This repository has been archived by the owner on Aug 5, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 5
/
ffi.lua
214 lines (190 loc) · 4.44 KB
/
ffi.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
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
local ffi = require 'ffi'
local cdefs = [[
void MKLNN_RealSpatialConvolution_forward(
THMKLRealTensor *input,
THMKLRealTensor *output,
THRealTensor *weight,
THRealTensor *bias,
THLongTensor *primitives,
int initOk,
int kW,
int kH,
int dW,
int dH,
int padW,
int padH,
int group);
void MKLNN_RealSpatialConvolution_bwdData(
THMKLRealTensor *input,
THMKLRealTensor *gradOutput,
THMKLRealTensor *gradInput,
THRealTensor *weight,
THRealTensor *bias,
THLongTensor *primitives,
int initOk,
int kW,
int kH,
int dW,
int dH,
int padW,
int padH,
int group);
void MKLNN_RealSpatialConvolution_bwdFilter(
THMKLRealTensor *input,
THMKLRealTensor *gradOutput,
THRealTensor *gradWeight,
THRealTensor *gradBias,
THLongTensor *primitives,
int initOk,
int kW,
int kH,
int dW,
int dH,
int padW,
int padH,
float scale,
int group);
void MKLNN_RealThreshold_updateGradInput(
THMKLRealTensor *input,
THMKLRealTensor *gradOutput,
THMKLRealTensor *gradInput,
float threshold,
bool inplace,
THLongTensor *primitives,
int initOk);
void MKLNN_RealThreshold_updateOutput(
THMKLRealTensor *input,
THMKLRealTensor *output,
float threshold,
float val,
bool inplace,
THLongTensor *primitives,
int initOk);
void MKLNN_RealSpatialMaxPooling_updateOutput(
THMKLRealTensor *input,
THMKLRealTensor *output,
int kW,
int kH,
int dW,
int dH,
int padW,
int padH,
bool ceil_mode,
THLongTensor *primitives,
int initOk);
void MKLNN_RealSpatialMaxPooling_updateGradInput(
THMKLRealTensor *input,
THMKLRealTensor *gradOutput,
THMKLRealTensor *gradInput,
int kW,
int kH,
int dW,
int dH,
int padW,
int padH,
bool ceil_mode,
THLongTensor *primitives,
int initOk);
void MKLNN_RealSpatialAveragePooling_updateOutput(
THMKLRealTensor *input,
THMKLRealTensor *output,
int kW,
int kH,
int dW,
int dH,
int padW,
int padH,
bool ceil_mode,
bool count_include_pad,
THLongTensor *primitives,
int initOk);
void MKLNN_RealSpatialAveragePooling_updateGradInput(
THMKLRealTensor *input,
THMKLRealTensor *gradOutput,
THMKLRealTensor *gradInput,
int kW,
int kH,
int dW,
int dH,
int padW,
int padH,
bool ceil_mode,
bool count_include_pad,
THLongTensor *primitives,
int initOk);
void MKLNN_RealBatchNormalization_updateOutput(
THMKLRealTensor *input,
THMKLRealTensor *output,
THRealTensor *weight,
THRealTensor *bias,
THRealTensor *running_mean,
THRealTensor *running_var,
bool train,
double momentum,
double eps,
THLongTensor *primitives,
int initOk);
void MKLNN_RealBatchNormalization_backward(
THMKLRealTensor *input,
THMKLRealTensor *gradOutput,
THMKLRealTensor *gradInput,
THRealTensor *gradWeight,
THRealTensor *gradBias,
THRealTensor *weight,
THRealTensor *running_mean,
THRealTensor *running_var,
bool train,
double scale,
double eps,
THLongTensor *primitives,
int initOk);
void MKLNN_RealCrossChannelLRN_updateOutput(
THMKLRealTensor *input,
THMKLRealTensor *output,
int size,
float alpha,
float beta,
float k,
THLongTensor *primitives,
int initOk);
void MKLNN_RealCrossChannelLRN_backward(
THMKLRealTensor *input,
THMKLRealTensor *gradOutput,
THMKLRealTensor *gradInput,
int size,
float alpha,
float beta,
float k,
THLongTensor *primitives,
int initOk);
void MKLNN_RealConcat_setupLongTensor(
THLongTensor * array,
THMKLRealTensor *input,
int index);
void MKLNN_RealConcat_updateOutput(
THLongTensor *inputarray,
THMKLRealTensor *output,
int moduleNum,
THLongTensor *primitives,
int initOk);
void MKLNN_RealConcat_backward_split(
THLongTensor *gradarray,
THMKLRealTensor *gradOutput,
int moduleNum,
THLongTensor *primitives,
int initOk);
void MKLNN_RealDropout_updateOutput(
THRealTensor *input,
THRealTensor *output,
double p);
]]
local Real2real = {
Float='float',
Double='double'
}
for Real, real in pairs(Real2real) do
local type_cdefs=cdefs:gsub('Real', Real):gsub('real', real)
ffi.cdef(type_cdefs)
end
local MKLENGINE_PATH = package.searchpath('libmklnn', package.cpath)
mklnn.C = ffi.load(MKLENGINE_PATH)