How long does it take for you to be comfortable with Haskell?

There is a very specific reason why quicksort ain't so quick in Haskell. It is a God-like example of an algorithm that has brilliant hackery woven into how it works - what I mean by hackery in this case is the kind of techniques that a true Haskell devotee would regard as unnecessarily dangerous and non-mathematical. The original implementation made every effort to break the rules that Haskell imposes on itself: the genuine quicksort works by overwriting storage slots with new information. This is very painful to do in Haskell, which finds it much easier to make whole new copies of existing information.

So although that naive two line Haskell version captures something of the essence of quicksort (it does the same number of key comparisons), it isn't really quicksort. It's missing a large portion of the genius that went into it, which took full advantage of the ability to tweak the state of existing values. So it makes large numbers of intermediate copies of pieces of the list.

Speculation: could a Haskell compiler analyze your code, applying the same reasoning as Hoare (inventor of quicksort) and figure out that it could optimize it by completely reimplementing it in a stateful way? Possibly.

The reason is that mathematical foundation of computers is different from mathematical foundation of Haskell and functional languages in general.

To get taste of problems compiler is facing... Translate Haskell program to C, keeping it as close to original as possible, then try to optimize that C code (without rewriting it from scratch in C-way)

It's not dumb, but functional languages are not made for performance, and concise mathematically simple notation doesn't come for free.