Alex Nitz <alex.nitz(a)ligo.org> writes:
> I was trying to understand the source of the pyopencl reduction kernel am
> not sure that the following line is correct.
> It seems to me that if the no_sync_size >= group_size the kernel may
> attempt to access local memory outside of the allocated amount. I have
> tested a case on my gtx580 and found that when group_size = no_sync_size =
> 32, there is indeed a kernel failure.
> I have attached a simple one line fix, but there might be a better way to
> do it.
You're right--thanks for pointing this out. Fixed in 4b3b520.