Closed Bug 1411988 Opened 7 years ago Closed 7 years ago

IE Maze solver speed regressed significantly over the last 2 weeks

Categories

(Core :: Web Painting, defect)

defect
Not set
normal

Tracking

()

RESOLVED FIXED
mozilla59
Tracking Status
firefox-esr52 --- unaffected
firefox56 --- unaffected
firefox57 --- unaffected
firefox58 --- fixed
firefox59 --- fixed

People

(Reporter: pascalc, Assigned: mattwoodrow)

References

Details

(Keywords: perf, regression, Whiteboard: [perf])

Mozilla/5.0 (X11; Linux x86_64; rv:58.0) Gecko/20100101 Firefox/58.0 ID:20171025230440

Go to https://testdrive-archive.azurewebsites.net/Performance/MazeSolver/Default.html and click on start (default 30x30 maze)

Firefox 56 on my machine takes 14s, the last nightly build with a fresh profile takes 33s.

I tested with mozregression and the last build giving ~14s was the one on 2017-10-17 then we jumped to 22s

Last good revision: 1498abcee79d232125fb44846849160ec3c1fa51
First bad revision: 940a0c98e46c363d97e9cc8ef448faec9ff725f2
Pushlog:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=1498abcee79d232125fb44846849160ec3c1fa51&tochange=940a0c98e46c363d97e9cc8ef448faec9ff725f2

Two bugs were listed as the cause: bug 1407815 and bug 1406727

Then 2 days ago there was another regression on this test that made us jump from 22s to 33s and for this one mozregression gave me this:

Last good revision: 71a01e882f44c3143b9b2dd777f988f58e7af862
First bad revision: 8990a530c1eaaec234bfb37b5860cde8a4c0bec7
Pushlog:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=71a01e882f44c3143b9b2dd777f988f58e7af862&tochange=8990a530c1eaaec234bfb37b5860cde8a4c0bec7

The only patch in this regression window is bug 1405146 which was also blocking bug 1407815 which caused the first perf regression.


I don't know if we care about this css performance test or not, Matt could you have a look since the regression seems to have been caused by the patches for retained display list? Thanks
Flags: needinfo?(matt.woodrow)
Component: Layout → Layout: Web Painting
Yeah, this is pretty bad on my Mac as well (13seconds in 57 -> 34seconds 58).
Flags: needinfo?(matt.woodrow)
Oops, sorry. Matt, can you take a look?
Flags: needinfo?(matt.woodrow)
Jet, any thoughts on this layout regression?
Flags: needinfo?(bugs)
These changes should only affect Nightly/Beta, and are going to be backed out soon.
Flags: needinfo?(matt.woodrow)
(In reply to Matt Woodrow (:mattwoodrow) from comment #4)
> These changes should only affect Nightly/Beta, and are going to be backed
> out soon.

Cool, thanks.
Flags: needinfo?(bugs)
Blocks: 1352499
Can we check if this issue still happened in Beta58 (this pref is off by default in beta)?
Flags: needinfo?(pascalc)
(In reply to Gerry Chang [:gchang] from comment #6)
> Can we check if this issue still happened in Beta58 (this pref is off by
> default in beta)?

Beta 5 of 58 takes 14s
Flags: needinfo?(pascalc)
59.0a1 (2017-11-26) (64-bit) took 14s too.
Fixed in bug 1406727.
Status: NEW → RESOLVED
Closed: 7 years ago
Depends on: 1406727
Resolution: --- → FIXED
Assignee: nobody → matt.woodrow
Target Milestone: --- → mozilla59
You need to log in before you can comment on or make changes to this bug.