Performing data flow testing in parallel

Testing is a critical and expensive part of software development. One testing methodology, data flow testing, uses the flow of data in a program to determine whether the program is adequately tested. We present a new approach that partitions the data flow testing workload into an appropriately sized granularity. The workload can be scheduled either statically or dynamically, and can be adapted to either a shared memory or a distributed memory environment. We implemented both uniprocessor and multiprocessor versions of our data flow tester in C for the Data General Corporation´s AViiON 5000 machine, experimented with a variety of programs, and obtained good speedup using the multiprocessor version over the uniprocessor version. Currently, we are expanding our tester to handle larger programs, and experimenting an machines with different architectures and a greater number of processors