I have a specific scheduling problem that I can solve easily using
FD.distribute (<1 second) but when I change the distribution strategy to
Schedule.firstsDist I left it running overnight with no solution. At
first (somewhat superficial) reading about these distributors I was kind
of expecting firstsDists to do better than that.
I was hoping some of you guru's might have suggestions on why this could
be so. In particular, I have some more challenging schedules of the
same class to solve (none of which I have been able to solve within a
few hours with either distribution method).
From my largely naive point of view, I was thinking that although the
problem seems somewhat like a job shop problem, there is only a single
resource, so distribution schemes that distribute based on overloaded
resources are going to have limited(!) choices. Also, the 3, relatively
long batches are ordered, limiting the opportunities for swapping tasks.
Its another case of the boys over in engineering hitting another brick
wall in their scheduler and me trying to leverage Oz to at least catch
up to them and hopefully overtake the performance of their C++ scheduler.
I have posted some files at:
- http://www.metrak.com/tmp/data_3.oz (schedule specification and
description)
- http://www.metrak.com/tmp/scheduler.oz (schedule compiler)
- http://www.metrak.com/tmp/data_3l.svg (graphical representation of a
solution)
I have been running these in the emacs environment, ie first feed
data_3.oz then scheduler.oz. At the bottom of scheduler.oz check the
commented parts for different output options.
paul sorenson
PS if there are any scheduling gurus who are located in Melbourne or
other parts of Australia who would be interested in offering some
consulting time to me, please let me know.
--
brewiki: http://brewiki.org/
Brewsta cross platform, open source brewing software:
http://brewsta.sourceforge.net
_________________________________________________________________________________
mozart-users mailing list
[email protected]
http://www.mozart-oz.org/mailman/listinfo/mozart-users
