stuartarchibald

@luk-f-a Recent work by @kc611 has been merged with #8031 which implements initial support, #8038 adding some standard distributions, #8040 adding more more advanced distributions, and there's a few more...

> Is there an issue that tracks the dispatch performance overhead problem (e.g. mentioned [here](https://github.com/numba/numba/pull/7900#discussion_r827973914))? I don't think so, please feel free to open one if it's important to your...

RE: https://github.com/numba/numba/issues/4499#issuecomment-1229235923 @kc611 I think this needs a `@lower_constant(types.NumPyRandomGeneratorType)` implementation for it to work. This would effectively "freeze" the state of `rng` into the generated code (as it's a global),...

> > RE: [#4499 (comment)](https://github.com/numba/numba/issues/4499#issuecomment-1229235923) @kc611 I think this needs a `@lower_constant(types.NumPyRandomGeneratorType)` implementation for it to work. This would effectively "freeze" the state of `rng` into the generated code (as...

> > Have been thinking about this a bit more. I think the "state" would have to be "copied" and the working parts of the state are: > > >...

Thanks for confirming @apmasell. This is a reproducer: ```python from numba import njit, prange import numpy as np @njit(parallel=True) def foo(arr, pred=None): for i in prange(1): if pred is not...

> @stuartarchibald @apmasell Seems fine to me. Great, thanks for taking a look @DrTodd13.

I guess there's two separate concerns here: 1. `distutils` is deprecated and replacement is required, `setuptools` could provide that. 2. That `setuptools` is potentially not needed at runtime in the...

> IIUC, we include the test suite with installs of Numba, with the expectation that you can run the tests on any installed Numba version - so if the tests...

@minrk @llimeht thanks for the input. I'll raise this at the Numba triage meeting today (and at the public meeting if there's no resolution found by the maintainers during triage)....