Created attachment 8753748 [details] [diff] [review] Patch The addEventListener micro-benchmark in bug 1273481 spends a lot of time under JS::GetScriptedCallerGlobal (called by GetIncumbentGlobal). GetScriptedCallerGlobal is mainly slow when dealing with Ion inline frames. We can add a fast path that bypasses this, though: if the top activation has at least one non-self-hosted, non-inline frame, we can use activation->compartment->global. This speeds up the addEventListener benchmark from ~2100 ms to ~1500 ms.
Comment on attachment 8753748 [details] [diff] [review] Patch Very nice