Adam

Can you provide the C# code for each of the IL code snippets? To be really sure how this ends up on the CPU, we'll also need to compare the...

> The last one is also compatible with the in keyword as you do not modify the vector you get in through the parameters (as in you can add the...

> Idk really I just noticed this when I looked at the IL and decided to measure it. Could be a good idea to try to reproduce it to see...

Okay, that really looks like there's an obvious perf gain from restructuring the operators that way. Very nice find 👍 If you're up for it, this could be a worthwhile...

> Also I fixed some trivial code duplication stuff (like vec * float and float * vec now use the same code instead of separate implementations that do the same...

> I didnt removed them I changed the reversed variant to use the logic of the non reversed variant in order to reduce code duplication. Ah, makes sense. If this...

Making vector or matrix types readonly would be quite a usability downgrade, since that would prevent field assignments. We can still avoid performance regressions though, if we use [the new...

I acknowledge the technical possibility, but that just reads terrible and wrong 😄 (Also, ref / out support, additional copies and potentially other stuff I'm not even thinking about -...

Yeah, I was eyeballing that possibility as well - definitely something to keep in mind.

This should technically be moved to [the docs repo](https://github.com/AdamsLair/duality-docs), though I guess that would hurt the general overview on works regarding the current milestone, so let's keep it here. The...