[Openmcl-devel] random statistical properties and period?

R. Matthew Emerson rme at clozure.com
Wed May 6 17:26:35 PDT 2009


On May 6, 2009, at 6:15 PM, Raffael Cavallaro wrote:

> What are the statistical properties and period of CCL's implementation
> of random?

I don't know the answer to this question directly, but CCL's  
implementation of RANDOM is based on the "minimal standard" generator  
as described in Park and Miller's CACM paper:

http://www.firstpr.com.au/dsp/rand31/p1192-park.pdf

CCL uses a multiplier of 48271 (which Park and Miller recommend in http://www.firstpr.com.au/dsp/rand31/p105-crawford.pdf) 
  instead of 16807.

One thing of note (that might be considered a bug) is that the 32-bit  
and 64-bit ports don't produce the same sequence of numbers.    
However, the 32-bit ports should produce results that agree with each  
other; likewise for the 64-bit ports.





More information about the Openmcl-devel mailing list