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

James jamesqf at charter.net
Mon Oct 13 11:03:42 CDT 2014


I've never actually measured it, but my impression is that gdb doesn't
slow down execution much (if at all), if you're just running a program
under it so as to get a backtrace on a crash.  Even a few breakpoints
don't slow it that much...

James

On Sun, 12 Oct 2014 23:29:35 -0700, Anatoly G <anatolyrishon at gmail.com>  
wrote:

> Thank you for an advice.
> Unfortunately I can't change input data rate (from external source). We  
> are
> working on slow down input rate. Using debugger cause execution slowdown  
> an
> we can't deal with current data rate.
>
> Regards,
> Anatoly.
>
> On Mon, Oct 13, 2014 at 8:03 AM, James <jamesqf at charter.net> wrote:
>
>> You could try running the master process under gdb.  If it terminates
>> abnormally - e.g. something like a segfault - you can then do a  
>> backtrace
>> to find out where the problem is happening.  Or if you get a "program
>> terminated normally" message, then you know that it's following some
>> code path that causes it to terminate.
>>
>> James
>>
>> On Sun, 12 Oct 2014 20:42:33 -0700, Anatoly G <anatolyrishon at gmail.com>
>> wrote:
>>
>>  I'm not sure about my current failure.
>>> I execute my application (MPI processes). Each process executes endless
>>> loop. After a couple of hours Master process fails (disappears - I  
>>> can't
>>> see it using 'ps' command). There is no core file. I'm not sure about  
>>> the
>>> reason of failure.
>>> It looks strange that single process (15 processes executed) fails w/o
>>> core
>>> file dropping.
>>> I suspect that I'm using MPICH in wrong way, which causes failure.
>>> But like I said before I'm not sure about failure.
>>>
>>> Regards,
>>> Anatoly.
>>>
>>>  _______________________________________________
>> discuss mailing list     discuss at mpich.org
>> To manage subscription options or unsubscribe:
>> https://lists.mpich.org/mailman/listinfo/discuss
>>


_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss


More information about the discuss mailing list