VIA Fails With KMS/3D, But Has Yet Another X Driver

[Adarion]

*sigh* Every year the same story.


By the way:
I DO use VIA GPUs. (as far as you can speak of "use" here).

I have an ECS G320 Laptop which is a nice machine in many terms but carries a CLE266 and I'm happy to have at least half-assed MPEG2 acceleration there. Furthermore I own a mainboard with a C7 Eden (nice thing) which has a CN700 and it's the same problem here. Some snapshots from openchrome even had strange artifacts during MPEG2 playback. It was intended back at the time to be a silent low power HTPC but that hasn't come to life yet. Sad, sad thing.
And sorry Luc, but your driver stopped compiling with xorg-server a good time ago, then Gentoo kicked the official package. Though I remember it might have been the first one I used some years ago on the CLE266.
But afair I never ever had real 3d, maybe "accelerated" glxgears which might really have been on the GPU due to a years old software in mesa but that wasn't much faster than running the stuff in a VESA driver in software rasterizer mode.
I should have gotten my money back while there was still time, but I am being a bearing person still hoping for things to improve.

On the article: Yes, it is sad that one of the smallest areas in Linux GPU stuff is so diverged. And I see no sense in drivers which cut out more and more features until the behave like a VESA driver. This is of no use to me. The few devs should stand together.
And yes, openchrome some corrections of typos in code comments recently. *cough*
But of course I don't blame the devs actually since it must be frustrating to have no real contact with VIA. No specs, no code, no support, nothing.


Well, the only thing I can do is keep with the things that work and worked for me in the past.

[Kristaps1]

I am using VIA CPUs (VIA Nano and VIA C7-M) and chipsets (VX800) on both of my computers. They are very good products in low power class. Unfortunately because of lack of open drivers, i am using Windows 7 on both machines. When i bought both computers, VIA gave to community signs, that chipset will become open-source supported, but it didn't happened so. Truly i am dissapointed.

The only reason, which i see? why we do not want to give their driver code to community is innovative architecture, but is it so? They already published programming guides, which gave to their competitors access to their architecture. So - where is the problem? Is it fear that process of driver development will go out of hands of VIA/S3? For corporate customers there always will be VIAs drivers. Is it fear that Intel or AMD will steal their ideas? Come on, out there already is high performance SOC CPUs, which can easy outperform VIAs CPU/chipsets! If they are legally binded with some strange exclusive agreement, which dictates that developing of Linux drivers is deligated to some outsource company, and that VIA is not allowed to give this code to other developers - so say that clearly, and change developer company, when realising next chipset.

[curaga]

The docs are bare register info, my unichrome code is the real source of info for writing suitable modesetting code.

And i personally find that GSoC results are usually overrated (everyone talks about it solving so many problems, in future, but no one really goes and evaluates such things publically; no-one compares up front expectations with actual results), and a proper modesetting driver is way too much work for GSoC anyway. VIA hw is not glint.

Would it still be too much work, if you consider it could be a port from your unichrome code?

[libv]

I was under the impression (almost) nobody uses VIA anymore.
http://www.phoronix.com/scan.php?pag..._results&num=1

No-one of the phoronix users, yes. Even though it is dwindling quickly, the actual VIA marketshare is higher than the figures revealed by the phoronix census.

[bug77]

I was under the impression (almost) nobody uses VIA anymore.

http://www.phoronix.com/scan.php?page=article&item=lgs_2010_results&num=1

[libv]

Considering there are some docs already out, why not set a GSoC target for Via KMS?

It would very likely be working before anything appears from Via...

The docs are bare register info, my unichrome code is the real source of info for writing suitable modesetting code.

And i personally find that GSoC results are usually overrated (everyone talks about it solving so many problems, in future, but no one really goes and evaluates such things publically; no-one compares up front expectations with actual results), and a proper modesetting driver is way too much work for GSoC anyway. VIA hw is not glint.

[curaga]

Considering there are some docs already out, why not set a GSoC target for Via KMS?

It would very likely be working before anything appears from Via...

[libv]

On desktop systems the bad via driver support does not hurt that extreme as you could add a dedicated card but for laptop users thats too bad currently. Most drivers are so stupid that you have to force h/v sync, screen res + display size even when they "work". It must be impossible for via to support they onboard solutions with kms with correct edid frequencies. I do not even speak of 3d...

Heh, my unichrome driver was the very first to do free modesetting. This was not what the openchrome people liked though (and also not what people like keithp and others liked).

But yes, this driver is many steps into the wrong direction, again.

[libv]

Clueless VIA shoots themselves in the foot yet again. Thanks, Michael!

This is not entirely VIA. This is Jon Nettleton, one of the least useful and the most noisy of the openchrome forkers. Also one of the people who was very keen to join in the "we love VIA for finally claiming they will work with open source, after more than 5 years of others trying to push there and after RadeonHD freeing ATI"-frenzy at openchrome.

I won't even bother to look into the commits, but the fact that the commits were made by root, makes the whole thing rather questionable, even from this very superficial pov.

Another Dodo.

[Kano]

On desktop systems the bad via driver support does not hurt that extreme as you could add a dedicated card but for laptop users thats too bad currently. Most drivers are so stupid that you have to force h/v sync, screen res + display size even when they "work". It must be impossible for via to support they onboard solutions with kms with correct edid frequencies. I do not even speak of 3d...