A Julia wrapper for the CUDA driver API.

This package aims to provide high-level wrappers for the functionality exposed by the CUDA driver API, and is meant for users who need high- or low-level access to the CUDA toolkit or the underlying hardware.

The package is built upon the low-level CUDA driver API, but that shouldn't make the Julia wrapper any harder to use. That said, it is a work-in-progress and does not offer the same functionality or convenience as the more popular CUDArt package, which is built upon the higher-level CUDA runtime API.



At the Julia REPL:

using CUDAdrv

# optionally

Building CUDAdrv might display error messages, indicating issues with your set-up. These messages can be cryptic as they happen too early for decent error handling to be loaded. However, please pay close attention to them as they might prevent CUDAdrv.jl from working properly! Some common issues:

Even if the build fails, CUDAdrv.jl should always be loadable. This simplifies use by downstream packages, until there is proper language support for conditional modules. You can check whether the package has been built properly by inspecting the CUDAdrv.configured global variable.