Date: Mon, 23 Jul 2007 09:48:56 -0500 (CDT) From: "Sean C. Farley" <scf@FreeBSD.org> To: Eric Anderson <anderson@FreeBSD.org> Cc: attilio@FreeBSD.org, freebsd-current@FreeBSD.org, Harald Schmalzbauer <h.schmalzbauer@omnisec.de> Subject: Re: kqemu crash (page fault) with -current Message-ID: <20070723094249.G9030@thor.farley.org> In-Reply-To: <46982BF7.3010207@freebsd.org> References: <200707131834.27131.h.schmalzbauer@omnisec.de> <4697CCEB.9080707@FreeBSD.org> <200707132155.43783.h.schmalzbauer@omnisec.de> <20070713172842.K26096@thor.farley.org> <46982BF7.3010207@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 13 Jul 2007, Eric Anderson wrote: > On 07/13/07 17:29, Sean C. Farley wrote: >> On Fri, 13 Jul 2007, Harald Schmalzbauer wrote: >> >>> I could install various OS, only when I enable -kernel-kqemu most >>> installer quit with page fault. >> >> As a side note, I found -kernel-kqemu to be slower for me than >> without it on STABLE. > > Using ubench from ports on -CURRENT, I see these kinds of numbers (for > CPU): > > Without any kqemu: 16000 -> 16800 > With kqemu, kernelmode: 180000 -> 182000 > With kqemu, usermode: 187000 -> 188000 > RAW CPU: 202000 -> 203000 Here is mine on -STABLE. It is amusing how much faster the raw CPU really is when you take these benchmark numbers into account. If I remember correctly from trying to find where QEMU was slow, kqemu user + kernel actually has significantly more context switches than just kqemu user. kqemu user ========== Ubench CPU: 74871 Ubench MEM: 8873 -------------------- Ubench AVG: 41872 kqemu user + kernel =================== Ubench CPU: 67768 Ubench MEM: 6520 -------------------- Ubench AVG: 37144 Raw === Ubench CPU: 44162 Ubench MEM: 42655 -------------------- Ubench AVG: 43408 Sean -- scf@FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070723094249.G9030>