Statistics
| Branch: | Revision:

ffmpeg / libavcodec / ppc / float_altivec.c @ 80ba1ddb

History | View | Annotate | Download (7.58 KB)

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 "libavcodec/dsputil.h"
22

    
23
#include "dsputil_altivec.h"
24
#include "util_altivec.h"
25

    
26
static void vector_fmul_altivec(float *dst, const float *src0, const float *src1, int len)
27
{
28
    int i;
29
    vector float d0, d1, s, zero = (vector float)vec_splat_u32(0);
30
    for(i=0; i<len-7; i+=8) {
31
        d0 = vec_ld(0, src0+i);
32
        s = vec_ld(0, src1+i);
33
        d1 = vec_ld(16, src0+i);
34
        d0 = vec_madd(d0, s, zero);
35
        d1 = vec_madd(d1, vec_ld(16,src1+i), zero);
36
        vec_st(d0, 0, dst+i);
37
        vec_st(d1, 16, dst+i);
38
    }
39
}
40

    
41
static void vector_fmul_reverse_altivec(float *dst, const float *src0,
42
                                        const float *src1, int len)
43
{
44
    int i;
45
    vector float d, s0, s1, h0, l0,
46
                 s2, s3, zero = (vector float)vec_splat_u32(0);
47
    src1 += len-4;
48
    for(i=0; i<len-7; i+=8) {
49
        s1 = vec_ld(0, src1-i);              // [a,b,c,d]
50
        s0 = vec_ld(0, src0+i);
51
        l0 = vec_mergel(s1, s1);             // [c,c,d,d]
52
        s3 = vec_ld(-16, src1-i);
53
        h0 = vec_mergeh(s1, s1);             // [a,a,b,b]
54
        s2 = vec_ld(16, src0+i);
55
        s1 = vec_mergeh(vec_mergel(l0,h0),   // [d,b,d,b]
56
                        vec_mergeh(l0,h0));  // [c,a,c,a]
57
                                             // [d,c,b,a]
58
        l0 = vec_mergel(s3, s3);
59
        d = vec_madd(s0, s1, zero);
60
        h0 = vec_mergeh(s3, s3);
61
        vec_st(d, 0, dst+i);
62
        s3 = vec_mergeh(vec_mergel(l0,h0),
63
                        vec_mergeh(l0,h0));
64
        d = vec_madd(s2, s3, zero);
65
        vec_st(d, 16, dst+i);
66
    }
67
}
68

    
69
static void vector_fmul_add_altivec(float *dst, const float *src0,
70
                                    const float *src1, const float *src2,
71
                                    int len)
72
{
73
    int i;
74
    vector float d, s0, s1, s2, t0, t1, edges;
75
    vector unsigned char align = vec_lvsr(0,dst),
76
                         mask = vec_lvsl(0, dst);
77

    
78
    for (i=0; i<len-3; i+=4) {
79
        t0 = vec_ld(0, dst+i);
80
        t1 = vec_ld(15, dst+i);
81
        s0 = vec_ld(0, src0+i);
82
        s1 = vec_ld(0, src1+i);
83
        s2 = vec_ld(0, src2+i);
84
        edges = vec_perm(t1 ,t0, mask);
85
        d = vec_madd(s0,s1,s2);
86
        t1 = vec_perm(d, edges, align);
87
        t0 = vec_perm(edges, d, align);
88
        vec_st(t1, 15, dst+i);
89
        vec_st(t0, 0, dst+i);
90
    }
91
}
92

    
93
static void vector_fmul_window_altivec(float *dst, const float *src0, const float *src1, const float *win, int len)
94
{
95
    vector float zero, t0, t1, s0, s1, wi, wj;
96
    const vector unsigned char reverse = vcprm(3,2,1,0);
97
    int i,j;
98

    
99
    dst += len;
100
    win += len;
101
    src0+= len;
102

    
103
    zero = (vector float)vec_splat_u32(0);
104

    
105
    for(i=-len*4, j=len*4-16; i<0; i+=16, j-=16) {
106
        s0 = vec_ld(i, src0);
107
        s1 = vec_ld(j, src1);
108
        wi = vec_ld(i, win);
109
        wj = vec_ld(j, win);
110

    
111
        s1 = vec_perm(s1, s1, reverse);
112
        wj = vec_perm(wj, wj, reverse);
113

    
114
        t0 = vec_madd(s0, wj, zero);
115
        t0 = vec_nmsub(s1, wi, t0);
116
        t1 = vec_madd(s0, wi, zero);
117
        t1 = vec_madd(s1, wj, t1);
118
        t1 = vec_perm(t1, t1, reverse);
119

    
120
        vec_st(t0, i, dst);
121
        vec_st(t1, j, dst);
122
    }
123
}
124

    
125
static void int32_to_float_fmul_scalar_altivec(float *dst, const int *src, float mul, int len)
126
{
127
    union {
128
        vector float v;
129
        float s[4];
130
    } mul_u;
131
    int i;
132
    vector float src1, src2, dst1, dst2, mul_v, zero;
133

    
134
    zero = (vector float)vec_splat_u32(0);
135
    mul_u.s[0] = mul;
136
    mul_v = vec_splat(mul_u.v, 0);
137

    
138
    for(i=0; i<len; i+=8) {
139
        src1 = vec_ctf(vec_ld(0,  src+i), 0);
140
        src2 = vec_ctf(vec_ld(16, src+i), 0);
141
        dst1 = vec_madd(src1, mul_v, zero);
142
        dst2 = vec_madd(src2, mul_v, zero);
143
        vec_st(dst1,  0, dst+i);
144
        vec_st(dst2, 16, dst+i);
145
    }
146
}
147

    
148

    
149
static vector signed short
150
float_to_int16_one_altivec(const float *src)
151
{
152
    vector float s0 = vec_ld(0, src);
153
    vector float s1 = vec_ld(16, src);
154
    vector signed int t0 = vec_cts(s0, 0);
155
    vector signed int t1 = vec_cts(s1, 0);
156
    return vec_packs(t0,t1);
157
}
158

    
159
static void float_to_int16_altivec(int16_t *dst, const float *src, long len)
160
{
161
    int i;
162
    vector signed short d0, d1, d;
163
    vector unsigned char align;
164
    if(((long)dst)&15) //FIXME
165
    for(i=0; i<len-7; i+=8) {
166
        d0 = vec_ld(0, dst+i);
167
        d = float_to_int16_one_altivec(src+i);
168
        d1 = vec_ld(15, dst+i);
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);
175
    }
176
    else
177
    for(i=0; i<len-7; i+=8) {
178
        d = float_to_int16_one_altivec(src+i);
179
        vec_st(d, 0, dst+i);
180
    }
181
}
182

    
183
static void
184
float_to_int16_interleave_altivec(int16_t *dst, const float **src,
185
                                  long len, int channels)
186
{
187
    int i;
188
    vector signed short d0, d1, d2, c0, c1, t0, t1;
189
    vector unsigned char align;
190
    if(channels == 1)
191
        float_to_int16_altivec(dst, src[0], len);
192
    else
193
        if (channels == 2) {
194
        if(((long)dst)&15)
195
        for(i=0; i<len-7; i+=8) {
196
            d0 = vec_ld(0, dst + i);
197
            t0 = float_to_int16_one_altivec(src[0] + i);
198
            d1 = vec_ld(31, dst + i);
199
            t1 = float_to_int16_one_altivec(src[1] + i);
200
            c0 = vec_mergeh(t0, t1);
201
            c1 = vec_mergel(t0, t1);
202
            d2 = vec_perm(d1, d0, vec_lvsl(0, dst + i));
203
            align = vec_lvsr(0, dst + i);
204
            d0 = vec_perm(d2, c0, align);
205
            d1 = vec_perm(c0, c1, align);
206
            vec_st(d0,  0, dst + i);
207
            d0 = vec_perm(c1, d2, align);
208
            vec_st(d1, 15, dst + i);
209
            vec_st(d0, 31, dst + i);
210
            dst+=8;
211
        }
212
        else
213
        for(i=0; i<len-7; i+=8) {
214
            t0 = float_to_int16_one_altivec(src[0] + i);
215
            t1 = float_to_int16_one_altivec(src[1] + i);
216
            d0 = vec_mergeh(t0, t1);
217
            d1 = vec_mergel(t0, t1);
218
            vec_st(d0,  0, dst + i);
219
            vec_st(d1, 16, dst + i);
220
            dst+=8;
221
        }
222
    } else {
223
        DECLARE_ALIGNED(16, int16_t, tmp)[len];
224
        int c, j;
225
        for (c = 0; c < channels; c++) {
226
            float_to_int16_altivec(tmp, src[c], len);
227
            for (i = 0, j = c; i < len; i++, j+=channels) {
228
                dst[j] = tmp[i];
229
            }
230
        }
231
   }
232
}
233

    
234
void float_init_altivec(DSPContext* c, AVCodecContext *avctx)
235
{
236
    c->vector_fmul = vector_fmul_altivec;
237
    c->vector_fmul_reverse = vector_fmul_reverse_altivec;
238
    c->vector_fmul_add = vector_fmul_add_altivec;
239
    c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_altivec;
240
    if(!(avctx->flags & CODEC_FLAG_BITEXACT)) {
241
        c->vector_fmul_window = vector_fmul_window_altivec;
242
        c->float_to_int16 = float_to_int16_altivec;
243
        c->float_to_int16_interleave = float_to_int16_interleave_altivec;
244
    }
245
}