Clustering and reassignment-based mapping strategy for message-passing architectures

• Published in: Journal of systems architecture Vol. 48; no. 8; pp. 267 - 283
• Main Authors: Senar, Miquel A., Ripoll, Ana, Cortés, Ana, Luque, Emilio
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.03.2003; Elsevier; Elsevier Sequoia S.A
• Subjects: Algorithms; Applied sciences; Clustering; Computer science; control theory; systems; Computer systems and distributed systems. User interface; Distributed processing; Exact sciences and technology; Graph partitioning; Heuristic; Heuristic algorithms; Mapping; Mapping problem; Scheduling; Software; Studies; Task allocation/assignment; Graph partitioning; Heuristic algorithms; Clustering; Mapping problem; Task allocation/assignment; Program graph; Compilation; Task scheduling; Trade off; Mapping; Parallel architectures; Computational complexity; Message passing; Parallel system; Parallel program
• ISSN: 1383-7621; 1873-6165
• DOI: 10.1016/S1383-7621(03)00011-0

A fundamental issue affecting the performance of a parallel application running on message-passing parallel systems is the assignment of tasks to processors. In this paper we present a compilation-time two stage mapping strategy (denoted as Task Allocation by Clustering, Reassignment and Embedding, TACRE) used for mapping arbitrary programs (modeled as task interaction graphs) onto message-passing parallel systems. The first stage is based on task clustering and task reassignment algorithms that contract the original task graph. The second stage takes the contracted graph and tries to well match the physical properties of the target system. The results shown that TACRE provides a good trade-off between mapping quality and computational complexity.