Journal of Northeastern University ›› 2013, Vol. 34 ›› Issue (6): 774-777.DOI: -

• Information & Control • Previous Articles     Next Articles

Research of Conflict Detection Algorithm in STM

LIU Ying, GAO Fuxiang   

  1. School of Information Science & Engineering, Northeastern University, Shenyang 110819, China.
  • Received:2013-01-04 Revised:2013-01-04 Online:2013-06-15 Published:2013-12-31
  • Contact: LIU Ying
  • About author:-
  • Supported by:
    -

Abstract: To improve the performance of the parallel program execution by transactional memory, RSTM was chosen as the experimental system. According to the situation of the application, specific library implementation was used for optimization. STAMP was the benchmark program suit developed by Stanford University. Its six typical applications were used in this experiment to test nine library implementations in wordbased STM. And the influence by data version management was evaluated in conflict detection. The experimental results showed that TML running on various applications had the best effect while SGLA had the worst one. In addition, compared TML with TML+Lazy, the eager version management had the better effect, because the conflict could be detected and avoided in time. But in the case that there were many aborts or livelocks, the performance might be declined.

Key words: software transactional memory (STM), signature, Bloom filter, data version management, conflict detection

CLC Number: