[mpich-discuss] MPI log via MPE on CRAY XC30 machine

Balaji, Pavan balaji at anl.gov
Wed Sep 17 13:20:19 CDT 2014

On Sep 17, 2014, at 1:18 PM, Milind Chabbi <Milind.Chabbi at rice.edu> wrote:
> That worked. Now, my second question, should I compile the entire
> application with mpecc or just using mpecc during the last linking
> step is sufficient?

Just the last linking step for static linking.  For dynamic linking, you can just preload the MPE library.

> Specifically, I have a scenario where I am already overriding some MPI
> functions such as MPI_Init, MPI_Finalize, and MPI_Barrier.
> So, I am trying to do a two step process:
> Step 1: Create a relocatable binary with my intercepted MPI calls:
> CC -Wl,-r -o foo_r foo.cpp  -nostdlib -Wl,--wrap=MPI_Init
> -Wl,--wrap=MPI_Finalize  -Wl,--wrap=MPI_Barrier  my_mpi_intercepts.o
> Step 2: Create the final executable by linking with MPECC
> $MPECC  -o foo foo_r  -mpilog
> When I do this, the generated executable does not produce any clog2 logs.

If you need multiple levels of profiler interception, you might want to consider PNMPI from Livermore.

  — Pavan

Pavan Balaji  ✉️

discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:

More information about the discuss mailing list