xref: /third_party/ffmpeg/libavcodec/mdct15.c (revision cabdff1a) 
1/*
2 * Copyright (c) 2013-2014 Mozilla Corporation
3 * Copyright (c) 2017 Rostislav Pehlivanov <atomnuker@gmail.com>
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
23 * @file
24 * Celt non-power of 2 iMDCT
25 */
26
27#include <float.h>
28#include <math.h>
29#include <stddef.h>
30#include <stdint.h>
31
32#include "config.h"
33
34#include "libavutil/attributes.h"
35#include "libavutil/error.h"
36
37#include "mdct15.h"
38
39#define FFT_FLOAT 1
40#include "fft-internal.h"
41
42#define CMUL3(c, a, b) CMUL((c).re, (c).im, (a).re, (a).im, (b).re, (b).im)
43
44av_cold void ff_mdct15_uninit(MDCT15Context **ps)
45{
46    MDCT15Context *s = *ps;
47
48    if (!s)
49        return;
50
51    ff_fft_end(&s->ptwo_fft);
52
53    av_freep(&s->pfa_prereindex);
54    av_freep(&s->pfa_postreindex);
55    av_freep(&s->twiddle_exptab);
56    av_freep(&s->tmp);
57
58    av_freep(ps);
59}
60
61static inline int init_pfa_reindex_tabs(MDCT15Context *s)
62{
63    int i, j;
64    const int b_ptwo = s->ptwo_fft.nbits; /* Bits for the power of two FFTs */
65    const int l_ptwo = 1 << b_ptwo; /* Total length for the power of two FFTs */
66    const int inv_1 = l_ptwo << ((4 - b_ptwo) & 3); /* (2^b_ptwo)^-1 mod 15 */
67    const int inv_2 = 0xeeeeeeef & ((1U << b_ptwo) - 1); /* 15^-1 mod 2^b_ptwo */
68
69    s->pfa_prereindex = av_malloc_array(15 * l_ptwo, sizeof(*s->pfa_prereindex));
70    if (!s->pfa_prereindex)
71        return 1;
72
73    s->pfa_postreindex = av_malloc_array(15 * l_ptwo, sizeof(*s->pfa_postreindex));
74    if (!s->pfa_postreindex)
75        return 1;
76
77    /* Pre/Post-reindex */
78    for (i = 0; i < l_ptwo; i++) {
79        for (j = 0; j < 15; j++) {
80            const int q_pre = ((l_ptwo * j)/15 + i) >> b_ptwo;
81            const int q_post = (((j*inv_1)/15) + (i*inv_2)) >> b_ptwo;
82            const int k_pre = 15*i + (j - q_pre*15)*(1 << b_ptwo);
83            const int k_post = i*inv_2*15 + j*inv_1 - 15*q_post*l_ptwo;
84            s->pfa_prereindex[i*15 + j] = k_pre << 1;
85            s->pfa_postreindex[k_post] = l_ptwo*j + i;
86        }
87    }
88
89    return 0;
90}
91
92/* Stride is hardcoded to 3 */
93static inline void fft5(FFTComplex *out, FFTComplex *in, FFTComplex exptab[2])
94{
95    FFTComplex z0[4], t[6];
96
97    t[0].re = in[3].re + in[12].re;
98    t[0].im = in[3].im + in[12].im;
99    t[1].im = in[3].re - in[12].re;
100    t[1].re = in[3].im - in[12].im;
101    t[2].re = in[6].re + in[ 9].re;
102    t[2].im = in[6].im + in[ 9].im;
103    t[3].im = in[6].re - in[ 9].re;
104    t[3].re = in[6].im - in[ 9].im;
105
106    out[0].re = in[0].re + in[3].re + in[6].re + in[9].re + in[12].re;
107    out[0].im = in[0].im + in[3].im + in[6].im + in[9].im + in[12].im;
108
109    t[4].re = exptab[0].re * t[2].re - exptab[1].re * t[0].re;
110    t[4].im = exptab[0].re * t[2].im - exptab[1].re * t[0].im;
111    t[0].re = exptab[0].re * t[0].re - exptab[1].re * t[2].re;
112    t[0].im = exptab[0].re * t[0].im - exptab[1].re * t[2].im;
113    t[5].re = exptab[0].im * t[3].re - exptab[1].im * t[1].re;
114    t[5].im = exptab[0].im * t[3].im - exptab[1].im * t[1].im;
115    t[1].re = exptab[0].im * t[1].re + exptab[1].im * t[3].re;
116    t[1].im = exptab[0].im * t[1].im + exptab[1].im * t[3].im;
117
118    z0[0].re = t[0].re - t[1].re;
119    z0[0].im = t[0].im - t[1].im;
120    z0[1].re = t[4].re + t[5].re;
121    z0[1].im = t[4].im + t[5].im;
122
123    z0[2].re = t[4].re - t[5].re;
124    z0[2].im = t[4].im - t[5].im;
125    z0[3].re = t[0].re + t[1].re;
126    z0[3].im = t[0].im + t[1].im;
127
128    out[1].re = in[0].re + z0[3].re;
129    out[1].im = in[0].im + z0[0].im;
130    out[2].re = in[0].re + z0[2].re;
131    out[2].im = in[0].im + z0[1].im;
132    out[3].re = in[0].re + z0[1].re;
133    out[3].im = in[0].im + z0[2].im;
134    out[4].re = in[0].re + z0[0].re;
135    out[4].im = in[0].im + z0[3].im;
136}
137
138static void fft15_c(FFTComplex *out, FFTComplex *in, FFTComplex *exptab, ptrdiff_t stride)
139{
140    int k;
141    FFTComplex tmp1[5], tmp2[5], tmp3[5];
142
143    fft5(tmp1, in + 0, exptab + 19);
144    fft5(tmp2, in + 1, exptab + 19);
145    fft5(tmp3, in + 2, exptab + 19);
146
147    for (k = 0; k < 5; k++) {
148        FFTComplex t[2];
149
150        CMUL3(t[0], tmp2[k], exptab[k]);
151        CMUL3(t[1], tmp3[k], exptab[2 * k]);
152        out[stride*k].re = tmp1[k].re + t[0].re + t[1].re;
153        out[stride*k].im = tmp1[k].im + t[0].im + t[1].im;
154
155        CMUL3(t[0], tmp2[k], exptab[k + 5]);
156        CMUL3(t[1], tmp3[k], exptab[2 * (k + 5)]);
157        out[stride*(k + 5)].re = tmp1[k].re + t[0].re + t[1].re;
158        out[stride*(k + 5)].im = tmp1[k].im + t[0].im + t[1].im;
159
160        CMUL3(t[0], tmp2[k], exptab[k + 10]);
161        CMUL3(t[1], tmp3[k], exptab[2 * k + 5]);
162        out[stride*(k + 10)].re = tmp1[k].re + t[0].re + t[1].re;
163        out[stride*(k + 10)].im = tmp1[k].im + t[0].im + t[1].im;
164    }
165}
166
167static void mdct15(MDCT15Context *s, float *dst, const float *src, ptrdiff_t stride)
168{
169    int i, j;
170    const int len4 = s->len4, len3 = len4 * 3, len8 = len4 >> 1;
171    const int l_ptwo = 1 << s->ptwo_fft.nbits;
172    FFTComplex fft15in[15];
173
174    /* Folding and pre-reindexing */
175    for (i = 0; i < l_ptwo; i++) {
176        for (j = 0; j < 15; j++) {
177            const int k = s->pfa_prereindex[i*15 + j];
178            FFTComplex tmp, exp = s->twiddle_exptab[k >> 1];
179            if (k < len4) {
180                tmp.re = -src[ len4 + k] + src[1*len4 - 1 - k];
181                tmp.im = -src[ len3 + k] - src[1*len3 - 1 - k];
182            } else {
183                tmp.re = -src[ len4 + k] - src[5*len4 - 1 - k];
184                tmp.im =  src[-len4 + k] - src[1*len3 - 1 - k];
185            }
186            CMUL(fft15in[j].im, fft15in[j].re, tmp.re, tmp.im, exp.re, exp.im);
187        }
188        s->fft15(s->tmp + s->ptwo_fft.revtab[i], fft15in, s->exptab, l_ptwo);
189    }
190
191    /* Then a 15xN FFT (where N is a power of two) */
192    for (i = 0; i < 15; i++)
193        s->ptwo_fft.fft_calc(&s->ptwo_fft, s->tmp + l_ptwo*i);
194
195    /* Reindex again, apply twiddles and output */
196    for (i = 0; i < len8; i++) {
197        const int i0 = len8 + i, i1 = len8 - i - 1;
198        const int s0 = s->pfa_postreindex[i0], s1 = s->pfa_postreindex[i1];
199
200        CMUL(dst[2*i1*stride + stride], dst[2*i0*stride], s->tmp[s0].re, s->tmp[s0].im,
201             s->twiddle_exptab[i0].im, s->twiddle_exptab[i0].re);
202        CMUL(dst[2*i0*stride + stride], dst[2*i1*stride], s->tmp[s1].re, s->tmp[s1].im,
203             s->twiddle_exptab[i1].im, s->twiddle_exptab[i1].re);
204    }
205}
206
207static void imdct15_half(MDCT15Context *s, float *dst, const float *src,
208                         ptrdiff_t stride)
209{
210    FFTComplex fft15in[15];
211    FFTComplex *z = (FFTComplex *)dst;
212    int i, j, len8 = s->len4 >> 1, l_ptwo = 1 << s->ptwo_fft.nbits;
213    const float *in1 = src, *in2 = src + (s->len2 - 1) * stride;
214
215    /* Reindex input, putting it into a buffer and doing an Nx15 FFT */
216    for (i = 0; i < l_ptwo; i++) {
217        for (j = 0; j < 15; j++) {
218            const int k = s->pfa_prereindex[i*15 + j];
219            FFTComplex tmp = { in2[-k*stride], in1[k*stride] };
220            CMUL3(fft15in[j], tmp, s->twiddle_exptab[k >> 1]);
221        }
222        s->fft15(s->tmp + s->ptwo_fft.revtab[i], fft15in, s->exptab, l_ptwo);
223    }
224
225    /* Then a 15xN FFT (where N is a power of two) */
226    for (i = 0; i < 15; i++)
227        s->ptwo_fft.fft_calc(&s->ptwo_fft, s->tmp + l_ptwo*i);
228
229    /* Reindex again, apply twiddles and output */
230    s->postreindex(z, s->tmp, s->twiddle_exptab, s->pfa_postreindex, len8);
231}
232
233static void postrotate_c(FFTComplex *out, FFTComplex *in, FFTComplex *exp,
234                         int *lut, ptrdiff_t len8)
235{
236    int i;
237
238    /* Reindex again, apply twiddles and output */
239    for (i = 0; i < len8; i++) {
240        const int i0 = len8 + i, i1 = len8 - i - 1;
241        const int s0 = lut[i0], s1 = lut[i1];
242
243        CMUL(out[i1].re, out[i0].im, in[s1].im, in[s1].re, exp[i1].im, exp[i1].re);
244        CMUL(out[i0].re, out[i1].im, in[s0].im, in[s0].re, exp[i0].im, exp[i0].re);
245    }
246}
247
248av_cold int ff_mdct15_init(MDCT15Context **ps, int inverse, int N, double scale)
249{
250    MDCT15Context *s;
251    double alpha, theta;
252    int len2 = 15 * (1 << N);
253    int len  = 2 * len2;
254    int i;
255
256    /* Tested and verified to work on everything in between */
257    if ((N < 2) || (N > 13))
258        return AVERROR(EINVAL);
259
260    s = av_mallocz(sizeof(*s));
261    if (!s)
262        return AVERROR(ENOMEM);
263
264    s->fft_n       = N - 1;
265    s->len4        = len2 / 2;
266    s->len2        = len2;
267    s->inverse     = inverse;
268    s->fft15       = fft15_c;
269    s->mdct        = mdct15;
270    s->imdct_half  = imdct15_half;
271    s->postreindex = postrotate_c;
272
273    if (ff_fft_init(&s->ptwo_fft, N - 1, s->inverse) < 0)
274        goto fail;
275
276    if (init_pfa_reindex_tabs(s))
277        goto fail;
278
279    s->tmp  = av_malloc_array(len, 2 * sizeof(*s->tmp));
280    if (!s->tmp)
281        goto fail;
282
283    s->twiddle_exptab = av_malloc_array(s->len4, sizeof(*s->twiddle_exptab));
284    if (!s->twiddle_exptab)
285        goto fail;
286
287    theta = 0.125f + (scale < 0 ? s->len4 : 0);
288    scale = sqrt(fabs(scale));
289    for (i = 0; i < s->len4; i++) {
290        alpha = 2 * M_PI * (i + theta) / len;
291        s->twiddle_exptab[i].re = cosf(alpha) * scale;
292        s->twiddle_exptab[i].im = sinf(alpha) * scale;
293    }
294
295    /* 15-point FFT exptab */
296    for (i = 0; i < 19; i++) {
297        if (i < 15) {
298            double theta = (2.0f * M_PI * i) / 15.0f;
299            if (!s->inverse)
300                theta *= -1;
301            s->exptab[i].re = cosf(theta);
302            s->exptab[i].im = sinf(theta);
303        } else { /* Wrap around to simplify fft15 */
304            s->exptab[i] = s->exptab[i - 15];
305        }
306    }
307
308    /* 5-point FFT exptab */
309    s->exptab[19].re = cosf(2.0f * M_PI / 5.0f);
310    s->exptab[19].im = sinf(2.0f * M_PI / 5.0f);
311    s->exptab[20].re = cosf(1.0f * M_PI / 5.0f);
312    s->exptab[20].im = sinf(1.0f * M_PI / 5.0f);
313
314    /* Invert the phase for an inverse transform, do nothing for a forward transform */
315    if (s->inverse) {
316        s->exptab[19].im *= -1;
317        s->exptab[20].im *= -1;
318    }
319
320#if ARCH_X86
321    ff_mdct15_init_x86(s);
322#endif
323
324    *ps = s;
325
326    return 0;
327
328fail:
329    ff_mdct15_uninit(&s);
330    return AVERROR(ENOMEM);
331}
332