[Haskell-cafe] Re: Haskell and .NET

[Neil Mitchell]

Hi

> > And strictness is not the biggest problem, a complete lack of any
> > optimisations is, but I'm working on that one too!
>
> It is great to hear that. Of course an optimiser will be beneficial
> too but I guess that even the benefit that the code generator can have
> from the strictness analyzer will make huge difference.

In the GHC back end I seem to remember strictness gives a 10%
performance improvement, but let motion gives a 30% improvement - its
quite likely that .NET will benefit more from strictness, but still
the general optimisation is probably going to be a bigger win.

I think basic strictness information gained about 5% in the Javascript
Yhc backend, so not great amounts.

The optimiser does generalised deforestation, and loads of other cool
tricks, so I'm hoping for at least 50% over the normal Yhc code :)

Thanks

Neil