GHC lazy eval optimization bug
dons at galois.com
Sun Feb 3 15:15:37 EST 2008
> I can confirm this performance bug with GHC version 6.6.1 on OS X PCC (G4)
> It performs correctly, lazily keeping 'primes' for all 4 repetitions, when
> using "ghc -Onot" but the the doSlow fails to retain 'primes' when
> compiling with "ghc -O" and "ghc -O2"
Sounds like the known issue of the simplifier eta-expanding `State#`
lambdas on purpose, which is good for IO, but risks re-computation. The
'calendar' nofib program exhibits this, iirc.
Try compiling with:
More information about the Glasgow-haskell-users