Global cost functions for test generation

The authors present a new cost function (GLOBAL) which is used to guide the decision process in test generation algorithms. By analyzing the global fanout structure of the circuit, GLOBAL guides the search for a test toward decisions that are least likely to result in conflict. The results obtained indicate that GLOBAL is a consistently superior guidance heuristic for test generation. Compared with previous cost functions, it provides greater fault coverage in approximately the same number of tests and usually in less CPU time. It appears to be more consistent and hence it can be reliably used on large circuits. GLOBAL identifies both lines whose justification is conflict free and lines whose justification is likely to lead to conflicts. Although its worst-case complexity is exponential in the number of stems, the analysis done by GLOBAL is performed only once, while its results are used at every decision step of an algorithm with a similar worst-case behavior. The results show that the time gained in test generation usually outweighs the time spent in preprocessing