Abstract
Solving large combinatorial optimisation problems is often time consuming, and
thus there is interest in accelerating current algorithms by building
application specific computers. This paper focuses on accelerating general
local search meta-heuristics, such as simulated annealing and tabu search, and
presents an architecture for this class of algorithms. As a design case study
we describe a specific machine which implements a simulated annealing based
algorithm for solving the Travelling Salesman Problem (TSP). This processor
which is built with XILINX field programmable logic and APTIX interconnection
matrices achieves a speedup of about 37 times over an IBM RS6000 workstation.
The paper highlights the use of reconfigurable memory as a key for obtaining
high performance.