Analysis and Implementation of the Asynchronous 2-Phase Single-Rail Protocol in the Balsa Synthesis System

Abstract

This thesis documents the design, verification and benchmarking of a new set of handshake components for the Balsa synthesis system. The handshake components employ the 2-phase single-rail asynchronous communication protocol. Balsa is the common name of a synthesis system and language that enable syntax-directed implementations of asynchronous circuits by connecting networks of handshake components. Handshake components enable an abstraction away from the gate level implementation, thus letting the designer focus entirely on functionality. To provide a background, an introduction to asynchronous systems and design styles is given, before the Balsa system and how it can help make asynchronous designs easier is explained. The general handshake component design process is gone through, in addition to particular challenges regarding specific components and evaluations of and possible improvements to certain components being highlighted. A collection of Balsa designs representing a diverse range of applications are used as benchmarking circuits for the 2-phase single-rail handshake components. Results of the benchmarks are presented, analyzed and compared to results from benchmarks of circuit implementations using the 4-phase single-rail protocol. Area estimations reveal an area overhead in implementations using the 2-phase single-rail protocol compared to 4-phase single-rail, which complies with expectations of complex control logic for the 2-phase single-rail protocol. Analysis of synthesis results also reveals that area is distributed on quite different handshake components for the two protocols because of differently optimized handshake components. Estimations of throughput and latency indicate a certain performance increase for circuits implemented using the 2-phase single-rail protocol, which can be seen as an expected, fortunate byproduct of the handshaking protocol's nature. Power consumption analyses performed on the benchmarking circuits indicate a substantial power and energy consumption decrease for the 2-phase single-rail protocol in the Atmel technology, but the numbers are incomplete and will need verification.Together the results indicate that circuits implemented using the 2-phase single-rail protocol will have potential benefits over the 4-phase single-rail protocol for certain applications and cost functions, and can become an important asset to the Balsa system.