FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
h264dsp.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (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 GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 #include "config.h"
22 #include "libavutil/attributes.h"
23 #include "libavutil/cpu.h"
24 #include "libavutil/intreadwrite.h"
25 #include "libavutil/ppc/cpu.h"
28 #include "libavcodec/h264data.h"
29 #include "libavcodec/h264dsp.h"
30 
31 #if HAVE_ALTIVEC
32 
33 /****************************************************************************
34  * IDCT transform:
35  ****************************************************************************/
36 
37 #define VEC_1D_DCT(vb0,vb1,vb2,vb3,va0,va1,va2,va3) \
38  /* 1st stage */ \
39  vz0 = vec_add(vb0,vb2); /* temp[0] = Y[0] + Y[2] */ \
40  vz1 = vec_sub(vb0,vb2); /* temp[1] = Y[0] - Y[2] */ \
41  vz2 = vec_sra(vb1,vec_splat_u16(1)); \
42  vz2 = vec_sub(vz2,vb3); /* temp[2] = Y[1].1/2 - Y[3] */ \
43  vz3 = vec_sra(vb3,vec_splat_u16(1)); \
44  vz3 = vec_add(vb1,vz3); /* temp[3] = Y[1] + Y[3].1/2 */ \
45  /* 2nd stage: output */ \
46  va0 = vec_add(vz0,vz3); /* x[0] = temp[0] + temp[3] */ \
47  va1 = vec_add(vz1,vz2); /* x[1] = temp[1] + temp[2] */ \
48  va2 = vec_sub(vz1,vz2); /* x[2] = temp[1] - temp[2] */ \
49  va3 = vec_sub(vz0,vz3) /* x[3] = temp[0] - temp[3] */
50 
51 #define VEC_TRANSPOSE_4(a0,a1,a2,a3,b0,b1,b2,b3) \
52  b0 = vec_mergeh( a0, a0 ); \
53  b1 = vec_mergeh( a1, a0 ); \
54  b2 = vec_mergeh( a2, a0 ); \
55  b3 = vec_mergeh( a3, a0 ); \
56  a0 = vec_mergeh( b0, b2 ); \
57  a1 = vec_mergel( b0, b2 ); \
58  a2 = vec_mergeh( b1, b3 ); \
59  a3 = vec_mergel( b1, b3 ); \
60  b0 = vec_mergeh( a0, a2 ); \
61  b1 = vec_mergel( a0, a2 ); \
62  b2 = vec_mergeh( a1, a3 ); \
63  b3 = vec_mergel( a1, a3 )
64 
65 #if HAVE_BIGENDIAN
66 #define vdst_load(d) \
67  vdst_orig = vec_ld(0, dst); \
68  vdst = vec_perm(vdst_orig, zero_u8v, vdst_mask);
69 #else
70 #define vdst_load(d) vdst = vec_vsx_ld(0, dst)
71 #endif
72 
73 #define VEC_LOAD_U8_ADD_S16_STORE_U8(va) \
74  vdst_load(); \
75  vdst_ss = (vec_s16) VEC_MERGEH(zero_u8v, vdst); \
76  va = vec_add(va, vdst_ss); \
77  va_u8 = vec_packsu(va, zero_s16v); \
78  va_u32 = vec_splat((vec_u32)va_u8, 0); \
79  vec_ste(va_u32, element, (uint32_t*)dst);
80 
81 static void h264_idct_add_altivec(uint8_t *dst, int16_t *block, int stride)
82 {
83  vec_s16 va0, va1, va2, va3;
84  vec_s16 vz0, vz1, vz2, vz3;
85  vec_s16 vtmp0, vtmp1, vtmp2, vtmp3;
86  vec_u8 va_u8;
87  vec_u32 va_u32;
88  vec_s16 vdst_ss;
89  const vec_u16 v6us = vec_splat_u16(6);
90  vec_u8 vdst, vdst_orig;
91  vec_u8 vdst_mask = vec_lvsl(0, dst);
92  int element = ((unsigned long)dst & 0xf) >> 2;
93  LOAD_ZERO;
94 
95  block[0] += 32; /* add 32 as a DC-level for rounding */
96 
97  vtmp0 = vec_ld(0,block);
98  vtmp1 = vec_sld(vtmp0, vtmp0, 8);
99  vtmp2 = vec_ld(16,block);
100  vtmp3 = vec_sld(vtmp2, vtmp2, 8);
101  memset(block, 0, 16 * sizeof(int16_t));
102 
103  VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
104  VEC_TRANSPOSE_4(va0,va1,va2,va3,vtmp0,vtmp1,vtmp2,vtmp3);
105  VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
106 
107  va0 = vec_sra(va0,v6us);
108  va1 = vec_sra(va1,v6us);
109  va2 = vec_sra(va2,v6us);
110  va3 = vec_sra(va3,v6us);
111 
112  VEC_LOAD_U8_ADD_S16_STORE_U8(va0);
113  dst += stride;
114  VEC_LOAD_U8_ADD_S16_STORE_U8(va1);
115  dst += stride;
116  VEC_LOAD_U8_ADD_S16_STORE_U8(va2);
117  dst += stride;
118  VEC_LOAD_U8_ADD_S16_STORE_U8(va3);
119 }
120 
121 #define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, d4, d5, d6, d7) {\
122  /* a0 = SRC(0) + SRC(4); */ \
123  vec_s16 a0v = vec_add(s0, s4); \
124  /* a2 = SRC(0) - SRC(4); */ \
125  vec_s16 a2v = vec_sub(s0, s4); \
126  /* a4 = (SRC(2)>>1) - SRC(6); */ \
127  vec_s16 a4v = vec_sub(vec_sra(s2, onev), s6); \
128  /* a6 = (SRC(6)>>1) + SRC(2); */ \
129  vec_s16 a6v = vec_add(vec_sra(s6, onev), s2); \
130  /* b0 = a0 + a6; */ \
131  vec_s16 b0v = vec_add(a0v, a6v); \
132  /* b2 = a2 + a4; */ \
133  vec_s16 b2v = vec_add(a2v, a4v); \
134  /* b4 = a2 - a4; */ \
135  vec_s16 b4v = vec_sub(a2v, a4v); \
136  /* b6 = a0 - a6; */ \
137  vec_s16 b6v = vec_sub(a0v, a6v); \
138  /* a1 = SRC(5) - SRC(3) - SRC(7) - (SRC(7)>>1); */ \
139  /* a1 = (SRC(5)-SRC(3)) - (SRC(7) + (SRC(7)>>1)); */ \
140  vec_s16 a1v = vec_sub( vec_sub(s5, s3), vec_add(s7, vec_sra(s7, onev)) ); \
141  /* a3 = SRC(7) + SRC(1) - SRC(3) - (SRC(3)>>1); */ \
142  /* a3 = (SRC(7)+SRC(1)) - (SRC(3) + (SRC(3)>>1)); */ \
143  vec_s16 a3v = vec_sub( vec_add(s7, s1), vec_add(s3, vec_sra(s3, onev)) );\
144  /* a5 = SRC(7) - SRC(1) + SRC(5) + (SRC(5)>>1); */ \
145  /* a5 = (SRC(7)-SRC(1)) + SRC(5) + (SRC(5)>>1); */ \
146  vec_s16 a5v = vec_add( vec_sub(s7, s1), vec_add(s5, vec_sra(s5, onev)) );\
147  /* a7 = SRC(5)+SRC(3) + SRC(1) + (SRC(1)>>1); */ \
148  vec_s16 a7v = vec_add( vec_add(s5, s3), vec_add(s1, vec_sra(s1, onev)) );\
149  /* b1 = (a7>>2) + a1; */ \
150  vec_s16 b1v = vec_add( vec_sra(a7v, twov), a1v); \
151  /* b3 = a3 + (a5>>2); */ \
152  vec_s16 b3v = vec_add(a3v, vec_sra(a5v, twov)); \
153  /* b5 = (a3>>2) - a5; */ \
154  vec_s16 b5v = vec_sub( vec_sra(a3v, twov), a5v); \
155  /* b7 = a7 - (a1>>2); */ \
156  vec_s16 b7v = vec_sub( a7v, vec_sra(a1v, twov)); \
157  /* DST(0, b0 + b7); */ \
158  d0 = vec_add(b0v, b7v); \
159  /* DST(1, b2 + b5); */ \
160  d1 = vec_add(b2v, b5v); \
161  /* DST(2, b4 + b3); */ \
162  d2 = vec_add(b4v, b3v); \
163  /* DST(3, b6 + b1); */ \
164  d3 = vec_add(b6v, b1v); \
165  /* DST(4, b6 - b1); */ \
166  d4 = vec_sub(b6v, b1v); \
167  /* DST(5, b4 - b3); */ \
168  d5 = vec_sub(b4v, b3v); \
169  /* DST(6, b2 - b5); */ \
170  d6 = vec_sub(b2v, b5v); \
171  /* DST(7, b0 - b7); */ \
172  d7 = vec_sub(b0v, b7v); \
173 }
174 
175 #if HAVE_BIGENDIAN
176 #define GET_2PERM(ldv, stv, d) \
177  ldv = vec_lvsl(0, d); \
178  stv = vec_lvsr(8, d);
179 #define dstv_load(d) \
180  vec_u8 hv = vec_ld( 0, d ); \
181  vec_u8 lv = vec_ld( 7, d); \
182  vec_u8 dstv = vec_perm( hv, lv, (vec_u8)perm_ldv );
183 #define dest_unligned_store(d) \
184  vec_u8 edgehv; \
185  vec_u8 bodyv = vec_perm( idstsum8, idstsum8, perm_stv ); \
186  vec_u8 edgelv = vec_perm( sel, zero_u8v, perm_stv ); \
187  lv = vec_sel( lv, bodyv, edgelv ); \
188  vec_st( lv, 7, d ); \
189  hv = vec_ld( 0, d ); \
190  edgehv = vec_perm( zero_u8v, sel, perm_stv ); \
191  hv = vec_sel( hv, bodyv, edgehv ); \
192  vec_st( hv, 0, d );
193 #else
194 
195 #define GET_2PERM(ldv, stv, d) {}
196 #define dstv_load(d) vec_u8 dstv = vec_vsx_ld(0, d)
197 #define dest_unligned_store(d)\
198  vec_u8 dst8 = vec_perm((vec_u8)idstsum8, dstv, vcprm(2,3,s2,s3));\
199  vec_vsx_st(dst8, 0, d)
200 #endif /* HAVE_BIGENDIAN */
201 
202 #define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel) { \
203  /* unaligned load */ \
204  dstv_load(dest); \
205  vec_s16 idct_sh6 = vec_sra(idctv, sixv); \
206  vec_u16 dst16 = (vec_u16)VEC_MERGEH(zero_u8v, dstv); \
207  vec_s16 idstsum = vec_adds(idct_sh6, (vec_s16)dst16); \
208  vec_u8 idstsum8 = vec_packsu(zero_s16v, idstsum); \
209  /* unaligned store */ \
210  dest_unligned_store(dest);\
211 }
212 
213 static void h264_idct8_add_altivec(uint8_t *dst, int16_t *dct, int stride)
214 {
215  vec_s16 s0, s1, s2, s3, s4, s5, s6, s7;
216  vec_s16 d0, d1, d2, d3, d4, d5, d6, d7;
217  vec_s16 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7;
218 
219  vec_u8 perm_ldv, perm_stv;
220  GET_2PERM(perm_ldv, perm_stv, dst);
221 
222  const vec_u16 onev = vec_splat_u16(1);
223  const vec_u16 twov = vec_splat_u16(2);
224  const vec_u16 sixv = vec_splat_u16(6);
225 
226  const vec_u8 sel = (vec_u8) {0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1};
227  LOAD_ZERO;
228 
229  dct[0] += 32; // rounding for the >>6 at the end
230 
231  s0 = vec_ld(0x00, (int16_t*)dct);
232  s1 = vec_ld(0x10, (int16_t*)dct);
233  s2 = vec_ld(0x20, (int16_t*)dct);
234  s3 = vec_ld(0x30, (int16_t*)dct);
235  s4 = vec_ld(0x40, (int16_t*)dct);
236  s5 = vec_ld(0x50, (int16_t*)dct);
237  s6 = vec_ld(0x60, (int16_t*)dct);
238  s7 = vec_ld(0x70, (int16_t*)dct);
239  memset(dct, 0, 64 * sizeof(int16_t));
240 
241  IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7,
242  d0, d1, d2, d3, d4, d5, d6, d7);
243 
244  TRANSPOSE8( d0, d1, d2, d3, d4, d5, d6, d7 );
245 
246  IDCT8_1D_ALTIVEC(d0, d1, d2, d3, d4, d5, d6, d7,
247  idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7);
248 
249  ALTIVEC_STORE_SUM_CLIP(&dst[0*stride], idct0, perm_ldv, perm_stv, sel);
250  ALTIVEC_STORE_SUM_CLIP(&dst[1*stride], idct1, perm_ldv, perm_stv, sel);
251  ALTIVEC_STORE_SUM_CLIP(&dst[2*stride], idct2, perm_ldv, perm_stv, sel);
252  ALTIVEC_STORE_SUM_CLIP(&dst[3*stride], idct3, perm_ldv, perm_stv, sel);
253  ALTIVEC_STORE_SUM_CLIP(&dst[4*stride], idct4, perm_ldv, perm_stv, sel);
254  ALTIVEC_STORE_SUM_CLIP(&dst[5*stride], idct5, perm_ldv, perm_stv, sel);
255  ALTIVEC_STORE_SUM_CLIP(&dst[6*stride], idct6, perm_ldv, perm_stv, sel);
256  ALTIVEC_STORE_SUM_CLIP(&dst[7*stride], idct7, perm_ldv, perm_stv, sel);
257 }
258 
259 #if HAVE_BIGENDIAN
260 #define DST_LD vec_ld
261 #else
262 #define DST_LD vec_vsx_ld
263 #endif
264 static av_always_inline void h264_idct_dc_add_internal(uint8_t *dst, int16_t *block, int stride, int size)
265 {
266  vec_s16 dc16;
267  vec_u8 dcplus, dcminus, v0, v1, v2, v3, aligner;
268  vec_s32 v_dc32;
269  LOAD_ZERO;
270  DECLARE_ALIGNED(16, int, dc);
271  int i;
272 
273  dc = (block[0] + 32) >> 6;
274  block[0] = 0;
275  v_dc32 = vec_lde(0, &dc);
276  dc16 = VEC_SPLAT16((vec_s16)v_dc32, 1);
277 
278  if (size == 4)
279  dc16 = VEC_SLD16(dc16, zero_s16v, 8);
280  dcplus = vec_packsu(dc16, zero_s16v);
281  dcminus = vec_packsu(vec_sub(zero_s16v, dc16), zero_s16v);
282 
283 #if HAVE_BIGENDIAN
284  aligner = vec_lvsr(0, dst);
285  dcplus = vec_perm(dcplus, dcplus, aligner);
286  dcminus = vec_perm(dcminus, dcminus, aligner);
287 #endif
288 
289  for (i = 0; i < size; i += 4) {
290  v0 = DST_LD(0, dst+0*stride);
291  v1 = DST_LD(0, dst+1*stride);
292  v2 = DST_LD(0, dst+2*stride);
293  v3 = DST_LD(0, dst+3*stride);
294 
295  v0 = vec_adds(v0, dcplus);
296  v1 = vec_adds(v1, dcplus);
297  v2 = vec_adds(v2, dcplus);
298  v3 = vec_adds(v3, dcplus);
299 
300  v0 = vec_subs(v0, dcminus);
301  v1 = vec_subs(v1, dcminus);
302  v2 = vec_subs(v2, dcminus);
303  v3 = vec_subs(v3, dcminus);
304 
305  VEC_ST(v0, 0, dst+0*stride);
306  VEC_ST(v1, 0, dst+1*stride);
307  VEC_ST(v2, 0, dst+2*stride);
308  VEC_ST(v3, 0, dst+3*stride);
309 
310  dst += 4*stride;
311  }
312 }
313 
314 static void h264_idct_dc_add_altivec(uint8_t *dst, int16_t *block, int stride)
315 {
316  h264_idct_dc_add_internal(dst, block, stride, 4);
317 }
318 
319 static void h264_idct8_dc_add_altivec(uint8_t *dst, int16_t *block, int stride)
320 {
321  h264_idct_dc_add_internal(dst, block, stride, 8);
322 }
323 
324 static void h264_idct_add16_altivec(uint8_t *dst, const int *block_offset,
325  int16_t *block, int stride,
326  const uint8_t nnzc[15 * 8])
327 {
328  int i;
329  for(i=0; i<16; i++){
330  int nnz = nnzc[ scan8[i] ];
331  if(nnz){
332  if(nnz==1 && block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
333  else h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
334  }
335  }
336 }
337 
338 static void h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset,
339  int16_t *block, int stride,
340  const uint8_t nnzc[15 * 8])
341 {
342  int i;
343  for(i=0; i<16; i++){
344  if(nnzc[ scan8[i] ]) h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
345  else if(block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
346  }
347 }
348 
349 static void h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset,
350  int16_t *block, int stride,
351  const uint8_t nnzc[15 * 8])
352 {
353  int i;
354  for(i=0; i<16; i+=4){
355  int nnz = nnzc[ scan8[i] ];
356  if(nnz){
357  if(nnz==1 && block[i*16]) h264_idct8_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
358  else h264_idct8_add_altivec(dst + block_offset[i], block + i*16, stride);
359  }
360  }
361 }
362 
363 static void h264_idct_add8_altivec(uint8_t **dest, const int *block_offset,
364  int16_t *block, int stride,
365  const uint8_t nnzc[15 * 8])
366 {
367  int i, j;
368  for (j = 1; j < 3; j++) {
369  for(i = j * 16; i < j * 16 + 4; i++){
370  if(nnzc[ scan8[i] ])
371  h264_idct_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
372  else if(block[i*16])
373  h264_idct_dc_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
374  }
375  }
376 }
377 
378 #define transpose4x16(r0, r1, r2, r3) { \
379  register vec_u8 r4; \
380  register vec_u8 r5; \
381  register vec_u8 r6; \
382  register vec_u8 r7; \
383  \
384  r4 = vec_mergeh(r0, r2); /*0, 2 set 0*/ \
385  r5 = vec_mergel(r0, r2); /*0, 2 set 1*/ \
386  r6 = vec_mergeh(r1, r3); /*1, 3 set 0*/ \
387  r7 = vec_mergel(r1, r3); /*1, 3 set 1*/ \
388  \
389  r0 = vec_mergeh(r4, r6); /*all set 0*/ \
390  r1 = vec_mergel(r4, r6); /*all set 1*/ \
391  r2 = vec_mergeh(r5, r7); /*all set 2*/ \
392  r3 = vec_mergel(r5, r7); /*all set 3*/ \
393 }
394 
395 static inline void write16x4(uint8_t *dst, int dst_stride,
396  register vec_u8 r0, register vec_u8 r1,
397  register vec_u8 r2, register vec_u8 r3) {
398  DECLARE_ALIGNED(16, unsigned char, result)[64];
399  uint32_t *src_int = (uint32_t *)result, *dst_int = (uint32_t *)dst;
400  int int_dst_stride = dst_stride/4;
401 
402  vec_st(r0, 0, result);
403  vec_st(r1, 16, result);
404  vec_st(r2, 32, result);
405  vec_st(r3, 48, result);
406  /* FIXME: there has to be a better way!!!! */
407  *dst_int = *src_int;
408  *(dst_int+ int_dst_stride) = *(src_int + 1);
409  *(dst_int+ 2*int_dst_stride) = *(src_int + 2);
410  *(dst_int+ 3*int_dst_stride) = *(src_int + 3);
411  *(dst_int+ 4*int_dst_stride) = *(src_int + 4);
412  *(dst_int+ 5*int_dst_stride) = *(src_int + 5);
413  *(dst_int+ 6*int_dst_stride) = *(src_int + 6);
414  *(dst_int+ 7*int_dst_stride) = *(src_int + 7);
415  *(dst_int+ 8*int_dst_stride) = *(src_int + 8);
416  *(dst_int+ 9*int_dst_stride) = *(src_int + 9);
417  *(dst_int+10*int_dst_stride) = *(src_int + 10);
418  *(dst_int+11*int_dst_stride) = *(src_int + 11);
419  *(dst_int+12*int_dst_stride) = *(src_int + 12);
420  *(dst_int+13*int_dst_stride) = *(src_int + 13);
421  *(dst_int+14*int_dst_stride) = *(src_int + 14);
422  *(dst_int+15*int_dst_stride) = *(src_int + 15);
423 }
424 
425 /** @brief performs a 6x16 transpose of data in src, and stores it to dst
426  @todo FIXME: see if we can't spare some vec_lvsl() by them factorizing
427  out of unaligned_load() */
428 #define readAndTranspose16x6(src, src_stride, r8, r9, r10, r11, r12, r13) {\
429  register vec_u8 r0 = unaligned_load(0, src); \
430  register vec_u8 r1 = unaligned_load( src_stride, src); \
431  register vec_u8 r2 = unaligned_load(2* src_stride, src); \
432  register vec_u8 r3 = unaligned_load(3* src_stride, src); \
433  register vec_u8 r4 = unaligned_load(4* src_stride, src); \
434  register vec_u8 r5 = unaligned_load(5* src_stride, src); \
435  register vec_u8 r6 = unaligned_load(6* src_stride, src); \
436  register vec_u8 r7 = unaligned_load(7* src_stride, src); \
437  register vec_u8 r14 = unaligned_load(14*src_stride, src); \
438  register vec_u8 r15 = unaligned_load(15*src_stride, src); \
439  \
440  r8 = unaligned_load( 8*src_stride, src); \
441  r9 = unaligned_load( 9*src_stride, src); \
442  r10 = unaligned_load(10*src_stride, src); \
443  r11 = unaligned_load(11*src_stride, src); \
444  r12 = unaligned_load(12*src_stride, src); \
445  r13 = unaligned_load(13*src_stride, src); \
446  \
447  /*Merge first pairs*/ \
448  r0 = vec_mergeh(r0, r8); /*0, 8*/ \
449  r1 = vec_mergeh(r1, r9); /*1, 9*/ \
450  r2 = vec_mergeh(r2, r10); /*2,10*/ \
451  r3 = vec_mergeh(r3, r11); /*3,11*/ \
452  r4 = vec_mergeh(r4, r12); /*4,12*/ \
453  r5 = vec_mergeh(r5, r13); /*5,13*/ \
454  r6 = vec_mergeh(r6, r14); /*6,14*/ \
455  r7 = vec_mergeh(r7, r15); /*7,15*/ \
456  \
457  /*Merge second pairs*/ \
458  r8 = vec_mergeh(r0, r4); /*0,4, 8,12 set 0*/ \
459  r9 = vec_mergel(r0, r4); /*0,4, 8,12 set 1*/ \
460  r10 = vec_mergeh(r1, r5); /*1,5, 9,13 set 0*/ \
461  r11 = vec_mergel(r1, r5); /*1,5, 9,13 set 1*/ \
462  r12 = vec_mergeh(r2, r6); /*2,6,10,14 set 0*/ \
463  r13 = vec_mergel(r2, r6); /*2,6,10,14 set 1*/ \
464  r14 = vec_mergeh(r3, r7); /*3,7,11,15 set 0*/ \
465  r15 = vec_mergel(r3, r7); /*3,7,11,15 set 1*/ \
466  \
467  /*Third merge*/ \
468  r0 = vec_mergeh(r8, r12); /*0,2,4,6,8,10,12,14 set 0*/ \
469  r1 = vec_mergel(r8, r12); /*0,2,4,6,8,10,12,14 set 1*/ \
470  r2 = vec_mergeh(r9, r13); /*0,2,4,6,8,10,12,14 set 2*/ \
471  r4 = vec_mergeh(r10, r14); /*1,3,5,7,9,11,13,15 set 0*/ \
472  r5 = vec_mergel(r10, r14); /*1,3,5,7,9,11,13,15 set 1*/ \
473  r6 = vec_mergeh(r11, r15); /*1,3,5,7,9,11,13,15 set 2*/ \
474  /* Don't need to compute 3 and 7*/ \
475  \
476  /*Final merge*/ \
477  r8 = vec_mergeh(r0, r4); /*all set 0*/ \
478  r9 = vec_mergel(r0, r4); /*all set 1*/ \
479  r10 = vec_mergeh(r1, r5); /*all set 2*/ \
480  r11 = vec_mergel(r1, r5); /*all set 3*/ \
481  r12 = vec_mergeh(r2, r6); /*all set 4*/ \
482  r13 = vec_mergel(r2, r6); /*all set 5*/ \
483  /* Don't need to compute 14 and 15*/ \
484  \
485 }
486 
487 // out: o = |x-y| < a
488 static inline vec_u8 diff_lt_altivec ( register vec_u8 x,
489  register vec_u8 y,
490  register vec_u8 a) {
491 
492  register vec_u8 diff = vec_subs(x, y);
493  register vec_u8 diffneg = vec_subs(y, x);
494  register vec_u8 o = vec_or(diff, diffneg); /* |x-y| */
495  o = (vec_u8)vec_cmplt(o, a);
496  return o;
497 }
498 
499 static inline vec_u8 h264_deblock_mask ( register vec_u8 p0,
500  register vec_u8 p1,
501  register vec_u8 q0,
502  register vec_u8 q1,
503  register vec_u8 alpha,
504  register vec_u8 beta) {
505 
506  register vec_u8 mask;
507  register vec_u8 tempmask;
508 
509  mask = diff_lt_altivec(p0, q0, alpha);
510  tempmask = diff_lt_altivec(p1, p0, beta);
511  mask = vec_and(mask, tempmask);
512  tempmask = diff_lt_altivec(q1, q0, beta);
513  mask = vec_and(mask, tempmask);
514 
515  return mask;
516 }
517 
518 // out: newp1 = clip((p2 + ((p0 + q0 + 1) >> 1)) >> 1, p1-tc0, p1+tc0)
519 static inline vec_u8 h264_deblock_q1(register vec_u8 p0,
520  register vec_u8 p1,
521  register vec_u8 p2,
522  register vec_u8 q0,
523  register vec_u8 tc0) {
524 
525  register vec_u8 average = vec_avg(p0, q0);
526  register vec_u8 temp;
527  register vec_u8 uncliped;
528  register vec_u8 ones;
529  register vec_u8 max;
530  register vec_u8 min;
531  register vec_u8 newp1;
532 
533  temp = vec_xor(average, p2);
534  average = vec_avg(average, p2); /*avg(p2, avg(p0, q0)) */
535  ones = vec_splat_u8(1);
536  temp = vec_and(temp, ones); /*(p2^avg(p0, q0)) & 1 */
537  uncliped = vec_subs(average, temp); /*(p2+((p0+q0+1)>>1))>>1 */
538  max = vec_adds(p1, tc0);
539  min = vec_subs(p1, tc0);
540  newp1 = vec_max(min, uncliped);
541  newp1 = vec_min(max, newp1);
542  return newp1;
543 }
544 
545 #define h264_deblock_p0_q0(p0, p1, q0, q1, tc0masked) { \
546  \
547  const vec_u8 A0v = vec_sl(vec_splat_u8(10), vec_splat_u8(4)); \
548  \
549  register vec_u8 pq0bit = vec_xor(p0,q0); \
550  register vec_u8 q1minus; \
551  register vec_u8 p0minus; \
552  register vec_u8 stage1; \
553  register vec_u8 stage2; \
554  register vec_u8 vec160; \
555  register vec_u8 delta; \
556  register vec_u8 deltaneg; \
557  \
558  q1minus = vec_nor(q1, q1); /* 255 - q1 */ \
559  stage1 = vec_avg(p1, q1minus); /* (p1 - q1 + 256)>>1 */ \
560  stage2 = vec_sr(stage1, vec_splat_u8(1)); /* (p1 - q1 + 256)>>2 = 64 + (p1 - q1) >> 2 */ \
561  p0minus = vec_nor(p0, p0); /* 255 - p0 */ \
562  stage1 = vec_avg(q0, p0minus); /* (q0 - p0 + 256)>>1 */ \
563  pq0bit = vec_and(pq0bit, vec_splat_u8(1)); \
564  stage2 = vec_avg(stage2, pq0bit); /* 32 + ((q0 - p0)&1 + (p1 - q1) >> 2 + 1) >> 1 */ \
565  stage2 = vec_adds(stage2, stage1); /* 160 + ((p0 - q0) + (p1 - q1) >> 2 + 1) >> 1 */ \
566  vec160 = vec_ld(0, &A0v); \
567  deltaneg = vec_subs(vec160, stage2); /* -d */ \
568  delta = vec_subs(stage2, vec160); /* d */ \
569  deltaneg = vec_min(tc0masked, deltaneg); \
570  delta = vec_min(tc0masked, delta); \
571  p0 = vec_subs(p0, deltaneg); \
572  q0 = vec_subs(q0, delta); \
573  p0 = vec_adds(p0, delta); \
574  q0 = vec_adds(q0, deltaneg); \
575 }
576 
577 #define h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0) { \
578  DECLARE_ALIGNED(16, unsigned char, temp)[16]; \
579  register vec_u8 alphavec; \
580  register vec_u8 betavec; \
581  register vec_u8 mask; \
582  register vec_u8 p1mask; \
583  register vec_u8 q1mask; \
584  register vector signed char tc0vec; \
585  register vec_u8 finaltc0; \
586  register vec_u8 tc0masked; \
587  register vec_u8 newp1; \
588  register vec_u8 newq1; \
589  \
590  temp[0] = alpha; \
591  temp[1] = beta; \
592  alphavec = vec_ld(0, temp); \
593  betavec = vec_splat(alphavec, 0x1); \
594  alphavec = vec_splat(alphavec, 0x0); \
595  mask = h264_deblock_mask(p0, p1, q0, q1, alphavec, betavec); /*if in block */ \
596  \
597  AV_COPY32(temp, tc0); \
598  tc0vec = vec_ld(0, (signed char*)temp); \
599  tc0vec = vec_mergeh(tc0vec, tc0vec); \
600  tc0vec = vec_mergeh(tc0vec, tc0vec); \
601  mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_s8(-1))); /* if tc0[i] >= 0 */ \
602  finaltc0 = vec_and((vec_u8)tc0vec, mask); /* tc = tc0 */ \
603  \
604  p1mask = diff_lt_altivec(p2, p0, betavec); \
605  p1mask = vec_and(p1mask, mask); /* if ( |p2 - p0| < beta) */ \
606  tc0masked = vec_and(p1mask, (vec_u8)tc0vec); \
607  finaltc0 = vec_sub(finaltc0, p1mask); /* tc++ */ \
608  newp1 = h264_deblock_q1(p0, p1, p2, q0, tc0masked); \
609  /*end if*/ \
610  \
611  q1mask = diff_lt_altivec(q2, q0, betavec); \
612  q1mask = vec_and(q1mask, mask); /* if ( |q2 - q0| < beta ) */\
613  tc0masked = vec_and(q1mask, (vec_u8)tc0vec); \
614  finaltc0 = vec_sub(finaltc0, q1mask); /* tc++ */ \
615  newq1 = h264_deblock_q1(p0, q1, q2, q0, tc0masked); \
616  /*end if*/ \
617  \
618  h264_deblock_p0_q0(p0, p1, q0, q1, finaltc0); \
619  p1 = newp1; \
620  q1 = newq1; \
621 }
622 
623 static void h264_v_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) {
624 
625  if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) >= 0) {
626  register vec_u8 p2 = vec_ld(-3*stride, pix);
627  register vec_u8 p1 = vec_ld(-2*stride, pix);
628  register vec_u8 p0 = vec_ld(-1*stride, pix);
629  register vec_u8 q0 = vec_ld(0, pix);
630  register vec_u8 q1 = vec_ld(stride, pix);
631  register vec_u8 q2 = vec_ld(2*stride, pix);
632  h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0);
633  vec_st(p1, -2*stride, pix);
634  vec_st(p0, -1*stride, pix);
635  vec_st(q0, 0, pix);
636  vec_st(q1, stride, pix);
637  }
638 }
639 
640 static void h264_h_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) {
641 
642  register vec_u8 line0, line1, line2, line3, line4, line5;
643  if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) < 0)
644  return;
645  readAndTranspose16x6(pix-3, stride, line0, line1, line2, line3, line4, line5);
646  h264_loop_filter_luma_altivec(line0, line1, line2, line3, line4, line5, alpha, beta, tc0);
647  transpose4x16(line1, line2, line3, line4);
648  write16x4(pix-2, stride, line1, line2, line3, line4);
649 }
650 
651 static av_always_inline
652 void weight_h264_W_altivec(uint8_t *block, int stride, int height,
653  int log2_denom, int weight, int offset, int w)
654 {
655  int y, aligned;
656  vec_u8 vblock;
657  vec_s16 vtemp, vweight, voffset, v0, v1;
658  vec_u16 vlog2_denom;
659  DECLARE_ALIGNED(16, int32_t, temp)[4];
660  LOAD_ZERO;
661 
662  offset <<= log2_denom;
663  if(log2_denom) offset += 1<<(log2_denom-1);
664  temp[0] = log2_denom;
665  temp[1] = weight;
666  temp[2] = offset;
667 
668  vtemp = (vec_s16)vec_ld(0, temp);
669 #if !HAVE_BIGENDIAN
670  vtemp =(vec_s16)vec_perm(vtemp, vtemp, vcswapi2s(0,1,2,3));
671 #endif
672  vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
673  vweight = vec_splat(vtemp, 3);
674  voffset = vec_splat(vtemp, 5);
675  aligned = !((unsigned long)block & 0xf);
676 
677  for (y = 0; y < height; y++) {
678  vblock = vec_ld(0, block);
679 
680  v0 = (vec_s16)VEC_MERGEH(zero_u8v, vblock);
681  v1 = (vec_s16)VEC_MERGEL(zero_u8v, vblock);
682 
683  if (w == 16 || aligned) {
684  v0 = vec_mladd(v0, vweight, zero_s16v);
685  v0 = vec_adds(v0, voffset);
686  v0 = vec_sra(v0, vlog2_denom);
687  }
688  if (w == 16 || !aligned) {
689  v1 = vec_mladd(v1, vweight, zero_s16v);
690  v1 = vec_adds(v1, voffset);
691  v1 = vec_sra(v1, vlog2_denom);
692  }
693  vblock = vec_packsu(v0, v1);
694  vec_st(vblock, 0, block);
695 
696  block += stride;
697  }
698 }
699 
700 static av_always_inline
701 void biweight_h264_W_altivec(uint8_t *dst, uint8_t *src, int stride, int height,
702  int log2_denom, int weightd, int weights, int offset, int w)
703 {
704  int y, dst_aligned, src_aligned;
705  vec_u8 vsrc, vdst;
706  vec_s16 vtemp, vweights, vweightd, voffset, v0, v1, v2, v3;
707  vec_u16 vlog2_denom;
708  DECLARE_ALIGNED(16, int32_t, temp)[4];
709  LOAD_ZERO;
710 
711  offset = ((offset + 1) | 1) << log2_denom;
712  temp[0] = log2_denom+1;
713  temp[1] = weights;
714  temp[2] = weightd;
715  temp[3] = offset;
716 
717  vtemp = (vec_s16)vec_ld(0, temp);
718 #if !HAVE_BIGENDIAN
719  vtemp =(vec_s16)vec_perm(vtemp, vtemp, vcswapi2s(0,1,2,3));
720 #endif
721  vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
722  vweights = vec_splat(vtemp, 3);
723  vweightd = vec_splat(vtemp, 5);
724  voffset = vec_splat(vtemp, 7);
725  dst_aligned = !((unsigned long)dst & 0xf);
726  src_aligned = !((unsigned long)src & 0xf);
727 
728  for (y = 0; y < height; y++) {
729  vdst = vec_ld(0, dst);
730  vsrc = vec_ld(0, src);
731 
732  v0 = (vec_s16)VEC_MERGEH(zero_u8v, vdst);
733  v1 = (vec_s16)VEC_MERGEL(zero_u8v, vdst);
734  v2 = (vec_s16)VEC_MERGEH(zero_u8v, vsrc);
735  v3 = (vec_s16)VEC_MERGEL(zero_u8v, vsrc);
736 
737  if (w == 8) {
738  if (src_aligned)
739  v3 = v2;
740  else
741  v2 = v3;
742  }
743 
744  if (w == 16 || dst_aligned) {
745  v0 = vec_mladd(v0, vweightd, zero_s16v);
746  v2 = vec_mladd(v2, vweights, zero_s16v);
747 
748  v0 = vec_adds(v0, voffset);
749  v0 = vec_adds(v0, v2);
750  v0 = vec_sra(v0, vlog2_denom);
751  }
752  if (w == 16 || !dst_aligned) {
753  v1 = vec_mladd(v1, vweightd, zero_s16v);
754  v3 = vec_mladd(v3, vweights, zero_s16v);
755 
756  v1 = vec_adds(v1, voffset);
757  v1 = vec_adds(v1, v3);
758  v1 = vec_sra(v1, vlog2_denom);
759  }
760  vdst = vec_packsu(v0, v1);
761  vec_st(vdst, 0, dst);
762 
763  dst += stride;
764  src += stride;
765  }
766 }
767 
768 #define H264_WEIGHT(W) \
769 static void weight_h264_pixels ## W ## _altivec(uint8_t *block, int stride, int height, \
770  int log2_denom, int weight, int offset) \
771 { \
772  weight_h264_W_altivec(block, stride, height, log2_denom, weight, offset, W); \
773 }\
774 static void biweight_h264_pixels ## W ## _altivec(uint8_t *dst, uint8_t *src, int stride, int height, \
775  int log2_denom, int weightd, int weights, int offset) \
776 { \
777  biweight_h264_W_altivec(dst, src, stride, height, log2_denom, weightd, weights, offset, W); \
778 }
779 
780 H264_WEIGHT(16)
781 H264_WEIGHT( 8)
782 #endif /* HAVE_ALTIVEC */
783 
785  const int chroma_format_idc)
786 {
787 #if HAVE_ALTIVEC
789  return;
790 
791  if (bit_depth == 8) {
792  c->h264_idct_add = h264_idct_add_altivec;
793  if (chroma_format_idc <= 1)
794  c->h264_idct_add8 = h264_idct_add8_altivec;
795  c->h264_idct_add16 = h264_idct_add16_altivec;
796  c->h264_idct_add16intra = h264_idct_add16intra_altivec;
797  c->h264_idct_dc_add= h264_idct_dc_add_altivec;
798  c->h264_idct8_dc_add = h264_idct8_dc_add_altivec;
799  c->h264_idct8_add = h264_idct8_add_altivec;
800  c->h264_idct8_add4 = h264_idct8_add4_altivec;
801  c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_altivec;
802  c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_altivec;
803 
804  c->weight_h264_pixels_tab[0] = weight_h264_pixels16_altivec;
805  c->weight_h264_pixels_tab[1] = weight_h264_pixels8_altivec;
806  c->biweight_h264_pixels_tab[0] = biweight_h264_pixels16_altivec;
807  c->biweight_h264_pixels_tab[1] = biweight_h264_pixels8_altivec;
808  }
809 #endif /* HAVE_ALTIVEC */
810 }
void(* h264_idct_add)(uint8_t *dst, int16_t *block, int stride)
Definition: h264dsp.h:80
else temp
Definition: vf_mcdeint.c:257
#define vec_s32
Definition: types_altivec.h:32
#define DECLARE_ALIGNED(n, t, v)
Definition: mem.h:53
GLfloat v0
Definition: opengl_enc.c:107
static const uint8_t q1[256]
Definition: twofish.c:96
H.264 DSP functions.
Macro definitions for various function/variable attributes.
void(* h264_idct_add16)(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 *8])
Definition: h264dsp.h:89
uint8_t
#define av_cold
Definition: attributes.h:74
void(* h264_idct_add8)(uint8_t **dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 *8])
Definition: h264dsp.h:95
h264_weight_func weight_h264_pixels_tab[4]
Definition: h264dsp.h:43
#define vec_s16
Definition: types_altivec.h:30
#define zero_s16v
Definition: types_altivec.h:43
ptrdiff_t size
Definition: opengl_enc.c:101
static double alpha(void *priv, double x, double y)
Definition: vf_geq.c:99
#define s2
Definition: regdef.h:39
static const uint16_t mask[17]
Definition: lzw.c:38
#define PPC_ALTIVEC(flags)
Definition: cpu.h:26
void(* h264_idct8_add)(uint8_t *dst, int16_t *block, int stride)
Definition: h264dsp.h:82
#define s0
Definition: regdef.h:37
#define vec_u16
Definition: types_altivec.h:29
#define H264_WEIGHT(W)
#define LOAD_ZERO
Definition: types_altivec.h:38
static const uint8_t offset[127][2]
Definition: vf_spp.c:92
static const uint8_t q0[256]
Definition: twofish.c:77
#define s5
Definition: regdef.h:42
static const uint8_t scan8[16 *3+3]
Definition: h264.h:1007
static void idct6(int pre_mant[6])
Calculate 6-point IDCT of the pre-mantissas.
Definition: eac3dec.c:173
h264_biweight_func biweight_h264_pixels_tab[4]
Definition: h264dsp.h:44
Context for storing H.264 DSP functions.
Definition: h264dsp.h:41
float y
void(* h264_idct_dc_add)(uint8_t *dst, int16_t *block, int stride)
Definition: h264dsp.h:84
int32_t
#define vec_u32
Definition: types_altivec.h:31
#define s4
Definition: regdef.h:41
#define s3
Definition: regdef.h:40
#define vec_u8
Definition: types_altivec.h:27
av_cold void ff_h264dsp_init_ppc(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
Definition: h264dsp.c:784
AVS_Value src
Definition: avisynth_c.h:482
H264 / AVC / MPEG4 part10 codec data table
void(* h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:47
BYTE int const BYTE int int int height
Definition: avisynth_c.h:676
void(* h264_idct8_add4)(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 *8])
Definition: h264dsp.h:92
#define s1
Definition: regdef.h:38
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:76
static int weight(int i, int blen, int offset)
Definition: diracdec.c:1299
void(* h264_idct8_dc_add)(uint8_t *dst, int16_t *block, int stride)
Definition: h264dsp.h:86
Contains misc utility macros and inline functions.
GLint GLenum GLboolean GLsizei stride
Definition: opengl_enc.c:105
static double c[64]
static unsigned bit_depth(uint64_t mask)
Definition: af_astats.c:128
static av_always_inline int diff(const uint32_t a, const uint32_t b)
void(* h264_idct_add16intra)(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 *8])
Definition: h264dsp.h:98
void(* h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:49
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> dc
#define s6
Definition: regdef.h:43
#define av_always_inline
Definition: attributes.h:37
#define stride
float min
#define zero_u8v
Definition: types_altivec.h:40
static int16_t block[64]
Definition: dct-test.c:110