Distributed Processing Environment

Simulation of local process disturbances is a computation- ally intensive task. The VLASIC (VLSI LAyout Simulation for Inte- grated Circuits) catastrophic fault yield simulator uses a Monte Carlo method that often requires tens of CPU hours to perform a simulation. In order to reduce the simulation time, we have developed DVLASIC, a distributed version of VLASIC, which is implemented on a network of 25 VAXstation 3200 workstations. DVLASIC achieves a speedup of 13.3 over VLASIC, with an efficiency of 89%. We describe the distrib- uted processing environment and implementation techniques used to obtain this speedup. Large speedups are difficult to obtain on the VLASIC problem due to the expense of Monte Carlo problem gener- ation, and the relatively small problem granularity. The distributed processing environment can be applied to many other CAD problems.