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.

constant_fuse_same_pass.h 2.5 kB

5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091
  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_CONSTANT_FUSE_SAME_PASS_H_
  17. #define GE_GRAPH_PASSES_CONSTANT_FUSE_SAME_PASS_H_
  18. #include <map>
  19. #include <set>
  20. #include <utility>
  21. #include <vector>
  22. #ifndef ONLY_COMPILE_OPEN_SRC
  23. #include "graph/aligned_ptr.h"
  24. #endif
  25. #include "graph/types.h"
  26. #include "inc/graph_pass.h"
  27. namespace ge {
  28. struct SameConstKey {
  29. int data_size;
  30. #ifndef ONLY_COMPILE_OPEN_SRC
  31. std::shared_ptr<AlignedPtr> aligned_ptr;
  32. #else
  33. const uint8_t *data;
  34. #endif
  35. DataType data_type;
  36. Format format;
  37. std::vector<int64_t> shape;
  38. public:
  39. bool operator< (const SameConstKey &key) const {
  40. if (data_size != key.data_size) {
  41. return data_size < key.data_size;
  42. }
  43. #ifndef ONLY_COMPILE_OPEN_SRC
  44. if (data_size != 0) {
  45. int ret = memcmp(aligned_ptr->Get(), key.aligned_ptr->Get(), data_size);
  46. if (ret != 0) {
  47. return ret < 0;
  48. }
  49. }
  50. #else
  51. int ret = memcmp(data, key.data, data_size);
  52. if (ret != 0) {
  53. return ret < 0;
  54. }
  55. #endif
  56. if (data_type != key.data_type) {
  57. return data_type < key.data_type;
  58. }
  59. if (format != key.format) {
  60. return format < key.format;
  61. }
  62. size_t shape_size = shape.size();
  63. if (shape_size != key.shape.size()) {
  64. return shape_size < key.shape.size();
  65. }
  66. for (size_t i = 0; i < shape_size; ++i) {
  67. if (shape.at(i) != key.shape.at(i)) {
  68. return shape.at(i) < key.shape.at(i);
  69. }
  70. }
  71. return false;
  72. }
  73. };
  74. class ConstantFuseSamePass : public GraphPass {
  75. public:
  76. Status Run(ge::ComputeGraphPtr graph) override;
  77. private:
  78. void GetFuseConstNodes(ComputeGraphPtr &graph,
  79. std::map<SameConstKey, std::vector<NodePtr>> &fuse_nodes);
  80. Status MoveOutDataEdges(NodePtr &src_node, NodePtr &dst_node);
  81. Status FuseConstNodes(ComputeGraphPtr &graph,
  82. std::map<SameConstKey, std::vector<NodePtr>> &fuse_nodes);
  83. };
  84. } // namespace ge
  85. #endif // GE_GRAPH_PASSES_CONSTANT_FUSE_SAME_PASS_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