FastECPP over MPI

Abstract

The FastECPP algorithm is currently the fastest approach to prove theprimality of general numbers, and has the additional benefit of creatingcertificates that can be checked independently and with a lower complexity.This article shows how by parallelising over a linear number of cores,its quartic time complexity becomes a cubic wallclock time complexity;and it presents the algorithmic choices of the FastECPP implementation inthe author's \cm\ software (https://www.multiprecision.org/cm/) which has been written with massive parallelisation over MPI in mind, and which has been used to establish a new primality record for the "repunit" $(10^{86453} - 1) / 9$.