Haskell doesn't assume any ordering, but nobody managed to make it scale across cores automatically. The problem is that automatic parallelism like you suggest creates lots of small tasks, but to go fast you need fewer large ones because of threading overhead.