Dav1d 0.9 Released With AVX2-Tuned 10b/12b Decode For Big Speed Boost

[brad0]

But you'll have to wait for browsers to update their fav1d version.

Chrome will be quick. Firefox will be slow AF.

[flashmozzg]

Chrome will be quick. Firefox will be slow AF.

Wouldn't know about that. FF was ok we the last few versions and it seems the [plan](https://bugzilla.mozilla.org/show_bug.cgi?id=1700452) is to do the next dav1d update in FF90 which is due in July. On the other hand, Chrome(ium) just updated to 0.8.2 in 90, so it might take a bit longer for the next update.

[Ronshere]

About


dav1d is a new open-source AV1 decoder developed by the VideoLAN and FFmpeg communities and sponsored by the Alliance for Open Media.
Goals

dav1d aims to be

• as fast as possible,
• small,
• very cross-platform.
• correctly threaded

Technical details

• Uses Meson and Ninja to build
• Written in C99
• Runs on Windows, Linux, macOS and Android
• Licensed under BSD 2-clause "Simplified" License
• As of October 2019, Dav1d is the fastest AV1 software decoder

Contributing


You can contribute to dav1d by writing C or ASM, additionally app integration and testing is wanted.

To contribute, check out the contribution document.

Most of the development discussion happens on IRC, in the #dav1d channel on Freenode.
The Code


The code can be found on the VideoLAN Gitlab:

git clone https://code.videolan.org/videolan/dav1d.gitSource


The source code tarball can be found on our ftp: dav1d tarballs.

[bug77]

Anyone else cringes when they read about 10bit encoding/decoding while at the same time realizes every video they see is banding galore?

[Azrael5]

Wouldn't know about that. FF was ok we the last few versions and it seems the [plan](https://bugzilla.mozilla.org/show_bug.cgi?id=1700452) is to do the next dav1d update in FF90 which is due in July. On the other hand, Chrome(ium) just updated to 0.8.2 in 90, so it might take a bit longer for the next update.

Is there a way to know which version of dav1d Firefox release supports? thanks

[flashmozzg]

Is there a way to know which version of dav1d Firefox release supports? thanks

Don't know if there is a way to do that from FF client, but according to bugtracker, current release (88) should have dav1d 0.8.2 (https://bugzilla.mozilla.org/show_bug.cgi?id=1688992) and the next one (89) will have 6c6d25d3: https://bugzilla.mozilla.org/show_bug.cgi?id=1694517

[usta]

just landed to manjaro linux

[brad0]

and the next one (89) will have 6c6d25d3: https://bugzilla.mozilla.org/show_bug.cgi?id=1694517

It never ceases to amaze me at how stupidly the Firefox guys do things.

[smitty3268]

It never ceases to amaze me at how stupidly the Firefox guys do things.

You're going to have to elaborate on this one.

[Toggleton]

Anyone else cringes when they read about 10bit encoding/decoding while at the same time realizes every video they see is banding galore?

Why? AV1 has 10bit in the main profile,so Hardware decoder need to support 10bit.
VP9 did not have that, 10bit was in VP9 profile 2 and not all devices did support it and so Youtube did only use 10bit for HDR content.
VVC/h266 does AFAIK go even so far and has only 10bit+ "VVC should support YCbCr 4:4:4, 4:2:2 and 4:2:0 with 10 to 16 bits per component," and i think AV2 will go the same route.

Netflix uses 10bit AV1 https://netflixtechblog.com/netflix-now-streaming-av1-on-android-d5264a515202 10bit on arm64 is nearly as fast as 8bit(Software decoding/dav1d).

Now that software decoding of 10bit AV1 will get more assembly on x86(so far only AVX2 but there is a SSE2 merge request in the dav1d repo) is it more likely that services like youtube could use it too.
As long as the source has not the banding already would that help to get rid of a big chunk of it.