[11:59] <mchehab> hi all
[11:59] <mchehab> (today I'm a bit short of time... I have just ~20 mins for discussions)
[11:59] <hverkuil> hi!
[12:00] <mchehab> btw, I won't be able to join our meeting next week... will be on a biztrip without Internet access
[12:01] <mchehab> hverkuil: I still have some patches from one of your PRs on my queue (the  ones after that /4 patch series we've discussed yesterday)
[12:01] <mchehab> I intend to handle it later today
[12:03] <mchehab> hverkuil, pinchartl, sailus, syoung: anything for today's discussion?
[12:03] <mchehab> syoung: I also intend to handle your DVB patch series today
[12:04] <hverkuil> ok. I think it's best to just drop the /4 series and apply everything else from that PR.
[12:05] <hverkuil> I'll likely make a new PR tomorrow for the staging rkisp1 driver.
[12:05] <hverkuil> v13 seems to be the lucky version :-)
[12:05] <mchehab> :-)
[12:06] <mchehab> as I won't have internet access next week, I hope not many PRs will arrive...
[12:06] <mchehab> as otherwise they may be applied only for 5.7
[12:07] <hverkuil> I get try to push that PR out today, if you prefer?
[12:07] <mchehab> (I'll likely apply  pending PRs on Jan, 20 - but if there are too many things, some stuff may end being postponed)
[12:07] <mchehab> tomorrow should be fine, if you can do it during the morning
[12:08] <hverkuil> are you back in Brazil, or still in Germany?
[12:08] <mchehab> still in DE
[12:09] <mchehab> anything else for today's meeting?
[12:09] <hverkuil> Not from me.
[12:10] <mchehab> ok. if someone pops up, I'll read the backlog and reply after lunch.]
[14:40] <pinchartl> hverkuil: regarding the rkisp1 driver
[14:40] <pinchartl> what's your opinion on how they handle ISP parameters ?
[14:40] <pinchartl> they're put in a buffer queued to a parameters queue
[14:41] <pinchartl> but it's then entirely consumed by the driver to write registers, the buffer isn't passed to the hardware
[14:47] <hverkuil> It's a valid choice, but not necessarily the best choice. I'd have chosen compound controls for this.
[14:47] <hverkuil> The only advantage of using a buffer is that can do zero-copy, whereas a control always involves a memcpy.
[14:48] <pinchartl> there are performance concerns, and the fact that we have no format support in the request API
[14:51] <hverkuil> Performance concerns with what? Using buffers or using controls?
[16:01] <mchehab> I suspect that the uapi for the driver need improvements... as it is now at staging, I'm blindly accepting it
[16:02] <mchehab> but I want to understand and see it properly documented before moving out of staging
[16:02] <mchehab> anyway, the pending PRs were applied
[16:16] <pinchartl> hverkuil: with using controls
[16:17] <pinchartl> mchehab: agreed, for staging it's fine