2 * SIMD optimized non-power-of-two MDCT functions
4 * Copyright (C) 2017 Rostislav Pehlivanov <atomnuker@gmail.com>
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 #include "libavutil/attributes.h"
26 #include "libavutil/x86/cpu.h"
27 #include "libavcodec/mdct15.h"
29 void ff_mdct15_postreindex_sse3(FFTComplex *out, FFTComplex *in, FFTComplex *exp, int *lut, ptrdiff_t len8);
30 void ff_mdct15_postreindex_avx2(FFTComplex *out, FFTComplex *in, FFTComplex *exp, int *lut, ptrdiff_t len8);
32 void ff_fft15_avx(FFTComplex *out, FFTComplex *in, FFTComplex *exptab, ptrdiff_t stride);
34 static void perm_twiddles(MDCT15Context *s)
39 /* 5-point FFT twiddles */
40 s->exptab[60].re = s->exptab[60].im = s->exptab[19].re;
41 s->exptab[61].re = s->exptab[61].im = s->exptab[19].im;
42 s->exptab[62].re = s->exptab[62].im = s->exptab[20].re;
43 s->exptab[63].re = s->exptab[63].im = s->exptab[20].im;
45 /* 15-point FFT twiddles */
46 for (k = 0; k < 5; k++) {
47 tmp[6*k + 0] = s->exptab[k + 0];
48 tmp[6*k + 2] = s->exptab[k + 5];
49 tmp[6*k + 4] = s->exptab[k + 10];
51 tmp[6*k + 1] = s->exptab[2 * (k + 0)];
52 tmp[6*k + 3] = s->exptab[2 * (k + 5)];
53 tmp[6*k + 5] = s->exptab[2 * k + 5 ];
56 for (k = 0; k < 6; k++) {
57 FFTComplex ac_exp[] = {
58 { tmp[6*1 + k].re, tmp[6*1 + k].re },
59 { tmp[6*2 + k].re, tmp[6*2 + k].re },
60 { tmp[6*3 + k].re, tmp[6*3 + k].re },
61 { tmp[6*4 + k].re, tmp[6*4 + k].re },
62 { tmp[6*1 + k].im, -tmp[6*1 + k].im },
63 { tmp[6*2 + k].im, -tmp[6*2 + k].im },
64 { tmp[6*3 + k].im, -tmp[6*3 + k].im },
65 { tmp[6*4 + k].im, -tmp[6*4 + k].im },
67 memcpy(s->exptab + 8*k, ac_exp, 8*sizeof(FFTComplex));
70 /* Specialcase when k = 0 */
71 for (k = 0; k < 3; k++) {
72 FFTComplex dc_exp[] = {
73 { tmp[2*k + 0].re, -tmp[2*k + 0].im },
74 { tmp[2*k + 0].im, tmp[2*k + 0].re },
75 { tmp[2*k + 1].re, -tmp[2*k + 1].im },
76 { tmp[2*k + 1].im, tmp[2*k + 1].re },
78 memcpy(s->exptab + 8*6 + 4*k, dc_exp, 4*sizeof(FFTComplex));
82 av_cold void ff_mdct15_init_x86(MDCT15Context *s)
84 int adjust_twiddles = 0;
85 int cpu_flags = av_get_cpu_flags();
87 if (EXTERNAL_SSE3(cpu_flags))
88 s->postreindex = ff_mdct15_postreindex_sse3;
90 if (ARCH_X86_64 && EXTERNAL_AVX(cpu_flags)) {
91 s->fft15 = ff_fft15_avx;
95 if (ARCH_X86_64 && EXTERNAL_AVX2_FAST(cpu_flags))
96 s->postreindex = ff_mdct15_postreindex_avx2;