Talos has detected a Firefox performance regression from push b658ca8e2aed84bbc9cbeb52b341e7d40588a8bd. As author of one of the patches included in that push, we need your help to address this regression. Regressions: 2% tp5n nonmain_normal_fileio windows7-32 opt 347,584,174.46 -> 356,073,110.46 2% tp5n nonmain_normal_fileio windows7-32 pgo e10s 348,593,349.62 -> 355,660,784.00 2% tp5n nonmain_normal_fileio windows7-32 pgo 349,798,966.67 -> 355,851,823.42 1% tp5n nonmain_normal_fileio windows7-32 opt e10s 349,238,495.67 -> 354,079,121.33 You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=6475 On the page above you can see an alert for each affected platform as well as a link to a graph showing the history of scores for this test. There is also a link to a treeherder page showing the Talos jobs in a pushlog format. To learn more about the regressing test(s), please see: https://wiki.mozilla.org/Buildbot/Talos/Tests For information on reproducing and debugging the regression, either on try or locally, see: https://wiki.mozilla.org/Buildbot/Talos/Running *** Please let us know your plans within 3 business days, or the offending patch(es) will be backed out! *** Our wiki page outlines the common responses and expectations: https://wiki.mozilla.org/Buildbot/Talos/RegressionBugsHandling
this is smaller, but looks to be consistent with the landing/backout/landing of thie patches in bug 1359653. nonmain == not the main thread normal == not startup or shutdown fileio == bytes read/write we only measure xperf on windows 7, so that is why this is only a win7 thing :) :kmag, can you look at your code in bug 1359653 and determine if we are increasing fileIO. Then determine if we can fix anything, if we need to backout, or accept the new level of fileio.
I believe this should be offset by a decrease in main fileio, which was the intention, and would be an improvement. It looks like that may be the case, but the numbers are pretty noisy for that changeset, so it's hard to tell.
The other possibility is that the additional nonmain normal IO comes from writing out the new cache file. We generally shouldn't do that for most sessions, unless something changes and we need to write out a new set of scripts. It's possible that's happening during talos runs, in which case I'd consider it an acceptable increase. But it might help to make sure we don't also write the same scripts to the main startup cache, too.
hmm, I don't see any improvements in main/nonmain startup/normal fileIO on e10s or non-e10s- I only looked at opt, not pgo. This is a small change though and luckily the other numbers didn't go up!
OK, in that case it probably is from writing out the new cache file. The numbers look about write for that. I'll try to confirm. We might be able to avoid that happening during tp5o tests.
You need to log in before you can comment on or make changes to this bug.