Sug Talk: AMD’s Derek Bouius On Enabling Generalized Gpu Support In Singularity

TL;DR: Accelerators showcase Singularity’s stance on integration over isolation. Here, AMD makes the case for a more-generalized support model for GPUs in the software.

SUG Series Introduction

The inaugural meeting of the Singularity User Group (SUG) was held March 12-13, 2019, at the San Diego Supercomputer Center (SDSC). The event attracted diverse representation from the international advanced computing community as conveyed through the post-event press release issued jointly by SDSC and Sylabs.

Over the course of the two-day event, over 20 talks were presented by members of the Singularity user, developer, and provider community. Because SUG generated a significant amount of interest, even from those who were unable to attend, we are sharing online each of the talks presented.

SUG Talk Introduction

Singularity favors integration over isolation; that critical containerization differentiator ensures it remains simple to make use of GPUs. With the availability of GPUs from AMD, Product Manager Derek Bouius makes the case for revisiting GPU support in Singularity. In fact, to address the more generalized set of requirements, Derek and his colleagues at AMD are poised to make contributions to the Singularity code base. The abstract for Derek’s contributed talk entitled Enabling Generalized GPU Support in Singularity is as follows:

Walk through design and code changes needed to enable general graphics card support for use in Singularity containers. Discuss some of the challenges of supporting dedicated acceleration devices in a server environment and highlight solutions using Singularity containers.

Derek’s talk from SUG can be found below and here. Enjoy!