Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
11 years ago

Channel from many work items to single task

Hi, I am trying to compile something that is computing something with many work items. They are spit out to a channel into a single work item. There is no way to know how many at compile time b...
Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
11 years ago

--- Quote Start ---

Hi,

I wished to have N work-items in the write kernel, and call the single task kernel reading N times.

I have found that with the static (defined) N, this seems to be faster.

Stephen

--- Quote End ---

Hi,

I actually am interested in doing so, too. You mean that with a static global size, it does work?

How do you define a static global size at compile time? I thought we only can specify it in the call to enqueueNDRangeKernel().