILTER_SIMPLE(vpn) \
458
459static const uint8_t subpel_filters[7][6] = {
460 { 0, 6, 123, 12, 1, 0 },
461 { 2, 11, 108, 36, 8, 1 },
462 { 0, 9, 93, 50, 6, 0 },
463 { 3, 16, 77, 77, 16, 3 },
464 { 0, 6, 50, 93, 9, 0 },
465 { 1, 8, 36, 108, 11, 2 },
466 { 0, 1, 12, 123, 6, 0 },
467};
468
469#define PUT_PIXELS(WIDTH) \
470static void put_vp8_pixels ## WIDTH ## _c(uint8_t *dst, ptrdiff_t dststride, \
471 const uint8_t *src, ptrdiff_t srcstride, \
472 int h, int x, int y) \
473{ \
474 int i; \
475 for (i = 0; i < h; i++, dst += dststride, src += srcstride) \
476 memcpy(dst, src, WIDTH); \
477}
478
479PUT_PIXELS(16)
480PUT_PIXELS(8)
481PUT_PIXELS(4)
482
483#define FILTER_6TAP(src, F, stride) \
484 cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] + \
485 F[0] * src[x - 2 * stride] + F[3] * src[x + 1 * stride] - \
486 F[4] * src[x + 2 * stride] + F[5] * src[x + 3 * stride] + 64) >> 7]
487
488#define FILTER_4TAP(src, F, stride) \
489 cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] + \
490 F[3] * src[x + 1 * stride] - F[4] * src[x + 2 * stride] + 64) >> 7]
491
492#define VP8_EPEL_H(SIZE, TAPS) \
493static void put_vp8_epel ## SIZE ## _h ## TAPS ## _c(uint8_t *dst, \
494 ptrdiff_t dststride, \
495 const uint8_t *src, \
496 ptrdiff_t srcstride, \
497 int h, int mx, int my) \
498{ \
499 const uint8_t *filter = subpel_filters[mx - 1]; \
500 const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
501 int x, y; \
502 for (y = 0; y < h; y++) { \
503 for (x = 0; x < SIZE; x++) \
504 dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1); \
505 dst += dststride; \
506 src += srcstride; \
507 } \
508}
509
510#define VP8_EPEL_V(SIZE, TAPS) \
511static void put_vp8_epel ## SIZE ## _v ## TAPS ## _c(uint8_t *dst, \
512 ptrdiff_t dststride, \
513 const uint8_t *src, \
514 ptrdiff_t srcstride, \
515 int h, int mx, int my) \
516{ \
517 const uint8_t *filter = subpel_filters[my - 1]; \
518 const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
519 int x, y; \
520 for (y = 0; y < h; y++) { \
521 for (x = 0; x < SIZE; x++) \
522 dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride); \
523 dst += dststride; \
524 src += srcstride; \
525 } \
526}
527
528#define VP8_EPEL_HV(SIZE, HTAPS, VTAPS) \
529static void \
530put_vp8_epel ## SIZE ## _h ## HTAPS ## v ## VTAPS ## _c(uint8_t *dst, \
531 ptrdiff_t dststride, \
532 const uint8_t *src, \
533 ptrdiff_t srcstride, \
534 int h, int mx, \
535 int my) \
536{ \
537 const uint8_t *filter = subpel_filters[mx - 1]; \
538 const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
539 int x, y; \
540 uint8_t tmp_array[(2 * SIZE + VTAPS - 1) * SIZE]; \
541 uint8_t *tmp = tmp_array; \
542 src -= (2 - (VTAPS == 4)) * srcstride; \
543 \
544 for (y = 0; y < h + VTAPS - 1; y++) { \
545 for (x = 0; x < SIZE; x++) \
546 tmp[x] = FILTER_ ## HTAPS ## TAP(src, filter, 1); \
547 tmp += SIZE; \
548 src += srcstride; \
549 } \
550 tmp = tmp_array + (2 - (VTAPS == 4)) * SIZE; \
551 filter = subpel_filters[my - 1]; \
552 \
553 for (y = 0; y < h; y++) { \
554 for (x = 0; x < SIZE; x++) \
555 dst[x] = FILTER_ ## VTAPS ## TAP(tmp, filter, SIZE); \
556 dst += dststride; \
557 tmp += SIZE; \
558 } \
559}
560
561VP8_EPEL_H(8, 4)
562VP8_EPEL_H(4, 4)
563VP8_EPEL_H(16, 6)
564VP8_EPEL_H(8, 6)
565VP8_EPEL_H(4, 6)
566VP8_EPEL_V(8, 4)
567VP8_EPEL_V(4, 4)
568VP8_EPEL_V(16, 6)
569VP8_EPEL_V(8, 6)
570VP8_EPEL_V(4, 6)
571
572VP8_EPEL_HV(8, 4, 4)
573VP8_EPEL_HV(4, 4, 4)
574VP8_EPEL_HV(8, 4, 6)
575VP8_EPEL_HV(4, 4, 6)
576VP8_EPEL_HV(8, 6, 4)
577VP8_EPEL_HV(4, 6, 4)
578VP8_EPEL_HV(16, 6, 6)
579VP8_EPEL_HV(8, 6, 6)
580VP8_EPEL_HV(4, 6, 6)
581
582#define VP8_BILINEAR(SIZE) \
583static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, ptrdiff_t dstride, \
584 const uint8_t *src, ptrdiff_t sstride, \
585 int h, int mx, int my) \
586{ \
587 int a = 8 - mx, b = mx; \
588 int x, y; \
589 for (y = 0; y < h; y++) { \
590 for (x = 0; x < SIZE; x++) \
591 dst[x] = (a * src[x] + b * src[x + 1] + 4) >> 3; \
592 dst += dstride; \
593 src += sstride; \
594 } \
595} \
596 \
597static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, ptrdiff_t dstride, \
598 const uint8_t *src, ptrdiff_t sstride, \
599 int h, int mx, int my) \
600{ \
601 int c = 8 - my, d = my; \
602 int x, y; \
603 for (y = 0; y < h; y++) { \
604 for (x = 0; x < SIZE; x++) \
605 dst[x] = (c * src[x] + d * src[x + sstride] + 4) >> 3; \
606 dst += dstride; \
607 src += sstride; \
608 } \
609} \
610 \
611static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst, \
612 ptrdiff_t dstride, \
613 const uint8_t *src, \
614 ptrdiff_t sstride, \
615 int h, int mx, int my) \
616{ \
617 int a = 8 - mx, b = mx; \
618 int c = 8 - my, d = my; \
619 int x, y; \
620 uint8_t tmp_array[(2 * SIZE + 1) * SIZE]; \
621 uint8_t *tmp = tmp_array; \
622 for (y = 0; y < h + 1; y++) { \
623 for (x = 0; x < SIZE; x++) \
624 tmp[x] = (a * src[x] + b * src[x + 1] + 4) >> 3; \
625 tmp += SIZE; \
626 src += sstride; \
627 } \
628 tmp = tmp_array; \
629 for (y = 0; y < h; y++) { \
630 for (x = 0; x < SIZE; x++) \
631 dst[x] = (c * tmp[x] + d * tmp[x + SIZE] + 4) >> 3; \
632 dst += dstride; \
633 tmp += SIZE; \
634 } \
635}
636
637VP8_BILINEAR(16)
638VP8_BILINEAR(8)
639VP8_BILINEAR(4)
640
641#define VP78_MC_FUNC(IDX, SIZE) \
642 dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
643 dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \
644 dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c; \
645 dsp->put_vp8_epel_pixels_tab[IDX][1][0] = put_vp8_epel ## SIZE ## _v4_c; \
646 dsp->put_vp8_epel_pixels_tab[IDX][1][1] = put_vp8_epel ## SIZE ## _h4v4_c; \
647 dsp->put_vp8_epel_pixels_tab[IDX][1][2] = put_vp8_epel ## SIZE ## _h6v4_c; \
648 dsp->put_vp8_epel_pixels_tab[IDX][2][0] = put_vp8_epel ## SIZE ## _v6_c; \
649 dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \
650 dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c
651
652#define VP78_BILINEAR_MC_FUNC(IDX, SIZE) \
653 dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
654 dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \
655 dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \
656 dsp->put_vp8_bilinear_pixels_tab[IDX][1][0] = put_vp8_bilinear ## SIZE ## _v_c; \
657 dsp->put_vp8_bilinear_pixels_tab[IDX][1][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
658 dsp->put_vp8_bilinear_pixels_tab[IDX][1][2] = put_vp8_bilinear ## SIZE ## _hv_c; \
659 dsp->put_vp8_bilinear_pixels_tab[IDX][2][0] = put_vp8_bilinear ## SIZE ## _v_c; \
660 dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
661 dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c
662
663av_cold void ff_vp78dsp_init(VP8DSPContext *dsp)
664{
665 dsp->put_vp8_epel_pixels_tab[0][0][0] = put_vp8_pixels16_c;
666 dsp->put_vp8_epel_pixels_tab[0][0][2] = put_vp8_epel16_h6_c;
667 dsp->put_vp8_epel_pixels_tab[0][2][0] = put_vp8_epel16_v6_c;
668 dsp->put_vp8_epel_pixels_tab[0][2][2] = put_vp8_epel16_h6v6_c;
669
670 VP78_MC_FUNC(1, 8);
671 VP78_MC_FUNC(2, 4);
672
673 VP78_BILINEAR_MC_FUNC(0, 16);
674 VP78_BILINEAR_MC_FUNC(1, 8);
675 VP78_BILINEAR_MC_FUNC(2, 4);
676
677#if ARCH_AARCH64
678 ff_vp78dsp_init_aarch64(dsp);
679#elif ARCH_ARM
680 ff_vp78dsp_init_arm(dsp);
681#elif ARCH_PPC
682 ff_vp78dsp_init_ppc(dsp);
683#elif ARCH_RISCV
684 ff_vp78dsp_init_riscv(dsp);
685#elif ARCH_X86 && HAVE_X86ASM
686 ff_vp78dsp_init_x86(dsp);
687#endif
688}
689
690#if CONFIG_VP7_DECODER
691LOOP_FILTERS(vp7)
692
693av_cold void ff_vp7dsp_init(VP8DSPContext *dsp)
694{
695 dsp->vp8_luma_dc_wht = vp7_luma_dc_wht_c;
696 dsp->vp8_luma_dc_wht_dc = vp7_luma_dc_wht_dc_c;
697 dsp->vp8_idct_add = vp7_idct_add_c;
698 dsp->vp8_idct_dc_add = vp7_idct_dc_add_c;
699 dsp->vp8_idct_dc_add4y = vp7_idct_dc_add4y_c;
700 dsp->vp8_idct_dc_add4uv = vp7_idct_dc_add4uv_c;
701
702 dsp->vp8_v_loop_filter16y = vp7_v_loop_filter16_c;
703 dsp->vp8_h_loop_filter16y = vp7_h_loop_filter16_c;
704 dsp->vp8_v_loop_filter8uv = vp7_v_loop_filter8uv_c;
705 dsp->vp8_h_loop_filter8uv = vp7_h_loop_filter8uv_c;
706
707 dsp->vp8_v_loop_filter16y_inner = vp7_v_loop_filter16_inner_c;
708 dsp->vp8_h_loop_filter16y_inner = vp7_h_loop_filter16_inner_c;
709 dsp->vp8_v_loop_filter8uv_inner = vp7_v_loop_filter8uv_inner_c;
710 dsp->vp8_h_loop_filter8uv_inner = vp7_h_loop_filter8uv_inner_c;
711
712 dsp->vp8_v_loop_filter_simple = vp7_v_loop_filter_simple_c;
713 dsp->vp8_h_loop_filter_simple = vp7_h_loop_filter_simple_c;
714
715#if ARCH_RISCV
716 ff_vp7dsp_init_riscv(dsp);
717#endif
718}
719#endif /* CONFIG_VP7_DECODER */
720
721#if CONFIG_VP8_DECODER
722LOOP_FILTERS(vp8)
723
724av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
725{
726 dsp->vp8_luma_dc_wht = vp8_luma_dc_wht_c;
727 dsp->vp8_luma_dc_wht_dc = vp8_luma_dc_wht_dc_c;
728 dsp->vp8_idct_add = vp8_idct_add_c;
729 dsp->vp8_idct_dc_add = vp8_idct_dc_add_c;
730 dsp->vp8_idct_dc_add4y = vp8_idct_dc_add4y_c;
731 dsp->vp8_idct_dc_add4uv = vp8_idct_dc_add4uv_c;
732
733 dsp->vp8_v_loop_filter16y = vp8_v_loop_filter16_c;
734 dsp->vp8_h_loop_filter16y = vp8_h_loop_filter16_c;
735 dsp->vp8_v_loop_filter8uv = vp8_v_loop_filter8uv_c;
736 dsp->vp8_h_loop_filter8uv = vp8_h_loop_filter8uv_c;
737
738 dsp->vp8_v_loop_filter16y_inner = vp8_v_loop_filter16_inner_c;
739 dsp->vp8_h_loop_filter16y_inner = vp8_h_loop_filter16_inner_c;
740 dsp->vp8_v_loop_filter8uv_inner = vp8_v_loop_filter8uv_inner_c;
741 dsp->vp8_h_loop_filter8uv_inner = vp8_h_loop_filter8uv_inner_c;
742
743 dsp->vp8_v_loop_filter_simple = vp8_v_loop_filter_simple_c;
744 dsp->vp8_h_loop_filter_simple = vp8_h_loop_filter_simple_c;
745
746#if ARCH_AARCH64
747 ff_vp8dsp_init_aarch64(dsp);
748#elif ARCH_ARM
749 ff_vp8dsp_init_arm(dsp);
750#elif ARCH_RISCV
751 ff_vp8dsp_init_riscv(dsp);
752#elif ARCH_X86 && HAVE_X86ASM
753 ff_vp8dsp_init_x86(dsp);
754#elif ARCH_MIPS
755 ff_vp8dsp_init_mips(dsp);
756#elif ARCH_LOONGARCH
757 ff_vp8dsp_init_loongarch(dsp);
758#endif
759}
760#endif /* CONFIG_VP8_DECODER */
dst
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition dsp.h:87
val
static double val(void *priv, double ch)
Definition aeval.c:77
i
#define i(width, name, range_min, range_max)
Definition cbs_h264.c:63
common.h
common internal and external API header
av_clip_uint8
#define av_clip_uint8
Definition common.h:106
FFABS
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition common.h:74
block
static int16_t block[64]
Definition dct.c:125
cm
#define cm
Definition dvbsubdec.c:40
a
int a
Definition h264pred_template.c:416
PUT_PIXELS
#define PUT_PIXELS(name)
Definition hevcdsp_init_neon.c:72
AV_ZERO64
#define AV_ZERO64(d)
Definition intreadwrite.h:666
VP8_EPEL_V
#define VP8_EPEL_V(SIZE, TAPS)
Definition vp8dsp.c:510
vp7_filter_common
static av_always_inline void vp7_filter_common(uint8_t *p, ptrdiff_t stride, int is4tap)
Definition vp8dsp.c:299
vp8_filter_common
static av_always_inline void vp8_filter_common(uint8_t *p, ptrdiff_t stride, int is4tap)
Definition vp8dsp.c:305
VP78_MC_FUNC
#define VP78_MC_FUNC(IDX, SIZE)
Definition vp8dsp.c:641
vp7_simple_limit
static av_always_inline int vp7_simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
Definition vp8dsp.c:311
filter_common
static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap, int is_vp7)
Definition vp8dsp.c:263
VP78_BILINEAR_MC_FUNC
#define VP78_BILINEAR_MC_FUNC(IDX, SIZE)
Definition vp8dsp.c:652
ff_vp78dsp_init
av_cold void ff_vp78dsp_init(VP8DSPContext *dsp)
Definition vp8dsp.c:663
PUT_PIXELS
#define PUT_PIXELS(WIDTH)
Definition vp8dsp.c:469
LOOP_FILTERS
#define LOOP_FILTERS(vpn)
Definition vp8dsp.c:452
subpel_filters
static const uint8_t subpel_filters[7][6]
Definition vp8dsp.c:459
MK_IDCT_DC_ADD4_C
#define MK_IDCT_DC_ADD4_C(name)
Definition vp8dsp.c:36
VP8_EPEL_HV
#define VP8_EPEL_HV(SIZE, HTAPS, VTAPS)
Definition vp8dsp.c:528
NORMAL_LIMIT
#define NORMAL_LIMIT(vpn)
E - limit at the macroblock edge I - limit for interior difference.
Definition vp8dsp.c:329
clip_int8
#define clip_int8(n)
Definition vp8dsp.c:261
LOAD_PIXELS
#define LOAD_PIXELS
Definition vp8dsp.c:251
VP8_EPEL_H
#define VP8_EPEL_H(SIZE, TAPS)
Definition vp8dsp.c:492
vp8_simple_limit
static av_always_inline int vp8_simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
Definition vp8dsp.c:318
filter_mbedge
static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
Definition vp8dsp.c:351
VP8_BILINEAR
#define VP8_BILINEAR(SIZE)
Definition vp8dsp.c:582
av_always_inline
#define av_always_inline
Definition attributes.h:72
av_cold
#define av_cold
Definition attributes.h:117
IS_VP8
#define IS_VP8(avctx)
Definition libvpxenc.c:52
w
uint8_t w
Definition llvidencdsp.c:39
FFMIN
#define FFMIN(a, b)
Definition macros.h:49
MAX_NEG_CROP
#define MAX_NEG_CROP
Definition mathops.h:31
ff_crop_tab
#define ff_crop_tab
Definition motionpixels_tablegen.c:26
c1
static const uint64_t c1
Definition murmur3.c:52
ff_vp8dsp_init_riscv
av_cold void ff_vp8dsp_init_riscv(VP8DSPContext *c)
Definition vp8dsp_init.c:118
ff_vp78dsp_init_riscv
av_cold void ff_vp78dsp_init_riscv(VP8DSPContext *c)
Definition vp8dsp_init.c:46
VP8DSPContext::vp8_h_loop_filter8uv_inner
void(* vp8_h_loop_filter8uv_inner)(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
Definition vp8dsp.h:65
VP8DSPContext::vp8_v_loop_filter_simple
void(* vp8_v_loop_filter_simple)(uint8_t *dst, ptrdiff_t stride, int flim)
Definition vp8dsp.h:69
VP8DSPContext::vp8_v_loop_filter16y
void(* vp8_v_loop_filter16y)(uint8_t *dst, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
Definition vp8dsp.h:48
VP8DSPContext::vp8_luma_dc_wht_dc
void(* vp8_luma_dc_wht_dc)(int16_t block[4][4][16], int16_t dc[16])
Definition vp8dsp.h:39
VP8DSPContext::vp8_idct_dc_add
void(* vp8_idct_dc_add)(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
Definition vp8dsp.h:41
VP8DSPContext::vp8_idct_dc_add4uv
void(* vp8_idct_dc_add4uv)(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)
Definition vp8dsp.h:44
VP8DSPContext::vp8_h_loop_filter16y
void(* vp8_h_loop_filter16y)(uint8_t *dst, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
Definition vp8dsp.h:50
VP8DSPContext::vp8_luma_dc_wht
void(* vp8_luma_dc_wht)(int16_t block[4][4][16], int16_t dc[16])
Definition vp8dsp.h:38
VP8DSPContext::vp8_v_loop_filter8uv_inner
void(* vp8_v_loop_filter8uv_inner)(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
Definition vp8dsp.h:62
VP8DSPContext::vp8_h_loop_filter8uv
void(* vp8_h_loop_filter8uv)(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
Definition vp8dsp.h:54
VP8DSPContext::vp8_h_loop_filter16y_inner
void(* vp8_h_loop_filter16y_inner)(uint8_t *dst, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
Definition vp8dsp.h:60
VP8DSPContext::vp8_idct_dc_add4y
void(* vp8_idct_dc_add4y)(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)
Definition vp8dsp.h:42
VP8DSPContext::vp8_h_loop_filter_simple
void(* vp8_h_loop_filter_simple)(uint8_t *dst, ptrdiff_t stride, int flim)
Definition vp8dsp.h:70
VP8DSPContext::vp8_v_loop_filter8uv
void(* vp8_v_loop_filter8uv)(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
Definition vp8dsp.h:52
VP8DSPContext::vp8_v_loop_filter16y_inner
void(* vp8_v_loop_filter16y_inner)(uint8_t *dst, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
Definition vp8dsp.h:58
VP8DSPContext::vp8_idct_add
void(* vp8_idct_add)(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
Definition vp8dsp.h:40
VP8DSPContext::put_vp8_epel_pixels_tab
vp8_mc_func put_vp8_epel_pixels_tab[3][3][3]
first dimension: 4-log2(width) second dimension: 0 if no vertical interpolation is needed; 1 4-tap ve...
Definition vp8dsp.h:80
stride
#define stride
Definition h264pred_template.c:536
tmp
static uint8_t tmp[40]
Definition aes_ctr.c:52
q1
static const uint8_t q1[256]
Definition twofish.c:100
q0
static const uint8_t q0[256]
Definition twofish.c:81
b1
static double b1(void *priv, double x, double y)
Definition vf_xfade.c:2034
a0
static double a0(void *priv, double x, double y)
Definition vf_xfade.c:2028
a2
static double a2(void *priv, double x, double y)
Definition vf_xfade.c:2030
a1
static double a1(void *priv, double x, double y)
Definition vf_xfade.c:2029
ff_vp7dsp_init_riscv
av_cold void ff_vp7dsp_init_riscv(VP8DSPContext *c)
Definition vp7dsp_init.c:42
vp8dsp.h
VP8 compatible video decoder.
ff_vp8dsp_init_x86
void ff_vp8dsp_init_x86(VP8DSPContext *c)
Definition vp8dsp_init.c:271
ff_vp78dsp_init_x86
void ff_vp78dsp_init_x86(VP8DSPContext *c)
Definition vp8dsp_init.c:241
ff_vp7dsp_init
void ff_vp7dsp_init(VP8DSPContext *c)
IS_VP7
#define IS_VP7
Definition vp8dsp.h:103
ff_vp8dsp_init
void ff_vp8dsp_init(VP8DSPContext *c)
ff_vp78dsp_init_ppc
av_cold void ff_vp78dsp_init_ppc(VP8DSPContext *c)
Definition vp8dsp_altivec.c:332
ff_vp78dsp_init_aarch64
av_cold void ff_vp78dsp_init_aarch64(VP8DSPContext *dsp)
Definition vp8dsp_init_aarch64.c:43
ff_vp8dsp_init_aarch64
av_cold void ff_vp8dsp_init_aarch64(VP8DSPContext *dsp)
Definition vp8dsp_init_aarch64.c:101
ff_vp78dsp_init_arm
av_cold void ff_vp78dsp_init_arm(VP8DSPContext *dsp)
Definition vp8dsp_init_arm.c:26
ff_vp8dsp_init_arm
av_cold void ff_vp8dsp_init_arm(VP8DSPContext *dsp)
Definition vp8dsp_init_arm.c:36
ff_vp8dsp_init_loongarch
av_cold void ff_vp8dsp_init_loongarch(VP8DSPContext *dsp)
Definition vp8dsp_init_loongarch.c:44
ff_vp8dsp_init_mips
av_cold void ff_vp8dsp_init_mips(VP8DSPContext *dsp)
Definition vp8dsp_init_mips.c:77
hev
static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)
Definition vp8dsp_mmi.c:732
Generated on Sun Aug 16 2026 20:20:53 for FFmpeg by doxygen 1.13.2