Using FPGAS as control support in MIMD executions

Abstract

In the wide field of parallel architectures, machines involving Fpgas on each node have appeared during the last years. Connecting these reconfigurable components opens new research horizons. Expensive control tasks required by distributed applications may then be accelerated using application specific hard-wired elements. A derivation from a standard algorithm for distributed simulation and its partial hardware implementation are presented in this paper, providing three magnitude orders speed-ups.