Russian Super-Computing Users Get Tired Of Catalyst, Start Looking At Open-Source AMD

[bridgman]

How's that going btw.? I suppose there's a lot of folks for whom development in the open is new. Have they gotten comfortable quickly?

It's gone pretty well, but some of the conceptual differences took a while to internalize (eg "first you release, THEN you bug fix" ).

[chithanh]

" And yet dear Linus Torvalds give nVidia a middle finger. "

Many people -- especially the ones who live in a universe where they think AMD invented Linux -- take what Linus did out of context. He wasn't complaining about the quality of Nvidia's driver stack. In fact, Torvalds has been known to use Nvidia hardware in at least some of his own systems.

Instead, he was complaining about his working relationship with Nvidia doing kernel development. Which is a fair complaint, but at the same time AMD is far from perfect in that regard too, and at least Nvidia's commercial drivers work more often than not.

Here is the full context: (From http://cpplover.blogspot.com/2012/06...ldsnvidia.html transcription of https://www.youtube.com/watch?v=MShbP3OpASA)

Question:

Okay, Ah so short back story. I got myself two years ago. A laptop that had two graphic cards. It had an Intel and an nVidia. It had a famous Optimus chip that was difficult to operate from linux. I saw at the beggining, I was like expecting to get supported at some point. It was kinda difficult at the beggining and full supports came something like half years ago for one project that's on github that's working pretty nice.

But what I saw that at some point I was expecting that maybe nVidia would kinda chip-in and do something for it. And they said flat out "No! We're not doing any support". And I was like "We're playing in the same sandbox why can't be nice to each other".

It's like... like... things like this that we cannot have the hardware producers think about the other stuff as well. They're like hard-set-on that you cannot truly think cooperate with them on.

What's your comment on this.

Linus:

Ah, I know exactly what you are talking about. And I'm very happy to say that it's... the exception rather than the rule.

And I'm also happy to very publicly point out that nVidia has been one of the worst troubled spot we've had with hardware manufactureres. And that is really sad. Because the nVidia tries to sell chips. A lot of chips into the Android market. And nVidia has been the single worst company we'd ever dealt with.

So nVidia! Fuck you! [Linus shows the finger to the camera]

So the issue is that stuff didn't work and NVidia could not be bothered to cooperate with those who tried to make it work.

[rob11311]

Instead, he was complaining about his working relationship with Nvidia doing kernel development. Which is a fair complaint, but at the same time AMD is far from perfect in that regard too, and at least Nvidia's commercial drivers work more often than not.

WTF!!! It was more about Nvidia's totally intransigent attitude, a non-working relationship requiring reverse engineering for their hardware.
Their binary blob drivers were found to cause plenty of system crashes, which is a reason the kernel sets a taint flag if you use them.

[utack]

Every single bit (0 or 1) in every computer at any location on Earth (and in space) is apolitical.

They are, but somehow Politicans think these generic data processing machines can be made "not for terrorists" or "not to do X", and so does DRM.

[Guest]

They are, but somehow Politicans think these generic data processing machines can be made "not for terrorists" or "not to do X", and so does DRM.

I just want to soft-prevent posts at Phoronix such as: If a person P belongs to community C then the person is automatically a bad person.

That's all.

[Herem]

That thing was always wondering and striking me, just think for a second about it: full time AMD (and Red Hat, and other) Mesa developers, having unlimited access to GPU specifications, making radeon driver which is superior in stability and catching up on performance, than full-time AMD developers, having the same access to GPU specifications.

I think the majority of the people on these forums understand and appreciate the commitment AMD has made towards open source development. I'm sure the effort which has been expended (including the time you and the other devs spend on these forums) will ultimately be beneficial to AMD's long term future.

[pal666]

That thing was always wondering and striking me, just think for a second about it: Mesa developers, with no specification, with pure reverse-engineering, in their spare time, making radeon driver which is superior in both performance and stability, than full-time AMD developers, having unlimited access to GPU specifications.

are you off your meds? mesa radeon is written by full-time AMD developers, having unlimited access to GPU specifications.

[bridgman]

This is probably a good place to point out that since "GPU specifications" are written at the start of the design process not the end they aren't always the magical resource that people expect. Ian Romanick summed it up really well in the description of his FOSDEM talk about simulation:

Hardware is funny stuff. It is often documented to work one way when it actually works a slightly different way. Different revisions of the hardware may have different bugs that require different sets of work-arounds. Programming it even slightly incorrectly can lead to software crashes or system hangs. Sometimes some versions of the hardware work fine, but the version not on the developer's desk crashes. Failure modes are often opaque and give no clues for fixing the problem. Writing robust, reliable software to run directly on hardware is hard.

But yeah, now that we are largely working in parallel with driver development for other OSes we have access to the same documentation and resources as the Catalyst developers, including the simulators and emulators. This is fairly recent though -- we have always had access but since there were a lot of years of catch-up work to be done it was pretty common for HW devs to have forgotten the details about the HW generation by the time we were trying to get drivers working on it.

Non-AMD developers obviously don't have direct access to the internal docs or people, but one of the things the AMD devs do is dig around internally on behalf of community developers to minimize the amount of reverse engineering and random poking required (or at least hold it down to the same level our internal devs are doing).

One non-obvious thing is that HW design for a product generation starts years before production, so most HW developers are working on products that will ship in 2H17 or 2018 and have to "think way back" when asked about a part that hasn't even launched yet.

[ElectricPrism]

"Many people -- especially the ones who live in a universe where they think AMD invented Linux -- take what Linus did out of context."

I know that and AMD invented Linux? Are you high or something? What the problem if we have the quality and they won't contribute to open source community as much as Linus wants?

Are you sure you read that right?

"Many people [...] live in a universe where they think AMD invented Linux"

So you both agree that's absurd.

[liam]

Well, old conditioned bogeyman images are hard to get rid off - and it's always the bad Russian,

never the opposite or other sides,

which is kind of weird if you think about it.

Anyway - why not work together for the greater good like in "Pacific Rim",

let's declare poverty, famine, slow progress on the technological sector our Kaijus (haha - loved that movie)

...what are you talking about?