ParallelKittens: Systematic and Practical Simplification of Multi-GPU AI Kernels