libavcodec/mdct.c

   1 /*
   2  * MDCT/IMDCT transforms
   3  * Copyright (c) 2002 Fabrice Bellard.
   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 #include "dsputil.h"
  22
  23 /**
  24  * @file mdct.c
  25  * MDCT/IMDCT transforms.
  26  */
  27
  28 // Generate a Kaiser-Bessel Derived Window.
  29 #define BESSEL_I0_ITER 50 // default: 50 iterations of Bessel I0 approximation
  30 void ff_kbd_window_init(float *window, float alpha, int n)
  31 {
  32    int i, j;
  33    double sum = 0.0, bessel, tmp;
  34    double local_window[n];
  35    double alpha2 = (alpha * M_PI / n) * (alpha * M_PI / n);
  36
  37    for (i = 0; i < n; i++) {
  38        tmp = i * (n - i) * alpha2;
  39        bessel = 1.0;
  40        for (j = BESSEL_I0_ITER; j > 0; j--)
  41            bessel = bessel * tmp / (j * j) + 1;
  42        sum += bessel;
  43        local_window[i] = sum;
  44    }
  45
  46    sum++;
  47    for (i = 0; i < n; i++)
  48        window[i] = sqrt(local_window[i] / sum);
  49 }
  50
  51 /**
  52  * init MDCT or IMDCT computation.
  53  */
  54 int ff_mdct_init(MDCTContext *s, int nbits, int inverse)
  55 {
  56     int n, n4, i;
  57     double alpha;
  58
  59     memset(s, 0, sizeof(*s));
  60     n = 1 << nbits;
  61     s->nbits = nbits;
  62     s->n = n;
  63     n4 = n >> 2;
  64     s->tcos = av_malloc(n4 * sizeof(FFTSample));
  65     if (!s->tcos)
  66         goto fail;
  67     s->tsin = av_malloc(n4 * sizeof(FFTSample));
  68     if (!s->tsin)
  69         goto fail;
  70
  71     for(i=0;i<n4;i++) {
  72         alpha = 2 * M_PI * (i + 1.0 / 8.0) / n;
  73         s->tcos[i] = -cos(alpha);
  74         s->tsin[i] = -sin(alpha);
  75     }
  76     if (ff_fft_init(&s->fft, s->nbits - 2, inverse) < 0)
  77         goto fail;
  78     return 0;
  79  fail:
  80     av_freep(&s->tcos);
  81     av_freep(&s->tsin);
  82     return -1;
  83 }
  84
  85 /* complex multiplication: p = a * b */
  86 #define CMUL(pre, pim, are, aim, bre, bim) \
  87 {\
  88     double _are = (are);\
  89     double _aim = (aim);\
  90     double _bre = (bre);\
  91     double _bim = (bim);\
  92     (pre) = _are * _bre - _aim * _bim;\
  93     (pim) = _are * _bim + _aim * _bre;\
  94 }
  95
  96 /**
  97  * Compute inverse MDCT of size N = 2^nbits
  98  * @param output N samples
  99  * @param input N/2 samples
 100  * @param tmp N/2 samples
 101  */
 102 void ff_imdct_calc(MDCTContext *s, FFTSample *output,
 103                    const FFTSample *input, FFTSample *tmp)
 104 {
 105     int k, n8, n4, n2, n, j;
 106     const uint16_t *revtab = s->fft.revtab;
 107     const FFTSample *tcos = s->tcos;
 108     const FFTSample *tsin = s->tsin;
 109     const FFTSample *in1, *in2;
 110     FFTComplex *z = (FFTComplex *)tmp;
 111
 112     n = 1 << s->nbits;
 113     n2 = n >> 1;
 114     n4 = n >> 2;
 115     n8 = n >> 3;
 116
 117     /* pre rotation */
 118     in1 = input;
 119     in2 = input + n2 - 1;
 120     for(k = 0; k < n4; k++) {
 121         j=revtab[k];
 122         CMUL(z[j].re, z[j].im, *in2, *in1, tcos[k], tsin[k]);
 123         in1 += 2;
 124         in2 -= 2;
 125     }
 126     ff_fft_calc(&s->fft, z);
 127
 128     /* post rotation + reordering */
 129     /* XXX: optimize */
 130     for(k = 0; k < n4; k++) {
 131         CMUL(z[k].re, z[k].im, z[k].re, z[k].im, tcos[k], tsin[k]);
 132     }
 133     for(k = 0; k < n8; k++) {
 134         output[2*k] = -z[n8 + k].im;
 135         output[n2-1-2*k] = z[n8 + k].im;
 136
 137         output[2*k+1] = z[n8-1-k].re;
 138         output[n2-1-2*k-1] = -z[n8-1-k].re;
 139
 140         output[n2 + 2*k]=-z[k+n8].re;
 141         output[n-1- 2*k]=-z[k+n8].re;
 142
 143         output[n2 + 2*k+1]=z[n8-k-1].im;
 144         output[n-2 - 2 * k] = z[n8-k-1].im;
 145     }
 146 }
 147
 148 /**
 149  * Compute MDCT of size N = 2^nbits
 150  * @param input N samples
 151  * @param out N/2 samples
 152  * @param tmp temporary storage of N/2 samples
 153  */
 154 void ff_mdct_calc(MDCTContext *s, FFTSample *out,
 155                   const FFTSample *input, FFTSample *tmp)
 156 {
 157     int i, j, n, n8, n4, n2, n3;
 158     FFTSample re, im, re1, im1;
 159     const uint16_t *revtab = s->fft.revtab;
 160     const FFTSample *tcos = s->tcos;
 161     const FFTSample *tsin = s->tsin;
 162     FFTComplex *x = (FFTComplex *)tmp;
 163
 164     n = 1 << s->nbits;
 165     n2 = n >> 1;
 166     n4 = n >> 2;
 167     n8 = n >> 3;
 168     n3 = 3 * n4;
 169
 170     /* pre rotation */
 171     for(i=0;i<n8;i++) {
 172         re = -input[2*i+3*n4] - input[n3-1-2*i];
 173         im = -input[n4+2*i] + input[n4-1-2*i];
 174         j = revtab[i];
 175         CMUL(x[j].re, x[j].im, re, im, -tcos[i], tsin[i]);
 176
 177         re = input[2*i] - input[n2-1-2*i];
 178         im = -(input[n2+2*i] + input[n-1-2*i]);
 179         j = revtab[n8 + i];
 180         CMUL(x[j].re, x[j].im, re, im, -tcos[n8 + i], tsin[n8 + i]);
 181     }
 182
 183     ff_fft_calc(&s->fft, x);
 184
 185     /* post rotation */
 186     for(i=0;i<n4;i++) {
 187         re = x[i].re;
 188         im = x[i].im;
 189         CMUL(re1, im1, re, im, -tsin[i], -tcos[i]);
 190         out[2*i] = im1;
 191         out[n2-1-2*i] = re1;
 192     }
 193 }
 194
 195 void ff_mdct_end(MDCTContext *s)
 196 {
 197     av_freep(&s->tcos);
 198     av_freep(&s->tsin);
 199     ff_fft_end(&s->fft);
 200 }