About PPCES

This one week online event will continue the tradition of previous annual week-long events that take place in Aachen every spring since 2001. We will cover the basics of parallel programming using OpenMP and MPI in Fortran and C/C++ and a first step towards performance tuning as well as current topics in AI/machine learning. Hands-on exercises for each topic will be included.

The contents of the courses are generally applicable but will be specialized towards CLAIX the compute cluster which is the current system installed at the RWTH’s IT Center. It might be helpful to read through the information which was provided during the HPC introduction on March 12 this year. This is especially true if you want to actively use CLAIX after this event.

OpenMP is a widely used approach for programming shared memory architectures, supported by most compilers nowadays. We will cover the basics of the programming paradigm as well as some advanced topics such as programming NUMA machines. The nodes of the RWTH Compute Cluster contain an increasing number of cores and thus we consider shared memory programming a vital alternative for applications that cannot be easily parallelized with MPI. We also expect a growing number of application codes to combine MPI and OpenMP for clusters of nodes with a growing number of cores.

The Message Passing Interface (MPI) is the de-facto standard for programming large HPC systems. We will introduce the basic concepts and give an overview of some advanced features. Also covered is hybrid parallelization, i.e., the combination of MPI and shared memory programming, which is gaining popularity as the number of cores per cluster node grows.

Machine Learning: We provide an overview to end-to-end deep learning with the latest version of Tensorflow/Keras. It covers the basic concepts to define models with Keras  and data pipelines with Tensorflow’s “Dataset”, and to visualize the results with Tensorboard while training. If training on one node or GPU is not enough,  we show how to scale up/out distributed training onto multiple compute nodes  and GPUs with Horovod. Furthermore, we provide an introduction to scikit-learn, with an overview of  different machine learning algorithms it provides and how to utilize it on GPUs  with H2O4GPU. The training courses consist of a hands-on exercises to be run directly on  RWTH infrastructure.

Guest Speakers

We are very happy to present two guest speakers:

• Ruud van der Pas (Oracle) on OpenMP
• Georg Zitzlsberger (IT4Innovations National Supercomputing Center at VSB – Technical University of Ostrava) on Machine Learning

Agenda

Day 1+2: OpenMP 

Monday, March 22

Day 3+4: MPI

Day 5: Machine Learning 

Seminar times will be 9:00-12:00 and 13:00-15:00.

This event is partially supported by The [Czech] Ministry of Education, Youth and Sports from the Large Infrastructures for Research, Experimental Development and Innovations project  “e-Infrastruktura CZ – LM2018140”

Prerequisites

Attendees of part I and II should be comfortable with C/C++ or Fortran programming in a Linux environment and interested in learning more about the technical details of application tuning and parallelization.
Participants of part III – machine learning –  will need some basic knowledge of Python.

All presentations will be given in English.

This event will be an online presentation.
All all parts of the tutorials will be accompanied by exercises.

Participants who have access to the RWTH identity management can use their own HPC account.
Those members of RWTH who do not yet have such an account can provide an HPC account here (https://sso.rwth-aachen.de/idp/profile/SAML2/Redirect/SSO?execution=e1s1) using the selfservice (Choose: Accounts und Kennwörter – Account anlegen – Hochleistungsrechnen)

External participants must provide themselves a Linux environment that contains an OpenMP compiler, a MPI library, or respectively a singularity environment.
For parts I and II  a Linux virtual machine will be sufficient.

For example on Ubuntu 20.04 LTS  the following commands can be used to install the necessary software für OpenMP and MPI:

# g++ is  available by default
sudo apt install gfortran # install  Fortran Compiler - if necessary
sudo apt-get install libomp-dev # install OpenMP libraries
sudo apt install mpich # install MPI library

Simple program examples can be compiled and executed by

g++ -fopenmp openmp-hello-world.cc;   ./a.out
gfortran -fopenmp openmp-hello-world.f90;  ./a.out
mpicc mpi_hello_world.c -o  ./a.out;  mpirun -np 2  ./a.out

For part III (ML) participants need to run singularity containers with access to one or more NVIDIA GPUs.

Course Material of PPCES 2021

OpenMP

Presentations

• Organization_PPCES2021
• 00-openmp-CT-welcome
• 01-openmp-CT-overview
• 02-openmp-CT-parallel_region
• 03-openmp-CT-worksharing
• 04-openmp-CT-scoping
• 05-openmp-CT-compilers
• 06-openmp-CT-welcome
• 07-openmp-CT-tasking_motivation
• 08-openmp-CT-tasking_model
• 09-openmp-CT-taskloop
• 10-openmp-CT-tasking_cutoff
• 11-openmp-CT-tasking_dependencies
• 12-openmp-CT-welcome
• 13-openmp-CT-NUMA
• 14-openmp-CT-tasking_affinity
• 15-openmp-CT-hybrid
• 16-openmp-CT-SIMD
• 17-openmp-CT-offloading

Exercises

• Exercises_OMP_2021
• openmp_exercises.zip

MPI

Presentations

• hpc.nrw-01-MPI-Overview
• hpc.nrw-02-MPI_Concepts
• hpc.nrw-03-Blocking_Point-to-Point_Communication
• hpc.nrw-04-Non-blocking-Point-to-Point_Communication
• hpc.nrw-05-Derived_Datatypes
• hpc.nrw-06-Blocking_Collective_Communication
• hpc.nrw-08-Hybrid_Programming
• hpc.nrw-09-Communicator_Handling
• hpc.nrw-XX-Advanced_Topics

Exercises

• ppces2021-MPI-labs-C.zip
• ppces2021-MPI-labs-Fortran.zip

Machine Learning

Presentations

• 00_Agenda
• 01a_Introduction_to_scikit-learn
• 01b_scikit-learn_Optimizations
• 03_scikit-learn_Hands-On
• 04_Introduction_Deep_Neural_Networks
• 05_Tensorflow_Keras
• 06_Multi-GPU_Horovod

Exercises

• 2021-ppces-ML-DL-instructions
• 2021-ppces-exercises-ML-DL.zip

Further Information

The OpenMP part is also available as online tutorial (including videos): https://hpc-wiki.info/hpc/OpenMP_in_Small_Bites