Hi all,
Here is my initial stab at a agenda for the media summit. As always, it
is subject to change and all times are guesstimates!
----------------------------------------------------------------------
v2:
- Reserve 45 minutes for Sakari's topic.
- Update "Subsystem development process" topic.
----------------------------------------------------------------------
8:45-9:15: get settled :-)
9:15-9:25: Hans: Quick introduction
9:25-10:15: Daniel Almeida: Rust V4L2 support
Discuss the potential blockers as well as the roadmap and priorities for
bindings. Present what is currently supported and what isn't in the Rust
code for V4L2. Identify potential candidates for new drivers that can be
written in Rust. Discuss about maintainership issues related to Rust in V4L2.
10:15-10:30: Break
10:30-11:15 Hans & Hsia-Jun Li: "Add extended fmt and buffer" patch series
https://patchwork.linuxtv.org/project/linux-media/cover/20230206043308.28365-1-ayaka@xxxxxxxxxxx/
We've been postponing the work on this, but I think we need to decide how to
proceed since pixel formats and single vs multi planar is getting to be a nightmare.
More details from Hsia-Jun Li:
1. v4l2 header would only maintain the codec format and pixel format in bus.
2. the pixel formats would be maintained by the DirectRender, those M
variant would not be supported in the new extend pixel format API.
3. The number of plane for a pixel format would also responds for its data
layout. Ex. NV12 = 2 planes(luma, chroma), I420 = 3 planes(Y, U, V).
4. Userspace that supports new extend API could access those driver didn't
adapt the new API, kernel would have a backward compatible layer. While
the opposite backward compatible is not offered (old API userspace can't
access the driver support the new API).
[optional part]
5. An alloc flag would be introduced for allocating those M variant buf.
https://lore.kernel.org/lkml/20230322105226.122467-1-randy.li@xxxxxxxxxxxxx/
6. Stateless codec format would be a modifier to the stateful codec format.
We could support different packing mode here.
11:15-11:30: Break
11:30-12:15: Andrzej Pietrasiewicz: Stateless Encoders: VP8
- Introduction to stateless encoders and previous work in this area
https://github.com/bootlin/linux/tree/hantro/h264-encoding-v5.11
https://lore.kernel.org/linux-arm-kernel/20230309125651.23911-1-andrzej.p@xxxxxxxxxxxxx/T/
- High level decisions
- Rate control
- VP8 uAPI
- Challenges
12:15-13:30: lunch
13:30-14:00: Ricardo Ribalda Delgado: KCAM update
I would like to share the progress on KCAM from our end.
I am also trying to collect all the requirements that we are getting
from vendors regarding V4L2.
14:00-14:45: Sakari: generic line-based metadata formats:
https://lore.kernel.org/linux-media/20230505215257.60704-1-sakari.ailus@xxxxxxxxxxxxxxx/T/#t
14:45-??:??: Mauro & sub-maintainers: Subsystem development process
How to improve it? How do we get more maintainers? It's not working very
well at the moment.
Mauro will present how the media CI works today.
Hans will present how the daily build scripts work.
Can this be harmonized?
Final short topic: Hans: Should we keep the media summit together with the ELCE conference?
Or organize it separately, and ask a company like Cisco, Intel or Collabora
to organize a room. What would work best? Anything to improve on this?
We have the room until 18:30, but (fingers crossed) I hope to finish before that.
The "Add extended fmt and buffer" and the "Subsystem development process" are
*very* hard to predict. I want to limit the initial discussion of the that first
topic to 45-60 minutes, and continue (if needed) in the afternoon after the discussion
about the subsystem development process. We will have see how it goes.
Please reply with corrections, questions, etc. to this email. I'll update the agenda
over time.
Regards,
Hans
