1/*
2 * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org>
3 *
4 * This file is part of FFmpeg.
5 *
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.
10 *
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.
15 *
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
19 */
20
21#include "util_altivec.h"
22#include "float_dsp_altivec.h"
23
24void ff_vector_fmul_altivec(float *dst, const float *src0, const float *src1,
25                            int len)
26{
27    int i;
28    vector float d0, d1, s, zero = (vector float)vec_splat_u32(0);
29    for (i = 0; i < len - 7; i += 8) {
30        d0 = vec_ld( 0, src0 + i);
31        s  = vec_ld( 0, src1 + i);
32        d1 = vec_ld(16, src0 + i);
33        d0 = vec_madd(d0, s, zero);
34        d1 = vec_madd(d1, vec_ld(16, src1 + i), zero);
35        vec_st(d0,  0, dst + i);
36        vec_st(d1, 16, dst + i);
37    }
38}
39
40void ff_vector_fmul_window_altivec(float *dst, const float *src0,
41                                   const float *src1, const float *win, int len)
42{
43    vector float zero, t0, t1, s0, s1, wi, wj;
44    const vector unsigned char reverse = vcprm(3, 2, 1, 0);
45    int i, j;
46
47    dst  += len;
48    win  += len;
49    src0 += len;
50
51    zero = (vector float)vec_splat_u32(0);
52
53    for (i = -len * 4, j = len * 4 - 16; i < 0; i += 16, j -= 16) {
54        s0 = vec_ld(i, src0);
55        s1 = vec_ld(j, src1);
56        wi = vec_ld(i, win);
57        wj = vec_ld(j, win);
58
59        s1 = vec_perm(s1, s1, reverse);
60        wj = vec_perm(wj, wj, reverse);
61
62        t0 = vec_madd(s0, wj, zero);
63        t0 = vec_nmsub(s1, wi, t0);
64        t1 = vec_madd(s0, wi, zero);
65        t1 = vec_madd(s1, wj, t1);
66        t1 = vec_perm(t1, t1, reverse);
67
68        vec_st(t0, i, dst);
69        vec_st(t1, j, dst);
70    }
71}
72
73void ff_vector_fmul_add_altivec(float *dst, const float *src0,
74                                const float *src1, const float *src2,
75                                int len)
76{
77    int i;
78    vector float d, s0, s1, s2, t0, t1, edges;
79    vector unsigned char align = vec_lvsr(0,dst),
80                         mask = vec_lvsl(0, dst);
81
82    for (i = 0; i < len - 3; i += 4) {
83        t0 = vec_ld(0, dst + i);
84        t1 = vec_ld(15, dst + i);
85        s0 = vec_ld(0, src0 + i);
86        s1 = vec_ld(0, src1 + i);
87        s2 = vec_ld(0, src2 + i);
88        edges = vec_perm(t1, t0, mask);
89        d = vec_madd(s0, s1, s2);
90        t1 = vec_perm(d, edges, align);
91        t0 = vec_perm(edges, d, align);
92        vec_st(t1, 15, dst + i);
93        vec_st(t0, 0, dst + i);
94    }
95}
96
97void ff_vector_fmul_reverse_altivec(float *dst, const float *src0,
98                                    const float *src1, int len)
99{
100    int i;
101    vector float d, s0, s1, h0, l0,
102                s2, s3, zero = (vector float) vec_splat_u32(0);
103
104    src1 += len-4;
105    for(i = 0; i < len - 7; i += 8) {
106        s1 = vec_ld(0, src1 - i);              // [a,b,c,d]
107        s0 = vec_ld(0, src0 + i);
108        l0 = vec_mergel(s1, s1);               // [c,c,d,d]
109        s3 = vec_ld(-16, src1 - i);
110        h0 = vec_mergeh(s1, s1);               // [a,a,b,b]
111        s2 = vec_ld(16, src0 + i);
112        s1 = vec_mergeh(vec_mergel(l0, h0),    // [d,b,d,b]
113                        vec_mergeh(l0, h0));   // [c,a,c,a]
114        // [d,c,b,a]
115        l0 = vec_mergel(s3, s3);
116        d = vec_madd(s0, s1, zero);
117        h0 = vec_mergeh(s3, s3);
118        vec_st(d, 0, dst + i);
119        s3 = vec_mergeh(vec_mergel(l0, h0),
120                        vec_mergeh(l0, h0));
121        d = vec_madd(s2, s3, zero);
122        vec_st(d, 16, dst + i);
123    }
124}
125