build_cinn_pass.cc 18.8 KB
Newer Older
J
jiangcheng 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/* Copyright (c) 2021 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/framework/paddle2cinn/build_cinn_pass.h"

17 18
#include <algorithm>
#include <iterator>
J
jiangcheng 已提交
19
#include <memory>
20
#include <regex>
J
jiangcheng 已提交
21 22 23
#include <string>
#include <unordered_map>
#include <unordered_set>
24
#include <utility>
J
jiangcheng 已提交
25 26
#include <vector>

27 28
#include "cinn/frontend/op_mapper_registry.h"
#include "cinn/frontend/op_mappers/use_op_mappers.h"
29 30
#include "gflags/gflags.h"
#include "glog/logging.h"
J
jiangcheng 已提交
31
#include "paddle/fluid/framework/ir/graph.h"
32
#include "paddle/fluid/framework/ir/graph_pattern_detector.h"
J
jiangcheng 已提交
33 34
#include "paddle/fluid/framework/ir/node.h"
#include "paddle/fluid/framework/ir/subgraph_detector.h"
35
#include "paddle/fluid/framework/op_proto_maker.h"
36
#include "paddle/fluid/framework/paddle2cinn/cinn_compiler.h"
37 38
#include "paddle/fluid/platform/enforce.h"
#include "paddle/fluid/platform/errors.h"
J
jiangcheng 已提交
39

40 41 42
DECLARE_string(allow_cinn_ops);
DECLARE_string(deny_cinn_ops);

J
jiangcheng 已提交
43 44 45 46 47 48 49 50 51
namespace paddle {
namespace framework {
namespace paddle2cinn {

using framework::ir::Graph;
using framework::ir::Node;

using GraphNodeVec = std::vector<Node*>;
using GraphNodeSet = std::unordered_set<Node*>;
52
using GraphNodeMap = std::unordered_map<Node*, Node*>;
J
jiangcheng 已提交
53

54
namespace {
55 56 57 58
// The delim(`;`) that is used to split the FLAGS_allow_cinn_ops
// & FLAGS_deny_cinn_ops.
constexpr char kDelim[] = ";";

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
const std::unordered_map<std::string, std::unordered_set<std::string>>
    kDenyParamMap = {{"batch_norm", {"ReserveSpace"}},
                     {"batch_norm_grad", {"ReserveSpace"}}};

std::unordered_set<std::string> GetDenyVarNames(const GraphNodeSet& cluster) {
  std::unordered_set<std::string> deny_var_set;

  auto get_debug_info = [](const std::unordered_set<std::string>& var_names) {
    std::string debug_info = "[";
    for (auto& var : var_names) {
      debug_info.append(var);
      debug_info.append(", ");
    }
    debug_info.append("]");
    return debug_info;
  };

  for (auto* op : cluster) {
    if (kDenyParamMap.count(op->Name())) {
      const auto* desc = op->Op();
      PADDLE_ENFORCE_NE(desc, nullptr,
                        platform::errors::PreconditionNotMet(
                            "The Op %s's OpDesc should not be NULL, which has "
                            "a parameter in kDenyParamMap.",
                            op->Name().c_str()));

      auto deny_param_names = kDenyParamMap.at(op->Name());
      VLOG(4) << "We found deny param " << get_debug_info(deny_param_names)
              << " in op [" << op->Name() << "].";

      for (const auto& param_name : deny_param_names) {
        if (desc->Inputs().count(param_name)) {
          const auto& arg_names = desc->Input(param_name);
          for (const auto& arg_name : arg_names) {
            deny_var_set.insert(arg_name);
            VLOG(4) << "deny param [" << param_name << "]'s argument name"
                    << " is [" << arg_name << "].";
          }
        }

        if (desc->HasOutput(param_name)) {
          const auto& arg_names = desc->Output(param_name);
          for (const auto& arg_name : arg_names) {
            deny_var_set.insert(arg_name);
            VLOG(4) << "deny param [" << param_name << "]'s argument name"
                    << " is [" << arg_name << "].";
          }
        }
      }
    }
  }

  VLOG(4) << "All deny var names are " << get_debug_info(deny_var_set);

  return deny_var_set;
}

116 117 118 119 120 121 122 123 124 125
std::unordered_set<std::string> StringSplit(const std::string& str,
                                            const std::string& delim) {
  std::regex reg(delim);
  std::unordered_set<std::string> elems{
      std::sregex_token_iterator(str.begin(), str.end(), reg, -1),
      std::sregex_token_iterator()};
  elems.erase("");
  return elems;
}

126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
int ExtractOpRole(const GraphNodeSet& cluster) {
  std::unordered_set<int> op_roles;
  std::string attr_name = OpProtoAndCheckerMaker::OpRoleAttrName();
  for (auto* n : cluster) {
    if (n->Op() && n->Op()->HasAttr(attr_name)) {
      op_roles.insert(BOOST_GET_CONST(int, n->Op()->GetAttr(attr_name)));
    }
  }
  if (op_roles.size() == 1U) {
    return *(op_roles.begin());
  } else {
    return static_cast<int>(OpRole::kNotSpecified);
  }
}

141 142
// Deal with subgraph's feed input var node:
// create a new input var node and it's feed op node
143 144 145
void AddFeedOpAndVar(const GraphNodeSet& feed_vars, const GraphNodeSet& cluster,
                     const GraphNodeMap& old_op2new_op,
                     const GraphNodeMap& old_var2new_var, Graph* graph) {
146 147 148 149 150 151 152
  for (auto* old_var : feed_vars) {
    // create feed op
    OpDesc desc;
    desc.SetType("feed");
    desc.SetOutput("Out", {old_var->Name()});
    auto op = graph->CreateOpNode(&desc);

153 154
    // get new feed var node
    auto* var = old_var2new_var.at(old_var);
155
    VLOG(4) << "Add Feed Op before: " << var->Name();
156 157

    // link feed op and feed var
158
    IR_NODE_LINK_TO(op, var);
159 160 161 162

    // link feed var to cluster op
    for (auto* old_op : old_var->outputs) {
      if (cluster.count(old_op)) {
163
        IR_NODE_LINK_TO(var, old_op2new_op.at(old_op));
164 165
      }
      // Do not need relink old op or old var here, they will be
166
      // fixed in RemoveSubGraphFromGraph, here we just deal with
167 168 169 170 171 172 173 174
      // new subgraph's node.
    }
  }
}

// Deal with subgraph's parameter var node:
// create a new input var node, it's data will get by scope,
// so it don't need feed op
175 176 177
void AddParamVar(const GraphNodeSet& param_vars, const GraphNodeSet& cluster,
                 const GraphNodeMap& old_op2new_op,
                 const GraphNodeMap& old_var2new_var, Graph* graph) {
178
  for (auto* old_var : param_vars) {
179
    auto* var = old_var2new_var.at(old_var);
180
    VLOG(4) << "Add Param Var Node: " << var->Name();
181 182 183

    for (auto* old_op : old_var->outputs) {
      if (cluster.count(old_op)) {
184
        IR_NODE_LINK_TO(var, old_op2new_op.at(old_op));
185 186 187 188 189 190 191
      }
    }
  }
}

// Deal with subgraph's outputs var node:
// create a new output var node and it's fetch op
192 193 194
void AddOutputVar(const GraphNodeSet& output_vars, const GraphNodeSet& cluster,
                  const GraphNodeMap& old_op2new_op,
                  const GraphNodeMap& old_var2new_var, Graph* graph) {
195
  for (auto* old_var : output_vars) {
196
    auto* var = old_var2new_var.at(old_var);
197
    VLOG(4) << "Add Output Var Node: " << var->Name();
198 199 200

    for (auto* old_op : old_var->inputs) {
      if (cluster.count(old_op)) {
201
        IR_NODE_LINK_TO(old_op2new_op.at(old_op), var);
202 203 204 205 206
      }
    }
  }
}

J
jiangcheng 已提交
207 208
// Create new subgraph with and op nodes are cluster nodes, and all
// var node are from internal nodes
209 210
std::unique_ptr<Graph> CreateNewSubGraph(const GraphNodeSet& cluster,
                                         const GraphNodeSet& cluster_internals,
211 212
                                         const GraphNodeSet& cluster_inputs,
                                         const GraphNodeSet& cluster_outputs) {
J
jiangcheng 已提交
213 214
  // Graph's constructor must has one parameter, and in our code,
  // the ProgramDesc is useless, so here we pass a temporary object.
215
  auto subgraph = std::make_unique<Graph>(framework::ProgramDesc());
J
jiangcheng 已提交
216

217
  GraphNodeMap old_op2new_op;
J
jiangcheng 已提交
218
  for (auto* op : cluster) {
219
    auto sub_node = subgraph->CreateOpNode(op->Op());
J
jiangcheng 已提交
220 221 222
    old_op2new_op[op] = sub_node;
  }

223
  GraphNodeMap old_var2new_var;
J
jiangcheng 已提交
224
  for (auto* var : cluster_internals) {
225 226 227 228 229
    PADDLE_ENFORCE_NOT_NULL(var->Var(),
                            platform::errors::PreconditionNotMet(
                                "The var desc of the node in cluster_internals "
                                "shouldn't be null."));
    auto* sub_node = subgraph->CreateVarNode(var->Var());
J
jiangcheng 已提交
230 231
    old_var2new_var[var] = sub_node;
  }
232 233 234 235 236 237 238 239 240 241 242 243
  for (auto* var : cluster_inputs) {
    if (var->Var()) {
      auto* sub_node = subgraph->CreateVarNode(var->Var());
      old_var2new_var[var] = sub_node;
    }
  }
  for (auto* var : cluster_outputs) {
    if (var->Var()) {
      auto* sub_node = subgraph->CreateVarNode(var->Var());
      old_var2new_var[var] = sub_node;
    }
  }
J
jiangcheng 已提交
244

245
  GraphNodeSet need_feed_vars;
246
  std::unordered_set<Node *> param_vars, output_vars;
J
jiangcheng 已提交
247 248 249 250 251
  // the subgraph is independently, so here we only need link
  // to the node in new subgraph, and discard the link to
  // out-graph.
  for (auto* op : cluster) {
    for (auto* var : op->inputs) {
252 253 254 255
      // one output var maybe an input of the cluster
      if (cluster_internals.count(var) ||
          (cluster_outputs.count(var) && old_var2new_var.count(var))) {
        IR_NODE_LINK_TO(old_var2new_var.at(var), old_op2new_op.at(op));
256
      } else if (cluster_inputs.count(var) && var->Var() != nullptr) {
257 258 259 260 261 262 263 264 265 266 267
        if (var->Var()->IsParameter()) {
          // Parameters have been preserved in scope, compared to feed var,
          // param just need add new var and don't need add feed op.
          // The var is used for check whether we need preserve the tensor
          // when transform paddle scope to CINN scope.
          param_vars.insert(var);
        } else {
          // When the var is subgraph input and the var is not parameter,
          // we need add a new feed op to feed the var.
          need_feed_vars.insert(var);
        }
J
jiangcheng 已提交
268 269 270 271
      }
    }
    for (auto* var : op->outputs) {
      if (cluster_internals.count(var)) {
272
        IR_NODE_LINK_TO(old_op2new_op.at(op), old_var2new_var.at(var));
273
      } else if (cluster_outputs.count(var) && var->Var() != nullptr) {
274 275 276 277
        // Create new output var node to guarantee the independency of
        // subgraph. In other words, the subgraph has no connection with
        // other graph, even the input graph.
        output_vars.insert(var);
J
jiangcheng 已提交
278 279 280 281
      }
    }
  }

282 283 284 285 286 287
  AddFeedOpAndVar(need_feed_vars, cluster, old_op2new_op, old_var2new_var,
                  subgraph.get());
  AddParamVar(param_vars, cluster, old_op2new_op, old_var2new_var,
              subgraph.get());
  AddOutputVar(output_vars, cluster, old_op2new_op, old_var2new_var,
               subgraph.get());
J
jiangcheng 已提交
288

289
  return subgraph;
J
jiangcheng 已提交
290 291 292 293
}

// This interface is used to classify all variables involved in a cluster into
// three types: inputs, outputs, and internals.
294 295 296
// The input node is some subgraph op's input but not any subgraph op's output.
// The output node is some subgraph op's output and some out-graph op's input.
// Specially, the internal node is a node that only used by subgraph, and
J
jiangcheng 已提交
297
// out-graph should not using this node at all.
298 299
// cluster_inputs & cluster_outputs & cluster_internals == NULL
// cluster_outputs | cluster_internals == all graph op's outputs node
300 301 302 303 304
void AnalyseClusterVariables(
    const GraphNodeSet& cluster,
    const std::unordered_set<std::string>& deny_var_set,
    GraphNodeSet* cluster_inputs, GraphNodeSet* cluster_outputs,
    GraphNodeSet* cluster_internals) {
J
jiangcheng 已提交
305 306
  // collecting all input and output of op
  for (auto* op_node : cluster) {
307
    const auto& op_name = op_node->Name();
J
jiangcheng 已提交
308
    for (auto* input_var_node : op_node->inputs) {
309 310 311 312
      if (!deny_var_set.count(input_var_node->Name())) {
        // ignore deny var node
        cluster_inputs->insert(input_var_node);
      }
J
jiangcheng 已提交
313 314
    }
    for (auto* output_var_node : op_node->outputs) {
315 316 317
      if (!deny_var_set.count(output_var_node->Name())) {
        cluster_outputs->insert(output_var_node);
      }
J
jiangcheng 已提交
318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
    }
  }
  // remove output node from cluster_inputs,
  // and add cluster_internals node
  for (auto* var_node : *cluster_outputs) {
    if (cluster_inputs->count(var_node) > 0) {
      // if a input node also exists in output list, remove
      cluster_inputs->erase(var_node);

      // the internal node is must an output node of sub-graph,
      // but not any input node of out-graph.
      bool is_only_used_internal = true;
      for (auto* next_op_node : var_node->outputs) {
        is_only_used_internal &= (cluster.count(next_op_node) > 0);
      }
      if (is_only_used_internal) {
        cluster_internals->insert(var_node);
      }
    }
  }

  // if a output node also exists in internal list, remove.
  for (auto* var_node : *cluster_internals) {
    cluster_outputs->erase(var_node);
  }
}

345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360
void AddLinkToCinnOp(const GraphNodeSet& cluster_inputs,
                     const GraphNodeSet& cluster_outputs, Node* cinn_op_node) {
  // add new link from cluster_inputs to cinn_op_node
  for (auto* var_node : cluster_inputs) {
    IR_NODE_LINK_TO(var_node, cinn_op_node);
  }

  // add new link from cinn_op_node to cluster_outputs
  for (auto* var_node : cluster_outputs) {
    IR_NODE_LINK_TO(cinn_op_node, var_node);
  }
}

void AddCinnOpToGraph(const GraphNodeSet& cluster,
                      const GraphNodeSet& cluster_inputs,
                      const GraphNodeSet& cluster_outputs,
361 362 363
                      const std::string& compilation_key,
                      const std::unordered_set<std::string>& deny_var_set,
                      Graph* graph) {
364 365 366
  // Add the cinn launch op
  framework::OpDesc cinn_op_desc;
  cinn_op_desc.SetType(kCinnLaunchOp);
367
  std::vector<std::string> input_names;
368

369
  std::for_each(cluster_inputs.begin(), cluster_inputs.end(),
370 371
                [&input_names, &deny_var_set](Node* n) {
                  if (n->Var() != nullptr && !deny_var_set.count(n->Name())) {
372 373 374
                    input_names.emplace_back(n->Name());
                  }
                });
375
  cinn_op_desc.SetInput("X", input_names);
376
  std::vector<std::string> output_names;
377
  std::for_each(cluster_outputs.begin(), cluster_outputs.end(),
378 379
                [&output_names, &deny_var_set](Node* n) {
                  if (n->Var() != nullptr && !deny_var_set.count(n->Name())) {
380 381 382
                    output_names.emplace_back(n->Name());
                  }
                });
383 384 385 386 387 388 389 390
  cinn_op_desc.SetOutput("Out", output_names);
  cinn_op_desc.SetAttr(kCompilationKey, compilation_key);
  cinn_op_desc.SetAttr(OpProtoAndCheckerMaker::OpRoleAttrName(),
                       ExtractOpRole(cluster));
  cinn_op_desc.Flush();
  auto* cinn_op_node = graph->CreateOpNode(&cinn_op_desc);
  // Add new links from or to the the cinn launch op node
  AddLinkToCinnOp(cluster_inputs, cluster_outputs, cinn_op_node);
391 392

  VLOG(4) << "Add op [" << kCinnLaunchOp << "] into graph.";
J
jiangcheng 已提交
393 394 395 396 397 398
}

// Removing cluster node and internals node from Graph
void RemoveSubGraphFromGraph(const GraphNodeSet& cluster,
                             const GraphNodeSet& cluster_internals,
                             Graph* graph) {
399 400 401 402 403 404
  const std::unordered_set<const Node*> const_cluster{cluster.cbegin(),
                                                      cluster.cend()};
  const std::unordered_set<const Node*> const_internals{
      cluster_internals.cbegin(), cluster_internals.cend()};
  ir::GraphSafeRemoveNodes(graph, const_cluster);
  ir::GraphSafeRemoveNodes(graph, const_internals);
J
jiangcheng 已提交
405 406
}

407
// Replacing Cinn subgraph to a cinn op node, whose op_type is
J
jiangcheng 已提交
408 409
// kCinnLaunchOp, and inputs ares cluster_inputs and outputs are
// cluster_outputs.
410
// Meanwhile, move all links of cluster to the cinn op.
411 412 413 414 415
void ReplaceSubGraphWithCinnOpNode(
    const GraphNodeSet& cluster, const GraphNodeSet& cluster_inputs,
    const GraphNodeSet& cluster_outputs, const GraphNodeSet& cluster_internals,
    const std::string& compilation_key,
    const std::unordered_set<std::string>& deny_var_set, Graph* graph) {
416 417
  // Add the cinn op node whose name is "kCinnLaunchOp" into graph
  AddCinnOpToGraph(cluster, cluster_inputs, cluster_outputs, compilation_key,
418
                   deny_var_set, graph);
419
  // Remove the cinn subgraph from graph
J
jiangcheng 已提交
420 421 422 423 424 425 426
  RemoveSubGraphFromGraph(cluster, cluster_internals, graph);
}

// Search all subgraphs which all op node supported by CINN,
// Here we using SubgraphDetector to detecte the subgraph that
// all of op node supported by CINN. We using OpMapperRegistry
// to check whether the op node supported by CINN.
427
void SearchAllSubgraphs(Graph* graph) {
428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445
  auto allow_ops = StringSplit(FLAGS_allow_cinn_ops, kDelim);
  auto deny_ops = StringSplit(FLAGS_deny_cinn_ops, kDelim);
  auto teller = [&allow_ops, &deny_ops](const Node* node) {
    bool registered = ::cinn::frontend::OpMapperRegistry::Global()->Find(
                          node->Name()) != nullptr;
    // if the op type is registered in CINN and allow_ops is not empty, return
    // true only when it is in allow_ops
    if (allow_ops.size()) {
      return registered && allow_ops.count(node->Name());
    }
    // if the op type is registered in CINN and deny_ops is not empty, return
    // true only when it is not in deny_ops
    if (deny_ops.size()) {
      return registered && !deny_ops.count(node->Name());
    }
    // if the user doesn't set FLAGS_allow_cinn_ops and FLAGS_deny_cinn_ops,
    // return true only when it is registered in CINN
    return registered;
J
jiangcheng 已提交
446
  };
447 448
  VLOG(4) << "The allowed Cinn Ops: " << FLAGS_allow_cinn_ops;
  VLOG(4) << "The denied Cinn Ops: " << FLAGS_deny_cinn_ops;
J
jiangcheng 已提交
449 450 451
  std::vector<GraphNodeVec> clusters =
      framework::ir::SubgraphDetector(graph, teller)();

452 453 454 455 456 457 458 459 460 461
  auto cluster_debug_info = [](const GraphNodeSet& cluster) {
    std::string res = "(";
    for (auto* node : cluster) {
      res.append(node->Name());
      res.append(", ");
    }
    res.append(")");
    return res;
  };

462
  auto* cinn_compiler = CinnCompiler::GetInstance();
J
jiangcheng 已提交
463
  for (const auto& node_vec : clusters) {
464
    // Classify var node to inputs, outputs, and internals.
J
jiangcheng 已提交
465 466
    GraphNodeSet cluster_set(node_vec.begin(), node_vec.end());

467 468
    auto deny_var_set = GetDenyVarNames(cluster_set);

J
jiangcheng 已提交
469
    GraphNodeSet cluster_inputs, cluster_outputs, cluster_internals;
470 471
    AnalyseClusterVariables(cluster_set, deny_var_set, &cluster_inputs,
                            &cluster_outputs, &cluster_internals);
472 473 474 475 476 477 478

    VLOG(4) << "Cluster Ops: " << cluster_debug_info(cluster_set);
    VLOG(4) << "Cluster input vars: " << cluster_debug_info(cluster_inputs);
    VLOG(4) << "Cluster output vars: " << cluster_debug_info(cluster_outputs);
    VLOG(4) << "Cluster internal vars: "
            << cluster_debug_info(cluster_internals);

479 480
    // Create a new subgraph according to the found cluster and
    // save it in CinnCompiler
481 482
    std::string compilation_key = cinn_compiler->AddGraph(CreateNewSubGraph(
        cluster_set, cluster_internals, cluster_inputs, cluster_outputs));
483 484
    VLOG(4) << "Compilation Key:\n"
            << cinn_compiler->ReadableKey(compilation_key);
485

486 487
    // Replace the found cluster to a new cinn op node
    ReplaceSubGraphWithCinnOpNode(cluster_set, cluster_inputs, cluster_outputs,
488 489
                                  cluster_internals, compilation_key,
                                  deny_var_set, graph);
J
jiangcheng 已提交
490 491
  }
}
492
}  // namespace
J
jiangcheng 已提交
493

494
void BuildCinnPass::ApplyImpl(Graph* graph) const { SearchAllSubgraphs(graph); }
J
jiangcheng 已提交
495 496 497 498 499 500

}  // namespace paddle2cinn
}  // namespace framework
}  // namespace paddle

REGISTER_PASS(build_cinn_pass, paddle::framework::paddle2cinn::BuildCinnPass);