That 3K and 9K platforms don't really have a notion of software switching. If it can't be programmed in the forwarding ASIC then it's not forwarded. There are a couple of minor exceptions but for standard unicast/multicast forwarding it's going to either go in hardware or not at all.
This, incidentally, is why the number of ACL's supported on these platforms is so limited (on the 3K in particular).
Also - packet size/volume/bandwidth doesn't really have an effect on CPU on these platforms but there is some PPS limit on the ASIC complex. I don't recall the specifics of the 3064 in particular but it should be at least in the close vicinity of line rate.
That said, instead of tracking CPU (...as would be the case on an older IOS box) you may want to pay close attention to buffer utilization...especially if you're running a speed mismatch on the box (i.e. 1G <-> 10G).