Skip to content

Latest commit

 

History

History
76 lines (50 loc) · 2.41 KB

README.md

File metadata and controls

76 lines (50 loc) · 2.41 KB
Raw

Knuth Fibonacci Lagged RNG

This folder contains the Fortran Fibonacci Lagged RNGs from Section 3.6 of Volume 2 of TAOCP (Semi-numerical Algorithms); slightly reformatted and changed (sorry Donald) from the versions downloadable from the code section of TAOCP site.

  • The KFLRNGD sub-folder contains the DOUBLE PRECISION version (producing numbers between 0.0 and 1.0).
  • The KFLRNGI sub-folder contains the INTEGER version.

Both versions exist in Fortran 77 (the original versions) and Fortran IV (66) (you can ask me why in private).

All versions have been revised to provide an API to make them readily usable in any code, as the C version available on the TAOCP site is.

Of course, all of this is for academic and hobbyist's purposes, as modern Fortran implementations (not to speak of other languages) already provide their own RNGs (e.g., check out the excellent xoshiro).

Yes, this is a bit of a retro computing hacking.

Building

To build the code and test it you can use cmake, make, or nmake; the CMakeLists.txt, unix.make, and windows.make can be used for each platform; note that the -build sub-directory is empty if you use cmake.

gfortran is assumed to be available.

The build tools will build the kflrng[di]-main and kflrng[di]-nrandom exectutables for you to test.

JCL for MVS 3.8j

The subfolders for Fortran IV (66) (named FIV) contain JCL to compile load modules on MVS 3.8j. (Not yet for the KFLRNGI subfolder; but it should be straighforward to reuse the JCL for KFLRNGD).

Usage

You can link in your code the object files obtained from kflrng[di]-rngnxt.f and knuth-frng{i,db}.f; the file rnaprt.f contains a subroutine to print a piece of an array for debugging purposes.

In order to get the next pseudo-random number you invoke the RNDNXI (INTEGER) or the RNDNXD (DOUBLE PRECISION, between 0.0 and 1.0) functions, of course after the proper declarations.

Notes

The various versions (INTEGER, DOUBLE PRECISION, Fortran 77, and Fortran IV) contain code duplications, but they are self-contained.

I am not much of a Fortran programmer; apart from Knuth's code, which must be left essentially untouched, all the rest can be modified, and I welcome suggestions.

Enjoy

2024-01-04, Athens, Greece

Enjoy