Open Bug 553055 Opened 10 years ago Updated 10 years ago
1-3% Ts regression after switching to Places
As noted in bug 351351 comment 29, we had a ~1-3% Ts regression from switching to Places: (In reply to bug 351351 comment 29) > Ts on cb-minibinus01 + ~40ms, + 2.6% > Ts on cb-xserve01 + ~5ms, + 0.8% > Ts on cb-xserve04 + ~20ms, + 3.1% I didn't record the actual numbers (in part because the Sunday night reboots on minus and x4 reset things somewhat, but they reset to a "base" level higher than the previous week's level), but cb-x1 is far more consistent and not showing any signs of shaking off its jump. [4:16pm] ardissone: hendy, smorgan: is the consensus on the Ts that there's nothing we can do to control the startup hit? [4:16pm] pinkerton: :( [4:16pm] smorgan: I don't know, I haven't profiled [4:16pm] smorgan: If hendy's core->objc loading code is currently run at startup, then there is something we can do [4:17pm] smorgan: (assuming that takes significant time, which I assume it does) [4:17pm] hendy: it's not run at startup, afaict [4:17pm] smorgan: Ah, bummer :( [4:18pm] ardissone: should we open a placeholder bug for having some profile that, or no? [4:19pm] smorgan: May as well I guess [4:19pm] ardissone: ok [4:19pm] smorgan: Worst case we'll profile startup and find nothing we can do differently [4:20pm] smorgan: Getting a current breakdown for how much time we spend on what during startup would be nice regardless of history changes :)
10 years ago
Depends on: 556428
You need to log in before you can comment on or make changes to this bug.