227

LibreCUDA – Launch CUDA code on Nvidia GPUs without the proprietary runtime (github.com)

submitted 4 months ago by BrikoX@lemmy.zip to c/opensource@lemmy.ml

12 comments fedilink hide all child comments

LibreCUDA is a project aimed at replacing the CUDA driver API to enable launching CUDA code on Nvidia GPUs without relying on the proprietary CUDA runtime. It achieves this by communicating directly with the hardware via ioctls, (specifically what Nvidia's open-gpu-kernel-modules refer to as the rmapi), as well as QMD, Nvidia's MMIO command queue structure. LibreCUDA is capable of uploading CUDA ELF binaries onto the GPU and launching them via the command queue.

you are viewing a single comment's thread
view the rest of the comments

[-] HumanPenguin@feddit.uk 4 points 4 months ago* (last edited 4 months ago)

Yeah any reverse engineering of closed source code takes time. It's a huge job on its own. Adding the need to avoid actions that may lead to legal issues.

Well yep, It's very likely this may never round to a perfect replacement product.

But it still has value. For starters, it will encourage new open source projects to use it rather than the propria try version, long before it's a direct replacement capable product.

So the effort is worth some excitement. At least a pat on the back and free beer for some of the guys trying.