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
31 #include "config.h"
32
33 #include "libavutil/attributes.h"
34 #include "libavutil/common.h"
35
36 #include "mdct15.h"
37
38 #define FFT_FLOAT 1
39 #include "fft-internal.h"
40
41 #define CMUL3(c, a, b) CMUL((c).re, (c).im, (a).re, (a).im, (b).re, (b).im)
42
ff_mdct15_uninit(MDCT15Context ** ps)43 av_cold void ff_mdct15_uninit(MDCT15Context **ps)
44 {
45 MDCT15Context *s = *ps;
46
47 if (!s)
48 return;
49
50 ff_fft_end(&s->ptwo_fft);
51
52 av_freep(&s->pfa_prereindex);
53 av_freep(&s->pfa_postreindex);
54 av_freep(&s->twiddle_exptab);
55 av_freep(&s->tmp);
56
57 av_freep(ps);
58 }
59
init_pfa_reindex_tabs(MDCT15Context * s)60 static inline int init_pfa_reindex_tabs(MDCT15Context *s)
61 {
62 int i, j;
63 const int b_ptwo = s->ptwo_fft.nbits; /* Bits for the power of two FFTs */
64 const int l_ptwo = 1 << b_ptwo; /* Total length for the power of two FFTs */
65 const int inv_1 = l_ptwo << ((4 - b_ptwo) & 3); /* (2^b_ptwo)^-1 mod 15 */
66 const int inv_2 = 0xeeeeeeef & ((1U << b_ptwo) - 1); /* 15^-1 mod 2^b_ptwo */
67
68 s->pfa_prereindex = av_malloc_array(15 * l_ptwo, sizeof(*s->pfa_prereindex));
69 if (!s->pfa_prereindex)
70 return 1;
71
72 s->pfa_postreindex = av_malloc_array(15 * l_ptwo, sizeof(*s->pfa_postreindex));
73 if (!s->pfa_postreindex)
74 return 1;
75
76 /* Pre/Post-reindex */
77 for (i = 0; i < l_ptwo; i++) {
78 for (j = 0; j < 15; j++) {
79 const int q_pre = ((l_ptwo * j)/15 + i) >> b_ptwo;
80 const int q_post = (((j*inv_1)/15) + (i*inv_2)) >> b_ptwo;
81 const int k_pre = 15*i + (j - q_pre*15)*(1 << b_ptwo);
82 const int k_post = i*inv_2*15 + j*inv_1 - 15*q_post*l_ptwo;
83 s->pfa_prereindex[i*15 + j] = k_pre << 1;
84 s->pfa_postreindex[k_post] = l_ptwo*j + i;
85 }
86 }
87
88 return 0;
89 }
90
91 /* Stride is hardcoded to 3 */
fft5(FFTComplex * out,FFTComplex * in,FFTComplex exptab[2])92 static inline void fft5(FFTComplex *out, FFTComplex *in, FFTComplex exptab[2])
93 {
94 FFTComplex z0[4], t[6];
95
96 t[0].re = in[3].re + in[12].re;
97 t[0].im = in[3].im + in[12].im;
98 t[1].im = in[3].re - in[12].re;
99 t[1].re = in[3].im - in[12].im;
100 t[2].re = in[6].re + in[ 9].re;
101 t[2].im = in[6].im + in[ 9].im;
102 t[3].im = in[6].re - in[ 9].re;
103 t[3].re = in[6].im - in[ 9].im;
104
105 out[0].re = in[0].re + in[3].re + in[6].re + in[9].re + in[12].re;
106 out[0].im = in[0].im + in[3].im + in[6].im + in[9].im + in[12].im;
107
108 t[4].re = exptab[0].re * t[2].re - exptab[1].re * t[0].re;
109 t[4].im = exptab[0].re * t[2].im - exptab[1].re * t[0].im;
110 t[0].re = exptab[0].re * t[0].re - exptab[1].re * t[2].re;
111 t[0].im = exptab[0].re * t[0].im - exptab[1].re * t[2].im;
112 t[5].re = exptab[0].im * t[3].re - exptab[1].im * t[1].re;
113 t[5].im = exptab[0].im * t[3].im - exptab[1].im * t[1].im;
114 t[1].re = exptab[0].im * t[1].re + exptab[1].im * t[3].re;
115 t[1].im = exptab[0].im * t[1].im + exptab[1].im * t[3].im;
116
117 z0[0].re = t[0].re - t[1].re;
118 z0[0].im = t[0].im - t[1].im;
119 z0[1].re = t[4].re + t[5].re;
120 z0[1].im = t[4].im + t[5].im;
121
122 z0[2].re = t[4].re - t[5].re;
123 z0[2].im = t[4].im - t[5].im;
124 z0[3].re = t[0].re + t[1].re;
125 z0[3].im = t[0].im + t[1].im;
126
127 out[1].re = in[0].re + z0[3].re;
128 out[1].im = in[0].im + z0[0].im;
129 out[2].re = in[0].re + z0[2].re;
130 out[2].im = in[0].im + z0[1].im;
131 out[3].re = in[0].re + z0[1].re;
132 out[3].im = in[0].im + z0[2].im;
133 out[4].re = in[0].re + z0[0].re;
134 out[4].im = in[0].im + z0[3].im;
135 }
136
fft15_c(FFTComplex * out,FFTComplex * in,FFTComplex * exptab,ptrdiff_t stride)137 static void fft15_c(FFTComplex *out, FFTComplex *in, FFTComplex *exptab, ptrdiff_t stride)
138 {
139 int k;
140 FFTComplex tmp1[5], tmp2[5], tmp3[5];
141
142 fft5(tmp1, in + 0, exptab + 19);
143 fft5(tmp2, in + 1, exptab + 19);
144 fft5(tmp3, in + 2, exptab + 19);
145
146 for (k = 0; k < 5; k++) {
147 FFTComplex t[2];
148
149 CMUL3(t[0], tmp2[k], exptab[k]);
150 CMUL3(t[1], tmp3[k], exptab[2 * k]);
151 out[stride*k].re = tmp1[k].re + t[0].re + t[1].re;
152 out[stride*k].im = tmp1[k].im + t[0].im + t[1].im;
153
154 CMUL3(t[0], tmp2[k], exptab[k + 5]);
155 CMUL3(t[1], tmp3[k], exptab[2 * (k + 5)]);
156 out[stride*(k + 5)].re = tmp1[k].re + t[0].re + t[1].re;
157 out[stride*(k + 5)].im = tmp1[k].im + t[0].im + t[1].im;
158
159 CMUL3(t[0], tmp2[k], exptab[k + 10]);
160 CMUL3(t[1], tmp3[k], exptab[2 * k + 5]);
161 out[stride*(k + 10)].re = tmp1[k].re + t[0].re + t[1].re;
162 out[stride*(k + 10)].im = tmp1[k].im + t[0].im + t[1].im;
163 }
164 }
165
mdct15(MDCT15Context * s,float * dst,const float * src,ptrdiff_t stride)166 static void mdct15(MDCT15Context *s, float *dst, const float *src, ptrdiff_t stride)
167 {
168 int i, j;
169 const int len4 = s->len4, len3 = len4 * 3, len8 = len4 >> 1;
170 const int l_ptwo = 1 << s->ptwo_fft.nbits;
171 FFTComplex fft15in[15];
172
173 /* Folding and pre-reindexing */
174 for (i = 0; i < l_ptwo; i++) {
175 for (j = 0; j < 15; j++) {
176 const int k = s->pfa_prereindex[i*15 + j];
177 FFTComplex tmp, exp = s->twiddle_exptab[k >> 1];
178 if (k < len4) {
179 tmp.re = -src[ len4 + k] + src[1*len4 - 1 - k];
180 tmp.im = -src[ len3 + k] - src[1*len3 - 1 - k];
181 } else {
182 tmp.re = -src[ len4 + k] - src[5*len4 - 1 - k];
183 tmp.im = src[-len4 + k] - src[1*len3 - 1 - k];
184 }
185 CMUL(fft15in[j].im, fft15in[j].re, tmp.re, tmp.im, exp.re, exp.im);
186 }
187 s->fft15(s->tmp + s->ptwo_fft.revtab[i], fft15in, s->exptab, l_ptwo);
188 }
189
190 /* Then a 15xN FFT (where N is a power of two) */
191 for (i = 0; i < 15; i++)
192 s->ptwo_fft.fft_calc(&s->ptwo_fft, s->tmp + l_ptwo*i);
193
194 /* Reindex again, apply twiddles and output */
195 for (i = 0; i < len8; i++) {
196 const int i0 = len8 + i, i1 = len8 - i - 1;
197 const int s0 = s->pfa_postreindex[i0], s1 = s->pfa_postreindex[i1];
198
199 CMUL(dst[2*i1*stride + stride], dst[2*i0*stride], s->tmp[s0].re, s->tmp[s0].im,
200 s->twiddle_exptab[i0].im, s->twiddle_exptab[i0].re);
201 CMUL(dst[2*i0*stride + stride], dst[2*i1*stride], s->tmp[s1].re, s->tmp[s1].im,
202 s->twiddle_exptab[i1].im, s->twiddle_exptab[i1].re);
203 }
204 }
205
imdct15_half(MDCT15Context * s,float * dst,const float * src,ptrdiff_t stride)206 static void imdct15_half(MDCT15Context *s, float *dst, const float *src,
207 ptrdiff_t stride)
208 {
209 FFTComplex fft15in[15];
210 FFTComplex *z = (FFTComplex *)dst;
211 int i, j, len8 = s->len4 >> 1, l_ptwo = 1 << s->ptwo_fft.nbits;
212 const float *in1 = src, *in2 = src + (s->len2 - 1) * stride;
213
214 /* Reindex input, putting it into a buffer and doing an Nx15 FFT */
215 for (i = 0; i < l_ptwo; i++) {
216 for (j = 0; j < 15; j++) {
217 const int k = s->pfa_prereindex[i*15 + j];
218 FFTComplex tmp = { in2[-k*stride], in1[k*stride] };
219 CMUL3(fft15in[j], tmp, s->twiddle_exptab[k >> 1]);
220 }
221 s->fft15(s->tmp + s->ptwo_fft.revtab[i], fft15in, s->exptab, l_ptwo);
222 }
223
224 /* Then a 15xN FFT (where N is a power of two) */
225 for (i = 0; i < 15; i++)
226 s->ptwo_fft.fft_calc(&s->ptwo_fft, s->tmp + l_ptwo*i);
227
228 /* Reindex again, apply twiddles and output */
229 s->postreindex(z, s->tmp, s->twiddle_exptab, s->pfa_postreindex, len8);
230 }
231
postrotate_c(FFTComplex * out,FFTComplex * in,FFTComplex * exp,int * lut,ptrdiff_t len8)232 static void postrotate_c(FFTComplex *out, FFTComplex *in, FFTComplex *exp,
233 int *lut, ptrdiff_t len8)
234 {
235 int i;
236
237 /* Reindex again, apply twiddles and output */
238 for (i = 0; i < len8; i++) {
239 const int i0 = len8 + i, i1 = len8 - i - 1;
240 const int s0 = lut[i0], s1 = lut[i1];
241
242 CMUL(out[i1].re, out[i0].im, in[s1].im, in[s1].re, exp[i1].im, exp[i1].re);
243 CMUL(out[i0].re, out[i1].im, in[s0].im, in[s0].re, exp[i0].im, exp[i0].re);
244 }
245 }
246
ff_mdct15_init(MDCT15Context ** ps,int inverse,int N,double scale)247 av_cold int ff_mdct15_init(MDCT15Context **ps, int inverse, int N, double scale)
248 {
249 MDCT15Context *s;
250 double alpha, theta;
251 int len2 = 15 * (1 << N);
252 int len = 2 * len2;
253 int i;
254
255 /* Tested and verified to work on everything in between */
256 if ((N < 2) || (N > 13))
257 return AVERROR(EINVAL);
258
259 s = av_mallocz(sizeof(*s));
260 if (!s)
261 return AVERROR(ENOMEM);
262
263 s->fft_n = N - 1;
264 s->len4 = len2 / 2;
265 s->len2 = len2;
266 s->inverse = inverse;
267 s->fft15 = fft15_c;
268 s->mdct = mdct15;
269 s->imdct_half = imdct15_half;
270 s->postreindex = postrotate_c;
271
272 if (ff_fft_init(&s->ptwo_fft, N - 1, s->inverse) < 0)
273 goto fail;
274
275 if (init_pfa_reindex_tabs(s))
276 goto fail;
277
278 s->tmp = av_malloc_array(len, 2 * sizeof(*s->tmp));
279 if (!s->tmp)
280 goto fail;
281
282 s->twiddle_exptab = av_malloc_array(s->len4, sizeof(*s->twiddle_exptab));
283 if (!s->twiddle_exptab)
284 goto fail;
285
286 theta = 0.125f + (scale < 0 ? s->len4 : 0);
287 scale = sqrt(fabs(scale));
288 for (i = 0; i < s->len4; i++) {
289 alpha = 2 * M_PI * (i + theta) / len;
290 s->twiddle_exptab[i].re = cosf(alpha) * scale;
291 s->twiddle_exptab[i].im = sinf(alpha) * scale;
292 }
293
294 /* 15-point FFT exptab */
295 for (i = 0; i < 19; i++) {
296 if (i < 15) {
297 double theta = (2.0f * M_PI * i) / 15.0f;
298 if (!s->inverse)
299 theta *= -1;
300 s->exptab[i].re = cosf(theta);
301 s->exptab[i].im = sinf(theta);
302 } else { /* Wrap around to simplify fft15 */
303 s->exptab[i] = s->exptab[i - 15];
304 }
305 }
306
307 /* 5-point FFT exptab */
308 s->exptab[19].re = cosf(2.0f * M_PI / 5.0f);
309 s->exptab[19].im = sinf(2.0f * M_PI / 5.0f);
310 s->exptab[20].re = cosf(1.0f * M_PI / 5.0f);
311 s->exptab[20].im = sinf(1.0f * M_PI / 5.0f);
312
313 /* Invert the phase for an inverse transform, do nothing for a forward transform */
314 if (s->inverse) {
315 s->exptab[19].im *= -1;
316 s->exptab[20].im *= -1;
317 }
318
319 if (ARCH_X86)
320 ff_mdct15_init_x86(s);
321
322 *ps = s;
323
324 return 0;
325
326 fail:
327 ff_mdct15_uninit(&s);
328 return AVERROR(ENOMEM);
329 }
330