User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Firefox/68.0

Steps to reproduce:

1. Start Firefox with a clean profile.
2. Play an AV1 video from yt.
3. Disable the RDD process, restart Firefox and repeat 2.

Observe the CPU usage in both steps 2 and 3.

Actual results:

When RDD is disabled, I can see almost the same CPU usage as VP9 for the same content (720p@24fps yt video). Observing the system monitor, I'm seeing ~5% for the Web content process regardless of the status of the RDD process. What changes when RDD is enabled is that, in addition to that ~5%, there is another ~11% used by RDD. The average total CPU usage rises from ~10% to ~25% (laptop with 4cores 8threads).

Expected results:

I would have expected AV1 playback to use about the same CPU regardless of the RDD process being enabled or disabled. Not sure if that's a valid expectation, but I thought it was better to file a bug, just in case.

This doesn't seem to be a regression. I wasn't sure if a performance profile would be useful or what threads to capture so I won't include one unless it's asked, in which case, please, specify the needed threads.

Extra CPU usage is to be expected with the RDD as now everything needs to be copied 2 times more (content->rdd->content->gpu)

My guess is that the extra overhead is basically what is described in bug 1539043

We're currently using AOM to decode on RDD, not DAV1D. Alex is working this in Bug 1538474.

I think we can make this one a dup of 1538474.