FFmpeg
filter_template.c
Go to the documentation of this file.
1 /*
2  * VVC filters DSP
3  *
4  * Copyright (C) 2022 Nuo Mi
5  *
6  * This file is part of FFmpeg.
7  *
8  * FFmpeg is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * FFmpeg is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with FFmpeg; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
24 
25 static void FUNC(lmcs_filter_luma)(uint8_t *_dst, ptrdiff_t dst_stride, const int width, const int height, const void *_lut)
26 {
27  const pixel *lut = _lut;
28  pixel *dst = (pixel*)_dst;
29  dst_stride /= sizeof(pixel);
30 
31  for (int y = 0; y < height; y++) {
32  for (int x = 0; x < width; x++)
33  dst[x] = lut[dst[x]];
34  dst += dst_stride;
35  }
36 }
37 
38 static av_always_inline int16_t FUNC(alf_clip)(pixel curr, pixel v0, pixel v1, int16_t clip)
39 {
40  return av_clip(v0 - curr, -clip, clip) + av_clip(v1 - curr, -clip, clip);
41 }
42 
43 static void FUNC(alf_filter_luma)(uint8_t *_dst, ptrdiff_t dst_stride, const uint8_t *_src, ptrdiff_t src_stride,
44  const int width, const int height, const int16_t *filter, const int16_t *clip, const int vb_pos)
45 {
46  const pixel *src = (pixel *)_src;
47  const int shift = 7;
48  const int offset = 1 << ( shift - 1 );
49  const int vb_above = vb_pos - 4;
50  const int vb_below = vb_pos + 3;
51 
52  dst_stride /= sizeof(pixel);
53  src_stride /= sizeof(pixel);
54 
55  for (int y = 0; y < height; y += ALF_BLOCK_SIZE) {
56  for (int x = 0; x < width; x += ALF_BLOCK_SIZE) {
57  const pixel *s0 = src + y * src_stride + x;
58  const pixel *s1 = s0 + src_stride;
59  const pixel *s2 = s0 - src_stride;
60  const pixel *s3 = s1 + src_stride;
61  const pixel *s4 = s2 - src_stride;
62  const pixel *s5 = s3 + src_stride;
63  const pixel *s6 = s4 - src_stride;
64 
65  for (int i = 0; i < ALF_BLOCK_SIZE; i++) {
66  pixel *dst = (pixel *)_dst + (y + i) * dst_stride + x;
67 
68  const pixel *p0 = s0 + i * src_stride;
69  const pixel *p1 = s1 + i * src_stride;
70  const pixel *p2 = s2 + i * src_stride;
71  const pixel *p3 = s3 + i * src_stride;
72  const pixel *p4 = s4 + i * src_stride;
73  const pixel *p5 = s5 + i * src_stride;
74  const pixel *p6 = s6 + i * src_stride;
75 
76  const int is_near_vb_above = (y + i < vb_pos) && (y + i >= vb_pos - 1);
77  const int is_near_vb_below = (y + i >= vb_pos) && (y + i <= vb_pos);
78  const int is_near_vb = is_near_vb_above || is_near_vb_below;
79 
80  if ((y + i < vb_pos) && ((y + i) > vb_above)) {
81  p1 = (y + i == vb_pos - 1) ? p0 : p1;
82  p3 = (y + i >= vb_pos - 2) ? p1 : p3;
83  p5 = (y + i >= vb_pos - 3) ? p3 : p5;
84 
85  p2 = (y + i == vb_pos - 1) ? p0 : p2;
86  p4 = (y + i >= vb_pos - 2) ? p2 : p4;
87  p6 = (y + i >= vb_pos - 3) ? p4 : p6;
88  } else if ((y + i >= vb_pos) && ((y + i) < vb_below)) {
89  p2 = (y + i == vb_pos ) ? p0 : p2;
90  p4 = (y + i <= vb_pos + 1) ? p2 : p4;
91  p6 = (y + i <= vb_pos + 2) ? p4 : p6;
92 
93  p1 = (y + i == vb_pos ) ? p0 : p1;
94  p3 = (y + i <= vb_pos + 1) ? p1 : p3;
95  p5 = (y + i <= vb_pos + 2) ? p3 : p5;
96  }
97 
98  for (int j = 0; j < ALF_BLOCK_SIZE; j++) {
99  int sum = 0;
100  const pixel curr = *p0;
101 
102  sum += filter[0] * FUNC(alf_clip)(curr, p5[+0], p6[+0], clip[0]);
103  sum += filter[1] * FUNC(alf_clip)(curr, p3[+1], p4[-1], clip[1]);
104  sum += filter[2] * FUNC(alf_clip)(curr, p3[+0], p4[+0], clip[2]);
105  sum += filter[3] * FUNC(alf_clip)(curr, p3[-1], p4[+1], clip[3]);
106  sum += filter[4] * FUNC(alf_clip)(curr, p1[+2], p2[-2], clip[4]);
107  sum += filter[5] * FUNC(alf_clip)(curr, p1[+1], p2[-1], clip[5]);
108  sum += filter[6] * FUNC(alf_clip)(curr, p1[+0], p2[+0], clip[6]);
109  sum += filter[7] * FUNC(alf_clip)(curr, p1[-1], p2[+1], clip[7]);
110  sum += filter[8] * FUNC(alf_clip)(curr, p1[-2], p2[+2], clip[8]);
111  sum += filter[9] * FUNC(alf_clip)(curr, p0[+3], p0[-3], clip[9]);
112  sum += filter[10] * FUNC(alf_clip)(curr, p0[+2], p0[-2], clip[10]);
113  sum += filter[11] * FUNC(alf_clip)(curr, p0[+1], p0[-1], clip[11]);
114 
115  if (!is_near_vb)
116  sum = (sum + offset) >> shift;
117  else
118  sum = (sum + (1 << ((shift + 3) - 1))) >> (shift + 3);
119  sum += curr;
120  dst[j] = CLIP(sum);
121 
122  p0++;
123  p1++;
124  p2++;
125  p3++;
126  p4++;
127  p5++;
128  p6++;
129  }
130  }
133  }
134  }
135 }
136 
137 static void FUNC(alf_filter_chroma)(uint8_t* _dst, ptrdiff_t dst_stride, const uint8_t* _src, ptrdiff_t src_stride,
138  const int width, const int height, const int16_t* filter, const int16_t* clip, const int vb_pos)
139 {
140  const pixel *src = (pixel *)_src;
141  const int shift = 7;
142  const int offset = 1 << ( shift - 1 );
143  const int vb_above = vb_pos - 2;
144  const int vb_below = vb_pos + 1;
145 
146  dst_stride /= sizeof(pixel);
147  src_stride /= sizeof(pixel);
148 
149  for (int y = 0; y < height; y += ALF_BLOCK_SIZE) {
150  for (int x = 0; x < width; x += ALF_BLOCK_SIZE) {
151  const pixel *s0 = src + y * src_stride + x;
152  const pixel *s1 = s0 + src_stride;
153  const pixel *s2 = s0 - src_stride;
154  const pixel *s3 = s1 + src_stride;
155  const pixel *s4 = s2 - src_stride;
156 
157  for (int i = 0; i < ALF_BLOCK_SIZE; i++) {
158  pixel *dst = (pixel *)_dst + (y + i) * dst_stride + x;
159 
160  const pixel *p0 = s0 + i * src_stride;
161  const pixel *p1 = s1 + i * src_stride;
162  const pixel *p2 = s2 + i * src_stride;
163  const pixel *p3 = s3 + i * src_stride;
164  const pixel *p4 = s4 + i * src_stride;
165 
166  const int is_near_vb_above = (y + i < vb_pos) && (y + i >= vb_pos - 1);
167  const int is_near_vb_below = (y + i >= vb_pos) && (y + i <= vb_pos);
168  const int is_near_vb = is_near_vb_above || is_near_vb_below;
169 
170  if ((y + i < vb_pos) && ((y + i) >= vb_above)) {
171  p1 = (y + i == vb_pos - 1) ? p0 : p1;
172  p3 = (y + i >= vb_pos - 2) ? p1 : p3;
173 
174  p2 = (y + i == vb_pos - 1) ? p0 : p2;
175  p4 = (y + i >= vb_pos - 2) ? p2 : p4;
176  } else if ((y + i >= vb_pos) && ((y + i) <= vb_below)) {
177  p2 = (y + i == vb_pos ) ? p0 : p2;
178  p4 = (y + i <= vb_pos + 1) ? p2 : p4;
179 
180  p1 = (y + i == vb_pos ) ? p0 : p1;
181  p3 = (y + i <= vb_pos + 1) ? p1 : p3;
182  }
183 
184  for (int j = 0; j < ALF_BLOCK_SIZE; j++) {
185  int sum = 0;
186  const pixel curr = *p0;
187 
188  sum += filter[0] * FUNC(alf_clip)(curr, p3[+0], p4[+0], clip[0]);
189  sum += filter[1] * FUNC(alf_clip)(curr, p1[+1], p2[-1], clip[1]);
190  sum += filter[2] * FUNC(alf_clip)(curr, p1[+0], p2[+0], clip[2]);
191  sum += filter[3] * FUNC(alf_clip)(curr, p1[-1], p2[+1], clip[3]);
192  sum += filter[4] * FUNC(alf_clip)(curr, p0[+2], p0[-2], clip[4]);
193  sum += filter[5] * FUNC(alf_clip)(curr, p0[+1], p0[-1], clip[5]);
194 
195  if (!is_near_vb)
196  sum = (sum + offset) >> shift;
197  else
198  sum = (sum + (1 << ((shift + 3) - 1))) >> (shift + 3);
199  sum += curr;
200  dst[j] = CLIP(sum);
201 
202  p0++;
203  p1++;
204  p2++;
205  p3++;
206  p4++;
207  }
208  }
209  }
210  }
211 }
212 
213 static void FUNC(alf_filter_cc)(uint8_t *_dst, ptrdiff_t dst_stride, const uint8_t *_luma, const ptrdiff_t luma_stride,
214  const int width, const int height, const int hs, const int vs, const int16_t *filter, const int vb_pos)
215 {
216  const ptrdiff_t stride = luma_stride / sizeof(pixel);
217 
218  dst_stride /= sizeof(pixel);
219 
220  for (int y = 0; y < height; y++) {
221  for (int x = 0; x < width; x++) {
222  int sum = 0;
223  pixel *dst = (pixel *)_dst + y * dst_stride + x;
224  const pixel *src = (pixel *)_luma + (y << vs) * stride + (x << hs);
225 
226  const pixel *s0 = src - stride;
227  const pixel *s1 = src;
228  const pixel *s2 = src + stride;
229  const pixel *s3 = src + 2 * stride;
230 
231  const int pos = y << vs;
232  if (!vs && (pos == vb_pos || pos == vb_pos + 1))
233  continue;
234 
235  if (pos == (vb_pos - 2) || pos == (vb_pos + 1))
236  s3 = s2;
237  else if (pos == (vb_pos - 1) || pos == vb_pos)
238  s3 = s2 = s0 = s1;
239 
240 
241  sum += filter[0] * (*s0 - *src);
242  sum += filter[1] * (*(s1 - 1) - *src);
243  sum += filter[2] * (*(s1 + 1) - *src);
244  sum += filter[3] * (*(s2 - 1) - *src);
245  sum += filter[4] * (*s2 - *src);
246  sum += filter[5] * (*(s2 + 1) - *src);
247  sum += filter[6] * (*s3 - *src);
248  sum = av_clip((sum + 64) >> 7, -(1 << (BIT_DEPTH - 1)), (1 << (BIT_DEPTH - 1)) - 1);
249  sum += *dst;
250  *dst = av_clip_pixel(sum);
251  }
252  }
253 }
254 
255 #define ALF_DIR_VERT 0
256 #define ALF_DIR_HORZ 1
257 #define ALF_DIR_DIGA0 2
258 #define ALF_DIR_DIGA1 3
259 
260 static void FUNC(alf_get_idx)(int *class_idx, int *transpose_idx, const int *sum, const int ac)
261 {
262  static const int arg_var[] = {0, 1, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 4 };
263 
264  int hv0, hv1, dir_hv, d0, d1, dir_d, hvd1, hvd0, sum_hv, dir1;
265 
266  dir_hv = sum[ALF_DIR_VERT] <= sum[ALF_DIR_HORZ];
267  hv1 = FFMAX(sum[ALF_DIR_VERT], sum[ALF_DIR_HORZ]);
268  hv0 = FFMIN(sum[ALF_DIR_VERT], sum[ALF_DIR_HORZ]);
269 
270  dir_d = sum[ALF_DIR_DIGA0] <= sum[ALF_DIR_DIGA1];
271  d1 = FFMAX(sum[ALF_DIR_DIGA0], sum[ALF_DIR_DIGA1]);
272  d0 = FFMIN(sum[ALF_DIR_DIGA0], sum[ALF_DIR_DIGA1]);
273 
274  //promote to avoid overflow
275  dir1 = (uint64_t)d1 * hv0 <= (uint64_t)hv1 * d0;
276  hvd1 = dir1 ? hv1 : d1;
277  hvd0 = dir1 ? hv0 : d0;
278 
279  sum_hv = sum[ALF_DIR_HORZ] + sum[ALF_DIR_VERT];
280  *class_idx = arg_var[av_clip_uintp2(sum_hv * ac >> (BIT_DEPTH - 1), 4)];
281  if (hvd1 * 2 > 9 * hvd0)
282  *class_idx += ((dir1 << 1) + 2) * 5;
283  else if (hvd1 > 2 * hvd0)
284  *class_idx += ((dir1 << 1) + 1) * 5;
285 
286  *transpose_idx = dir_d * 2 + dir_hv;
287 }
288 
289 static void FUNC(alf_classify)(int *class_idx, int *transpose_idx,
290  const uint8_t *_src, const ptrdiff_t _src_stride, const int width, const int height,
291  const int vb_pos, int *gradient_tmp)
292 {
293  int *grad;
294 
295  const int h = height + ALF_GRADIENT_BORDER * 2;
296  const int w = width + ALF_GRADIENT_BORDER * 2;
298  const int gstride = (w / ALF_GRADIENT_STEP) * ALF_NUM_DIR;
299 
300  const pixel *src = (const pixel *)_src;
301  const ptrdiff_t src_stride = _src_stride / sizeof(pixel);
302  src -= (ALF_GRADIENT_BORDER + 1) * src_stride + ALF_GRADIENT_BORDER;
303 
304  grad = gradient_tmp;
305  for (int y = 0; y < h; y += ALF_GRADIENT_STEP) {
306  const pixel *s0 = src + y * src_stride;
307  const pixel *s1 = s0 + src_stride;
308  const pixel *s2 = s1 + src_stride;
309  const pixel *s3 = s2 + src_stride;
310 
311  if (y == vb_pos) //above
312  s3 = s2;
313  else if (y == vb_pos + ALF_GRADIENT_BORDER)
314  s0 = s1;
315 
316  for (int x = 0; x < w; x += ALF_GRADIENT_STEP) {
317  //two points a time
318  const pixel *a0 = s0 + x;
319  const pixel *p0 = s1 + x;
320  const pixel *b0 = s2 + x;
321  const int val0 = (*p0) << 1;
322 
323  const pixel *a1 = s1 + x + 1;
324  const pixel *p1 = s2 + x + 1;
325  const pixel *b1 = s3 + x + 1;
326  const int val1 = (*p1) << 1;
327 
328  grad[ALF_DIR_VERT] = FFABS(val0 - *a0 - *b0) + FFABS(val1 - *a1 - *b1);
329  grad[ALF_DIR_HORZ] = FFABS(val0 - *(p0 - 1) - *(p0 + 1)) + FFABS(val1 - *(p1 - 1) - *(p1 + 1));
330  grad[ALF_DIR_DIGA0] = FFABS(val0 - *(a0 - 1) - *(b0 + 1)) + FFABS(val1 - *(a1 - 1) - *(b1 + 1));
331  grad[ALF_DIR_DIGA1] = FFABS(val0 - *(a0 + 1) - *(b0 - 1)) + FFABS(val1 - *(a1 + 1) - *(b1 - 1));
332  grad += ALF_NUM_DIR;
333  }
334  }
335 
336  for (int y = 0; y < height ; y += ALF_BLOCK_SIZE ) {
337  int start = 0;
339  int ac = 2;
340  if (y + ALF_BLOCK_SIZE == vb_pos) {
342  ac = 3;
343  } else if (y == vb_pos) {
345  ac = 3;
346  }
347  for (int x = 0; x < width; x += ALF_BLOCK_SIZE) {
348  const int xg = x / ALF_GRADIENT_STEP;
349  const int yg = y / ALF_GRADIENT_STEP;
350  int sum[ALF_NUM_DIR] = { 0 };
351 
352  grad = gradient_tmp + (yg + start) * gstride + xg * ALF_NUM_DIR;
353  //todo: optimize this loop
354  for (int i = start; i < end; i++) {
355  for (int j = 0; j < size; j++) {
356  sum[ALF_DIR_VERT] += grad[ALF_DIR_VERT];
357  sum[ALF_DIR_HORZ] += grad[ALF_DIR_HORZ];
360  grad += ALF_NUM_DIR;
361  }
362  grad += gstride - size * ALF_NUM_DIR;
363  }
364  FUNC(alf_get_idx)(class_idx, transpose_idx, sum, ac);
365 
366  class_idx++;
367  transpose_idx++;
368  }
369  }
370 
371 }
372 
373 static void FUNC(alf_recon_coeff_and_clip)(int16_t *coeff, int16_t *clip,
374  const int *class_idx, const int *transpose_idx, const int size,
375  const int16_t *coeff_set, const uint8_t *clip_idx_set, const uint8_t *class_to_filt)
376 {
377  const static int index[][ALF_NUM_COEFF_LUMA] = {
378  { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 },
379  { 9, 4, 10, 8, 1, 5, 11, 7, 3, 0, 2, 6 },
380  { 0, 3, 2, 1, 8, 7, 6, 5, 4, 9, 10, 11 },
381  { 9, 8, 10, 4, 3, 7, 11, 5, 1, 0, 2, 6 },
382  };
383 
384  const int16_t clip_set[] = {
385  1 << BIT_DEPTH, 1 << (BIT_DEPTH - 3), 1 << (BIT_DEPTH - 5), 1 << (BIT_DEPTH - 7)
386  };
387 
388  for (int i = 0; i < size; i++) {
389  const int16_t *src_coeff = coeff_set + class_to_filt[class_idx[i]] * ALF_NUM_COEFF_LUMA;
390  const uint8_t *clip_idx = clip_idx_set + class_idx[i] * ALF_NUM_COEFF_LUMA;
391 
392  for (int j = 0; j < ALF_NUM_COEFF_LUMA; j++) {
393  const int idx = index[transpose_idx[i]][j];
394  *coeff++ = src_coeff[idx];
395  *clip++ = clip_set[clip_idx[idx]];
396  }
397  }
398 }
399 
400 #undef ALF_DIR_HORZ
401 #undef ALF_DIR_VERT
402 #undef ALF_DIR_DIGA0
403 #undef ALF_DIR_DIGA1
404 
405 // line zero
406 #define P7 pix[-8 * xstride]
407 #define P6 pix[-7 * xstride]
408 #define P5 pix[-6 * xstride]
409 #define P4 pix[-5 * xstride]
410 #define P3 pix[-4 * xstride]
411 #define P2 pix[-3 * xstride]
412 #define P1 pix[-2 * xstride]
413 #define P0 pix[-1 * xstride]
414 #define Q0 pix[0 * xstride]
415 #define Q1 pix[1 * xstride]
416 #define Q2 pix[2 * xstride]
417 #define Q3 pix[3 * xstride]
418 #define Q4 pix[4 * xstride]
419 #define Q5 pix[5 * xstride]
420 #define Q6 pix[6 * xstride]
421 #define Q7 pix[7 * xstride]
422 #define P(x) pix[(-(x)-1) * xstride]
423 #define Q(x) pix[(x) * xstride]
424 
425 // line three. used only for deblocking decision
426 #define TP7 pix[-8 * xstride + 3 * ystride]
427 #define TP6 pix[-7 * xstride + 3 * ystride]
428 #define TP5 pix[-6 * xstride + 3 * ystride]
429 #define TP4 pix[-5 * xstride + 3 * ystride]
430 #define TP3 pix[-4 * xstride + 3 * ystride]
431 #define TP2 pix[-3 * xstride + 3 * ystride]
432 #define TP1 pix[-2 * xstride + 3 * ystride]
433 #define TP0 pix[-1 * xstride + 3 * ystride]
434 #define TQ0 pix[0 * xstride + 3 * ystride]
435 #define TQ1 pix[1 * xstride + 3 * ystride]
436 #define TQ2 pix[2 * xstride + 3 * ystride]
437 #define TQ3 pix[3 * xstride + 3 * ystride]
438 #define TQ4 pix[4 * xstride + 3 * ystride]
439 #define TQ5 pix[5 * xstride + 3 * ystride]
440 #define TQ6 pix[6 * xstride + 3 * ystride]
441 #define TQ7 pix[7 * xstride + 3 * ystride]
442 #define TP(x) pix[(-(x)-1) * xstride + 3 * ystride]
443 #define TQ(x) pix[(x) * xstride + 3 * ystride]
444 
445 #define FP3 pix[-4 * xstride + 1 * ystride]
446 #define FP2 pix[-3 * xstride + 1 * ystride]
447 #define FP1 pix[-2 * xstride + 1 * ystride]
448 #define FP0 pix[-1 * xstride + 1 * ystride]
449 #define FQ0 pix[0 * xstride + 1 * ystride]
450 #define FQ1 pix[1 * xstride + 1 * ystride]
451 #define FQ2 pix[2 * xstride + 1 * ystride]
452 #define FQ3 pix[3 * xstride + 1 * ystride]
453 
455 
456 static void FUNC(loop_filter_luma_large)(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride, const int32_t tc,
457  const uint8_t no_p, const uint8_t no_q, const uint8_t max_len_p, const uint8_t max_len_q)
458 {
459  for (int d = 0; d < 4; d++) {
460  const int p6 = P6;
461  const int p5 = P5;
462  const int p4 = P4;
463  const int p3 = P3;
464  const int p2 = P2;
465  const int p1 = P1;
466  const int p0 = P0;
467  const int q0 = Q0;
468  const int q1 = Q1;
469  const int q2 = Q2;
470  const int q3 = Q3;
471  const int q4 = Q4;
472  const int q5 = Q5;
473  const int q6 = Q6;
474  int m;
475  if (max_len_p == 5 && max_len_q == 5)
476  m = (p4 + p3 + 2 * (p2 + p1 + p0 + q0 + q1 + q2) + q3 + q4 + 8) >> 4;
477  else if (max_len_p == max_len_q)
478  m = (p6 + p5 + p4 + p3 + p2 + p1 + 2 * (p0 + q0) + q1 + q2 + q3 + q4 + q5 + q6 + 8) >> 4;
479  else if (max_len_p + max_len_q == 12)
480  m = (p5 + p4 + p3 + p2 + 2 * (p1 + p0 + q0 + q1) + q2 + q3 + q4 + q5 + 8) >> 4;
481  else if (max_len_p + max_len_q == 8)
482  m = (p3 + p2 + p1 + p0 + q0 + q1 + q2 + q3 + 4) >> 3;
483  else if (max_len_q == 7)
484  m = (2 * (p2 + p1 + p0 + q0) + p0 + p1 + q1 + q2 + q3 + q4 + q5 + q6 + 8) >> 4;
485  else
486  m = (p6 + p5 + p4 + p3 + p2 + p1 + 2 * (q2 + q1 + q0 + p0) + q0 + q1 + 8) >> 4;
487  if (!no_p) {
488  const int refp = (P(max_len_p) + P(max_len_p - 1) + 1) >> 1;
489  if (max_len_p == 3) {
490  P0 = p0 + av_clip(((m * 53 + refp * 11 + 32) >> 6) - p0, -(tc * 6 >> 1), (tc * 6 >> 1));
491  P1 = p1 + av_clip(((m * 32 + refp * 32 + 32) >> 6) - p1, -(tc * 4 >> 1), (tc * 4 >> 1));
492  P2 = p2 + av_clip(((m * 11 + refp * 53 + 32) >> 6) - p2, -(tc * 2 >> 1), (tc * 2 >> 1));
493  } else if (max_len_p == 5) {
494  P0 = p0 + av_clip(((m * 58 + refp * 6 + 32) >> 6) - p0, -(tc * 6 >> 1), (tc * 6 >> 1));
495  P1 = p1 + av_clip(((m * 45 + refp * 19 + 32) >> 6) - p1, -(tc * 5 >> 1), (tc * 5 >> 1));
496  P2 = p2 + av_clip(((m * 32 + refp * 32 + 32) >> 6) - p2, -(tc * 4 >> 1), (tc * 4 >> 1));
497  P3 = p3 + av_clip(((m * 19 + refp * 45 + 32) >> 6) - p3, -(tc * 3 >> 1), (tc * 3 >> 1));
498  P4 = p4 + av_clip(((m * 6 + refp * 58 + 32) >> 6) - p4, -(tc * 2 >> 1), (tc * 2 >> 1));
499  } else {
500  P0 = p0 + av_clip(((m * 59 + refp * 5 + 32) >> 6) - p0, -(tc * 6 >> 1), (tc * 6 >> 1));
501  P1 = p1 + av_clip(((m * 50 + refp * 14 + 32) >> 6) - p1, -(tc * 5 >> 1), (tc * 5 >> 1));
502  P2 = p2 + av_clip(((m * 41 + refp * 23 + 32) >> 6) - p2, -(tc * 4 >> 1), (tc * 4 >> 1));
503  P3 = p3 + av_clip(((m * 32 + refp * 32 + 32) >> 6) - p3, -(tc * 3 >> 1), (tc * 3 >> 1));
504  P4 = p4 + av_clip(((m * 23 + refp * 41 + 32) >> 6) - p4, -(tc * 2 >> 1), (tc * 2 >> 1));
505  P5 = p5 + av_clip(((m * 14 + refp * 50 + 32) >> 6) - p5, -(tc * 1 >> 1), (tc * 1 >> 1));
506  P6 = p6 + av_clip(((m * 5 + refp * 59 + 32) >> 6) - p6, -(tc * 1 >> 1), (tc * 1 >> 1));
507  }
508  }
509  if (!no_q) {
510  const int refq = (Q(max_len_q) + Q(max_len_q - 1) + 1) >> 1;
511  if (max_len_q == 3) {
512  Q0 = q0 + av_clip(((m * 53 + refq * 11 + 32) >> 6) - q0, -(tc * 6 >> 1), (tc * 6 >> 1));
513  Q1 = q1 + av_clip(((m * 32 + refq * 32 + 32) >> 6) - q1, -(tc * 4 >> 1), (tc * 4 >> 1));
514  Q2 = q2 + av_clip(((m * 11 + refq * 53 + 32) >> 6) - q2, -(tc * 2 >> 1), (tc * 2 >> 1));
515  } else if (max_len_q == 5) {
516  Q0 = q0 + av_clip(((m * 58 + refq * 6 + 32) >> 6) - q0, -(tc * 6 >> 1), (tc * 6 >> 1));
517  Q1 = q1 + av_clip(((m * 45 + refq * 19 + 32) >> 6) - q1, -(tc * 5 >> 1), (tc * 5 >> 1));
518  Q2 = q2 + av_clip(((m * 32 + refq * 32 + 32) >> 6) - q2, -(tc * 4 >> 1), (tc * 4 >> 1));
519  Q3 = q3 + av_clip(((m * 19 + refq * 45 + 32) >> 6) - q3, -(tc * 3 >> 1), (tc * 3 >> 1));
520  Q4 = q4 + av_clip(((m * 6 + refq * 58 + 32) >> 6) - q4, -(tc * 2 >> 1), (tc * 2 >> 1));
521  } else {
522  Q0 = q0 + av_clip(((m * 59 + refq * 5 + 32) >> 6) - q0, -(tc * 6 >> 1), (tc * 6 >> 1));
523  Q1 = q1 + av_clip(((m * 50 + refq * 14 + 32) >> 6) - q1, -(tc * 5 >> 1), (tc * 5 >> 1));
524  Q2 = q2 + av_clip(((m * 41 + refq * 23 + 32) >> 6) - q2, -(tc * 4 >> 1), (tc * 4 >> 1));
525  Q3 = q3 + av_clip(((m * 32 + refq * 32 + 32) >> 6) - q3, -(tc * 3 >> 1), (tc * 3 >> 1));
526  Q4 = q4 + av_clip(((m * 23 + refq * 41 + 32) >> 6) - q4, -(tc * 2 >> 1), (tc * 2 >> 1));
527  Q5 = q5 + av_clip(((m * 14 + refq * 50 + 32) >> 6) - q5, -(tc * 1 >> 1), (tc * 1 >> 1));
528  Q6 = q6 + av_clip(((m * 5 + refq * 59 + 32) >> 6) - q6, -(tc * 1 >> 1), (tc * 1 >> 1));
529  }
530 
531  }
532  pix += ystride;
533  }
534 }
535 
536 static void FUNC(vvc_loop_filter_luma)(uint8_t* _pix, ptrdiff_t _xstride, ptrdiff_t _ystride,
537  const int32_t *_beta, const int32_t *_tc, const uint8_t *_no_p, const uint8_t *_no_q,
538  const uint8_t *_max_len_p, const uint8_t *_max_len_q, const int hor_ctu_edge)
539 {
540  const ptrdiff_t xstride = _xstride / sizeof(pixel);
541  const ptrdiff_t ystride = _ystride / sizeof(pixel);
542 
543  for (int i = 0; i < 2; i++) {
544 #if BIT_DEPTH < 10
545  const int tc = (_tc[i] + (1 << (9 - BIT_DEPTH))) >> (10 - BIT_DEPTH);
546 #else
547  const int tc = _tc[i] << (BIT_DEPTH - 10);
548 #endif
549  if (tc) {
550  pixel* pix = (pixel*)_pix + i * 4 * ystride;
551  const int dp0 = abs(P2 - 2 * P1 + P0);
552  const int dq0 = abs(Q2 - 2 * Q1 + Q0);
553  const int dp3 = abs(TP2 - 2 * TP1 + TP0);
554  const int dq3 = abs(TQ2 - 2 * TQ1 + TQ0);
555  const int d0 = dp0 + dq0;
556  const int d3 = dp3 + dq3;
557  const int tc25 = ((tc * 5 + 1) >> 1);
558 
559  const int no_p = _no_p[i];
560  const int no_q = _no_q[i];
561 
562  int max_len_p = _max_len_p[i];
563  int max_len_q = _max_len_q[i];
564 
565  const int large_p = (max_len_p > 3 && !hor_ctu_edge);
566  const int large_q = max_len_q > 3;
567 
568  const int beta = _beta[i] << BIT_DEPTH - 8;
569  const int beta_3 = beta >> 3;
570  const int beta_2 = beta >> 2;
571 
572  if (large_p || large_q) {
573  const int dp0l = large_p ? ((dp0 + abs(P5 - 2 * P4 + P3) + 1) >> 1) : dp0;
574  const int dq0l = large_q ? ((dq0 + abs(Q5 - 2 * Q4 + Q3) + 1) >> 1) : dq0;
575  const int dp3l = large_p ? ((dp3 + abs(TP5 - 2 * TP4 + TP3) + 1) >> 1) : dp3;
576  const int dq3l = large_q ? ((dq3 + abs(TQ5 - 2 * TQ4 + TQ3) + 1) >> 1) : dq3;
577  const int d0l = dp0l + dq0l;
578  const int d3l = dp3l + dq3l;
579  const int beta53 = beta * 3 >> 5;
580  const int beta_4 = beta >> 4;
581  max_len_p = large_p ? max_len_p : 3;
582  max_len_q = large_q ? max_len_q : 3;
583 
584  if (d0l + d3l < beta) {
585  const int sp0l = abs(P3 - P0) + (max_len_p == 7 ? abs(P7 - P6 - P5 + P4) : 0);
586  const int sq0l = abs(Q0 - Q3) + (max_len_q == 7 ? abs(Q4 - Q5 - Q6 + Q7) : 0);
587  const int sp3l = abs(TP3 - TP0) + (max_len_p == 7 ? abs(TP7 - TP6 - TP5 + TP4) : 0);
588  const int sq3l = abs(TQ0 - TQ3) + (max_len_q == 7 ? abs(TQ4 - TQ5 - TQ6 + TQ7) : 0);
589  const int sp0 = large_p ? ((sp0l + abs(P3 - P(max_len_p)) + 1) >> 1) : sp0l;
590  const int sp3 = large_p ? ((sp3l + abs(TP3 - TP(max_len_p)) + 1) >> 1) : sp3l;
591  const int sq0 = large_q ? ((sq0l + abs(Q3 - Q(max_len_q)) + 1) >> 1) : sq0l;
592  const int sq3 = large_q ? ((sq3l + abs(TQ3 - TQ(max_len_q)) + 1) >> 1) : sq3l;
593  if (sp0 + sq0 < beta53 && abs(P0 - Q0) < tc25 &&
594  sp3 + sq3 < beta53 && abs(TP0 - TQ0) < tc25 &&
595  (d0l << 1) < beta_4 && (d3l << 1) < beta_4) {
596  FUNC(loop_filter_luma_large)(pix, xstride, ystride, tc, no_p, no_q, max_len_p, max_len_q);
597  continue;
598  }
599  }
600  }
601  if (d0 + d3 < beta) {
602  if (max_len_p > 2 && max_len_q > 2 &&
603  abs(P3 - P0) + abs(Q3 - Q0) < beta_3 && abs(P0 - Q0) < tc25 &&
604  abs(TP3 - TP0) + abs(TQ3 - TQ0) < beta_3 && abs(TP0 - TQ0) < tc25 &&
605  (d0 << 1) < beta_2 && (d3 << 1) < beta_2) {
606  FUNC(loop_filter_luma_strong)(pix, xstride, ystride, tc, tc << 1, tc * 3, no_p, no_q);
607  } else {
608  int nd_p = 1;
609  int nd_q = 1;
610  if (max_len_p > 1 && max_len_q > 1) {
611  if (dp0 + dp3 < ((beta + (beta >> 1)) >> 3))
612  nd_p = 2;
613  if (dq0 + dq3 < ((beta + (beta >> 1)) >> 3))
614  nd_q = 2;
615  }
616  FUNC(loop_filter_luma_weak)(pix, xstride, ystride, tc, beta, no_p, no_q, nd_p, nd_q);
617  }
618  }
619  }
620  }
621 }
622 
623 static void FUNC(loop_filter_chroma_strong)(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride,
624  const int size, const int32_t tc, const uint8_t no_p, const uint8_t no_q)
625 {
626  for (int d = 0; d < size; d++) {
627  const int p3 = P3;
628  const int p2 = P2;
629  const int p1 = P1;
630  const int p0 = P0;
631  const int q0 = Q0;
632  const int q1 = Q1;
633  const int q2 = Q2;
634  const int q3 = Q3;
635  if (!no_p) {
636  P0 = av_clip((p3 + p2 + p1 + 2 * p0 + q0 + q1 + q2 + 4) >> 3, p0 - tc, p0 + tc);
637  P1 = av_clip((2 * p3 + p2 + 2 * p1 + p0 + q0 + q1 + 4) >> 3, p1 - tc, p1 + tc);
638  P2 = av_clip((3 * p3 + 2 * p2 + p1 + p0 + q0 + 4) >> 3, p2 - tc, p2 + tc );
639  }
640  if (!no_q) {
641  Q0 = av_clip((p2 + p1 + p0 + 2 * q0 + q1 + q2 + q3 + 4) >> 3, q0 - tc, q0 + tc);
642  Q1 = av_clip((p1 + p0 + q0 + 2 * q1 + q2 + 2 * q3 + 4) >> 3, q1 - tc, q1 + tc);
643  Q2 = av_clip((p0 + q0 + q1 + 2 * q2 + 3 * q3 + 4) >> 3, q2 - tc, q2 + tc);
644  }
645  pix += ystride;
646  }
647 }
648 
649 static void FUNC(loop_filter_chroma_strong_one_side)(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride,
650  const int size, const int32_t tc, const uint8_t no_p, const uint8_t no_q)
651 {
652  for (int d = 0; d < size; d++) {
653  const int p1 = P1;
654  const int p0 = P0;
655  const int q0 = Q0;
656  const int q1 = Q1;
657  const int q2 = Q2;
658  const int q3 = Q3;
659  if (!no_p) {
660  P0 = av_clip((3 * p1 + 2 * p0 + q0 + q1 + q2 + 4) >> 3, p0 - tc, p0 + tc);
661  }
662  if (!no_q) {
663  Q0 = av_clip((2 * p1 + p0 + 2 * q0 + q1 + q2 + q3 + 4) >> 3, q0 - tc, q0 + tc);
664  Q1 = av_clip((p1 + p0 + q0 + 2 * q1 + q2 + 2 * q3 + 4) >> 3, q1 - tc, q1 + tc);
665  Q2 = av_clip((p0 + q0 + q1 + 2 * q2 + 3 * q3 + 4) >> 3, q2 - tc, q2 + tc);
666  }
667  pix += ystride;
668  }
669 }
670 
671 static void FUNC(vvc_loop_filter_chroma)(uint8_t *_pix, const ptrdiff_t _xstride, const ptrdiff_t _ystride,
672  const int32_t *_beta, const int32_t *_tc, const uint8_t *_no_p, const uint8_t *_no_q,
673  const uint8_t *_max_len_p, const uint8_t *_max_len_q, const int shift)
674 {
675  const ptrdiff_t xstride = _xstride / sizeof(pixel);
676  const ptrdiff_t ystride = _ystride / sizeof(pixel);
677  const int size = shift ? 2 : 4;
678  const int end = 8 / size; // 8 samples a loop
679 
680  for (int i = 0; i < end; i++) {
681 #if BIT_DEPTH < 10
682  const int tc = (_tc[i] + (1 << (9 - BIT_DEPTH))) >> (10 - BIT_DEPTH);
683 #else
684  const int tc = _tc[i] << (BIT_DEPTH - 10);
685 #endif
686  if (tc) {
687  pixel *pix = (pixel *)_pix + i * size * ystride;
688  const uint8_t no_p = _no_p[i];
689  const uint8_t no_q = _no_q[i];
690 
691  const int beta = _beta[i] << (BIT_DEPTH - 8);
692  const int beta_3 = beta >> 3;
693  const int beta_2 = beta >> 2;
694 
695  const int tc25 = ((tc * 5 + 1) >> 1);
696 
697  uint8_t max_len_p = _max_len_p[i];
698  uint8_t max_len_q = _max_len_q[i];
699 
700  if (!max_len_p || !max_len_q)
701  continue;
702 
703  if (max_len_q == 3){
704  const int p1n = shift ? FP1 : TP1;
705  const int p2n = max_len_p == 1 ? p1n : (shift ? FP2 : TP2);
706  const int p0n = shift ? FP0 : TP0;
707  const int q0n = shift ? FQ0 : TQ0;
708  const int q1n = shift ? FQ1 : TQ1;
709  const int q2n = shift ? FQ2 : TQ2;
710  const int p3 = max_len_p == 1 ? P1 : P3;
711  const int p2 = max_len_p == 1 ? P1 : P2;
712  const int p1 = P1;
713  const int p0 = P0;
714  const int dp0 = abs(p2 - 2 * p1 + p0);
715  const int dq0 = abs(Q2 - 2 * Q1 + Q0);
716 
717  const int dp1 = abs(p2n - 2 * p1n + p0n);
718  const int dq1 = abs(q2n - 2 * q1n + q0n);
719  const int d0 = dp0 + dq0;
720  const int d1 = dp1 + dq1;
721 
722  if (d0 + d1 < beta) {
723  const int p3n = max_len_p == 1 ? p1n : (shift ? FP3 : TP3);
724  const int q3n = shift ? FQ3 : TQ3;
725  const int dsam0 = (d0 << 1) < beta_2 && (abs(p3 - p0) + abs(Q0 - Q3) < beta_3) &&
726  abs(p0 - Q0) < tc25;
727  const int dsam1 = (d1 << 1) < beta_2 && (abs(p3n - p0n) + abs(q0n - q3n) < beta_3) &&
728  abs(p0n - q0n) < tc25;
729  if (!dsam0 || !dsam1)
730  max_len_p = max_len_q = 1;
731  } else {
732  max_len_p = max_len_q = 1;
733  }
734  }
735 
736  if (max_len_p == 3 && max_len_q == 3)
737  FUNC(loop_filter_chroma_strong)(pix, xstride, ystride, size, tc, no_p, no_q);
738  else if (max_len_q == 3)
739  FUNC(loop_filter_chroma_strong_one_side)(pix, xstride, ystride, size, tc, no_p, no_q);
740  else
741  FUNC(loop_filter_chroma_weak)(pix, xstride, ystride, size, tc, no_p, no_q);
742  }
743  }
744 }
745 
746 static void FUNC(vvc_h_loop_filter_chroma)(uint8_t *pix, ptrdiff_t stride,
747  const int32_t *beta, const int32_t *tc, const uint8_t *no_p, const uint8_t *no_q,
748  const uint8_t *max_len_p, const uint8_t *max_len_q, int shift)
749 {
750  FUNC(vvc_loop_filter_chroma)(pix, stride, sizeof(pixel), beta, tc,
751  no_p, no_q, max_len_p, max_len_q, shift);
752 }
753 
754 static void FUNC(vvc_v_loop_filter_chroma)(uint8_t *pix, ptrdiff_t stride,
755  const int32_t *beta, const int32_t *tc, const uint8_t *no_p, const uint8_t *no_q,
756  const uint8_t *max_len_p, const uint8_t *max_len_q, int shift)
757 {
758  FUNC(vvc_loop_filter_chroma)(pix, sizeof(pixel), stride, beta, tc,
759  no_p, no_q, max_len_p, max_len_q, shift);
760 }
761 
762 static void FUNC(vvc_h_loop_filter_luma)(uint8_t *pix, ptrdiff_t stride,
763  const int32_t *beta, const int32_t *tc, const uint8_t *no_p, const uint8_t *no_q,
764  const uint8_t *max_len_p, const uint8_t *max_len_q, const int hor_ctu_edge)
765 {
766  FUNC(vvc_loop_filter_luma)(pix, stride, sizeof(pixel), beta, tc,
767  no_p, no_q, max_len_p, max_len_q, hor_ctu_edge);
768 }
769 
770 static void FUNC(vvc_v_loop_filter_luma)(uint8_t *pix, ptrdiff_t stride,
771  const int32_t *beta, const int32_t *tc, const uint8_t *no_p, const uint8_t *no_q,
772  const uint8_t *max_len_p, const uint8_t *max_len_q, const int hor_ctu_edge)
773 {
774  FUNC(vvc_loop_filter_luma)(pix, sizeof(pixel), stride, beta, tc,
775  no_p, no_q, max_len_p, max_len_q, hor_ctu_edge);
776 }
777 
778 static int FUNC(vvc_loop_ladf_level)(const uint8_t *_pix, const ptrdiff_t _xstride, const ptrdiff_t _ystride)
779 {
780  const pixel *pix = (pixel *)_pix;
781  const ptrdiff_t xstride = _xstride / sizeof(pixel);
782  const ptrdiff_t ystride = _ystride / sizeof(pixel);
783  return (P0 + TP0 + Q0 + TQ0) >> 2;
784 }
785 
786 static int FUNC(vvc_h_loop_ladf_level)(const uint8_t *pix, ptrdiff_t stride)
787 {
788  return FUNC(vvc_loop_ladf_level)(pix, stride, sizeof(pixel));
789 }
790 
791 static int FUNC(vvc_v_loop_ladf_level)(const uint8_t *pix, ptrdiff_t stride)
792 {
793  return FUNC(vvc_loop_ladf_level)(pix, sizeof(pixel), stride);
794 }
795 
796 #undef P7
797 #undef P6
798 #undef P5
799 #undef P4
800 #undef P3
801 #undef P2
802 #undef P1
803 #undef P0
804 #undef Q0
805 #undef Q1
806 #undef Q2
807 #undef Q3
808 #undef Q4
809 #undef Q5
810 #undef Q6
811 #undef Q7
812 
813 #undef TP7
814 #undef TP6
815 #undef TP5
816 #undef TP4
817 #undef TP3
818 #undef TP2
819 #undef TP1
820 #undef TP0
821 #undef TQ0
822 #undef TQ1
823 #undef TQ2
824 #undef TQ3
825 #undef TQ4
826 #undef TQ5
827 #undef TQ6
828 #undef TQ7
829 
830 static void FUNC(ff_vvc_lmcs_dsp_init)(VVCLMCSDSPContext *const lmcs)
831 {
832  lmcs->filter = FUNC(lmcs_filter_luma);
833 }
834 
835 static void FUNC(ff_vvc_lf_dsp_init)(VVCLFDSPContext *const lf)
836 {
837  lf->ladf_level[0] = FUNC(vvc_h_loop_ladf_level);
838  lf->ladf_level[1] = FUNC(vvc_v_loop_ladf_level);
839  lf->filter_luma[0] = FUNC(vvc_h_loop_filter_luma);
840  lf->filter_luma[1] = FUNC(vvc_v_loop_filter_luma);
841  lf->filter_chroma[0] = FUNC(vvc_h_loop_filter_chroma);
842  lf->filter_chroma[1] = FUNC(vvc_v_loop_filter_chroma);
843 }
844 
845 static void FUNC(ff_vvc_sao_dsp_init)(VVCSAODSPContext *const sao)
846 {
847  for (int i = 0; i < FF_ARRAY_ELEMS(sao->band_filter); i++)
848  sao->band_filter[i] = FUNC(sao_band_filter);
849  for (int i = 0; i < FF_ARRAY_ELEMS(sao->edge_filter); i++)
850  sao->edge_filter[i] = FUNC(sao_edge_filter);
851  sao->edge_restore[0] = FUNC(sao_edge_restore_0);
852  sao->edge_restore[1] = FUNC(sao_edge_restore_1);
853 }
854 
855 static void FUNC(ff_vvc_alf_dsp_init)(VVCALFDSPContext *const alf)
856 {
857  alf->filter[LUMA] = FUNC(alf_filter_luma);
858  alf->filter[CHROMA] = FUNC(alf_filter_chroma);
859  alf->filter_cc = FUNC(alf_filter_cc);
860  alf->classify = FUNC(alf_classify);
861  alf->recon_coeff_and_clip = FUNC(alf_recon_coeff_and_clip);
862 }
_dst
uint8_t * _dst
Definition: dsp.h:52
q1
static const uint8_t q1[256]
Definition: twofish.c:100
av_clip
#define av_clip
Definition: common.h:100
TP0
#define TP0
Definition: filter_template.c:433
LUMA
#define LUMA
Definition: filter.c:31
FP1
#define FP1
Definition: filter_template.c:447
loop_filter_chroma_strong
static void FUNC() loop_filter_chroma_strong(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride, const int size, const int32_t tc, const uint8_t no_p, const uint8_t no_q)
Definition: filter_template.c:623
av_clip_uintp2
#define av_clip_uintp2
Definition: common.h:124
alf_get_idx
static void FUNC() alf_get_idx(int *class_idx, int *transpose_idx, const int *sum, const int ac)
Definition: filter_template.c:260
w
uint8_t w
Definition: llviddspenc.c:38
FQ0
#define FQ0
Definition: filter_template.c:449
TQ6
#define TQ6
Definition: filter_template.c:440
ALF_DIR_VERT
#define ALF_DIR_VERT
Definition: filter_template.c:255
filter
void(* filter)(uint8_t *src, int stride, int qscale)
Definition: h263dsp.c:29
Q5
#define Q5
Definition: filter_template.c:419
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
TQ1
#define TQ1
Definition: filter_template.c:435
vvc_loop_filter_chroma
static void FUNC() vvc_loop_filter_chroma(uint8_t *_pix, const ptrdiff_t _xstride, const ptrdiff_t _ystride, const int32_t *_beta, const int32_t *_tc, const uint8_t *_no_p, const uint8_t *_no_q, const uint8_t *_max_len_p, const uint8_t *_max_len_q, const int shift)
Definition: filter_template.c:671
sao_edge_restore_1
static void FUNC() sao_edge_restore_1(uint8_t *_dst, const uint8_t *_src, ptrdiff_t stride_dst, ptrdiff_t stride_src, const SAOParams *sao, const int *borders, int _width, int _height, int c_idx, const uint8_t *vert_edge, const uint8_t *horiz_edge, const uint8_t *diag_edge)
Definition: h2656_sao_template.c:131
loop_filter_luma_strong
static void FUNC() loop_filter_luma_strong(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride, const int32_t tc, const int32_t tc2, const int tc3, const uint8_t no_p, const uint8_t no_q)
Definition: h2656_deblock_template.c:25
ff_vvc_lmcs_dsp_init
static void FUNC() ff_vvc_lmcs_dsp_init(VVCLMCSDSPContext *const lmcs)
Definition: filter_template.c:830
_src
uint8_t ptrdiff_t const uint8_t * _src
Definition: dsp.h:52
Q1
#define Q1
Definition: filter_template.c:415
sao_edge_restore_0
static void FUNC() sao_edge_restore_0(uint8_t *_dst, const uint8_t *_src, ptrdiff_t stride_dst, ptrdiff_t stride_src, const SAOParams *sao, const int *borders, int _width, int _height, int c_idx, const uint8_t *vert_edge, const uint8_t *horiz_edge, const uint8_t *diag_edge)
Definition: h2656_sao_template.c:81
FP3
#define FP3
Definition: filter_template.c:445
Q3
#define Q3
Definition: filter_template.c:417
b1
static double b1(void *priv, double x, double y)
Definition: vf_xfade.c:2034
TP5
#define TP5
Definition: filter_template.c:428
P
#define P(x)
Definition: filter_template.c:422
vvc_v_loop_filter_chroma
static void FUNC() vvc_v_loop_filter_chroma(uint8_t *pix, ptrdiff_t stride, const int32_t *beta, const int32_t *tc, const uint8_t *no_p, const uint8_t *no_q, const uint8_t *max_len_p, const uint8_t *max_len_q, int shift)
Definition: filter_template.c:754
P1
#define P1
Definition: filter_template.c:412
ALF_GRADIENT_BORDER
#define ALF_GRADIENT_BORDER
Definition: ctu.h:85
ALF_DIR_HORZ
#define ALF_DIR_HORZ
Definition: filter_template.c:256
TQ3
#define TQ3
Definition: filter_template.c:437
vvc_v_loop_ladf_level
static int FUNC() vvc_v_loop_ladf_level(const uint8_t *pix, ptrdiff_t stride)
Definition: filter_template.c:791
FF_ARRAY_ELEMS
#define FF_ARRAY_ELEMS(a)
Definition: sinewin_tablegen.c:29
alf_filter_cc
static void FUNC() alf_filter_cc(uint8_t *_dst, ptrdiff_t dst_stride, const uint8_t *_luma, const ptrdiff_t luma_stride, const int width, const int height, const int hs, const int vs, const int16_t *filter, const int vb_pos)
Definition: filter_template.c:213
clip
clip
Definition: af_crystalizer.c:122
sao_band_filter
static void FUNC() sao_band_filter(uint8_t *_dst, const uint8_t *_src, ptrdiff_t stride_dst, ptrdiff_t stride_src, const int16_t *sao_offset_val, int sao_left_class, int width, int height)
Definition: h2656_sao_template.c:24
ALF_DIR_DIGA1
#define ALF_DIR_DIGA1
Definition: filter_template.c:258
loop_filter_luma_large
static void FUNC() loop_filter_luma_large(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride, const int32_t tc, const uint8_t no_p, const uint8_t no_q, const uint8_t max_len_p, const uint8_t max_len_q)
Definition: filter_template.c:456
Q4
#define Q4
Definition: filter_template.c:418
VVCALFDSPContext
Definition: dsp.h:157
TP2
#define TP2
Definition: filter_template.c:431
vvc_h_loop_filter_luma
static void FUNC() vvc_h_loop_filter_luma(uint8_t *pix, ptrdiff_t stride, const int32_t *beta, const int32_t *tc, const uint8_t *no_p, const uint8_t *no_q, const uint8_t *max_len_p, const uint8_t *max_len_q, const int hor_ctu_edge)
Definition: filter_template.c:762
loop_filter_chroma_weak
static void FUNC() loop_filter_chroma_weak(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride, const int size, const int32_t tc, const uint8_t no_p, const uint8_t no_q)
Definition: h2656_deblock_template.c:83
q0
static const uint8_t q0[256]
Definition: twofish.c:81
FFABS
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition: common.h:74
TQ5
#define TQ5
Definition: filter_template.c:439
ALF_DIR_DIGA0
#define ALF_DIR_DIGA0
Definition: filter_template.c:257
pixel
uint8_t pixel
Definition: tiny_ssim.c:41
grad
static double grad(int hash, double x, double y, double z)
Definition: perlin.c:42
Q6
#define Q6
Definition: filter_template.c:420
P0
#define P0
Definition: filter_template.c:413
h2656_deblock_template.c
ff_vvc_alf_dsp_init
static void FUNC() ff_vvc_alf_dsp_init(VVCALFDSPContext *const alf)
Definition: filter_template.c:855
TQ2
#define TQ2
Definition: filter_template.c:436
FQ1
#define FQ1
Definition: filter_template.c:450
ALF_NUM_DIR
#define ALF_NUM_DIR
Definition: ctu.h:87
abs
#define abs(x)
Definition: cuda_runtime.h:35
VVCSAODSPContext
Definition: dsp.h:146
Q0
#define Q0
Definition: filter_template.c:414
index
int index
Definition: gxfenc.c:90
alf_classify
static void FUNC() alf_classify(int *class_idx, int *transpose_idx, const uint8_t *_src, const ptrdiff_t _src_stride, const int width, const int height, const int vb_pos, int *gradient_tmp)
Definition: filter_template.c:289
Q2
#define Q2
Definition: filter_template.c:416
height
#define height
Definition: dsp.h:85
shift
static int shift(int a, int b)
Definition: bonk.c:261
dst
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition: dsp.h:83
P6
#define P6
Definition: filter_template.c:407
VVCLMCSDSPContext
Definition: dsp.h:132
size
int size
Definition: twinvq_data.h:10344
vvc_h_loop_ladf_level
static int FUNC() vvc_h_loop_ladf_level(const uint8_t *pix, ptrdiff_t stride)
Definition: filter_template.c:786
ALF_BLOCK_SIZE
#define ALF_BLOCK_SIZE
Definition: ctu.h:76
TP4
#define TP4
Definition: filter_template.c:429
FQ3
#define FQ3
Definition: filter_template.c:452
TQ0
#define TQ0
Definition: filter_template.c:434
TP1
#define TP1
Definition: filter_template.c:432
TQ
#define TQ(x)
Definition: filter_template.c:443
a0
static double a0(void *priv, double x, double y)
Definition: vf_xfade.c:2028
offset
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf offset
Definition: writing_filters.txt:86
alf_filter_luma
static void FUNC() alf_filter_luma(uint8_t *_dst, ptrdiff_t dst_stride, const uint8_t *_src, ptrdiff_t src_stride, const int width, const int height, const int16_t *filter, const int16_t *clip, const int vb_pos)
Definition: filter_template.c:43
TP3
#define TP3
Definition: filter_template.c:430
TP
#define TP(x)
Definition: filter_template.c:442
alf_filter_chroma
static void FUNC() alf_filter_chroma(uint8_t *_dst, ptrdiff_t dst_stride, const uint8_t *_src, ptrdiff_t src_stride, const int width, const int height, const int16_t *filter, const int16_t *clip, const int vb_pos)
Definition: filter_template.c:137
ff_vvc_lf_dsp_init
static void FUNC() ff_vvc_lf_dsp_init(VVCLFDSPContext *const lf)
Definition: filter_template.c:835
P3
#define P3
Definition: filter_template.c:410
vvc_h_loop_filter_chroma
static void FUNC() vvc_h_loop_filter_chroma(uint8_t *pix, ptrdiff_t stride, const int32_t *beta, const int32_t *tc, const uint8_t *no_p, const uint8_t *no_q, const uint8_t *max_len_p, const uint8_t *max_len_q, int shift)
Definition: filter_template.c:746
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:256
vvc_loop_filter_luma
static void FUNC() vvc_loop_filter_luma(uint8_t *_pix, ptrdiff_t _xstride, ptrdiff_t _ystride, const int32_t *_beta, const int32_t *_tc, const uint8_t *_no_p, const uint8_t *_no_q, const uint8_t *_max_len_p, const uint8_t *_max_len_q, const int hor_ctu_edge)
Definition: filter_template.c:536
FP0
#define FP0
Definition: filter_template.c:448
av_always_inline
#define av_always_inline
Definition: attributes.h:49
P2
#define P2
Definition: filter_template.c:411
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
alf_clip
static av_always_inline int16_t FUNC() alf_clip(pixel curr, pixel v0, pixel v1, int16_t clip)
Definition: filter_template.c:38
ff_vvc_sao_dsp_init
static void FUNC() ff_vvc_sao_dsp_init(VVCSAODSPContext *const sao)
Definition: filter_template.c:845
TQ7
#define TQ7
Definition: filter_template.c:441
stride
#define stride
Definition: h264pred_template.c:537
sao_edge_filter
static void FUNC() sao_edge_filter(uint8_t *_dst, const uint8_t *_src, ptrdiff_t stride_dst, const int16_t *sao_offset_val, int eo, int width, int height)
Definition: h2656_sao_template.c:50
ALF_NUM_COEFF_LUMA
#define ALF_NUM_COEFF_LUMA
Definition: ps.h:167
CHROMA
@ CHROMA
Definition: vf_waveform.c:49
av_clip_pixel
#define av_clip_pixel(a)
Definition: bit_depth_template.c:98
pos
unsigned int pos
Definition: spdifenc.c:414
loop_filter_chroma_strong_one_side
static void FUNC() loop_filter_chroma_strong_one_side(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride, const int size, const int32_t tc, const uint8_t no_p, const uint8_t no_q)
Definition: filter_template.c:649
FQ2
#define FQ2
Definition: filter_template.c:451
FUNC
#define FUNC(a)
Definition: bit_depth_template.c:104
TQ4
#define TQ4
Definition: filter_template.c:438
BIT_DEPTH
#define BIT_DEPTH
Definition: dsp_init.c:38
loop_filter_luma_weak
static void FUNC() loop_filter_luma_weak(pixel *pix, const ptrdiff_t xstride, const ptrdiff_t ystride, const int32_t tc, const int32_t beta, const uint8_t no_p, const uint8_t no_q, const int nd_p, const int nd_q)
Definition: h2656_deblock_template.c:52
P7
#define P7
Definition: filter_template.c:406
ALF_GRADIENT_STEP
#define ALF_GRADIENT_STEP
Definition: ctu.h:84
Q7
#define Q7
Definition: filter_template.c:421
CLIP
@ CLIP
Definition: qdrw.c:37
VVCLFDSPContext
Definition: dsp.h:136
P4
#define P4
Definition: filter_template.c:409
P5
#define P5
Definition: filter_template.c:408
int32_t
int32_t
Definition: audioconvert.c:56
Q
#define Q(x)
Definition: filter_template.c:423
TP6
#define TP6
Definition: filter_template.c:427
coeff
static const double coeff[2][5]
Definition: vf_owdenoise.c:80
alf_recon_coeff_and_clip
static void FUNC() alf_recon_coeff_and_clip(int16_t *coeff, int16_t *clip, const int *class_idx, const int *transpose_idx, const int size, const int16_t *coeff_set, const uint8_t *clip_idx_set, const uint8_t *class_to_filt)
Definition: filter_template.c:373
FP2
#define FP2
Definition: filter_template.c:446
b0
static double b0(void *priv, double x, double y)
Definition: vf_xfade.c:2033
a1
static double a1(void *priv, double x, double y)
Definition: vf_xfade.c:2029
h
h
Definition: vp9dsp_template.c:2070
width
#define width
Definition: dsp.h:85
vvc_v_loop_filter_luma
static void FUNC() vvc_v_loop_filter_luma(uint8_t *pix, ptrdiff_t stride, const int32_t *beta, const int32_t *tc, const uint8_t *no_p, const uint8_t *no_q, const uint8_t *max_len_p, const uint8_t *max_len_q, const int hor_ctu_edge)
Definition: filter_template.c:770
src
#define src
Definition: vp8dsp.c:248
TP7
#define TP7
Definition: filter_template.c:426
vvc_loop_ladf_level
static int FUNC() vvc_loop_ladf_level(const uint8_t *_pix, const ptrdiff_t _xstride, const ptrdiff_t _ystride)
Definition: filter_template.c:778
h2656_sao_template.c
lmcs_filter_luma
static void FUNC() lmcs_filter_luma(uint8_t *_dst, ptrdiff_t dst_stride, const int width, const int height, const void *_lut)
Definition: filter_template.c:25