1261273 - MSE with webm/vp9 should be automatically enabled on platform with VP9 hardware decoder

Status: NEW

(Core :: Audio/Video: Playback, enhancement, P3)

Description

[Jean-Yves Avenard [:jya]]

We currently override the media.mediasource.webm.enabled value if:
1- mp4/h264 isn't available
2- h264 isn't hardware accelerated
3- We can decode vp9 quickly enough.

We should also enable it on machines that have hardware VP9 capabilities (some haswell and later intel, or android devices)

Comment 2

[Anthony Jones (:ajones, :kentuckyfriedtakahe, :k17e)]

Mass change P2 -> P3

Comment 3

[Anthony Jones (:ajones, :kentuckyfriedtakahe, :k17e)]

Jean-Yves - is this still an issue?

Comment 4

[Jean-Yves Avenard [:jya]]

yes it is... it should be done. The enabling of MediaCapabilities however may alleviate the problem, as following bug 1409664, will be able to properly detect if hardware decoding is enabled or not

Removing the dependency on 1399019, as facebook doesn't use vp9 but h264.

Comment 5

[theod]

@jya Actually, this is blocking bug 1399019 in part, since *some* Facebook videos do play vp9 :) Not all videos have vp9 encodings, but some do. Specifically, most 360 videos do play VP9 by default on Desktop. I'm on the 360 videos team @ FB and we just disabled VP9 delivery on Firefox Windows after discovering it wouldn't play (had previously disabled it for MAC only following bug 1403412, but we did not know that it was now disabled on Windows as well.

Comment 6

[Alastor Wu [:alwu]]

For the reference, Intel started fully supporting VP9 hardware decoding since the version 7 (Ice Lake) [1], and Nvidia started from GP102 [2]. We've already forced to use VP9 on Android if it has hardware support [3], so this bug is for other platfroms.

[1]
https://en.m.wikipedia.org/wiki/Intel_Quick_Sync_Video
https://software.intel.com/content/www/us/en/develop/articles/encode-and-decode-capabilities-for-7th-generation-intel-core-processors-and-newer.html
[2] https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new
[3] https://searchfox.org/mozilla-central/rev/4e87b5392eafe1f1d49017e76f7317b06ec0b1d8/dom/media/mediasource/MediaSource.cpp#82-83

Comment 7

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

For reference we may already be doing this on non-Android platfoms[0].

[0] https://searchfox.org/mozilla-central/rev/4e87b5392eafe1f1d49017e76f7317b06ec0b1d8/modules/libpref/init/StaticPrefList.yaml#7296

Comment 8

[Alastor Wu [:alwu]]

Yes, but currently we would override the preference here [1], so this bug is used to force enabling VP9 for those desktops which have ability to decode VP9 in hardware.

[1] https://searchfox.org/mozilla-central/rev/4e87b5392eafe1f1d49017e76f7317b06ec0b1d8/dom/media/mediasource/MediaSource.cpp#128-129

Comment 9

[Alastor Wu [:alwu]]

Matt, I wonder if we can query this information from the graphic in current Firefox? Although I found this [1], it seems only to be used to block certain mobile devices. If not, is that a good place to extend the support and non-support list? Do we have any similar check that I can take a reference?
Thank you so much.

[1] https://searchfox.org/mozilla-central/rev/4e87b5392eafe1f1d49017e76f7317b06ec0b1d8/widget/android/GfxInfo.cpp#581-595

Comment 10

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

(In reply to Alastor Wu [:alwu] from comment #8)

Yes, but currently we would override the preference here [1], so this bug is used to force enabling VP9 for those desktops which have ability to decode VP9 in hardware.

[1] https://searchfox.org/mozilla-central/rev/4e87b5392eafe1f1d49017e76f7317b06ec0b1d8/dom/media/mediasource/MediaSource.cpp#128-129

Ah, thanks. Makes me wonder if we should even bother with the pref. I recall looking at similar cases around webm, and having a pref to toggle functionality that doesn't actually toggle it seems like a foot gun.

Comment 11

[Jean-Yves Avenard [:jya]]

Afaik, vp9 is now always enabled now that we have media capabilities enabled