Date: Wed, 29 Aug 2001 10:38:54 -0700 (PDT) From: Matt Dillon <dillon@earth.backplane.com> To: Peter Wemm <peter@wemm.org> Cc: Brooks Davis <brooks@one-eyed-alien.net>, Julian Elischer <julian@elischer.org>, Trent Nelson <nelsont@wa.switch.aust.com>, freebsd-arch@FreeBSD.ORG Subject: Re: Paper detailing blocking heuristics based on Scheduler Activiations. Message-ID: <200108291738.f7THcsT47493@earth.backplane.com> References: <20010829122003.A2A7C3906@overcee.netplex.com.au>
next in thread | previous in thread | raw e-mail | index | archive | help
:Going direct worked for me: : :fetch ftp://ftp.cs.rochester.edu/pub/papers/systems/93.PPoPP.Using_scheduler_information_barrier_synchr_performance.ps.Z : :Thanks.. Ok, I printed out and read this paper. I think there are some serious flaws in the paper. The assumptions being made are so specialized (such as assuming that the 'work' section for each thread between barrier points is about the same length), that I don't think the results can be applied generally. Also, the test programs appear to be very simple and thus probably fairly small - which means that there are major assumptions being made in regards to cache effects when spinning verses blocking. It seems to devolve down into essentially non-preemptive cooperative scheduling, and I think the last two graphs pretty much prove that. -Matt To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200108291738.f7THcsT47493>