"Outperforming Imperative with Pure Functional Languages" by Richard Feldman

> Functional programming experts say that their code is efficient because the compiler translates it to an efficient representation on the CPU. They have TCO and other things. But they don't mutate data and hence every function call creates a new data in memory taking up space. How do they overcome that? By letting the compiler make assumptions. For example, consecutive map calls can be converted into a single one. If the data isn't shared with something else, the compiler can mutate it. MUCH easier to paralellize maps. As a starting point for those nitty grittys, this talk is probably as good as any [this video]

If you are interested, «why yet another programming language?». The unique selling point of Roc is clever optimization to convert purely functional source code to deep-imperative fast machine code, while keeping all the correctness of functional algorithms. See this video of Richard Feldman for details — «Outperforming Imperative with Pure Functional Languages»: [this video] Among those clever optimizations: - static reference counting (no GC, like in Rust, but with no borrowing and borrow problems); - stack allocation of the data with no external links; - hidden («opportunistic») mutability even in the cases, where Haskell or Lisp will copy the value.