CopiedBlock should be 16kB
Created attachment 275551 [details] Patch
Comment on attachment 275551 [details] Patch r=me
Comment on attachment 275551 [details] Patch cq+
Comment on attachment 275551 [details] Patch Clearing flags on attachment: 275551 Committed r199016: <http://trac.webkit.org/changeset/199016>
All reviewed patches have been landed. Closing bug.
Something really curious happened with this: https://arewefastyet.com/#machine=29&view=single&suite=kraken&subtest=beat-detection Any idea how to get back this win without the other regressions?
Rolled out in <http://trac.webkit.org/changeset/199032>. "Regressed Octane and Kraken on the perf bots."
> Any idea how to get back this win without the other regressions? I might be able to bring it back. I didn't quite do it the best way the first time because I forgot to consider aligned allocation requests, which still take the slow path. I don't understand that graph though -- what does it say?
Committed r199125: <http://trac.webkit.org/changeset/199125>
Rolled out in <http://trac.webkit.org/changeset/199145>. "Perf bots are down, so I can't re-land this right now."
On Octane, there's no regression. On Kraken: 1% speedup on Mac Pro 1% speedup on iPhone (+/- !%) 1% regression on MacBook Pro 1% regression on MacBook Air That's enough of a mixed review that I think I'll leave this alone for now.