Merge commit 'fd9212f2edfe9b107c3c08ba2df5fd2cba5ab9e3'
[ffmpeg.git] / libavcodec / x86 / synth_filter_init.c
1 /*
2  * Copyright (c) 2012-2014 Christophe Gisquet <christophe.gisquet@gmail.com>
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 #include "libavutil/attributes.h"
22 #include "libavutil/cpu.h"
23 #include "libavutil/x86/cpu.h"
24 #include "libavcodec/synth_filter.h"
25
26 #define SYNTH_FILTER_FUNC(opt)                                                 \
27 void ff_synth_filter_inner_##opt(float *synth_buf_ptr, float synth_buf2[32],   \
28                                  const float window[512],                      \
29                                  float out[32], intptr_t offset, float scale); \
30 static void synth_filter_##opt(FFTContext *imdct,                              \
31                                float *synth_buf_ptr, int *synth_buf_offset,    \
32                                float synth_buf2[32], const float window[512],  \
33                                float out[32], const float in[32], float scale) \
34 {                                                                              \
35     float *synth_buf= synth_buf_ptr + *synth_buf_offset;                       \
36                                                                                \
37     imdct->imdct_half(imdct, synth_buf, in);                                   \
38                                                                                \
39     ff_synth_filter_inner_##opt(synth_buf, synth_buf2, window,                 \
40                                 out, *synth_buf_offset, scale);                \
41                                                                                \
42     *synth_buf_offset = (*synth_buf_offset - 32) & 511;                        \
43 }                                                                              \
44
45 #if HAVE_X86ASM
46 #if ARCH_X86_32
47 SYNTH_FILTER_FUNC(sse)
48 #endif
49 SYNTH_FILTER_FUNC(sse2)
50 SYNTH_FILTER_FUNC(avx)
51 SYNTH_FILTER_FUNC(fma3)
52 #endif /* HAVE_X86ASM */
53
54 av_cold void ff_synth_filter_init_x86(SynthFilterContext *s)
55 {
56 #if HAVE_X86ASM
57     int cpu_flags = av_get_cpu_flags();
58
59 #if ARCH_X86_32
60     if (EXTERNAL_SSE(cpu_flags)) {
61         s->synth_filter_float = synth_filter_sse;
62     }
63 #endif
64     if (EXTERNAL_SSE2(cpu_flags)) {
65         s->synth_filter_float = synth_filter_sse2;
66     }
67     if (EXTERNAL_AVX_FAST(cpu_flags)) {
68         s->synth_filter_float = synth_filter_avx;
69     }
70     if (EXTERNAL_FMA3_FAST(cpu_flags)) {
71         s->synth_filter_float = synth_filter_fma3;
72     }
73 #endif /* HAVE_X86ASM */
74 }