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.

ge_prof.h 2.3 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869
  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. #ifndef INC_EXTERNAL_GE_GE_PROF_H_
  17. #define INC_EXTERNAL_GE_GE_PROF_H_
  18. #include <map>
  19. #include <string>
  20. #include <vector>
  21. #include "ge/ge_api_error_codes.h"
  22. namespace ge {
  23. enum ProfDataTypeConfig {
  24. kProfAcl = 0x0001,
  25. kProfTaskTime = 0x0002,
  26. kProfAiCoreMetrics = 0x0004,
  27. kProfAicpuTrace = 0x0008,
  28. kProfModelExecute = 0x0010,
  29. kProfRuntimeApi = 0x0020,
  30. kProfRuntimeTrace = 0x0040,
  31. kProfScheduleTimeline = 0x0080,
  32. kProfScheduleTrace = 0x0100,
  33. kProfAiVectorCoreMetrics = 0x0200,
  34. kProfSubtaskTime = 0x0400,
  35. kProfTrainingTrace = 0x0800,
  36. kProfHcclTrace = 0x1000,
  37. kProfDataProcess = 0x2000,
  38. kProfTaskTrace = 0x3842,
  39. kProfModelLoad = 0x8000000000000000
  40. };
  41. enum ProfilingAicoreMetrics {
  42. kAicoreArithmaticThroughput = 0,
  43. kAicorePipeline = 1,
  44. kAicoreSynchronization = 2,
  45. kAicoreMemory = 3,
  46. kAicoreInternalMemory = 4,
  47. kAicoreStall = 5,
  48. kAicoreMetricsAll = 255 // only for op_trace
  49. };
  50. typedef struct ProfAicoreEvents ProfAicoreEvents;
  51. typedef struct aclgrphProfConfig aclgrphProfConfig;
  52. Status aclgrphProfInit(const char *profiler_path, uint32_t length);
  53. Status aclgrphProfFinalize();
  54. aclgrphProfConfig *aclgrphProfCreateConfig(uint32_t *deviceid_list, uint32_t device_nums,
  55. ProfilingAicoreMetrics aicore_metrics, ProfAicoreEvents *aicore_events,
  56. uint64_t data_type_config);
  57. Status aclgrphProfDestroyConfig(aclgrphProfConfig *profiler_config);
  58. Status aclgrphProfStart(aclgrphProfConfig *profiler_config);
  59. Status aclgrphProfStop(aclgrphProfConfig *profiler_config);
  60. } // namespace ge
  61. #endif // INC_EXTERNAL_GE_GE_PROF_H_

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