1061885 - [Skia] Do not demote for non-standard compositing ops

Status: NEW

(Core :: Graphics: Canvas2D, defect)

Description

[George Wright (:gw280) (needinfo me!)]

We currently demote to a software canvas if we detect a "non-standard" compositing operator is used. This is because at the time, the GPU backend for Skia didn't support these ops very well.

We should see if this is still an issue, and if so, fix it.

https://tbpl.mozilla.org/?tree=Try&rev=ff8322348d04

Comment 1

[George Wright (:gw280) (needinfo me!)]

Attachment: remove-compositing-fallback.patch

Looks like try was green. Go ahead and land?

Comment 2

[James Willcox (:snorp) ([e-mail]) (he/him)]

Comment on attachment 8483606 [details] [diff] [review]
remove-compositing-fallback.patch

Review of attachment 8483606 [details] [diff] [review]:
-----------------------------------------------------------------

Cool. I wonder if we're actually testing these with SkiaGL, though. That got disabled for a lot of cases, didn't it? Can you try to smoke test it with an actual example?

Comment 3

[Milan Sreckovic [:milan] (needinfo for best results)]

Right - we probably want to give QA heads up on this, as it may need a lot of manual testing on different devices.  We're removing the "demotion" to software, which means we may get more overall memory usage, better performance... with this change?

Comment 4

[George Wright (:gw280) (needinfo me!)]

Oh good point. I will test it locally and see what happens.

Milan: yep, possibly more overall memory usage, but no spike when we demote to software. Better performance definitely.

Comment 5

[George Wright (:gw280) (needinfo me!)]

http://codepen.io/adobe/full/nmfic/ looks fine on my device. Let's land this.

Comment 6

[George Wright (:gw280) (needinfo me!)]

(also: buttery smooth!)

Comment 7

[George Wright (:gw280) (needinfo me!)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/0fc176518638