In bug 1342854 we increased the timer coalesce limit from 1 to 5. This makes the page animation run better in desktop, but it actually makes things worse in fennec. On android I see: * In FF51 with no timer yielding (no coalesce limit) the audio is janked the entire time. * In FF52 beta build with yielding after every timer (limit 1) the audio is not janked at all and the page renders smoothly (and quickly) * In FF53 aurora build with limit 5 (effecitvely 6) the audio alternates between playing and being janked Allowing 3 consecutive callbacks seems to work best on my nexus 5x. Barring another site to tune for I think this is a reasonable value. JS is going to run slower, so allowing fewer callbacks makes sense.
I think we should tune this in FF53, but not worth uplifting to FF52.
Created attachment 8842900 [details] [diff] [review] Tune dom.timeout.max_consecutive_callbacks on fennec. r=smaug This tunes the fennec pref down to 3 based on my local testing. I also added the pref to the default all.js pref file to make it easier for people to find the default on other platforms.
Comment on attachment 8842900 [details] [diff] [review] Tune dom.timeout.max_consecutive_callbacks on fennec. r=smaug nexus is some high-end-ish phone, right? Should we ask someone to try out lower end phone too?
(In reply to Olli Pettay [:smaug] from comment #3) > nexus is some high-end-ish phone, right? Should we ask someone to try out > lower end phone too? Nexus 5x lower end than you would think for the price due to some poor thermal design. Its second core is shutdown most of the time AIUI. I think, however, if we really want to discriminate between high and end low end devices on the same platform we need the time budget approach to limiting callbacks. The simple limit pref is too course to cover the range of possibilities. I'll file that follow-up now
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/mozilla-inbound/rev/7facce0be4a3 Tune dom.timeout.max_consecutive_callbacks on fennec. r=smaug