Automatic error diagnosis and correction for RTL designs

Abstract

Recent improvements in design verification strive to automate the error-detection process and greatly enhance engineers' ability to detect functional errors. However, the process of diagnosing the cause of these errors and fixing them remains difficult and requires significant ad-hoc manual effort. Our work proposes improvements to this aspect of verification by presenting novel constructs and algorithms to automate the error-repair process at the Register-Transfer Level (RTL), where most development occurs. Our contributions include a new RTL error model and scalable error-repair algorithms. Empirical results show that our solution can diagnose and correct errors in just a handful of minutes even for complex designs o/up to several thousand lines of RTL code in minutes. This demonstrates the superior scalability and efficiency of our approach compared to previous work.