| 1 | /* |
| 2 | * Copyright (c) 2013-2014 Mozilla Corporation |
| 3 | * |
| 4 | * This file is part of FFmpeg. |
| 5 | * |
| 6 | * FFmpeg is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU Lesser General Public |
| 8 | * License as published by the Free Software Foundation; either |
| 9 | * version 2.1 of the License, or (at your option) any later version. |
| 10 | * |
| 11 | * FFmpeg is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 14 | * Lesser General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU Lesser General Public |
| 17 | * License along with FFmpeg; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 19 | */ |
| 20 | |
| 21 | /** |
| 22 | * @file |
| 23 | * Celt non-power of 2 iMDCT |
| 24 | */ |
| 25 | |
| 26 | #include <float.h> |
| 27 | #include <math.h> |
| 28 | #include <stddef.h> |
| 29 | |
| 30 | #include "config.h" |
| 31 | |
| 32 | #include "libavutil/attributes.h" |
| 33 | #include "libavutil/common.h" |
| 34 | |
| 35 | #include "avfft.h" |
| 36 | #include "opus.h" |
| 37 | #include "opus_imdct.h" |
| 38 | |
| 39 | // minimal iMDCT size to make SIMD opts easier |
| 40 | #define CELT_MIN_IMDCT_SIZE 120 |
| 41 | |
| 42 | // complex c = a * b |
| 43 | #define CMUL3(cre, cim, are, aim, bre, bim) \ |
| 44 | do { \ |
| 45 | cre = are * bre - aim * bim; \ |
| 46 | cim = are * bim + aim * bre; \ |
| 47 | } while (0) |
| 48 | |
| 49 | #define CMUL(c, a, b) CMUL3((c).re, (c).im, (a).re, (a).im, (b).re, (b).im) |
| 50 | |
| 51 | // complex c = a * b |
| 52 | // d = a * conjugate(b) |
| 53 | #define CMUL2(c, d, a, b) \ |
| 54 | do { \ |
| 55 | float are = (a).re; \ |
| 56 | float aim = (a).im; \ |
| 57 | float bre = (b).re; \ |
| 58 | float bim = (b).im; \ |
| 59 | float rr = are * bre; \ |
| 60 | float ri = are * bim; \ |
| 61 | float ir = aim * bre; \ |
| 62 | float ii = aim * bim; \ |
| 63 | (c).re = rr - ii; \ |
| 64 | (c).im = ri + ir; \ |
| 65 | (d).re = rr + ii; \ |
| 66 | (d).im = -ri + ir; \ |
| 67 | } while (0) |
| 68 | |
| 69 | av_cold void ff_celt_imdct_uninit(CeltIMDCTContext **ps) |
| 70 | { |
| 71 | CeltIMDCTContext *s = *ps; |
| 72 | int i; |
| 73 | |
| 74 | if (!s) |
| 75 | return; |
| 76 | |
| 77 | for (i = 0; i < FF_ARRAY_ELEMS(s->exptab); i++) |
| 78 | av_freep(&s->exptab[i]); |
| 79 | |
| 80 | av_freep(&s->twiddle_exptab); |
| 81 | |
| 82 | av_freep(&s->tmp); |
| 83 | |
| 84 | av_freep(ps); |
| 85 | } |
| 86 | |
| 87 | static void celt_imdct_half(CeltIMDCTContext *s, float *dst, const float *src, |
| 88 | ptrdiff_t stride, float scale); |
| 89 | |
| 90 | av_cold int ff_celt_imdct_init(CeltIMDCTContext **ps, int N) |
| 91 | { |
| 92 | CeltIMDCTContext *s; |
| 93 | int len2 = 15 * (1 << N); |
| 94 | int len = 2 * len2; |
| 95 | int i, j; |
| 96 | |
| 97 | if (len2 > CELT_MAX_FRAME_SIZE || len2 < CELT_MIN_IMDCT_SIZE) |
| 98 | return AVERROR(EINVAL); |
| 99 | |
| 100 | s = av_mallocz(sizeof(*s)); |
| 101 | if (!s) |
| 102 | return AVERROR(ENOMEM); |
| 103 | |
| 104 | s->fft_n = N - 1; |
| 105 | s->len4 = len2 / 2; |
| 106 | s->len2 = len2; |
| 107 | |
| 108 | s->tmp = av_malloc(len * 2 * sizeof(*s->tmp)); |
| 109 | if (!s->tmp) |
| 110 | goto fail; |
| 111 | |
| 112 | s->twiddle_exptab = av_malloc(s->len4 * sizeof(*s->twiddle_exptab)); |
| 113 | if (!s->twiddle_exptab) |
| 114 | goto fail; |
| 115 | |
| 116 | for (i = 0; i < s->len4; i++) { |
| 117 | s->twiddle_exptab[i].re = cos(2 * M_PI * (i + 0.125 + s->len4) / len); |
| 118 | s->twiddle_exptab[i].im = sin(2 * M_PI * (i + 0.125 + s->len4) / len); |
| 119 | } |
| 120 | |
| 121 | for (i = 0; i < FF_ARRAY_ELEMS(s->exptab); i++) { |
| 122 | int N = 15 * (1 << i); |
| 123 | s->exptab[i] = av_malloc(sizeof(*s->exptab[i]) * FFMAX(N, 19)); |
| 124 | if (!s->exptab[i]) |
| 125 | goto fail; |
| 126 | |
| 127 | for (j = 0; j < N; j++) { |
| 128 | s->exptab[i][j].re = cos(2 * M_PI * j / N); |
| 129 | s->exptab[i][j].im = sin(2 * M_PI * j / N); |
| 130 | } |
| 131 | } |
| 132 | |
| 133 | // wrap around to simplify fft15 |
| 134 | for (j = 15; j < 19; j++) |
| 135 | s->exptab[0][j] = s->exptab[0][j - 15]; |
| 136 | |
| 137 | s->imdct_half = celt_imdct_half; |
| 138 | |
| 139 | if (ARCH_AARCH64) |
| 140 | ff_celt_imdct_init_aarch64(s); |
| 141 | |
| 142 | *ps = s; |
| 143 | |
| 144 | return 0; |
| 145 | fail: |
| 146 | ff_celt_imdct_uninit(&s); |
| 147 | return AVERROR(ENOMEM); |
| 148 | } |
| 149 | |
| 150 | static void fft5(FFTComplex *out, const FFTComplex *in, ptrdiff_t stride) |
| 151 | { |
| 152 | // [0] = exp(2 * i * pi / 5), [1] = exp(2 * i * pi * 2 / 5) |
| 153 | static const FFTComplex fact[] = { { 0.30901699437494745, 0.95105651629515353 }, |
| 154 | { -0.80901699437494734, 0.58778525229247325 } }; |
| 155 | |
| 156 | FFTComplex z[4][4]; |
| 157 | |
| 158 | CMUL2(z[0][0], z[0][3], in[1 * stride], fact[0]); |
| 159 | CMUL2(z[0][1], z[0][2], in[1 * stride], fact[1]); |
| 160 | CMUL2(z[1][0], z[1][3], in[2 * stride], fact[0]); |
| 161 | CMUL2(z[1][1], z[1][2], in[2 * stride], fact[1]); |
| 162 | CMUL2(z[2][0], z[2][3], in[3 * stride], fact[0]); |
| 163 | CMUL2(z[2][1], z[2][2], in[3 * stride], fact[1]); |
| 164 | CMUL2(z[3][0], z[3][3], in[4 * stride], fact[0]); |
| 165 | CMUL2(z[3][1], z[3][2], in[4 * stride], fact[1]); |
| 166 | |
| 167 | out[0].re = in[0].re + in[stride].re + in[2 * stride].re + in[3 * stride].re + in[4 * stride].re; |
| 168 | out[0].im = in[0].im + in[stride].im + in[2 * stride].im + in[3 * stride].im + in[4 * stride].im; |
| 169 | |
| 170 | out[1].re = in[0].re + z[0][0].re + z[1][1].re + z[2][2].re + z[3][3].re; |
| 171 | out[1].im = in[0].im + z[0][0].im + z[1][1].im + z[2][2].im + z[3][3].im; |
| 172 | |
| 173 | out[2].re = in[0].re + z[0][1].re + z[1][3].re + z[2][0].re + z[3][2].re; |
| 174 | out[2].im = in[0].im + z[0][1].im + z[1][3].im + z[2][0].im + z[3][2].im; |
| 175 | |
| 176 | out[3].re = in[0].re + z[0][2].re + z[1][0].re + z[2][3].re + z[3][1].re; |
| 177 | out[3].im = in[0].im + z[0][2].im + z[1][0].im + z[2][3].im + z[3][1].im; |
| 178 | |
| 179 | out[4].re = in[0].re + z[0][3].re + z[1][2].re + z[2][1].re + z[3][0].re; |
| 180 | out[4].im = in[0].im + z[0][3].im + z[1][2].im + z[2][1].im + z[3][0].im; |
| 181 | } |
| 182 | |
| 183 | static void fft15(CeltIMDCTContext *s, FFTComplex *out, const FFTComplex *in, ptrdiff_t stride) |
| 184 | { |
| 185 | const FFTComplex *exptab = s->exptab[0]; |
| 186 | FFTComplex tmp[5]; |
| 187 | FFTComplex tmp1[5]; |
| 188 | FFTComplex tmp2[5]; |
| 189 | int k; |
| 190 | |
| 191 | fft5(tmp, in, stride * 3); |
| 192 | fft5(tmp1, in + stride, stride * 3); |
| 193 | fft5(tmp2, in + 2 * stride, stride * 3); |
| 194 | |
| 195 | for (k = 0; k < 5; k++) { |
| 196 | FFTComplex t1, t2; |
| 197 | |
| 198 | CMUL(t1, tmp1[k], exptab[k]); |
| 199 | CMUL(t2, tmp2[k], exptab[2 * k]); |
| 200 | out[k].re = tmp[k].re + t1.re + t2.re; |
| 201 | out[k].im = tmp[k].im + t1.im + t2.im; |
| 202 | |
| 203 | CMUL(t1, tmp1[k], exptab[k + 5]); |
| 204 | CMUL(t2, tmp2[k], exptab[2 * (k + 5)]); |
| 205 | out[k + 5].re = tmp[k].re + t1.re + t2.re; |
| 206 | out[k + 5].im = tmp[k].im + t1.im + t2.im; |
| 207 | |
| 208 | CMUL(t1, tmp1[k], exptab[k + 10]); |
| 209 | CMUL(t2, tmp2[k], exptab[2 * k + 5]); |
| 210 | out[k + 10].re = tmp[k].re + t1.re + t2.re; |
| 211 | out[k + 10].im = tmp[k].im + t1.im + t2.im; |
| 212 | } |
| 213 | } |
| 214 | |
| 215 | /* |
| 216 | * FFT of the length 15 * (2^N) |
| 217 | */ |
| 218 | static void fft_calc(CeltIMDCTContext *s, FFTComplex *out, const FFTComplex *in, |
| 219 | int N, ptrdiff_t stride) |
| 220 | { |
| 221 | if (N) { |
| 222 | const FFTComplex *exptab = s->exptab[N]; |
| 223 | const int len2 = 15 * (1 << (N - 1)); |
| 224 | int k; |
| 225 | |
| 226 | fft_calc(s, out, in, N - 1, stride * 2); |
| 227 | fft_calc(s, out + len2, in + stride, N - 1, stride * 2); |
| 228 | |
| 229 | for (k = 0; k < len2; k++) { |
| 230 | FFTComplex t; |
| 231 | |
| 232 | CMUL(t, out[len2 + k], exptab[k]); |
| 233 | |
| 234 | out[len2 + k].re = out[k].re - t.re; |
| 235 | out[len2 + k].im = out[k].im - t.im; |
| 236 | |
| 237 | out[k].re += t.re; |
| 238 | out[k].im += t.im; |
| 239 | } |
| 240 | } else |
| 241 | fft15(s, out, in, stride); |
| 242 | } |
| 243 | |
| 244 | static void celt_imdct_half(CeltIMDCTContext *s, float *dst, const float *src, |
| 245 | ptrdiff_t stride, float scale) |
| 246 | { |
| 247 | FFTComplex *z = (FFTComplex *)dst; |
| 248 | const int len8 = s->len4 / 2; |
| 249 | const float *in1 = src; |
| 250 | const float *in2 = src + (s->len2 - 1) * stride; |
| 251 | int i; |
| 252 | |
| 253 | for (i = 0; i < s->len4; i++) { |
| 254 | FFTComplex tmp = { *in2, *in1 }; |
| 255 | CMUL(s->tmp[i], tmp, s->twiddle_exptab[i]); |
| 256 | in1 += 2 * stride; |
| 257 | in2 -= 2 * stride; |
| 258 | } |
| 259 | |
| 260 | fft_calc(s, z, s->tmp, s->fft_n, 1); |
| 261 | |
| 262 | for (i = 0; i < len8; i++) { |
| 263 | float r0, i0, r1, i1; |
| 264 | |
| 265 | CMUL3(r0, i1, z[len8 - i - 1].im, z[len8 - i - 1].re, s->twiddle_exptab[len8 - i - 1].im, s->twiddle_exptab[len8 - i - 1].re); |
| 266 | CMUL3(r1, i0, z[len8 + i].im, z[len8 + i].re, s->twiddle_exptab[len8 + i].im, s->twiddle_exptab[len8 + i].re); |
| 267 | z[len8 - i - 1].re = scale * r0; |
| 268 | z[len8 - i - 1].im = scale * i0; |
| 269 | z[len8 + i].re = scale * r1; |
| 270 | z[len8 + i].im = scale * i1; |
| 271 | } |
| 272 | } |