Mutation testing is a fault-based method for testing software that is computationally expensive. Mothra is an interpreter-based mutation testing system that is centered around an interpreter. This paper presents a parallel implementation of Mothra's interpreter on a MIMD machine. The parallel interpreter, HyperMothra, is implemented on a sixteen processor Intel iPSC/2 hypercube. Our goal was to demonstrate that the expense of software testing schemes such as mutation can be reduced by using parallel processing, and we demonstrate this by measuring the performance gains of the parallel interpreter over the Mothra interpreter. Results are presented using ten test programs, three different static work distribution schemes, and various numbers of processors. On our test programs, we found that our parallel interpreter achieved almost linear speedup over Mothra's sequential interpreter. With larger, faster high-performance computers available, mutation testing can be done at significantly less expense.
Back to my home page.