You should have a look to the SPF / Dijkstra compute from OSPF. It is a good fit for GPUs.
My 2 cents, Le 21 déc. 2015 19:23, "Donald Sharp" <sha...@cumulusnetworks.com> a écrit : > They are in table.c: route_node_match and route_node_lookup. I suspect > there will need to be some fairly serious rewrite of the code in order to > take advantage of GPU's. > > The reality though is that I don't think these are the current bottlenecks > in handling large #'s of routes and peers in the code. I would point you > towards perf as a starting point here. > > donald > > On Mon, Dec 21, 2015 at 3:18 AM, M.Saeed Ansari <ansari.r...@gmail.com> > wrote: > >> Dear Friends, >> >> I want to do some of CPU consumer task of Quagga with GPU. >> I need to know where are longest prefix match and other data-flow task >> functions and how can i change their code? >> I want to mix OpenCL with Quagga code to do tasks with GPU in this >> project. >> >> Regards, >> Saeed >> >> _______________________________________________ >> Quagga-dev mailing list >> Quagga-dev@lists.quagga.net >> https://lists.quagga.net/mailman/listinfo/quagga-dev >> > > > _______________________________________________ > Quagga-dev mailing list > Quagga-dev@lists.quagga.net > https://lists.quagga.net/mailman/listinfo/quagga-dev >
_______________________________________________ Quagga-dev mailing list Quagga-dev@lists.quagga.net https://lists.quagga.net/mailman/listinfo/quagga-dev