I introduced a performance regression in http://trac.webkit.org/changeset/54499. It can be fixed by providing a split similar to what had existed previously: one path for objects that might be constructed in a WorkerContext, and a fast path for objects that are never in a WorkerContext.
Created attachment 48512 [details] patch Performance on the DOM benchmark suite with this patch is comparable to a build with 54499 reverted, and dromaeo performance appears to be slightly better.
Thanks a lot, Nate. Only one thing---instantiate in worker context tells me that it must be a worker context, however it looks like it's possibly a worker context. If it's correct, could you give some other name. The ugly idea I have is in possible worker context.
(In reply to comment #2) > Thanks a lot, Nate. > > Only one thing---instantiate in worker context tells me that it must be a > worker context, however it looks like it's possibly a worker context. > > If it's correct, could you give some other name. The ugly idea I have is in > possible worker context. Yeah, I debated the name. Perhaps instantiateV8ObjectMaybeInWorkerContext?
(In reply to comment #3) > (In reply to comment #2) > > Thanks a lot, Nate. > > > > Only one thing---instantiate in worker context tells me that it must be a > > worker context, however it looks like it's possibly a worker context. > > > > If it's correct, could you give some other name. The ugly idea I have is in > > possible worker context. > > Yeah, I debated the name. Perhaps instantiateV8ObjectMaybeInWorkerContext? Yep, something like that. Thanks.
Comment on attachment 48512 [details] patch ok. I was so happy for this code to be gone. Oh well.
http://trac.webkit.org/changeset/54680