Distributed fault simulation using vector set partitioning

To explore the potential speedup to be obtained through parallelism, a mathematical model for the performance of fault simulation is developed.A framework for performing fault simulation in a distributed environment is developed based on the assumption that fault simulation of different vectors can be run independently on different processors. This can be done using any uniprocessor fault simulator and requires only a fairly simple and low bandwidth communication mechanism. This communication mechanism distributes the list of faults detected by each simulator to all other simulators so that the simulation can be controlled based on global criteria such as the fault coverage obtained by the distributed simulator, and the amount of work to be done by each processor is dynamically reduced through fault dropping and fault-free dropping based on the coverage obtained by all processors.Finally, a distributed simulator is implemented using the above mechanism on a network of workstations using the uniprocessor fault simulator Tulip