7 Jun
2005
7 Jun
'05
10:44 a.m.
On Tue, 2005-06-07 at 10:52 +0100, Simon Peyton-Jones wrote:
There is definitely a performance bug somewhere in GHC, and we're collecting cases that show it up. It'd be ideal if you, or someone else, could snip out a standalone module or two that demonstrates the problem, rather than all of WASH. If not, we'll try to tackle it directly.
I've got a happy grammar that takes ghc many minutes and hundreds of Mb to typecheck if I leave off the types for all the productions. If I add types for all the productions it's pretty quick and uses very little memory. But I think this is a known issue (it's mentioned in the happy manual) and not the one you were interested in. Am I right? I was just suprised that it is such a dramatic effect! Duncan