[00:31] *** BrianG61UK has joined #linux-media [00:32] *** BrianG61UK_ has joined #linux-media [00:33] *** camus has joined #linux-media [00:36] *** camus has quit IRC (Remote host closed the connection) [00:37] *** camus1 has quit IRC (Ping timeout: 480 seconds) [00:37] *** camus has joined #linux-media [00:39] *** camus has quit IRC (Remote host closed the connection) [00:40] *** camus has joined #linux-media [01:40] *** Inky1003 has quit IRC (Remote host closed the connection) [02:05] *** camus1 has joined #linux-media [02:07] *** camus has quit IRC (Ping timeout: 480 seconds) [02:59] *** camus1 has quit IRC (Remote host closed the connection) [03:00] *** camus has joined #linux-media [04:12] *** miqztee has joined #linux-media [04:29] *** camus1 has joined #linux-media [04:34] *** camus has quit IRC (Ping timeout: 480 seconds) [05:06] *** jm_h has joined #linux-media [05:20] *** NiksDev has joined #linux-media [05:43] *** svarbanov has joined #linux-media [07:19] *** ao2 has joined #linux-media [07:29] *** GBenji has joined #linux-media [08:20] *** miqztee has quit IRC (Quit: Leaving) [08:58] *** djrscally has joined #linux-media [09:18] *** camus has joined #linux-media [09:23] *** camus1 has quit IRC (Ping timeout: 480 seconds) [09:51] <hverkuil> ezequielg: ping [11:34] <pinchartl> hverkuil: ping [11:35] <hverkuil> pinchartl: pong [11:37] <pinchartl> I had a question for you the other day, about the control framework [11:37] <pinchartl> did you see it ? [11:37] <hverkuil> No, I didn't. It probably wasn't logged either since I was doing maintenance to my server. [11:38] <pinchartl> ok [11:38] <pinchartl> so in a nutsheel [11:38] <pinchartl> nutshell [11:38] <pinchartl> I have two controls [11:38] <pinchartl> V4L2_CID_VBLANK and V4L2_CID_EXPOSURE [11:38] <pinchartl> the EXPOSURE limits depend on the VBLANK value [11:39] <pinchartl> as sensors require EXPOSURE < total height (height + VBLANK) - sensor-specific margin [11:39] <pinchartl> changing VBLANK changes the limits of EXPOSURE [11:39] <pinchartl> if I set both controls in a single VIDIOC_S_EXT_CTRLS [11:40] <pinchartl> the value of EXPOSURE gets clamped to the current limits, not the limits corresponding to the new VBLANK value [11:40] <pinchartl> is that fixable in the control framework ? [11:40] <hverkuil> I now remember reading about this, I think you asked this last week (or even longer ago) as well, right? [11:40] <hverkuil> Sorry, it's been very busy lately. [11:41] <pinchartl> no worries [11:45] <hverkuil> Actually, I think you can do this today already. First of all the two controls have to be in the same control cluster: they depend on one another, so they have to be validated atomically. [11:46] <hverkuil> Secondly, you can create your own type_ops. If you don't specify type_ops, then std_type_ops (defined in v4l2-ctrls-core.c) is used, but you can supply your own. [11:47] <pinchartl> ok [11:47] <hverkuil> Unfortunately, it looks like the four standard ops are marked static in v4l2-ctrls-core.c, it would be really nice if that can be changed so that in your driver you can still use std_equal, std_init and std_log, and only need to supply your own validate. [11:48] <hverkuil> But that would be a simple change, of course. [11:48] <pinchartl> not very nice to implement in sensor drivers, but doable [11:48] <hverkuil> Is this a one-off, or will this be true for all sensor drivers? [11:48] <pinchartl> v4l2_subdev is really the wrong level of abstraction for camera sensors, there's way too much boilerplate code [11:48] <pinchartl> it should be true for all sensor drivers [11:50] <hverkuil> OK, then a standard sensor_type_ops helper might be best, you can just use that. [11:51] <hverkuil> You may have to experiment a bit to see what works best. [11:52] <pinchartl> thanks for your feedback [11:52] <hverkuil> I never tried something like this before :-) [11:56] <pinchartl> you should work more in userspace, it's impressive how many kernel shortcomings one can uncover :-) [11:56] <pinchartl> the vblank control, for instance, is absolutely awful to work with [11:56] <pinchartl> most sensors have a maximum total vertical size, it's a fixed value [11:57] <pinchartl> to the maximum vblank value depends on the height of the configured format [11:57] <pinchartl> s/to/so/ [12:15] <hverkuil> Not my control :-) [12:19] *** svarbanov has quit IRC (Ping timeout: 480 seconds) [12:22] *** camus1 has joined #linux-media [12:26] *** camus has quit IRC (Ping timeout: 480 seconds) [12:32] *** svarbanov has joined #linux-media [13:01] *** camus1 has quit IRC (Remote host closed the connection) [13:06] *** camus1 has joined #linux-media [13:11] *** NiksDev has quit IRC (Remote host closed the connection) [13:11] *** NiksDev has joined #linux-media [14:13] *** b-rad has quit IRC (Remote host closed the connection) [14:48] *** camus1 has quit IRC (Remote host closed the connection) [14:48] *** camus has joined #linux-media [15:11] *** ao2 has quit IRC (Quit: Leaving) [15:20] <ndufresne> pinchartl: not so easy to figure-out that V4L2_PIX_FMT_NV12MT_16X16 and V4L2_PIX_FMT_HM12 are identical pixel format, but one is mplane, the other is single allocation [15:20] <ndufresne> I clearly missed that one, why didn't you follow the existing naming, and just call it V4L2_PIX_FMT_NV12T_16X16 [15:24] <ezequielg> hverkuil: ping [15:24] <ezequielg> i mean, pong... pang, peng. [15:25] <ezequielg> sorry for the delay :) [15:25] <pinchartl> ndufresne: did I add those formats ? I don't recall :-) [15:25] <ezequielg> fwiw, it's never too late, uh? i think we can improve documentation a bit. [15:26] <hverkuil> ezequielg: simple question: just verifying that '[PATCH v2 00/10] hantro: Enable H.264 VDPU2' is indeed superseding '[PATCH 00/12] hantro: Enable H.264 VDPU2 (Odroid Advance Go)'? [15:26] <ezequielg> yes. [15:27] <ezequielg> sorry, forgot to mark as superseeding. [15:27] <hverkuil> OK, thanks. Marked as such in patchwork. [15:27] <hverkuil> I need to start doing patch reviews again now that rc1 is merged... [15:27] <ezequielg> speaking of Hantro, as I'm typing I'm testing Andrzej's VP9 Hantro G2. So looks like we may land VP9 veeery soon. [15:28] <hverkuil> nice [15:55] *** camus has quit IRC (Remote host closed the connection) [15:56] *** camus has joined #linux-media [16:40] *** gouchi has joined #linux-media [16:41] *** NiksDev2 has joined #linux-media [16:42] *** NiksDev has quit IRC (Remote host closed the connection) [16:51] *** ao2 has joined #linux-media [16:59] *** NiksDev2 has quit IRC (Ping timeout: 480 seconds) [17:26] *** camus has quit IRC (Read error: Connection reset by peer) [17:27] *** camus has joined #linux-media [17:46] *** svarbanov has quit IRC (Ping timeout: 480 seconds) [17:55] *** NiksDev has joined #linux-media [17:59] *** GBenji has left [18:00] <ndufresne> pinchartl: ah, I see, you simply moved it, who added that then ... [18:01] <ndufresne> it predates 2012, when the entire uapi got reworked massively [18:03] <ndufresne> ezequielg: funny, doc for cx2341x explains the format, HM stands for Hauppauge Macroblock, a miss-nommer for 'Conexant Macroblock' [18:04] <ndufresne> and these card have MPEG2 decoders, something H264 decoders in it [18:04] <ezequielg> who knows if it's all the same layout. [18:04] <ezequielg> it's even more funny than a sunxi driver declares that form. [18:04] <ndufresne> who knows if it's the same codec lol [18:04] <ezequielg> *pixelformat [18:04] <ezequielg> :P [18:05] <ndufresne> but 16x16 was also supported by TI magical rotator [18:05] <ndufresne> it still magical to me, it suppose to do zero-copy 90deg rotation [18:05] <ndufresne> with some memory bus feature iirc [19:31] <jernej> ezequielg: VP9 hantro G2? that's very good to hear [19:31] <jernej> ping me for porting on Allwinner H6 [19:33] <ezequielg> jernej: ping. [19:33] <ezequielg> :) [19:33] <ezequielg> please start porting? :) [19:33] * ezequielg runs [19:34] <jernej> uh, I'm not following ML lately [19:34] <jernej> it's there already? [19:34] <jernej> currently I'm hacking SDIO support for wifi drivers, so... next week? [19:35] <ezequielg> when you say Allwinner H6, I bet it's some cedrus variant? [19:37] <jernej> no, H6 has secondary VPU, just for VP9 [19:37] <jernej> and it's stripped down G2 variant [19:38] <jernej> so in theory only H6 specifics need to be added, like clocks and reset lines [19:39] <ezequielg> oh la la! [19:39] <ezequielg> are there any specs to read? [19:40] <ezequielg> with info about the VPU I mean [19:42] <jernej> nope [19:42] <jernej> I dumped few cfg regs [19:43] <jernej> it may also support hevc, but vendor code is using it strictly for vp9 [19:48] <ezequielg> super good. [19:49] <ezequielg> my colleague Andrzej is driving all the effort, together with ndufresne and dwlsalmeida on the gstreamer side. [19:49] <ezequielg> We had the rkvdec (the old variant) working and we wanted to have hantro g2 as well, so the vp9 controls got wider adoption. [19:50] <ezequielg> The driver is working well, and looks fairly clean. Adding some quirks to support a stripped down version would be great. [19:50] <ezequielg> i'm all for gaining more users. [20:07] <jernej> sure :) [20:07] <jernej> is code already public? [20:08] <ezequielg> no, but it's about to be. [20:10] <jernej> that's why I said ping me :) [20:14] <ezequielg> we'll Cc you. [20:16] <jernej> thanks! [20:29] <ndufresne> jernej: perhaps considering giving ezequielg your working email, I notice you have an email addr that bounce these days [20:30] <jernej> ndufresne: it was fixed in drm-misc a long time now [20:30] <jernej> I thought it already propagated to stable [20:31] <jernej> I even added alias to .mailmap [20:31] <ezequielg> I'll track that emil and use it [20:31] <ezequielg> *email [20:31] <ezequielg> I add mails manually, sorry about that. [20:32] <jernej> it's in stable already: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/.mailmap?h=v5.13.1#n163 [20:33] <ndufresne> ++ [20:34] *** jm_h has quit IRC (Remote host closed the connection) [20:43] <ezequielg> I was under the impression that only applied to some script and tools, such as get_maintainer. Sometimes you want to Cc specific people. In those cases, I got to git-log and use whatever latest mail I see there. [20:48] *** ao2 has quit IRC (Quit: Leaving) [20:51] *** camus1 has joined #linux-media [20:51] *** camus has quit IRC (Remote host closed the connection) [21:37] *** gouchi has quit IRC (Remote host closed the connection) [22:56] *** jernej has quit IRC (Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net) [22:57] *** jernej has joined #linux-media [22:58] *** jernej has quit IRC (Remote host closed the connection) [23:02] <djrscally> Quick question; does it make sense to have a sensor configured with a horizontal timing size that's lower than the output width? [23:03] *** b-rad has joined #linux-media [23:03] *** jernej has joined #linux-media [23:32] *** b-rad has quit IRC (Ping timeout: 480 seconds) [23:35] *** djrscally has quit IRC (Ping timeout: 480 seconds) [23:41] *** b-rad has joined #linux-media