[Aces-support] Parallel Programming course IAP 12.950

[Constantinos Evangelinos]

Please excuse the mass mailing - I hope this proves to be of interest to at 
least some of the students or researchers that are computationally active and 
feel the need to parallelize their codes (or maintain and enhance already 
parallel codes they are working with). The class starts this Wednesday and 
you can register in class if you prefer. Listeners are also very welcome 
though taking the course for credit should not be avoided.

The class will use the latest integrated development environment/debugger for 
parallel computing (PTP - http://www.eclipse.org/ptp) and students will be 
able to use either a virtual machine the class will provide or accounts on 
the Amazon Elastic Compute Cloud (Amazon EC2), in both cases with all the 
software setup for them. Should one prefer it, a student can also use the 
Athena dialup machines or the ACESgrid compute clusters.

Here is the class announcement:
http://student.mit.edu/searchiap/fs-12-950.html

Parallel Programming for Multicore Machines Using OpenMP and MPI
Constantinos Evangelinos
Wed Jan 23, Fri Jan 25, Mon Jan 28, Tue Jan 29, Thu Jan 31, 02-04:00pm, 
54-1510

Pre-register on WebSIS and attend first class.
Limited to 30 participants.
Listeners allowed, space permitting
Prereq: Good programming skills in C/C++ or Fortran
Level: H 2 units Standard A - F Grading Can be repeated for credit   

You just got a brand new quad core machine but don't know how to use it to 
solve your research problems? Then it might be time to consider using MPI 
and/or OpenMP to recode your applications in parallel.

OpenMP is a standard API for shared memory parallel programming that allows 
for incremental parallelization of your codes while providing the constructs 
necessary for performance tuning; it has demonstrated scaling on up to 8
processor cores and is well suited for parallel execution on multicore chips. 
OpenMP use demonstrated with model scientific codes.

MPI, the Message Passing Interface is a highly successful, universal API 
standard for distributed memory parallel programming. Used in applications 
scaling to thousands of processors, MPI may appear daunting. This course will
start with basic principles.

Web: http://eapsweb.mit.edu/courses/iap.html
Contact: Constantinos Evangelinos, 54-1518, x3-5259, ce107 at MIT.EDU
-- 
Dr. Constantinos Evangelinos
Department of Earth, Atmospheric and Planetary Sciences
Massachusetts Institute of Technology