Libav
mdct_template.c
Go to the documentation of this file.
1 /*
2  * MDCT/IMDCT transforms
3  * Copyright (c) 2002 Fabrice Bellard
4  *
5  * This file is part of Libav.
6  *
7  * Libav 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  * Libav 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 Libav; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 #include <stdlib.h>
23 #include <string.h>
24 #include "libavutil/common.h"
25 #include "libavutil/mathematics.h"
26 #include "fft.h"
27 #include "fft-internal.h"
28 
34 #if FFT_FLOAT
35 # define RSCALE(x) (x)
36 #else
37 # define RSCALE(x) ((x) >> 1)
38 #endif
39 
43 av_cold int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale)
44 {
45  int n, n4, i;
46  double alpha, theta;
47  int tstep;
48 
49  memset(s, 0, sizeof(*s));
50  n = 1 << nbits;
51  s->mdct_bits = nbits;
52  s->mdct_size = n;
53  n4 = n >> 2;
55 
56  if (ff_fft_init(s, s->mdct_bits - 2, inverse) < 0)
57  goto fail;
58 
62 
63 #if FFT_FLOAT
64  if (ARCH_AARCH64)
66  if (ARCH_ARM)
68  if (ARCH_PPC)
70  if (ARCH_X86)
72  s->mdct_calcw = s->mdct_calc;
73 #else
75  if (ARCH_ARM)
77 #endif
78 
79  s->tcos = av_malloc(n/2 * sizeof(FFTSample));
80  if (!s->tcos)
81  goto fail;
82 
83  switch (s->mdct_permutation) {
84  case FF_MDCT_PERM_NONE:
85  s->tsin = s->tcos + n4;
86  tstep = 1;
87  break;
89  s->tsin = s->tcos + 1;
90  tstep = 2;
91  break;
92  default:
93  goto fail;
94  }
95 
96  theta = 1.0 / 8.0 + (scale < 0 ? n4 : 0);
97  scale = sqrt(fabs(scale));
98  for(i=0;i<n4;i++) {
99  alpha = 2 * M_PI * (i + theta) / n;
100  s->tcos[i*tstep] = FIX15(-cos(alpha) * scale);
101  s->tsin[i*tstep] = FIX15(-sin(alpha) * scale);
102  }
103  return 0;
104  fail:
105  ff_mdct_end(s);
106  return -1;
107 }
108 
115 void ff_imdct_half_c(FFTContext *s, FFTSample *output, const FFTSample *input)
116 {
117  int k, n8, n4, n2, n, j;
118  const uint16_t *revtab = s->revtab;
119  const FFTSample *tcos = s->tcos;
120  const FFTSample *tsin = s->tsin;
121  const FFTSample *in1, *in2;
122  FFTComplex *z = (FFTComplex *)output;
123 
124  n = 1 << s->mdct_bits;
125  n2 = n >> 1;
126  n4 = n >> 2;
127  n8 = n >> 3;
128 
129  /* pre rotation */
130  in1 = input;
131  in2 = input + n2 - 1;
132  for(k = 0; k < n4; k++) {
133  j=revtab[k];
134  CMUL(z[j].re, z[j].im, *in2, *in1, tcos[k], tsin[k]);
135  in1 += 2;
136  in2 -= 2;
137  }
138  s->fft_calc(s, z);
139 
140  /* post rotation + reordering */
141  for(k = 0; k < n8; k++) {
142  FFTSample r0, i0, r1, i1;
143  CMUL(r0, i1, z[n8-k-1].im, z[n8-k-1].re, tsin[n8-k-1], tcos[n8-k-1]);
144  CMUL(r1, i0, z[n8+k ].im, z[n8+k ].re, tsin[n8+k ], tcos[n8+k ]);
145  z[n8-k-1].re = r0;
146  z[n8-k-1].im = i0;
147  z[n8+k ].re = r1;
148  z[n8+k ].im = i1;
149  }
150 }
151 
157 void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input)
158 {
159  int k;
160  int n = 1 << s->mdct_bits;
161  int n2 = n >> 1;
162  int n4 = n >> 2;
163 
164  ff_imdct_half_c(s, output+n4, input);
165 
166  for(k = 0; k < n4; k++) {
167  output[k] = -output[n2-k-1];
168  output[n-k-1] = output[n2+k];
169  }
170 }
171 
178 {
179  int i, j, n, n8, n4, n2, n3;
180  FFTDouble re, im;
181  const uint16_t *revtab = s->revtab;
182  const FFTSample *tcos = s->tcos;
183  const FFTSample *tsin = s->tsin;
184  FFTComplex *x = (FFTComplex *)out;
185 
186  n = 1 << s->mdct_bits;
187  n2 = n >> 1;
188  n4 = n >> 2;
189  n8 = n >> 3;
190  n3 = 3 * n4;
191 
192  /* pre rotation */
193  for(i=0;i<n8;i++) {
194  re = RSCALE(-input[2*i+n3] - input[n3-1-2*i]);
195  im = RSCALE(-input[n4+2*i] + input[n4-1-2*i]);
196  j = revtab[i];
197  CMUL(x[j].re, x[j].im, re, im, -tcos[i], tsin[i]);
198 
199  re = RSCALE( input[2*i] - input[n2-1-2*i]);
200  im = RSCALE(-input[n2+2*i] - input[ n-1-2*i]);
201  j = revtab[n8 + i];
202  CMUL(x[j].re, x[j].im, re, im, -tcos[n8 + i], tsin[n8 + i]);
203  }
204 
205  s->fft_calc(s, x);
206 
207  /* post rotation */
208  for(i=0;i<n8;i++) {
209  FFTSample r0, i0, r1, i1;
210  CMUL(i1, r0, x[n8-i-1].re, x[n8-i-1].im, -tsin[n8-i-1], -tcos[n8-i-1]);
211  CMUL(i0, r1, x[n8+i ].re, x[n8+i ].im, -tsin[n8+i ], -tcos[n8+i ]);
212  x[n8-i-1].re = r0;
213  x[n8-i-1].im = i0;
214  x[n8+i ].re = r1;
215  x[n8+i ].im = i1;
216  }
217 }
218 
220 {
221  av_freep(&s->tcos);
222  ff_fft_end(s);
223 }
void * av_malloc(size_t size)
Allocate a block of size bytes with alignment suitable for all memory accesses (including vectors if ...
Definition: mem.c:62
float FFTDouble
Definition: fft.h:39
av_cold int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale)
init MDCT or IMDCT computation.
Definition: mdct_template.c:43
float re
Definition: fft.c:69
#define ARCH_PPC
Definition: config.h:24
FFTSample re
Definition: avfft.h:38
av_cold void ff_mdct_init_aarch64(FFTContext *s)
Definition: mdct_init.c:29
void ff_imdct_half_c(FFTContext *s, FFTSample *output, const FFTSample *input)
Compute the middle half of the inverse MDCT of size N = 2^nbits, thus excluding the parts that can be...
void av_freep(void *arg)
Free a memory block which has been allocated with av_malloc(z)() or av_realloc() and set the pointer ...
Definition: mem.c:202
#define av_cold
Definition: attributes.h:66
#define CMUL(dre, dim, are, aim, bre, bim)
Definition: fft-internal.h:59
void ff_mdct_calc_c(FFTContext *s, FFTSample *out, const FFTSample *input)
Compute MDCT of size N = 2^nbits.
#define FIX15(a)
Definition: fft-internal.h:45
#define ARCH_X86
Definition: config.h:33
void(* mdct_calcw)(struct FFTContext *s, FFTDouble *output, const FFTSample *input)
Definition: fft.h:95
av_cold void ff_mdct_end(FFTContext *s)
void(* mdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input)
Definition: fft.h:94
void(* imdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input)
Definition: fft.h:92
float FFTSample
Definition: avfft.h:35
#define fail()
Definition: checkasm.h:80
void ff_mdct_init_x86(FFTContext *s)
Definition: mdct_init.c:27
void ff_mdct_init_ppc(FFTContext *s)
Definition: mdct_init.c:143
Definition: fft.h:73
FFTSample * tsin
Definition: fft.h:82
#define ARCH_ARM
Definition: config.h:14
#define ff_fft_init
Definition: fft.h:132
#define RSCALE(x)
Definition: mdct_template.c:35
int mdct_bits
Definition: fft.h:79
float im
Definition: fft.c:69
void(* imdct_half)(struct FFTContext *s, FFTSample *output, const FFTSample *input)
Definition: fft.h:93
av_cold void ff_mdct_fixed_init_arm(FFTContext *s)
FFTSample im
Definition: avfft.h:38
common internal and external API header
av_cold void ff_mdct_init_arm(FFTContext *s)
Definition: mdct_init_arm.c:33
void ff_mdct_calcw_c(FFTContext *s, FFTDouble *output, const FFTSample *input)
Definition: mdct_fixed.c:23
#define ff_fft_end
Definition: fft.h:133
void(* fft_calc)(struct FFTContext *s, FFTComplex *z)
Do a complex FFT with the parameters defined in ff_fft_init().
Definition: fft.h:91
FFTSample * tcos
Definition: fft.h:81
#define ARCH_AARCH64
Definition: config.h:12
FILE * out
Definition: movenc.c:54
uint16_t * revtab
Definition: fft.h:76
static uint32_t inverse(uint32_t v)
find multiplicative inverse modulo 2 ^ 32
Definition: asfcrypt.c:35
int mdct_size
Definition: fft.h:78
enum mdct_permutation_type mdct_permutation
Definition: fft.h:97
void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input)
Compute inverse MDCT of size N = 2^nbits.