Now that generations are competitive, each one is taking around 3-4 hours to complete on an AMD tri-core PC. I looked into CUDA to see if it would increase the speed and reduce generation time.
CUDA is Compute Unified Device Architecture. It allows tasks to be passed to the graphics card so that the GPU's ability to run massively parallel tasks could be utilised.
This would have been ideal for the task of playing many games of poker, the downside is that all work would have to be written in and compiled in a special language for the graphics card to run. I have already written alot of code in Java and didn't like the idea of replicating it.
So although CUDA seems great, I don't think it would be ideal for this task.