Multicore Meets Petascale: The Catalyst for a Software Revolution

Kathy Yelick

EECS, UC Berkeley and National Energy Research Scientific Computing Center, LBNL
Slides

Petascale systems will soon be available to the computational science community at multiple sites. These systems will represent a variety of architectural models, but with one common component, which is an increasing reliance on multicore technology as the building block for these machines. This shift towards on-chip parallelism brings into question the message-passing programming model that has dominated high-end programming for the past decade. In this talk I will describe some of the challenges that application scientists and algorithm developers will face in utilizing future Petascale systems. I will also present an alternative to message passing called Partitioned Global Address Space (PGAS) languages, describe some of the performance and scaling advantages, and also advocate a new notion of high end programming in which software is designed up-front to be adaptable to current and future systems.