Segmentation Violation (Multi-Directed Continuous Gravitational Wave search CV v1.00 (AVX) x86_64-pc-linux-gnu)

Carsten Milkau
Carsten Milkau
Joined: 3 Oct 06
Posts: 3
Credit: 11414521
RAC: 0
Topic 205177

I always get segmentation violations after a few seconds with application "Multi-Directed Continuous Gravitational Wave search CV v1.00 (AVX) x86_64-pc-linux-gnu". Example:

https://einsteinathome.org/de/task/606314126

My CPU is Intel i7 920, that is indeed x86_64, but it does not support AVX (relevant?).

Richie
Richie
Joined: 7 Mar 14
Posts: 656
Credit: 1702989778
RAC: 0

There's something about

There's something about segmentation violation in general:

https://www.novell.com/support/kb/doc.php?id=7001662

Your system has 24GB RAM. Could it be all of that might not be not operating with ultimate stability under load? Has the system been stable with everything else? Many of the tasks have failed at 0.00 sec, so the problem could well be something else.

If it's easy to open the box, I would try if reducing RAM amount has any effect on that problem.

That AVX-thing is not relevant, because application checks if the computer is able to run AVX-extensions. If there is no capability, then it will switch to SSE.

Carsten Milkau
Carsten Milkau
Joined: 3 Oct 06
Posts: 3
Credit: 11414521
RAC: 0

Yes, memory is very stable,

Yes, memory is very stable, I've been using it in this configuration for years. Im also crunching LHC@home / VirtualLHC@home, which employ virtualbox using quite a substantial amount of RAM (6-8 VMs usually use about 10-16 GB of allocated memory and 6-10 GB of disk cache), and those tasks have been running just fine in all that time.

This problem is

  • very specific to the mentioned application (other applications of Einstein@home or other projects run properly) and
  • perfectly reproducible (every single WU crashes the app within seconds, there is no single success in dozens and dozens of WU)

It appears very likely it is a problem with this application.

Christian Beer
Christian Beer
Joined: 9 Feb 05
Posts: 595
Credit: 127258630
RAC: 355427

Did you recently do a kernel

Did you recently do a kernel update on this host? This looks like the problem we had on Debian when they switched of vsyscall emulation. See the discussion in vsyscall is now disabled on latest Linux distros. Debian did revert the change in the meantime because it was causing more problems than it solved.

Carsten Milkau
Carsten Milkau
Joined: 3 Oct 06
Posts: 3
Credit: 11414521
RAC: 0

Ugh, forgot about vsyscall.

Ugh, forgot about vsyscall. No I didn't update my kernel, but maybe this application is newer or has been backported to an older version of libc? I had dropped vsyscall support several months ago already, there was only one or two boinc apps actually relying on it. Guess that's three now ^^ I'll just disable that app, then.

 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.