AMD GPU (ROCm) programming in Julia

• Julia 1.9+ (Navi 3 GPUs require Julia 1.10 or higher).
• ROCm 5.3+, which means only GPUs that are supported by these versions will work.
• 64-bit Linux and Windows are supported, see requirements section in the docs.

For Julia 1.11 set JULIA_LLVM_ARGS="-opaque-pointers" to enable opaque pointers and use system-wide device libraries, instead of patched from artifacts.

AMDGPU.jl can be installed with the Julia package manager. From the Julia REPL, type ] to enter the Pkg REPL mode and run:

pkg> add AMDGPU

Or, equivalently, via the Pkg API:

julia> import Pkg; Pkg.add("AMDGPU")

To ensure that everything is working you can run tests for the package with

pkg> test AMDGPU

Or specifying a subset of tests to run:

julia> using Pkg

julia> Pkg.test("AMDGPU"; test_args=["core", "kernelabstractions"])

Full list of tests to run can be obtained with --list argument:

julia> Pkg.test("AMDGPU"; test_args=["--list"])

Usage questions can be posted on the Julia Discourse forum under the GPU domain and/or in the #gpu channel of the Julia Slack.

Contributions are very welcome, as are feature requests and suggestions. Please open an issue if you encounter any problems.

AMDGPU would not have been possible without the work by Tim Besard and contributors to CUDA.jl and LLVM.jl.

AMDGPU.jl is licensed under the MIT License.