CPUs are pipelined though so you can easily imagine many slots being operated on at once. Perhaps a better metric is the relative cost of the slot assignment ag

CPUs are pipelined though so you can easily imagine many slots being operated on at once. Perhaps a better metric is the relative cost of the slot assignment against the the cost of the read/write operation. Not to mention the fact that elements may be in L1, in which case the penalty of additional cycles may actually be felt.