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.

sub_kernel.h 1.5 kB

5 years ago
5 years ago
5 years ago
5 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445
  1. /**
  2. * Copyright 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 GE_GRAPH_PASSES_FOLDING_KERNEL_SUB_KERNEL_H_
  17. #define GE_GRAPH_PASSES_FOLDING_KERNEL_SUB_KERNEL_H_
  18. #include <vector>
  19. #include "inc/kernel.h"
  20. #include "common/fp16_t.h"
  21. namespace ge {
  22. class SubKernel : public Kernel {
  23. public:
  24. Status Compute(const ge::OpDescPtr attr, const std::vector<ge::ConstGeTensorPtr> &input,
  25. vector<ge::GeTensorPtr> &v_output) override;
  26. private:
  27. std::vector<int8_t> y_data_int8_t_;
  28. std::vector<int16_t> y_data_int16_t_;
  29. std::vector<int32_t> y_data_int32_t_;
  30. std::vector<int64_t> y_data_int64_t_;
  31. std::vector<uint8_t> y_data_uint8_t_;
  32. std::vector<uint16_t> y_data_uint16_t_;
  33. std::vector<uint32_t> y_data_uint32_t_;
  34. std::vector<uint64_t> y_data_uint64_t_;
  35. std::vector<fp16_t> y_data_fp16_t_;
  36. std::vector<float> y_data_float_;
  37. std::vector<double> y_data_double_;
  38. };
  39. } // namespace ge
  40. #endif // GE_GRAPH_PASSES_FOLDING_KERNEL_SUB_KERNEL_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