February 8th, 2009, 02:09 PM
Random Number Generation
Hello, I'm a physicist working on some numerical stuff. I need a good random number generator package that generates numbers from basic distributions (for my current application I just need uniform and gamma). The catch is I need something with an exceptionally long period because my application may call for very, very large quantities of random numbers (I would probably be doing on the order of thousands of runs with each run involving around a billion random numbers). The simplest, most user friendly thing I've found online is Newran. While it is nicely laid out and easy to install, it admits that its random number generation becomes suspicious around 100 million generations. The question is, what should I use? If there is something known as a "standard" in the C++ numerical community then that would be perfect because this library will probably be used to produce results that will be in an academic journal. Hence a well known library is preferable.
I read that there is something called the Technical Report 1 (or something like that) for C++ which is going to extend the standard library and include lots of very good random number generators, and further it says that there are already implementations of it. In fact, it says that g++ 4 already includes this stuff! (this is the compiler I use, I'm working in Linux). Except I can't seem to find information on it with g++.
I also tried a library called Boost which implements many of the recommended things from Technical Report 1, I played around with the website for a bit but I found the presentation much less clear than Newran and I'm not sure if its what I want or not.
So, any suggestions?
PS I'm in physics, not CS so installing libraries and getting things to link correctly is still a huge pain for me, so a library that is very easy to install/good installation instructions counts for a lot.
Tags for this Thread
Click Here to Expand Forum to Full Width
This a Codeguru.com survey!