Discussion:
[OMPI users] <sub.> Request to debug the code
Nitu
2017-11-17 23:20:42 UTC
Permalink
I've been trying to find the reason for the error,but unable to do so.It
is showing Signal: segmentation fault and Signal code: Address not
mapped.Please help me debug the code.

Thanks,
Nitu
2017-11-17 23:31:06 UTC
Permalink
-------- Forwarded Message --------
Subject: <sub.> Request to debug the code
Date: Sat, 18 Nov 2017 04:50:42 +0530
From: Nitu <***@students.iiit.ac.in>
To: ***@lists.open-mpi.org

I've been trying to find the reason for the error,but unable to do so.It
is showing Signal: segmentation fault and Signal code: Address not
mapped.Please help me debug the code.

Thanks,
Gilles Gouaillardet
2017-11-18 00:51:32 UTC
Permalink
The easiest way to debug a code is to use a parallel debugger such as
DDT or TotalView.
Note your vendor might provide its own parallel debugger.

You only need to compile your application with the '-g -O0' flag, run
it under the debugger, and wait for the crash.


An other option is to do post mortem debugging.
Most sites have a default core size equal to zero, so you need to
ulimit -c unlimited
and then run your application
then you can
gdb <binary name> <core file name>
the 'bt' command can be used to visualize the stack and understand
where the crash occurred.


I do not know how many MPI tasks you plan to use, but so far it does
not crash with 2 MPI taks.
With other task counts (i tested 1, 3 and 4), there are several
crashes/errors that are caused by your application
(e.g. the root cause is not Open MPI)



Cheers,

Gilles

On Fri, Nov 17, 2017 at 4:31 PM, Nitu
Post by Nitu
-------- Forwarded Message --------
Subject: <sub.> Request to debug the code
Date: Sat, 18 Nov 2017 04:50:42 +0530
I've been trying to find the reason for the error,but unable to do so.It is
showing Signal: segmentation fault and Signal code: Address not
mapped.Please help me debug the code.
Thanks,
_______________________________________________
users mailing list
https://lists.open-mpi.org/mailman/listinfo/users
Jeff Squyres (jsquyres)
2017-11-18 15:26:30 UTC
Permalink
Nitu --

We actually try hard not to do students' homework for them on this list. We are more than willing to *help*, but please don't just send your program to us and say "fix it for me." Remember that we are volunteers on this list; people are inspired to help others when it is obvious that they have first tried to help themselves.

For example: a good bug report provides all kinds of information, such as:

- what are you trying to do / what is the program supposed to do?

- what exactly is going wrong?
- be as specific as possible
- what error messages do you see?
- what parts are you confident *are* working correctly?
- ...etc.

- what environment are you working in?
- what kind of machine(s) are you using?
- what operating system are you using?
- what network are you using?
- how many machines are you using?
- what version of Open MPI are you using?
- ...etc.

In short: without more information, we cannot help you.
Post by Gilles Gouaillardet
The easiest way to debug a code is to use a parallel debugger such as
DDT or TotalView.
Note your vendor might provide its own parallel debugger.
You only need to compile your application with the '-g -O0' flag, run
it under the debugger, and wait for the crash.
An other option is to do post mortem debugging.
Most sites have a default core size equal to zero, so you need to
ulimit -c unlimited
and then run your application
then you can
gdb <binary name> <core file name>
the 'bt' command can be used to visualize the stack and understand
where the crash occurred.
I do not know how many MPI tasks you plan to use, but so far it does
not crash with 2 MPI taks.
With other task counts (i tested 1, 3 and 4), there are several
crashes/errors that are caused by your application
(e.g. the root cause is not Open MPI)
Cheers,
Gilles
On Fri, Nov 17, 2017 at 4:31 PM, Nitu
Post by Nitu
-------- Forwarded Message --------
Subject: <sub.> Request to debug the code
Date: Sat, 18 Nov 2017 04:50:42 +0530
I've been trying to find the reason for the error,but unable to do so.It is
showing Signal: segmentation fault and Signal code: Address not
mapped.Please help me debug the code.
Thanks,
_______________________________________________
users mailing list
https://lists.open-mpi.org/mailman/listinfo/users
_______________________________________________
users mailing list
https://lists.open-mpi.org/mailman/listinfo/users
--
Jeff Squyres
***@cisco.com
Loading...