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 all 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.

Content Levels: Beginners = 0:00h (0%) + Intermediate = 1:30h (10%) + Advanced = 13:30h (90%)

1st day – 22 June 2022

08:45      Join online
09:00      Welcome
09:05      Motivation
09:15      Introduction
09:30      Programming Models
09:35      MPI + OpenMP
10:00      Practical (how to compile and start)
10:30      Break
10:45      continue: MPI + OpenMP
11:30      Break
11:45      continue: MPI + OpenMP
12:30      Practical (how to do pinning)
13:00      Lunch
14:00      Practical (hybrid through OpenMP parallelization)
15:30      Q & A, Discussion
16:00      End of first day

2nd day – 23 June 2022

08:45   Join online
09:00    Overlapping Communication and Computation
09:30    Practical (taskloops)
10:30    Break
10:45    MPI + OpenMP Conclusions
11:00    MPI + Accelerators
11:30    Tools
11:45    Break
12:00    Programming Models (continued)
12:05    MPI + MPI-3.0 Shared Memory
13:00    Lunch
14:00    Practical (replicated data)
15:30    Q & A, Discussion
16:00    End of second day

3rd day – 24 June 2022

08:45    Join online
09:00    MPI Memory Models and Synchronization
09:40    Pure MPI
10:00    Break
10:15    Recap - MPI Virtual Topologies
10:45    Topology Optimization
11:15    Break
11:30    Practical/Demo (application aware Cartesian topology)
12:30    Topology Optimization (Wrap up)
12:45    Conclusions
13:00    Lunch
14:00    Finish the hands-on labs, Discussion, Q & A, Feedback
16:00    End of third day (course)

Prerequisites:

Basic MPI and OpenMP knowledge as presented in our VSC courses on MPI and OpenMP.
For the hands-on sessions you should know Unix/Linux and either C/C++ or Fortran.

Hands-on labs:

The hands-on labs will be done on a cluster provided by LRZ.

Lecturers:

Georg Hager (NHR@FAU, Uni. Erlangen), Rolf Rabenseifner (HLRS, Uni. Stuttgart), Claudia Blaas-Schenner (VSC Research Center, TU Wien)

Language: English

Prices and Eligibility:

This course is a PRACE Training Center (PTC) event. Therefore, the course is open and free of charge for people from academia and industry from the Member States (MS) of the European Union (EU) and Associated/Other Countries to the Horizon 2020 programme.