1. Fall 2005
1.1. Wednesday, September 14, 2005, 2:30pm
Living in a Peta-flop world
Back in the "day," high performance computing (HPC) meant that you had access to a Cray-1 supercomputer capable of an amazing 80 million floating point operations a second (MFLOPS). To get access to this startlingly fast resource, you needed to recast all of your (FORTRAN) programs into vectorized form to take advantage of the underlying hardware. Fast forward to the 1990's when the first gigaflop machines appeared, again with specialized, vastly parallel hardware and programming needs (e.g. Connection Machine's CM-5). The first teraflop machines in the late 90's were clusters with hundreds of processors requiring MPI libraries and SPMD style programming.
Now we stand on the threshold of petaflop machines with hundreds of thousands of processors. The old models of programming are yet again insufficient to deal with the new architectures! Dr. Miller will talk about ways to cope. He will discuss Livermore's Petascale Simulation Initiative which is developing new MPMD programming models; pyMPI, his parallel tool for interacting with thousands of processors; and his hobby project -- Science by FlashMob, which lets scientists cheaply and quickly create small to moderate sized parallel computer clusters.
Dr. Miller is a computational scientist for the Center for Applied Scientific Computing at Lawrence Livermore National Laboratory and a Consulting Professor at Stanford University's Institute for Computational and Mathematical Engineering. He specializes in the creation and understanding of parallel programming languages.
1.2. September 15, 2005: Flashmob Supercomputing Event, 5:30pm
5:30pm to 9pm in Thomas Great Hall.
http://membership.acs.org/C/COMP/flashmob.html
2. Spring 2006
Zach Dodds, Pomona College
