[04:01] <tfiga> sailus: speaking of documentation, do you know if there is anything coming to improve the IPU3 IMGU documentation?
[08:48] <sailus> hverkuil: Ping?
[08:48] <hverkuil> sailus: pong
[08:48] <sailus> How do you do?
[08:48] <hverkuil> fine, thank you.
[08:49] <sailus> About the stm32-dcmi driver --- there's are more pipeline elements being added to a driver that used to be a plain V4L2 driver.
[08:50] <sailus> AFAIR there's a CSI-2 to parallel converter that is a separate device.
[08:51] <sailus> It's not the property of the original device though.
[08:52] <hverkuil> so?
[08:52] <sailus> That's in principle the case for all parallel bridges; if you add a CSI-2 to parallel converter, or a sensor driver that exposes more than one sub-device, MC support in the driver will be needed to support new use cases.
[08:54] <sailus> I haven't reviewed the latest set, but at least the earlier ones did not add that.
[09:02] <hverkuil> AFAICT the driver propagates the format along the pipeline, using the MC topology to discover the pipeline. It does assume a simple linear pipeline, so nothing complex.
[09:25] <sailus> v3 introduces MC support indeed.
[09:25] <sailus> The semantics of the S_FMT IOCTL isn't quite what we used to expect from MC-enabled drivers.
[09:27] <hverkuil> It's not a MC-enabled driver. Yes, it uses the MC internally (it needs the topology after all), but seen from userspace it is a regular non-MC video device.
[09:32] <sailus> Ugh.
[09:33] <sailus> I guess it's possible to do it that way.
[09:34] <hverkuil> Fully switching to an MC would mean ABI breakage, and for a simple low-end SoC like this it would be total overkill. The MC is basically used in a similar way as card info structs are for USB and PCI devices.
[12:26] <pH5> ezequielg: do you have a current ffmpeg branch corresponding to the hantro h264 v4 series?
[13:34] <ezequielg> pH5: yes, give me a sec. Hm, should've added that to the cover.
[18:40] <ezequielg> paulk-leonov: hverkuil: do you think the start code adjusts with our recent discussion?
[18:40] <ezequielg> sorry it took me so long, partially, i wanted to make sure the interface made sense, and worked on both cedrus and rockchip.
[18:48] <hverkuil> ezequielg: sorry, haven't looked at your series. It might be Monday before I get around to it.