Closed Bug 1262885 Opened 9 years ago Closed 9 years ago

Investigate performance impact of disabling e10s when d3d9 is enabled

Categories

(Core :: Graphics, defect, P4)

defect

Tracking

()

RESOLVED FIXED
Tracking Status
e10s + ---

People

(Reporter: gw280, Unassigned)

Details

(Whiteboard: [gfx-noted])

Follow up from bug 1237769 As per Jeff's comment: "Milan: can we compare performance of two configurations of Firefox: A) d3d9 is disabled, e10s is enabled B) d3d9 is enabled, e10s is disabled I want to know what performance trade-offs there are, if any."
tracking-e10s: --- → +
Priority: -- → P4
Right - not the highest urgency, as we know we're not regressing anything for the XP users at this point.
Whiteboard: [gfx-noted]
here is a compare view of data for the disabling and the push prior: https://treeherder.mozilla.org/perf.html#/compare?originalProject=mozilla-inbound&originalRevision=453f314bd7ae&newProject=mozilla-inbound&newRevision=a3f8a4e83cdd&framework=1&filter=xp this is pretty much in-line with the alerts we recieved: https://treeherder.mozilla.org/perf.html#/alerts?id=768 as this is just disabling e10s mode on certain OS/hardware configs, then we are not really seeing regressions. maybe we need to ask ourselves how we can test e10s with d3d9 on windows xp? or maybe this is a good sanity check for us? with no d3d9, then e10s tests on windowsxp seems pointless.
Joel, do the talos test machines get d3d9 in non-e10s? Are we comparing non-d3d9/e10s to d3d9/non-e10s or non-d3d9/e10s to non-d3d9/non-e10s here?
the regression is non-d3d9/e10s <-> non-d3d9/non-e10s (the patch) for non-e10s mode, I don't believe we have d3d9 available on the winxp machines currently.
odd, this has made tart bi-modal on winxp e10s
with the landing of bug 1262954, all our perf changes here are reversed. I would prefer to mark this as fixed.
Works for me. We can reopen if/when it changes.
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.