VIA Publishes 2D/3D Documentation, Partners With OpenChrome

[phoronix]

Phoronix: VIA Publishes 2D/3D Documentation, Partners With OpenChrome

Earlier this year VIA announced they wanted to join the open-source bandwagon by establishing an open-source driver development initiative, releasing documentation and source-code, and to better engage with the Linux community at large. They have made a few small steps over the past few months, but today they have made their largest open-source contribution yet by releasing four programming documentation guides that cover the video, 2D, and 3D programming for their Chrome 9 graphics processor. In addition, they are now partnering with the community-spawned OpenChrome developers.

http://www.phoronix.com/vr.php?view=13134

[IanW]

Waitaminute!

First Adobe release a 64-bit Linux Flash player, now this happens THE SAME WEEK? :eek:

Next you'll be telling us Duke Nukem Forever got released, with a LINUX CLIENT! :rolleyes:

[conholster]

Why do you always use the "bandwagon" word? From my experience it has a negative meaning and is usually used to describe copycat bands (that record crappy music).

[Eragon]

Well, now Nvidia is the only one left to "open up", but they already have a good closed-source driver. (which is why I just bought a 512 MB GeForce 8 to replace the crappy Nvidia card that came with my comp).

Plus, with gallium 3d merget into nouveau, we can look forward to a working 3d nouveau driver, altough to satisfy me, it has to be exactly as stable in games as the nvidia closed-source one, also with propietary games.

[Kano]

Well as usually the date of releasing the specs is not the same date of a working driver... So no need to jump on the desk.

[Svartalf]

Well as usually the date of releasing the specs is not the same date of a working driver... So no need to jump on the desk.

Heh... Depends on whether or not the info's...

1) All there...
2) Complicated or not...

I'm checking into the "all there" part right now. The 3D document's only 157 pages in size for the VX700 3D programming doc (First thing I noticed, they did it under a Creative Commons license...definitely a change of pace...)- if it's all there, it means it's a somewhat simpler chip (explains the performance deltas, no?).

We'll see. It might be up and going against the old DRI framework quickly. I'd rather see a Gallium backend myself, but that may have to wait until the final cuts of the Intel Gallium3D support show up for that one.

[libv]

Why do you always use the "bandwagon" word? From my experience it has a negative meaning and is usually used to describe copycat bands (that record crappy music).

Wellll...

In oktober 2007 I was once again contacted by yet another VIA person (this was not the first time, but all of such contacts were an effective waste of time) who wondered what ideas i had for VIAs open source strategy. This was of course many months before their marketing team accidentally put all of the current stuff in motion. The first thing i said was that i was now working for SUSE, and i was actively helping a hardware vendor doing the right thing, becausei was of course very heavily involved with the AMD project, and them making docs available...

This was followed by a long silence at the other side of the line, something which was very telling.

So bandwagon is a very correct choice of words from where i sit.

[Svartalf]

This was followed by a long silence at the other side of the line, something which was very telling.

So bandwagon is a very correct choice of words from where i sit.

I think they might have caught a clue with Harald over there. I'm not done yet with the perusal of the register specs, but it's looking remotely possible that they gave out enough info to at least get a full-on GL 1.3 level driver done with the info provided. A proper 1.5/2.0 capable renderer would probably have to rely on the Gallium3D framework and the LLVM to produce CPU-centric Vetex shader support at the least, based on the cursory reading here.

So far, it's looking a lot better than the story we got last pass from them. This may actually be a useful doc release from them.

[jeffro-tull]

It is also a bit odd that their leading priorities are multi-head and RandR support, when those aren't really huge sought after features for IGP customers compared to say improving the 3D support or improving video acceleration.

I know why they're worried about RandR and multi-head support. Their Nano reference boards are finally making it to OEM's. Meaning there should finally be a "second wave" of VIA-powered netbooks. I think I speak for all netbook owners when I say it's nice when the VGA-out actually works.

[calica]

I know why they're worried about RandR and multi-head support. Their Nano reference boards are finally making it to OEM's. Meaning there should finally be a "second wave" of VIA-powered netbooks. I think I speak for all netbook owners when I say it's nice when the VGA-out actually works.

That is exactly the reason. I'm rather disappointed that Michael couldn't see it. Maybe he's too desktop-centric?

In anycase, 3d is still important for netbooks. Compiz effects are incredibly useful on small screens.

[jeffro-tull]

In anycase, 3d is still important for netbooks. Compiz effects are incredibly useful on small screens.

Really?

Granted, I'm not at all familiar with Compiz. I've used Kwin's effects (KDE4.1.x), but they seem to be more "hey, look what this little guy can do!" rather than actually useful. If you know something I don't, then by all means...

[myxal]

I definitely agree with jeffro-tull...

I, for one, welcome the RandR-oriented efforts. Sure, Compiz is cool and all that, but if you can't even set the resolution and/or rotate without restarting X, that's a major annoyance/obstacle to ordinary desktop users right there.
Not being able to run compiz (I can't believe someone would suggest 3D games for a VIA card) is hardly an issue in comparison.

EDIT: Can someone explain to me what the difference between open- and uni-chrome is? The openchrome wiki mentions a fork of an experimental branch in unichrome, but no details as to why it was forked or what feature difference there is, if any.

Eragon: I hate to break it to you, but there's still SiS - quite a popular chipset/IGP on TONS of cheap (not just "good-deal cheap", I mean "bottom-of-the-barrel cheap") desktops and laptops. The situation here is even worse than nVidia - you have an open 2D driver and a "quantum state" binary 3D driver which is known to exist, but no one has managed to obtain it (the person who claims to have developed it is allegedly barred from distributing it, while the SiS website tells you to go to the OEM vendor who, as expected have no clue about some Linux/X binary 3D driver for SiS graphics). Lastly, there are some 3rd party efforts to add 3D to the open driver, where the mileage varies too much to be useful.

[Eragon]

I definitely agree with jeffro-tull...

Eragon: I hate to break it to you, but there's still SiS - quite a popular chipset/IGP on TONS of cheap (not just "good-deal cheap", I mean "bottom-of-the-barrel cheap") desktops and laptops. The situation here is even worse than nVidia - you have an open 2D driver and a "quantum state" binary 3D driver which is known to exist, but no one has managed to obtain it (the person who claims to have developed it is allegedly barred from distributing it, while the SiS website tells you to go to the OEM vendor who, as expected have no clue about some Linux/X binary 3D driver for SiS graphics). Lastly, there are some 3rd party efforts to add 3D to the open driver, where the mileage varies too much to be useful.

Yeah I forgot about SiS. However, I think the situation with Nvidia is about perfect... It's a lot easier to install / update their closed-source driver then the open-source radion driver, and the 3d performance is much more stable. IMHO, the open-source drivers for ATI r500 and below hardware simply aren't good enough yet if you want to game: not all games run on them, and those that do often aren't stable.
Getting them to work is a pain in the ****.

Moral of the story: I rather have the easy-to-install, working, nvidia closed driver :)

[calica]

Really?

Granted, I'm not at all familiar with Compiz. I've used Kwin's effects (KDE4.1.x), but they seem to be more "hey, look what this little guy can do!" rather than actually useful. If you know something I don't, then by all means...

The expose like window switcher is nice with maximized windows. On desktops I prefer a little overlap so I can switch windows without using the taskbar. I do agree 3D is pretty low priority but luckily 3d desktop effects aren't too demanding and a basic driver is sufficient.

[myxal]

...

Moral of the story: I rather have the easy-to-install, working, nvidia closed driver :)
Well, yeah.. somewhat. I hope you realize that's a false dichotomy - the PITA that comes with upgrading opensource drivers is not inherent to them being opensource.

Also, I'd like to know just how many Linux users use the system for games (and no, a survey on Phoronix doesn't count as unbiased)? Apart from Compiz, which I can do without, the only apps requiring 3D I use are googleearth and graphic output of various BOINC projects. I find Google Maps a sufficient replacement for the former in most cases (streetview still isn't available for my area as are many 3D building models) and can completely live without the latter.

Truth be told, if I couldn't live without every last feature the hardware offers, I'd just bite the bullet and *gulp* use Windows(C)(R)(TM). It's annoying to see even opensource projects giving out ready-to-go packages for Windows users while expecting the distro maintainers to either catch up or users compiling the stuff themselves (I'm pointing at you, VLC!)

[sloggerKhan]

It is also a bit odd that their leading priorities are multi-head and RandR support, when those aren't really huge sought after features for IGP customers compared to say improving the 3D support or improving video acceleration.




Actually, if VIA graphics are in use on laptops and netbooks, I can see why randr and multihead output could be a priority. As low power laptop user, being able to hotplug a projector for a presentation is many times more important than being able to play movies or 3-D apps.
Oops, repeat idea.

[Svartalf]

Indeed... I'd say RandR support would be highest priority, followed by scaling, followed by video playback, followed by 3D in that order.

It's the most critical to least- and it's the strongest to the weakest plays for these chips.

UniChrome/UniChrome2 are, at best, OpenGL 1.3 capable cores- no shaders, fixed functionality, and it looks like no vertex processing present.

Chrome9 looks to be akin to a GMA900- fragment shader support and fobbing the vertex path off on to the CPU. Chrome9's got some potential, but it won't be fully realized without Gallium3D.

I'd be worrying about RandR first too... ;)

[curaga]

Can someone explain to me what the difference between open- and uni-chrome is? The openchrome wiki mentions a fork of an experimental branch in unichrome, but no details as to why it was forked or what feature difference there is, if any.Unichrome is clean and stable, but supports less chips and has less functionality (no XvMC, no Via's own mpeg4 accel). Openchrome has these, but might be buggier.

Oh, and the SiS driver, don't even talk about it. It has not been updated for the later chipsets, and while it will modeset, it has frequent 2d glitches (random lines on screen, sometimes small, sometimes over the whole screen like a spider web). Vesa at least handles them fine, but forces 60HZ, awful for crt's.

[schlobinux]

I think they might have caught a clue with Harald over there. I'm not done yet with the perusal of the register specs, but it's looking remotely possible that they gave out enough info to at least get a full-on GL 1.3 level driver done with the info provided. A proper 1.5/2.0 capable renderer would probably have to rely on the Gallium3D framework and the LLVM to produce CPU-centric Vetex shader support at the least, based on the cursory reading here.

So far, it's looking a lot better than the story we got last pass from them. This may actually be a useful doc release from them.

The docs are a big and visible step forward, not cheap talk. Unfortunately, it lacks informations on the shaders, but openchrome requested these informations from VIA. They are currently gathering and checking them for 3rd party IP before a future public release.

It takes time and a lot of diplomacy to build a healthy relationship with a manufacturer. Harald is doing a great job at driving VIA toward better understanding and support of the opensource community and this indeed helps openchrome and other opensource project a lot.
On the other hand, I certainly believe constant mud-throwing is not helping the community a single bit, libv. I prefer to follow a more pragmatic approach.

[schlobinux]

Unichrome is clean and stable, but supports less chips and has less functionality (no XvMC, no Via's own mpeg4 accel). Openchrome has these, but might be buggier.

Openchrome supports all VIA IGPs, unichrome only supports few of them. Also, XvMC is a key feature, especially when attached to a low power but also low consumption and low thermal envelope CPU.

[libv]

Openchrome supports all VIA IGPs, unichrome only supports few of them. Also, XvMC is a key feature, especially when attached to a low power but also low consumption and low thermal envelope CPU.

Hrm... But what were the important things you said are highest on your TODO list again? Aren't those a really close match for what i wanted to achieve with unichrome and the things the people who ran to opechrome refused to prioritise as such and therefor forked away?