> Thus, in this single case when a Rust implementation is more or less faster than C, the performance difference is not about better compiler, but about more efficient structure of the program, which allows a compiler to optimize code better.

This is strange takeaway.... I would say that's almost the only thing that matters. They compared one Rust compiler to 3 C++ compilers and picked the best result? Who does that in practice - who compiles their codebase with 3 different compilers and picks the most efficient one for each object file? Also, the compiler can often improve, where as the language itself is much more difficult to improve - the fact that the language lets you to write a more efficient (and more readable!) structure is crucial.

Also, they seem to misunderstand the point of Rust entirely. The main point of Rust is "safety" (w/o sacrificing performance, yes - but safety was the primary design goal). And for a good reason! Systems programming is more about safe systems than it is about fast systems - fast buggy system programs are useless. The authors decry the loss of goto saying it's "good for juniors but too limiting for professionals" - as if professionals aren't humans too! I'm sorry to say that, but whenever I saw that attitude before - it was with programmers that greatly overestimated their skill level.