Possibly due to the lack of LTO, Link-Time Optimization. Rust standard library and user program get separately compiled and needs a special procedure to inline

Possibly due to the lack of LTO, Link-Time Optimization. Rust standard library and user program get separately compiled and needs a special procedure to inline them out. I couldn't find Cargo.toml used in this benchmark but it ought to contain `lto = true` somewhere in the profile section to get the optimal performance especially for this kind of tight loops. (It makes compile slower, unfortunately, and no profile currently turns that on by default.)