Skip to content

Latest commit

 

History

History
56 lines (40 loc) · 2.98 KB

windows_build.md

File metadata and controls

56 lines (40 loc) · 2.98 KB
Raw

Windows build

  1. Install Microsoft Visual Studio (2017 or later). Make sure the option "Desktop development with C++" is selected (you can add it later if not).

  2. Install git for windows - this can be used to get lc0 but is also needed for meson. If you haven't downloaded lc0, you can do it now following the instructions in the README(https://github.com/LeelaChessZero/lc0/blob/master/README.md).

  3. GPU users with nVIDIA cards (and "compute capability" 3.0 or higher) can build with CUDA/CuDNN.

  • Install CUDA (v10.0 is fine for Visual Studio 2017, newer is needed for Visual Studio 2019) and then
  • install the appropriate cuDNN.

  1. GPU users with recent Windows 10 installations can build with DirectX 12, this only requires updated SDK headers (that may already be available in Visual Studio).

  2. CPU users may want to install a BLAS library. This can be either OpenBLAS, Intel MKL or Intel DNNL. This is optional since the Eigen library can be used without installing anything, but probably with worse performance.

  • For OpenBLAS go here, you need a binary package with a filename of the form OpenBLAS-version-Win64-int32.zip, they are not available for all versions, which you just unpack at a location of your choise (but not inside the lc0 directory).
  • For Intel MKL go here, where you need to register. After installation don't forget to run mklvars.bat intel64 to set up the paths to the dlls.
  • For Intel DNNL go here. Note that not all releases have binaries available, you want dnnl_win_*_cpu_vcomp.zip.
  1. For OpenCL you also need to install OpenCL developer libraries.
  • For AMD cards the AMD APP SDK 3.0 seems to be the appropriate one, to be installed after the card drivers. This is not currently available on the AMD website, but links to a signed installer are available in the AMD community forum.
  • For nVIDIA cards it is included in the CUDA toolkit.

  1. Install Python3 - be sure to check the box to add python to the path.

  2. Install Meson: pip3 install --upgrade meson

  3. Edit build.cmd:

  • At the top, set to true and false the variables for the backends you want to build.
  • Then set the paths for the build dependencies.
    • Note: for OPENCL_INCLUDE_PATH you don't want the directory containing opencl.h, but one level higher (the one containing CL).

  1. Run build.cmd. It will ask permission to delete the build directory, then generate MSVS project and pause.

  2. Hit Enter to build it.

  3. Resulting binary will be build/lc0.exe

Alternatively you can

  1. open generated solution build/lc0.sln in Visual Studio and build yourself.