Path: utzoo!mnetor!uunet!husc6!rutgers!rochester!PT.CS.CMU.EDU!K.GP.CS.CMU.EDU!lindsay
From: lindsay@K.GP.CS.CMU.EDU (Donald Lindsay)
Newsgroups: comp.arch
Subject: Re: Single tasking the wave of the future?
Message-ID: <511@PT.CS.CMU.EDU>
Date: 14 Dec 87 17:35:33 GMT
References: <201@PT.CS.CMU.EDU> <388@sdcjove.CAM.UNISYS.COM> <988@edge.UUCP> <1227@sugar.UUCP> <151@sdeggo.UUCP> <1423@cuuxb.ATT.COM> <439@xyzzy.UUCP> <440@xyzzy.UUCP> <36083@sun.uucp> <18@amelia.nas.nasa.gov> <2341@encore.UUCP>
Sender: netnews@PT.CS.CMU.EDU
Organization: Carnegie-Mellon University, CS/RI
Lines: 17


Multitasking gives flexibility, which is considered virtuous.

Interestingly, there are cases when you get a performance win by simulating
a multiprocessor on a uniprocessor. For example, suppose you are searching a
big tree of choices while solving a Travelling Salesman type problem.  This
has the nice property that there is variance in the individual compute
times. So, if you have a (simulated) multiprocessor, some of the
computations finish earlier than their peers. This gives the overall search
new data (about lower bounds) which may allow you to kill some of the
computations that haven't finished yet. This can save work, not only
directly, but also by preventing the killed searches from generating
offspring. 


-- 
	Don		lindsay@k.gp.cs.cmu.edu    CMU Computer Science