Your code is fast – if you're lucky

Recently, while I was working on an optimized Quicksort implementation, I came across a rather interesting quirk. Modern compilers (especially Clang) optimize loops using fast, branch-free instruction...