axiom-developer
[Top][All Lists]
Advanced

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

[Axiom-developer] lattice


From: root
Subject: [Axiom-developer] lattice
Date: Sun, 17 Aug 2003 15:05:28 -0400

re: lattice

The lattice is quite subtle in places. The hardest part is the size
of the thing. ~1100 domains in ~375 files. Every time I set out to
do anything like change the authors which takes less than 1 minute
per file it turns into a several hour task. Some of the tasks took
over an hour per domain. I've been at this task since January.
I have much "back-checking" to do but I can do that after upload.
(e.g. I have to check that the bootstrap code is current, a painfully
slow process).

I did a partial analysis years ago. It helped me find the most
frequently loaded domains which are now preloaded at build time.
I wish I had kept the data from that analysis because it
would have come in handy now. I've separated out the scaffolding
data into a file called Lattice.pamphlet. This contains information
that allows us to compute several interesting answers to questions
like: what's the "thickness" of the "main trunk" of the tree of
files holding up the lattice. I observed over the last few months
that there are only a few key files that hold the whole lattice up.
I'm also thinking of writing a lattice domain as I can think of
several operations that might be interesting to compute. Also of
interest is that the lattice data tells us where optimizations will
have the most effect.

As a side-effect the Makefile.pamphlet has shrunk. Once I add in
David's change it should shrink again. You still won't be happy
with it though :-)

re: interval.spad.pamphlet

Your interval.spad.pamphlet file is not a trivial rewrite as far
as I can see. The file diffs are almost nowhere the same. My general
rule for "authorship" is to include the names of people who make
"intellectual changes" (that is, not just typos but rewrites or
handling special cases, etc) as authors. Credit is easy to share.
Once the file has more than 5 authors it'll switch over to the
generic "The Axiom Team". Currently 5 authors is the longest list.
 
I'm also looking to rewrite the .as files. It simplifies things
if the basic algebra is all .spad rather than a mixture. I've
ignored the .as files in the build for this reason. If you're
feeling ambitious you could look at the other few .as files.

re: value stack overflow

Yes, it is very strange that I get a value stack overflow because
I used to build the databases. Indeed, I rebuilt the databases for
the hand-built version. I'm going to back down a level in the GCL
build and see if that fixes it. If so I'll upload the world using
the backlevel lisp until we get a chance to debug it.

re savannah 

I'm working hard to get the savannah site set up so everybody is
working from the same set of mistakes :-) However, there is no
point in uploading mistakes I can easily fix. It is a tradeoff
but I'm leaning heavily toward the upload side now that the world
can be built from scratch.

Tim
address@hidden
address@hidden







reply via email to

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