Exploring PGAS Communication for Heterogeneous Clusters with FPGAs

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.

A multithreaded CUDA and OpenMP based power‐aware programming framework for multi‐node GPU systems

“…Several compliant languages and libraries have been discussed in Reference 16 including: original PGAS languages—CAF, Titanium, UPC; HPCS PGAS languages—Chapel, X10, Fortress; Retrospective PGAS languages—HPF, ZPL and GA as well as XCalableMP (XMP)—PGAS extension for C and Fortran. Notable recent examples include using PCJ for HPC systems, 17 big data processing, 18 clouds 19 as well as Shoal for clusters of processors and FPGAs 20 . HPX 21 is a C++ library developed for concurrency and parallelism that supports parallel, concurrent and distributed functions for general purpose programming, in particular active global address space (AGAS) that allows moving objects between nodes without changing addresses.…”