xBGAS: An Address Space Extension for Scalable High-Performance Computing

Authors: Xi Wang (Texas Tech University); John D. Leidel (Tactical Computing Laboratories); Brody Williams (Texas Tech University); Alan Ehret, Miguel Mark, and Michel Kinsy (Texas A&M University); and Yong Chen (Texas Tech University)

Abstract: The tremendous expansion of data volume has driven the transition from monolithic architectures towards systems integrated with discrete and distributed subcomponents in modern scalable high-performance computing (HPC). As such, multi-layered software infrastructures have become essential to bridge the gap between heterogeneous commodity devices. Operations across synthesized components with divergent interfaces, however, inevitably lead to redundant software footprints and undesired latency. A scalable and unified computing platform, capable of supporting efficient interactions among individual components, is therefore desired for large-scale data-intensive applications.

In this work, we introduce xBGAS, a micro-architecture extension to the RISC-V instruction set architecture (ISA) that offers a scalable, global address space for high-performance, shared memory computing. The xBGAS extension provides native ISA-level support for direct accesses to remote shared data blocks by mapping remote objects into a system’s extended address space. We perform both software and hardware evaluations of the xBGAS design.

Best Poster Finalist (BP): no

Poster: PDF
Poster summary: PDF


Back to Poster Archive Listing