Unless we were compiling I think ~100s, just doing the naive thing would actually have the same runtime performance