Clarifications On Poulsbo's Gallium3D Driver

[phoronix]

Phoronix: Clarifications On Poulsbo's Gallium3D Driver

Yesterday we reported on a new Linux driver coming for Intel's Poulsbo chipset that is currently notorious on Linux. This graphics processor is found in many Atom-powered netbooks, but its binary driver is a mess...

http://www.phoronix.com/vr.php?view=NzY2NA

[elanthis]

If it's not Open, it's not interesting as a driver. I'm all for (quality) proprietry software on Linux, especially in spaces with no FOSS counterparts of equal quality (eg games), so long as it is not part of the core stack.

[stan]

Splitting a binary blob into three binary "plugins" is like rearranging the decks on the Titanic. The fact that Intel is using PowerVR for their next-gen chipsets shows they're not serious about open source software. Else they would have used something that they could open source, like their in-house chipsets.

I hope the three-blobbed DRM gets rejected from the kernel, again.

[hubick]

At this point my very expensive Panasonic CF-U1 Menlow MID is a BRICK because I don't have time to sort all this crap out. I'm pissed at Intel because I thought their name on the video chip meant it would Just Work under Linux. At this point, I will settle for distro's (Fedora in my case) shipping with any open source driver that can set the damn screen resolution right, and let me run a web browser and my desktop. If I tolerated closed source crap, I would just stick Vista back on the thing.

[mgc8]

I guess this goes to show that being "open" and "free" are not top priorities for any company, not even Intel, and everything they do is just for PR. This seemed promising in the beginning, yet after the details settled it's sad, sad, sad (once for each binary blob that is needed for the "open" part to function)... There must certainly be a lot of high-quality intellectual "property" being protected here -- nVidia and ATI must be shaking at the thought of the awesome PowerVR chip, and probably spending millions to find it's well-guarded secrets :-/

I hope the three-blobbed DRM gets rejected from the kernel, again.

I hope they have the decency to not even try again. They should admit the fact that it's a blob and deal with it clearly like nVidia does, don't try to make it seem open when it obviously isn't.

Seeing this only makes my respect and admiration for AMD and what they did to grow each day (that coming from someone who's not using any of their products atm). Yes, they're in it for the PR as well, but at least they clearly state what is open, what is closed, what could be opened in the future and so on. This whole Poulsbo mess is unbecoming for a company like Intel.

[MostAwesomeDude]

I don't work for Intel.

Intel has always intended to release open-source drivers for this chipset; they've been working behind closed doors for months upon months trying to get Imagination to release docs. The Poulsbo situation may be several heaping gobs of shit, but Intel's been working hard to increase the documented-code-to-shit ratio. Blaming them for this shit is taking a simple and untrue view of the situation.

[V!NCENT]

I don't work for Intel.

Intel has always intended to release open-source drivers for this chipset; they've been working behind closed doors for months upon months trying to get Imagination to release docs. The Poulsbo situation may be several heaping gobs of shit, but Intel's been working hard to increase the documented-code-to-shit ratio. Blaming them for this shit is taking a simple and untrue view of the situation.

First of all thanks for the insight.

I never heared of a company called Imagination, but if I was in charge of Imagination I would think twice about not listening to a huge customer like Intel because it's not like there will be any other serious customers for that kind of shitty GPU's, ever.

[rohcQaH]

makes me wonder why intel used those GPUs in the first place, if driver support is aproblem. Why didn't they install their own GPUs?

[Kano]

When you sum it up, the only opensource parts are the DRM and and the 2d driver. I guess they think thats enough to push the DRM mainline ;)

[zoomblab]

I don't care much whether my drivers are closed or open. I just want them to be solid.

Interestingly, in that regard, I think that if all drivers were closed, the kernel developers would be forced to keep stable ABI and APIs. That in turn would mean that I wouldn't have to pray for everything to work as well as before whenever I have to upgrade to a new distro version. It would also mean that the drivers are developed by the same entity that made the hardware in the first place and not some random guy from another organization.

[Kazade]

First of all thanks for the insight.

I never heared of a company called Imagination, but if I was in charge of Imagination I would think twice about not listening to a huge customer like Intel because it's not like there will be any other serious customers for that kind of shitty GPU's, ever.
Imagination Technologies, they (amongst other things) designed the GPU in the Dreamcast. I went for an Interview there (worst interview ever, seriously, I messed up so bad) and I'm glad I didn't get it, it was only afterward that I heard how anti-opensource they are.

Their whole company is based on selling IP to companies like Intel, they don't product any hardware, they just design it.

[RoboJ1M]

Yeah, they are like ARM now.
They used to make their own hardware though. (KyroII FTW!!)

I believe I agree with the Most Awesome Dude, I find it very likely that it's just a very slow process getting it all done.

And Imagination are very small, I doubt they have the resources to throw at it like AMD/ATI does.

Oh and, "it's not like there will be any other serious customers for that kind of shitty GPU's, ever".

Imagination are VERY like ARM in that they design embedded GPUs that are very VERY low voltage and you probably already own several in your various mobile phones, PDAs, handheld consoles, etc. They may not make it and sell it to you but their global volumes have to be pretty big these days.

Finally: I'm with the Just Works camp for now. I'll have it working first, open source eventually. Because eventually it will all be open source.

J.

[cb88]

Just out of curiosity but will this driver also work for the Pandora(http://openpandora.org/) or is there no arm version of the binaries?

[val-gaav]

I don't care much whether my drivers are closed or open. I just want them to be solid.

...and working with a latest Xorg, kernel etc. which is really not possible with any blob. Nvidia is the best here but the support here is also not from day one.

The other thing is legacy drivers as it is now for ATI r100-r500 ... In this case nvidia is once again the best giving updates also for old legaqcy drivers, however for example my riva tnt2 will not work with new Xorg and nvidia blob. Sure it's an old card but it's in a PC that does not need a better card...

Interestingly, in that regard, I think that if all drivers were closed, the kernel developers would be forced to keep stable ABI and APIs.


Here you have Linus opinion about it :
http://www.phoronix.com/forums/showpost.php?p=78838&postcount=78
and here opinion about why having stable ABI and API is not so good idea :
http://lxr.linux.no/#linux+v2.6.31/Documentation/stable_api_nonsense.txt

[Remco]

This is Intel's fault. They accepted crap. I know this, because they are selling crap. Even now, they are trying to trick me into buying useless plastic with their advertisements.

But regardless of whose fault it is, I'm not buying Intel until this Poulsbo hardware is long gone. They can expect me to look at their hardware again in at least 5 years.

[zoomblab]

@val-gaav
Like with everything in life there are benefits/costs. Linus is basically saying "in any case I am gonna go ahead and do MY THING" which sounds pretty irresponsible to me. That is not the way to make friends and it shows.

[Remco]

@val-gaav
Like with everything in life there are benefits/costs. Linus is basically saying "in any case I am gonna go ahead and do MY THING" which sounds pretty irresponsible to me. That is not the way to make friends and it shows.
How does it show? Hardware support now is better than ever. Linux kernel development is a continuous explosion in rate of change and number of developers.

[Max Spain]

My 2c: Intel is putting its brand name on these products. If they fail to deliver, it is INTEL'S problem, not Imagination whatever company because Intel and Imagination would have signed a contract they both agreed upon. If Imagination is not fulfilling their end, they would have breached their contract, and Intel would not be using any more of their products. If Intel didn't stipulate OSS friendly provisions, then Intel didn't care about their customers.

If it is true that Intel is going to continue using their chips, then the problem is obvious (http://www.motivatedphotos.com/?id=44745).

[V!NCENT]

If Intel didn't stipulate OSS friendly provisions, then Intel didn't care about their customers.
Well at least Intel is open sourcing what they can and they are, according to MostAwesomeDude, putting a lot of effort into it.

I just hope that none of this Imagination Property is inside Larrabee :(

[gururise]

Like most others, I just want a working, solid driver for the Poulsbo chipset on Linux! Does anyone know if this will make it into the Moblin 2.1 final release???

[mgc8]

I don't care much whether my drivers are closed or open. I just want them to be solid.
Interestingly, in that regard, I think that if all drivers were closed, the kernel developers would be forced to keep stable ABI and APIs. That in turn would mean that I wouldn't have to pray for everything to work as well as before whenever I have to upgrade to a new distro version. It would also mean that the drivers are developed by the same entity that made the hardware in the first place and not some random guy from another organization.

And it would also mean that you'd be running Windows instead of Linux. Seriosuly, what's the point of a free OS if the drivers are closed-source?! Just look at the kernel source, more than half of that is drivers! If all of that was closed-source, you'd be left with a pathetic shim of a kernel that wouldn't deserve to be called "open" or "free". Open-source drivers are a prerequisite for a free OS -- if you don't care, you can certainly use any of the alternatives which follow different philosophies.

All this crap about "intellectual property" when it comes to a silly little piece of code that does basic interfacing is aggravating. Yes, custom drivers with performance-optimized features are useful in specific areas like 3D, however keeping it all secret is just corporate paranoia. But what if their competitors found out about...? guess what, they already know all there is to know. You're just screwing legitimate customers.

@Max Spain -- you said it!

[jonwil]

Knowing how anti-open source Imagination are, I doubt that even if Intel said "we guarantee to purchase/make a million PowerVR chips (i.e. pay royalties for a million PowerVR cores) but only if you will open your specs", it wouldn't be enough to make Imagination open up.

[Apopas]

And it would also mean that you'd be running Windows instead of Linux. Seriosuly, what's the point of a free OS if the drivers are closed-source?! Just look at the kernel source, more than half of that is drivers! If all of that was closed-source, you'd be left with a pathetic shim of a kernel that wouldn't deserve to be called "open" or "free". Open-source drivers are a prerequisite for a free OS -- if you don't care, you can certainly use any of the alternatives which follow different philosophies.

All this crap about "intellectual property" when it comes to a silly little piece of code that does basic interfacing is aggravating. Yes, custom drivers with performance-optimized features are useful in specific areas like 3D, however keeping it all secret is just corporate paranoia. But what if their competitors found out about...? guess what, they already know all there is to know. You're just screwing legitimate customers.

@Max Spain -- you said it!
Exactly!:)

[Dard]

Personally, I really like to use open source drivers, but if any CS drivers are good, I am willing to use them, even though I would rather use an open source driver.

However, the current Poulsbo driver is pretty much the ultimate example against CS drivers you can find.
Usable only on one kernel version, either VESA slow and stable or fast and totally unstable. (with MigrationHeuristic greedy it crashes every time I wake up my notebook from suspend)
And I don't even care about any 3D. Sure, it's nice, but not necessary. But even 2D is buggy.
And you can't even ask anybody, because there is no support on either Linux or Windows. Oh yeah, Windows, the drivers there are also bad. I only tried to run two very old games on it, one using OpenGL and one DirectX. The DirectX one didn't run correctly and there isn't even OpenGL anywhere to be found.

So, the Poulsbo drivers are really a mess and it looked like it would never change.

Now there is information that things may improve.
I had already given up hope that there would be a fixed driver for Poulsbo on Linx *ever*!
Right now I am happy for every support, being closed or open source. If I get *something* I will not complain that I don't get *everything*.

If it will ever get so far, I will not complain anymore. Maybe we get a working 3D driver, maybe we don't. If the CS driver is good, I will use it. If it isn't, I won't. There are a lot of very good CS drivers out there, so "Closed Source" and "good" doesn't have to be mutally exclusive.
Right now I don't have the choice, I have to use a bad driver. Hopefully at least the 2D will get better.

All I'm trying to say is: Those were mostly good news. Don't make it sound too bad.

[gururise]

Any news on when we will see a release version of this? Perhaps the Moblin people have a timeline as to when it will be included in Moblin!? I've been dying to try Mobilin on my new Atom Z530 netbook.