Floating-point CORDIC Coprocessor with Dynamic Iterations

In this paper a detailed description is provided for computing of sine, cosine and atan for both single and double precision floating point using dynamic iterations of CORDIC algorithm. The iterations needed for computing a function are decided in real time based on the input value. A detailed emphasis is given on the validity and advantages of this method. Also, the computation of the above functions is done by the linear terms of the Taylor series expansion in their respective valid domains both for the ease of implementation and to confine the number of maximum iterations. The size of the hardware (shifters, adders, atan table) required for the above implementation is derived and justified. The results of these theoretical derivations are coded in VHDL and simulation was carried out. The floating-point results match with Intel processor outputs up to 1 LS bit mismatch. The hardware module of CORDIC was interfaced to custom processor which is having micro code-based control unit. New set of instructions were incorporated in the microcode to support the trigonometric operations. With this compute module hardware implementation, a 17-times reduction in the clock cycle required for the computation of the above functions was achieved.