Date: Sat, 14 May 2005 15:08:25 +0200 From: des@des.no (=?iso-8859-1?q?Dag-Erling_Sm=F8rgrav?=) To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: "Drew B. \[Security Expertise/Freelance Security research\]." <d4rkstorm@gmail.com> Subject: Re: FreeBSD Security Advisory FreeBSD-SA-05:09.htt [REVISED] Message-ID: <86fywqdkfq.fsf@xps.des.no> In-Reply-To: <94145.1116037219@critter.freebsd.dk> References: <94145.1116037219@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
"Poul-Henning Kamp" <phk@phk.freebsd.dk> writes: > The correct (technical) workaround (IMO) is to restrict HTT to be > used only for threads from the same process. > > The political problem is that if all operating systems do that, > Intel has a pretty dud feature on their hands, and they are not > particularly eager to accept that fact. No. Intel themselves recommend only scheduling threads from the same address space on the same physical core, simply because it improves cache efficiency. Scheduling threads from separate processes on the same physical core can actually *reduce* performance due to cache churn. Fixing our scheduler to avoid scheduling threads from different processes on the same physical core is likely to increase performance across the board, because heavily multithreaded software will get a bigger boost and single-threaded software will be penalized less than it is today. DES --=20 Dag-Erling Sm=F8rgrav - des@des.no
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?86fywqdkfq.fsf>