Stan Seibert

Since the CPU version of the code-path with Hemi is inherently single-threaded, I imagined that the atomic operations when compiling for the CPU would literally be things like: ``` c++...

That makes this harder, since I can't think of a generic way to implement the atomic that works for any multi-threading situation. If Hemi specifically only supported OpenMP for multi-threading...

Related issue is how to handle bfloat16, which seems to be gaining traction and will have limited AVX-512 support on very recent Intel CPU architectures, and is also used on...

I ticked the `half` box for you.

It also looks like LLVM 11 adds bfloat16 support: https://github.com/llvm/llvm-project/commit/8c24f33158d8 The type name seems to be `bfloat`.

@esc: I think this might be a different issue. The script reproducer above isn't running the Numba test suite with pytest, it is importing numba and pytest into the same...

I'm trying to reproduce this bug, but need some additional details, since I think the cause is something subtle: * What channel did you get the Numba 0.55.2 package and...

I don't see the bug on Windows 10 with Numba 0.55.1 and pytest 7.1.2, both from the Anaconda main channel, but that is not exactly the configuration listed above, so...

OK, if it is true that the error is related to our old vendored copy of appdirs (from https://github.com/ActiveState/appdirs), then I suspect that we probably want to update that, given...

OK, I have access to a Windows 11 system now, but I have not been able to reproduce the above crash bug. I've created environments both ways: ``` conda create...