How to reproduce: * Load a webpage from scratch * Wait to see the page complete rendering Expected: * Fennec native is within ballpark of performance of Google Chrome for Android or Android's native browser. Actual: * Fennec native is substantially worse. Quoting from bug 723361 (currently private): I went ahead and wrote the necessary code to do a cross-browser startup comparison using Eideticker. I decided to create a benchmark that determined the first frame of an HDMI capture which represents a more-or-less finished version of a page -- in this case the new york times. You can translate that into "# of seconds to see completed webpage" by dividing it by 60. Unfortunately the news here is not particularly good: Browser,result1,result2,result3,average com.android.chrome,199,206,195,200 = 3.33s avg org.mozilla.firefox,375,365,385,375 = 6.25s avg org.mozilla.firefox_beta,423,400,420,414 = 6.9s avg com.google.android.browser,221,185,195,200 = 3.33s avg What this benchmark seems to show is that the Fennec beta is actually marginally *worse* than the current released XUL fennec, as well as being worse than all the other browsers overall. Looking at the captures myself, I can see that the story isn't actually as bad as the numbers initially suggest-- although it takes longer to actually get a completed rendered web page with the new fennec, we do see the browser chrome much earlier with the new beta (so the user experience is better).
See bug 760152 for some already committed work on this bug (which unfortunately doesn't seem to help much).
My experience with Fennec as a user gives the feeling that network access sucks badly. I'm not even talking about startup here, I'm talking about getting any page displayed.
(In reply to Mike Hommey [:glandium] from comment #2) > My experience with Fennec as a user gives the feeling that network access > sucks badly. +1. I filed bug 758338 for this issue and plan to spend some time investigating this.
You need to log in before you can comment on or make changes to this bug.