ABD - The alternative: Simulation

The alternative is extensive simulation with overnight testing for regressions.

Can either write a RTL or ESL yes/no automaton as part of the test bench.

Spool outputs to file and diff against golden with PERL script.

Downfall of simulation: it's non-exhaustive and time consuming.

ABD benefits (and challenges):

• Completeness (how to define this?)
• Scalability (tools limited in practice?),
• Rare corner situations (unusual conjunctions of events) are covered.

But: Simulations

• are needed for performance analysis and general design confidence,
• can generate some production test vectors.
• can be partly formal: using bus monitors for dynamic validation and Specman/VERA constrained pattern generators for stimulus.