Some days ago the VideoLAN and Ffmpeg communities unveiled the release of the new version of the library dav1d 0.6.0. This is an alternative free decoder implementation of the AV1 video encoding format.
The dav1d library supports all the features of AV1, including advanced downsampling types and all color depth control parameters declared in the specification (8, 10 and 12 bit). The library was tested on a large collection of files in AV1 format.
About the dav1d decoder
The video codec AV1 was developed by Open Media Alliance. (AOMedia), in which companies such as Mozilla, Google, Microsoft, Intel, ARM, NVIDIA, IBM, Cisco, Amazon, Netflix, AMD, VideoLAN, CCN and Realtek are represented
AV1 is positioned as a free access video encoding format that does not require payment of fees, which is notably superior to H.264 and VP9 in terms of compression.
The reference decoder for AV1 is great, but it's a research codebase, so it has a lot to improve. That is why the VideoLAN, VLC and FFmpeg communities began work on a new decoder, sponsored by the Alliance of Open Media, to create the reference optimized decoder for AV1.
A key feature of dav1d is its focus on achieving performance decoding rate possible and ensure high-quality multi-threaded operation.
The library's work was tested on a large collection of files in AV1 format. The key feature of dav1d is its focus on achieving the highest possible performance decoding and ensure high-quality work in a multithreaded mode.
The goal of this new decoder is:
- To be small
- be as fast as possible
- Provide cross-platform support
- Correctly threaded,
- Free and (actually) open source.
The dav1d project code is written in programming language C(C99) and it also has assembler inserts (NASM / GAS) and is distributed under the BSD license. The decoder counts with implemented support for x86, x86_64, ARMv7 and ARMv8 architectures and Linux, Windows, macOS, Android, and iOS operating systems.
Main new features of dav1d 0.6.0
This new version of the decoder dav1d 0.6.0 gets to fix some bugs that were present in the previous version, plus the developers have implemented optimizations specific ARM64 architecture They cover many operations when working with 10 and 12 bit color depths.
Also highlights the work that was done to add optimizations based on SSSE3 instructions for digital noise reduction as well as instruction-based optimizations AVX2 for the msac_adapt16 operation.
Other optimizations implemented in this new edition are the loop, cdef, and msac restore operations enhancements for ARM64 and also improved AVX2 optimizations for cdef_filter.
On the other hand, it is mentioned in the announcement that the developers worked on adding optimizations based on AVX-512 instructions for the prep_bilin, prep_8tap, cdef_filter and mc_avg / w_avg / mask operations.
For the part of the corrections it is mentioned that fixed rare discrepancies in behavior with the decoder reference AV1 and that an improvement was implemented in the itxfm and cdef_filter operations in C.
Finally if you want to know more about it about this new version and also about the project of this decoder, you can check the details In the following link.
How to install dav1d decoder on Ubuntu and derivatives?
For those who are interested in being able to install this decoder in their systems, they can do so by following the instructions we share below.
Generally for most Linux distributions, the guys from the VideoLan project, offer the decoder package by means of a Snap package.
So to install it by this means, it only requires that your distribution has support for this type of package.
In a terminal they just have to execute the following command:
sudo snap install dav1d --edge