Jste zde

Parallel Programming with MPI

Termín: 
Po, 14.12.2015 9:30 - Út, 15.12.2015 15:30
Uzávěrka registrace: 
Po, 07.12.2015 23:30
Místo: 
VŠB - Technical University Ostrava, IT4Innovations building, room 207
Lektor: 
Pavan Balaji (Argonne National Laboratory, USA)
Úroveň: 
Basic/Intermediate (day 1) and Intermediate/Advanced (day 2)
Jazyk: 
English

Annotation

Message Passing Interface (MPI) is the most prominent programming model for parallel programming on distributed memory systems.  Virtually every supercomputer in the world provides a tuned implementation of MPI, and practically every large distributed-memory application uses MPI either directly or indirectly.  While MPI has evolved from a bare-minimum message passing functionality to an feature and functionality rich programming model, most users are not aware of its capabilities and consequently tend to use the minimal set of functionality. 

This tutorial focuses on introducing various concepts in MPI to the attendees.  The first day, having the basic/intermediate level, will focus on basic material including understanding the core concepts in MPI, and how to program with MPI.  On the second day, characterized as intermediate/advanced (or even towards the end of the first day already), we shall focus on more advanced material including those introduced in MPI-2.2 and MPI-3.  Towards the end of the tutorial, we shall spend some time discussing the different proposals that are being pursued for MPI-4.

About the tutor

Dr. Pavan Balaji is a Computer Scientist and Group Lead at Argonne National Laboratory.  His team develops the MPICH implementation of MPI, which is used on nine out of the top 10 supercomputers in the world.  Dr. Balaji also chairs the hybrid programming working group within the MPI Forum, and contributes to almost every aspect of the MPI standard.  He is an author of the MPI-2.1, MPI-2.2, MPI-3.0 and MPI-3.1 standards, and is heavily involved in the preparation of the upcoming MPI-4 standard.  He is also an editor of the "Programming Models for Parallel Computing" book that introduces all of the commonly used programming models in parallel computing in a nutshell.

Preliminary schedule

Monday December 14,  2015
09:30-10:00

registration

10:00-11:30

Introductory concepts in MPI

11:30-13:00

lunch break

13:00-14:30

Point-to-point communication

14:30-15:00

coffee break

15:00-16:30

Group communication and Derived Datatypes

16:30-17:00

coffee break

17:00-18:00

Nonblocking collectives

 

Tuesday December 15,  2015
  
09:00-10:30

One-sided communication

10:30-11:00

coffee break

11:00-12:45

Hybrid programming

12:45-14:00

lunch break

14:00-15:30

Network locality and Topology

Prerequisites

Notebook to access the remote system for the hands-on part.

Registration

Obligatory registration - registration form here; deadline see above or exhausted course capacity.

Fees

The event is provided free of charge for the participants.

Capacity

30 attendees

Practicalities

  • See a page on transport and accommodation (in Czech) how to get to the campus of  VŠB - Technical University Ostrava and to the IT4Innovations building.
  • Participants without the IT4Innovations card please arrive early enough to settle the formalities with obtaining an entry permit.
  • System documentation is available at http://support.it4i.cz/docs.
Přílohy: