Task partitionings for parallel triangular solver on a MIMD computer

Abstract

Considers a parallel triangular solver on a distributed-memory MIMD computer. Three task partitioning methods are discussed, with both task assignment and task scheduling. Their estimated times are provided by using a performance model and a parallel performance evaluation methodology. The optimal task granularities are deduced by the analysis of their performance. Experiences on a transputer-based multicomputer are given.