NOVA: Coordinated Test Selection and Bayes-Optimized Constrained Randomization for Accelerated Coverage Closure

TL;DR

<3-5 sentence high-level summary> NOVA addresses the high cost of RTL verification by jointly optimizing which test categories to run and how to generate stimuli for coverage. It introduces coverage-aware test selection and a Bayes-optimized, parameterized CRV solver, combined with a clustering-based objective and dimensionality-reduction techniques. Through warm-up-guided clustering, ensemble-based novelty scoring, and SAASBO-driven solver tuning, NOVA achieves substantial speedups in coverage convergence across multiple designs, outperforming both random and manually engineered baselines as well as state-of-the-art learning-based methods. The framework is designed to integrate with existing CRV workflows and demonstrates practical gains in industrial-like verification settings.

Abstract

Functional verification relies on large simulation-based regressions. Traditional test selection relies on static test features and overlooks actual coverage behavior, wasting substantial simulation time, while constrained random stimuli generation depends on manually crafted distributions that are difficult to design and often ineffective. We present NOVA, a framework that coordinates coverage-aware test selection with Bayes-optimized constrained randomization. NOVA extracts fine-grained coverage features to filter redundant tests and modifies the constraint solver to expose parameterized decision strategies whose settings are tuned via Bayesian optimization to maximize coverage growth. Across multiple RTL designs, NOVA achieves up to a 2.82$\times$ coverage convergence speedup without requiring human-crafted heuristics.

Figures (6)

• Figure 1: Overview of NOVA
• Figure 2: Test Selection Strategy
• Figure 3: Parameterized Solver
• Figure 4: Overall speedups of NOVA over isolation forest and random selection baselines across all benchmarks.
• Figure 5: Bayes-optimized constrained randomization consistently outperforms human-engineered randomization across both UART and AXI4 benchmarks.