Random Generation of Test Instances for Logic Optimizers

The attempt of using random test circuits for evaluating the performance of logic optimizers like SIS is apparently new. To generate "reasonable" random circuits, we propose the random applications of several transformation rules to an initial circuit instead of the obvious method, random placement of connections. A preliminary experiment has been conducted on SIS's responses against such random circuits. SIS shows considerably different performances for different circuits generated from the same original circuit.