mpi Error

[Ezat Shokrani]

Dear SimVascular developers,

I have installed the previous versions of the SimVascular and svsolver (SimVascular-linux-x64-2021.06.15.deb & svsolver-linux-x64-2021.06.15.deb) on a virtual machine and Ubuntu 18.04 platform. I set up my case completely. However, when I pressed the "run simulation" button, I got the mpi error.
I do not know which version of OpenMPI I should install on my virtual machine in order to be compatible with the svsolver and therefore, I can run my simulation. I would be wondering if you could guide me on this issue.

Best regard,
Ezat Shokrani

[David Parker]

Hi Ezat,

Please upgrade SV and svSolver to the September 2021 release.

The September 2021 version of svSolver was built using GCC 7.5 and OpenMPI 2.1.

Cheers,
Dave

[Ezat Shokrani]

Dear Dr. David Parker,

Thanks for your response.
I have encountered some questions regarding parallel computing. I would be wondering if you could guide me.

Firstly, I do not know how many cores are optimized in order to run my case study on a virtual machine and whether or not there are any limitations for the SimVascular software to be run with high computational resources.

Secondly, I am not familiar with using the cluster to run my case study. I do not know how I should install it on a local cluster (I should use the binary files or must be complied from the source code).

Finally, which command must be used in order to run the case on a cluster?


I am sorry if my message gets too long
Thanks for your help and taking time answering my questions.

Best regards,
Ezat Shokrani

[David Parker]

Hi Ezat,

Firstly, I do not know how many cores are optimized in order to run my case study on a virtual machine and whether or not there are any limitations for the SimVascular software to be run with high computational resources.

svSolver uses MPI to execute in parallel. When executing the solver using mpiexec -np N you give the number of processors N (i.e. cores) to use. The solver will use N cores. You need to determine how many cores to use for the size (number of elements) used in your mesh.

Secondly, I am not familiar with using the cluster to run my case study. I do not know how I should install it on a local cluster (I should use the binary files or must be complied from the source code).

You will need to build svSolver on your cluster if it is not using Ubuntu. The cluster administrator should be able to help you do that.

Finally, which command must be used in order to run the case on a cluster?

After building svSolver, for example in the BUILD directory, you will have a svSolver/BUILD/svSolver-build/bin directory that contains the following four executables

Code: Select all

svpost*
svpre*
svsolver*
svsolver-nompi*

The svsolver executable is used to perform a simulation. You can run a simulation with MPI using the following command

Code: Select all

mpiexec -np 4 svSolver/BUILD/svSolver-build/binsvsolver solver.inp

Note that if your cluster has a resource manager like Slurm you will need to create a job script to execute the solver.

Cheers,
Dave