I am not confident I fully understand the question but in case this clarification is helpful: In terms of comparing implementing tones (complex or real) with the CORDIC or an NCO (accumulator and look-up table) my first consideration is if I have more memory or more time. If memory and resources are precious but we have the time for multiple iterations, the CORDIC is a prime consideration given it is an iterative approach with minimum resources to rotate a complex phasor by any angle (so if we keep rotating at a constant rate, we can produce Sines and Cosines as needed). If we have memory and want to operate at the fastest possible rate (we don't have time), then the one-shot approach of looking up a Sine and Cosine value from a look-up table while we increment through phase values with an accumulator is an attractive appraoch. Of course we have many ways to reduce that memory requirement, moving the boundary of decision further, such as quarter-cycle storage, interpolation, and memory compression algorithms specific to NCO implementations including the Sunderland algorithm and Hutchinson algorithm. Still if we want the lowest possible power in the smallest possible real-estate, and we are operating at relatively low clock rates, the CORDIC is very compelling and relevant. Thank you Jack Volder.