On Mon, Aug 29, 2016 at 10:00:33AM -0700, Bart Van Assche wrote: > Was it really necessary to remove these two member variables? If the per-PD > global MR is ever removed all the code that uses pd->unsafe_global_rkey > will have to be touched again. No, it won't. I actually have a tree where it's gone, but as it so far only supports mlx4 I didn't want to bother anyone with it. > Keeping the two global_mr variables would > reduce the number of load instructions the CPU has to perform while mapping > SRP requests. We can add pointers to the pd in the same places, that should give you the same number of load instructions. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html