This is a "it depends" answer, and not just on the performance of the router or switch.
For example, if the had an ACL with 500 ACEs, probably a big performance difference if 99.9% of the processing stopped with the first ACE vs. falling all the way through the ACL.
Some IOSs had implemented a feature called TurboACL (or compiled ACLs), which (from some literature) found an ACE in a "flat" time equivalent of processing 5 ACEs regardless of ACL size.
Some router IOSs, I suspect, might "optimize" the ACEs into a reduced number, physically, to improve ACL processing.
I.e.
if your ACEs were something like:
access-list 1 permit host 10.0.0.0
access-list 1 permit host 10.0.0.1
access-list 1 permit host 10.0.0.2
access-list 1 permit host 10.0.0.3
access-list 1 permit host 10.0.0.4
access-list 1 permit host 10.0.0.5
access-list 1 permit host 10.0.0.6
access-list 1 permit host 10.0.0.7
possibly what's actually matched against might be:
access-list 1 permit standard 10.0.0.0 0.0.0.7
or perhaps having
access-list 1 permit host 10.0.0.0
access-list 1 permit host 10.0.0.1
access-list 1 permit host 10.0.0.2
access-list 1 permit host 10.0.0.3
access-list 1 permit host 10.0.0.4
access-list 1 permit host 10.0.0.5
access-list 1 permit host 10.0.0.6
access-list 1 permit host 10.0.0.7
might match against
access-list 1 deny host 10.0.0.6
access-list 1 permit standard 10.0.0.0 0.0.0.7
Switches, I believe, do something like the forgoing when placing ACLs into TCAM. Processing time should be the same, regardless of the ACL length but different switches have different TCAM resources. Too large/many ACLs can overflow such resources, and you would then have a major performance problem.
As, again, this is a "it depends" answer, it's very difficult to predict just how any particular router or switch will deal with an ACL without knowing much more. About the only thing you might infer, a "faster/better" device will likely have more capacity dealing with a large ACL.
