blob: 04396b46d6c96999a6797540001f3fcb0a91fc31
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 | |
22 | #include <stdlib.h> |
23 | #include <string.h> |
24 | #include "libavutil/common.h" |
25 | #include "libavutil/libm.h" |
26 | #include "libavutil/mathematics.h" |
27 | #include "fft.h" |
28 | #include "fft-internal.h" |
29 | |
30 | /** |
31 | * @file |
32 | * MDCT/IMDCT transforms. |
33 | */ |
34 | |
35 | #if FFT_FLOAT |
36 | # define RSCALE(x) (x) |
37 | #else |
38 | #if FFT_FIXED_32 |
39 | # define RSCALE(x) (((x) + 32) >> 6) |
40 | #else /* FFT_FIXED_32 */ |
41 | # define RSCALE(x) ((x) >> 1) |
42 | #endif /* FFT_FIXED_32 */ |
43 | #endif |
44 | |
45 | /** |
46 | * init MDCT or IMDCT computation. |
47 | */ |
48 | av_cold int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale) |
49 | { |
50 | int n, n4, i; |
51 | double alpha, theta; |
52 | int tstep; |
53 | |
54 | memset(s, 0, sizeof(*s)); |
55 | n = 1 << nbits; |
56 | s->mdct_bits = nbits; |
57 | s->mdct_size = n; |
58 | n4 = n >> 2; |
59 | s->mdct_permutation = FF_MDCT_PERM_NONE; |
60 | |
61 | if (ff_fft_init(s, s->mdct_bits - 2, inverse) < 0) |
62 | goto fail; |
63 | |
64 | s->tcos = av_malloc_array(n/2, sizeof(FFTSample)); |
65 | if (!s->tcos) |
66 | goto fail; |
67 | |
68 | switch (s->mdct_permutation) { |
69 | case FF_MDCT_PERM_NONE: |
70 | s->tsin = s->tcos + n4; |
71 | tstep = 1; |
72 | break; |
73 | case FF_MDCT_PERM_INTERLEAVE: |
74 | s->tsin = s->tcos + 1; |
75 | tstep = 2; |
76 | break; |
77 | default: |
78 | goto fail; |
79 | } |
80 | |
81 | theta = 1.0 / 8.0 + (scale < 0 ? n4 : 0); |
82 | scale = sqrt(fabs(scale)); |
83 | for(i=0;i<n4;i++) { |
84 | alpha = 2 * M_PI * (i + theta) / n; |
85 | #if FFT_FIXED_32 |
86 | s->tcos[i*tstep] = lrint(-cos(alpha) * 2147483648.0); |
87 | s->tsin[i*tstep] = lrint(-sin(alpha) * 2147483648.0); |
88 | #else |
89 | s->tcos[i*tstep] = FIX15(-cos(alpha) * scale); |
90 | s->tsin[i*tstep] = FIX15(-sin(alpha) * scale); |
91 | #endif |
92 | } |
93 | return 0; |
94 | fail: |
95 | ff_mdct_end(s); |
96 | return -1; |
97 | } |
98 | |
99 | /** |
100 | * Compute the middle half of the inverse MDCT of size N = 2^nbits, |
101 | * thus excluding the parts that can be derived by symmetry |
102 | * @param output N/2 samples |
103 | * @param input N/2 samples |
104 | */ |
105 | void ff_imdct_half_c(FFTContext *s, FFTSample *output, const FFTSample *input) |
106 | { |
107 | int k, n8, n4, n2, n, j; |
108 | const uint16_t *revtab = s->revtab; |
109 | const FFTSample *tcos = s->tcos; |
110 | const FFTSample *tsin = s->tsin; |
111 | const FFTSample *in1, *in2; |
112 | FFTComplex *z = (FFTComplex *)output; |
113 | |
114 | n = 1 << s->mdct_bits; |
115 | n2 = n >> 1; |
116 | n4 = n >> 2; |
117 | n8 = n >> 3; |
118 | |
119 | /* pre rotation */ |
120 | in1 = input; |
121 | in2 = input + n2 - 1; |
122 | for(k = 0; k < n4; k++) { |
123 | j=revtab[k]; |
124 | CMUL(z[j].re, z[j].im, *in2, *in1, tcos[k], tsin[k]); |
125 | in1 += 2; |
126 | in2 -= 2; |
127 | } |
128 | s->fft_calc(s, z); |
129 | |
130 | /* post rotation + reordering */ |
131 | for(k = 0; k < n8; k++) { |
132 | FFTSample r0, i0, r1, i1; |
133 | CMUL(r0, i1, z[n8-k-1].im, z[n8-k-1].re, tsin[n8-k-1], tcos[n8-k-1]); |
134 | CMUL(r1, i0, z[n8+k ].im, z[n8+k ].re, tsin[n8+k ], tcos[n8+k ]); |
135 | z[n8-k-1].re = r0; |
136 | z[n8-k-1].im = i0; |
137 | z[n8+k ].re = r1; |
138 | z[n8+k ].im = i1; |
139 | } |
140 | } |
141 | |
142 | /** |
143 | * Compute inverse MDCT of size N = 2^nbits |
144 | * @param output N samples |
145 | * @param input N/2 samples |
146 | */ |
147 | void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input) |
148 | { |
149 | int k; |
150 | int n = 1 << s->mdct_bits; |
151 | int n2 = n >> 1; |
152 | int n4 = n >> 2; |
153 | |
154 | ff_imdct_half_c(s, output+n4, input); |
155 | |
156 | for(k = 0; k < n4; k++) { |
157 | output[k] = -output[n2-k-1]; |
158 | output[n-k-1] = output[n2+k]; |
159 | } |
160 | } |
161 | |
162 | /** |
163 | * Compute MDCT of size N = 2^nbits |
164 | * @param input N samples |
165 | * @param out N/2 samples |
166 | */ |
167 | void ff_mdct_calc_c(FFTContext *s, FFTSample *out, const FFTSample *input) |
168 | { |
169 | int i, j, n, n8, n4, n2, n3; |
170 | FFTDouble re, im; |
171 | const uint16_t *revtab = s->revtab; |
172 | const FFTSample *tcos = s->tcos; |
173 | const FFTSample *tsin = s->tsin; |
174 | FFTComplex *x = (FFTComplex *)out; |
175 | |
176 | n = 1 << s->mdct_bits; |
177 | n2 = n >> 1; |
178 | n4 = n >> 2; |
179 | n8 = n >> 3; |
180 | n3 = 3 * n4; |
181 | |
182 | /* pre rotation */ |
183 | for(i=0;i<n8;i++) { |
184 | re = RSCALE(-input[2*i+n3] - input[n3-1-2*i]); |
185 | im = RSCALE(-input[n4+2*i] + input[n4-1-2*i]); |
186 | j = revtab[i]; |
187 | CMUL(x[j].re, x[j].im, re, im, -tcos[i], tsin[i]); |
188 | |
189 | re = RSCALE( input[2*i] - input[n2-1-2*i]); |
190 | im = RSCALE(-input[n2+2*i] - input[ n-1-2*i]); |
191 | j = revtab[n8 + i]; |
192 | CMUL(x[j].re, x[j].im, re, im, -tcos[n8 + i], tsin[n8 + i]); |
193 | } |
194 | |
195 | s->fft_calc(s, x); |
196 | |
197 | /* post rotation */ |
198 | for(i=0;i<n8;i++) { |
199 | FFTSample r0, i0, r1, i1; |
200 | CMUL(i1, r0, x[n8-i-1].re, x[n8-i-1].im, -tsin[n8-i-1], -tcos[n8-i-1]); |
201 | CMUL(i0, r1, x[n8+i ].re, x[n8+i ].im, -tsin[n8+i ], -tcos[n8+i ]); |
202 | x[n8-i-1].re = r0; |
203 | x[n8-i-1].im = i0; |
204 | x[n8+i ].re = r1; |
205 | x[n8+i ].im = i1; |
206 | } |
207 | } |
208 | |
209 | av_cold void ff_mdct_end(FFTContext *s) |
210 | { |
211 | av_freep(&s->tcos); |
212 | ff_fft_end(s); |
213 | } |
214 |