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#include "config.h"
20
21#include "libavutil/attributes.h"
22#include "libavutil/cpu.h"
23#include "libavutil/float_dsp.h"
24#include "cpu.h"
25#include "asm.h"
26
27void ff_vector_fmul_sse(float *dst, const float *src0, const float *src1,
28                        int len);
29void ff_vector_fmul_avx(float *dst, const float *src0, const float *src1,
30                        int len);
31
32void ff_vector_dmul_sse2(double *dst, const double *src0, const double *src1,
33                         int len);
34void ff_vector_dmul_avx(double *dst, const double *src0, const double *src1,
35                        int len);
36
37void ff_vector_fmac_scalar_sse(float *dst, const float *src, float mul,
38                               int len);
39void ff_vector_fmac_scalar_avx(float *dst, const float *src, float mul,
40                               int len);
41void ff_vector_fmac_scalar_fma3(float *dst, const float *src, float mul,
42                                int len);
43
44void ff_vector_fmul_scalar_sse(float *dst, const float *src, float mul,
45                               int len);
46
47void ff_vector_dmac_scalar_sse2(double *dst, const double *src, double mul,
48                                int len);
49void ff_vector_dmac_scalar_avx(double *dst, const double *src, double mul,
50                               int len);
51void ff_vector_dmac_scalar_fma3(double *dst, const double *src, double mul,
52                                int len);
53
54void ff_vector_dmul_scalar_sse2(double *dst, const double *src,
55                                double mul, int len);
56void ff_vector_dmul_scalar_avx(double *dst, const double *src,
57                               double mul, int len);
58
59void ff_vector_fmul_window_sse(float *dst, const float *src0,
60                               const float *src1, const float *win, int len);
61
62void ff_vector_fmul_add_sse(float *dst, const float *src0, const float *src1,
63                            const float *src2, int len);
64void ff_vector_fmul_add_avx(float *dst, const float *src0, const float *src1,
65                            const float *src2, int len);
66void ff_vector_fmul_add_fma3(float *dst, const float *src0, const float *src1,
67                             const float *src2, int len);
68
69void ff_vector_fmul_reverse_sse(float *dst, const float *src0,
70                                const float *src1, int len);
71void ff_vector_fmul_reverse_avx(float *dst, const float *src0,
72                                const float *src1, int len);
73void ff_vector_fmul_reverse_avx2(float *dst, const float *src0,
74                                 const float *src1, int len);
75
76float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order);
77
78void ff_butterflies_float_sse(float *av_restrict src0, float *av_restrict src1, int len);
79
80av_cold void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp)
81{
82    int cpu_flags = av_get_cpu_flags();
83
84    if (EXTERNAL_SSE(cpu_flags)) {
85        fdsp->vector_fmul = ff_vector_fmul_sse;
86        fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_sse;
87        fdsp->vector_fmul_scalar = ff_vector_fmul_scalar_sse;
88        fdsp->vector_fmul_window = ff_vector_fmul_window_sse;
89        fdsp->vector_fmul_add    = ff_vector_fmul_add_sse;
90        fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_sse;
91        fdsp->scalarproduct_float = ff_scalarproduct_float_sse;
92        fdsp->butterflies_float   = ff_butterflies_float_sse;
93    }
94    if (EXTERNAL_SSE2(cpu_flags)) {
95        fdsp->vector_dmul = ff_vector_dmul_sse2;
96        fdsp->vector_dmac_scalar = ff_vector_dmac_scalar_sse2;
97        fdsp->vector_dmul_scalar = ff_vector_dmul_scalar_sse2;
98    }
99    if (EXTERNAL_AVX_FAST(cpu_flags)) {
100        fdsp->vector_fmul = ff_vector_fmul_avx;
101        fdsp->vector_dmul = ff_vector_dmul_avx;
102        fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_avx;
103        fdsp->vector_dmul_scalar = ff_vector_dmul_scalar_avx;
104        fdsp->vector_dmac_scalar = ff_vector_dmac_scalar_avx;
105        fdsp->vector_fmul_add    = ff_vector_fmul_add_avx;
106        fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_avx;
107    }
108    if (EXTERNAL_AVX2_FAST(cpu_flags)) {
109        fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_avx2;
110    }
111    if (EXTERNAL_FMA3_FAST(cpu_flags)) {
112        fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_fma3;
113        fdsp->vector_fmul_add    = ff_vector_fmul_add_fma3;
114        fdsp->vector_dmac_scalar = ff_vector_dmac_scalar_fma3;
115    }
116}
117