Comparison of parallelized radix-2 and radix-4 scalable Montgomery multipliers

Radix‐16 CSA‐based low‐latency non‐Montgomery modular multiplier

“…In each kernel cycle of Montgomery modular multiplication with [19], it shifts operands $$M$$ and $$A$$ 1 bit left $$n/w$$ times with temporary results $$Z:= Z+ q_i\cdot M+ b_i\cdot A$$, and then right shift the result $$n/w$$ bits 1 time at the end of the kernel cycle with $$Z:= Z/2^{n/w}$$ [20], that is, performing the division of $$2^{n/w}$$ in $$(Z+ q_i\cdot M+ b_i\cdot A)/2^{n/w}$$ from serial operation to one‐step operation. In this way, the latency between processing elements is decreased from 2 clock cycles to 1 clock cycle.…”