[Webkit-unassigned] [Bug 98694] [Performance] Speed-up DOM tree traversal on ARM platforms

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Thu Oct 11 17:10:47 PDT 2012


https://bugs.webkit.org/show_bug.cgi?id=98694





--- Comment #17 from Benjamin Poulain <benjamin at webkit.org>  2012-10-11 17:11:28 PST ---
>     - Not making NULL check for PLD addresse seems to have very negative impact on both X86 and ARM based architectures.

This is very surprising for ARM. In the design document, it says PLD has been designed explicitly with the branch issue in mind. Loading an invalid address has no impact, while the branch will cost cycles on misspredictions.

Have you only tested on Krait or did you get to try on a "regular" implementation of ARMv7? Because krait is more the exception than the norm.

Have you tried having Node::updatePrefetchTarget() implemented but being an NOOP? Because Node::updatePrefetchTarget() influence the cache so you are not measuring the PLD alone.

Very nice work in any case.
I think we may have to tweak the stride size by ARM implementation.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.



More information about the webkit-unassigned mailing list