You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For the ElementwiseKernel, there is no mechanism to determine the work group size before invocation. Sometimes, this is a problem, e.g. if you want to allocate local memory proportional to the expected work group size.
What is the best pattern to overcome this limitation? Would it be a good idea to incorporate a mechanism for controlling the work group size before invocation into ElementwiseKernel?
Thanks in advance,
Johannes
The text was updated successfully, but these errors were encountered:
I'm sort of against that. ElementwiseKernel is intended for simple things. As soon as you allocate local memory, what you're doing is no longer simple, to my mind. Specifically, workgroups are an unexposed implementation detail of ElementwiseKernel. As soon as you want the level of control you describe, I feel like you should graduate to just a plain old kernel invocation.
I'm sort of against that. ElementwiseKernel is intended for simple
things. As soon as you allocate local memory, what you're doing is no
longer simple, to my mind. Specifically, workgroups are an unexposed
implementation detail of ElementwiseKernel. As soon as you want the level
of control you describe, I feel like you should graduate to just a plain
old kernel invocation.
Sorry,
Andreas
—
Reply to this email directly or view it on GitHubhttps://github.com//issues/26#issuecomment-32796813
.
For the ElementwiseKernel, there is no mechanism to determine the work group size before invocation. Sometimes, this is a problem, e.g. if you want to allocate local memory proportional to the expected work group size.
What is the best pattern to overcome this limitation? Would it be a good idea to incorporate a mechanism for controlling the work group size before invocation into ElementwiseKernel?
Thanks in advance,
Johannes
The text was updated successfully, but these errors were encountered: