Performance Analysis and Verification of Elastic Circuits under Process Variations

Abstract

The diversity of designs and optimization methods for elastic systems makes evaluating their performance challenging. This paper proposes a method based on xMAS and high-level modeling to analyze the performance, functional and timing verification of regular and early evaluation synchronous elastic circuits while considering process variations. The xMAS framework provides modularity, precise semantics, and executable models, enhancing formal verification and high-level analysis capabilities over existing approaches. The proposed platform calculates the throughput value, which is the most critical performance factor in elastic circuits. The power, delay, and PDP of all early evaluation elastic components are evaluated under process variations and compared to those of regular elastic circuits. The results indicate that early evaluation properties increase the sensitivity of circuit components to process variations, making their performance less predictable. Modeling results of the Elastic DLX microprocessor highlight these findings by demonstrating that process variations can cause a 26% reduction in throughput and lead to a 0.2% chance of synchronization errors between data and control signals. These findings underscore the critical need to account for process variations when designing and verifying early evaluation elastic circuits to maintain performance reliability.