[Ffmpeg-devel-irc] ffmpeg-devel.log.20180910

[burek]

[00:26:51 CEST] <jamrial> jkqxz: \o/
[00:30:23 CEST] <jkqxz> I'm sure there are plenty more of my stupid bugs for anyone so inclined to find :)
[00:30:57 CEST] <jkqxz> The biggest problem is that it hasn't been run on a large variety of streams - having official test streams checking every path would be incredibly useful.
[00:44:24 CEST] <j-b> that's a large patch...
[00:58:32 CEST] <BtbN> btw., YouTube seems to have rolled out AV1
[00:59:16 CEST] <BtbN> oh, seems like I'm too late with that
[00:59:44 CEST] <jamrial> yeah, but so far only one confirmed video
[01:00:45 CEST] <jamrial> and it was 480p resolution and lower only. 1080p still h264 and vp9 only
[01:00:59 CEST] <BtbN> They are probably still transcoding
[01:04:48 CEST] <jamrial> the 480p av1 stream in that video said something like created on sep 5. I hope that's when the encoder was compiled, and not when the stream was created then :p
[01:05:19 CEST] <j-b> Does it even follow the mp4 spec?
[01:12:26 CEST] <jamrial> looks like it does
[01:12:30 CEST] <atomnuker> its not present in other versions, also the date makes no sense, 9th of may?
[01:12:49 CEST] <jamrial> atomnuker: sep 5
[01:13:35 CEST] <atomnuker> welp, if they can't format their date properly what hope does that leave for the file to be compliant...
[01:14:11 CEST] <jamrial> creation_time   : 2018-09-06T02:27:56.000000Z
[01:14:12 CEST] <jamrial> handler_name    : ISO Media file produced by Google Inc. Created on: 09/05/2018.
[01:28:23 CEST] <j-b> stupid US date format
[01:34:10 CEST] <atomnuker> j-b: is it true in spoken french mostly the 24h time format is used instead of am/pm or just ambiguous 6:30
[09:07:28 CEST] <atomnuker> whoa, avs is ieee standardized
[09:08:04 CEST] <nevcairiel> apparently thats not worth much anymore these days if a ripped-off hevc can get that mark =p
[09:33:56 CEST] <j-b> atomnuker: true.
[09:34:22 CEST] <j-b> atomnuker: we say "eighteen hour thirty" (dix-huit heure trente) for 18:30
[09:35:07 CEST] <j-b> ever heard about NICE codec?
[12:02:59 CEST] <durandal_1707> are there ac-4 files in wild?
[12:03:24 CEST] <JEEB> not that I've seen. only the DASH-IF samples are something I've seen in real life
[12:12:10 CEST] <cone-165> ffmpeg 03Roman Arzumanyan 07master:90ba17652aa1: avfilter: add nvidia NPP based transpose filter
[12:51:11 CEST] <kierank> durandal_1707: don't do ac4 we don't want to support dolby
[12:52:27 CEST] <nevcairiel> we do multimedia, not politics
[12:53:16 CEST] <kierank> get dolby to write it if they want people to use their crappy format
[12:57:01 CEST] <JEEB> I just don't see the use case for AC-4 yet :P
[12:57:10 CEST] <JEEB> so I'm not sure if it's worth the effort at this point
[15:26:35 CEST] <BtbN> wtf
[15:33:06 CEST] <durandal_1707> anybody have Closed Captions specification?
[16:44:34 CEST] <kierank> durandal_1707: "Those patches are nonsense, ffmpeg doesn't have decoder for this type of closed captions." eh?
[16:45:33 CEST] <durandal_1707> kierank: sorry, but consulting is not free at this time
[16:45:42 CEST] <kierank> ffmpeg has cea-608 decoder
[16:45:43 CEST] <kierank> iiuc
[16:46:34 CEST] <durandal_1707> samples uses different bitstream
[17:12:34 CEST] <durandal_1707> worst spammer ever
[17:20:09 CEST] <BtbN> Is that some kind of shitty bot?
[17:21:48 CEST] <gnafu> Are the reports at least legitimate in the sense that these are features that are missing but that could be implemented?
[17:21:57 CEST] <gnafu> Or is it nonsensical stuff?
[17:22:13 CEST] <BtbN> QSV on Linux is just shit and nobody should use it.
[17:25:02 CEST] <gnafu> Aah.
[17:25:36 CEST] <jkqxz> My guess is that some junior Intel person who doesn't really know what they're doing has been told "make bugs against FFmpeg for missing features in libmfx support".
[17:27:14 CEST] <jkqxz> I'll go through them later - I think they are mostly "patch welcome".
[17:30:59 CEST] <jamrial> he's trying a bunch of options, probably from some other command line encoder, that obviously don't exist on ffmpeg
[17:33:53 CEST] <jkqxz> They do sound like they would make sense, though.  (Well, some.  I have no idea what "refpicmarkrep" or "userawref" would be meant to do.)
[21:35:32 CEST] <cone-038> ffmpeg 03Paul B Mahol 07master:df05fd3131fd: avcodec/loco: switch to planar rgb format
[22:01:04 CEST] <jkqxz> Urgh, anonymous unions.  Is it worse to add a configure test to duck the problem or to add pointless members called "_" everywhere?
[22:01:52 CEST] <atomnuker> what uses them?
[22:03:38 CEST] <jkqxz> AV1 CBS.  (The outstanding caption-and-related-stuff patch for H.264 had the same problem, and I should probably send that again at some point.)
[22:05:16 CEST] <philipl> Drop support for ancient compilers. :-)
[22:07:21 CEST] <atomnuker> ^^please
[22:08:13 CEST] <atomnuker> I wouldn't mind a configure check to disable a component for now though, it'll just add more incentive to drop support for compilers without them
[22:08:15 CEST] <jkqxz> I'm not really in favour of that in general, but not supporting new features on them seems reasonable.
[22:08:34 CEST] <jamrial> jkqxz: you're using named unions in other modules
[22:08:51 CEST] <jamrial> like cbs_h2645 the payload union in H265RawSEIPayload
[22:09:16 CEST] <jamrial> just name one in AV1RawOBU something like obu
[22:34:44 CEST] <durandal_1707> i improved decoder significally locally, but will not share it with the world
[22:35:37 CEST] <atomnuker> which one?
[22:36:15 CEST] <gnafu> durandal_1707: Will you release it at VDD?
[22:36:37 CEST] <durandal_1707> atomnuker: ilbc,h264,hevc,prores,vp9,vp8,vp7,vp6,vp5,vp4,vp3,vp2,vp1,vp0,vp-1
[22:36:43 CEST] <durandal_1707> gnafu: sure, if you pay me
[22:37:58 CEST] <gnafu> durandal_1707: Tempting.
[22:41:02 CEST] <durandal_1707> decoding speed: 714x vs 960x
[22:45:37 CEST] <atomnuker> we have an ilbc decoder?
[22:46:06 CEST] <atomnuker> oh, as an external library
[22:48:02 CEST] <durandal_1707> atomnuker: ilbc: wrapper vs native
[22:48:36 CEST] <atomnuker> we don't have a native ilbc decoder, its a voip thing that came in 2003 and died when opus hit 1.0
[22:49:47 CEST] <wbs> it's still more widely used than the wideband codec from the same company, iSAC
[22:50:10 CEST] <durandal_1707> atomnuker: i adopted one
[22:50:23 CEST] <atomnuker> cool
[22:51:21 CEST] <atomnuker> the only thing you can speed up in vp* that they all have in common until vp6 is the entropy decoding system, and I've taken a look at it without much luck
[22:51:52 CEST] <atomnuker> oh and vp9 loopfilter avx2 simd which still hasn't happened and its taking the most time for decoding currently
[23:18:46 CEST] <cone-038> ffmpeg 03Jason Stevens 07master:69f1853a2831: libavcodec/dnxhd: change ff_dnxhd_get_hr_frame_size to avpriv_
[23:18:47 CEST] <cone-038> ffmpeg 03Jason Stevens 07master:563204493991: libavformat/mxfenc: add missing dnxhr mxfcontainer essence ULs
[23:32:44 CEST] <durandal_1707> atomnuker: can you ask atrac9 guy does he work on binka?
[23:40:48 CEST] <atomnuker> durandal_1707: haven't found a good way to contact him other than to open an issue https://github.com/Thealexbarney/LibAtrac9 and ask him
[00:00:00 CEST] --- Tue Sep 11 2018