Skip to content

Latest commit

 

History

History
49 lines (34 loc) · 1.36 KB

gpu.md

File metadata and controls

49 lines (34 loc) · 1.36 KB
Raw

GPU

Nanos is the only unikernel with native GPU support. This is sometimes necessary for blockchain or AI based workloads. The GPU support targets most common cloud and consumer Nvidia based GPUs.

On the cloud you can specify the number and GPU type via:

{
    "RunConfig": {
        "GPUs": 1,
        "GPUType": "nvidia-tesla-t4"
    }
}

You must also include the 'gpu_nvidia' klib and include the Nvidia firwmware:

  "Klibs": ["gpu_nvidia"],
  "Dirs": ["nvidia"]

Nightly binary builds are built and uploaded to:

https://storage.googleapis.com/nanos/release/nightly/gpu-nvidia-x86_64.tar.gz

Note: For companies over 50 employees using for commercial purposes a paid subscription is required at https://nanovms.com/services/subscription for using binary builds NanoVMs builds. If you don't wish to sign up for that you can build the source yourself.

For more information checkout the tutorial at https://nanovms.com/dev/tutorials/gpu-accelerated-computing-nanos-unikernels and nanovms/ops#1528 .

OnPrem Instructions

  • VT-d (Intel) or IOMMU (AMD) must be enabled in bios

  • Add the proper kernel command line argument:

Intel: intel_iommu=on iommu=pt

AMD: amd_iommu=on iommu=pt