ONE - On-device Neural Engine
Loading...
Searching...
No Matches
BackendContext.cc
Go to the documentation of this file.
1/*
2 * Copyright (c) 2020 Samsung Electronics Co., Ltd. All Rights Reserved
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17#include "BackendContext.h"
18
19#include "TensorBuilder.h"
20#include "KernelGenerator.h"
21#include "util/logging.h"
22#include "ir/Index.h"
23#include "ir/OperandIndexMap.h"
27
29
30namespace onert
31{
32namespace backend
33{
34namespace cpu
35{
36
38{
40 data().op_order, tensor_builder->getSharedMemoryOperandIndexes());
41}
42
44{
45 FunctionMap ret;
46
48 tensor_builder->getSharedMemoryOperandIndexes());
49
50 // TODO: Change type of tensor_registry field to TensorRegistry
51 auto tensor_registry_concreted =
52 nnfw::misc::polymorphic_downcast<basic::TensorRegistry *>(tensor_registry.get());
53 basic::initSharedMemoryConsts(graph()->operands(), external_operands(), tensor_registry_concreted,
54 tensor_builder->getSharedMemoryOperandIndexes());
55
56 for (auto &&op_ind : _data.op_order)
57 {
58 auto fn_seq = kernel_gen->generate(op_ind);
59 ret.emplace(op_ind, std::move(fn_seq));
60 }
61
62 // NOTE For memory optimization, we want to free some operand data
63 const_cast<ir::Graph &>(*_data.graph)
64 .operands()
65 .iterate([&](const ir::OperandIndex &, ir::Operand &obj) { obj.releaseData(); });
66
67 for (auto &&it : ret)
68 {
69 auto &fn_seq = it.second;
70 fn_seq->iterate([&](exec::IFunction &ifunc) { ifunc.prepare(); });
71 }
72
73 return ret;
74}
75
76} // namespace cpu
77} // namespace backend
78} // namespace onert
std::shared_ptr< ITensorRegistry > tensor_registry
const ContextData & data() const
const ir::Graph * graph() const
const util::Set< ir::OperandIndex > & external_operands() const
ITensorRegistry * genTensors() override
std::shared_ptr< KernelGenerator > kernel_gen
std::shared_ptr< TensorBuilder > tensor_builder
virtual void prepare()
Definition IFunction.h:30
void initSharedMemoryConsts(const ir::Operands &operands, const util::Set< ir::OperandIndex > &external_operands, TensorRegistry *tensor_registry, const ir::OperandIndexMap< ir::OperandIndex > &shared_memory_operands_map)
ITensorRegistry * genTensors(const std::shared_ptr< T_TensorBuilder > &tensor_builder, const ir::Graph &graph, const util::Set< ir::OperandIndex > &external_operands, const std::shared_ptr< ITensorRegistry > &tensor_registry, const std::vector< onert::ir::OperationIndex > &op_order, const ir::OperandIndexMap< ir::OperandIndex > &shared_memory_operand_idx)
void initConsts(const ir::Operands &operands, const util::Set< ir::OperandIndex > &external_operands, ITensorRegistry *tensor_registry, const ir::OperandIndexMap< ir::OperandIndex > &shared_memory_operands_map)
std::unordered_map< ir::OperationIndex, std::unique_ptr< exec::FunctionSequence > > FunctionMap
std::vector< onert::ir::OperationIndex > op_order
std::unique_ptr< ir::Graph > graph