Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ncar!husc6!ogccse!johnr From: johnr@ogccse.ogc.edu (John Roberts) Newsgroups: comp.arch Subject: Re: Speed Summary: Amdahl's Law & parallelism Message-ID: <2566@ogccse.ogc.edu> Date: 5 May 89 07:09:19 GMT References: <13089@paris.ics.uci.edu> <2187@wpi.wpi.edu> Distribution: comp Organization: Oregon Graduate Center, Beaverton, OR Lines: 28 In article <2187@wpi.wpi.edu>, lfoard@wpi.wpi.edu (Lawrence C Foard) writes: > > Why isn't parallel computing used more? I can't think of many problems... Most people quote Amdahl's Law which essentially says that some portion of a program is sequential in nature and the amount of time it takes to execute that sequential code is the lower limit on the amount of time it will take the program to execute. Other's have disagreed, saying that you would rewrite the program (eliminating the sequential code) for parallel systems. This lead to Mohler's law which said that you'd scale up the parallelizable part of the program and solve larger problems. There are real parallel systems out there. The Cogent XTM system from Cogent Research is one example (of a distributed memory system). For around $20K you get two Inmos T800 transputers each with 4 Mbytes of RAM, a 90 Mbyte hard drive, an 800K floppy drive and a 1024 by 808 by 256 grey scale monitor. However the most important thing about the system is that it is scalable! You can attach a "resource server" with more processors (and memory). The operating system has built into it a communication system called "Linda" which makes parallel programming much easier than other parallel machines. The operating system is highly compatible with Unix System V. The future of computing lies in parallel programming. John Roberts johnr@ogccse.ogc.edu