| > though, I think this is a great oprotunity to improve ghc's optimizer. | | Huh, that would be the best thing with the whole shootout endeavour.. Yes indeed. One thing that would be really helpful, as a first step, would be to identify a bunch of concrete examples that GHC "should" have optimised, or perhaps "could" have optimised, and attach them to the GHC wiki somewhere. One place would be here http://haskell.org/haskellwiki/Performance or perhaps somewhere on the developers wiki, or as a task. http://hackage.haskell.org/trac/ghc I'm not promising that anything will happen fast (unless someone else does it!) but it's always much more likely to happen if there's a specific place that advertises low-hanging performance fruit. This is closely related to the "advice to programmers" that Don circulated yesterday, in which he explained how he optimised one particular case. But not quite the same. Simon