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
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
|
/*
* Copyright (c) 2019 Samsung Electronics Co., Ltd. All Rights Reserved
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "ConstantFoldHelper.h"
#include <cassert>
#include <sstream>
#include <string>
namespace
{
// TODO this may need to be moved to loco
bool same_shape(const loco::TensorShape *lhs, const loco::TensorShape *rhs)
{
if (lhs->rank() != rhs->rank())
return false;
for (uint32_t r = 0; r < lhs->rank(); r++)
{
if (lhs->dim(r).value() != rhs->dim(r).value())
return false;
}
return true;
}
} // namespace
namespace moco
{
TFConst *new_const(loco::Graph *graph, loco::TensorShape &tensor_shape, const loco::DataType &dtype)
{
assert(dtype == loco::DataType::S32 || dtype == loco::DataType::FLOAT32);
auto const_node = graph->nodes()->create<TFConst>();
const_node->dtype(dtype);
const_node->rank(tensor_shape.rank());
// Calc number of elements for target node and set shape
uint32_t num_elements = 1;
for (uint32_t r = 0; r < tensor_shape.rank(); r++)
{
const_node->dim(r) = tensor_shape.dim(r);
assert(const_node->dim(r).known());
num_elements = num_elements * const_node->dim(r).value();
}
if (dtype == loco::DataType::S32)
const_node->size<loco::DataType::S32>(num_elements);
else if (dtype == loco::DataType::FLOAT32)
const_node->size<loco::DataType::FLOAT32>(num_elements);
// give name for this node from address to be unique
std::ostringstream oss;
oss << "Const_" << (void *)const_node;
const_node->name(oss.str());
return const_node;
}
} // namespace moco
namespace moco
{
template <> int32_t scalar_from_const<int32_t>(const TFConst *tfconst)
{
assert(tfconst->rank() == 0 || tfconst->rank() == 1);
assert(tfconst->dtype() == loco::DataType::S32);
return tfconst->at<loco::DataType::S32>(0);
}
template <> float scalar_from_const<float>(const TFConst *tfconst)
{
assert(tfconst->rank() == 0 || tfconst->rank() == 1);
assert(tfconst->dtype() == loco::DataType::FLOAT32);
return tfconst->at<loco::DataType::FLOAT32>(0);
}
bool valid_shape_for_constfold_binary_op(const loco::TensorShape &lhs, const loco::TensorShape &rhs)
{
// scalar
if (lhs.rank() == 0 || rhs.rank() == 0)
return true;
// same as scalar
if (lhs.rank() == 1 && lhs.dim(0).value() == 1)
return true;
if (rhs.rank() == 1 && rhs.dim(0).value() == 1)
return true;
// for elementwise binary operation
return ::same_shape(&lhs, &rhs);
}
} // namespace moco
namespace moco
{
float BinaryFunc::apply(float, float) const
{
throw std::runtime_error{"F32 is not supported yet"};
}
int32_t BinaryFunc::apply(int32_t, int32_t) const
{
throw std::runtime_error{"S32 is not supported yet"};
}
} // namespace moco
namespace
{
void apply_binary_s32(const moco::TFConst *lhs, int32_t rhs, moco::TFConst *output,
const moco::BinaryFunc &f)
{
assert(lhs->dtype() == loco::DataType::S32);
assert(same_shape(lhs, output));
uint32_t nume = num_elements(lhs);
for (uint32_t e = 0; e < nume; e++)
{
output->at<loco::DataType::S32>(e) = f.apply(lhs->at<loco::DataType::S32>(e), rhs);
}
}
void apply_binary_f32(const moco::TFConst *lhs, float rhs, moco::TFConst *output,
const moco::BinaryFunc &f)
{
assert(lhs->dtype() == loco::DataType::FLOAT32);
assert(same_shape(lhs, output));
uint32_t nume = num_elements(lhs);
for (uint32_t e = 0; e < nume; e++)
{
output->at<loco::DataType::FLOAT32>(e) = f.apply(lhs->at<loco::DataType::FLOAT32>(e), rhs);
}
}
void apply_binary_s32(const moco::TFConst *lhs, const moco::TFConst *rhs, moco::TFConst *output,
const moco::BinaryFunc &f)
{
assert(same_shape(output, lhs));
assert(same_shape(output, rhs));
assert(output->dtype() == lhs->dtype());
assert(output->dtype() == rhs->dtype());
uint32_t nume = num_elements(lhs);
for (uint32_t e = 0; e < nume; e++)
{
output->at<loco::DataType::S32>(e) =
f.apply(lhs->at<loco::DataType::S32>(e), rhs->at<loco::DataType::S32>(e));
}
}
void apply_binary_f32(const moco::TFConst *lhs, const moco::TFConst *rhs, moco::TFConst *output,
const moco::BinaryFunc &f)
{
assert(same_shape(output, lhs));
assert(same_shape(output, rhs));
assert(output->dtype() == lhs->dtype());
assert(output->dtype() == rhs->dtype());
uint32_t nume = num_elements(lhs);
for (uint32_t e = 0; e < nume; e++)
{
output->at<loco::DataType::FLOAT32>(e) =
f.apply(lhs->at<loco::DataType::FLOAT32>(e), rhs->at<loco::DataType::FLOAT32>(e));
}
}
} // namespace
namespace moco
{
template <>
void apply_binary<int32_t>(const moco::TFConst *x_const, const moco::TFConst *y_const,
moco::TFConst *output_const, const moco::BinaryFunc &f)
{
auto x_shape = moco::tensor_shape(x_const);
auto y_shape = moco::tensor_shape(y_const);
if (y_shape.rank() == 0 || y_shape.rank() == 1)
{
auto rhs = scalar_from_const<int32_t>(y_const);
apply_binary_s32(x_const, rhs, output_const, f);
}
else if (x_shape.rank() == 0 || x_shape.rank() == 1)
{
auto rhs = scalar_from_const<int32_t>(x_const);
apply_binary_s32(y_const, rhs, output_const, f);
}
else
{
apply_binary_f32(x_const, y_const, output_const, f);
}
}
template <>
void apply_binary<float>(const moco::TFConst *x_const, const moco::TFConst *y_const,
moco::TFConst *output_const, const moco::BinaryFunc &f)
{
auto x_shape = moco::tensor_shape(x_const);
auto y_shape = moco::tensor_shape(y_const);
if (y_shape.rank() == 0 || y_shape.rank() == 1)
{
auto rhs = scalar_from_const<float>(y_const);
apply_binary_f32(x_const, rhs, output_const, f);
}
else if (x_shape.rank() == 0 || x_shape.rank() == 1)
{
auto rhs = scalar_from_const<float>(x_const);
apply_binary_f32(y_const, rhs, output_const, f);
}
else
{
apply_binary_f32(x_const, y_const, output_const, f);
}
}
} // namespace moco
|
