[R] simple questions about R
edd at debian.org
Thu Nov 5 02:24:04 CET 1998
Following up on the RNG discussions, I suggest the (GPL'ed) C code for the
'Mersenne Twister'. This RNG has a couple of quite impressibe features, see
the quote below. The quote is from a second implementation by Shawn Cokus
which adds even more speed. The code is available via the quoted URL. The
algorithm was published in one of the ACM journals early this year.
// This is the ``Mersenne Twister'' random number generator MT19937, which
// generates pseudorandom integers uniformly distributed in 0..(2^32 - 1)
// starting from any odd seed in 0..(2^32 - 1). This version is a recode
// by Shawn Cokus (Cokus at math.washington.edu) on March 8, 1998 of a version by
// Takuji Nishimura (who had suggestions from Topher Cooper and Marc Rieffel in
// July-August 1997).
// According to the URL <http://www.math.keio.ac.jp/~matumoto/emt.html>
// (and paraphrasing a bit in places), the Mersenne Twister is ``designed
// with consideration of the flaws of various existing generators,'' has
// a period of 2^19937 - 1, gives a sequence that is 623-dimensionally
// equidistributed, and ``has passed many stringent tests, including the
// die-hard test of G. Marsaglia and the load test of P. Hellekalek and
// S. Wegenkittl.'' It is efficient in memory usage (typically using 2506
// to 5012 bytes of static data, depending on data type sizes, and the code
// is quite short as well). It generates random numbers in batches of 624
// at a time, so the caching and pipelining of modern systems is exploited.
// It is also divide- and mod-free.
Linux is not only free; it is, arguably, a better operating system, offering
a degree of stability and an ability to scale up that NT cannot match.
-- The Economist, Oct 3, 1998
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch
More information about the R-help