Closed Bug 1192196 Opened 9 years ago Closed 8 years ago

PAINT_RASTERIZE_TIME histogram has regressed on the 25th of July

Categories

(Core :: Panning and Zooming, defect)

Product:
Core
Component:
Panning and Zooming
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED WONTFIX

People

(Reporter: rvitillo, Unassigned)

References

Details

Benoit: can you forward this to the right person?
Flags: needinfo?(bgirard)
My first guess without looking at the range was APZ and it's right in the range. Bug 1157745. Not 100% sure but it's fairly likely.

It make sense, we're having to rasterize a lot more. I'm already working on mitigating the regressions from APZ but by some metrics it's going to be worse. This might just be one of these metrics.

In any case it's an indication that the measurement is effective. It would be bad if it didn't notice a major change like APZ.
Depends on: apz-windows
Flags: needinfo?(bgirard) → needinfo?(bugmail.mozilla)
Component: Layout → Panning and Zooming
Is it possible to show the alerts graph in comment 0 split by platform? If it were APZ I would expect the regression in that range to be limited to windows desktop, and I would expect a similar regression to show up for OS X desktop on July 22 or 23.

But yes I agree with BenWa that this is most likely a result of turning on APZ, and is pretty much expected behaviour. We can monitor this as BenWa lands bug 1186662 to see how much it helps but other than that this is likely a WONTFIX.
Blocks: apz-desktop
Flags: needinfo?(bugmail.mozilla) → needinfo?(rvitillo)
(In reply to Kartikaya Gupta (email:kats@mozilla.com) from comment #3)
> Is it possible to show the alerts graph in comment 0 split by platform? 

That's not possible in the alerts graph but you can segment the histogram by platform within the telemetry dashboard [1].

[1] http://telemetry.mozilla.org/new-pipeline/dist.html#!compare=os&cumulative=0&end_date=2015-07-25&max_channel_version=nightly%252F42&measure=PAINT_RASTERIZE_TIME&min_channel_version=null&product=Firefox&sanitize=1&sort_keys=submissions&start_date=2015-07-25&table=0&trim=1&use_submission_date=0
Flags: needinfo?(rvitillo)
Thanks. I poked around with that (specifically e10s vs non-e10s data on windows) and am pretty convinced that this regression is from APZ. The OS X data seems to be a bit noisier and it's harder to tell there but it seems to line up with the APZ landing there as well.

I'll leave this open until bug 1186662 lands and we can see how it impacts things.
Blocks: apz-windows
Depends on: 1186662
No longer depends on: apz-windows
(In reply to Kartikaya Gupta (email:kats@mozilla.com) from comment #5)
> I'll leave this open until bug 1186662 lands and we can see how it impacts
> things.

I don't expect bug 1186662 to improve the score measurably. It will improve rasterize time during important user interactions but we're still going to be painting the full displayport eventually which this metric will pick up.
There's nothing more to do here.
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.