aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/fft.c
blob: 1c63f6889ff3e8ec782168fedc142c7955b4181c (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
/*
 * FFT/IFFT transforms
 * Copyright (c) 2002 Fabrice Bellard.
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file fft.c
 * FFT/IFFT transforms.
 */

#include "dsputil.h"

/**
 * The size of the FFT is 2^nbits. If inverse is TRUE, inverse FFT is
 * done
 */
int ff_fft_init(FFTContext *s, int nbits, int inverse)
{
    int i, j, m, n;
    float alpha, c1, s1, s2;

    s->nbits = nbits;
    n = 1 << nbits;

    s->exptab = av_malloc((n / 2) * sizeof(FFTComplex));
    if (!s->exptab)
        goto fail;
    s->revtab = av_malloc(n * sizeof(uint16_t));
    if (!s->revtab)
        goto fail;
    s->inverse = inverse;

    s2 = inverse ? 1.0 : -1.0;

    for(i=0;i<(n/2);i++) {
        alpha = 2 * M_PI * (float)i / (float)n;
        c1 = cos(alpha);
        s1 = sin(alpha) * s2;
        s->exptab[i].re = c1;
        s->exptab[i].im = s1;
    }
    s->fft_calc = ff_fft_calc_c;
    s->exptab1 = NULL;

    /* compute constant table for HAVE_SSE version */
#if (defined(HAVE_MMX) && (defined(HAVE_BUILTIN_VECTOR) || defined(HAVE_MM3DNOW))) || defined(HAVE_ALTIVEC)
    {
        int has_vectors = 0;

#if defined(HAVE_MMX)
#ifdef HAVE_MM3DNOW
        has_vectors = mm_support() & (MM_3DNOW | MM_3DNOWEXT | MM_SSE | MM_SSE2);
#else
        has_vectors = mm_support() & (MM_SSE | MM_SSE2);
#endif
#endif
#if defined(HAVE_ALTIVEC) && !defined(ALTIVEC_USE_REFERENCE_C_CODE)
        has_vectors = mm_support() & MM_ALTIVEC;
#endif
        if (has_vectors) {
            int np, nblocks, np2, l;
            FFTComplex *q;

            np = 1 << nbits;
            nblocks = np >> 3;
            np2 = np >> 1;
            s->exptab1 = av_malloc(np * 2 * sizeof(FFTComplex));
            if (!s->exptab1)
                goto fail;
            q = s->exptab1;
            do {
                for(l = 0; l < np2; l += 2 * nblocks) {
                    *q++ = s->exptab[l];
                    *q++ = s->exptab[l + nblocks];

                    q->re = -s->exptab[l].im;
                    q->im = s->exptab[l].re;
                    q++;
                    q->re = -s->exptab[l + nblocks].im;
                    q->im = s->exptab[l + nblocks].re;
                    q++;
                }
                nblocks = nblocks >> 1;
            } while (nblocks != 0);
            av_freep(&s->exptab);
#if defined(HAVE_MMX)
#ifdef HAVE_MM3DNOW
            if (has_vectors & MM_3DNOWEXT)
                /* 3DNowEx for Athlon(XP) */
                s->fft_calc = ff_fft_calc_3dn2;
            else if (has_vectors & MM_3DNOW)
                /* 3DNow! for K6-2/3 */
                s->fft_calc = ff_fft_calc_3dn;
#endif
#ifdef HAVE_BUILTIN_VECTOR
            if (has_vectors & MM_SSE2)
                /* SSE for P4/K8 */
                s->fft_calc = ff_fft_calc_sse;
            else if ((has_vectors & MM_SSE) &&
                     s->fft_calc == ff_fft_calc_c)
                /* SSE for P3 */
                s->fft_calc = ff_fft_calc_sse;
#endif
#else /* HAVE_MMX */
            s->fft_calc = ff_fft_calc_altivec;
#endif
        }
    }
#endif

    /* compute bit reverse table */

    for(i=0;i<n;i++) {
        m=0;
        for(j=0;j<nbits;j++) {
            m |= ((i >> j) & 1) << (nbits-j-1);
        }
        s->revtab[i]=m;
    }
    return 0;
 fail:
    av_freep(&s->revtab);
    av_freep(&s->exptab);
    av_freep(&s->exptab1);
    return -1;
}

/* butter fly op */
#define BF(pre, pim, qre, qim, pre1, pim1, qre1, qim1) \
{\
  FFTSample ax, ay, bx, by;\
  bx=pre1;\
  by=pim1;\
  ax=qre1;\
  ay=qim1;\
  pre = (bx + ax);\
  pim = (by + ay);\
  qre = (bx - ax);\
  qim = (by - ay);\
}

#define MUL16(a,b) ((a) * (b))

#define CMUL(pre, pim, are, aim, bre, bim) \
{\
   pre = (MUL16(are, bre) - MUL16(aim, bim));\
   pim = (MUL16(are, bim) + MUL16(bre, aim));\
}

/**
 * Do a complex FFT with the parameters defined in ff_fft_init(). The
 * input data must be permuted before with s->revtab table. No
 * 1.0/sqrt(n) normalization is done.
 */
void ff_fft_calc_c(FFTContext *s, FFTComplex *z)
{
    int ln = s->nbits;
    int j, np, np2;
    int nblocks, nloops;
    register FFTComplex *p, *q;
    FFTComplex *exptab = s->exptab;
    int l;
    FFTSample tmp_re, tmp_im;

    np = 1 << ln;

    /* pass 0 */

    p=&z[0];
    j=(np >> 1);
    do {
        BF(p[0].re, p[0].im, p[1].re, p[1].im,
           p[0].re, p[0].im, p[1].re, p[1].im);
        p+=2;
    } while (--j != 0);

    /* pass 1 */


    p=&z[0];
    j=np >> 2;
    if (s->inverse) {
        do {
            BF(p[0].re, p[0].im, p[2].re, p[2].im,
               p[0].re, p[0].im, p[2].re, p[2].im);
            BF(p[1].re, p[1].im, p[3].re, p[3].im,
               p[1].re, p[1].im, -p[3].im, p[3].re);
            p+=4;
        } while (--j != 0);
    } else {
        do {
            BF(p[0].re, p[0].im, p[2].re, p[2].im,
               p[0].re, p[0].im, p[2].re, p[2].im);
            BF(p[1].re, p[1].im, p[3].re, p[3].im,
               p[1].re, p[1].im, p[3].im, -p[3].re);
            p+=4;
        } while (--j != 0);
    }
    /* pass 2 .. ln-1 */

    nblocks = np >> 3;
    nloops = 1 << 2;
    np2 = np >> 1;
    do {
        p = z;
        q = z + nloops;
        for (j = 0; j < nblocks; ++j) {
            BF(p->re, p->im, q->re, q->im,
               p->re, p->im, q->re, q->im);

            p++;
            q++;
            for(l = nblocks; l < np2; l += nblocks) {
                CMUL(tmp_re, tmp_im, exptab[l].re, exptab[l].im, q->re, q->im);
                BF(p->re, p->im, q->re, q->im,
                   p->re, p->im, tmp_re, tmp_im);
                p++;
                q++;
            }

            p += nloops;
            q += nloops;
        }
        nblocks = nblocks >> 1;
        nloops = nloops << 1;
    } while (nblocks != 0);
}

/**
 * Do the permutation needed BEFORE calling ff_fft_calc()
 */
void ff_fft_permute(FFTContext *s, FFTComplex *z)
{
    int j, k, np;
    FFTComplex tmp;
    const uint16_t *revtab = s->revtab;

    /* reverse */
    np = 1 << s->nbits;
    for(j=0;j<np;j++) {
        k = revtab[j];
        if (k < j) {
            tmp = z[k];
            z[k] = z[j];
            z[j] = tmp;
        }
    }
}

void ff_fft_end(FFTContext *s)
{
    av_freep(&s->revtab);
    av_freep(&s->exptab);
    av_freep(&s->exptab1);
}