CUDA

• Stanford CUDA lectures
• ✨ Just enough CUDA to be dangerous
  • The book I was recommended for learning CUDA when I first showed up at PyTorch: Programming Massively Parallel Processors
    • available from UniMi and on Drive
  • The environment variable that makes CUDA synchronous is CUDA_LAUNCH_BLOCKING=1. cuda-memcheck is also useful for debugging CUDA problems https://docs.nvidia.com/cuda/cuda-memcheck/index.html
• All about NVIDIA GPUs - PyTorch Developer Podcast
• PyTorch CUDA Semantics
  • CUDA semantics — PyTorch 2.6 documentation
• 1. Why CUDA Compatibility — CUDA Compatibility
• CUDA Compatibility

Snippets

CUDA version:

nvidia-smi | grep "CUDA Version" | awk '{print $9}'

CUDA driver version:

nvidia-smi | grep "Driver Version" | awk '{print $3}'