All of that is very dependent on the compiler used to compile it. Using a modern gcc with -march=native will probably have it be somewhere near tcc for performance, at least when I tried it a few years ago. If you compile it with itself maybe much worse? Also this would change based on your processor and so on.