Skip to main content
FOSDEM PWA

Review of kernel and user-space Neural Processing Unit (NPU) chips support on Linux

UD2.120 (Chavanne) | Day 1 | 18:15 - 18:35 | Speakers: Jakov Petrina Trnski

Review of kernel and user-space Neural Processing Unit (NPU) chips support on Linux
A picture of a devroom at FOSDEM 2024
Open in browser

Notes

Abstract

In the last 10 years there's been a hardware race to build the best application-specific integrated circuit (ASIC) for both machine learning training and inference i.e. AI accelerators. What started with vision processing units (VPUs) went through tensor PUs (TPUs) and now we are dealing with neural processing units (NPUs). What's next?

This talk will take a systematic look at the different hardware platforms for AI acceleration, but with a focus on the software stacks that support them on Linux. We'll take a look how individual vendors approached their ASICs from the kernel side, and how they exposed the acceleration functionality for user-space.

Is it all proprietary or is there liberté? We'll find out together!

Attachments

Speakers

Jakov Petrina Trnski

With almost a decade of experience, Jakov is a firmware engineer with a specialization in embedded Linux systems.

As a lead engineer and architect, he has successfully deployed Linux solutions across a diverse range of networking-focused hardware platforms as well as developed new approaches to resolving long-standing challenges. His expertise spans the entire stack, from bootloaders and small kernel work to designing custom firmware build systems competing with Yocto, Buildroot, and OpenWrt.

As an active contributor to the open-source community, Jakov has worked on projects such as Gentoo, NETCONF / Sysrepo, and the kernel. He previously served as a technical steering contributor to the DENT Project and has shared his insights regarding open-source technologies at conferences including FOSDEM, All Systems Go!, and DORS/CLUC.

Notice: The placeholder video image is licensed under CC BY-SA 4.0. The original image can be found hereChanges made to the image are: Cropped the image to a new ratio, part of the image was cut off.