On 28/11/13 20:26, Rémi Fontan wrote:
Hi,
would you know what is the cost of implementing the double dispatch
technique as described on following link? (section "What if I want
overloading?")
http://smallcultfollowing.com/babysteps/blog/2012/10/04/refining-traits-slash-impls/
does using traits means making used of something similar to vtable? Or
would the compiler optimise those double dispatch to a very optimise code?
The motivation is to have multiple implementation of mathematical
operators for struct like vector 2d, 3d, matrices, and having them as
efficient as possible is important.
cheers,
Rémi
Generics are always monomorphised to be static dispatch; explicit trait
objects (~Trait, &Trait, etc) are the only times that one gets dynamic
dispatch (using a vtable). So, trait methods be purely compile time if
you never box the types into trait objects, i.e.
fn static_with_generics<T: Trait>(x: &T) { ... }
fn dynamic_with_vtables(x: &Trait) { ... }
The layered traits like in the example in that blog-post are the same as
normal generics, just the compiler has to do a little more work (but
this is only at compile time).
Huon
_______________________________________________
Rust-dev mailing list
[email protected]
https://mail.mozilla.org/listinfo/rust-dev
