-
Notifications
You must be signed in to change notification settings - Fork 18
/
cufft.lua
151 lines (131 loc) · 4.13 KB
/
cufft.lua
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
-- Lua wrapper for cufft functions
require 'libspectralnet'
cufft = cufft or {}
function cufft.fft1d(input, output)
local nSamples = input:size(1)
local nPlanes = input:size(2)
local N = input:size(3)
local M = input:size(4)
input:resize(nSamples*nPlanes*N, M)
output:resize(nSamples*nPlanes*N, M/2+1, 2)
libspectralnet.fft1d_r2c(input, output)
input:resize(nSamples, nPlanes, N, M)
output:resize(nSamples, nPlanes, N, M/2+1, 2)
end
function cufft.ifft1d(input, output)
local nSamples = output:size(1)
local nPlanes = output:size(2)
local N = output:size(3)
local M = output:size(4)
input:resize(nSamples*nPlanes*N, M/2+1, 2)
output:resize(nSamples*nPlanes*N, M)
libspectralnet.fft1d_c2r(input,output)
output:div(M)
input:resize(nSamples, nPlanes, N, M/2+1, 2)
output:resize(nSamples, nPlanes, N, M)
end
function cufft.fft2d_r2c(input,output,debug)
local debug = debug or false
local nSamples = input:size(1)
local nPlanes = input:size(2)
local N = input:size(3)
local M = input:size(4)
input:resize(nSamples*nPlanes, N, M)
if debug then
output:resize(nSamples*nPlanes, N, M, 2)
local dft1 = cufft.dftmatrix(N,1)
local dft2 = cufft.dftmatrix(M,1)
for i = 1,output:size(1) do
local input2 = torch.zeros(N,M,2)
input2:select(3,1):copy(input[i])
input2 = complex.mm(input2,dft2)
input2 = complex.mm(dft1,input2)
output[i]:copy(input2)
end
output:resize(nSamples, nPlanes, N, M, 2)
else
output:resize(nSamples*nPlanes, N, M/2+1, 2)
libspectralnet.fft2d_r2c(input, output)
output:resize(nSamples, nPlanes, N, M/2+1, 2)
end
input:resize(nSamples, nPlanes, N, M)
--print(input:isContiguous())
end
function cufft.fft2d_c2r(input, output, debug)
local debug = debug or false
local nSamples = output:size(1)
local nPlanes = output:size(2)
local N = output:size(3)
local M = output:size(4)
output:resize(nSamples*nPlanes, N, M)
--print(input:isContiguous())
if debug then
input:resize(nSamples*nPlanes, N, M, 2)
local dft1 = cufft.dftmatrix(N,-1)
local dft2 = cufft.dftmatrix(M,-1)
for i = 1,input:size(1) do
local input2 = torch.zeros(N,M,2)
input2:select(3,1):copy(input[i])
input2 = complex.mm(input2,dft2)
input2 = complex.mm(dft1,input2)
output[i]:copy(input2)
end
input:resize(nSamples,nPlanes, N, M, 2)
else
input:resize(nSamples*nPlanes, N, M/2+1, 2)
libspectralnet.fft2d_c2r(input,output)
input:resize(nSamples, nPlanes, N, M/2+1, 2)
end
output:div(M*N)
output:resize(nSamples, nPlanes, N, M)
end
function cufft.fft2d_c2c(input,output,dir,debug)
local dir = dir or 1
local debug = debug or false
local nSamples = output:size(1)
local nPlanes = output:size(2)
local N = output:size(3)
local M = output:size(4)
input:resize(nSamples*nPlanes, N, M, 2)
output:resize(nSamples*nPlanes, N, M, 2)
if debug then
print('warning, slow')
local dft1 = cufft.dftmatrix(N,dir)
local dft2 = cufft.dftmatrix(M,dir)
for i = 1,input:size(1) do
local input2 = input[i]:clone():double()
input2 = complex.mm(input2,dft2)
input2 = complex.mm(dft1,input2)
output[i]:copy(input2)
end
else
libspectralnet.fft2d_c2c(input,output,dir)
end
input:resize(nSamples, nPlanes, N, M, 2)
if dir == -1 then
output:div(M*N)
end
output:resize(nSamples, nPlanes, N, M, 2)
end
function cufft.fft2dsingle(input,dir)
local dir = dir or 1
local dft1 = cufft.dftmatrix(input:size(1),dir)
local dft2 = cufft.dftmatrix(input:size(2),dir)
local output = complex.mm(input,dft2)
output = complex.mm(dft1,output)
return output
end
function cufft.dftmatrix(n,dir)
local dir = dir or 1
local dft = torch.Tensor(n,n,2)
local real = dft:select(3,1)
local imag = dft:select(3,2)
for i = 1,n do
for j = 1,n do
local theta = 2*math.pi*(i-1)*(j-1)/n
real[i][j] = math.cos(theta)
imag[i][j] = -dir*math.sin(theta)
end
end
return dft
end