ActorISx: Exploiting Asynchrony for Scalable High-Performance Integer Sort

Integer Sorting (ISx) represents a fundamental computational kernel in high-performance computing, serving as both a practical utility and a key benchmark for evaluating data movement and sorting performance at scale. In this paper, we present a novel implementation of ISx using HClib-Actor, a distributed and asynchronous actor-based programming system. This work demonstrates how actor-based programming models can effectively scale data-intensive applications like integer sorting, which have traditionally been implemented using message passing paradigms and suffer from substantial communication overhead and synchronization bottlenecks. We perform scalability studies on the NERSC Perlmutter supercomputer up to 16K cores, achieving comparable or superior performance to traditional MPI and OpenSHMEM implementations while providing a more intuitive programming model. Our results demonstrate weak scaling efficiencies, addressing key limitations of traditional implementations while showcasing the potential of actor-based approaches for highly scalable applications.