4:30 PM-6:30 PM
Room: Executive Salon 3
Monte Carlo computations are "naturally parallel" and can be run effectively on architectures ranging from MPP systems with ultra-low latency interconnection networks to machines volunteered over the world-wide web. The speakers in this minisymposium will describe a scalable library for pseudorandom number generation, that supports distributed memory applications and a new version of the cycle scavenger that supports this same functionality on networks of workstations. They will also describe new support for high performance Monte Carlo using the Java programming language, and present results on the dynamic creation of the highly touted "Mersenne Twister" random number generators for parallel and distributed applications.
Organizer: Michael Mascagni
University of Southern Mississippi
MMD, 1/5/99