Please select the most relevant MFiX category: | Installation | How to | Bug report | Share | for this topic.
How to
I am currently working on a DEM simulation, where the grid and particle numbers have reached the millions level. Due to this, a single node with a 32-core processor is unable to handle the computation. I would like to apply for multiple nodes for parallel computing. I am using Linux and MFIX version 21.1.4. Currently, I am still using the DMP calculation method, and the theoretical maximum speedup is only 0.2. Could anyone suggest tutorials or resources that I can use to learn more about setting up and running parallel simulations on multiple nodes?
Hi, kkzhou, welcome to the MFiX forum.
Where did you get the theoretical maximum speedup of 0.2?
MFiX does not support any parallel computation methods other than SMP and DMP, and SMP is limited to a single node. For computation on a cluster, DMP is what you need. You may need to make the grid coarse or reduce the number of particles to match the resources available.
A good online tutorial for MPI is Message Passing Interface (MPI) | LLNL HPC Tutorials
Two good books on the subject are “Using MPI” and “Using Advanced MPI” by Gropp and Lusk Using MPI and Using Advanced MPI
if you’ve got access to significant resources (say >~ 64 cpus) and are interested in Np > 10^6 DEM, you might want to consider using MFIX-Exa