On Mon, Oct 25, 2010 at 07:24:00AM +0000, Simon Peyton-Jones wrote:
> Now that *is* strange. The change only affects Lint, ant it affects Lint only 
> by gathering the top-level names (done already) and checking that "M.x" 
> doesn't occur twice (for any M, x). I would not have expected that to double 
> space usage!
> 
> Might you investigate a bit more with your profiled compiler?

Aha, top of the profile:

findDupsEq                     ListSetOps            14.1   35.6
lintCoreBindings               CoreLint               7.3    0.1

I imagine the quadratic behaviour of findDupsEq is the problem, with
most of the lintCoreBindings time being spent in eq_ext. So I guess we
want a findDupsOrd, which sorts first?


Thanks
Ian

_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc

Reply via email to