On Fri, Sep 19, 2008 at 11:27 PM, Robert Iakobashvili <[EMAIL PROTECTED]> wrote: >> It occurs to me that libevent could reduce the number >> of such system calls transparently by using >> EPOLLET and caching readiness status in userspace. > > This may be particularly useful, when > working on demultiplexing of tens and hundred thousand of sockets > with even more events, like e.g. curl-loader or ngnix are doing. > > Have you tried to benchmark it and see the gains?
Nope, haven't had time. I do think it's worth trying. Of course, apps that can are better off using EV_ET explicitly; then there's even less wasted motion. (My proposal reduces unneeded libevent <-> kernel interaction, but EV_ET also reduces unneeded app <-> libevent interaction. The problem with EV_ET is that it requires apps to be aware of it.) - Dan _______________________________________________ Libevent-users mailing list Libevent-users@monkey.org http://monkeymail.org/mailman/listinfo/libevent-users
