A collection of high-performance CUDA kernels and experiments for learning and optimizing GPU compute primitives.