Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!gem.mps.ohio-state.edu!brutus.cs.uiuc.edu!lll-winken!vette!brooks From: brooks@vette.llnl.gov (Eugene Brooks) Newsgroups: comp.arch Subject: Re: 1000000x1000000 Matrix (was: linpack) Message-ID: <36621@lll-winken.LLNL.GOV> Date: 23 Oct 89 20:44:02 GMT References: <9089@batcomputer.tn.cornell.edu> <46500082@uxe.cso.uiuc.edu> <9118@batcomputer.tn.cornell.edu> <36553@lll-winken.LLNL.GOV> <9123@batcomputer.tn.cornell.edu> Sender: usenet@lll-winken.LLNL.GOV Reply-To: brooks@maddog.llnl.gov (Eugene Brooks) Organization: Lawrence Livermore National Laboratory Lines: 56 In article <9123@batcomputer.tn.cornell.edu> kahn@batcomputer.tn.cornell.edu (Shahin Kahn) writes: >> Their real nightmares have KILLER MICROS in them. >YOUR nightmares, maybe. I have pleasant dreams about KILLER MICROS. Let me give you a quote from an engineer at CRI, I will withhold the name to protect the innocent. "We have lost the scalar performance battle to the microprocessor... We will never recover it." Clearly, you are not worried, but the fellow from CRI is quite worried. >All I asked was to request everyone to stop using the >single processor XMP as the definition. Comparing CPU to CPU, the YMP is only 30% faster than the XMP. Twice as many are supplied in a box. There is no significant difference between their performances on scalar code that won't be completely overrun by the next wave of killer micros in 6 months. Any crap about 8 YMP CPUS being lots faster than one XMP CPU will only be answered by having to deal with 1024 KILLER MICROS. >> FOR MY APPLICATION > >BINGO! For Y O U R application! That's the whole point. >Funny we should start from a general red-alert about the attack of >killerr micros and end up where we all knew about: "It depends on >the application". Whether or not the current KILLER MICROS actually OUTRUN current supercomputers is application dependent. It is not, however, a set of measure zero and the set of such applications is growing. KILLER MICROS are more cost effective for ALL APPLICATIONS, however, with the supercomputers only getting close to BREAK EVEN on codes which run both the adder and multiplier on every cycle. Highly parallel machines also applicable to these embarrarsingly parallel problems and the KILLER MICRO powered multiprocessors will provide both more total horsepower and more total main memory. > >Yes, if your application is "pagemaker" you redefine supercomputing >by going to a mac-ii-ci !! I see no need to attempt an insult, my application is not "pagemaker". It is getting difficult to carry on a sensible discussion with such stuff leaking in. I will not reply to postings containing this sort of abusive retoric in the future. >But it's too early for the supers to pack up and go home. WAY too early. Supers have been sent home with their tails between their legs for scalar code by the KILLER MICROS. It will only be a matter 5 years before the same thing happens for vectorized code. We probably won't call them micros, out of respect for the dearly departed, but micros is what they will be. brooks@maddog.llnl.gov, brooks@maddog.uucp