I'm just curious how come the techempower benchmarks are showing mitol with a ~20% improvement over the default implementation, while the readme is claiming a 3x improvement?
https://www.techempower.com/benchmarks/#section=data-r15&hw=ph&test=plaintext&l=h4s9h9