[Gcl-devel] Re: inlining

[Camm Maguire]

Greetings!

This should not be too difficult, but I think it would require a
modification we've discussed before -- keeping the lisp source around
post compilation.

In fact, I don't see whay a cl compiler could not do as in scheme as
mentioned in one of your other posts and automatically inline mutual
recursion too.

Currently, GCL will inline lambdas with onla a small subset of the
possible lambda list special symbols (e.g. &optional et.al.) present. 

Take care,

Robert Boyer <address@hidden> writes:

> It seems to me that ideally and eventually GCL could provide
> the possibility of inlining calls to a user's Lisp function,
> say foo, for the mere expression (declare (inline foo)).
> 
> I am not sure whether this has ever been done really well
> for any programming language, but it is something that
> programmers do in their heads all the time, but with many
> mistakes.  In a nice, clean mathematical setting such as the
> lambda calculus, it is a piece of cake, to inline,
> logically, in principle.  But as soon as one starts
> manipulating bound variables, historically terrible mistakes
> occur.  Throw in order of evaluation issues, keywords, and
> all the glories of Common Lisp and you have something that
> is almost unthinkably hard in the general case.  In the old
> days, when every reference to memory cost a uniform time,
> say, one microsecond, inline optimizations could be
> fabulously effective.  Now, with so many different caches,
> pipes, etc., I don't really know.
> 
> Bob
> 
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah