CUDA programming in Julia
This repository hosts a Julia package that bundles functionality from several other packages for CUDA programming, and provides high-level documentation and tutorials for effectively using CUDA GPUs from Julia. The documentation is accessible at juliagpu.gitlab.io.
CUDA.jl includes functionality from the following packages:
- CUDAdrv.jl: interface to the CUDA driver
- CUDAnative.jl: kernel programming capabilities
- CuArrays.jl: GPU array abstraction
For details on the APIs that these packages expose, refer to the associated documentation.
Versioning of this package follows SemVer as used by the Julia package manager: Depending on a specific major version of CUDA.jl should guarantee that your application will not break, as long as it only uses functionality from the package's public API. For CUDA.jl, this API includes certain non-exported functions and macros that would otherwise clash with implementations in Julia. Refer to src/CUDA.jl for more details.