In this work we develop a generalization of the CORDIC algorithm for a
ny radix in three coordinate systems, linear, circular and hyperbolic.
We carry out a comparative study between different radixes at the num
ber of additions level, due to the fact that the complexity in additio
ns determines the total hardware associated with the implementation of
the algorithm. Radix 4 minimizes the number of additions, therefore w
e propose a high speed CORDIC processor based on this radix 4. We have
developed new criteria for the vectorization and rotation modes and w
e introduce a new technique for the compensation of a non-constant sca
le factor, multifactor compensation. The processor we propose is of a
general and pipelined character and uses redundant arithmetic (signed
digit) in order to reduce the delay in each stage, parallelizing the o
peration of the adders using a competititive implementation (reduction
of the number of stages) with respect to radix 2 implementations whic
h have recently appeared in the literature.