Boost logo

Boost :

From: troyer (troyer_at_[hidden])
Date: 2000-05-29 03:12:32


Jens Maurer wrote:
>
> I would like to ask for a formal review of my random number
> library as per the new library submission guidelines.

I do not know about the formal review process but since I am
extensively using random numbers (1000 CPU-years of Monte Carlo
simulations in the past years) I want to add my comments.

>
> The library is available in the egroups "files" section
> at http://www.egroups.com/files/boost/random/random.zip

I was glad to see that someone has spent the time to implement
such a library - a task that I thought about a lot but always
delayed for a later time. Just today I wanted to finally start
working on it more intensely and was so glad to see that someone
has done most of the work already!

While I have not looked in detail at all parts of the library yet
I found some problems that need to be discussed

1.) problems with exponential random number generator
2.) incorrect spherical random numbers
3.) problems with floating-point generators
4.) performance issues of the spherical random number generator

Furthermore I suggest

4.) adding fast lagged Fibonacci generators
5.) providing wrapper classes that allow compile time polymorphism

Before elaborating on these points however I want to inquire
if this is the right forum to discuss these points, or if the
discussion should be done privately or on another list.

Matthias Troyer


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk