Xref: utzoo sci.math:5824 comp.sources.wanted:6527
Path: utzoo!attcan!uunet!lll-winken!ames!mailrus!uflorida!gatech!gitpyr!news
From: news@pyr.gatech.EDU (News Administrator)
Newsgroups: sci.math,comp.sources.wanted
Subject: Re: random numbers
Summary: Fast Normal Genarators.
Keywords: normal distribution, Gaussian
Message-ID: <7379@pyr.gatech.EDU>
Date: 25 Feb 89 20:52:10 GMT
References: <10471@ut-emx.UUCP> <734@fcs280s.ncifcrf.gov> <44881@linus.UUCP>
Reply-To: naras@stat.fsu.edu (B. Narasimhan)
Distribution: na
Organization: Dept. of Statistics, Florida State Univ.
Lines: 12

The algorithm gains its speed by having a fast part 99% of the time (a multiplication and a comparison) and a slow part 1% of the time. This algorithm is not suited to vector machines though because of the condition checks. In fact any rejection method suffers from the same problem. I have coded this algorithm bot in Fortran and 8086 assembler and I'll mail them upon request. One gains a very good knowledge about the cost of functions like Log etc, even when you have a co-processor by using the probability







 integral transformation to generate an exponetial and then generating the same by the above method.
B. Narasimhan e-mail: naras@stat.fsu.edu on internet.
From: naras@stat.uucp (B. Narasimhan)
Path: stat!naras


-- 
News Administrator -- Don Deal
Georgia Institute of Technology, Atlanta GA 30332
...!{akgua,allegra,amd,hplabs,ihnp4,masscomp,ut-ngp}!gatech!gitpyr!news