218 VLSI Test Principles and Architectures 4.8 ADVANCED SIMULATION-BASED ATPG 4.8.1 Seeding the GA with Helpful Sequences Genetic algorithms have been shown to be effective for test generation in the above discussion. However, for some difficult faults, the previous GA-based methods may still underperform the deterministic ATPGs. For such faults, it may be helpful to embed certain individuals in the initial population to guide the GA. This is called seeding . For example, suppose a fault has been excited and propagated to one or more flip-flops in a sequential circuit, and now the GA attempts to drive the fault-effect from those flip-flops to a primary output. If there are previously known sequences that were successful in propagating fault-effects from a similar set of flip-flops, then seeding these sequences into the initial population may tremendously help the GA. The DIGATE [Hsiao 1996a] [Hsiao 1998] and the STRATEGATE test generators [Hsiao 1997] [Hsiao 2000] aggressively apply seeding of useful sequences for the GA. When there are no such sequences available, both DIGATE and STRATEGATE try
This is the end of the preview. Sign up
access the rest of the document.