Test Generation: A Symbolic Approach

Automatic test pattern generation (ATPG), based on random methods, is widely applied in different flavors to detect functional faults. In particular, genetic algo- rithms, which can be considered part of the random-based category, work quite well for a quick exploration of the test patterns space, achieving good fault cover- age in a short time. However, a certain number of faults are hard to detect for ran- dom-based approaches. On the contrary, such faults can be easy to detect for other strategies, as for example symbolic test pattern generation. Nevertheless, faults that are easy to detect for random-based ATPG may be, indeed, hard to detect for symbolic techniques. Addressing this class of faults, a symbolic ATPG can require longer execution time than a random one, and in some cases it can be unable to generate test sequences for some faults. This work shows how a hybrid approach, where symbolic ATPG is applied after a random-based ATPG, may represent a valuable solution to achieve a very high fault coverage keeping low the execution time. The testing methodology is implemented in a highly flexible functional veri- fication framework that is based on a high-level fault model and a test generation strategy applicable for both random-based and symbolic-based approaches. The applicability and the efficiency of the functional testing framework presented have been confirmed by the benchmarks analyzed.