The ati petition

[Alex Deucher]

On Sat, 04 Sep 2004 19:04:30 +0100, Hamie <hamish at travellingkiwi.com> wrote:
> Alex Deucher wrote:
> 
> >On Sat, 04 Sep 2004 11:20:56 +0100, Hamie <hamish at travellingkiwi.com> wrote:
> >
> >
> >>Alan Cox wrote:
> >>
> >>
> >>
> >>>On Gwe, 2004-09-03 at 16:23, Gene Heskett wrote:
> >>>
> >>>
> >>>
> >>>
> >>>>The machine?  AMD 2800XP.  Biostar M7-NCD-Pro Motherboard.  A gig of
> >>>>DDR333 dual channel ram, and an Extacy 9200SE 128 meg card.  I
> >>>>certainly ought to be able to run tuxracer.
> >>>>
> >>>>
> >>>>
> >>>>
> >>>9200 3D works in CVS Xorg.
> >>>
> >>>Their 2D in CVS works up to the very latest PCI Express hardware - which
> >>>I think counts as pretty damned good support.
> >>>
> >>>
> >>>
> >>>
> >>>
> >>I'm not sure what the PCI Express hardware is sorry. But anything r3xx
> >>based (Radeon 9600, 9800 etc) doesn't have squat as far as acceleration
> >>goes AFAIK (I'd be grateful... extreemly grateful... If I was wrong...
> >>But unfortunatley, I don't think I am). AT least in open source... And
> >>that's the main problem.
> >>
> >>
> >
> >They provide full 2D accerlation for all their currently available
> >cards as well as full support for both crtcs, non-bios mode setting,
> >and DFPs, LCDs, and CRTs.  The only thing that requires the use of
> >their binary module is 3D acceleration for their current chips.  Ati
> >is no different than any other chip vender at this point.  With the
> >exception of intel no other vendor has released opensource drivers or
> >databooks for current generation 3D chips. When you look the feature
> >set they support in opensource it outweighs what almost evey other
> >vendor provides.
> >
> >
> 
> As Alan just informed me... From reading the sources, I'd assumed that
> the 2D required DRI as well as the 3D.
> 
> >
> >
> >>ATI probably THINK they're doing a good job with supporting Unix users
> >>and their fglrx closed source drivers. And for SOME users (OK. The
> >>majority) it probably works quite well. That's desktop people who don't
> >>use ACPI for power management, and who never need to suspend their
> >>machines etc..
> >>
> >>
> >
> >Part of the problem is X (xorg/xfree86) has no support for ACPI.  They
> >are merely working within the framework provided by the windowing
> >system.  No X driver supports ACPI.  Full ACPI would probably require
> >at least some sort of kernel component and for the most part X is
> >completely userspace. Do any of the other binary-only drivers support
> >power management at all?  Until X gets proper power management support
> >don't look for things to really improve.
> >
> >
> >
> 
> It's not power management that's required for this particular problem...
> It's recovery from suspend/resume...
> 
> Well, OK. Power management really is required, if only to have the damn
> fan not run 100% of the time on my laptop, but just managing to not
> crash & burn on resume (Or even when returning from text mode) would be
> nice... Heck if they managed that I could use the fglrx drivers &
> probably wouldn't even care so much. But unlike the Open Source
> solution, there's nothing I can do beyond filling out a form on their
> site that gets completely ignored... I know it gets ignored, because
> I've filled it out for EVERY release they've made in the last 6 months
> since I got my r50p. And they've never fixed a single problem. Nor have
> they even acknowledged it.

Their drivers use the X infrastructure.  
X has weak power management support.  
their drivers have weak power management support.

They don't have to provide you with a driver. you could be stuck with
windows only.

> 
> 
> >>Unfortunatly they have IMO missed the point... There is more than one
> >>windowing system (i.e. XOrg & XFree) and although their drivers probably
> >>work (Where they do work) in both systems at the moment, they may not
> >>tomorrow... Plus they work in a manner where the ATI developers decree
> >>they should work. It's hard to belive they even recognise that the
> >>drivers are actually used on laptops for example. Purely because on a
> >>laptop they are not the best shall we say at recovering from
> >>suspending... They also don't work with frame buffers... And I like
> >>200x75 text mode...
> >>
> >>if ATI would just release the programming information there'd be any
> >>number of people willing to write GOOD and STABLE drivers that would
> >>work for ALL situations... And if they didn't work, someone (Besides the
> >>original author, who may not even know we have a problem after all)
> >>could fix them. Because all the information would be there...
> >>
> >>
> >
> >Instead of complaining why not try and develop a power management system for X?
> >
> >
> 
> Because it's not needed to fix the problem? if it was, then none of the
> XOrg drivers would recover from suspend/resume... And they do... The
> r200 does fine I believe...
> 
> >
> >
> >>Their model may work in Windoze, but even there the VENDOR usually does
> >>the drivers... At least as far as IBM laptops go... ATI probably never
> >>even deal with laptops in the windoze world. And thus would lack the
> >>experience required to truely appreciate what's required.
> >>
> >>
> >
> >Most laptops drivers are tweaked by the laptop OEMs because they tend
> >to all be wired slightly differently.  different laptops support
> >different panel sizes and output types (dvi/hd15/tv/etc.) as well as
> >wiring different GPIOs to different laptop hardware so things like fn
> >key combos works.  The laptop oems customize all of that to give their
> >laptop the edge. As such they need to tweak ati's reference drivers to
> >fully support the way they wire everything up.
> >
> >
> >
> >>Of course if I'm wrong I'm sorry.... I wish I was & that the ATI drivers
> >>were the best in the world (Because after all I have a laptop with an
> >>M10/rv350 in it). But unless ATI release the INFORMATION for CURRENT
> >>chipsets (Heck the 3xx isn't even state of the art any more, thats the
> >>4xx isn't it?) taht's never going to happen.
> >>
> >>
> >
> >I too would like to see full open source drivers for the 3d component
> >of current chips, but whether we have specs are not there are still
> >quite a few infrastructural changes that need to happen to X in order
> >to fully utilize every feature of the chips.  Plus even if they
> >released they specs, who'd write the driver?  Writing a 3D driver is
> >no trivial task.  Most developers work on X part time.
> >
> >
> >
> 
> Well... There's Vladimir Dergachev for a start. Or are you implying that
> only people employed by ATI would ever have the smarts or time to doit?
> Well heck, the same goes for an OS I guess Linus has been wasting his
> time hasn't he?

I'm not implying that only ATI empyees can write driver code.  I've
written quite a bit of code for radeon.  the problem is I have to work
a regular job all day, eat, sleep and have a life.  that doesn't
always leave a lot of time for coding.  MOST other x developers fall
into this camp as well.   I'm not saying it can't be done, but it's a
lot of work. there are databooks available for quite a few cards out
there and yet no one has written driver for them.  Why? lack of time,
lack of hardware, etc.  just because the specs open up does not mean
we'll magically have perfect drivers shortly there after.

> 
> I'm a little confused... Why do you think I should be grateful to ATI
> for NOT supporting the 3D open source drivers?

They do support opensource 3d drivers, just not yet for your chip. 
I'd like to hope one day they will, but right now they don't.  Bad
mouthing them is no going to help our cause.  Why should they even
provide a binary driver, much less full specs if all you are going to
do is bitch about it.


Alex

> 
> 
> H
>