Bug 88888 added an optimization to avoid walking such subtrees during painting but it didn't patch the hit testing path. On http://dglazkov.github.com/performance-tests/biggrid.html, hit testing is suffering from the same over-head as painting: merely traversing a lot of non-self-painting-layers is proving costly. This means hit testing would benefit from the same optimization.
Created attachment 148191 [details] Proposed performance change 1.
Comment on attachment 148191 [details] Proposed performance change 1. Looks reasonable.
Thanks Eric, this optimization is still valid and would be superseded by bug 89899. I don't have much time to investigate the other bug so let's just land this in the meantime.
Created attachment 169280 [details] Patch for landing
Comment on attachment 169280 [details] Patch for landing Clearing flags on attachment: 169280 Committed r131665: <http://trac.webkit.org/changeset/131665>
All reviewed patches have been landed. Closing bug.