On Tue, Mar 11, 2014 at 6:23 PM, Antti Kantee <[email protected]> wrote: >> Is there a way to run the driver (or similar) under native NetBSD, >> still curious how it compares. > > Do you mean "native" as in "does a syscall to the host kernel"? I guess > as a quick hack you could take if_virt and comment out the hypercalls. > Sending should work exactly the same, netstat -bi might even tell you > how much was sent. > > Though, if you have 10GigE, you might just as well stuff things to the > NIC. If we assume a user packet framework can send in less than 100 > cycles, sending from within the kernel can't be much more expensive (if > at all). > > Those would be very interesting numbers (and hopefully not very > depressing ones ;)
I meant actually on NetBSD vs rump. Clearly it should be faster in theory, but would be good to test if so or there is other overhead that can be removed. >> A buildrump.sh option to compile locks_up would be useful I guess, or >> runtime switch. > > Yea, it might be. I've just never gotten around figuring out the > performance and size impact of runtime selection. A compile time easy config option would be just as good. It depends how it selects at runtime what the overhead would be. Justin ------------------------------------------------------------------------------ Learn Graph Databases - Download FREE O'Reilly Book "Graph Databases" is the definitive new guide to graph databases and their applications. Written by three acclaimed leaders in the field, this first edition is now available. Download your free book today! http://p.sf.net/sfu/13534_NeoTech _______________________________________________ rumpkernel-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/rumpkernel-users
