Naive Kernel GFLOPs vs. Matrix Size