[06:07] *** Tex_Nick has left "In Linux, We Trust" [07:17] <hverkuil> mchehab: just as a reminder (making sure it won't be forgotten): https://patchwork.linuxtv.org/patch/33908/ [07:29] <hverkuil> nohous: ping [07:29] <nohous> hverkuil: pong [07:31] <hverkuil> I did some more digging this week in how HDMI handles deep color and things like YCbCr 4:2:2 and 4:2:0. [07:31] <hverkuil> This might be of interest for you. [07:32] <hverkuil> Deep color (10/12/16 bits per color component) is actually done by increasing the TMDS data clock. [07:32] <hverkuil> The dv timings struct contains a clock frequency, but the thing to remember is that this is the pixel clock frequency, not the data clock frequency. [07:33] <nohous> allright, but this is hidden by the adv chip i presume [07:33] <hverkuil> So for 10 bit color the data clock frequency will be increased a bit to handle this. Obviously there is a max frequency that the hw can handle and that is encoded in the EDID so the transmitter will now the max data clock rate. [07:34] <hverkuil> This is indeed handled by the adv (well, I think. We don't yet have deep color support since nobody needed it.) [07:34] <nohous> yeah, i've came to something similar with sdi [07:35] <nohous> the memory format is now independent of the data format on the sdi line [07:35] <nohous> and i have some mapping routines that calculate sdi mode (hd, 3g, 6g) out of timing / sdi pixel format information [07:36] <hverkuil> The HDMI handling of YCbCr 4:2:2 is weird: it is using the same data clock as for YCbCr 4:4:4 but the Y, Cb and Cr component are transferred as 12 bit values. [07:36] <hverkuil> That's fixed, there is no 8, 10 or 16 bit YCbCr 4:2:2. [07:36] <nohous> aha, interesting [07:37] <hverkuil> YCbCr 4:2:0 is really reducing the bandwidth and is packing the data smartly. This does support all bit depths, again by increasing the data clock with the pixel clock being constant. [07:39] <nohous> is 4:2:0 defined past hdmi 1.4? because i don't see it in specs [07:40] <hverkuil> It's part of HDMI 2.0, but it can be used with HDMI 1.4 receivers (provided they can unpack it). The reason 4:2:0 was added to HDMI was to support 4kp30 with 1.4 rated cables (max pixelclock of 148.5 MHz). [07:41] <hverkuil> Since most sources are already using 4:2:0 (H264/265, MPEG, etc) and at fps rates <= 30 (max NTSC framerate), this is OK. [07:42] <hverkuil> This is way older/cheaper 4k TVs can get away with HDMI 1.4 inputs. [07:42] <hverkuil> s/way/why/ [07:43] <nohous> yeah, we've once reworked stock cheap 4k TV to have dual dvi input, because it only had hdmi 1.4 if i recall correctly ;-) [07:43] <nohous> and the application required 4:4:4 rgb [08:03] <hverkuil> posciak: ping [11:02] <mchehab> hverkuil: I'm placing the slides using during the media summit at: https://linuxtv.org/downloads/presentations/media_summit_2016_san_diego/ [11:02] <mchehab> I guess the only things missed there are your slides [11:03] <mchehab> did you send it already? If not, could you please send them to me? [11:05] <hverkuil> I don't believe I had any slides. [11:05] <mchehab> ah, ok [11:06] <hverkuil> The only thing I showed was one slide of my CEC presentation (I think). [11:06] <mchehab> IMHO, it would be good to post your CEC slides there, as otherwise the discussions about CEC seem incomplete ;) [11:07] <mchehab> also, you assumed there that people interested on CEC listened to your presentation [11:09] <hverkuil> https://docs.google.com/viewer?url=http%3A%2F%2Fevents.linuxfoundation.org%2Fsites%2Fevents%2Ffiles%2Fslides%2Fcec-wide_1.pdf [11:25] <mchehab> hmm.... I guess pinchartl also presented some slides there for the request API... [11:25] <mchehab> pinchartl: could you please send me the slide deck you used at the Media Summit? [11:26] <hverkuil> mchehab: these should be available from events.linuxfoundation.org, just like mine. [11:26] <hverkuil> http://events.linuxfoundation.org/events/embedded-linux-conference/program/slides [11:27] <hverkuil> Hmm, it looks like Laurent didn't upload his slides. [11:28] <hverkuil> Note that his slides are useless without the video, which isn't available yet. [11:35] <mchehab> I'll keep his slides out for this draft [11:42] <mchehab> https://linuxtv.org/news.php?entry=2016-04-20.mchehab [11:47] <mchehab> the group photo was too dark and blank balance is wrong. I did a quick fix just to make it look a little better, but it is not 100%. If someone wants to try improving it, I can send the original images took during the meeting. [11:56] <larsc> the bright background destroys the auto balance [13:22] <hverkuil> Ah! ELCE is three days again instead of two and moved to Oct 11-13. [13:29] <larsc> and you can buy a mini-summit slot for only 20k$! we better start saving up [13:32] <hverkuil> Co-located with OpenIoT again. Makes sense. [13:34] <hverkuil> 11-13 is Tue-Thu, so having a summit on Monday would work quite well. [13:35] <pinchartl> hverkuil: I did upload my slides, I don't know why they didn't make it to the linux foundation website [13:35] <pinchartl> mchehab: they're also available from http://ideasonboard.org/talks/20160405-elc.pdf [13:37] <pinchartl> larsc: I wasn't even sure if you were joking :-) [14:02] * headless notices that LinuxCon is still Oct 4-6 [14:03] <larsc> that's the point. they wanted to make them separate [14:04] <larsc> otherwise ELCE would have been only 2 days [14:04] <larsc> now it is 3 [14:04] <headless> hm [14:04] <hverkuil> And they added OpenIoT together with ELCE. [14:32] <mchehab> pinchartl: thanks! slides uploaded [14:32] <mchehab> and pointed at the web draft version of the report [16:04] *** benjiG has left [16:23] *** awalls has left [18:48] <mchehab> hverkuil: I'll apply your smatch patches directly... [18:48] <mchehab> having less errors on my console always helps my workflow ;) [19:44] <shuah> mchehab: I see use-after-free in media_ioctl+0xf0/0x130 [media] on Linux 4.6-rc4 - fyi [19:45] <mchehab> yes, you should see, because the fixes didn't went to it [19:45] <shuah> this is when I unbind uvc_driver while app is running calling ioctl MEDIA_IOC_DEVICE_INFO in a loop with 10 sec sleep in between [19:45] <mchehab> they're not applied on my tree yet [19:45] <shuah> okay good [19:47] <_franck_> mchehab: hi, have you read this: http://www.spinics.net/lists/linux-media/msg98847.html ? [20:17] <shuah> pinchartl, sailus: I noticed your names don't pupup in getmaintainer scripts on media controller files (media_devnode.c) for example. I am adding your names to the patches - let me know if you don't want to ne bothered [20:21] *** msev- has left "Leaving" [20:25] <pinchartl> shuah: that's a good point. maybe Sakari and I should be added to MAINTAINERS for those files ? [20:37] <shuah> yeah probably - that way you don't have to sift through the patches