ACM Home Page
Please provide us with feedback. Feedback
Transactional memory and the birthday paradox
Full text PdfPdf (106 KB)
Source ACM Symposium on Parallel Algorithms and Architectures archive
Proceedings of the nineteenth annual ACM symposium on Parallel algorithms and architectures table of contents
San Diego, California, USA
SESSION: Brief announcements III: parallel computing table of contents
Pages: 303 - 304  
Year of Publication: 2007
ISBN:978-1-59593-667-7
Authors
Craig Zilles  University of Illinois at Urbana-Champaign
Ravi Rajwar  Intel Corporation
Sponsors
ACM: Association for Computing Machinery
SIGACT: ACM Special Interest Group on Algorithms and Computation Theory
SIGARCH: ACM Special Interest Group on Computer Architecture
Publisher
ACM  New York, NY, USA
Bibliometrics
Downloads (6 Weeks): 8,   Downloads (12 Months): 57,   Citation Count: 4
Additional Information:

abstract   references   cited by   index terms   collaborative colleagues  

Tools and Actions: Request Permissions Request Permissions    Review this Article  
DOI Bookmark: Use this link to bookmark this Article: http://doi.acm.org/10.1145/1248377.1248428
What is a DOI?

ABSTRACT

Many word-based Software TransactionalMemory systems (STMs) have been proposed using tagless ownership tables, where read and write permissions are granted at the granularity of all addresses that map to a given ownership table entry. This optimization to reduce overhead potentially results in false conflicts. Using address traces from a multithreaded program, we demonstrate that the frequency of these false conflicts grows superlinearly with both the TM data footprint and concurrency and that increasing the size of the ownership table results in only a sub-linear reduction in conflict rate. These somewhat surprising relationships have a theoretical foundation that is also responsible for the (naively) unintuitive statistical result generally referred to as the "Birthday Paradox." We present an analytical model based on random population of an ownership table by concurrently executing transactions that correctly predicts the trends in measured data. These results call into question the viability of such an optimization that can undermine the scalability and concurrency claims of software transactional memory.




Collaborative Colleagues:
Craig Zilles: colleagues
Ravi Rajwar: colleagues