Some basic problems on CUDA

Hello, everyone~

I am a rookie in CUDA programming. I start learning it by reading the programming guide, although that is really tedious to do that. Some problems poped up in the process,

How to match the programming model with the hardware?

The number of the multiprocessor in one graphics card is determinded by the card itself? If so, where can I find this information.

Any better learning curve in study this programming language?

Big thanks in advance~ :rolleyes:

The programming guide is the best place to learn in my opinion. It is basically THE source of all the useful information.

If you are looking for a table of hardware capabilities of the various GPUs, check Appendix A.

Take a look at the first reply in this thread:

There’s really not much learning curve if you know C. If not, then that’s a different issue.

The programming model matches pretty directly to the hardware. You might want to check out articles about G80 or G200 architecture, such as the ones on Beyond3D. These are very helpful for understanding how GPUs work.

You can have info of the GPU in the deviceQuery example, execute it and it will tell you usefull info as maxthreads per block, max blocks of yor GPU, memory per block…