Parallel Computing
Introduction
The following four types of parallel computing are available for you to consider depending on the nature of your computational work and your preference.
Model | Description |
---|---|
OpenMP | Running a program by multithreading method on multiple cores within a compute node |
MPI | Running a program on multiple cores either within a compute node or crossing multiple compute nodes |
GPU | Running a program by multithreading method on multiple GPU cores within a GPU compute node |
To learn more about the GPU computing and Grid computing, please check at the relevant website.
Parallel Computing on HPC Clusters
OpenMP and MPI parallel computing are supported on the HPC clusters. Licensed softwares such as Abaqus, Amber, Gaussian, Fluent and Material Studio can be run in parallel mode on the HPC clusters as well.
To run parallel OpenMP/MPI program and those softwares, you shall submit the jobs to an LSF parallel batch queues. You can list out the LSF parallel queues by:
# hpc queue ======== Status of Parallel Queues ======== Queue : Running : Waiting flexi : 11 : 194 parallel20 : 37 : 33 parallel24 : 54 : 115 parallel12 : 177 : 14 parallel8 : 6 : 0 ------------------------------------------- === Status of Serial/Openmp/App Queues ==== Queue : Running : Waiting gpu : 6 : 0 openmp : 12 : 6 serial : 72 : 198 short : 51 : 183 volta_gpu : 9 : 0 -------------------------------------------
The details and the instructions for running OpenMP/MPI program and licensed softwares on the HPC clusters can be found below.
Number | Description |
---|---|
Abaqus | Finite Element Analysis (FEA) Software |
Amber | Molecular mechanical force fields and simulation program |
Gaussian | ab initio quantum mechanics |
Fluent | Generic Computational Fluid Dynamics (CFD) package FAQS |
MPI Codes | MPI parallel program computing |
OpenMP Codes | OpenMP parallel program computing |
Gromacs | Molecular dynamics package |
Quantum Espresso | Molecule Structure Calculation and Nanoscale Material Modeling |