Hi, there is a parameter in function ' clEnqueueNDRangeKernel
', called 'global_work_offset'. Can anybody explain it to me, please? (I have read the specifications, but still feel confused). It's better to give me an example or something like that.
Many thanks.
Originally posted by: haibo031031 Hi, there is a parameter in function ' clEnqueueNDRangeKernel
', called 'global_work_offset'. Can anybody explain it to me, please? (I have read the specifications, but still feel confused). It's better to give me an example or something like that.Many thanks.
Suppose you have two 1D array of 100 elements and you want to add only last 50 elements.
just specifing offset as 50 will add all last 50 element by element.
Originally posted by: genaganna Originally posted by: haibo031031 Hi, there is a parameter in function 'clEnqueueNDRangeKernel
', called 'global_work_offset'. Can anybody explain it to me, please? (I have read the specifications, but still feel confused). It's better to give me an example or something like that.
Many thanks.
Suppose you have two 1D array of 100 elements and you want to add only last 50 elements.
just specifing offset as 50 will add all last 50 element by element.
Thanks for your reply. That is : there are 100 elements, and we use 100 work-items to process them. When I want to all the last 50 elements, just specify offset as 50. At this moment, the total number of work-items/threads is still 100 or 50?
Originally posted by: haibo031031
Thanks for your reply. That is : there are 100 elements, and we use 100 work-items to process them. When I want to all the last 50 elements, just specify offset as 50. At this moment, the total number of work-items/threads is still 100 or 50?
Total number threads/work-items are 50 which is equivalent to (global_work_items - offset).