1/*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19#define TX_FLOAT
20#include "libavutil/tx_priv.h"
21#include "libavutil/attributes.h"
22#include "libavutil/x86/cpu.h"
23
24#include "config.h"
25
26TX_DECL_FN(fft2,      sse3)
27TX_DECL_FN(fft4_fwd,  sse2)
28TX_DECL_FN(fft4_inv,  sse2)
29TX_DECL_FN(fft8,      sse3)
30TX_DECL_FN(fft8_ns,   sse3)
31TX_DECL_FN(fft8,      avx)
32TX_DECL_FN(fft8_ns,   avx)
33TX_DECL_FN(fft16,     avx)
34TX_DECL_FN(fft16_ns,  avx)
35TX_DECL_FN(fft16,     fma3)
36TX_DECL_FN(fft16_ns,  fma3)
37TX_DECL_FN(fft32,     avx)
38TX_DECL_FN(fft32_ns,  avx)
39TX_DECL_FN(fft32,     fma3)
40TX_DECL_FN(fft32_ns,  fma3)
41TX_DECL_FN(fft_sr,    fma3)
42TX_DECL_FN(fft_sr_ns, fma3)
43TX_DECL_FN(fft_sr,    avx2)
44TX_DECL_FN(fft_sr_ns, avx2)
45
46#define DECL_INIT_FN(basis, interleave)                                        \
47static av_cold int b ##basis## _i ##interleave(AVTXContext *s,                 \
48                                               const FFTXCodelet *cd,          \
49                                               uint64_t flags,                 \
50                                               FFTXCodeletOptions *opts,       \
51                                               int len, int inv,               \
52                                               const void *scale)              \
53{                                                                              \
54    const int inv_lookup = opts ? opts->invert_lookup : 1;                     \
55    ff_tx_init_tabs_float(len);                                                \
56    if (cd->max_len == 2)                                                      \
57        return ff_tx_gen_ptwo_revtab(s, inv_lookup);                           \
58    else                                                                       \
59        return ff_tx_gen_split_radix_parity_revtab(s, inv_lookup,              \
60                                                   basis, interleave);         \
61}
62
63DECL_INIT_FN(8, 0)
64DECL_INIT_FN(8, 2)
65
66const FFTXCodelet * const ff_tx_codelet_list_float_x86[] = {
67    TX_DEF(fft2,     FFT,  2,  2, 2, 0, 128, NULL,  sse3, SSE3, AV_TX_INPLACE, 0),
68    TX_DEF(fft2,     FFT,  2,  2, 2, 0, 192, b8_i0, sse3, SSE3, AV_TX_INPLACE | FF_TX_PRESHUFFLE, 0),
69    TX_DEF(fft4_fwd, FFT,  4,  4, 2, 0, 128, NULL,  sse2, SSE2, AV_TX_INPLACE | FF_TX_FORWARD_ONLY, 0),
70    TX_DEF(fft4_fwd, FFT,  4,  4, 2, 0, 192, b8_i0, sse2, SSE2, AV_TX_INPLACE | FF_TX_PRESHUFFLE, 0),
71    TX_DEF(fft4_inv, FFT,  4,  4, 2, 0, 128, NULL,  sse2, SSE2, AV_TX_INPLACE | FF_TX_INVERSE_ONLY, 0),
72    TX_DEF(fft8,     FFT,  8,  8, 2, 0, 128, b8_i0, sse3, SSE3, AV_TX_INPLACE, 0),
73    TX_DEF(fft8_ns,  FFT,  8,  8, 2, 0, 192, b8_i0, sse3, SSE3, AV_TX_INPLACE | FF_TX_PRESHUFFLE, 0),
74    TX_DEF(fft8,     FFT,  8,  8, 2, 0, 256, b8_i0, avx,  AVX,  AV_TX_INPLACE, AV_CPU_FLAG_AVXSLOW),
75    TX_DEF(fft8_ns,  FFT,  8,  8, 2, 0, 320, b8_i0, avx,  AVX,  AV_TX_INPLACE | FF_TX_PRESHUFFLE,
76           AV_CPU_FLAG_AVXSLOW),
77    TX_DEF(fft16,    FFT, 16, 16, 2, 0, 256, b8_i2, avx,  AVX,  AV_TX_INPLACE, AV_CPU_FLAG_AVXSLOW),
78    TX_DEF(fft16_ns, FFT, 16, 16, 2, 0, 320, b8_i2, avx,  AVX,  AV_TX_INPLACE | FF_TX_PRESHUFFLE,
79           AV_CPU_FLAG_AVXSLOW),
80    TX_DEF(fft16,    FFT, 16, 16, 2, 0, 288, b8_i2, fma3, FMA3, AV_TX_INPLACE, AV_CPU_FLAG_AVXSLOW),
81    TX_DEF(fft16_ns, FFT, 16, 16, 2, 0, 352, b8_i2, fma3, FMA3, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
82           AV_CPU_FLAG_AVXSLOW),
83
84#if ARCH_X86_64
85    TX_DEF(fft32,    FFT, 32, 32, 2, 0, 256, b8_i2, avx,  AVX,  AV_TX_INPLACE, AV_CPU_FLAG_AVXSLOW),
86    TX_DEF(fft32_ns, FFT, 32, 32, 2, 0, 320, b8_i2, avx,  AVX,  AV_TX_INPLACE | FF_TX_PRESHUFFLE,
87           AV_CPU_FLAG_AVXSLOW),
88    TX_DEF(fft32,    FFT, 32, 32, 2, 0, 288, b8_i2, fma3, FMA3, AV_TX_INPLACE, AV_CPU_FLAG_AVXSLOW),
89    TX_DEF(fft32_ns, FFT, 32, 32, 2, 0, 352, b8_i2, fma3, FMA3, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
90           AV_CPU_FLAG_AVXSLOW),
91    TX_DEF(fft_sr,    FFT, 64, 131072, 2, 0, 288, b8_i2, fma3,  FMA3,  0, AV_CPU_FLAG_AVXSLOW),
92    TX_DEF(fft_sr_ns, FFT, 64, 131072, 2, 0, 352, b8_i2, fma3,  FMA3,  AV_TX_INPLACE | FF_TX_PRESHUFFLE,
93           AV_CPU_FLAG_AVXSLOW),
94#if HAVE_AVX2_EXTERNAL
95    TX_DEF(fft_sr,    FFT, 64, 131072, 2, 0, 320, b8_i2, avx2, AVX2, 0,
96           AV_CPU_FLAG_AVXSLOW | AV_CPU_FLAG_SLOW_GATHER),
97    TX_DEF(fft_sr_ns, FFT, 64, 131072, 2, 0, 384, b8_i2, avx2, AVX2, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
98           AV_CPU_FLAG_AVXSLOW | AV_CPU_FLAG_SLOW_GATHER),
99#endif
100#endif
101
102    NULL,
103};
104