FFmpeg
Loading...
Searching...
No Matches
h264dsp_altivec.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
23#include <stdint.h>
24#include <string.h>
25
26#include "libavutil/attributes.h"
27#include "libavutil/intreadwrite.h"
28#include "libavutil/mem_internal.h"
29#include "libavutil/ppc/util_altivec.h"
30
31#include "libavcodec/h264dec.h"
32#include "libavcodec/h264dsp.h"
33
34#include "h264dsp_altivec.h"
35
36/****************************************************************************
37 * IDCT transform:
38 ****************************************************************************/
39
40#define VEC_1D_DCT(vb0,vb1,vb2,vb3,va0,va1,va2,va3) \
41 /* 1st stage */ \
42 vz0 = vec_add(vb0,vb2); /* temp[0] = Y[0] + Y[2] */ \
43 vz1 = vec_sub(vb0,vb2); /* temp[1] = Y[0] - Y[2] */ \
44 vz2 = vec_sra(vb1,vec_splat_u16(1)); \
45 vz2 = vec_sub(vz2,vb3); /* temp[2] = Y[1].1/2 - Y[3] */ \
46 vz3 = vec_sra(vb3,vec_splat_u16(1)); \
47 vz3 = vec_add(vb1,vz3); /* temp[3] = Y[1] + Y[3].1/2 */ \
48 /* 2nd stage: output */ \
49 va0 = vec_add(vz0,vz3); /* x[0] = temp[0] + temp[3] */ \
50 va1 = vec_add(vz1,vz2); /* x[1] = temp[1] + temp[2] */ \
51 va2 = vec_sub(vz1,vz2); /* x[2] = temp[1] - temp[2] */ \
52 va3 = vec_sub(vz0,vz3) /* x[3] = temp[0] - temp[3] */
53
54#define VEC_TRANSPOSE_4(a0,a1,a2,a3,b0,b1,b2,b3) \
55 b0 = vec_mergeh( a0, a0 ); \
56 b1 = vec_mergeh( a1, a0 ); \
57 b2 = vec_mergeh( a2, a0 ); \
58 b3 = vec_mergeh( a3, a0 ); \
59 a0 = vec_mergeh( b0, b2 ); \
60 a1 = vec_mergel( b0, b2 ); \
61 a2 = vec_mergeh( b1, b3 ); \
62 a3 = vec_mergel( b1, b3 ); \
63 b0 = vec_mergeh( a0, a2 ); \
64 b1 = vec_mergel( a0, a2 ); \
65 b2 = vec_mergeh( a1, a3 ); \
66 b3 = vec_mergel( a1, a3 )
67
68#if HAVE_BIGENDIAN
69#define vdst_load(d) \
70 vdst_orig = vec_ld(0, dst); \
71 vdst = vec_perm(vdst_orig, zero_u8v, vdst_mask);
72#else
73#define vdst_load(d) vdst = vec_vsx_ld(0, dst)
74#endif
75
76#define VEC_LOAD_U8_ADD_S16_STORE_U8(va) \
77 vdst_load(); \
78 vdst_ss = (vec_s16) VEC_MERGEH(zero_u8v, vdst); \
79 va = vec_add(va, vdst_ss); \
80 va_u8 = vec_packsu(va, zero_s16v); \
81 va_u32 = vec_splat((vec_u32)va_u8, 0); \
82 vec_ste(va_u32, element, (uint32_t*)dst);
83
84static void h264_idct_add_altivec(uint8_t *dst, int16_t *block, ptrdiff_t stride)
85{
86 vec_s16 va0, va1, va2, va3;
87 vec_s16 vz0, vz1, vz2, vz3;
88 vec_s16 vtmp0, vtmp1, vtmp2, vtmp3;
89 vec_u8 va_u8;
90 vec_u32 va_u32;
91 vec_s16 vdst_ss;
92 const vec_u16 v6us = vec_splat_u16(6);
93 vec_u8 vdst, vdst_orig;
94 vec_u8 vdst_mask = vec_lvsl(0, dst);
95 int element = ((unsigned long)dst & 0xf) >> 2;
96 LOAD_ZERO;
97
98 block[0] += 32; /* add 32 as a DC-level for rounding */
99
100 vtmp0 = vec_ld(0,block);
101 vtmp1 = vec_sld(vtmp0, vtmp0, 8);
102 vtmp2 = vec_ld(16,block);
103 vtmp3 = vec_sld(vtmp2, vtmp2, 8);
104 memset(block, 0, 16 * sizeof(int16_t));
105
106 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
107 VEC_TRANSPOSE_4(va0,va1,va2,va3,vtmp0,vtmp1,vtmp2,vtmp3);
108 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
109
110 va0 = vec_sra(va0,v6us);
111 va1 = vec_sra(va1,v6us);
112 va2 = vec_sra(va2,v6us);
113 va3 = vec_sra(va3,v6us);
114
115 VEC_LOAD_U8_ADD_S16_STORE_U8(va0);
116 dst += stride;
117 VEC_LOAD_U8_ADD_S16_STORE_U8(va1);
118 dst += stride;
119 VEC_LOAD_U8_ADD_S16_STORE_U8(va2);
120 dst += stride;
121 VEC_LOAD_U8_ADD_S16_STORE_U8(va3);
122}
123
124#define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, d4, d5, d6, d7) {\
125 /* a0 = SRC(0) + SRC(4); */ \
126 vec_s16 a0v = vec_add(s0, s4); \
127 /* a2 = SRC(0) - SRC(4); */ \
128 vec_s16 a2v = vec_sub(s0, s4); \
129 /* a4 = (SRC(2)>>1) - SRC(6); */ \
130 vec_s16 a4v = vec_sub(vec_sra(s2, onev), s6); \
131 /* a6 = (SRC(6)>>1) + SRC(2); */ \
132 vec_s16 a6v = vec_add(vec_sra(s6, onev), s2); \
133 /* b0 = a0 + a6; */ \
134 vec_s16 b0v = vec_add(a0v, a6v); \
135 /* b2 = a2 + a4; */ \
136 vec_s16 b2v = vec_add(a2v, a4v); \
137 /* b4 = a2 - a4; */ \
138 vec_s16 b4v = vec_sub(a2v, a4v); \
139 /* b6 = a0 - a6; */ \
140 vec_s16 b6v = vec_sub(a0v, a6v); \
141 /* a1 = SRC(5) - SRC(3) - SRC(7) - (SRC(7)>>1); */ \
142 /* a1 = (SRC(5)-SRC(3)) - (SRC(7) + (SRC(7)>>1)); */ \
143 vec_s16 a1v = vec_sub( vec_sub(s5, s3), vec_add(s7, vec_sra(s7, onev)) ); \
144 /* a3 = SRC(7) + SRC(1) - SRC(3) - (SRC(3)>>1); */ \
145 /* a3 = (SRC(7)+SRC(1)) - (SRC(3) + (SRC(3)>>1)); */ \
146 vec_s16 a3v = vec_sub( vec_add(s7, s1), vec_add(s3, vec_sra(s3, onev)) );\
147 /* a5 = SRC(7) - SRC(1) + SRC(5) + (SRC(5)>>1); */ \
148 /* a5 = (SRC(7)-SRC(1)) + SRC(5) + (SRC(5)>>1); */ \
149 vec_s16 a5v = vec_add( vec_sub(s7, s1), vec_add(s5, vec_sra(s5, onev)) );\
150 /* a7 = SRC(5)+SRC(3) + SRC(1) + (SRC(1)>>1); */ \
151 vec_s16 a7v = vec_add( vec_add(s5, s3), vec_add(s1, vec_sra(s1, onev)) );\
152 /* b1 = (a7>>2) + a1; */ \
153 vec_s16 b1v = vec_add( vec_sra(a7v, twov), a1v); \
154 /* b3 = a3 + (a5>>2); */ \
155 vec_s16 b3v = vec_add(a3v, vec_sra(a5v, twov)); \
156 /* b5 = (a3>>2) - a5; */ \
157 vec_s16 b5v = vec_sub( vec_sra(a3v, twov), a5v); \
158 /* b7 = a7 - (a1>>2); */ \
159 vec_s16 b7v = vec_sub( a7v, vec_sra(a1v, twov)); \
160 /* DST(0, b0 + b7); */ \
161 d0 = vec_add(b0v, b7v); \
162 /* DST(1, b2 + b5); */ \
163 d1 = vec_add(b2v, b5v); \
164 /* DST(2, b4 + b3); */ \
165 d2 = vec_add(b4v, b3v); \
166 /* DST(3, b6 + b1); */ \
167 d3 = vec_add(b6v, b1v); \
168 /* DST(4, b6 - b1); */ \
169 d4 = vec_sub(b6v, b1v); \
170 /* DST(5, b4 - b3); */ \
171 d5 = vec_sub(b4v, b3v); \
172 /* DST(6, b2 - b5); */ \
173 d6 = vec_sub(b2v, b5v); \
174 /* DST(7, b0 - b7); */ \
175 d7 = vec_sub(b0v, b7v); \
176}
177
178#if HAVE_BIGENDIAN
179#define GET_2PERM(ldv, stv, d) \
180 ldv = vec_lvsl(0, d); \
181 stv = vec_lvsr(8, d);
182#define dstv_load(d) \
183 vec_u8 hv = vec_ld( 0, d ); \
184 vec_u8 lv = vec_ld( 7, d); \
185 vec_u8 dstv = vec_perm( hv, lv, (vec_u8)perm_ldv );
186#define dest_unligned_store(d) \
187 vec_u8 edgehv; \
188 vec_u8 bodyv = vec_perm( idstsum8, idstsum8, perm_stv ); \
189 vec_u8 edgelv = vec_perm( sel, zero_u8v, perm_stv ); \
190 lv = vec_sel( lv, bodyv, edgelv ); \
191 vec_st( lv, 7, d ); \
192 hv = vec_ld( 0, d ); \
193 edgehv = vec_perm( zero_u8v, sel, perm_stv ); \
194 hv = vec_sel( hv, bodyv, edgehv ); \
195 vec_st( hv, 0, d );
196#else
197
198#define GET_2PERM(ldv, stv, d) {}
199#define dstv_load(d) vec_u8 dstv = vec_vsx_ld(0, d)
200#define dest_unligned_store(d)\
201 vec_u8 dst8 = vec_perm((vec_u8)idstsum8, dstv, vcprm(2,3,s2,s3));\
202 vec_vsx_st(dst8, 0, d)
203#endif /* HAVE_BIGENDIAN */
204
205#define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel) { \
206 /* unaligned load */ \
207 dstv_load(dest); \
208 vec_s16 idct_sh6 = vec_sra(idctv, sixv); \
209 vec_u16 dst16 = (vec_u16)VEC_MERGEH(zero_u8v, dstv); \
210 vec_s16 idstsum = vec_adds(idct_sh6, (vec_s16)dst16); \
211 vec_u8 idstsum8 = vec_packsu(zero_s16v, idstsum); \
212 /* unaligned store */ \
213 dest_unligned_store(dest);\
214}
215
216static void h264_idct8_add_altivec(uint8_t *dst, int16_t *dct, ptrdiff_t stride)
217{
218 vec_s16 s0, s1, s2, s3, s4, s5, s6, s7;
219 vec_s16 d0, d1, d2, d3, d4, d5, d6, d7;
220 vec_s16 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7;
221
222 vec_u8 perm_ldv, perm_stv;
223 GET_2PERM(perm_ldv, perm_stv, dst);
224
225 const vec_u16 onev = vec_splat_u16(1);
226 const vec_u16 twov = vec_splat_u16(2);
227 const vec_u16 sixv = vec_splat_u16(6);
228
229 const vec_u8 sel = (vec_u8) {0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1};
230 LOAD_ZERO;
231
232 dct[0] += 32; // rounding for the >>6 at the end
233
234 s0 = vec_ld(0x00, (int16_t*)dct);
235 s1 = vec_ld(0x10, (int16_t*)dct);
236 s2 = vec_ld(0x20, (int16_t*)dct);
237 s3 = vec_ld(0x30, (int16_t*)dct);
238 s4 = vec_ld(0x40, (int16_t*)dct);
239 s5 = vec_ld(0x50, (int16_t*)dct);
240 s6 = vec_ld(0x60, (int16_t*)dct);
241 s7 = vec_ld(0x70, (int16_t*)dct);
242 memset(dct, 0, 64 * sizeof(int16_t));
243
244 IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7,
245 d0, d1, d2, d3, d4, d5, d6, d7);
246
247 TRANSPOSE8( d0, d1, d2, d3, d4, d5, d6, d7 );
248
249 IDCT8_1D_ALTIVEC(d0, d1, d2, d3, d4, d5, d6, d7,
250 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7);
251
252 ALTIVEC_STORE_SUM_CLIP(&dst[0*stride], idct0, perm_ldv, perm_stv, sel);
253 ALTIVEC_STORE_SUM_CLIP(&dst[1*stride], idct1, perm_ldv, perm_stv, sel);
254 ALTIVEC_STORE_SUM_CLIP(&dst[2*stride], idct2, perm_ldv, perm_stv, sel);
255 ALTIVEC_STORE_SUM_CLIP(&dst[3*stride], idct3, perm_ldv, perm_stv, sel);
256 ALTIVEC_STORE_SUM_CLIP(&dst[4*stride], idct4, perm_ldv, perm_stv, sel);
257 ALTIVEC_STORE_SUM_CLIP(&dst[5*stride], idct5, perm_ldv, perm_stv, sel);
258 ALTIVEC_STORE_SUM_CLIP(&dst[6*stride], idct6, perm_ldv, perm_stv, sel);
259 ALTIVEC_STORE_SUM_CLIP(&dst[7*stride], idct7, perm_ldv, perm_stv, sel);
260}
261
262#if HAVE_BIGENDIAN
263#define DST_LD vec_ld
264#else
265#define DST_LD vec_vsx_ld
266#endif
267static av_always_inline void h264_idct_dc_add_internal(uint8_t *dst, int16_t *block, ptrdiff_t stride, int size)
268{
269 vec_s16 dc16;
270 vec_u8 dcplus, dcminus, v0, v1, v2, v3, aligner;
271 vec_s32 v_dc32;
272 LOAD_ZERO;
273 DECLARE_ALIGNED(16, int, dc);
274 int i;
275
276 dc = (block[0] + 32) >> 6;
277 block[0] = 0;
278 v_dc32 = vec_lde(0, &dc);
279 dc16 = VEC_SPLAT16((vec_s16)v_dc32, 1);
280
281 if (size == 4)
282 dc16 = VEC_SLD16(dc16, zero_s16v, 8);
283 dcplus = vec_packsu(dc16, zero_s16v);
284 dcminus = vec_packsu(vec_sub(zero_s16v, dc16), zero_s16v);
285
286#if HAVE_BIGENDIAN
287 aligner = vec_lvsr(0, dst);
288 dcplus = vec_perm(dcplus, dcplus, aligner);
289 dcminus = vec_perm(dcminus, dcminus, aligner);
290#endif
291
292 for (i = 0; i < size; i += 4) {
293 v0 = DST_LD(0, dst+0*stride);
294 v1 = DST_LD(0, dst+1*stride);
295 v2 = DST_LD(0, dst+2*stride);
296 v3 = DST_LD(0, dst+3*stride);
297
298 v0 = vec_adds(v0, dcplus);
299 v1 = vec_adds(v1, dcplus);
300 v2 = vec_adds(v2, dcplus);
301 v3 = vec_adds(v3, dcplus);
302
303 v0 = vec_subs(v0, dcminus);
304 v1 = vec_subs(v1, dcminus);
305 v2 = vec_subs(v2, dcminus);
306 v3 = vec_subs(v3, dcminus);
307
308 VEC_ST(v0, 0, dst+0*stride);
309 VEC_ST(v1, 0, dst+1*stride);
310 VEC_ST(v2, 0, dst+2*stride);
311 VEC_ST(v3, 0, dst+3*stride);
312
313 dst += 4*stride;
314 }
315}
316
317static void h264_idct_dc_add_altivec(uint8_t *dst, int16_t *block, ptrdiff_t stride)
318{
319 h264_idct_dc_add_internal(dst, block, stride, 4);
320}
321
322static void h264_idct8_dc_add_altivec(uint8_t *dst, int16_t *block, ptrdiff_t stride)
323{
324 h264_idct_dc_add_internal(dst, block, stride, 8);
325}
326
327static void h264_idct_add16_altivec(uint8_t *dst, const int *block_offset,
328 int16_t *block, ptrdiff_t stride,
329 const uint8_t nnzc[5 * 8])
330{
331 int i;
332 for(i=0; i<16; i++){
333 int nnz = nnzc[ scan8[i] ];
334 if(nnz){
335 if(nnz==1 && block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
336 else h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
337 }
338 }
339}
340
341static void h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset,
342 int16_t *block, ptrdiff_t stride,
343 const uint8_t nnzc[5 * 8])
344{
345 int i;
346 for(i=0; i<16; i++){
347 if(nnzc[ scan8[i] ]) h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
348 else if(block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
349 }
350}
351
352static void h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset,
353 int16_t *block, ptrdiff_t stride,
354 const uint8_t nnzc[5 * 8])
355{
356 int i;
357 for(i=0; i<16; i+=4){
358 int nnz = nnzc[ scan8[i] ];
359 if(nnz){
360 if(nnz==1 && block[i*16]) h264_idct8_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
361 else h264_idct8_add_altivec(dst + block_offset[i], block + i*16, stride);
362 }
363 }
364}
365
366static void h264_idct_add8_altivec(uint8_t **dest, const int *block_offset,
367 int16_t *block, ptrdiff_t stride,
368 const uint8_t nnzc[15 * 8])
369{
370 int i, j;
371 for (j = 1; j < 3; j++) {
372 for(i = j * 16; i < j * 16 + 4; i++){
373 if(nnzc[ scan8[i] ])
374 h264_idct_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
375 else if(block[i*16])
376 h264_idct_dc_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
377 }
378 }
379}
380
381#define transpose4x16(r0, r1, r2, r3) { \
382 register vec_u8 r4; \
383 register vec_u8 r5; \
384 register vec_u8 r6; \
385 register vec_u8 r7; \
386 \
387 r4 = vec_mergeh(r0, r2); /*0, 2 set 0*/ \
388 r5 = vec_mergel(r0, r2); /*0, 2 set 1*/ \
389 r6 = vec_mergeh(r1, r3); /*1, 3 set 0*/ \
390 r7 = vec_mergel(r1, r3); /*1, 3 set 1*/ \
391 \
392 r0 = vec_mergeh(r4, r6); /*all set 0*/ \
393 r1 = vec_mergel(r4, r6); /*all set 1*/ \
394 r2 = vec_mergeh(r5, r7); /*all set 2*/ \
395 r3 = vec_mergel(r5, r7); /*all set 3*/ \
396}
397
398static inline void write16x4(uint8_t *dst, int dst_stride,
399 register vec_u8 r0, register vec_u8 r1,
400 register vec_u8 r2, register vec_u8 r3) {
401 DECLARE_ALIGNED(16, unsigned char, result)[64];
402 uint32_t *src_int = (uint32_t *)result;
403
404 vec_st(r0, 0, result);
405 vec_st(r1, 16, result);
406 vec_st(r2, 32, result);
407 vec_st(r3, 48, result);
408 /* FIXME: there has to be a better way!!!! */
409 AV_WN32(dst, AV_RN32A(src_int + 0));
410 AV_WN32(dst + dst_stride, AV_RN32A(src_int + 1));
411 AV_WN32(dst + 2 * dst_stride, AV_RN32A(src_int + 2));
412 AV_WN32(dst + 3 * dst_stride, AV_RN32A(src_int + 3));
413 AV_WN32(dst + 4 * dst_stride, AV_RN32A(src_int + 4));
414 AV_WN32(dst + 5 * dst_stride, AV_RN32A(src_int + 5));
415 AV_WN32(dst + 6 * dst_stride, AV_RN32A(src_int + 6));
416 AV_WN32(dst + 7 * dst_stride, AV_RN32A(src_int + 7));
417 AV_WN32(dst + 8 * dst_stride, AV_RN32A(src_int + 8));
418 AV_WN32(dst + 9 * dst_stride, AV_RN32A(src_int + 9));
419 AV_WN32(dst + 10 * dst_stride, AV_RN32A(src_int + 10));
420 AV_WN32(dst + 11 * dst_stride, AV_RN32A(src_int + 11));
421 AV_WN32(dst + 12 * dst_stride, AV_RN32A(src_int + 12));
422 AV_WN32(dst + 13 * dst_stride, AV_RN32A(src_int + 13));
423 AV_WN32(dst + 14 * dst_stride, AV_RN32A(src_int + 14));
424 AV_WN32(dst + 15 * dst_stride, AV_RN32A(src_int + 15));
425}
426
427/** @brief performs a 6x16 transpose of data in src, and stores it to dst
428 @todo FIXME: see if we can't spare some vec_lvsl() by them factorizing
429 out of unaligned_load() */
430#define readAndTranspose16x6(src, src_stride, r8, r9, r10, r11, r12, r13) {\
431 register vec_u8 r0 = unaligned_load(0, src); \
432 register vec_u8 r1 = unaligned_load( src_stride, src); \
433 register vec_u8 r2 = unaligned_load(2* src_stride, src); \
434 register vec_u8 r3 = unaligned_load(3* src_stride, src); \
435 register vec_u8 r4 = unaligned_load(4* src_stride, src); \
436 register vec_u8 r5 = unaligned_load(5* src_stride, src); \
437 register vec_u8 r6 = unaligned_load(6* src_stride, src); \
438 register vec_u8 r7 = unaligned_load(7* src_stride, src); \
439 register vec_u8 r14 = unaligned_load(14*src_stride, src); \
440 register vec_u8 r15 = unaligned_load(15*src_stride, src); \
441 \
442 r8 = unaligned_load( 8*src_stride, src); \
443 r9 = unaligned_load( 9*src_stride, src); \
444 r10 = unaligned_load(10*src_stride, src); \
445 r11 = unaligned_load(11*src_stride, src); \
446 r12 = unaligned_load(12*src_stride, src); \
447 r13 = unaligned_load(13*src_stride, src); \
448 \
449 /*Merge first pairs*/ \
450 r0 = vec_mergeh(r0, r8); /*0, 8*/ \
451 r1 = vec_mergeh(r1, r9); /*1, 9*/ \
452 r2 = vec_mergeh(r2, r10); /*2,10*/ \
453 r3 = vec_mergeh(r3, r11); /*3,11*/ \
454 r4 = vec_mergeh(r4, r12); /*4,12*/ \
455 r5 = vec_mergeh(r5, r13); /*5,13*/ \
456 r6 = vec_mergeh(r6, r14); /*6,14*/ \
457 r7 = vec_mergeh(r7, r15); /*7,15*/ \
458 \
459 /*Merge second pairs*/ \
460 r8 = vec_mergeh(r0, r4); /*0,4, 8,12 set 0*/ \
461 r9 = vec_mergel(r0, r4); /*0,4, 8,12 set 1*/ \
462 r10 = vec_mergeh(r1, r5); /*1,5, 9,13 set 0*/ \
463 r11 = vec_mergel(r1, r5); /*1,5, 9,13 set 1*/ \
464 r12 = vec_mergeh(r2, r6); /*2,6,10,14 set 0*/ \
465 r13 = vec_mergel(r2, r6); /*2,6,10,14 set 1*/ \
466 r14 = vec_mergeh(r3, r7); /*3,7,11,15 set 0*/ \
467 r15 = vec_mergel(r3, r7); /*3,7,11,15 set 1*/ \
468 \
469 /*Third merge*/ \
470 r0 = vec_mergeh(r8, r12); /*0,2,4,6,8,10,12,14 set 0*/ \
471 r1 = vec_mergel(r8, r12); /*0,2,4,6,8,10,12,14 set 1*/ \
472 r2 = vec_mergeh(r9, r13); /*0,2,4,6,8,10,12,14 set 2*/ \
473 r4 = vec_mergeh(r10, r14); /*1,3,5,7,9,11,13,15 set 0*/ \
474 r5 = vec_mergel(r10, r14); /*1,3,5,7,9,11,13,15 set 1*/ \
475 r6 = vec_mergeh(r11, r15); /*1,3,5,7,9,11,13,15 set 2*/ \
476 /* Don't need to compute 3 and 7*/ \
477 \
478 /*Final merge*/ \
479 r8 = vec_mergeh(r0, r4); /*all set 0*/ \
480 r9 = vec_mergel(r0, r4); /*all set 1*/ \
481 r10 = vec_mergeh(r1, r5); /*all set 2*/ \
482 r11 = vec_mergel(r1, r5); /*all set 3*/ \
483 r12 = vec_mergeh(r2, r6); /*all set 4*/ \
484 r13 = vec_mergel(r2, r6); /*all set 5*/ \
485 /* Don't need to compute 14 and 15*/ \
486 \
487}
488
489// out: o = |x-y| < a
490static inline vec_u8 diff_lt_altivec ( register vec_u8 x,
491 register vec_u8 y,
492 register vec_u8 a) {
493
494 register vec_u8 diff = vec_subs(x, y);
495 register vec_u8 diffneg = vec_subs(y, x);
496 register vec_u8 o = vec_or(diff, diffneg); /* |x-y| */
497 o = (vec_u8)vec_cmplt(o, a);
498 return o;
499}
500
501static inline vec_u8 h264_deblock_mask ( register vec_u8 p0,
502 register vec_u8 p1,
503 register vec_u8 q0,
504 register vec_u8 q1,
505 register vec_u8 alpha,
506 register vec_u8 beta) {
507
508 register vec_u8 mask;
509 register vec_u8 tempmask;
510
511 mask = diff_lt_altivec(p0, q0, alpha);
512 tempmask = diff_lt_altivec(p1, p0, beta);
513 mask = vec_and(mask, tempmask);
514 tempmask = diff_lt_altivec(q1, q0, beta);
515 mask = vec_and(mask, tempmask);
516
517 return mask;
518}
519
520// out: newp1 = clip((p2 + ((p0 + q0 + 1) >> 1)) >> 1, p1-tc0, p1+tc0)
521static inline vec_u8 h264_deblock_q1(register vec_u8 p0,
522 register vec_u8 p1,
523 register vec_u8 p2,
524 register vec_u8 q0,
525 register vec_u8 tc0) {
526
527 register vec_u8 average = vec_avg(p0, q0);
528 register vec_u8 temp;
529 register vec_u8 unclipped;
530 register vec_u8 ones;
531 register vec_u8 max;
532 register vec_u8 min;
533 register vec_u8 newp1;
534
535 temp = vec_xor(average, p2);
536 average = vec_avg(average, p2); /*avg(p2, avg(p0, q0)) */
537 ones = vec_splat_u8(1);
538 temp = vec_and(temp, ones); /*(p2^avg(p0, q0)) & 1 */
539 unclipped = vec_subs(average, temp); /*(p2+((p0+q0+1)>>1))>>1 */
540 max = vec_adds(p1, tc0);
541 min = vec_subs(p1, tc0);
542 newp1 = vec_max(min, unclipped);
543 newp1 = vec_min(max, newp1);
544 return newp1;
545}
546
547#define h264_deblock_p0_q0(p0, p1, q0, q1, tc0masked) { \
548 \
549 const vec_u8 A0v = vec_sl(vec_splat_u8(10), vec_splat_u8(4)); \
550 \
551 register vec_u8 pq0bit = vec_xor(p0,q0); \
552 register vec_u8 q1minus; \
553 register vec_u8 p0minus; \
554 register vec_u8 stage1; \
555 register vec_u8 stage2; \
556 register vec_u8 vec160; \
557 register vec_u8 delta; \
558 register vec_u8 deltaneg; \
559 \
560 q1minus = vec_nor(q1, q1); /* 255 - q1 */ \
561 stage1 = vec_avg(p1, q1minus); /* (p1 - q1 + 256)>>1 */ \
562 stage2 = vec_sr(stage1, vec_splat_u8(1)); /* (p1 - q1 + 256)>>2 = 64 + (p1 - q1) >> 2 */ \
563 p0minus = vec_nor(p0, p0); /* 255 - p0 */ \
564 stage1 = vec_avg(q0, p0minus); /* (q0 - p0 + 256)>>1 */ \
565 pq0bit = vec_and(pq0bit, vec_splat_u8(1)); \
566 stage2 = vec_avg(stage2, pq0bit); /* 32 + ((q0 - p0)&1 + (p1 - q1) >> 2 + 1) >> 1 */ \
567 stage2 = vec_adds(stage2, stage1); /* 160 + ((p0 - q0) + (p1 - q1) >> 2 + 1) >> 1 */ \
568 vec160 = vec_ld(0, &A0v); \
569 deltaneg = vec_subs(vec160, stage2); /* -d */ \
570 delta = vec_subs(stage2, vec160); /* d */ \
571 deltaneg = vec_min(tc0masked, deltaneg); \
572 delta = vec_min(tc0masked, delta); \
573 p0 = vec_subs(p0, deltaneg); \
574 q0 = vec_subs(q0, delta); \
575 p0 = vec_adds(p0, delta); \
576 q0 = vec_adds(q0, deltaneg); \
577}
578
579#define h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0) { \
580 DECLARE_ALIGNED(16, unsigned char, temp)[16]; \
581 register vec_u8 alphavec; \
582 register vec_u8 betavec; \
583 register vec_u8 mask; \
584 register vec_u8 p1mask; \
585 register vec_u8 q1mask; \
586 register vector signed char tc0vec; \
587 register vec_u8 finaltc0; \
588 register vec_u8 tc0masked; \
589 register vec_u8 newp1; \
590 register vec_u8 newq1; \
591 \
592 temp[0] = alpha; \
593 temp[1] = beta; \
594 alphavec = vec_ld(0, temp); \
595 betavec = vec_splat(alphavec, 0x1); \
596 alphavec = vec_splat(alphavec, 0x0); \
597 mask = h264_deblock_mask(p0, p1, q0, q1, alphavec, betavec); /*if in block */ \
598 \
599 AV_COPY32(temp, tc0); \
600 tc0vec = vec_ld(0, (signed char*)temp); \
601 tc0vec = vec_mergeh(tc0vec, tc0vec); \
602 tc0vec = vec_mergeh(tc0vec, tc0vec); \
603 mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_s8(-1))); /* if tc0[i] >= 0 */ \
604 finaltc0 = vec_and((vec_u8)tc0vec, mask); /* tc = tc0 */ \
605 \
606 p1mask = diff_lt_altivec(p2, p0, betavec); \
607 p1mask = vec_and(p1mask, mask); /* if ( |p2 - p0| < beta) */ \
608 tc0masked = vec_and(p1mask, (vec_u8)tc0vec); \
609 finaltc0 = vec_sub(finaltc0, p1mask); /* tc++ */ \
610 newp1 = h264_deblock_q1(p0, p1, p2, q0, tc0masked); \
611 /*end if*/ \
612 \
613 q1mask = diff_lt_altivec(q2, q0, betavec); \
614 q1mask = vec_and(q1mask, mask); /* if ( |q2 - q0| < beta ) */\
615 tc0masked = vec_and(q1mask, (vec_u8)tc0vec); \
616 finaltc0 = vec_sub(finaltc0, q1mask); /* tc++ */ \
617 newq1 = h264_deblock_q1(p0, q1, q2, q0, tc0masked); \
618 /*end if*/ \
619 \
620 h264_deblock_p0_q0(p0, p1, q0, q1, finaltc0); \
621 p1 = newp1; \
622 q1 = newq1; \
623}
624
625static void h264_v_loop_filter_luma_altivec(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0) {
626
627 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) >= 0) {
628 register vec_u8 p2 = vec_ld(-3*stride, pix);
629 register vec_u8 p1 = vec_ld(-2*stride, pix);
630 register vec_u8 p0 = vec_ld(-1*stride, pix);
631 register vec_u8 q0 = vec_ld(0, pix);
632 register vec_u8 q1 = vec_ld(stride, pix);
633 register vec_u8 q2 = vec_ld(2*stride, pix);
634 h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0);
635 vec_st(p1, -2*stride, pix);
636 vec_st(p0, -1*stride, pix);
637 vec_st(q0, 0, pix);
638 vec_st(q1, stride, pix);
639 }
640}
641
642static void h264_h_loop_filter_luma_altivec(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0) {
643
644 register vec_u8 line0, line1, line2, line3, line4, line5;
645 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) < 0)
646 return;
647 readAndTranspose16x6(pix-3, stride, line0, line1, line2, line3, line4, line5);
648 h264_loop_filter_luma_altivec(line0, line1, line2, line3, line4, line5, alpha, beta, tc0);
649 transpose4x16(line1, line2, line3, line4);
650 write16x4(pix-2, stride, line1, line2, line3, line4);
651}
652
653static av_always_inline
654void weight_h264_W_altivec(uint8_t *block, ptrdiff_t stride, int height,
655 int log2_denom, int weight, int offset, int w)
656{
657 int y, aligned;
658 vec_u8 vblock;
659 vec_s16 vtemp, vweight, voffset, v0, v1;
660 vec_u16 vlog2_denom;
661 DECLARE_ALIGNED(16, int32_t, temp)[4];
662 LOAD_ZERO;
663
664 offset *= 1 << log2_denom;
665 if(log2_denom) offset += 1<<(log2_denom-1);
666 temp[0] = log2_denom;
667 temp[1] = weight;
668 temp[2] = offset;
669
670 vtemp = (vec_s16)vec_ld(0, temp);
671#if !HAVE_BIGENDIAN
672 vtemp =(vec_s16)vec_perm(vtemp, vtemp, vcswapi2s(0,1,2,3));
673#endif
674 vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
675 vweight = vec_splat(vtemp, 3);
676 voffset = vec_splat(vtemp, 5);
677 aligned = !((unsigned long)block & 0xf);
678
679 for (y = 0; y < height; y++) {
680 vblock = vec_ld(0, block);
681
682 v0 = (vec_s16)VEC_MERGEH(zero_u8v, vblock);
683 v1 = (vec_s16)VEC_MERGEL(zero_u8v, vblock);
684
685 if (w == 16 || aligned) {
686 v0 = vec_mladd(v0, vweight, zero_s16v);
687 v0 = vec_adds(v0, voffset);
688 v0 = vec_sra(v0, vlog2_denom);
689 }
690 if (w == 16 || !aligned) {
691 v1 = vec_mladd(v1, vweight, zero_s16v);
692 v1 = vec_adds(v1, voffset);
693 v1 = vec_sra(v1, vlog2_denom);
694 }
695 vblock = vec_packsu(v0, v1);
696 vec_st(vblock, 0, block);
697
698 block += stride;
699 }
700}
701
702static av_always_inline
703void biweight_h264_W_altivec(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height,
704 int log2_denom, int weightd, int weights, int offset, int w)
705{
706 int y, dst_aligned, src_aligned;
707 vec_u8 vsrc, vdst;
708 vec_s16 vtemp, vweights, vweightd, voffset, v0, v1, v2, v3;
709 vec_u16 vlog2_denom;
710 DECLARE_ALIGNED(16, int32_t, temp)[4];
711 LOAD_ZERO;
712
713 offset = ((offset + 1) | 1) * (1 << log2_denom);
714 temp[0] = log2_denom+1;
715 temp[1] = weights;
716 temp[2] = weightd;
717 temp[3] = offset;
718
719 vtemp = (vec_s16)vec_ld(0, temp);
720#if !HAVE_BIGENDIAN
721 vtemp =(vec_s16)vec_perm(vtemp, vtemp, vcswapi2s(0,1,2,3));
722#endif
723 vlog2_denom = (vec_u16)vec_splat(vtemp, 1);
724 vweights = vec_splat(vtemp, 3);
725 vweightd = vec_splat(vtemp, 5);
726 voffset = vec_splat(vtemp, 7);
727 dst_aligned = !((unsigned long)dst & 0xf);
728 src_aligned = !((unsigned long)src & 0xf);
729
730 for (y = 0; y < height; y++) {
731 vdst = vec_ld(0, dst);
732 vsrc = vec_ld(0, src);
733
734 v0 = (vec_s16)VEC_MERGEH(zero_u8v, vdst);
735 v1 = (vec_s16)VEC_MERGEL(zero_u8v, vdst);
736 v2 = (vec_s16)VEC_MERGEH(zero_u8v, vsrc);
737 v3 = (vec_s16)VEC_MERGEL(zero_u8v, vsrc);
738
739 if (w == 8) {
740 if (src_aligned)
741 v3 = v2;
742 else
743 v2 = v3;
744 }
745
746 if (w == 16 || dst_aligned) {
747 v0 = vec_mladd(v0, vweightd, zero_s16v);
748 v2 = vec_mladd(v2, vweights, zero_s16v);
749
750 v0 = vec_adds(v0, v2);
751 v0 = vec_adds(v0, voffset);
752 v0 = vec_sra(v0, vlog2_denom);
753 }
754 if (w == 16 || !dst_aligned) {
755 v1 = vec_mladd(v1, vweightd, zero_s16v);
756 v3 = vec_mladd(v3, vweights, zero_s16v);
757
758 v1 = vec_adds(v1, v3);
759 v1 = vec_adds(v1, voffset);
760 v1 = vec_sra(v1, vlog2_denom);
761 }
762 vdst = vec_packsu(v0, v1);
763 vec_st(vdst, 0, dst);
764
765 dst += stride;
766 src += stride;
767 }
768}
769
770#define H264_WEIGHT(W) \
771static void weight_h264_pixels ## W ## _altivec(uint8_t *block, ptrdiff_t stride, int height, \
772 int log2_denom, int weight, int offset) \
773{ \
774 weight_h264_W_altivec(block, stride, height, log2_denom, weight, offset, W); \
775}\
776static void biweight_h264_pixels ## W ## _altivec(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, \
777 int log2_denom, int weightd, int weights, int offset) \
778{ \
779 biweight_h264_W_altivec(dst, src, stride, height, log2_denom, weightd, weights, offset, W); \
780}
781
782H264_WEIGHT(16)
783H264_WEIGHT( 8)
784
785av_cold void ff_h264dsp_init_altivec(H264DSPContext *c, const int bit_depth,
786 const int chroma_format_idc)
787{
788 if (bit_depth == 8) {
789 c->idct_add = h264_idct_add_altivec;
790 if (chroma_format_idc <= 1)
791 c->idct_add8 = h264_idct_add8_altivec;
792 c->idct_add16 = h264_idct_add16_altivec;
793 c->idct_add16intra = h264_idct_add16intra_altivec;
794 c->idct_dc_add = h264_idct_dc_add_altivec;
795 c->idct8_dc_add = h264_idct8_dc_add_altivec;
796 c->idct8_add = h264_idct8_add_altivec;
797 c->idct8_add4 = h264_idct8_add4_altivec;
798 c->v_loop_filter_luma = h264_v_loop_filter_luma_altivec;
799 c->h_loop_filter_luma = h264_h_loop_filter_luma_altivec;
800
801 c->weight_pixels_tab[0] = weight_h264_pixels16_altivec;
802 c->weight_pixels_tab[1] = weight_h264_pixels8_altivec;
803 c->biweight_pixels_tab[0] = biweight_h264_pixels16_altivec;
804 c->biweight_pixels_tab[1] = biweight_h264_pixels8_altivec;
805 }
806}
dst
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition dsp.h:97
dct
static void dct(AudioRNNContext *s, float *out, const float *in)
Definition af_arnndn.c:1017
bit_depth
static void bit_depth(AudioStatsContext *s, const uint64_t *const mask, uint8_t *depth)
Definition af_astats.c:246
int32_t
int32_t
Definition audioconvert.c:58
i
#define i(width, name, range_min, range_max)
Definition cbs_h264.c:63
min
#define min(a, b)
Definition cuda_runtime.h:34
max
#define max(a, b)
Definition cuda_runtime.h:33
block
static int16_t block[64]
Definition dct.c:125
idct6
static void idct6(int pre_mant[6])
Calculate 6-point IDCT of the pre-mantissas.
Definition eac3dec.c:165
scan8
static const uint8_t scan8[16 *3+3]
Definition h264_parse.h:40
h264dec.h
H.264 / AVC / MPEG-4 part10 codec.
h264dsp.h
H.264 DSP functions.
ff_h264dsp_init_altivec
av_cold void ff_h264dsp_init_altivec(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
Definition h264dsp_altivec.c:785
h264_deblock_q1
static vec_u8 h264_deblock_q1(register vec_u8 p0, register vec_u8 p1, register vec_u8 p2, register vec_u8 q0, register vec_u8 tc0)
Definition h264dsp_altivec.c:521
VEC_LOAD_U8_ADD_S16_STORE_U8
#define VEC_LOAD_U8_ADD_S16_STORE_U8(va)
Definition h264dsp_altivec.c:76
GET_2PERM
#define GET_2PERM(ldv, stv, d)
Definition h264dsp_altivec.c:198
h264_idct8_add4_altivec
static void h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, int16_t *block, ptrdiff_t stride, const uint8_t nnzc[5 *8])
Definition h264dsp_altivec.c:352
biweight_h264_W_altivec
static av_always_inline void biweight_h264_W_altivec(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, int log2_denom, int weightd, int weights, int offset, int w)
Definition h264dsp_altivec.c:703
h264_loop_filter_luma_altivec
#define h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0)
Definition h264dsp_altivec.c:579
readAndTranspose16x6
#define readAndTranspose16x6(src, src_stride, r8, r9, r10, r11, r12, r13)
performs a 6x16 transpose of data in src, and stores it to dst
Definition h264dsp_altivec.c:430
ALTIVEC_STORE_SUM_CLIP
#define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel)
Definition h264dsp_altivec.c:205
h264_deblock_mask
static vec_u8 h264_deblock_mask(register vec_u8 p0, register vec_u8 p1, register vec_u8 q0, register vec_u8 q1, register vec_u8 alpha, register vec_u8 beta)
Definition h264dsp_altivec.c:501
diff_lt_altivec
static vec_u8 diff_lt_altivec(register vec_u8 x, register vec_u8 y, register vec_u8 a)
Definition h264dsp_altivec.c:490
h264_idct_add8_altivec
static void h264_idct_add8_altivec(uint8_t **dest, const int *block_offset, int16_t *block, ptrdiff_t stride, const uint8_t nnzc[15 *8])
Definition h264dsp_altivec.c:366
h264_v_loop_filter_luma_altivec
static void h264_v_loop_filter_luma_altivec(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0)
Definition h264dsp_altivec.c:625
h264_idct_add16intra_altivec
static void h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset, int16_t *block, ptrdiff_t stride, const uint8_t nnzc[5 *8])
Definition h264dsp_altivec.c:341
VEC_1D_DCT
#define VEC_1D_DCT(vb0, vb1, vb2, vb3, va0, va1, va2, va3)
Definition h264dsp_altivec.c:40
h264_idct8_dc_add_altivec
static void h264_idct8_dc_add_altivec(uint8_t *dst, int16_t *block, ptrdiff_t stride)
Definition h264dsp_altivec.c:322
H264_WEIGHT
#define H264_WEIGHT(W)
Definition h264dsp_altivec.c:770
transpose4x16
#define transpose4x16(r0, r1, r2, r3)
Definition h264dsp_altivec.c:381
h264_idct8_add_altivec
static void h264_idct8_add_altivec(uint8_t *dst, int16_t *dct, ptrdiff_t stride)
Definition h264dsp_altivec.c:216
IDCT8_1D_ALTIVEC
#define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, d4, d5, d6, d7)
Definition h264dsp_altivec.c:124
VEC_TRANSPOSE_4
#define VEC_TRANSPOSE_4(a0, a1, a2, a3, b0, b1, b2, b3)
Definition h264dsp_altivec.c:54
weight_h264_W_altivec
static av_always_inline void weight_h264_W_altivec(uint8_t *block, ptrdiff_t stride, int height, int log2_denom, int weight, int offset, int w)
Definition h264dsp_altivec.c:654
h264_h_loop_filter_luma_altivec
static void h264_h_loop_filter_luma_altivec(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0)
Definition h264dsp_altivec.c:642
h264_idct_dc_add_altivec
static void h264_idct_dc_add_altivec(uint8_t *dst, int16_t *block, ptrdiff_t stride)
Definition h264dsp_altivec.c:317
h264_idct_add_altivec
static void h264_idct_add_altivec(uint8_t *dst, int16_t *block, ptrdiff_t stride)
Definition h264dsp_altivec.c:84
h264_idct_add16_altivec
static void h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, int16_t *block, ptrdiff_t stride, const uint8_t nnzc[5 *8])
Definition h264dsp_altivec.c:327
write16x4
static void write16x4(uint8_t *dst, int dst_stride, register vec_u8 r0, register vec_u8 r1, register vec_u8 r2, register vec_u8 r3)
Definition h264dsp_altivec.c:398
h264_idct_dc_add_internal
static av_always_inline void h264_idct_dc_add_internal(uint8_t *dst, int16_t *block, ptrdiff_t stride, int size)
Definition h264dsp_altivec.c:267
DST_LD
#define DST_LD
Definition h264dsp_altivec.c:265
h264dsp_altivec.h
H264_WEIGHT
#define H264_WEIGHT(W)
Definition h264dsp_template.c:32
a
int a
Definition h264pred_template.c:416
weights
static const int weights[]
Definition hevc_pel.c:32
alpha
static const int16_t alpha[]
Definition ilbcdata.h:55
AV_WN32
#define AV_WN32(p, v)
Definition intreadwrite.h:372
AV_RN32A
#define AV_RN32A(p)
Definition intreadwrite.h:522
offset
unsigned offset
Definition libaomenc.c:763
attributes.h
Macro definitions for various function/variable attributes.
av_always_inline
#define av_always_inline
Definition attributes.h:72
av_cold
#define av_cold
Definition attributes.h:117
w
uint8_t w
Definition llvidencdsp.c:39
mask
static const uint16_t mask[17]
Definition lzw.c:38
mem_internal.h
DECLARE_ALIGNED
#define DECLARE_ALIGNED(n, t, v)
Declare a variable that is aligned in memory.
Definition mem_internal.h:104
pix
enum AVPixelFormat pix
Definition ohcodec.c:55
weight
const h264_weight_func weight
Definition h264dsp_init.c:33
H264DSPContext
Context for storing H.264 DSP functions.
Definition h264dsp.h:42
stride
#define stride
Definition h264pred_template.c:536
src
#define src
Definition vp8dsp.c:248
height
#define height
Definition dsp.h:99
size
int size
Definition twinvq_data.h:10344
q1
static const uint8_t q1[256]
Definition twofish.c:100
q0
static const uint8_t q0[256]
Definition twofish.c:81
util_altivec.h
Contains misc utility macros and inline functions.
vec_s16
#define vec_s16
Definition util_altivec.h:37
vec_u8
#define vec_u8
Definition util_altivec.h:34
zero_s16v
#define zero_s16v
Definition util_altivec.h:50
zero_u8v
#define zero_u8v
Definition util_altivec.h:47
vec_u32
#define vec_u32
Definition util_altivec.h:38
vec_u16
#define vec_u16
Definition util_altivec.h:36
LOAD_ZERO
#define LOAD_ZERO
Definition util_altivec.h:45
vec_s32
#define vec_s32
Definition util_altivec.h:39
temp
else temp
Definition vf_mcdeint.c:275
diff
static av_always_inline int diff(const struct color_info *a, const struct color_info *b, const int trans_thresh)
Definition vf_paletteuse.c:166
c
static double c[64]
Definition vsrc_mptestsrc.c:89
Generated on Sat Nov 7 2026 19:21:02 for FFmpeg by doxygen 1.13.2