Speed up matrix multiply
mdekstrand opened this issue · comments
Right now our Numba matrix multiply code is decent, but its asymptotic performance is worse than either SciPy or MKL. It's fine to get us started, but it would be nice to implement a more performant algorithm.