> 1) are anyone pursuing this line of work, and
>Sadly, Urban left us for industry (Carlstedt Research and Technology), where
>Thomas Johnsson is also on sabbatical at the moment. So nobody here is
>pursuing this at the moment.
> 2) is the software available?
>Ask Urban: [EMAIL PROTECTED] ought to reach him still.
My former student Karl-Filip Faxen did his PhD thesis on very similar
compilation methods for lazy languages (Karl-Filip, do you follow the
Haskell mailing list?), and he also got impressive speedups for a set of
admittedly small benchmarks. His compiler does not implement full Haskell,
but there should be no problems in principle to extend it to full Haskell
and his compilation techniques are in no way dependent on this
restriction. He can be reached at [EMAIL PROTECTED] He is still working on his
compiler and he has a paper in the upcoming ICFP conference.
I think Karl-Filip's and Urban's methods are very interesting and the way to
go for efficient compilation of lazy languages. The key to cope with large
programs is probably to have an intelligent strategy for function cloning,
which clones functions to increase the potential to optimise by
specialisation, but only for calls where there is a potential to gain from
this in order to avoid the potential code explosion.
Bj�rn Lisper
