Support for `std::complex<SIMD>`?

[ax3l]

Hi,

Does this library already support std::complex for SIMD types? :)

[ax3l]

@mattkretz I probably just need a more detailed example. Looking at #34, it looks like you support complex numbers by vectorizing over the tuple of each complex or so via simdize.

[mattkretz]

The PR added a check to make sure that std::complex is not reflectable and thus simdize refuses to work with it. The intent here is that complex itself should become a vectorizable type with additional semantics over simdize.

Now, I've been implementing simd<complex> since the start of the year for C++26's std::simd. For lack of time (or manpower, however you want to put it), I won't be able to easily bring this back into vir-simd. It should be possible to somehow hack it in though.

What I believe would be simpler to do is a specialization of a user-defined complex class template for simd value-types. That's almost a straight copy of std::complex with a few adjustments for the few places where it needs branching. But again. Manpower/time issue, since I need to focus on C++26 std::simd.