-
Notifications
You must be signed in to change notification settings - Fork 16
/
THCLNN.lua
170 lines (156 loc) · 4.86 KB
/
THCLNN.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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
-- from THCUNN.lua:
local ffi = require 'ffi'
local THNN = require 'nn.THNN'
local THCLNN = {}
-- load libTHCLNN
local libthclnn_searchpath = package.searchpath('libTHCLNN', package.cpath)
print('libthclnn_searchpath', libthclnn_searchpath)
THCLNN.C = ffi.load(libthclnn_searchpath)
local THCLNN_h = [[
typedef void THClState;
TH_API void THNN_ClAbs_updateOutput(
THClState *state,
THClTensor *input,
THClTensor *output);
TH_API void THNN_ClAbs_updateGradInput(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradInput);
TH_API void THNN_ClELU_updateOutput(
THClState *state,
THClTensor *input,
THClTensor *output,
float alpha,
bool inplace);
TH_API void THNN_ClELU_updateGradInput(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradInput,
THClTensor *output,
float alpha,
bool inplace);
TH_API void THNN_ClTanh_updateOutput(
THClState *state,
THClTensor *input,
THClTensor *output);
TH_API void THNN_ClTanh_updateGradInput(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradInput,
THClTensor *output);
TH_API void THNN_ClSpatialConvolutionMM_updateOutput(
THClState *state,
THClTensor *input,
THClTensor *output,
THClTensor *weight,
THClTensor *bias,
THClTensor *columns,
THClTensor *ones,
int kW, int kH,
int dW, int dH,
int padW, int padH);
TH_API void THNN_ClSpatialConvolutionMM_updateGradInput(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradInput,
THClTensor *weight,
THClTensor *bias,
THClTensor *columns,
THClTensor *ones,
int kW, int kH,
int dW, int dH,
int padW, int padH);
TH_API void THNN_ClSpatialConvolutionMM_accGradParameters(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradWeight,
THClTensor *gradBias,
THClTensor *columns,
THClTensor *ones,
int kW, int kH,
int dW, int dH,
int padW, int padH,
float scale);
TH_API void THNN_ClSpatialAveragePooling_updateOutput(
THClState *state,
THClTensor *input,
THClTensor *output,
int kW, int kH,
int dW, int dH,
int padW, int padH,
bool ceil_mode,
bool count_include_pad);
TH_API void THNN_ClSpatialAveragePooling_updateGradInput(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradInput,
int kW, int kH,
int dW, int dH,
int padW, int padH,
bool ceil_mode,
bool count_include_pad);
TH_API void THNN_ClSpatialMaxPooling_updateOutput(
THClState *state,
THClTensor *input,
THClTensor *output,
THClTensor *indices,
int kW, int kH,
int dW, int dH,
int padW, int padH,
bool ceil_mode);
TH_API void THNN_ClSpatialMaxPooling_updateGradInput(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradInput,
THClTensor *indices,
int kW, int kH,
int dW, int dH,
int padW, int padH,
bool ceil_mode);
TH_API void THNN_ClSoftMax_updateOutput(
THClState *state,
THClTensor *input,
THClTensor *output);
TH_API void THNN_ClSoftMax_updateGradInput(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradInput,
THClTensor *output);
TH_API void THNN_ClSpatialUpSamplingNearest_updateOutput(
THClState *state,
THClTensor *input,
THClTensor *output,
int scale_factor);
TH_API void THNN_ClSpatialUpSamplingNearest_updateGradInput(
THClState *state,
THClTensor *input,
THClTensor *gradOutput,
THClTensor *gradInput,
int scale_factor);
]]
local preprocessed = string.gsub(THCLNN_h, 'TH_API ', '')
ffi.cdef(preprocessed)
local THClState_ptr = ffi.typeof('THClState*')
function THCLNN.getState()
return THClState_ptr(cltorch.getState());
end
local function extract_function_names(s)
local t = {}
for n in string.gmatch(s, 'TH_API void THNN_Cl([%a%d_]+)') do
t[#t+1] = n
end
return t
end
-- build function table
local function_names = extract_function_names(THCLNN_h)
THNN.kernels['torch.ClTensor'] = THNN.bind(THCLNN.C, function_names, 'Cl', THCLNN.getState)
torch.getmetatable('torch.ClTensor').THNN = THNN.kernels['torch.ClTensor']
return THCLNN