Although probably very slow in pure Python the benefit of not using numpy is that the code can potentially be JIT compiled via numba or pypy to speed things up. Just curious if this has been explored at all and how it benchmarks against numpy.
Agains numpy LA is about ~100 slower. Other algorithms like solvers are comparable with SciPy. I never tried numba. If you try nlib with Numba let me know.
Although probably very slow in pure Python the benefit of not using numpy is that the code can potentially be JIT compiled via
numba
orpypy
to speed things up. Just curious if this has been explored at all and how it benchmarks againstnumpy
.