[lkb] Re: GLB types in LKB

[Stephan Oepen]

hei igjen!

on 29-sep-04, Nicholas Yates <nyates at ulb.ac.be> wrote:

> I was wondering if someone could help me find some detailed
> information of how the automatic glbtype creation works in LKB. Is
> there any publication or stuff I could look up apart from the
> comments in the source code. I would like to implement something
> similar in a prolog program.

the semi-lattice embedding (aka introduction of necessary glb types to
yield a BCPO) in the LKB is the same as in PET: it is discussed in the
MSc thesis of Ulrich Callmeier, the PET developer (section 3.2.2):

  http://www.coli.uni-sb.de/~uc/thesis/thesis.pdf

the two Kiefer et al. papers that Dag Schmidtke referred to earlier on
this list, on the other hand, discuss optimizations to /run-time/ type
unification.  the LKB and PET, incidentally, use the caching scheme in
run-time glb computation (plus a specialized, perfect hash function);
again, see the Callmeier (2001) thesis for details.

                                                all the best  -  oe

nb: i had attempted this posting several weeks ago, but because i was
subscribed to the list with an out-of-date email address, the posting
bounced.  using the old list manager, it was a bit tedious to work out
my subscription address and correct it, hence i waited for the list to
move to the DELPH-IN Mailman first :-).

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+++ Universitetet i Oslo (ILF); Boks 1102 Blindern; 0317 Oslo; (+47) 2285 7989
+++     CSLI Stanford; Ventura Hall; Stanford, CA 94305; (+1 650) 723 0515
+++       --- oe at csli.stanford.edu; oe at hf.uio.no; stephan at oepen.net ---
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++