1 /*
2 * Copyright (c) 2015 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_vsx.h"
23
ff_vector_fmul_vsx(float * dst,const float * src0,const float * src1,int len)24 void ff_vector_fmul_vsx(float *dst,
25 const float *src0, const float *src1,
26 int len)
27 {
28 int i;
29 vec_f d0, d1, zero = (vec_f)vec_splat_u32(0);
30 for (i = 0; i < len - 7; i += 8) {
31 d0 = vec_vsx_ld( 0, src0 + i);
32 d1 = vec_vsx_ld(16, src0 + i);
33 d0 = vec_madd(d0, vec_vsx_ld( 0, src1 + i), zero);
34 d1 = vec_madd(d1, vec_vsx_ld(16, src1 + i), zero);
35 vec_vsx_st(d0, 0, dst + i);
36 vec_vsx_st(d1, 16, dst + i);
37 }
38 }
39
ff_vector_fmul_window_vsx(float * dst,const float * src0,const float * src1,const float * win,int len)40 void ff_vector_fmul_window_vsx(float *dst, const float *src0,
41 const float *src1, const float *win,
42 int len)
43 {
44 vec_f zero, t0, t1, s0, s1, wi, wj;
45 const vec_u8 reverse = vcprm(3, 2, 1, 0);
46 int i, j;
47
48 dst += len;
49 win += len;
50 src0 += len;
51
52 zero = (vec_f)vec_splat_u32(0);
53
54 for (i = -len * 4, j = len * 4 - 16; i < 0; i += 16, j -= 16) {
55 s0 = vec_vsx_ld(i, src0);
56 s1 = vec_vsx_ld(j, src1);
57 wi = vec_vsx_ld(i, win);
58 wj = vec_vsx_ld(j, win);
59
60 s1 = vec_perm(s1, s1, reverse);
61 wj = vec_perm(wj, wj, reverse);
62
63 t0 = vec_madd(s0, wj, zero);
64 t0 = vec_nmsub(s1, wi, t0);
65 t1 = vec_madd(s0, wi, zero);
66 t1 = vec_madd(s1, wj, t1);
67 t1 = vec_perm(t1, t1, reverse);
68
69 vec_vsx_st(t0, i, dst);
70 vec_vsx_st(t1, j, dst);
71 }
72 }
73
ff_vector_fmul_add_vsx(float * dst,const float * src0,const float * src1,const float * src2,int len)74 void ff_vector_fmul_add_vsx(float *dst, const float *src0,
75 const float *src1, const float *src2,
76 int len)
77 {
78 int i;
79 vec_f d, s0, s1, s2;
80
81 for (i = 0; i < len - 3; i += 4) {
82 s0 = vec_vsx_ld(0, src0 + i);
83 s1 = vec_vsx_ld(0, src1 + i);
84 s2 = vec_vsx_ld(0, src2 + i);
85 d = vec_madd(s0, s1, s2);
86 vec_vsx_st(d, 0, dst + i);
87 }
88 }
89
ff_vector_fmul_reverse_vsx(float * dst,const float * src0,const float * src1,int len)90 void ff_vector_fmul_reverse_vsx(float *dst, const float *src0,
91 const float *src1, int len)
92 {
93 int i;
94 vec_f d, s0, s1, h0, l0, s2, s3;
95 vec_f zero = (vec_f)vec_splat_u32(0);
96
97 src1 += len - 4;
98 for (i = 0; i < len - 7; i += 8) {
99 s1 = vec_vsx_ld(0, src1 - i); // [a,b,c,d]
100 s0 = vec_vsx_ld(0, src0 + i);
101 l0 = vec_mergel(s1, s1); // [c,c,d,d]
102 s3 = vec_vsx_ld(-16, src1 - i);
103 h0 = vec_mergeh(s1, s1); // [a,a,b,b]
104 s2 = vec_vsx_ld(16, src0 + i);
105 s1 = vec_mergeh(vec_mergel(l0, h0), // [d,b,d,b]
106 vec_mergeh(l0, h0)); // [c,a,c,a]
107 // [d,c,b,a]
108 l0 = vec_mergel(s3, s3);
109 d = vec_madd(s0, s1, zero);
110 h0 = vec_mergeh(s3, s3);
111 vec_vsx_st(d, 0, dst + i);
112 s3 = vec_mergeh(vec_mergel(l0, h0),
113 vec_mergeh(l0, h0));
114 d = vec_madd(s2, s3, zero);
115 vec_vsx_st(d, 16, dst + i);
116 }
117 }
118