2 * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org>
4 * This file is part of FFmpeg.
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.
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.
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
21 #include "../dsputil.h"
23 #include "gcc_fixes.h"
25 #include "dsputil_altivec.h"
27 static void vector_fmul_altivec(float *dst, const float *src, int len)
30 vector float d0, d1, s, zero = (vector float)vec_splat_u32(0);
31 for(i=0; i<len-7; i+=8) {
32 d0 = vec_ld(0, dst+i);
34 d1 = vec_ld(16, dst+i);
35 d0 = vec_madd(d0, s, zero);
36 d1 = vec_madd(d1, vec_ld(16,src+i), zero);
38 vec_st(d1, 16, dst+i);
42 static void vector_fmul_reverse_altivec(float *dst, const float *src0,
43 const float *src1, int len)
46 vector float d, s0, s1, h0, l0,
47 s2, s3, zero = (vector float)vec_splat_u32(0);
49 for(i=0; i<len-7; i+=8) {
50 s1 = vec_ld(0, src1-i); // [a,b,c,d]
51 s0 = vec_ld(0, src0+i);
52 l0 = vec_mergel(s1, s1); // [c,c,d,d]
53 s3 = vec_ld(-16, src1-i);
54 h0 = vec_mergeh(s1, s1); // [a,a,b,b]
55 s2 = vec_ld(16, src0+i);
56 s1 = vec_mergeh(vec_mergel(l0,h0), // [d,b,d,b]
57 vec_mergeh(l0,h0)); // [c,a,c,a]
59 l0 = vec_mergel(s3, s3);
60 d = vec_madd(s0, s1, zero);
61 h0 = vec_mergeh(s3, s3);
63 s3 = vec_mergeh(vec_mergel(l0,h0),
65 d = vec_madd(s2, s3, zero);
70 static void vector_fmul_add_add_altivec(float *dst, const float *src0,
71 const float *src1, const float *src2,
72 int src3, int len, int step)
75 vector float d, s0, s1, s2, t0, t1, edges;
76 vector unsigned char align = vec_lvsr(0,dst),
77 mask = vec_lvsl(0, dst);
80 #if 0 //FIXME: there is still something wrong
83 vector float d0, d1, s3, t2;
84 vector unsigned int sel =
85 vec_mergeh(vec_splat_u32(-1), vec_splat_u32(0));
87 for (i=0,y=0; i<len-3; i+=4,y+=8) {
89 s0 = vec_ld(0,src0+i);
90 s1 = vec_ld(0,src1+i);
91 s2 = vec_ld(0,src2+i);
93 // t0 = vec_ld(0, dst+y); //[x x x|a]
94 // t1 = vec_ld(16, dst+y); //[b c d|e]
95 t2 = vec_ld(31, dst+y); //[f g h|x]
97 d = vec_madd(s0,s1,s2); // [A B C D]
103 d0 = vec_perm(t0, t1, mask); // [a b c d]
105 d0 = vec_sel(vec_mergeh(d, d), d0, sel); // [A b B d]
107 edges = vec_perm(t1, t0, mask);
109 t0 = vec_perm(edges, d0, align); // [x x x|A]
111 t1 = vec_perm(d0, edges, align); // [b B d|e]
113 vec_stl(t0, 0, dst+y);
115 d1 = vec_perm(t1, t2, mask); // [e f g h]
117 d1 = vec_sel(vec_mergel(d, d), d1, sel); // [C f D h]
119 edges = vec_perm(t2, t1, mask);
121 t1 = vec_perm(edges, d1, align); // [b B d|C]
123 t2 = vec_perm(d1, edges, align); // [f D h|x]
125 vec_stl(t1, 16, dst+y);
129 vec_stl(t2, 31, dst+y);
135 if (step == 1 && src3 == 0)
136 for (i=0; i<len-3; i+=4) {
137 t1 = vec_ld(15, dst+i);
138 s0 = vec_ld(0, src0+i);
139 s1 = vec_ld(0, src1+i);
140 s2 = vec_ld(0, src2+i);
141 edges = vec_perm(t1 ,t0, mask);
142 d = vec_madd(s0,s1,s2);
143 t1 = vec_perm(d, edges, align);
144 t0 = vec_perm(edges, d, align);
145 vec_st(t1, 15, dst+i);
146 vec_st(t0, 0, dst+i);
150 ff_vector_fmul_add_add_c(dst, src0, src1, src2, src3, len, step);
153 void float_to_int16_altivec(int16_t *dst, const float *src, int len)
157 vector signed int t0, t1;
158 vector signed short d0, d1, d;
159 vector unsigned char align;
160 if(((long)dst)&15) //FIXME
161 for(i=0; i<len-7; i+=8) {
162 s0 = vec_ld(0, src+i);
163 s1 = vec_ld(16, src+i);
165 d0 = vec_ld(0, dst+i);
167 d1 = vec_ld(15, dst+i);
168 d = vec_packs(t0,t1);
169 d1 = vec_perm(d1, d0, vec_lvsl(0,dst+i));
170 align = vec_lvsr(0, dst+i);
171 d0 = vec_perm(d1, d, align);
172 d1 = vec_perm(d, d1, align);
173 vec_st(d0, 0, dst+i);
174 vec_st(d1,15, dst+i);
177 for(i=0; i<len-7; i+=8) {
178 s0 = vec_ld(0, src+i);
179 s1 = vec_ld(16, src+i);
182 d = vec_packs(t0,t1);
187 void float_init_altivec(DSPContext* c, AVCodecContext *avctx)
189 c->vector_fmul = vector_fmul_altivec;
190 c->vector_fmul_reverse = vector_fmul_reverse_altivec;
191 c->vector_fmul_add_add = vector_fmul_add_add_altivec;
192 if(!(avctx->flags & CODEC_FLAG_BITEXACT))
193 c->float_to_int16 = float_to_int16_altivec;