ZLUDA Has Been Seeing New Activity For CUDA On AMD GPUs

Back in February I wrote about AMD having quietly funded the effort for a drop-in CUDA implementation for AMD GPUs built atop the ROCm library. This was an incarnation of ZLUDA that originally began as a CUDA implementation for Intel GPUs using oneAPI Level Zero. While AMD discontinued funding ZLUDA development earlier this year, this CUDA implementation for AMD GPUs is continuing to see some new code activity.

ZLUDA came to mind yesterday and I was delighted to notice there's been a stream of new commits. ZLUDA has seen new code as recently as last week. Some of the recent ZLUDA development has focused on code generation improvements, compatibility with older versions of Rust, build improvements, Blender 4.x support, and other fixes/changes.

ZLUDA lead developer Andrzej Janik confirmed there is still some development taking place while continuing to pursue sponsorship for the project to allow the work to continue.

It's great seeing the open-source ZLUDA continue as there's still plenty of software out there targeting NVIDIA CUDA but not AMD ROCm or having less performant AMD code paths as shown in some of the prior ZLUDA benchmarks. Here's to hoping a bright future ahead for ZLUDA can still be achieved. Those wanting to track its progress can do so via ZLUDA on GitHub.