Idea: Optimising compile-time filter chains #272
Labels
area/filters
Related to Quilkin's filter API.
area/performance
Anything to do with Quilkin being slow, or making it go faster.
kind/design
Proposal discussing new features / fixes and how they should be implemented
priority/low
Issues that don't need to be addressed in the near term.
Someone recently wrote this crate (
zero_v
) as a prototype of an approach to optimise out vtables when you know the collection of dynamic objects that you’re going to be running. They have a benchmark you can run and the gains are pretty substantial from what I can see (~100ns to ~20-30ns).I asked about making this runtime-friendly so that you could still have this when you update at runtime (such as what we do), and it seems relatively straightforward, as we’re already doing the enum approach for configuration.
The text was updated successfully, but these errors were encountered: