Parallel and Distributed Processing Symposium, International (2001)
San Francisco, California, USA
Apr. 23, 2001 to Apr. 27, 2001
<p>This work introduces a new technique that enables SDSMs to categorize dynamically and accurately memory sharing patterns in both classes of regular and irregular applications. The categorization is carried out automatically at run-time on a per-page basis, requiring no user or compiler assistance.</p> <p>We evaluate the potential benefits of our technique using execution-driven simulations of 8 applications running on TreadMarks on a network of 8 workstations. Surprisingly, we found that producer-consumer(s) and migratory are the dominant patterns even in irregular applications. Preliminary results suggest that the categorization technique we propose is a promising option to further improve the performance of current adaptive SDSM systems.</p>
Adaptive software DSM systems, memory sharing patterns, latency-hiding techniques
M. C. Castro and C. d. Amorim, "Efficient Categorization of Memory Sharing Patterns in Software DSM Systems," Parallel and Distributed Processing Symposium, International(IPDPS), San Francisco, California, USA, 2001, pp. 10063a.