[Prev][Next][Index][Thread]

Re: HARLEQUIN DYLAN and C FFI



On Wed, 2 Feb 2000, Brian Rogoff wrote:
> On Wed, 2 Feb 2000, Jason Trenouth wrote:
> > On Wed, 2 Feb 2000 01:00:02 -0500 (EST), "Shawn" <shawn@anarchy-arts.com>
> > wrote:
> > >     ... I thought that Harlequin Dylan would allow me to step away
> > > from using C/C++ for my development (PC Games) but it seems to have
> > > floated off into obscurity. 
> > 
> > That would be a shame because we're still here.
> 
> Jason,
>     I was under the impression that Harlequin Dylan was no more and that 
> the rights were sold to Functional Objects, Inc. Could you expand on your 
> statement a little? ...

I believe I may do that for him: "we" here is roughly "Functional Objects
and some ex-Harlequin-Dylan people helping out" (including Jason, who's
still at Hqn, and me, who's not :-).  HD has effectively become FD, which
is still in beta, but is equally still alive.

> ... The Dylan situation is confusing (as well as tragic :-() and since
> the FO web site isn't very informative it isn't surprising that people
> suspect the worst. 

I'm not in an official position to give anything like release dates, I can
only repeat that "we're still here".  But we'll all be happier when the
2.0 release actually arrives :-)

> ...  I don't think dynamic languages can give you the same small memory
> footprint or high performance as static languages. That was supposed to
> be the point of the type declarations and sealing in Dylan, right?

I suppose you're broadly right, though I haven't seen any numeric
comparisons beyond simple benchmarks.  Dylan is supposed to be able to
improve speed and space over Lisp (and Smalltalk?), but not necessarily
get quite as small as C++ (certainly not as C).  OTOH, it definitely
improves expressibility over C (and I'd say over C++).

The theory is that you can do high-level stuff in a dynamic language like
Dylan and low-level stuff in C etc.  As for the practice, I believe "Crash
Bandicoot" (or similar) on the PlayStation has its high-level map and
actor control written in a home-grown Lisp interpreter ... and I'm sure
someone at Harlequin had an "interacting with DOOM" demo ;-)  But we've
yet to see large-scale performance comparisons in any app area (unless you
count HD/FD itself -- vs. JBuilder/MSVC++, anyone?), let alone
high-performance areas like games.

HTH,
Hugh






Follow-Ups: References: