Intel® Clusters and HPC Technology

Intel MPI and Infiniband: MTU optimisation doc ?


We have a little cluster:
- master
- 8 nodes with 12 cores pro nodes
- Infiniband connectx DDR 2X
- linux CentOS 5.5
- Infiniband stack from OFED
- Intel MPI

At the moment we have a normal basic install of the infiniband packages. So we are in connected mode with a MTU of ~65k. Has anybody info, documentation, links on the optimization of the MTU ? We are using it for Computational Fluid Dynamics computations.

Thx in advance,
Best regards,

Intel MPI 4 - Port Issues on secondary smpd process


I have got the intel mpi working when I have the firewalls turned off, however, when I turn on firewalls I have an issue;

I have unblocked port 8678 on both machines and the request is received on the second machine, however another smpd process is started which uses a different port each time and it is this port this is being block.

Does anyone know how to keep this fixed with a setting or something. Below is the debug information:

Error during installation cluster Studio 2011

Dear all,

I have downloaded the Cluster Studio 2011 XE to evaluate it, but during the installation (no other program is running) the following message appears:

"Microsoft visual Studio 2005 is currenntly running. Installation cannot continue while MVS2005 software s running. Suggestion: Exit Microsoft Visual Studio .NET 2005 software and run the installation again"

After pressing "ok", the installation procedure continues for several minutes until an error message advises that the installation is failed.

what is the purpose of the -rmk option to mpiexec.hydra

I can't figure out either from the documentation or from experiment what the -rmk option is supposed to do. I use a cluster that is shared via Torque (and Maui) and I see that the only supported argument to -rmk is pbs so I'm curious to find out. I thought it might supply the list of allocated nodes, but my experiments don't show that. And it doesn't help to accurately report the resources used on remote nodes... so what's it for?Ed

mpdboot for a cluster

Dear all,I have following lines in my job script:# Number of cores:#SBATCH --nodes=8 --ntasks-per-node=8## Set up job environmentsource/site/bin/jobsetup#startmpd:mpdboot -n 64## Run the program:../bin/sem3dcaf ../input/test_nproc64_sf2.psemI amtryingto run a coarray program with 64 cores (8cores per node). But I could not correctly start mpd. I am new to intel cluster toolkit. I would be grateful for the suggestion.Thanks.

Mixed C++/Fortran MPI + OpenMp and ILP64

I am writing a mixed c++/Fortran code that uses MPI, OpenMP, and 64 bit integers. I am having trouble determining which libraries to link.

1. The getting started guide says to

"Add impi.lib and impicxx.lib (Release) or impid.lib and impicxxd.lib (Debug) to your target link command for C++ applications. Link application with impimt.lib (Release) impidmt.lib (Debug) for multithreading."

It is not clear whether 'impimt.lib' replaces 'impi.lib' and/or 'impicxx.lib'. I am guessing that impimt.lib replaces impi.lib but not impicxx.lib. I am not sure though.

$SLURM_NODELIST not interpreted correctly


I've recently setup a new cluster that uses slurm for resource allocation. Upon starting a new mpi job with:
$ salloc -n 32 sh
$ mpirun -np 32 -nolocal a.out

I get the following errors:

failed to connect to the socket (sock2): {socket.gaierror, (-2, 'Name or service not known')}. Probable reason: host "node2" is invalid
(mpdboot 494): failed to connect to the socket (sock2): {socket.error, (9, 'Bad file descriptor')}. Probable reason: host "node2" is invalid

MPI + mixed C++/Fortran library issue

I am trying to use MPI in a mixed C++/Fortran code. The main routine is C++ which calls Fortran subroutines. I have successfully compiled the code in VS2008 using the Microsoft C++ compiler and Intel Fortran 11.0.074. If I comment out the Fortran subroutine calls, the C++ code loads and runs. MPI nodes are reported properly by the main C++ routine. However, when I add back the Fortran subroutine calls, the code crashes during the program startup. I see the following clues.

1. When running with mpiexec, no program output at all is produced.

Subscribe to Intel® Clusters and HPC Technology