graph.cc 9.0 KB
Newer Older
X
Xin Pan 已提交
1
/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
X
start  
Xin Pan 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14

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. */

X
Xin Pan 已提交
15 16 17
#include <algorithm>
#include <unordered_set>

X
start  
Xin Pan 已提交
18
#include "paddle/fluid/framework/ir/graph.h"
X
Xin Pan 已提交
19
#include "paddle/fluid/framework/op_proto_maker.h"
20 21
#include "paddle/fluid/framework/program_desc.h"
#include "paddle/fluid/framework/var_desc.h"
X
start  
Xin Pan 已提交
22 23

namespace paddle {
X
Xin Pan 已提交
24
namespace framework {
X
Xin Pan 已提交
25
namespace ir {
X
Xin Pan 已提交
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
std::vector<std::string> FindDistTrainSendVars(
    const std::vector<ir::Node *> &nodes) {
  std::vector<std::string> send_vars;
  // since parameters are all in block 0,
  // it's enough to only scan send ops in block 0
  for (auto &node : nodes) {
    auto op_vars = node->Op()->InputArgumentNames();
    send_vars.reserve(send_vars.size() +
                      std::distance(op_vars.begin(), op_vars.end()));
    send_vars.insert(send_vars.end(), op_vars.begin(), op_vars.end());
  }
  return send_vars;
}

std::vector<std::string> FindDistTrainRecvVars(
    const std::vector<ir::Node *> &nodes) {
  std::vector<std::string> recv_vars;
  for (auto &node : nodes) {
    auto op_vars = node->Op()->OutputArgumentNames();
    recv_vars.reserve(recv_vars.size() +
                      std::distance(op_vars.begin(), op_vars.end()));
    recv_vars.insert(recv_vars.end(), op_vars.begin(), op_vars.end());
  }
  return recv_vars;
}

bool IsDistTrainOp(ir::Node *node, const std::vector<std::string> &send_vars,
                   const std::vector<std::string> &recv_vars) {
  if (send_vars.size() == 0 || recv_vars.size() == 0) {
    return false;
  }

  /**
   * Check any of opvars contains `.block` and in sendvars
   */
  auto checker = [](const std::vector<std::string> &opvars,
                    const std::vector<std::string> &rpc_vars) -> bool {
    for (auto &var : opvars) {
      // a variable name with the suffix `.block` means it's a splited
      // variable by (DistributeTranspiler)
      // [python/paddle/fluid/transpiler/distribute_transpiler.py]
      if (var.find(".block") != std::string::npos &&
          std::find(rpc_vars.begin(), rpc_vars.end(), var) != rpc_vars.end()) {
        return true;
      }
    }
    return false;
  };

  std::vector<std::string> input_var_names;
  std::vector<std::string> output_var_names;
  for (ir::Node *input : node->inputs) {
    input_var_names.push_back(input->Name());
  }
  for (ir::Node *output : node->outputs) {
    output_var_names.push_back(output->Name());
  }

  return checker(output_var_names, send_vars) ||
         checker(input_var_names, recv_vars);
}

X
clean  
Xin Pan 已提交
89
Graph::Graph(const ProgramDesc &program) : program_(program) {
Q
qiaolongfei 已提交
90
  VLOG(3) << "block in program:" << program_.Size();
91 92 93 94 95
  std::unordered_map<std::string, VarDesc *> all_vars;
  for (auto *var : program.Block(0).AllVars()) {
    all_vars.emplace(var->Name(), var);
  }

X
Xin Pan 已提交
96
  std::map<std::string, std::vector<ir::Node *>> var_nodes;
97
  for (auto *op : program.Block(0).AllOps()) {
X
clean  
Xin Pan 已提交
98
    ir::Node *node = CreateOpNode(op);
X
Xin Pan 已提交
99 100
    // For input args, reuse the same var name if it was created before.
    // Otherwise, create a new one.
101 102
    for (auto &each_var_name : op->InputArgumentNames()) {
      ir::Node *var = nullptr;
X
Xin Pan 已提交
103
      if (var_nodes.find(each_var_name) != var_nodes.end()) {
X
Xin Pan 已提交
104
        var = var_nodes.at(each_var_name).back();
X
Xin Pan 已提交
105
      } else if (all_vars.count(each_var_name) != 0) {
X
clean  
Xin Pan 已提交
106
        var = CreateVarNode(all_vars.at(each_var_name));
X
Xin Pan 已提交
107
        var_nodes[each_var_name].push_back(var);
108
      } else {
X
Xin Pan 已提交
109 110 111
        // Operation input var can be optional (dispensable). Which means
        // the operation doesn't really need the var at runtime. In this
        // case, the no-existed var is ready at the beginning.
X
polish  
Xin Pan 已提交
112
        var = CreateEmptyNode(each_var_name, ir::Node::Type::kVariable);
X
Xin Pan 已提交
113
        var_nodes[each_var_name].push_back(var);
114 115 116 117
      }
      node->inputs.push_back(var);
      var->outputs.push_back(node);
    }
X
Xin Pan 已提交
118
    // For output args, always create a new var.
119
    for (auto &each_var_name : op->OutputArgumentNames()) {
X
Xin Pan 已提交
120 121 122 123 124 125 126 127 128
      ir::Node *var = nullptr;
      if (all_vars.count(each_var_name) != 0) {
        var = CreateVarNode(all_vars.at(each_var_name));
      } else {
        // Operation output vars can be @EMPTY@. For example, while_grad
        // can have multi @EMPTY@ outputs with no VarDesc.
        // TODO(panyx0718): Add a test.
        var = CreateEmptyNode(each_var_name, ir::Node::Type::kVariable);
      }
X
Xin Pan 已提交
129
      var_nodes[each_var_name].push_back(var);
130 131 132 133
      node->outputs.push_back(var);
      var->inputs.push_back(node);
    }
  }
X
Xin Pan 已提交
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

  std::vector<ir::Node *> send_ops;
  ir::Node *send_bar = nullptr;
  std::vector<ir::Node *> recv_ops;
  ir::Node *fetch_bar = nullptr;
  for (ir::Node *node : Nodes()) {
    if (node->Name() == "send") {
      send_ops.push_back(node);
    } else if (node->Name() == "send_barrier") {
      PADDLE_ENFORCE(!send_bar, "only has one send barrier");
      send_bar = node;
    } else if (node->Name() == "recv") {
      recv_ops.push_back(node);
    } else if (node->Name() == "fetch_barrier") {
      PADDLE_ENFORCE(!fetch_bar, "only has one fetch barrier");
      fetch_bar = node;
    }
  }
  if (send_bar) {
    for (ir::Node *send : send_ops) {
      ir::Node *dep_var = CreateControlDepVar();
      send->outputs.push_back(dep_var);
      dep_var->inputs.push_back(send);
      send_bar->inputs.push_back(dep_var);
      dep_var->outputs.push_back(send_bar);
    }
    for (ir::Node *recv : recv_ops) {
      ir::Node *dep_var = CreateControlDepVar();
      recv->inputs.push_back(dep_var);
      dep_var->outputs.push_back(recv);
      send_bar->outputs.push_back(dep_var);
      dep_var->inputs.push_back(send_bar);
    }
  }
  if (fetch_bar) {
    for (ir::Node *recv : recv_ops) {
      ir::Node *dep_var = CreateControlDepVar();
      recv->outputs.push_back(dep_var);
      dep_var->inputs.push_back(recv);
      fetch_bar->inputs.push_back(dep_var);
      dep_var->outputs.push_back(fetch_bar);
    }
  }
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191

  std::vector<std::string> send_vars = FindDistTrainSendVars(send_ops);
  std::vector<std::string> recv_vars = FindDistTrainRecvVars(recv_ops);
  for (ir::Node *node : Nodes()) {
    if (IsDistTrainOp(node, send_vars, recv_vars)) {
      if (fetch_bar && node->Name() == "concat") {
        ir::Node *dep_var = CreateControlDepVar();
        fetch_bar->outputs.push_back(dep_var);
        dep_var->inputs.push_back(fetch_bar);
        node->inputs.push_back(dep_var);
        dep_var->outputs.push_back(node);
      }
    }
  }

X
polish  
Xin Pan 已提交
192
  /**
193 194 195 196 197
   * We should handle write after read(WAR) and write after write(WAW) here.
   * Because some of the operators of the program can be executed parallelly.
   * So, to make the program running in the right order, we should add the
   * dependence of WAR and WAW.
   *
X
polish  
Xin Pan 已提交
198 199 200
   *
   * https://en.wikipedia.org/wiki/Hazard_(computer_architecture)#Write_after_read_(WAR)
   */
X
Xin Pan 已提交
201

X
Xin Pan 已提交
202 203 204 205 206 207 208 209 210 211 212 213
  for (auto &var : var_nodes) {
    auto &versions = var.second;
    if (versions.size() <= 1) continue;

    auto it_new = versions.rbegin();
    auto it_old = versions.rbegin();
    ++it_old;
    for (; it_old != versions.rend(); it_new = it_old, ++it_old) {
      ir::Node *write_op =
          (*it_new)->inputs.empty() ? nullptr : (*it_new)->inputs[0];
      const auto &read_ops = (*it_old)->outputs;

214 215 216 217 218
      PADDLE_ENFORCE(write_op, "The write_op should not be empty.");

      // Add write after write dependence
      ir::Node *upstream_op =
          (*it_old)->inputs.empty() ? nullptr : (*it_old)->inputs[0];
X
Xin Pan 已提交
219 220
      // TODO(zcd): Add a test.
      if (upstream_op && upstream_op != write_op) {
221 222 223 224 225 226 227
        ir::Node *dep_var = CreateControlDepVar();
        write_op->inputs.push_back(dep_var);
        upstream_op->outputs.push_back(dep_var);
        dep_var->outputs.push_back(write_op);
        dep_var->inputs.push_back(upstream_op);
      }

X
Xin Pan 已提交
228 229 230 231 232 233
      for (auto *read_op : read_ops) {
        // Manually add a dependency var from read_op to write_op;
        if (read_op == write_op) {
          // Read Write is the same op.
          continue;
        }
X
Xin Pan 已提交
234 235 236 237 238 239 240 241 242 243 244
        // 2 ops might have been connected via other vars.
        bool has_dep = false;
        for (ir::Node *r_out : read_op->outputs) {
          for (ir::Node *w_in : write_op->inputs) {
            if (r_out == w_in) {
              has_dep = true;
              break;
            }
          }
        }
        if (has_dep) continue;
X
Xin Pan 已提交
245

X
Xin Pan 已提交
246
        ir::Node *dep_var = CreateControlDepVar();
X
Xin Pan 已提交
247 248 249 250 251 252 253
        read_op->outputs.push_back(dep_var);
        dep_var->inputs.push_back(read_op);
        write_op->inputs.push_back(dep_var);
        dep_var->outputs.push_back(write_op);
      }
    }
  }
X
better  
Xin Pan 已提交
254
}
X
Xin Pan 已提交
255 256 257 258

bool IsControlDepVar(const ir::Node &var) {
  return var.Name().find(ir::Node::kControlDepVarName) != std::string::npos;
}
X
Xin Pan 已提交
259
}  // namespace ir
X
Xin Pan 已提交
260
}  // namespace framework
X
start  
Xin Pan 已提交
261
}  // namespace paddle