This paper presents a framework for high-level design validation using an efficient register-transfer level (RTL) automatic test pattern generator (ATPG). The RTL ATPG algorithm first generates a test environment for each validation objective, which includes variable assignments, conditional statements, and arithmetic expressions in the hardware description language (HDL) description. The test environment for a given validation objective is a set of symbolic conditions that allow for full controllability and observability of that objective. After the RTL ATPG terminates, a back-end translator intelligently translates the test environments into validation vectors by filling in the necessary values. Since the observability of error effect is naturally handled by the RTL ATPG algorithm, this approach is superior to most existing validation methods, which only focus on the excitation of HDL constructs. A set of heuristics is proposed to utilize high-level circuit information to enhance the RTL ATPG algorithm and to maximize the validation efficiency. The RTL ATPG algorithm is also coupled with an improved RTL validation-coverage metric, which can help users to gain a higher degree of confidence on the quality of generated validation vectors. The usage of the coverage metric also results in the generation of compact vector sets. Experimental results on academic and industrial benchmark circuits demonstrate that our method is able to obtain very high design-error coverage in short execution times.

We propose a new dynamic method of abstraction, which can be applied during successive steps of the model checking algorithm to further reduce the model produced by traditional static abstraction methods. This is facilitated by information gathered from an analysis of the proof of unsatisfiability of SAT-based bounded model checking problems formulated on the original design. The dynamic abstraction effectively allows the model checker to work with smaller abstract models. Experiments on several industrial benchmarks demonstrate that dynamic abstraction can significantly improve both the performance and the capacity of typical abstraction refinement flows.

Bounded model checking (BMC) based on Boolean satisfiability (SAT) methods has recently gained popularity as a viable alternative to BDD-based techniques for verifying large designs. This work proposes a number of conceptually simple, but extremely effective, optimizations for enhancing the performance of SAT-based BMC flows. The key ideas include: (1) a novel idea to combine SAT-based inductive reasoning and BMC; (2) clever orchestration of variable ordering and learned information in an incremental framework for BMC; and (3) BMC-specific ordering strategies for the SAT solver. Our experiments, conducted on a wide range of industrial designs, show that the proposed optimizations consistently provide between 1-2 orders of magnitude speedup and can be extremely useful in enhancing the efficacy of typical SAT-BMC tools.

We present a framework for high-level design validation using an efficient register-transfer level (RTL) automatic test pattern generator (ATPG). The RTL ATPG generates the test environments for validation targets, which include variable assignments, conditional statements, and arithmetic expressions in the HDL description. A test environment is a set of conditions that allow for full controllability and observability of the validation target. Each test environment is then translated to validation vectors by filling in the unspecified values in the environment. Since the observability of error effect is naturally handled by our ATPG, our approach is superior to methods that only focus on the excitation of HDL descriptions. The experimental results on ITC99 benchmark circuits and an industrial circuit demonstrate that very high design error coverage can be obtained in a small CPU times.

We present an efficient register-transfer level automatic test pattern generation (ATPG) algorithm. First, our ATPG generates a series of sequential justification and propagation paths for each RTL primitive via a deterministic branch-and-bound search process, called a test environment. Then the precomputed test vectors for the RTL primitives are plugged into the generated test environments to form gate-level test vectors. We augment a 9-valued algebra to efficiently represent the justification and propagation objectives at the RT Level. Our ATPG automatically extracts any finite state machine (FSM) from the circuit, constructs the state transition graph (STG), and uses high-level information to guide the search process. We propose new static methods to identify embedded counter structures, and we use implication-based techniques and static learning to find the FSM traversal sequences sufficient to control the counters. Finally, a simulation-based RTL extension is added to augment the deterministic test set in a few cases when there is additional room for the improvement in fault coverage. Experimental results show that our new deterministic RTL techniques achieve several orders of magnitude reduction of test generation time without compromising fault coverage when compared to gatelevel ATPG tools. Our ATPG also outperforms a recently reported simulation-based high-level ATPG tool in terms of both fault coverage and CPU time. 1.