Configurable Computing Machines (CCMs) have been able to provide orders of
magnitude increases in execution rates for applications such as image proce
ssing, signal processing, and automatic target recognition. This paper desc
ribes the use of CCMs to accelerate complex, large-scale scientific computa
tions. These applications present a challenge for CCMs because of their lar
ge size, hundreds of thousands of lines of code, and the unstructured natur
e of the computations. This paper describes strategies for accelerating sci
entific computations on CCMs and demonstrates the effectiveness of one such
strategy on the Annapolis Micro Systems WildForce board. Results from this
implementation are analyzed. (C) 2000 Elsevier Science Ltd. All rights res
erved.