HPC Lecture Series

An Introduction to High-Performance Computing: A Lecture Series

From January to April 2009, a series of lectures on high-performance computing (HPC) will be available via Access Grid (AG), originating from the University of Alberta. Over several hours of lectures, with roughly one lecture per week, it is hoped that the audience will get a more thorough understanding of the theory, practice, and interdisciplinary nature of HPC than is typically possible in the more-standard one day or half-day workshop format.

These lectures are part of a course offered at the Univeristy of Alberta by Paul Lu (Associate Professor, Department of Computing Science). The course, CMPUT 497, is an introduction to high-performance computing (HPC), clusters, and parallel programming. For the undergraduate students registered in the course, they will learn about the historical context of HPC, some fundamental ideas (e.g. Amdahl's Law), and programming with Pthreads and MPI. For more information please visit: http://www.cs.ualberta.ca/~paullu/C497/.

Some of the background and conceptual topics may be of interest to a wider audience and will be offered via AG.

**Please note: All are welcome to attend, but you must RSVP to the appropriate university contact located on the Seminar Locations page.

Tuesday, January 20, 2009 - 10:00 PST / 11:00 MST / 12:00 CST

Let's Get Real: Amdahl's Law and the Cold Reality of Parallel Performance
Summary: Promises are often made about the performance of clusters and parallel systems. The decades-old Amdahl's Law provides some sobering insights into what is often achievable in practice. The news is not all bad, but...

Tuesday, February 3, 2009 - 10:00 PST / 11:00 MST / 12:00 CST

The Landscape of Parallel Computing Research: A View from Berkeley
Summary: In late 2006, several prominent researchers from UC Berkeley published a position paper about the future of parallel computing. The paper contains many interesting points worthy of discussion and debate. Download the paper for this seminar here.

Tuesday, February 10, 2009 - 10:00 PST / 11:00 MST / 12:00 CST

The Need for Diversity in HPC
Summary: A monoculture has developed in high-performance computing: Linux, commodity x86, commodity networks. Although these clusters have great price-performance, it would be short-sighted to put all of our eggs in that basket, especially with the emergence of manycore/multicore systems, GP-GPUs, etc. as key technologies.

Tuesday, February 24, 2009 - 10:00 PST / 11:00 MST / 12:00 CST

Basic Do's and Don'ts in Parallel Programming: Granularity is Key
Summary: It is hard enough to get parallel applications to execute correctly.But, improving parallel performance can be difficult if one doesn't have at least a basic plan of attack. For the beginner-to-intermediate parallel programmer, we will outline a basic, simple strategy for performance optimization. Many of these tips are very subjective and personal, but they will be based on fundamental ideas from performance evaluation and tuning.

Tuesday, March 10, 2009 - 10:00 PDT / 11:00 MDT / 11:00 CST / 12:00 CDT

The Cluster Challenge: Undergraduate Education and Training in HPC
Summary: The Cluster Challenge at Supercomputing 2007 and 2008 (and planned for 2009) provides an interesting datapoint into the state of undergraduate education and training in HPC. We provide some stories and observations drawn from coaching the University of Alberta team that placed First in 2007 and Third in 2008.

Tuesday, March 17, 2009 -10:00 PDT / 11:00 MDT / 11:00 CST / 12:00 CDT

The Large Hadron Collider (LHC) Grid
Guest Speaker: Dr. Bryan Caron, ATLAS Project, University of Alberta