Earticle

Debugging multithreaded or parallel programs is a hard problem due to unintended non-deterministic results of the program, called data races. Previous race detection techniques have limitations of coverage for synchronization constructs in detecting first races for parallel programs with random synchronizations. We present, in this paper, an efficient two-pass algorithm to detect the first races in a particular execution of parallel program, by eliminating independent accesses to make a performance gain and candidate accesses, and then analyzing the traced information with candidate accesses by replaying the program. Therefore, our technique makes the race detection more practical in debugging parallel programs which have random synchronizations.

Detecting data races in multi-threaded programs is a challenging problem in debugging, because the races could reveal nondeterministic program behavior in execution of such pro- grams. Nondeterministic runtime eects of a race are hard to identify and it is dicult to decide whether the reported data races can appear or not in the actual program executions. Previous techniques for detecting races cannot provide method to locate rst races in par- allel programs with random synchronization including lock/unlock mechnism. This paper presents an algorithm which extracts rst races by replaying the program and checking con- currency between sequenced traced data and candidate accesses, which are from a particular execution of parallel programs. We also present the correctness of our algorithm by showing that all the rst races are included in the traced accesses which are composed of key accesses.