Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!network.ucsd.edu!news.service.uci.edu!ucivax!news.claremont.edu!uunet!olivea!pagesat!news.cerf.net!nic.cerf.net!speakez From: speakez@nic.cerf.net (Robert Crowe) Newsgroups: comp.os.386bsd.questions Subject: Re: usefulness of memory-cache in multitasking systems Date: 1 Aug 1993 03:49:36 GMT Organization: CERFnet Dial n' CERF Customer Lines: 21 Message-ID: <23fegg$fuk@news.cerf.net> References: <AgKP52600WB7MsjVgA@andrew.cmu.edu> NNTP-Posting-Host: nic.cerf.net In article <AgKP52600WB7MsjVgA@andrew.cmu.edu> "Alex R.N. Wetmore" <aw2t+@andrew.cmu.edu> writes: >I've been sort of wondering how useful a cache would be in a >multitasking systems (specifically NetBSD). It seems that every time >there was a process swap all of a sudden there would be a different >piece of code that would be best off in the cache. I suppose that there >are parts of the kernel that will always do better off in the cache >(like the scheduler), but does it make a huge difference in system >performance. > I've not actually counted the average number of instructions between task switching, but I'm sure it's pretty high (> 10,000?). That leaves an awfull lot of instructions to make use of the cache. Empirically, I have definately noticed a difference. Just ask the people out there with faulty cache controllers! > >alex > Bob Crowe. bob@speakez.com