Concurrency-oriented optimization for low-power asynchronous systems

We introduce new architectural optimizations for low-power asynchronous systems, such as Tangram-based systems of van Berkel et al. (1994). Our goal is to reduce power consumption by improving system concurrency. We introduce two new sequencer designs, with greater concurrency than existing ones, that provide the opportunity for substantial power savings through voltage scaling. To safely accommodate this added concurrency, new latch designs are presented, for both dual-rail and single-rail implementations.