Calling vectorize with parallel target inside njit

[ehsantn]

Reporting a bug

• I am using the latest released version of Numba (most recent is visible in
  the change log (https://github.com/numba/numba/blob/master/CHANGE_LOG).
• I have included below a minimal working reproducer (if you are unsure how
  to write one see http://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports).

Calling a vectorize function with the parallel target inside an njit function fails:

import numpy as np
import numba as nb

# @nb.vectorize([nb.float64(nb.float64, nb.float64)]) # works
@nb.vectorize([nb.float64(nb.float64, nb.float64)], target="parallel")
def g(a, b):
    return (a * b + 1)

@nb.njit
def f(A, B):
    return g(A, B)

print(f(np.ones(4), np.zeros(4)))

[stuartarchibald]

Thanks for the report. I think this is because the standard target for @vectorize would return a Numba DUFunc instance whereas the parallel target would return a NumPy ufunc instance. The former can be consumed by the compiler, the latter cannot.

import numba as nb

@nb.vectorize([nb.float64(nb.float64, nb.float64)])
def gs(a, b):
    return (a * b + 1)

print(type(gs))

@nb.vectorize([nb.float64(nb.float64, nb.float64)], target="parallel")
def gp(a, b):
    return (a * b + 1)

print(type(gp))

gives:

<class 'numba.npyufunc.dufunc.DUFunc'>
<class 'numpy.ufunc'>

related: #2469