Mercurial > libavcodec.hg
comparison ppc/float_altivec.c @ 7674:e1876d3200ee libavcodec
Introduce float_to_int16_one_altivec
| author | lu_zero |
|---|---|
| date | Sat, 23 Aug 2008 21:02:30 +0000 |
| parents | f7cbb7733146 |
| children | ad4bf45b9b63 |
comparison
equal
deleted
inserted
replaced
| 7673:adcd246127eb | 7674:e1876d3200ee |
|---|---|
| 147 } | 147 } |
| 148 else | 148 else |
| 149 ff_vector_fmul_add_add_c(dst, src0, src1, src2, src3, len, step); | 149 ff_vector_fmul_add_add_c(dst, src0, src1, src2, src3, len, step); |
| 150 } | 150 } |
| 151 | 151 |
| 152 | |
| 153 static vector signed short | |
| 154 float_to_int16_one_altivec(const float *src) | |
| 155 { | |
| 156 vector float s0 = vec_ld(0, src); | |
| 157 vector float s1 = vec_ld(16, src); | |
| 158 vector signed int t0 = vec_cts(s0, 0); | |
| 159 vector signed int t1 = vec_cts(s1, 0); | |
| 160 return vec_packs(t0,t1); | |
| 161 } | |
| 162 | |
| 152 void float_to_int16_altivec(int16_t *dst, const float *src, int len) | 163 void float_to_int16_altivec(int16_t *dst, const float *src, int len) |
| 153 { | 164 { |
| 154 int i; | 165 int i; |
| 155 vector float s0, s1; | 166 vector float s0, s1; |
| 156 vector signed int t0, t1; | 167 vector signed int t0, t1; |
| 157 vector signed short d0, d1, d; | 168 vector signed short d0, d1, d; |
| 158 vector unsigned char align; | 169 vector unsigned char align; |
| 159 if(((long)dst)&15) //FIXME | 170 if(((long)dst)&15) //FIXME |
| 160 for(i=0; i<len-7; i+=8) { | 171 for(i=0; i<len-7; i+=8) { |
| 161 s0 = vec_ld(0, src+i); | |
| 162 s1 = vec_ld(16, src+i); | |
| 163 t0 = vec_cts(s0, 0); | |
| 164 d0 = vec_ld(0, dst+i); | 172 d0 = vec_ld(0, dst+i); |
| 165 t1 = vec_cts(s1, 0); | 173 d = float_to_int16_one_altivec(src+i); |
| 166 d1 = vec_ld(15, dst+i); | 174 d1 = vec_ld(15, dst+i); |
| 167 d = vec_packs(t0,t1); | |
| 168 d1 = vec_perm(d1, d0, vec_lvsl(0,dst+i)); | 175 d1 = vec_perm(d1, d0, vec_lvsl(0,dst+i)); |
| 169 align = vec_lvsr(0, dst+i); | 176 align = vec_lvsr(0, dst+i); |
| 170 d0 = vec_perm(d1, d, align); | 177 d0 = vec_perm(d1, d, align); |
| 171 d1 = vec_perm(d, d1, align); | 178 d1 = vec_perm(d, d1, align); |
| 172 vec_st(d0, 0, dst+i); | 179 vec_st(d0, 0, dst+i); |
| 173 vec_st(d1,15, dst+i); | 180 vec_st(d1,15, dst+i); |
| 174 } | 181 } |
| 175 else | 182 else |
| 176 for(i=0; i<len-7; i+=8) { | 183 for(i=0; i<len-7; i+=8) { |
| 177 s0 = vec_ld(0, src+i); | 184 d = float_to_int16_one_altivec(src+i); |
| 178 s1 = vec_ld(16, src+i); | |
| 179 t0 = vec_cts(s0, 0); | |
| 180 t1 = vec_cts(s1, 0); | |
| 181 d = vec_packs(t0,t1); | |
| 182 vec_st(d, 0, dst+i); | 185 vec_st(d, 0, dst+i); |
| 183 } | 186 } |
| 184 } | 187 } |
| 185 | 188 |
| 186 void float_init_altivec(DSPContext* c, AVCodecContext *avctx) | 189 void float_init_altivec(DSPContext* c, AVCodecContext *avctx) |