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.

opr_impl.cpp 1.2 kB

123456789101112131415161718192021222324252627282930313233343536
  1. /**
  2. * \file dnn/src/cuda/eye/opr_impl.cpp
  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. #include "src/cuda/eye/opr_impl.h"
  12. #include "src/cuda/eye/eye.cuh"
  13. #include "src/cuda/utils.h"
  14. namespace megdnn {
  15. namespace cuda {
  16. void EyeImpl::exec(_megdnn_tensor_out dst, _megdnn_workspace workspace)
  17. {
  18. check_exec(dst.layout, workspace.size);
  19. #define cb(DType) \
  20. if (dst.layout.dtype.enumv() == DTypeTrait<DType>::enumv) { \
  21. using ctype = typename DTypeTrait<DType>::ctype; \
  22. eye::exec_internal<ctype>(dst.ptr<ctype>(), \
  23. dst.layout.shape[0], dst.layout.shape[1], \
  24. param().k, \
  25. cuda_stream(handle())); \
  26. }
  27. MEGDNN_FOREACH_COMPUTING_DTYPE(cb)
  28. #undef cb
  29. }
  30. } // namespace cuda
  31. } // namespace megdnn
  32. // vim: syntax=cpp.doxygen foldmethod=marker foldmarker=f{{{,f}}}

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