FPGAs for accelerating HPC engineering workloads: the why and the how

Research output: Other contribution

Abstract

Scientists and engineers are continually demanding the ability to generate more detailed results at reduced time to solution. Little wonder then that, in an attempt to meet such ambition, the future role that novel hardware architectures such as Field Programmable Gate Arrays (FPGAs) can play in High Performance Computing (HPC) is of great interest. In this whitepaper we argue that there are four major benefits that FPGAs can provide to HPC workloads and describe the importance of designing HPC codes from the perspective of dataflow for FPGAs. We describe three rules for achieving good performance for HPC codes on the FPGA, determining whether a general design approach will deliver acceptable performance before designing top down and then optimising bottom up.The result of this whitepaper is an understanding of where FPGAs can benefit HPC workloads and a design methodology for developing high performance dataflow codes more widely.
Original languageEnglish
TypeWhitepaper
Number of pages13
Publication statusPublished - 20 May 2022

Fingerprint

Dive into the research topics of 'FPGAs for accelerating HPC engineering workloads: the why and the how'. Together they form a unique fingerprint.

Cite this