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 5.1 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151
  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. ARM_COMMON = 3,
  27. ARMV7 = 4,
  28. AARCH64 = 5,
  29. CUDA = 6,
  30. };
  31. protected:
  32. Handle(megcoreComputingHandle_t computing_handle, HandleType type);
  33. public:
  34. /**
  35. * \brief Create a MegDNN handle from a MegCore Computing handle.
  36. *
  37. * \param[in] computing_handle MegCore computing handle. Please note
  38. * that computing_handle would not be released when this Handle is
  39. * destructed
  40. * \param[in] debug_level
  41. * Applicable for CPU computing handle.
  42. * 0 means taking the fastest possible code path; it may contains
  43. * platform-specific instructions such as SSE for x86_64 or NEON for
  44. * armv7v7.
  45. * 1 means taking the fastest possible code path without
  46. * platform-specific instructions in C++ code. Note that the compiled
  47. * binary file still contains platform-specific codes.
  48. * 2 means taking the naive code path. Performance is severely
  49. * hampered, but it is less error-prone since the internal
  50. * implementation is rather straightforward.
  51. *
  52. * **Debug level 1 and 2 should not be used in productions.**
  53. */
  54. static std::unique_ptr<Handle> make(
  55. megcoreComputingHandle_t computing_handle,
  56. int debug_level = 0);
  57. #if MEGDNN_WITH_CUDA
  58. static std::unique_ptr<Handle> make_cuda_handle(
  59. megcoreComputingHandle_t computing_handle);
  60. template <typename opr>
  61. std::unique_ptr<opr> create_cuda_operator();
  62. #endif
  63. virtual ~Handle();
  64. /*!
  65. * \brief Get the underlying megcore computing handle.
  66. */
  67. megcoreComputingHandle_t megcore_computing_handle() const {
  68. return m_computing_handle;
  69. }
  70. /*!
  71. * \brief set a callback function to be invoked when this handle is
  72. * destructed, so associated resources can be released (e.g.
  73. * computing handle)
  74. *
  75. * This function can be called at most once.
  76. */
  77. void set_destructor(const thin_function<void()> &d);
  78. /*!
  79. * \brief set a callback to be invoked when an operator is destructed
  80. * \param[in,out] cb the callback function; it would be set to the
  81. * previous callback function
  82. */
  83. void set_opr_destruct_callback(thin_function<void(OperatorBase*)> &cb) {
  84. cb.swap(m_on_opr_destructed);
  85. }
  86. void on_opr_destructed(OperatorBase* opr);
  87. /**
  88. * \brief Create operator of Opr type.
  89. */
  90. template <typename Opr>
  91. std::unique_ptr<Opr> create_operator();
  92. /*
  93. * =============================================================
  94. * Users should call functions below to query memory requirement.
  95. * =============================================================
  96. */
  97. /**
  98. * \brief The internal data pointer of TensorND should be aligned to
  99. * alignment_requirement() in bytes.
  100. */
  101. virtual size_t alignment_requirement() const;
  102. //! get alignment in bytes for rows of image 2D tensor format
  103. virtual size_t image2d_pitch_alignment() const;
  104. HandleType type() const {
  105. return m_handle_type;
  106. }
  107. /**
  108. * \brief Check is the layout satisfy cross device copy constraint.
  109. * 1. The handle of the src and the dst is the same kind
  110. * 2. The dst is continguous.
  111. */
  112. virtual bool check_cross_dev_copy_constraint(const TensorLayout &src);
  113. private:
  114. static constexpr uint32_t ALIVE_MAGIC = 0x8595e9d2u;
  115. volatile uint32_t m_alive_magic = ALIVE_MAGIC;
  116. megcoreComputingHandle_t m_computing_handle;
  117. const HandleType m_handle_type;
  118. thin_function<void()> m_destructor;
  119. thin_function<void(OperatorBase*)> m_on_opr_destructed;
  120. Handle() = delete;
  121. Handle(const Handle &rhs) = delete;
  122. Handle &operator=(const Handle &rhs) = delete;
  123. };
  124. } // namespace megdnn
  125. #include "megdnn/internal/visibility_epilogue.h"
  126. // 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