Siu Kwan Lam

We need something like this to guard other loops from reusing the same list variable: ```diff diff --git a/numba/core/inline_closurecall.py b/numba/core/inline_closurecall.py index 6eb11a16d..124fe1e72 100644 --- a/numba/core/inline_closurecall.py +++ b/numba/core/inline_closurecall.py @@ -1053,6 +1053,19...

Categorizing this as a feature request that is awaiting changes in NumPy. Numpy has plans to improve ufunc dispatch (https://github.com/numpy/numpy/pull/16723)

@JoshuaB47, Thank you for debugging this. Your assessment is correct. I attempted a more general fix so that the constant matches the dtype: ```diff diff --git a/numba/np/arrayobj.py b/numba/np/arrayobj.py index 12f985c5c..3eef01c56...

I can confirm that uncommenting the last line changes the printed output before it

I believe the problem is coming from `InlineClosureCallPass` debug output contains this: ``` InlineClosureCallPass: start inline arraycall fix_array_assign: found SetItem: $24load_attr.9[$const26.10] = $const20.7 find_array_def: $24load_attr.9getattr(value=vinst, attr=var) ``` No reason for...

I can reproduce the bug

> Trying to re-iterate and re-read your proposal. > > 1. Expose different locators via env variable in `numba.core.config` > 2. Create a new instance of locator that does exactly...

I brought this PR up in the Numba developer meeting today. The feedbacks from other maintainers are: - Need to support loading CacheLocator classes that are not baked into Numba....

@futurewasfree, thank you for the contribution. I've approved the PR. The remaining CI failures will be revolved when https://github.com/numba/numba/issues/10161 is fixed. For future reference, avoid force-push once the review has...

I can confirm this problem. It is a long standing issue that ufunc are not fully usable in njit code. Only `target='cpu'` (the default) can be used in njit.