Online / 5 & 6 February 2022


LibVF.IO: vGPU & SR-IOV on Consumer GPUs using Nim

LibVF.IO is bringing vGPU functionality that's historically been restricted to the enterprise datacenter segment to normal consumer GPUs running free open source desktop operating systems.

I'd like to showcase LibVF.IO's new LIME Runtime feature (Lime Is Mediated Emulation) and do a deep dive on open source vGPU technology in general.

Some of the topics I would like to cover are the VFIO-Mdev API, the differences between Virgl vs. Hardware GPU virtualization (SR-IOV) vs. Software GPU virtualization (soft Mdev), various common GPU driver virtualization implementations (Intel i915, AMD GPU-IOV Module, and Nvidia), then talk about the benefits of hardware virtualization of GPUs for projects like Qubes OS and the ongoing effort to run LibVF.IO with Xen for use with Qubes. I'd like to share LibVF.IO's vision to bring vGPU SR-IOV functionality out of the datacenter and to most consumer devices running free software. Finally I'd like to explore what a world might look like if we're successful in bringing the benefits vGPU SR-IOV functionality to every day people running free software (native performance backwards compatibility via virtualization, hardware isolation, & workload balancing).


Photo of Arthur Rasmusson Arthur Rasmusson