Calling cuSparse library on Tesla A100 with CUDA11.1 is much slower than that on Tesla P100 with CUDA9.0

Do mind filing a bug? How to report a bug