Matt Woodrow (:mattwoodrow) Reporter
	Description • 6 years ago

We currently record the start timestamp for CONTENT_FULL_PAINT_TIME at the start of the LayerManager transaction, which makes sense as a paint metric, but doesn't capture layout/style flushes and rAF.

The time that we don't currently record does contribute to whether we miss vsync for the composite, and affects CONTENT_FRAME_TIME (which records from the refresh driver tick start).

It's currently hard to know what values of CONTENT_FULL_PAINT_TIME contribute to a missed vsync, as we have to guess how long we might have spent doing layout etc.

We could also consider recording both metrics using the actual vsync timestamp, to account for cases where the refresh driver ticked late, though that should matter less.

What do you think Jeff?

	Jeff Muizelaar [:jrmuizel]
	Comment 1 • 6 years ago

I think I'd prefer to have both metrics. It's nice having CONTENT_FULL_PAINT_TIME being just an extended version of CONTENT_PAINT_TIME and I'd rather not lose that. I do agree having the full time from refresh driver start will be helpful for interpreting other times.