Ralf Hinze <[EMAIL PROTECTED]> writes:
> | Is this true in practice? That is, are there programs which have
> | different asymptotic running times when compiled under ghc or hbc than
> | when running under Hugs?
> |
> | It would actually surprise me if there were; I'm having a hard time
> | imagining a realistic optimization that would do this. (Which could
> | easily be a failure of my imagination.)
>
> What about common subexpression elimination? AFAIK neither Hugs nor GHC
> nor HBC implement CSE, but if they did the asymptotic running time
> could sometimes radically change.
OK, so my imagination is severely lacking. :-)
I'm still curious about my first question, though, about the specific
optimizations included in ghc and hbc. If in fact they don't do CSE,
are there optimizations which they do perform which would change the
asymptotic running time?
Carl Witty
[EMAIL PROTECTED]
