Joakim:
Of course, that's meaningless if they were run on CPUs that varied a lot in ability,
The CPU is the same (mine) but the Haskell timings is imprecise and can't be relied upon (perhaps I have to recompute it). (Rosettacode has a policy of not showing the run-time of programs. They are probably making an exception for me because of some reasons and because they are gentle with me, but I understand their reasons are also good.)
but am I right in guessing that you wrote the Haskell version too?
The Haskell entry was written by someone else, but later I have cleaned up it a little in several ways.
If so, I do think it means something that the simple version for D was so much faster, but depending on the possibly easy optimizations missed by the Haskell implementor, maybe not much.
There are so many differences between the way Haskell manages lists and trunks lazily compared to how D manages ranges... very different tradeoffs on many different levels. The size of this newsgroup post is not enough to even list them :-)
Yes, the skill of the contributors and amount of time spent matters a lot, no doubt.
It's not just a matter of skill of the contributors and amount of time spent, it's also first of all a matter of how much semantically clean you want to write the code, how many abstractions you accept to remove from the implementation, how much you care for many factors more than performance, etc.
Bye, bearophile
