Bug 104323 – 10% createNodes regression.

• Status: RESOLVED
• Resolution: WONTFIX
• Component: DOM

[Ojan Vafai]

10% regression: http://build.chromium.org/f/chromium/perf/linux-release-webkit-latest/dom_perf/report.html?rev=171397&graph=CreateNodes&trace=score&history=100

Hard to be sure, but the regression range looks like http://trac.webkit.org/log/?verbose=on&rev=136605&stop_rev=136590.

Could it possibly be http://trac.webkit.org/changeset/136590/ ?

[Adam Barth]

I'm sorry that I won't be available to follow up on these regressions.  Please feel free to roll out any patches of mine that you suspect as causing a regression.

[Kentaro Hara]

I'll take it over. Enjoy your vacation:)

[Kentaro Hara]

(In reply to comment #0)
> 10% regression: http://build.chromium.org/f/chromium/perf/linux-release-webkit-latest/dom_perf/report.html?rev=171397&graph=CreateNodes&trace=score&history=100
> 
> Hard to be sure, but the regression range looks like http://trac.webkit.org/log/?verbose=on&rev=136605&stop_rev=136590.
> 
> Could it possibly be http://trac.webkit.org/changeset/136590/ ?

As far as I see, r136590 won't be the culprit. The patch removes one member variable and if checks, which will slightly improve performance in theory. Also the patch won't affect performance of CreateNodes.

It looks a bit hard to identify a regression range.

[Emil A Eklund]

Any luck tracking this down? We are still worse of than before :(

[Ahmad Saleem]

This report is too old to be actionable at this point. We might have already actioned / fixed this. Apple will also have internal performance tracking as well (which might have this coverage).

Marking this as 'RESOLVED WONTFIX'. Please reopen if you think we still can do something here.