[Top][All Lists]

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

Re: [Axiom-developer] Building Axiom twice

From: Martin Rubey
Subject: Re: [Axiom-developer] Building Axiom twice
Date: 25 Nov 2006 09:58:21 +0100
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.4

Dear Waldek,

Waldek Hebisch <address@hidden> writes:

> The main reason is that NTL uses algorithm with better asymptitic complexity.
> So, Axiom also needs faster algorithms and here build time matters very much.

I disagree here. When developing new algebra, I will not immediately integrate
it into the system, i.e., build time doesn't matter at all.

What does matter, and in fact, it is a pain currently, is time needed by the
compiler. That's why I switched to Aldor.

The compiler has the misfeature of taking an incredibly long time on certain
functions, and the amount of time needed changes when you shift the function
definition towards the end of the package, say. For an example, try to compile
my guessing package, with ord1, ord2, deg1, deg2 defined at other places (i.e.,
just before they are needed). I suspect that the reason is that SPAD (the
compiler) remembers imports (since they are mostly done implicitely in SPAD)
and has to try them all before succeeding.

I think time should be mostly spent on documenting and improving compiler, but
I must say that I'm looking forward to the point where adding algebra is
possible :-) 

(currently, it is not possible to add my guessing package, since we would need
to ship a new database as well... As you know, I think the solution would be to
start with an almost empty database and rebuild the database after each layer.)

All the best,


reply via email to

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