Hi Johannes, I'm very curious to what has gone amiss on your system. What Linux distro are you using? Did you build the OpenThreads/OSG yourself?
Robert. On Thu, Oct 28, 2010 at 7:31 PM, Johannes Scholz <[email protected]> wrote: > Hi, > > we are currently working on a routine that deforms large vertex arrays on the > fly. To speed this up we implemented multithreading using OpenThreads. > > Simple example: > > We have a point cloud of e.g. 10 million points. The calculation is done on > e.g. 4 processors. > The first thread is assigned to processor 0 using setProcessorAffinity(0) and > works on the first 2.5 million points. On the second thread > setProcessorAffinity(1) is applied. The other points are equivalently handled > by processors 2 and 3. Performance scales by about 90% per doubled processor > count. > > This works fine on WINDOWS 32-Bit. > > When testing this on 64bit Linux we noticed that setProcessorAffinity returns > -1. Performance does not scale at all. All operations seem to be executed on > one CPU. > I already learned that using pthreads on linux setProcessorAffinity won't > work at all? The operation system does not seem to automatically apply our > threads to different cpus. > > So what can you suggest to distribute our calculations to multiple cpus on > LINUX? > > > Thank you! > > Cheers, > Johannes > > ------------------ > Read this topic online here: > http://forum.openscenegraph.org/viewtopic.php?p=33205#33205 > > > > > > _______________________________________________ > osg-users mailing list > [email protected] > http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org > _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
