1/*
2 * VP9 compatible video decoder
3 *
4 * Copyright (C) 2013 Ronald S. Bultje <rsbultje gmail com>
5 * Copyright (C) 2013 Clément Bœsch <u pkh me>
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 "libavutil/common.h"
25#include "bit_depth_template.c"
26#include "vp9dsp.h"
27
28#if BIT_DEPTH != 12
29
30// FIXME see whether we can merge parts of this (perhaps at least 4x4 and 8x8)
31// back with h264pred.[ch]
32
33static void vert_4x4_c(uint8_t *_dst, ptrdiff_t stride,
34                       const uint8_t *left, const uint8_t *_top)
35{
36    pixel *dst = (pixel *) _dst;
37    const pixel *top = (const pixel *) _top;
38    pixel4 p4 = AV_RN4PA(top);
39
40    stride /= sizeof(pixel);
41    AV_WN4PA(dst + stride * 0, p4);
42    AV_WN4PA(dst + stride * 1, p4);
43    AV_WN4PA(dst + stride * 2, p4);
44    AV_WN4PA(dst + stride * 3, p4);
45}
46
47static void vert_8x8_c(uint8_t *_dst, ptrdiff_t stride,
48                       const uint8_t *left, const uint8_t *_top)
49{
50    pixel *dst = (pixel *) _dst;
51    const pixel *top = (const pixel *) _top;
52    pixel4 p4a = AV_RN4PA(top + 0);
53    pixel4 p4b = AV_RN4PA(top + 4);
54    int y;
55
56    stride /= sizeof(pixel);
57    for (y = 0; y < 8; y++) {
58        AV_WN4PA(dst + 0, p4a);
59        AV_WN4PA(dst + 4, p4b);
60        dst += stride;
61    }
62}
63
64static void vert_16x16_c(uint8_t *_dst, ptrdiff_t stride,
65                         const uint8_t *left, const uint8_t *_top)
66{
67    pixel *dst = (pixel *) _dst;
68    const pixel *top = (const pixel *) _top;
69    pixel4 p4a = AV_RN4PA(top +  0);
70    pixel4 p4b = AV_RN4PA(top +  4);
71    pixel4 p4c = AV_RN4PA(top +  8);
72    pixel4 p4d = AV_RN4PA(top + 12);
73    int y;
74
75    stride /= sizeof(pixel);
76    for (y = 0; y < 16; y++) {
77        AV_WN4PA(dst +  0, p4a);
78        AV_WN4PA(dst +  4, p4b);
79        AV_WN4PA(dst +  8, p4c);
80        AV_WN4PA(dst + 12, p4d);
81        dst += stride;
82    }
83}
84
85static void vert_32x32_c(uint8_t *_dst, ptrdiff_t stride,
86                         const uint8_t *left, const uint8_t *_top)
87{
88    pixel *dst = (pixel *) _dst;
89    const pixel *top = (const pixel *) _top;
90    pixel4 p4a = AV_RN4PA(top +  0);
91    pixel4 p4b = AV_RN4PA(top +  4);
92    pixel4 p4c = AV_RN4PA(top +  8);
93    pixel4 p4d = AV_RN4PA(top + 12);
94    pixel4 p4e = AV_RN4PA(top + 16);
95    pixel4 p4f = AV_RN4PA(top + 20);
96    pixel4 p4g = AV_RN4PA(top + 24);
97    pixel4 p4h = AV_RN4PA(top + 28);
98    int y;
99
100    stride /= sizeof(pixel);
101    for (y = 0; y < 32; y++) {
102        AV_WN4PA(dst +  0, p4a);
103        AV_WN4PA(dst +  4, p4b);
104        AV_WN4PA(dst +  8, p4c);
105        AV_WN4PA(dst + 12, p4d);
106        AV_WN4PA(dst + 16, p4e);
107        AV_WN4PA(dst + 20, p4f);
108        AV_WN4PA(dst + 24, p4g);
109        AV_WN4PA(dst + 28, p4h);
110        dst += stride;
111    }
112}
113
114static void hor_4x4_c(uint8_t *_dst, ptrdiff_t stride,
115                      const uint8_t *_left, const uint8_t *top)
116{
117    pixel *dst = (pixel *) _dst;
118    const pixel *left = (const pixel *) _left;
119
120    stride /= sizeof(pixel);
121    AV_WN4PA(dst + stride * 0, PIXEL_SPLAT_X4(left[3]));
122    AV_WN4PA(dst + stride * 1, PIXEL_SPLAT_X4(left[2]));
123    AV_WN4PA(dst + stride * 2, PIXEL_SPLAT_X4(left[1]));
124    AV_WN4PA(dst + stride * 3, PIXEL_SPLAT_X4(left[0]));
125}
126
127static void hor_8x8_c(uint8_t *_dst, ptrdiff_t stride,
128                      const uint8_t *_left, const uint8_t *top)
129{
130    pixel *dst = (pixel *) _dst;
131    const pixel *left = (const pixel *) _left;
132    int y;
133
134    stride /= sizeof(pixel);
135    for (y = 0; y < 8; y++) {
136        pixel4 p4 = PIXEL_SPLAT_X4(left[7 - y]);
137
138        AV_WN4PA(dst + 0, p4);
139        AV_WN4PA(dst + 4, p4);
140        dst += stride;
141    }
142}
143
144static void hor_16x16_c(uint8_t *_dst, ptrdiff_t stride,
145                        const uint8_t *_left, const uint8_t *top)
146{
147    pixel *dst = (pixel *) _dst;
148    const pixel *left = (const pixel *) _left;
149    int y;
150
151    stride /= sizeof(pixel);
152    for (y = 0; y < 16; y++) {
153        pixel4 p4 = PIXEL_SPLAT_X4(left[15 - y]);
154
155        AV_WN4PA(dst +  0, p4);
156        AV_WN4PA(dst +  4, p4);
157        AV_WN4PA(dst +  8, p4);
158        AV_WN4PA(dst + 12, p4);
159        dst += stride;
160    }
161}
162
163static void hor_32x32_c(uint8_t *_dst, ptrdiff_t stride,
164                        const uint8_t *_left, const uint8_t *top)
165{
166    pixel *dst = (pixel *) _dst;
167    const pixel *left = (const pixel *) _left;
168    int y;
169
170    stride /= sizeof(pixel);
171    for (y = 0; y < 32; y++) {
172        pixel4 p4 = PIXEL_SPLAT_X4(left[31 - y]);
173
174        AV_WN4PA(dst +  0, p4);
175        AV_WN4PA(dst +  4, p4);
176        AV_WN4PA(dst +  8, p4);
177        AV_WN4PA(dst + 12, p4);
178        AV_WN4PA(dst + 16, p4);
179        AV_WN4PA(dst + 20, p4);
180        AV_WN4PA(dst + 24, p4);
181        AV_WN4PA(dst + 28, p4);
182        dst += stride;
183    }
184}
185
186#endif /* BIT_DEPTH != 12 */
187
188static void tm_4x4_c(uint8_t *_dst, ptrdiff_t stride,
189                     const uint8_t *_left, const uint8_t *_top)
190{
191    pixel *dst = (pixel *) _dst;
192    const pixel *left = (const pixel *) _left;
193    const pixel *top = (const pixel *) _top;
194    int y, tl = top[-1];
195
196    stride /= sizeof(pixel);
197    for (y = 0; y < 4; y++) {
198        int l_m_tl = left[3 - y] - tl;
199
200        dst[0] = av_clip_pixel(top[0] + l_m_tl);
201        dst[1] = av_clip_pixel(top[1] + l_m_tl);
202        dst[2] = av_clip_pixel(top[2] + l_m_tl);
203        dst[3] = av_clip_pixel(top[3] + l_m_tl);
204        dst += stride;
205    }
206}
207
208static void tm_8x8_c(uint8_t *_dst, ptrdiff_t stride,
209                     const uint8_t *_left, const uint8_t *_top)
210{
211    pixel *dst = (pixel *) _dst;
212    const pixel *left = (const pixel *) _left;
213    const pixel *top = (const pixel *) _top;
214    int y, tl = top[-1];
215
216    stride /= sizeof(pixel);
217    for (y = 0; y < 8; y++) {
218        int l_m_tl = left[7 - y] - tl;
219
220        dst[0] = av_clip_pixel(top[0] + l_m_tl);
221        dst[1] = av_clip_pixel(top[1] + l_m_tl);
222        dst[2] = av_clip_pixel(top[2] + l_m_tl);
223        dst[3] = av_clip_pixel(top[3] + l_m_tl);
224        dst[4] = av_clip_pixel(top[4] + l_m_tl);
225        dst[5] = av_clip_pixel(top[5] + l_m_tl);
226        dst[6] = av_clip_pixel(top[6] + l_m_tl);
227        dst[7] = av_clip_pixel(top[7] + l_m_tl);
228        dst += stride;
229    }
230}
231
232static void tm_16x16_c(uint8_t *_dst, ptrdiff_t stride,
233                       const uint8_t *_left, const uint8_t *_top)
234{
235    pixel *dst = (pixel *) _dst;
236    const pixel *left = (const pixel *) _left;
237    const pixel *top = (const pixel *) _top;
238    int y, tl = top[-1];
239
240    stride /= sizeof(pixel);
241    for (y = 0; y < 16; y++) {
242        int l_m_tl = left[15 - y] - tl;
243
244        dst[ 0] = av_clip_pixel(top[ 0] + l_m_tl);
245        dst[ 1] = av_clip_pixel(top[ 1] + l_m_tl);
246        dst[ 2] = av_clip_pixel(top[ 2] + l_m_tl);
247        dst[ 3] = av_clip_pixel(top[ 3] + l_m_tl);
248        dst[ 4] = av_clip_pixel(top[ 4] + l_m_tl);
249        dst[ 5] = av_clip_pixel(top[ 5] + l_m_tl);
250        dst[ 6] = av_clip_pixel(top[ 6] + l_m_tl);
251        dst[ 7] = av_clip_pixel(top[ 7] + l_m_tl);
252        dst[ 8] = av_clip_pixel(top[ 8] + l_m_tl);
253        dst[ 9] = av_clip_pixel(top[ 9] + l_m_tl);
254        dst[10] = av_clip_pixel(top[10] + l_m_tl);
255        dst[11] = av_clip_pixel(top[11] + l_m_tl);
256        dst[12] = av_clip_pixel(top[12] + l_m_tl);
257        dst[13] = av_clip_pixel(top[13] + l_m_tl);
258        dst[14] = av_clip_pixel(top[14] + l_m_tl);
259        dst[15] = av_clip_pixel(top[15] + l_m_tl);
260        dst += stride;
261    }
262}
263
264static void tm_32x32_c(uint8_t *_dst, ptrdiff_t stride,
265                       const uint8_t *_left, const uint8_t *_top)
266{
267    pixel *dst = (pixel *) _dst;
268    const pixel *left = (const pixel *) _left;
269    const pixel *top = (const pixel *) _top;
270    int y, tl = top[-1];
271
272    stride /= sizeof(pixel);
273    for (y = 0; y < 32; y++) {
274        int l_m_tl = left[31 - y] - tl;
275
276        dst[ 0] = av_clip_pixel(top[ 0] + l_m_tl);
277        dst[ 1] = av_clip_pixel(top[ 1] + l_m_tl);
278        dst[ 2] = av_clip_pixel(top[ 2] + l_m_tl);
279        dst[ 3] = av_clip_pixel(top[ 3] + l_m_tl);
280        dst[ 4] = av_clip_pixel(top[ 4] + l_m_tl);
281        dst[ 5] = av_clip_pixel(top[ 5] + l_m_tl);
282        dst[ 6] = av_clip_pixel(top[ 6] + l_m_tl);
283        dst[ 7] = av_clip_pixel(top[ 7] + l_m_tl);
284        dst[ 8] = av_clip_pixel(top[ 8] + l_m_tl);
285        dst[ 9] = av_clip_pixel(top[ 9] + l_m_tl);
286        dst[10] = av_clip_pixel(top[10] + l_m_tl);
287        dst[11] = av_clip_pixel(top[11] + l_m_tl);
288        dst[12] = av_clip_pixel(top[12] + l_m_tl);
289        dst[13] = av_clip_pixel(top[13] + l_m_tl);
290        dst[14] = av_clip_pixel(top[14] + l_m_tl);
291        dst[15] = av_clip_pixel(top[15] + l_m_tl);
292        dst[16] = av_clip_pixel(top[16] + l_m_tl);
293        dst[17] = av_clip_pixel(top[17] + l_m_tl);
294        dst[18] = av_clip_pixel(top[18] + l_m_tl);
295        dst[19] = av_clip_pixel(top[19] + l_m_tl);
296        dst[20] = av_clip_pixel(top[20] + l_m_tl);
297        dst[21] = av_clip_pixel(top[21] + l_m_tl);
298        dst[22] = av_clip_pixel(top[22] + l_m_tl);
299        dst[23] = av_clip_pixel(top[23] + l_m_tl);
300        dst[24] = av_clip_pixel(top[24] + l_m_tl);
301        dst[25] = av_clip_pixel(top[25] + l_m_tl);
302        dst[26] = av_clip_pixel(top[26] + l_m_tl);
303        dst[27] = av_clip_pixel(top[27] + l_m_tl);
304        dst[28] = av_clip_pixel(top[28] + l_m_tl);
305        dst[29] = av_clip_pixel(top[29] + l_m_tl);
306        dst[30] = av_clip_pixel(top[30] + l_m_tl);
307        dst[31] = av_clip_pixel(top[31] + l_m_tl);
308        dst += stride;
309    }
310}
311
312#if BIT_DEPTH != 12
313
314static void dc_4x4_c(uint8_t *_dst, ptrdiff_t stride,
315                     const uint8_t *_left, const uint8_t *_top)
316{
317    pixel *dst = (pixel *) _dst;
318    const pixel *left = (const pixel *) _left;
319    const pixel *top = (const pixel *) _top;
320    pixel4 dc = PIXEL_SPLAT_X4((left[0] + left[1] + left[2] + left[3] +
321                                top[0] + top[1] + top[2] + top[3] + 4) >> 3);
322
323    stride /= sizeof(pixel);
324    AV_WN4PA(dst + stride * 0, dc);
325    AV_WN4PA(dst + stride * 1, dc);
326    AV_WN4PA(dst + stride * 2, dc);
327    AV_WN4PA(dst + stride * 3, dc);
328}
329
330static void dc_8x8_c(uint8_t *_dst, ptrdiff_t stride,
331                     const uint8_t *_left, const uint8_t *_top)
332{
333    pixel *dst = (pixel *) _dst;
334    const pixel *left = (const pixel *) _left;
335    const pixel *top = (const pixel *) _top;
336    pixel4 dc = PIXEL_SPLAT_X4
337        ((left[0] + left[1] + left[2] + left[3] + left[4] + left[5] +
338          left[6] + left[7] + top[0] + top[1] + top[2] + top[3] +
339          top[4] + top[5] + top[6] + top[7] + 8) >> 4);
340    int y;
341
342    stride /= sizeof(pixel);
343    for (y = 0; y < 8; y++) {
344        AV_WN4PA(dst + 0, dc);
345        AV_WN4PA(dst + 4, dc);
346        dst += stride;
347    }
348}
349
350static void dc_16x16_c(uint8_t *_dst, ptrdiff_t stride,
351                       const uint8_t *_left, const uint8_t *_top)
352{
353    pixel *dst = (pixel *) _dst;
354    const pixel *left = (const pixel *) _left;
355    const pixel *top = (const pixel *) _top;
356    pixel4 dc = PIXEL_SPLAT_X4
357        ((left[0] + left[1] + left[2] + left[3] + left[4] + left[5] + left[6] +
358          left[7] + left[8] + left[9] + left[10] + left[11] + left[12] +
359          left[13] + left[14] + left[15] + top[0] + top[1] + top[2] + top[3] +
360          top[4] + top[5] + top[6] + top[7] + top[8] + top[9] + top[10] +
361          top[11] + top[12] + top[13] + top[14] + top[15] + 16) >> 5);
362    int y;
363
364    stride /= sizeof(pixel);
365    for (y = 0; y < 16; y++) {
366        AV_WN4PA(dst +  0, dc);
367        AV_WN4PA(dst +  4, dc);
368        AV_WN4PA(dst +  8, dc);
369        AV_WN4PA(dst + 12, dc);
370        dst += stride;
371    }
372}
373
374static void dc_32x32_c(uint8_t *_dst, ptrdiff_t stride,
375                       const uint8_t *_left, const uint8_t *_top)
376{
377    pixel *dst = (pixel *) _dst;
378    const pixel *left = (const pixel *) _left;
379    const pixel *top = (const pixel *) _top;
380    pixel4 dc = PIXEL_SPLAT_X4
381        ((left[0] + left[1] + left[2] + left[3] + left[4] + left[5] + left[6] +
382          left[7] + left[8] + left[9] + left[10] + left[11] + left[12] +
383          left[13] + left[14] + left[15] + left[16] + left[17] + left[18] +
384          left[19] + left[20] + left[21] + left[22] + left[23] + left[24] +
385          left[25] + left[26] + left[27] + left[28] + left[29] + left[30] +
386          left[31] + top[0] + top[1] + top[2] + top[3] + top[4] + top[5] +
387          top[6] + top[7] + top[8] + top[9] + top[10] + top[11] + top[12] +
388          top[13] + top[14] + top[15] + top[16] + top[17] + top[18] + top[19] +
389          top[20] + top[21] + top[22] + top[23] + top[24] + top[25] + top[26] +
390          top[27] + top[28] + top[29] + top[30] + top[31] + 32) >> 6);
391    int y;
392
393    stride /= sizeof(pixel);
394    for (y = 0; y < 32; y++) {
395        AV_WN4PA(dst +  0, dc);
396        AV_WN4PA(dst +  4, dc);
397        AV_WN4PA(dst +  8, dc);
398        AV_WN4PA(dst + 12, dc);
399        AV_WN4PA(dst + 16, dc);
400        AV_WN4PA(dst + 20, dc);
401        AV_WN4PA(dst + 24, dc);
402        AV_WN4PA(dst + 28, dc);
403        dst += stride;
404    }
405}
406
407static void dc_left_4x4_c(uint8_t *_dst, ptrdiff_t stride,
408                          const uint8_t *_left, const uint8_t *top)
409{
410    pixel *dst = (pixel *) _dst;
411    const pixel *left = (const pixel *) _left;
412    pixel4 dc = PIXEL_SPLAT_X4((left[0] + left[1] + left[2] + left[3] + 2) >> 2);
413
414    stride /= sizeof(pixel);
415    AV_WN4PA(dst + stride * 0, dc);
416    AV_WN4PA(dst + stride * 1, dc);
417    AV_WN4PA(dst + stride * 2, dc);
418    AV_WN4PA(dst + stride * 3, dc);
419}
420
421static void dc_left_8x8_c(uint8_t *_dst, ptrdiff_t stride,
422                          const uint8_t *_left, const uint8_t *top)
423{
424    pixel *dst = (pixel *) _dst;
425    const pixel *left = (const pixel *) _left;
426    pixel4 dc = PIXEL_SPLAT_X4
427        ((left[0] + left[1] + left[2] + left[3] +
428          left[4] + left[5] + left[6] + left[7] + 4) >> 3);
429    int y;
430
431    stride /= sizeof(pixel);
432    for (y = 0; y < 8; y++) {
433        AV_WN4PA(dst + 0, dc);
434        AV_WN4PA(dst + 4, dc);
435        dst += stride;
436    }
437}
438
439static void dc_left_16x16_c(uint8_t *_dst, ptrdiff_t stride,
440                            const uint8_t *_left, const uint8_t *top)
441{
442    pixel *dst = (pixel *) _dst;
443    const pixel *left = (const pixel *) _left;
444    pixel4 dc = PIXEL_SPLAT_X4
445        ((left[0] + left[1] + left[2] + left[3] + left[4] + left[5] +
446          left[6] + left[7] + left[8] + left[9] + left[10] + left[11] +
447          left[12] + left[13] + left[14] + left[15] + 8) >> 4);
448    int y;
449
450    stride /= sizeof(pixel);
451    for (y = 0; y < 16; y++) {
452        AV_WN4PA(dst +  0, dc);
453        AV_WN4PA(dst +  4, dc);
454        AV_WN4PA(dst +  8, dc);
455        AV_WN4PA(dst + 12, dc);
456        dst += stride;
457    }
458}
459
460static void dc_left_32x32_c(uint8_t *_dst, ptrdiff_t stride,
461                            const uint8_t *_left, const uint8_t *top)
462{
463    pixel *dst = (pixel *) _dst;
464    const pixel *left = (const pixel *) _left;
465    pixel4 dc = PIXEL_SPLAT_X4
466        ((left[0] + left[1] + left[2] + left[3] + left[4] + left[5] +
467          left[6] + left[7] + left[8] + left[9] + left[10] + left[11] +
468          left[12] + left[13] + left[14] + left[15] + left[16] + left[17] +
469          left[18] + left[19] + left[20] + left[21] + left[22] + left[23] +
470          left[24] + left[25] + left[26] + left[27] + left[28] + left[29] +
471          left[30] + left[31] + 16) >> 5);
472    int y;
473
474    stride /= sizeof(pixel);
475    for (y = 0; y < 32; y++) {
476        AV_WN4PA(dst +  0, dc);
477        AV_WN4PA(dst +  4, dc);
478        AV_WN4PA(dst +  8, dc);
479        AV_WN4PA(dst + 12, dc);
480        AV_WN4PA(dst + 16, dc);
481        AV_WN4PA(dst + 20, dc);
482        AV_WN4PA(dst + 24, dc);
483        AV_WN4PA(dst + 28, dc);
484        dst += stride;
485    }
486}
487
488static void dc_top_4x4_c(uint8_t *_dst, ptrdiff_t stride,
489                         const uint8_t *left, const uint8_t *_top)
490{
491    pixel *dst = (pixel *) _dst;
492    const pixel *top = (const pixel *) _top;
493    pixel4 dc = PIXEL_SPLAT_X4((top[0] + top[1] + top[2] + top[3] + 2) >> 2);
494
495    stride /= sizeof(pixel);
496    AV_WN4PA(dst + stride * 0, dc);
497    AV_WN4PA(dst + stride * 1, dc);
498    AV_WN4PA(dst + stride * 2, dc);
499    AV_WN4PA(dst + stride * 3, dc);
500}
501
502static void dc_top_8x8_c(uint8_t *_dst, ptrdiff_t stride,
503                         const uint8_t *left, const uint8_t *_top)
504{
505    pixel *dst = (pixel *) _dst;
506    const pixel *top = (const pixel *) _top;
507    pixel4 dc = PIXEL_SPLAT_X4
508        ((top[0] + top[1] + top[2] + top[3] +
509          top[4] + top[5] + top[6] + top[7] + 4) >> 3);
510    int y;
511
512    stride /= sizeof(pixel);
513    for (y = 0; y < 8; y++) {
514        AV_WN4PA(dst + 0, dc);
515        AV_WN4PA(dst + 4, dc);
516        dst += stride;
517    }
518}
519
520static void dc_top_16x16_c(uint8_t *_dst, ptrdiff_t stride,
521                           const uint8_t *left, const uint8_t *_top)
522{
523    pixel *dst = (pixel *) _dst;
524    const pixel *top = (const pixel *) _top;
525    pixel4 dc = PIXEL_SPLAT_X4
526        ((top[0] + top[1] + top[2] + top[3] + top[4] + top[5] +
527          top[6] + top[7] + top[8] + top[9] + top[10] + top[11] +
528          top[12] + top[13] + top[14] + top[15] + 8) >> 4);
529    int y;
530
531    stride /= sizeof(pixel);
532    for (y = 0; y < 16; y++) {
533        AV_WN4PA(dst +  0, dc);
534        AV_WN4PA(dst +  4, dc);
535        AV_WN4PA(dst +  8, dc);
536        AV_WN4PA(dst + 12, dc);
537        dst += stride;
538    }
539}
540
541static void dc_top_32x32_c(uint8_t *_dst, ptrdiff_t stride,
542                           const uint8_t *left, const uint8_t *_top)
543{
544    pixel *dst = (pixel *) _dst;
545    const pixel *top = (const pixel *) _top;
546    pixel4 dc = PIXEL_SPLAT_X4
547        ((top[0] + top[1] + top[2] + top[3] + top[4] + top[5] +
548          top[6] + top[7] + top[8] + top[9] + top[10] + top[11] +
549          top[12] + top[13] + top[14] + top[15] + top[16] + top[17] +
550          top[18] + top[19] + top[20] + top[21] + top[22] + top[23] +
551          top[24] + top[25] + top[26] + top[27] + top[28] + top[29] +
552          top[30] + top[31] + 16) >> 5);
553    int y;
554
555    stride /= sizeof(pixel);
556    for (y = 0; y < 32; y++) {
557        AV_WN4PA(dst +  0, dc);
558        AV_WN4PA(dst +  4, dc);
559        AV_WN4PA(dst +  8, dc);
560        AV_WN4PA(dst + 12, dc);
561        AV_WN4PA(dst + 16, dc);
562        AV_WN4PA(dst + 20, dc);
563        AV_WN4PA(dst + 24, dc);
564        AV_WN4PA(dst + 28, dc);
565        dst += stride;
566    }
567}
568
569#endif /* BIT_DEPTH != 12 */
570
571static void dc_128_4x4_c(uint8_t *_dst, ptrdiff_t stride,
572                         const uint8_t *left, const uint8_t *top)
573{
574    pixel *dst = (pixel *) _dst;
575    pixel4 val = PIXEL_SPLAT_X4(128 << (BIT_DEPTH - 8));
576
577    stride /= sizeof(pixel);
578    AV_WN4PA(dst + stride * 0, val);
579    AV_WN4PA(dst + stride * 1, val);
580    AV_WN4PA(dst + stride * 2, val);
581    AV_WN4PA(dst + stride * 3, val);
582}
583
584static void dc_128_8x8_c(uint8_t *_dst, ptrdiff_t stride,
585                         const uint8_t *left, const uint8_t *top)
586{
587    pixel *dst = (pixel *) _dst;
588    pixel4 val = PIXEL_SPLAT_X4(128 << (BIT_DEPTH - 8));
589    int y;
590
591    stride /= sizeof(pixel);
592    for (y = 0; y < 8; y++) {
593        AV_WN4PA(dst + 0, val);
594        AV_WN4PA(dst + 4, val);
595        dst += stride;
596    }
597}
598
599static void dc_128_16x16_c(uint8_t *_dst, ptrdiff_t stride,
600                           const uint8_t *left, const uint8_t *top)
601{
602    pixel *dst = (pixel *) _dst;
603    pixel4 val = PIXEL_SPLAT_X4(128 << (BIT_DEPTH - 8));
604    int y;
605
606    stride /= sizeof(pixel);
607    for (y = 0; y < 16; y++) {
608        AV_WN4PA(dst +  0, val);
609        AV_WN4PA(dst +  4, val);
610        AV_WN4PA(dst +  8, val);
611        AV_WN4PA(dst + 12, val);
612        dst += stride;
613    }
614}
615
616static void dc_128_32x32_c(uint8_t *_dst, ptrdiff_t stride,
617                           const uint8_t *left, const uint8_t *top)
618{
619    pixel *dst = (pixel *) _dst;
620    pixel4 val = PIXEL_SPLAT_X4(128 << (BIT_DEPTH - 8));
621    int y;
622
623    stride /= sizeof(pixel);
624    for (y = 0; y < 32; y++) {
625        AV_WN4PA(dst +  0, val);
626        AV_WN4PA(dst +  4, val);
627        AV_WN4PA(dst +  8, val);
628        AV_WN4PA(dst + 12, val);
629        AV_WN4PA(dst + 16, val);
630        AV_WN4PA(dst + 20, val);
631        AV_WN4PA(dst + 24, val);
632        AV_WN4PA(dst + 28, val);
633        dst += stride;
634    }
635}
636
637static void dc_127_4x4_c(uint8_t *_dst, ptrdiff_t stride,
638                         const uint8_t *left, const uint8_t *top)
639{
640    pixel *dst = (pixel *) _dst;
641    pixel4 val = PIXEL_SPLAT_X4((128 << (BIT_DEPTH - 8)) - 1);
642
643    stride /= sizeof(pixel);
644    AV_WN4PA(dst + stride * 0, val);
645    AV_WN4PA(dst + stride * 1, val);
646    AV_WN4PA(dst + stride * 2, val);
647    AV_WN4PA(dst + stride * 3, val);}
648
649static void dc_127_8x8_c(uint8_t *_dst, ptrdiff_t stride,
650                         const uint8_t *left, const uint8_t *top)
651{
652    pixel *dst = (pixel *) _dst;
653    pixel4 val = PIXEL_SPLAT_X4((128 << (BIT_DEPTH - 8)) - 1);
654    int y;
655
656    stride /= sizeof(pixel);
657    for (y = 0; y < 8; y++) {
658        AV_WN4PA(dst + 0, val);
659        AV_WN4PA(dst + 4, val);
660        dst += stride;
661    }
662}
663
664static void dc_127_16x16_c(uint8_t *_dst, ptrdiff_t stride,
665                           const uint8_t *left, const uint8_t *top)
666{
667    pixel *dst = (pixel *) _dst;
668    pixel4 val = PIXEL_SPLAT_X4((128 << (BIT_DEPTH - 8)) - 1);
669    int y;
670
671    stride /= sizeof(pixel);
672    for (y = 0; y < 16; y++) {
673        AV_WN4PA(dst +  0, val);
674        AV_WN4PA(dst +  4, val);
675        AV_WN4PA(dst +  8, val);
676        AV_WN4PA(dst + 12, val);
677        dst += stride;
678    }
679}
680
681static void dc_127_32x32_c(uint8_t *_dst, ptrdiff_t stride,
682                           const uint8_t *left, const uint8_t *top)
683{
684    pixel *dst = (pixel *) _dst;
685    pixel4 val = PIXEL_SPLAT_X4((128 << (BIT_DEPTH - 8)) - 1);
686    int y;
687
688    stride /= sizeof(pixel);
689    for (y = 0; y < 32; y++) {
690        AV_WN4PA(dst +  0, val);
691        AV_WN4PA(dst +  4, val);
692        AV_WN4PA(dst +  8, val);
693        AV_WN4PA(dst + 12, val);
694        AV_WN4PA(dst + 16, val);
695        AV_WN4PA(dst + 20, val);
696        AV_WN4PA(dst + 24, val);
697        AV_WN4PA(dst + 28, val);
698        dst += stride;
699    }
700}
701
702static void dc_129_4x4_c(uint8_t *_dst, ptrdiff_t stride,
703                         const uint8_t *left, const uint8_t *top)
704{
705    pixel *dst = (pixel *) _dst;
706    pixel4 val = PIXEL_SPLAT_X4((128 << (BIT_DEPTH - 8)) + 1);
707
708    stride /= sizeof(pixel);
709    AV_WN4PA(dst + stride * 0, val);
710    AV_WN4PA(dst + stride * 1, val);
711    AV_WN4PA(dst + stride * 2, val);
712    AV_WN4PA(dst + stride * 3, val);
713}
714
715static void dc_129_8x8_c(uint8_t *_dst, ptrdiff_t stride,
716                         const uint8_t *left, const uint8_t *top)
717{
718    pixel *dst = (pixel *) _dst;
719    pixel4 val = PIXEL_SPLAT_X4((128 << (BIT_DEPTH - 8)) + 1);
720    int y;
721
722    stride /= sizeof(pixel);
723    for (y = 0; y < 8; y++) {
724        AV_WN4PA(dst + 0, val);
725        AV_WN4PA(dst + 4, val);
726        dst += stride;
727    }
728}
729
730static void dc_129_16x16_c(uint8_t *_dst, ptrdiff_t stride,
731                           const uint8_t *left, const uint8_t *top)
732{
733    pixel *dst = (pixel *) _dst;
734    pixel4 val = PIXEL_SPLAT_X4((128 << (BIT_DEPTH - 8)) + 1);
735    int y;
736
737    stride /= sizeof(pixel);
738    for (y = 0; y < 16; y++) {
739        AV_WN4PA(dst +  0, val);
740        AV_WN4PA(dst +  4, val);
741        AV_WN4PA(dst +  8, val);
742        AV_WN4PA(dst + 12, val);
743        dst += stride;
744    }
745}
746
747static void dc_129_32x32_c(uint8_t *_dst, ptrdiff_t stride,
748                           const uint8_t *left, const uint8_t *top)
749{
750    pixel *dst = (pixel *) _dst;
751    pixel4 val = PIXEL_SPLAT_X4((128 << (BIT_DEPTH - 8)) + 1);
752    int y;
753
754    stride /= sizeof(pixel);
755    for (y = 0; y < 32; y++) {
756        AV_WN4PA(dst +  0, val);
757        AV_WN4PA(dst +  4, val);
758        AV_WN4PA(dst +  8, val);
759        AV_WN4PA(dst + 12, val);
760        AV_WN4PA(dst + 16, val);
761        AV_WN4PA(dst + 20, val);
762        AV_WN4PA(dst + 24, val);
763        AV_WN4PA(dst + 28, val);
764        dst += stride;
765    }
766}
767
768#if BIT_DEPTH != 12
769
770#if BIT_DEPTH == 8
771#define memset_bpc memset
772#else
773static inline void memset_bpc(uint16_t *dst, int val, int len) {
774    int n;
775    for (n = 0; n < len; n++) {
776        dst[n] = val;
777    }
778}
779#endif
780
781#define DST(x, y) dst[(x) + (y) * stride]
782
783static void diag_downleft_4x4_c(uint8_t *_dst, ptrdiff_t stride,
784                                const uint8_t *left, const uint8_t *_top)
785{
786    pixel *dst = (pixel *) _dst;
787    const pixel *top = (const pixel *) _top;
788    int a0 = top[0], a1 = top[1], a2 = top[2], a3 = top[3],
789        a4 = top[4], a5 = top[5], a6 = top[6], a7 = top[7];
790
791    stride /= sizeof(pixel);
792    DST(0,0) = (a0 + a1 * 2 + a2 + 2) >> 2;
793    DST(1,0) = DST(0,1) = (a1 + a2 * 2 + a3 + 2) >> 2;
794    DST(2,0) = DST(1,1) = DST(0,2) = (a2 + a3 * 2 + a4 + 2) >> 2;
795    DST(3,0) = DST(2,1) = DST(1,2) = DST(0,3) = (a3 + a4 * 2 + a5 + 2) >> 2;
796    DST(3,1) = DST(2,2) = DST(1,3) = (a4 + a5 * 2 + a6 + 2) >> 2;
797    DST(3,2) = DST(2,3) = (a5 + a6 * 2 + a7 + 2) >> 2;
798    DST(3,3) = a7;  // note: this is different from vp8 and such
799}
800
801#define def_diag_downleft(size) \
802static void diag_downleft_##size##x##size##_c(uint8_t *_dst, ptrdiff_t stride, \
803                                              const uint8_t *left, const uint8_t *_top) \
804{ \
805    pixel *dst = (pixel *) _dst; \
806    const pixel *top = (const pixel *) _top; \
807    int i, j; \
808    pixel v[size - 1]; \
809\
810    stride /= sizeof(pixel); \
811    for (i = 0; i < size - 2; i++) \
812        v[i] = (top[i] + top[i + 1] * 2 + top[i + 2] + 2) >> 2; \
813    v[size - 2] = (top[size - 2] + top[size - 1] * 3 + 2) >> 2; \
814\
815    for (j = 0; j < size; j++) { \
816        memcpy(dst + j*stride, v + j, (size - 1 - j) * sizeof(pixel)); \
817        memset_bpc(dst + j*stride + size - 1 - j, top[size - 1], j + 1); \
818    } \
819}
820
821def_diag_downleft(8)
822def_diag_downleft(16)
823def_diag_downleft(32)
824
825static void diag_downright_4x4_c(uint8_t *_dst, ptrdiff_t stride,
826                                 const uint8_t *_left, const uint8_t *_top)
827{
828    pixel *dst = (pixel *) _dst;
829    const pixel *top = (const pixel *) _top;
830    const pixel *left = (const pixel *) _left;
831    int tl = top[-1], a0 = top[0], a1 = top[1], a2 = top[2], a3 = top[3],
832        l0 = left[3], l1 = left[2], l2 = left[1], l3 = left[0];
833
834    stride /= sizeof(pixel);
835    DST(0,3) = (l1 + l2 * 2 + l3 + 2) >> 2;
836    DST(0,2) = DST(1,3) = (l0 + l1 * 2 + l2 + 2) >> 2;
837    DST(0,1) = DST(1,2) = DST(2,3) = (tl + l0 * 2 + l1 + 2) >> 2;
838    DST(0,0) = DST(1,1) = DST(2,2) = DST(3,3) = (l0 + tl * 2 + a0 + 2) >> 2;
839    DST(1,0) = DST(2,1) = DST(3,2) = (tl + a0 * 2 + a1 + 2) >> 2;
840    DST(2,0) = DST(3,1) = (a0 + a1 * 2 + a2 + 2) >> 2;
841    DST(3,0) = (a1 + a2 * 2 + a3 + 2) >> 2;
842}
843
844#define def_diag_downright(size) \
845static void diag_downright_##size##x##size##_c(uint8_t *_dst, ptrdiff_t stride, \
846                                               const uint8_t *_left, const uint8_t *_top) \
847{ \
848    pixel *dst = (pixel *) _dst; \
849    const pixel *top = (const pixel *) _top; \
850    const pixel *left = (const pixel *) _left; \
851    int i, j; \
852    pixel v[size + size - 1]; \
853\
854    stride /= sizeof(pixel); \
855    for (i = 0; i < size - 2; i++) { \
856        v[i           ] = (left[i] + left[i + 1] * 2 + left[i + 2] + 2) >> 2; \
857        v[size + 1 + i] = (top[i]  + top[i + 1]  * 2 + top[i + 2]  + 2) >> 2; \
858    } \
859    v[size - 2] = (left[size - 2] + left[size - 1] * 2 + top[-1] + 2) >> 2; \
860    v[size - 1] = (left[size - 1] + top[-1] * 2 + top[ 0] + 2) >> 2; \
861    v[size    ] = (top[-1] + top[0]  * 2 + top[ 1] + 2) >> 2; \
862\
863    for (j = 0; j < size; j++) \
864        memcpy(dst + j*stride, v + size - 1 - j, size * sizeof(pixel)); \
865}
866
867def_diag_downright(8)
868def_diag_downright(16)
869def_diag_downright(32)
870
871static void vert_right_4x4_c(uint8_t *_dst, ptrdiff_t stride,
872                             const uint8_t *_left, const uint8_t *_top)
873{
874    pixel *dst = (pixel *) _dst;
875    const pixel *top = (const pixel *) _top;
876    const pixel *left = (const pixel *) _left;
877    int tl = top[-1], a0 = top[0], a1 = top[1], a2 = top[2], a3 = top[3],
878        l0 = left[3], l1 = left[2], l2 = left[1];
879
880    stride /= sizeof(pixel);
881    DST(0,3) = (l0 + l1 * 2 + l2 + 2) >> 2;
882    DST(0,2) = (tl + l0 * 2 + l1 + 2) >> 2;
883    DST(0,0) = DST(1,2) = (tl + a0 + 1) >> 1;
884    DST(0,1) = DST(1,3) = (l0 + tl * 2 + a0 + 2) >> 2;
885    DST(1,0) = DST(2,2) = (a0 + a1 + 1) >> 1;
886    DST(1,1) = DST(2,3) = (tl + a0 * 2 + a1 + 2) >> 2;
887    DST(2,0) = DST(3,2) = (a1 + a2 + 1) >> 1;
888    DST(2,1) = DST(3,3) = (a0 + a1 * 2 + a2 + 2) >> 2;
889    DST(3,0) = (a2 + a3 + 1) >> 1;
890    DST(3,1) = (a1 + a2 * 2 + a3 + 2) >> 2;
891}
892
893#define def_vert_right(size) \
894static void vert_right_##size##x##size##_c(uint8_t *_dst, ptrdiff_t stride, \
895                                           const uint8_t *_left, const uint8_t *_top) \
896{ \
897    pixel *dst = (pixel *) _dst; \
898    const pixel *top = (const pixel *) _top; \
899    const pixel *left = (const pixel *) _left; \
900    int i, j; \
901    pixel ve[size + size/2 - 1], vo[size + size/2 - 1]; \
902\
903    stride /= sizeof(pixel); \
904    for (i = 0; i < size/2 - 2; i++) { \
905        vo[i] = (left[i*2 + 3] + left[i*2 + 2] * 2 + left[i*2 + 1] + 2) >> 2; \
906        ve[i] = (left[i*2 + 4] + left[i*2 + 3] * 2 + left[i*2 + 2] + 2) >> 2; \
907    } \
908    vo[size/2 - 2] = (left[size - 1] + left[size - 2] * 2 + left[size - 3] + 2) >> 2; \
909    ve[size/2 - 2] = (top[-1] + left[size - 1] * 2 + left[size - 2] + 2) >> 2; \
910\
911    ve[size/2 - 1] = (top[-1] + top[0] + 1) >> 1; \
912    vo[size/2 - 1] = (left[size - 1] + top[-1] * 2 + top[0] + 2) >> 2; \
913    for (i = 0; i < size - 1; i++) { \
914        ve[size/2 + i] = (top[i] + top[i + 1] + 1) >> 1; \
915        vo[size/2 + i] = (top[i - 1] + top[i] * 2 + top[i + 1] + 2) >> 2; \
916    } \
917\
918    for (j = 0; j < size / 2; j++) { \
919        memcpy(dst +  j*2     *stride, ve + size/2 - 1 - j, size * sizeof(pixel)); \
920        memcpy(dst + (j*2 + 1)*stride, vo + size/2 - 1 - j, size * sizeof(pixel)); \
921    } \
922}
923
924def_vert_right(8)
925def_vert_right(16)
926def_vert_right(32)
927
928static void hor_down_4x4_c(uint8_t *_dst, ptrdiff_t stride,
929                           const uint8_t *_left, const uint8_t *_top)
930{
931    pixel *dst = (pixel *) _dst;
932    const pixel *top = (const pixel *) _top;
933    const pixel *left = (const pixel *) _left;
934    int l0 = left[3], l1 = left[2], l2 = left[1], l3 = left[0],
935        tl = top[-1], a0 = top[0], a1 = top[1], a2 = top[2];
936
937    stride /= sizeof(pixel);
938    DST(2,0) = (tl + a0 * 2 + a1 + 2) >> 2;
939    DST(3,0) = (a0 + a1 * 2 + a2 + 2) >> 2;
940    DST(0,0) = DST(2,1) = (tl + l0 + 1) >> 1;
941    DST(1,0) = DST(3,1) = (a0 + tl * 2 + l0 + 2) >> 2;
942    DST(0,1) = DST(2,2) = (l0 + l1 + 1) >> 1;
943    DST(1,1) = DST(3,2) = (tl + l0 * 2 + l1 + 2) >> 2;
944    DST(0,2) = DST(2,3) = (l1 + l2 + 1) >> 1;
945    DST(1,2) = DST(3,3) = (l0 + l1 * 2 + l2 + 2) >> 2;
946    DST(0,3) = (l2 + l3 + 1) >> 1;
947    DST(1,3) = (l1 + l2 * 2 + l3 + 2) >> 2;
948}
949
950#define def_hor_down(size) \
951static void hor_down_##size##x##size##_c(uint8_t *_dst, ptrdiff_t stride, \
952                                         const uint8_t *_left, const uint8_t *_top) \
953{ \
954    pixel *dst = (pixel *) _dst; \
955    const pixel *top = (const pixel *) _top; \
956    const pixel *left = (const pixel *) _left; \
957    int i, j; \
958    pixel v[size * 3 - 2]; \
959\
960    stride /= sizeof(pixel); \
961    for (i = 0; i < size - 2; i++) { \
962        v[i*2       ] = (left[i + 1] + left[i + 0] + 1) >> 1; \
963        v[i*2    + 1] = (left[i + 2] + left[i + 1] * 2 + left[i + 0] + 2) >> 2; \
964        v[size*2 + i] = (top[i - 1] + top[i] * 2 + top[i + 1] + 2) >> 2; \
965    } \
966    v[size*2 - 2] = (top[-1] + left[size - 1] + 1) >> 1; \
967    v[size*2 - 4] = (left[size - 1] + left[size - 2] + 1) >> 1; \
968    v[size*2 - 1] = (top[0]  + top[-1] * 2 + left[size - 1] + 2) >> 2; \
969    v[size*2 - 3] = (top[-1] + left[size - 1] * 2 + left[size - 2] + 2) >> 2; \
970\
971    for (j = 0; j < size; j++) \
972        memcpy(dst + j*stride, v + size*2 - 2 - j*2, size * sizeof(pixel)); \
973}
974
975def_hor_down(8)
976def_hor_down(16)
977def_hor_down(32)
978
979static void vert_left_4x4_c(uint8_t *_dst, ptrdiff_t stride,
980                            const uint8_t *left, const uint8_t *_top)
981{
982    pixel *dst = (pixel *) _dst;
983    const pixel *top = (const pixel *) _top;
984    int a0 = top[0], a1 = top[1], a2 = top[2], a3 = top[3],
985        a4 = top[4], a5 = top[5], a6 = top[6];
986
987    stride /= sizeof(pixel);
988    DST(0,0) = (a0 + a1 + 1) >> 1;
989    DST(0,1) = (a0 + a1 * 2 + a2 + 2) >> 2;
990    DST(1,0) = DST(0,2) = (a1 + a2 + 1) >> 1;
991    DST(1,1) = DST(0,3) = (a1 + a2 * 2 + a3 + 2) >> 2;
992    DST(2,0) = DST(1,2) = (a2 + a3 + 1) >> 1;
993    DST(2,1) = DST(1,3) = (a2 + a3 * 2 + a4 + 2) >> 2;
994    DST(3,0) = DST(2,2) = (a3 + a4 + 1) >> 1;
995    DST(3,1) = DST(2,3) = (a3 + a4 * 2 + a5 + 2) >> 2;
996    DST(3,2) = (a4 + a5 + 1) >> 1;
997    DST(3,3) = (a4 + a5 * 2 + a6 + 2) >> 2;
998}
999
1000#define def_vert_left(size) \
1001static void vert_left_##size##x##size##_c(uint8_t *_dst, ptrdiff_t stride, \
1002                                          const uint8_t *left, const uint8_t *_top) \
1003{ \
1004    pixel *dst = (pixel *) _dst; \
1005    const pixel *top = (const pixel *) _top; \
1006    int i, j; \
1007    pixel ve[size - 1], vo[size - 1]; \
1008\
1009    stride /= sizeof(pixel); \
1010    for (i = 0; i < size - 2; i++) { \
1011        ve[i] = (top[i] + top[i + 1] + 1) >> 1; \
1012        vo[i] = (top[i] + top[i + 1] * 2 + top[i + 2] + 2) >> 2; \
1013    } \
1014    ve[size - 2] = (top[size - 2] + top[size - 1] + 1) >> 1; \
1015    vo[size - 2] = (top[size - 2] + top[size - 1] * 3 + 2) >> 2; \
1016\
1017    for (j = 0; j < size / 2; j++) { \
1018        memcpy(dst +  j*2      * stride, ve + j, (size - j - 1) * sizeof(pixel)); \
1019        memset_bpc(dst +  j*2      * stride + size - j - 1, top[size - 1], j + 1); \
1020        memcpy(dst + (j*2 + 1) * stride, vo + j, (size - j - 1) * sizeof(pixel)); \
1021        memset_bpc(dst + (j*2 + 1) * stride + size - j - 1, top[size - 1], j + 1); \
1022    } \
1023}
1024
1025def_vert_left(8)
1026def_vert_left(16)
1027def_vert_left(32)
1028
1029static void hor_up_4x4_c(uint8_t *_dst, ptrdiff_t stride,
1030                         const uint8_t *_left, const uint8_t *top)
1031{
1032    pixel *dst = (pixel *) _dst;
1033    const pixel *left = (const pixel *) _left;
1034    int l0 = left[0], l1 = left[1], l2 = left[2], l3 = left[3];
1035
1036    stride /= sizeof(pixel);
1037    DST(0,0) = (l0 + l1 + 1) >> 1;
1038    DST(1,0) = (l0 + l1 * 2 + l2 + 2) >> 2;
1039    DST(0,1) = DST(2,0) = (l1 + l2 + 1) >> 1;
1040    DST(1,1) = DST(3,0) = (l1 + l2 * 2 + l3 + 2) >> 2;
1041    DST(0,2) = DST(2,1) = (l2 + l3 + 1) >> 1;
1042    DST(1,2) = DST(3,1) = (l2 + l3 * 3 + 2) >> 2;
1043    DST(0,3) = DST(1,3) = DST(2,2) = DST(2,3) = DST(3,2) = DST(3,3) = l3;
1044}
1045
1046#define def_hor_up(size) \
1047static void hor_up_##size##x##size##_c(uint8_t *_dst, ptrdiff_t stride, \
1048                                       const uint8_t *_left, const uint8_t *top) \
1049{ \
1050    pixel *dst = (pixel *) _dst; \
1051    const pixel *left = (const pixel *) _left; \
1052    int i, j; \
1053    pixel v[size*2 - 2]; \
1054\
1055    stride /= sizeof(pixel); \
1056    for (i = 0; i < size - 2; i++) { \
1057        v[i*2    ] = (left[i] + left[i + 1] + 1) >> 1; \
1058        v[i*2 + 1] = (left[i] + left[i + 1] * 2 + left[i + 2] + 2) >> 2; \
1059    } \
1060    v[size*2 - 4] = (left[size - 2] + left[size - 1] + 1) >> 1; \
1061    v[size*2 - 3] = (left[size - 2] + left[size - 1] * 3 + 2) >> 2; \
1062\
1063    for (j = 0; j < size / 2; j++) \
1064        memcpy(dst + j*stride, v + j*2, size * sizeof(pixel)); \
1065    for (j = size / 2; j < size; j++) { \
1066        memcpy(dst + j*stride, v + j*2, (size*2 - 2 - j*2) * sizeof(pixel)); \
1067        memset_bpc(dst + j*stride + size*2 - 2 - j*2, left[size - 1], \
1068                   2 + j*2 - size); \
1069    } \
1070}
1071
1072def_hor_up(8)
1073def_hor_up(16)
1074def_hor_up(32)
1075
1076#undef DST
1077
1078#endif /* BIT_DEPTH != 12 */
1079
1080#if BIT_DEPTH != 8
1081void ff_vp9dsp_intrapred_init_10(VP9DSPContext *dsp);
1082#endif
1083#if BIT_DEPTH != 10
1084static
1085#endif
1086av_cold void FUNC(ff_vp9dsp_intrapred_init)(VP9DSPContext *dsp)
1087{
1088#define init_intra_pred_bd_aware(tx, sz) \
1089    dsp->intra_pred[tx][TM_VP8_PRED]          = tm_##sz##_c; \
1090    dsp->intra_pred[tx][DC_128_PRED]          = dc_128_##sz##_c; \
1091    dsp->intra_pred[tx][DC_127_PRED]          = dc_127_##sz##_c; \
1092    dsp->intra_pred[tx][DC_129_PRED]          = dc_129_##sz##_c
1093
1094#if BIT_DEPTH == 12
1095    ff_vp9dsp_intrapred_init_10(dsp);
1096#define init_intra_pred(tx, sz) \
1097    init_intra_pred_bd_aware(tx, sz)
1098#else
1099    #define init_intra_pred(tx, sz) \
1100    dsp->intra_pred[tx][VERT_PRED]            = vert_##sz##_c; \
1101    dsp->intra_pred[tx][HOR_PRED]             = hor_##sz##_c; \
1102    dsp->intra_pred[tx][DC_PRED]              = dc_##sz##_c; \
1103    dsp->intra_pred[tx][DIAG_DOWN_LEFT_PRED]  = diag_downleft_##sz##_c; \
1104    dsp->intra_pred[tx][DIAG_DOWN_RIGHT_PRED] = diag_downright_##sz##_c; \
1105    dsp->intra_pred[tx][VERT_RIGHT_PRED]      = vert_right_##sz##_c; \
1106    dsp->intra_pred[tx][HOR_DOWN_PRED]        = hor_down_##sz##_c; \
1107    dsp->intra_pred[tx][VERT_LEFT_PRED]       = vert_left_##sz##_c; \
1108    dsp->intra_pred[tx][HOR_UP_PRED]          = hor_up_##sz##_c; \
1109    dsp->intra_pred[tx][LEFT_DC_PRED]         = dc_left_##sz##_c; \
1110    dsp->intra_pred[tx][TOP_DC_PRED]          = dc_top_##sz##_c; \
1111    init_intra_pred_bd_aware(tx, sz)
1112#endif
1113
1114    init_intra_pred(TX_4X4,   4x4);
1115    init_intra_pred(TX_8X8,   8x8);
1116    init_intra_pred(TX_16X16, 16x16);
1117    init_intra_pred(TX_32X32, 32x32);
1118
1119#undef init_intra_pred
1120#undef init_intra_pred_bd_aware
1121}
1122
1123#define itxfm_wrapper(type_a, type_b, sz, bits, has_dconly) \
1124static void type_a##_##type_b##_##sz##x##sz##_add_c(uint8_t *_dst, \
1125                                                    ptrdiff_t stride, \
1126                                                    int16_t *_block, int eob) \
1127{ \
1128    int i, j; \
1129    pixel *dst = (pixel *) _dst; \
1130    dctcoef *block = (dctcoef *) _block, tmp[sz * sz], out[sz]; \
1131\
1132    stride /= sizeof(pixel); \
1133    if (has_dconly && eob == 1) { \
1134        const int t  = ((((dctint) block[0] * 11585 + (1 << 13)) >> 14) \
1135                                            * 11585 + (1 << 13)) >> 14; \
1136        block[0] = 0; \
1137        for (i = 0; i < sz; i++) { \
1138            for (j = 0; j < sz; j++) \
1139                dst[j * stride] = av_clip_pixel(dst[j * stride] + \
1140                                                (bits ? \
1141                                                 (int)(t + (1U << (bits - 1))) >> bits : \
1142                                                 t)); \
1143            dst++; \
1144        } \
1145        return; \
1146    } \
1147\
1148    for (i = 0; i < sz; i++) \
1149        type_a##sz##_1d(block + i, sz, tmp + i * sz, 0); \
1150    memset(block, 0, sz * sz * sizeof(*block)); \
1151    for (i = 0; i < sz; i++) { \
1152        type_b##sz##_1d(tmp + i, sz, out, 1); \
1153        for (j = 0; j < sz; j++) \
1154            dst[j * stride] = av_clip_pixel(dst[j * stride] + \
1155                                            (bits ? \
1156                                             (int)(out[j] + (1U << (bits - 1))) >> bits : \
1157                                             out[j])); \
1158        dst++; \
1159    } \
1160}
1161
1162#define itxfm_wrap(sz, bits) \
1163itxfm_wrapper(idct,  idct,  sz, bits, 1) \
1164itxfm_wrapper(iadst, idct,  sz, bits, 0) \
1165itxfm_wrapper(idct,  iadst, sz, bits, 0) \
1166itxfm_wrapper(iadst, iadst, sz, bits, 0)
1167
1168#define IN(x) ((dctint) in[(x) * stride])
1169
1170static av_always_inline void idct4_1d(const dctcoef *in, ptrdiff_t stride,
1171                                      dctcoef *out, int pass)
1172{
1173    dctint t0, t1, t2, t3;
1174
1175    t0 = ((IN(0) + IN(2)) * 11585 + (1 << 13)) >> 14;
1176    t1 = ((IN(0) - IN(2)) * 11585 + (1 << 13)) >> 14;
1177    t2 = (IN(1) *  6270 - IN(3) * 15137 + (1 << 13)) >> 14;
1178    t3 = (IN(1) * 15137 + IN(3) *  6270 + (1 << 13)) >> 14;
1179
1180    out[0] = t0 + t3;
1181    out[1] = t1 + t2;
1182    out[2] = t1 - t2;
1183    out[3] = t0 - t3;
1184}
1185
1186static av_always_inline void iadst4_1d(const dctcoef *in, ptrdiff_t stride,
1187                                       dctcoef *out, int pass)
1188{
1189    dctint t0, t1, t2, t3;
1190
1191    t0 =  5283 * IN(0) + 15212 * IN(2) +  9929 * IN(3);
1192    t1 =  9929 * IN(0) -  5283 * IN(2) - 15212 * IN(3);
1193    t2 = 13377 * (IN(0) - IN(2) + IN(3));
1194    t3 = 13377 * IN(1);
1195
1196    out[0] = (t0 + t3      + (1 << 13)) >> 14;
1197    out[1] = (t1 + t3      + (1 << 13)) >> 14;
1198    out[2] = (t2           + (1 << 13)) >> 14;
1199    out[3] = (t0 + t1 - t3 + (1 << 13)) >> 14;
1200}
1201
1202itxfm_wrap(4, 4)
1203
1204static av_always_inline void idct8_1d(const dctcoef *in, ptrdiff_t stride,
1205                                      dctcoef *out, int pass)
1206{
1207    dctint t0, t0a, t1, t1a, t2, t2a, t3, t3a, t4, t4a, t5, t5a, t6, t6a, t7, t7a;
1208
1209    t0a = ((IN(0) + IN(4)) * 11585 + (1 << 13)) >> 14;
1210    t1a = ((IN(0) - IN(4)) * 11585 + (1 << 13)) >> 14;
1211    t2a = (IN(2) *  6270 - IN(6) * 15137 + (1 << 13)) >> 14;
1212    t3a = (IN(2) * 15137 + IN(6) *  6270 + (1 << 13)) >> 14;
1213    t4a = (IN(1) *  3196 - IN(7) * 16069 + (1 << 13)) >> 14;
1214    t5a = (IN(5) * 13623 - IN(3) *  9102 + (1 << 13)) >> 14;
1215    t6a = (IN(5) *  9102 + IN(3) * 13623 + (1 << 13)) >> 14;
1216    t7a = (IN(1) * 16069 + IN(7) *  3196 + (1 << 13)) >> 14;
1217
1218    t0  = t0a + t3a;
1219    t1  = t1a + t2a;
1220    t2  = t1a - t2a;
1221    t3  = t0a - t3a;
1222    t4  = t4a + t5a;
1223    t5a = t4a - t5a;
1224    t7  = t7a + t6a;
1225    t6a = t7a - t6a;
1226
1227    t5  = ((t6a - t5a) * 11585 + (1 << 13)) >> 14;
1228    t6  = ((t6a + t5a) * 11585 + (1 << 13)) >> 14;
1229
1230    out[0] = t0 + t7;
1231    out[1] = t1 + t6;
1232    out[2] = t2 + t5;
1233    out[3] = t3 + t4;
1234    out[4] = t3 - t4;
1235    out[5] = t2 - t5;
1236    out[6] = t1 - t6;
1237    out[7] = t0 - t7;
1238}
1239
1240static av_always_inline void iadst8_1d(const dctcoef *in, ptrdiff_t stride,
1241                                       dctcoef *out, int pass)
1242{
1243    dctint t0, t0a, t1, t1a, t2, t2a, t3, t3a, t4, t4a, t5, t5a, t6, t6a, t7, t7a;
1244
1245    t0a = 16305 * IN(7) +  1606 * IN(0);
1246    t1a =  1606 * IN(7) - 16305 * IN(0);
1247    t2a = 14449 * IN(5) +  7723 * IN(2);
1248    t3a =  7723 * IN(5) - 14449 * IN(2);
1249    t4a = 10394 * IN(3) + 12665 * IN(4);
1250    t5a = 12665 * IN(3) - 10394 * IN(4);
1251    t6a =  4756 * IN(1) + 15679 * IN(6);
1252    t7a = 15679 * IN(1) -  4756 * IN(6);
1253
1254    t0 = (t0a + t4a + (1 << 13)) >> 14;
1255    t1 = (t1a + t5a + (1 << 13)) >> 14;
1256    t2 = (t2a + t6a + (1 << 13)) >> 14;
1257    t3 = (t3a + t7a + (1 << 13)) >> 14;
1258    t4 = (t0a - t4a + (1 << 13)) >> 14;
1259    t5 = (t1a - t5a + (1 << 13)) >> 14;
1260    t6 = (t2a - t6a + (1 << 13)) >> 14;
1261    t7 = (t3a - t7a + (1 << 13)) >> 14;
1262
1263    t4a = 15137U * t4 +  6270U * t5;
1264    t5a =  6270U * t4 - 15137U * t5;
1265    t6a = 15137U * t7 -  6270U * t6;
1266    t7a =  6270U * t7 + 15137U * t6;
1267
1268    out[0] =   t0 + t2;
1269    out[7] = -(t1 + t3);
1270    t2     =   t0 - t2;
1271    t3     =   t1 - t3;
1272
1273    out[1] = -((dctint)((1U << 13) + t4a + t6a) >> 14);
1274    out[6] =   (dctint)((1U << 13) + t5a + t7a) >> 14;
1275    t6     =   (dctint)((1U << 13) + t4a - t6a) >> 14;
1276    t7     =   (dctint)((1U << 13) + t5a - t7a) >> 14;
1277
1278    out[3] = -((dctint)((t2 + t3) * 11585U + (1 << 13)) >> 14);
1279    out[4] =   (dctint)((t2 - t3) * 11585U + (1 << 13)) >> 14;
1280    out[2] =   (dctint)((t6 + t7) * 11585U + (1 << 13)) >> 14;
1281    out[5] = -((dctint)((t6 - t7) * 11585U + (1 << 13)) >> 14);
1282}
1283
1284itxfm_wrap(8, 5)
1285
1286static av_always_inline void idct16_1d(const dctcoef *in, ptrdiff_t stride,
1287                                       dctcoef *out, int pass)
1288{
1289    dctint t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15;
1290    dctint t0a, t1a, t2a, t3a, t4a, t5a, t6a, t7a;
1291    dctint t8a, t9a, t10a, t11a, t12a, t13a, t14a, t15a;
1292
1293    t0a  = (dctint)((IN(0) + IN(8)) * 11585U + (1 << 13)) >> 14;
1294    t1a  = (dctint)((IN(0) - IN(8)) * 11585U + (1 << 13)) >> 14;
1295    t2a  = (dctint)(IN(4)  *  6270U - IN(12) * 15137U + (1 << 13)) >> 14;
1296    t3a  = (dctint)(IN(4)  * 15137U + IN(12) *  6270U + (1 << 13)) >> 14;
1297    t4a  = (dctint)(IN(2)  *  3196U - IN(14) * 16069U + (1 << 13)) >> 14;
1298    t7a  = (dctint)(IN(2)  * 16069U + IN(14) *  3196U + (1 << 13)) >> 14;
1299    t5a  = (dctint)(IN(10) * 13623U - IN(6)  *  9102U + (1 << 13)) >> 14;
1300    t6a  = (dctint)(IN(10) *  9102U + IN(6)  * 13623U + (1 << 13)) >> 14;
1301    t8a  = (dctint)(IN(1)  *  1606U - IN(15) * 16305U + (1 << 13)) >> 14;
1302    t15a = (dctint)(IN(1)  * 16305U + IN(15) *  1606U + (1 << 13)) >> 14;
1303    t9a  = (dctint)(IN(9)  * 12665U - IN(7)  * 10394U + (1 << 13)) >> 14;
1304    t14a = (dctint)(IN(9)  * 10394U + IN(7)  * 12665U + (1 << 13)) >> 14;
1305    t10a = (dctint)(IN(5)  *  7723U - IN(11) * 14449U + (1 << 13)) >> 14;
1306    t13a = (dctint)(IN(5)  * 14449U + IN(11) *  7723U + (1 << 13)) >> 14;
1307    t11a = (dctint)(IN(13) * 15679U - IN(3)  *  4756U + (1 << 13)) >> 14;
1308    t12a = (dctint)(IN(13) *  4756U + IN(3)  * 15679U + (1 << 13)) >> 14;
1309
1310    t0  = t0a  + t3a;
1311    t1  = t1a  + t2a;
1312    t2  = t1a  - t2a;
1313    t3  = t0a  - t3a;
1314    t4  = t4a  + t5a;
1315    t5  = t4a  - t5a;
1316    t6  = t7a  - t6a;
1317    t7  = t7a  + t6a;
1318    t8  = t8a  + t9a;
1319    t9  = t8a  - t9a;
1320    t10 = t11a - t10a;
1321    t11 = t11a + t10a;
1322    t12 = t12a + t13a;
1323    t13 = t12a - t13a;
1324    t14 = t15a - t14a;
1325    t15 = t15a + t14a;
1326
1327    t5a  = (dctint)((t6 - t5) * 11585U + (1 << 13)) >> 14;
1328    t6a  = (dctint)((t6 + t5) * 11585U + (1 << 13)) >> 14;
1329    t9a  = (dctint)(  t14 *  6270U - t9  * 15137U  + (1 << 13)) >> 14;
1330    t14a = (dctint)(  t14 * 15137U + t9  *  6270U  + (1 << 13)) >> 14;
1331    t10a = (dctint)(-(t13 * 15137U + t10 *  6270U) + (1 << 13)) >> 14;
1332    t13a = (dctint)(  t13 *  6270U - t10 * 15137U  + (1 << 13)) >> 14;
1333
1334    t0a  = t0   + t7;
1335    t1a  = t1   + t6a;
1336    t2a  = t2   + t5a;
1337    t3a  = t3   + t4;
1338    t4   = t3   - t4;
1339    t5   = t2   - t5a;
1340    t6   = t1   - t6a;
1341    t7   = t0   - t7;
1342    t8a  = t8   + t11;
1343    t9   = t9a  + t10a;
1344    t10  = t9a  - t10a;
1345    t11a = t8   - t11;
1346    t12a = t15  - t12;
1347    t13  = t14a - t13a;
1348    t14  = t14a + t13a;
1349    t15a = t15  + t12;
1350
1351    t10a = (dctint)((t13  - t10)  * 11585U + (1 << 13)) >> 14;
1352    t13a = (dctint)((t13  + t10)  * 11585U + (1 << 13)) >> 14;
1353    t11  = (dctint)((t12a - t11a) * 11585U + (1 << 13)) >> 14;
1354    t12  = (dctint)((t12a + t11a) * 11585U + (1 << 13)) >> 14;
1355
1356    out[ 0] = t0a + t15a;
1357    out[ 1] = t1a + t14;
1358    out[ 2] = t2a + t13a;
1359    out[ 3] = t3a + t12;
1360    out[ 4] = t4  + t11;
1361    out[ 5] = t5  + t10a;
1362    out[ 6] = t6  + t9;
1363    out[ 7] = t7  + t8a;
1364    out[ 8] = t7  - t8a;
1365    out[ 9] = t6  - t9;
1366    out[10] = t5  - t10a;
1367    out[11] = t4  - t11;
1368    out[12] = t3a - t12;
1369    out[13] = t2a - t13a;
1370    out[14] = t1a - t14;
1371    out[15] = t0a - t15a;
1372}
1373
1374static av_always_inline void iadst16_1d(const dctcoef *in, ptrdiff_t stride,
1375                                        dctcoef *out, int pass)
1376{
1377    dctint t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15;
1378    dctint t0a, t1a, t2a, t3a, t4a, t5a, t6a, t7a;
1379    dctint t8a, t9a, t10a, t11a, t12a, t13a, t14a, t15a;
1380
1381    t0  = IN(15) * 16364U + IN(0)  *   804U;
1382    t1  = IN(15) *   804U - IN(0)  * 16364U;
1383    t2  = IN(13) * 15893U + IN(2)  *  3981U;
1384    t3  = IN(13) *  3981U - IN(2)  * 15893U;
1385    t4  = IN(11) * 14811U + IN(4)  *  7005U;
1386    t5  = IN(11) *  7005U - IN(4)  * 14811U;
1387    t6  = IN(9)  * 13160U + IN(6)  *  9760U;
1388    t7  = IN(9)  *  9760U - IN(6)  * 13160U;
1389    t8  = IN(7)  * 11003U + IN(8)  * 12140U;
1390    t9  = IN(7)  * 12140U - IN(8)  * 11003U;
1391    t10 = IN(5)  *  8423U + IN(10) * 14053U;
1392    t11 = IN(5)  * 14053U - IN(10) *  8423U;
1393    t12 = IN(3)  *  5520U + IN(12) * 15426U;
1394    t13 = IN(3)  * 15426U - IN(12) *  5520U;
1395    t14 = IN(1)  *  2404U + IN(14) * 16207U;
1396    t15 = IN(1)  * 16207U - IN(14) *  2404U;
1397
1398    t0a  = (dctint)((1U << 13) + t0 + t8 ) >> 14;
1399    t1a  = (dctint)((1U << 13) + t1 + t9 ) >> 14;
1400    t2a  = (dctint)((1U << 13) + t2 + t10) >> 14;
1401    t3a  = (dctint)((1U << 13) + t3 + t11) >> 14;
1402    t4a  = (dctint)((1U << 13) + t4 + t12) >> 14;
1403    t5a  = (dctint)((1U << 13) + t5 + t13) >> 14;
1404    t6a  = (dctint)((1U << 13) + t6 + t14) >> 14;
1405    t7a  = (dctint)((1U << 13) + t7 + t15) >> 14;
1406    t8a  = (dctint)((1U << 13) + t0 - t8 ) >> 14;
1407    t9a  = (dctint)((1U << 13) + t1 - t9 ) >> 14;
1408    t10a = (dctint)((1U << 13) + t2 - t10) >> 14;
1409    t11a = (dctint)((1U << 13) + t3 - t11) >> 14;
1410    t12a = (dctint)((1U << 13) + t4 - t12) >> 14;
1411    t13a = (dctint)((1U << 13) + t5 - t13) >> 14;
1412    t14a = (dctint)((1U << 13) + t6 - t14) >> 14;
1413    t15a = (dctint)((1U << 13) + t7 - t15) >> 14;
1414
1415    t8   = t8a  * 16069U + t9a  *  3196U;
1416    t9   = t8a  *  3196U - t9a  * 16069U;
1417    t10  = t10a *  9102U + t11a * 13623U;
1418    t11  = t10a * 13623U - t11a *  9102U;
1419    t12  = t13a * 16069U - t12a *  3196U;
1420    t13  = t13a *  3196U + t12a * 16069U;
1421    t14  = t15a *  9102U - t14a * 13623U;
1422    t15  = t15a * 13623U + t14a *  9102U;
1423
1424    t0   = t0a + t4a;
1425    t1   = t1a + t5a;
1426    t2   = t2a + t6a;
1427    t3   = t3a + t7a;
1428    t4   = t0a - t4a;
1429    t5   = t1a - t5a;
1430    t6   = t2a - t6a;
1431    t7   = t3a - t7a;
1432    t8a  = (dctint)((1U << 13) + t8  + t12) >> 14;
1433    t9a  = (dctint)((1U << 13) + t9  + t13) >> 14;
1434    t10a = (dctint)((1U << 13) + t10 + t14) >> 14;
1435    t11a = (dctint)((1U << 13) + t11 + t15) >> 14;
1436    t12a = (dctint)((1U << 13) + t8  - t12) >> 14;
1437    t13a = (dctint)((1U << 13) + t9  - t13) >> 14;
1438    t14a = (dctint)((1U << 13) + t10 - t14) >> 14;
1439    t15a = (dctint)((1U << 13) + t11 - t15) >> 14;
1440
1441    t4a  = t4 * 15137U + t5 *  6270U;
1442    t5a  = t4 *  6270U - t5 * 15137U;
1443    t6a  = t7 * 15137U - t6 *  6270U;
1444    t7a  = t7 *  6270U + t6 * 15137U;
1445    t12  = t12a * 15137U + t13a *  6270U;
1446    t13  = t12a *  6270U - t13a * 15137U;
1447    t14  = t15a * 15137U - t14a *  6270U;
1448    t15  = t15a *  6270U + t14a * 15137U;
1449
1450    out[ 0] =   t0 + t2;
1451    out[15] = -(t1 + t3);
1452    t2a     =   t0 - t2;
1453    t3a     =   t1 - t3;
1454    out[ 3] = -((dctint)((1U << 13) + t4a + t6a) >> 14);
1455    out[12] =   (dctint)((1U << 13) + t5a + t7a) >> 14;
1456    t6      =   (dctint)((1U << 13) + t4a - t6a) >> 14;
1457    t7      =   (dctint)((1U << 13) + t5a - t7a) >> 14;
1458    out[ 1] = -(t8a + t10a);
1459    out[14] =   t9a + t11a;
1460    t10     =   t8a - t10a;
1461    t11     =   t9a - t11a;
1462    out[ 2] =   (dctint)((1U << 13) + t12 + t14) >> 14;
1463    out[13] = -((dctint)((1U << 13) + t13 + t15) >> 14);
1464    t14a    =   (dctint)((1U << 13) + t12 - t14) >> 14;
1465    t15a    =   (dctint)((1U << 13) + t13 - t15) >> 14;
1466
1467    out[ 7] = (dctint)(-(t2a  + t3a)  * 11585U  + (1 << 13)) >> 14;
1468    out[ 8] = (dctint)( (t2a  - t3a)  * 11585U  + (1 << 13)) >> 14;
1469    out[ 4] = (dctint)( (t7   + t6)   * 11585U  + (1 << 13)) >> 14;
1470    out[11] = (dctint)( (t7   - t6)   * 11585U  + (1 << 13)) >> 14;
1471    out[ 6] = (dctint)( (t11  + t10)  * 11585U  + (1 << 13)) >> 14;
1472    out[ 9] = (dctint)( (t11  - t10)  * 11585U  + (1 << 13)) >> 14;
1473    out[ 5] = (dctint)(-(t14a + t15a) * 11585U  + (1 << 13)) >> 14;
1474    out[10] = (dctint)( (t14a - t15a) * 11585U  + (1 << 13)) >> 14;
1475}
1476
1477itxfm_wrap(16, 6)
1478
1479static av_always_inline void idct32_1d(const dctcoef *in, ptrdiff_t stride,
1480                                       dctcoef *out, int pass)
1481{
1482    dctint t0a  = (dctint)((IN(0) + IN(16)) * 11585U         + (1 << 13)) >> 14;
1483    dctint t1a  = (dctint)((IN(0) - IN(16)) * 11585U         + (1 << 13)) >> 14;
1484    dctint t2a  = (dctint)(IN( 8) *  6270U - IN(24) * 15137U + (1 << 13)) >> 14;
1485    dctint t3a  = (dctint)(IN( 8) * 15137U + IN(24) *  6270U + (1 << 13)) >> 14;
1486    dctint t4a  = (dctint)(IN( 4) *  3196U - IN(28) * 16069U + (1 << 13)) >> 14;
1487    dctint t7a  = (dctint)(IN( 4) * 16069U + IN(28) *  3196U + (1 << 13)) >> 14;
1488    dctint t5a  = (dctint)(IN(20) * 13623U - IN(12) *  9102U + (1 << 13)) >> 14;
1489    dctint t6a  = (dctint)(IN(20) *  9102U + IN(12) * 13623U + (1 << 13)) >> 14;
1490    dctint t8a  = (dctint)(IN( 2) *  1606U - IN(30) * 16305U + (1 << 13)) >> 14;
1491    dctint t15a = (dctint)(IN( 2) * 16305U + IN(30) *  1606U + (1 << 13)) >> 14;
1492    dctint t9a  = (dctint)(IN(18) * 12665U - IN(14) * 10394U + (1 << 13)) >> 14;
1493    dctint t14a = (dctint)(IN(18) * 10394U + IN(14) * 12665U + (1 << 13)) >> 14;
1494    dctint t10a = (dctint)(IN(10) *  7723U - IN(22) * 14449U + (1 << 13)) >> 14;
1495    dctint t13a = (dctint)(IN(10) * 14449U + IN(22) *  7723U + (1 << 13)) >> 14;
1496    dctint t11a = (dctint)(IN(26) * 15679U - IN( 6) *  4756U + (1 << 13)) >> 14;
1497    dctint t12a = (dctint)(IN(26) *  4756U + IN( 6) * 15679U + (1 << 13)) >> 14;
1498    dctint t16a = (dctint)(IN( 1) *   804U - IN(31) * 16364U + (1 << 13)) >> 14;
1499    dctint t31a = (dctint)(IN( 1) * 16364U + IN(31) *   804U + (1 << 13)) >> 14;
1500    dctint t17a = (dctint)(IN(17) * 12140U - IN(15) * 11003U + (1 << 13)) >> 14;
1501    dctint t30a = (dctint)(IN(17) * 11003U + IN(15) * 12140U + (1 << 13)) >> 14;
1502    dctint t18a = (dctint)(IN( 9) *  7005U - IN(23) * 14811U + (1 << 13)) >> 14;
1503    dctint t29a = (dctint)(IN( 9) * 14811U + IN(23) *  7005U + (1 << 13)) >> 14;
1504    dctint t19a = (dctint)(IN(25) * 15426U - IN( 7) *  5520U + (1 << 13)) >> 14;
1505    dctint t28a = (dctint)(IN(25) *  5520U + IN( 7) * 15426U + (1 << 13)) >> 14;
1506    dctint t20a = (dctint)(IN( 5) *  3981U - IN(27) * 15893U + (1 << 13)) >> 14;
1507    dctint t27a = (dctint)(IN( 5) * 15893U + IN(27) *  3981U + (1 << 13)) >> 14;
1508    dctint t21a = (dctint)(IN(21) * 14053U - IN(11) *  8423U + (1 << 13)) >> 14;
1509    dctint t26a = (dctint)(IN(21) *  8423U + IN(11) * 14053U + (1 << 13)) >> 14;
1510    dctint t22a = (dctint)(IN(13) *  9760U - IN(19) * 13160U + (1 << 13)) >> 14;
1511    dctint t25a = (dctint)(IN(13) * 13160U + IN(19) *  9760U + (1 << 13)) >> 14;
1512    dctint t23a = (dctint)(IN(29) * 16207U - IN( 3) *  2404U + (1 << 13)) >> 14;
1513    dctint t24a = (dctint)(IN(29) *  2404U + IN( 3) * 16207U + (1 << 13)) >> 14;
1514
1515    dctint t0  = t0a  + t3a;
1516    dctint t1  = t1a  + t2a;
1517    dctint t2  = t1a  - t2a;
1518    dctint t3  = t0a  - t3a;
1519    dctint t4  = t4a  + t5a;
1520    dctint t5  = t4a  - t5a;
1521    dctint t6  = t7a  - t6a;
1522    dctint t7  = t7a  + t6a;
1523    dctint t8  = t8a  + t9a;
1524    dctint t9  = t8a  - t9a;
1525    dctint t10 = t11a - t10a;
1526    dctint t11 = t11a + t10a;
1527    dctint t12 = t12a + t13a;
1528    dctint t13 = t12a - t13a;
1529    dctint t14 = t15a - t14a;
1530    dctint t15 = t15a + t14a;
1531    dctint t16 = t16a + t17a;
1532    dctint t17 = t16a - t17a;
1533    dctint t18 = t19a - t18a;
1534    dctint t19 = t19a + t18a;
1535    dctint t20 = t20a + t21a;
1536    dctint t21 = t20a - t21a;
1537    dctint t22 = t23a - t22a;
1538    dctint t23 = t23a + t22a;
1539    dctint t24 = t24a + t25a;
1540    dctint t25 = t24a - t25a;
1541    dctint t26 = t27a - t26a;
1542    dctint t27 = t27a + t26a;
1543    dctint t28 = t28a + t29a;
1544    dctint t29 = t28a - t29a;
1545    dctint t30 = t31a - t30a;
1546    dctint t31 = t31a + t30a;
1547
1548    t5a  = (dctint)((t6 - t5) * 11585U             + (1 << 13)) >> 14;
1549    t6a  = (dctint)((t6 + t5) * 11585U             + (1 << 13)) >> 14;
1550    t9a  = (dctint)(  t14 *  6270U - t9  * 15137U  + (1 << 13)) >> 14;
1551    t14a = (dctint)(  t14 * 15137U + t9  *  6270U  + (1 << 13)) >> 14;
1552    t10a = (dctint)(-(t13 * 15137U + t10 *  6270U) + (1 << 13)) >> 14;
1553    t13a = (dctint)(  t13 *  6270U - t10 * 15137U  + (1 << 13)) >> 14;
1554    t17a = (dctint)(  t30 *  3196U - t17 * 16069U  + (1 << 13)) >> 14;
1555    t30a = (dctint)(  t30 * 16069U + t17 *  3196U  + (1 << 13)) >> 14;
1556    t18a = (dctint)(-(t29 * 16069U + t18 *  3196U) + (1 << 13)) >> 14;
1557    t29a = (dctint)(  t29 *  3196U - t18 * 16069U  + (1 << 13)) >> 14;
1558    t21a = (dctint)(  t26 * 13623U - t21 *  9102U  + (1 << 13)) >> 14;
1559    t26a = (dctint)(  t26 *  9102U + t21 * 13623U  + (1 << 13)) >> 14;
1560    t22a = (dctint)(-(t25 *  9102U + t22 * 13623U) + (1 << 13)) >> 14;
1561    t25a = (dctint)(  t25 * 13623U - t22 *  9102U  + (1 << 13)) >> 14;
1562
1563    t0a  = t0   + t7;
1564    t1a  = t1   + t6a;
1565    t2a  = t2   + t5a;
1566    t3a  = t3   + t4;
1567    t4a  = t3   - t4;
1568    t5   = t2   - t5a;
1569    t6   = t1   - t6a;
1570    t7a  = t0   - t7;
1571    t8a  = t8   + t11;
1572    t9   = t9a  + t10a;
1573    t10  = t9a  - t10a;
1574    t11a = t8   - t11;
1575    t12a = t15  - t12;
1576    t13  = t14a - t13a;
1577    t14  = t14a + t13a;
1578    t15a = t15  + t12;
1579    t16a = t16  + t19;
1580    t17  = t17a + t18a;
1581    t18  = t17a - t18a;
1582    t19a = t16  - t19;
1583    t20a = t23  - t20;
1584    t21  = t22a - t21a;
1585    t22  = t22a + t21a;
1586    t23a = t23  + t20;
1587    t24a = t24  + t27;
1588    t25  = t25a + t26a;
1589    t26  = t25a - t26a;
1590    t27a = t24  - t27;
1591    t28a = t31  - t28;
1592    t29  = t30a - t29a;
1593    t30  = t30a + t29a;
1594    t31a = t31  + t28;
1595
1596    t10a = (dctint)((t13  - t10)  * 11585U           + (1 << 13)) >> 14;
1597    t13a = (dctint)((t13  + t10)  * 11585U           + (1 << 13)) >> 14;
1598    t11  = (dctint)((t12a - t11a) * 11585U           + (1 << 13)) >> 14;
1599    t12  = (dctint)((t12a + t11a) * 11585U           + (1 << 13)) >> 14;
1600    t18a = (dctint)(  t29  *  6270U - t18  * 15137U  + (1 << 13)) >> 14;
1601    t29a = (dctint)(  t29  * 15137U + t18  *  6270U  + (1 << 13)) >> 14;
1602    t19  = (dctint)(  t28a *  6270U - t19a * 15137U  + (1 << 13)) >> 14;
1603    t28  = (dctint)(  t28a * 15137U + t19a *  6270U  + (1 << 13)) >> 14;
1604    t20  = (dctint)(-(t27a * 15137U + t20a *  6270U) + (1 << 13)) >> 14;
1605    t27  = (dctint)(  t27a *  6270U - t20a * 15137U  + (1 << 13)) >> 14;
1606    t21a = (dctint)(-(t26  * 15137U + t21  *  6270U) + (1 << 13)) >> 14;
1607    t26a = (dctint)(  t26  *  6270U - t21  * 15137U  + (1 << 13)) >> 14;
1608
1609    t0   = t0a + t15a;
1610    t1   = t1a + t14;
1611    t2   = t2a + t13a;
1612    t3   = t3a + t12;
1613    t4   = t4a + t11;
1614    t5a  = t5  + t10a;
1615    t6a  = t6  + t9;
1616    t7   = t7a + t8a;
1617    t8   = t7a - t8a;
1618    t9a  = t6  - t9;
1619    t10  = t5  - t10a;
1620    t11a = t4a - t11;
1621    t12a = t3a - t12;
1622    t13  = t2a - t13a;
1623    t14a = t1a - t14;
1624    t15  = t0a - t15a;
1625    t16  = t16a + t23a;
1626    t17a = t17  + t22;
1627    t18  = t18a + t21a;
1628    t19a = t19  + t20;
1629    t20a = t19  - t20;
1630    t21  = t18a - t21a;
1631    t22a = t17  - t22;
1632    t23  = t16a - t23a;
1633    t24  = t31a - t24a;
1634    t25a = t30  - t25;
1635    t26  = t29a - t26a;
1636    t27a = t28  - t27;
1637    t28a = t28  + t27;
1638    t29  = t29a + t26a;
1639    t30a = t30  + t25;
1640    t31  = t31a + t24a;
1641
1642    t20  = (dctint)((t27a - t20a) * 11585U + (1 << 13)) >> 14;
1643    t27  = (dctint)((t27a + t20a) * 11585U + (1 << 13)) >> 14;
1644    t21a = (dctint)((t26  - t21 ) * 11585U + (1 << 13)) >> 14;
1645    t26a = (dctint)((t26  + t21 ) * 11585U + (1 << 13)) >> 14;
1646    t22  = (dctint)((t25a - t22a) * 11585U + (1 << 13)) >> 14;
1647    t25  = (dctint)((t25a + t22a) * 11585U + (1 << 13)) >> 14;
1648    t23a = (dctint)((t24  - t23 ) * 11585U + (1 << 13)) >> 14;
1649    t24a = (dctint)((t24  + t23 ) * 11585U + (1 << 13)) >> 14;
1650
1651    out[ 0] = t0   + t31;
1652    out[ 1] = t1   + t30a;
1653    out[ 2] = t2   + t29;
1654    out[ 3] = t3   + t28a;
1655    out[ 4] = t4   + t27;
1656    out[ 5] = t5a  + t26a;
1657    out[ 6] = t6a  + t25;
1658    out[ 7] = t7   + t24a;
1659    out[ 8] = t8   + t23a;
1660    out[ 9] = t9a  + t22;
1661    out[10] = t10  + t21a;
1662    out[11] = t11a + t20;
1663    out[12] = t12a + t19a;
1664    out[13] = t13  + t18;
1665    out[14] = t14a + t17a;
1666    out[15] = t15  + t16;
1667    out[16] = t15  - t16;
1668    out[17] = t14a - t17a;
1669    out[18] = t13  - t18;
1670    out[19] = t12a - t19a;
1671    out[20] = t11a - t20;
1672    out[21] = t10  - t21a;
1673    out[22] = t9a  - t22;
1674    out[23] = t8   - t23a;
1675    out[24] = t7   - t24a;
1676    out[25] = t6a  - t25;
1677    out[26] = t5a  - t26a;
1678    out[27] = t4   - t27;
1679    out[28] = t3   - t28a;
1680    out[29] = t2   - t29;
1681    out[30] = t1   - t30a;
1682    out[31] = t0   - t31;
1683}
1684
1685itxfm_wrapper(idct, idct, 32, 6, 1)
1686
1687static av_always_inline void iwht4_1d(const dctcoef *in, ptrdiff_t stride,
1688                                      dctcoef *out, int pass)
1689{
1690    int t0, t1, t2, t3, t4;
1691
1692    if (pass == 0) {
1693        t0 = IN(0) >> 2;
1694        t1 = IN(3) >> 2;
1695        t2 = IN(1) >> 2;
1696        t3 = IN(2) >> 2;
1697    } else {
1698        t0 = IN(0);
1699        t1 = IN(3);
1700        t2 = IN(1);
1701        t3 = IN(2);
1702    }
1703
1704    t0 += t2;
1705    t3 -= t1;
1706    t4 = (t0 - t3) >> 1;
1707    t1 = t4 - t1;
1708    t2 = t4 - t2;
1709    t0 -= t1;
1710    t3 += t2;
1711
1712    out[0] = t0;
1713    out[1] = t1;
1714    out[2] = t2;
1715    out[3] = t3;
1716}
1717
1718itxfm_wrapper(iwht, iwht, 4, 0, 0)
1719
1720#undef IN
1721#undef itxfm_wrapper
1722#undef itxfm_wrap
1723
1724static av_cold void vp9dsp_itxfm_init(VP9DSPContext *dsp)
1725{
1726#define init_itxfm(tx, sz) \
1727    dsp->itxfm_add[tx][DCT_DCT]   = idct_idct_##sz##_add_c; \
1728    dsp->itxfm_add[tx][DCT_ADST]  = iadst_idct_##sz##_add_c; \
1729    dsp->itxfm_add[tx][ADST_DCT]  = idct_iadst_##sz##_add_c; \
1730    dsp->itxfm_add[tx][ADST_ADST] = iadst_iadst_##sz##_add_c
1731
1732#define init_idct(tx, nm) \
1733    dsp->itxfm_add[tx][DCT_DCT]   = \
1734    dsp->itxfm_add[tx][ADST_DCT]  = \
1735    dsp->itxfm_add[tx][DCT_ADST]  = \
1736    dsp->itxfm_add[tx][ADST_ADST] = nm##_add_c
1737
1738    init_itxfm(TX_4X4,   4x4);
1739    init_itxfm(TX_8X8,   8x8);
1740    init_itxfm(TX_16X16, 16x16);
1741    init_idct(TX_32X32,  idct_idct_32x32);
1742    init_idct(4 /* lossless */, iwht_iwht_4x4);
1743
1744#undef init_itxfm
1745#undef init_idct
1746}
1747
1748static av_always_inline void loop_filter(pixel *dst, int E, int I, int H,
1749                                         ptrdiff_t stridea, ptrdiff_t strideb,
1750                                         int wd)
1751{
1752    int i, F = 1 << (BIT_DEPTH - 8);
1753
1754    E <<= (BIT_DEPTH - 8);
1755    I <<= (BIT_DEPTH - 8);
1756    H <<= (BIT_DEPTH - 8);
1757    for (i = 0; i < 8; i++, dst += stridea) {
1758        int p7, p6, p5, p4;
1759        int p3 = dst[strideb * -4], p2 = dst[strideb * -3];
1760        int p1 = dst[strideb * -2], p0 = dst[strideb * -1];
1761        int q0 = dst[strideb * +0], q1 = dst[strideb * +1];
1762        int q2 = dst[strideb * +2], q3 = dst[strideb * +3];
1763        int q4, q5, q6, q7;
1764        int fm = FFABS(p3 - p2) <= I && FFABS(p2 - p1) <= I &&
1765                 FFABS(p1 - p0) <= I && FFABS(q1 - q0) <= I &&
1766                 FFABS(q2 - q1) <= I && FFABS(q3 - q2) <= I &&
1767                 FFABS(p0 - q0) * 2 + (FFABS(p1 - q1) >> 1) <= E;
1768        int flat8out, flat8in;
1769
1770        if (!fm)
1771            continue;
1772
1773        if (wd >= 16) {
1774            p7 = dst[strideb * -8];
1775            p6 = dst[strideb * -7];
1776            p5 = dst[strideb * -6];
1777            p4 = dst[strideb * -5];
1778            q4 = dst[strideb * +4];
1779            q5 = dst[strideb * +5];
1780            q6 = dst[strideb * +6];
1781            q7 = dst[strideb * +7];
1782
1783            flat8out = FFABS(p7 - p0) <= F && FFABS(p6 - p0) <= F &&
1784                       FFABS(p5 - p0) <= F && FFABS(p4 - p0) <= F &&
1785                       FFABS(q4 - q0) <= F && FFABS(q5 - q0) <= F &&
1786                       FFABS(q6 - q0) <= F && FFABS(q7 - q0) <= F;
1787        }
1788
1789        if (wd >= 8)
1790            flat8in = FFABS(p3 - p0) <= F && FFABS(p2 - p0) <= F &&
1791                      FFABS(p1 - p0) <= F && FFABS(q1 - q0) <= F &&
1792                      FFABS(q2 - q0) <= F && FFABS(q3 - q0) <= F;
1793
1794        if (wd >= 16 && flat8out && flat8in) {
1795            dst[strideb * -7] = (p7 + p7 + p7 + p7 + p7 + p7 + p7 + p6 * 2 +
1796                                 p5 + p4 + p3 + p2 + p1 + p0 + q0 + 8) >> 4;
1797            dst[strideb * -6] = (p7 + p7 + p7 + p7 + p7 + p7 + p6 + p5 * 2 +
1798                                 p4 + p3 + p2 + p1 + p0 + q0 + q1 + 8) >> 4;
1799            dst[strideb * -5] = (p7 + p7 + p7 + p7 + p7 + p6 + p5 + p4 * 2 +
1800                                 p3 + p2 + p1 + p0 + q0 + q1 + q2 + 8) >> 4;
1801            dst[strideb * -4] = (p7 + p7 + p7 + p7 + p6 + p5 + p4 + p3 * 2 +
1802                                 p2 + p1 + p0 + q0 + q1 + q2 + q3 + 8) >> 4;
1803            dst[strideb * -3] = (p7 + p7 + p7 + p6 + p5 + p4 + p3 + p2 * 2 +
1804                                 p1 + p0 + q0 + q1 + q2 + q3 + q4 + 8) >> 4;
1805            dst[strideb * -2] = (p7 + p7 + p6 + p5 + p4 + p3 + p2 + p1 * 2 +
1806                                 p0 + q0 + q1 + q2 + q3 + q4 + q5 + 8) >> 4;
1807            dst[strideb * -1] = (p7 + p6 + p5 + p4 + p3 + p2 + p1 + p0 * 2 +
1808                                 q0 + q1 + q2 + q3 + q4 + q5 + q6 + 8) >> 4;
1809            dst[strideb * +0] = (p6 + p5 + p4 + p3 + p2 + p1 + p0 + q0 * 2 +
1810                                 q1 + q2 + q3 + q4 + q5 + q6 + q7 + 8) >> 4;
1811            dst[strideb * +1] = (p5 + p4 + p3 + p2 + p1 + p0 + q0 + q1 * 2 +
1812                                 q2 + q3 + q4 + q5 + q6 + q7 + q7 + 8) >> 4;
1813            dst[strideb * +2] = (p4 + p3 + p2 + p1 + p0 + q0 + q1 + q2 * 2 +
1814                                 q3 + q4 + q5 + q6 + q7 + q7 + q7 + 8) >> 4;
1815            dst[strideb * +3] = (p3 + p2 + p1 + p0 + q0 + q1 + q2 + q3 * 2 +
1816                                 q4 + q5 + q6 + q7 + q7 + q7 + q7 + 8) >> 4;
1817            dst[strideb * +4] = (p2 + p1 + p0 + q0 + q1 + q2 + q3 + q4 * 2 +
1818                                 q5 + q6 + q7 + q7 + q7 + q7 + q7 + 8) >> 4;
1819            dst[strideb * +5] = (p1 + p0 + q0 + q1 + q2 + q3 + q4 + q5 * 2 +
1820                                 q6 + q7 + q7 + q7 + q7 + q7 + q7 + 8) >> 4;
1821            dst[strideb * +6] = (p0 + q0 + q1 + q2 + q3 + q4 + q5 + q6 * 2 +
1822                                 q7 + q7 + q7 + q7 + q7 + q7 + q7 + 8) >> 4;
1823        } else if (wd >= 8 && flat8in) {
1824            dst[strideb * -3] = (p3 + p3 + p3 + 2 * p2 + p1 + p0 + q0 + 4) >> 3;
1825            dst[strideb * -2] = (p3 + p3 + p2 + 2 * p1 + p0 + q0 + q1 + 4) >> 3;
1826            dst[strideb * -1] = (p3 + p2 + p1 + 2 * p0 + q0 + q1 + q2 + 4) >> 3;
1827            dst[strideb * +0] = (p2 + p1 + p0 + 2 * q0 + q1 + q2 + q3 + 4) >> 3;
1828            dst[strideb * +1] = (p1 + p0 + q0 + 2 * q1 + q2 + q3 + q3 + 4) >> 3;
1829            dst[strideb * +2] = (p0 + q0 + q1 + 2 * q2 + q3 + q3 + q3 + 4) >> 3;
1830        } else {
1831            int hev = FFABS(p1 - p0) > H || FFABS(q1 - q0) > H;
1832
1833            if (hev) {
1834                int f = av_clip_intp2(p1 - q1, BIT_DEPTH - 1), f1, f2;
1835                f = av_clip_intp2(3 * (q0 - p0) + f, BIT_DEPTH - 1);
1836
1837                f1 = FFMIN(f + 4, (1 << (BIT_DEPTH - 1)) - 1) >> 3;
1838                f2 = FFMIN(f + 3, (1 << (BIT_DEPTH - 1)) - 1) >> 3;
1839
1840                dst[strideb * -1] = av_clip_pixel(p0 + f2);
1841                dst[strideb * +0] = av_clip_pixel(q0 - f1);
1842            } else {
1843                int f = av_clip_intp2(3 * (q0 - p0), BIT_DEPTH - 1), f1, f2;
1844
1845                f1 = FFMIN(f + 4, (1 << (BIT_DEPTH - 1)) - 1) >> 3;
1846                f2 = FFMIN(f + 3, (1 << (BIT_DEPTH - 1)) - 1) >> 3;
1847
1848                dst[strideb * -1] = av_clip_pixel(p0 + f2);
1849                dst[strideb * +0] = av_clip_pixel(q0 - f1);
1850
1851                f = (f1 + 1) >> 1;
1852                dst[strideb * -2] = av_clip_pixel(p1 + f);
1853                dst[strideb * +1] = av_clip_pixel(q1 - f);
1854            }
1855        }
1856    }
1857}
1858
1859#define lf_8_fn(dir, wd, stridea, strideb) \
1860static void loop_filter_##dir##_##wd##_8_c(uint8_t *_dst, \
1861                                           ptrdiff_t stride, \
1862                                           int E, int I, int H) \
1863{ \
1864    pixel *dst = (pixel *) _dst; \
1865    stride /= sizeof(pixel); \
1866    loop_filter(dst, E, I, H, stridea, strideb, wd); \
1867}
1868
1869#define lf_8_fns(wd) \
1870lf_8_fn(h, wd, stride, 1) \
1871lf_8_fn(v, wd, 1, stride)
1872
1873lf_8_fns(4)
1874lf_8_fns(8)
1875lf_8_fns(16)
1876
1877#undef lf_8_fn
1878#undef lf_8_fns
1879
1880#define lf_16_fn(dir, stridea) \
1881static void loop_filter_##dir##_16_16_c(uint8_t *dst, \
1882                                        ptrdiff_t stride, \
1883                                        int E, int I, int H) \
1884{ \
1885    loop_filter_##dir##_16_8_c(dst, stride, E, I, H); \
1886    loop_filter_##dir##_16_8_c(dst + 8 * stridea, stride, E, I, H); \
1887}
1888
1889lf_16_fn(h, stride)
1890lf_16_fn(v, sizeof(pixel))
1891
1892#undef lf_16_fn
1893
1894#define lf_mix_fn(dir, wd1, wd2, stridea) \
1895static void loop_filter_##dir##_##wd1##wd2##_16_c(uint8_t *dst, \
1896                                                  ptrdiff_t stride, \
1897                                                  int E, int I, int H) \
1898{ \
1899    loop_filter_##dir##_##wd1##_8_c(dst, stride, E & 0xff, I & 0xff, H & 0xff); \
1900    loop_filter_##dir##_##wd2##_8_c(dst + 8 * stridea, stride, E >> 8, I >> 8, H >> 8); \
1901}
1902
1903#define lf_mix_fns(wd1, wd2) \
1904lf_mix_fn(h, wd1, wd2, stride) \
1905lf_mix_fn(v, wd1, wd2, sizeof(pixel))
1906
1907lf_mix_fns(4, 4)
1908lf_mix_fns(4, 8)
1909lf_mix_fns(8, 4)
1910lf_mix_fns(8, 8)
1911
1912#undef lf_mix_fn
1913#undef lf_mix_fns
1914
1915static av_cold void vp9dsp_loopfilter_init(VP9DSPContext *dsp)
1916{
1917    dsp->loop_filter_8[0][0] = loop_filter_h_4_8_c;
1918    dsp->loop_filter_8[0][1] = loop_filter_v_4_8_c;
1919    dsp->loop_filter_8[1][0] = loop_filter_h_8_8_c;
1920    dsp->loop_filter_8[1][1] = loop_filter_v_8_8_c;
1921    dsp->loop_filter_8[2][0] = loop_filter_h_16_8_c;
1922    dsp->loop_filter_8[2][1] = loop_filter_v_16_8_c;
1923
1924    dsp->loop_filter_16[0] = loop_filter_h_16_16_c;
1925    dsp->loop_filter_16[1] = loop_filter_v_16_16_c;
1926
1927    dsp->loop_filter_mix2[0][0][0] = loop_filter_h_44_16_c;
1928    dsp->loop_filter_mix2[0][0][1] = loop_filter_v_44_16_c;
1929    dsp->loop_filter_mix2[0][1][0] = loop_filter_h_48_16_c;
1930    dsp->loop_filter_mix2[0][1][1] = loop_filter_v_48_16_c;
1931    dsp->loop_filter_mix2[1][0][0] = loop_filter_h_84_16_c;
1932    dsp->loop_filter_mix2[1][0][1] = loop_filter_v_84_16_c;
1933    dsp->loop_filter_mix2[1][1][0] = loop_filter_h_88_16_c;
1934    dsp->loop_filter_mix2[1][1][1] = loop_filter_v_88_16_c;
1935}
1936
1937#if BIT_DEPTH != 12
1938
1939static av_always_inline void copy_c(uint8_t *dst, ptrdiff_t dst_stride,
1940                                    const uint8_t *src, ptrdiff_t src_stride,
1941                                    int w, int h)
1942{
1943    do {
1944        memcpy(dst, src, w * sizeof(pixel));
1945
1946        dst += dst_stride;
1947        src += src_stride;
1948    } while (--h);
1949}
1950
1951static av_always_inline void avg_c(uint8_t *_dst, ptrdiff_t dst_stride,
1952                                   const uint8_t *_src, ptrdiff_t src_stride,
1953                                   int w, int h)
1954{
1955    pixel *dst = (pixel *) _dst;
1956    const pixel *src = (const pixel *) _src;
1957
1958    dst_stride /= sizeof(pixel);
1959    src_stride /= sizeof(pixel);
1960    do {
1961        int x;
1962
1963        for (x = 0; x < w; x += 4)
1964            AV_WN4PA(&dst[x], rnd_avg_pixel4(AV_RN4PA(&dst[x]), AV_RN4P(&src[x])));
1965
1966        dst += dst_stride;
1967        src += src_stride;
1968    } while (--h);
1969}
1970
1971#define fpel_fn(type, sz) \
1972static void type##sz##_c(uint8_t *dst, ptrdiff_t dst_stride, \
1973                         const uint8_t *src, ptrdiff_t src_stride, \
1974                         int h, int mx, int my) \
1975{ \
1976    type##_c(dst, dst_stride, src, src_stride, sz, h); \
1977}
1978
1979#define copy_avg_fn(sz) \
1980fpel_fn(copy, sz) \
1981fpel_fn(avg,  sz)
1982
1983copy_avg_fn(64)
1984copy_avg_fn(32)
1985copy_avg_fn(16)
1986copy_avg_fn(8)
1987copy_avg_fn(4)
1988
1989#undef fpel_fn
1990#undef copy_avg_fn
1991
1992#endif /* BIT_DEPTH != 12 */
1993
1994#define FILTER_8TAP(src, x, F, stride) \
1995    av_clip_pixel((F[0] * src[x + -3 * stride] + \
1996                   F[1] * src[x + -2 * stride] + \
1997                   F[2] * src[x + -1 * stride] + \
1998                   F[3] * src[x + +0 * stride] + \
1999                   F[4] * src[x + +1 * stride] + \
2000                   F[5] * src[x + +2 * stride] + \
2001                   F[6] * src[x + +3 * stride] + \
2002                   F[7] * src[x + +4 * stride] + 64) >> 7)
2003
2004static av_always_inline void do_8tap_1d_c(uint8_t *_dst, ptrdiff_t dst_stride,
2005                                          const uint8_t *_src, ptrdiff_t src_stride,
2006                                          int w, int h, ptrdiff_t ds,
2007                                          const int16_t *filter, int avg)
2008{
2009    pixel *dst = (pixel *) _dst;
2010    const pixel *src = (const pixel *) _src;
2011
2012    dst_stride /= sizeof(pixel);
2013    src_stride /= sizeof(pixel);
2014    do {
2015        int x;
2016
2017        for (x = 0; x < w; x++)
2018            if (avg) {
2019                dst[x] = (dst[x] + FILTER_8TAP(src, x, filter, ds) + 1) >> 1;
2020            } else {
2021                dst[x] = FILTER_8TAP(src, x, filter, ds);
2022            }
2023
2024        dst += dst_stride;
2025        src += src_stride;
2026    } while (--h);
2027}
2028
2029#define filter_8tap_1d_fn(opn, opa, dir, ds) \
2030static av_noinline void opn##_8tap_1d_##dir##_c(uint8_t *dst, ptrdiff_t dst_stride, \
2031                                                const uint8_t *src, ptrdiff_t src_stride, \
2032                                                int w, int h, const int16_t *filter) \
2033{ \
2034    do_8tap_1d_c(dst, dst_stride, src, src_stride, w, h, ds, filter, opa); \
2035}
2036
2037filter_8tap_1d_fn(put, 0, v, src_stride / sizeof(pixel))
2038filter_8tap_1d_fn(put, 0, h, 1)
2039filter_8tap_1d_fn(avg, 1, v, src_stride / sizeof(pixel))
2040filter_8tap_1d_fn(avg, 1, h, 1)
2041
2042#undef filter_8tap_1d_fn
2043
2044static av_always_inline void do_8tap_2d_c(uint8_t *_dst, ptrdiff_t dst_stride,
2045                                          const uint8_t *_src, ptrdiff_t src_stride,
2046                                          int w, int h, const int16_t *filterx,
2047                                          const int16_t *filtery, int avg)
2048{
2049    int tmp_h = h + 7;
2050    pixel tmp[64 * 71], *tmp_ptr = tmp;
2051    pixel *dst = (pixel *) _dst;
2052    const pixel *src = (const pixel *) _src;
2053
2054    dst_stride /= sizeof(pixel);
2055    src_stride /= sizeof(pixel);
2056    src -= src_stride * 3;
2057    do {
2058        int x;
2059
2060        for (x = 0; x < w; x++)
2061            tmp_ptr[x] = FILTER_8TAP(src, x, filterx, 1);
2062
2063        tmp_ptr += 64;
2064        src += src_stride;
2065    } while (--tmp_h);
2066
2067    tmp_ptr = tmp + 64 * 3;
2068    do {
2069        int x;
2070
2071        for (x = 0; x < w; x++)
2072            if (avg) {
2073                dst[x] = (dst[x] + FILTER_8TAP(tmp_ptr, x, filtery, 64) + 1) >> 1;
2074            } else {
2075                dst[x] = FILTER_8TAP(tmp_ptr, x, filtery, 64);
2076            }
2077
2078        tmp_ptr += 64;
2079        dst += dst_stride;
2080    } while (--h);
2081}
2082
2083#define filter_8tap_2d_fn(opn, opa) \
2084static av_noinline void opn##_8tap_2d_hv_c(uint8_t *dst, ptrdiff_t dst_stride, \
2085                                           const uint8_t *src, ptrdiff_t src_stride, \
2086                                           int w, int h, const int16_t *filterx, \
2087                                           const int16_t *filtery) \
2088{ \
2089    do_8tap_2d_c(dst, dst_stride, src, src_stride, w, h, filterx, filtery, opa); \
2090}
2091
2092filter_8tap_2d_fn(put, 0)
2093filter_8tap_2d_fn(avg, 1)
2094
2095#undef filter_8tap_2d_fn
2096
2097#define filter_fn_1d(sz, dir, dir_m, type, type_idx, avg) \
2098static void avg##_8tap_##type##_##sz##dir##_c(uint8_t *dst, ptrdiff_t dst_stride, \
2099                                              const uint8_t *src, ptrdiff_t src_stride, \
2100                                              int h, int mx, int my) \
2101{ \
2102    avg##_8tap_1d_##dir##_c(dst, dst_stride, src, src_stride, sz, h, \
2103                            ff_vp9_subpel_filters[type_idx][dir_m]); \
2104}
2105
2106#define filter_fn_2d(sz, type, type_idx, avg) \
2107static void avg##_8tap_##type##_##sz##hv_c(uint8_t *dst, ptrdiff_t dst_stride, \
2108                                           const uint8_t *src, ptrdiff_t src_stride, \
2109                                           int h, int mx, int my) \
2110{ \
2111    avg##_8tap_2d_hv_c(dst, dst_stride, src, src_stride, sz, h, \
2112                       ff_vp9_subpel_filters[type_idx][mx], \
2113                       ff_vp9_subpel_filters[type_idx][my]); \
2114}
2115
2116#if BIT_DEPTH != 12
2117
2118#define FILTER_BILIN(src, x, mxy, stride) \
2119    (src[x] + ((mxy * (src[x + stride] - src[x]) + 8) >> 4))
2120
2121static av_always_inline void do_bilin_1d_c(uint8_t *_dst, ptrdiff_t dst_stride,
2122                                           const uint8_t *_src, ptrdiff_t src_stride,
2123                                           int w, int h, ptrdiff_t ds, int mxy, int avg)
2124{
2125    pixel *dst = (pixel *) _dst;
2126    const pixel *src = (const pixel *) _src;
2127
2128    dst_stride /= sizeof(pixel);
2129    src_stride /= sizeof(pixel);
2130    do {
2131        int x;
2132
2133        for (x = 0; x < w; x++)
2134            if (avg) {
2135                dst[x] = (dst[x] + FILTER_BILIN(src, x, mxy, ds) + 1) >> 1;
2136            } else {
2137                dst[x] = FILTER_BILIN(src, x, mxy, ds);
2138            }
2139
2140        dst += dst_stride;
2141        src += src_stride;
2142    } while (--h);
2143}
2144
2145#define bilin_1d_fn(opn, opa, dir, ds) \
2146static av_noinline void opn##_bilin_1d_##dir##_c(uint8_t *dst, ptrdiff_t dst_stride, \
2147                                                 const uint8_t *src, ptrdiff_t src_stride, \
2148                                                 int w, int h, int mxy) \
2149{ \
2150    do_bilin_1d_c(dst, dst_stride, src, src_stride, w, h, ds, mxy, opa); \
2151}
2152
2153bilin_1d_fn(put, 0, v, src_stride / sizeof(pixel))
2154bilin_1d_fn(put, 0, h, 1)
2155bilin_1d_fn(avg, 1, v, src_stride / sizeof(pixel))
2156bilin_1d_fn(avg, 1, h, 1)
2157
2158#undef bilin_1d_fn
2159
2160static av_always_inline void do_bilin_2d_c(uint8_t *_dst, ptrdiff_t dst_stride,
2161                                           const uint8_t *_src, ptrdiff_t src_stride,
2162                                           int w, int h, int mx, int my, int avg)
2163{
2164    pixel tmp[64 * 65], *tmp_ptr = tmp;
2165    int tmp_h = h + 1;
2166    pixel *dst = (pixel *) _dst;
2167    const pixel *src = (const pixel *) _src;
2168
2169    dst_stride /= sizeof(pixel);
2170    src_stride /= sizeof(pixel);
2171    do {
2172        int x;
2173
2174        for (x = 0; x < w; x++)
2175            tmp_ptr[x] = FILTER_BILIN(src, x, mx, 1);
2176
2177        tmp_ptr += 64;
2178        src += src_stride;
2179    } while (--tmp_h);
2180
2181    tmp_ptr = tmp;
2182    do {
2183        int x;
2184
2185        for (x = 0; x < w; x++)
2186            if (avg) {
2187                dst[x] = (dst[x] + FILTER_BILIN(tmp_ptr, x, my, 64) + 1) >> 1;
2188            } else {
2189                dst[x] = FILTER_BILIN(tmp_ptr, x, my, 64);
2190            }
2191
2192        tmp_ptr += 64;
2193        dst += dst_stride;
2194    } while (--h);
2195}
2196
2197#define bilin_2d_fn(opn, opa) \
2198static av_noinline void opn##_bilin_2d_hv_c(uint8_t *dst, ptrdiff_t dst_stride, \
2199                                            const uint8_t *src, ptrdiff_t src_stride, \
2200                                            int w, int h, int mx, int my) \
2201{ \
2202    do_bilin_2d_c(dst, dst_stride, src, src_stride, w, h, mx, my, opa); \
2203}
2204
2205bilin_2d_fn(put, 0)
2206bilin_2d_fn(avg, 1)
2207
2208#undef bilin_2d_fn
2209
2210#define bilinf_fn_1d(sz, dir, dir_m, avg) \
2211static void avg##_bilin_##sz##dir##_c(uint8_t *dst, ptrdiff_t dst_stride, \
2212                                      const uint8_t *src, ptrdiff_t src_stride, \
2213                                      int h, int mx, int my) \
2214{ \
2215    avg##_bilin_1d_##dir##_c(dst, dst_stride, src, src_stride, sz, h, dir_m); \
2216}
2217
2218#define bilinf_fn_2d(sz, avg) \
2219static void avg##_bilin_##sz##hv_c(uint8_t *dst, ptrdiff_t dst_stride, \
2220                                   const uint8_t *src, ptrdiff_t src_stride, \
2221                                   int h, int mx, int my) \
2222{ \
2223    avg##_bilin_2d_hv_c(dst, dst_stride, src, src_stride, sz, h, mx, my); \
2224}
2225
2226#else
2227
2228#define bilinf_fn_1d(a, b, c, d)
2229#define bilinf_fn_2d(a, b)
2230
2231#endif
2232
2233#define filter_fn(sz, avg) \
2234filter_fn_1d(sz, h, mx, regular, FILTER_8TAP_REGULAR, avg) \
2235filter_fn_1d(sz, v, my, regular, FILTER_8TAP_REGULAR, avg) \
2236filter_fn_2d(sz,        regular, FILTER_8TAP_REGULAR, avg) \
2237filter_fn_1d(sz, h, mx, smooth,  FILTER_8TAP_SMOOTH,  avg) \
2238filter_fn_1d(sz, v, my, smooth,  FILTER_8TAP_SMOOTH,  avg) \
2239filter_fn_2d(sz,        smooth,  FILTER_8TAP_SMOOTH,  avg) \
2240filter_fn_1d(sz, h, mx, sharp,   FILTER_8TAP_SHARP,   avg) \
2241filter_fn_1d(sz, v, my, sharp,   FILTER_8TAP_SHARP,   avg) \
2242filter_fn_2d(sz,        sharp,   FILTER_8TAP_SHARP,   avg) \
2243bilinf_fn_1d(sz, h, mx,                               avg) \
2244bilinf_fn_1d(sz, v, my,                               avg) \
2245bilinf_fn_2d(sz,                                      avg)
2246
2247#define filter_fn_set(avg) \
2248filter_fn(64, avg) \
2249filter_fn(32, avg) \
2250filter_fn(16, avg) \
2251filter_fn(8,  avg) \
2252filter_fn(4,  avg)
2253
2254filter_fn_set(put)
2255filter_fn_set(avg)
2256
2257#undef filter_fn
2258#undef filter_fn_set
2259#undef filter_fn_1d
2260#undef filter_fn_2d
2261#undef bilinf_fn_1d
2262#undef bilinf_fn_2d
2263
2264#if BIT_DEPTH != 8
2265void ff_vp9dsp_mc_init_10(VP9DSPContext *dsp);
2266#endif
2267#if BIT_DEPTH != 10
2268static
2269#endif
2270av_cold void FUNC(ff_vp9dsp_mc_init)(VP9DSPContext *dsp)
2271{
2272#if BIT_DEPTH == 12
2273    ff_vp9dsp_mc_init_10(dsp);
2274#else /* BIT_DEPTH == 12 */
2275
2276#define init_fpel(idx1, idx2, sz, type) \
2277    dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] = type##sz##_c; \
2278    dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = type##sz##_c; \
2279    dsp->mc[idx1][FILTER_8TAP_SHARP  ][idx2][0][0] = type##sz##_c; \
2280    dsp->mc[idx1][FILTER_BILINEAR    ][idx2][0][0] = type##sz##_c
2281
2282#define init_copy_avg(idx, sz) \
2283    init_fpel(idx, 0, sz, copy); \
2284    init_fpel(idx, 1, sz, avg)
2285
2286    init_copy_avg(0, 64);
2287    init_copy_avg(1, 32);
2288    init_copy_avg(2, 16);
2289    init_copy_avg(3,  8);
2290    init_copy_avg(4,  4);
2291
2292#undef init_copy_avg
2293#undef init_fpel
2294
2295#endif /* BIT_DEPTH == 12 */
2296
2297#define init_subpel1_bd_aware(idx1, idx2, idxh, idxv, sz, dir, type) \
2298    dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][idxh][idxv] = type##_8tap_smooth_##sz##dir##_c; \
2299    dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][idxh][idxv] = type##_8tap_regular_##sz##dir##_c; \
2300    dsp->mc[idx1][FILTER_8TAP_SHARP  ][idx2][idxh][idxv] = type##_8tap_sharp_##sz##dir##_c
2301
2302#if BIT_DEPTH == 12
2303#define init_subpel1 init_subpel1_bd_aware
2304#else
2305#define init_subpel1(idx1, idx2, idxh, idxv, sz, dir, type) \
2306    init_subpel1_bd_aware(idx1, idx2, idxh, idxv, sz, dir, type); \
2307    dsp->mc[idx1][FILTER_BILINEAR    ][idx2][idxh][idxv] = type##_bilin_##sz##dir##_c
2308#endif
2309
2310#define init_subpel2(idx, idxh, idxv, dir, type) \
2311    init_subpel1(0, idx, idxh, idxv, 64, dir, type); \
2312    init_subpel1(1, idx, idxh, idxv, 32, dir, type); \
2313    init_subpel1(2, idx, idxh, idxv, 16, dir, type); \
2314    init_subpel1(3, idx, idxh, idxv,  8, dir, type); \
2315    init_subpel1(4, idx, idxh, idxv,  4, dir, type)
2316
2317#define init_subpel3(idx, type) \
2318    init_subpel2(idx, 1, 1, hv, type); \
2319    init_subpel2(idx, 0, 1, v, type); \
2320    init_subpel2(idx, 1, 0, h, type)
2321
2322    init_subpel3(0, put);
2323    init_subpel3(1, avg);
2324
2325#undef init_subpel1
2326#undef init_subpel2
2327#undef init_subpel3
2328#undef init_subpel1_bd_aware
2329}
2330
2331static av_always_inline void do_scaled_8tap_c(uint8_t *_dst, ptrdiff_t dst_stride,
2332                                              const uint8_t *_src, ptrdiff_t src_stride,
2333                                              int w, int h, int mx, int my,
2334                                              int dx, int dy, int avg,
2335                                              const int16_t (*filters)[8])
2336{
2337    int tmp_h = (((h - 1) * dy + my) >> 4) + 8;
2338    pixel tmp[64 * 135], *tmp_ptr = tmp;
2339    pixel *dst = (pixel *) _dst;
2340    const pixel *src = (const pixel *) _src;
2341
2342    dst_stride /= sizeof(pixel);
2343    src_stride /= sizeof(pixel);
2344    src -= src_stride * 3;
2345    do {
2346        int x;
2347        int imx = mx, ioff = 0;
2348
2349        for (x = 0; x < w; x++) {
2350            tmp_ptr[x] = FILTER_8TAP(src, ioff, filters[imx], 1);
2351            imx += dx;
2352            ioff += imx >> 4;
2353            imx &= 0xf;
2354        }
2355
2356        tmp_ptr += 64;
2357        src += src_stride;
2358    } while (--tmp_h);
2359
2360    tmp_ptr = tmp + 64 * 3;
2361    do {
2362        int x;
2363        const int16_t *filter = filters[my];
2364
2365        for (x = 0; x < w; x++)
2366            if (avg) {
2367                dst[x] = (dst[x] + FILTER_8TAP(tmp_ptr, x, filter, 64) + 1) >> 1;
2368            } else {
2369                dst[x] = FILTER_8TAP(tmp_ptr, x, filter, 64);
2370            }
2371
2372        my += dy;
2373        tmp_ptr += (my >> 4) * 64;
2374        my &= 0xf;
2375        dst += dst_stride;
2376    } while (--h);
2377}
2378
2379#define scaled_filter_8tap_fn(opn, opa) \
2380static av_noinline void opn##_scaled_8tap_c(uint8_t *dst, ptrdiff_t dst_stride, \
2381                                            const uint8_t *src, ptrdiff_t src_stride, \
2382                                            int w, int h, int mx, int my, int dx, int dy, \
2383                                            const int16_t (*filters)[8]) \
2384{ \
2385    do_scaled_8tap_c(dst, dst_stride, src, src_stride, w, h, mx, my, dx, dy, \
2386                     opa, filters); \
2387}
2388
2389scaled_filter_8tap_fn(put, 0)
2390scaled_filter_8tap_fn(avg, 1)
2391
2392#undef scaled_filter_8tap_fn
2393
2394#undef FILTER_8TAP
2395
2396#define scaled_filter_fn(sz, type, type_idx, avg) \
2397static void avg##_scaled_##type##_##sz##_c(uint8_t *dst, ptrdiff_t dst_stride, \
2398                                           const uint8_t *src, ptrdiff_t src_stride, \
2399                                           int h, int mx, int my, int dx, int dy) \
2400{ \
2401    avg##_scaled_8tap_c(dst, dst_stride, src, src_stride, sz, h, mx, my, dx, dy, \
2402                        ff_vp9_subpel_filters[type_idx]); \
2403}
2404
2405#if BIT_DEPTH != 12
2406
2407static av_always_inline void do_scaled_bilin_c(uint8_t *_dst, ptrdiff_t dst_stride,
2408                                               const uint8_t *_src, ptrdiff_t src_stride,
2409                                               int w, int h, int mx, int my,
2410                                               int dx, int dy, int avg)
2411{
2412    pixel tmp[64 * 129], *tmp_ptr = tmp;
2413    int tmp_h = (((h - 1) * dy + my) >> 4) + 2;
2414    pixel *dst = (pixel *) _dst;
2415    const pixel *src = (const pixel *) _src;
2416
2417    dst_stride /= sizeof(pixel);
2418    src_stride /= sizeof(pixel);
2419    do {
2420        int x;
2421        int imx = mx, ioff = 0;
2422
2423        for (x = 0; x < w; x++) {
2424            tmp_ptr[x] = FILTER_BILIN(src, ioff, imx, 1);
2425            imx += dx;
2426            ioff += imx >> 4;
2427            imx &= 0xf;
2428        }
2429
2430        tmp_ptr += 64;
2431        src += src_stride;
2432    } while (--tmp_h);
2433
2434    tmp_ptr = tmp;
2435    do {
2436        int x;
2437
2438        for (x = 0; x < w; x++)
2439            if (avg) {
2440                dst[x] = (dst[x] + FILTER_BILIN(tmp_ptr, x, my, 64) + 1) >> 1;
2441            } else {
2442                dst[x] = FILTER_BILIN(tmp_ptr, x, my, 64);
2443            }
2444
2445        my += dy;
2446        tmp_ptr += (my >> 4) * 64;
2447        my &= 0xf;
2448        dst += dst_stride;
2449    } while (--h);
2450}
2451
2452#define scaled_bilin_fn(opn, opa) \
2453static av_noinline void opn##_scaled_bilin_c(uint8_t *dst, ptrdiff_t dst_stride, \
2454                                             const uint8_t *src, ptrdiff_t src_stride, \
2455                                             int w, int h, int mx, int my, int dx, int dy) \
2456{ \
2457    do_scaled_bilin_c(dst, dst_stride, src, src_stride, w, h, mx, my, dx, dy, opa); \
2458}
2459
2460scaled_bilin_fn(put, 0)
2461scaled_bilin_fn(avg, 1)
2462
2463#undef scaled_bilin_fn
2464
2465#undef FILTER_BILIN
2466
2467#define scaled_bilinf_fn(sz, avg) \
2468static void avg##_scaled_bilin_##sz##_c(uint8_t *dst, ptrdiff_t dst_stride, \
2469                                        const uint8_t *src, ptrdiff_t src_stride, \
2470                                        int h, int mx, int my, int dx, int dy) \
2471{ \
2472    avg##_scaled_bilin_c(dst, dst_stride, src, src_stride, sz, h, mx, my, dx, dy); \
2473}
2474
2475#else
2476
2477#define scaled_bilinf_fn(a, b)
2478
2479#endif
2480
2481#define scaled_filter_fns(sz, avg) \
2482scaled_filter_fn(sz,        regular, FILTER_8TAP_REGULAR, avg) \
2483scaled_filter_fn(sz,        smooth,  FILTER_8TAP_SMOOTH,  avg) \
2484scaled_filter_fn(sz,        sharp,   FILTER_8TAP_SHARP,   avg) \
2485scaled_bilinf_fn(sz,                                      avg)
2486
2487#define scaled_filter_fn_set(avg) \
2488scaled_filter_fns(64, avg) \
2489scaled_filter_fns(32, avg) \
2490scaled_filter_fns(16, avg) \
2491scaled_filter_fns(8,  avg) \
2492scaled_filter_fns(4,  avg)
2493
2494scaled_filter_fn_set(put)
2495scaled_filter_fn_set(avg)
2496
2497#undef scaled_filter_fns
2498#undef scaled_filter_fn_set
2499#undef scaled_filter_fn
2500#undef scaled_bilinf_fn
2501
2502#if BIT_DEPTH != 8
2503void ff_vp9dsp_scaled_mc_init_10(VP9DSPContext *dsp);
2504#endif
2505#if BIT_DEPTH != 10
2506static
2507#endif
2508av_cold void FUNC(ff_vp9dsp_scaled_mc_init)(VP9DSPContext *dsp)
2509{
2510#define init_scaled_bd_aware(idx1, idx2, sz, type) \
2511    dsp->smc[idx1][FILTER_8TAP_SMOOTH ][idx2] = type##_scaled_smooth_##sz##_c; \
2512    dsp->smc[idx1][FILTER_8TAP_REGULAR][idx2] = type##_scaled_regular_##sz##_c; \
2513    dsp->smc[idx1][FILTER_8TAP_SHARP  ][idx2] = type##_scaled_sharp_##sz##_c
2514
2515#if BIT_DEPTH == 12
2516    ff_vp9dsp_scaled_mc_init_10(dsp);
2517#define init_scaled(a,b,c,d) init_scaled_bd_aware(a,b,c,d)
2518#else
2519#define init_scaled(idx1, idx2, sz, type) \
2520    init_scaled_bd_aware(idx1, idx2, sz, type); \
2521    dsp->smc[idx1][FILTER_BILINEAR    ][idx2] = type##_scaled_bilin_##sz##_c
2522#endif
2523
2524#define init_scaled_put_avg(idx, sz) \
2525    init_scaled(idx, 0, sz, put); \
2526    init_scaled(idx, 1, sz, avg)
2527
2528    init_scaled_put_avg(0, 64);
2529    init_scaled_put_avg(1, 32);
2530    init_scaled_put_avg(2, 16);
2531    init_scaled_put_avg(3,  8);
2532    init_scaled_put_avg(4,  4);
2533
2534#undef init_scaled_put_avg
2535#undef init_scaled
2536#undef init_scaled_bd_aware
2537}
2538
2539av_cold void FUNC(ff_vp9dsp_init)(VP9DSPContext *dsp)
2540{
2541    FUNC(ff_vp9dsp_intrapred_init)(dsp);
2542    vp9dsp_itxfm_init(dsp);
2543    vp9dsp_loopfilter_init(dsp);
2544    FUNC(ff_vp9dsp_mc_init)(dsp);
2545    FUNC(ff_vp9dsp_scaled_mc_init)(dsp);
2546}
2547