Hi, Martin Wegner wrote: > Hello, > > we, three students, want to write a multithreading Java program that > uses eclipse prolog. > > From the universirty we can use a server with four CPUs and we want to > use them. > > It is only possible to create one EmbeddedEclipse object and when four > threads use this object for prolog commands only one processor is used > by the JVM. > > > So we tried OutOfProcessEclipse and every thread has now its own > OutOfProcessEclipse object. So all four CPUs can be used by the JVM but > we had a big problem with this. In some cases the threads wants to > execute a lot of small prolog command so that we had a very huge > communication overhead which kills our performance under a single core > solution :( > > Is it possible that an EmbeddedEclipse object can use more than one > processor with the JVM? > > If I understand you correctly, you want to run ECLiPSe in parallel -- ECLiPSe is currently single threaded, and does not support concurrent or parallel execution, so it does not make sense to run an ECLiPSe process on more than one processor. With OutOfProcessEclipse, each ECLiPSe process is independent of the others that you run, so this is probably not what you want. Are you saying that when you have a single embedded ECLiPSe process, you can't run the Java threads on more than one processor? You obviously can't run the ECLiPSe on more than one process, but does this also limit the other Java threads? ECLiPSe did have support for some form of or-parallelism (finding the multiple branches of the search tree in parallel) many years ago, and the code is still in the source, but I don't think it has been tested in recent years, and features developed since (such as the Java interface), does not really take this parallel execution into account and probably will not work with it. > Is it possible to "tune" OutOfProcessEclipse to minimize the > communication overhead with some eclipse option tweaks? > > Thanks a lot! > > I think the main overhead with using out-of-process (or remote) ECLiPSe is the use of the sockets for transmitting the data between ECLiPSe and Java. There seems to be some (rather large) minimum time to send anything through the socket. So it is quite expensive if you transmit only small amounts of data, e.g via the queues. You can reduce this overhead somewhat by transmitting larger amounts of data each time. Howver, as I said, I am not sure if this is what you want, as each out-of-process ECLiPSe you set up is not connected to the others. Cheers, Kish > Greetings! > > Martin > > ------------------------------------------------------------------------- > Check out the new SourceForge.net Marketplace. > It's the best place to buy or sell services for > just about anything Open Source. > http://sourceforge.net/services/buy/index.php > _______________________________________________ > ECLiPSe-CLP-Users mailing list > ECLiPSe-CLP-Users_at_lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/eclipse-clp-users >Received on Tue Jun 10 2008 - 06:02:30 CEST
This archive was generated by hypermail 2.3.0 : Wed Sep 25 2024 - 15:13:20 CEST