1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
|
/*
* Copyright (C) 2020 Linux Studio Plugins Project <https://lsp-plug.in/>
* (C) 2020 Vladimir Sadovnikov <sadko4u@gmail.com>
*
* This file is part of lsp-dsp-lib
* Created on: 31 мар. 2020 г.
*
* lsp-dsp-lib is free software: you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* any later version.
*
* lsp-dsp-lib is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with lsp-dsp-lib. If not, see <https://www.gnu.org/licenses/>.
*/
#ifndef PRIVATE_DSP_ARCH_X86_SSE_FFT_NORMALIZE_H_
#define PRIVATE_DSP_ARCH_X86_SSE_FFT_NORMALIZE_H_
#ifndef PRIVATE_DSP_ARCH_X86_SSE_IMPL
#error "This header should not be included directly"
#endif /* PRIVATE_DSP_ARCH_X86_SSE_IMPL */
namespace lsp
{
namespace sse
{
void normalize_fft3(float *dre, float *dim, const float *re, const float *im, size_t rank)
{
IF_ARCH_X86(
float k = 1.0f/(1 << rank);
size_t count = 1 << rank, off = 0;
);
ARCH_X86_ASM(
// x8 blocks
__ASM_EMIT ("shufps $0x00, %%xmm0, %%xmm0") // xmm0 = k
__ASM_EMIT32("subl $8, %[count]")
__ASM_EMIT64("sub $8, %[count]")
__ASM_EMIT ("movaps %%xmm0, %%xmm1")
__ASM_EMIT ("jb 2f")
__ASM_EMIT ("1:")
__ASM_EMIT ("movups 0x00(%[s_re], %[off]), %%xmm4")
__ASM_EMIT ("movups 0x10(%[s_re], %[off]), %%xmm5")
__ASM_EMIT ("movups 0x00(%[s_im], %[off]), %%xmm6")
__ASM_EMIT ("movups 0x10(%[s_im], %[off]), %%xmm7")
__ASM_EMIT ("mulps %%xmm0, %%xmm4")
__ASM_EMIT ("mulps %%xmm1, %%xmm5")
__ASM_EMIT ("mulps %%xmm0, %%xmm6")
__ASM_EMIT ("mulps %%xmm1, %%xmm7")
__ASM_EMIT ("movups %%xmm4, 0x00(%[d_re], %[off])")
__ASM_EMIT ("movups %%xmm5, 0x10(%[d_re], %[off])")
__ASM_EMIT ("movups %%xmm6, 0x00(%[d_im], %[off])")
__ASM_EMIT ("movups %%xmm7, 0x10(%[d_im], %[off])")
__ASM_EMIT ("add $0x20, %[off]")
__ASM_EMIT32("subl $8, %[count]")
__ASM_EMIT64("sub $8, %[count]")
__ASM_EMIT ("jae 1b")
__ASM_EMIT ("2:")
: [off] "+r" (off), [count] __ASM_ARG_RW(count),
[k] "+Yz" (k)
: [s_re] "r" (re), [s_im] "r" (im),
[d_re] "r" (dre), [d_im] "r" (dim)
: "cc", "memory",
"%xmm1",
"%xmm4", "%xmm5", "%xmm6", "%xmm7"
);
}
void normalize_fft2(float *re, float *im, size_t rank)
{
IF_ARCH_X86(
float k = 1.0f/(1 << rank);
size_t count = 1 << rank, off = 0;
);
ARCH_X86_ASM(
// x8 blocks
__ASM_EMIT ("shufps $0x00, %%xmm0, %%xmm0") // xmm0 = k
__ASM_EMIT ("sub $8, %[count]")
__ASM_EMIT ("movaps %%xmm0, %%xmm1")
__ASM_EMIT ("jb 2f")
__ASM_EMIT ("1:")
__ASM_EMIT ("movups 0x00(%[d_re], %[off]), %%xmm4")
__ASM_EMIT ("movups 0x10(%[d_re], %[off]), %%xmm5")
__ASM_EMIT ("movups 0x00(%[d_im], %[off]), %%xmm6")
__ASM_EMIT ("movups 0x10(%[d_im], %[off]), %%xmm7")
__ASM_EMIT ("mulps %%xmm0, %%xmm4")
__ASM_EMIT ("mulps %%xmm1, %%xmm5")
__ASM_EMIT ("mulps %%xmm0, %%xmm6")
__ASM_EMIT ("mulps %%xmm1, %%xmm7")
__ASM_EMIT ("movups %%xmm4, 0x00(%[d_re], %[off])")
__ASM_EMIT ("movups %%xmm5, 0x10(%[d_re], %[off])")
__ASM_EMIT ("movups %%xmm6, 0x00(%[d_im], %[off])")
__ASM_EMIT ("movups %%xmm7, 0x10(%[d_im], %[off])")
__ASM_EMIT ("add $0x20, %[off]")
__ASM_EMIT ("sub $8, %[count]")
__ASM_EMIT ("jae 1b")
__ASM_EMIT ("2:")
: [off] "+r" (off), [count] "+r" (count),
[k] "+Yz" (k)
: [d_re] "r" (re), [d_im] "r" (im)
: "cc", "memory",
"%xmm1",
"%xmm4", "%xmm5", "%xmm6", "%xmm7"
);
}
}
}
#endif /* PRIVATE_DSP_ARCH_X86_SSE_FFT_NORMALIZE_H_ */
|
