Libav
hevcdsp_template.c
Go to the documentation of this file.
1 /*
2  * HEVC video decoder
3  *
4  * Copyright (C) 2012 - 2013 Guillaume Martres
5  *
6  * This file is part of Libav.
7  *
8  * Libav is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * Libav is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with Libav; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
23 #include "get_bits.h"
24 #include "hevc.h"
25 
26 #include "bit_depth_template.c"
27 
28 static void FUNC(put_pcm)(uint8_t *_dst, ptrdiff_t stride, int size,
29  GetBitContext *gb, int pcm_bit_depth)
30 {
31  int x, y;
32  pixel *dst = (pixel *)_dst;
33 
34  stride /= sizeof(pixel);
35 
36  for (y = 0; y < size; y++) {
37  for (x = 0; x < size; x++)
38  dst[x] = get_bits(gb, pcm_bit_depth) << (BIT_DEPTH - pcm_bit_depth);
39  dst += stride;
40  }
41 }
42 
43 static av_always_inline void FUNC(add_residual)(uint8_t *_dst, int16_t *res,
44  ptrdiff_t stride, int size)
45 {
46  int x, y;
47  pixel *dst = (pixel *)_dst;
48 
49  stride /= sizeof(pixel);
50 
51  for (y = 0; y < size; y++) {
52  for (x = 0; x < size; x++) {
53  dst[x] = av_clip_pixel(dst[x] + *res);
54  res++;
55  }
56  dst += stride;
57  }
58 }
59 
60 static void FUNC(add_residual4x4)(uint8_t *_dst, int16_t *res,
61  ptrdiff_t stride)
62 {
63  FUNC(add_residual)(_dst, res, stride, 4);
64 }
65 
66 static void FUNC(add_residual8x8)(uint8_t *_dst, int16_t *res,
67  ptrdiff_t stride)
68 {
69  FUNC(add_residual)(_dst, res, stride, 8);
70 }
71 
72 static void FUNC(add_residual16x16)(uint8_t *_dst, int16_t *res,
73  ptrdiff_t stride)
74 {
75  FUNC(add_residual)(_dst, res, stride, 16);
76 }
77 
78 static void FUNC(add_residual32x32)(uint8_t *_dst, int16_t *res,
79  ptrdiff_t stride)
80 {
81  FUNC(add_residual)(_dst, res, stride, 32);
82 }
83 
84 static void FUNC(dequant)(int16_t *coeffs)
85 {
86  int shift = 13 - BIT_DEPTH;
87 #if BIT_DEPTH <= 13
88  int offset = 1 << (shift - 1);
89 #else
90  int offset = 0;
91 #endif
92  int x, y;
93 
94  for (y = 0; y < 4 * 4; y += 4) {
95  for (x = 0; x < 4; x++)
96  coeffs[y + x] = (coeffs[y + x] + offset) >> shift;
97  }
98 }
99 
100 #define SET(dst, x) (dst) = (x)
101 #define SCALE(dst, x) (dst) = av_clip_int16(((x) + add) >> shift)
102 
103 #define TR_4x4_LUMA(dst, src, step, assign) \
104  do { \
105  int c0 = src[0 * step] + src[2 * step]; \
106  int c1 = src[2 * step] + src[3 * step]; \
107  int c2 = src[0 * step] - src[3 * step]; \
108  int c3 = 74 * src[1 * step]; \
109  \
110  assign(dst[2 * step], 74 * (src[0 * step] - \
111  src[2 * step] + \
112  src[3 * step])); \
113  assign(dst[0 * step], 29 * c0 + 55 * c1 + c3); \
114  assign(dst[1 * step], 55 * c2 - 29 * c1 + c3); \
115  assign(dst[3 * step], 55 * c0 + 29 * c2 - c3); \
116  } while (0)
117 
118 static void FUNC(transform_4x4_luma)(int16_t *coeffs)
119 {
120  int i;
121  int shift = 7;
122  int add = 1 << (shift - 1);
123  int16_t *src = coeffs;
124 
125  for (i = 0; i < 4; i++) {
126  TR_4x4_LUMA(src, src, 4, SCALE);
127  src++;
128  }
129 
130  shift = 20 - BIT_DEPTH;
131  add = 1 << (shift - 1);
132  for (i = 0; i < 4; i++) {
133  TR_4x4_LUMA(coeffs, coeffs, 1, SCALE);
134  coeffs += 4;
135  }
136 }
137 
138 #undef TR_4x4_LUMA
139 
140 #define TR_4(dst, src, dstep, sstep, assign, end) \
141  do { \
142  const int e0 = transform[8 * 0][0] * src[0 * sstep] + \
143  transform[8 * 2][0] * src[2 * sstep]; \
144  const int e1 = transform[8 * 0][1] * src[0 * sstep] + \
145  transform[8 * 2][1] * src[2 * sstep]; \
146  const int o0 = transform[8 * 1][0] * src[1 * sstep] + \
147  transform[8 * 3][0] * src[3 * sstep]; \
148  const int o1 = transform[8 * 1][1] * src[1 * sstep] + \
149  transform[8 * 3][1] * src[3 * sstep]; \
150  \
151  assign(dst[0 * dstep], e0 + o0); \
152  assign(dst[1 * dstep], e1 + o1); \
153  assign(dst[2 * dstep], e1 - o1); \
154  assign(dst[3 * dstep], e0 - o0); \
155  } while (0)
156 
157 #define TR_8(dst, src, dstep, sstep, assign, end) \
158  do { \
159  int i, j; \
160  int e_8[4]; \
161  int o_8[4] = { 0 }; \
162  for (i = 0; i < 4; i++) \
163  for (j = 1; j < end; j += 2) \
164  o_8[i] += transform[4 * j][i] * src[j * sstep]; \
165  TR_4(e_8, src, 1, 2 * sstep, SET, 4); \
166  \
167  for (i = 0; i < 4; i++) { \
168  assign(dst[i * dstep], e_8[i] + o_8[i]); \
169  assign(dst[(7 - i) * dstep], e_8[i] - o_8[i]); \
170  } \
171  } while (0)
172 
173 #define TR_16(dst, src, dstep, sstep, assign, end) \
174  do { \
175  int i, j; \
176  int e_16[8]; \
177  int o_16[8] = { 0 }; \
178  for (i = 0; i < 8; i++) \
179  for (j = 1; j < end; j += 2) \
180  o_16[i] += transform[2 * j][i] * src[j * sstep]; \
181  TR_8(e_16, src, 1, 2 * sstep, SET, 8); \
182  \
183  for (i = 0; i < 8; i++) { \
184  assign(dst[i * dstep], e_16[i] + o_16[i]); \
185  assign(dst[(15 - i) * dstep], e_16[i] - o_16[i]); \
186  } \
187  } while (0)
188 
189 #define TR_32(dst, src, dstep, sstep, assign, end) \
190  do { \
191  int i, j; \
192  int e_32[16]; \
193  int o_32[16] = { 0 }; \
194  for (i = 0; i < 16; i++) \
195  for (j = 1; j < end; j += 2) \
196  o_32[i] += transform[j][i] * src[j * sstep]; \
197  TR_16(e_32, src, 1, 2 * sstep, SET, end / 2); \
198  \
199  for (i = 0; i < 16; i++) { \
200  assign(dst[i * dstep], e_32[i] + o_32[i]); \
201  assign(dst[(31 - i) * dstep], e_32[i] - o_32[i]); \
202  } \
203  } while (0)
204 
205 #define IDCT_VAR4(H) \
206  int limit2 = FFMIN(col_limit + 4, H)
207 #define IDCT_VAR8(H) \
208  int limit = FFMIN(col_limit, H); \
209  int limit2 = FFMIN(col_limit + 4, H)
210 #define IDCT_VAR16(H) IDCT_VAR8(H)
211 #define IDCT_VAR32(H) IDCT_VAR8(H)
212 
213 #define IDCT(H) \
214 static void FUNC(idct_ ## H ## x ## H )(int16_t *coeffs, \
215  int col_limit) \
216 { \
217  int i; \
218  int shift = 7; \
219  int add = 1 << (shift - 1); \
220  int16_t *src = coeffs; \
221  IDCT_VAR ## H(H); \
222  \
223  for (i = 0; i < H; i++) { \
224  TR_ ## H(src, src, H, H, SCALE, limit2); \
225  if (limit2 < H && i%4 == 0 && !!i) \
226  limit2 -= 4; \
227  src++; \
228  } \
229  \
230  shift = 20 - BIT_DEPTH; \
231  add = 1 << (shift - 1); \
232  for (i = 0; i < H; i++) { \
233  TR_ ## H(coeffs, coeffs, 1, 1, SCALE, limit); \
234  coeffs += H; \
235  } \
236 }
237 
238 #define IDCT_DC(H) \
239 static void FUNC(idct_ ## H ## x ## H ## _dc)(int16_t *coeffs) \
240 { \
241  int i, j; \
242  int shift = 14 - BIT_DEPTH; \
243  int add = 1 << (shift - 1); \
244  int coeff = (((coeffs[0] + 1) >> 1) + add) >> shift; \
245  \
246  for (j = 0; j < H; j++) { \
247  for (i = 0; i < H; i++) { \
248  coeffs[i + j * H] = coeff; \
249  } \
250  } \
251 }
252 
253 IDCT( 4)
254 IDCT( 8)
255 IDCT(16)
256 IDCT(32)
257 IDCT_DC( 4)
258 IDCT_DC( 8)
259 IDCT_DC(16)
260 IDCT_DC(32)
261 #undef TR_4
262 #undef TR_8
263 #undef TR_16
264 #undef TR_32
265 
266 #undef SET
267 #undef SCALE
268 #undef ADD_AND_SCALE
269 
270 static void FUNC(sao_band_filter)(uint8_t *_dst, uint8_t *_src,
271  ptrdiff_t stride, SAOParams *sao,
272  int *borders, int width, int height,
273  int c_idx, int class)
274 {
275  pixel *dst = (pixel *)_dst;
276  pixel *src = (pixel *)_src;
277  int offset_table[32] = { 0 };
278  int k, y, x;
279  int chroma = !!c_idx;
280  int shift = BIT_DEPTH - 5;
281  int *sao_offset_val = sao->offset_val[c_idx];
282  int sao_left_class = sao->band_position[c_idx];
283  int init_y = 0, init_x = 0;
284 
285  stride /= sizeof(pixel);
286 
287  switch (class) {
288  case 0:
289  if (!borders[2])
290  width -= (8 >> chroma) + 2;
291  if (!borders[3])
292  height -= (4 >> chroma) + 2;
293  break;
294  case 1:
295  init_y = -(4 >> chroma) - 2;
296  if (!borders[2])
297  width -= (8 >> chroma) + 2;
298  height = (4 >> chroma) + 2;
299  break;
300  case 2:
301  init_x = -(8 >> chroma) - 2;
302  width = (8 >> chroma) + 2;
303  if (!borders[3])
304  height -= (4 >> chroma) + 2;
305  break;
306  case 3:
307  init_y = -(4 >> chroma) - 2;
308  init_x = -(8 >> chroma) - 2;
309  width = (8 >> chroma) + 2;
310  height = (4 >> chroma) + 2;
311  break;
312  }
313 
314  dst = dst + (init_y * stride + init_x);
315  src = src + (init_y * stride + init_x);
316  for (k = 0; k < 4; k++)
317  offset_table[(k + sao_left_class) & 31] = sao_offset_val[k + 1];
318  for (y = 0; y < height; y++) {
319  for (x = 0; x < width; x++)
320  dst[x] = av_clip_pixel(src[x] + offset_table[src[x] >> shift]);
321  dst += stride;
322  src += stride;
323  }
324 }
325 
327  ptrdiff_t stride, SAOParams *sao,
328  int *borders, int width, int height,
329  int c_idx)
330 {
331  FUNC(sao_band_filter)(dst, src, stride, sao, borders,
332  width, height, c_idx, 0);
333 }
334 
336  ptrdiff_t stride, SAOParams *sao,
337  int *borders, int width, int height,
338  int c_idx)
339 {
340  FUNC(sao_band_filter)(dst, src, stride, sao, borders,
341  width, height, c_idx, 1);
342 }
343 
345  ptrdiff_t stride, SAOParams *sao,
346  int *borders, int width, int height,
347  int c_idx)
348 {
349  FUNC(sao_band_filter)(dst, src, stride, sao, borders,
350  width, height, c_idx, 2);
351 }
352 
353 static void FUNC(sao_band_filter_3)(uint8_t *_dst, uint8_t *_src,
354  ptrdiff_t stride, SAOParams *sao,
355  int *borders, int width, int height,
356  int c_idx)
357 {
358  FUNC(sao_band_filter)(_dst, _src, stride, sao, borders,
359  width, height, c_idx, 3);
360 }
361 
362 static void FUNC(sao_edge_filter_0)(uint8_t *_dst, uint8_t *_src,
363  ptrdiff_t stride, SAOParams *sao,
364  int *borders, int _width, int _height,
365  int c_idx, uint8_t vert_edge,
366  uint8_t horiz_edge, uint8_t diag_edge)
367 {
368  int x, y;
369  pixel *dst = (pixel *)_dst;
370  pixel *src = (pixel *)_src;
371  int chroma = !!c_idx;
372  int *sao_offset_val = sao->offset_val[c_idx];
373  int sao_eo_class = sao->eo_class[c_idx];
374  int init_x = 0, init_y = 0, width = _width, height = _height;
375 
376  static const int8_t pos[4][2][2] = {
377  { { -1, 0 }, { 1, 0 } }, // horizontal
378  { { 0, -1 }, { 0, 1 } }, // vertical
379  { { -1, -1 }, { 1, 1 } }, // 45 degree
380  { { 1, -1 }, { -1, 1 } }, // 135 degree
381  };
382  static const uint8_t edge_idx[] = { 1, 2, 0, 3, 4 };
383 
384 #define CMP(a, b) ((a) > (b) ? 1 : ((a) == (b) ? 0 : -1))
385 
386  stride /= sizeof(pixel);
387 
388  if (!borders[2])
389  width -= (8 >> chroma) + 2;
390  if (!borders[3])
391  height -= (4 >> chroma) + 2;
392 
393  dst = dst + (init_y * stride + init_x);
394  src = src + (init_y * stride + init_x);
395  init_y = init_x = 0;
396  if (sao_eo_class != SAO_EO_VERT) {
397  if (borders[0]) {
398  int offset_val = sao_offset_val[0];
399  int y_stride = 0;
400  for (y = 0; y < height; y++) {
401  dst[y_stride] = av_clip_pixel(src[y_stride] + offset_val);
402  y_stride += stride;
403  }
404  init_x = 1;
405  }
406  if (borders[2]) {
407  int offset_val = sao_offset_val[0];
408  int x_stride = width - 1;
409  for (x = 0; x < height; x++) {
410  dst[x_stride] = av_clip_pixel(src[x_stride] + offset_val);
411  x_stride += stride;
412  }
413  width--;
414  }
415  }
416  if (sao_eo_class != SAO_EO_HORIZ) {
417  if (borders[1]) {
418  int offset_val = sao_offset_val[0];
419  for (x = init_x; x < width; x++)
420  dst[x] = av_clip_pixel(src[x] + offset_val);
421  init_y = 1;
422  }
423  if (borders[3]) {
424  int offset_val = sao_offset_val[0];
425  int y_stride = stride * (height - 1);
426  for (x = init_x; x < width; x++)
427  dst[x + y_stride] = av_clip_pixel(src[x + y_stride] + offset_val);
428  height--;
429  }
430  }
431  {
432  int y_stride = init_y * stride;
433  int pos_0_0 = pos[sao_eo_class][0][0];
434  int pos_0_1 = pos[sao_eo_class][0][1];
435  int pos_1_0 = pos[sao_eo_class][1][0];
436  int pos_1_1 = pos[sao_eo_class][1][1];
437 
438  int y_stride_0_1 = (init_y + pos_0_1) * stride;
439  int y_stride_1_1 = (init_y + pos_1_1) * stride;
440  for (y = init_y; y < height; y++) {
441  for (x = init_x; x < width; x++) {
442  int diff0 = CMP(src[x + y_stride], src[x + pos_0_0 + y_stride_0_1]);
443  int diff1 = CMP(src[x + y_stride], src[x + pos_1_0 + y_stride_1_1]);
444  int offset_val = edge_idx[2 + diff0 + diff1];
445  dst[x + y_stride] = av_clip_pixel(src[x + y_stride] + sao_offset_val[offset_val]);
446  }
447  y_stride += stride;
448  y_stride_0_1 += stride;
449  y_stride_1_1 += stride;
450  }
451  }
452 
453  {
454  // Restore pixels that can't be modified
455  int save_upper_left = !diag_edge && sao_eo_class == SAO_EO_135D && !borders[0] && !borders[1];
456  if (vert_edge && sao_eo_class != SAO_EO_VERT)
457  for (y = init_y+save_upper_left; y< height; y++)
458  dst[y*stride] = src[y*stride];
459  if(horiz_edge && sao_eo_class != SAO_EO_HORIZ)
460  for(x = init_x+save_upper_left; x<width; x++)
461  dst[x] = src[x];
462  if(diag_edge && sao_eo_class == SAO_EO_135D)
463  dst[0] = src[0];
464  }
465 
466 #undef CMP
467 }
468 
469 static void FUNC(sao_edge_filter_1)(uint8_t *_dst, uint8_t *_src,
470  ptrdiff_t stride, SAOParams *sao,
471  int *borders, int _width, int _height,
472  int c_idx, uint8_t vert_edge,
473  uint8_t horiz_edge, uint8_t diag_edge)
474 {
475  int x, y;
476  pixel *dst = (pixel *)_dst;
477  pixel *src = (pixel *)_src;
478  int chroma = !!c_idx;
479  int *sao_offset_val = sao->offset_val[c_idx];
480  int sao_eo_class = sao->eo_class[c_idx];
481  int init_x = 0, init_y = 0, width = _width, height = _height;
482 
483  static const int8_t pos[4][2][2] = {
484  { { -1, 0 }, { 1, 0 } }, // horizontal
485  { { 0, -1 }, { 0, 1 } }, // vertical
486  { { -1, -1 }, { 1, 1 } }, // 45 degree
487  { { 1, -1 }, { -1, 1 } }, // 135 degree
488  };
489  static const uint8_t edge_idx[] = { 1, 2, 0, 3, 4 };
490 
491 #define CMP(a, b) ((a) > (b) ? 1 : ((a) == (b) ? 0 : -1))
492 
493  stride /= sizeof(pixel);
494 
495  init_y = -(4 >> chroma) - 2;
496  if (!borders[2])
497  width -= (8 >> chroma) + 2;
498  height = (4 >> chroma) + 2;
499 
500  dst = dst + (init_y * stride + init_x);
501  src = src + (init_y * stride + init_x);
502  init_y = init_x = 0;
503  if (sao_eo_class != SAO_EO_VERT) {
504  if (borders[0]) {
505  int offset_val = sao_offset_val[0];
506  int y_stride = 0;
507  for (y = 0; y < height; y++) {
508  dst[y_stride] = av_clip_pixel(src[y_stride] + offset_val);
509  y_stride += stride;
510  }
511  init_x = 1;
512  }
513  if (borders[2]) {
514  int offset_val = sao_offset_val[0];
515  int x_stride = width - 1;
516  for (x = 0; x < height; x++) {
517  dst[x_stride] = av_clip_pixel(src[x_stride] + offset_val);
518  x_stride += stride;
519  }
520  width--;
521  }
522  }
523  {
524  int y_stride = init_y * stride;
525  int pos_0_0 = pos[sao_eo_class][0][0];
526  int pos_0_1 = pos[sao_eo_class][0][1];
527  int pos_1_0 = pos[sao_eo_class][1][0];
528  int pos_1_1 = pos[sao_eo_class][1][1];
529 
530  int y_stride_0_1 = (init_y + pos_0_1) * stride;
531  int y_stride_1_1 = (init_y + pos_1_1) * stride;
532  for (y = init_y; y < height; y++) {
533  for (x = init_x; x < width; x++) {
534  int diff0 = CMP(src[x + y_stride], src[x + pos_0_0 + y_stride_0_1]);
535  int diff1 = CMP(src[x + y_stride], src[x + pos_1_0 + y_stride_1_1]);
536  int offset_val = edge_idx[2 + diff0 + diff1];
537  dst[x + y_stride] = av_clip_pixel(src[x + y_stride] + sao_offset_val[offset_val]);
538  }
539  y_stride += stride;
540  y_stride_0_1 += stride;
541  y_stride_1_1 += stride;
542  }
543  }
544 
545  {
546  // Restore pixels that can't be modified
547  int save_lower_left = !diag_edge && sao_eo_class == SAO_EO_45D && !borders[0];
548  if(vert_edge && sao_eo_class != SAO_EO_VERT)
549  for(y = init_y; y< height-save_lower_left; y++)
550  dst[y*stride] = src[y*stride];
551  if(horiz_edge && sao_eo_class != SAO_EO_HORIZ)
552  for(x = init_x+save_lower_left; x<width; x++)
553  dst[(height-1)*stride+x] = src[(height-1)*stride+x];
554  if(diag_edge && sao_eo_class == SAO_EO_45D)
555  dst[stride*(height-1)] = src[stride*(height-1)];
556  }
557 
558 #undef CMP
559 }
560 
561 static void FUNC(sao_edge_filter_2)(uint8_t *_dst, uint8_t *_src,
562  ptrdiff_t stride, SAOParams *sao,
563  int *borders, int _width, int _height,
564  int c_idx, uint8_t vert_edge,
565  uint8_t horiz_edge, uint8_t diag_edge)
566 {
567  int x, y;
568  pixel *dst = (pixel *)_dst;
569  pixel *src = (pixel *)_src;
570  int chroma = !!c_idx;
571  int *sao_offset_val = sao->offset_val[c_idx];
572  int sao_eo_class = sao->eo_class[c_idx];
573  int init_x = 0, init_y = 0, width = _width, height = _height;
574 
575  static const int8_t pos[4][2][2] = {
576  { { -1, 0 }, { 1, 0 } }, // horizontal
577  { { 0, -1 }, { 0, 1 } }, // vertical
578  { { -1, -1 }, { 1, 1 } }, // 45 degree
579  { { 1, -1 }, { -1, 1 } }, // 135 degree
580  };
581  static const uint8_t edge_idx[] = { 1, 2, 0, 3, 4 };
582 
583 #define CMP(a, b) ((a) > (b) ? 1 : ((a) == (b) ? 0 : -1))
584 
585  stride /= sizeof(pixel);
586 
587  init_x = -(8 >> chroma) - 2;
588  width = (8 >> chroma) + 2;
589  if (!borders[3])
590  height -= (4 >> chroma) + 2;
591 
592  dst = dst + (init_y * stride + init_x);
593  src = src + (init_y * stride + init_x);
594  init_y = init_x = 0;
595  if (sao_eo_class != SAO_EO_HORIZ) {
596  if (borders[1]) {
597  int offset_val = sao_offset_val[0];
598  for (x = init_x; x < width; x++)
599  dst[x] = av_clip_pixel(src[x] + offset_val);
600  init_y = 1;
601  }
602  if (borders[3]) {
603  int offset_val = sao_offset_val[0];
604  int y_stride = stride * (height - 1);
605  for (x = init_x; x < width; x++)
606  dst[x + y_stride] = av_clip_pixel(src[x + y_stride] + offset_val);
607  height--;
608  }
609  }
610  {
611  int y_stride = init_y * stride;
612  int pos_0_0 = pos[sao_eo_class][0][0];
613  int pos_0_1 = pos[sao_eo_class][0][1];
614  int pos_1_0 = pos[sao_eo_class][1][0];
615  int pos_1_1 = pos[sao_eo_class][1][1];
616 
617  int y_stride_0_1 = (init_y + pos_0_1) * stride;
618  int y_stride_1_1 = (init_y + pos_1_1) * stride;
619  for (y = init_y; y < height; y++) {
620  for (x = init_x; x < width; x++) {
621  int diff0 = CMP(src[x + y_stride], src[x + pos_0_0 + y_stride_0_1]);
622  int diff1 = CMP(src[x + y_stride], src[x + pos_1_0 + y_stride_1_1]);
623  int offset_val = edge_idx[2 + diff0 + diff1];
624  dst[x + y_stride] = av_clip_pixel(src[x + y_stride] + sao_offset_val[offset_val]);
625  }
626  y_stride += stride;
627  y_stride_0_1 += stride;
628  y_stride_1_1 += stride;
629  }
630  }
631 
632  {
633  // Restore pixels that can't be modified
634  int save_upper_right = !diag_edge && sao_eo_class == SAO_EO_45D && !borders[1];
635  if(vert_edge && sao_eo_class != SAO_EO_VERT)
636  for(y = init_y+save_upper_right; y< height; y++)
637  dst[y*stride+width-1] = src[y*stride+width-1];
638  if(horiz_edge && sao_eo_class != SAO_EO_HORIZ)
639  for(x = init_x; x<width-save_upper_right; x++)
640  dst[x] = src[x];
641  if(diag_edge && sao_eo_class == SAO_EO_45D)
642  dst[width-1] = src[width-1];
643  }
644 #undef CMP
645 }
646 
647 static void FUNC(sao_edge_filter_3)(uint8_t *_dst, uint8_t *_src,
648  ptrdiff_t stride, SAOParams *sao,
649  int *borders, int _width, int _height,
650  int c_idx, uint8_t vert_edge,
651  uint8_t horiz_edge, uint8_t diag_edge)
652 {
653  int x, y;
654  pixel *dst = (pixel *)_dst;
655  pixel *src = (pixel *)_src;
656  int chroma = !!c_idx;
657  int *sao_offset_val = sao->offset_val[c_idx];
658  int sao_eo_class = sao->eo_class[c_idx];
659  int init_x = 0, init_y = 0, width = _width, height = _height;
660 
661  static const int8_t pos[4][2][2] = {
662  { { -1, 0 }, { 1, 0 } }, // horizontal
663  { { 0, -1 }, { 0, 1 } }, // vertical
664  { { -1, -1 }, { 1, 1 } }, // 45 degree
665  { { 1, -1 }, { -1, 1 } }, // 135 degree
666  };
667  static const uint8_t edge_idx[] = { 1, 2, 0, 3, 4 };
668 
669 #define CMP(a, b) ((a) > (b) ? 1 : ((a) == (b) ? 0 : -1))
670 
671  stride /= sizeof(pixel);
672 
673  init_y = -(4 >> chroma) - 2;
674  init_x = -(8 >> chroma) - 2;
675  width = (8 >> chroma) + 2;
676  height = (4 >> chroma) + 2;
677 
678 
679  dst = dst + (init_y * stride + init_x);
680  src = src + (init_y * stride + init_x);
681  init_y = init_x = 0;
682 
683  {
684  int y_stride = init_y * stride;
685  int pos_0_0 = pos[sao_eo_class][0][0];
686  int pos_0_1 = pos[sao_eo_class][0][1];
687  int pos_1_0 = pos[sao_eo_class][1][0];
688  int pos_1_1 = pos[sao_eo_class][1][1];
689 
690  int y_stride_0_1 = (init_y + pos_0_1) * stride;
691  int y_stride_1_1 = (init_y + pos_1_1) * stride;
692 
693  for (y = init_y; y < height; y++) {
694  for (x = init_x; x < width; x++) {
695  int diff0 = CMP(src[x + y_stride], src[x + pos_0_0 + y_stride_0_1]);
696  int diff1 = CMP(src[x + y_stride], src[x + pos_1_0 + y_stride_1_1]);
697  int offset_val = edge_idx[2 + diff0 + diff1];
698  dst[x + y_stride] = av_clip_pixel(src[x + y_stride] + sao_offset_val[offset_val]);
699  }
700  y_stride += stride;
701  y_stride_0_1 += stride;
702  y_stride_1_1 += stride;
703  }
704  }
705 
706  {
707  // Restore pixels that can't be modified
708  int save_lower_right = !diag_edge && sao_eo_class == SAO_EO_135D;
709  if(vert_edge && sao_eo_class != SAO_EO_VERT)
710  for(y = init_y; y< height-save_lower_right; y++)
711  dst[y*stride+width-1] = src[y*stride+width-1];
712  if(horiz_edge && sao_eo_class != SAO_EO_HORIZ)
713  for(x = init_x; x<width-save_lower_right; x++)
714  dst[(height-1)*stride+x] = src[(height-1)*stride+x];
715  if(diag_edge && sao_eo_class == SAO_EO_135D)
716  dst[stride*(height-1)+width-1] = src[stride*(height-1)+width-1];
717  }
718 #undef CMP
719 }
720 
721 #undef SET
722 #undef SCALE
723 #undef TR_4
724 #undef TR_8
725 #undef TR_16
726 #undef TR_32
727 
728 static av_always_inline void
729 FUNC(put_hevc_qpel_pixels)(int16_t *dst, ptrdiff_t dststride,
730  uint8_t *_src, ptrdiff_t _srcstride,
731  int width, int height, int mx, int my,
732  int16_t* mcbuffer)
733 {
734  int x, y;
735  pixel *src = (pixel *)_src;
736  ptrdiff_t srcstride = _srcstride / sizeof(pixel);
737 
738  dststride /= sizeof(*dst);
739  for (y = 0; y < height; y++) {
740  for (x = 0; x < width; x++)
741  dst[x] = src[x] << (14 - BIT_DEPTH);
742  src += srcstride;
743  dst += dststride;
744  }
745 }
746 
747 #define QPEL_FILTER_1(src, stride) \
748  (1 * -src[x - 3 * stride] + \
749  4 * src[x - 2 * stride] - \
750  10 * src[x - stride] + \
751  58 * src[x] + \
752  17 * src[x + stride] - \
753  5 * src[x + 2 * stride] + \
754  1 * src[x + 3 * stride])
755 
756 #define QPEL_FILTER_2(src, stride) \
757  (1 * -src[x - 3 * stride] + \
758  4 * src[x - 2 * stride] - \
759  11 * src[x - stride] + \
760  40 * src[x] + \
761  40 * src[x + stride] - \
762  11 * src[x + 2 * stride] + \
763  4 * src[x + 3 * stride] - \
764  1 * src[x + 4 * stride])
765 
766 #define QPEL_FILTER_3(src, stride) \
767  (1 * src[x - 2 * stride] - \
768  5 * src[x - stride] + \
769  17 * src[x] + \
770  58 * src[x + stride] - \
771  10 * src[x + 2 * stride] + \
772  4 * src[x + 3 * stride] - \
773  1 * src[x + 4 * stride])
774 
775 
776 #define PUT_HEVC_QPEL_H(H) \
777 static void FUNC(put_hevc_qpel_h ## H)(int16_t *dst, ptrdiff_t dststride, \
778  uint8_t *_src, ptrdiff_t _srcstride, \
779  int width, int height, \
780  int16_t* mcbuffer) \
781 { \
782  int x, y; \
783  pixel *src = (pixel*)_src; \
784  ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
785  \
786  dststride /= sizeof(*dst); \
787  for (y = 0; y < height; y++) { \
788  for (x = 0; x < width; x++) \
789  dst[x] = QPEL_FILTER_ ## H(src, 1) >> (BIT_DEPTH - 8); \
790  src += srcstride; \
791  dst += dststride; \
792  } \
793 }
794 
795 #define PUT_HEVC_QPEL_V(V) \
796 static void FUNC(put_hevc_qpel_v ## V)(int16_t *dst, ptrdiff_t dststride, \
797  uint8_t *_src, ptrdiff_t _srcstride, \
798  int width, int height, \
799  int16_t* mcbuffer) \
800 { \
801  int x, y; \
802  pixel *src = (pixel*)_src; \
803  ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
804  \
805  dststride /= sizeof(*dst); \
806  for (y = 0; y < height; y++) { \
807  for (x = 0; x < width; x++) \
808  dst[x] = QPEL_FILTER_ ## V(src, srcstride) >> (BIT_DEPTH - 8); \
809  src += srcstride; \
810  dst += dststride; \
811  } \
812 }
813 
814 #define PUT_HEVC_QPEL_HV(H, V) \
815 static void FUNC(put_hevc_qpel_h ## H ## v ## V)(int16_t *dst, \
816  ptrdiff_t dststride, \
817  uint8_t *_src, \
818  ptrdiff_t _srcstride, \
819  int width, int height, \
820  int16_t* mcbuffer) \
821 { \
822  int x, y; \
823  pixel *src = (pixel*)_src; \
824  ptrdiff_t srcstride = _srcstride / sizeof(pixel); \
825  \
826  int16_t tmp_array[(MAX_PB_SIZE + 7) * MAX_PB_SIZE]; \
827  int16_t *tmp = tmp_array; \
828  \
829  dststride /= sizeof(*dst); \
830  src -= ff_hevc_qpel_extra_before[V] * srcstride; \
831  \
832  for (y = 0; y < height + ff_hevc_qpel_extra[V]; y++) { \
833  for (x = 0; x < width; x++) \
834  tmp[x] = QPEL_FILTER_ ## H(src, 1) >> (BIT_DEPTH - 8); \
835  src += srcstride; \
836  tmp += MAX_PB_SIZE; \
837  } \
838  \
839  tmp = tmp_array + ff_hevc_qpel_extra_before[V] * MAX_PB_SIZE; \
840  \
841  for (y = 0; y < height; y++) { \
842  for (x = 0; x < width; x++) \
843  dst[x] = QPEL_FILTER_ ## V(tmp, MAX_PB_SIZE) >> 6; \
844  tmp += MAX_PB_SIZE; \
845  dst += dststride; \
846  } \
847 }
848 
855 PUT_HEVC_QPEL_HV(1, 1)
856 PUT_HEVC_QPEL_HV(1, 2)
857 PUT_HEVC_QPEL_HV(1, 3)
858 PUT_HEVC_QPEL_HV(2, 1)
859 PUT_HEVC_QPEL_HV(2, 2)
860 PUT_HEVC_QPEL_HV(2, 3)
861 PUT_HEVC_QPEL_HV(3, 1)
862 PUT_HEVC_QPEL_HV(3, 2)
863 PUT_HEVC_QPEL_HV(3, 3)
864 
865 #define QPEL(W) \
866 static void FUNC(put_hevc_qpel_pixels_ ## W)(int16_t *dst, ptrdiff_t dststride, \
867  uint8_t *src, ptrdiff_t srcstride, \
868  int height, int mx, int my, \
869  int16_t *mcbuffer) \
870 { \
871  FUNC(put_hevc_qpel_pixels)(dst, dststride, src, srcstride, W, height, \
872  mx, my, mcbuffer); \
873 } \
874  \
875 static void FUNC(put_hevc_qpel_h_ ## W)(int16_t *dst, ptrdiff_t dststride, \
876  uint8_t *src, ptrdiff_t srcstride, \
877  int height, int mx, int my, \
878  int16_t *mcbuffer) \
879 { \
880  if (mx == 1) \
881  FUNC(put_hevc_qpel_h1)(dst, dststride, src, srcstride, W, height, mcbuffer); \
882  else if (mx == 2) \
883  FUNC(put_hevc_qpel_h2)(dst, dststride, src, srcstride, W, height, mcbuffer); \
884  else \
885  FUNC(put_hevc_qpel_h3)(dst, dststride, src, srcstride, W, height, mcbuffer); \
886 } \
887  \
888 static void FUNC(put_hevc_qpel_v_ ## W)(int16_t *dst, ptrdiff_t dststride, \
889  uint8_t *src, ptrdiff_t srcstride, \
890  int height, int mx, int my, \
891  int16_t *mcbuffer) \
892 { \
893  if (my == 1) \
894  FUNC(put_hevc_qpel_v1)(dst, dststride, src, srcstride, W, height, mcbuffer); \
895  else if (my == 2) \
896  FUNC(put_hevc_qpel_v2)(dst, dststride, src, srcstride, W, height, mcbuffer); \
897  else \
898  FUNC(put_hevc_qpel_v3)(dst, dststride, src, srcstride, W, height, mcbuffer); \
899 } \
900  \
901 static void FUNC(put_hevc_qpel_hv_ ## W)(int16_t *dst, ptrdiff_t dststride, \
902  uint8_t *src, ptrdiff_t srcstride, \
903  int height, int mx, int my, \
904  int16_t *mcbuffer) \
905 { \
906  if (my == 1) { \
907  if (mx == 1) \
908  FUNC(put_hevc_qpel_h1v1)(dst, dststride, src, srcstride, W, height, mcbuffer); \
909  else if (mx == 2) \
910  FUNC(put_hevc_qpel_h2v1)(dst, dststride, src, srcstride, W, height, mcbuffer); \
911  else \
912  FUNC(put_hevc_qpel_h3v1)(dst, dststride, src, srcstride, W, height, mcbuffer); \
913  } else if (my == 2) { \
914  if (mx == 1) \
915  FUNC(put_hevc_qpel_h1v2)(dst, dststride, src, srcstride, W, height, mcbuffer); \
916  else if (mx == 2) \
917  FUNC(put_hevc_qpel_h2v2)(dst, dststride, src, srcstride, W, height, mcbuffer); \
918  else \
919  FUNC(put_hevc_qpel_h3v2)(dst, dststride, src, srcstride, W, height, mcbuffer); \
920  } else { \
921  if (mx == 1) \
922  FUNC(put_hevc_qpel_h1v3)(dst, dststride, src, srcstride, W, height, mcbuffer); \
923  else if (mx == 2) \
924  FUNC(put_hevc_qpel_h2v3)(dst, dststride, src, srcstride, W, height, mcbuffer); \
925  else \
926  FUNC(put_hevc_qpel_h3v3)(dst, dststride, src, srcstride, W, height, mcbuffer); \
927  } \
928 }
929 
930 QPEL(64)
931 QPEL(48)
932 QPEL(32)
933 QPEL(24)
934 QPEL(16)
935 QPEL(12)
936 QPEL(8)
937 QPEL(4)
938 
939 static inline void FUNC(put_hevc_epel_pixels)(int16_t *dst, ptrdiff_t dststride,
940  uint8_t *_src, ptrdiff_t _srcstride,
941  int width, int height, int mx, int my,
942  int16_t* mcbuffer)
943 {
944  int x, y;
945  pixel *src = (pixel *)_src;
946  ptrdiff_t srcstride = _srcstride / sizeof(pixel);
947 
948  dststride /= sizeof(*dst);
949  for (y = 0; y < height; y++) {
950  for (x = 0; x < width; x++)
951  dst[x] = src[x] << (14 - BIT_DEPTH);
952  src += srcstride;
953  dst += dststride;
954  }
955 }
956 
957 #define EPEL_FILTER(src, stride) \
958  (filter_0 * src[x - stride] + \
959  filter_1 * src[x] + \
960  filter_2 * src[x + stride] + \
961  filter_3 * src[x + 2 * stride])
962 
963 static inline void FUNC(put_hevc_epel_h)(int16_t *dst, ptrdiff_t dststride,
964  uint8_t *_src, ptrdiff_t _srcstride,
965  int width, int height, int mx, int my,
966  int16_t* mcbuffer)
967 {
968  int x, y;
969  pixel *src = (pixel *)_src;
970  ptrdiff_t srcstride = _srcstride / sizeof(pixel);
971  const int16_t *filter = ff_hevc_epel_coeffs[mx - 1];
972  int8_t filter_0 = filter[0];
973  int8_t filter_1 = filter[1];
974  int8_t filter_2 = filter[2];
975  int8_t filter_3 = filter[3];
976  dststride /= sizeof(*dst);
977  for (y = 0; y < height; y++) {
978  for (x = 0; x < width; x++)
979  dst[x] = EPEL_FILTER(src, 1) >> (BIT_DEPTH - 8);
980  src += srcstride;
981  dst += dststride;
982  }
983 }
984 
985 static inline void FUNC(put_hevc_epel_v)(int16_t *dst, ptrdiff_t dststride,
986  uint8_t *_src, ptrdiff_t _srcstride,
987  int width, int height, int mx, int my,
988  int16_t* mcbuffer)
989 {
990  int x, y;
991  pixel *src = (pixel *)_src;
992  ptrdiff_t srcstride = _srcstride / sizeof(pixel);
993  const int16_t *filter = ff_hevc_epel_coeffs[my - 1];
994  int8_t filter_0 = filter[0];
995  int8_t filter_1 = filter[1];
996  int8_t filter_2 = filter[2];
997  int8_t filter_3 = filter[3];
998 
999  dststride /= sizeof(*dst);
1000  for (y = 0; y < height; y++) {
1001  for (x = 0; x < width; x++)
1002  dst[x] = EPEL_FILTER(src, srcstride) >> (BIT_DEPTH - 8);
1003  src += srcstride;
1004  dst += dststride;
1005  }
1006 }
1007 
1008 static inline void FUNC(put_hevc_epel_hv)(int16_t *dst, ptrdiff_t dststride,
1009  uint8_t *_src, ptrdiff_t _srcstride,
1010  int width, int height, int mx, int my,
1011  int16_t* mcbuffer)
1012 {
1013  int x, y;
1014  pixel *src = (pixel *)_src;
1015  ptrdiff_t srcstride = _srcstride / sizeof(pixel);
1016  const int16_t *filter_h = ff_hevc_epel_coeffs[mx - 1];
1017  const int16_t *filter_v = ff_hevc_epel_coeffs[my - 1];
1018  int8_t filter_0 = filter_h[0];
1019  int8_t filter_1 = filter_h[1];
1020  int8_t filter_2 = filter_h[2];
1021  int8_t filter_3 = filter_h[3];
1022  int16_t tmp_array[(MAX_PB_SIZE + 3) * MAX_PB_SIZE];
1023  int16_t *tmp = tmp_array;
1024 
1025  dststride /= sizeof(*dst);
1026  src -= EPEL_EXTRA_BEFORE * srcstride;
1027 
1028  for (y = 0; y < height + EPEL_EXTRA; y++) {
1029  for (x = 0; x < width; x++)
1030  tmp[x] = EPEL_FILTER(src, 1) >> (BIT_DEPTH - 8);
1031  src += srcstride;
1032  tmp += MAX_PB_SIZE;
1033  }
1034 
1035  tmp = tmp_array + EPEL_EXTRA_BEFORE * MAX_PB_SIZE;
1036  filter_0 = filter_v[0];
1037  filter_1 = filter_v[1];
1038  filter_2 = filter_v[2];
1039  filter_3 = filter_v[3];
1040  for (y = 0; y < height; y++) {
1041  for (x = 0; x < width; x++)
1042  dst[x] = EPEL_FILTER(tmp, MAX_PB_SIZE) >> 6;
1043  tmp += MAX_PB_SIZE;
1044  dst += dststride;
1045  }
1046 }
1047 
1048 #define EPEL(W) \
1049 static void FUNC(put_hevc_epel_pixels_ ## W)(int16_t *dst, ptrdiff_t dststride, \
1050  uint8_t *src, ptrdiff_t srcstride, \
1051  int height, int mx, int my, \
1052  int16_t *mcbuffer) \
1053 { \
1054  FUNC(put_hevc_epel_pixels)(dst, dststride, src, srcstride, \
1055  W, height, mx, my, mcbuffer); \
1056 } \
1057 static void FUNC(put_hevc_epel_h_ ## W)(int16_t *dst, ptrdiff_t dststride, \
1058  uint8_t *src, ptrdiff_t srcstride, \
1059  int height, int mx, int my, \
1060  int16_t *mcbuffer) \
1061 { \
1062  FUNC(put_hevc_epel_h)(dst, dststride, src, srcstride, \
1063  W, height, mx, my, mcbuffer); \
1064 } \
1065 static void FUNC(put_hevc_epel_v_ ## W)(int16_t *dst, ptrdiff_t dststride, \
1066  uint8_t *src, ptrdiff_t srcstride, \
1067  int height, int mx, int my, \
1068  int16_t *mcbuffer) \
1069 { \
1070  FUNC(put_hevc_epel_v)(dst, dststride, src, srcstride, \
1071  W, height, mx, my, mcbuffer); \
1072 } \
1073 static void FUNC(put_hevc_epel_hv_ ## W)(int16_t *dst, ptrdiff_t dststride, \
1074  uint8_t *src, ptrdiff_t srcstride, \
1075  int height, int mx, int my, \
1076  int16_t *mcbuffer) \
1077 { \
1078  FUNC(put_hevc_epel_hv)(dst, dststride, src, srcstride, \
1079  W, height, mx, my, mcbuffer); \
1080 }
1081 
1082 EPEL(32)
1083 EPEL(24)
1084 EPEL(16)
1085 EPEL(12)
1086 EPEL(8)
1087 EPEL(6)
1088 EPEL(4)
1089 EPEL(2)
1090 
1091 static av_always_inline void
1092 FUNC(put_unweighted_pred)(uint8_t *_dst, ptrdiff_t _dststride,
1093  int16_t *src, ptrdiff_t srcstride,
1094  int width, int height)
1095 {
1096  int x, y;
1097  pixel *dst = (pixel *)_dst;
1098  ptrdiff_t dststride = _dststride / sizeof(pixel);
1099 
1100  int shift = 14 - BIT_DEPTH;
1101 #if BIT_DEPTH < 14
1102  int offset = 1 << (shift - 1);
1103 #else
1104  int offset = 0;
1105 #endif
1106  srcstride /= sizeof(*src);
1107  for (y = 0; y < height; y++) {
1108  for (x = 0; x < width; x++)
1109  dst[x] = av_clip_pixel((src[x] + offset) >> shift);
1110  dst += dststride;
1111  src += srcstride;
1112  }
1113 }
1114 
1115 static av_always_inline void
1116 FUNC(put_unweighted_pred_avg)(uint8_t *_dst, ptrdiff_t _dststride,
1117  int16_t *src1, int16_t *src2,
1118  ptrdiff_t srcstride,
1119  int width, int height)
1120 {
1121  int x, y;
1122  pixel *dst = (pixel *)_dst;
1123  ptrdiff_t dststride = _dststride / sizeof(pixel);
1124 
1125  int shift = 14 + 1 - BIT_DEPTH;
1126 #if BIT_DEPTH < 14
1127  int offset = 1 << (shift - 1);
1128 #else
1129  int offset = 0;
1130 #endif
1131 
1132  srcstride /= sizeof(*src1);
1133  for (y = 0; y < height; y++) {
1134  for (x = 0; x < width; x++)
1135  dst[x] = av_clip_pixel((src1[x] + src2[x] + offset) >> shift);
1136  dst += dststride;
1137  src1 += srcstride;
1138  src2 += srcstride;
1139  }
1140 }
1141 
1142 static av_always_inline void
1143 FUNC(weighted_pred)(uint8_t denom, int16_t wlxFlag, int16_t olxFlag,
1144  uint8_t *_dst, ptrdiff_t _dststride,
1145  int16_t *src, ptrdiff_t srcstride,
1146  int width, int height)
1147 {
1148  int shift, log2Wd, wx, ox, x, y, offset;
1149  pixel *dst = (pixel *)_dst;
1150  ptrdiff_t dststride = _dststride / sizeof(pixel);
1151 
1152  shift = 14 - BIT_DEPTH;
1153  log2Wd = denom + shift;
1154  offset = 1 << (log2Wd - 1);
1155  wx = wlxFlag;
1156  ox = olxFlag * (1 << (BIT_DEPTH - 8));
1157 
1158  srcstride /= sizeof(*src);
1159  for (y = 0; y < height; y++) {
1160  for (x = 0; x < width; x++) {
1161  if (log2Wd >= 1) {
1162  dst[x] = av_clip_pixel(((src[x] * wx + offset) >> log2Wd) + ox);
1163  } else {
1164  dst[x] = av_clip_pixel(src[x] * wx + ox);
1165  }
1166  }
1167  dst += dststride;
1168  src += srcstride;
1169  }
1170 }
1171 
1172 static av_always_inline void
1174  int16_t wl0Flag, int16_t wl1Flag,
1175  int16_t ol0Flag, int16_t ol1Flag,
1176  uint8_t *_dst, ptrdiff_t _dststride,
1177  int16_t *src1, int16_t *src2,
1178  ptrdiff_t srcstride,
1179  int width, int height)
1180 {
1181  int shift, log2Wd, w0, w1, o0, o1, x, y;
1182  pixel *dst = (pixel *)_dst;
1183  ptrdiff_t dststride = _dststride / sizeof(pixel);
1184 
1185  shift = 14 - BIT_DEPTH;
1186  log2Wd = denom + shift;
1187  w0 = wl0Flag;
1188  w1 = wl1Flag;
1189  o0 = ol0Flag * (1 << (BIT_DEPTH - 8));
1190  o1 = ol1Flag * (1 << (BIT_DEPTH - 8));
1191 
1192  srcstride /= sizeof(*src1);
1193  for (y = 0; y < height; y++) {
1194  for (x = 0; x < width; x++)
1195  dst[x] = av_clip_pixel((src1[x] * w0 + src2[x] * w1 +
1196  ((o0 + o1 + 1) << log2Wd)) >> (log2Wd + 1));
1197  dst += dststride;
1198  src1 += srcstride;
1199  src2 += srcstride;
1200  }
1201 }
1202 
1203 #define PUT_PRED(w) \
1204 static void FUNC(put_unweighted_pred_ ## w)(uint8_t *dst, ptrdiff_t dststride, \
1205  int16_t *src, ptrdiff_t srcstride, \
1206  int height) \
1207 { \
1208  FUNC(put_unweighted_pred)(dst, dststride, src, srcstride, w, height); \
1209 } \
1210 static void FUNC(put_unweighted_pred_avg_ ## w)(uint8_t *dst, ptrdiff_t dststride, \
1211  int16_t *src1, int16_t *src2, \
1212  ptrdiff_t srcstride, int height) \
1213 { \
1214  FUNC(put_unweighted_pred_avg)(dst, dststride, src1, src2, srcstride, w, height); \
1215 } \
1216 static void FUNC(put_weighted_pred_ ## w)(uint8_t denom, int16_t weight, int16_t offset, \
1217  uint8_t *dst, ptrdiff_t dststride, \
1218  int16_t *src, ptrdiff_t srcstride, int height) \
1219 { \
1220  FUNC(weighted_pred)(denom, weight, offset, \
1221  dst, dststride, src, srcstride, w, height); \
1222 } \
1223 static void FUNC(put_weighted_pred_avg_ ## w)(uint8_t denom, int16_t weight0, int16_t weight1, \
1224  int16_t offset0, int16_t offset1, \
1225  uint8_t *dst, ptrdiff_t dststride, \
1226  int16_t *src1, int16_t *src2, \
1227  ptrdiff_t srcstride, int height) \
1228 { \
1229  FUNC(weighted_pred_avg)(denom, weight0, weight1, offset0, offset1, \
1230  dst, dststride, src1, src2, srcstride, w, height); \
1231 }
1232 
1233 PUT_PRED(64)
1234 PUT_PRED(48)
1235 PUT_PRED(32)
1236 PUT_PRED(24)
1237 PUT_PRED(16)
1238 PUT_PRED(12)
1239 PUT_PRED(8)
1240 PUT_PRED(6)
1241 PUT_PRED(4)
1242 PUT_PRED(2)
1243 
1244 // line zero
1245 #define P3 pix[-4 * xstride]
1246 #define P2 pix[-3 * xstride]
1247 #define P1 pix[-2 * xstride]
1248 #define P0 pix[-1 * xstride]
1249 #define Q0 pix[0 * xstride]
1250 #define Q1 pix[1 * xstride]
1251 #define Q2 pix[2 * xstride]
1252 #define Q3 pix[3 * xstride]
1253 
1254 // line three. used only for deblocking decision
1255 #define TP3 pix[-4 * xstride + 3 * ystride]
1256 #define TP2 pix[-3 * xstride + 3 * ystride]
1257 #define TP1 pix[-2 * xstride + 3 * ystride]
1258 #define TP0 pix[-1 * xstride + 3 * ystride]
1259 #define TQ0 pix[0 * xstride + 3 * ystride]
1260 #define TQ1 pix[1 * xstride + 3 * ystride]
1261 #define TQ2 pix[2 * xstride + 3 * ystride]
1262 #define TQ3 pix[3 * xstride + 3 * ystride]
1263 
1265  ptrdiff_t _xstride, ptrdiff_t _ystride,
1266  int beta, int *_tc,
1267  uint8_t *_no_p, uint8_t *_no_q)
1268 {
1269  int d, j;
1270  pixel *pix = (pixel *)_pix;
1271  ptrdiff_t xstride = _xstride / sizeof(pixel);
1272  ptrdiff_t ystride = _ystride / sizeof(pixel);
1273 
1274  beta <<= BIT_DEPTH - 8;
1275 
1276  for (j = 0; j < 2; j++) {
1277  const int dp0 = abs(P2 - 2 * P1 + P0);
1278  const int dq0 = abs(Q2 - 2 * Q1 + Q0);
1279  const int dp3 = abs(TP2 - 2 * TP1 + TP0);
1280  const int dq3 = abs(TQ2 - 2 * TQ1 + TQ0);
1281  const int d0 = dp0 + dq0;
1282  const int d3 = dp3 + dq3;
1283  const int tc = _tc[j] << (BIT_DEPTH - 8);
1284  const int no_p = _no_p[j];
1285  const int no_q = _no_q[j];
1286 
1287  if (d0 + d3 >= beta) {
1288  pix += 4 * ystride;
1289  continue;
1290  } else {
1291  const int beta_3 = beta >> 3;
1292  const int beta_2 = beta >> 2;
1293  const int tc25 = ((tc * 5 + 1) >> 1);
1294 
1295  if (abs(P3 - P0) + abs(Q3 - Q0) < beta_3 && abs(P0 - Q0) < tc25 &&
1296  abs(TP3 - TP0) + abs(TQ3 - TQ0) < beta_3 && abs(TP0 - TQ0) < tc25 &&
1297  (d0 << 1) < beta_2 && (d3 << 1) < beta_2) {
1298  // strong filtering
1299  const int tc2 = tc << 1;
1300  for (d = 0; d < 4; d++) {
1301  const int p3 = P3;
1302  const int p2 = P2;
1303  const int p1 = P1;
1304  const int p0 = P0;
1305  const int q0 = Q0;
1306  const int q1 = Q1;
1307  const int q2 = Q2;
1308  const int q3 = Q3;
1309  if (!no_p) {
1310  P0 = p0 + av_clip(((p2 + 2 * p1 + 2 * p0 + 2 * q0 + q1 + 4) >> 3) - p0, -tc2, tc2);
1311  P1 = p1 + av_clip(((p2 + p1 + p0 + q0 + 2) >> 2) - p1, -tc2, tc2);
1312  P2 = p2 + av_clip(((2 * p3 + 3 * p2 + p1 + p0 + q0 + 4) >> 3) - p2, -tc2, tc2);
1313  }
1314  if (!no_q) {
1315  Q0 = q0 + av_clip(((p1 + 2 * p0 + 2 * q0 + 2 * q1 + q2 + 4) >> 3) - q0, -tc2, tc2);
1316  Q1 = q1 + av_clip(((p0 + q0 + q1 + q2 + 2) >> 2) - q1, -tc2, tc2);
1317  Q2 = q2 + av_clip(((2 * q3 + 3 * q2 + q1 + q0 + p0 + 4) >> 3) - q2, -tc2, tc2);
1318  }
1319  pix += ystride;
1320  }
1321  } else { // normal filtering
1322  int nd_p = 1;
1323  int nd_q = 1;
1324  const int tc_2 = tc >> 1;
1325  if (dp0 + dp3 < ((beta + (beta >> 1)) >> 3))
1326  nd_p = 2;
1327  if (dq0 + dq3 < ((beta + (beta >> 1)) >> 3))
1328  nd_q = 2;
1329 
1330  for (d = 0; d < 4; d++) {
1331  const int p2 = P2;
1332  const int p1 = P1;
1333  const int p0 = P0;
1334  const int q0 = Q0;
1335  const int q1 = Q1;
1336  const int q2 = Q2;
1337  int delta0 = (9 * (q0 - p0) - 3 * (q1 - p1) + 8) >> 4;
1338  if (abs(delta0) < 10 * tc) {
1339  delta0 = av_clip(delta0, -tc, tc);
1340  if (!no_p)
1341  P0 = av_clip_pixel(p0 + delta0);
1342  if (!no_q)
1343  Q0 = av_clip_pixel(q0 - delta0);
1344  if (!no_p && nd_p > 1) {
1345  const int deltap1 = av_clip((((p2 + p0 + 1) >> 1) - p1 + delta0) >> 1, -tc_2, tc_2);
1346  P1 = av_clip_pixel(p1 + deltap1);
1347  }
1348  if (!no_q && nd_q > 1) {
1349  const int deltaq1 = av_clip((((q2 + q0 + 1) >> 1) - q1 - delta0) >> 1, -tc_2, tc_2);
1350  Q1 = av_clip_pixel(q1 + deltaq1);
1351  }
1352  }
1353  pix += ystride;
1354  }
1355  }
1356  }
1357  }
1358 }
1359 
1360 static void FUNC(hevc_loop_filter_chroma)(uint8_t *_pix, ptrdiff_t _xstride,
1361  ptrdiff_t _ystride, int *_tc,
1362  uint8_t *_no_p, uint8_t *_no_q)
1363 {
1364  int d, j, no_p, no_q;
1365  pixel *pix = (pixel *)_pix;
1366  ptrdiff_t xstride = _xstride / sizeof(pixel);
1367  ptrdiff_t ystride = _ystride / sizeof(pixel);
1368 
1369  for (j = 0; j < 2; j++) {
1370  const int tc = _tc[j] << (BIT_DEPTH - 8);
1371  if (tc <= 0) {
1372  pix += 4 * ystride;
1373  continue;
1374  }
1375  no_p = _no_p[j];
1376  no_q = _no_q[j];
1377 
1378  for (d = 0; d < 4; d++) {
1379  int delta0;
1380  const int p1 = P1;
1381  const int p0 = P0;
1382  const int q0 = Q0;
1383  const int q1 = Q1;
1384  delta0 = av_clip((((q0 - p0) * 4) + p1 - q1 + 4) >> 3, -tc, tc);
1385  if (!no_p)
1386  P0 = av_clip_pixel(p0 + delta0);
1387  if (!no_q)
1388  Q0 = av_clip_pixel(q0 - delta0);
1389  pix += ystride;
1390  }
1391  }
1392 }
1393 
1394 static void FUNC(hevc_h_loop_filter_chroma)(uint8_t *pix, ptrdiff_t stride,
1395  int *tc, uint8_t *no_p,
1396  uint8_t *no_q)
1397 {
1398  FUNC(hevc_loop_filter_chroma)(pix, stride, sizeof(pixel), tc, no_p, no_q);
1399 }
1400 
1401 static void FUNC(hevc_v_loop_filter_chroma)(uint8_t *pix, ptrdiff_t stride,
1402  int *tc, uint8_t *no_p,
1403  uint8_t *no_q)
1404 {
1405  FUNC(hevc_loop_filter_chroma)(pix, sizeof(pixel), stride, tc, no_p, no_q);
1406 }
1407 
1408 static void FUNC(hevc_h_loop_filter_luma)(uint8_t *pix, ptrdiff_t stride,
1409  int beta, int *tc, uint8_t *no_p,
1410  uint8_t *no_q)
1411 {
1412  FUNC(hevc_loop_filter_luma)(pix, stride, sizeof(pixel),
1413  beta, tc, no_p, no_q);
1414 }
1415 
1416 static void FUNC(hevc_v_loop_filter_luma)(uint8_t *pix, ptrdiff_t stride,
1417  int beta, int *tc, uint8_t *no_p,
1418  uint8_t *no_q)
1419 {
1420  FUNC(hevc_loop_filter_luma)(pix, sizeof(pixel), stride,
1421  beta, tc, no_p, no_q);
1422 }
1423 
1424 #undef P3
1425 #undef P2
1426 #undef P1
1427 #undef P0
1428 #undef Q0
1429 #undef Q1
1430 #undef Q2
1431 #undef Q3
1432 
1433 #undef TP3
1434 #undef TP2
1435 #undef TP1
1436 #undef TP0
1437 #undef TQ0
1438 #undef TQ1
1439 #undef TQ2
1440 #undef TQ3
static void FUNC() add_residual16x16(uint8_t *_dst, int16_t *res, ptrdiff_t stride)
int size
static void FUNC() hevc_h_loop_filter_luma(uint8_t *pix, ptrdiff_t stride, int beta, int *tc, uint8_t *no_p, uint8_t *no_q)
static void FUNC() sao_edge_filter_3(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride, SAOParams *sao, int *borders, int _width, int _height, int c_idx, uint8_t vert_edge, uint8_t horiz_edge, uint8_t diag_edge)
#define IDCT_DC(H)
static unsigned int get_bits(GetBitContext *s, int n)
Read 1-25 bits.
Definition: get_bits.h:228
static void FUNC() put_hevc_epel_h(int16_t *dst, ptrdiff_t dststride, uint8_t *_src, ptrdiff_t _srcstride, int width, int height, int mx, int my, int16_t *mcbuffer)
#define P1
#define PUT_HEVC_QPEL_HV(H, V)
static av_always_inline void FUNC() add_residual(uint8_t *_dst, int16_t *res, ptrdiff_t stride, int size)
#define Q0
#define BIT_DEPTH
static void FUNC() sao_edge_filter_1(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride, SAOParams *sao, int *borders, int _width, int _height, int c_idx, uint8_t vert_edge, uint8_t horiz_edge, uint8_t diag_edge)
static void FUNC() sao_band_filter_1(uint8_t *dst, uint8_t *src, ptrdiff_t stride, SAOParams *sao, int *borders, int width, int height, int c_idx)
#define MAX_PB_SIZE
Definition: hevc.h:59
static void FUNC() add_residual4x4(uint8_t *_dst, int16_t *res, ptrdiff_t stride)
int stride
Definition: mace.c:144
static void FUNC() add_residual32x32(uint8_t *_dst, int16_t *res, ptrdiff_t stride)
#define av_clip_pixel(a)
#define Q3
#define TQ1
#define P0
uint8_t
const int16_t ff_hevc_epel_coeffs[7][16]
Definition: hevcdsp.c:92
#define TQ2
static av_always_inline void FUNC() weighted_pred(uint8_t denom, int16_t wlxFlag, int16_t olxFlag, uint8_t *_dst, ptrdiff_t _dststride, int16_t *src, ptrdiff_t srcstride, int width, int height)
bitstream reader API header.
#define EPEL_EXTRA_BEFORE
Definition: hevc.h:71
#define EPEL(W)
#define TP3
static void FUNC() hevc_v_loop_filter_chroma(uint8_t *pix, ptrdiff_t stride, int *tc, uint8_t *no_p, uint8_t *no_q)
#define src
Definition: vp8dsp.c:254
#define P3
#define FUNC(a)
static av_always_inline void FUNC() put_unweighted_pred(uint8_t *_dst, ptrdiff_t _dststride, int16_t *src, ptrdiff_t srcstride, int width, int height)
static void FUNC() put_pcm(uint8_t *_dst, ptrdiff_t stride, int size, GetBitContext *gb, int pcm_bit_depth)
static void FUNC() hevc_v_loop_filter_luma(uint8_t *pix, ptrdiff_t stride, int beta, int *tc, uint8_t *no_p, uint8_t *no_q)
static av_always_inline void FUNC() weighted_pred_avg(uint8_t denom, int16_t wl0Flag, int16_t wl1Flag, int16_t ol0Flag, int16_t ol1Flag, uint8_t *_dst, ptrdiff_t _dststride, int16_t *src1, int16_t *src2, ptrdiff_t srcstride, int width, int height)
#define PUT_PRED(w)
static void FUNC() sao_band_filter(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride, SAOParams *sao, int *borders, int width, int height, int c_idx, int class)
#define TP2
static void FUNC() add_residual8x8(uint8_t *_dst, int16_t *res, ptrdiff_t stride)
#define EPEL_FILTER(src, stride)
static void filter(MpegAudioContext *s, int ch, const short *samples, int incr)
Definition: mpegaudioenc.c:307
static av_always_inline void FUNC() put_hevc_qpel_pixels(int16_t *dst, ptrdiff_t dststride, uint8_t *_src, ptrdiff_t _srcstride, int width, int height, int mx, int my, int16_t *mcbuffer)
#define QPEL(W)
#define IDCT(H)
#define SCALE(dst, x)
#define pixel
int offset_val[3][5]
SaoOffsetVal.
Definition: hevcdsp.h:36
static int width
Definition: utils.c:156
#define src1
Definition: h264pred.c:139
static void FUNC() put_hevc_epel_v(int16_t *dst, ptrdiff_t dststride, uint8_t *_src, ptrdiff_t _srcstride, int width, int height, int mx, int my, int16_t *mcbuffer)
#define Q1
#define TP0
#define PUT_HEVC_QPEL_H(H)
static void FUNC() hevc_loop_filter_chroma(uint8_t *_pix, ptrdiff_t _xstride, ptrdiff_t _ystride, int *_tc, uint8_t *_no_p, uint8_t *_no_q)
static void FUNC() sao_edge_filter_2(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride, SAOParams *sao, int *borders, int _width, int _height, int c_idx, uint8_t vert_edge, uint8_t horiz_edge, uint8_t diag_edge)
#define EPEL_EXTRA
Definition: hevc.h:73
static void FUNC() put_hevc_epel_hv(int16_t *dst, ptrdiff_t dststride, uint8_t *_src, ptrdiff_t _srcstride, int width, int height, int mx, int my, int16_t *mcbuffer)
static void FUNC() sao_band_filter_3(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride, SAOParams *sao, int *borders, int width, int height, int c_idx)
static void FUNC() sao_band_filter_0(uint8_t *dst, uint8_t *src, ptrdiff_t stride, SAOParams *sao, int *borders, int width, int height, int c_idx)
static av_always_inline void FUNC() put_unweighted_pred_avg(uint8_t *_dst, ptrdiff_t _dststride, int16_t *src1, int16_t *src2, ptrdiff_t srcstride, int width, int height)
int height
Definition: gxfenc.c:72
#define CMP(a, b)
#define PUT_HEVC_QPEL_V(V)
#define TP1
static void FUNC() sao_band_filter_2(uint8_t *dst, uint8_t *src, ptrdiff_t stride, SAOParams *sao, int *borders, int width, int height, int c_idx)
#define TR_4x4_LUMA(dst, src, step, assign)
static void FUNC() put_hevc_epel_pixels(int16_t *dst, ptrdiff_t dststride, uint8_t *_src, ptrdiff_t _srcstride, int width, int height, int mx, int my, int16_t *mcbuffer)
#define P2
static const uint8_t offset_table[]
Definition: escape130.c:41
#define TQ3
static uint8_t tmp[8]
Definition: des.c:38
static void FUNC() transform_4x4_luma(int16_t *coeffs)
static void FUNC() hevc_h_loop_filter_chroma(uint8_t *pix, ptrdiff_t stride, int *tc, uint8_t *no_p, uint8_t *no_q)
static void FUNC() hevc_loop_filter_luma(uint8_t *_pix, ptrdiff_t _xstride, ptrdiff_t _ystride, int beta, int *_tc, uint8_t *_no_p, uint8_t *_no_q)
static void FUNC() dequant(int16_t *coeffs)
static void FUNC() sao_edge_filter_0(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride, SAOParams *sao, int *borders, int _width, int _height, int c_idx, uint8_t vert_edge, uint8_t horiz_edge, uint8_t diag_edge)
#define av_always_inline
Definition: attributes.h:40
#define TQ0
#define Q2