artefacts on 855 graphic

artefacts on 855 graphic

[Alexey Fisher]

Hallo all,

i know it is known issue that 855 is not really working with current
driver. But i wont to know if there is work in progress or any interest
of back reports?

I just had access to the laptop of my friend. It is in really bad shape:
by default vesa driver is used - even simplest video take 100% cpu
if i force intel driver, compiz make really wired artefacts - each
window can just disappear
if i disable compiz. it will sort of work, but has some other artefats -
part of old interface mixed with new, to make it disappear i need to
minimize and maximize window.

Regards,
	Alexey

Re: artefacts on 855 graphic

[Daniel Vetter]

On Sat, Nov 27, 2010 at 10:07:24AM +0100, Alexey Fisher wrote:
> Hallo all,
> 
> i know it is known issue that 855 is not really working with current
> driver. But i wont to know if there is work in progress or any interest
> of back reports?

Well, it's i8xx in general that's in a very sorry state with gem.
Unfortunately there's no easy fix available, even within intel no one
knows anymore how to work-around these problems (hw people designed these
chips approx 8 years ago and moved on). I have ideas that might fix these
problems. Unfortunately this requires rather massive code rewriting. But
I've been (very) slowly moving towards this in the past few months. Don't
hold your breath, though.

Meanwhile Chris Wilson's shadowfb support should give you Xv accel +
reasonable fast 2d (cpu-rendered, but the gpu was never really faster for
2d on these chips, anyway).

-Daniel
-- 
Daniel Vetter
Mail: daniel@ffwll.ch
Mobile: +41 (0)79 365 57 48

Re: artefacts on 855 graphic

[Alan W. Irwin]

On 2010-11-27 15:49+0100 Daniel Vetter wrote:

> On Sat, Nov 27, 2010 at 10:07:24AM +0100, Alexey Fisher wrote:
>> Hallo all,
>>
>> i know it is known issue that 855 is not really working with current
>> driver. But i wont to know if there is work in progress or any interest
>> of back reports?
>
> Well, it's i8xx in general that's in a very sorry state with gem.
> Unfortunately there's no easy fix available, even within intel no one
> knows anymore how to work-around these problems (hw people designed these
> chips approx 8 years ago and moved on). I have ideas that might fix these
> problems. Unfortunately this requires rather massive code rewriting. But
> I've been (very) slowly moving towards this in the past few months. Don't
> hold your breath, though.
>
> Meanwhile Chris Wilson's shadowfb support should give you Xv accel +
> reasonable fast 2d (cpu-rendered, but the gpu was never really faster for
> 2d on these chips, anyway).

Frankly, it harms Intel's Linux reputation that this regression in 3D
support for old chips has been allowed to develop.  Shuttle was kind
enough back in 2004 to donate two of their shuttle boxes to the LUG I
happened to belong to at that time.  Those boxes had Intel Extreme
Graphics 2 chipsets (I assume 855GM's), and low-end 3D games worked
well with the Intel driver then according to my own experience as well
as the wonderful Linux reviews that particular shuttle box (the
SB62G2) was getting at the time.  I have felt positive about Shuttle
and Intel ever since that good experience, but this regression in 3D
support for old hardware is giving me second thoughts about Intel.

I realize it would probably take some modest additional personnel
resources from Intel to support those old devices (probably with a
completely separate and minimally maintained driver that was forked
from the last edition of the Intel driver that worked properly for
those devices since that old hardware appears not to be compatible
with GEM). The allocation of such personnel resources from Intel would
give their Linux reputation a much-needed boost since extended
software support times are a big selling point for hardware that is
run with open-source device drivers.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: artefacts on 855 graphic

[Clemens Eisserer]

Hi Alan,

> Frankly, it harms Intel's Linux reputation that this regression in 3D
> support for old chips has been allowed to develop.

Well, I also own a 855GM powered laptop, but I have to add that I
understand intel's descision to not devote a lot development resources
to it anymore. 3D on that hardware doesn't make a lot of sence except
you need to run legacy apps that depend on it, and about running
compiz - even their next-gen chip i915 doesn't run Vista's Aero
composition manager.

Furthermore its not different from what other GPU vendors do with
legacy hardware.
Take nvidia for example - their old chips don't get driver updates
anymore, and the more-or-less official recommendation is to use nv
instead. I have to use the reverse engineered nouveau driver on that
machines (and I am really happy with it) to get at least useable 2D.
And for AMD, the HD2100 powered mainboard I bought exactly one year
ago doesn't receive any driver updates for windows anymore too, one
year after I bought it.

- Clemens

Re: artefacts on 855 graphic

[Alan W. Irwin]

On 2010-11-28 00:01+0100 Clemens Eisserer wrote:

> Hi Alan,
>
>> Frankly, it harms Intel's Linux reputation that this regression in 3D
>> support for old chips has been allowed to develop.
>
> Well, I also own a 855GM powered laptop, but I have to add that I
> understand intel's descision to not devote a lot development resources
> to it anymore.

Hi Clemens:

Yeah, I agree that Intel should not devote huge resources to this.
That was why I was talking about Intel supporting a minimally
maintained fork of an old working driver that did have good 3D support
for legacy Intel hardware.

> 3D on that hardware doesn't make a lot of sence ....

I disagree with you there.  I know from experience that 855GM hardware
would still be fairly powerful with the right software support.
Low-end 3D games worked well on that Shuttle box for me, and
it also received good reviews for 3D elsewhere with the caveat of not
expecting much for demanding high-end 3D games.  I didn't try it for
3D desktop effects at the time (I am not sure they even existed in
2004) but assuming the software and hardware demands of current 3D
desktop effects are similar to those of low-end 3D games, then a
minimally maintained fork of the old Intel driver that resurrected the
3D responsiveness I remember for that hardware should be more than
adequate to support 3D desktop effects.

> Furthermore its not different from what other GPU vendors do with
> legacy hardware.
> Take nvidia for example - their old chips don't get driver updates
> anymore, and the more-or-less official recommendation is to use nv
> instead. I have to use the reverse engineered nouveau driver on that
> machines (and I am really happy with it) to get at least useable 2D.
> And for AMD, the HD2100 powered mainboard I bought exactly one year
> ago doesn't receive any driver updates for windows anymore too, one
> year after I bought it.

Yup, this is the known downside of proprietary software drivers, but
why would Intel want to copy that bad practice?  People like me want
to be free to "use up and wear out" our computer devices rather than
being the victim of forced hardware upgrades.  That's one of the big
advantages of open-source driver software from a user's perspective. 
For example, the open-source nouveau for Nvidia and Radeon for AMD/ATI
drivers do support legacy hardware. Why would Intel not want to follow
that good open-source practice?

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: artefacts on 855 graphic

[Clemens Eisserer]

Hi again,

> Yeah, I agree that Intel should not devote huge resources to this.
> That was why I was talking about Intel supporting a minimally
> maintained fork of an old working driver that did have good 3D support
> for legacy Intel hardware.

I guess a big problem with Linux are ever changing APIs/ABIs. If
you've created a driver for Windows Vista back in 2006 - you still
have no problems using the same binary with Windows-7, including Aero
and DirectX.

If you've created a ddx/mesa stack in 2006, it needs tons of work to
function properly with a modern distribution.
There are still releases of intel-2.4 and old mesa versions out, which
I guess would work fine when ported to latest xorg/kernel releases,
the question is simply who will maintain this stuff.

Good luck, Clemens

Re: artefacts on 855 graphic

[Rémi Cardona]

Le 27/11/2010 10:07, Alexey Fisher a écrit :
> Hallo all,
> 
> i know it is known issue that 855 is not really working with current
> driver. But i wont to know if there is work in progress or any interest
> of back reports?

One of the bugs where 8xx issues are tracked is this one :

https://bugs.freedesktop.org/27187

In a nutshell, 8xx are in a sad state because the actual hardware was
poorly designed *and* it was validated using the available Windows
version at the time : XP.

The XP graphics stack does not push the driver/hardware like the modern
X stack does so all those bugs went unnoticed until Vista and DRI2/GEM
came out.

As for Chris's shadowfb effort, it seems to be held up in Limbo since
the 8xx user base is getting smaller everyday.

As far as my personal experience goes, the current git stack works ok on
my 855GM, albeit with a couple of rendering glitches with heavy Flash
applets. YMMV.

I'm of course just as annoyed as you to see (once) popular hardware with
such uneven levels of support after all these years.

Cheers,

Rémi