Open deepinthebuild opened 7 years ago
I'm down for this; but I'm not gonna make it a priority, just because I think it could take a lot of effort to implement under the hood.
Having done some profiling, though, it does look like this is close to the only big performance optimisation we have left – Rope.split()
does spend a majority of it's time in &str.to_owned()
. Avoiding splits on small strings might help us here...
We should change the behavior of
insert
etc. to not split leaves that are below a certain size threshold. It might help with performance due to cache friendliness.