- cross-posted to:
- linux
- linux@lemmy.ml
- linux@lemmy.world
- rust@lemmit.online
- cross-posted to:
- linux
- linux@lemmy.ml
- linux@lemmy.world
- rust@lemmit.online
cross-posted from: https://lemmy.world/post/20535629
ZLUDA is a compatibility layer for Nividia’s CUDA on other processors
I don’t understand the history here. I thought AMD said “don’t do it”. Now there’s funding from a third party to " just do it"?
There were murmurs that NVIDIA had told AMD to stop funding the project, probably under threat of a lawsuit. Is this project not in a similar position again?
History:
- ZLUDA starts as a project to make CUDA work on Intel GPUs, with funding from Intel.
- Intel pulls funding, author manages to get funding from AMD instead.
- Development of a new version targetting AMD GPUs happens under closed doors with the informal agreement that the source code will be publicly released if AMD pulls funding.
- After a couple of years, AMD pulls funding and the source code for the new version is released.
- Development continues in the open for a few months, albeit at a slowed pace.
- AMD goes back on their word, claims previous agreement wasn’t legally binding and asks that ZLUDA source code be taken down.
- Author reverts codebase to its pre-AMD state, looks for new source of funding.
- ZLUDA’s Third Life
- Anything regarding NVIDIA involvement is pure speculation and should be treated as such.
Oh great, thanks for the run down! That clears things up 👍
It would be great I’d this got to a point where one could “just” buy any GPU without worrying about CUDA. Really amazing project.
Who do we think is ZLUDA’s new backer? 10 bucks says it’s Valve. Tbh I don’t really believe that myself since Valve has been pretty transparent in its funding, but who else stands to gain as much?
What exactly does Valve stand to gain at all from funding a CUDA compatibility layer targetting mainly machine learning software? They’re a video game company. Arguably the most gaming-centric thing CUDA is used for was explicitly discarded in the blog post (“Raytracing is gone”).
Machine learning is massive now and there are many companies who could be interested in funding this kind of project. I’m pretty skeptical it’s possible to make any good guesses with what little info we have.
I was thinking physics processing. Cloth/soft body sim, smoke/dust and other particle physics, potentially water physics, and so on.
No, I bet it is some AI firm. nVidia is currently milking most of them dry and I would not be surprised if it is either Microsoft or Meta