Recently developing a systems API. Did a benchmark. A simple response of pong to an http request. Mojolicious Vs go.
Perl was not fast. By a BIG order of magnitude.
The only thing you need to know when it comes to optimising code is very simple: You can never make code run faster. You can only make it do less.
I think with modern processor architectures there are some edge cases where you can shuffle calls around, keep the same runtime complexity, but have things go faster. For example, moving a conditional outside an inner loop function call to inside the function can cause the entire function to be loaded into cache, decreasing runtime.
I have had fun with Go, but I really hope they add something less verbose for error handling in a future version.