MathGroup Archive 1996

[Date Index] [Thread Index] [Author Index]

Search the Archive

linear congruential generator

  • To: mathgroup at
  • Subject: [mg4770] linear congruential generator
  • From: cabana at (David Cabana)
  • Date: Fri, 13 Sep 1996 13:54:52 -0400
  • Organization: Florida Internet Corp
  • Sender: owner-wri-mathgroup at

This is in response to a recent posting regarding (among other things), a
linear congrential generator.  Here is what I got when I ran the proposed

x[0] = 1;
x[n+1] := Mod[(7^7*x[n]), 2^31 -1]//N
u[n] := x[n]/(2^31-1)
Table[x[n], {n,1,10}]

{x[1], x[2], x[3], x[4], x[5], x[6], x[7], x[8], x[9], x[10]} 

Here is roughly the same code, with some changes to make it work. 

x[0] = 1;
x[n_] := Mod[(7^7 * x[n-1]), 2^31 - 1]
u[n_] := x[n]/(2^31-1) //N
Table[x[n], {n,1,10}]

Notice the changes in the definitions of x and u. The underscores are
critical. I also expressed the recursion a bit differently.  The generator
must use exact integer arithmetic to be correct, so I moved the //N to the
definition of u.

There is another problem. As written, this program is very slow. The
reason is that to compute  x[100] it must compute x[99], x[98], ..., x[1].
Then to compute x[101], it must compute x[100], x[99], x[98], ... x[1]. It
computes the same thing over and over again.  There are various fixes to
this. One is to use memoization, say by writing 
   x[n_] := x[n] = Mod[(7^7 * x[n-1]), 2^31 - 1]
instead of 
   x[n_] := Mod[(7^7 * x[n-1]), 2^31 - 1]

Another is to take an entirely different approach:
generator[n_]:= Mod[7^7 * n, 2^31 - 1]
uniformVariates[n_, seed_] := 
   NestList[generator, seed, n] / (2^31 - 1) //N

I did not run timing tests, but I suspect the 2nd approach is quicker.

David Cabana   cabana at


  • Prev by Date: Graphic for Conic Sections
  • Next by Date: functional code
  • Previous by thread: Graphic for Conic Sections
  • Next by thread: functional code