Exploring Dynamic Reconfigurable CORDIC Co-Processors Tightly Coupled with a VLIW-SIMD Soft-Processor Architecture.

Most of embedded signal processing applications require the computation of complex arithmetic operations. Tightly coupled with a host processor, the CORDIC algorithm is one possible solution to establish a processing base, which is capable of performing a wide spectrum of operations with high precision. A software implementation as well as fully programmable and dedicated hardware implementations of the CORDIC algorithm are compared with each other in this paper. Especially, the use of non-programmable, area-efficient modules with a pre-determined functionality is particularly suited for use in dynamical partial reconfiguration (DPR) systems. This paper discusses the different aspects and design issues of different implementation strategies and shows that the initial reconfiguration overhead introduced by DPR systems is negligible in contrast to the speedup gained by the reconfiguration of the hardware accelerator when computing the same function more than 51 times consecutively (depending on the arithmetic function to be computed).