cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

sandyandr
Adept I

Closed-loop kernel with OpenCL 2.0 - is it possible?

I know, the question may seem strange, but still: is it possible to make (and use) a kernel, which works in a closed-loop manner? I can foresee only problems with graphics here, but I can run it on my secondary GPU. What for? I just want to get rid of even smallest overhead from the kernel queuing. The kernel execution should be controlled by host over fine-grained SVM (and thru atomic operations from the inside) only. Is it possible, how do you think?

0 Likes
3 Replies
maxdz8
Elite

It's not like piping is overhead free. I wouldn't expect it to burn in flames but I cannot see much cases for that type of usage. Please try to make a more realistic example. From this 10'000 ft view it might make sense or not, I personally would not consider it even worth trying just for the sake of it - it is not standard dispatch model.

Host-side kernel enqueuing is a relatively simple operation and the hardware dispatcher has some overhead as well.

0 Likes

The general case: suppose, you have some rules, in accordance to which you should generate a lot of information very quickly from some constantly incoming data. Though it seems you're right about overhead, the suggested approach could give a kernel much greater advantage here - consistent local memory (CU) - how do you think? I'm planning to try it, anyway.

0 Likes
maxdz8
Elite

Please do try it. It's very interesting.

0 Likes