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.

extra_impl_helper.cpp 1.6 kB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647
  1. /**
  2. * \file test/common/extra_impl_helper.cpp
  3. * MegEngine is Licensed under the Apache License, Version 2.0 (the "License")
  4. *
  5. * Copyright (c) 2014-2021 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. #include "test/common/extra_impl_helper.h"
  12. namespace megdnn {
  13. namespace test {
  14. template <>
  15. std::function<void(const TensorNDArray&)> extra_impl_helper<AddUpdate>(
  16. Handle* h, const AddUpdate::Param& p) {
  17. auto impl = [](const TensorNDArray& tensors, Handle* h, const AddUpdate::Param& p) {
  18. auto fp32_opr = h->create_operator<AddUpdate>();
  19. auto type_cvt = h->create_operator<TypeCvt>();
  20. fp32_opr->param() = p;
  21. TensorNDArray fp32_tensors;
  22. for (size_t i = 0; i < tensors.size(); ++i) {
  23. auto layout = tensors[i].layout;
  24. layout.dtype = dtype::Float32();
  25. fp32_tensors.emplace_back(malloc(layout.span().dist_byte()), layout);
  26. type_cvt->exec(tensors[i], fp32_tensors[i]);
  27. }
  28. fp32_opr->exec(fp32_tensors[0], fp32_tensors[1]);
  29. type_cvt->exec(fp32_tensors[0], tensors[0]);
  30. for (size_t i = 0; i < tensors.size(); ++i) {
  31. free(fp32_tensors[i].raw_ptr);
  32. }
  33. };
  34. return std::bind(impl, std::placeholders::_1, h, std::cref(p));
  35. }
  36. } // namespace test
  37. } // namespace megdnn
  38. // 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