ir_pass_manager.cc 5.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
// Copyright (c) 2018 PaddlePaddle Authors. 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 "paddle/fluid/inference/analysis/ir_pass_manager.h"
16
#include <memory>
17
#include <string>
18
#include <unordered_map>
19 20
#include <unordered_set>
#include <utility>
L
luotao1 已提交
21
#include <vector>
Y
Yan Chunwei 已提交
22
#include "paddle/fluid/framework/ir/fuse_pass_base.h"
23 24
#include "paddle/fluid/framework/ir/graph.h"
#include "paddle/fluid/framework/scope.h"
25 26
#include "paddle/fluid/inference/analysis/argument.h"
#include "paddle/fluid/inference/analysis/ir_passes/subgraph_detector.h"
Y
Yan Chunwei 已提交
27
#include "paddle/fluid/string/pretty_log.h"
28 29 30 31

namespace paddle {
namespace inference {
namespace analysis {
Y
Yan Chunwei 已提交
32 33 34
using string::PrettyLogEndl;
using string::PrettyLog;
using string::Style;
35

36 37 38 39 40 41 42 43 44 45 46
IRPassManager::IRPassManager(Argument *argument) {
  ARGUMENT_CHECK_FIELD(argument, main_program);
  graph_ = std::unique_ptr<Graph>(new Graph(argument->main_program()));
  if (argument->Has("scope")) {
    graph_->Set(framework::ir::kParamScopeAttr,
                new framework::Scope *(
                    const_cast<framework::Scope *>(&argument->scope())));
  }

  ARGUMENT_CHECK_FIELD(argument, ir_analysis_passes);
  CreatePasses(argument, argument->ir_analysis_passes());
47 48
}

49 50
void IRPassManager::CreatePasses(Argument *argument,
                                 const std::vector<std::string> &passes) {
51
  std::string pre_pass;
L
luotao1 已提交
52
  int pass_num = 0;
53
  for (const std::string &pass_name : passes) {
54
    auto pass = framework::ir::PassRegistry::Instance().Get(pass_name);
55

56
    if (pass_name == "graph_viz_pass") {
L
luotao1 已提交
57 58 59
      std::string dot_file_path = std::to_string(pass_num) + "_ir_" +
                                  (pre_pass.empty() ? "origin" : pre_pass) +
                                  ".dot";
60
      pass->Set("graph_viz_path", new std::string(std::move(dot_file_path)));
L
luotao1 已提交
61
      pass_num++;
62
    } else if (pass_name == "mkldnn_placement_pass") {
63 64 65
      pass->Set("mkldnn_enabled_op_types",
                new std::unordered_set<std::string>(
                    argument->mkldnn_enabled_op_types()));
66 67 68
    } else if (pass_name == "cpu_quantize_pass") {
      pass->Set("quant_var_scales",
                new VarQuantScale(argument->quant_var_scales()));
69 70 71 72 73 74 75 76
    }

    if (pass_name == "anakin_subgraph_pass") {
      pass->Set("program",
                new framework::ProgramDesc *(&argument->main_program()));
    }

    if (pass_name == "tensorrt_subgraph_pass") {
77 78
      pass->Set("workspace_size", new int(argument->tensorrt_workspace_size()));
      pass->Set("max_batch_size", new int(argument->tensorrt_max_batch_size()));
79 80
      pass->Set("min_subgraph_size",
                new int(argument->tensorrt_min_subgraph_size()));
N
nhzlx 已提交
81 82
      pass->Set("program",
                new framework::ProgramDesc *(&argument->main_program()));
N
nhzlx 已提交
83

N
nhzlx 已提交
84
      bool enable_int8 = argument->tensorrt_precision_mode() ==
85
                         AnalysisConfig::Precision::kInt8;
N
nhzlx 已提交
86 87

      pass->Set("enable_int8", new bool(enable_int8));
N
nhzlx 已提交
88 89 90 91 92 93 94
      std::string model_opt_cache_dir =
          argument->Has("model_dir")
              ? argument->model_dir()
              : GetDirRoot(argument->model_program_path());
      pass->Set(
          "model_opt_cache_dir",
          new std::string(GetOrCreateModelOptCacheDir(model_opt_cache_dir)));
N
nhzlx 已提交
95
      pass->Set("gpu_device_id", new int(argument->gpu_device_id()));
N
nhzlx 已提交
96 97
      pass->Set("use_static_engine",
                new bool(argument->tensorrt_use_static_engine()));
98 99
    }

100
    pre_pass = pass_name;
101 102

    passes_.emplace_back(std::move(pass));
103 104 105
  }
}

106 107 108 109 110 111 112
std::unique_ptr<Graph> IRPassManager::Apply(std::unique_ptr<Graph> graph) {
  if (passes_.empty()) {
    return graph;
  }
  PADDLE_ENFORCE(graph.get());
  // Apply all the passes
  for (const auto &pass : passes_) {
Y
Yan Chunwei 已提交
113 114 115
    if (pass->Type() != "graph_viz_pass") {
      PrettyLogEndl(Style::H2(), "--- Running IR pass [%s]", pass->Type());
    }
116 117
    graph = pass->Apply(std::move(graph));
  }
G
Gabor Buella 已提交
118
  return graph;
119 120 121
}

framework::proto::ProgramDesc IRPassManager::AcquireProgram(
N
nhzlx 已提交
122
    std::unique_ptr<Graph> *graph, ProgramDesc *program) const {
123 124 125
  auto pass =
      framework::ir::PassRegistry::Instance().Get("graph_to_program_pass");

N
nhzlx 已提交
126 127
  // Direct using ProgramDesc desc(argument->main_program()) may cause
  // incomplete copies of information.
N
nhzlx 已提交
128
  ProgramDesc desc;
N
nhzlx 已提交
129
  desc.CopyFrom(*program->Proto());
130 131 132 133 134 135
  pass->SetNotOwned("program", &desc);
  auto *the_graph = graph->release();
  *graph = pass->Apply(std::unique_ptr<Graph>(the_graph));
  return *desc.Proto();
}

136 137 138
}  // namespace analysis
}  // namespace inference
}  // namespace paddle