1. 07 1月, 2022 1 次提交
  2. 03 11月, 2021 1 次提交
    • M
      feat(whl/api/lar): enable megengine dll on Windows · 25ec2530
      Megvii Engine Team 提交于
      1: reduce python whl package size
      2: unify api link logic on all OS
      3: add option: MGE_WINDOWS_BUILD_WITH_STATIC_CRT
          --- default OFF
          --- if build CRT(vc runtime) with STATIC with megengine.dll
              some CRT api will crash, for example, flush, so if you
              build with static megengine, and do not want to install CRT
              you can set MGE_WINDOWS_BUILD_WITH_STATIC_CRT TRUE
          --- how to install CRT:
              https://docs.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-160
              install VC_redist.x64.exe
      4: rename megengine_export to megengine_shared(only export needed symbols ),
         caused by runtime symbols conflict with pytorch
      
      GitOrigin-RevId: 93d8d80f29dcf03c00b18e8b708b26df9232e3df
      25ec2530
  3. 19 10月, 2021 1 次提交
  4. 22 3月, 2021 1 次提交
  5. 25 8月, 2020 1 次提交
  6. 19 6月, 2020 2 次提交
    • M
      feat(build): install CMake config module and pkg-config descriptor · 786afef4
      Megvii Engine Team 提交于
      Also, upgrade to CMake 3.13.
      
      The commit also contains significant refactors, as otherwise it is not
      possible to properly export target `megengine` to
      MegEngine-targets.cmake:
      
      1. Optionally use system provided Flatbuffers.
      2. Optionally use system provided MKL-DNN (Tested with Debian).
      3. Refactor megbrain and megdnn targets into object libraries.
      4. Set different path in BUILD_INTERFACE and INSTALL_INTERFACE of
         various target_include_directories.
      5. Specify PUBLIC/PRIVATE on various target_link_libraries.
      
      GitOrigin-RevId: df118a879e12acdc42b56bdd7c7c71b34b0effc3
      786afef4
    • M
      fix(dnn/cuda-stub): simplify and use proper search paths · ea6bfe6c
      Megvii Engine Team 提交于
      Removed the `access()` call before `dlopen()`.
      It was copy-pasted from the opencl-stub, does not make sense here, and
      prevents `dlopen()` from loading `libcuda.so` from non-default path.
      
      Updated the name of the library providing CUDA Driver API on different
      platforms, these are harvested from the following file in a CUDA
      install:
      samples/6_Advanced/matrixMulDynlinkJIT/cuda_drvapi_dynlink.c
      GitOrigin-RevId: ed43cab8c8b4a9b9f8baaa958e8b5123b72dd179
      ea6bfe6c
  7. 25 3月, 2020 1 次提交