Architectural considerations for application-specific counterflow pipelines

Application-specific processor design is a promising approach for meeting the performance and cost goals of a system. Application-specific processors are especially promising for embedded systems (e.g., digital cameras, cellular phones, etc.) where a small increase in performance and decrease in cost can have a large impact on a product's viability. Sproull, Sutherland and Molnar (see IEEE Design and Test of Computers, vol. 11, no. 3, p. 48-59, 1994) have proposed a new pipeline organization called the Counterflow Pipeline (CFP). This paper evaluates CFP design alternatives and shows that the CFP is an ideal architecture for fast, low-cost design of high-performance processors customized for computation-intensive embedded applications. First, we describe why CFP's are particularly well-suited to realizing application-specific processors. Second we describe how a CFP tailored to an application can be constructed automatically. Third, we present measurements that evaluate CFP design trade-offs and show that CFP's provide speculative and out-of-order execution, and register renaming that is matched to an application. Fourth, we show that asynchronous counterflow pipelines achieve high-performance by reducing the average execution latency of instructions over synchronous implementations. Finally, we demonstrate that custom CFP's achieve cycles per instruction measurements that are competitive with 4-way superscalar out-of-order processors at a potentially low design complexity.