Help! research work on CUDA

Hi how are you?. I am doing a research work on CUDA and I want to resolve some doubts. When the CPU throws a kernel-cuda, OS-level CPU, what is created? A kernel thread, a user thread, a process or a lightweight process?. Another question is for CUDA thread blocks. For each block of threads CUDA, the CUDA API What is created? A process for each CUDA threads block? Threads level CUDA GPU API, what kinds of threads are?

Thank you and sorry if my english is not very good.

Hi how are you?. I am doing a research work on CUDA and I want to resolve some doubts. When the CPU throws a kernel-cuda, OS-level CPU, what is created? A kernel thread, a user thread, a process or a lightweight process?. Another question is for CUDA thread blocks. For each block of threads CUDA, the CUDA API What is created? A process for each CUDA threads block? Threads level CUDA GPU API, what kinds of threads are?

Thank you and sorry if my english is not very good.

To answer your question regarding thread blocks: GPU “threads” should not be confused with CPU threads as you know them! They are native to the hardware. There’s no “process” to speak of on a thread block level. Please read the programming guide that’s included with the CUDA toolkit, it gives an excellent introduction into the CUDA memory and execution models, far better than I would be able to explain ;). I’m not sure what you mean with “CPU throws a kernel-cuda, OS-level CPU, what is created?” so I can’t answer that.