NVIDIA CUDA Toolkit SDK for Workshop

NVIDIA CUDA is a parallel computing platform and programming model for NVIDIA GPUs. This SDK installs the CUDA Toolkit inside the workshop during setup-base using the official NVIDIA CUDA APT repository, placing everything under /usr/local/cuda.

---

Reference workshop

A minimal workshop:

# workshop.yaml
name: cuda-dev
base: [email protected]
sdks:
  - name: cuda-toolkit
    channel: 12.9/stable

actions:
  check-gpu: |
    nvidia-smi
  compile-sample: |
    cd /project
    nvcc -o sample sample.cu
    ./sample

This demonstrates GPU access and CUDA compilation inside a workshop.

Available tracks

Track	CUDA series	Platforms
12.9/stable	CUDA 12.9.x	[email protected], [email protected]
12.8/stable	CUDA 12.8.x	[email protected], [email protected]
12.6/stable	CUDA 12.6.x	[email protected], [email protected]
11.8/stable	CUDA 11.8.x (frozen)	[email protected] only

---

Using the SDK

Prerequisites, project layout

The SDK installs the NVIDIA CUDA Toolkit from the official NVIDIA APT repository and automatically configures PATH and library paths. An NVIDIA GPU with a compatible driver must be available on the host.

No specific project layout is required. Place your .cu files in the project directory.

Run CUDA samples

To build a project or run CUDA examples in the workshop:

git clone https://github.com/NVIDIA/cuda-samples.git
workshop launch
workshop shell

Inside the workshop:

cd /project/cuda-samples/Samples/1_Utilities/deviceQuery
make
./deviceQuery

Verify GPU access

Once the workshop is ready:

workshop shell
nvidia-smi

This shows detected NVIDIA GPUs, driver version, and CUDA version. To verify the compiler:

nvcc --version

---

Plugs (resources this SDK consumes)

gpu

• Interface: gpu
• Purpose: Grants access to NVIDIA GPU hardware on the host.

Slots (resources this SDK provides)

This SDK doesn't define any slots.

---

Documentation and guidance

• CUDA Toolkit documentation
• CUDA samples
• Workshop documentation

---

Community and support

• NVIDIA developer community: NVIDIA Developer Forums
• Workshop forum: Discourse
• Please review our Code of Conduct before participating.

---

Contributions

All contributions, including code, documentation updates, and issue reports, are welcome!

• See CONTRIBUTING.md for guidelines.
• Open issues or pull requests on the official repository.

---

License and copyright

Copyright 2026 Canonical Ltd.

This program is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License version 2.1 (LGPLv2.1) as published by the Free Software Foundation.

NVIDIA CUDA Toolkit is subject to the NVIDIA CUDA Toolkit EULA.