Is it intentional that the benchmarks include not only running the program itself but also compiling it? e.g. in the linked source code, the bench.sh includes the compilation step which is known to be slow in rust:

    #!/usr/bin/env bash

    set -e
    run() {
      cargo build --release 2> /dev/null
      ./target/release/rust
    }

    run;

Sure, if you run it many times in succession the compiler won't do much but the benchmarking script (run.js) doesn't really indicate that and the blog post also doesn't mention that.

EDIT: I was just being stupid, don't mind me. The times were taken within each language and not externally.