42 #define VEC_1D_DCT(vb0,vb1,vb2,vb3,va0,va1,va2,va3) \ 44 vz0 = vec_add(vb0,vb2); \ 45 vz1 = vec_sub(vb0,vb2); \ 46 vz2 = vec_sra(vb1,vec_splat_u16(1)); \ 47 vz2 = vec_sub(vz2,vb3); \ 48 vz3 = vec_sra(vb3,vec_splat_u16(1)); \ 49 vz3 = vec_add(vb1,vz3); \ 51 va0 = vec_add(vz0,vz3); \ 52 va1 = vec_add(vz1,vz2); \ 53 va2 = vec_sub(vz1,vz2); \ 54 va3 = vec_sub(vz0,vz3) 56 #define VEC_TRANSPOSE_4(a0,a1,a2,a3,b0,b1,b2,b3) \ 57 b0 = vec_mergeh( a0, a0 ); \ 58 b1 = vec_mergeh( a1, a0 ); \ 59 b2 = vec_mergeh( a2, a0 ); \ 60 b3 = vec_mergeh( a3, a0 ); \ 61 a0 = vec_mergeh( b0, b2 ); \ 62 a1 = vec_mergel( b0, b2 ); \ 63 a2 = vec_mergeh( b1, b3 ); \ 64 a3 = vec_mergel( b1, b3 ); \ 65 b0 = vec_mergeh( a0, a2 ); \ 66 b1 = vec_mergel( a0, a2 ); \ 67 b2 = vec_mergeh( a1, a3 ); \ 68 b3 = vec_mergel( a1, a3 ) 71 #define vdst_load(d) \ 72 vdst_orig = vec_ld(0, dst); \ 73 vdst = vec_perm(vdst_orig, zero_u8v, vdst_mask); 75 #define vdst_load(d) vdst = vec_vsx_ld(0, dst) 78 #define VEC_LOAD_U8_ADD_S16_STORE_U8(va) \ 80 vdst_ss = (vec_s16) VEC_MERGEH(zero_u8v, vdst); \ 81 va = vec_add(va, vdst_ss); \ 82 va_u8 = vec_packsu(va, zero_s16v); \ 83 va_u32 = vec_splat((vec_u32)va_u8, 0); \ 84 vec_ste(va_u32, element, (uint32_t*)dst); 90 vec_s16 vtmp0, vtmp1, vtmp2, vtmp3;
94 const vec_u16 v6us = vec_splat_u16(6);
96 vec_u8 vdst_mask = vec_lvsl(0, dst);
97 int element = ((
unsigned long)dst & 0xf) >> 2;
102 vtmp0 = vec_ld(0,block);
103 vtmp1 = vec_sld(vtmp0, vtmp0, 8);
104 vtmp2 = vec_ld(16,block);
105 vtmp3 = vec_sld(vtmp2, vtmp2, 8);
106 memset(block, 0, 16 *
sizeof(int16_t));
108 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
109 VEC_TRANSPOSE_4(va0,va1,va2,va3,vtmp0,vtmp1,vtmp2,vtmp3);
110 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
112 va0 = vec_sra(va0,v6us);
113 va1 = vec_sra(va1,v6us);
114 va2 = vec_sra(va2,v6us);
115 va3 = vec_sra(va3,v6us);
117 VEC_LOAD_U8_ADD_S16_STORE_U8(va0);
119 VEC_LOAD_U8_ADD_S16_STORE_U8(va1);
121 VEC_LOAD_U8_ADD_S16_STORE_U8(va2);
123 VEC_LOAD_U8_ADD_S16_STORE_U8(va3);
126 #define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, d4, d5, d6, d7) {\ 128 vec_s16 a0v = vec_add(s0, s4); \ 130 vec_s16 a2v = vec_sub(s0, s4); \ 132 vec_s16 a4v = vec_sub(vec_sra(s2, onev), s6); \ 134 vec_s16 a6v = vec_add(vec_sra(s6, onev), s2); \ 136 vec_s16 b0v = vec_add(a0v, a6v); \ 138 vec_s16 b2v = vec_add(a2v, a4v); \ 140 vec_s16 b4v = vec_sub(a2v, a4v); \ 142 vec_s16 b6v = vec_sub(a0v, a6v); \ 145 vec_s16 a1v = vec_sub( vec_sub(s5, s3), vec_add(s7, vec_sra(s7, onev)) ); \ 148 vec_s16 a3v = vec_sub( vec_add(s7, s1), vec_add(s3, vec_sra(s3, onev)) );\ 151 vec_s16 a5v = vec_add( vec_sub(s7, s1), vec_add(s5, vec_sra(s5, onev)) );\ 153 vec_s16 a7v = vec_add( vec_add(s5, s3), vec_add(s1, vec_sra(s1, onev)) );\ 155 vec_s16 b1v = vec_add( vec_sra(a7v, twov), a1v); \ 157 vec_s16 b3v = vec_add(a3v, vec_sra(a5v, twov)); \ 159 vec_s16 b5v = vec_sub( vec_sra(a3v, twov), a5v); \ 161 vec_s16 b7v = vec_sub( a7v, vec_sra(a1v, twov)); \ 163 d0 = vec_add(b0v, b7v); \ 165 d1 = vec_add(b2v, b5v); \ 167 d2 = vec_add(b4v, b3v); \ 169 d3 = vec_add(b6v, b1v); \ 171 d4 = vec_sub(b6v, b1v); \ 173 d5 = vec_sub(b4v, b3v); \ 175 d6 = vec_sub(b2v, b5v); \ 177 d7 = vec_sub(b0v, b7v); \ 181 #define GET_2PERM(ldv, stv, d) \ 182 ldv = vec_lvsl(0, d); \ 183 stv = vec_lvsr(8, d); 184 #define dstv_load(d) \ 185 vec_u8 hv = vec_ld( 0, d ); \ 186 vec_u8 lv = vec_ld( 7, d); \ 187 vec_u8 dstv = vec_perm( hv, lv, (vec_u8)perm_ldv ); 188 #define dest_unligned_store(d) \ 190 vec_u8 bodyv = vec_perm( idstsum8, idstsum8, perm_stv ); \ 191 vec_u8 edgelv = vec_perm( sel, zero_u8v, perm_stv ); \ 192 lv = vec_sel( lv, bodyv, edgelv ); \ 193 vec_st( lv, 7, d ); \ 194 hv = vec_ld( 0, d ); \ 195 edgehv = vec_perm( zero_u8v, sel, perm_stv ); \ 196 hv = vec_sel( hv, bodyv, edgehv ); \ 200 #define GET_2PERM(ldv, stv, d) {} 201 #define dstv_load(d) vec_u8 dstv = vec_vsx_ld(0, d) 202 #define dest_unligned_store(d)\ 203 vec_u8 dst8 = vec_perm((vec_u8)idstsum8, dstv, vcprm(2,3,s2,s3));\ 204 vec_vsx_st(dst8, 0, d) 207 #define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel) { \ 210 vec_s16 idct_sh6 = vec_sra(idctv, sixv); \ 211 vec_u16 dst16 = (vec_u16)VEC_MERGEH(zero_u8v, dstv); \ 212 vec_s16 idstsum = vec_adds(idct_sh6, (vec_s16)dst16); \ 213 vec_u8 idstsum8 = vec_packsu(zero_s16v, idstsum); \ 215 dest_unligned_store(dest);\ 218 static void h264_idct8_add_altivec(
uint8_t *dst, int16_t *dct,
int stride)
221 vec_s16 d0, d1, d2, d3, d4, d5, d6, d7;
222 vec_s16 idct0, idct1, idct2, idct3, idct4, idct5,
idct6, idct7;
224 vec_u8 perm_ldv, perm_stv;
225 GET_2PERM(perm_ldv, perm_stv, dst);
227 const vec_u16 onev = vec_splat_u16(1);
228 const vec_u16 twov = vec_splat_u16(2);
229 const vec_u16 sixv = vec_splat_u16(6);
231 const vec_u8 sel = (
vec_u8) {0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1};
236 s0 = vec_ld(0x00, (int16_t*)dct);
237 s1 = vec_ld(0x10, (int16_t*)dct);
238 s2 = vec_ld(0x20, (int16_t*)dct);
239 s3 = vec_ld(0x30, (int16_t*)dct);
240 s4 = vec_ld(0x40, (int16_t*)dct);
241 s5 = vec_ld(0x50, (int16_t*)dct);
242 s6 = vec_ld(0x60, (int16_t*)dct);
243 s7 = vec_ld(0x70, (int16_t*)dct);
244 memset(dct, 0, 64 *
sizeof(int16_t));
246 IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7,
247 d0, d1, d2, d3, d4, d5, d6, d7);
249 TRANSPOSE8( d0, d1, d2, d3, d4, d5, d6, d7 );
251 IDCT8_1D_ALTIVEC(d0, d1, d2, d3, d4, d5, d6, d7,
252 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7);
254 ALTIVEC_STORE_SUM_CLIP(&dst[0*stride], idct0, perm_ldv, perm_stv, sel);
255 ALTIVEC_STORE_SUM_CLIP(&dst[1*stride], idct1, perm_ldv, perm_stv, sel);
256 ALTIVEC_STORE_SUM_CLIP(&dst[2*stride], idct2, perm_ldv, perm_stv, sel);
257 ALTIVEC_STORE_SUM_CLIP(&dst[3*stride], idct3, perm_ldv, perm_stv, sel);
258 ALTIVEC_STORE_SUM_CLIP(&dst[4*stride], idct4, perm_ldv, perm_stv, sel);
259 ALTIVEC_STORE_SUM_CLIP(&dst[5*stride], idct5, perm_ldv, perm_stv, sel);
260 ALTIVEC_STORE_SUM_CLIP(&dst[6*stride], idct6, perm_ldv, perm_stv, sel);
261 ALTIVEC_STORE_SUM_CLIP(&dst[7*stride], idct7, perm_ldv, perm_stv, sel);
265 #define DST_LD vec_ld 267 #define DST_LD vec_vsx_ld 272 vec_u8 dcplus, dcminus,
v0, v1, v2, v3, aligner;
278 dc = (block[0] + 32) >> 6;
280 v_dc32 = vec_lde(0, &
dc);
281 dc16 = VEC_SPLAT16((
vec_s16)v_dc32, 1);
289 aligner = vec_lvsr(0, dst);
290 dcplus = vec_perm(dcplus, dcplus, aligner);
291 dcminus = vec_perm(dcminus, dcminus, aligner);
294 for (i = 0; i <
size; i += 4) {
295 v0 = DST_LD(0, dst+0*stride);
296 v1 = DST_LD(0, dst+1*stride);
297 v2 = DST_LD(0, dst+2*stride);
298 v3 = DST_LD(0, dst+3*stride);
300 v0 = vec_adds(v0, dcplus);
301 v1 = vec_adds(v1, dcplus);
302 v2 = vec_adds(v2, dcplus);
303 v3 = vec_adds(v3, dcplus);
305 v0 = vec_subs(v0, dcminus);
306 v1 = vec_subs(v1, dcminus);
307 v2 = vec_subs(v2, dcminus);
308 v3 = vec_subs(v3, dcminus);
310 VEC_ST(v0, 0, dst+0*stride);
311 VEC_ST(v1, 0, dst+1*stride);
312 VEC_ST(v2, 0, dst+2*stride);
313 VEC_ST(v3, 0, dst+3*stride);
319 static void h264_idct_dc_add_altivec(
uint8_t *dst, int16_t *block,
int stride)
321 h264_idct_dc_add_internal(dst, block, stride, 4);
324 static void h264_idct8_dc_add_altivec(
uint8_t *dst, int16_t *block,
int stride)
326 h264_idct_dc_add_internal(dst, block, stride, 8);
329 static void h264_idct_add16_altivec(
uint8_t *dst,
const int *block_offset,
330 int16_t *block,
int stride,
335 int nnz = nnzc[
scan8[i] ];
337 if(nnz==1 && block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
338 else h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
343 static void h264_idct_add16intra_altivec(
uint8_t *dst,
const int *block_offset,
344 int16_t *block,
int stride,
349 if(nnzc[ scan8[i] ]) h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride);
350 else if(block[i*16]) h264_idct_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
354 static void h264_idct8_add4_altivec(
uint8_t *dst,
const int *block_offset,
355 int16_t *block,
int stride,
359 for(i=0; i<16; i+=4){
360 int nnz = nnzc[ scan8[i] ];
362 if(nnz==1 && block[i*16]) h264_idct8_dc_add_altivec(dst + block_offset[i], block + i*16, stride);
363 else h264_idct8_add_altivec(dst + block_offset[i], block + i*16, stride);
368 static void h264_idct_add8_altivec(
uint8_t **dest,
const int *block_offset,
369 int16_t *block,
int stride,
373 for (j = 1; j < 3; j++) {
374 for(i = j * 16; i < j * 16 + 4; i++){
376 h264_idct_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
378 h264_idct_dc_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride);
383 #define transpose4x16(r0, r1, r2, r3) { \ 384 register vec_u8 r4; \ 385 register vec_u8 r5; \ 386 register vec_u8 r6; \ 387 register vec_u8 r7; \ 389 r4 = vec_mergeh(r0, r2); \ 390 r5 = vec_mergel(r0, r2); \ 391 r6 = vec_mergeh(r1, r3); \ 392 r7 = vec_mergel(r1, r3); \ 394 r0 = vec_mergeh(r4, r6); \ 395 r1 = vec_mergel(r4, r6); \ 396 r2 = vec_mergeh(r5, r7); \ 397 r3 = vec_mergel(r5, r7); \ 400 static inline void write16x4(
uint8_t *dst,
int dst_stride,
404 uint32_t *src_int = (uint32_t *)result, *dst_int = (uint32_t *)dst;
405 int int_dst_stride = dst_stride/4;
407 vec_st(r0, 0, result);
408 vec_st(r1, 16, result);
409 vec_st(r2, 32, result);
410 vec_st(r3, 48, result);
413 *(dst_int+ int_dst_stride) = *(src_int + 1);
414 *(dst_int+ 2*int_dst_stride) = *(src_int + 2);
415 *(dst_int+ 3*int_dst_stride) = *(src_int + 3);
416 *(dst_int+ 4*int_dst_stride) = *(src_int + 4);
417 *(dst_int+ 5*int_dst_stride) = *(src_int + 5);
418 *(dst_int+ 6*int_dst_stride) = *(src_int + 6);
419 *(dst_int+ 7*int_dst_stride) = *(src_int + 7);
420 *(dst_int+ 8*int_dst_stride) = *(src_int + 8);
421 *(dst_int+ 9*int_dst_stride) = *(src_int + 9);
422 *(dst_int+10*int_dst_stride) = *(src_int + 10);
423 *(dst_int+11*int_dst_stride) = *(src_int + 11);
424 *(dst_int+12*int_dst_stride) = *(src_int + 12);
425 *(dst_int+13*int_dst_stride) = *(src_int + 13);
426 *(dst_int+14*int_dst_stride) = *(src_int + 14);
427 *(dst_int+15*int_dst_stride) = *(src_int + 15);
433 #define readAndTranspose16x6(src, src_stride, r8, r9, r10, r11, r12, r13) {\ 434 register vec_u8 r0 = unaligned_load(0, src); \ 435 register vec_u8 r1 = unaligned_load( src_stride, src); \ 436 register vec_u8 r2 = unaligned_load(2* src_stride, src); \ 437 register vec_u8 r3 = unaligned_load(3* src_stride, src); \ 438 register vec_u8 r4 = unaligned_load(4* src_stride, src); \ 439 register vec_u8 r5 = unaligned_load(5* src_stride, src); \ 440 register vec_u8 r6 = unaligned_load(6* src_stride, src); \ 441 register vec_u8 r7 = unaligned_load(7* src_stride, src); \ 442 register vec_u8 r14 = unaligned_load(14*src_stride, src); \ 443 register vec_u8 r15 = unaligned_load(15*src_stride, src); \ 445 r8 = unaligned_load( 8*src_stride, src); \ 446 r9 = unaligned_load( 9*src_stride, src); \ 447 r10 = unaligned_load(10*src_stride, src); \ 448 r11 = unaligned_load(11*src_stride, src); \ 449 r12 = unaligned_load(12*src_stride, src); \ 450 r13 = unaligned_load(13*src_stride, src); \ 453 r0 = vec_mergeh(r0, r8); \ 454 r1 = vec_mergeh(r1, r9); \ 455 r2 = vec_mergeh(r2, r10); \ 456 r3 = vec_mergeh(r3, r11); \ 457 r4 = vec_mergeh(r4, r12); \ 458 r5 = vec_mergeh(r5, r13); \ 459 r6 = vec_mergeh(r6, r14); \ 460 r7 = vec_mergeh(r7, r15); \ 463 r8 = vec_mergeh(r0, r4); \ 464 r9 = vec_mergel(r0, r4); \ 465 r10 = vec_mergeh(r1, r5); \ 466 r11 = vec_mergel(r1, r5); \ 467 r12 = vec_mergeh(r2, r6); \ 468 r13 = vec_mergel(r2, r6); \ 469 r14 = vec_mergeh(r3, r7); \ 470 r15 = vec_mergel(r3, r7); \ 473 r0 = vec_mergeh(r8, r12); \ 474 r1 = vec_mergel(r8, r12); \ 475 r2 = vec_mergeh(r9, r13); \ 476 r4 = vec_mergeh(r10, r14); \ 477 r5 = vec_mergel(r10, r14); \ 478 r6 = vec_mergeh(r11, r15); \ 482 r8 = vec_mergeh(r0, r4); \ 483 r9 = vec_mergel(r0, r4); \ 484 r10 = vec_mergeh(r1, r5); \ 485 r11 = vec_mergel(r1, r5); \ 486 r12 = vec_mergeh(r2, r6); \ 487 r13 = vec_mergel(r2, r6); \ 493 static inline vec_u8 diff_lt_altivec (
register vec_u8 x,
498 register vec_u8 diffneg = vec_subs(y, x);
499 register vec_u8 o = vec_or(diff, diffneg);
500 o = (
vec_u8)vec_cmplt(o, a);
504 static inline vec_u8 h264_deblock_mask (
register vec_u8 p0,
514 mask = diff_lt_altivec(p0, q0, alpha);
515 tempmask = diff_lt_altivec(p1, p0, beta);
516 mask = vec_and(mask, tempmask);
517 tempmask = diff_lt_altivec(q1, q0, beta);
518 mask = vec_and(mask, tempmask);
524 static inline vec_u8 h264_deblock_q1(
register vec_u8 p0,
530 register vec_u8 average = vec_avg(p0, q0);
532 register vec_u8 unclipped;
538 temp = vec_xor(average, p2);
539 average = vec_avg(average, p2);
540 ones = vec_splat_u8(1);
541 temp = vec_and(temp, ones);
542 unclipped = vec_subs(average, temp);
543 max = vec_adds(p1, tc0);
544 min = vec_subs(p1, tc0);
545 newp1 = vec_max(min, unclipped);
546 newp1 = vec_min(max, newp1);
550 #define h264_deblock_p0_q0(p0, p1, q0, q1, tc0masked) { \ 552 const vec_u8 A0v = vec_sl(vec_splat_u8(10), vec_splat_u8(4)); \ 554 register vec_u8 pq0bit = vec_xor(p0,q0); \ 555 register vec_u8 q1minus; \ 556 register vec_u8 p0minus; \ 557 register vec_u8 stage1; \ 558 register vec_u8 stage2; \ 559 register vec_u8 vec160; \ 560 register vec_u8 delta; \ 561 register vec_u8 deltaneg; \ 563 q1minus = vec_nor(q1, q1); \ 564 stage1 = vec_avg(p1, q1minus); \ 565 stage2 = vec_sr(stage1, vec_splat_u8(1)); \ 566 p0minus = vec_nor(p0, p0); \ 567 stage1 = vec_avg(q0, p0minus); \ 568 pq0bit = vec_and(pq0bit, vec_splat_u8(1)); \ 569 stage2 = vec_avg(stage2, pq0bit); \ 570 stage2 = vec_adds(stage2, stage1); \ 571 vec160 = vec_ld(0, &A0v); \ 572 deltaneg = vec_subs(vec160, stage2); \ 573 delta = vec_subs(stage2, vec160); \ 574 deltaneg = vec_min(tc0masked, deltaneg); \ 575 delta = vec_min(tc0masked, delta); \ 576 p0 = vec_subs(p0, deltaneg); \ 577 q0 = vec_subs(q0, delta); \ 578 p0 = vec_adds(p0, delta); \ 579 q0 = vec_adds(q0, deltaneg); \ 582 #define h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0) { \ 583 DECLARE_ALIGNED(16, unsigned char, temp)[16]; \ 584 register vec_u8 alphavec; \ 585 register vec_u8 betavec; \ 586 register vec_u8 mask; \ 587 register vec_u8 p1mask; \ 588 register vec_u8 q1mask; \ 589 register vector signed char tc0vec; \ 590 register vec_u8 finaltc0; \ 591 register vec_u8 tc0masked; \ 592 register vec_u8 newp1; \ 593 register vec_u8 newq1; \ 597 alphavec = vec_ld(0, temp); \ 598 betavec = vec_splat(alphavec, 0x1); \ 599 alphavec = vec_splat(alphavec, 0x0); \ 600 mask = h264_deblock_mask(p0, p1, q0, q1, alphavec, betavec); \ 602 AV_COPY32(temp, tc0); \ 603 tc0vec = vec_ld(0, (signed char*)temp); \ 604 tc0vec = vec_mergeh(tc0vec, tc0vec); \ 605 tc0vec = vec_mergeh(tc0vec, tc0vec); \ 606 mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_s8(-1))); \ 607 finaltc0 = vec_and((vec_u8)tc0vec, mask); \ 609 p1mask = diff_lt_altivec(p2, p0, betavec); \ 610 p1mask = vec_and(p1mask, mask); \ 611 tc0masked = vec_and(p1mask, (vec_u8)tc0vec); \ 612 finaltc0 = vec_sub(finaltc0, p1mask); \ 613 newp1 = h264_deblock_q1(p0, p1, p2, q0, tc0masked); \ 616 q1mask = diff_lt_altivec(q2, q0, betavec); \ 617 q1mask = vec_and(q1mask, mask); \ 618 tc0masked = vec_and(q1mask, (vec_u8)tc0vec); \ 619 finaltc0 = vec_sub(finaltc0, q1mask); \ 620 newq1 = h264_deblock_q1(p0, q1, q2, q0, tc0masked); \ 623 h264_deblock_p0_q0(p0, p1, q0, q1, finaltc0); \ 628 static void h264_v_loop_filter_luma_altivec(
uint8_t *pix,
int stride,
int alpha,
int beta, int8_t *tc0) {
630 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) >= 0) {
631 register vec_u8 p2 = vec_ld(-3*stride, pix);
632 register vec_u8 p1 = vec_ld(-2*stride, pix);
633 register vec_u8 p0 = vec_ld(-1*stride, pix);
634 register vec_u8 q0 = vec_ld(0, pix);
635 register vec_u8 q1 = vec_ld(stride, pix);
636 register vec_u8 q2 = vec_ld(2*stride, pix);
637 h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0);
638 vec_st(p1, -2*stride, pix);
639 vec_st(p0, -1*stride, pix);
641 vec_st(q1, stride, pix);
645 static void h264_h_loop_filter_luma_altivec(
uint8_t *pix,
int stride,
int alpha,
int beta, int8_t *tc0) {
647 register vec_u8 line0, line1, line2, line3, line4, line5;
648 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) < 0)
650 readAndTranspose16x6(pix-3, stride, line0, line1, line2, line3, line4, line5);
651 h264_loop_filter_luma_altivec(line0, line1, line2, line3, line4, line5, alpha, beta, tc0);
652 transpose4x16(line1, line2, line3, line4);
653 write16x4(pix-2, stride, line1, line2, line3, line4);
657 void weight_h264_W_altivec(
uint8_t *block,
int stride,
int height,
667 offset <<= log2_denom;
668 if(log2_denom) offset += 1<<(log2_denom-1);
669 temp[0] = log2_denom;
675 vtemp =(
vec_s16)vec_perm(vtemp, vtemp, vcswapi2s(0,1,2,3));
677 vlog2_denom = (
vec_u16)vec_splat(vtemp, 1);
678 vweight = vec_splat(vtemp, 3);
679 voffset = vec_splat(vtemp, 5);
680 aligned = !((
unsigned long)block & 0xf);
682 for (y = 0; y <
height; y++) {
683 vblock = vec_ld(0, block);
688 if (w == 16 || aligned) {
690 v0 = vec_adds(v0, voffset);
691 v0 = vec_sra(v0, vlog2_denom);
693 if (w == 16 || !aligned) {
695 v1 = vec_adds(v1, voffset);
696 v1 = vec_sra(v1, vlog2_denom);
698 vblock = vec_packsu(v0, v1);
699 vec_st(vblock, 0, block);
707 int log2_denom,
int weightd,
int weights,
int offset,
int w)
709 int y, dst_aligned, src_aligned;
711 vec_s16 vtemp, vweights, vweightd, voffset,
v0, v1, v2, v3;
716 offset = ((offset + 1) | 1) << log2_denom;
717 temp[0] = log2_denom+1;
724 vtemp =(
vec_s16)vec_perm(vtemp, vtemp, vcswapi2s(0,1,2,3));
726 vlog2_denom = (
vec_u16)vec_splat(vtemp, 1);
727 vweights = vec_splat(vtemp, 3);
728 vweightd = vec_splat(vtemp, 5);
729 voffset = vec_splat(vtemp, 7);
730 dst_aligned = !((
unsigned long)dst & 0xf);
731 src_aligned = !((
unsigned long)src & 0xf);
733 for (y = 0; y <
height; y++) {
734 vdst = vec_ld(0, dst);
735 vsrc = vec_ld(0, src);
749 if (w == 16 || dst_aligned) {
753 v0 = vec_adds(v0, voffset);
754 v0 = vec_adds(v0, v2);
755 v0 = vec_sra(v0, vlog2_denom);
757 if (w == 16 || !dst_aligned) {
761 v1 = vec_adds(v1, voffset);
762 v1 = vec_adds(v1, v3);
763 v1 = vec_sra(v1, vlog2_denom);
765 vdst = vec_packsu(v0, v1);
766 vec_st(vdst, 0, dst);
773 #define H264_WEIGHT(W) \ 774 static void weight_h264_pixels ## W ## _altivec(uint8_t *block, ptrdiff_t stride, int height, \ 775 int log2_denom, int weight, int offset) \ 777 weight_h264_W_altivec(block, stride, height, log2_denom, weight, offset, W); \ 779 static void biweight_h264_pixels ## W ## _altivec(uint8_t *dst, uint8_t *src, ptrdiff_t stride, int height, \ 780 int log2_denom, int weightd, int weights, int offset) \ 782 biweight_h264_W_altivec(dst, src, stride, height, log2_denom, weightd, weights, offset, W); \ 790 const int chroma_format_idc)
796 if (bit_depth == 8) {
798 if (chroma_format_idc <= 1)
static float alpha(float a)
Memory handling functions.
static const uint8_t q1[256]
Macro definitions for various function/variable attributes.
h264_weight_func weight_h264_pixels_tab[4]
void(* h264_idct_add16intra)(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 *8])
#define DECLARE_ALIGNED(n, t, v)
Declare a variable that is aligned in memory.
void(* h264_idct_add)(uint8_t *dst, int16_t *block, int stride)
void(* h264_idct8_dc_add)(uint8_t *dst, int16_t *block, int stride)
static int aligned(int val)
static const uint16_t mask[17]
#define PPC_ALTIVEC(flags)
static const uint8_t offset[127][2]
static const uint8_t q0[256]
static void idct6(int pre_mant[6])
Calculate 6-point IDCT of the pre-mantissas.
void(* h264_idct_add16)(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 *8])
h264_biweight_func biweight_h264_pixels_tab[4]
Context for storing H.264 DSP functions.
static void bit_depth(AudioStatsContext *s, uint64_t mask, uint64_t imask, AVRational *depth)
H.264 / AVC / MPEG-4 part10 codec.
void(* h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
av_cold void ff_h264dsp_init_ppc(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
void(* h264_idct8_add)(uint8_t *dst, int16_t *block, int stride)
static const uint8_t scan8[16 *3+3]
void(* h264_idct8_add4)(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 *8])
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
static int weight(int i, int blen, int offset)
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi - 0x80) *(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi - 0x80) *(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(const int16_t *) pi >> 8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t, *(const int16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t, *(const int16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(const int32_t *) pi >> 24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t, *(const int32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t, *(const int32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(const float *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(const float *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(const float *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(const double *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(const double *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(const double *) pi *(1U<< 31)))) #define SET_CONV_FUNC_GROUP(ofmt, ifmt) static void set_generic_function(AudioConvert *ac) { } void ff_audio_convert_free(AudioConvert **ac) { if(! *ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);} AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enum AVSampleFormat out_fmt, enum AVSampleFormat in_fmt, int channels, int sample_rate, int apply_map) { AudioConvert *ac;int in_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) return NULL;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);return NULL;} return ac;} 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;} else if(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;else ac->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);return ac;} int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in) { int use_generic=1;int len=in->nb_samples;int p;if(ac->dc) { av_log(ac->avr, AV_LOG_TRACE, "%d samples - audio_convert: %s to %s (dithered)\", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));return ff_convert_dither(ac-> dc
Contains misc utility macros and inline functions.
void(* h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
static av_always_inline int diff(const uint32_t a, const uint32_t b)
void(* h264_idct_dc_add)(uint8_t *dst, int16_t *block, int stride)
void(* h264_idct_add8)(uint8_t **dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 *8])