2 * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org>
4 * This file is part of Libav.
6 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "libavcodec/fmtconvert.h"
23 #include "libavutil/ppc/util_altivec.h"
24 #include "libavutil/mem.h"
25 #include "dsputil_altivec.h"
27 static void int32_to_float_fmul_scalar_altivec(float *dst, const int *src,
35 vector float src1, src2, dst1, dst2, mul_v, zero;
37 zero = (vector float)vec_splat_u32(0);
39 mul_v = vec_splat(mul_u.v, 0);
41 for (i = 0; i < len; i += 8) {
42 src1 = vec_ctf(vec_ld(0, src+i), 0);
43 src2 = vec_ctf(vec_ld(16, src+i), 0);
44 dst1 = vec_madd(src1, mul_v, zero);
45 dst2 = vec_madd(src2, mul_v, zero);
46 vec_st(dst1, 0, dst+i);
47 vec_st(dst2, 16, dst+i);
52 static vector signed short float_to_int16_one_altivec(const float *src)
54 vector float s0 = vec_ld(0, src);
55 vector float s1 = vec_ld(16, src);
56 vector signed int t0 = vec_cts(s0, 0);
57 vector signed int t1 = vec_cts(s1, 0);
58 return vec_packs(t0,t1);
61 static void float_to_int16_altivec(int16_t *dst, const float *src, long len)
64 vector signed short d0, d1, d;
65 vector unsigned char align;
66 if (((long)dst) & 15) { //FIXME
67 for (i = 0; i < len - 7; i += 8) {
68 d0 = vec_ld(0, dst+i);
69 d = float_to_int16_one_altivec(src + i);
70 d1 = vec_ld(15, dst+i);
71 d1 = vec_perm(d1, d0, vec_lvsl(0, dst + i));
72 align = vec_lvsr(0, dst + i);
73 d0 = vec_perm(d1, d, align);
74 d1 = vec_perm(d, d1, align);
75 vec_st(d0, 0, dst + i);
76 vec_st(d1, 15, dst + i);
79 for (i = 0; i < len - 7; i += 8) {
80 d = float_to_int16_one_altivec(src + i);
81 vec_st(d, 0, dst + i);
86 #define VSTE_INC(dst, v, elem, inc) do { \
87 vector signed short s = vec_splat(v, elem); \
92 static void float_to_int16_stride_altivec(int16_t *dst, const float *src,
96 vector signed short d, s;
98 for (i = 0; i < len - 7; i += 8) {
99 d = float_to_int16_one_altivec(src + i);
100 VSTE_INC(dst, d, 0, stride);
101 VSTE_INC(dst, d, 1, stride);
102 VSTE_INC(dst, d, 2, stride);
103 VSTE_INC(dst, d, 3, stride);
104 VSTE_INC(dst, d, 4, stride);
105 VSTE_INC(dst, d, 5, stride);
106 VSTE_INC(dst, d, 6, stride);
107 VSTE_INC(dst, d, 7, stride);
111 static void float_to_int16_interleave_altivec(int16_t *dst, const float **src,
112 long len, int channels)
115 vector signed short d0, d1, d2, c0, c1, t0, t1;
116 vector unsigned char align;
119 float_to_int16_altivec(dst, src[0], len);
122 if (((long)dst) & 15) {
123 for (i = 0; i < len - 7; i += 8) {
124 d0 = vec_ld(0, dst + i);
125 t0 = float_to_int16_one_altivec(src[0] + i);
126 d1 = vec_ld(31, dst + i);
127 t1 = float_to_int16_one_altivec(src[1] + i);
128 c0 = vec_mergeh(t0, t1);
129 c1 = vec_mergel(t0, t1);
130 d2 = vec_perm(d1, d0, vec_lvsl(0, dst + i));
131 align = vec_lvsr(0, dst + i);
132 d0 = vec_perm(d2, c0, align);
133 d1 = vec_perm(c0, c1, align);
134 vec_st(d0, 0, dst + i);
135 d0 = vec_perm(c1, d2, align);
136 vec_st(d1, 15, dst + i);
137 vec_st(d0, 31, dst + i);
141 for (i = 0; i < len - 7; i += 8) {
142 t0 = float_to_int16_one_altivec(src[0] + i);
143 t1 = float_to_int16_one_altivec(src[1] + i);
144 d0 = vec_mergeh(t0, t1);
145 d1 = vec_mergel(t0, t1);
146 vec_st(d0, 0, dst + i);
147 vec_st(d1, 16, dst + i);
152 for (i = 0; i < channels; i++)
153 float_to_int16_stride_altivec(dst + i, src[i], len, channels);
158 void ff_fmt_convert_init_altivec(FmtConvertContext *c, AVCodecContext *avctx)
160 c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_altivec;
161 if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
162 c->float_to_int16 = float_to_int16_altivec;
163 c->float_to_int16_interleave = float_to_int16_interleave_altivec;