|
|
|
|
|
by alex_muscar
1081 days ago
|
|
That is just untrue. Compilers tend to generate better code for imperative algorithms because current architectures implement a fundamentally imperative model. Yes, "sufficiently advanced" compilers can generate code that has similar performance to imperative algorithms from functional ones. But that requires some heavy lifting static analysis and optimisation passes. Have a look here: https://godbolt.org/z/sPT3snWE4. Only if you remove `#[inline(never)]` from `add1` will you get efficient code for the functional algorithm. This is a trivial example, but it ilustrates some of the nuances. |
|