I don't like how the author has labeled tables as "Linux" and "Mac", when really most of the differences between those columns are the result of the compiler used, and to a lesser extent, the fact that the Mac was a local machine while the "Linux" tests were done on a remote server.

A much more useful comparison would keep everything constant except the single variable that is different. This could have been done by utilizing the same hardware, and only using a different compiler. Since both GCC & Clang work on both linux & mac, there is no excuse.