[an error occurred while processing this directive]

Introduction to MPI: a Hands-on Tutorial

Primary tabs

Modern-day large-scale numerical simulations in science and engineering are often carried out on massively parallel supercomputers which consist of hundreds or even thousands of compute nodes. Computer programs written for such simulations need to be able to communicate data between different compute nodes. Such communications are commonly implemented through the Message Passing Interface, or MPI. In this tutorial, MPI basics such as point-to-point and collective communications will be introduced. The students will also get a chance to have hands-on exercises to turn serial example C programs into parallel programs using MPI. 
 
PREREQUISITE
Students should have experience writing C programs, be comfortable in a Unix/Linux programming environment, and know how to use a text editor. Specifically your computer should be capable of running ssh, either natively with MacIntosh/Linux boxes, or via putty or the equivalent on Windows systems. You should also know how to use ssh. Additionally, it is strongly recommended that you have a working knowledge of vi, emacs, or other command line editor. Finally, it is recommended that Windows users have functioning X server software on their Windows laptop, and know how to use it. We recommend Xming: http://sourceforge.net/projects/xming/files/latest/download?source=typ_redirect
 
There will not be time during the workshop to debug problems relating to the above items, so please make certain that you have the software and skills cited above, and that the software is working properly on your laptop.
 
Agenda
08:30 - 09:00 am Breakfast
09:00 - 10:15 am Lecture I: Introduction to MPI
10:15 - 10:45 am Coffee Break
10:45 - 12:00 pm Lecture II: Point-to-Point Communications
12:00 - 01:00 pm Lunch
01:00 - 02:15 pm Lecture III: Collective Communications
02:15 - 02:45 pm Coffee Break
02:45 - 04:00 pm Lecture IV: Other topics in MPI
 

Bio

Meifeng Lin received her PhD in Theoretical Particle Physics from Columbia University, and she is currently an Assistant Computational Scientist at the Computational Science Center at Brookhaven National Laboratory. Prior to coming to the lab she worked at Argonne National Laboratory and the Center for Computational Sciences at Boston University. She completed two postdocs at Yale and at MIT. Her areas of interest are computational lattice gauge theory; first-principles hadron structure calculations; HPC; large-scale data mining and analysis; scalable, adaptive and multilevel algorithms; and theoretical nuclear and particle physics.
Meifeng Lin

Speaker

Meifeng Lin

Date

Thursday, January 21, 2016

Time

8:30 am - 4:00 pm

Location

IACS Seminar Room

Registration

Registration