Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!wuarchive!sdd.hp.com!spool.mu.edu!uunet!cbmvax!cbmehq!babylon!rbabel From: rbabel@babylon.rmt.sub.org (Ralph Babel) Newsgroups: comp.sys.amiga.hardware Subject: Re: Speed Trials: ICD Advantage 2000 Message-ID: <07075.AA07075@babylon.rmt.sub.org> Date: 26 Mar 91 19:29:30 GMT References: <2807@enea.se> <91079.154628IO92257@MAINE.BITNET> <1991Mar25.174202.14540@mintaka.lcs.mit.edu> <07027.AA07027@babylon.rmt.sub.org> Reply-To: cbmvax.commodore.com!cbmehq!babylon!rbabel (Ralph Babel) Lines: 20 In article , ben@epmooch.UUCP (Rev. Ben A. Mesander) writes: >> Remember that ICD's caching algorithm uses deferred writes! >> Just imagine a system crash right before CMD_UPDATE ... > > Yeah, that's the case with UNIX, too though. It should be fine as long as > you don't go snapping your Amiga on and off like a toaster or something. > Let's just say I haven't had much problem with deferred writes on my Sun-3 True. But on a Sun, an illegal pointer (for example) would cause a kernel function to return EFAULT. Maybe you'll get a coredump if Really Bad Things happen. But a KERNEL PANIC or a complete system lockup is much less likely to occur on a UNIX machine than a GURU (or a crash) on an Amiga. I don't see any problems with deferred writes on a system with full MMU protection and separation between kernel and user space, but I wouldn't recommend it for an Amiga. Ralph