[Top][All Lists]

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

RE: [Axiom-developer] Re: lisp

From: Bill Page
Subject: RE: [Axiom-developer] Re: lisp
Date: Tue, 6 Apr 2004 17:06:34 -0400


It seems to me that you are talking about something else. A different issue.
Fast link is just a GCL thing isn't it? It just means that GCL produces more
efficient C code that gets compiled into a faster binary. This is not an
algebra "fixed point" in the sense that I meant it.

What I am talking about is the situation where the lisp generated by AXIOM
for a bootstrapped module might not be the same as the original lisp code
that was used for the bootstrap. This could be the case if changes are made
in any of the algebra modules in the loop in which the bootstrap module
occurs - a kind of feedback - because the dependencies are not really a
lattice, they form a directed graph with cycles. The bootstrap method is a
way to break the cycles so that what we end up compiling is a lattice of
dependent code. As I understand it, the lisp code for the current bootstrap
was obtained by you from a previous version of AXIOM or maybe even some of
it was hand-compiled. The code that gets generated in the end might depend
on the initial choice of bootstrap LISP code. This is not desirable because
REALLY the AXIOM algebra is supposed to be completely specified by the SPAD
code and all LISP should be derived from that.

To get correct LISP code for all modules in the loop you have to take the
generated lisp and feed it back into a recompile and check again whether the
generated lisp is the same as the bootstrap. As far as I know we (at least
I) have never checked that all the generated lisp for the bootstrapped
modules is equivalent to the original bootstrap code.

Am I clearly describing what I mean?

Bill Page.

> -----Original Message-----
> From: root [mailto:address@hidden 
> Sent: Tuesday, April 06, 2004 4:56 PM
> To: address@hidden
> Cc: address@hidden; address@hidden; address@hidden
> Subject: Re: [Axiom-developer] Re: lisp
> re: fixed point. Close but not quite the issue. The reason 
> for the second compile is performance. The first compile will 
> generate .fn files. If they exist the GCL compiler can use 
> the type information from the .fn files to generate better code.
> Tim

reply via email to

[Prev in Thread] Current Thread [Next in Thread]