1.0-7667 for Linux x86-64 released

[ctilburg]

regarding lockups...

I recently upgraded from FC3 to FC4 and started getting lockups... first on
DOOM3, then I found it happend very reliably running the Scimark2 C
version compiled with gcc 2.95.3. Compiling this benchmark with FC4
stock gcc (4.0.1) or gcc32 (3.2.3) did not result in a lockup. Stock FC4
gcc has also failed to produce other working packages built from source;
they segfault, but work when compiled with gcc32.

JIC: an attempt to re-configure and re-compile gcc 2.95.3 compiler was made,
since it was built under FC2 or FC3. This resulted in a make bootstrap error
looking for lowlevellock.h via /usr/include/bits/stdio-lock.h... !

Since vendors are very conservative about compilers, I would hazard to
guess that the switch to gcc 4.0.1 that happened at FC4 could be causing
problems, and/or that there might be some trouble with some possible libc
(or other) surgery that has resulted in lowlevellock.h disappearing.

At the very least, perhaps a full re-compile of the libraries and driver with
FC4 gcc32 (3.2.3). at least, would be in order, and possibly fix this problem.

[gilboa]

Interesting.
I'm playing Doom3 on my dual Opteron machines (FC4 x86-64) and I've only seen it lock once. (I couldn't SSH into the machine in-order to see if it Xid's)
I've downloaded Scimark2 C source from http://math.nist.gov/scimark2/scimark2c.zip and I'll try and run it later today. (If I have the time.)

I doubt that it's a GCC 4.01 vs 2.95. (Though it may result from GCC4 botching the kernel or the GL library.)

Cheers,

[ctilburg]

FYI: I am running an XP3000... its not a 64 bit amd, but the root cause/issue
may be the same/related.

I have re-posted in the x86 thread.

[ctilburg]

Since gcc 2.95.3 cannot be completely bootstrapped under FC4,
if you want to try my 32 bit binary, see the attachment at:

https://bugzilla.redhat.com/bugzilla....cgi?id=162314

This lockup problem may be agp related, since another system
running an identical FC4 software installation, but without Nvidia's
drivers, not running X, not using an AGP card (and not even an
nvidia card (S3): its a server) does not have this lockup problem
with the scimark2 binary.

It looks as if the whole system locks up just as scimark2 is trying
to write the results into the xterm window... which makes me
wonder about lowlevellock.h...

[gilboa]

Seems to be running just fine here.
I'll try again later.

[tji]

My problem appears to be permissions related.. XvMC works when I run mythfrontend as root. Anyone know what permissions come into play when using XvMC?

Originally Posted by tji

Does anyone have XvMC working with this driver under x86-64 mode??

I am tryint to use XvMC with MythTV and/or MPlayer, to have better playback of HDTV content. But, XvMC will not work with either.

With Mplayer, I get an error saying:

"FATAL: Cannot initialize video driver."

Setting XVMC_DEBUG outputs a bit more:

NV-XVMC (ERROR): Authentication failed


What authentication is failing?!? Is there some other step needed to get XvMC running?

[gilboa]

ctilburg,

Tried your binary again.
Works fine on my Opteron workstation.

[libertytrek]

Would it be possibel fro these to be released as simple tar files in addition to the run installer?

When I am installing SuSE SLES9, I need to be able to feed the drivers to the installer, and I'd like at a minimum to be able to put the Ethermet drivers on a floppy so I can get installed with network support.

I'm not a Linux/Unix wiz, and I happen to use a Windows Workstation, so cannot simply 'extract' the files as instructed.

[Lithorus]

Originally Posted by libertytrek

Would it be possibel fro these to be released as simple tar files in addition to the run installer?

When I am installing SuSE SLES9, I need to be able to feed the drivers to the installer, and I'd like at a minimum to be able to put the Ethermet drivers on a floppy so I can get installed with network support.

I'm not a Linux/Unix wiz, and I happen to use a Windows Workstation, so cannot simply 'extract' the files as instructed.

This thread is NOT about the network drivers, but the graphics driver.

Anyway id the SUSE installer uses a newer kernel, support for the nvidia network should be included.

Edit:
Which version of SUSE are we talking about here?
I would guess 9.2 and newer(9.3 is the latest) would have support for your network.

[ctilburg]

gilboa,

You have to stick it in a shell loop; one run is not sufficient.

It also doesn't seem to affect my other systems, only one.

It also does *NOT* involve nvidia's drivers since running scimark2 at
the console, after a fresh boot, doesn't even have their driver loaded.

If you can get this to happen on 64 bit FC4, please add to:

https://bugzilla.redhat.com/bugzilla....cgi?id=164976

I was not reporting the bug correctly.

[gilboa]

I'm on a FC4/x86-64.

I'll try running it in a loop and see if it helps.

[gilboa]

No go.
I ran it in a loop for 15 minutes.
Nada.

You sure your computer isn't just over-heating due to insufficient CPU/case cooling?