comparison ppc/float_altivec.c @ 8365:d33b47d1f4c1 libavcodec

add AltiVec implementation of int32_to_float_fmul_scalar
author gpoirier
date Wed, 17 Dec 2008 09:47:06 +0000
parents 8313d5901c5c
children 93a3020d9636
comparison
equal deleted inserted replaced
8364:4a01f7144da5 8365:d33b47d1f4c1
145 vec_st(t1, 15, dst+i); 145 vec_st(t1, 15, dst+i);
146 vec_st(t0, 0, dst+i); 146 vec_st(t0, 0, dst+i);
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 }
151
152
153 static void int32_to_float_fmul_scalar_altivec(float *dst, const int *src, float mul, int len)
154 {
155 union {
156 vector float v;
157 float s[4];
158 } mul_u;
159 int i;
160 vector float src1, src2, dst1, dst2, mul_v, zero;
161
162 zero = (vector float)vec_splat_u32(0);
163 mul_u.s[0] = mul;
164 mul_v = vec_splat(mul_u.v, 0);
165
166 for(i=0; i<len; i+=8) {
167 src1 = vec_ctf(vec_ld(0, src+i), 0);
168 src2 = vec_ctf(vec_ld(16, src+i), 0);
169 dst1 = vec_madd(src1, mul_v, zero);
170 dst2 = vec_madd(src2, mul_v, zero);
171 vec_st(dst1, 0, dst+i);
172 vec_st(dst2, 16, dst+i);
173 }
150 } 174 }
151 175
152 176
153 static vector signed short 177 static vector signed short
154 float_to_int16_one_altivec(const float *src) 178 float_to_int16_one_altivec(const float *src)
238 void float_init_altivec(DSPContext* c, AVCodecContext *avctx) 262 void float_init_altivec(DSPContext* c, AVCodecContext *avctx)
239 { 263 {
240 c->vector_fmul = vector_fmul_altivec; 264 c->vector_fmul = vector_fmul_altivec;
241 c->vector_fmul_reverse = vector_fmul_reverse_altivec; 265 c->vector_fmul_reverse = vector_fmul_reverse_altivec;
242 c->vector_fmul_add_add = vector_fmul_add_add_altivec; 266 c->vector_fmul_add_add = vector_fmul_add_add_altivec;
267 c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_altivec;
243 if(!(avctx->flags & CODEC_FLAG_BITEXACT)) { 268 if(!(avctx->flags & CODEC_FLAG_BITEXACT)) {
244 c->float_to_int16 = float_to_int16_altivec; 269 c->float_to_int16 = float_to_int16_altivec;
245 c->float_to_int16_interleave = float_to_int16_interleave_altivec; 270 c->float_to_int16_interleave = float_to_int16_interleave_altivec;
246 } 271 }
247 } 272 }