Specifying and Compiling Applications for RaPiD

E cient, deeply pipelined implementations exist for a wide variety of important computation-intensive applications, and many special-purpose hardware machines have been built that take advantage of these pipelined computation structures. While these implementations achieve high performance, this comes at the expense of exibility. On the other hand, exible architectures proposed thus far have not been very e cient. RaPiD is a recon gurable pipelined datapath architecture designed to provide a combination of performance and exibility for a variety of applications. It uses a combination of static and dynamic control to e ciently implement pipelined computations. This control, however, is very complicated; specifying a computation's control circuitry directly would be prohibitively di cult. This paper describes how speci cations of a pipelined computation in a suitably high-level languageare compiled into the control required to implement that computation in the RaPiD architecture. T...

• Type: Conference
• Year: 1998
• Where: FCCM
• Authors: Darren C. Cronquist, Paul Franklin, Stefan G. Berg, Carl Ebeling
• Keywords: Dynamic Control, FCCM 1998, Gurable Pipelined Datapath, Pipelined Implementations, VLSI