News AMD asks developer to take down open source ZLUDA, dev vows to rebuild his project

Status
Not open for further replies.
The hard part will be implementing features that are not similar enough to what was done with AMD to be considered a breach. Other projects have done it when the original open source license is changed to "business source licenses" (Consul for example), but the chance of legal action usually reduces the feature set devs will implement in its place.
 
IDK how they can backtrack..if theres proof AMD said he could then that should be legal. Not devs fault if that person sending email didnt ask fact is company agreed to it.
Legal or not doesn't matter when you're an individual with limited funds facing the risk of company lawyers. It's not a fair playing field any more.
 
If they have an e-mail from AMD, that is a legal document. If AMD's legal department now want to stop the person working on the code, that is a different legal document. It may supersede the original one, but would not invalidate it.
It's probably a bit more in depth than that, along the lines of "As we provided funding, as co-owners of the code we withdraw the right to publish the code or create any new code based on the funded code" which is why the dev has to go back to the pre-AMD code and fork it.
 
This smells more like NV talking to AMD and saying you have your propriety stuff we have ours, we should avoid stepping on each other's toes, stifling innovation, compatibility and user choice, basic cartel practices!
 
In the way nVidia worded the new EULA for CUDA this is self-explanatory.

I'll have to assume a couple things:
1- The fella didn't disclose his way of doing this translation layer was doing reverse-engineering of the CUDA code.
2- AMD accepted it as a work piece which would not be doing anything against any EULA or TOS from nVidia.

As you may, or may not, know reverse-engineering code is illegal. Why? No idea, but it just is. I'm guessing because there's copyright involved in using code you're not licenced to when doing whatever you need to. This is, among some other subtle things, why "Piracy" is illegal: you're not just disabling things, but also reverse engineering code to know how to do it.

It sucks, but I can't blame AMD for trying to distance itself from this before nVidia's lawyers build up a case against them. The question now is: will this be enough? I still think nVidia may have grounds for a trial if this fella can testify that AMD accepted to fund him if he disclosed he would reverse-engineer code. That would be super nasty.

Regards.
 
  • Like
Reactions: -Fran-
This claim is overly broard to the point of falsehood.

Furthermore, Andrzej Janik is a resident of the EU, not the US, so is afforded even more protections for reverse-engineering software, even if the EULA says 'no'.

Both are moot, in that AMD can simply browbeat compliance via the threat of litigation costs even if AMD do not legally have a leg to stand on.
From your link:

If your access to the code or computer system you are studying is conditioned upon agreeing to any contractual terms (e.g. End User License Agreements (EULA), terms of service notices (TOS), terms of use notices (TOU), a non-disclosure agreement (NDA), developers agreement or API agreement), you are at greater legal risk if your research activities do not comply with their stated terms and conditions. You should talk to a lawyer before agreeing to any terms and before studying any software distributed with such terms and conditions, even if you have come into possession of that code without agreeing to anything.

That's exactly what I mentioned.

I do take your point, but I have to insist that AMD, with good reason, is being overly cauteous.

Regards.
 
This claim is overly broard to the point of falsehood.

Furthermore, Andrzej Janik is a resident of the EU, not the US, so is afforded even more protections for reverse-engineering software, even if the EULA says 'no'.

Both are moot, in that AMD can simply browbeat compliance via the threat of litigation costs even if AMD do not legally have a leg to stand on.
Absolutely agreed. There are no software patents in Europe.

Further, in the European Union, reverse engineering is permitted – under the verdict of the Court of Justice issued in 2012 – for creating computer programs with operation analogous to the original program.

In the United States, even if an artifact or process is protected by trade secrets, reverse-engineering the artifact or process is often lawful if it has been legitimately obtained. Since the project's reverse engineering was from a time it was legitimately obtained, it should still be lawful. Only features created after the license change would be covered by the EULA, and only for reverse engineerimg in the US. As shown above, the reverse engineering can be done legally in Europe and that reverse engineering (since it's lawfully obtained) would presumably be protected by the US law.

Looks like this is reliant on squeezing those who can't afford lawyers, as it does not seem to be legitimate. Looks like a latterday DECSS Jon is required.

(For those unfamiliar with the reference, a European guy reverse-engineered CSS Encryption. The MPAA and RIAA arranged for the guy to be violently abused by the authorities, since the law didn't forbid the reverse engineering, so they couldn't legally do anything. It... kinda backfired on them.)
 
  • Like
Reactions: -Fran-
We wouldn't need ZLUDA if ROCm wasn't, you know, 100% completely worthless and universally unsupported by devs on PC.
I don't think the Mac Versions of Adobe products had ROCm support for GPU-acceleration either, not even in the era when AMD was making exclusive workstation GPUs specifically designed for people running adobe products on Mac.

But as is it is, using ZLUDA is the only thing making AMD cards even remotely viable as a "it might only be 1/4 the performance as Nvidia, but at least it's half the cost" alternative for AI on PC.

Some people with Radeon gaming PCs might be willing to dual boot Linux for AI work, but it's probably not a lot - and anybody building a dedicated AI box is just buying Nvidia.
 
So in other news, Nvidia doesnt want Anything to upset their good times selling gpus to people wanting to do AI work.
 
Status
Not open for further replies.