Wow. That php program is fast! I was skeptical for much the same reasons as the other commentators. But I took the rust version, packed the strings into u128, eliminated the allocations in the loop, --release, lto, codegen-units=1, target-cpu=native (basically all the tricks to make it fast). I tried FxHashMap, AHashMap, and hashbrown, but they didn't make much difference with the other optimizations (and FxHashMap was significantly slower).

Even with all that, the php version was still 2x faster. And it truly is due to the hashtable implementation itself.