a simple Particle Swarm from the Mersenne Twister example - suggestions for performance improvement?

Accelerated Computing CUDA CUDA Programming and Performance
1

I have programmed a simple PSO search in 3 dimensions. (code :[attachment=23580:pso9.zip])
At every iteration, I am updating particles by using a modified version of the RandomGPU from Mercenne twister SDK example, I get the following performance compared to normal c code (see jpeg file for performance detail at [attachment=23581:pso_performance.jpg]) : around 10 times faster.
I am now wondering what could be done to improve on that…
Any suggestions?

2

Need to post details about your system etc. Try to use rng routines from last sdk. They may be faster.

3

Need to post details about your system etc. Try to use rng routines from last sdk. They may be faster.