Search⌘ K

Learn to Use HPC Systems and Supercomputers

0%

1.Supercomputers and HPC clusters

A Little bit of Supercomputing history Supercomputing examples HPC cluster computers Benefits of using cluster computing

2.Components of a HPC system

Components of a HPC cluster Login node(s)Compute node(s)Master node(s)Storage node(s)

3.HPC software stack

Access to HPC Data Transfer HPC software list HPC software modules Job Schedulers

4.PBS - Portable Batch System

Introduction to PBS PBS basic commands PBS `qsub`PBS `qstat`PBS `qdel` command PBS `qalter`PBS job states PBS variables A simple PBS job script PBS interactive jobs PBS arrays PBS Matlab example

5.SLURM -Workload Manager

Introduction to Slurm Slurm commands A simple Slurm job Slurm distrbuted MPI and GPU jobs Slurm multi-threaded OpenMP jobs Slurm interactive jobs Slurm array jobs Slurm job dependencies

6.Parallel programming - OpenMP

OpenMP OpenMP basics Open MP - clauses OpenMP - worksharing constructs OpenMP- Hello world!Open MP - reduction and parallel `for-loop`OpenMP - section parallelization OpenMP vector addition

7.Parallel programming - MPI

MPI - Message Passing Interface MPI program stucture MPI - hello world!MPI send/ receive MPI `ping-pong` send and receive example

8.Parallel programming - GPU and CUDA

GPUs - graphics processing units GPU Programming - CUDA CUDA - hello world!CUDA - vector additon demo

9.Conclusions and references

Conclusion References

Learn to Use HPC Systems and Supercomputers

Slurm distrbuted MPI and GPU jobs

We'll cover the following...

A MPI job can considered as a cross-node and multi-process job. Sample Slurm MPI job script would look like below:

Shell

#!/bin/bash
#SBATCH --job-name=MyJob
#SBATCH --account=username
#SBATCH --ntasks=32
#SBATCH --ntasks-per-node=16
#SBATCH --cpus-per-tasks=1
module load openmpi
mpiexec <program>

The script ...