search

hyperion-rt / hyperion

Hyperion Radiative Transfer Code
http://www.hyperion-rt.org
BSD 2-Clause "Simplified" License
52 stars 26 forks source link

Code not running in HPC cluster #217

Open suman-phy opened 5 years ago

suman-phy commented 5 years ago

When I tried to run the code in serial in a HPC cluster, it's running quite well. However, when I tried to run it in mpi mode, it was simply not running. Can you help me about any specific reason?

astrofrog commented 5 years ago

How did you install Hyperion? How are you running it with MPI? And are you getting any error messages?

suman-phy commented 5 years ago

I have tried installing hyperion using both the commands in the following link:

https://anaconda.org/conda-forge/hyperion

I set the code for arbitrary model with a spherical source and grid with a self defined density. For mpi running, I've tried both m.run('model.rtout', mpi=True, n_processes=) & mpiexec -n hyperion_sph_mpi model.rtin model.rtout.

The code is running fine under mpi mode with 4 processses in my desktop. However, in HPC only serial mode is running. While running under mpi mode in HPC, there is no error message or anything in the out file, however the status shows that program is running. Below is a script file I tried in HPC

!/bin/tcsh

SBATCH --ntasks-per-node=32

mpiexec -n 32 hyperion_sph_mpi a.rtin a.rtout

I couldn't figure out if I'm doing anything wrong. Also, I'm new to HPC. Kindly help me.

astrofrog commented 4 years ago

@suman-phy - sorry for my delayed response! I'm also not sure what could be happening, but this might be an issue specific to your HPC set-up. Have you tried asking the administrators of the cluster?