Optimizing The New Rust Trait Solver
This year, we're stabilizing Rust's next trait solver. lcnr talked about this at length at EuroRust last year. An important blocker has been its performance. Often the answer to "why is the old solver so fast" is "well, it is unsound..." Counterintuitively, micro-optimizations don't matter much: where we actually gain speed is the algorithmic level. For example, just by being better at figuring out when it's correct to reuse previously cached results, suddenly we gain 60% on some benchmarks. In this talk I'll tell you all about it, and much more!
