Cuda

CS336 Notes: Lecture 6 - Kernels and Triton

Writing efficient GPU kernels with Triton: profiling, benchmarking, kernel fusion, and when to hand-optimize versus using torch.compile.