Switch to `nalgebra` everywhere about hotham

[Ralith]

If you want physics, you'll probably use rapier, which means you'll have nalgebra in your dependency tree regardless. It's very powerful but some people find the verbose/obscure type errors from the highly generic stuff offputting, and you might get better compile times if you touch it as little as possible. On the whole I like it, personally.

[kanerogers]

If you want physics, you'll probably use rapier, which means you'll have nalgebra in your dependency tree regardless. It's very powerful but some people find the verbose/obscure type errors from the highly generic stuff offputting, and you might get better compile times if you touch it as little as possible. On the whole I like it, personally.

Good point re: rapier. We're already using it and I'm having to do conversions between cgmath and nalgebra in places anyway which is a bit gross. I have to admit I'd prefer something a bit less generic, but there's (probably) a bigger performance penalty switching around between libraries than just sticking with nalgebra.

[Ralith]

there's (probably) a bigger performance penalty switching around between libraries than just sticking with nalgebra.

I'm actually skeptical of that. The layouts are identical, so LLVM will see that it's a simple memcpy, and it'll bake down to no different than a regular Copy. Might matter if it prevents you from passing matrices by reference in a major way, but otherwise it's probably insignificant. Of course, the cognitive overhead of switching between them all the time is its own thing.

[kanerogers]

There are a couple places where we pass vectors/matrices by reference so it does force us to do unnecessary copies (or an evil mem::transmute), but yes, the cognitive overhead is probably the real problem. :)

…

On Wed, Sep 1, 2021 at 10:51 AM Benjamin Saunders ***@***.***> wrote: there's (probably) a bigger performance penalty switching around between libraries than just sticking with nalgebra. I'm actually skeptical of that. The layouts are identical, so LLVM will see that it's a simple memcpy, and it'll bake down to no different than a regular Copy. Might matter if it prevents you from passing matrices by reference in a major way, but otherwise it's probably insignificant. Of course, the cognitive overhead of switching between them all the time is its own thing. — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#46 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAPNXZ22PU4ENGVP3R6HT33T7V2JDANCNFSM5B3FXG5Q> .