Cost of mutual exclusion with spin locks on multi-core CPUs

Authors:
Sándor Juhász;Ákos Dudás;Tamás Schrádi
Affiliations:
Budapest University of Technology and Economics, Department of Automation and Applied Informatics, Hungary;Budapest University of Technology and Economics, Department of Automation and Applied Informatics, Hungary;Budapest University of Technology and Economics, Department of Automation and Applied Informatics, Hungary
Venue:
BICA'12 Proceedings of the 5th WSEAS congress on Applied Computing conference, and Proceedings of the 1st international conference on Biologically Inspired Computation
Year:
2012

Citing 6
Cited 0

Synchronization Algorithms for Shared-Memory Multiprocessors

Computer
Algorithms for scalable synchronization on shared-memory multiprocessors

ACM Transactions on Computer Systems (TOCS)
Parallelization of a Dynamic Unstructured Algorithm Using Three Leading Programming Paradigms

IEEE Transactions on Parallel and Distributed Systems
The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors

IEEE Transactions on Parallel and Distributed Systems
Shared-memory mutual exclusion: major research trends since 1986

Distributed Computing - Papers in celebration of the 20th anniversary of PODC
Real-Time Synchronization on Multiprocessors: To Block or Not to Block, to Suspend or Spin?

RTAS '08 Proceedings of the 2008 IEEE Real-Time and Embedded Technology and Applications Symposium

Quantified Score

Hi-index	0.00

Visualization

Abstract

Multi-core CPUs promise great performance increase in applications and algorithms that can take advantage of them. Such algorithms, offering high level of parallelism, must protect data structures from concurrent modifications possibly violating data integrity. Synchronization mechanisms offer protection from such erroneous behavior by limiting data modifications to a single thread. In shared memory systems, such as multi-core CPUs, synchronization is cheap as threads can communicate with each other via locks placed in the main system memory. These locks, however, present as performance overhead not only by limiting the level of parallelism but also by changing the characteristics of the application. By investigating various implementations of mutual exclusion mechanisms with the use of spin locks an overview is presented of synchronization options in multi-core systems and the performance of several spin locks is evaluated.