FFmpeg
h264dsp.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2016 Martin Storsjo
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License as published by
8  * the Free Software Foundation; either version 2 of the License, or
9  * (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14  * GNU General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License along
17  * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
18  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
19  */
20 
21 #include <string.h>
22 #include "checkasm.h"
23 #include "libavcodec/h264dsp.h"
24 #include "libavcodec/h264data.h"
25 #include "libavcodec/h264_parse.h"
26 #include "libavutil/common.h"
27 #include "libavutil/intreadwrite.h"
28 #include "libavutil/mem_internal.h"
29 
30 static const uint32_t pixel_mask[5] = { 0xffffffff, 0x01ff01ff, 0x03ff03ff, 0x0fff0fff, 0x3fff3fff };
31 static const uint32_t pixel_mask_lf[3] = { 0xff0fff0f, 0x01ff000f, 0x03ff000f };
32 
33 #define SIZEOF_PIXEL ((bit_depth + 7) / 8)
34 #define SIZEOF_COEF (2 * ((bit_depth + 7) / 8))
35 #define PIXEL_STRIDE 16
36 
37 #define randomize_buffers(idx) \
38  do { \
39  int x, y; \
40  uint32_t mask = pixel_mask[(idx)]; \
41  for (y = 0; y < sz; y++) { \
42  for (x = 0; x < PIXEL_STRIDE; x += 4) { \
43  AV_WN32A(src + y * PIXEL_STRIDE + x, rnd() & mask); \
44  AV_WN32A(dst + y * PIXEL_STRIDE + x, rnd() & mask); \
45  } \
46  for (x = 0; x < sz; x++) { \
47  if (bit_depth == 8) { \
48  coef[y * sz + x] = src[y * PIXEL_STRIDE + x] - \
49  dst[y * PIXEL_STRIDE + x]; \
50  } else { \
51  ((int32_t *)coef)[y * sz + x] = \
52  ((uint16_t *)src)[y * (PIXEL_STRIDE/2) + x] - \
53  ((uint16_t *)dst)[y * (PIXEL_STRIDE/2) + x]; \
54  } \
55  } \
56  } \
57  } while (0)
58 
59 #define dct4x4_impl(size, dctcoef) \
60 static void dct4x4_##size(dctcoef *coef) \
61 { \
62  int i, y, x; \
63  dctcoef tmp[16]; \
64  for (i = 0; i < 4; i++) { \
65  const int z0 = coef[i*4 + 0] + coef[i*4 + 3]; \
66  const int z1 = coef[i*4 + 1] + coef[i*4 + 2]; \
67  const int z2 = coef[i*4 + 0] - coef[i*4 + 3]; \
68  const int z3 = coef[i*4 + 1] - coef[i*4 + 2]; \
69  tmp[i + 4*0] = z0 + z1; \
70  tmp[i + 4*1] = 2*z2 + z3; \
71  tmp[i + 4*2] = z0 - z1; \
72  tmp[i + 4*3] = z2 - 2*z3; \
73  } \
74  for (i = 0; i < 4; i++) { \
75  const int z0 = tmp[i*4 + 0] + tmp[i*4 + 3]; \
76  const int z1 = tmp[i*4 + 1] + tmp[i*4 + 2]; \
77  const int z2 = tmp[i*4 + 0] - tmp[i*4 + 3]; \
78  const int z3 = tmp[i*4 + 1] - tmp[i*4 + 2]; \
79  coef[i*4 + 0] = z0 + z1; \
80  coef[i*4 + 1] = 2*z2 + z3; \
81  coef[i*4 + 2] = z0 - z1; \
82  coef[i*4 + 3] = z2 - 2*z3; \
83  } \
84  for (y = 0; y < 4; y++) { \
85  for (x = 0; x < 4; x++) { \
86  const int64_t scale[] = { 13107 * 10, 8066 * 13, 5243 * 16 }; \
87  const int idx = (y & 1) + (x & 1); \
88  coef[y*4 + x] = (coef[y*4 + x] * scale[idx] + (1 << 14)) >> 15; \
89  } \
90  } \
91 }
92 
93 #define DCT8_1D(src, srcstride, dst, dststride) do { \
94  const int a0 = (src)[srcstride * 0] + (src)[srcstride * 7]; \
95  const int a1 = (src)[srcstride * 0] - (src)[srcstride * 7]; \
96  const int a2 = (src)[srcstride * 1] + (src)[srcstride * 6]; \
97  const int a3 = (src)[srcstride * 1] - (src)[srcstride * 6]; \
98  const int a4 = (src)[srcstride * 2] + (src)[srcstride * 5]; \
99  const int a5 = (src)[srcstride * 2] - (src)[srcstride * 5]; \
100  const int a6 = (src)[srcstride * 3] + (src)[srcstride * 4]; \
101  const int a7 = (src)[srcstride * 3] - (src)[srcstride * 4]; \
102  const int b0 = a0 + a6; \
103  const int b1 = a2 + a4; \
104  const int b2 = a0 - a6; \
105  const int b3 = a2 - a4; \
106  const int b4 = a3 + a5 + (a1 + (a1 >> 1)); \
107  const int b5 = a1 - a7 - (a5 + (a5 >> 1)); \
108  const int b6 = a1 + a7 - (a3 + (a3 >> 1)); \
109  const int b7 = a3 - a5 + (a7 + (a7 >> 1)); \
110  (dst)[dststride * 0] = b0 + b1; \
111  (dst)[dststride * 1] = b4 + (b7 >> 2); \
112  (dst)[dststride * 2] = b2 + (b3 >> 1); \
113  (dst)[dststride * 3] = b5 + (b6 >> 2); \
114  (dst)[dststride * 4] = b0 - b1; \
115  (dst)[dststride * 5] = b6 - (b5 >> 2); \
116  (dst)[dststride * 6] = (b2 >> 1) - b3; \
117  (dst)[dststride * 7] = (b4 >> 2) - b7; \
118 } while (0)
119 
120 #define dct8x8_impl(size, dctcoef) \
121 static void dct8x8_##size(dctcoef *coef) \
122 { \
123  int i, x, y; \
124  dctcoef tmp[64]; \
125  for (i = 0; i < 8; i++) \
126  DCT8_1D(coef + i, 8, tmp + i, 8); \
127  \
128  for (i = 0; i < 8; i++) \
129  DCT8_1D(tmp + 8*i, 1, coef + i, 8); \
130  \
131  for (y = 0; y < 8; y++) { \
132  for (x = 0; x < 8; x++) { \
133  static const int scale[] = { \
134  13107 * 20, 11428 * 18, 20972 * 32, \
135  12222 * 19, 16777 * 25, 15481 * 24, \
136  }; \
137  static const int idxmap[] = { \
138  0, 3, 4, 3, \
139  3, 1, 5, 1, \
140  4, 5, 2, 5, \
141  3, 1, 5, 1, \
142  }; \
143  const int idx = idxmap[(y & 3) * 4 + (x & 3)]; \
144  coef[y*8 + x] = ((int64_t)coef[y*8 + x] * \
145  scale[idx] + (1 << 17)) >> 18; \
146  } \
147  } \
148 }
149 
150 dct4x4_impl(16, int16_t)
151 dct4x4_impl(32, int32_t)
152 
153 dct8x8_impl(16, int16_t)
154 dct8x8_impl(32, int32_t)
155 
156 static void dct4x4(int16_t *coef, int bit_depth)
157 {
158  if (bit_depth == 8)
159  dct4x4_16(coef);
160  else
161  dct4x4_32((int32_t *) coef);
162 }
163 
164 static void dct8x8(int16_t *coef, int bit_depth)
165 {
166  if (bit_depth == 8) {
167  dct8x8_16(coef);
168  } else {
169  dct8x8_32((int32_t *) coef);
170  }
171 }
172 
173 
174 static void check_idct(void)
175 {
176  static const int depths[5] = { 8, 9, 10, 12, 14 };
177  LOCAL_ALIGNED_16(uint8_t, src, [8 * 8 * 2]);
178  LOCAL_ALIGNED_16(uint8_t, dst, [8 * 8 * 2]);
179  LOCAL_ALIGNED_16(uint8_t, dst0, [8 * 8 * 2]);
180  LOCAL_ALIGNED_16(uint8_t, dst1_base, [8 * 8 * 2 + 32]);
181  LOCAL_ALIGNED_16(int16_t, coef, [8 * 8 * 2]);
182  LOCAL_ALIGNED_16(int16_t, subcoef0, [8 * 8 * 2]);
183  LOCAL_ALIGNED_16(int16_t, subcoef1, [8 * 8 * 2]);
185  int bit_depth, sz, align, dc, i;
186  declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *dst, int16_t *block, int stride);
187 
188  for (i = 0; i < FF_ARRAY_ELEMS(depths); i++) {
189  bit_depth = depths[i];
191 
192  for (dc = 0; dc <= 2; dc++) {
193  for (sz = 4; sz <= 8; sz += 4) {
194  void (*idct)(uint8_t *, int16_t *, int) = NULL;
195  const char fmts[3][28] = {
196  "h264_idct%d_add_%dbpp", "h264_idct%d_dc_add_%dbpp",
197  "h264_add_pixels%d_%dbpp",
198  };
199 
201 
202  if (sz == 4)
203  dct4x4(coef, bit_depth);
204  else
205  dct8x8(coef, bit_depth);
206 
207  switch ((sz << 2) | dc) {
208  case (4 << 2) | 0: idct = h.h264_idct_add; break;
209  case (4 << 2) | 1: idct = h.h264_idct_dc_add; break;
210  case (4 << 2) | 2: idct = h.h264_add_pixels4_clear; break;
211  case (8 << 2) | 0: idct = h.h264_idct8_add; break;
212  case (8 << 2) | 1: idct = h.h264_idct8_dc_add; break;
213  case (8 << 2) | 2: idct = h.h264_add_pixels8_clear; break;
214  }
215 
216  if (check_func(idct, fmts[dc], sz, bit_depth)) {
217  for (align = 0; align < 16; align += sz * SIZEOF_PIXEL) {
218  uint8_t *dst1 = dst1_base + align;
219  if (dc) {
220  memset(subcoef0, 0, sz * sz * SIZEOF_COEF);
221  memcpy(subcoef0, coef, SIZEOF_COEF);
222  } else {
223  memcpy(subcoef0, coef, sz * sz * SIZEOF_COEF);
224  }
225  memcpy(dst0, dst, sz * PIXEL_STRIDE);
226  memcpy(dst1, dst, sz * PIXEL_STRIDE);
227  memcpy(subcoef1, subcoef0, sz * sz * SIZEOF_COEF);
228  call_ref(dst0, subcoef0, PIXEL_STRIDE);
229  call_new(dst1, subcoef1, PIXEL_STRIDE);
230  if (memcmp(dst0, dst1, sz * PIXEL_STRIDE) ||
231  memcmp(subcoef0, subcoef1, sz * sz * SIZEOF_COEF))
232  fail();
233  bench_new(dst1, subcoef1, sz * SIZEOF_PIXEL);
234  }
235  }
236  }
237  }
238  }
239 }
240 
241 static void check_idct_multiple(void)
242 {
243  LOCAL_ALIGNED_16(uint8_t, dst_full, [16 * 16 * 2]);
244  LOCAL_ALIGNED_16(int16_t, coef_full, [16 * 16 * 2]);
245  LOCAL_ALIGNED_16(uint8_t, dst0, [16 * 16 * 2]);
246  LOCAL_ALIGNED_16(uint8_t, dst1, [16 * 16 * 2]);
247  LOCAL_ALIGNED_16(int16_t, coef0, [16 * 16 * 2]);
248  LOCAL_ALIGNED_16(int16_t, coef1, [16 * 16 * 2]);
249  LOCAL_ALIGNED_16(uint8_t, nnzc, [15 * 8]);
251  int bit_depth, i, y, func;
252  declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]);
253 
254  for (bit_depth = 8; bit_depth <= 10; bit_depth++) {
256  for (func = 0; func < 3; func++) {
257  void (*idct)(uint8_t *, const int *, int16_t *, int, const uint8_t[]) = NULL;
258  const char *name;
259  int sz = 4, intra = 0;
260  int block_offset[16] = { 0 };
261  switch (func) {
262  case 0:
263  idct = h.h264_idct_add16;
264  name = "h264_idct_add16";
265  break;
266  case 1:
267  idct = h.h264_idct_add16intra;
268  name = "h264_idct_add16intra";
269  intra = 1;
270  break;
271  case 2:
272  idct = h.h264_idct8_add4;
273  name = "h264_idct8_add4";
274  sz = 8;
275  break;
276  }
277  memset(nnzc, 0, 15 * 8);
278  memset(coef_full, 0, 16 * 16 * SIZEOF_COEF);
279  for (i = 0; i < 16 * 16; i += sz * sz) {
280  uint8_t src[8 * 8 * 2];
281  uint8_t dst[8 * 8 * 2];
282  int16_t coef[8 * 8 * 2];
283  int index = i / sz;
284  int block_y = (index / 16) * sz;
285  int block_x = index % 16;
286  int offset = (block_y * 16 + block_x) * SIZEOF_PIXEL;
287  int nnz = rnd() % 3;
288 
290  if (sz == 4)
291  dct4x4(coef, bit_depth);
292  else
293  dct8x8(coef, bit_depth);
294 
295  for (y = 0; y < sz; y++)
296  memcpy(&dst_full[offset + y * 16 * SIZEOF_PIXEL],
297  &dst[PIXEL_STRIDE * y], sz * SIZEOF_PIXEL);
298 
299  if (nnz > 1)
300  nnz = sz * sz;
301  memcpy(&coef_full[i * SIZEOF_COEF/sizeof(coef[0])],
302  coef, nnz * SIZEOF_COEF);
303 
304  if (intra && nnz == 1)
305  nnz = 0;
306 
307  nnzc[scan8[i / 16]] = nnz;
308  block_offset[i / 16] = offset;
309  }
310 
311  if (check_func(idct, "%s_%dbpp", name, bit_depth)) {
312  memcpy(coef0, coef_full, 16 * 16 * SIZEOF_COEF);
313  memcpy(coef1, coef_full, 16 * 16 * SIZEOF_COEF);
314  memcpy(dst0, dst_full, 16 * 16 * SIZEOF_PIXEL);
315  memcpy(dst1, dst_full, 16 * 16 * SIZEOF_PIXEL);
316  call_ref(dst0, block_offset, coef0, 16 * SIZEOF_PIXEL, nnzc);
317  call_new(dst1, block_offset, coef1, 16 * SIZEOF_PIXEL, nnzc);
318  if (memcmp(dst0, dst1, 16 * 16 * SIZEOF_PIXEL) ||
319  memcmp(coef0, coef1, 16 * 16 * SIZEOF_COEF))
320  fail();
321  bench_new(dst1, block_offset, coef1, 16 * SIZEOF_PIXEL, nnzc);
322  }
323  }
324  }
325 }
326 
327 
328 static void check_loop_filter(void)
329 {
330  LOCAL_ALIGNED_16(uint8_t, dst, [32 * 16 * 2]);
331  LOCAL_ALIGNED_16(uint8_t, dst0, [32 * 16 * 2]);
332  LOCAL_ALIGNED_16(uint8_t, dst1, [32 * 16 * 2]);
334  int bit_depth;
335  int alphas[36], betas[36];
336  int8_t tc0[36][4];
337 
338  declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *pix, ptrdiff_t stride,
339  int alpha, int beta, int8_t *tc0);
340 
341  for (bit_depth = 8; bit_depth <= 10; bit_depth++) {
342  int i, j, a, c;
343  uint32_t mask = pixel_mask_lf[bit_depth - 8];
345  for (i = 35, a = 255, c = 250; i >= 0; i--) {
346  alphas[i] = a << (bit_depth - 8);
347  betas[i] = (i + 1) / 2 << (bit_depth - 8);
348  tc0[i][0] = tc0[i][3] = (c + 6) / 10;
349  tc0[i][1] = (c + 7) / 15;
350  tc0[i][2] = (c + 9) / 20;
351  a = a*9/10;
352  c = c*9/10;
353  }
354 
355 #define CHECK_LOOP_FILTER(name, align, idc) \
356  do { \
357  if (check_func(h.name, #name #idc "_%dbpp", bit_depth)) { \
358  for (j = 0; j < 36; j++) { \
359  intptr_t off = 8 * 32 + (j & 15) * 4 * !align; \
360  for (i = 0; i < 1024; i+=4) { \
361  AV_WN32A(dst + i, rnd() & mask); \
362  } \
363  memcpy(dst0, dst, 32 * 16 * 2); \
364  memcpy(dst1, dst, 32 * 16 * 2); \
365  \
366  call_ref(dst0 + off, 32, alphas[j], betas[j], tc0[j]); \
367  call_new(dst1 + off, 32, alphas[j], betas[j], tc0[j]); \
368  if (memcmp(dst0, dst1, 32 * 16 * SIZEOF_PIXEL)) { \
369  fprintf(stderr, #name #idc ": j:%d, alpha:%d beta:%d " \
370  "tc0:{%d,%d,%d,%d}\n", j, alphas[j], betas[j], \
371  tc0[j][0], tc0[j][1], tc0[j][2], tc0[j][3]); \
372  fail(); \
373  } \
374  bench_new(dst1 + off, 32, alphas[j], betas[j], tc0[j]);\
375  } \
376  } \
377  } while (0)
378 
385 
389 #undef CHECK_LOOP_FILTER
390  }
391 }
392 
393 static void check_loop_filter_intra(void)
394 {
395  LOCAL_ALIGNED_16(uint8_t, dst, [32 * 16 * 2]);
396  LOCAL_ALIGNED_16(uint8_t, dst0, [32 * 16 * 2]);
397  LOCAL_ALIGNED_16(uint8_t, dst1, [32 * 16 * 2]);
399  int bit_depth;
400  int alphas[36], betas[36];
401 
402  declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *pix, ptrdiff_t stride,
403  int alpha, int beta);
404 
405  for (bit_depth = 8; bit_depth <= 10; bit_depth++) {
406  int i, j, a;
407  uint32_t mask = pixel_mask_lf[bit_depth - 8];
409  for (i = 35, a = 255; i >= 0; i--) {
410  alphas[i] = a << (bit_depth - 8);
411  betas[i] = (i + 1) / 2 << (bit_depth - 8);
412  a = a*9/10;
413  }
414 
415 #define CHECK_LOOP_FILTER(name, align, idc) \
416  do { \
417  if (check_func(h.name, #name #idc "_%dbpp", bit_depth)) { \
418  for (j = 0; j < 36; j++) { \
419  intptr_t off = 8 * 32 + (j & 15) * 4 * !align; \
420  for (i = 0; i < 1024; i+=4) { \
421  AV_WN32A(dst + i, rnd() & mask); \
422  } \
423  memcpy(dst0, dst, 32 * 16 * 2); \
424  memcpy(dst1, dst, 32 * 16 * 2); \
425  \
426  call_ref(dst0 + off, 32, alphas[j], betas[j]); \
427  call_new(dst1 + off, 32, alphas[j], betas[j]); \
428  if (memcmp(dst0, dst1, 32 * 16 * SIZEOF_PIXEL)) { \
429  fprintf(stderr, #name #idc ": j:%d, alpha:%d beta:%d\n", \
430  j, alphas[j], betas[j]); \
431  fail(); \
432  } \
433  bench_new(dst1 + off, 32, alphas[j], betas[j]); \
434  } \
435  } \
436  } while (0)
437 
444 
448 #undef CHECK_LOOP_FILTER
449  }
450 }
451 
453 {
454  check_idct();
456  report("idct");
457 
459  report("loop_filter");
460 
462  report("loop_filter_intra");
463 }
CHECK_LOOP_FILTER
#define CHECK_LOOP_FILTER(name, align, idc)
func
int(* func)(AVBPrint *dst, const char *in, const char *arg)
Definition: jacosubdec.c:68
declare_func_emms
#define declare_func_emms(cpu_flags, ret,...)
Definition: checkasm.h:190
name
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 default minimum maximum flags name is the option name
Definition: writing_filters.txt:88
mem_internal.h
check_idct
static void check_idct(void)
Definition: h264dsp.c:174
mask
int mask
Definition: mediacodecdec_common.c:154
h264_parse.h
SIZEOF_PIXEL
#define SIZEOF_PIXEL
Definition: h264dsp.c:33
check_func
#define check_func(func,...)
Definition: checkasm.h:184
h264_v_loop_filter_luma_intra
static void FUNCC() h264_v_loop_filter_luma_intra(uint8_t *pix, ptrdiff_t stride, int alpha, int beta)
Definition: h264dsp_template.c:218
call_ref
#define call_ref(...)
Definition: checkasm.h:199
bit_depth
static void bit_depth(AudioStatsContext *s, const uint64_t *const mask, uint8_t *depth)
Definition: af_astats.c:246
h264_h_loop_filter_chroma_mbaff_intra
static void FUNCC() h264_h_loop_filter_chroma_mbaff_intra(uint8_t *pix, ptrdiff_t stride, int alpha, int beta)
Definition: h264dsp_template.c:317
fail
#define fail()
Definition: checkasm.h:193
pixel_mask
static const uint32_t pixel_mask[5]
Definition: h264dsp.c:30
h264_h_loop_filter_luma_mbaff_intra
static void FUNCC() h264_h_loop_filter_luma_mbaff_intra(uint8_t *pix, ptrdiff_t stride, int alpha, int beta)
Definition: h264dsp_template.c:226
checkasm.h
scan8
static const uint8_t scan8[16 *3+3]
Definition: h264_parse.h:40
idct
static void idct(int16_t block[64])
Definition: 4xm.c:167
h264_h_loop_filter_chroma_mbaff
static void FUNCC() h264_h_loop_filter_chroma_mbaff(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp_template.c:272
h264_h_loop_filter_luma_intra
static void FUNCC() h264_h_loop_filter_luma_intra(uint8_t *pix, ptrdiff_t stride, int alpha, int beta)
Definition: h264dsp_template.c:222
rnd
#define rnd()
Definition: checkasm.h:177
checkasm_check_h264dsp
void checkasm_check_h264dsp(void)
Definition: h264dsp.c:452
FF_ARRAY_ELEMS
#define FF_ARRAY_ELEMS(a)
Definition: sinewin_tablegen.c:29
intreadwrite.h
LOCAL_ALIGNED_16
#define LOCAL_ALIGNED_16(t, v,...)
Definition: mem_internal.h:128
h264data.h
call_new
#define call_new(...)
Definition: checkasm.h:302
NULL
#define NULL
Definition: coverity.c:32
h264dsp.h
index
int index
Definition: gxfenc.c:90
c
Undefined Behavior In the C some operations are like signed integer dereferencing freed accessing outside allocated Undefined Behavior must not occur in a C it is not safe even if the output of undefined operations is unused The unsafety may seem nit picking but Optimizing compilers have in fact optimized code on the assumption that no undefined Behavior occurs Optimizing code based on wrong assumptions can and has in some cases lead to effects beyond the output of computations The signed integer overflow problem in speed critical code Code which is highly optimized and works with signed integers sometimes has the problem that often the output of the computation does not c
Definition: undefined.txt:32
dct8x8
static void dct8x8(int16_t *coef, int bit_depth)
Definition: h264dsp.c:164
dct8x8_impl
#define dct8x8_impl(size, dctcoef)
Definition: h264dsp.c:120
H264DSPContext
Context for storing H.264 DSP functions.
Definition: h264dsp.h:42
dc
Tag MUST be and< 10hcoeff half pel interpolation filter coefficients, hcoeff[0] are the 2 middle coefficients[1] are the next outer ones and so on, resulting in a filter like:...eff[2], hcoeff[1], hcoeff[0], hcoeff[0], hcoeff[1], hcoeff[2] ... the sign of the coefficients is not explicitly stored but alternates after each coeff and coeff[0] is positive, so ...,+,-,+,-,+,+,-,+,-,+,... hcoeff[0] is not explicitly stored but found by subtracting the sum of all stored coefficients with signs from 32 hcoeff[0]=32 - hcoeff[1] - hcoeff[2] - ... a good choice for hcoeff and htaps is htaps=6 hcoeff={40,-10, 2} an alternative which requires more computations at both encoder and decoder side and may or may not be better is htaps=8 hcoeff={42,-14, 6,-2}ref_frames minimum of the number of available reference frames and max_ref_frames for example the first frame after a key frame always has ref_frames=1spatial_decomposition_type wavelet type 0 is a 9/7 symmetric compact integer wavelet 1 is a 5/3 symmetric compact integer wavelet others are reserved stored as delta from last, last is reset to 0 if always_reset||keyframeqlog quality(logarithmic quantizer scale) stored as delta from last, last is reset to 0 if always_reset||keyframemv_scale stored as delta from last, last is reset to 0 if always_reset||keyframe FIXME check that everything works fine if this changes between framesqbias dequantization bias stored as delta from last, last is reset to 0 if always_reset||keyframeblock_max_depth maximum depth of the block tree stored as delta from last, last is reset to 0 if always_reset||keyframequant_table quantization tableHighlevel bitstream structure:==============================--------------------------------------------|Header|--------------------------------------------|------------------------------------|||Block0||||split?||||yes no||||......... intra?||||:Block01 :yes no||||:Block02 :....... ..........||||:Block03 ::y DC ::ref index:||||:Block04 ::cb DC ::motion x :||||......... :cr DC ::motion y :||||....... ..........|||------------------------------------||------------------------------------|||Block1|||...|--------------------------------------------|------------ ------------ ------------|||Y subbands||Cb subbands||Cr subbands||||--- ---||--- ---||--- ---|||||LL0||HL0||||LL0||HL0||||LL0||HL0|||||--- ---||--- ---||--- ---||||--- ---||--- ---||--- ---|||||LH0||HH0||||LH0||HH0||||LH0||HH0|||||--- ---||--- ---||--- ---||||--- ---||--- ---||--- ---|||||HL1||LH1||||HL1||LH1||||HL1||LH1|||||--- ---||--- ---||--- ---||||--- ---||--- ---||--- ---|||||HH1||HL2||||HH1||HL2||||HH1||HL2|||||...||...||...|||------------ ------------ ------------|--------------------------------------------Decoding process:=================------------|||Subbands|------------||||------------|Intra DC||||LL0 subband prediction ------------|\ Dequantization ------------------- \||Reference frames|\ IDWT|------- -------|Motion \|||Frame 0||Frame 1||Compensation . OBMC v -------|------- -------|--------------. \------> Frame n output Frame Frame<----------------------------------/|...|------------------- Range Coder:============Binary Range Coder:------------------- The implemented range coder is an adapted version based upon "Range encoding: an algorithm for removing redundancy from a digitised message." by G. N. N. Martin. The symbols encoded by the Snow range coder are bits(0|1). The associated probabilities are not fix but change depending on the symbol mix seen so far. bit seen|new state ---------+----------------------------------------------- 0|256 - state_transition_table[256 - old_state];1|state_transition_table[old_state];state_transition_table={ 0, 0, 0, 0, 0, 0, 0, 0, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 190, 191, 192, 194, 194, 195, 196, 197, 198, 199, 200, 201, 202, 202, 204, 205, 206, 207, 208, 209, 209, 210, 211, 212, 213, 215, 215, 216, 217, 218, 219, 220, 220, 222, 223, 224, 225, 226, 227, 227, 229, 229, 230, 231, 232, 234, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 248, 0, 0, 0, 0, 0, 0, 0};FIXME Range Coding of integers:------------------------- FIXME Neighboring Blocks:===================left and top are set to the respective blocks unless they are outside of the image in which case they are set to the Null block top-left is set to the top left block unless it is outside of the image in which case it is set to the left block if this block has no larger parent block or it is at the left side of its parent block and the top right block is not outside of the image then the top right block is used for top-right else the top-left block is used Null block y, cb, cr are 128 level, ref, mx and my are 0 Motion Vector Prediction:=========================1. the motion vectors of all the neighboring blocks are scaled to compensate for the difference of reference frames scaled_mv=(mv *(256 *(current_reference+1)/(mv.reference+1))+128)> the median of the scaled top and top right vectors is used as motion vector prediction the used motion vector is the sum of the predictor and(mvx_diff, mvy_diff) *mv_scale Intra DC Prediction block[y][x] dc[1]
Definition: snow.txt:400
dst
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition: dsp.h:83
h264_v_loop_filter_chroma_intra
static void FUNCC() h264_v_loop_filter_chroma_intra(uint8_t *pix, ptrdiff_t stride, int alpha, int beta)
Definition: h264dsp_template.c:309
align
static const uint8_t *BS_FUNC() align(BSCTX *bc)
Skip bits to a byte boundary.
Definition: bitstream_template.h:411
dct4x4_impl
#define dct4x4_impl(size, dctcoef)
Definition: h264dsp.c:59
a
The reader does not expect b to be semantically here and if the code is changed by maybe adding a a division or other the signedness will almost certainly be mistaken To avoid this confusion a new type was SUINT is the C unsigned type but it holds a signed int to use the same example SUINT a
Definition: undefined.txt:41
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
check_idct_multiple
static void check_idct_multiple(void)
Definition: h264dsp.c:241
h264_h_loop_filter_luma
static void FUNCC() h264_h_loop_filter_luma(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp_template.c:156
check_loop_filter_intra
static void check_loop_filter_intra(void)
Definition: h264dsp.c:393
report
#define report
Definition: checkasm.h:196
check_loop_filter
static void check_loop_filter(void)
Definition: h264dsp.c:328
SIZEOF_COEF
#define SIZEOF_COEF
Definition: h264dsp.c:34
bench_new
#define bench_new(...)
Definition: checkasm.h:373
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:256
PIXEL_STRIDE
#define PIXEL_STRIDE
Definition: h264dsp.c:35
common.h
randomize_buffers
#define randomize_buffers(idx)
Definition: h264dsp.c:37
stride
#define stride
Definition: h264pred_template.c:537
dct4x4
static void dct4x4(int16_t *coef)
Definition: vp8dsp.c:47
ff_h264dsp_init
av_cold void ff_h264dsp_init(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
Definition: h264dsp.c:66
AV_CPU_FLAG_MMX
#define AV_CPU_FLAG_MMX
standard MMX
Definition: cpu.h:30
h264_v_loop_filter_chroma
static void FUNCC() h264_v_loop_filter_chroma(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp_template.c:264
h264_h_loop_filter_chroma
static void FUNCC() h264_h_loop_filter_chroma(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp_template.c:268
h264_h_loop_filter_chroma_intra
static void FUNCC() h264_h_loop_filter_chroma_intra(uint8_t *pix, ptrdiff_t stride, int alpha, int beta)
Definition: h264dsp_template.c:313
alpha
static const int16_t alpha[]
Definition: ilbcdata.h:55
int32_t
int32_t
Definition: audioconvert.c:56
block
The exact code depends on how similar the blocks are and how related they are to the block
Definition: filter_design.txt:207
h264_h_loop_filter_luma_mbaff
static void FUNCC() h264_h_loop_filter_luma_mbaff(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp_template.c:160
h
h
Definition: vp9dsp_template.c:2070
h264_v_loop_filter_luma
static void FUNCC() h264_v_loop_filter_luma(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp_template.c:152
src
#define src
Definition: vp8dsp.c:248
pixel_mask_lf
static const uint32_t pixel_mask_lf[3]
Definition: h264dsp.c:31