Building the ruleset takes very long and i want to run multiple solver per core, because with only one solver the core has only 70% load. Building the ruleset multiple times parallel increases the prepare time alot. Am 23.06.2010 15:39, schrieb Wit Jakuczun: > W dniu 2010-06-23 15:11, Christian Wirth pisze: >> I think i found the error now, but i don't know how to fix it. >> The problem is, as mentioned, that i can't do the preparation for every >> solver because they take too long. > Why? It should take exactly the same time as for one solver if done in > parallel. As I understood you correctly, queries to the solver are > completely independent. So the workflow should be like this: > 1) In parallel build ruleset for each solver > 2) Send queries independently to each solver as necessary > > The other solution whould be to implement parallel solver on ECLiPSe > side but I do not know whether ECLiPSe support this... Neither I have > done this. > >> I'm currently copying the information >> via java, but that destroys the variable information, and they >> have to >> be preserved ? How to copy data in complete ? > If you have some free (or clp) variables then it is not possible to > copy them. > > Best regards >Received on Wed Jun 23 2010 - 14:45:41 CEST
This archive was generated by hypermail 2.2.0 : Thu Feb 02 2012 - 02:31:58 CET