In email with subject: [fricas-devel] Re: Alert, heavy patch committed ...
On Wed, Sep 10, 2008 at 2:30 AM, Martin Rubey wrote:
>
> Did you use --with-algebra-optimization="((speed 3) (safety 0))" ?
> When you do, the resulting friCAS should be quite a bit faster...
>
I used only the defaults in configure. I will try --with-algebra-optimization
next.
> I am *very* impressed by the robustness of Peter and Ralf's aldor
> interface! And it's wonderful to be able to build it just using
> --enable-aldor. This uncovers quickly many little bugs.
>
Yes, it's great.
My next objective is to make Aldor available to Sage users via the
Axiom interface in Sage. Recently Mike Hansen has made some major
improvements to the 'axiom.py' interface that will likely be available
in the next release of Sage. And I recently added the ablity to run
the Axiom library compiler from the Axiom interface in Sage, e.g. like
this:
sage: axiom.compile("""
)abbrev domain TEST Test
Test(): with ...
""")
This is Spad code by default but it would already work to compile
aldor code if an appropriate file name is provided:
sage: axiom.compile("""
Test(): with { .. }
== add {
...
}
""", filename="test.as")
(Unlike Spad, file names are significant when compiling Aldor code in Axiom.)
So what I would like to do in produce an update to the Sage
'fricas-1.0.3.spkg' optional package to include building the Aldor
interface if Aldor is present.
The Sage people, however are very concerned about the build time for
packages. To solve that problem for FriCAS Waldek implemented the
cached-Lisp approach. The spkg distribution of FriCAS includes
machine-generated architecture-independent Lisp code. When the package
is installed by a Sage user, FriCAS is built from the cached Lisp in
about 5 to 10 minutes (compared to 1 to 2 hours from raw source).
The build time for the Aldor interface will be an issue unless we can
find a way to do something similar for it. I know that Aldor is able
to generate high-level architecture-independent intermediate code and
even Lisp code, but I am not sure if the make scripts written by Ralf
make caching of this intermediate code or Lisp possible. Building the
Aldor interface currently takes about 30 minutes. We know that this
can be reduced at the expense of a larger library file by optimizing
the use of 'ar'. (Maybe this is due to a bug in 'ar'? Or is it a
problem in the way we are using 'ar'?)
It would be nice to reduce the build time for the Aldor interface to
something under 5 minutes.
With the work being done by Pippijn on the Aldor compiler, it may soon
also be possible to package Aldor itself as an optional Sage package.
Because of licensing issues the FriCAS and Aldor package for Sage must
currently remain separate. I am doubtful however if compiling Aldor
from source can be made much less than 10 or 15 minutes in total. So
maybe it is a good thing that they remain separate packages anyway.
Regards,
Bill Page.
_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org
signature.asc (196 bytes) 