Date: Fri, 27 Nov 2009 07:55:09 +0200 From: Manolis Kiagias <sonicy@otenet.gr> To: questions@freebsd.org Subject: Re: HTT on Atom (Was: FreeBSD 8.0 and Atheros AzureWave wireless chipset) Message-ID: <4B0F69BD.9050503@otenet.gr> In-Reply-To: <200911270218.TAA17345@lariat.net> References: <200911260629.XAA08100@lariat.net> <alpine.BSF.2.00.0911260822290.63225@wonkity.com> <200911262320.QAA16364@lariat.net> <4B0F103D.2070607@otenet.gr> <200911270218.TAA17345@lariat.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Brett Glass wrote: > At 04:33 PM 11/26/2009, Manolis Kiagias wrote: > >> Though it seems hyperthreading is improved on the Atom and there is no >> penalty for leaving it on. > > Is there really no penalty? With HZ=1000 there are double the clock > interrupts to be serviced at least. And as I understand it the Atom > has less redundant hardware, so there are less likely to be unused > resources available to the second thread. I am seeing substantially > faster compiles with the SMP option commented out of the kernel. > > --Brett > My tests involved building a custom kernel - I never tried without SMP, just without hyperthreading and there was no appreciable difference. Using -j3 in make kernel, the kernel is built in just about 40 minutes. Without -j same procedure lasts 55 minutes. (I am using an Atom 330 which is dual core) On a Pentium 4 with HTT, -j actually results in a somewhat slower build.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4B0F69BD.9050503>