AMD lawyers claw back CUDA compatibility layer ZLUDA (www.theregister.com)
from Alphane_Moon@lemmy.world to technology@lemmy.world on 11 Aug 2024 06:23
https://lemmy.world/post/18514349

#technology

threaded - newest

seaQueue@lemmy.world on 11 Aug 2024 07:06 next collapse

If anyone’s holding a copy of this repo as of its last published state before the rollback hit me up, I’d love to have a copy of it stashed away to play with.

git bundle should be able to dump the entire repo as a single archive if you have a cloned copy sitting somewhere on disk.

verdantbanana@lemmy.world on 11 Aug 2024 08:24 collapse
ms_lane@lemmy.world on 11 Aug 2024 07:40 next collapse

AMD Doing everything they can to make sure Intel and nVidia stay on top.

boreengreen@lemm.ee on 11 Aug 2024 10:40 next collapse

Well, it does say that Nvidia does not allow a translation layer like this.

tabular@lemmy.world on 11 Aug 2024 11:57 collapse

Nvidia may be using an EULA to try and make people not use a translation layer, but if the EULA doesn’t apply or the consequences of breaking it don’t prevent you continuing then what Nvidia wants means diddly.

I don’t use CUDA or Nvidia so I don’t know but Google release Android Studio and have an EULA saying you can’t do bla bla bla. But Android Studio is open source so if I don’t use their binary and compile it myself then (as far as I know) their EULA doesn’t apply (only the open source license used before they added an EULA on top of it for distribution).

DarkThoughts@fedia.io on 11 Aug 2024 13:40 collapse

An EULA is an End User License Agreement. It has no legal authority over a customer who does not even use an nvidia product, let alone a company.

tabular@lemmy.world on 11 Aug 2024 14:14 collapse

Perhaps not even when you use an Nvidia product like if I buy Nvidia hardware but don’t use their software (i.e. use open source drivers instead). I don’t know enough about CUDA to say if you’re not using Nvidia software (normally, the topic discusses a reverse-engineered one which doesn’t infringe on Nvidia’s copyright of their software).

datelmd5sum@lemmy.world on 11 Aug 2024 13:27 next collapse

Intel are on top?

sugar_in_your_tea@sh.itjust.works on 11 Aug 2024 13:58 next collapse

No, but that’s not AMD’s fault.

zik@lemmy.world on 11 Aug 2024 20:13 collapse

Not in reliability…

But they’re probably still selling more CPUs to your average buyer who always buys Intel, doesn’t read tech news and never even heard about the controversy.

datelmd5sum@lemmy.world on 11 Aug 2024 21:01 collapse

And they’re still somehow generating twice the revenue with Xeons vs. what AMD does with EPYCs. Who keeps buying all these Xeons!?

sunzu@kbin.run on 12 Aug 2024 02:38 collapse

Amd dominanace started with 5000 series 4 years ago, it takes time for corpos to change vendor like that I would assume. So it takes years to play out.

Static_Rocket@lemmy.world on 13 Aug 2024 03:26 collapse

This is a short term loss for a potential long term improvement. By eliminating dependency on translation APIs they can force the use of more open solutions like oneAPI which is even getting buy-in from companies like Imagination.

Keeping cuda alive is a bad idea.

verdantbanana@lemmy.world on 11 Aug 2024 08:21 collapse

old news but there is new news on this topic

docs.scale-lang.com

github.com/AdaptiveCpp/AdaptiveCpp