You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

task_compile_engine.cc 1.7 kB

1234567891011121314151617181920212223242526272829303132333435363738394041
  1. /**
  2. * Copyright 2019-2020 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. #include "hybrid/executor/worker/task_compile_engine.h"
  17. #include "init/gelib.h"
  18. #include "hybrid/node_executor/node_executor.h"
  19. namespace ge {
  20. namespace hybrid {
  21. Status TaskCompileEngine::Compile(NodeState &node_state, GraphExecutionContext *context) {
  22. const auto &node_item = *node_state.GetNodeItem();
  23. GE_CHECK_NOTNULL(context);
  24. RECORD_COMPILE_EVENT(context, node_item.NodeName().c_str(), "[Compile] Start");
  25. GE_CHK_RT_RET(rtCtxSetCurrent(context->rt_gen_context));
  26. if (context->ge_context != nullptr) {
  27. GetThreadLocalContext() = *context->ge_context;
  28. }
  29. shared_ptr<NodeTask> kernel_task;
  30. auto ret = node_item.node_executor->CompileTask(*context->model, node_item.node, kernel_task);
  31. RECORD_COMPILE_EVENT(context, node_state.GetName().c_str(), "[Compile] End");
  32. GE_CHK_STATUS_RET(ret, "Failed to create task for node: %s", node_item.NodeName().c_str());
  33. node_state.SetKernelTask(kernel_task);
  34. GELOGI("Compiling node %s successfully", node_state.GetName().c_str());
  35. return SUCCESS;
  36. }
  37. } // namespace hybrid
  38. } // namespace ge

图引擎模块(GE)是MindSpore的一个子模块,其代码由C++实现,位于前端模块ME和底层硬件之间,起到承接作用。图引擎模块以ME下发的图作为输入,然后进行一系列的深度图优化操作,最后输出一张可以在底层硬件上高效运行的图。GE针对昇腾AI处理器的硬件结构特点,做了特定的优化工作,以此来充分发挥出昇腾AI处理器的强大算力。在进行模型训练/推理时,GE会被自动调用而用户并不感知。GE主要由GE API和GE Core两部分组成,详细的架构图如下所示