[webkit-dev] Optimizing DOMNode _nodeWith

[Alexey Proskuryakov]

On 29.04.2006 00:53, "Matt Gough" <devlists at softchaos.com> wrote:

>> Generally, we consider a repeatable 1-3% performance improvement to
>> be worth it. And I think your change is a cleaner design anyway. I
>> encourage you to put it up on bugzilla.
> 
> OK, but I'm not sure if I should provide a test case and if so, what
> it would test (and how)? (My patch runs all the current tests without
> any issues)

  Performance tweaks don't need test cases of their own.

  One of the BuildBot slaves runs an Apple internal suite of performance
tests for each checkin, presenting the results in this form:
<http://build.webkit.org/results/page-layout-test-mac-os-x/1227/PerformanceR
eportSummary.xml>. It's not possible to run these tests without checking in
a patch (for non-Apple people, that is), but at least one can promptly
rollback a patch that causes a performance regression :)

- WBR, Alexey Proskuryakov