Debugging

Jump to bottom Edit New page

Axel Huebl edited this page Apr 10, 2014 · 5 revisions

You are here: Home > Developer Documentation > Debugging

This page collects some useful hints about how to debug a hybrid (CUDA + device) parallel (MPI) application.

MPI + Valgrind

Use the OpenMPI supressions list

mpiexec ... valgrind --suppressions=$MPI_ROOT/share/openmpi/openmpi-valgrind.supp picongpu ...

MPI + GDB

Trigger start and back trace afterwards:

mpiexec ... gdb -ex r -ex tb --args picongpu ...

(might need X11 and xterm -e before the gdb call?)

CUDA + Valgrind

Check this thread (cuda 2.0) and cudagrind (-> cuda-memcheck)