WSL2 Setup for Systems Development

A practical WSL2 setup for low-level systems and GPU work.

Essentials

• Use WSL2 (not WSL1) for real Linux syscalls and GPU passthrough.
• Keep your project files inside the Linux filesystem (~/), not on /mnt/c — cross-filesystem I/O is painfully slow.
• Configure .wslconfig to cap memory and enable localhostForwarding.

GPU / CUDA

• Install the Windows NVIDIA driver only; the CUDA toolkit goes inside WSL.
• Verify with nvidia-smi from the WSL shell.

Dev loop

• ccache + a warm build dir makes iterating on PyTorch source builds bearable.

Related: cuda-debugging, performance-reading-list.