Re: [mpich-discuss] MPICH 3.1 fails doesn't create core file.

12 Oct 2014

      Calling MPI_Abort is not an error that would cause the OS to dump a core. For that, you'll need to do something within your application that would normally cause it to dump a core. MPICH itself won't do it for you. You can add a divide by zero error to your own code and you'll get a core file of your app. What specifically within MPICH are you trying to debug?

Thanks,
Wesley
...
On Oct 12, 2014, at 10:09 PM, Anatoly G <[email protected]> wrote:
Thank you.
My OS (Kubuntu 14.04).
ulimit is already set to unlimited.
When I generate exception (abort() function or divide by zero) I always get a core file,
but when MPICH fails (for example with MPI_Abort) no core file created.
May be I need to change any other settings?
Regards,
Anatoly.
...
On Sun, Oct 12, 2014 at 4:47 PM, Lu, Huiwei <[email protected]> wrote:
The core file is created by the OS, not MPICH. What’s the output of ‘ulimit -c’ on your machine? Could you set ‘ulimit -c unlimited’ and try again?
—
Huiwei
On Oct 12, 2014, at 7:35 AM, Anatoly G <[email protected]> wrote:
...
Dear MPICH.
I'm using MPICH3.1
After application run of 4-5 hours Master process fails, but it doesn't creates a core file.
When I used MPI_Abort (in the past) it not creates core file too.
Can I make MPICH create core in any fail (of course if it's MPI fail).
Regards,
Anatoly.
_______________________________________________
discuss mailing list     [email protected]
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss
_______________________________________________
discuss mailing list     [email protected]
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss
_______________________________________________
discuss mailing list     [email protected]
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss
_______________________________________________
discuss mailing list     [email protected]
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss

Wesley Bland

tags

participants (1)