Computation errors since months
2 Jun 2008 1:07:48 UTC
Topic 193707
(moderation:
Hi folks!
I have a serious problem with einstein@home, again. Using Boinc version 5.10.45 I got no problems with the current seti@home app (5.27) but einstein (S5R3_4.38) creates computation errors for each and every package it tries to calculate. (code 38)
See the recent tasks.
Though it doesn't matter: system is Linux from 2.6.23 to 2.6.25. (Arch Linux) Boinc is from the community-repo.
How can I find out more about the reasons for that error?
Computation errors since months
)
I could only check this one (The link you gave is accessible only by you :-). The key to the problem seems to be:
APP DEBUG: Application caught signal 8.Perhaps you find more infos in the BOINC Wiki or the BOINC FAQ Service. (There's also an Unofficial Wiki, but I don't have the url :-)
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
Heat and faulty memory are
)
Heat and faulty memory are possible reasons too.
Any blue screens lately?
MfG
Michael
Team Linux Users Everywhere
Gundolf Jahn wrote:Perhaps
)
Splendid, neither error code 38 nor signal 8 are explained on this page. But I can tell that they exist... Maybe I find something if I take a closer look.
Sorry, didn't know about the links. But you get them by my nick anyway, I guess.
Never had any BSoD on Linux, do you? I didn't have kernel panics, as well. And since Seti works properly, I would suspect more something like a versions incompatibility. Last time it was such thing.
RE: Michael Karlinsky
)
Tried to make a snappy remark. Failed, as soon as I realised you are using Linux too....
Michael
Team Linux Users Everywhere
free_outerrim wrote:Michael
)
You could paint them blue. ;-)
Perhaps that [url=http://www.nabble.com/CONFIG_PREEMPT-causes-corruption-of-application's-FPU-stack-td17293854.html]this discussion[/url] helps you out. Let me know and I'll adapt it into the FAQs.
Ageless wrote:Perhaps that
)
Well, Arch Linux uses vanilla Kernels...
But if my problem would depend on the issue in this thead, I would have the problem since April 2007. According to DistroWatch the 2.6.20 kernel was introduced there into the Distro. But that is not the case. I don't know exactly when it occurs the fist time, but it must be less than 4 months. I guess the maintainer guys haven't used this CONFIG_PREEMPT thing before. But I don't compile the kernel for my self, so I'm not so sure about it.
The bugtracker doesn't show the issue, yet. I start a thread there. If that brings any additional infos I'll write it here.
RE: Splendid, neither error
)
Always start looking @home: man 7 signal
Signal 8 is SIGFPE (floating point exception). To find out more about the FPU status word value (0xb8c1 == 0b1011_1000_1100_0001), look here. As to the actual cause: heat (OC?) is a good suspect.
Regards,
Waldi
RE: Ageless wrote:Perhaps
)
So it would be interesting to know if your kernel uses the CONFIG_PREEMPT setting.
Maybe a .config or similar ASCII file can be found in /usr/src/linux or a directory with a similar name? Some Linuxes also support /proc/config.gz , so you can try
cat /proc/config.gz | gunzip - | grep CONFIG_PREEMPT
CU
Bikeman
Paper Moon wrote:Always start
)
Ok, I hope I'll remember this next time, thanks.
How about that:
That may be it, doesn't it?
RE: How about that:$ grep
)
It may, if you know how to build a kernel and this fixes the problem for you.
It may not be an option in all distro kernels, that's something that needs to be tested as well.
And the applications need to be checked, as this is only another workaround hinting to the real problem.