1/*
2 * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org>
3 *
4 * This file is part of Libav.
5 *
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.
10 *
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.
15 *
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
19 */
20
21#include "libavcodec/fmtconvert.h"
22
23#include "dsputil_altivec.h"
24#include "util_altivec.h"
25
26static void int32_to_float_fmul_scalar_altivec(float *dst, const int *src, float mul, int len)
27{
28    union {
29        vector float v;
30        float s[4];
31    } mul_u;
32    int i;
33    vector float src1, src2, dst1, dst2, mul_v, zero;
34
35    zero = (vector float)vec_splat_u32(0);
36    mul_u.s[0] = mul;
37    mul_v = vec_splat(mul_u.v, 0);
38
39    for(i=0; i<len; i+=8) {
40        src1 = vec_ctf(vec_ld(0,  src+i), 0);
41        src2 = vec_ctf(vec_ld(16, src+i), 0);
42        dst1 = vec_madd(src1, mul_v, zero);
43        dst2 = vec_madd(src2, mul_v, zero);
44        vec_st(dst1,  0, dst+i);
45        vec_st(dst2, 16, dst+i);
46    }
47}
48
49
50static vector signed short
51float_to_int16_one_altivec(const float *src)
52{
53    vector float s0 = vec_ld(0, src);
54    vector float s1 = vec_ld(16, src);
55    vector signed int t0 = vec_cts(s0, 0);
56    vector signed int t1 = vec_cts(s1, 0);
57    return vec_packs(t0,t1);
58}
59
60static void float_to_int16_altivec(int16_t *dst, const float *src, long len)
61{
62    int i;
63    vector signed short d0, d1, d;
64    vector unsigned char align;
65    if(((long)dst)&15) //FIXME
66    for(i=0; i<len-7; i+=8) {
67        d0 = vec_ld(0, dst+i);
68        d = float_to_int16_one_altivec(src+i);
69        d1 = vec_ld(15, dst+i);
70        d1 = vec_perm(d1, d0, vec_lvsl(0,dst+i));
71        align = vec_lvsr(0, dst+i);
72        d0 = vec_perm(d1, d, align);
73        d1 = vec_perm(d, d1, align);
74        vec_st(d0, 0, dst+i);
75        vec_st(d1,15, dst+i);
76    }
77    else
78    for(i=0; i<len-7; i+=8) {
79        d = float_to_int16_one_altivec(src+i);
80        vec_st(d, 0, dst+i);
81    }
82}
83
84static void
85float_to_int16_interleave_altivec(int16_t *dst, const float **src,
86                                  long len, int channels)
87{
88    int i;
89    vector signed short d0, d1, d2, c0, c1, t0, t1;
90    vector unsigned char align;
91    if(channels == 1)
92        float_to_int16_altivec(dst, src[0], len);
93    else
94        if (channels == 2) {
95        if(((long)dst)&15)
96        for(i=0; i<len-7; i+=8) {
97            d0 = vec_ld(0, dst + i);
98            t0 = float_to_int16_one_altivec(src[0] + i);
99            d1 = vec_ld(31, dst + i);
100            t1 = float_to_int16_one_altivec(src[1] + i);
101            c0 = vec_mergeh(t0, t1);
102            c1 = vec_mergel(t0, t1);
103            d2 = vec_perm(d1, d0, vec_lvsl(0, dst + i));
104            align = vec_lvsr(0, dst + i);
105            d0 = vec_perm(d2, c0, align);
106            d1 = vec_perm(c0, c1, align);
107            vec_st(d0,  0, dst + i);
108            d0 = vec_perm(c1, d2, align);
109            vec_st(d1, 15, dst + i);
110            vec_st(d0, 31, dst + i);
111            dst+=8;
112        }
113        else
114        for(i=0; i<len-7; i+=8) {
115            t0 = float_to_int16_one_altivec(src[0] + i);
116            t1 = float_to_int16_one_altivec(src[1] + i);
117            d0 = vec_mergeh(t0, t1);
118            d1 = vec_mergel(t0, t1);
119            vec_st(d0,  0, dst + i);
120            vec_st(d1, 16, dst + i);
121            dst+=8;
122        }
123    } else {
124        DECLARE_ALIGNED(16, int16_t, tmp)[len];
125        int c, j;
126        for (c = 0; c < channels; c++) {
127            float_to_int16_altivec(tmp, src[c], len);
128            for (i = 0, j = c; i < len; i++, j+=channels) {
129                dst[j] = tmp[i];
130            }
131        }
132   }
133}
134
135void ff_fmt_convert_init_altivec(FmtConvertContext *c, AVCodecContext *avctx)
136{
137    c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_altivec;
138    if(!(avctx->flags & CODEC_FLAG_BITEXACT)) {
139        c->float_to_int16 = float_to_int16_altivec;
140        c->float_to_int16_interleave = float_to_int16_interleave_altivec;
141    }
142}
143