They're not very conclusive though, because the testing isn't fine-grained enough for these microbenchmarks.