Statistics
| Branch: | Revision:

ffmpeg / libavcodec / fft.c @ ea937d01

History | View | Annotate | Download (6.05 KB)

1 bb6f5690 Fabrice Bellard
/*
2
 * FFT/IFFT transforms
3
 * Copyright (c) 2002 Fabrice Bellard.
4
 *
5
 * This library is free software; you can redistribute it and/or
6
 * modify it under the terms of the GNU Lesser General Public
7
 * License as published by the Free Software Foundation; either
8
 * version 2 of the License, or (at your option) any later version.
9
 *
10
 * This library is distributed in the hope that it will be useful,
11
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13
 * Lesser General Public License for more details.
14
 *
15
 * You should have received a copy of the GNU Lesser General Public
16
 * License along with this library; if not, write to the Free Software
17
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
18
 */
19
#include "dsputil.h"
20
21
/**
22
 * The size of the FFT is 2^nbits. If inverse is TRUE, inverse FFT is
23
 * done 
24
 */
25
int fft_init(FFTContext *s, int nbits, int inverse)
26
{
27
    int i, j, m, n;
28
    float alpha, c1, s1, s2;
29
    
30
    s->nbits = nbits;
31
    n = 1 << nbits;
32
33
    s->exptab = av_malloc((n / 2) * sizeof(FFTComplex));
34
    if (!s->exptab)
35
        goto fail;
36
    s->revtab = av_malloc(n * sizeof(uint16_t));
37
    if (!s->revtab)
38
        goto fail;
39
    s->inverse = inverse;
40
41
    s2 = inverse ? 1.0 : -1.0;
42
        
43
    for(i=0;i<(n/2);i++) {
44
        alpha = 2 * M_PI * (float)i / (float)n;
45
        c1 = cos(alpha);
46
        s1 = sin(alpha) * s2;
47
        s->exptab[i].re = c1;
48
        s->exptab[i].im = s1;
49
    }
50
    s->fft_calc = fft_calc_c;
51
    s->exptab1 = NULL;
52
53
    /* compute constant table for HAVE_SSE version */
54 8d268a7d Fabrice Bellard
#if (defined(HAVE_MMX) && defined(HAVE_BUILTIN_VECTOR)) || defined(HAVE_ALTIVEC)
55
    {
56 db40a39a Michael Niedermayer
        int has_vectors = 0;
57 8d268a7d Fabrice Bellard
58
#if defined(HAVE_MMX)
59
        has_vectors = mm_support() & MM_SSE;
60 e629ab68 Romain Dolbeau
#endif
61 db40a39a Michael Niedermayer
#if defined(HAVE_ALTIVEC) && !defined(ALTIVEC_USE_REFERENCE_C_CODE)
62 e629ab68 Romain Dolbeau
        has_vectors = mm_support() & MM_ALTIVEC;
63 8d268a7d Fabrice Bellard
#endif
64
        if (has_vectors) {
65
            int np, nblocks, np2, l;
66
            FFTComplex *q;
67
            
68
            np = 1 << nbits;
69
            nblocks = np >> 3;
70
            np2 = np >> 1;
71
            s->exptab1 = av_malloc(np * 2 * sizeof(FFTComplex));
72
            if (!s->exptab1)
73
                goto fail;
74
            q = s->exptab1;
75
            do {
76
                for(l = 0; l < np2; l += 2 * nblocks) {
77
                    *q++ = s->exptab[l];
78
                    *q++ = s->exptab[l + nblocks];
79 bb6f5690 Fabrice Bellard
80 8d268a7d Fabrice Bellard
                    q->re = -s->exptab[l].im;
81
                    q->im = s->exptab[l].re;
82
                    q++;
83
                    q->re = -s->exptab[l + nblocks].im;
84
                    q->im = s->exptab[l + nblocks].re;
85
                    q++;
86
                }
87
                nblocks = nblocks >> 1;
88
            } while (nblocks != 0);
89
            av_freep(&s->exptab);
90
#if defined(HAVE_MMX)
91
            s->fft_calc = fft_calc_sse;
92
#else
93
            s->fft_calc = fft_calc_altivec;
94
#endif
95
        }
96 bb6f5690 Fabrice Bellard
    }
97
#endif
98
99
    /* compute bit reverse table */
100
101
    for(i=0;i<n;i++) {
102
        m=0;
103
        for(j=0;j<nbits;j++) {
104
            m |= ((i >> j) & 1) << (nbits-j-1);
105
        }
106
        s->revtab[i]=m;
107
    }
108
    return 0;
109
 fail:
110
    av_freep(&s->revtab);
111
    av_freep(&s->exptab);
112
    av_freep(&s->exptab1);
113
    return -1;
114
}
115
116
/* butter fly op */
117
#define BF(pre, pim, qre, qim, pre1, pim1, qre1, qim1) \
118
{\
119
  FFTSample ax, ay, bx, by;\
120
  bx=pre1;\
121
  by=pim1;\
122
  ax=qre1;\
123
  ay=qim1;\
124
  pre = (bx + ax);\
125
  pim = (by + ay);\
126
  qre = (bx - ax);\
127
  qim = (by - ay);\
128
}
129
130
#define MUL16(a,b) ((a) * (b))
131
132
#define CMUL(pre, pim, are, aim, bre, bim) \
133
{\
134
   pre = (MUL16(are, bre) - MUL16(aim, bim));\
135
   pim = (MUL16(are, bim) + MUL16(bre, aim));\
136
}
137
138
/**
139
 * Do a complex FFT with the parameters defined in fft_init(). The
140
 * input data must be permuted before with s->revtab table. No
141
 * 1.0/sqrt(n) normalization is done.  
142
 */
143
void fft_calc_c(FFTContext *s, FFTComplex *z)
144
{
145
    int ln = s->nbits;
146
    int        j, np, np2;
147
    int        nblocks, nloops;
148
    register FFTComplex *p, *q;
149
    FFTComplex *exptab = s->exptab;
150
    int l;
151
    FFTSample tmp_re, tmp_im;
152
153
    np = 1 << ln;
154
155
    /* pass 0 */
156
157
    p=&z[0];
158
    j=(np >> 1);
159
    do {
160
        BF(p[0].re, p[0].im, p[1].re, p[1].im, 
161
           p[0].re, p[0].im, p[1].re, p[1].im);
162
        p+=2;
163
    } while (--j != 0);
164
165
    /* pass 1 */
166
167
    
168
    p=&z[0];
169
    j=np >> 2;
170
    if (s->inverse) {
171
        do {
172
            BF(p[0].re, p[0].im, p[2].re, p[2].im, 
173
               p[0].re, p[0].im, p[2].re, p[2].im);
174
            BF(p[1].re, p[1].im, p[3].re, p[3].im, 
175
               p[1].re, p[1].im, -p[3].im, p[3].re);
176
            p+=4;
177
        } while (--j != 0);
178
    } else {
179
        do {
180
            BF(p[0].re, p[0].im, p[2].re, p[2].im, 
181
               p[0].re, p[0].im, p[2].re, p[2].im);
182
            BF(p[1].re, p[1].im, p[3].re, p[3].im, 
183
               p[1].re, p[1].im, p[3].im, -p[3].re);
184
            p+=4;
185
        } while (--j != 0);
186
    }
187
    /* pass 2 .. ln-1 */
188
189
    nblocks = np >> 3;
190
    nloops = 1 << 2;
191
    np2 = np >> 1;
192
    do {
193
        p = z;
194
        q = z + nloops;
195
        for (j = 0; j < nblocks; ++j) {
196
            BF(p->re, p->im, q->re, q->im,
197
               p->re, p->im, q->re, q->im);
198
            
199
            p++;
200
            q++;
201
            for(l = nblocks; l < np2; l += nblocks) {
202
                CMUL(tmp_re, tmp_im, exptab[l].re, exptab[l].im, q->re, q->im);
203
                BF(p->re, p->im, q->re, q->im,
204
                   p->re, p->im, tmp_re, tmp_im);
205
                p++;
206
                q++;
207
            }
208
209
            p += nloops;
210
            q += nloops;
211
        }
212
        nblocks = nblocks >> 1;
213
        nloops = nloops << 1;
214
    } while (nblocks != 0);
215
}
216
217
/**
218
 * Do the permutation needed BEFORE calling fft_calc()
219
 */
220
void fft_permute(FFTContext *s, FFTComplex *z)
221
{
222
    int j, k, np;
223
    FFTComplex tmp;
224
    const uint16_t *revtab = s->revtab;
225
    
226
    /* reverse */
227
    np = 1 << s->nbits;
228
    for(j=0;j<np;j++) {
229
        k = revtab[j];
230
        if (k < j) {
231
            tmp = z[k];
232
            z[k] = z[j];
233
            z[j] = tmp;
234
        }
235
    }
236
}
237
238
void fft_end(FFTContext *s)
239
{
240
    av_freep(&s->revtab);
241
    av_freep(&s->exptab);
242
    av_freep(&s->exptab1);
243
}