1685582 - rasterflood_svg is about 1% slower with (Fission + WebRender) than (e10s + WebRender) on Linux (not Windows)

Status: RESOLVED INCOMPLETE

(Core :: Graphics: WebRender, defect, P3)

Description

[Chris Peterson [:cpeterson]]

Attachment: Perfherder_Screenshot.png

https://treeherder.mozilla.org/perfherder/graphs?highlightAlerts=1&highlightChangelogData=1&series=autoland,2812903,1,1&series=mozilla-central,2820203,1,1&series=mozilla-central,2819203,1,1&series=mozilla-central,2821000,1,1&timerange=5184000&zoom=1608928249976,1610135334739,9064.956017134888,9293.779707547674

This problem affects Linux but not Windows.

Comment 1

[Jeff Muizelaar [:jrmuizel]]

This is not expected.

Comment 2

[Chris Peterson [:cpeterson]]

Oops. This is actually just a 0.4% difference (~40 ms difference when running a ~9160 ms test), not 4%. I'll leave this bug open (with a low priority) because the difference is real and obvious in the Perfherder graphs.

Comment 3

[Neha Kochar [:neha]]

Eyeballing the graphs at https://treeherder.mozilla.org/perfherder/graphs?highlightAlerts=1&highlightChangelogData=1&series=autoland,2812903,1,1&series=mozilla-central,2820203,1,1&timerange=2592000 looks like ~2% regression with Fission.

Comment 4

[Jim Mathies [:jimm]]

Both of these are webrender, so we're not thinking this has anything to do with graphics. Also the regression is very small., so we're not prioritizing this. If the fission team would like any help, please reach out.

Comment 5

[Chris Peterson [:cpeterson]]

(In reply to Jim Mathies [:jimm] from comment #4)

Both of these are webrender, so we're not thinking this has anything to do with graphics. Also the regression is very small., so we're not prioritizing this. If the fission team would like any help, please reach out.

In that case, this bug doesn't need to block shipping Fission. Deferring to Fission Future.

At this time, I only see about a 1% regression, but it is very consistent.

Comment 6

[Glenn Watson [:gw]]

Do we still want to keep this open? What's a more appropriate component than gfx:wr for it?

Comment 7

[Chris Peterson [:cpeterson]]

(In reply to Glenn Watson [:gw] from comment #6)

Do we still want to keep this open? What's a more appropriate component than gfx:wr for it?

I'll close this bug. It's no longer relevant.