Word Level Predicate Generation for GoldMine

Abstract

GoldMine is an automatic assertion generation tool that uses data mining and static analysis to generate bit level assertions for a register transfer level (RTL) design. These bit level assertions can be verbose, and consequently less usable. Recent studies show that word level assertions have higher expressiveness and less redundancy than bit level assertions. This thesis presents an implementation of a word level predicate generation engine for GoldMine. The implementation instruments RTL source code to record which conditional statements will be executed during simulation. The implementation reads the simulation data to generate concrete word level predicates. GoldMine utilizes the generated predicates to mine word level assertions.