AN IMPLEMENTATIONTECHNIQUEOF MULTI-CYCLEDARITHMETICFUNCTIONS FOR A DYNAMICALLYRECONFIGURABLEPROCESSOR

Dynamically Reconfigurable Processor (DRP) released by NEC Electronics is expected to have potential for high de­ gree of parallel processing. Applications for DRP are de­ scribed in C language, and parallelism in the source code is automatically extracted by a compiler. On the other hand, it is also important to optimize descriptions so that the po­ tential performance of the device is effectively brought out. In this paper, arithmetic algorithms and an optimized coding technique to efficiently implement applications with multi­ cycled arithmetic functions on DRP are discussed, focus­ ing on the required number of the states. In this technique, the same kind of multi-cycled functions are aggregated into single functions, and arithmetic algorithms whose behav­ ior is steady on operand values are utilized. The effects of the technique are evaluated with fixed-point arithmetic func­ tions and polynomial arithmetic functions over a finite field, showing 2.68",,3.09 times performance improvement with­ out large increase in the number of states nor severe degra­ dation of the frequency.

[1]  Hideharu Amano,et al.  Implementation of active direction-pass filter on dynamically reconfigurable processor , 2005, 2005 IEEE/RSJ International Conference on Intelligent Robots and Systems.

[2]  Takayuki Sugawara,et al.  Dynamically Reconfigurable Processor Implemented with IPFlex's DAPDNA Technology , 2004, IEICE Trans. Inf. Syst..

[3]  Markus Weinhardt,et al.  PACT XPP—A Self-Reconfigurable Data Processing Architecture , 2003, The Journal of Supercomputing.

[4]  Vitit Kantabutra A new algorithm for division in hardware , 1996, Proceedings International Conference on Computer Design. VLSI in Computers and Processors.