Hello again fxos, Previously [1] I had laid out a comparison of our v2.2 performance baseline with data available for master. This was done against the baseline we had at the time, the 319MB Flame, but the effort is now to switch to the 512MB Flame as the new baseline.
We have revamped our v2.2 performance baseline for the 512MB Flame, which also lets us use an updated v2.2 Gecko and Gaia, updated base build, longer rests between application launches, and performing GCs before memory capturing. The result is an overall more stable baseline to compare against. v2.2 -> master comparison: https://gist.github.com/eliperelman/b81a27b7f7f3b3f75f49 Thanks to some amazing statistical insight from Stas, we are now also introducing a different calculation to ground our baseline in: 95% Upper Bound [2]. This is now the calculation used in Raptor v3.x instead of 95th percentile. The dashboards [3] have been updated to display the new baseline for the 512MB Flame and now use the 95% Upper Bound for charting metrics. If you have any questions about these metrics and how they are derived, or would like assistance in testing Raptor with your own apps, please feel free to reach out. Thanks! Eli Perelman P.S. Huge thanks to :gandalf, :stas, and :rwood for making this new baseline possible! [1] https://groups.google.com/d/msg/mozilla.dev.fxos/HeGci2jormU/a3rQZcSvAwAJ [2] I'll let Stas elaborate on what that means if he feels inclined. :) [3] https://raptor.mozilla.org/
_______________________________________________ dev-fxos mailing list [email protected] https://lists.mozilla.org/listinfo/dev-fxos
