in example of bug 732872 it takes up to 11% because of URI computation. Cached URI usage should help.

Comment on attachment 625024 [details] [diff] [review] patch I'm not familiar with Link. Jonas and Boris should know it better, based on blame.

Comment on attachment 625024 [details] [diff] [review] patch I won't have time to get to this anytime soon. Checking with bz.

Comment on attachment 625024 [details] [diff] [review] patch Wouldn't it make more sense to move the cache check into IsLink()?

(In reply to Boris Zbarsky (:bz) from comment #3) > Wouldn't it make more sense to move the cache check into IsLink()? IsLink is a virtual method and takes an argument we don't need what makes us a little bit slower than we could be. Good example how minor stuffs can make a difference I've got is bug 756381 where we lost 3% perf on nsCOMPtr things.

(In reply to Boris Zbarsky (:bz) from comment #3) > Comment on attachment 625024 [details] [diff] [review] > patch > > Wouldn't it make more sense to move the cache check into IsLink()? in case if I misread you: I can fix IsLink() while I'm here but I would avoid to use it for IsHTMLFocusable implementation.

Comment on attachment 625024 [details] [diff] [review] patch OK, r=me. I'm still planning to rip out all this stuff anyway...

(In reply to Boris Zbarsky (:bz) from comment #6) > OK, r=me. I'm still planning to rip out all this stuff anyway... ok, cool since these stuffs are crazy a little bit (I guess because old and new code share the same room). Then I'm leaving IsLink untouched for now.