Accelerated montgomery multiplication using plural multipliers

Abstract

Montgomery multipliers and methods modular multiply a residue multiplicand by a residue multiplier to obtain a residue product, using a scalar multiplier, a first vector multiplier and a second vector multiplier. A controller is configured to control the scalar multiplier, the first vector multiplier and the second vector multiplier, to overlap scalar multiplies using a selected digit of the multiplier and vector multiplies using a modulus and the multiplicand. The scalar multiplier is configured to multiply a least significant digit of the multiplicand by a first selected digit of the multiplier, to produce a scalar multiplier output. The first vector multiplier is configured to multiply the scalar multiplier output by a modulus, to produce a first vector multiplier output. The second vector multiplier is configured to multiply a second selected digit of the multiplier by the multiplicand, to produce a second vector multiplier output. An accumulator is configured to add the first vector multiplier output, and the second vector multiplier output, to produce a product output. The latency of Montgomery multiplication thereby can be reduced to nearly the latency of a single scalar multiplication.