[Top][All Lists]

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

Re: [Axiom-developer] CCL maintenance.

From: Stephen Wilson
Subject: Re: [Axiom-developer] CCL maintenance.
Date: 30 May 2007 20:36:38 -0400
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.4

Gabriel Dos Reis <address@hidden> writes:

> On Wed, 30 May 2007, Stephen Wilson wrote:
> | Gabriel Dos Reis <address@hidden> writes:
> | > The slowness of GCL is another irritating aspect of working on current
> | > Axiom code base.
> | > 
> | > Please, don't tell me you're not concerned about wasting developer's time.
> | 
> | Oh no. Though GCL may not be super-duper-fast when it comes to compile
> | times, there are may opportunities for speeding up the build which at
> | the same time have other benefits.  For example, eliminating dead
> | code.  I looked at the email archives w.r.t saturn and the majority of
> | the posts were discussing how to remove the connection to the system so
> | that a windows build was possible.
> I don't know if that is of any help but builds and
> works with $saturn set to the same value as for Unix systems.

Ok, noted.  Thanks for the insight.

> ...
> | Im sure you could aticipate this, but I suspect that getting Boot out
> | of the picture and eliminating (understand, this is just my
> | perspective) a significant layer of indirection would _significantly_
> | reduce compile times.  I would not be surprised if a lisp rewrite saw
> | the SPAD compiler build time drop to 15-20 minutes (I think this is a
> | conservative estimate).  If done properly, the algebra build could
> | drop significantly as well.
> | 
> | In short, I dont blame GCL for axioms long build times.
> In all my measurements, the Boot-to-Lisp translation takes up non-significant
> part of the build.  All the measurements I've done show the compilation of
> Lisp codes  take major proportion.  I don't think it is because of the quality
> of Lisp generated.

I have looked at a lot of the boot generated lisp and I too do not see
significant problems with the code produced from the perspective of
lisp being an assembly language (of course, there is always room for
mechanical optimization) However, I am confident that the manner in
which things are expressed is far from optimal.  Im quite convinced
that human lispers could do a better job of expressing the system in
lisp than the boot compiler ever will.  

Of course, I might be wrong.  Well just have to wait and see :)

> As you know I don't believe in clean, easy modifiable, and maintainable Axiom
> written in Lisp.  Lisp is just too low level.  For one thing, the verbosity
> (I'm not talking of parenthesis) obscures the main ideas; its primitivity
> drags further down.  But, I've said enough on the subject already.

I believe we are polar opposites in the way we see Lisp. 

Im not out to convert anyone.


reply via email to

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