[Trilinos-Users] Inquiry about which C++ std library to use when linking Trilinos 10.6 on MacOS (Snow Leopard) with an application using Intel compilers v 11

[Bartlett, Roscoe A]

Aytekin,

The MPI errors you are seeing and the link warnings you show don't look they are related (but anything is possible).

Do you have any way to debug this MPI error?  The error message does not look very helpful.

- Ross




From: Aytekin Gel [mailto:aike at alpemi.com]
Sent: Wednesday, December 22, 2010 10:35 AM
To: Bartlett, Roscoe A
Cc: trilinos-users at software.sandia.gov; Hu, Jonathan
Subject: Re: [Trilinos-Users] Inquiry about which C++ std library to use when linking Trilinos 10.6 on MacOS (Snow Leopard) with an application using Intel compilers v 11

Thank you Ross for your reply.

It links and creates the executable successfully but the execution fails complaining about MPI as shown below:


 Time =   0.0000      Dt =  0.20000E-03    CPU time left =     0.000 s

[mackozmik.local:28176] *** An error occurred in MPI_comm_size

[mackozmik.local:28176] *** on communicator MPI_COMM_WORLD

[mackozmik.local:28176] *** MPI_ERR_COMM: invalid communicator

[mackozmik.local:28176] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)

--------------------------------------------------------------------------

mpirun has exited due to process rank 0 with PID 28176 on

node mackozmik.local exiting without calling "finalize". This may

have caused other processes in the application to be

terminated by signals sent by mpirun (as reported here).

--------------------------------------------------------------------------



So I was trying to rule out the warnings during link stage for the cause of this failure as Jonathan indicated some concern about these warnings. Again the same source code runs ok on jaguarpg at OLCF so I don't think there is a problem with the code itself and MPI works ok on the new system (i.e., MacBook) because when I run the same executable without using Trilinos solvers (it uses our own BiCGSTAB implementation) it runs ok with the installed OpenMPI. Something seems to be happening in Trilinos interface or beyond.

Thanks,

Aytekin





On Wed, Dec 22, 2010 at 9:50 AM, Bartlett, Roscoe A <rabartl at sandia.gov<mailto:rabartl at sandia.gov>> wrote:
Aytekin,

I don't think that I have never seen this type of link warning/error about 'visibility' before.

Does the code actually link and run okay?  If it does, I would not be too concerned.  Some systems generate strange warnings that seem to be harmless.

- Ross


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://software.sandia.gov/pipermail/trilinos-users/attachments/20101222/5e99f64c/attachment.html