From 91b259e643a170cafdddf5da47c7ee64598b5b2d Mon Sep 17 00:00:00 2001 From: Roy Shilkrot Date: Tue, 12 Sep 2023 01:14:02 -0400 Subject: [PATCH] readme --- README.md | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/README.md b/README.md index 303fa59..16b3703 100644 --- a/README.md +++ b/README.md @@ -77,3 +77,22 @@ Use the CI scripts again, for example: ``` The build should exist in the `./release` folder off the root. You can manually install the files in the OBS directory. + +#### Building with CUDA support on Windows + +To build with CUDA support on Windows, you need to install the CUDA toolkit from NVIDIA. The CUDA toolkit is available for download from [here](https://developer.nvidia.com/cuda-downloads). + +After installing the CUDA toolkit, you need to set variables to point CMake to the CUDA toolkit installation directory. For example, if you have installed the CUDA toolkit in `C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4`, you need to set `CUDA_TOOLKIT_ROOT_DIR` to `C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4` and `LOCALVOCAL_WITH_CUDA` to `ON` when running `.github/scripts/Build-Windows.ps1`. + +For example +```powershell +> .github/scripts/Build-Windows.ps1 -Target x64 -ExtraCmakeFlags "-D LOCALVOCAL_WITH_CUDA=ON -D CUDA_TOOLKIT_ROOT_DIR='C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4'" +``` + +You will need to copy a few CUDA .dll files to the location of the plugin .dll for it to run. The required .dll files from CUDA (which are located in the `bin` folder of the CUDA toolkit installation directory) are: + +- `cudart64_NN.dll` +- `cublas64_NN.dll` +- `cublasLt64_NN.dll` + +where `NN` is the CUDA version number. For example, if you have installed CUDA 11.4 then `NN` is likely `11`.