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.

handle.h 5.0 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148
  1. /**
  2. * \file dnn/include/megdnn/handle.h
  3. * MegEngine is Licensed under the Apache License, Version 2.0 (the "License")
  4. *
  5. * Copyright (c) 2014-2020 Megvii Inc. All rights reserved.
  6. *
  7. * Unless required by applicable law or agreed to in writing,
  8. * software distributed under the License is distributed on an
  9. * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  10. */
  11. #pragma once
  12. #include "megcore.h"
  13. #include "megdnn/config/config.h"
  14. #include "megdnn/basic_types.h"
  15. #include <functional>
  16. #include <memory>
  17. #include "megdnn/internal/visibility_prologue.h"
  18. namespace megdnn {
  19. class OperatorBase;
  20. class Handle {
  21. public:
  22. enum class HandleType {
  23. NAIVE = 0,
  24. FALLBACK = 1,
  25. X86 = 2,
  26. CUDA = 6,
  27. };
  28. protected:
  29. Handle(megcoreComputingHandle_t computing_handle, HandleType type);
  30. public:
  31. /**
  32. * \brief Create a MegDNN handle from a MegCore Computing handle.
  33. *
  34. * \param[in] computing_handle MegCore computing handle. Please note
  35. * that computing_handle would not be released when this Handle is
  36. * destructed
  37. * \param[in] debug_level
  38. * Applicable for CPU computing handle.
  39. * 0 means taking the fastest possible code path; it may contains
  40. * platform-specific instructions such as SSE for x86_64 or NEON for
  41. * armv7v7.
  42. * 1 means taking the fastest possible code path without
  43. * platform-specific instructions in C++ code. Note that the compiled
  44. * binary file still contains platform-specific codes.
  45. * 2 means taking the naive code path. Performance is severely
  46. * hampered, but it is less error-prone since the internal
  47. * implementation is rather straightforward.
  48. *
  49. * **Debug level 1 and 2 should not be used in productions.**
  50. */
  51. static std::unique_ptr<Handle> make(
  52. megcoreComputingHandle_t computing_handle,
  53. int debug_level = 0);
  54. #if MEGDNN_WITH_CUDA
  55. static std::unique_ptr<Handle> make_cuda_handle(
  56. megcoreComputingHandle_t computing_handle);
  57. template <typename opr>
  58. std::unique_ptr<opr> create_cuda_operator();
  59. #endif
  60. virtual ~Handle();
  61. /*!
  62. * \brief Get the underlying megcore computing handle.
  63. */
  64. megcoreComputingHandle_t megcore_computing_handle() const {
  65. return m_computing_handle;
  66. }
  67. /*!
  68. * \brief set a callback function to be invoked when this handle is
  69. * destructed, so associated resources can be released (e.g.
  70. * computing handle)
  71. *
  72. * This function can be called at most once.
  73. */
  74. void set_destructor(const thin_function<void()> &d);
  75. /*!
  76. * \brief set a callback to be invoked when an operator is destructed
  77. * \param[in,out] cb the callback function; it would be set to the
  78. * previous callback function
  79. */
  80. void set_opr_destruct_callback(thin_function<void(OperatorBase*)> &cb) {
  81. cb.swap(m_on_opr_destructed);
  82. }
  83. void on_opr_destructed(OperatorBase* opr);
  84. /**
  85. * \brief Create operator of Opr type.
  86. */
  87. template <typename Opr>
  88. std::unique_ptr<Opr> create_operator();
  89. /*
  90. * =============================================================
  91. * Users should call functions below to query memory requirement.
  92. * =============================================================
  93. */
  94. /**
  95. * \brief The internal data pointer of TensorND should be aligned to
  96. * alignment_requirement() in bytes.
  97. */
  98. virtual size_t alignment_requirement() const;
  99. //! get alignment in bytes for rows of image 2D tensor format
  100. virtual size_t image2d_pitch_alignment() const;
  101. HandleType type() const {
  102. return m_handle_type;
  103. }
  104. /**
  105. * \brief Check is the layout satisfy cross device copy constraint.
  106. * 1. The handle of the src and the dst is the same kind
  107. * 2. The dst is continguous.
  108. */
  109. virtual bool check_cross_dev_copy_constraint(const TensorLayout &src);
  110. private:
  111. static constexpr uint32_t ALIVE_MAGIC = 0x8595e9d2u;
  112. volatile uint32_t m_alive_magic = ALIVE_MAGIC;
  113. megcoreComputingHandle_t m_computing_handle;
  114. const HandleType m_handle_type;
  115. thin_function<void()> m_destructor;
  116. thin_function<void(OperatorBase*)> m_on_opr_destructed;
  117. Handle() = delete;
  118. Handle(const Handle &rhs) = delete;
  119. Handle &operator=(const Handle &rhs) = delete;
  120. };
  121. } // namespace megdnn
  122. #include "megdnn/internal/visibility_epilogue.h"
  123. // vim: syntax=cpp.doxygen

MegEngine 安装包中集成了使用 GPU 运行代码所需的 CUDA 环境,不用区分 CPU 和 GPU 版。 如果想要运行 GPU 程序,请确保机器本身配有 GPU 硬件设备并安装好驱动。 如果你想体验在云端 GPU 算力平台进行深度学习开发的感觉,欢迎访问 MegStudio 平台

Contributors (1)