2 ms·Learning CUDA by optimizing matrix-vector multiplication for cuBLAS-like perf2 points by rrampage 2y ago