• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /**
2  * Copyright 2020-2021 Huawei Technologies Co., Ltd
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 <vector>
18 #include <algorithm>
19 #include <memory>
20 #include "tools/converter/legacy_optimizer/graph/subgraph_tensor_pass.h"
21 #include "src/common/log_adapter.h"
22 #include "src/common/utils.h"
23 #include "tools/common/graph_util.h"
24 #include "include/errorcode.h"
25 #include "schema/inner/model_generated.h"
26 #include "src/common/log_util.h"
27 
28 namespace mindspore {
29 namespace lite {
IsUsing(schema::MetaGraphT * graph,const uint32_t & tensor_idx)30 bool SubgraphTensorPass::IsUsing(schema::MetaGraphT *graph, const uint32_t &tensor_idx) {
31   for (const auto &node : graph->nodes) {
32     if (IsContain<uint32_t>(node->inputIndex, tensor_idx)) {
33       return true;
34     }
35     if (IsContain<uint32_t>(node->outputIndex, tensor_idx)) {
36       return true;
37     }
38   }
39   for (const auto &subgraph : graph->subGraph) {
40     if (IsContain<uint32_t>(subgraph->inputIndices, tensor_idx)) {
41       return true;
42     }
43     if (IsContain<uint32_t>(subgraph->outputIndices, tensor_idx)) {
44       return true;
45     }
46   }
47   return false;
48 }
49 
UpdateTensorIdx(schema::MetaGraphT * graph,const uint32_t & tensor_idx)50 STATUS SubgraphTensorPass::UpdateTensorIdx(schema::MetaGraphT *graph, const uint32_t &tensor_idx) {
51   for (const auto &subgraph : graph->subGraph) {
52     UpdateVec<uint32_t>(&(subgraph->inputIndices), tensor_idx);
53     UpdateVec<uint32_t>(&(subgraph->outputIndices), tensor_idx);
54   }
55   for (const auto &node : graph->nodes) {
56     UpdateVec<uint32_t>(&(node->inputIndex), tensor_idx);
57     UpdateVec<uint32_t>(&(node->outputIndex), tensor_idx);
58   }
59   UpdateVec<uint32_t>(&(graph->inputIndex), tensor_idx);
60   UpdateVec<uint32_t>(&(graph->outputIndex), tensor_idx);
61   return RET_OK;
62 }
63 
RemoveUselessTensors(schema::MetaGraphT * graph)64 STATUS SubgraphTensorPass::RemoveUselessTensors(schema::MetaGraphT *graph) {
65   for (auto it = graph->allTensors.begin(); it != graph->allTensors.end();) {
66     uint32_t idx = it - graph->allTensors.begin();
67     if (IsUsing(graph, idx)) {
68       it++;
69     } else {
70       it = graph->allTensors.erase(it);
71       UpdateTensorIdx(graph, idx);
72     }
73   }
74   return RET_OK;
75 }
76 
SyncMainGraphInputAndOutput(schema::MetaGraphT * graph)77 STATUS SubgraphTensorPass::SyncMainGraphInputAndOutput(schema::MetaGraphT *graph) {
78   MS_ASSERT(graph->subGraph.size() > 0);
79   graph->subGraph[0]->inputIndices.assign(graph->inputIndex.begin(), graph->inputIndex.end());
80   return RET_OK;
81 }
82 
Run(schema::MetaGraphT * graph)83 STATUS SubgraphTensorPass::Run(schema::MetaGraphT *graph) {
84   CHECK_NULL_RETURN(graph);
85 
86   int ret = RemoveUselessTensors(graph);
87   if (ret != RET_OK) {
88     MS_LOG(ERROR) << "RemoveUselessTensors failed, ret: " << ret;
89     return ret;
90   }
91 
92   ret = SetSubgraphTensorIndices(graph);
93   if (ret != RET_OK) {
94     MS_LOG(ERROR) << "SetSubgraphTensorIndices failed, ret: " << ret;
95     return ret;
96   }
97 
98   ret = SyncMainGraphInputAndOutput(graph);
99   if (ret != RET_OK) {
100     MS_LOG(ERROR) << "SetSubgraphTensorIndices failed, ret: " << ret;
101     return ret;
102   }
103 
104   return RET_OK;
105 }
106 }  // namespace lite
107 }  // namespace mindspore
108