Re: [webrtc-pc] scaleResolutionDownBy's odd-resolution requirement trips up HW encoders (#2802)

There are more reasons why the send resolution could be decimated than just scaleResolutionDownBy. fmtp parameters like max-fs, or cpu overuse detection.

The reason I rounded down to the closest alignment in Firefox was mainly to avoid a regression, and we apply all of the constraints above in the same place, meaning this change was simplest.

We could round up instead. But should the spec mention other reasons the resolution could get affected? Imagine a width 700, scaleDown 2.0, alignment 16, and max-fs being lower than what the calculated width 352 leads to.

> Would providing a maximum (smaller) resolution instead be better? If so, we'd need to address what happens if realignment drops one dimension to 0.

Regardless of approach for rounding to alignment, even regardless of scaleResolutionDownBy, the source resolution may be lower than the alignment requirement and that would have to be handled similarly.

GitHub Notification of comment by Pehrsons
Please view or discuss this issue at using your GitHub account

Sent via github-notify-ml as configured in

Received on Wednesday, 14 December 2022 08:29:55 UTC