dominik_g

global_work_offset in OpenCL 1.1

Discussion created by dominik_g on Jan 12, 2011
Latest reply on Jan 12, 2011 by dominik_g

With OpenCL 1.1 the global_work_offset parameter for clEnqueueNDRange can be used to specify an offset for calculating the global ID of a work-item.

In my understanding this should mean that both get_global_id() and get_group_id() should be affected by the offset. However, it seems like the AMD implementation only considers the offset for get_global_id() but not for get_group_id().

Unfortunately I don't have access to NVIDIA's implementation of OpenCL 1.1 to test what they do...

Outcomes