laffra@serc.nl (Chris Laffra) (10/10/90)
Question: Does anybody out there has any experience with, or pointers to parallel solving of constraint languages problems? I read the book by Wm. Leler in which he indicates that the subject of parallelism is an area yet to be discovered. Has anyone tried to solve constraints in parallel? What are the problems? Is it faster? Is it worthwile? Can you combine constraints with a parallel object-oriented language? Chris Laffra SERC-The Netherlands
hawley@uunet.UU.NET (David John Hawley) (10/11/90)
In article <10852@hubcap.clemson.edu> laffra@serc.nl (Chris Laffra) writes: >Question: > >Does anybody out there has any experience with, or pointers to >parallel solving of constraint languages problems? There were some initial results published by the CHIP (constraint logic programming language) group at ECRC, on combining finite-domain constraints (i.e. CSP) with the PEPSys parallel logic programming language/system. I've heard rumours that this work has been abandoned because of poor performance. A visiting researcher did some work on single and all-solutions search wrt CSP problems, and got quite reasonable results for systems on the order of 10s of processors. I don't remember the details. For constraint solving using a symbolic algebra approach (for example, to solve polynomial equations), there has been some work on parallelizing the Buchberger algorithm, reported in "Computer Algebra and Parallelism", Dora & Fitch, Academic Press. The results are not encouraging. The CAL group at ICOT is very interested in this approach, and so we are continuing to explore it's parallization. >Can you combine constraints with a parallel object-oriented >language? There is some work being done at Xerox PARC by Kahn & Saraswat on this, based on the message-based programming paradigm often used with concurrent logic languages. However, the class of constraints considered maybe different from the usual conception. To the extent that concurrent logic languages can be considered to be object-oriented, both ourselves and some people at SICS are working in this area. --------------------------- David Hawley, ICOT, 4th Lab csnet: hawley%icot.jp@relay.cs.net uucp:{enea,inria,mit-eddie,ukc}!icot!hawley ICOT, 1-4-28 Mita, Minato-ku, Tokyo 108 JAPAN. TEL/FAX {81-3-456-}2514/1618
chassin@csinn.uucp (Jacques Chassin de Kergommeaux) (10/22/90)
In article <10906@hubcap.clemson.edu> kddlab!icot32.icot.or.jp!hawley@uunet.UU.NET (David John Hawley) writes: >In article <10852@hubcap.clemson.edu> laffra@serc.nl (Chris Laffra) writes: >>Question: >> >>Does anybody out there has any experience with, or pointers to >>parallel solving of constraint languages problems? > >There were some initial results published by the CHIP (constraint >logic programming language) group at ECRC, on combining finite-domain >constraints (i.e. CSP) with the PEPSys parallel logic programming >language/system. I've heard rumours that this work has been abandoned >because of poor performance. > The work you are mentioning was performed by Pascal van Hentenryck from the CHIP team with the support of myself from the PEPSys team. Some results have been published in the 6th. ICLP at Lisboa, in 1989. In addition to the published results, many other *real* programs have been run. Contrary to the "rumours", performances were *excellent*. Later on, the Sequent Balance prototype was ported on Sequent Symmetry by Pierre Heuze', who used it to develop a secondary structure prediction program, with again *excellent* results (published at ECRC, as ElipSys research report number 10, in december 1989). The fact that Pascal Van Hentenryck is now at Brown (pvh@cs.brown.edu), Pierre Heuze' at Paris and myself at Grenoble is perhaps a better explanation for this work being abandoned than poor performances. Next time you should check real information instead of reproducing uncontrolled "rumours". Jacques Chassin de Kergommeaux CAP-GEMINI-Innovation 7, Chemin du Vieux Che^ne, ZIRST, F-38240 Meylan, France. Tel: (+33) 76 76 47 55 Fax: (+33) 76 41 06 29 E-mail: chassin@capsogeti.fr