A PGAS Communication Library for Heterogeneous Clusters

TL;DR

This paper introduces Shoal, a PGAS communication library for heterogeneous clusters of processors and FPGAs, enabling flexible application migration and demonstrating promising performance and ease of use.

Contribution

The work presents Shoal, the first PGAS library supporting heterogeneous hardware, facilitating application portability and dynamic topology deployment.

Findings

Shoal achieves low latency and high throughput in microbenchmarks.

The Jacobi method implementation shows improved run times across platforms.

The library demonstrates feasibility of PGAS models for heterogeneous clusters.

Abstract

This work presents a heterogeneous communication library for clusters of processors and FPGAs. This library, Shoal, supports the Partitioned Global Address Space (PGAS) memory model for applications. PGAS is a shared memory model for clusters that creates a distinction between local and remote memory access. Through Shoal and its common application programming interface for hardware and software, applications can be more freely migrated to the optimal platform and deployed onto dynamic cluster topologies. The library is tested using a thorough suite of microbenchmarks to establish latency and throughput performance. We also show an implementation of the Jacobi iterative method that demonstrates the ease with which applications can be moved between platforms to yield faster run times. Through this work, we have demonstrated the feasibility of using a PGAS programming model for multi-node heterogeneous platforms.