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
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
|
/*
* 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_FASTCONV_IRESTORE_H_
#define PRIVATE_DSP_ARCH_X86_SSE_FASTCONV_IRESTORE_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
{
static inline void fastconv_restore_internal(float *dst, float *tmp, size_t rank)
{
// Prepare for butterflies
size_t last = size_t(1) << rank;
size_t items = last << 1;
size_t n = 8;
size_t bs = n << 1;
const float *wk = XFFT_W;
const float *ak = XFFT_A;
// Iterate butterflies
while (n < last)
{
for (size_t p=0; p<items; p += bs)
{
// Set initial values of pointers
float *a = &tmp[p];
float *b = &a[n];
size_t k = n;
ARCH_X86_ASM
(
__ASM_EMIT("movaps 0x00(%[ak]), %%xmm6") /* xmm6 = rak[i] */
__ASM_EMIT("movaps 0x10(%[ak]), %%xmm7") /* xmm7 = iak[i] */
:
: [ak] "r"(ak)
: "%xmm6", "%xmm7"
);
ARCH_X86_ASM
(
// __ASM_EMIT(".align 16")
__ASM_EMIT("1:")
__ASM_EMIT("movups 0x00(%[a]), %%xmm0") /* xmm0 = ra[i] */
__ASM_EMIT("movups 0x10(%[a]), %%xmm1") /* xmm1 = ia[i] */
__ASM_EMIT("movups 0x00(%[b]), %%xmm2") /* xmm2 = rb[i] */
__ASM_EMIT("movups 0x10(%[b]), %%xmm3") /* xmm3 = ib[i] */
__ASM_EMIT("movaps %%xmm2, %%xmm4") /* xmm4 = rb[i] */
__ASM_EMIT("movaps %%xmm3, %%xmm5") /* xmm5 = ib[i] */
__ASM_EMIT("mulps %%xmm6, %%xmm2") /* xmm2 = rak[i]*rb[i] */
__ASM_EMIT("mulps %%xmm7, %%xmm4") /* xmm4 = iak[i]*rb[i] */
__ASM_EMIT("mulps %%xmm6, %%xmm3") /* xmm3 = rak[i]*ib[i] */
__ASM_EMIT("mulps %%xmm7, %%xmm5") /* xmm5 = iak[i]*ib[i] */
__ASM_EMIT("addps %%xmm4, %%xmm3") /* xmm3 = rak[i]*ib[i] + iak[i]*rb[i] = ic[i] */
__ASM_EMIT("subps %%xmm5, %%xmm2") /* xmm2 = rak[i]*rb[i] - iak[i]*ib[i] = rc[i] */
__ASM_EMIT("movaps %%xmm0, %%xmm4") /* xmm4 = ra[i] */
__ASM_EMIT("movaps %%xmm1, %%xmm5") /* xmm5 = ia[i] */
__ASM_EMIT("subps %%xmm2, %%xmm0") /* xmm0 = ra[i] - rc[i] */
__ASM_EMIT("subps %%xmm3, %%xmm1") /* xmm1 = ia[i] - ic[i] */
__ASM_EMIT("addps %%xmm4, %%xmm2") /* xmm2 = ra[i] + rc[i] */
__ASM_EMIT("addps %%xmm5, %%xmm3") /* xmm3 = ia[i] + ic[i] */
__ASM_EMIT("movups %%xmm2, 0x00(%[a])")
__ASM_EMIT("movups %%xmm3, 0x10(%[a])")
__ASM_EMIT("movups %%xmm0, 0x00(%[b])")
__ASM_EMIT("movups %%xmm1, 0x10(%[b])")
__ASM_EMIT("add $0x20, %[a]")
__ASM_EMIT("add $0x20, %[b]")
__ASM_EMIT("sub $8, %[k]")
__ASM_EMIT("jz 2f")
/* Rotate angle */
__ASM_EMIT("movaps 0x00(%[wk]), %%xmm0") /* xmm0 = rw */
__ASM_EMIT("movaps 0x10(%[wk]), %%xmm1") /* xmm1 = iw */
__ASM_EMIT("movaps %%xmm0, %%xmm2") /* xmm2 = rw */
__ASM_EMIT("movaps %%xmm1, %%xmm3") /* xmm3 = iw */
__ASM_EMIT("mulps %%xmm6, %%xmm3") /* xmm3 = ra * iw */
__ASM_EMIT("mulps %%xmm7, %%xmm1") /* xmm1 = ia * iw */
__ASM_EMIT("mulps %%xmm0, %%xmm6") /* xmm6 = ra * rw */
__ASM_EMIT("mulps %%xmm2, %%xmm7") /* xmm7 = ia * rw */
__ASM_EMIT("subps %%xmm1, %%xmm6") /* xmm6 = ra * rw - ia * iw */
__ASM_EMIT("addps %%xmm3, %%xmm7") /* xmm7 = ra * iw + ia * rw */
__ASM_EMIT("jmp 1b")
__ASM_EMIT("2:")
: [a] "+r"(a), [b] "+r"(b), [k] "+r" (k)
: [wk] "r"(wk)
: "cc", "memory",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7"
);
}
wk += 8;
ak += 8;
n <<= 1;
bs <<= 1;
}
if (n < items)
{
// ONE LARGE CYCLE
// Set initial values of pointers
float kn = 1.0f / last;
size_t k = n;
ARCH_X86_ASM
(
__ASM_EMIT("movaps 0x00(%[ak]), %%xmm6") /* xmm6 = rak[i] */
__ASM_EMIT("movaps 0x10(%[ak]), %%xmm7") /* xmm7 = iak[i] */
__ASM_EMIT("shufps $0x00, %%xmm0, %%xmm0") /* xmm0 = kn */
__ASM_EMIT("movaps %%xmm0, %%xmm1") /* xmm1 = kn */
: "+Yz" (kn)
: [ak] "r" (ak)
: "%xmm1", "%xmm6", "%xmm7"
);
ARCH_X86_ASM
(
// __ASM_EMIT(".align 16")
__ASM_EMIT("1:")
__ASM_EMIT("movups 0x00(%[tmp]), %%xmm4") /* xmm4 = ra[i] */
__ASM_EMIT("movups 0x00(%[tmp],%[n],4), %%xmm2") /* xmm2 = rb[i] */
__ASM_EMIT("movups 0x10(%[tmp],%[n],4), %%xmm3") /* xmm3 = ib[i] */
__ASM_EMIT("mulps %%xmm6, %%xmm2") /* xmm2 = rak[i]*rb[i] */
__ASM_EMIT("mulps %%xmm7, %%xmm3") /* xmm3 = iak[i]*ib[i] */
__ASM_EMIT("movaps %%xmm4, %%xmm5") /* xmm5 = ra[i] */
__ASM_EMIT("subps %%xmm3, %%xmm2") /* xmm2 = rak[i]*rb[i] - iak[i]*ib[i] = rc[i] */
__ASM_EMIT("addps %%xmm2, %%xmm4") /* xmm4 = ra[i] + rc[i] */
__ASM_EMIT("subps %%xmm2, %%xmm5") /* xmm5 = ra[i] - rc[i] */
__ASM_EMIT("mulps %%xmm0, %%xmm4") /* xmm4 = kn*(ra[i] + rc[i]) */
__ASM_EMIT("mulps %%xmm1, %%xmm5") /* xmm5 = kn*(ra[i] - rc[i]) */
__ASM_EMIT("movups 0x00(%[dst]), %%xmm2")
__ASM_EMIT("movups 0x00(%[dst], %[n], 2), %%xmm3")
__ASM_EMIT("addps %%xmm4, %%xmm2")
__ASM_EMIT("addps %%xmm5, %%xmm3")
__ASM_EMIT("movups %%xmm2, 0x00(%[dst])")
__ASM_EMIT("movups %%xmm3, 0x00(%[dst],%[n],2)")
__ASM_EMIT("add $0x20, %[tmp]")
__ASM_EMIT("add $0x10, %[dst]")
__ASM_EMIT("sub $8, %[k]")
__ASM_EMIT("jz 2f")
/* Rotate angle */
__ASM_EMIT("movaps 0x00(%[wk]), %%xmm4") /* xmm4 = rw */
__ASM_EMIT("movaps 0x10(%[wk]), %%xmm5") /* xmm5 = iw */
__ASM_EMIT("movaps %%xmm4, %%xmm2") /* xmm2 = rw */
__ASM_EMIT("movaps %%xmm5, %%xmm3") /* xmm3 = iw */
__ASM_EMIT("mulps %%xmm6, %%xmm3") /* xmm3 = ra * iw */
__ASM_EMIT("mulps %%xmm7, %%xmm5") /* xmm5 = ia * iw */
__ASM_EMIT("mulps %%xmm4, %%xmm6") /* xmm6 = ra * rw */
__ASM_EMIT("mulps %%xmm2, %%xmm7") /* xmm7 = ia * rw */
__ASM_EMIT("subps %%xmm5, %%xmm6") /* xmm6 = ra * rw - ia * iw */
__ASM_EMIT("addps %%xmm3, %%xmm7") /* xmm7 = ra * iw + ia * rw */
__ASM_EMIT("jmp 1b")
__ASM_EMIT("2:")
: [tmp] "+r"(tmp), [dst] "+r"(dst), [n] "+r"(n), [k] "+r" (k)
: [wk] "r"(wk)
: "cc", "memory",
"%xmm0", "%xmm5", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7"
);
}
else
{
// Add real result to the target (ignore complex result)
float kn = 1.0f / last;
// Unpack 4x split complex to 4x real
ARCH_X86_ASM
(
__ASM_EMIT("shufps $0x00, %%xmm0, %%xmm0") /* xmm0 = kn */
__ASM_EMIT("movups 0x00(%[tmp]), %%xmm1") /* xmm0 = s[i] */
__ASM_EMIT("movups 0x00(%[dst]), %%xmm2")
__ASM_EMIT("mulps %%xmm0, %%xmm1")
__ASM_EMIT("addps %%xmm1, %%xmm2")
__ASM_EMIT("movups %%xmm2, 0x00(%[dst])")
: "+Yz" (kn)
: [tmp] "r" (tmp), [dst] "r" (dst)
: "cc", "memory",
"%xmm1", "%xmm2"
);
}
}
}
}
#endif /* PRIVATE_DSP_ARCH_X86_SSE_FASTCONV_IRESTORE_H_ */
|
