Hi all > Using a thread to receive listed for input, and another thread to work > on reducing the tree, then this should work nicely. Perhaps a third > thread to send out messages produced by the tree reducer... > > Tomas, I guess this design is similar to the one you though of when we > discussed alternatives for the SIMAP runtime?
Yes, that was very much my idea. Drop rounds and split communication out in one or two threads, so that communication and computation could be parallelized. I'm not sure if the above is the best idea, but implementing rounds explicitly means "wasting" potential communication time on processing (and vice versa), which is bad. Of course this means a new proof of security (how is that coming along?). /Tomas _______________________________________________ viff-devel mailing list (http://viff.dk/) viff-devel@viff.dk http://lists.viff.dk/listinfo.cgi/viff-devel-viff.dk