Super HN

   CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication Through RL (github.com) 11 points by dzign 35 minutes ago