From: hjstein@bfr.co.il (Harvey J. Stein) Newsgroups: comp.lang.fortran,comp.parallel.pvm Subject: Re: fortran77 programming Date: 04 Dec 1998 11:14:01 +0200 Organization: Unspecified Organization Message-Id: References: <3660D892.E1EDBF12@est.it> <3665044D.A0F57E59@kings.uq.edu.au> <3665E491.7993355@flash.net> Cc: hjstein@bfr.co.il Xref: ukc comp.lang.fortran:61810 comp.parallel.pvm:7824 Craig Burley writes: > Very loosely (since comp.lang.fortran isn't the place for detailed > examination of the arguments), any software system S split into > threads T1 through Tn run on parallel system MP can be at least as > efficiently run on an equivalent-hardware uniprocessor-system SP by > modifying the operating environment for S to emulate parallel > execution of T1 through Tn using timesharing. And since SP needn't > arbitrate for access to shared memories as MP must (at *some* > level), that emulation isn't necessarily going to be slower, though > it might be faster. > > In this case, the real-time-critical thread needs only half of the > single-CPU's time to be serviced according to its scheduling needs > (else it couldn't meet those needs on the dual-CPU system). The > other half of the time can be used to service the timesharing > system generally, and the "slower process" specifically. This is not the case because of context switch overhead. -- Harvey J. Stein BFM Financial Research hjstein@bfr.co.il