1255936 - [e10s] tsvgr_opacity is about 65% slower on 32-bit Windows XP and 7 compared to non-e10s

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect, P4)

Description

[Chris Peterson [:cpeterson]]

tsvgr_opacity is about 65% slower on 32-bit Windows XP and 7, but 36% faster on 64-bit Windows 8:

https://treeherder.mozilla.org/perf.html#/e10s?filter=tsvgr_opacity

windows7-32 pgo: non-e10s = 532.70 ± 1.48%, e10s = 893.49 ± 1.36%, delta = 67.73%
windows8-64 pgo: non-e10s = 417.46 ± 1.02%, e10s = 265.16 ± 0.55%, delta = -36.48%
windowsxp pgo: non-e10s = 525.70 ± 1.80%, e10s = 865.08 ± 2.42%, delta = 64.56%

@ Avi: who can help debug this tsvgr_opacity regression?

Comment 1

[Avi Halachmi (:avih)]

Most probably the "fix" to the test at bug 1250717 make the results go from false negative to false positive. It's investigated either there or at bug 1254628.

IMO the numbers of tsvgr_opacity are not reliable for now (and probably weren't before too, especially on e10s).

mstange was hoping to investigate it this week.

Comment 2

[Jonathan Watt [:jwatt]]

This now looks like it's WORKSFORME.

https://treeherder.mozilla.org/perf.html#/graphs?series=%5Bmozilla-central,f22a87e9898beb0c7dc5fefec8267c3a9ad89a8b,1,1%5D&series=%5Bmozilla-central,3bfe93820de5fd84b3a3d997670b1689a9a70839,1,1%5D&series=%5Bmozilla-central,197ee23fabf38297ead2dbbf086ececa2552de87,1,1%5D&series=%5Bmozilla-central,2f31eebc1e873e4111c12c1ff2aa9154d2d8b0be,1,1%5D

Comment 3

[Gregory Pappas [:gregp]]

Closing per comment 2