@tenderlove @jhawthorn @flavorjones Ruby is definitely faster than most folks think and YJIT is doing *serious* work here! 🎉

Reading this has me curious about a few things, though. A pure-Crystal benchmark for this method runs in 12ms on my machine, 80ms in Crystal without `--release`, and 112ms with Ruby+YJIT. John mentioned FFI overhead, but I also wonder if this gem doesn't (can't?) enable LLVM optimizations on the Crystal code.