1254877 - Compare basic compositor video performance with D3D11 on WARP

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Jeff Muizelaar [:jrmuizel]]

We should know if we're doing better or worse.

Comment 1

[Jeff Muizelaar [:jrmuizel]]

Ethan, can you remeasure with a current Nightly and post the results here?

Comment 2

[Ethan Lin[:ethlin]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #1)
> Ethan, can you remeasure with a current Nightly and post the results here?

No problem, I am on a day off today and I will update the results when I go to the office.

Comment 3

[Ethan Lin[:ethlin]]

I update the result I got yesterday. I will do more tests on different devices and videos with latest Nightly when I go to the office.

Test video:
  1920x1080@60fps (https://www.youtube.com/watch?v=QzVGf-K2heg)

Firefox:
  nightly 48.0a1 (2016-03-14)

PC info:
  i7-3770, nVidia GT630

Result:
   Basic compositor  -   60 fps
   Warp              -   30 fps

Comment 4

[Ethan Lin[:ethlin]]

I did test on PC & laptop with different video.

Firefox Nightly 48.0a1 (2016-03-16) 

Device Info:
  PC     - i7-3770,  nVidia GT630, 1920x1080
  Laptop - i7-2640M, nVidia 4200M, 1600x900 

Video1:
  1920x1080@60fps
  https://www.youtube.com/watch?v=QzVGf-K2heg

Video2:
  3840x2160@30fps
  https://www.youtube.com/watch?v=I7CycuMpQ3w

Video3
  3840x2160@60fps
  https://www.youtube.com/watch?v=aqz-KE-bpKQ

Result:
+--------+-----------+---------+--------------+-------------+
|        | PC Basic  | PC Warp | Laptop Basic | Laptop Warp |
+-----------------------------------------------------------+
| video1 | 60        | 30      | 50           | 30(*)       |
+-----------------------------------------------------------+
| video2 | 30        | 30      | 15           | 15          |
+-----------------------------------------------------------+
| video3 | 25        | 30(*)   | 15           | 15(*)       |
+--------+-----------+---------+--------------+-------------+

The (*) means the video intermittently paused when playing.

Comment 5

[Ethan Lin[:ethlin]]

By the result in comment 4, we can see that the performance of basic compositor is generally better than d3d11 warp. And there is a video intermittent suspension problem with d3d11 warp.
Jeff, is there anything you want to do further research?

Comment 6

[Jeff Muizelaar [:jrmuizel]]

Nope that looks good to me.

Comment 7

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

Anthony, this would be useful to do on a larger array of devices, video resolutions, Windows versions.  It has to be a new enough nightly, but if you can put together a plan, we could run Betabreakers as well as OneAndDone.

Comment 8

[u279076]

(In reply to Milan Sreckovic [:milan] from comment #7)
> Anthony, this would be useful to do on a larger array of devices, video
> resolutions, Windows versions.  It has to be a new enough nightly, but if
> you can put together a plan, we could run Betabreakers as well as OneAndDone.

I will work with Ethan off this bug report to get something set up.

Comment 9

[Ethan Lin[:ethlin]]

It should be done, so I close the bug.