3 * Copyright (c) 2002 Fabrice Bellard.
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.
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.
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
22 * The size of the FFT is 2^nbits. If inverse is TRUE, inverse FFT is
25 int fft_init(FFTContext *s, int nbits, int inverse)
28 float alpha, c1, s1, s2;
33 s->exptab = av_malloc((n / 2) * sizeof(FFTComplex));
36 s->revtab = av_malloc(n * sizeof(uint16_t));
41 s2 = inverse ? 1.0 : -1.0;
43 for(i=0;i<(n/2);i++) {
44 alpha = 2 * M_PI * (float)i / (float)n;
50 s->fft_calc = fft_calc_c;
53 /* compute constant table for HAVE_SSE version */
54 #if (defined(HAVE_MMX) && defined(HAVE_BUILTIN_VECTOR)) || defined(HAVE_ALTIVEC)
59 has_vectors = mm_support() & MM_SSE;
61 #if defined(HAVE_ALTIVEC)
62 has_vectors = mm_support() & MM_ALTIVEC;
65 int np, nblocks, np2, l;
71 s->exptab1 = av_malloc(np * 2 * sizeof(FFTComplex));
76 for(l = 0; l < np2; l += 2 * nblocks) {
78 *q++ = s->exptab[l + nblocks];
80 q->re = -s->exptab[l].im;
81 q->im = s->exptab[l].re;
83 q->re = -s->exptab[l + nblocks].im;
84 q->im = s->exptab[l + nblocks].re;
87 nblocks = nblocks >> 1;
88 } while (nblocks != 0);
91 s->fft_calc = fft_calc_sse;
93 s->fft_calc = fft_calc_altivec;
99 /* compute bit reverse table */
103 for(j=0;j<nbits;j++) {
104 m |= ((i >> j) & 1) << (nbits-j-1);
110 av_freep(&s->revtab);
111 av_freep(&s->exptab);
112 av_freep(&s->exptab1);
117 #define BF(pre, pim, qre, qim, pre1, pim1, qre1, qim1) \
119 FFTSample ax, ay, bx, by;\
130 #define MUL16(a,b) ((a) * (b))
132 #define CMUL(pre, pim, are, aim, bre, bim) \
134 pre = (MUL16(are, bre) - MUL16(aim, bim));\
135 pim = (MUL16(are, bim) + MUL16(bre, aim));\
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.
143 void fft_calc_c(FFTContext *s, FFTComplex *z)
148 register FFTComplex *p, *q;
149 FFTComplex *exptab = s->exptab;
151 FFTSample tmp_re, tmp_im;
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);
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);
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);
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);
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);
212 nblocks = nblocks >> 1;
213 nloops = nloops << 1;
214 } while (nblocks != 0);
218 * Do the permutation needed BEFORE calling fft_calc()
220 void fft_permute(FFTContext *s, FFTComplex *z)
224 const uint16_t *revtab = s->revtab;
238 void fft_end(FFTContext *s)
240 av_freep(&s->revtab);
241 av_freep(&s->exptab);
242 av_freep(&s->exptab1);