30 #define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
31 #define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
33 #define OP_U8_ALTIVEC PUT_OP_U8_ALTIVEC
34 #define PREFIX_h264_chroma_mc8_altivec put_h264_chroma_mc8_altivec
35 #define PREFIX_h264_chroma_mc8_num altivec_put_h264_chroma_mc8_num
36 #define PREFIX_h264_qpel16_h_lowpass_altivec put_h264_qpel16_h_lowpass_altivec
37 #define PREFIX_h264_qpel16_h_lowpass_num altivec_put_h264_qpel16_h_lowpass_num
38 #define PREFIX_h264_qpel16_v_lowpass_altivec put_h264_qpel16_v_lowpass_altivec
39 #define PREFIX_h264_qpel16_v_lowpass_num altivec_put_h264_qpel16_v_lowpass_num
40 #define PREFIX_h264_qpel16_hv_lowpass_altivec put_h264_qpel16_hv_lowpass_altivec
41 #define PREFIX_h264_qpel16_hv_lowpass_num altivec_put_h264_qpel16_hv_lowpass_num
44 #undef PREFIX_h264_chroma_mc8_altivec
45 #undef PREFIX_h264_chroma_mc8_num
46 #undef PREFIX_h264_qpel16_h_lowpass_altivec
47 #undef PREFIX_h264_qpel16_h_lowpass_num
48 #undef PREFIX_h264_qpel16_v_lowpass_altivec
49 #undef PREFIX_h264_qpel16_v_lowpass_num
50 #undef PREFIX_h264_qpel16_hv_lowpass_altivec
51 #undef PREFIX_h264_qpel16_hv_lowpass_num
53 #define OP_U8_ALTIVEC AVG_OP_U8_ALTIVEC
54 #define PREFIX_h264_chroma_mc8_altivec avg_h264_chroma_mc8_altivec
55 #define PREFIX_h264_chroma_mc8_num altivec_avg_h264_chroma_mc8_num
56 #define PREFIX_h264_qpel16_h_lowpass_altivec avg_h264_qpel16_h_lowpass_altivec
57 #define PREFIX_h264_qpel16_h_lowpass_num altivec_avg_h264_qpel16_h_lowpass_num
58 #define PREFIX_h264_qpel16_v_lowpass_altivec avg_h264_qpel16_v_lowpass_altivec
59 #define PREFIX_h264_qpel16_v_lowpass_num altivec_avg_h264_qpel16_v_lowpass_num
60 #define PREFIX_h264_qpel16_hv_lowpass_altivec avg_h264_qpel16_hv_lowpass_altivec
61 #define PREFIX_h264_qpel16_hv_lowpass_num altivec_avg_h264_qpel16_hv_lowpass_num
64 #undef PREFIX_h264_chroma_mc8_altivec
65 #undef PREFIX_h264_chroma_mc8_num
66 #undef PREFIX_h264_qpel16_h_lowpass_altivec
67 #undef PREFIX_h264_qpel16_h_lowpass_num
68 #undef PREFIX_h264_qpel16_v_lowpass_altivec
69 #undef PREFIX_h264_qpel16_v_lowpass_num
70 #undef PREFIX_h264_qpel16_hv_lowpass_altivec
71 #undef PREFIX_h264_qpel16_hv_lowpass_num
73 #define H264_MC(OPNAME, SIZE, CODETYPE) \
74 static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## CODETYPE (uint8_t *dst, uint8_t *src, int stride){\
75 OPNAME ## pixels ## SIZE ## _ ## CODETYPE(dst, src, stride, SIZE);\
78 static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){ \
79 DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
80 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
81 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
84 static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
85 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(dst, src, stride, stride);\
88 static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
89 DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
90 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
91 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+1, half, stride, stride, SIZE);\
94 static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
95 DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
96 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
97 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
100 static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
101 OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(dst, src, stride, stride);\
104 static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
105 DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
106 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
107 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+stride, half, stride, stride, SIZE);\
110 static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
111 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
112 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
113 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
114 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
115 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
118 static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
119 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
120 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
121 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
122 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
123 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
126 static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
127 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
128 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
129 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
130 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
131 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
134 static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
135 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
136 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
137 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
138 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
139 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
142 static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
143 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
144 OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(dst, tmp, src, stride, SIZE, stride);\
147 static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
148 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
149 DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
150 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
151 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
152 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
153 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
156 static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
157 DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
158 DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
159 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
160 put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
161 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
162 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
165 static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
166 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
167 DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
168 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
169 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
170 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
171 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
174 static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## CODETYPE(uint8_t *dst, uint8_t *src, int stride){\
175 DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
176 DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
177 DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
178 put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
179 put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
180 OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
184 const uint8_t * src2,
int dst_stride,
185 int src_stride1,
int h)
188 vec_u8 a,
b, d, tmp1, tmp2,
mask, mask_, edges, align;
190 mask_ = vec_lvsl(0, src2);
192 for (i = 0; i < h; i++) {
194 tmp1 = vec_ld(i * src_stride1, src1);
195 mask = vec_lvsl(i * src_stride1, src1);
196 tmp2 = vec_ld(i * src_stride1 + 15, src1);
198 a = vec_perm(tmp1, tmp2, mask);
200 tmp1 = vec_ld(i * 16, src2);
201 tmp2 = vec_ld(i * 16 + 15, src2);
203 b = vec_perm(tmp1, tmp2, mask_);
205 tmp1 = vec_ld(0, dst);
206 mask = vec_lvsl(0, dst);
207 tmp2 = vec_ld(15, dst);
211 edges = vec_perm(tmp2, tmp1, mask);
213 align = vec_lvsr(0, dst);
215 tmp2 = vec_perm(d, edges, align);
216 tmp1 = vec_perm(edges, d, align);
218 vec_st(tmp2, 15, dst);
219 vec_st(tmp1, 0 , dst);
226 const uint8_t * src2,
int dst_stride,
227 int src_stride1,
int h)
230 vec_u8 a,
b, d, tmp1, tmp2,
mask, mask_, edges, align;
232 mask_ = vec_lvsl(0, src2);
234 for (i = 0; i < h; i++) {
236 tmp1 = vec_ld(i * src_stride1, src1);
237 mask = vec_lvsl(i * src_stride1, src1);
238 tmp2 = vec_ld(i * src_stride1 + 15, src1);
240 a = vec_perm(tmp1, tmp2, mask);
242 tmp1 = vec_ld(i * 16, src2);
243 tmp2 = vec_ld(i * 16 + 15, src2);
245 b = vec_perm(tmp1, tmp2, mask_);
247 tmp1 = vec_ld(0, dst);
248 mask = vec_lvsl(0, dst);
249 tmp2 = vec_ld(15, dst);
251 d = vec_avg(vec_perm(tmp1, tmp2, mask), vec_avg(a, b));
253 edges = vec_perm(tmp2, tmp1, mask);
255 align = vec_lvsr(0, dst);
257 tmp2 = vec_perm(d, edges, align);
258 tmp1 = vec_perm(edges, d, align);
260 vec_st(tmp2, 15, dst);
261 vec_st(tmp1, 0 , dst);
280 #define VEC_1D_DCT(vb0,vb1,vb2,vb3,va0,va1,va2,va3) \
282 vz0 = vec_add(vb0,vb2); \
283 vz1 = vec_sub(vb0,vb2); \
284 vz2 = vec_sra(vb1,vec_splat_u16(1)); \
285 vz2 = vec_sub(vz2,vb3); \
286 vz3 = vec_sra(vb3,vec_splat_u16(1)); \
287 vz3 = vec_add(vb1,vz3); \
289 va0 = vec_add(vz0,vz3); \
290 va1 = vec_add(vz1,vz2); \
291 va2 = vec_sub(vz1,vz2); \
292 va3 = vec_sub(vz0,vz3)
294 #define VEC_TRANSPOSE_4(a0,a1,a2,a3,b0,b1,b2,b3) \
295 b0 = vec_mergeh( a0, a0 ); \
296 b1 = vec_mergeh( a1, a0 ); \
297 b2 = vec_mergeh( a2, a0 ); \
298 b3 = vec_mergeh( a3, a0 ); \
299 a0 = vec_mergeh( b0, b2 ); \
300 a1 = vec_mergel( b0, b2 ); \
301 a2 = vec_mergeh( b1, b3 ); \
302 a3 = vec_mergel( b1, b3 ); \
303 b0 = vec_mergeh( a0, a2 ); \
304 b1 = vec_mergel( a0, a2 ); \
305 b2 = vec_mergeh( a1, a3 ); \
306 b3 = vec_mergel( a1, a3 )
308 #define VEC_LOAD_U8_ADD_S16_STORE_U8(va) \
309 vdst_orig = vec_ld(0, dst); \
310 vdst = vec_perm(vdst_orig, zero_u8v, vdst_mask); \
311 vdst_ss = (vec_s16) vec_mergeh(zero_u8v, vdst); \
312 va = vec_add(va, vdst_ss); \
313 va_u8 = vec_packsu(va, zero_s16v); \
314 va_u32 = vec_splat((vec_u32)va_u8, 0); \
315 vec_ste(va_u32, element, (uint32_t*)dst);
321 vec_s16 vtmp0, vtmp1, vtmp2, vtmp3;
325 const vec_u16 v6us = vec_splat_u16(6);
327 vec_u8 vdst_mask = vec_lvsl(0, dst);
328 int element = ((
unsigned long)dst & 0xf) >> 2;
333 vtmp0 = vec_ld(0,block);
334 vtmp1 = vec_sld(vtmp0, vtmp0, 8);
335 vtmp2 = vec_ld(16,block);
336 vtmp3 = vec_sld(vtmp2, vtmp2, 8);
338 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
340 VEC_1D_DCT(vtmp0,vtmp1,vtmp2,vtmp3,va0,va1,va2,va3);
342 va0 = vec_sra(va0,v6us);
343 va1 = vec_sra(va1,v6us);
344 va2 = vec_sra(va2,v6us);
345 va3 = vec_sra(va3,v6us);
356 #define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, d4, d5, d6, d7) {\
358 vec_s16 a0v = vec_add(s0, s4); \
360 vec_s16 a2v = vec_sub(s0, s4); \
362 vec_s16 a4v = vec_sub(vec_sra(s2, onev), s6); \
364 vec_s16 a6v = vec_add(vec_sra(s6, onev), s2); \
366 vec_s16 b0v = vec_add(a0v, a6v); \
368 vec_s16 b2v = vec_add(a2v, a4v); \
370 vec_s16 b4v = vec_sub(a2v, a4v); \
372 vec_s16 b6v = vec_sub(a0v, a6v); \
375 vec_s16 a1v = vec_sub( vec_sub(s5, s3), vec_add(s7, vec_sra(s7, onev)) ); \
378 vec_s16 a3v = vec_sub( vec_add(s7, s1), vec_add(s3, vec_sra(s3, onev)) );\
381 vec_s16 a5v = vec_add( vec_sub(s7, s1), vec_add(s5, vec_sra(s5, onev)) );\
383 vec_s16 a7v = vec_add( vec_add(s5, s3), vec_add(s1, vec_sra(s1, onev)) );\
385 vec_s16 b1v = vec_add( vec_sra(a7v, twov), a1v); \
387 vec_s16 b3v = vec_add(a3v, vec_sra(a5v, twov)); \
389 vec_s16 b5v = vec_sub( vec_sra(a3v, twov), a5v); \
391 vec_s16 b7v = vec_sub( a7v, vec_sra(a1v, twov)); \
393 d0 = vec_add(b0v, b7v); \
395 d1 = vec_add(b2v, b5v); \
397 d2 = vec_add(b4v, b3v); \
399 d3 = vec_add(b6v, b1v); \
401 d4 = vec_sub(b6v, b1v); \
403 d5 = vec_sub(b4v, b3v); \
405 d6 = vec_sub(b2v, b5v); \
407 d7 = vec_sub(b0v, b7v); \
410 #define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel) { \
412 vec_u8 hv = vec_ld( 0, dest ); \
413 vec_u8 lv = vec_ld( 7, dest ); \
414 vec_u8 dstv = vec_perm( hv, lv, (vec_u8)perm_ldv ); \
415 vec_s16 idct_sh6 = vec_sra(idctv, sixv); \
416 vec_u16 dst16 = (vec_u16)vec_mergeh(zero_u8v, dstv); \
417 vec_s16 idstsum = vec_adds(idct_sh6, (vec_s16)dst16); \
418 vec_u8 idstsum8 = vec_packsu(zero_s16v, idstsum); \
421 vec_u8 bodyv = vec_perm( idstsum8, idstsum8, perm_stv );\
422 vec_u8 edgelv = vec_perm( sel, zero_u8v, perm_stv ); \
423 lv = vec_sel( lv, bodyv, edgelv ); \
424 vec_st( lv, 7, dest ); \
425 hv = vec_ld( 0, dest ); \
426 edgehv = vec_perm( zero_u8v, sel, perm_stv ); \
427 hv = vec_sel( hv, bodyv, edgehv ); \
428 vec_st( hv, 0, dest ); \
433 vec_s16 d0, d1, d2, d3, d4, d5, d6, d7;
434 vec_s16 idct0, idct1, idct2, idct3, idct4, idct5,
idct6, idct7;
436 vec_u8 perm_ldv = vec_lvsl(0, dst);
437 vec_u8 perm_stv = vec_lvsr(8, dst);
439 const vec_u16 onev = vec_splat_u16(1);
440 const vec_u16 twov = vec_splat_u16(2);
441 const vec_u16 sixv = vec_splat_u16(6);
443 const vec_u8 sel = (
vec_u8) {0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1};
448 s0 = vec_ld(0x00, (int16_t*)dct);
449 s1 = vec_ld(0x10, (int16_t*)dct);
450 s2 = vec_ld(0x20, (int16_t*)dct);
451 s3 = vec_ld(0x30, (int16_t*)dct);
452 s4 = vec_ld(0x40, (int16_t*)dct);
453 s5 = vec_ld(0x50, (int16_t*)dct);
454 s6 = vec_ld(0x60, (int16_t*)dct);
455 s7 = vec_ld(0x70, (int16_t*)dct);
458 d0, d1, d2, d3, d4, d5, d6, d7);
463 idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7);
478 vec_u8 dcplus, dcminus,
v0, v1, v2, v3, aligner;
483 dc = (block[0] + 32) >> 6;
484 dc16 = vec_splat((
vec_s16) vec_lde(0, &dc), 1);
491 aligner = vec_lvsr(0, dst);
492 dcplus = vec_perm(dcplus, dcplus, aligner);
493 dcminus = vec_perm(dcminus, dcminus, aligner);
495 for (i = 0; i <
size; i += 4) {
496 v0 = vec_ld(0, dst+0*stride);
497 v1 = vec_ld(0, dst+1*stride);
498 v2 = vec_ld(0, dst+2*stride);
499 v3 = vec_ld(0, dst+3*stride);
501 v0 = vec_adds(v0, dcplus);
502 v1 = vec_adds(v1, dcplus);
503 v2 = vec_adds(v2, dcplus);
504 v3 = vec_adds(v3, dcplus);
506 v0 = vec_subs(v0, dcminus);
507 v1 = vec_subs(v1, dcminus);
508 v2 = vec_subs(v2, dcminus);
509 v3 = vec_subs(v3, dcminus);
511 vec_st(v0, 0, dst+0*stride);
512 vec_st(v1, 0, dst+1*stride);
513 vec_st(v2, 0, dst+2*stride);
514 vec_st(v3, 0, dst+3*stride);
533 int nnz = nnzc[
scan8[i] ];
551 for(i=0; i<16; i+=4){
552 int nnz = nnzc[
scan8[i] ];
562 for (j = 1; j < 3; j++) {
563 for(i = j * 16; i < j * 16 + 4; i++){
572 #define transpose4x16(r0, r1, r2, r3) { \
573 register vec_u8 r4; \
574 register vec_u8 r5; \
575 register vec_u8 r6; \
576 register vec_u8 r7; \
578 r4 = vec_mergeh(r0, r2); \
579 r5 = vec_mergel(r0, r2); \
580 r6 = vec_mergeh(r1, r3); \
581 r7 = vec_mergel(r1, r3); \
583 r0 = vec_mergeh(r4, r6); \
584 r1 = vec_mergel(r4, r6); \
585 r2 = vec_mergeh(r5, r7); \
586 r3 = vec_mergel(r5, r7); \
589 static inline void write16x4(uint8_t *dst,
int dst_stride,
593 uint32_t *src_int = (uint32_t *)result, *dst_int = (uint32_t *)dst;
594 int int_dst_stride = dst_stride/4;
596 vec_st(r0, 0, result);
597 vec_st(r1, 16, result);
598 vec_st(r2, 32, result);
599 vec_st(r3, 48, result);
602 *(dst_int+ int_dst_stride) = *(src_int + 1);
603 *(dst_int+ 2*int_dst_stride) = *(src_int + 2);
604 *(dst_int+ 3*int_dst_stride) = *(src_int + 3);
605 *(dst_int+ 4*int_dst_stride) = *(src_int + 4);
606 *(dst_int+ 5*int_dst_stride) = *(src_int + 5);
607 *(dst_int+ 6*int_dst_stride) = *(src_int + 6);
608 *(dst_int+ 7*int_dst_stride) = *(src_int + 7);
609 *(dst_int+ 8*int_dst_stride) = *(src_int + 8);
610 *(dst_int+ 9*int_dst_stride) = *(src_int + 9);
611 *(dst_int+10*int_dst_stride) = *(src_int + 10);
612 *(dst_int+11*int_dst_stride) = *(src_int + 11);
613 *(dst_int+12*int_dst_stride) = *(src_int + 12);
614 *(dst_int+13*int_dst_stride) = *(src_int + 13);
615 *(dst_int+14*int_dst_stride) = *(src_int + 14);
616 *(dst_int+15*int_dst_stride) = *(src_int + 15);
622 #define readAndTranspose16x6(src, src_stride, r8, r9, r10, r11, r12, r13) {\
623 register vec_u8 r0 = unaligned_load(0, src); \
624 register vec_u8 r1 = unaligned_load( src_stride, src); \
625 register vec_u8 r2 = unaligned_load(2* src_stride, src); \
626 register vec_u8 r3 = unaligned_load(3* src_stride, src); \
627 register vec_u8 r4 = unaligned_load(4* src_stride, src); \
628 register vec_u8 r5 = unaligned_load(5* src_stride, src); \
629 register vec_u8 r6 = unaligned_load(6* src_stride, src); \
630 register vec_u8 r7 = unaligned_load(7* src_stride, src); \
631 register vec_u8 r14 = unaligned_load(14*src_stride, src); \
632 register vec_u8 r15 = unaligned_load(15*src_stride, src); \
634 r8 = unaligned_load( 8*src_stride, src); \
635 r9 = unaligned_load( 9*src_stride, src); \
636 r10 = unaligned_load(10*src_stride, src); \
637 r11 = unaligned_load(11*src_stride, src); \
638 r12 = unaligned_load(12*src_stride, src); \
639 r13 = unaligned_load(13*src_stride, src); \
642 r0 = vec_mergeh(r0, r8); \
643 r1 = vec_mergeh(r1, r9); \
644 r2 = vec_mergeh(r2, r10); \
645 r3 = vec_mergeh(r3, r11); \
646 r4 = vec_mergeh(r4, r12); \
647 r5 = vec_mergeh(r5, r13); \
648 r6 = vec_mergeh(r6, r14); \
649 r7 = vec_mergeh(r7, r15); \
652 r8 = vec_mergeh(r0, r4); \
653 r9 = vec_mergel(r0, r4); \
654 r10 = vec_mergeh(r1, r5); \
655 r11 = vec_mergel(r1, r5); \
656 r12 = vec_mergeh(r2, r6); \
657 r13 = vec_mergel(r2, r6); \
658 r14 = vec_mergeh(r3, r7); \
659 r15 = vec_mergel(r3, r7); \
662 r0 = vec_mergeh(r8, r12); \
663 r1 = vec_mergel(r8, r12); \
664 r2 = vec_mergeh(r9, r13); \
665 r4 = vec_mergeh(r10, r14); \
666 r5 = vec_mergel(r10, r14); \
667 r6 = vec_mergeh(r11, r15); \
671 r8 = vec_mergeh(r0, r4); \
672 r9 = vec_mergel(r0, r4); \
673 r10 = vec_mergeh(r1, r5); \
674 r11 = vec_mergel(r1, r5); \
675 r12 = vec_mergeh(r2, r6); \
676 r13 = vec_mergel(r2, r6); \
686 register vec_u8 diff = vec_subs(x, y);
687 register vec_u8 diffneg = vec_subs(y, x);
688 register vec_u8 o = vec_or(diff, diffneg);
689 o = (
vec_u8)vec_cmplt(o, a);
705 mask = vec_and(mask, tempmask);
707 mask = vec_and(mask, tempmask);
719 register vec_u8 average = vec_avg(p0, q0);
727 temp = vec_xor(average, p2);
728 average = vec_avg(average, p2);
729 ones = vec_splat_u8(1);
730 temp = vec_and(temp, ones);
731 uncliped = vec_subs(average, temp);
732 max = vec_adds(p1, tc0);
733 min = vec_subs(p1, tc0);
734 newp1 = vec_max(min, uncliped);
735 newp1 = vec_min(max, newp1);
739 #define h264_deblock_p0_q0(p0, p1, q0, q1, tc0masked) { \
741 const vec_u8 A0v = vec_sl(vec_splat_u8(10), vec_splat_u8(4)); \
743 register vec_u8 pq0bit = vec_xor(p0,q0); \
744 register vec_u8 q1minus; \
745 register vec_u8 p0minus; \
746 register vec_u8 stage1; \
747 register vec_u8 stage2; \
748 register vec_u8 vec160; \
749 register vec_u8 delta; \
750 register vec_u8 deltaneg; \
752 q1minus = vec_nor(q1, q1); \
753 stage1 = vec_avg(p1, q1minus); \
754 stage2 = vec_sr(stage1, vec_splat_u8(1)); \
755 p0minus = vec_nor(p0, p0); \
756 stage1 = vec_avg(q0, p0minus); \
757 pq0bit = vec_and(pq0bit, vec_splat_u8(1)); \
758 stage2 = vec_avg(stage2, pq0bit); \
759 stage2 = vec_adds(stage2, stage1); \
760 vec160 = vec_ld(0, &A0v); \
761 deltaneg = vec_subs(vec160, stage2); \
762 delta = vec_subs(stage2, vec160); \
763 deltaneg = vec_min(tc0masked, deltaneg); \
764 delta = vec_min(tc0masked, delta); \
765 p0 = vec_subs(p0, deltaneg); \
766 q0 = vec_subs(q0, delta); \
767 p0 = vec_adds(p0, delta); \
768 q0 = vec_adds(q0, deltaneg); \
771 #define h264_loop_filter_luma_altivec(p2, p1, p0, q0, q1, q2, alpha, beta, tc0) { \
772 DECLARE_ALIGNED(16, unsigned char, temp)[16]; \
773 register vec_u8 alphavec; \
774 register vec_u8 betavec; \
775 register vec_u8 mask; \
776 register vec_u8 p1mask; \
777 register vec_u8 q1mask; \
778 register vector signed char tc0vec; \
779 register vec_u8 finaltc0; \
780 register vec_u8 tc0masked; \
781 register vec_u8 newp1; \
782 register vec_u8 newq1; \
786 alphavec = vec_ld(0, temp); \
787 betavec = vec_splat(alphavec, 0x1); \
788 alphavec = vec_splat(alphavec, 0x0); \
789 mask = h264_deblock_mask(p0, p1, q0, q1, alphavec, betavec); \
791 *((int *)temp) = *((int *)tc0); \
792 tc0vec = vec_ld(0, (signed char*)temp); \
793 tc0vec = vec_mergeh(tc0vec, tc0vec); \
794 tc0vec = vec_mergeh(tc0vec, tc0vec); \
795 mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_s8(-1))); \
796 finaltc0 = vec_and((vec_u8)tc0vec, mask); \
798 p1mask = diff_lt_altivec(p2, p0, betavec); \
799 p1mask = vec_and(p1mask, mask); \
800 tc0masked = vec_and(p1mask, (vec_u8)tc0vec); \
801 finaltc0 = vec_sub(finaltc0, p1mask); \
802 newp1 = h264_deblock_q1(p0, p1, p2, q0, tc0masked); \
805 q1mask = diff_lt_altivec(q2, q0, betavec); \
806 q1mask = vec_and(q1mask, mask); \
807 tc0masked = vec_and(q1mask, (vec_u8)tc0vec); \
808 finaltc0 = vec_sub(finaltc0, q1mask); \
809 newq1 = h264_deblock_q1(p0, q1, q2, q0, tc0masked); \
812 h264_deblock_p0_q0(p0, p1, q0, q1, finaltc0); \
819 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) >= 0) {
820 register vec_u8 p2 = vec_ld(-3*stride, pix);
821 register vec_u8 p1 = vec_ld(-2*stride, pix);
822 register vec_u8 p0 = vec_ld(-1*stride, pix);
823 register vec_u8 q0 = vec_ld(0, pix);
824 register vec_u8 q1 = vec_ld(stride, pix);
825 register vec_u8 q2 = vec_ld(2*stride, pix);
827 vec_st(p1, -2*stride, pix);
828 vec_st(p0, -1*stride, pix);
830 vec_st(q1, stride, pix);
836 register vec_u8 line0, line1, line2, line3, line4, line5;
837 if ((tc0[0] & tc0[1] & tc0[2] & tc0[3]) < 0)
842 write16x4(pix-2, stride, line1, line2, line3, line4);
847 int log2_denom,
int weight,
int offset,
int w)
856 offset <<= log2_denom;
857 if(log2_denom) offset += 1<<(log2_denom-1);
858 temp[0] = log2_denom;
862 vtemp = (
vec_s16)vec_ld(0, temp);
863 vlog2_denom = (
vec_u16)vec_splat(vtemp, 1);
864 vweight = vec_splat(vtemp, 3);
865 voffset = vec_splat(vtemp, 5);
866 aligned = !((
unsigned long)block & 0xf);
868 for (y = 0; y <
height; y++) {
869 vblock = vec_ld(0, block);
874 if (w == 16 || aligned) {
876 v0 = vec_adds(v0, voffset);
877 v0 = vec_sra(v0, vlog2_denom);
879 if (w == 16 || !aligned) {
881 v1 = vec_adds(v1, voffset);
882 v1 = vec_sra(v1, vlog2_denom);
884 vblock = vec_packsu(v0, v1);
885 vec_st(vblock, 0, block);
893 int log2_denom,
int weightd,
int weights,
int offset,
int w)
895 int y, dst_aligned, src_aligned;
897 vec_s16 vtemp, vweights, vweightd, voffset,
v0, v1, v2, v3;
902 offset = ((offset + 1) | 1) << log2_denom;
903 temp[0] = log2_denom+1;
908 vtemp = (
vec_s16)vec_ld(0, temp);
909 vlog2_denom = (
vec_u16)vec_splat(vtemp, 1);
910 vweights = vec_splat(vtemp, 3);
911 vweightd = vec_splat(vtemp, 5);
912 voffset = vec_splat(vtemp, 7);
913 dst_aligned = !((
unsigned long)dst & 0xf);
914 src_aligned = !((
unsigned long)src & 0xf);
916 for (y = 0; y <
height; y++) {
917 vdst = vec_ld(0, dst);
918 vsrc = vec_ld(0, src);
932 if (w == 16 || dst_aligned) {
936 v0 = vec_adds(v0, voffset);
937 v0 = vec_adds(v0, v2);
938 v0 = vec_sra(v0, vlog2_denom);
940 if (w == 16 || !dst_aligned) {
944 v1 = vec_adds(v1, voffset);
945 v1 = vec_adds(v1, v3);
946 v1 = vec_sra(v1, vlog2_denom);
948 vdst = vec_packsu(v0, v1);
949 vec_st(vdst, 0, dst);
956 #define H264_WEIGHT(W) \
957 static void ff_weight_h264_pixels ## W ## _altivec(uint8_t *block, int stride, int height, \
958 int log2_denom, int weight, int offset){ \
959 weight_h264_W_altivec(block, stride, height, log2_denom, weight, offset, W); \
961 static void ff_biweight_h264_pixels ## W ## _altivec(uint8_t *dst, uint8_t *src, int stride, int height, \
962 int log2_denom, int weightd, int weights, int offset){ \
963 biweight_h264_W_altivec(dst, src, stride, height, log2_denom, weightd, weights, offset, W); \
970 const int high_bit_depth = avctx->bits_per_raw_sample > 8;
973 if (!high_bit_depth) {
974 c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec;
975 c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec;
977 #define dspfunc(PFX, IDX, NUM) \
978 c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \
979 c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_altivec; \
980 c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_altivec; \
981 c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_altivec; \
982 c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_altivec; \
983 c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_altivec; \
984 c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_altivec; \
985 c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_altivec; \
986 c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_altivec; \
987 c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_altivec; \
988 c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_altivec; \
989 c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_altivec; \
990 c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_altivec; \
991 c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_altivec; \
992 c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_altivec; \
993 c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_altivec
1005 if (bit_depth == 8) {
1007 if (chroma_format_idc <= 1)