Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
Title of Dissertation: Scalable Synchronization in Shared Memory Multiprocessing Systems
 

Summary: Abstract
Title of Dissertation: Scalable Synchronization in Shared Memory
Multiprocessing Systems
Jae-Heon Yang, Doctor of Philosophy, 1994
Dissertation directed by: Assistant Professor James H. Anderson
Department of Computer Science
It is our thesis that scalable synchronization can be achieved with only mini-
mal hardware support, speci cally read/write atomicity. This is contrary to the
conventional viewpoint that stronger hardware support is required for scalable
synchronization such support not only requires additional cost, but also leads
to portability problems.
As evidence in support of our thesis, we present a new scalable mutual ex-
clusion algorithm based on read and write instructions. The performance of this
algorithm is better than prior mutual exclusion algorithms based on read/write
atomicity, and even rivals that of the fastest mutual exclusion algorithms that
require stronger primitives. Our algorithm is based on the technique of local
spinning, i.e., busy-waiting on variables that are locally-accessible to the wait-
ing process. Local-spinning minimizes remote accesses of shared memory, which
tend to dominate performance under heavy contention.
An obvious question left open by the algorithm described above is whether

  

Source: Anderson, James - Department of Computer Science, University of North Carolina at Chapel Hill

 

Collections: Computer Technologies and Information Sciences