[1337p337]

Definitely performance is a nice benefit. I don't use AdBlock; I have a few thousand lines in my /etc/hosts that map various bad actors (vast chunks of Javascript that doesn't do anything useful, 1x1 .gifs, ads, annoying widgets, etc.) to 255.255.255.255 (NXDOMAIN), and use dnsmasq on that machine to get that benefit across to the rest of the local network, regardless of browser. The thing that usually triggers another bout of "block these hosts" is if I notice that my ostensibly idle browser is burning CPU. I don't want to wait several seconds for a page to render, and I certainly don't want to waste laptop battery time on refreshing an iframe with "what people are saying about this on Twitter" or feeding someone's analytics server. Half the time a site runs really slowly, it's usually doing some nonsense that I'd be happier without, or my browser is waiting to download something that I don't need to see.

The total amount of Javascript in kB is one issue, but you don't need to hit the level of opcodes, I think. Counting the nodes in the parsetree would be useful, as well as measuring the amount of memory that is allocated by the Javascript, determining whether it runs once or continuously or on some trigger (e.g., scrolling, hovering, mouse movement). If you ran a profiler and measured the amount of time spent in the Javascript bits of your runtime (e.g., how long to parse the JS, how long to eval it, how often it runs, throw in some memory profiling) you could probably get some interesting numbers.