Cleanup CONTENT_FRAME_TIME recording and try to record the same data for WR and non-WR where possible

RESOLVED FIXED in Firefox 66

Status

()

enhancement
RESOLVED FIXED
5 months ago
5 months ago

People

(Reporter: mattwoodrow, Assigned: mattwoodrow)

Tracking

Trunk
mozilla66
Points:
---

Firefox Tracking Flags

(firefox66 fixed)

Details

Attachments

(4 attachments)

Assignee

Description

5 months ago
We currently implement all the code for CFT twice, and do slightly different things for each.

We should share code, and try to do the same as much as possible.
Assignee

Comment 1

5 months ago
This shouldn't change any behaviour, just puts the code into a separate function so that it's easier to follow.
Assignee

Comment 2

5 months ago
This also results in us getting the CONTENT_FRAME_TIME profiler markers added for non-WR.

Depends on D15794
Assignee

Comment 3

5 months ago
We're switching to prefering the vsync version, so making these diagnostic equivalents use that timing should be more useful.
I think leaving the profiler marker (which uses ms, not vsync%) showing just the graphics component still makes sense.

Depends on D15795
Assignee

Comment 4

5 months ago
This will let us get the MissedCompositeLow/Mid/Long variants in the CFT_REASON telemetry output.

Depends on D15796

Comment 5

5 months ago
Pushed by mwoodrow@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/39ab423f06ef
Move CONTENT_FRAME_TIME calculation code into CompositorBridgeParent. r=jrmuizel
https://hg.mozilla.org/integration/autoland/rev/dad482c7ff92
Use shared CONTENT_FRAME_TIME calculation for non-WebRender. r=jrmuizel
https://hg.mozilla.org/integration/autoland/rev/89f55d26acaf
Record SVG and UPLOAD specific variants of CONTENT_FRAME_TIME using the vsync based timings. r=jrmuizel
https://hg.mozilla.org/integration/autoland/rev/65266c77b815
Pass FULL_PAINT_TIME value to CONTENT_FRAME_TIME calculation for non-WR. r=jrmuizel
You need to log in before you can comment on or make changes to this bug.