Making asynchronous parallelism safe for the world
POPL '90 Proceedings of the 17th ACM SIGPLAN-SIGACT symposium on Principles of programming languages
An empirical comparison of monitoring algorithms for access anomaly detection
PPOPP '90 Proceedings of the second ACM SIGPLAN symposium on Principles & practice of parallel programming
Detecting access anomalies in programs with critical sections
PADD '91 Proceedings of the 1991 ACM/ONR workshop on Parallel and distributed debugging
On-the-fly detection of data races for programs with nested fork-join parallelism
Proceedings of the 1991 ACM/IEEE conference on Supercomputing
What are race conditions?: Some issues and formalizations
ACM Letters on Programming Languages and Systems (LOPLAS)
Online data-race detection via coherency guarantees
OSDI '96 Proceedings of the second USENIX symposium on Operating systems design and implementation
Efficient detection of determinacy races in Cilk programs
Proceedings of the ninth annual ACM symposium on Parallel algorithms and architectures
Eraser: a dynamic data race detector for multi-threaded programs
Proceedings of the sixteenth ACM symposium on Operating systems principles
Detecting data races in Cilk programs that use locks
Proceedings of the tenth annual ACM symposium on Parallel algorithms and architectures
Applications of Path Compression on Balanced Trees
Journal of the ACM (JACM)
Communications of the ACM
| Hi-index | 0.00 |
When either of two concurrent accesses of the same data is not in its critical section, a "race condition" occurs. Previous race-detection techniques are only applicable to parallel programs without using critical sections, or to parallel programs with critical sections implemented by mutex locks. This paper presents a race-detection algorithm, called the Protect algorithm, for parallel programs where critical sections are defined in a higher-level programming construct - "critical region". Both the time and space complexity of the Protect algorithm is improved over ones of the previous race-detection algorithms. If the control-flow of a parallel execution on a given input is deterministic, then a single-run of any race-detection algorithm guarantees to find out all races. In the presence of nondeterministic control-flow, a single-run of any race-detection algorithm cannot find out all races. We present another algorithm, called the Alter algorithm, which either guarantees to detect any nondeterministic control-flow in a program using critical regions, or certifies that the program is control-flow deterministic, thereafter all races can be detected by the Protect algorithm.