Why is the Grid Z Dimension limited to only 1? is this a HW or SW limit? and is it going to change in the future?
Having a higher Z dim in the grid would allow to perform batch kernels more easily (I’m working with a large number of very small images)
It’s a hw limit on pre-Fermi arch. Fermi allows 3d grids.
Thanks! Running to the store to buy one!!! :-)
Before running, make sure Fermi’s z-limit is not ‘2’ or something like that ;-)
Actually, after checking it turns out the hardware supports this, but CUDA doesn’t currently.
You should still run to the store to buy a Fermi, it’s not hard to calculate you own thread indices!