Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ncar!husc6!m2c!wpi!lfoard From: lfoard@wpi.wpi.edu (Lawrence C Foard) Newsgroups: comp.arch Subject: Speed Message-ID: <2187@wpi.wpi.edu> Date: 3 May 89 17:41:35 GMT References: <13089@paris.ics.uci.edu> <150@dg.dg.com> Reply-To: lfoard@wpi.wpi.edu (Lawrence C Foard) Distribution: comp Organization: Worcester Polytechnic Institute, Worcester, MA. USA Lines: 25 Why isn't parallel computing used more? I can't think of many problems that couldn't be done in parallel if the hardware/software to do it was done well. We have an Encore here at WPI and although most programs don't make use of the multiple processors directly simple things like being able to do parallel make greatly increase the speed. When you double a processors speed it will cost much much more (depending how close you are to the current limit), but doubling the number of processors only doubles the cost. For example 30 ARM chips will cost around $1000 and have a total of around 200MIPS! With some clever programming it should be possible to make use of this, for example you could have the ultimate game machine. 3D hidden surfaces using Z buffering can be done mostly in Parallel. Imagine flight simulator with real shaded 3D surfaces. Any numeric problem where a cell only depends on its neighbors can also easily be broken up, for example solving Laplaces equation. Or even better drawing Mandelbrot sets. Since most applications that need speed can be made parallel it would seem logical to spend more effort finding cheap ways to implement parallel processing rather than trying to squeezy every last drop out of a single processor. P.S. What important (time consuming) problems can't make use of parallel processing. -- Disclaimer: My school does not share my views about FORTRAN. FORTRAN does not share my views about my school.