最近看了这篇文章：

https://pointersgonewild.com/2025-08-06-out-fibbing-cpython-with-the-plush-interpreter/

> My PhD advisor seemed to have an obsession with the recursive Fibonacci
> microbenchmark, or just fib for short. It was almost like a ritual. Every
> time he would bring up some kind of compiler or language implementation, one
> of his first goals was to get to the point where he could run this tiny
> program, and if you were working on some kind of compiler, he would ask you
> how fast it could run fib. Sometimes he would also ask to compare the
> performance of your compiler against his Scheme compiler on this
> microbenchmark, and you couldn't tell if he was making some kind of a joke or
> if he was trying to compete with you, or both.
>
>   fun fib(n)
>   {
>       if (n < 2)
>           return n;
>   
>       return fib(n - 1) + fib(n - 2);
>   }
>
> It's well-known that microbenchmarks are not representative of the overall
> performance of language implementations. There are many reasons for that, the
> most obvious being that they tend to overemphasize certain aspects of your
> implementation while completely disregarding others. In this case, the fib
> microbenchmark heavily emphasizes how fast you can execute function calls.
> This is not a bad thing to measure though, as function calls are central to
> modern programming languages, and they can also be a significant source of
> overhead.


里面提到作者的博士导师特别喜欢用这个fib函数来比较编程寓言的性能。我在写自己的Lisp解释器的时候，为
了跟其它编程语言比较性能，也大量使用了这个fib函数测试性能，连代码都一模一样。