Closed
Bug 1352442
Opened 7 years ago
Closed 7 years ago
Do not needlessly clear the compositor background before drawing
Categories
(Core :: Graphics: Layers, enhancement)
Core
Graphics: Layers
Tracking
()
Tracking | Status | |
---|---|---|
firefox55 | --- | fixed |
People
(Reporter: bas.schouten, Assigned: bas.schouten)
References
Details
Attachments
(1 file)
Right now we always clear the invalid region of the backbuffer in CompositorD3D11::BeginFrame, there's not a real good reason to do that when we're going to overwrite that with new content. We pass an opaque region to BeginFrame already and can use that to restrict the area that we clear.
Comment hidden (mozreview-request) |
Comment 2•7 years ago
|
||
mozreview-review |
Comment on attachment 8853446 [details] Bug 1352442: Only clear transparent regions of the back buffer on BeginFrame. https://reviewboard.mozilla.org/r/125536/#review128122
Attachment #8853446 -
Flags: review?(matt.woodrow) → review+
Pushed by bschouten@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/58181c34aa51 Only clear transparent regions of the back buffer on BeginFrame. r=mattwoodrow
Comment 4•7 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/58181c34aa51
Status: ASSIGNED → RESOLVED
Closed: 7 years ago
status-firefox55:
--- → fixed
Resolution: --- → FIXED
Target Milestone: --- → mozilla55
Comment 5•7 years ago
|
||
it looks like this patch had a huge win on performance according to talos: == Change summary for alert #5801 (as of March 31 2017 18:51 UTC) == Improvements: 9% tscrollx summary windows8-64 pgo 2.73 -> 2.49 8% tscrollx summary windows7-32 pgo 2.76 -> 2.53 7% tp5o_scroll summary windows8-64 pgo e10s1.99 -> 1.85 7% tp5o_scroll summary windows7-32 opt e10s2.41 -> 2.23 7% tp5o_scroll summary windows8-64 opt e10s2.02 -> 1.88 6% tscrollx summary windows8-64 pgo e10s 2.46 -> 2.3 6% tscrollx summary windows8-64 opt e10s 2.43 -> 2.29 6% glterrain summary windows7-32 opt e10s5.18 -> 4.87 5% tpaint linux64 pgo 239.09 -> 226.9 3% tp5o_scroll summary windows7-32 pgo 2.54 -> 2.46 3% tpaint linux64 opt 286.64 -> 277.97 2% damp summary linux64 pgo 283.87 -> 277.88 For up to date results, see: https://treeherder.mozilla.org/perf.html#/alerts?id=5801
Comment 6•7 years ago
|
||
Is this something we'd consider backporting to 54 given that it fixed at least one known correctness issue, or should we let it ride the 55 train?
Flags: needinfo?(bas)
Assignee | ||
Comment 7•7 years ago
|
||
(In reply to Joel Maher ( :jmaher) from comment #5) > it looks like this patch had a huge win on performance according to talos: > == Change summary for alert #5801 (as of March 31 2017 18:51 UTC) == > > Improvements: > > 9% tscrollx summary windows8-64 pgo 2.73 -> 2.49 > 8% tscrollx summary windows7-32 pgo 2.76 -> 2.53 > 7% tp5o_scroll summary windows8-64 pgo e10s1.99 -> 1.85 > 7% tp5o_scroll summary windows7-32 opt e10s2.41 -> 2.23 > 7% tp5o_scroll summary windows8-64 opt e10s2.02 -> 1.88 > 6% tscrollx summary windows8-64 pgo e10s 2.46 -> 2.3 > 6% tscrollx summary windows8-64 opt e10s 2.43 -> 2.29 > 6% glterrain summary windows7-32 opt e10s5.18 -> 4.87 > 5% tpaint linux64 pgo 239.09 -> 226.9 > 3% tp5o_scroll summary windows7-32 pgo 2.54 -> 2.46 > 3% tpaint linux64 opt 286.64 -> 277.97 > 2% damp summary linux64 pgo 283.87 -> 277.88 > > For up to date results, see: > https://treeherder.mozilla.org/perf.html#/alerts?id=5801 The linux ones must be flukes. This patch has no affect on Linux. The Windows improvements are believeable.
Assignee | ||
Comment 8•7 years ago
|
||
(In reply to [Out of Office Until 24-April] Ryan VanderMeulen [:RyanVM] from comment #6) > Is this something we'd consider backporting to 54 given that it fixed at > least one known correctness issue, or should we let it ride the 55 train? This should be a low-risk patch to backport. I'll leave it up to release folks to decide if the risk and disruption of any backport is justified at this point in the cycle.
Flags: needinfo?(bas)
Updated•7 years ago
|
Whiteboard: [qf:p1]
Updated•2 years ago
|
Performance Impact: --- → P1
Whiteboard: [qf:p1]
You need to log in
before you can comment on or make changes to this bug.
Description
•