Newbie questions How do I get started?

Hi, all.

I am very interested in CUDA, OpenCL, and multiprocessing, particularly on the Mac.

I did parallel processing in the early 1980’s on a CRAY 1S for oil field simulation. Things have certainly changed a lot in that time. I am thinging that a new Mac PRO + CUDA or OpenCL will give me more power on my desktop!

I am specifically interested in all of this with the focus on AI for rules evaluation and neural network training.

Here are some questions.

  1. What kind of introductory reading would be a good start?

  2. Which aspects of programming are or are not suitable for this? In the old days, we used Cray’s FORTRAN compiler which would identify DO looping constructs and perform that processing in parallel. This seems to be more of a division of processing at the programming level, where my application would invoke processing on the GPU. I guess the question is how to identify which parts of processing to do on the GPU. Is this limited to math only? Or logic processing surrounding the math as well?

  3. Is this the appropriate place for me to even ask these questions? Or is there another forum somewhere else that is a better fit?

Thanks!

Pete

  1. I’d start with the programming guide, the CUDA podcasts, and the UIUC course (which is probably a little out of date with GT200 out but it’s still valuable info): http://www.nvidia.com/object/cuda_education.html

  2. Depends on the application. This will probably make more sense once you have a better handle on the programming model, so don’t fret.

  3. Yeah, this is probably the best place.