Date: Tue, 15 Nov 2011 13:45:10 -0800 From: Garrett Cooper <yanegomi@gmail.com> To: Hans Petter Selasky <hselasky@c2i.net> Cc: "svn-src-head@freebsd.org" <svn-src-head@freebsd.org>, mdf@freebsd.org, "svn-src-all@freebsd.org" <svn-src-all@freebsd.org>, "src-committers@freebsd.org" <src-committers@freebsd.org> Subject: Re: svn commit: r227541 - head/sys/dev/usb/controller Message-ID: <1A1FDBC6-90A1-4C4A-B2C6-62B40FD2B523@gmail.com> In-Reply-To: <201111152224.27868.hselasky@c2i.net> References: <201111152048.pAFKmvNC016452@svn.freebsd.org> <CAMBSHm8HXCfF3yVp9fFwAAThL6Aaki=iuV7-Qtz3g7tpsm-VLw@mail.gmail.com> <CAMBSHm9%2Bhsa0fvVvdLkFZkMoAQ6vkspWpvsfV1%2BJPtJe72J9cQ@mail.gmail.com> <201111152224.27868.hselasky@c2i.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Nov 15, 2011, at 1:24 PM, Hans Petter Selasky wrote: >>=20 >> ... and I also just remembered that I have seen recommendations that, >> when FreeBSD is used as a virtual machine, hz should be set to 100 so >> that the virtual interrupt overhead is reduced. Those two >> recommendations are at odds with each other. >=20 > USB in virtual machines is a completely different story. All the = timing=20 > critical stuff is handled by the real USB host operating system. The = OS inside=20 > the VM just has to respond to some "virtual" USB events. Timing does = not=20 kern.hz=3D100 (!=3D 1000) recommendation matters because: 1. It causes too many interrupts on multicore systems, which folks = reported caused performance degradation. 2. In VMs it caused clock skew IIRC. I could dig up sources if you need them. Thanks, -Garrett=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1A1FDBC6-90A1-4C4A-B2C6-62B40FD2B523>