CUDA forum reorganization

Ok ok we all know you just started CUDA, but I think the CUDA forums need some reorganization before they start to be flooded!

For example:

  1. CUDA announcements and news ( and allow people to comment news, don’t lock it like is atm! )

  2. CUDA general discussion ( post here themes non related to the other sections )

  3. CUDA specific implementation and drivers
    3.1 Windows
    3.2 Linux
    3.3 MacOS
    3.4 Solaris
    3.5 blah blah

  4. CUDA programming
    4.1 C SDK
    4.2 Java SDK
    4.3 .NET SDK
    4.4 Interface and theorical design discussion

  5. CUDA bug submit ( do something like Microsoft’s connect… you post a bug and you can track it, others can comment, etc )

  6. CUDA suggestions and feedback

  7. Applications using CUDA ( you could post your application info here. Something like… here is my new raytracer renderer using CUDA or here is my protein folding using cuda… )

I tried this simple test because I have problem which need around 100,000 random number for reach thread for one run. Then I found if I use more threads, around 10000 #threads*#blocks, it will crash.
I have tried different random number generator and several different implementations, but it always crashes when the #threads*#blocks reaches some not big number, sO I am wondering if there are some problem with the random number gererator code or the GPU will fail with too much calculations for each thread if #thread*#bloack reach some not big number.

THanks a lot for any comments!