Overview

This course has been cancelled due to corona virus.

For information see also the LRZ course page

Most HPC systems are clusters of shared memory nodes. Such SMP nodes can be small multi-core CPUs up to large many-core CPUs. Parallel programming may combine the distributed memory parallelization on the node interconnect (e.g., with MPI) with the shared memory parallelization inside of each node (e.g., with OpenMP or MPI-3.0 shared memory). This course analyzes the strengths and weaknesses of several parallel programming models on clusters of SMP nodes. Multi-socket-multi-core systems in highly parallel environments are given special consideration. MPI-3.0 has introduced a new shared memory programming interface, which can be combined with inter-node MPI communication. It can be used for direct neighbor accesses similar to OpenMP or for direct halo copies, and enables new hybrid programming models. These models are compared with various hybrid MPI+OpenMP approaches and pure MPI. Numerous case studies and micro-benchmarks demonstrate the performance-related aspects of hybrid programming.

Hands-on sessions are included on both days. Tools for hybrid programming such as thread/process placement support and performance analysis are presented in a "how-to" section. This course provides scientific training in Computational Science, and in addition, the scientific exchange of the participants among themselves. The course is a PRACE Advanced Training Center event. It is organized by LRZ in cooperation with HLRS, RRZE, and VSC (Vienna Scientific Cluster).

Agenda & Content

Agenda & Content (preliminary)

1st day

09:30   Registration
10:00    Welcome
10:05    Motivation
10:15    Introduction
10:45    Programming Models
10:50 - MPI + OpenMP
11:30 Coffee Break
11:50 - continue: MPI + OpenMP
12:40 Practical (how to compile and start)
13:00   Lunch
14:40 Practical (continued)
15:00 Practical (hybrid through OpenMP parallelization)
16:00 Coffee Break
16:20 - Overlapping Communication and Computation
16:40 Practical (taskloops)
17:20 - MPI + OpenMP Conclusions
17:30 - MPI + Accelerators
17:45    Tools
18:00 End of first day

19:00 Social Event at Gasthof Neuwirt (self paying)

2nd day

09:00    Programming Models (continued)
09:05 - MPI + MPI-3.0 Shared Memory
09:45 Practical (replicated data)
10:30 Coffee break
10:50 continue: Practical (replicated data)
11:50 - MPI Memory Models and Synchronization
12:30 Lunch
13:30 - Pure MPI
13:50 - Topology Optimization
14:30 Coffee Break
14:50 Practical (application aware Cartesian topology)
15:45 - Topology Optimization (Wrap up)
16:00    Conclusions
16:15    Q & A
16:30 End of second day (course)

Prerequisites

Basic MPI and OpenMP knowledge as presented in the LRZ course "Parallel programming of High Performance Systems". For the hands-on sessions you should know Unix/Linux and either C/C++ or Fortran in particular.

Language

The course language is English.

Course material

See http://tiny.cc/MPIX-LRZ

Teacher

Dr. habil. Georg Hager (RRZE/HPC, Uni. Erlangen), Dr. Rolf Rabenseifner (Stuttgart), Dr. Claudia Blaas-Schenner and Dr. Irene Reichl (VSC Team, TU Wien)

Registration & Further Information

Registration via https://events.prace-ri.eu/event/902/registrations/634/

For further information, see also course pages

at LRZ: http://www.lrz.de/services/compute/courses/2020-04-20_hhyp1s20/ (valid only before the course)
and after the course at: http://www.lrz.de/services/compute/courses/archive/2020/2020-04-20_hhyp1s20/
at PRACE: http://events.prace-ri.eu/e/LRZ-2020-HHYP1S20
and at VSC (Vienna Scientific Cluster) http://vsc.ac.at/training/2020/HY-LRZ

Deadline

for registration is Apr. 5, 2020.

Fee

This course is a PRACE Advanced Training Center event. Therefore, the course is free of charge for all participants from the EU or from PRACE-member-countries.