Performance Results for Nested Parallelism on Multicore Architectures

Speaker: Professor David Walker,

School of Computer Science, Cardiff University.

Date/Time: 3rd December 2008, 13:00-14:00.

Location: Room G74, Philip Lyle Building.

Map: http://www.info.rdg.ac.uk/maps/maps-display.asp

Abstract:

This talk will present performance results for a molecular dynamics simulation on a number of cluster computers containing multicore processors. A comparison will be made between a pure MPI implementation, and a hybrid implementation that uses MPI to communicate data between cluster nodes and OpenMP to share data between the cores within a node. This latter case is an example of nested parallelism. The performance results will be interpreted in the context of a performance model that takes into account the hierarchical memory architecture of the target computers.