<!-- Some styling for better description lists --><style type='text/css'>dt { font-weight: bold;float: left;display:inline;margin-right: 1em} dd { display:block; margin-left: 2em}</style>

   ***: svarbanov has joined #linux-media
   <br> eelstrebor has quit IRC (Quit: Ex-Chat)
   <br> NiksDev2 has quit IRC (Ping timeout: 480 seconds)
   <br> NiksDev has joined #linux-media
   <br> NiksDev2 has joined #linux-media
   <br> jm_h has joined #linux-media
   <br> GBenji has joined #linux-media
   <br> camus1 has joined #linux-media
   hverkuil: <u>mchehab</u>: ping (dvb question for you)
   ***: camus has quit IRC (Ping timeout: 480 seconds)
   <br> camus has joined #linux-media
   <br> camus1 has quit IRC (Read error: Connection reset by peer)
   <br> camus1 has joined #linux-media
   <br> camus1 has quit IRC (Read error: Connection reset by peer)
   <br> camus has quit IRC (Ping timeout: 480 seconds)
   <br> camus has joined #linux-media
   <br> djrscally has joined #linux-media
   <br> epoll has quit IRC (Ping timeout: 480 seconds)
   <br> epoll has joined #linux-media
   <br> NiksDev has quit IRC (Remote host closed the connection)
   <br> NiksDev has joined #linux-media
   <br> GBenji has left
   <br> GBenji has joined #linux-media
   <br> paulk has joined #linux-media
   <br> paulk has quit IRC ()
   <br> paulk has joined #linux-media
   <br> paulk has quit IRC (Ping timeout: 480 seconds)
   <br> GBenji has left
   ezequielg: <u>ndufresne</u>: jernej: are guys around?
   <br> i'm looking at VP9 uAPI with andrzej_p.
   <br> I don't really like how the VP9 API between userspace and kernel, is mixed with the hardware interface. IOW, the current patchset is passing values such as enum v4l2_vp9_segment_feature feature to the harware.
   <br> I think it'd be less brittle to decouple the interfaces completely.
   jernej: <u>ezequielg</u>: I am, but not for long
   <br> I'm not familiar with VP9, but I dislike enums too. For API, defines are better IMO.
   ezequielg: OK, so that makes two of us.
   <br> my other dislike is avoid writing values that come from the API such as V4L2_VP9_SEG_LVL_ALT_Q, into the hw.
   jernej: not sure I understand you here
   <br> let me check
   ezequielg: the Hantro driver seems to be passing the values of stuff like V4L2_VP9_SEG_LVL_ALT_Q, directly to hw.
   <br> see update_feat_and_flag()
   jernej: ah, I see
   <br> I never looked at driver that closely, but this part of API always bugged me as unusual
   ezequielg: well, we can still ask for changes :)
   jernej: anyway, I have to go, bbl
   ezequielg: ttys
   <br> <u>andrzej_p</u>: I'll think about this some more.
   <br> (afk)
   ***: eelstrebor has joined #linux-media
   <br> gouchi has joined #linux-media
   ndufresne: <u>ezequielg</u>: seems trivial changes for having a switch that translate to avoid possible random values from user space
   <br> though, you will have to better write down your rational, I have to do a lot of guessing here when reading you
   <br> I haven't checked the API you mention, and neither you or jernej seems to bring a rational why enum vs define is better, it's like saying now using C language would be better
   <br> <u>ezequielg</u>: note that we will likely mark the G2 VP9 as DOT NOT MERGE, the only purpose was to proof the API, but you said you will nack it anyway
   <br> * not using
   ezequielg: why would we not merge it?
   <br> <u>ndufresne</u>: ^
   jernej: <u>ndufresne</u>: I and ezequielg talked about enums vs macros before, there just wasn't official response about that
   ndufresne: <u>ezequielg</u>: you said that, you said this cannot be merged till the control block driver issue is resolved
   ezequielg: I said that about VP9?
   ndufresne: perhaps it was for HEVC, but its the same chip, why would that be different ?
   ezequielg: OK, so let's summarize. For VP9 we are proposing a stable uAPI, and there's a GStreamer MR that goes with it.
   <br> That's 2 out of 3. Yes, you need CTRL BLK for G2...
   <br> For HEVC, you don't have either of that: no stable uAPI, no GStreamer MR and no CTRL BLK for G2.
   <br> So I'm thinking, OK, we maybe could merge Hantro (which BTW, is potentially useful for some Allwinner?)... given it's 2/3 :-)
   <br> I mean, 2 out of 3.
   ndufresne: you'll have to be more clear about your NAKs then, your seem very against anything G2 from my reading of it
   ezequielg: anything against G2 HEVC.
   jernej: <u>ezequielg</u>: what would be useful for some Allwinner?
   ezequielg: Didn't you say there was some Allwinner that had some G2-like?
   jernej: ah, that
   <br> yes, H6 has G2 with VP9 only
   ndufresne: a G2 from middle age ;-P
   <br> <u>ezequielg</u>: I think H2 G2 will take time, a lot of delta to figure-out
   <br> *H6
   jernej: true, I aligned some registers but still without luck
   <br> it doesn't lock up though, what I already consider partial success :)
   ndufresne: or we need to find a way to not rely on luck
   ezequielg: also, I'm not opposed to "anything G2 HEVC". My main feedback was... let's make sure we work towards an HEVC uAPI patch that can take the uAPI out of staging, unblocking GStreamer/FFmpeg. There were other patchsets for G2 HEVC, which I didn't NAK, I just gave feedback which wasn't addressed.
   jernej: <u>ndufresne</u>: I have zero documentation, so searching for deltas is all that's left - and that requires luck :)
   ndufresne: (sorry, HEVC is not on my priority list, so I just ignore all this)
   <br> yeah, and here I only got documentation that says a) might have been different prior to 7XXX, not enough to help
   <br> a working implementation we could spy on would be nice
   jernej: well, there are vendor binaries, but they are pain to use
   ezequielg: <u>ndufresne</u>: I'd like hverkuil opinion here, but I'd say it'd be cool to merge G2 VP9, and then be able to say "oh, this is only missing CTRL-BLK", as a way to encourage NXP to fix it.
   ***: camus1 has joined #linux-media
   <br> camus has quit IRC (Ping timeout: 480 seconds)
   ndufresne: <u>ezequielg</u>: ok, well, I relaly don't mind, that focus is on the API anyway, I'm sure there have been much worst driver being merged in the past that no one could ever know it was unusable
   ***: jm_h has quit IRC (Remote host closed the connection)
   <br> gouchi has quit IRC (Remote host closed the connection)
   <br> camus has joined #linux-media
   <br> camus1 has quit IRC (Ping timeout: 480 seconds)
   <br> djrscally has quit IRC (Ping timeout: 480 seconds)
   <br> NiksDev2 has quit IRC (Ping timeout: 480 seconds)
   <br> NiksDev has quit IRC (Ping timeout: 480 seconds)