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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 * FFT/IFFT transforms.
28 * The size of the FFT is 2^nbits. If inverse is TRUE, inverse FFT is
31 int ff_fft_init(FFTContext *s, int nbits, int inverse)
34 float alpha, c1, s1, s2;
39 s->exptab = av_malloc((n / 2) * sizeof(FFTComplex));
42 s->revtab = av_malloc(n * sizeof(uint16_t));
47 s2 = inverse ? 1.0 : -1.0;
49 for(i=0;i<(n/2);i++) {
50 alpha = 2 * M_PI * (float)i / (float)n;
56 s->fft_calc = ff_fft_calc_c;
57 s->imdct_calc = ff_imdct_calc;
60 /* compute constant table for HAVE_SSE version */
61 #if defined(HAVE_MMX) \
62 || (defined(HAVE_ALTIVEC) && !defined(ALTIVEC_USE_REFERENCE_C_CODE))
64 int has_vectors = mm_support();
68 if (has_vectors & MM_3DNOWEXT) {
69 /* 3DNowEx for K7/K8 */
70 s->imdct_calc = ff_imdct_calc_3dn2;
71 s->fft_calc = ff_fft_calc_3dn2;
72 } else if (has_vectors & MM_3DNOW) {
73 /* 3DNow! for K6-2/3 */
74 s->fft_calc = ff_fft_calc_3dn;
75 } else if (has_vectors & MM_SSE) {
77 s->imdct_calc = ff_imdct_calc_sse;
78 s->fft_calc = ff_fft_calc_sse;
81 if (has_vectors & MM_ALTIVEC)
82 s->fft_calc = ff_fft_calc_altivec;
85 if (s->fft_calc != ff_fft_calc_c) {
86 int np, nblocks, np2, l;
92 s->exptab1 = av_malloc(np * 2 * sizeof(FFTComplex));
97 for(l = 0; l < np2; l += 2 * nblocks) {
99 *q++ = s->exptab[l + nblocks];
101 q->re = -s->exptab[l].im;
102 q->im = s->exptab[l].re;
104 q->re = -s->exptab[l + nblocks].im;
105 q->im = s->exptab[l + nblocks].re;
108 nblocks = nblocks >> 1;
109 } while (nblocks != 0);
110 av_freep(&s->exptab);
115 /* compute bit reverse table */
119 for(j=0;j<nbits;j++) {
120 m |= ((i >> j) & 1) << (nbits-j-1);
126 av_freep(&s->revtab);
127 av_freep(&s->exptab);
128 av_freep(&s->exptab1);
133 #define BF(pre, pim, qre, qim, pre1, pim1, qre1, qim1) \
135 FFTSample ax, ay, bx, by;\
146 #define MUL16(a,b) ((a) * (b))
148 #define CMUL(pre, pim, are, aim, bre, bim) \
150 pre = (MUL16(are, bre) - MUL16(aim, bim));\
151 pim = (MUL16(are, bim) + MUL16(bre, aim));\
155 * Do a complex FFT with the parameters defined in ff_fft_init(). The
156 * input data must be permuted before with s->revtab table. No
157 * 1.0/sqrt(n) normalization is done.
159 void ff_fft_calc_c(FFTContext *s, FFTComplex *z)
164 register FFTComplex *p, *q;
165 FFTComplex *exptab = s->exptab;
167 FFTSample tmp_re, tmp_im;
176 BF(p[0].re, p[0].im, p[1].re, p[1].im,
177 p[0].re, p[0].im, p[1].re, p[1].im);
188 BF(p[0].re, p[0].im, p[2].re, p[2].im,
189 p[0].re, p[0].im, p[2].re, p[2].im);
190 BF(p[1].re, p[1].im, p[3].re, p[3].im,
191 p[1].re, p[1].im, -p[3].im, p[3].re);
196 BF(p[0].re, p[0].im, p[2].re, p[2].im,
197 p[0].re, p[0].im, p[2].re, p[2].im);
198 BF(p[1].re, p[1].im, p[3].re, p[3].im,
199 p[1].re, p[1].im, p[3].im, -p[3].re);
211 for (j = 0; j < nblocks; ++j) {
212 BF(p->re, p->im, q->re, q->im,
213 p->re, p->im, q->re, q->im);
217 for(l = nblocks; l < np2; l += nblocks) {
218 CMUL(tmp_re, tmp_im, exptab[l].re, exptab[l].im, q->re, q->im);
219 BF(p->re, p->im, q->re, q->im,
220 p->re, p->im, tmp_re, tmp_im);
228 nblocks = nblocks >> 1;
229 nloops = nloops << 1;
230 } while (nblocks != 0);
234 * Do the permutation needed BEFORE calling ff_fft_calc()
236 void ff_fft_permute(FFTContext *s, FFTComplex *z)
240 const uint16_t *revtab = s->revtab;
254 void ff_fft_end(FFTContext *s)
256 av_freep(&s->revtab);
257 av_freep(&s->exptab);
258 av_freep(&s->exptab1);