FastTrack: efficient and precise dynamic race detection

  • Authors:
  • Cormac Flanagan;Stephen N. Freund

  • Affiliations:
  • University of California at Santa Cruz, Santa Cruz, CA;Williams College, Williamstown, MA

  • Venue:
  • Communications of the ACM
  • Year:
  • 2010

Quantified Score

Hi-index 48.23

Visualization

Abstract

Multithreaded programs are notoriously prone to race conditions. Prior work developed precise dynamic race detectors that never report false alarms. However, these checkers employ expensive data structures, such as vector clocks (VCs), that result in significant performance overhead. This paper exploits the insight that the full generality of VCs is not necessary in most cases. That is, we can replace VCs with an adaptive lightweight representation that, for almost all operations of the target program, requires constant space and supports constant-time operations. Experimental results show that the resulting race detection algorithm is over twice as fast as prior precise race detectors, with no loss of precision.