Branch-and-bound algorithms on a hypercube

Abstract

The parallel implementation of branch-and-bound algorithms on multiprocessors has received an increasing amount of attention in recent years. This paper describes research currently being conducted at Clemson University. We study the job scheduling problem: given n jobs, each with processing time, pi, and due date, di, schedule these jobs on a single machine such that total tardiness is minimized. Total tardiness, T, is defined as: T = n ∑ i=1 wi * max { 0, Ci - di } where Ci is the completion time and wi is the weight of job i. The hypercube used is a Floating Point Systems T-20: sixteen INMOS Transputers, synchronous communication, programmed in OCCAM. We describe an implementation focusing specifically on efficient communication and load balancing algorithms.