I'm currently trying to do a MPI_Comm_Spawn of a "non-MPI" application which its launched via a script (well, its an MPI application which handles the connection by itselfr).
But my parent process is blocking on MPI_Comm_Spawn call, i guess that's because my script won't do any MPI_Init.
Is there a way to configure MPI_Comm_Spawn so it doesn't block? (or doesn't expect MPI_Init, like the ompi_non_mpi in open-mpi MPI_Info structure which it's passed to MPI_Comm_Spawn).
On the cluster of our University (nodes with 2 hexacores proc.) we get strange values for the "Time Use" in our Job Manager (Torque/Maui).
Which programs compute the "Time Use" value (the job manager or Intel MPI) ? I think Torque. But does Torque receive information from Intel MPI ?
Where can I find more Informations about the computations of "Time Use" and "Elapsed Time" ?
Thx a lot,
Hopefully this is in the right place. I recently acquired 5 1U rackmount server nodes from a business and would like to cluster them for my own private use for MPI programming. (Research for my degree mostly)
I have not successfully set up a cluster before, and am having issues figuring out where to start. I've looked at many older guides on clustering, but many leave out important setup details or utilize software that is long retired. I was wondering if anybody had any suggested material/advice for getting this mini-cluster set up.
My hardware includes:
The Intel® MPI Library includes two utilities for improving MPI performance. The Automatic Tuning utility allows you to tune runtime parameters to match your system and applications. The Intel® Trace Analyzer and Collector allows you to instrument and profile your MPI programs to improve parallelism. I will be presenting a webinar on December 12 discussing how to use these capabilities to improve your MPI application's performance. For more information, please see http://software.intel.com/node/333491, or you can
Two questions, first on the Cluster Studio installation.
In the document, software "expect" is needed to make the ssh connectivity between nodes of the cluster. However, in my current cluster, all my nodes can ssh to each other without password. I wonder whether I can skip this "expect" requirement, since I do not want to complicate the existing ssh environment.
Second question, on using Intel MPI
I've installed intel mpi, use mpi-selector to choose intel mpi (installed in impi folder in /opt/intel/), but when I launch a job with
mpirun -n 8 ./HEART_MR
where HEART_MR is the executable, I obtain 8 processes running each one on a core, when I want a single process running on 8 cores. If I use openmpi, things happen how I expect to be.
Can you help me?
My machine has an eight core intel Xeon processor.
The Intel® Cluster Studio XE 2013 for Linux* and Windows* combines all Intel® Parallel Studio XE and Intel® Cluster Tools into a single package. This multi-component software toolkit contains the core libraries and tools to efficiently develop, optimize, run, and distribute parallel applications for clusters with Intel processors. This package is for cluster users who develop on and build for IA-32 and Intel® 64 architectures on Linux* and Windows*, as well as customers running over the Intel® Xeon Phi™ coprocessor on Linux*.
I am licensed user of Cluster Studio XE 2012, and now the version 2013 is out. I plann to run /opt/intel/icsxe/2012.0.032/uninstall.sh to remove the old version, then install the new version.
However I have one more question. Since the Cluster Studio XE contains many components such as mkl, mpi, tbb, icc etc, and we get updates of these component throughout the year. What would be the recommended procedure of updating seperate components for an already installed Cluster Studio XE?
Thanks for any advice.
I am using Intel MPI 4.0.3.006 to run my application, and found that the generated outputs were different from those generated using older version of MPI. I tried with MPICH2-1.4 and got the same wrong outputs. Has anyone seen similar problem and how can I fix it?