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
|
/*
* 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 "moco/Pass/Passes/ConstantFoldPack.h"
#include "ConstantFoldHelper.h"
#include "TensorPackEnumerator.h"
#include <moco/IR/Nodes/TFPack.h>
#include <moco/IR/Nodes/TFConst.h>
#include <moco/Support/NodeAs.h>
#include <loco/IR/TensorShape.h>
#include <oops/UserExn.h>
#include <cassert>
#include <vector>
namespace
{
bool valid_axis_range(int32_t output_rank, int32_t pack_axis)
{
// check axis range in [-r-1, r+1)
assert(output_rank > 0);
return (-output_rank <= pack_axis) && (pack_axis < output_rank);
}
bool constantfold_pack(moco::TFPack *node)
{
// check if all the inputs are Const
std::vector<moco::TFConst *> input_nodes;
uint32_t num = node->N();
for (uint32_t index = 0; index < num; ++index)
{
auto in = dynamic_cast<moco::TFConst *>(node->values(index));
if (in == nullptr)
return false;
input_nodes.push_back(in);
}
assert(input_nodes.size() == num);
// check if all inputs have same shape and dtype
auto input_0 = input_nodes.at(0);
auto shape_0 = moco::tensor_shape(input_0);
auto dtype_0 = input_0->dtype();
if (dtype_0 != loco::DataType::S32 && dtype_0 != loco::DataType::FLOAT32)
{
// TODO support other types
assert(false);
return false;
}
for (uint32_t index = 1; index < num; ++index)
{
auto input_i = input_nodes.at(index);
auto shape_i = moco::tensor_shape(input_i);
auto dtype_i = input_i->dtype();
if (!(shape_0 == shape_i))
return false;
if (dtype_0 != dtype_i)
return false;
}
int32_t output_rank = static_cast<int32_t>(shape_0.rank() + 1);
int32_t pack_axis = node->axis();
if (!valid_axis_range(output_rank, pack_axis))
{
throw oops::UserExn("axis is out of range: ", node->name());
}
if (pack_axis < 0)
{
pack_axis = output_rank + pack_axis;
}
// define output shape
loco::TensorShape output_shape;
output_shape.rank(output_rank);
for (int32_t r = 0, s = 0; r < output_rank; ++r)
{
if (r == pack_axis)
{
output_shape.dim(r).set(num);
}
else
{
output_shape.dim(r).set(shape_0.dim(s++).value());
}
}
auto graph = node->graph();
// create new constant
auto output_const = moco::new_const(graph, output_shape, input_0->dtype());
moco::TensorPackEnumerator etor;
etor.shape(shape_0, output_shape);
etor.axis(pack_axis);
for (etor.start(); etor.valid(); etor.advance())
{
uint32_t inp_num = etor.inp_num();
uint32_t inp_element = etor.inp_element();
uint32_t out_element = etor.out_element();
auto inp_const = input_nodes[inp_num];
if (input_0->dtype() == loco::DataType::S32)
{
int32_t val = inp_const->at<loco::DataType::S32>(inp_element);
output_const->at<loco::DataType::S32>(out_element) = val;
}
else if (input_0->dtype() == loco::DataType::FLOAT32)
{
float val = inp_const->at<loco::DataType::FLOAT32>(inp_element);
output_const->at<loco::DataType::FLOAT32>(out_element) = val;
}
}
// replace
loco::replace(node).with(output_const);
return true;
}
} // namespace
namespace moco
{
/**
* @note This will Replace TFPack with TFConst when inputs are TFConst
*
* Before
* A --- TFPack --- C
* B --/
* After
* A --- TFPack
* B --/
* TFConst ---------- C
* Where
* A, B : inputs of TFPack
* C : a node that uses TFPack as an input
* TFPack is disconnected from C
* Nodes are drawn multiple times to simplify the diagram
*/
bool ConstantFoldPack::run(loco::Graph *graph)
{
bool changed = false;
for (auto node : loco::active_nodes(loco::output_nodes(graph)))
{
if (auto pack_node = as<moco::TFPack>(node))
{
if (constantfold_pack(pack_node))
changed = true;
}
}
return changed;
}
} // namespace moco
|
