option nogil=True not required for ufuncs and gufuncs

[pradkrish]

Closes #1317
nogil=True option is not required for ufuncs and gufuncs. Print a UserWarning if it is provided.

[gmarkall]

CUDA ufuncs don't release the GIL, for various reasons (some Python code needed around to setup arguments, kernel launch asynchrony, etc.). When trying to use nogil=True an error is produced. For example:

@vectorize(nogil=True, target='cuda')
def f(x, y):
    return x + y

Gives:

KeyError: "cuda vectorize target does not support option: 'nogil'"

This is all fine with the functionality in this PR: However, I think the message in the warning is a little misleading - perhaps "ufuncs and gufuncs always release the GIL with the CPU target" would be more accurate?

[pradkrish]

This is all fine with the functionality in this PR: However, I think the message in the warning is a little misleading - perhaps "ufuncs and gufuncs always release the GIL with the CPU target" would be more accurate?

Thanks for the review, I have modified the warning.

[stuartarchibald]

Thanks for the PR, just a couple of minor things to resolve else good to go! Thanks again!

[stuartarchibald]

Suggest using numba.errors.NumbaWarning as the warning category such that it's easier for users to filter out things from Numba.

[stuartarchibald]

Suggested change

	warnings.warn("nogil option is not required, ufuncs and gufuncs "
	"always releases GIL for CPU target")
	warnings.warn("The 'nogil' option is not required, ufuncs and gufuncs "
	"always release the GIL on the CPU target")

[stuartarchibald]

Would be good to check guvectorize as well given the error message mentions it?

[esc]

@pradkrish just enquiring about the state of this PR, where is it at?

[github-actions]

This pull request is marked as stale as it has had no activity in the past 3 months. Please respond to this comment if you're still interested in working on this. Many thanks!