Save a per-character branch in the HTML parser
Created attachment 62479 [details] Patch
Comment on attachment 62479 [details] Patch So I'm confused by the m_data.clear()s being moved to .clear(). When is ::clear() called? I guess either the branch isn't actually that hot, or branch prediction is just winning out here.
> So I'm confused by the m_data.clear()s being moved to .clear(). When is ::clear() called? Clear is called by the HTMLDocumentParser after each token. It reset the token. Before we were using lazy clearing where we delayed zeroing everything out until we started a new token. I want to get rid of the branch, so i made this eager. > I guess either the branch isn't actually that hot, or branch prediction is just winning out here. Probably. Or maybe moving the clear is costing us. In any case, I think this is prettier.
Comment on attachment 62479 [details] Patch Rejecting patch 62479 from commit-queue. Failed to run "['WebKitTools/Scripts/run-webkit-tests', '--no-launch-safari', '--exit-after-n-failures=1', '--ignore-tests', 'compositing', '--quiet']" exit_code: 1 Running build-dumprendertree Compiling Java tests make: Nothing to be done for `default'. Running tests from /Users/eseidel/Projects/CommitQueue/LayoutTests Testing 20715 test cases. animations/play-state.html -> failed Exiting early after 1 failures. 123 tests run. 12.72s total testing time 122 test cases (99%) succeeded 1 test case (<1%) had incorrect layout Full output: http://queues.webkit.org/results/3623012
Comment on attachment 62479 [details] Patch Bug 42821, again.
Comment on attachment 62479 [details] Patch Rejecting patch 62479 from commit-queue. Failed to run "['WebKitTools/Scripts/run-webkit-tests', '--no-launch-safari', '--exit-after-n-failures=1', '--ignore-tests', 'compositing', '--quiet']" exit_code: 1 Running build-dumprendertree Compiling Java tests make: Nothing to be done for `default'. Running tests from /Users/eseidel/Projects/CommitQueue/LayoutTests Testing 20715 test cases. animations/play-state.html -> failed Exiting early after 1 failures. 123 tests run. 11.55s total testing time 122 test cases (99%) succeeded 1 test case (<1%) had incorrect layout Full output: http://queues.webkit.org/results/3572410
Comment on attachment 62479 [details] Patch Broken animation test. :(
The failure rate on that test is out of control. Can we do something about it?
Comment on attachment 62479 [details] Patch Clearing flags on attachment: 62479 Committed r64010: <http://trac.webkit.org/changeset/64010>
All reviewed patches have been landed. Closing bug.