I take issue with this benchmark since comparing it to Haskell is not illuminating since the Cython cdef function is straight C and being called as a foreign function. If you do the same in Haskell via the Haskell C FFI it ends up being roughly the same as the C code ( not surprisingly ).

Having written quite a bit of Cython, I refuse to touch it anymore given how wildly undefined it's behavior can be with intermingling Python and C semantics.