I doubt it's a big enough perf problem to block the landing, so I'm splitting this out into a separate bug. I added a dynamic check to StoreBuffer::CellPtrEdge::trace() to determine whether we're looking at a string or an object, and I *think* it can be done away with by making a separate store buffer for each. (Though I also think that it may be easier in the JIT to not distinguish during insertion?) Also, I have a mostly redundant IsInsideNursery check that could be removed with some refactoring.
status-firefox57: --- → affected
Priority: -- → P3
You need to log in before you can comment on or make changes to this bug.