Forwarded message: -------------------Date: Fri, 25 Jan 2002 04:23:54 -0500
From: digisin@earthlink.net (digisin)
Subject: Re: sdl with nvidia = bugs
Mr. Latinga,
I tried your suggestions and can report that I have no problems when I
run an sdl app that doesn’t use
sound while simultaneously running a program that does. I also ran xmms
with a visualization plugin
that uses sdl and had no problems. The audio problems only occur when
going to fullscreen mode
in an sdl program.
So, I had previously upgraded to 2.4.17 in the hopes that it would
alleviate my problems, but to no effect 
I had even tried patching the kernel with Ingo Mulnar’s scheduler patch,
and Robert Love’s pre-emptive
kernel patch (not only just to fix the problem mind you ;)).
Unfortunately the problem persisted.
Finally I tried Michael Callahan’s mjc3 patch for 2.4.17-2.4.18-pre?
kernels on a clean 2.4.17,
since it included not only all of the ‘speed improvement’ patches but
fixes as well.
After I built it, I updated the sblive drivers from
opensource.creative.com’s CVS.
I also ‘downgraded’ to NVIDIA_kernel-1.0-2313 instead of 2314 AND
removed a double define in
it’s Makefile: -D_LOOSE_KERNEL_NAMES was defined twice in the same line.
Lastly I noticed that after all this, I still had problems - not with
sdl apps that I built myself, but with
most that came with my SuSE Linux 6.3 pro. Not only sdl apps this
time… in particular with ANY
app that has Alsa linked into it! Quite strange… so I removed all alsa
libs and sdl libs, rebuilt the
sdl libs without alsa support - only with oss - and finally I only have
problems with two programs now:
the wm kde 2.2.2 and Cannon Smash. The later uses sdl, and I believe the
problems with kde to be
caused by alsa being linked in … but I am wary of recompiling it since
kde 3 is just around the corner 
(March 13th?). I could not compile Cannon Smash :-\ I have NO CLUE as to
what SuSE Labs did to
get it to compile, I tried the version with my distro., the latest
version, tried with the default bundled gcc
and with gcc 3.0.3 that I hand-built – have problems when it links at
the end. Even the source rpm
provided by SuSE does not have the patch that they MUST have used … wtf.
Ok, so after all this work and what-not, finally I read this
http://theregister.co.uk/content/3/23817.html .
You might have seen something similar other places (i.e. slashdot et. al).
The strange thing is that I do have an athlon supposedly affected by
this ‘bug’, but it doesn’t cause lockups
or system-hangs … just audio corruption from my sblive and only when
going to fullscreen 3D.
What makes it even more bizarre is the fact that Alsa was what was
causing me problems…
btw - those cvs sblive drivers have HardWare surround sound decoding
(both active and passive) !!!
To sum it all up - I don’t think that the mem=nopentium option is a
valid fix. Sure it works, but it
only cures the SYMPTOM NOT THE PROBLEM. I share AMD’s view on this one:
AMD concludes:- “Our conclusion is that the operating system is creating
coherency problems within
the system by creating cacheable translation to AGP GART-mapped physical
memory…
When the cache-line eviction occurs the stale data written to physical
memory has fatal side effects.”
Easy proof of this is how the ‘bug’ does not affect BSD - due mainly to
the way it handles memory
(or so I’ve heard).
I don’t suppose you know Alan Cox or someone else in the kernel dev team
and would care to throw
that at them and ask for a cure for the problem - not just the symptom;
on behalf of meager linux
users everywhere such as myself?
I apologize for the length and depth of this email, and want to thank
you again for all of your time
and your efforts towards the open source community.
Sam Lantinga wrote:
Hi Jonas, thanks for sending me the mail, but it definitely sounds like
driver/kernel issues to me. You might try updating to the latest kernel.
Does the sound pop happen with an SDL application that doesn’t use sound,
but with another program playing sound?In the SDL test directory is a program “testgl” which you can use to test
this out - it doesn’t use sound at all.Please let me know how it turns out!
Good luck,
-Sam Lantinga, Software Engineer, Blizzard Entertainment
------------------- end forwarded message.
