[Ffmpeg-devel-irc] ffmpeg-devel.log.20191110

burek burek at teamnet.rs
Mon Nov 11 03:05:03 EET 2019


[03:44:08 CET] <taliho> has anyone ever proposed to have a function that discovers pix_fmts at runtime in AVCodec?
[03:44:54 CET] <taliho> this seems important in at least v4l2m2m devices, where the available pix_fmts will depend on the hardware device (so .pix_fmts is not set for v4l2m2m encoders)
[03:45:52 CET] <taliho> at the meoment v4l2m2m encoding just sets it's input format, i.e. nv21 which can be completely different to input frames avctx->pix_fmt (i.e yuv420p)
[03:46:20 CET] <taliho> in fact there's not even an error/warning - just seems to encode yuv420p as nv21 
[03:46:36 CET] <taliho> there is at least one ticket on trac about this
[03:48:27 CET] <taliho> I'm going to add error/warning.. seems easy enough. But would like to work on something that discovers .pix_fmts dynamically. That way the user doesn't even have to set correct -pix_fmts on cli  
[03:49:20 CET] <taliho> anyone have any thoughts about this? 
[03:53:30 CET] <taliho> :jamrial ^^ would appreciate your feedback 
[05:47:45 CET] <taliho> is that call still happening?  
[06:08:13 CET] <kierank> https://mensuel.framapad.org/p/neF434T1a0?lang=en
[06:11:02 CET] <jdarnley> I'm listening to the gitlab stuff
[06:11:13 CET] <jdarnley> thrn I'll be back for the argument at 1500
[06:12:16 CET] <jdarnley> Oh.  A "sergio" I arrived with today wanted to ask about input buffer size for streams only being in bytes, not time.
[06:20:52 CET] <jdarnley> kierank: are you the pink/purple making the notes there?
[06:21:10 CET] <jdarnley> FYI, you can set your name in the top right menu
[06:21:39 CET] <kierank> Yes, just too busy talking
[06:39:29 CET] <kurosu> "But to be more inclusive, the meeting will be joinable from the distance with Skype/Hangouts/Something-similar and live on IRC" <- is this gonna happen or?
[06:42:47 CET] <jdarnley> Beats me if something official will be done.
[06:43:02 CET] <jdarnley> But I am willing to relay some things
[06:43:23 CET] <jdarnley> sorry should be "official"
[06:49:10 CET] <kurosu> To be fair that sounded unpractical
[06:50:06 CET] <j-b> kurosu: we'll try.
[06:50:07 CET] <j-b> come on
[07:11:50 CET] <kierank> please join #ffmpeg-meeting if you want to participate
[07:12:07 CET] <jdarnley> Google hangout somewhere too
[07:14:22 CET] <jdarnley> https://hangouts.google.com/call/xKeJl8WHuMJ67Ue-y0l4AEEM
[07:14:48 CET] <j-b> jamrial: ping
[07:14:54 CET] <j-b> michaelni: ping
[07:15:16 CET] <jamrial> j-b: pong
[07:15:45 CET] <j-b> jamrial: please join
[07:15:52 CET] <BBB> jamrial: if you want to participate in the meeting, please join #ffmpeg-meeting
[07:16:08 CET] <BBB> there's a video stream and we can also participate on IRC
[07:16:38 CET] <jamrial> ok
[07:36:28 CET] <mjbshaw> Is there a specific room for the meeting?
[07:38:10 CET] <kurosu> please join #ffmpeg-meeting if you want to participate
[07:38:39 CET] <kurosu> there's a google hangout too
[07:38:48 CET] <jdarnley> https://hangouts.google.com/call/xKeJl8WHuMJ67Ue-y0l4AEEM
[07:38:51 CET] <mjbshaw> (Carl Eugen)
[07:38:51 CET] <mjbshaw> #join ffmpeg-meeting
[07:40:30 CET] <mjbshaw> TY!
[10:20:56 CET] <jdarnley> ringo (Sergio) is looking for a mentor for libavformat input/output new features
[10:21:03 CET] <jdarnley> He wants to add RIST
[10:21:29 CET] <jdarnley> sergio at ammirata.net
[10:26:58 CET] <jdarnley> Will explain more later
[10:27:01 CET] Action: jdarnley goes afk
[10:49:16 CET] <JEEB> jdarnley: cool.
[13:35:50 CET] <az_>  hello guys. i decode rtsp h264 with option FF_THREAD_SLICE with function avcodec_decode_video2(). Call graph led me into decode_simple_internal() where that is: https://pastebin.com/wxgrKshg. so it appears that decoding occur indeed in one thread. Why is this done that way? Bear in mind that function ff_slice_thread_init() creates threads. Despite that everything goes one thread as i mentioned
[14:35:53 CET] <hero100> az_: does the stream encoded in multi-slices ?
[14:36:31 CET] <az_> how would i know this?
[14:38:20 CET] <az_> I think it doesn't matter though since call graph has no other path and leads definitely to one thread processing
[14:50:14 CET] <hero100> az_: if latency doesn't matter, try thread_frame
[14:51:15 CET] <az_> the problem is about making ffmpeg use my thread pool, not its own
[14:59:34 CET] <cone-743> ffmpeg 03Derek Buitenhuis 07master:d8bf24459b69: avcodec: Add librav1e encoder
[15:40:28 CET] <jdarnley> az_ If you want slice thread decoding then you will need a picture with slices.  If you have that then I don't know what's going on.
[18:29:32 CET] <durandal_1707> fg
[18:31:00 CET] <mkver> fg what?
[19:14:27 CET] <microchip_> mkver: fg on Linux is used to continue a paused background process. Not sure if he means that
[19:15:02 CET] <mkver> Thanks for the info.
[19:15:05 CET] <iive> to bring a process to the foreground, it may be paused or running inthe background.
[19:16:19 CET] <microchip_> Ctrl+Z also pauses a process. I use it sometimes to pause my encodes and continue them with fg
[00:00:00 CET] --- Mon Nov 11 2019


More information about the Ffmpeg-devel-irc mailing list