1/*
2 * SIMD optimized non-power-of-two MDCT functions
3 *
4 * Copyright (C) 2017 Rostislav Pehlivanov <atomnuker@gmail.com>
5 *
6 * This file is part of FFmpeg.
7 *
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.
12 *
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.
17 *
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
21 */
22
23#include <string.h>
24
25#include "config.h"
26
27#include "libavutil/attributes.h"
28#include "libavutil/x86/cpu.h"
29#include "libavcodec/mdct15.h"
30
31void ff_mdct15_postreindex_sse3(FFTComplex *out, FFTComplex *in, FFTComplex *exp, int *lut, ptrdiff_t len8);
32void ff_mdct15_postreindex_avx2(FFTComplex *out, FFTComplex *in, FFTComplex *exp, int *lut, ptrdiff_t len8);
33
34void ff_fft15_avx(FFTComplex *out, FFTComplex *in, FFTComplex *exptab, ptrdiff_t stride);
35
36static void perm_twiddles(MDCT15Context *s)
37{
38    int k;
39    FFTComplex tmp[30];
40
41    /* 5-point FFT twiddles */
42    s->exptab[60].re = s->exptab[60].im = s->exptab[19].re;
43    s->exptab[61].re = s->exptab[61].im = s->exptab[19].im;
44    s->exptab[62].re = s->exptab[62].im = s->exptab[20].re;
45    s->exptab[63].re = s->exptab[63].im = s->exptab[20].im;
46
47    /* 15-point FFT twiddles */
48    for (k = 0; k < 5; k++) {
49        tmp[6*k + 0] = s->exptab[k +  0];
50        tmp[6*k + 2] = s->exptab[k +  5];
51        tmp[6*k + 4] = s->exptab[k + 10];
52
53        tmp[6*k + 1] = s->exptab[2 * (k + 0)];
54        tmp[6*k + 3] = s->exptab[2 * (k + 5)];
55        tmp[6*k + 5] = s->exptab[2 *  k + 5 ];
56    }
57
58    for (k = 0; k < 6; k++) {
59        FFTComplex ac_exp[] = {
60            { tmp[6*1 + k].re,  tmp[6*1 + k].re },
61            { tmp[6*2 + k].re,  tmp[6*2 + k].re },
62            { tmp[6*3 + k].re,  tmp[6*3 + k].re },
63            { tmp[6*4 + k].re,  tmp[6*4 + k].re },
64            { tmp[6*1 + k].im, -tmp[6*1 + k].im },
65            { tmp[6*2 + k].im, -tmp[6*2 + k].im },
66            { tmp[6*3 + k].im, -tmp[6*3 + k].im },
67            { tmp[6*4 + k].im, -tmp[6*4 + k].im },
68        };
69        memcpy(s->exptab + 8*k, ac_exp, 8*sizeof(FFTComplex));
70    }
71
72    /* Specialcase when k = 0 */
73    for (k = 0; k < 3; k++) {
74        FFTComplex dc_exp[] = {
75            { tmp[2*k + 0].re, -tmp[2*k + 0].im },
76            { tmp[2*k + 0].im,  tmp[2*k + 0].re },
77            { tmp[2*k + 1].re, -tmp[2*k + 1].im },
78            { tmp[2*k + 1].im,  tmp[2*k + 1].re },
79        };
80        memcpy(s->exptab + 8*6 + 4*k, dc_exp, 4*sizeof(FFTComplex));
81    }
82}
83
84av_cold void ff_mdct15_init_x86(MDCT15Context *s)
85{
86    int adjust_twiddles = 0;
87    int cpu_flags = av_get_cpu_flags();
88
89    if (EXTERNAL_SSE3(cpu_flags))
90        s->postreindex = ff_mdct15_postreindex_sse3;
91
92#if ARCH_X86_64
93    if (EXTERNAL_AVX(cpu_flags)) {
94        s->fft15 = ff_fft15_avx;
95        adjust_twiddles = 1;
96    }
97
98    if (EXTERNAL_AVX2_FAST(cpu_flags))
99        s->postreindex = ff_mdct15_postreindex_avx2;
100#endif
101
102    if (adjust_twiddles)
103        perm_twiddles(s);
104}
105