Open MPI for Debian
-------------------
InfiniBand support
------------------
InfiniBand and UCX support is build into the package and enabled by default.
On Debian, OpenMPI sets the following:
OMPI_MCA_btl_base_warn_component_unused=0
btl = ^uct,openib
pml = ^ucx
osc = ^ucx
This is set in /etc/openmpi/openmpi-mca-params.conf and turns off warnings that OpenFabric is configured
but not used on the system. Many or most users of Debian will not have OpenFabric present on their systems
and turning this off prevents spurious warnings that were breaking higher-level libraries and
applications.
However the Libfabric EFA provider currently does not support applications calling fork()
when memory registrations are handled by the EFA provider. This is to prevent applications
from encountering potential data corruption or crashes when operating in this mode.
Most applications, including the majority of MPI and NCCL applications, will still function
with this constraint. Some applications do require fork() support, however,
these applications do not require memory registrations to be handled by the EFA provider.
If you wish to use OpenIB and know the application to be safe, comment out the exclude line.
Please see https://github.com/ofiwg/libfabric/issues/6332 for details.
Similarly UCX is disabled due to spurious warnings about missing devices on systems without RDMA.
Please enable as required.
MPI Alternatives
----------------
The current Open MPI version of Debian uses the Debian Alternatives system.
Open MPI is considered the default MPI implementation in Debian. MPICH2 is
also available. You can check which MPI is the default on your system with
update-alternatives --list mpirun
and list all installed implementations with
update-alternatives --display mpirun
If you have several implementations installed and Open MPI is not the default
on your system, you can enable it explicitly with
update-alternatives --set mpirun /usr/bin/mpirun.openmpi
-- Alastair McKinstry <mckinstry@debian.org>, Sat, 9 Jan 2021 13:34:35 +0000
Generated by dwww version 1.15 on Tue Jul 2 04:44:53 CEST 2024.