Random Number Package

This package has functors for generating random deviates with uniform, discrete, finite, exponential, gamma, normal, and Poisson distributions. The code for measuring the performance of the algorithms is in stlib/performance/numerical/random. There are makefiles for running the performance code and gnuplot scripts for generating graphs in the stlib/results/numerical/random directory. For the performance results presented herein, the testing code was compiled with GNU g++ 4.0 using the flags: -O3 -funroll-loops -fstrict-aliasing. I ran the tests on a Mac Mini with a 1.66 GHz Intel Core Duo processor and 512 MB DDR2 SDRAM.

Some of the generators use the GNU Scientific Library. (Namely the the ones with a "Gsl" suffix.) To use these, you must have GSL installed on your system. Also, you should define appropriate environment variables so the compiler can find the GSL header files and libraries. For example, with the GNU compiler define CPATH to the include directory and LIBRARY_PATH to the library directory. If you do not use these generators, you don't need to have GSL installed.

Links:

• Mersenne Twister Home Page
• http://random.mat.sbg.ac.at/