Software Implementation of Modular Reduction by Pseudo-mersenne Primes

Modern cryptosystems allow the use of operation in prime fields with special kind of modules that can speed up the prime field operation: multiplication, squaring, exponentiation. The authors took into account in the optimizations: the CPU architecture and the multiplicity of the degree of the modulus in relation to the machine word width. As example, shown adopted module reduction algorithms hard-coded for modern CPU in special form of pseudo-Mersenne prime used in MAC algorithm Poly1305, - in electronic signature algorithm EdDSA and - in short message encryption algorithm DSTU 9041. These algorithms have been software implemented on both 32-bit and 64-bit platforms and compared with Barrett modular reduction algorithm for different pseudo-Mersenne and generalized-Mersenne modules. Timings for proposed and Barrett algorithms for different modules are presented and discussed.