This dev container provides the toolchain expected by cuda-oxide:

• Ubuntu 24.04
• CUDA Toolkit 13.0
• LLVM 21 with NVPTX support
• Clang 21 resource headers for bindgen
• Rust nightly-2026-08-28 with rust-src, rustc-dev, rust-analyzer, rustfmt and clippy. llvm-tools is not preinstalled in the image; rustup adds it on your first cargo command in the checkout, because rust-toolchain.toml names it. The image ships LLVM 21 and CUDA_OXIDE_LLC already points at llc-21, so nothing waits on it.

Open the repository in a devcontainer-aware editor and choose "Reopen in Container". The container requests GPU access with --gpus=all and uses updateRemoteUserUID so generated Cargo artifacts and exported PTX files stay writable from the host checkout.

Inside the container:

cargo oxide doctor
cargo oxide run vecadd

The CUDA Toolkit is provided by the container image; it does not need to be installed on the host. The host must provide an NVIDIA GPU, a driver compatible with CUDA 13.0, and the NVIDIA Container Toolkit so Docker can expose the GPU to the container.

If the host driver is too old, GPU commands such as nvidia-smi, cargo oxide doctor, or cargo oxide run vecadd will fail inside the container. Update the host NVIDIA driver rather than installing a different CUDA Toolkit in the container.

The image does not set LD_LIBRARY_PATH; the NVIDIA Container Toolkit should provide the host driver libraries at runtime.