Multiword atomic read/write registers on multiprocessor systems

Authors:
Andreas Larsson;Anders Gidenstam;Phuong H. Ha;Marina Papatriantafilou;Philippas Tsigas
Affiliations:
Chalmers University of Technology, Göteborg, Sweden;Max Planck Institute for Computer Science;University of Tromsø;Chalmers University of Technology, Göteborg, Sweden;Chalmers University of Technology, Göteborg, Sweden
Venue:
Journal of Experimental Algorithmics (JEA)
Year:
2009

Citing 10
Cited 0

Constructing 1-writer multireader multivalued atomic variables from regular variables

Journal of the ACM (JACM)
Simple extensions of 1-writer atomic variable constructions to multiwriter ones

Acta Informatica
How to share concurrent wait-free variables

Journal of the ACM (JACM)
Evaluating the performance of non-blocking synchronization on shared-memory multiprocessors

Proceedings of the 2001 ACM SIGMETRICS international conference on Measurement and modeling of computer systems
Bounded concurrent timestamp systems using vector clocks

Journal of the ACM (JACM)
Modern Operating Systems

Modern Operating Systems
Parallel Computer Architecture: A Hardware/Software Approach

Parallel Computer Architecture: A Hardware/Software Approach
Operating System Concepts

Operating System Concepts
Integrating non-blocking synchronisation in parallel applications: performance advantages and methodologies

WOSP '02 Proceedings of the 3rd international workshop on Software and performance
Shared Memory Consistency Models: A Tutorial

Computer

Quantified Score

Hi-index	0.00

Visualization

Abstract

Modern multiprocessor systems offer advanced synchronization primitives, built in hardware, to support the development of efficient parallel algorithms. In this article, we develop a simple and efficient algorithm, the READERSFIELD algorithm, for atomic registers (variables) of arbitrary length. The simplicity and better complexity of the algorithm is achieved via the utilization of two such common synchronization primitives. In this article, we also experimentally evaluate the performance of our algorithm, together with lock-based approaches and a practical, previously previously known algorithm that is based that is based only on read and write primitives. The experimental evaluation is performed on three well-known parallel architectures. This evaluation clearly shows that both algorithms are practical and that as the size of the register increases the READERSFIELD algorithm performs better, following its analytical evaluation.