2000-11-29 00:32:38

by Petter Sundl?f

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

Thanks to some nice people in #NVIDIA I found what seems to be a
solution; compile with processor type as "K6". No segfaults, lost
terminfo or disabled consoles.

So are there issues with the K7 processor code? Bleh, never mind, I have
no idea what I am talking about.

Original bug report: http://findus.dhs.org/~odd/nvidia.txt


2000-11-29 00:42:53

by Alan

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

> Thanks to some nice people in #NVIDIA I found what seems to be a
> solution; compile with processor type as "K6". No segfaults, lost
> terminfo or disabled consoles.
>
> So are there issues with the K7 processor code? Bleh, never mind, I have
> no idea what I am talking about.

The K7 optimisations are not used for I/O space accessess. Or shouldnt be,
but the nvidia code is unreadable so they may have done so

2000-11-29 00:52:27

by Petter Sundl?f

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

Alan Cox wrote:
> The K7 optimisations are not used for I/O space accessess. Or shouldnt be,
> but the nvidia code is unreadable so they may have done so

OK. I believe at least one of the NVIDIA developers read this list, so
hopefully they can look at what can be done on their side.
But seeing as how they've made no release for circa three months, and
considering the closed nature of their drivers, I doubt it'll be fixed
any time soon.

Oh well.

2000-11-29 01:10:09

by Dan Hollis

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

On Wed, 29 Nov 2000, Alan Cox wrote:
> > Thanks to some nice people in #NVIDIA I found what seems to be a
> > solution; compile with processor type as "K6". No segfaults, lost
> > terminfo or disabled consoles.
> > So are there issues with the K7 processor code? Bleh, never mind, I have
> > no idea what I am talking about.
> The K7 optimisations are not used for I/O space accessess. Or shouldnt be,
> but the nvidia code is unreadable so they may have done so

Dont forget the nvidia driver is completely SMP broken. As in, trash your
filesystems broken.

-Dan

2000-11-29 01:40:59

by J.A. Magallon

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]


On Wed, 29 Nov 2000 01:39:56 Dan Hollis wrote:
>
> Dont forget the nvidia driver is completely SMP broken. As in, trash your
> filesystems broken.
>

Not so broken. I use it under SMP 2.2.18-pre23 and works fine. But under 2.4
hangs. So I think it is something that changed between 2.2 and 2.4 (locking
granularity???). And just hangs your box, does not corrupt anything.

It is a pitty that any piece of copylefted code prevents nVivia of giving all
of the drivers in source form.

--
Juan Antonio Magallon Lacarta #> cd /pub
mailto:[email protected] #> more beer

Linux 2.2.18-pre23-vm #3 SMP Wed Nov 22 22:33:53 CET 2000 i686 unknown

2000-11-29 01:59:06

by Dan Hollis

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

On Wed, 29 Nov 2000, J . A . Magallon wrote:
> On Wed, 29 Nov 2000 01:39:56 Dan Hollis wrote:
> > Dont forget the nvidia driver is completely SMP broken. As in, trash your
> > filesystems broken.
> Not so broken. I use it under SMP 2.2.18-pre23 and works fine.

Try unreal tournament. Locks up hard during the intro animation. It's been
listed for months in the nvidia FAQ as a known bug (#6.5.7) with no fix.

http://x54.deja.com/getdoc.xp?AN=687585074&CONTEXT=975461056.438435863&hitnum=2
http://x54.deja.com/getdoc.xp?AN=690445861&CONTEXT=975461056.438435863&hitnum=3

-Dan

2000-11-29 02:19:03

by J.A. Magallon

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]


On Wed, 29 Nov 2000 02:29:04 Dan Hollis wrote:
> On Wed, 29 Nov 2000, J . A . Magallon wrote:
> > On Wed, 29 Nov 2000 01:39:56 Dan Hollis wrote:
> > > Dont forget the nvidia driver is completely SMP broken. As in, trash your
> > > filesystems broken.
> > Not so broken. I use it under SMP 2.2.18-pre23 and works fine.
>
> Try unreal tournament. Locks up hard during the intro animation. It's been
> listed for months in the nvidia FAQ as a known bug (#6.5.7) with no fix.
>

Not exactly UT, but I have tried quake3, heretic2, and descent3. And in
the serious group, OpenGL Performer and recently Inventor. I have not
tried to leave q3a in demo mode for two days, but everything else works
fine. And Performer now (v2.4) is SMP'ing.

--
Juan Antonio Magallon Lacarta #> cd /pub
mailto:[email protected] #> more beer

Linux 2.2.18-pre23-vm #3 SMP Wed Nov 22 22:33:53 CET 2000 i686 unknown

2000-11-29 02:31:51

by Kiril Vidimce

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

On Tue, 28 Nov 2000, Dan Hollis wrote:
> Dont forget the nvidia driver is completely SMP broken. As in, trash your
> filesystems broken.

Not true. It works for us with no problems on a number of SMP boxes
running 2.2.{14,16}. I don't know about 2.4.x.

KV
--
___________________________________________________________________
Studio Tools [email protected]
Pixar Animation Studios http://www.pixar.com/

2000-11-29 02:39:53

by Alan

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

> On Tue, 28 Nov 2000, Dan Hollis wrote:
> > Dont forget the nvidia driver is completely SMP broken. As in, trash your
> > filesystems broken.
>
> Not true. It works for us with no problems on a number of SMP boxes
> running 2.2.{14,16}. I don't know about 2.4.x.

Dan is not the only one to report it totally trashing a machine and file systems
SMP. So I suspect there is something there , but I don't know what (or care).
I've seen other demos of bugs in the nv driver, long standing ones and
reading the mangled code you can see bugs even in their mangled code
without looking too hard.

2000-11-29 02:55:24

by Kiril Vidimce

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

On Wed, 29 Nov 2000, Alan Cox wrote:
> > On Tue, 28 Nov 2000, Dan Hollis wrote:
> > > Dont forget the nvidia driver is completely SMP broken. As in, trash your
> > > filesystems broken.
> >
> > Not true. It works for us with no problems on a number of SMP boxes
> > running 2.2.{14,16}. I don't know about 2.4.x.
>
> Dan is not the only one to report it totally trashing a machine and file systems
> SMP. So I suspect there is something there , but I don't know what (or care).
> I've seen other demos of bugs in the nv driver, long standing ones and
> reading the mangled code you can see bugs even in their mangled code
> without looking too hard.

I've never seen such thing as code without bugs. In my experience,
the NVIDIA drivers are by far the most complete and solid 3D drivers
under Linux.

KV
--
___________________________________________________________________
Studio Tools [email protected]
Pixar Animation Studios http://www.pixar.com/

2000-11-29 03:03:25

by Alan

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

> I've never seen such thing as code without bugs. In my experience,
> the NVIDIA drivers are by far the most complete and solid 3D drivers
> under Linux.

You are welcome to your opinion. I've got this great bridge to sell you too


2000-11-29 03:19:00

by Kiril Vidimce

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

On Wed, 29 Nov 2000, Alan Cox wrote:
> > I've never seen such thing as code without bugs. In my experience,
> > the NVIDIA drivers are by far the most complete and solid 3D drivers
> > under Linux.
>
> You are welcome to your opinion. I've got this great bridge to sell you too

I don't see the need for sarcasm. If you feel that you can demonstrate
that the above is untrue, by all means do so. I am sure there is more
than one person that would want to know what's currently the best 3D
Linux configuration out there.

KV
--
___________________________________________________________________
Studio Tools [email protected]
Pixar Animation Studios http://www.pixar.com/

2000-11-29 03:42:27

by Kiril Vidimce

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

On Wed, 29 Nov 2000, Alan Cox wrote:
> > I've never seen such thing as code without bugs. In my experience,
> > the NVIDIA drivers are by far the most complete and solid 3D drivers
> > under Linux.
>
> You are welcome to your opinion. I've got this great bridge to sell you too

BTW, in case this is not clear, I am not saying that there are no problems
with the drivers. In fact, I have some issues with them but have reasonable
workarounds for the moment. However, if there are bugs, that doesn't mean
that the drivers are completely broken.

KV
--
___________________________________________________________________
Studio Tools [email protected]
Pixar Animation Studios http://www.pixar.com/

2000-11-29 10:14:59

by Yoann Vandoorselaere

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

Kiril Vidimce <[email protected]> writes:

> On Wed, 29 Nov 2000, Alan Cox wrote:
> > > I've never seen such thing as code without bugs. In my experience,
> > > the NVIDIA drivers are by far the most complete and solid 3D drivers
> > > under Linux.
> >
> > You are welcome to your opinion. I've got this great bridge to sell you too
>
> I don't see the need for sarcasm. If you feel that you can demonstrate
> that the above is untrue, by all means do so. I am sure there is more
> than one person that would want to know what's currently the best 3D
> Linux configuration out there.

It was already discussed too much time...
A good argument against what they do is that they don't even respect
existing standard (see DRI). Also, from what I seen, the Radeon card
have almost the same performance as the nvidia one... and we have an
open source driver... at least.

Ps : and this is going offtopic...

--
-- Yoann http://www.mandrakesoft.com/~yoann/
Murphy's law : If anything can go wrong, it will.
O'Tool's commentary : Murphy was an optimist.

2000-11-29 13:43:15

by Android

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

> > I've never seen such thing as code without bugs. In my experience,
> > the NVIDIA drivers are by far the most complete and solid 3D drivers
> > under Linux.
>
> You are welcome to your opinion. I've got this great bridge to sell you
too

The most stable of all video drivers under Linux has to be the standard VGA
driver.
As for SVGA, then VESA wins that territory. The problem with specific
SVGA/Accelorator
drivers is that the details of the video card operation is kept secret, and
what information
is known is scarce. Much of it is found by experimentation, and dare I say,
reverse engineering.
In most cases, you're lucky you get a working driver at all. The same holds
true for sound cards.
So, unless the company has fully released the details of their NVIDIA card
to whomever wrote
the Linux driver, then don't think for a moment that it is crash-free. Just
one wrong bit sent to
an incorrect port, and it's time to reset your machine. I don't need to
mention data loss.

-- Ted


2000-11-29 15:25:14

by Eric W. Biederman

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]

"Android" <[email protected]> writes:

> > > I've never seen such thing as code without bugs. In my experience,
> > > the NVIDIA drivers are by far the most complete and solid 3D drivers
> > > under Linux.
> >
> > You are welcome to your opinion. I've got this great bridge to sell you
> too
>

> So, unless the company has fully released the details of their NVIDIA card
> to whomever wrote
> the Linux driver, then don't think for a moment that it is crash-free.
It would be very odd if NVIDIA didn't release the details to themselves.
What they haven't done is released good source code.

Eric

2000-11-29 17:28:52

by James Simmons

[permalink] [raw]
Subject: Re: XFree 4.0.1/NVIDIA 0.9-5/2.4.0-testX/11 woes [solved]


> I've never seen such thing as code without bugs. In my experience,
> the NVIDIA drivers are by far the most complete and solid 3D drivers
> under Linux.

I have to agree. Opensource or not the NVIDIA drivers are the best I have
seen as well.