AMD Has No Plans To Suspend Catalyst For Linux

[bridgman]

What plans are those? Probably there are no comments about that because people are not aware of that to begin with...

I wasn't able to get out to AFDS this year so don't know the exact words Phil used (will try to find a video link tomorrow) but if you're familiar with FSA/HSA the announcement was basically about providing the full Linux execution stack (compiler/runtime/kernel drivers) in open source form, except for one commercial third party piece (the C++ parser front end). Slides and abbreviated transcript at :

http://www.slideshare.net/hsafoundation/afds2012-the-programmers-guide-to-a-universe-of-possibility-heterogeneous-system-architecture

Slide 30 IIRC. The presentation was basically a "heads up" for developers and ISVs. There was some media coverage but mostly summarized as "will also be available on Linux"

[smitty3268]

I'm more curious about what kind of black magic of graphics driver development (science?) has not been revealed by now in white papers, verbally, subtle hints, etc. Surely the fast-path "trickery" for specific programs can't explain all the performance difference... Being this frustrated, I can only imagine the open source devs'.

I think a lot has to do with just managing memory/states better - every time the card flushes data to the CPU or vice versa it introduces a huge delay, and I think the proprietary drivers are just much smarter at minimizing the number of those that are required.

That's a tricky thing to optimize, because you have to look at the whole flow of how the driver is working, you can't just sit down and optimize 1 function or 1 file. In fact, a lot is probably tied into the cross-driver bits of Mesa rather than even 1 specific hardware driver.

I imagine that effects Intel much less since their GPU is sitting right on the die along with the CPU, and also because it is relatively much less powerful - which means delays are both shorter and less likely to be noticed in the first place.

[edgar_wibeau]

nobody seems to have commented on the open source plans discussed at AFDS yet, which is surprising

I've also been waiting for at least a mention on phoronix front page. That was a mjor commitment end to mee sounded like at least OpenCL & stuff (HSA bolt) will fully be supported in open source under Linux, which, to me at least, sounds absolutely amazing! Of course it'll have to be because otherwise AMD APUs wouldn't be fully Linux compatible anymore, at least in the way it's meant to be (free and fully open - and what is Linux without this?), but still: Great news!

halfmanhalfamazing: First thing i thought when I read the headline (fglrx not being dropped) was what a pity That's because it's just not usable for me, hasn't been for years (e.g. [worst example:] switching video streams using xinelib just crashes X, always has, with FGLRX). So dump this useless sh*t and fully go for the radeon driver. Which is relatively slow, but stable. But of course I know that it takes some time before it can (potentially) completely take over.

[enrico.tagliavini]

I'm sorry I will not be constructive but AMD should just stop make stupid decision kicking its user in the butt. And I talk by a OS-indipendent point of view. On my laptop I'm not able to update to windows 8, as far as I understand, couse they will not add support for new OSes. It is very logical to update a 1 year old pc to the lastest version of windows. Of course I'm not intrested, but the fact that I just can't because AMD likes to tag my 1 year old harware as legacy is just stupid! This is even more true for GNU/Linux of course given it is updated really more frequently.

In the end it doesn't really matter if AMD want to stop the development of fglrx given the problem is far upper in the Marketing team which has just shoot in its feet another time.

Sorry again for the rant, but this really pissed me off

[entropy]

The support for ATI Radeon X1000 (R500) through Radeon HD 4000 series (R700) now lives in an AMD Catalyst legacy branch that may see updates once in a while.

So that means that R500 will eventually see a "legacy" update for the first time since Catalyst 9.3?
That would come really unexpected..

[blackiwid]

OK, I have to ask.. how can you possibly look at market trends and our own presentations (nobody seems to have commented on the open source plans discussed at AFDS yet, which is surprising) then conclude that "fglrx getting more important" means that the open source driver is becoming less important ?

so you plan to hire some new developers soon? Nice ok than I with that.

Or do you see another way to make the fglrx faster better (or is it just getting more important and nothing results in that?) and also then develop the open driver faster.... because even if you would develop in the same speed in the future the free drivers and accelerate the development of the catalyst it would be fatal, because even if it takes very long in linux the free drivers have to replace the closed ones.

I did never hear that amd will switch to the free drivers in the long run, but I heared here the oposite that the free drivers are only a plaything or so, that they never will replace the blobs and thats not acceptable.

Did all the other OSes disappear from the market while I was travelling ?

yes tell that your coworkers so they know it only exist linux now, they only should write the drivers for that os because the other ceased to exist ^^


BTW I was not on AFDS I dont even know that it exists and I dont live in america (its there isnt it?) I can only comment what the press (phoronix) writes here, if thats not complete its not me to complain ^^

[droidhacker]

There is no source link to publicly cite.

BAHAHAHAHA!!
More super high quality journalism out of phoronix.

Also, you are MISREPORTING the "rumor".
The rumor isn't discontinuation of the driver to focus on open source... its discontinuation of R600/R700 to focus on open source.
Oh wait... that's exactly what AMD said, isn't it?


Quite frankly, they SHOULD discontinue blob drivers on ALL operating systems.

[Kano]

I do not really like news with unknown sources. Hard to verify, maybe brigdman can put some light onto this issue? Who is Jammy Zhou and why should that be important what he says? He is definitely not the ceo, but what does he?

[its-techs]

So no mention what so ever on how you obtained this quote? Why should we believe you? Something like "in an email to blah he mention" or "via phone interview person X said" would have sufficed. Glad to see you're working so hard at keeping phrononix a respectible news source that works hard to cite sources and provides external link for more info.

On that note was there ever even a rumor that catalyst was gonna be dropped for linux? You would think that would be be news. Nope just another excuse for Michael to half ass an article about nothing.

I was the one that posed the question to AMD, and both Michael and I saw the answer to the question because it was on a closed list. The reason Michael is not talking more about this is because the entire closed list is under NDA and there are legal issues involved. I should also add that the context of the question was such that Jammy Zhou understood that his reply to this specific question would be shared.

The reason I'm here now is to provide confirmation on his behalf to refute accusations of this type.

[Kano]

And WHO answered it?