> Spreading FUD about the language they compare against

Yeah, that's a big part of why I'm not a fan. They never reflect the actual 
experience of using a language.

> FWIW we actually do pretty well on the compilation speed: 
> [https://vlang.io/compilation_speed](https://vlang.io/compilation_speed). I 
> find it incredibly unlikely that this benchmark hasn't been gamed in V's 
> favour though, 0.6s?

It's not gamed, but the small print has been omitted. Basically, V has two 
compilation modes. One is a dumb tcc-style compiler that doesn't really 
optimize anything, the other is to compile to C (like Nim). You get the fast 
compilation speed by basically having a very fast backend and a language with 
simple semantics.

Problems:

  * It's really not that much of a benefit, as long as you have incremental 
compilation. The test compiles massive standalone files that you will rarely 
see in practice and which may actually trigger non-linear behavior that 
compiler writers don't care about, because nobody will write such code.
  * If you need optimization during development, you lose the fast compilation 
speed. Many programs require at least some optimization to run at acceptable 
speeds even while debugging them.
  * The fairly large tradeoffs between compilation speed and code performance 
may encourage bad programming practices, such as manual inlining to keep the 
compilation speed of the dumb compiler and being able to live without 
optimization.


Reply via email to