https://bugs.documentfoundation.org/show_bug.cgi?id=167979

--- Comment #8 from Eyal Rozenberg <[email protected]> ---
(In reply to Samuel Fogh from comment #7)
> This patch might fix it. Global worker size was not divisible by local 
> worker size, which could cause CL_INVALID_WORK_GROUP_SIZE.

In CUDA that's true, in OpenCL, the overall NDRange dimensions don't generally
have to be divisible by the workgroup size.

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to