GPU CPU Communication

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi all,
I have a little idea about openCL.I did some coding for matrix multiplication using openCL on GPU.
Now I am doing a project on GPU-CPU Communication where CPU will port some task on GPU.
After porting the task to GPU,CPU will continue with it’s normal execution until GPU complete the task.
I want to integrate this code at linux kernel level.So whether such thing is possible using linux kernel
and openCL kernel. I will be extremely thankful if you help me.
Also I am posting thing message on "OpenCL " Forum.
thanks in advance.

–Regards
Ashwin Vasani

2

Not possible. OpenCL and CUDA use APIs which are only exposed in user space. There is no way kernel space code can directly call either API. The only way it could be done would be with a user space helper process.

3

Not possible. OpenCL and CUDA use APIs which are only exposed in user space. There is no way kernel space code can directly call either API. The only way it could be done would be with a user space helper process.