I’m sure there is someone with more technical knowledge on this, but in a nutshell content blockers add a bit of overhead to Safari as it is basically just string matching that needs to be executed fast, for every page load. This is something the 64-bit processors are much better at. Content blockers were designed for speed and efficiency so it makes sense that they would restrict it to the hardware that actually delivers that.
Also note that Apple does enforce a size limit for the block list, so I suppose there needs to be a baseline where the performance remains adequate if that limit is hit. So even though ‘light’ adblockers could run just fine on 32-bit devices, maybe the performance is bad on large websites with an extensive block list. This does have an impact on the user experience and may be of particular concern to Apple.
Perhaps.
I use Lightbeam on Firefox on my rMBP. It shows me what other sites are linked to by the sites I go to. Sometimes, just landing on a home page for a site will result in that site going to 20 or more other (third-party) sites. Some of those are for retrieving content to display, some are adservers, some are analytics or tracking sites. From what I've seen, most of these links are in the last 2 categories.
Safari on IOS would experience the same linking phenomena as I see with Firefox running OSX on my rMBP.
So I tend to think that Safari performance on IOS9 is far more affected by these link outs to other sites than it is to do the processing to decide which sites to block. Does the content blocking in IOS9 block all these other nefarious sites, like the trackers that don't result in ads, or is it just the adservers that get blocked?
I know there was a significant performance tax on the A5 in my iPad3 because of the Retina display handling needed. But I can't believe that a few filtering calculations could be all that taxing to it.