PROR compaction scheme for larger circuits and longer vectors with deterministic ATPG

Reverse order restoration ROR techniques have found great use in sequential automatic test pattern generation ATPG, esp. spectral and perturbation-based ATPG. This paper deals with improving ROR for that purpose. We introduce parallel-fault multipass 2-level polynomial reverse order restoration PROR algorithms with constant complexity of the form H(n)G(n) + c where H(n) is the number of vectors to be released this iteration and G(n) is the attenuation factor. In PROR H(n) = n and G(n) here is 1−H(n)/no vectors with c about 1 to 64, where n is the number of iterations, the fault has been hanging around in the compactor and k is the polynomial complexity of the algorithm, in each iteration. k is variant and can take on any real value. We also divide the vector length such that it has a maximum of 10000 units at best, so that if the vector length is greater than 10000, it is still rounded to 10000 by considering a bigger chunk size. We also added PODEM of backtrace limit 30 on the last vector to get a faster and better quality test set. On contrast to algorithms which do not have results reported on large circuits and on longer vector lengths, we showcase our algorithm. For example, on b22, we achieved a 94.36% FC in 1.31 days with VL 146476.