Quote Originally Posted by jai
it means that values are automatically assigned to cores of gpu ? It means kernel automatically runs on next core whose value is --> previous core's id + 1?
Yes you got it.

I think this picture is better than the last one i posted and will show how it works:

One workitem is done on one processor. Every workitem get's a unique id(get_global_id()). The items inside a workgroup are executed in parallel. And this is done for all items.

You could use http://www.fixstars.com/en/opencl/bo...ng-the-kernel/ this page for clarification. Its quite helpful.