1667974 - 0.72 - 4.53% cnn PerceptualSpeedIndex / cnn SpeedIndex / ebay-kleinanzeigen-search LastVisualChange / ebay-kleinanzeigen-search fcp (android-hw-g5-7-0-arm7-api-16-shippable) regression on push f8cda577ac2f0eb74b5707f5f8140d656994628f (Wed August 26 2020)

Status: RESOLVED WONTFIX

(Core :: Layout: Text and Fonts, defect)

Description

[Florin Strugariu [:Bebe]]

Perfherder has detected a browsertime performance regression from push f8cda577ac2f0eb74b5707f5f8140d656994628f. As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

5% ebay-kleinanzeigen-search fcp android-hw-g5-7-0-arm7-api-16-shippable opt cold 1,191.79 -> 1,245.75
4% ebay-kleinanzeigen-search LastVisualChange android-hw-g5-7-0-arm7-api-16-shippable opt cold 3,930.48 -> 4,098.83
4% cnn PerceptualSpeedIndex android-hw-g5-7-0-arm7-api-16-shippable opt cold 4,818.87 -> 5,002.00
3% cnn SpeedIndex android-hw-g5-7-0-arm7-api-16-shippable opt cold 5,377.43 -> 5,562.17
1% ebay-kleinanzeigen-search fcp android-hw-g5-7-0-arm7-api-16-shippable opt cold 1,231.02 -> 1,239.83

Details of the alert can be found in the alert summary, including links to graphs and comparisons for each of the affected tests. Please follow our guide to handling regression bugs and let us know your plans within 3 business days, or the offending patch(es) will be backed out in accordance with our regression policy.

For more information on performance sheriffing please see our FAQ.

Comment 1

[Jonathan Kew [:jfkthame]]

Bug 1371386 made our font selection more reliable and correct, but that does involve doing additional work in some cases, so it's not surprising there could be some small performance regressions as a result. We may be having to check more fonts in order to ensure we make the correct selection for the given content.

I notice that the regressions reported here are all "cold", which I assume means they're occurring with a newly-launched browser, so no font information has yet been cached. Once the browser has had a chance to cache much of the relevant data, the impact is likely to be much less; this is supported by the fact that only "cold" tests are reported here.

So this is the price of ensuring we get correct font selection across a wider range of content and font configurations. IMO we should accept this and resolve WONTFIX here; :m_kato, what do you think?

Comment 2

[Makoto Kato [:m_kato]]

(In reply to Jonathan Kew (:jfkthame) from comment #1)

Bug 1371386 made our font selection more reliable and correct, but that does involve doing additional work in some cases, so it's not surprising there could be some small performance regressions as a result. We may be having to check more fonts in order to ensure we make the correct selection for the given content.

Yes. To select color font or non-color font, we have to read internal font data for color glyph support and we don't know current font supports full-emoji area. So I'm not surprised for this perf data.

I notice that the regressions reported here are all "cold", which I assume means they're occurring with a newly-launched browser, so no font information has yet been cached. Once the browser has had a chance to cache much of the relevant data, the impact is likely to be much less; this is supported by the fact that only "cold" tests are reported here.

So this is the price of ensuring we get correct font selection across a wider range of content and font configurations. IMO we should accept this and resolve WONTFIX here; :m_kato, what do you think?

Agreed. we should mark as WONTFIX for this.