Marcin Pączkowski

How's this coming together @elgiano ?

Thanks @elgiano ! Skimming through it it looks great! I think it would be beneficial to make this more discoverable. Can you add a link in `News_3.14?.schelp`? Or elsewhere, if...

> OT: I noticed that in News, we still have one "known issue" that has been resolved for rc2 (View.setBackgroundImage not working). I could remove it in this PR while...

> Any ideas what's going on? I tried restarting it again, to no avail. It seems that the server hangs and prevents new servers from running after `TestCoreUGens: testbufugens`. ```...

I can confirm this issue. This is not good :/ I can also see a difference eg in `2.sqrt` on the server: ```supercollider { DC.ar(2).sqrt.poll }.play; // UGen(UnaryOpUGen): 1.41421 -...

It seems that the problem occurs in the `nova-simd` class here: https://github.com/timblechmann/nova-simd/blob/master/detail/vec_math.hpp#L550-L558 I can see NaNs here when I set a breakpoint there. Also, I'm not sure if `VecType::gen_sign_mask();` is...

@timblechmann do you have any ideas how to solve this?

@timblechmann indeed `signed_sqrt_tests` fail, see https://github.com/timblechmann/nova-simd/issues/11. Do you have any suggestions how this can be fixed?

Additionally: `.reciprocal` is also broken. ```supercollider x = {DC.ar(2).reciprocal.poll; Silent.ar()}.play; // UGen(UnaryOpUGen): 0.499998 ``` A workaround is coming up.

https://github.com/supercollider/supercollider/pull/5901 fixes the behavior on M1. I would consider keeping this issue open for now and close it once the issue is resolved in nova-simd. At that time the workaround...